Children’s English: Principles-and-Parameters Perspectives

Andrew Radford, University of Essex, 10 May 2005

List of contents
Chapter heading Page

1 Language acquisition and principles of Universal Grammar………………………2

2 Parameters and parameter-setting………………………………………………....20

3 The structure-building model……………………………………………………..37

4 The truncation model……………………………………………………………..56

5 The continuity model.…………………………………………………………….69

6 The underspecification model……………………………………………………83

7 The syntactic development of bilingual children…………….…………………105

8 Language mixing in bilingual children……….………………….……………..117

9 References………………………………………………………………….…..134

10 Suggested assignment and dissertation topics………………………………..145

11 Appendix: The Claire files……………………………………………………148

 1 Language acquisition and principles of Universal Grammar
The Kuczaj files on the CHILDES data-base1 show that (when asked by his linguist
father how babies learn to talk) a boy called Abe (at age 3 years 9 months) replied
‘I think they…they hear their mother and father talking to them and they listen,
and that’s how…how they learn how to talk.’ And when his father went on to ask
‘Does somebody teach babies how to talk, or do they just learn how to talk?’ Abe
replied ‘They just learn’.

1.1 Introduction
Children2 generally produce their first recognisable word (e.g. Mama or Dada) by the time
of their first birthday. For the next 6 months or so, there is little apparent evidence of grammatical
development in their speech production, although the child’s productive vocabulary typically increases
by about five words a month until it reaches around 30 words at age 1;6 (1 year and 6 months).
Throughout this single-word stage, children’s utterances typically comprise single words spoken in
isolation: e.g. a child may say Apple when reaching for an apple, or Up when wanting to climb up onto
her mother’s knee. During this single-word stage, it is difficult to find any clear evidence of the
acquisition of grammar, in that children do not make productive use of inflections (e.g. they don’t add
the plural -s ending to nouns, or the past tense -d ending to verbs), and don’t productively combine
words together to form two- and three-word utterances.
At around the age of 1;6 (though with considerable variation from one child to another), we find the
first visible signs of the acquisition of grammar: children start to make productive use of inflections
(e.g. using plural nouns like doggies alongside the singular form doggy, and inflected verb forms like
going/gone alongside the uninflected verb form go), and similarly start to produce elementary two-
and three-word utterances such as Want Teddy, Eating cookie, Daddy gone office, etc. From this point
on, there is a rapid expansion in their grammatical development, until by the age of around 30 months
they have usually acquired most of the inflections and core grammatical constructions used in English,
and are able to produce adult-like sentences such as Where’s Mummy gone? What’s Daddy doing? Can
we go to the zoo, Daddy? etc. (though occasional morphological and syntactic errors persist until the
age of four years or so – e.g. We goed there with Daddy, What we can do? etc.).
So, the central phenomenon which any theory of language acquisition must seek to explain is this:
how is it that after a long drawn-out period during which there is no obvious sign of grammatical
development, at around age 1;6 there is a sudden spurt as multiword speech starts to emerge, and a
phenomenal growth in grammatical development then takes place over the next year? This uniformity
and (once the spurt has started) rapidity in the pattern of children’s linguistic development are the
central facts which a theory of language acquisition must seek to explain. But how?

1.2 Evidence for an innate Language Faculty

Someone whose approach to the question of how children acquire grammar has had
enormous influence over the past 50 years is Noam Chomsky, who for most of this time has been
researching at MIT (the Massachusetts Institute of Technology, in Cambridge Massachussetts).
Although Chomsky’s own work has been concerned with developing theoretical models of language
(not directly with language acquisition), his work has nonetheless had a profound impact on
acquisition research. He maintains that the most plausible explanation for the uniformity and rapidity
of first language acquisition is to posit that the course of acquisition is determined by a biologically
endowed innate Language Faculty (or language acquisition program, to borrow a computer software
metaphor) within the brain, which provides children with a genetically transmitted algorithm (i.e. set
of procedures) for developing a grammar, on the basis of their linguistic experience (i.e. on the basis
of the speech input they receive). The way in which Chomsky visualises the acquisition process can be
1
http://childes.psy.cmu.edu
2
Throughout, my main concern will be with children acquiring English as a first language. Initially, I shall focus
on monolingual children, before turning to look at bilinguals in the final two chapters. The material on pages 1-3
is (with minor amendments) taken from Radford (2004a), pp.6-9.
represented schematically as in (1) below (where L is the language being acquired):

(1) Experience  Language  Grammar

of L Faculty of L

Children acquiring a language will observe people around them using the language, and the set of
expressions in the language which a child hears (and the contexts in which they are used) in the course
of acquiring the language constitute the child’s linguistic experience of the language. This experience
serves as input to the child’s language faculty, which provides the child with an algorithm (i.e. set of
procedures) for (subconsciously) analysing the experience and devising a grammar of the language
being acquired. Thus, the input to the language faculty is the child’s experience, and the output of the
language faculty is a grammar of the language being acquired.
The hypothesis that the course of language acquisition is determined by an innate language faculty
is known popularly as the Innateness Hypothesis. Chomsky maintains that the ability to speak and
acquire languages is unique to human beings, and that grammars of natural languages incorporate
principles which are also unique to humans and which reflect the nature of the human mind:

Whatever evidence we do have seems to me to support the view that the ability to acquire
and use language is a species-specific human capacity, that there are very deep and restrictive
principles that determine the nature of human language and are rooted in the specific
character of the human mind. (Chomsky 1972, p. 102)
Moreover, he notes, language acquisition is an ability which all humans possess, independently of
their general intelligence:
Even at low levels of intelligence, at pathological levels, we find a command of language that
is totally unattainable by an ape that may, in other respects, surpass a human imbecile in
problem-solving activity and other adaptive behaviour. (Chomsky 1972, p. 10)
In addition, the apparent uniformity in the types of grammars developed by different speakers of the
same language suggests that children have genetic guidance in the task of constructing a grammar of
their native language:
We know that the grammars that are in fact constructed vary only slightly among speakers
of the same language, despite wide variations not only in intelligence but also in the
conditions under which language is acquired. (Chomsky 1972, p. 79)
Furthermore, the rapidity of acquisition (once the grammar spurt has started) also points to genetic
guidance in grammar construction:
Otherwise it is impossible to explain how children come to construct grammars...under the
given conditions of time and access to data. (Chomsky 1972, p. 113)
(The sequence ‘under...data’ means simply ‘in so short a time, and on the basis of such limited
linguistic experience.’) What makes the uniformity and rapidity of acquisition even more remarkable
is the fact that the child’s linguistic experience is often degenerate (i.e. imperfect), since it is based
on the linguistic performance of adult speakers, and this may be a poor reflection of their competence:
A good deal of normal speech consists of false starts, disconnected phrases, and other
deviations from idealised competence. (Chomsky 1972, p. 158)
If much of the speech input which children receive is ungrammatical (because of performance errors),
how is it that they can use this degenerate experience to develop a (competence) grammar which
specifies how to form grammatical sentences? Chomsky’s answer is to draw the following analogy:
Descartes asks: how is it when we see a sort of irregular figure drawn in front of us we see it
as a triangle? He observes, quite correctly, that there’s a disparity between the data presented
to us and the percept that we construct. And he argues, I think quite plausibly, that we see the
figure as a triangle because there’s something about the nature of our minds which makes the
 image of a triangle easily constructible by the mind. (Chomsky 1968, p. 687)
The obvious implication is that in much the same way as we are genetically predisposed to analyse
shapes (however irregular) as having specific geometrical properties, so too we are genetically
predisposed to analyse sentences (however ungrammatical) are having specific grammatical
properties. (For evaluation of this degenerate input argument, see Pullum and Scholz 2002, Thomas
2002, Sampson 2002, Fodor and Crowther 2002, Lasnik and Uriagereka 2002, Legate and Yang 2002,
Crain and Pietroski 2002, and Scholz and Pullum 2002.)
A further argument Chomsky uses in support of the Innateness Hypothesis relates to the fact that
language acquisition is an entirely subconscious and involuntary activity (in the sense that you can't
consciously choose whether or not to acquire your native language – though you can choose whether
or not you wish to learn chess, for example). It is also an activity which is largely unguided (in the
sense that parents don’t teach children to talk):
Children acquire...languages quite successfully even though no special care is taken to teach
them and no special attention is given to their progress. (Chomsky 1965, pp. 200-1)
The implication is that we don’t learn to have a native language, any more than we learn to have arms
or legs; the ability to acquire a native language is part of our genetic endowment – just like the ability
to learn to walk.
Studies of language acquisition lend empirical support to the innateness hypothesis. Research has
suggested that there is a critical period for the acquisition of syntax, in the sense that children who
learn a given language before the age of around 7 years achieve native competence in it, whereas those
acquire a (first or second) language after that age rarely manage to achieve native-like syntactic
competence: see Lenneberg (1967), Hurford (1991) and Smith (1998, 1999) for discussion. A
particularly poignant example of this is a child called Genie (See Curtiss 1977, Rymer 1993), who was
deprived of speech input and kept locked up on her own in a room until age 13. When eventually taken
into care and exposed to intensive language input, her vocabulary grew enormously, but her syntax
never developed. This suggests that the acquisition of syntax is determined by an innate ‘language
acquisition programme’ which is in effect switched off (or gradually atrophies) at around the onset of
puberty. (For further discussion of the Innateness Hypothesis, see Antony and Hornstein 2002.)

1.3 Universal Grammar and the acquisition of syntax

If (as Chomsky claims) human beings are biologically endowed with an innate language
faculty, an obvious question to ask is what is the nature of the language faculty. An important point to
note in this regard is that children can in principle acquire any natural language as their native
language (e.g. Afghan orphans brought up by English-speaking foster parents in an English-speaking
community acquire English as their first language). It therefore follows that the language faculty must
incorporate a theory of Universal Grammar/UG which enables the child to develop a grammar of
any natural language on the basis of suitable linguistic experience of the language (i.e. sufficient
speech input). Experience of a particular language L (examples of words, phrases and sentences in L
which the child hears produced by native speakers of L in particular contexts) serves as input to the
child’s language faculty which incorporates a theory of Universal Grammar providing the child with a
procedure for developing a grammar of L.
If the acquisition of a grammar is indeed controlled by a genetically endowed Language Faculty
incorporating a theory of UG, then it follows that certain aspects of child (and adult) competence are
known without experience, and hence must be part of the genetic information about language with
which we are biologically endowed at birth. Such aspects of language would not have to be learned,
precisely because they form part of the child’s genetic inheritance. If we make the (plausible)
assumption that the language faculty does not vary significantly from one (normal) human being to
another, those aspects of language which are universal will be determined by innate Principles of
Universal Grammar (= UG principles) genetically wired into the brain. Over the past three decades,
research into the syntax of a wide range of languages has led Chomsky and his collaborators to posit a
wide range of putatively universal syntactic principles. In order to give a flavour of the kind of
principles involved, I will briefly present a simplified discussion of the syntactic structure of one
sentence and show it would be analysed from a Chomskyan perspective. In subsequent sections, I will
look at some of the principles underlying the analysis, and ask whether the same principles also
operate in child grammars.
In order to illustrate the nature of adult syntactic structures and the different types of operation
involved in their formation, let’s look at how the following wh-question is formed:
(2) What has Jim not understood?
The wh-word what has the (semantic/thematic) function of being the complement of the verb
UNDERSTAND, as we can see from the corresponding echo question in (3) below:

(3) Jim has not understood what?

The fact that what is positioned immediately after understood in (3) suggests that it is indeed the
(direct object) complement of the verb UNDERSTAND, given that complements are generally positioned
after their associated verbs in English. In much the same way, Jim has the function of being the
(semantic/thematic) subject of the verb UNDERSTAND, in the sense that Jim is the person whose
understanding is in question. To use traditional semantic terminology, UNDERSTAND is a predicate,
and Jim and what are its two arguments (Jim being the subject argument of the predicate, and what
being its object argument). If we assume (in accordance with the VP-Internal Subject Hypothesis
adopted in work over the past twenty years) that that constituents originate in a position which reflects
their semantic function, this means that what originates as the complement of the verb UNDERSTAND
and Jim as its subject. Accordingly, sentence (2) is derived as follows.
The verb UNDERSTAND merges (i.e. ‘combines’) with its complement what and its subject Jim to
form the verb phrase Jim understand what which has the syntactic structure shown in simplified form
below3:
(4) [VP Jim [V UNDERSTAND] what]
Because sentence (2) is negative, the verb phrase in (4) is then negated. This is done by merging a null
negative head (conventionally denoted as NEG) with the verb phrase in (4) and with the negative
particle not, forming the NEGP (= Negative Phrase) in (5) below4:
(5) [NEGP not [NEG ø] [VP Jim [V UNDERSTAND] what]]
The resulting NEGP constituent is then merged with a T (i.e. tense-marked auxiliary) constituent
containing the perfect-aspect auxiliary HAVE, forming the TP shown below:
(6) [TP [T HAVE] [NEGP not [NEG ø] [VP Jim [V UNDERSTAND] what]]]
Under the analysis of verb morphology in Adger (2003), the verb UNDERSTAND will agree in aspect
with the perfect-aspect auxiliary HAVE, and so will be assigned a perfect-aspect feature (leading to it
eventually being spelled out in the morphophonology as the perfect participle understood). The
auxiliary HAVE will itself agree (in person and number) with the third person singular expression Jim,
and so (if used in a present-tense context) is eventually spelled out as the third-person-singular
3
Here (and throughout), I simplify syntactic structure in numerous ways in order to simplify exposition for those
who may not be entirely familiar with recent work in theoretical syntax. These include: (i) showing only (some)
heads and phrases, not intermediate (single-bar) constituents; (ii) assuming that verb phrases have a simple,
single-layered structure rather than the complex, multi-layered shell structure (comprising an inner VP core and
an outer vP shell) outlined in chapter 9 of Radford (2004a, 2004b); and (iii) not providing details of the structure
of expressions which are not directly relevant to the discussion at hand. A convention adopted throughout is that
small capitals (e.g. for UNDERSTAND) are used to indicate that the precise phonetic form of the relevant item has
not yet been determined at the relevant stage of derivation, so that the notation UNDERSTAND means ‘an
appropriate form of the verb UNDERSTAND which will be determined at a later stage of derivation’.
4
The analysis of negation outlined here follows that in §5.7 of Radford (2004a, 2004b). A
key assumption made there is that the negative participle not is the specifier rather than
the head of the NEGP constituent containing it. One reason for assuming this is in order to
account for how BE (in its use as a copular verb) can raise from V through NEG into T in
sentences like It is not raining any more; if not occupied the head NEG position in NEGP,
movement of BE into NEG would be blocked (because the NEG position would be occupied
by not and not is not a clitic and hence does not allow another item to attach to it), and
such sentences would wrongly be predicted to be ungrammatical.
present-tense form has. Because a T constituent requires a syntactic subject of its own, the T
constituent containing HAVE attracts the closest noun or pronoun expression to move to become its
subject5; accordingly, Jim moves from being the semantic subject of UNDERSTAND to becoming the
syntactic subject of HAVE in the manner shown by the arrow below6:
(7) [TP Jim [T has] [NEGP not [NEG ø] [VP Jim [V understood] what]]]

The next stage in the derivation involves converting the relevant structure into a question. This is done
by merging the TP in (7) with a null interrogative C constituent (C being a clause-type marker known
as a complementiser, and serving to indicate whether a clause is declarative, interrogative, imperative
or declarative in force). Following an idea suggested by Baker (1970) and adopted in Chomsky
(1995), let’s suppose that a TP is converted into a question by being merged (i.e.combined) with a C
constituent containing an abstract (i.e. null/invisible) question particle Q. Merging this question
particle with the TP in (7) will form the CP in (8) below:
(8) [CP [C Q] [TP Jim [T has] [NEGP not [NEG ø] [VP Jim [V understood] what]]]]
C in main-clause questions attracts a finite T constituent beneath it to move into C to attach/adjoin to
the question particle Q, and so C in (8) attracts the T constituent has to move into C and adjoin to Q (a
movement operation traditionally known as auxiliary inversion and referred to in more recent work
as T-to-C movement): if we suppose that Q (like the Latin question article -ne) is a suffix, then it
follows that the auxiliary has will end up positioned immediately to the left of Q. In a wh-question, C
also attracts the closest wh-word to move into the specifier position in front of C, at the lefthand edge
of CP (a movement operation traditionally known as wh-movement). The relevant movement
operations are arrowed in (9) below7:

(9) [CP What [C has+Q] [TP Jim [T has] [NEGP not [NEG ø] [VP Jim [V understood] what]]]]

The auxiliary has can optionally cliticise (i.e. attach itself) to the immediately preceding wh-word
what in informal styles, so deriving What’s Jim not understood?
Our discussion of the derivation of sentence (2) What has Jim not understood? can be used to
illustrate a number of UG principles (i.e. principles of Universal Grammar) which determine the
nature of syntactic structure. For example, Chomsky in recent work (1999, p.2) has posited a
Structural Uniformity Principle to the effect that phrases and sentences have a uniform (and hence
universal) structure across languages (in the sense that each constituent of a given type has the same
structure in all languages). For example, we might suppose that all sentences universally have a
CP+TP+(NEGP+) VP structure8, and that this is a matter of conceptual necessity, since sentences must
contain a C constituent to mark clause-type (for example, the complementiser that marks the structure
5
In Chomsky’s recent work (1998, 1999, 2001, 2002, 2004), this is assumed to be a consequence of T having a
feature (called an EPP feature) which requires T to attract the closest expression with which it agrees in
person/number to become its subject.
6
Each of the words in (7) is shown in its orthographic spellout form (i.e. in the form it has
in the English spelling system) because the spellout form can be determined at this
stage, once the relevant agreement operations have applied. A convention adopted
throughout is that strikethrough is used to indicate the position which was occupied by a
moved constituent before it moved.
7
Derivations are simplified throughout by ignoring the possibility that a wh-expression which moves out of a
verb phrase moves to the edge of the verb phrase before moving to the edge of CP: see chapter 10 of Radford
(2004a, 2004b) for the rationale behind such a two-step approach to wh-movement.
8
The parentheses around NEGP are intended to mean that a negative clause will contain a NEGP constituent
positioned between TP and VP – though a positive clause will not contain a NEGP constituent. It should be noted
that Chomsky claims that there are a class of defective complement clauses which lack the CP layer found in
other clauses – e.g. the bracketed complement clauses in examples (19a, b) discussed later in the text.
it introduces as a finite declarative (i.e. statement-making) complement clause, and if as a finite yes-no
question), and must contain a T constituent to specify the (e.g. present or past) time at which an event
takes place, and must contain a V constituent to specify the nature of the relevant event. Likewise, a
negative clause must (again as a matter of conceptual necessity) contain a NEGP constituent marking
the fact that the clause is negative.
In the various sections below, we shall look at a number of UG principles, discuss how they
operate in adult English, and produce evidence that they also operate in child grammars.

1.4 The Copy Principle

Just as there are principles which determine the nature of syntactic structure (like the
Structural Uniformity Principle discussed at the end of the previous section), so too there are
principles which determine how syntactic operations (like movement and agreement) work. One such
(tacitly assumed in our discussion in the previous section) is the following principle governing how
movement works:
(10) Copy Principle
Movement is a copying operation by which a copy of a moved constituent is merged in a new
position at the edge (= in the head or specifier position) of a functional projection 9, with
spellout conditions determining which of the copies will have an overt spellout (and so be
pronounced) and which will have a null spellout (and so be silent)
Although we did not discuss spellout in the previous section, we can suppose that it is subject to the
following condition in languages like English (see Abels 2001, Bošković 2001, Chomsky 2005b and
Hiraira 2005):
(11) Spellout Condition
Material is spelled out overtly on the highest copy containing it, and generally has a null
spellout on other copies (unless a particular lower copy is required to have an overt spellout
to satisfy some independent condition)
To see how (11) works, consider our earlier structure (9) repeated as (12) below:
(12) [CP What [C has+Q] [TP Jim [T has] [NEGP not [NEG ø] [VP Jim [V understood] what]]]]
The condition in (11) that only the highest/leftmost copy of a moved constituent is overtly spelled out
correctly predicts that the leftmost occurrences of what, has and Jim will be the only copies which are
overtly spelled out – all the other copies having a null spellout (and so being silent).
If (as Chomsky claims) UG principles are genetically wired into an innate
Language Faculty, it follows that children do not have to learn principles like the
Copy Principle. It also follows that all the structures children produce (throughout
the various stages of acquisition they go through) will be constrained by (i.e.
obey) the relevant principles. An obvious question to ask, therefore, is whether
there is any empirical evidence that UG principles like the Copy Principle are
indeed operative in child grammars.
An experimental study by Guasti, Thornton and Wexler (1995) provides evidence that
this is the case. They undertook an experimental study of children’s questions using prompts such as ‘I
hear the snail doesn’t like some things to eat. Ask him what’ (the target response being What don’t you
like to eat?’). They report that ‘Children’s positive questions were, for the most part, grammatical’,
and that ‘Children’s negative declaratives were adult in form’. However, they also report that all the
children in their study produced negative questions (like those below) which involve auxiliary
copying:
(13)(a) Why could Snoopy couldn’t fit in the boat? (Kathy 4;0)
(b) What did he didn’t wanna bring to school? (Darrell 4;1)
(c) What kind of break do you don’t like? (Rosy 3;10)
(d) How can Ernie can’t sit? (Emily 4;2)

9
A functional projection is a constituent like CP, TP or DP whose head is a functor (i.e. functional morpheme).
Each of the sentences in (13) contains an inverted auxiliary in C, and an overt copy of the auxiliary in
T which has the contracted negative particle clitic n’t attached to it. How does overt auxiliary copying
come about – and why do the children use it in negative questions but not generally in other questions
or other negatives?
By way of background, it should be noted that n’t in English is a clitic (a reduced form of the word
not) and that we find two different types of clitic (i.e. contracted word form) across languages. Some
clitics are syntactic clitics which attach to a particular host in the syntax, and then move along with the
host if the host undergoes syntactic movement. The negative clitic n’t is a syntactic clitic in adult
English, in that it attaches to a preceding finite T constituent (e.g. a present- or past-tense auxiliary) in
the syntax and then can move along with T if T undergoes inversion and moves to C – as in sentences
like:
(14)(a) He doesn’t understand syntax (b) Doesn’t he understand syntax?
A second type of clitic are PF clitics – i.e. clitics which attach to a suitable overt host which ends up
being adjacent to them in the PF component (i.e. morphophonology), after syntactic movement
operations have applied. The contracted form ’s of is functions as a PF clitic, as we can see from
sentences like:
(15)(a) Who do you think’s telling the truth? (b) *Who’s do you think telling the truth?
If (contrary to what we claim here) ’s were a syntactic clitic, it would attach to its subject who in the
syntax and then undergo wh-movement along with who, so wrongly predicting that sentences like
(15b) are grammatical. But in fact ’s is a PF clitic, and attaches to the closest overt item which
precedes it at PF – this being the verb think in the case of (15a).
Since UG allows for two types of clitic, children are clearly going to have to learn whether a given
clitic they come across in the language they are acquiring is a syntactic clitic or a PF clitic. Suppose
that children acquiring English sometimes wrongly assume that n’t is (or can function as) a PF clitic.
If so, n’t will remain in situ in the syntax, so that a sentence like (13a) will have the syntactic structure
below (simplified by showing only copies of the inverted auxiliary could):
(16) [CP Why [C could+Q] [TP Snoopy [T could] [NegP n’t [Neg ø] [VP [V fit] in the boat]]]]
In particular, auxiliary inversion will result in a copy of the auxiliary could in T
being placed in C. In accordance with the Spellout Condition (11), we might
expect the higher/lefthand copy of the auxiliary could to have an overt spellout,
and the righthand copy to have a null spellout – as in (17) below:
(17) [CP Why [C could+Q] [TP Snoopy [T could] [NegP n’t [Neg ø] [VP [V fit] in the boat]]]]
But the resulting structure would leave the clitic unattached, in violation of the
following condition:
(18) Attachment Condition
A clitic/affix cannot remain unattached and must be attached to an appropriate overt host
However, note that our Spellout Condition (11) says that a lower copy generally
has a null spellout unless ‘required to have an overt spellout for some
independent reason’. In the case of a syntactic structure like (16), there is an
‘independent reason’ to give the lower/righthand occurrence of could an overt
spellout, since the negative clitic n’t requires an immediately adjacent overt
auxiliary host to attach to (in order for the clitic not to remain unattached and
thereby violate the Attachment Constraint). Accordingly, both occurrences in
could in (16) are given an overt spellout – as in (13a) Why could Snoopy couldn’t
fit in the boat? And sentences like (13a) in turn provide us with independent
evidence that UG principles like the Copy Principle operate in child grammars.

1.5 The Phase Impenetrability Condition

 A considerable amount of evidence has been uncovered over the last few decades that UG
incorporates principles which constrain (i.e. limit/restrict) how far a constituent can be moved in any
single movement operation. One such principle developed by Chomsky in recent work
(1998, 1999, 2001, 2002, 2004) is the Phase Impenetrability Condition. We can
illustrate the kind of phenomena which it seeks to account for by considering why
an active sentence like (19a) has the passive counterpart (19b), but an active
sentence like (20a) does not have a passive counterpart like (20b):
(19)(a) We intend [this to happen] (b) This is intended [to happen]
(20)(a) We intend [for this to happen] (b) *This is intended [for to happen]
In varities of English like mine, the verb intend can take an infinitive complement with or without the
complementiser for, but only allows passivisation of the (italicised) complement-clause subject this
when for is not used. In order to try and undertand why this should be, let’s look at the derivation of
(19b) and (20b) in turn.
In recent work, infinitival to has been taken to be a non-finite tense particle (belonging to the
category T) which is often future in meaning – as we see from the fact that (19b) can be crudely
paraphrased as ‘It is intended that this will happen at some stage in the future’. Given the assumption
that to is an infinitival T constituent, let’s suppose that we have reached a stage in the derivation of
(19b) at which we have formed the infinitival TP this to happen. This TP is then merged with the
(passive participle) verb intended, forming the verb phrase/VP intended this to happen, and this VP is
then merged with the auxiliary BE10, forming the TP structure shown below11:
(21) [TP [T BE] [VP [V intended] [TP this [T to] happen]]]
The T constituent BE searches for a constituent to agree in person and number with, and locates the
pronoun this (with the result that via agreement, BE is marked as a third person singular form and so
eventually spelled out as the form is). But T also has the property that it requires a syntactic subject of
its own, and so looks to see if there is any suitable expression which it can attract to become its
subject/specifier, locating this as a suitable candidate. T therefore attracts the pronoun this to become
its subject (and specifier), resulting in the structure shown in simplified form below:
(22) [TP This [T is] [VP [V intended] [TP this [T to] happen]]]
This type of movement operation is traditionally termed passivisation (though is known in work over
the past 30 years as A-movement).
Although this can be passivised (i.e. be moved from being the subject of to and become the subject
of is) in a sentence like (19b), passivisation is not possible in for-infinitive structures like (20b) *This
is intended for to happen. Why should this be? The answer lies is another UG principle developed by
Chomsky in recent work, which he terms the Phase Impenetrability Condition/PIC. Since
understanding PIC requires a prior understanding of the notion of phase developed in Chomsky (1998,
1999, 2001, 2002, 2004), let’s first take a brief look at what phases are.
Chomsky claims that the Language Faculty can only process limited amounts of syntactic structure
at one time – and, more specifically, can only hold a limited amount of syntactic structure in its ‘active
memory’ (Chomsky 1999, p.9). In order to ensure a ‘reduction of computational burden’ (1999, p.9)
Chomsky proposes that ‘the derivation of EXP[ressions] proceeds by phase’ (ibid.), so that syntactic
structures are built up one phase at a time. He maintains (1999, p.9) that phases are ‘propositional’ in
nature, and hence include CPs. His rationale for taking CP to be a phase is that CP represents a
complete clausal complex (including a specification of clause type – i.e. whether a clause is
declarative or interrogative etc.). He further proposes that syntactic operations are constrained by a UG
principle which can be paraphrased informally as follows:
(23) Phase Impenetrability Condition/PIC
10
Recall that capital letters are used to represent a word whose precise form has not been determined as yet (in
this case because BE hasn’t yet found an expression to agree with).
11
Although most clauses are CPs, some non-finite complement clauses (known as defective clauses) are simply
TPs which lack the CP layer found in canonical (i.e. ordinary) clauses. The complement of intend in the use
illustrated here is a defective clause of this kind, and hence a TP.
 Any expression which is below the head of a phase is impenetrable to (and so cannot agree
with or by attracted by) any constituent above the phase 12
Since CPs are phases, and since (in terms of a labelled bracketings above can be thought of informally
as meaning ‘to the left of’ and below ‘to the right of’), what (23) amounts to more informally is:
(24) Anything below (= to the right of) C is impenetrable to anything above (= to the left) of CP
Let’s see how PIC helps us account for the ungrammaticality of structures like (20b).
As before, let’s suppose that we have reached a stage of derivation at which we have formed the
infinitival TP this to happen. The infinitival complementiser for subsequently merges with this TP to
form the CP for this to happen. The (passive participle) verb intended then merges with this CP to
form the verb phrase/VP intended for this to happen, and this is subsequently merged with a T
constituent containing the auxiliary BE, so forming the TP in (25) below:
(25) [TP [T BE] [VP [V intended] [CP [C for] [TP this [T to] happen]]]]
As in the case of our earlier structure (21), the auxiliary BE searches for a noun/pronoun expression
which it can agree with and which it can attract to become its subject. The only noun/pronoun
expression in the structure is this. But the problem is that the Phase Impenetrability Condition/PIC
(23/24) makes this impenetrable (i.e. inaccessible) to BE in (25), for the following reason. Since all
CPs are phases, it follows that the CP for this to happen must be a phase. Now, the auxiliary BE is
positioned above/to the left of this CP phase (in the sense that BE occurs to the left of the CP for this to
happen), and the pronoun this is positioned below/to the right of the head C constituent for of the CP
phase: hence this is impenetrable to BE because PIC tells us that no item above/to the left of a phase
like CP can attract or agree with any expression below/to the right of a phase head like the
complementiser for. In other words, PIC prevents BE from agreeing with this and attracting this to
become its subject – and correctly predicts that (20b) *This is were intended for to happen is
ungrammatical.
The wider implication of our discussion of passivisation in sentences like (19b, 20b) is that the
Phase Impenetrability Condition/PIC constrains (i.e. places restrictions on) the operation of
movement and agreement, ensuring that both are local operations, in the sense that a head can only
agree with or attract a local (i.e. nearby) expression. PIC turns out to have far-reaching implications
for how (what would appear to be) long-distance movement operations work – as we can illustrate in
relation to the following sentence
(26) Where is it thought that he will go?
The locative adverb (i.e. adverb of place) where originates as the complement of go (as we see by
comparing (26) with the echo question ‘It is thought that he will go where?’), but subsequently
undergoes wh-movement and is thereby moved to the front of the overall sentence. Given our earlier
assumption that such movement comes about because an interrogative C constituent attracts an
interrogative wh-word to become its specifier, the derivation of (26) might at first sight seem to be
straightforward, and to proceed along the following lines. Suppose that (via a series of merger and
agreement operations) we have reached a stage of derivation where we have formed the TP It is
thought that he has gone where. This is then merged with an interrogative C constituent (containing a
null question particle Q) which attracts the present-tense auxiliary is to move into C, and which also
attracts the wh-word where to become its specifier. On this view, wh-movement applies in a single
step, as shown by the lower arrow in the structure below (the upper arrow showing T-to-C movement,
i.e. auxiliary inversion, and the lower arrow showing wh-movement):

(27) [CP Where [C is+Q] [TP it [T is] thought [CP [C that] [TP he [T will] go where]]]]

12
The relations above and below can be defined more formally in terms of the relation c-command in ways
which I will not explore here. Since I am using labelled bracketing to represent structure here, above can be
equated with ‘to the left of’ and below with ‘to the right of’.
However, closer analysis of (27) shows that (although movement of is from T to C is unproblematic),
the Phase Impenetrability Condition (23/24) blocks direct movement of where from being the
complement of go to becoming the specifier of the C constituent containing the inverted auxiliary is.
This is because (27) contains the complement clause CP that he will go where (this being the
complement of thought), and a CP is a phase. To see why this poses a problem, suppose that we have
reached the stage of derivation at which the TP it is thought that he will go where has been merged
with an interrogative C containing a question particle Q which attracts the auxiliary is to move from T
to C, so forming the structure shown in simplified form below:
(28) [CP [C is+Q] [TP it [T is] thought [CP [C that] [TP he [T will] go where]]]]
Since (26) is a wh-question, the main-clause C constituent containing the inverted auxiliary is requires
a wh-word as its specifier. But the Phase Impenetrability Condition (23/24) tells us that where is too
far away to be attracted by [C is], because the complement clause that he will go where is a CP and so a
phase, and [C is] is positioned to the left of the that-clause CP, and where is positioned to the right of
the head C constituent that of the CP. In other words, PIC makes where impenetrable to [C is], with the
result that [C is] cannot attract where to move directly to the front of the overall sentence and thereby
become the specifier of [C is].
The conclusion we reach is that where is ‘too far away’ from [C is] to be attracted by it: or, in
simpler terms, where cannot move directly to the front of the overall sentence in a single step in the
manner shown in (27) above. So how, then, does where get to the front of the sentence? An answer
which dates back to work in the 1970s is to suppose that any C constituent can attract a wh-word to
become its specifier. From this assumption, it follows that the complementiser that which heads the
complement-clause CP that he will go where can attract a wh-word like where. Let’s suppose that this
is the case. Suppose also that we have reached a stage of derivation at which we have formed the TP
he will go where, and that we now merge this with the C/complementiser that to form that he will go
where. Suppose too that a C constituent of any kind can attract a wh-word to become its specifier. This
means that the complementiser that introducing the complement clause can attract the wh-word where
to become its specifier, so triggering the wh-movement operation shown by the arrow in the simplified
structure below:
(29) [CP where [C that] he will go where]

The derivation now proceeds with the CP in (29) being merged with the (passive participle) verb
thought forming the verb phrase/VP thought where that he will go. This VP in turn merges with the
present-tense auxiliary BE forming BE thought where that he will go, and the resulting expression is
merged with the expletive subject it, which agrees with the T constituent BE and thereby forms the TP
it is thought where that he will go. This TP is subsequently merged with an interrogative C constituent
(containing an abstract question particle Q) which attracts the auxiliary is to move into C, so forming
the CP structure shown below:
(30) [CP [C is+Q] [TP it [T is] thought [CP where [C that] [TP he [T will] go where]]]]
But C in a wh-question also attracts a local (i.e. nearby) wh-word to become its specifier. The question
which now arises is whether where is close enough to [C is] for [C is] to be able to attract where to
become its specifier. This amounts to asking whether the Phase Impenetrability Condition/PIC (23/24)
makes where impenetrable to [C is]. Recall that PIC in effect says that anything to the right of C is
impenetrable to anything to the left of CP. [C is] occupies a position to the left of the CP where that he
will go, but where does not occupy a position to the right of the C that (on the contrary, where is
positioned to the left of that.) The bottom line is that PIC does not prevent [C is] from attracting where
to become its specifier, so that where can move from being the specifier of [C that] to becoming the
specifier of [C is]. This means that movement of where from being the complement of go to becoming
the specifier of [C is] takes place in two successive steps – as shown by the arrows in the simplified
structure in (31) below:
(31) [CP Where [C is+Q] [TP it [T is] thought [CP where [C that] [TP he [T will] go where]]]]
To use the conventional technical term, the Phase Impenetrability Condition (a principle of UG)
requires wh-movement to apply in a (multiple-step) successive-cyclic fashion in complex sentences
(i.e. sentences containing more than one CP), with the wh-word being attracted to become the specifier
first of the closest C above the wh-phrase, then of the next closest C … and so on. Since English
allows a complementiser like that to have a null spellout (under certain circumstances), alongside
that-structures like (31) above, we also find parallel structures like (32) below in which that has been
given a null spellout (symbolised as that):
(32) [CP Where [C is] [TP it [T is] thought [CP where [C that] [TP he [T will] go where]]]]
This shows us that functional heads like C (and also T) can sometimes be empty/null (i.e. not contain
any overtly spelled out item) under certain conditions.
The upshot of our discussion above is that (in consequence of the Phase Impenetrability
Condition/PIC), operations like wh-movement apply in a successive-cyclic (one-CP-at-a-time) fashion
in adult grammars. If PIC is a principle of UG genetically wired into the brain, we should expect to
find evidence that long-distance questions13 in child English involve successive-cyclic movement from
one spec-CP position to the next highest spec-CP position. In this connection, it is interesting to note
that Ros Thornton (1995) reports a boy called AJ (at age 5;4) producing wh-copying questions such as
the following:
(33) How much do you think how much the bad guy stole?
Here, the wh-expression how much originates as the complement of the verb stole. It first moves to the
bold-printed spec-CP position at the front of the complement clause in (34) below, and then from there
moves on to the italicised spec-CP position at the front of the main clause:
(34) [CP How much [C do+Q] you think [CP how much [C ø] the bad guy stole how much]]
The fact that an overt copy of the moved wh-expression how much appears at the front of both the
complement clause as well as at the front of the main clause provides evidence that the Copy Principle
(10) operates in child grammars. However, the fact that the child spells out the bold-printed copy of
how much overtly suggests that the child has not yet mastered the spellout conditions operating in
English and wrongly assumes it is possible to give an overt spellout to any copy of a wh-expression on
the edge of a CP. Interestingly, there are adult languages which seem to allow this kind of multiple
spellout of moved wh-expressions on the edge of CP, as the following examples illustrate:
(35)(a) Van watter mense het jy gedink van watter mense praat ek?
Of what people have you thought of what people talk I
‘Which people did you think I was talking about?’ (Afrikaans, cited in Hong 2004)
(b) Kas o Demiri mislenola kas i Arìfa dikhla?
Whom the Demir think whom the Arifa saw?
‘Whom does Demir think Arifa saw?’ (Romani, adapted from McDaniel 1989, p569, fn.5)
(c) Wen glaubst du, wen Peter meint, wen Susi heiratet?
Who believe you who Peter thinks who Susi marries?
‘Who do you believe Peter thinks that Susi is marrying?’ (German, Felser 2004, p.563)
As (35c) illustrates, in cases of long-distance wh-movement out of more than one complement clause,
a copy of a moved wh-pronoun appears at the beginning of each clause/CP. This suggests that spellout
conditions are parameterised – e.g. there is a Spellout Parameter which allows languages to differ
(e.g.) in respect of whether they allow single spellout of the highest copy in a wh-chain or multiple
spellout of any/every copy at the edge of a clause. However, of more direct relevance to our discussion
in this chapter is that child sentences like (33) provide evidence that UG principles like the Copy
Principle and the Phase Impenetrability Condition operate in child grammars (PIC forcing the
wh-expression how much to move one clause at a time, in order to avoid violation of PIC).

13
These are questions involving movement of a wh-expression out of a lower clause into a higher clause
 A rather different type of non-adult wh-question structure is reported in an experimental study
which Thornton and Crain (1994) conducted on 21 three- and four-year-old children. They acted out
short vignettes with toys in front of the children and then attempted to elicit wh-questions by prompts
such as ‘We know Cookie Monster likes someone. Ask Ratty who he thinks’ (the target question being
Who do you think Cookie Monster likes?). They report that in contexts where adults would produce a
long-distance wh-question like (36a) below, some of the children in their study produced so-called
partial movement wh-questions like (36b):
(36)(a) Where do you think this froggy lives? (= target adult structure)
(b) What do you think where this froggy lives? (= child’s structure)
In the adult structure (36a), the wh-adverb where (which originates as the complement of lives) first
moves into the bold-printed spec-CP position in the complement clause, and then moves into the
italicised spec-CP position in the main clause, as in (37) below:
(37) [CP Where [C do+Q] you think [CP where [C ø] this froggy lives where]]
Where must move one clause at a time, in order not to violate the Phase Impenetrability Condition
(23/24). In conformity with the adult English spellout condition (11), only the highest/leftmost copy of
where is overtly spelled out in the resulting adult sentence (36a) Where do you think this froggy lives?
But now consider what happens in the corresponding child structure (37). Here, the wh-adverb
where moves only as far as the spec-CP position at the front of the complement clause. But because
where has (semantic) scope over the main clause as well 14, the wh-word what is inserted in the
spec-CP position in the main clause as a scope-marker, resulting in the structure (38) below:
(38) [CP What [C do+Q] you think [CP where [C ø] this froggy lives where]]
Only the (bold-printed) copy of where on the edge of CP is overtly spelled out, as we would expect.
Thornton and Crain argue that partial movement structures like (38) provide evidence that
wh-movement applies in a successive-cyclic (one-clause-at-a-time) fashion in child grammars, since
children who produce such structures manage to perform only one of the two wh-movements
necessary to derive an adult long-distance question structure such as (37). And since it is the Phase
Impenetrability Condition which forces wh-expressions to move one clause at a time, we can conclude
that partial movement structures like (38) provide further evidence that PIC operates in child
grammars15.

1.6 The Agreement Principle

In our earlier discussion of passivisation in structures like (21) above, repeated below:
(39) [TP [T BE] [VP [V intended] [TP this [T to] happen]]]
we noted that the T constituent BE attracts the pronoun this to become its subject, ultimately deriving
the structure associated with (19b) This is intended to happen. However, it is interesting to note that
[T BE] also agrees with the very same pronoun this that it attracts to become its subject. Chomsky
(1998, 1999, 2001) argues that it is no coincidence that T attracts a constituent that it agrees with. On
the contrary, he maintains that it is in the very nature of attraction that a head like T attracts
(movement of) the closest constituent which it agrees with in respect of some grammatical feature/s:
e.g. in (39), BE attracts the pronoun this with which it agrees in person and number to become its
subject/specifier. We can capture Chomsky’s insight in terms of the following UG principle:
(40) Agreement Principle
A head attracts an expression it agrees with to become its specifier.
However, a principle like (40) has important implications for how wh-movement works. More
specifically, it means that C agrees with any wh-word which it attracts. Since a wh-word (such as

14
In the sense that the question asks where you think the frog lives, not where the frog actually does live.
15
Partial movement questions are also found in some adult languages – e.g. in German dialects in sentences such
as Was glaubt Hans mit wem Jakob jetz spricht? ‘What thinks Hans with whom Jakob now speaks?’ (i.e. ‘Who
does Hans think Jakob is now speaking to?’)
who/what/where/when) carries a wh-feature, it seems reasonable to suppose that a C which attracts a
wh-word agrees in respect of this wh-feature with the wh-word it attracts, and that C is thereby
wh-marked (i.e. assigned a wh-feature). In languages like Irish and Chamorro with a richer
morphology than English, the wh-marking of a complementiser which attracts a wh-word has an overt
morphological reflex, in the sense that the complementiser has a special form in wh-clauses (so that in
place of the equivalent of that we find the equivalent of what). Wh-marking of a complementiser
which attracts a wh-word is not overtly signalled in Standard English (where complementisers are
invariable and so that is either spelled out as that or given a null spellout irrespective of whether
wh-marked or not), but it is overtly marked in the kind of variety of British English (spoken e.g. in the
East End of London) which shows the that/what alternation illustrated below:
(41) SPEAKER A: I reckon that Den done it (= ‘I think that Dennis did it’)
SPEAKER B: No way – he ain’t got the bottle (= ‘No way – he hasn’t got the guts/courage’)
SPEAKER A: So who d’you reckon what done it? (= ‘So who do you think that did it?’)
Here, what seems to be a wh-marked variant of that. In the second utterance produced by speaker A,
the wh-word who originates as the subject of done it and is then attracted to become the specifier of
that (with that agreeing in wh-ness with who and thereby being wh-marked – this being marked by the
use of a WH subscript), resulting in the CP structure shown below 16:
(42) [CP who [C THATWH] [TP who [T ø] done it]
At a later stage of derivation, the main-clause C attracts who to become its specifier, ultimately
resulting in Who do you reckon what done it? (with the wh-marked complementiser being spelled out
as what). The fact that the complementiser that is spelled out as its wh-counterpart what provides us
with evidence that wh-movement in (38) involves a form of wh-agreement between C and the
wh-word it attracts – and, more generally, lends empirical support to the Agreement Principle
(specifying that a functional head agrees with an expression which it attracts to become its specifier).
But do we have any evidence that the Agreement Principle operates in child grammars? In this
connection, it is interesting to note that Thornton (1995) reports a boy called AJ (aged 5;4) producing
long-distance wh-questions like (43b) below in contexts where adults would use questions like (43a):
(43)(a) Which drink do you think that the ghost drank? (= adult question)
(b) Which drink do you think what the ghost drank? (= child question)
Why should AJ use what in a context where adults would use the complementiser that? We can
provide a principled answer to this question along the following lines.
Suppose we have reached a stage in the derivation of the sentences in (43) at which we have
merged the the finite declarative complementiser THAT with the expression the ghost drank which
drink to form the CP shown in skeletal form below:
(44) [CP [C THAT] the ghost drank which drink]
The complementiser THAT attracts the wh-expression which drink to become its specifier. Because the
Agreement Principle specifies that a head agrees with an expression which it attracts to become its
specifier, the complementiser agrees with the wh-expression which drink and thereby becomes
wh-marked (i.e. gets to carry a wh-feature, indicated by the use of a WH subscript below), so resulting
in the structure (45) below:
(45) [CP which drink [C THATWH] the ghost drank which drink]
The derivation then continues until the main clause CP is formed, with the main-clause C position
being filled by the auxiliary do and attracting a copy of which drink to move to the specifier position
within CP (and agreeing in wh-ness with it), so deriving the structure shown in simplified form below:
(46) [CP Which drink [C doWH+Q] you think [CP which drink [C THATWH] the ghost drank which drink]]
In standard varieties of adult English, a wh-feature on a constituent in C has a null spellout, so that
wh-agreement is invisible (and hence e.g. THAT is spelled out as that). But the child in question (AJ)
spells out the complementiser THAT as what when it is wh-marked resulting in (43b) Which drink do
16
The head T position of the TP contains no overt auxiliary, so is empty/null.
you think what the ghost drank? The use of the wh-marked form what here provides empirical
evidence that the Agreement Principle operates in child grammars.
In fact, some children seem to mark not only agreement in wh-ness between C and a wh-
expression which it attracts but also agreement in other features. In this connection, consider the
following wh-question reported by Thornton (1995, p.151) to have been produced by a girl called
Tiffany at age 4;9:
(47) Which Smurf do you think who has roller skates on?
Here, it appears that the complementiser THAT which typically introduces a clause used as the
complement of the verb THINK (e.g. in I think that this Smurf has roller skates on) ends up being
spelled out as who. How does this come about? It would appear that the complementiser THAT agrees
not only in wh-ness with which Smurf but also in (animate) gender, and hence is spelled out as the
animate form who (rather than the inanimate/default form what). Needless to say, sentences like (47)
provide further empirical evidence for the operation of the Agreement Principle in child grammars.

1.7 The Pied Piping Principle

Simple wh-questions in adult English like those below involve movement of an italicised
wh-word to the specifier position within CP:
(48)(a) Which have you chosen? (b) What will she wear?
In sentences like (48), the head C constituent of CP attracts a wh-word like which or what to move on
its own to the specifier position within CP. However, in sentences like those below, movement of a
wh-word on its own leads to ungrammaticality:
(49)(a) Which photo of Mary have you chosen? (b) *Which have you chosen photo of Mary?
(50)(a) What dress will she wear? (b) *What will she wear dress?
Sentences (49) and (50) illustrate a phenomenon termed pied-piping by Ross (1967). Ross observed
that when a wh-quantifier like which or what is used to modify a noun expression (such as the noun
phrase photo of her in 49a or the noun dress in 49b), the moved wh-word pied-pipes (i.e. carries along
with it) the expression which it modifies (in much the same way, as the Pied-Piper is reputed to have
induced the children of the rat-infested village of Hamelin to follow him out of the village by playing
his pipe). In order to see what triggers the pied-piping of modified material along with a preposed
wh-word, let’s look at why the noun phrase photo of Mary is pied-piped along with which in sentence
(49a) Which photo of Mary have you chosen?
Suppose we have reached a stage in the derivation of (49a) at which we have formed the clause
you have chosen which photo of Mary and merged it with an interrogative C containing a question
particle Q to form the structure shown in simplified form below (where we take which to be a
Quantifier/Q and which photo of Mary to be a Quantifier Phrase/QP):
(51) [CP [C Q] you have chosen [QP [Q which] photo of Mary]]
C attracts the present-tense auxiliary have to attach to the question particle Q. Of more relevance to
our concerns here is the fact that C also attracts a wh-word to move to the specifier position within CP.
Following Chomsky (1995, p.262), let us suppose that movement operations are governed by a
principle which we can outline informally as follows17:
(52) Pied-Piping Principle
A word which cannot move on its own pied-pipes along with it the minimum material required
to ensure convergence (i.e. to ensure that the resulting structure is well-formed)
In a structure like (51), C attracts the wh-word which to become its specifier. But because only a
maximal projection (i.e. a constituent which is the largest expression headed by a particular word) can
be a specifier and which in (51) is not a maximal projection (on the contrary, which is the head
quantifier of a larger quantifier phrase, and the maximal projection of which is the quantifier phrase
which photo of Mary), instead of the quantifier which moving on its own, the whole quantifier phrase
17
This principle is referred to in chapter 6 of Radford (2004a, 2004b) as the Convergence Principle
which picture of Mary must move to the specifier position within CP, so deriving the structure shown
in highly simplified form below:
(53) [CP Which photo of Mary [C have] you have chosen which photo of Mary]
Because the quantifier which cannot move on its own to spec-CP (because only a maximal projection
can occupy a specifier position), the Pied Piping Principle (52) tells that we have to move the whole
quantifier phrase which photo of Mary, and thereby pied-pipe the noun phrase photo of Mary along
with the moved wh-quantifier which.
If (as Chomsky suggests) pied-piping is governed by innate UG principles, we would expect to
find that children’s earliest wh-questions obey these principles – and hence that children will never
move which or what on its own without pied-piping any expression that it modifies. In order to test
whether this is indeed the case, I examined the first 150 of the Abe files from the Kuczaj corpus on the
CHILDES database (when Abe was aged between 2;5 and 3;11), looking for structures where
which/what modify a following noun expression. In the relevant files, Abe produced 36 interrogative
clauses like those in (54) below containing an (italicised) quantifier phrase headed by what, and 17
like those in (55) containing an (italicised) quantifier phrase headed by which:
(54) What kind is it? (2;5.0) What way you make candles? (2;10.3)
Those, what kind they are? (2;10.30) What number I’m gonna be on my birthday? (2;11.30)
What kind of fish are you talking about? (3;3.1) And what kind of bug was it? (3;6.16)
What else should I put away? (3;6.22) What channel is it on? (3;9.23)
We’ll have to look in the TV guide…and…and see what other day it’s on (3;10.15)
(55) Which book you wanta took over there? (2;10.22) Which way does this do? (2;11.25)
Which one do you think could fit? (3;1.11) Which of the plants we got? (3;2.7)
Could I show you which tree I cutted some of the bark off? (3;5.6)
Which snake did he say was in the United States? (3;8.22)
Which one do you want me to empty? (3;11.11)
All of the relevant structures showed obligatory pied-piping of the modified material along with the
preposed wh-word like what/which. This provides us with evidence that the Pied Piping Principle
operates in child grammars.

1.7 Whose questions

In the same way as which or what pied-pipe any noun expression they modify along with
them when they move, so too the possessive/genitive wh-pronoun whose pied-pipes a modified noun
expression along with it when it moves, as we see from contrasts such as:
(56)(a) Whose book have you borrowed? (b) *Whose have you borrowed book?
At first sight, it might seem as if pied-piping with whose should be treated in the same was as with
which/what. And yet whose-phrases are different in structure from which-/what-phrases, as we will
see.
Chomsky (1995, p.263) argues that a noun expression like whose book is a DP/Determiner Phrase
whose head is the affixal determiner ’s, whose specifier is the possessor who and whose complement is
the possessum (i.e. possessed noun expression) book: the fact that the wh-pronoun has the
orthographic (spelling) form whose rather than who’s is a mere quirk of English orthography (i.e. an
idiosyncrasy of the English spelling system), on this view. Given Chomsky’s analysis, sentence (56a)
will have the structure (57) below at the point where the clause you have borrowed whose book is
merged with an interrogative C (containing an abstract question affix Q):
(57) [CP [C Q] you have borrowed [DP who [D ’s] book]]
Given that only a maximal projection can be a specifier, the Pied-Piping Principle (52) in effect tells us
to move the smallest possible maximal projection containing the wh-word who. Under the analysis in
(60), the smallest such constituent is who, since who is a maximal projection by virtue of being the
largest expression headed by the word who. And yet, fronting who alone results in the structure (58)
below, which is ungrammatical in standard varieties of English:
(58) *[CP Who [C have+Q] you have borrowed [DP who [D ’s] book]]
Why is the resulting structure ungrammatical? Chomsky (1995, p.263) offers an answer along the
following lines. It is in the nature of affixes that they can only attach to an overt host, and the affix ’s
has the additional property that (in standard varieties of English) it can only attach to its own specifier
(i.e. to the possessor). Since its possessor specifier in (58) is a null copy of who and an affix requires
an overt host, the affix ’s is left stranded (with no overt host to attach to), so violating the Attachment
Constraint (18) and leading to ungrammaticality.
Since preposing of who on its own is not permitted in standard varieties, we might expect that (in
accordance with the Pied Piping Principle), the next largest expression containing who will be
preposed, namely the edge of the relevant projection comprising the specifier who and the head ’s.
However, movement of the string who’s (orthographically whose) also causes the derivation to crash –
as we see from the ungrammaticality of (56b) *Whose have you borrowed book? Why should this be?
The answer is that only a complete maximal projection can move to a specifier position, and who’s is
not a complete maximal projection. Rather, who’s is a subpart of the maximal projection who’s book
(which, under Chomsky’s analysis, is a DP which is the maximal projection of the determiner ’s).
Since movement of both who and who’s on their own are blocked, the Pied Piping Principle (52) tells
us to try and prepose the next largest expression containing who, namely the maximal projection (DP)
who’s book. This results in the structure shown in simplified form in (59) below:
(59) [CP Who’s book [C have+Q] you have borrowed who’s book]
and yields the grammatical outcome (56a) Whose book have you borrowed?
If (as Chomsky maintains) the syntax of whose-questions in adult English is determined by innate
principles of UG (the Attachment Constraint and the Pied Piping Principle), we should expect to find
that the same principle operating in whose questions in child English. In this connection, it is
interesting to consider results from an elicitation study of 12 children aged 4;5 to 6;0 by Gavruseva
and Thornton (2001). They report that 10 of the 12 children in their study produced pied-piping
questions like (60) below:
(60) Whose cat do you think Spiderman saved?
At first sight, this might seem to lead us to the straightforward conclusion that children’s whose
questions are constrained by the same UG principles that operate in adult grammars.
However, Gavruseva and Thornton report that 10 of the 12 children in their study also produced
split question structures like (61) below in which the wh-word who ends up in a different position
from the rest of the material in the DP containing it:
(61) Who do you think’s sunglasses Pocahontas tried on?
(= ‘Whose sunglasses do you think Pocahontas tried on?’)
At first sight, it might seem as if split question structures like (65) serve to empirically falsify the
claim that the syntax of children’s wh-possessor questions is constrained by UG principles. But closer
reflection suggests that such a conclusion may be hasty. Let’s see why.
Suppose we have reached a stage in the derivation of (61) at which we have formed the clause
Pocohontas tried on whose sunglasses and merged it with a null complementiser (a null counterpart of
that) to form the CP shown in simplified form below:
(62) [CP [C ø] Pocahontas tried on [DP who [D ’s ] sunglasses]]
The null complementiser triggers movement of the whole wh-expression who’s sunglasses, with the
result that this moves to the specifier position in CP, so deriving the structure:
(63) [CP [DP who [D ’s ] sunglasses] [C ø] Pocohontas tried on [DP who [D ’s ] sunglasses]]
The derivation then proceeds until we have formed the main-clause CP structure in (64) below
(simplified by showing only copies of wh-expressions, and abbreviated by using P for Pocahontas):
(64) [CP [C do+Q] you think [CP [DP who [D ’s ] sunglasses] [C ø] P tried on [DP who [D ’s ] sunglasses]]]
The C constituent introducing the main clause then attracts who to move on its own to become its
specifier, so deriving the structure shown below:
(65) [CP Who [C do+Q] you think [CP [DP who [D ’s ] sunglasses] [C ø] P tried on [DP who [D ’s ] sunglasses]]]
Deletion of all but the highest/leftmost copy of each constituent in accordance with the Spellout
Condition (11) then derives:
(66) [CP Who [C do+Q] you think [CP [DP who [D ’s ] sunglasses] [C ø] P tried on [DP who [D ’s ] sunglasses]]]
But (if UG principles operate in child grammars) wouldn’t we expect the resulting
derivation to crash, by virtue of violating the Attachment Constraint (18)? After
all, the affix ’s will remain unattached since it cannot attach to the pronoun who
immediately preceding it, because this is a null copy (and an affix can only attach
to an overt constituent). At this point, we might feel driven to the conclusion that
since children produce structures like (66) which seemingly contain an
unattached affix, UG principles like the Attachment Constraint simply don’t
operate in child grammars.
And yet, closer reflection suggests that there is an alternative (more plausible) way of acconting
for why children produce split who-possessives like (61) Who do you think’s sunglasses Pocahontas
tried on? Let us suppose that affixation works in a rather different way in child grammars than in adult
grammars. More specifically, let’s suppose that although for both adults and children, the affix must
attach to the closest overt word preceding it, adults only allow the affix to attach to its own specifier,
whereas children allow the affix to attach to whatever is the first overt word preceding it. In the light
of the different ways in which affixation works in adult and child grammars, let’s return to consider
what happens in the case of a structure like (66) above, repeated as (67) below:
(67) [CP Who [C do+Q] you think [CP [DP who [D ’s ] sunglasses] [C ø] P tried on [DP who [D ’s ] sunglasses]]]
In adult English the affix ’s can only attach to its own specifier, but since this is
the null pronoun who in (67) and an affix requires an overt host, a structure like
(67) is ungrammatical in adult English by virtue of violating the Attachment
Constraint (18). By contrast, children who allow the affix ’s to attach to any kind
of host will attach the affix to the verb think in (68), so providing a host for the
affix and conforming to the Attachment Constraint (18). Interestingly, Gavruseva
and Thornton report that 4 of the 12 adults in their study also produced split who-
questions like Who do you think’s food he didn’t eat? – underlining the view that
these are UG-constrained structures.

1.8 Summary
We began this chapter in §1.1 by noting that children generally start combining words
together to produce two-word utterances at around the age of 1;6, and that there is then a rapid,
uniform growth in the acquisition of grammar over the next 12 months which sees them acquire a
wide range of inflections and structures. In §1.2 we saw that Chomsky explains this rapid and uniform
growth by positing that the course of language acquisition is determined by a biologically endowed
innate Language Faculty which provides children with an algorithm for developing a grammar of any
language they are exposed to. In §1.3, we discussed Chomsky’s view that the Language Faculty
incorporates a set of Principles of Universal Grammar (= UG principles) which are innately wired into
the brain (and so do not have to be learned) and which constrain the range of syntactic structures and
syntactic operations found in natural (i.e. human) languages. We noted that if UG principles are wired
into the Language Faculty, we expect to find them operating in child grammars as well as adult
grammars; and in subsequent sections, we went on to present evidence that this is so. In §1.4 we saw
how the Copy Principle determines that moved constituents leave behind a copy which generally has a
null spellout, but which can be given an overt spellout if this is required for independent reasons. We
went on to show that children’s auxiliary-copying structures like Why could Snoopy couldn’t fit in the
boat? provide empirical evidence that the Copy Principle operates in child grammars (with the
righthand copy of could being overtly spelled out in order to provide a host for the clitic n’t). In §1.5
we noted that Chomsky’s Phase Impenetrability Condition/PIC (which specifies that anything below C
is impenetrable to anything above CP) forces wh-expressions to move one clause at a time in adult
long-distance wh-questions (first to the front of the lowest CP, then to the front of the next lowest CP,
and so on). We argued that children’s wh-copying questions like How much do you think how much
the bad guy stole? and their partial movement questions like What do you think where this froggy
lives? provide evidence that PIC also operates in child grammars (and forces wh-expressions to move
one clause at a time). In §1.6 we outlined the Agreement Principle (which specifies that a head can
only attract an expression it agrees with to become its specifier), and saw that this implies that C must
agree in wh-ness with any wh-expression which it attracts (with the consequence that the
complementiser THAT is spelled out as the wh-marked form what in non-standard adult varieties in
questions like Who d’you reckon what done it?). We went on to argue that children’s use of what in
wh-questions in which adults would use that (e.g. in questions like Which drink do you think what the
ghost drank?) provides empirical evidence that the Agreement Principle also operates in child
grammars. In §1.7 we saw how the Pied Piping Principle accounts for the noun phrase photo of Mary
being pied-piped along with the wh-quantifier which in an adult English sentence like Which photo of
Mary have you chosen? and we presented evidence that the principle also operates in child grammars.
In §1.8, we saw how the Attachment Constraint prevents movement of the possessor who on its own in
adult questions like *Who have you borrowed’s book? (since this would mean that the possessive affix
’s is not attached to a possessor host), and how the Pied Piping Principle accounts for why the whole
DP who’s book has to be moved (resulting in Whose book have you borrowed?). However, we noted
that children produce split questions like Who do you think’s sunglasses Pocahontas tried on? We
argued that such children allow the ’s affix to attach to whatever is the closest overt word preceding it
(in this case, the verb think) and hence that their structures satisfy the Attachment Constraint.

Workbook section
Exercise 1.1 Children’s questions
Below are examples of various types of non-adult-like questions produced by children (the examples
being taken from Stromswold 1990, Thornton and Crain 1994, Thornton 1995, and Gavruseva and
Thornton 2001). Discuss the derivation of each of the sentences, and whether they provide evidence
for the operation of UG principles in child grammars (and if so, which principles and why). Highlight
similarities and differences between each child sentence and its adult counterpart. Identify any
sentences which appear to pose a potential problem for the view that children’s syntax is
UG-constrained, discuss the nature of the problem and see if you can think of any way of overcoming
it. Where the name/initials and age of the child producing the relevant sentences are reported in the
relevant studies, they are shown in parentheses; where not, a reference is included to the article
reporting the relevant sentence.
1. Is the clock is working? (Shem, 2;5)
2. Do she don’t need that one? (Adam 3;6)
3. Which chairs did the queen didn’t sit in? (AJ 5;4)
4. Which Smurf the big man couldn’t pick up? (AJ 5;4)
5. Which dinosaur that Grover didn’t ride on? (AJ 5;4)
6. Which mouse what the cat didn’t see? (AJ 5;4)
7. What do you think that’s under these? (AJ 5;4)
8. Who do you think what’s under there? (AJ 5;4)
9. Which animal do you think what really says “woof woof” (Tiffany 4;9)
10. Who do you think who Cookie Monster likes? (Thornton and Crain, 1994, p.218)
11. Which mouse do you think who the cat chased? (AJ 5;4)
12. How do you think how Cookie Monster got this cookie? (AJ 5;4)
13. What do you think what the mouse ate? (AJ 5;4)
14. What do you think who ate the cheese? (AJ 5;4)
15. What do you think which animal says “woof woof”? (Tiffany 4;9)
16. What do you think how the boy paddled the boat? (AJ 5;4)
17. Whose ears did she didn’t tickle? (Kathy 4;6)18
18. Whose do you think lunch the baboon made? (Tori 5;8)
19. Who do you think’s Spiderman saved cat? (Gab 5;11)
20. Whose do you think’s Gonzo rubbed down horse? (Gab 5;11)
As in the main text, use simplified labelled bracketings to represent structure, and show only the
structure of the CP constituents in each clause, along with the internal structure of any wh-phrase (like
which mouse or whose lunch) that the sentence contains.

2 Parameters and parameter-setting

2.1 Introduction
In the previous chapter, we looked at Chomsky’s claim that human beings are genetically
endowed with an innate Language Faculty which incorporates a set of universal principles which
guide the child in constructing a grammar. However, it clearly cannot be the case that all aspects of the
grammar of languages are universal; if this were so, all natural language grammars would be the same
and there would be no grammatical learning involved in language acquisition (i.e. no need for
children to learn anything about the grammar of the language they are acquiring), only lexical
learning (viz. learning the lexical items/words in the language and their idiosyncratic linguistic
properties, e.g. whether a given item has an irregular plural or past tense form). But although (as we
saw in the previous chapter) there are universal principles which operate in the grammars of natural
languages, there also seem to be language-particular aspects of grammar which children have to learn
18
The adult sentences corresponding to 17-20 would be: 17 Whose ears didn’t she tickle? 18/20 Whose cat do
you think Spiderman saved? 19 Whose lunch do you think the baboon made?
as part of the task of acquiring their native language. Thus, language acquisition involves not only
lexical learning but also grammatical learning. In this chapter, we take a closer look at the nature of the
grammatical learning involved, and what it tells us about the language acquisition process.

2.2 Parameters and the Head Position Parameter

Clearly, grammatical learning is not going to involve learning those aspects of grammar
which are determined by universal (hence innate) principles. Rather, grammatical learning will be
limited to those parameters (i.e. dimensions or aspects) of grammar which are subject to language-
particular variation (and hence vary from one language to another). In other words, grammatical
learning will be limited to parametrised aspects of grammar (i.e. those aspects of grammar which are
subject to parametric variation from one language to another). One way to determine what children
have to learn about grammar is to look at how the grammars of different adult languages (and
language varieties) vary.
Perhaps the most immediately obvious way in which languages differ from each other is in respect
of certain aspects of word order. For example, there is considerable word order variation between
languages with respect to the relative position of heads and complements within phrases. A principle
of UG (the Headedness Principle) requires that every phrase must have a head word which
determines the nature of the overall phrase. For example, an expression such as students of Philosophy
is a plural noun phrase because its head word (i.e. the key word in the phrase whose nature determines
the properties of the overall phrase) is the plural noun students: the noun students (and not the noun
Philosophy) is the head word because the phrase students of Philosophy denotes kinds of student, not
kinds of Philosophy. The following expression of Philosophy which combines with the head noun
students to form the noun phrase students of Philosophy functions as the complement of the noun
students. In much the same way, an expression such as in the kitchen is a prepositional phrase which
comprises the head preposition in and its complement the kitchen. Likewise, an expression such as
stay with me is a verb phrase which comprises the head verb stay and its complement with me. And
similarly, an expression such as fond of fast food is an adjectival phrase formed by merging the head
adjective fond with its complement of fast food.
In English all heads (whether nouns, verbs, prepositions, or adjectives etc.) normally precede their
complements; however, there are also languages like Korean in which all heads normally follow their
complements. In informal terms, we can say that English is a head-initial (or head-first) language,
whereas Korean is a head-final (or head-last) language. The differences between the two types of
language can be illustrated by comparing the English examples in (1) below with their Korean
counterparts in (2):
(1)(a) Close the door (b) desire for change
(2)(a) Muneul dadara (b) byunhwa-edaehan galmang
Door close change-for desire
In the English verb phrase close the door in (1a), the head verb close precedes its complement the
door; if we suppose that the door is a determiner phrase, then the head of the phrase (= the determiner
the) precedes its complement (= the noun door). Likewise, in the English noun phrase desire for
change in (1b), the head noun desire precedes its complement for change; the complement for change
is in turn a prepositional phrase in which the head preposition for likewise precedes its complement
change. Since English consistently positions heads before complements, it is a head-first language. By
contrast, we find precisely the opposite ordering in Korean. In the verb phrase muneul dadara
(literally ‘door close’) in (2a), the head verb dadara ‘close’ follows its complement muneul ‘door’;
likewise, in the noun phrase byunhwa-edaehan galmang (literally ‘change-for desire’) in (2b) the head
noun galmang ‘desire’ follows its complement byunhwa-edaehan ‘change-for’; the expression
byunhwa-edaehan ‘change-for’ is in turn a prepositional phrase whose head preposition edaehan
‘for/about’ follows its complement byunhwa ‘change’ (so that edaehan might more appropriately be
called a postposition; prepositions and postpositions are differents kinds of adposition). Since Korean
consistently positions heads after their complements, it is a head-final language. Given that English is
head-initial and Korean head-final, it is clear that the relative positioning of heads with respect to their
complements is one word-order parameter along which languages differ; the relevant parameter is
termed the Head Position Parameter.
It should be noted, however, that word-order variation in respect of the relative positioning of heads
and complements falls within narrowly circumscribed limits. There are many logically possible types
of word order variation which just don’t seem to occur in natural languages. For example, we might
imagine that in a given language some verbs would precede and others follow their complements, so
that (e.g.) if two new hypothetical verbs like scrunge and plurg were coined in English, then scrunge
might take a following complement, and plurg a preceding complement. And yet, this doesn’t ever
seem to happen: rather all verbs typically occupy the same position in a given language with respect to
a given type of complement19.
What this suggests is that there are constraints (i.e. restrictions) on the range of parametric variation
found across languages in respect of the relative ordering of heads and complements. It would seem as
if there are only two different possibilities which UG allows for: a given type of structure in a given
language must either be head-initial (with relevant heads positioned before their complements), or
head-final (with relevant heads positioned after their complements). Many other logically possible
orderings of heads with respect to complements appear not to be found in natural language grammars.
The obvious question to ask is why this should be. The answer given by Parameter Theory is that the
language faculty imposes genetic constraints on the range of parametric variation permitted in natural
language grammars. In the case of the Head Position Parameter (i.e. the parameter which determines
the relative positioning of heads with respect to their complements), the language faculty allows only a
binary set of possibilities – namely that a given kind of structure in a given language is either
consistently head-initial or consistently head-final.

2.3 Head Strength Parameter

Chomsky (1995) argues that much of the parametric variation in languages is associated
with properties of functional heads20. One such type of variation concerns whether a given type of
functional head is strong or weak. We can illustrate what it means for a head to be strong in relation to
the syntax of the bracketed TP constituent in the (colloquial) French sentences below:

(3)(a) Je crois [CP [C qu’] [TP ils [T ont] pas [VP [V compris] les instructions]]]
I think that they have not understood the instructions
(b) Je crois [CP [C qu’] [TP ils [T comprennent] pas [VP [V comprennent] les instructions]]]
I think that they understand not understand the instructions
A finite T is strong in French, and this means that the head T position of TP in a finite clause can either
be filled by merging (= directly positioning) an auxiliary like ont ‘have’ in T as in (3a), or by
attracting a main verb like comprennent ‘understand’ to move from the head V position of VP into the
head T position of CP, as in (3b). By contrast, a finite T is weak in present-day English, and this means
that T can only be filled by directly merging an auxiliary in T, not by attracting a main verb to move
out of V into T. Hence, the English counterpart of (3b) is (4a) below, not (4b):
(4)(a) I think [CP [C that] [TP they [T do] not [VP [V understand] the instructions]]
(b) *I think [CP [C that] [TP they [T understand] not [VP [V understand] the instructions]]
Generalising at this point, we can say that there is parametric variation between French and English in

19
A complication overlooked here in the interest of expository simplicity is that there are languages which
position some types of head before their complements, and other types of head after their complements: for
example, German positions verbal heads after their complements (as shown by the fact that the verb is positioned
at the end of subordinate clauses in German) but other types of head before their complement. Complications
arise relating to movement of verbs in main clauses in German, but I will not go into these here.
20
Functional heads are morphemes which serve as the head of a phrase, and which serve to mark purely
grammatical properties. So, C is a functional head because it is the head of CP and serves to mark clause type.
Likewise, T is a functional category because it is the head of TP and serves to mark tense.
relation to whether a finite T is a strong head or not 21. Since a strong T can trigger verb movement (i.e.
movement of a main/lexical verb from V to T), the relevant Head Strength Parameter is also known
as the Verb Movement Parameter.
In much the same way, we find parametric variation with respect to whether or not the functional
head C (which marks clause type) is strong or weak. In this respect, it is instructive to compare the
derivation of the colloquial Standard English/SE wh-question in (5a) below with that of its African
American English/AAE counterpart in (5b):
(5)(a) Where am I gonna go? (b) Where I’m gonna go?
Suppose that we reach a stage in the derivation of these sentences at which we have formed the TP
shown in highly simplified form in (6) below:
(6) [TP I [T am] gonna go where]
The next stage of derivation in Standard English/SE is to merge the TP in (6) with an interrogative C
containing an abstract question particle Q which attracts the auxiliary am to move from T into the head
C position of CP, and which also attracts the wh-word where to move into the specifier position within
CP, so deriving the structure shown in simplified form below (with arrows showing the movements
which take place):

(7) [CP Where [C am+Q] [TP I [T am] gonna go where]]

The fact that C attracts the auxiliary am to move from T into C means that C is strong in main-clause
questions in English. Within a structure like (7), the inverted auxiliary am occupies the head C position
of CP, the wh-expression where is the specifier of C, and the TP I am gonna go where is the
complement of C.
But now consider what happens when we reach the stage of derivation shown in (6) above in
African American English/AAE. As before, the TP in (6) is merged with an interrogative C containing
an abstract question particle. But C in wh-questions in AAE is weak, with the result that C does not
attract the auxiliary am in T to move to C, so that C simply contains the null question particle Q and
the auxiliary am remains in situ in T. However, C (like its counterpart in SE) attracts the wh-word
where to become its specifier, so deriving the structure shown in highly simplified form below (with
the arrow showing the movement that takes place):
(8) [CP Where [C Q] [TP I [T am] gonna go where]]

Subsequently the auxiliary am cliticises onto its subject I to form Where I’m gonna go?
What our discussion in this section suggests is that languages differ in respect of whether or not a
given type of functional head is strong or weak. This in turn leads us to the conclusion that there is a
Head Strength Parameter/HSP which determines whether a given type of functional head in a given
language is strong (and hence can attract a particular kind of head), or weak. Thus, T is strong in finite
clauses in French, but weak in English; and C is strong in main-clause wh-questions in Standard
English, but weak in the corresponding structure in African American English. What is also interesting
is that the Head Strength Parameter (like the Head Position Parameter) appears to be binary in
nature, in the sense that it allows for only two possibilities: e.g. C is either strong in a given type of
clause and hence attracts any item of the relevant type (i.e. any present or past tense auxiliary), or C is
weak and attracts no item of the relevant type. So, for example, we find no variety of English in which
C can attract will but not would, or can but not could. It would seem that the range of parametric

21
T can be argued to contain a strong tense/agreement affix requiring a host verb to attach to, and this
requirement can be satisfied in French either by directly merging an auxiliary with the affix in T, or by moving a
main/lexical verb from V to T to adjoin to the affix. By contrast, T in English contains a weak tense/agreement
affix, and this can be attached to a verbal host either by directly merging an auxiliary with the affix in T, or by
lowering the affix onto the main verb via the morphological operation of Affix Hopping.
variation found with respect to HSP is limited to just two possibilities: viz. a given type of head (like
C) either does or doesn’t attract a particular other type of head (e.g. a present/past tense auxiliary).

2.4 Wh-Parameter
Our brief discussion of wh-questions in the previous section highlighted one particular way
in which the syntax of questions varies in a parametric fashion from one type of language to another.
But if we compare wh-questions in English and Chinese, a further pattern of parametric variation
emerges – as we see by comparing the two sentences below:
(9)(a) What do you think he will say?
(b) Ni xiangxin ta hui shuo shenme
You think he will say what?
In wh-questions in English, the italicised wh-expression is moved to the beginning of the sentence
(attracted by the head C constituent of CP as we saw in the previous section), as is the case with what
in (9a). By contrast, in Chinese, the italicised wh-expression does not move to the front of the
sentence, but rather remains in situ (i.e. in the same place as would be occupied by a corresponding
non-interrogative expression), so that shenme ‘what’ is positioned after the verb shuo ‘say’ because it
is the (direct object) complement of the verb, and complements of the relevant type are normally
positioned after their verbs in Chinese. Thus, another parameter of variation between languages is the
wh-parameter – a parameter which determines whether wh-expressions are moved to the front of the
overall interrogative structure containing them (resulting in a wh-initial structure like (9a) above) or
whether they remain in situ (resulting in a wh-in-situ structure like (9b) above). Interestingly, the
wh-parameter also appears to be binary in nature, in that it allows for only two possibilities – viz. a
language either does or doesn’t allow wh-movement (i.e. movement of wh-expressions to the front of
the sentence). Many other possibilities for wh-movement just don’t seem to occur in natural
languages: for example, there is no language in which the counterpart of who undergoes wh-fronting
but not the counterpart of what (e.g. no language in which it is OK to say Who did you see? but not
What did you see?). Likewise, there is no language in which wh-complements of some verbs can
undergo fronting, but not wh-complements of other verbs (e.g. no language in which it is OK to say
What did he drink? but not What did he eat?). It would seem that the range of parametric variation
found with respect to wh-fronting is limited to just two possibilities: viz. a language either does or
doesn’t allow wh-expressions to be systematically fronted 22.
But what kind of property does C have which enables it to attract a wh-expression to become its
specifier in English, but not in Chinese? In recent work, Chomsky (1998, 1999, 2001) has suggested
that C in wh-clauses in English has (what he calls) an EPP feature which requires C to have an
Extended Phrasal Projection (hence EPP) into a CP containing wh-expression as its specifier. On this
view, Chinese and English would differ in that C has a (wh-attracting) EPP feature in wh-questions in
English, but not in Chinese.

2.5 Null Subject Parameter

So far, all the parameters we have looked at have involved word-order. A different kind of
parameter which we shall look at in this section is the Null Subject Parameter – a parameter which
determines whether subject pronouns can be null (i.e. ‘silent’ or ‘unpronounced’) in finite clauses.
Before looking at this parameter, however, it should be pointed out that
there is one type of finite clause in English which allows a particular kind of null
subject – namely imperative clauses: these are clauses used to issue an order or
22
However, it should be noted that a number of complications are overlooked here in the interest of simplifying
exposition: e.g. some languages like English allow only one wh-expression to be fronted in this way, whereas
others allow more than one wh-expression to be fronted; see Bošković (2002) for a recent account. One way of
accounting for this is to suppose that two different subparameters are involved, one determining whether there is
wh-movement or not in a given language, and another determining whether (in a language with wh-movement)
single or multiple wh-expressions are preposed. An additional complication is posed by the fact that
wh-movement appears to be optional in some languages, either in main clauses, or in main and complement
clauses alike: see Denham 2000, and Cheng and Rooryck 2000.
request, and they have the property that they can generally be introduced by
please. As the examples below illustrate, an imperative sentence in English can
have an overt subject which is either a second person expression like you, or a
third person expression like anyone:
(10)(a) Don’t you lose your nerve! (b) Don’t anyone lose their
nerve!
A second person (you) subject can have a null spellout (i.e. can be
unpronounced) in imperatives in English, but not a third person subject like
anyone, as the contrast in (11) below shows (where strikethrough indicates that
the relevant subject pronoun is ‘silent’):
(11)(a) Don’t you lose your nerve! (b) *Don’t anyone lose
their nerve!
One way of describing this is to say that English is a language which allows a null
second-person imperative subject (i.e. allows a you subject to be
unpronounced in imperatives).
By contrast, English does not generally allow subject pronouns to be null (i.e. ‘omitted’) in other
types of finite clause, as we see from the contrast below:
(12)(a) Was she sure that they had finished? (b) *Was sure that had finished?
In (12a), she is the subject of an interrogative (= question-asking) finite main clause (containing the
finite past-tense copular verb was) and cannot be null, while they is the subject of a declarative
(= statement-making) complement clause (containing the finite past-tense auxiliary verb has) and
likewise cannot be null. Interestingly, however, the corresponding sentence in Italian allows the
subject of both clauses to be null, as we see from (13) below:
(13) Era sicura che avevano finito?
Was sure that had finished? (= ‘Was she sure that they had finished?’)
As these examples illustrate, the two languages differ in that in finite (declarative or interrogative)
clauses where English uses a nominative subject pronoun like she/they, Italian drops the subject
pronoun (so that Italian is traditionally said to be a pro-drop language). However, it is clear that we
cannot say that the two clauses in (13) are simply subjectless and don’t have a subject of any kind. On
the contrary, the main clause must have a null counterpart of she as its subject since verbs and
adjectives agree with their subjects in Italian and the verb era ‘is’ is a third person singular form and
the adjective sicura ‘sure’ is a feminine singular form. Likewise, the che ‘that’ clause must have a null
counterpart of they as its subject if we are to account for why the auxiliary avevano ‘had’ is a third
person plural form. The more general conclusion to be drawn from sentences like (13) is that
languages like Italian allow any finite (auxiliary or main) verb in any kind of clause (whether
interrogative or declarative, main clause or complement clause, for example) to have a null subject
pronoun which can be thought of as a silent or invisible counterpart of nominative personal pronouns
like I/we/you/he/she/it/they in English. This null subject pronoun is conventionally designated as pro,
so that (13) has the structure shown in highly simplified form below:
(14) pro3FSG era sicura che pro3PL avevano finito?
pro3FSG was sure that pro3PL had finished?’
where the first occurrence of pro is a null third person feminine singular nominative pronoun (i.e. a
silent counterpart of English she), and the second is a null third person plural nominative pronoun (i.e.
a null counterpart of English they). Note that it isn’t the case that Italian has no nominative subject
pronouns at all: rather, nominative pronouns in Italian are strong pronouns which mark emphasis or
contrast – as in the sentence below (where italics mark contrastive stress):
(15) Lui parla francese, però lei no
He speaks French, but she no (= ‘He speaks French, but she doesn’t’)
By contrast, nominative pronouns in English can serve as both strong (emphatic/contrastive) and
weak23 (unemphatic/non-contrastive) pronouns, with some phonological differences between the two:
for example, in standard varieties of English he can lose its h and be reduced to ’e when used as a
weak pronoun (as in The trouble with Jim is that e’s lazy), but not (in my variety of English) when
used as a strong pronoun (e.g. when used contrastively in ‘HE speaks French but SHE doesn’t’, where
capitals mark contrastive stress).
We can describe the differences between Italian and English illustrated above by saying that Italian
is a null subject language/NSL (i.e. the kind of language which allows any kind of finite verb in any
kind of clause to have a null subject) whereas English is a non-null subject language/non-NSL. More
generally, there appears to be parametric variation between languages as to whether they use a null
pronoun in contexts were English would use a weak nominative subject pronoun. Since pro in Italian
occurs in finite clauses where adult English would use a weak nominative pronoun, we can refer to
pro as a null nominative pronoun, or finite null subject. The relevant parameter (termed the Null
Subject Parameter) would appear to be a binary one, with only two possible settings for any
language, viz. A given language L either does or doesn’t allow a weak nominative pronoun in a finite
clause to have a null spellout (i.e. to be ‘silent’)24.
An obvious question to ask is why finite (main and auxiliary) verbs should allow a null pro subject
in Italian but not in English. A traditional answer is that finite (auxiliary and main) verbs carry a rich
enough set of agreement inflections to enable identification of the person/number properties of a null
subject in Italian, but not in English (because English has an impoverished inflectional morphology).
For example, since had in (12) does not overtly inflect for person or number in English, if we omit the
subject of had, there will be no morphological clues to help us identify whether the missing subject is
I/we/you/he/she/it/they. By contrast, since its Italian counterpart in (13) overtly inflects for person and
number, the agreement morphology on the verb helps us identify the ‘silent’ subject of avevo as a null
1.SG (first person singular) pronoun, of avevi as a 2.SG subject, of aveva as a 3.SG subject, of avevamo
as a 1.PL subject, or avevate as a 2.PL subject, and of avevano as a 3.PL subject (and all other finite
auxiliaries and main verbs carry a similarly rich set of agreement inflections in Italian). If this is so,
then the Null Subject Parameter reduces to a morphological property – viz. whether finite verbs carry
rich enough agreement morphology to identify the person/number properties of a null subject 25.
We can generalise our discussion of parameters thus far in the following terms. If the Head
Position Parameter, the Head Strength Parameter, the Wh-Parameter and the Null Subject
Parameter all allow a binary choice of settings, it seems implausible that binarity could be an
accidental property of these particular parameters. Rather, it seems much more likely that it is an
inherent property of parameters that they constrain the range of structural variation between languages,
and limit it to a simple binary choice. Generalising still further, it seems possible that all grammatical
variation between languages can be characterised in terms of a set of parameters, and that for each
parameter, the Language Faculty specifies a binary choice of possible values for the parameter.

2.6 How parameters are set

The theory of parameters outlined in the previous section has important implications for a
theory of language acquisition. If all grammatical variation can be characterised in terms of a series of
parameters with binary settings, it follows that the only grammatical learning which children have to
undertake in relation to the syntactic properties of the relevant class of constructions is to determine
(on the basis of their linguistic experience) which of the two alternative settings for each parameter is
23
As should be obvious, the terms strong/weak are used here in an entirely different sense (relating to whether a
pronoun is stressed/emphatic/contrastive in use or not) from that employed in relation to movement (where a
functional head is said to be strong if it has the ability to attract another head).
24
A complication glossed over here is posed by languages in which only some types of finite clause can have
null subjects, or only some types of subject pronoun can be null: see Vainikka and Levy 1999 and the collection
of papers in Jaeggli and Safir 1989 for illustration and discussion.
25
The claim that null subjects are morphologically identified might seem to be called into question by the fact
that languages like Chinese (in which verbs carry no agreement morphology whatever) nonetheless allow null
subjects. However, a crucial difference between Italian and Chinese is that Italian only allows subjects (not
objects) to be null, whereas Chinese allows both subjects and objects to be null when they are topics (i.e. when
they denote entities familiar from the discourse context). So, more properly, languages like Chinese can be said
to be topic-drop (or null topic) languages rather than subject-drop (or null subject) languages.
the appropriate one for the language being acquired. So, for example, children have to learn whether
the native language they are acquiring is a head-initial or head-final language, whether it is a language
in which a finite T is strong or weak, whether a finite T does or doesn’t allow a null subject, whether C
is strong (and so can attract a tense-marked auxiliary) or weak, whether C (has an EPP feature which)
can attract a wh-expression or not...and so on for all the other parameters along which languages vary.
Of course, children also face the formidable task of lexical learning – i.e. building up their vocabulary
in the relevant language, learning what words mean and what range of forms they have (e.g. whether
they are regular or irregular in respect of their morphology), what kinds of structures they can be used
in and so on. On this view, the acquisition of grammar involves the twin tasks of lexical learning and
parameter-setting.
This leads us to the following view of the language acquisition process. The central task which the
child faces in acquiring a language is to construct a grammar of the language. The innate Language
Faculty incorporates (i) a set of universal grammatical principles, and (ii) a set of grammatical
parameters which impose severe constraints on the range of grammatical variation permitted in natural
languages (perhaps limiting variation to binary choices). Since universal principles don’t have to be
learned, the child’s syntactic learning task is limited to that of parameter-setting (i.e. determining an
appropriate setting for each of the relevant grammatical parameters). For obvious reasons, the theory
outlined here (developed by Chomsky in work dating back to Chomsky 1981) is known as Principles-
and-Parameters Theory/PPT.
The PPT model clearly has important implications for the nature of the language acquisition
process, since it vastly reduces the complexity of the acquisition task which children face. PPT
hypothesises that grammatical properties which are determined by principles of Universal Grammar
will not have to be learned by the child, since UG principles are wired into the Language Faculty and
hence part of the child’s genetic endowment: on the contrary, all the child has to learn about grammar
are those grammatical properties which are subject to parametric variation across languages.
Moreover, the child’s learning task will be further simplified if it turns out (as research since 1980 has
suggested) that the values which a parameter can have fall within a narrowly specified range, perhaps
characterisable in terms of a series of binary choices.
However, an important question posed by the parameter-setting model of acquisition is just how
children come to arrive at the appropriate setting for a given parameter, and what kind(s) of evidence
they make use of in setting parameters. As Chomsky notes (1981, pp. 8-9), there are two types of
evidence which we might expect to be available to the language learner in principle, namely positive
evidence and negative evidence. Positive evidence comprises a set of observed expressions
illustrating a particular phenomenon: for example, if a child hears a phrase like with Daddy and
‘knows’ that with is a preposition and Daddy is its complement, the phrase in question provides the
child with positive evidence that prepositions precede their complements in English (i.e. that English
has head-initial word order in prepositional phrases). By contrast, negative evidence might come from
the correction of children’s errors by other speakers of the language. However, (contrary to what is
often imagined) correction plays a fairly insignificant role in language acquisition, for two reasons.
Firstly, correction is relatively infrequent: adults simply don’t correct all the errors children make (if
they did, children would soon become inhibited and discouraged from speaking). Secondly, children
are notoriously unresponsive to correction, as the following dialogue (from McNeill 1966, p. 69)
illustrates:
(16) CHILD: Nobody don’t like me
ADULT: No, say: ‘Nobody likes me’
CHILD: Nobody don’t like me
(8 repetitions of this dialogue)
ADULT: No, now listen carefully. Say ‘Nobody likes me’
CHILD: Oh, nobody don’t likes me

As Hyams (1986, p.91) notes: ‘Negative evidence in the form of parental disapproval or overt
corrections has no discernible effect on the child’s developing syntactic ability.’ (See McNeill 1966,
Brown, Cazden and Bellugi 1968, Brown and Hanlon 1970, Braine 1971, Bowerman 1988, Morgan
and Travis 1989, and Marcus 1993 for further evidence in support of this conclusion.)
Rather than say that children rely on direct negative evidence, we might instead imagine that they
learn from indirect negative evidence (i.e. evidence relating to the non-occurrence of certain types of
structure). Suppose that a child’s experience includes no examples of structures in which heads follow
their complements (e.g. no prepositional phrases like *dinner after in which the head preposition after
follows its complement dinner, and no verb phrases such as *cake eat in which the head verb eat
follows its complement cake). On the basis of such indirect negative evidence (i.e. evidence based on
the non-occurrence of head-final structures), the child might infer that English is not a language which
allows head-final word order.
Although it might seem natural to suppose that indirect negative evidence plays some role in the
acquisition process, there are potential learnability problems posed by any such claim. After all, the
fact that a given construction does not occur in a given chunk of the child’s experience does not
provide conclusive evidence that the structure is ungrammatical, since it may well be that the non-
occurrence of the relevant structure in the relevant chunk of experience is an accidental (rather than a
systematic) gap. Thus, the child would need to process a very large (in principle, infinite) chunk of
experience in order to be sure that non-occurrence reflects ungrammaticality. It seems implausible to
suppose that children store massive chunks of experience in this way and search through it for
negative evidence about the non-occurrence of certain types of structure. In any case, given the
assumption that parameters are binary and single-valued, negative evidence becomes entirely
unnecessary: after all, once the child hears a prepositional phrase like with Daddy in which the head
preposition with precedes its complement Daddy, the child will have positive evidence that English
allows head-initial order in prepositional phrases; and given the assumptions that the Head Position
Parameter is a binary one and that each parameter allows only a single setting, then it follows (as a
matter of logical necessity) that if English allows head-initial prepositional phrases, it will not allow
head-final prepositional phrases. Thus, in order for the child to know that English doesn’t allow head-
final prepositional phrases, the child does not need negative evidence from the non-occurrence of such
structures, but rather can rely on positive evidence from the occurrence of the converse order in head-
initial structures (on the assumption that if a given structure is head-initial, UG specifies that it cannot
be head-final). And, as we have already noted, a minimal amount of positive evidence is required in
order to identify English as a uniformly head-initial language (i.e. a language in which all heads
precede their complements). Learnability considerations such as these have led Chomsky (1986, p.55)
to conclude that ‘There is good reason to believe that children learn language from positive evidence
only.’ The claim that children do not make use of negative evidence in setting parameters is known as
the No-Negative-Evidence Hypothesis; it is a hypothesis which is widely assumed in current
acquisition research.

2.7 Parameter setting in early child grammars

The assumption that acquiring the grammar of a language involves the relatively simple
task of using positive evidence to set a number of (binary) grammatical parameters provides a way of
accounting for the fact that the acquisition of specific parameters appears to be a remarkably rapid and
error-free process in young children. For example, young children acquiring English as their native
language seem to set the Head Position Parameter at its appropriate head-initial setting from the very
earliest multiword utterances they produce (at around age 1;6), and seem to know (tacitly, not
explicitly, of course) that English is a head-initial language. Accordingly, the earliest verb phrases and
prepositional phrases produced by young children acquiring English consistently show verbs and
prepositions positioned before their complements, as structures such as the following indicate
(produced by a young boy called Jem (= James) at age 1;8; head verbs are italicised in (17a) and head
prepositions in (17b), and their complements are in non-italic print):
(17)(a) Touch heads. Cuddle book. Want crayons. Want malteser. Open door. Want biscuit.
Bang bottom. See cats. Sit down
(b) On Mummy. To lady. Without shoe. With potty. In keyhole. In school. On carpet.
On box. With crayons. To mummy
The obvious conclusion to be drawn from structures like those in (17) is that children like Jem
consistently position heads before their complements from the very earliest multiword utterances they
produce. They do not use different orders for different words of the same type (e.g. they don’t position
the verb see after its complement but the verb want before its complement), or for different types of
words (e.g. they don’t position verbs before and prepositions after their complements).
By contrast, Korean children correctly position heads after their complements from the very onset
of multiword speech, as the following examples of complement+verb structures in child Korean (from
Lee 1999) illustrate:
(18)(a) Unga haessta [Nohen 1;9] (b) Wungdong haessci [Younguk 2;4]
Shit did Exercise did
‘I had a shit’ ‘I took some exercise’
This suggests that such children already ‘know’ that Korean is a head-final language.
A natural question to ask at this point is how we can provide a principled explanation for the fact
that from the very onset of multiword speech we find English children correctly positioning heads
before their complements (and Korean children correctly positioning heads after complements) The
Principles-and-Parameters model enables us to provide an explanation for why children manage to
learn the relative ordering of heads and complements in such a rapid and error-free fashion. The
answer provided by the model is that learning this aspect of word order involves the comparatively
simple task of setting a binary parameter at its appropriate value. This task will be a relatively
straightforward one if the language faculty tells the child that the only possible choice is for a given
type of structure in a given language to be uniformly head-initial or uniformly head-final. Given such
an assumption, the child could set the parameter correctly on the basis of minimal linguistic
experience. For example, once the child is able to parse (i.e. grammatically analyse) an adult utterance
such as Help Daddy and knows that it contains a verb phrase comprising the head verb help and its
complement Daddy, then (on the assumption that the language faculty specifies that all heads of a
given type behave uniformly with regard to whether they are positioned before or after their
complements), the child will automatically know that all verbs in English are canonically (i.e.
normally) positioned before their complements.
In much the same way, there seems to be evidence that children correctly set the Wh-parameter at
the value appropriate to the language they are acquiring from the very earliest wh-questions they
produce. Guasti (2000) uses data from the Adam, Eve, Nina and Sarah files on the CHILDES data-
base to argue that (from the very earliest wh-questions they produce), English children correctly move
wh-words to the front of the interrogative clause, producing the five different types of wh-question in
(19) below (the numbers in parentheses giving the raw number and frequency of the relevant type of
construction in the four sets of files which Guasti examines):
(19)(a) Where (i)s Daddy going? (1440/52%)
(b) Where Daddy go? (652/24%)
(c) Where Daddy going? (354/13%)
(d) Where Daddy (i)s going? (84/3%)
(e) Where Daddy goes/went? (237/9%)
Guasti’s findings suggest that children set the Wh-Parameter at the value appropriate for English from
the outset and so ‘know’ that an interrogative C (has an EPP feature which) attracts an interrogative
wh-expression to move to the front of the relevant clause (to become the specifier of C).
In much the same way, we have evidence that the Head Strength Parameter is correctly set from a
very early stage. Amy Pierce (1992) studied transcripts from four French-speaking children aged
between 1;8 and 2;2. From looking at the negative sentences they produced, she found evidence that
they already ‘knew’ that a finite T is strong in French and so can attract an (italicised) main/lexical
verb to raise out of the head V position in VP, across the negative particle pas ‘not’ into the head T
position in TP, so resulting in sentences such as (20) below:
(20)(a) Marche pas (b) Ça tourne pas
Works not (= ‘It doesn’t work’) It turns not (= ‘It won’t turn’)
216/225 of the finite negative clauses produced by the children showed the finite verb being correctly
raised to the T position in front of the negative particle. Likewise, the children also knew that a non-
finite T is weak in French and so non-finite verbs remain in situ in the head V position of VP, and
hence follow the negative particle pas ‘not’ as in the sentences below (where INF denotes a verb in the
infinitive form):
(21)(a) Pas casser (b) Pas rouler en vélo
Not breakINF ‘No break’ Not rollINF on bike ‘Not go around on a bike’
122/124 of the non-finite negative clauses the children produced showed the non-finite verb remaining
in situ in the head V position of VP, following the negative particle. Pierce’s data suggest that French
children correctly set the Head Strength Parameter from the outset, and so ‘know’ that a finite (but not
a non-finite) T is strong and hence can attract a main verb to move from V to T.
In much the same vein, Poeppel and Wexler (1993) argue (from a study of a German child called
Andreas at age 2;1) that there is evidence that German children know from a very young age that T
and C are always strong in finite main clauses, with the result that finite verbs raise through T into C
(with some other constituent moving into the specifier position within CP), resulting in structures such
as that shown in skeletal form below:
(22) [CP Eine Fase [C hab] ich] (Andreas 2;1)
A vase have I
Movement of an (italicised) finite verb through T into C results in so-called V2/verb-second word
order, in that the verb ends up as the second constituent in the sentence – in the case of (22), following
the preposed object Eine Fase ‘a vase’. Poeppel and Wexler report that Andreas used V2 word order in
197/208 of the finite main clauses which he produced. The obvious conclusion to draw from such data
is that Andreas has correctly set the Head Strength Parameter for the relevant functional heads, and
‘knows’ that T is always strong in finite clauses in German, and that C is likewise strong in finite main
clauses.
Further support for the claim that children correctly set the Head Strength Parameter at its
appropriate value from the outset comes from a study of 12 American children by Stromswold (1990).
She found that in positive questions containing an overt auxiliary, the children inverted the auxiliary
93.4% of the time. Since auxiliaries move to C in interrogative structures like that shown in highly
simplified form below:
(23) [CP Where [C has] Daddy gone]?
this suggests that the children already know that C is strong in main-clause questions in English.
What all of this suggests is that from their very earliest multiword utterances, children set
parameters correctly, so that (e.g.) even one-year old English children ‘know’ that English is a
head-initial language (and so always position heads before their complements), a wh-initial language
(so that C in an interrogative clause attracts a wh-expression to move to initial position at the front of
the clause), a weak T language in which T is always weak (and so cannot attract a main verb to move
to T) – and (in main-clause questions) a strong C language (i.e. a language in which C in main
interrogative clauses attracts an auxiliary in T to move to C). We might therefore expect to find
evidence that from the very earliest multiword utterances they produce, English children also set the
Null Subject Parameter at the appropriate ‘no null subjects’ value and hence know that English is not a
null subject language – and so (for example) does not allow a subject pronoun to be ‘dropped’ in a
main clause. And yet, as we shall see in the next section, this does not appear to be so.
2.8 Setting the Null Subject Parameter
In highly influential research in the 1980s, Nina Hyams (1986, 1987,
1989) came up with an interesting finding which had an enormous impact on
acquisition research. She observed that young children acquiring a non-null
subject language like English initially go through a null subject stage (which
typically lasts until around the age of 2;1) during which they frequently produce
sentences with null subjects (e.g. saying Want ice-cream rather than I want ice-
cream), and that only when they enter the next stage of development (at around
age 2;2) do they stop using null subjects and correctly analyse English as a non-
null-subject language. Hyams also observed that children acquiring a genuine
null subject language like Italian correctly identify it as a null subject language
from the very outset, and that at all stages of their development Italian children
make adult-like use of null subjects. Hyams concluded that both children
acquiring null subject languages and those acquiring non-null-subject languages
go through an initial stage in which they analyse the language as a null-subject
language and so make frequent use of null subjects. Why should this be?
The answer given by Hyams in her (1986) book26 is that the innate Language Faculty provides a
universal default setting for the Null Subject Parameter/NSP: in other words, parameters like NSP
come with a preset initial default value determined by UG (i.e. by the theory of Universal Grammar
innately wired into the Language Faculty of the brain). In the case of NSP, the initial value specified
by UG is positive [+NS] (i.e. ‘allows a finite verb to have a null subject’). In other words, children are
genetically programmed to assume initially that the language they are acquiring is a null subject
language which allows any finite verb in any kind of clause to have a null subject, and only to change
the setting of the parameter to the corresponding negative value [-NS] (i.e. ‘doesn’t allow a finite verb
to have a null subject’) if they find clear evidence that the language they are acquiring doesn’t allow
finite verbs to have null subjects. On this view, the learning task which children face is to decide (on
the basis of their experience) whether or not to re-set the Null Subject Parameter: e.g. English children
have to re-set the parameter, whereas Italian children do not. This leads Hyams to develop a
parameter-resetting model of acquisition in which children don’t have to learn how to set parameters
like NSP but rather whether to reset them.
The parameter-resetting model raises important questions about the mechanism by which children
learn to re-set parameters – e.g. how and why English children come to reset the Null Subject
Parameter from its pre-set (UG-determined) initial setting [+NS] ‘allows null subjects’ to the [-NS]
‘doesn’t allow null subjects’ value appropriate for adult English. Hyams argued that particular types of
structure serve as triggers for parameter-resetting, claiming that ‘Triggers are...data...incompatible
with the child’s current grammar’ (Hyams 1989, fn 16). For example, she noted that an important
typological difference between null-subject languages (NSLs) and non-null-subject languages (non-
NSLs) is that non-NSLs have expletive subject pronouns, whereas NSLs like Italian do not. In this
connection, compare the English sentence (24a) below with its Italian counterpart (24b):
(24)(a) It is evident that there have remained several problems
(b) È evidente che sono rimasti parecchi problemi
Is evident that are remained several problems
The pronouns it and there in (24a) are said to be expletive pronouns in that they seem to have no
intrinsic semantic properties (e.g. it doesn’t refer to a specific inanimate entity, and there doesn’t refer
to a specific place) – as we see from the fact that we cannot replace it in (24a) by a referential pronoun
like this/that or by an interrogative pronoun like what? nor replace there by here or by an interrogative
pronoun like where? (cf. *What is evident that there have remained several problems?/*Where is it
evident that have remained several problems?). The fact that English requires us to use expletive
subject pronouns in sentences like (24a), Hyams claims, is a reflection of the fact that English is a non-
null-subject language which therefore requires every finite (auxiliary or main) verb to have an overt
subject – leading to the use of ‘meaningless’ expletive subject pronouns in sentences like (24a). By
contrast, a null-subject language like Italian lacks expletive pronouns, for the obvious reason that it
has no requirement for a finite verb to have an overt subject – as we see from the absence of any
counterpart of it/there in (24b). Hyams (1986) suggested that English children typically acquire
expletive subject pronouns at around two years of age, and that the acquisition of these pronouns
serves as an automatic trigger for resetting the Null Subject Parameter (the idea being that UG tells
the child that any language with expletive pronouns is a Non-Null-Subject language and hence does
not allow Italian-style null pro subjects). This has become known as the Trigger Hypothesis.
Hyams’ Triggered Parameter Resetting/TPR model of the acquisition of the Null Subject
Parameter proved to be problematic in ways which you can explore for yourself in relation to some of
the exercise material at the end of the chapter. This led her to develop an alternative Morphological
Uniformity/MU account of the acquisition of the Null Subject Parameter in later work (Hyams 1992).
The starting point for her revised model was an account of the Null Subject Parameter in adult

26
Here, I present a highly simplified and relatively non-technical account of Hyams’ work. For alternative
textbook accounts of her work, see the relevant chapters of Atkinson (1992), O’Grady (1997) and Guasti (2002).
For a critique of Hyams’ early work, see Valian (1991).
grammars developed by Jaeggli and Safir (1989). They observed that Null Subject Languages/NSLs
like Italian and Chinese differ from Non-Null Subject Languages/non-NSLs like English in that verbs
in NSLs are uniform in respect of the agreement inflections carried by finite verbs whereas those in
non-NSLs are non-uniform. We can illustrate this difference in terms of forms of the verb speak and
its Italian and Chinese counterparts which would be used in a present-tense context (i.e. in a context
where English would use a present-tense verb, e.g. in a sentence like I speak French), which are
shown in tabular form below:

25 Verb forms used in present-tense contexts in Italian, Chinese and English

Form Italian Chinese English
1.SG parlo shuo speak
2.SG parli shuo speak
3.SG parla shuo speaks
1.PL parliamo shuo speak
2.PL parlate shuo speak
3.PL parlano shuo speak

As the table above shows, the relevant verb forms in Italian are morphologically uniform in that all
are of the form STEM+AFFIX (the relevant agreement affixes being italicised). Likewise, the
corresponding verb forms in Chinese are also morphologically uniform, in that they all comprise a
‘bare’ (uninflected) verb stem. But the corresponding verb forms in English are morphologically non-
uniform in that the form speaks comprises the stem speak and the affix -s, whereas the form speak
comprises only a bare stem27. Observing that Italian and Chinese are Null Subject Languages and are
also morphologically uniform, whereas English is a Non-Null Subject Language and is non-uniform,
Jaeggli and Safir concluded that principles of Universal Grammar wired into the innate Language
Faculty determine that morphologically uniform languages allow null subjects whereas non-uniform
languages do not.
Hyams (1992) noted that English children typically don’t acquire tense inflections (like present
tense -s and past tense -d) until around two years of age, and that in contexts where adult use a tense-
inflected verb, one-year-olds typically use a bare verb (saying e.g. Daddy play with me in contexts
where an adult would say Daddy plays/played with me). She concluded from this that the verb forms
produced by one-year-old English children in finite contexts are morphologically uniform (in that, like
their Chinese counterparts, they are all bare forms). If UG specifies that morphologically uniform
languages allow null subjects but non-uniform languages do not, then (concluded Hyams) verb
morphology (viz. whether finite verbs are uniformly inflected or not) serves as a trigger which
automatically sets the Null Subject Parameter at the appropriate value. Thus, one-year old English
children who have not yet acquired third-person-singular present-tense -s will have a uniform (bare)
verb morphology that serves as a trigger which automatically sets NSP at the [+NS] ‘allows null
subjects’ value. But once children acquire third-person-singular present-tense -s at around two years of
age, their verb morphology becomes non-uniform, and this triggers an automatic re-setting of NSP to
the [-NS] ‘doesn’t allow null subjects’ value appropriate for adult English.
2.9 Child null subjects as a reflex of adult null subjects

27
Of course, bare verbs in English could be taken to carry a null affix. If so, a uniform language would be
defined as one which uniformly uses only overt affixes, or only null affixes (not a combination of the two).
 The core assumption underlying (either version of) Hyams’ parameter-resetting model is that
children acquiring a language like English go through an initial stage when they wrongly
assume that English allows the kinds of null subject found in Italian or Chinese – so that
English children initially mis-set the Null Subject Parameter, and have to subsequently
reset it. However, research by a number of other linguists has questioned this approach,
and instead argued that the kinds of null subject used by children acquiring English are
simply a reflex of the kinds of null subjects found in adult English. For example,
Guilfoyle (1994) and O’Grady et al (1989) observed that the null-subject sentences
produced by young children are typically non-finite, in that they contain no auxiliary or
verb inflected for tense (e.g. no present/past tense auxiliary like is/has and no present/past
tense main verb like goes/went). The significance of their observation is that non-finite
clauses in adult English (and perhaps universally) allow a null subject – as we can
illustrate in relation to sentences like those below (where PRO designates the null
subject):
(26)(a) It’s important [PRO to prepare myself properly for the exam]
(b) There’s no point in [PRO blaming yourself for what happened]
The bracketed clauses in (26) are non-finite in that they contain no present or past tense verb or
auxiliary; since reflexives like myself/yourself require a local antecedent (i.e. must refer back to a
nearby expression within the bracketed clause containing them), the bracketed clause in each example
must contain a subject pronoun which can serve as the antecedent of the reflexive. But since the
bracketed clause does not contain an overt subject pronoun, it must contain a null subject pronoun of
some kind. The relevant kind of null subject pronoun cannot occur in finite clauses, as we see from:
(27) It’s important [that I/*PRO should prepare myself properly for the exam]
Since the type of subject pronoun found in sentences like (27) occurs only in non-finite clauses, we
can refer to it as a non-finite null subject pronoun and term it ‘big PRO28’ (in order to differentiate it
from the finite null subject pronoun found in finite clauses in languages like Italian, conventionally
termed ‘little pro’.) In (26a) PRO refers to the speaker and so is a first person singular pronoun which
can accordingly serve as the antecedent of the first person singular reflexive myself. By contrast, in
(26b), PRO refers to the person being spoken to and so is a second person singular pronoun which can
therefore serve as the antecedent of the second person singular reflexive pronoun yourself.
In the light of the assumption that non-finite clauses can have a non-finite PRO null subject,
consider child null-subject sentences such as the following (from Radford 1990):
(28) Coming to rubbish. Gone out. Want one (Bethan 1;8)
The verb form coming is a progressive participle, and hence a non-finite verb.
Likewise, gone is a perfect participle, and so also a non-finite verb. And since the
verb want carries no inflection of any kind, it too may be a non-finite form (e.g.
the infinitive form found after infinitival to in adult sentences like ‘Is it so wrong to
want happiness?’). If all three verbs in (28) are non-finite (participle or infinitive)
forms, then it is plausible to maintain that each has a non-finite PRO subject, as
shown in highly simplified form below:
(29) PRO coming to rubbish. PRO gone out. PRO want one
It goes without saying that if the null subjects found in early child English all
occur in non-finite clauses, then children’s null subjects can be taken to be
instances of PRO (i.e. the type of null subject found in non-finite clauses in adult English). There
would then be no evidence that English children ever mis-set the Null Subject Parameter and wrongly
assume that English allows finite clauses to have the kind of finite null pro subject found in Italian,
since children’s null-subject clauses would be non-finite and hence allow the kind of null PRO subject
found in non-finite clauses in adult English.

28
‘Big’ in the sense that it is conventionally written in capital letters.
 However, while it is plausible to suppose that some of the null subjects
produced by English children are non-finite PRO subjects (occurring in non-finite
clauses), it seems implausible to analyse all null-subject sentences produced by
English children in this way. For example, Sano and Hyams (1994) report that
56.5% of the d-inflected verbs produced by Adam (in the Brown corpus on the
CHILDES data-base) from age 2;3 to 3;0 had null subjects, citing examples such
as the following:
(30) Goed on that way. Dropped a rubber band. Slapped Becca and Rachel
If (as seems likely) the verb forms in (30) are past tense forms, they are finite and
so cannot in principle have a non-finite PRO subject. Moreover, it is relatively
common to find young children using null subjects with finite auxiliaries, as
examples such as the following (produced by a different boy called Adam at age
2;2 in the course of a single 45-minute recording) show:
(31) Don’t know (x 1429: used in reply to questions like ‘What’s this?’, ‘Is it a
train?’ etc.)
Don’t paint that (= ‘I didn’t paint that’). Don’t work (x 3 = ‘It doesn’t
work’)
Don’t wanna draw on this one. Does (response to ‘Yeah, it does’). Won’t
(response to ‘Does
it work?’). Can’t knock them down. Can’t get it out. Can’t stroke me now.
Can’t (x 3: reply to
questions like ‘Can you see anything, Adam?’)
Since sentences like those in (31) contain a present-tense auxiliary like
does/don’t/won’t/can’t, the relevant clauses are clearly finite and thus cannot be
analysed as having a non-finite PRO subject. So what kind of null subject do they
contain?
An interesting answer to this question is offered by Luigi Rizzi (1994). Rizzi notes that in rapid
colloquial speech adults often truncate English sentences by not pronouncing a weak (e.g. short,
unstressed) word at the very beginning of a sentence. This can lead to adults truncating a weak
(unstressed, non-contrastive) subject pronoun when it is the first word in a sentence, as in the
examples below (produced by mothers in conversation with their children, from the Bates files on the
CHILDES data-base; the omitted subject pronoun is indicated in parentheses)
(32)(a) Think they’re finished (= I; mother talking to Ed at 2;4)
(b) Guess that’s all, huh? (= I; mother talking to Mandy at 2;4)
(c) Don’t know (= I; mother talking to Frank at 2;4)
(d) Almost had it (= you; mother talking to Wanda at 1;8)
(e) Goes in this hole (= it; mother talking to Zeke at 1;8, after he takes object from mother)
(f) Looks like a piece of pie (= it; mother talking to Betty at 2;4)
(g) No. Won’t come off (= it; mother talking to Frank at 2;4)
(h) Won’t work, will it? (= it; mother talking to Paula at 2;4)
(i) Doesn’t want any? (= he; mother talking to Jane at 2;4)
In each of the above sentences, the (italicised) weak subject pronoun has been truncated (i.e. omitted).
However, this type of omission is only possible where the subject is the first word in the sentence: for
example, we could not truncate you in (33) below because it is the third word in the sentence – hence
the ungrammaticality of:
(33)(a) What would you like to drink? (b) *What would like to drink?
Truncation can also affect other kinds of sentence-initial word – for example, the wh-word what can
be truncated in rapid speech when unstressed in a wh-question such as (34a) below (truncation being
marked by strikethrough), and as can be truncated in (34b)

29
The notation ‘x 14’ means that Adam produced this utterance 14 times in the relevant recording.
(34)(a) What time is it? (b) As soon as I get home, I’ll ring my dad
The truncation analysis can account for null-subject sentences like Can’t get it out in (31) by
supposing that the weak pronoun I can be truncated (i.e. given a null spellout) by virtue of being the
first word in the sentence.
It may well be that prosodic factors play some role in accounting for why weak sentence-initial
words (including weak subjects) can be truncated. In this context, it is interesting to note that Gerken
(1991) suggests that young children tend to omit unstressed syllables in iambic feet (i.e. metrical feet
which consisting of a weak+strong syllable) more frequently than in trochaic (strong+weak) feet.
More specifically, she claims that young children tend to omit function words (including pronouns and
determiners like a/the) when they are unstressed and occur in iambic feet. The percentages in (35)
below show the rate of omission of unstressed function words in two target sentences on a sentence
repetition task given to a group of English children with a mean age of 2;3 (with square brackets
indicating metrical feet, and italics marking unstressed words):
(35)(a) [She39% kissed] [the28% dog] (b) [The39% dog] [kissed her0%]
Sentence (35a) contains two iambic feet, and in each case the italicised unstressed word is omitted by
the children around a third of the time. Sentence (35b) contains an iambic foot followed by a trochaic
foot: the unstressed determiner the in the iambic foot is frequently omitted, but the unstressed object
pronoun her in the trochaic foot is never omitted. Thus, under Gerken’s account, omission of subject
pronouns is attributable to phonological rather than grammatical factors.
The overall significance of our discussion in this section is as follows. The crucial assumption
made by Nina Hyams is that English-speaking children go through a stage in which they use a kind of
null subject found in Italian (or Chinese), but not found in adult English. If this is true, it means that
the relevant children do indeed initially mis-set the Null Subject Parameter. But given that adult
English allows three types of null subject (null second person imperative you subjects, null non-finite
PRO subjects, and null truncated sentence-initial subjects), the question which arises is whether
children acquiring English simply produce the three types of null subject found in adult English, or
whether they also produce a fourth type of null subject (e.g. the kind of finite null pro subject found in
Italian). This is a question which the material in the workbook section is designed to help you answer.

Workbook section

2.1 On expletives as triggers

Nina Hyams (1986, p.94) notes that children are exposed to sentences containing expletive subjects
from birth. She also notes (1989, fn.17) that Hebrew is a null subject language which has an expletive
subject pronoun ze ‘it’. Ingham (1992) reports on an intensive longitudinal study of a two-year-old girl
and notes that she acquired expletives several months after the end of the null-subject stage. Valian
(1991) reports that the American children in her study used (albeit sporadically) expletive subject
pronouns during the null-subject stage. What questions/problems do these observations raise for
Hyams’ Triggered Parameter Resetting model?

2.2 Distribution and frequency of null subjects

Virginia Valian (1991) reported on a comparative study of null subjects produced by Italian and
(English-speaking) American children. One of her findings was that while the American children in
her study frequently produced null subjects in finite main clauses, they never did so in finite
subordinate clauses (none of the 132 finite subordinate clauses produced by the English children she
studied had a null subject). She also reported the following mean group figures on the frequency of
null subjects in clauses containing a (non-imperative) verb. The children with a mean MLU of 1.8 (and
a mean age of 2;0) showed 31% mean use of null subjects; those with a mean MLU of 2.5 (and a mean
age of 2;5) showed 11% mean use of null subjects; those with a mean MLU of 3.4 (and a mean age of
2.5) showed 7% mean use of null subjects. Those with a mean MLU of 4.2 (and a mean age of 2;7)
showed 5% use of null subjects. The English-speaking adults in her study showed 2%-4% use of null
subjects. The Italian children in her study showed 70% mean use of null subjects throughout. Discuss
the significance of these findings for the acquisition of NSP.

2.3 Children’s use of overt subjects

Virginia Valian (1991, p.45) reports that in her study of American children using null subjects, ‘more
than half their subjects’ were overt pronouns. What is the potential significance of this observation –
and why is it unfortunate that Valian doesn’t differentiate between strong and weak pronouns, and
hence includes demonstrative and interrogatives in her count as well as personal pronouns?

2.4 Null Subjects Languages and verb morphology

William O’Grady (1997, p.88) notes that (adult) Persian and Wichita are null-subject languages.
Persian has the following system of agreement suffixes on regular finite (present tense) verbs:
1.SG = -am 2.SG = -i 3.SG = ø 1.PL = im 2.PL = id 3.PL = and
Wichita has the following system of agreement prefixes (for transitive verbs with a 3.sg direct object):
1/2/3.SG = ø- 1/2/3.DUAL30 = hi- 1/2.PL = ra.k/hi- 3.PL = hi-
Discuss the implications of O’Grady’s observations for trigger-based accounts of the setting of NSP.

2.5 Sporadic use of verb inflections

Studies of the acquisition of English verb morphology suggest that the acquisition of 3.SG.PRES -s is a
protracted process, with one- and two-year-olds sometimes using -s and sometimes omitting it in
contexts where adults use it (e.g. alternating between Daddy like ice-cream and Daddy likes ice-
cream). For example, de Villiers and de Villiers (1973) report that Joshua at age 1;9 showed 12.5% use
of -s in obligatory contexts. What issues does this raise for the Morphological Uniformity account of
the Null Subject Parameter?

2.6 Weissenborn’s study of French children

Juergen Weissenborn (1992, pp.280-281) reports a group of children aged between 1;10 and 2;2
acquiring French (a non-null subject language) producing null-subject sentences such as the following:
a) Peux le faire b) Ai mangé des quettes c) Est sale
Can it do Have eaten some pancakes Is dirty
‘I can do it’ ‘I have eaten some pancakes’ ‘It is dirty’
d) A fait e) Est tombé f) Veux manger
Has done Is fallen Want eat
‘He has done it’ ‘He has fallen’ ‘I want to eat’
Weissenborn also reports (p.285) that none of the French children in his study used null subjects in
finite complement clauses. Discuss the potential significance of Weissenborn’s observations for
accounts of children’s null subjects.

2.7 Children’s wh-questions

Roeper and Rohrbacher (2000) report that two-year-old English children alternate between finite main-
clause wh-questions (like Where’s Daddy going?/Where’s Daddy gone?/Where did Daddy go?) and
non-finite main-clause wh-questions like Where Daddy going/gone/go? They note that while 49% of
the non-finite wh-questions produced by the children in their study were null-subject questions like
Where going/gone/go?, virtually none (only 5%) of the finite wh-questions they produced had null
subjects. Similarly, Crisma (1992) reports that only 1% of the wh-questions produced by the French
children in their study had a null subject. By contrast, Guasti (1996) reports that Italian children show
60% use of null subjects in wh-questions. Discuss the potential significance of these findings for

30
A dual form denotes two entities, and a plural form more than two (e.g. both is a dual quantifier in English,
whereas all can serves as a plural quantifier)
accounts of children’s null subjects.

2.8 The topic-drop analysis of null subjects

In a (1994) paper not discussed in the main text, Nina Hyams proposed a further
account of the null subjects used by English children which sees them as being
the result of the same type of Topic Drop phenomenon as is found in Chinese.
Languages like Chinese which have no (person/number) agreement morphology
allow a pronoun to be null if it is a discourse topic (e.g. if it refers to some familiar
entity previously mentioned in the discourse). For example, in reply to a question
like ‘Did John see Mary?’, English would say ‘Yes, he saw him’, but Chinese can
omit either the subject pronoun (and say ‘Yes, saw him’) or the object pronoun
and say ‘Yes, he saw’ (or can indeed omit both pronouns and say ‘Yes, saw’).
Hyams (1994) suggested that English children at the null-subject stage may
(mistakenly) assume that English is a topic-drop language like Chinese, and so
allows a subject to be null when it is a discourse topic (i.e. when it represents
familiar information). A study of a group of 2-4 year-old English and Chinese
children by Wang et al. (1992) showed 47% use of null subjects and 23% use of
null objects in Chinese, compared to 33% use of null subjects and 4% use of null
objects in English. Discuss the potential significance of these two sets of findings
for Hyams’ topic-drop analysis of English children’s null subjects.

2.9 An alternative processing account of null subjects

Paul Bloom (1970) suggested a processing account of null subjects, to the effect that processing
constraints on utterance length (i.e. a restriction on the length of utterances that children can process)
lead children to omit subjects in longer sentences. In this connection, consider the data in the table
below, which shows the mean number of words in the verb phrases produced by three English-
speaking children in sentences with and without a subject:
Mean number of words in verb phrases/VP in sentences with/without a subject
Child Age VP with a subject VP with no subject
Adam 2;3-2;7 2.3 2.6
Eve 1;6-1;10 2.0 2.7
Sarah 2;3-2;7 1.8 2.5
To what extent are the findings reported in the table consistent with Bloom’s processing account of
null subjects? (For a critique of Bloom’s work, see Hyams and Wexler 1993.)

2.10 Wh-movement and auxiliary-inversion in wh-questions

Consider the extent to which Guasti’s findings about wh-questions in child English (reported in (19) in
the main text) lend support to the claim that children set parameters correctly from the very outset.
Consider also the implications of a longitudinal study by Labov and Labov (1978) of their daughter
Jessie which showed that Jessie never failed to prepose wh-expressions in wh-questions, but only
sporadically inverted auxiliaries in main-clause questions (her frequency of auxiliary inversion rising
steadily from around 10% inversion at three-and-a-half years of age to around 60% inversion at four-
and-a-half years of age, according to Labov and Labov (1978, p. 22).

2.11 How children parse sentences

Janet Dean Fodor (1998) argues that UG must ensure that young children are able to parse (and
acquire) languages with distinct parameter settings – e.g. both a wh-initial language like English and a
wh-in-situ language like Korean. What implications (if any) does her observation have for Hyams’
(1986) claim that (at least some) parameters have a universal (UG-determined) fixed initial setting?
 3 The structure-building model

3.1 Introduction
One of the most intriguing questions which theories of acquisition seek to answer is how
children start to combine words together to form syntactic structures, and what is the nature of the
initial syntactic structures (i.e. the earliest multiword combinations) produced by one-year old
children. In this connection, consider the differences between adult and child speech illustrated by the
dialogue below between a mother and a one-year-old child:
(1) MOTHER: What’s the man doing? CHILD: Man driving car MOTHER: Yes, the man’s driving a car
If we compare the child’s utterance Man driving car with its adult counterpart The man’s driving a car
we see that both contain the nouns man/car and the verb driving, but that they differ in that the child’s
utterance lacks the determiners the/a and the auxiliary is. Since nouns and verbs are lexical (or
substantive) categories whereas determiners and auxiliaries are functional categories, what this might
lead us to conclude is that the earliest syntactic structures produced by one-year-olds are projections of
(i.e. are built up by combining) lexical categories, and that functional categories are acquired at a later
stage of acquisition. This is the core idea underlying the structure-building model of acquisition
developed in the 1980s and 1990s in a number of acquisition studies (including Radford 1986, 1988,
1990, 1995; Lebeaux 1987; Guilfoyle and Noonan 1988, 1992; Kazman 1988; Tsimpli 1992; Vainikka
1994; Powers 1996). The key assumption underlying the model is that one-year-old children initially
go through a lexical/prefunctional stage in the acquisition of syntax during which they produce
structures which contain only lexical categories like N (= nouns), V (= verbs), A (= adjectives), and P
(= prepositions). At around two years of age (several months earlier/later in the case of fast/slow
developers) they enter a later functional stage at which they start to produce structures containing
functional categories like T (= tense-/agreement-marking auxiliary), C (= clause-typing
particle/complementiser) and D (= determiner). On this view, children gradually build more and more
complex syntactic structures in the course of their linguistic development, starting with simple lexical
structures and then gradually building up more and more complex functional structures.

3.2 Absence of T in children’s initial clauses

If the earliest syntactic structures produced by one-year-old children are prefunctional in
nature, we should expect to find that children’s initial clauses lack the T constituent which forms an
integral part of the structure of all clauses containing verbs in adult English. T is generally considered
to be the locus of tense and subject-agreement features in clauses, as we can illustrate in relation to
adult English sentences such as the following:
(2)(a) Daddy is working (b) Daddy has gone to work
(c) Daddy is nice/at work/a doctor (d) Daddy plays with me
Each of the clauses in (2) contains a T constituent which carries (present) tense and (third person
singular) agreement features. The T position is occupied by the progressive auxiliary is in (2a), the
perfect auxiliary has in (2b), and the copula is in (2c)31. So, for example, (2b) contains a TP with the
structure shown in simplified form below:
(3) [TP Daddy [T has] [VP [V gone] to work]]
If we suppose that all adult clauses contain a T constituent which carries tense and agreement features,
then (2d) will also contain a T constituent marking tense and agreement. But since there is no overt
auxiliary like is or has in sentences like (2d), T simply contains an abstract third person singular
present tense affix (below denoted as Af) which (whenever there is no auxiliary in T for it to attach to)
is lowered onto the main verb by a morphological operation traditionally known as Affix Hopping –
as shown by the arrow in (4) below:
(4) [TP Daddy [T Af3.SG.PR] [VP [V play ] with me]]

The resulting verb+affix structure play+Af3.SG.PR is ultimately spelled out as plays.

The conclusion to be drawn from the discussion in the previous paragraph is that all finite clauses
in adult English contain a T affix which is the locus of tense and agreement features: this is attached to
the auxiliary in T if there is one, and lowered onto the main verb by Affix Hopping if there is not.
Now, if (as claimed under the structure-building model) children’s initial clauses are prefunctional
structures which contain no functional categories and hence no T constituent, and if T is the locus of
finiteness (i.e. tense and agreement marking), then we should expect to find that children’s initial
clauses will contain neither auxiliaries nor main verbs inflected for tense and agreement. In this
connection, it is interesting to note that in contexts where adults use a structure containing a finite
(present or past tense) auxiliary, one-year-old children use a simple structure containing a non-finite
(tenseless) main verb without an auxiliary – as the following examples illustrate 32:
(5)(a) Mummy doing dinner. Teddy crying (Daniel 1;10)
(b) Daddy taken biscuit. Man got coat on (Daniel 1;10)
The adult counterparts of structures like (5a) would contain the finite T constituent is plus a non-finite
V like doing/crying in the progressive participle form, while the adult counterpart of (5b) would
contain the T finite constituent has plus a non-finite verb like taken/got in the perfect participle form:
however, the child’s sentence in each case simply contains a non-finite verb, with no auxiliary.
Likewise, in contexts where adults use the tense-/agreement-marking copula BE to link a subject to a
an adjectival or prepositional predicate, one-year-olds will typically omit the copula: cf,
(6) Geraint naughty (Bethan 1;9) Mommy busy (Kathryn 1;9)
Wayne in bedroom (Daniel 1;9) Stephen in school (Jem 1;11)
Such verbless clauses are often referred to as small clauses: the adult counterpart of child sentences
like (6) will contain the finite T constituent is, but the corresponding structure produced by the
relevant children contains no is and so can be assumed to lack a T constituent. And as the dialogues in
(7) below illustrate, in contexts where adults use a clause containing a (bold-printed) auxiliary or main
verb marked for past or present tense, children use an (italicised) bare/uninflected infinitive verb form:
(7)(a) ADULT: What did you draw? CHILD: Hayley draw boat (Hayley 1;8)
(b) ADULT: What does Ashley do? CHILD: Ashley do pee, Ashley do poo (Jem 1;11)
If T is the locus of the tense and agreement features of main verbs, and if children’s initial clauses
contain no T constituent, it follows that the verbs they produce in finite contexts will be bare non-finite
verb forms like draw/do not marked for tense or agreement.
Further evidence of the absence of T in children’s initial clauses comes from the fact that in
contexts where adults use a complement clause containing infinitival to, children use a bare (to-less)
infinitive, as we see from the italicised complement clauses in the examples below:
(8)(a) Want [Teddy drink] (Daniel 1;7) (b) Want [lady get chocolate] (Daniel 1;11)
31
BE is said to be a copula (i.e. linking verb) in such sentences because it serves to link the subject Daddy with
an adjectival/prepositional/nominal expression like nice/at work/a doctor.
32
Most of the argumentation and illustrative data presented here are from Radford (1990).
 (c) Want [open door] (Daniel 1;8) (d) Want [sit on knee] (Daniel 1;11)
Given the assumption in much recent work that infinitival to is a non-finite tense particle belonging to
the category T (and often future in interpretation), the absence of infinitival to in the bracketed
complement clauses in sentences like (8) is consistent with the assumption that early child clauses
contain no T constituent.
If (as assumed here) adult clauses contain a T constituent but child clauses do not, an obvious
question which arises concerns the nature of the initial clauses produced by children. Radford (1990)
suggests that these clauses are simple verb phrases/VPs which contain no T and hence no TP either.
On this view, sentences like (5a) Mummy doing dinner, (5b) Daddy taken biscuit and (7a) Hayley
draw boat would be VPs with the respective structures shown in simplified form below:
(9) [VP Mummy [V doing] dinner] [VP Daddy [V taken] biscuit] [VP Hayley [V draw] boat]
Likewise, the bracketed infinitive complement in (8b) Want [lady get chocolate] would be a VP with
the simplified structure shown by the labelled bracketing in (10) below:
(10) Want [VP lady [V get] chocolate]
In each of the relevant verb phrases, the verb doing/taken/draw/get would be the head of the overall
VP structure, the preceding noun Mummy/Daddy/Hayley/lady would be its subject/specifier, and the
following noun dinner/biscuit/boat/chocolate would be its complement. The verbs produced by
children at this stage would all be non-finite (and so not marked for tense or agreement), comprising
either a bare infinitive form like draw/get or a progressive participle like doing, or a perfect participle
like taken.
Radford (1990) suggests that children’s null-copula sentences like (6) may also be VPs, but headed
by a null copular verb, so that a sentence like Geraint naughty would have a structure along the lines
of (11) below:
(11) [VP Geraint [V ø] naughty]
The copula would serve the semantic function of expressing predication (i.e. indicating that the
property of being naughty is predicated of Geraint), but would have a null spellout rather than being
spelled out as is because the form is is marked for tense and agreement, and tense/agreement are
properties of the T constituent which one-year-olds have typically not yet acquired. Note that a copula
like is can also receive a null spellout under certain circumstances in adult English – e.g. in so-called
‘Mad magazine sentences’ like that italicised in (12) below:
(12) SPEAKER A: I reckon Gary is drunk SPEAKER B: Gary drunk? I don’t believe it!
In (12), SPEAKER B’s small clause Gary drunk appears to echo SPEAKER A’s finite clause Gary is
drunk, and yet contains no overt counterpart of the copula is (suggesting that the copula gets a null
spellout in ‘Mad magazine’ sentences). If so, this lends added plausibility to the claim that verbless
structures such as (11) contain a null counterpart of the copular verb BE.
An interesting implication of the VP analysis of children’s initial clauses in (9,10,11) is that
subjects occupy a ‘lower’ position (= subject position within VP) than in adult English (where they
occupy the subject position within TP). This might at first sight seem to involve positing undesirable
discontinuity between the structure of adult and child clauses. However (as we saw in chapter 1), work
in theoretical syntax over the past two decades has argued that subjects originate internally within VP
(as semantic arguments33 of the relevant verb) and are subsequently raised into spec-TP (in order to
satisfy the requirement for T to have a syntactic subject of its own). An interesting piece of evidence in
support of the claim that subjects originate internally within VP in adult English comes from so-called
floating quantifiers. In this connection, consider the following alternation:
(13)(a) All the students have passed the exam (b) The students have all passed the exam
In both sentences, the quantifier all modifies the students. Generally, a quantifier immediately
precedes the expression it modifies, as in (13a). But in (13b), the quantifier all is left floating in the
middle of the sentence, separated from the expression the students which it modifies? How does the
33
The term argument (of a verb) denotes an expression which is the ‘semantic subject’ or ‘semantic
complement’ of the verb.
quantifier all come to be separated from the students in (13b)? The answer given by Dominique
Sportiche (1988) is that the expression all the students originates as the subject of the verb phrase
[VP all the students passed the exam], and that subsequently this VP is merged (i.e. combined) with the
present tense T auxiliary have. The requirement for a T constituent (like have) to have a syntactic
subject of its own34 is satisfied either by moving all the students out of VP to become the subject of
have and thereby deriving (13a) All the students have passed the exam, or by moving the students on
its own to become the subject of have and so deriving (13b) The students have all passed the exam,
with all left stranded/floating in the subject position within VP.
On this view (known as the VP-Internal Subject Hypothesis/VPISH), subjects are initially
merged (i.e. positioned) internally within the verb phrase in adult grammars, and are subsequently
moved into subject position within TP in order to satisfy the requirement for T to have a subject. What
this means is that the VP-internal position of subjects in children’s initial clauses directly reflects the
initial position occupied by subjects in adult grammars (before the subject moves to occupy the
specifier position within TP). The prefunctional analysis of children’s initial clauses thus involves
essential structural continuity with adult grammars (since in both cases, subjects originate within the
verb phrase). More technically, we could say that children’s initial clauses are direct projections of
argument structure (e.g. of a verb with its subject and complement arguments).
An interesting piece of evidence that subjects are positioned within the verb phrase in child
grammars comes from the syntax of negation. In this connection, consider the structure of the
bracketed complement clause in an adult English sentence like (14) below:
(14) I suspect [CP [C that] [TP the students [T have] not [VP all [V passed] the exam]]
In (14), the students occupies the subject position within TP, and the floating quantifier all occupies
the subject position within VP. Of particular interest to us is the fact that the negative particle not
precedes the floating quantifier all which is the subject of VP, and follows the T constituent have. If
children’s initial clauses are VPs, and the subjects of those clauses are in spec-VP, then we should
expect to find that children’s earliest negative sentences will show not positioned in front of the
subject (because the subject is in spec-VP and so should follow the negative in the same way as all is
the subject of VP in (14) and follows not). And this does indeed appear to be case in sentences like
those below:
(15) No Mommy giving baby Sarah milk. Not Fraser read it. No Fraser drink all tea (Eve 1;9)
In each of the sentences in (15), the negative particle no(t) precedes the subject Mommy/Fraser –
precisely as we would expect if the negative is positioned immediately to the left of VP as in (14), and
if the subject occupies the specifier position within VP 35. Radford (1994) suggests that not in such
sentences may be a verb phrase adjunct – i.e. the kind of expression which can be attached to a verb
phrase to expand it into an even larger verb phrase. If so, a sentence like No Fraser drink all tea will
have a structure along the lines shown in simplified form below:
(16) [VP No [VP Fraser [V drink] all tea]
The overall structure would be a verb phrase headed by the verb drink, with all tea serving as the
complement of drink, Fraser serving as its subject/specifier, and no serving as an adjunct to the verb
phrase Fraser drink all tea.

3.3 Absence of C in children’s initial clauses

34
In recent theoretical work, Chomsky has suggested that this is because T has an [EPP] feature requiring it to
have an Extended Phrasal Projection into a TP containing an expression carrying a person feature as its
subject/specifier. Since both the students and all the students are third person expressions, making either of them
into the subject of have satisfies this requirement.
35
Déprez and Pierce (1994, p. 61) report that of the earliest negative sentences produced
by Eve at ages 18-21 months, Peter at 23-25 months and Nina at age 23-25 months, 96%
(71/74) contained sentence-initial negatives.
 If (as the structure-building model claims) the earliest syntactic structures produced by
children contain no functional categories, then we should expect to find that not only do
the earliest clauses produced by one-year-olds contain no T constituent, but also that they
contain no C constituent. In this connection, it is interesting to note that the complement
clauses produced by one-year-old children don’t contain complementiser/C constituents
like that/if/for36. Typical examples of complement clauses produced by one-year-olds are
bracketed below:
(17)(a) Want [Teddy drink] (Daniel 1;7) (b) Want [baby talking] (Hayley 1;8)
(c) Want [have money] (Daniel 1;7) (d) Want [car out] (Daniel 1;9)
The bracketed complement clauses in (17a-c) seem to be simple verb phrases containing a non-finite
verb (and no T or C constituent), while the complement in (17d) appears to be a (seemingly verbless)
small clause. The fact that one-year-olds do not produce complement clauses containing an overt
complementiser suggests that their clauses are prefunctional in nature and hence contain no C
constituent.
Moreover, children’s initial yes-no questions typically do not show auxiliary inversion – as we see
from the examples below:
(18) Neil sit? Neil sit that? Man taste it? Mommy read? Mommy turn? (= ‘Can Mommy turn it?’)
Come down? Sit down? Read? Mommy celery? (= ‘Does Mommy want celery?’) That
Mommy eggnog? (= ‘Is that Mommy’s eggnog?’) [Eve 1;6]
In adult English, yes-no questions involve auxiliary inversion – i.e. movement of an auxiliary from T
to C. Accordingly, the adult counterpart of Mommy turn would be a sentence such as Can Mummy turn
it? which involves the arrowed T-to-C movement operation shown below (with can adjoining to an
abstract question particle Q):
(19) [CP [C Can+Q] [TP Mummy [T can] [VP [V turn] it]]

But Eve’s wh-question Mommy turn? shows no auxiliary inversion – and indeed contains no auxiliary
at all. Why should this be? The answer given by the structure-building model is that children’s initial
clauses are simple verb phrases, so that Eve’s utterance Mommy turn has the structure shown in
simplified form in (20) below (where ø denotes a null constituent which is the child’s counterpart of
the adult pronoun it):
(20) [VP Mommy [V turn] ø]
The absence of auxiliary inversion (i.e. T-to-C movement) can be accounted for within the structure-
building model by supposing that children’s earliest sentences are simple VPs, and hence (by virtue of
containing neither a T constituent nor a C constituent) do not contain an inverted auxiliary (i.e. an
auxiliary which moves from T to C). The interrogative force of sentences like (20) is marked simply
by intonation37.
A question which arises if children’s initial clauses are simple VPs is how we account for early
child wh-questions such as the following:
(21) Where Papa go? Where put? (= ‘Where did you put them?’) Where stool? Where Eve pencil?
Where crayon? Who come? What happen? What doing, Mommy? (Eve 1;8-1;9)
One answer consistent with the structure-building model is to suppose that in a child wh-question like
Where Papa go? the wh-word originates as the complement of the verb go and is subsequently
preposed to a position within the verb phrase in front of the subject Papa. Given the assumption made
by Chomsky in recent work that heads allow multiple specifiers, we can suppose that Papa is the first
(inner) specifier of the verb go, and that where is its second (outer) specifier, so that Where Papa go?

36
These items function as complementisers in sentences like I admitted that I was wrong, I doubt if she is
coming, She’s keen for him to be there.
37
If the force of an utterance is marked by a feature carried by the head of the overall structure, the head V of the
overall VP will carry an interrogative force feature.
is an extended verb phrase which involves the arrowed wh-movement operation shown below:

(22) [VP Where Papa [V go] where]

(with where indicating the original position occupied by where before it moved).
At first sight, an analysis along the lines of (22) might seem to involve an intrinsically undesirable
discontinuity between child and adult grammars, since wh-words are assumed to move into the
specifier position within CP in adult grammars, but to move into the specifier position within VP in
children’s initial grammars. However, one way of countering this objection is to suppose that
wh-movement in adult and child grammars alike involves moving a wh-expression into the specifier
position within the highest clausal constituent in the sentence – this being CP in adult English, and VP
in child English. Equivalently, we might suppose that a wh-word has to move to occupy clause-initial
position (perhaps because only a clause beginning with a wh-word is interpreted as a wh-question in
English): if clauses are CPs in adult English but VPs in early child English, it therefore follows that
wh-words will move to spec-CP in adult English, but to spec-VP in child English. Moreover, it should
be noted that in his (1986) Barriers monograph and in more recent (1998, 1999, 2001, 2004) work on
Minimalism, Chomsky has argued that a wh-expression contained within VP first moves to the
outermost specifier position within VP before moving on to occupy the specifier position within CP:
see Radford 2004, chapter 10 for an account in terms of phases38. This means that the spec-VP analysis
of child wh-questions is far from implausible from a theoretical point of view.

3.4 Children’s initial nominals

Having looked at the structure of children’s initial clauses, we now turn to look at the
structure of children’s initial nominals – i.e. at the structure of the earliest noun and pronoun
expressions which they produce. Theoretical work in syntax over the past 3 decades (e.g. by Abney
1987 and Longobardi 1994) has argued that argument noun expressions (i.e. noun expressions used as
subjects or complements) in adult grammars have the status of DP/Determiner Phrase constituents, so
that in a sentence such as (23a) below the italicised noun expressions are DPs which comprise a
determiner (viz. the indefinite article a or the definite article the) modifying a noun like
earthquake/houses), and so have the respective structures shown in (23b,c):
(23)(a) An earthquake had destroyed the houses
(b) [DP [D an] [N earthquake]] (c) [DP [D the] [N houses]]
If (as the structure-building analysis claims), children’s initial syntactic structures lack functional
categories, we would expect to find that children’s initial argument noun expressions lack the DP layer
found in their adult counterparts. In this connection, consider the structure of the italicised noun
expressions in (24) below:
(24)(a) Hayley draw boat. Turn page (Hayley 1;8)
(b) Dog barking. Want ball (Bethan 1;9)
(c) Open can. Man drive truck (Allison 1;10)
Each of the italicised expressions is a singular count noun used as a (subject or complement)
argument of the associated verb, and would correspond to a DP headed by a singular determiner (like
a/the) in adult English – so that e.g. the adult counterpart of Man drive truck would be The man drives
a truck. The fact that one-year-old children omit determiners in obligatory contexts (i.e. contexts
where they are obligatory in adult grammars) is consistent with the assumption that children’s initial
nominals contain no determiners and hence are not DPs (expressions like those italicised in (24) being
38
A potential technical complication is posed by the fact that Chomsky takes transitive (but not intransitive) verb
phrases to be phases, and hence to trigger movement to an outer specifier position in the verb phrase. However,
if root projections/structures are phases, a sentence which comprises an intransitive verb phrase could be argued
to be a phase, and hence to allow wh-movement to a specifier position within the verb phrase. Moreover, Legate
(2003) and Radford (2004b) argue that wh-phrases move to the edge of the verb phrase even in intransitive
clauses. The relevant theoretical issues are highly complex, and it would not be appropriate to delve further into
them here.
simple nouns).
Note that it doesn’t seem to be the case that production constraints on the length of structures they
are able to produce prevent children at this stage from modifying nouns, since they produce nominals
in which a noun is modified by one or more (italicised) adjectives, as we see from the examples
below:
(25) Blue pen. Brown horsey. Black hair. Nice fire. Big castle. Little cup. High mountain. Hot bath.
Naughty cow. Bad pen. Dirty boots. Big heavy book (Jem 1;9)
The fact that one-year-old children produce adjective+noun structures but not determiner+noun
structures is consistent with the central claim made by the prefunctional analysis that children’s
earliest nominals contain lexical categories (like nouns and adjectives) but not functional categories
(like determiners).
In addition to lacking the functional category determiner, early child nominals also lack other
functors, including case particles. English has two types of genitive case particle, one being the
so-called Anglo-Saxon genitive morpheme ’s found in structures such as the book’s cover and the other
being the so-called Norman case particle of found in structures like the cover of the book.
However, the earliest nominals produced by one-year-old children show no use of genitive ’s in
obligatory contexts – as the following examples illustrate, where the italicised possessor is a bare
nominal lacking the genitive ’s suffix which would be required in adult English:
(26) Mommy haircurl(er); mommy cottage cheese; mommy milk; mommy
hangnail; mommy
vegetable; mommy pigtail; mommy sock; mommy slipper; Kathryn sock;
Kathryn shoe;
Wendy cottage cheese; baby cottage cheese; cat cottage cheese;
Jocelyn cheek; baby milk;
tiger tail; sheep ear (Kathryn 1;9, from Bloom 1970).
Likewise, de Villiers and de Villiers (1973) report 0% use of genitive ’s by the two one-year-olds for
whom they scored this in their study.
In much the same way, one-year-old children’s initial nominals show no use of genitive of in
obligatory contexts, as we see from the fact that the italicised bare nominals in the examples below are
not preceded by the genitive case particle of which would be required in adult English:
(27)(a) Want piece bar (Daniel 1;8) (b) Want cup tea (Jenny 1;9)
(c) Eve piece celery (Eve 1;9) (d) Cup tea ready (John 1;10)
If we follow Abney (1987) in supposing that genitive morphemes like ’s and of belong to the
functional category K of case particle, the conclusion we can draw is that the earliest noun expressions
produced by young children are lexical projections (i.e. structures formed out of lexical categories like
nouns) which lack functional categories like D/determiner and K/case particle.
Additional evidence relating to the absence of functional categories in children’s initial grammars
comes from the observation made by Bloom et al (1978) that one-year-olds have a nominal speech
style and use noun expressions in contexts where adults use personal pronouns like I/you/he etc.
Examples of this nominal speech style are given below:
(28) Baby eat cookies. Help baby. Put away Allison bag. Mommy eat cookie. Get Mommy cookie.
Eat Mommy cookie (Allison 1;10)

When referring to herself, Allison does not use the personal pronoun I but rather nouns like baby and
Allison. When talking and referring to her mother, she does not use the personal pronoun you but
rather the noun Mommy. What is the significance of the fact that Alison does not use personal
pronouns but instead uses nouns? The answer is that personal pronouns are function words which
encode grammatical properties such as person, number, gender and case (he for example being a third
person singular masculine nominative pronoun): in recent work, personal pronouns are analysed as
belonging to the category D with two uses, as we can see from sentences like You’re all the same, you
linguists, where the first occurrence of you is a pronoun (more precisely a D-pronoun) and the second
is a determiner modifying the noun linguists). Given that personal pronouns are determiners,
the fact that one-year-old children like Allison use nouns in contexts where adults use personal
pronouns is consistent with their being at a prefunctional stage where they have not yet acquired
functional categories like D (and so have not acquired D-pronouns like I/you/he etc.)
However, although children at the prefunctional stage lack personal pronouns (= D-pronouns),
they do make use of the pronoun one, as the following example illustrates:
(29) Nice yellow pen, nice one (Jem, 1;11)
What is the status of one in such sentences? The answer is that one is the kind of pronoun which
replaces a count noun (e.g. replacing the noun photo in a sentence like If you take a photo of me, I’ll
take another one of you) and so can be said to be an N-pronoun. Since an N-pronoun is a pronominal
noun and so belongs to the lexical category N/noun, we would expect to find that children at the
prefunctional stage have acquired one.
The idea that children at the prefunctional stage acquire N-pronouns like one offers us an
interesting account of their use of null subjects in sentences such as those below, where the italicised
verb has a null subject:
(30) (a) Want one. Gone out. Coming to rubbish (Bethan 1;8)
(b) Want Mummy come. Pee in potty (Jem 1;9)
(c) Find Mommy. Taste cereal (Kendall 1;10)
Radford (1990) suggests that the ‘missing’ subject in such sentences is a null noun which refers
directly to an entity in the domain of discourse (whether to the person speaking, the person being
spoken to, or some other person or thing). He argues that adults use a null noun to refer to an object in
the domain of discourse in certain circumstances – e.g. in the following example:
(31) Shall I take the red [N ø] or the blue [N ø]? (said by a woman looking at two dresses in a shop)
If children do indeed use null nouns to refer to entities in the domain of discourse, we should expect
these null nouns to occur not only as subjects, but also as objects. Indeed, we should expect to find that
both the subject and the object of a verb can be a null noun – as in the italicised example in the
sequence below:
(32) Danny want bar. Want (Daniel 1;10, reaching for a bar of chocolate)
Under the null-noun analysis, the two utterances produced by Daniel in (32) would have a parallel
structure, and would each be verb phrases containing a noun subject, a verb and a noun object – and so
would comprise the string (= sequence of constituents) shown in (33) below:
(33)(a) [N Danny] [V want] [N bar] (b) [N ø] [V want] [N ø]
The ‘understood’ subject and object nouns Danny and bar are not pronounced in the second utterance
in (32, 33) because they are mentioned in the first utterance and so taken to be implicit. Radford
(1990) offers a generalised account of children’s use of null constituents. He suggests that production
constraints (viz. a desire to minimise the complexity of the structures they produce) lead young
children to give a null spellout to redundant material (i.e. material which can readily be identified
from the discourse context). Radford claims that not just nouns but also other types of constituent can
be given a null spellout when their content is discourse identifiable. In this connection, consider the
reply given by Daniel (at age 1;8) to the question below put to him by an adult:
(34) ADULT: Who’s got the sweeties? DANIEL: Wayne sweeties
Since Daniel is at the prefunctional stage, we would not expect his reply to contain the T constituent
(ha)s: rather, we would expect him to produce the verb phrase Wayne got sweeties. But in his reply,
Daniel omits the verb got. Why? The answer would seem to be that got is ellipsed (i.e. given a null
spellout/omitted) by Daniel because it is mentioned in the previous utterance in the discourse, and so
can be taken to be implicit (i.e. identifiable from the discourse setting). A similar process of ellipsis
(i.e. omission of previously mentioned material) is found in adult English in co-ordinate sentences
such as Mary ate a banana and Jane ate an apple (with strikethrough indicating material which is
omitted because it is redundant).
 3.5 Explaining why lexical categories are acquired before functional categories
Given that the ultimate goal of any theory is explanation, a crucial
question which the structure-building analysis has to answer is why the earliest
syntactic structures produced by one-year-olds lack functional categories – e.g.
why children’s initial clauses are VPs which lack the TP and CP layers of functional
superstructure found in adult English clauses. One possible account (suggested
by Harald Clahsen in relation to his own work on the acquisition of German: see
e.g. Clahsen, Penke and Parodi 1994) is a lexical learning account which holds
that the structures produced by children at any given stage of development will
be projections of (i.e. will be formed by combining) the lexical items (i.e. words)
which they have acquired. If (as Radford 1990 argues) children acquire
contentives (i.e. content words) before functors, then (since content words
belong to lexical categories like noun, verb, adjective and preposition), it follows
that children’s initial syntactic structures will be lexical in nature, lacking any
functional superstructure. Of course, this account raises the question of why
contentives are acquired before functors: traditional answers given to this
question are that functors are late acquired because of their lack of
perceptual/acoustic salience (Gleitman and Wanner 1982, p.17), or their greater
cognitive/semantic complexity (Hyams 1986, p. 82), or their greater grammatical
complexity (Radford 1990, pp. 264-266), or the fact that they are subject to
parametric variation across languages (Chomsky 1989 proposed the Functional
Parametrisation Hypothesis under which languages differ in respect of the
range of functional categories that they employ). A further possibility is that
functors carry uninterpretable features, and these caused acquisition problems.
A second type of explanation (cf. Radford 1990, pp. 266-268) is a structural
(more specifically, teleological) one. We might argue that it is in the nature of
the grammatical structure being acquired that some parts of the structure must
be ‘in place’ before others can develop. Chomsky maintains that syntactic
structures are derived (i.e. formed) in a bottom-up fashion, so that (in deriving a
simple clause), the VP layer is built up first, then (in a negative sentence) the
NEGP layer, then the TP layer, and finally the CP layer (See the discussion of the
derivation of What has Jim not understood? in §1.3 for exemplification). It might
therefore be argued that (as a matter of conceptual necessity) children must first
learn to form VP before they can form TP, and must learn to form TP before they
can form CP. In other words, we might conclude that it is in the (bottom-up)
nature of the system that children will first acquire VP, then TP and then CP.
A third type of explanation which has been offered is a maturational one
(See e.g. Cinque 1988). Under this account, the limited processing capacity of
one-year-olds means that they only have the biological capability of forming
comparatively simple lexical structures. The onset of the functional period (at
around the time of their second birthday) coincides with dendritic development in
Broca’s area, and with an increase in neural connectivity (for relevant studies,
see Simonds and Scheibel 1989 and Greenfield 1991): this provides children with
the increased processing capacity required to acquire functional superstructure.
Wakefield and Wilcox (1995, p. 645) argue that the transition from the lexical to
the functional stage coincides with an increased production of myelin at around
the time of the child’s second birthday which ‘permits the efficient conduction of
nerve impulses over long distances, allowing precise integration of information
from widely separated regions of the brain’. If so, this suggests that there may be
a biological basis for the transition from the prefunctional to the functional stage
in language acquisition.

3.6 Questions about the Empirical Adequacy of the Structure-Building Model

Although the central idea of the structure-building model that children initially start out by
producing simple prefunctional structures and then gradually come to build more and more complex
layers of functional structure seems intuitively plausible, a number of questions about the adequacy of
the structure-building model have been raised by its critics. One type of objection (which we look at in
this section) is empirical in nature and questions the factual basis of the claim that one-year-old
English children initially go through a prefunctional stage at which they never use functors: a number
of studies have sought to show that children do indeed use functors (albeit infrequently) from the very
earliest syntactic structures they produce.
Radford (1990) acknowledges that alongside copula-less sentences like (35a) below, some
one-year-olds sporadically produce sentences like (35b) containing the contracted copula form ’s:
(35)(a) That spoon. (b) That’s Teddy (Daniel 1;8)
Critics of the prefunctional analysis would claim that if ’s is a third person singular present tense form
in (35b), and if T is the locus of tense and agreement inflections, then (35b) must contain a TP with the
highly simplified structure shown in (36) below:
(36) [TP That [T ’s] Teddy]
If this is so, and if UG incorporates Rizzi’s (2000) Categorial Uniformity Principle to the effect that
all expressions of the same type (e.g. all declarative clauses 39) belong to the same syntactic category
and have the same syntactic structure, then it follows that copula-less sentences like (35a) should be
analysed as containing a TP like that in (37) below:
(37) [TP That [T ø] Teddy]
with the T constituent receiving a null spellout (rather than being spelled out as (i)s) – perhaps because
of phonological problems posed by producing a triple-consonant cluster like tst, or perhaps because of
processing problems (e.g. problems in retrieving the irregular verb form is from the child’s mental
lexicon). On this view, there would be structural continuity between the child’s sentences in (35a, b)
and their adult counterparts, since both would have the same syntactic structure (and, more
particularly, both would contain a TP and hence be functional rather than prefunctional structures). It
is generally assumed in work on acquisition that (all things being equal) it is desirable to maximise
continuity between child and adult grammars, since any account of children’s sentences which
assumes substantial discontinuity between adult and child structures raises the question of why
children initially acquire the ‘wrong’ syntactic structure, and how they subsequently come to acquire
(and replace the ‘wrong’ structure by) the ‘right’ structure. Supporters of the alternative continuity
view of acquisition would claim that from the outset, children’s earliest multiword sentences have
essentially the same syntactic structure as the corresponding adult sentences, except that some functors
receive a null spellout in early child grammars (either because children have not learned how to spell
them out, or because they have problems in retrieving the appropriate inflected forms from their
lexicon, or because production problems make it difficult for them to pronounce certain sound
sequences).
The reply given by Radford (1990) to this type of objection is that since ’s is the only form of the
copula produced by children at the relevant stage (e.g. they do not also produce forms like am/are/was
etc.), ’s can be regarded as an invariable verb-form, and the structures in (35) can be analysed as
simple (pre-functional) verb phrases with the respective structures shown in (38) below:
(38)(a) [VP That [V ’s] Teddy] (b) [VP That [V ø] Teddy]
One question raised by this analysis, however, is why the copula should sometimes be overtly spelled
out as ’s, and why it should sometimes have a null spellout (though of course possible answers may lie
in retrieval problems, articulatory problems and so forth). An alternative possibility is to suppose that
overt-copula structures like That’s Teddy are indeed functional structures containing a T constituent,
and that the relevant child is in a transitional prefunctional-functional phase (between the
prefunctional and functional stages) during which he alternates between producing prefunctional VP
structures like (39)(a) below and functional TP structures like (39b):

39
A declarative clause is one used to make a statement: hence It is raining is a declarative (statement-making)
clause, whereas Is it raining? is an interrogative (question-asking) clause.
(39)(a) [VP That [V ø] Teddy] (b) [TP That [T ’s] Teddy]
This would in effect mean that during this transitional phase, children are bilingual and alternate
between using an earlier prefunctional grammar and a later functional grammar.
A further point to be noted is that the (1973) study by de Villiers and de Villiers showed that two of
the three one-year-old children in their study for whom they had sufficient data to score their
suppliance of the contracted copula form ’s showed 0% use of the copula, and the third scored only
3.5% use – suggesting that the claim that one-year-olds go through a prefunctional stage during which
they make no use of the copula ’s does indeed have empirical support40.
A parallel criticism of the prefunctional analysis of children’s initial nominals is presented by Abu-
Akel and Bailey (2000) – henceforth AAB. Basing their study on a corpus of 17 English children
studied longitudinally at 3-month intervals between ages 1;6 and 3;6, AAB argue that even in the
earliest corpora there is evidence of the children making sporadic use of the definite/indefinite articles
the/a (thus falsifying the claim that children initially go through a prefunctional stage during which
they make no use of functional categories like D/determiner). More specifically, they claim (ibid, p.54)
‘There is no stage at which children exclusively leave out all determiners in obligatory contexts’;
instead, they maintain that ‘The use of determiners in the early stages in optional’ and that ‘when
present, they are used correctly’ (ibid.). For example, AAB report a mean suppliance rate of 19% for
articles in obligatory contexts by the children in their study at age 1;6 (so that 19% of the time the
children would produce nominals like e.g. the/a car and 81% of the time they would produce bare
nominals like car). If we suppose that when the children produce a nominal like the/a car they are
producing a DP of the form (40a/b) below, considerations of structural symmetry might lead us to
suppose that when they say simply car, they are producing a DP of the form (40c):
(40)(a) [DP [D the] car] (b) [DP [D a] car] (c) [DP [D ø] car]
A number of factors might lead to the omission of the determiner in structures like (40c) – for
example, difficulty in retrieving function words from the lexicon, or a failure to specify the determiner
for the definiteness/indefiniteness feature marked by the use of the/a, or phonological production
problems with short reduced vowels like schwa (or interdental fricative consonants like đ) and so on.
However, AAB’s arguments are not entirely persuasive, for several reasons. For one thing, their
study shows that the mean use of articles gradually decreases as we go further back in the child’s
development, with the mean article suppliance rate being 67% at age 2;3, 40% at age 2;0, 32% at age
1;9 and 19% at age 1;6. This raises the question of whether if we looked at the same children (say)
three months earlier, their mean suppliance rate would have fallen to nil (0%). A second problem with
AAB’s study is that it is based on mean group scores (i.e. figures averaged across a group of 17
different children). But the problem with group means is they mask individual differences: for
example, a mean group score of 19% article suppliance could come about if some slow learners in the
group showed 0% article suppliance, and some fast learners showed around 40% article suppliance.
From a methodological viewpoint, therefore, the question which needs to be asked is whether
individual children show evidence of going through a prefunctional stage in their production of
nominals during which they have an article suppliance rate of 0% (i.e. they make no use of definite or
indefinite articles). This is not a question which AAB either ask or answer. However, Ohara (2001)
sought to answer this question by undertaking a quantitative analysis of the first four Allison Bloom
files on CHILDES when Allison was aged between 1;4 and 1;10. She found that Allison produced 548
singular count noun expressions, and not a single one contained a definite or indefinite article. This
provides strong evidence that Allison had not yet aquired the functional category D of determiner.
A further research finding which might be used to attempt to undermine the claim that children’s
initial noun expressions are prefunctional structures which lack functors is the observation by
Courtney Cazden (1986) that a detailed longitudinal study of a girl called Eve (from the Roger Brown
files on the CHILDES data base) showed that far from not using the genitive case suffix ’s at all in
possessive structures, Eve between ages 1;6 and 2;0 sporadically used genitive ’s in 13% of obligatory
contexts. The inference which opponents of the structure-building model would invite us to draw is
40
A further point to bear in mind is that some early uses of the copula e.g. in expressions like What’s that?
(which often gets reduced to osat, osa or even sa) may be routines – i.e. expressions which are memorised as a
single unit rather than decomposed into separate constituent parts.
that this provides us with strong empirical evidence there is no true prefunctional stage at which
functors like genitive ’s are entirely absent, but rather than such functors are sporadically used from the
outset. And yet, there is a crucial methodological weakness in the relevant type of argumentation.
Cazden’s figure of 13% is a mean score based on the study of a number of different (weekly)
recordings of Eve made at different stages of development. Vainikka (1994) estimates that Eve
undergoes the transition from the prefunctional to the functional stage at around age 1;8 (much earlier
than most children, since Eve is a fast developer). This being so, what Cazden is doing is combining
figures from the earlier prefunctional stage with figures from the later functional stage, and thereby
producing a mean score which has the effect of masking potential stage differences. In actual fact, if
we look at the first two files in the Eve corpus studied by Courtney Cazden (both files being recorded
when Eve was age 1;6), we find that Eve produced the range of non-imitative possessive nominals in
(41a) in file 1, and those in (40b)) in file 241:
(41)(a) Fraser coffee (x2). Fraser hat (x6). Eve pencil (x3). Eve seat (x3). Eve light. Eve horsie
(x7). Eve letter. Mommy book. Mommy pencil. Mommy letter. Dollie celery (x2). Dollie
shoe. Horsie eye (x4). Cromer briefcase. Lynn letter. Papa tray (x2). Papa breakfast.
(b) Eve lunch. Eve soup. Eve stool. Eve cracker. A Mommy soup. That Mommy soup. That
Mommy eggnog? Mommy stool (x2). Mommy sandwich. Mommy umbrella? (x2). Papa
umbrella? Papa cracker. Fraser coffee (x2). Cromer coffee
Of the 54 non-imitative possessive structures produced by Eve at 1;6, not a single one contained an
example of possessive ’s – consistent with the view that Eve has not yet acquired the functional
category K of case-particle. Ohara’s (2001) study of Allison from 1;4 to 1;10 likewise showed that
none of her possessive structures contained genitive ’s.
The overall conclusion to be drawn from our discussion here is that the issue of whether it is truly
the case that children’s initial syntactic structures lack functors cannot be resolved by quantitative
studies based on mean figures derived from aggregating scores from different children (or from
multiple recordings of the same child at different stages of development), since such studies may mask
individual and stage differences. Rather, only careful analysis of individual recordings of individual
children can provide us with an authoritative answer to the question of whether or not the
prefunctional stage exists.

3.7 Questions about the Universality of the Structure-Building Model

A further type of question which has been raised about the structure-building model
concerns its potential universality – in other words, whether empirical studies of the acquisition of
languages other than English provides evidence that one-year-old children go through a prefunctional
stage. The reason why it is important to look at languages other than English becomes immediately
obvious if we look at the morphosyntax of verbs in English. If an English child says Want apple, it is
unclear whether the verb want is a bare/uninflected infinitive form (not marked for tense or
agreement) or whether it is a first person singular present tense form – for the obvious reason that
infinitive and 1.SG. present verb forms are homophonous in English. Moreover, since main verbs in
English remain in situ (i.e. they never move from V to T), we can’t use syntactic evidence (from the
position of the verb in a clause) to determine whether a verb in English is finite or not. But if we look
at the acquisition of languages like French, Italian or Spanish where finite verbs are morphologically
and syntactically distinct from non-finite verbs (in that finite verbs carry different inflections from
non-finite verbs, and finite verbs raise to T whereas non-finite verbs remain in situ in the head V
position of VP), we have a much more transparent way of testing the claim made in the structure-
building model that the very earliest clauses produced by young children are uniformly non-finite.

41
Fraser/Cromer are surnames, so the corresponding adult form would be Mr Fraser’s/Mr Cromer’s. The
notation (x3) indicates that a particular structure occurred 3 times in the relevant file. It is a methodological
convention in acquisition research that imitative utterances (i.e. utterances which are exact repetitions of
something said by someone else) should be excluded in principle from any analysis of the utterances produced
by children, since they may be more of a reflection of the child’s ability to hold fragments of adult sentences in
their short term memory (and then repeat them in a parrot-like fashion) than of their syntactic abilities.
 Nina Hyams (1994, p. 22) argues that if we look at the acquisition of (e.g.
Romance or Germanic) languages with a richer verb morphology and verb
syntax, ‘We see that children acquire certain inflectional elements at a very early
age, from the beginning of their multiword utterances. Moreover, they control
syntactic operations such as verb raising and verb second (V2) 42, which are
dependent on the presence of functional heads’. Hyams’ claims would appear to
be borne out by numerous empirical studies arguing that in the initial stages of
the acquisition of other languages such as French (cf. e.g. Pierce 1992), German
(cf. e.g. Poeppel and Wexler 1993) and Italian (cf. e.g. Guasti 1992), one-year-old
children already differentiate finite from nonfinite verbs both in respect of their
morphology and in respect of their syntax (in that e.g. finite verbs are positioned
before negatives and nonfinite verbs after negatives, and clitic pronouns attach
to the left of finite verbs but to the right of nonfinite verbs).
A study of this ilk is Déprez & Pierce (1994), who report that one-year olds acquiring French
position finite (but not nonfinite) verbs in front of the negative particle pas ‘not’ (just like adults) – as
illustrated by the examples in (41) below:
(42)(a) Pas casser ‘Not break’ [Daniel 1;8] (b) Marche pas ‘Works not’ [Daniel 1;8]
In (42a), casser ‘break’ is a non-finite (infinitive) verb form and follows the negative particle pas ‘not’
– as would be expected if the overall structure were simply a negative verb phrase. But in (42b), the
verb marche ‘works’ is a finite (present-tense) form and precedes the negative particle pas ‘not’. How
does this come about? Déprez and Pierce argue that the verb+negative word-order in finite clauses like
(42b) comes about by virtue of the finite verb raising from the head V position of VP across the
intervening negative pas ‘not’ into the head T position in TP, in the manner shown in the partial clause
structure in (43) below:

(43) ...[T marche] pas [V marche]...

Déprez & Pierce conclude that sentences like (42) provide strong evidence that one-year-old French
children have acquired the morphological and syntactic differences between finite and non-finite
verbs, and that movement of finite verbs from V to T in structures like (43) provides clear empirical
evidence that their earliest clauses contain a T constituent (and hence cannot be prefunctional
structures).
However, a word of caution needs to be sounded. If we look at the original
study by Amy Pierce (1989) which provides the data for the Déprez and Pierce
study, we find the following observation made about one of the three French
children (Nathalie) in the study. In the very first recording of Nathalie at age 1;9;3
(designated by Pierce as N1/T1), Nathalie uses only nonfinite verb forms, not
finite verbs. Thus, Pierce (1989, p. 41) observes that ‘Nathalie at N1 lacks tensed
verbs’, and concludes that there is an ‘absence of inflected forms in the very
early stages’ (op. cit. p. 42), noting that ‘In Nathalie at T1, then, we catch a
glimpse of a stage in French acquisition before verb raising to tense sets in.’
Thus, children like Nathalie appear to provide us with a crucial ‘glimpse’ (to use
Pierce’s own word) of an earlier prefunctional stage. We might reach a similar
conclusion about Child Dutch on the basis of Wijnen et al (2001)’s claim that the
earliest utterances produced by a group of Dutch children they studied contained
only non-finite clauses. Adone’s (1993) study of the acquisition of Mauritian
Creole (a French-based creole in which finiteness is marked by
42
The assumption underlying this claim is that in finite main clauses in German, the verb may raise from V
through T into C, and hence come to occupy second position in the clause, following the specifier position within
CP (which is typically filled by a preposed subject, complement or adjunct): this is referred to as V2 since it
results in the verb being the second constituent in the clause.
tense/aspect/modality auxiliaries) argues that the youngest child in her study
(Laura, aged 1;9) produced no auxiliaries at all (omitting them 33 times in
obligatory contexts), leading Adone to the conclusion that Laura has ‘a syntax
without functional projections’ (Adone 1993, p. 8).
Further evidence of a prefunctional stage comes from a study of the
acquisition of Welsh by Aldridge, Borsley and Clack (1995). Welsh is a language
which generally has Verb+Subject word order, as in:
(44) Mi brynodd Gwyn geffyl
PRT bought Gwyn horse (PRT = clause-introducing particle)
‘Gwyn bought a horse’),
where Mi is an optional clause-introducing particle. One account of Welsh clause
structure is to suppose that the subject occupies initial position within the verb
phrase, but that VP is combined with a preceding T constituent which is filled
either by positioning an auxiliary directly in T, or by moving the main verb from V
to T, and that clause-initial particles are complementisers. On this view, (44) will
have a derivation along the lines of (45) below (with the arrow showing
movement of the main verb from V to T):
(45) [CP [C Mi] [TP [T brynoddbought] [VP Gwyn [V brynodd] geffylhorse]]]

A longitudinal study of a Welsh boy called Kevin between 1;6 and 2;6 by Aldridge,
Borsley and Clack (1995) reports that up to age 1;11, clauses produced by Kevin
were Subject+Verb(+Complement) structures like:
(46)(a) Dyn isda yfana (b) Mam cosi Lisa
Man sit there (Kevin 1;10) Mum tickle Lisa (Kevin
1;11)
During this period, Kevin produced no examples of Verb-Subject order and no
examples of auxiliaries or clause-introducing particles: verb-subject and auxiliary
structures only started to appear at age
2;0-2;1. Aldridge, Borsley and Clack conclude that the earliest clause structures
produced by Kevin are simple verb phrases which entirely lack functional
categories like T and C – so that sentences like (46a,b) have the respective
structures shown below:
(47)(a) [VP Dynman [V isdasit] yfanathere] (b) [VP MamMum [V cositickle] Lisa]
Their research leads them to the more general conclusion that the initial
structures produced by children acquiring Welsh are prefunctional.
Nevertheless, research like that of Déprez and Pierce on child French does appear to suggest that
functional categories like T may be acquired a few months earlier in languages like French than in
English. If this is true, why should it be? The answer may be that the morphosyntactic properties of T
are more visible in French than in English, firstly because French has a richer system of tense and
agreement inflections on finite verbs than English, and secondly because finite verbs move from V to
T in French (but remain in situ in English), and this movement is directly visible in relation to the
position of verbs with respect to verb-modifying adverbs – as we see from sentence pairs such as:
(48)(a) Marie really likes Paul (b) Marie aime bien Paul
Marie likes well Paul
Verb-modifying adverbs like those italicised in (48) are generally positioned at the left edge of the
verb phrase, and so (because verbs remain in situ in English), the verb likes follows the adverb really
in (48a). But the fact that the French verb aime ‘likes’ in (48b) is positioned in front of the adverb bien
‘well’ provides children with clear syntactic evidence of the existence of a T constituent in French,
since the assumption that the verb aime ‘likes’ moves from V to T provides a straightforward account
of how the verb aime comes to be positioned in front of the adverb bien which modifies it.
The more general conclusion which this might lead us to is that we should expect a functional
category like T to be acquired considerably earlier in languages in which T is a strong head which
triggers verb movement and in which T is associated with a rich system of tense/agreement affixes
than in a language like English in which T is syntactically weak and morphologically impoverished (T
often being spelled out as a null affix in English, as in sentences like I/We/You/They speak French,
where the verb speak carries a null affix marking its invisible tense and agreement properties).
Accordingly, it may well be that children’s initial clauses are universally VPs, and that they only
project VP into a larger TP structure once they have overwhelming evidence for positing a T
constituent. Because T has greater morphological and syntactic visibility in French than in English, the
transition between the VP stage and the TP stage would be expected to take place substantially earlier
in languages like French where functors are more visible (i.e. have a more transparent morphosyntax)
than in languages like English where they have a more opaque morphosyntax.

3.8 Theoretical questions raised by the prefunctional analysis

A third type of critical question which needs to be asked about the structure-building model
is whether the claim that children’s initial syntactic structures are pure projections of lexical categories
and entirely lack functional categories is tenable from a theoretical perspective. After all, many of the
structure-building analyses of children’s initial syntactic structures were developed in work dating
back to the late 1980s and were consequently based on a theoretical model of syntax developed more
than two decades ago, and hence it is reasonable to ask whether the assumptions made there are
compatible with more recent theoretical work in syntax.
As a case in point, consider theoretical work by Memo Cinque (1994, 1999) on the syntax of
modifiers (e.g. adjectives modifying nouns, and adverbs modifying verbs). Cinque posits that
principles of UG determine that the modification relation between a modifier and the expression which
it modifies is mediated by a functional head. On this view, an adjective which modifies a following
noun will serve as the specifier of a null functional head F which takes the noun as its complement and
the modifying adjective(s) as its specifier(s). This means that, a nominal like blue pen (produced by
Jem at 1;11) might have a structure along the lines shown in simplified form in (44) below:
(49) [FP blue [F ø] pen]
F here denotes a null functional head which has a noun complement (pen) and an adjectival specifier
(blue); F has the semantic property of expressing modification, so that (49) has an interpretation
paraphraseable as ‘pen which as the property of being blue’. An obvious implication of the analysis in
(49) is that early child nominals which contain an adjective modifying a noun must in principle
contain a null functor of some kind.
Much the same could be said of early child possessive nominals like Mommy milk in (26) above
which contain a possessor Mommy modifying a possessum/possessed noun milk. If we adopt an
approach to possessive modifiers which is parallel to that of Cinque (1994), possessive modification
will be mediated by a null functional head marking possession (below symbolised as Poss) which will
take the possessum as its complement and the possessor as its specifier. This will mean that a
possessive nominal like Fraser hat will have the (simplified) structure shown below:
(50) [PossP Fraser [Poss ø] hat]
The overall structure will thus be a PossP ‘Possessive Phrase’ which comprises a Poss constituent
whose complement is the possessum hat and whose specifier is the possessor Fraser. A structure like
(50) would appear to be incompatible with the claim that early child clauses are prefunctional
structures, since (50) contains the null possession-marking functor Poss.
The logic of the argumentation presented here is that if UG principles require modification to be
mediated by a functional head, and if children’s initial grammars obey UG principles, then adjective-
modified nominals like (49) and possessor-modified nominals like (50) must of necessity contain an
abstract functor to link the modifier to the expression which it modifies (e.g. to link the the adjective
blue to the noun pen which it modifies, and to link the possessor Fraser to the possessum hat). It
would then follow that theoretical considerations make it make it untenable to maintain that children
go through an initial prefunctional stage when their utterances contain no functors of any kind.
 There are several types of response which can be given to this kind of theoretical objection to the
prefunctional analysis. One is that alternative theoretical analyses of modifiers are available which
don’t require the postulation of abstract functional categories (e.g. treating adjectives as adjuncts to the
noun expressions which they modify, as in Radford 1990, and likewise treating possessors as nominal
adjuncts as in Adger 2003). Another is to question the extent to which (even if we adopt an analysis of
modification in which the modifier is the specifier and the modified expression the complement of a
null head), the head which mediates the modification relation can truly be said to be a functor. In this
connection, consider the function of the Poss head in a structure like (50) [PossP Fraser [Poss ø] hat]. The
essential function of Poss in a structure like (50) is to mark the noun Fraser as having the semantic
role of a POSSESSOR, and the noun hat as having the semantic role of POSSESSUM. Such semantic roles
have been termed thematic roles in work since the 1960s, and the general assumption which has been
made is that only lexical heads can assign a thematic role to their complements and specifiers, not
functional heads (e.g. verbs but not auxiliaries). If the only role fulfilled by Poss in a structure like
(50) is to mark the possession relation between the POSSESSOR and the POSSESSUM, and if Poss is not
responsible for the assignment of functional features (e.g. does not assign genitive case to the
possessor Fraser), it is hard to see how Poss can be taken to be a true functor.
A crucial question which the discussion here raises concerns the defining characteristics of
functors. One possibility is to relate this to the distinction drawn in Chomsky (1995) between
interpretable features (i.e. features which contribute to determining both the form and meaning of a
word) and uninterpretable features (i.e. features which contribute to determining form but not
meaning). We might then define functors as items which carry uninterpretable features. From this
perspective, Poss would not be a functional head because the only feature it carries is an interpretable
feature marking possession (not e.g. an uninterpretable feature marking genitive case) 43. What this
would then suggest is that the initial syntactic structures produced by children at the prefunctional
stage contain only items carrying interpretable features, and no items carrying uninterpretable features.
So, for example, an early child utterance like Doggies chasing man car (‘The doggies are chasing the
man’s car’) would comprise the noun doggy (carrying an interpretable plural-number feature), the verb
chasing (carrying an interpretable progressive-aspect feature), the nouns man and car (each of which
carries an interpretable singular-number feature), and an abstract Poss morpheme (carrying an
interpretable feature marking possession). The auxiliary are would be missing (since although it
encodes an interpretable tense feature, it also encodes uninterpretable agreement features), and so too
would the genitive case particle ’s (case being an uninterpretable feature).
The idea that the prefunctional stage is in fact a stage at which structures contain only items
carrying interpretable features provides an interesting perspective on the observation made in Radford
(1990) that many one-year-olds who otherwise seem to be at the prefunctional stage nonetheless use
personal pronouns. But as Radford notes, the earliest pronouns they seem to use are accusatives (not
nominatives), and they use accusative pronouns not only as objects (e.g. as in Help me!) but also as
subjects – as the examples below illustrate:
(51) Me got bean (Stefan 1;5). Me do (Helen 1;6). Me talk. Me look (Stephen 1;7)
Me do it. Me do. Me come in. Me want. Me help. Me get it (Bethan 1;8)

Radford maintains that the accusative form of the pronoun is the only form used by the children at the
relevant stage, and hence that the relevant pronouns are caseless forms (because there is no case
contrast between accusative pronouns like me and nominative pronouns like I if the children
concerned have not acquired nominatives like I). If so, this means that the relevant pronouns carry
only interpretable person/number (and in the case of him/her/it gender) features, but no uninterpretable
case feature. If we assume that it is in the nature of functors that they carry uninterpretable features, it
follows that these early proto-pronouns used by children at the prefunctional stage are not true
functors (because they do not carry the uninterpretable case feature carried by functor pronouns).

43
Much the same reply ould be given to any claim that sentences containing a verb inflected for progressive or
perfect aspect (as with utterances like Mummy doing dinner or Daddy taken biscuit produced by Daniel at age
1;10) must in principle contain an AspP/Aspect Phrase projection. If such a projection exists, it would seem to
carry only interpretable features.
 A further type of structure which might be thought to pose theoretical problems for the
prefunctional analysis are intransitive SUBJECT+VERB structures such as the following (from Radford
1990, p.148):
(52)(a) Him gone (Hayley 1;8)
(b) Daddy come. Daddy coming (Helen 1;9)
(c) Mummy come. Mummy coming (Jem 1;9)
(d) Biscuit gone. Hammer gone. Car gone (Angharad 1;10)
Come and go belong to a special class of verbs known as unaccusatives (See §7.5 of Radford 2004a,
or §7.6 of Radford 2004b for a description of their properties). They are so called because their
apparent subjects originate as their complements, and are not assigned the accusative case which the
complement of a verb would normally be expected to receive – as can be seen from Icelandic
structures like (53) below (since Icelandic is a language in which quantifiers and nouns overtly inflect
for nominative case):
(53) Þad hafa komið nokkrirNOM gestirNOM
There have come some guests
The fact that the italicised expression follows the verb in structures like (53) suggests that the subjects
of unaccusative verb orginate as their complements. Accordingly, if child structures like Daddy
coming were simple verb phrases (as the prefunctional analysis claims), we might expect to find
children at the prefunctional stage producing verb-initial (verb+complement) structures like:
(54) [VP [V Coming] Daddy]
in which the noun Daddy follows the unacusative verb coming by virtue of being its complement. But
instead (as we see from examples like those in (52) above), they typically produce verb-final structures
like Daddy coming. Proponents of the functional analysis would argue that such verb-final structures
can only be the result of the complement of the verb being raised into the specifier position in TP, in
the manner shown by the arrow below:
(55) [TP Daddy [T ø ] [VP [V coming] Daddy]

And an analysis along the lines of (55) would clearly require us to suppose that the children’s
unaccusative clauses contain a TP constituent, with Daddy moving from complement position within
the VP to subject/specifier position within TP. Such an assumption would clearly be incompatible with
the idea that children’s initial clauses are prefunctional VPs.
The key claim being made here is that the prefunctional analysis cannot provide a principled
account of subject-verb word order in unaccusative clauses like those in (52). But is this really true?
After all, suppose that children develop a linearization condition (i.e. word order rule) such as the
following:
(56) The highest argument44 of a verb is positioned in front of the verb, and other arguments after it
What (56) would predict in relation to an unaccusative structure like (50) above is that because Daddy
is the only (and hence the highest) argument of the verb coming, it must be linearised (i.e. positioned)
in front of the verb, so that the structure the child produces is not (55) above, but rather (57) below:
(57) [VP Daddy [V coming]]
And if we define the subject of a verb as its highest argument, we can then say that Daddy in (57) is
the subject of the verb coming. In short, there is no reason to suppose that the prefunctional analysis
cannot provide a principled account of the Subject+Verb word-order typically found in children’s
unaccusative clauses.

44
The question of which argument of a verb is projected as its highest argument (and hence subject) is widely
thought to be determined by a thematic hierarchy of some kind – e.g. an AGENT argument (if there is one) will be
projected higher than a THEME/PATIENT argument, which in turn will be projected higher than a GOAL argument
(and so on): see Radford (2004a, p.377) for one such hierarchy.
 It may seem confusing to find that different studies come to different conclusions about whether
children’s initial syntactic structures do or don’t contain functional categories – and the obvious
question to ask is why this should be. The reason is that there is a dearth of intensive longitudinal
studies of very young children (from the moment they produce their first one-word utterances at
around the time of their first birthday), and you might ask why this should be so. The answer is that it
is extremely difficult to get reliable data on the speech production of one-year-olds. On the one hand,
they are not suitable subjects for traditional experimental studies (They are too young for
experimenters to perform e.g. elicitation tasks or grammaticality judgment tasks on them), while on
the other they are far from ideal subjects for naturalistic studies (since very young children tend to say
very little, and much of what they do say may be unintelligible). What we need are more dense corpus
studies of children’s speech production from the onset of the one-word stage on (a dense corpus study
being one which records several hours of a child’s speech output every day over one or two years).

Workbook section
3.1 Kendall’s initial clauses
In Appendix C to her (1973) book, Melissa Bowerman reports that a girl called Kendall produced
utterances such as the following (at age 1;11) in contexts where adults would use a clause of the form
Subject+Verb+Complement. (Items written as a single word are claimed by Bowerman to be treated by
Kendall as single words. Parentheses provide contextual information, or show Bowerman’s paraphrase
of the child’s intended meaning in single inverted commas.)
1. Mommy sew doggy 2. Kimmy ride bike 3. Kendall turn page 4. Kendall break
5. Kendall see Kendall (looking at a picture of herself). 6. Ben swim pool
7. Mommy lady (response to ‘What’s Mommy?’) 8. Kendall bath (‘Kendall takes a bath’)
9. Kendall book (‘Kendall reads a book’) 10. Kendall spider (‘Kendall looked at a spider’)
11. Read book 12. Bite finger 13. Open lotion 14. Doggy sew (‘Sew doggy’)
15. Kimmy kick (‘Kick Kimmy’) 16. Kendall pickup (‘Pick up Kendall’)
17. Doggie lookit (‘look at doggy) 18. Hug Mommy (‘Mommy hugs’)
19. See Kendall (‘Kendall sees’) 20. Mommy hit Kendall (‘Kendall hit Mommy’)
21. Where doggy go? 22. Where pillow go?
To what extent are such utterances consistent with the prefunctional analysis of early child English?
Since Complement+Verb+Subject word-order in main clauses in languages like German has been
argued to be the result of the verb moving from V through T into C (and the complement moving to
spec-CP), to what extent do sentences like 20 provide support for positing functional projections in
Kendall’s grammar? Consider an altermative prefunctional analysis of sentences like 20 on which
some children at the prefunctional stage have not yet learned some of the one of the word-order rules
in English (one such rule being that complements/objects are positioned after heads, and another that
specifiers/subjects are positioned on the opposite side of heads from complements).

3.2 Eve’s initial clause structures

The first file in the Eve corpus on the CHILDES data-base shows Eve (at age 1;6) producing a range
of clauses with overt subjects including the following:
1. Doll eat celery 2. Mommy fix (Mother replies ‘Want me to fix it?’). 3. Papa have it (Mother
replies ‘Papa had it’) 4. That radio (response to mother asking ‘What is that?’). 5. Come car.
6. Car coming. 7. Fish are swimming (reply to ‘What’re the fish doing?) 8. Man no taste it.
9. Block broke. 10. I did it
Discuss the extent to which the relevant utterances are consistent with the claim that one-year-olds go
through an initial prefunctional stage.

3.3 Eve’s earliest infinitive structures

The utterances below represent the earliest examples of potential (italicised) infinitive clauses in the
first 5 Eve files on the CHILDES data-base.
(a) Want watch (file 2, 1;6). (b) I want Mommy read (file 2, 1;6)
(c) I don’t want to (file 3, 1;7) (d) Want Mommy read (file 3, 1;7)
(e) Want swing (file 4, 1;7) (f) Want crack nut (file 4, 1;7)
(g) To help (<In reply to the question: ‘Well, what do you want me to do?’ - file 4, 1;7)
(h) Want find top (file 5, 1;8)
Discuss the potential significance of such examples for the structure-building model.

3.4 Eve’s initial wh-questions

Alongside wh-questions like those in (21) in the main text, Eve (at ages 1;8-1;9) also produced
wh-questions like those below in which the wh-word and/or the subject appear to be missing:
What doing, Mommy? People doing? Doing, Mama? (Eve 1;8-1;9)
Luigi Rizzi (2000) argues that many adult languages allow root specifiers (i.e. constituents which are
the specifiers of the highest phrase/projection in a given tree structure) to have a null spellout under
certain conditions, and that similar phenomenon is found in child grammars. Discuss the possible
implications of Rizzi’s analysis for the prefunctional analysis of wh-questions like those above. In
analysing the relevant questions, ignore the vocative expressions Mommy/Mama.

3.5 Garman et al’s study

Garman et al (2004) report on the linguistic development of a one-year-old English-Latvian bilingual
child called Maija. They report that she produced the following English verb forms each month (the
superscript number indicating the number of times each form was produced during the relevant
recording sessions, and the items being listed in alphabetical order):
1;4 = get1;
1;5 = come3; eating1; gone1; sit1;
1;6 = come2; go1; play1; sit4;
1;7 = are2; come1; doing1; get1; (i)s3; look3;
1;8 = come4; coming1; eat1; go2; got3; (i)s9; look2; open2; opening1; play1; playing3; sleeping1;
1;9 = get1; (i)s5; doing2 eating2;
1;10 = are1; do2; eating1; go4; goes5; (i)s5; looking3; sit5; sitting1; sleeping1;
1;11 = come3; do2; eating1; gone1; going1; (i)s23; look2; play1; playing3; sit2; sleep1; was1;
Discuss the possible implications of the relevant data for the structure-building model.
3.6 Children’s initial negatives
Since work by Pollock (1989) on the syntax of negation, it has generally been assumed that a negative
clause like the that-clause italicised in ‘I feel that he is not telling the truth’ would have a structure
along the lines shown in simplified form below (with strikethrough indicating that he originates as the
subject of the verb telling, before being raised up to become the subject of is):
[CP [C that] [TP he [T is] [NEGP not [NEG ø] [VP he [V telling] the truth]]]]
On this view, negation of a verb phrase is mediated by an abstract NEG head whose complement is the
following verb phrase and whose specifier is the negative particle not. Discuss the potential
implications of Pollock’s work for the analysis of children’s negatives like No Fraser drink all tea
presented in the main text.

3.7 Stromswold’s analysis of children’s initial negatives

Karin Stromswold (1996) suggested that early child negative sentences like No lamb have it (produced
by Nina at age 2;0) have the structure shown in simplified form below (in which no is taken to be a
negative complementiser, i.e. a clause-type-marking particle):
[CP [C No] [TP lamb [T ø] [VP [V have] it]]]
Discuss the implications of Stromswold’s analysis, and highlight any potential weaknesses.

3.8 Children’s medial negatives

Alongside initial negatives (i.e. sentences in which the negative particle is the first word in the
sentence, one-year-olds are also reported to produce medial negatives (i.e. sentences in which the
negative is in the middle of the sentence and follows the subject) like those below:
Man no go in there. Kathryn no fix this. Kathryn no like celery (Kathryn 1;10)
Wayne not eating it. Wayne not eat it (Daniel 1;11)
Radford (1990) claims that children at the prefunctional stage treat no(t) as an item which can be
adjoined to any V-projection (i.e. verbal expression), hence to either a VP or a V-bar 45. What problems
(if any) might such an analysis pose? What would be the implications of extending Pollock’s NEGP
analysis of negation to such structures?

4. The Truncation Model

4.1 Introduction
In work spanning almost a decade, Luigi Rizzi (1992, 1994a/b, 2000) has argued that young
children go through a stage (generally lasting until around their third birthday, but in some cases
lasting rather longer) in which they alternate between producing full adult-like functional structures
(e.g. clauses containing CP and TP) and truncated structures in which they omit one of more of the
layers of functional superstructure found in the corresponding adult structures (either the CP layer of
clause structure, or both the CP and TP layers): for obvious reasons, this has become known as the
Truncation Model. Before introducing the truncation model, let’s clarify the assumptions made by
Rizzi about the nature of syntactic structure in adult grammars.

4.2 Assumptions which Rizzi makes about adult structures

In line with recent work by Chomsky, Rizzi assumes that all canonical (i.e. ‘normal’) adult
45
Within the split projection analysis of the syntax of verb phrases proposed by Larson (1988, 1990), Hale and
Keyser (1991, 1993, 1994) and Chomsky (1995), the negative particle no(t) could alternatively be taken to
modify either the lower lexical verb phrase or the upper light verb phrase – however, I shall not go into relevant
technical detail here.
clauses are CPs. This is a traditional assumption about the syntax of interrogative clauses, as we can
illustrate in relation to an adult wh-question like:
(1) Which lecture might you not attend?
Given Chomsky’s assumption that sentences are derived (i.e. formed) in a bottom-up fashion, sentence
(1) will have the following derivation (simplified for expository purposes). The verb attend is merged
(i.e. ‘combined’) with its complement which lecture and its subject you to form the verb phrase
[VP you [V attend] which lecture]. Under the analysis of negation in Pollock (1989), an abstract NEG
(= negation) head then merges with the resulting VP, and with the particle not, forming the NEGATION
PHRASE/NEGP constituent [NEGP not [NEG ø] [VP you [V attend] which lecture]], in which the verb phrase
is the complement of NEG, and not is its specifier. The resulting NEGP is then merged with the (past
tense) T constituent might46, and since every T constituent requires a syntactic subject of its own,
[T might] attracts the pronoun you to become the subject of might, so forming the TP [TP You [T might]
[NEGP not [NEG ø] [VP you [V attend] which lecture]]], with strikethrough indicating the position occupied
by a constituent before it moved (in this case, showing that you was the subject of attend before
becoming the subject of might). The resulting TP is then merged with an interrogative C
(= clause-type-marking constituent, containing an abstract question particle Q which marks the clause
as interrogative in type) and this attracts the (past tense) T constituent might to move from T to C, and
also attracts the wh-expression which lecture to move into the specifier position within CP, so deriving
the CP structure shown below:
(2) [CP Which lecture [C might+Q] [TP you [T might] [NEGP not [NEG ø] [VP you [V attend] which lecture]]]]
As is self-evident, the resulting clause has a CP+TP+(NEGP+)VP structure (the NEGP constituent
here being placed in parentheses to show that it is only present in negative clauses, so that non-
negative questions have a CP+TP+VP structure).
Rizzi further supposes that all declarative clauses (like that italicised in (3) below) are likewise
CPs:
(3) Don’t pretend that there have not been any problems!47
The italicised clause in (3a) is a complement clause which is declarative in type (in that it serves to
mark a statement) and is a CP with the structure shown in simplified form below:
(4) [CP [C that] [TP there [T have] [NEP not [NEG ø] [VP [V been] any problems]]]
In other words, a declarative clause like (4) has essentially the same CP+TP+(NEGP+)VP structure as
an interrogative clause like (2) (with NEGP again only being present in negative clauses, non-negative
clauses having a simpler CP+TP+VP structure).
But what of declarative main clauses like that below?
(5) There have not been any problems
Rizzi (2000) posits that principles of Universal Grammar determine the structure of clauses, and that
one such principle is the following:
(6) Categorial Uniformity Principle
All expressions of the same type belong to the same category
This principle implies that every declarative clause in a given language must have the same categorial
status (and structure) as every other declarative clause. Hence, since declarative complement clauses
like that italicised in (3) are CPs, declarative main clauses like that in (5) must also be CPs. This being
so, (4b) will have the structure (7) below:
(7) [CP [C ø] [TP there [T have] [NEP not [NEG ø] [VP [V been] any problems]]]

46
From a morphological point of view, might is the past tense of may, and contains the same past-tense -t suffix
that we find in irregular past tense verbs like went, sent, bought, caught etc.
47
There are two clauses in (3): the main (don’t! clause) is imperative in type, and the complement (that) clause
is declarative.
The overall clause structure will thus be a CP headed by a null clause-type-marking C constituent
which serves to mark the clause as being declarative in type. The CP analysis of main-clause
declaratives in (7) (which is also adopted in recent work by Chomsky) is made all the more plausible
by the fact that there are languages like Arabic which use an overt clause-type-marking C constituent
to introduce main declarative clauses in sentences such as:
(8) ?inna lwalada qad taraka lbayta
That the.boy did leave the.house (= ‘The boy left the house’)
Since interrogatives are also CPs, this leads to the greater generalisation that all canonical (i.e.
‘normal’) clauses are CPs in adult English48 - an assumption which both Rizzi and Chomsky make.

4.3 Truncation in child clauses

Rizzi posits that young children sometimes project functional projections (like CP and TP)
and sometimes truncate (i.e. omit) them. So, when producing non-negative clauses, children
sometimes may project simply a VP on its own; sometimes they may project a VP further into a
TP+VP structure; and sometimes they may project a TP still further into a complete (adultlike)
CP+TP+VP structure. Likewise, when producing negative clauses, children may sometimes project
only a NEGP+VP structure; sometimes they may project NEGP further into a TP+NEGP+VP
structure; and sometimes they may project TP yet further into the full CP+TP+NEGP+VP structure
found in canonical adult negative clauses. As should be apparent, there are some similarities with the
prefunctional analysis discussed in the previous chapter (which sees the earliest clauses produced by
young children as being simple VPs which lack the TP and CP layers of functional structure found in
adult clauses). Indeed, it is possible that truncation is a phenomenon associated with a later stage of
development, when children are moving out of the prefunctional stage, and are beginning to produce
functional TP and CP structures (albeit only sometimes projecting clauses beyond VP into TP, and only
sometimes projecting them still further beyond TP into CP).
An important (structural continuity) assumption made by Rizzi is that there is continuity between
adult and child structures in certain respects. For example, Rizzi posits that principles of UG (innately
wired into the Language Faculty) provide a universal CP>TP>VP template for the structure of
(non-negative) clauses: this means that (universally) any CP contains a TP beneath it, and that any TP
contains a VP beneath it: and this in turn means that a child may produce a VP structure, or a TP+VP
structure, or a CP+TP+VP structure, but will not produce (e.g.) a CP+VP or a TP+CP+VP structure
because these are not in conformity with the structural template provided by UG. A second assumption
made by Rizzi is that there is continuity between the syntax of constituents in adult and child
grammars. So, for example, since preposed wh-expressions move to to the specifier position within CP
in adult English (as indeed which lecture does in (2) above), then preposed wh-expressions must move
to spec-CP in child English as well. Likewise, just as subjects move out of the verb phrase into the
specifier position within TP in adult grammars, they are likewise assumed to move into the specifier
position within TP in child grammars (in clauses containing a TP). In much the same way, the system
of case assignment which operates in adult grammars is assumed to operate in child grammars.
Rizzi (1992, p.103) notes that empirical support for the truncation analysis comes from a study of
the wh-questions in the Adam files on the CHILDES data base. Although most of Adam’s
wh-questions (like those in (9) below) show wh-initial word order, at the same time he also produces
wh-in-situ questions like (10) and (11) below:
(9)(a) What shall we shall have? (Adam 3;2, file 25)
(b) How d’you take then off? (Adam 3;6, file 33)
(10)(a) They are for who? (Adam 3;2, file 25)
(b) He may do what to me? (Adam 3;6, file 33)
(11)(a) Use dat for what? (Adam 3;2, file 25)

48
The only types of clause not assumed to be CPs in Chomsky’s recent work are a small class of complement
clauses like those italicised in ‘I consider John to be unsuitable’ (termed a defective clause by Chomsky) and ‘I
consider John unsuitable’ (traditionally termed a small clause). The relevant types of complement clause are
said to be defective in that they do not contain the CP layer found in other clauses.
 (b) Doing what? (Adam 3;8, file 35)
One account of this type of the relevant alternation is to suppose that sentences like (9) are full
CP+TP+VP structures, with the italicised wh-word moving to the specifier position within CP;
sentences like (10) are TP+VP structures (with the CP layer truncated) containing a finite T constituent
are/may, and sentences like (11) are simple non-finite VP structures (with both the CP and TP layers
truncated). It may be that Adam’s yes-no questions like those below (produced at age 3;6, file 33)
provide similar evidence for truncation:
(12)(a) Is dis part of it? (Adam 3;6, file 33)
(b) Dis is the part? (Adam 3;6, file 33)
(c) Go on here? (= ‘Does it go on here?’ Adam 3;8, file 35)
Thus, (12a) may be a full CP+TP+VP structure (containing the inverted auxiliary is in the head C
position of CP), (12b) a finite TP+VP structure (with the CP layer truncated), and (12c) a non-finite
VP structure. If the analysis of sentences like (9-12) outlined here is along the right lines, Adam’s
interrogative clauses provide empirical support for the truncation analysis.
Further potential evidence in support of the truncation model comes from the fact that children at
the truncation stage are reported to alternate between initial and medial negatives, as the following
sentences from the Adam and Eve corpora on the CHILDES data-base 49 illustrate:
(13)(a) No dat blast off (Adam 2;11, file 19) (b) Dat no blast off (Adam 2;11, file 19)
(14)(a) Not Fraser read it (Eve 1;9, file 8) (b) Fraser not see him (Eve 2;0, file 14)
If we suppose that there is a universal hierarchy of projections (with NEGP positioned between TP and
VP), and that children may truncate one or more layers of functional structure (either CP, or both CP
and TP), this leads us to the following analysis. Initial negatives are structures in which the clause is
projected only as far as NEGP, whereas medial negatives are structures in which the clause is
projected beyond NEGP into TP. On this view, (13a) No dat blast off is a NEG+VP structure of the
form (15a) below, while (13b) Dat no blast off is a TP+NEGP+VP structure of the form (15b):
(15)(a) [NEGP No [NEG ø] [VP dat [V blast] off]]
(b) [TP Dat [T ø] [NEGP no [NEG ø] [VP dat [V blast] off]]]

If (in line with Rizzi’s structural continuity assumption) we suppose that T has the same requirement
for a subject of its own in child grammars as in adult grammars, then it follows that there cannot be a
TP layer of structure in (15a) or else the subject dat would raise across the negative no to become the
specifier of T and move into the position it occupies in (15b). This would mean that Adam truncates
both the CP and TP layers of functional structure in (15a), but only truncates the CP layer in (15b). On
this view, negative sentences provide us with further evidence that children go through a truncation
stage at which they alternate between projecting and truncating (i.e. omitting) functional categories
like CP and TP.
In much the same way, the truncation analysis would provide us with a way of accounting for the
fact that children (like Adam) go through a stage where they alternate between using and omitting to in
(italicised) infinitive clauses – as we see from the examples below (all produced by Adam at age 3;6,
the data being from the Brown corpus on the CHILDES data-base):
(16)(a) I don’t want it to break (b) I want them be straight like that
(17)(a) I gonna turn it fast (b) I gon go the other way
The italicised infinitive clauses in (16a, 17a) contain infinitival to (with going to being reduced to
gonna in (17a), as in colloquial adult English), and so appear to contain a TP headed by the infinitival
T to. But the italicised infinitive clauses in (16b, 17b) lack infinitival to and hence would appear to be
simple verb phrases/VPs which contain no TP. If so, this suggests that Adam (at the relevant stage)
alternates between producing infinitival TPs and infinitival VPs.

49
Dat is Adam’s counterpart of that (and dis of this). It may be of interest to note that Adam is a slow developer
and Eve a fast developer.
 We could suggest a similar truncation analysis of sentence pairs like the following (both produced
by Adam at age 3;6):
(18)(a) My head’s hurting (b) My head hurting
The clause in (18a) seems to contain a finite TP headed by the present-tense T auxiliary (i)s; but the
auxiliariless counterpart (18b) could simply be a non-finite VP (with the TP layer truncated).
The truncation analysis also provides an interesting way of accounting for the fact that children at
the relevant stage seem to alternate between producing nominative and accusative subjects, as the
examples below illustrate50:
(19)(a) Him go on there (b) He goes around (Angharad 1;10)
(20)(a) Me get apple (b) I don’t know (Jem 2;4)
(21)(a) Me having a house (b) Can I have it? (Michelle 2;5)
(22)(a) Him wanting to walk (b) She’s waking up (Hannah 2;6)
The accusative-subject structures in the (a) examples in (19-22) seem to be non-finite clauses, in that
they contain infinitive forms like go/get and progressive participle forms like having/wanting. If we
take them to be non-finite VPs, (19a/21a) will have the respective structures shown in (23a/b) below:
(23)(a) [VP Him [V go] on there] (b) [VP Me [V having] a house]
The structures in (23) are thus truncated clauses, lacking the TP and CP layers of functional structure
found in adult English. The verbs in them are non-finite because the relevant structures are VPs which
lack the TP layer of structure which contains the T constituent that is the locus of finiteness in adult
English. Let us suppose that the case-system operating in child grammars at this stage is essentially the
same as that operating in adult grammars, viz:
(24) A nominal (i.e. noun or pronoun expression) is assigned
(i) nominative case if in the domain51 of a nominative-case-assigning head (a finite T)
(ii) genitive case if in the domain of a genitive-case-assigning head 52
(iii) accusative case if in the domain of an accusative-case-assigning head (e.g. a transitive
V, P or C), or if not in the domain of any case-assigning head 53
It follows from (24) that the subjects of truncated VP-clauses like those in (23) will be assigned
accusative case by virtue of not falling within the domain of any case assigner – because there is no
head above/to the left of the subject which can case-mark the subject.
By contrast, the (b) examples in (19-22) all seem to be finite clauses, since (19b) contains the finite
verb goes, and (20b/21b/22b) contain the finite auxiliaries don’t/can/(i)s. It follows that the relevant
examples must project beyond VP into (at least) a TP+VP structure. So, for example, (22b) She’s
waking up will be derived as follows. The verb waking will merge with its complement up and its
subject she to form the verb phrase [VP she [V waking] up]. The resulting VP will then merge with the T
constituent (i)s which will agree (in person and number with) and assign nominative case to the
subject she in accordance with (24i), and (because T requires a subject of its own) will attract she to
become the subject of (i)s, so forming the TP structure in (25) below (with strikethrough indicating the
original position occupied by she before it moved):
(25) [TP She [T (i)s] [VP she [V waking] up]]
In the case of interrogative structures like (21b) Can I have it? the clause structure will project even
further into a CP+TP+VP structure. More specifically, (21b) will be derived as follows. The verb have
merges with its complement it and its subject I to form the VP I have it; this VP is then merged with
the finite (present-tense) T constituent can which assigns nominative case to the subject I in
50
The examples (but not the analysis) are taken from Aldridge (1989).
51
More fully, ‘in the c-command domain of’ (i.e. c-commanded by) a head of the relevant kind
52
Abney (1987) argued that possessors are assigned genitive case by a null determiner ‘higher up’ in the nominal
containing them. So, in a structure like [DP [D ø] [PossP John [Poss ø] car]], John is assigned genitive case by a null
determiner above (and to the left of) it.
53
The case assigned to a noun or pronoun expression which does not fall within the domain of any case assigner
is termed the default case: the default case is accusative in English, but nominative in German (Schütze 2001).
accordance with (24i), and (because every T requires a subject) attracts I to move to become the
subject of can, forming the TP [TP I [T can] [VP I [V have] it]]. The resulting TP is merged with an
interrogative C containing a null question particle Q which attracts can to move from T to C and attach
to Q, so deriving the CP structure:
(26) [CP [C Can+Q] [TP I [T can] [VP I [V have] it]]]
On this view, clauses with accusative subjects are VPs (with the subject in spec-VP), but those with
nominative subjects are TPs or CPs (with the subject in spec-TP).
An immediate consequence of the truncation analysis is that two-year-olds alternate between
producing finite and non-finite main clauses – e.g. between a finite clause like (19b) He goes around
and an infinitive clause like (19a) Him go on there. Since a main clause is one which is the root (i.e.
highest constituent) in a structure, such clauses are known as root clauses. And accordingly, Rizzi
notes that one of the major phenomena which truncation gives rise to is root infinitives (main-clauses
containing an infinitival verb-form) like Him go in there. These do not occur in adult English because
root clauses in adult English are generally required to be finite 54.

4.4 Null auxiliaries and null subjects in finite main clauses in adult English
Rizzi (1992, fn. 11) observes that a ‘semantically empty’ auxiliary (i.e. a non-modal
auxiliary like DO/BE/HAVE) can be null in adult English root-clause questions. Supporting evidence for
this claim comes from the fact that the Bates files on the CHILDES data-base show mothers talking to
children aged 1;8 and 2;4 omitting the parenthesised auxiliary forms in yes-no questions like (27a) and
wh-questions like (27b):
(27)(a) They having dinner? (are). You doing a little dance? (are). That a chair? (is).
It snowing outside today? (is). You found another chair? (have). Ya need a tissue? (do)
That look like a dog to you? (does).
(b) Where you going? (are). What she doing? (is). What they got in them (have).
How they sit (do). Then what she do? (did)
To see why, consider the structure of sentence (27a) They having dinner? This is shown in highly
simplified form below:
(28) [CP [C Are] [TP they [T are] having dinner]]
The auxiliary are originates in the head T position of TP and from there raises to the head C position
of CP. Rizzi (2000) posits that principles of UG (relating to the identification of null constituents, as
we shall see below) allow for the possibility that:
(29) A constituent at the edge (i.e. in the head or specifier position of) a root projection 55 can
receive a null spellout under certain conditions
Because CP is the root (i.e. the topmost/leftmost constituent) of the clause in (28), and the preposed
auxiliary are occupies the head C position of CP, are is therefore a root head and can consequently
receive a null spellout in accordance with (29) by virtue of being positioned at the edge of CP – as is
implicit in the remark by Guasti & Rizzi (1996, p.289) that ‘A null aux is possible in the head of the
root’.
However, it is not only auxiliaries which can sometimes be null in finite clauses in English, but
also subjects. Rizzi observes that a subject pronoun can receive a null spellout in a finite declarative
main clause in informal styles of spoken adult English, and in diary styles of written English – so that

54
Two types of non-finite main/root clauses found in adult English are why questions like Why worry about it?
and so-called ‘Mad magazine’ sentences like that italicised in ‘Gary drunk? I don’t believe it!’ Strictly speaking,
it would be more accurate to say that truncation of the TP layer in clause structure gives rise to root non-finite
clauses, since utterances like Me having a house contain a progressive participle, not an infinitive.
55
The term projection can be thought of informally as meaning ‘structure formed by combining two or more
words or expressions’. The root of a syntactic tree is the topmost node in the tree (corresponding to the
constituent representd by the leftmost bracket in a labelled bracketing). So CP is the root projection in a structure
like (28) because it is the leftmost bracketed constituent in the structure.
the italicised nominative pronouns in sentences like (30a) below can receive a null spellout (denoted
as ø) in informal or diary styles, as shown in (30b):
(30)(a) I can’t find my pen. It must be on my desk
(b) ø can’t find my pen. ø must be on my desk
Rizzi notes, however, that the subject of a declarative complement clause cannot receive a null
spellout in adult English – as we see from the fact that the main-clause subject I can have a null
spellout in (31) below, but not the complement-clause subject it:
(31)(a) I think it must be on my desk
(b) ø think it must be on my desk
(c) *ø think ø must be on my desk
Why should it be that the main clause subject can be null in sentences like (31), but not the
complement clause subject? Rizzi invokes truncation as a mechanism to account for the relevant
facts. More particularly, he makes the following assumption:
(32) The CP layer of structure in a root/main declarative clause can be truncated (i.e. can simply
not be projected) in informal style in spoken adult English (and diary style in written English)
Let’s see how his truncation analysis works.
In formal styles of adult English, declarative clauses are CPs headed by a null complementiser (for
the reasons given earlier), so that a clause like It must be on my desk has the structure shown below:
(33) [CP [C ø] [TP It [T must] [VP [V be] on my desk]]]
The subject it occupies the specifier position within TP, but because TP is not the root of the clause
(because TP is contained within a CP, and it is this CP which is the root of the clause), it is not in a
root edge position (i.e. it does not occupy head or specifier position at the edge of a root/main clause),
and hence (29) does not allow the subject to be null in formal styles of English. But, argues Rizzi, in
informal styles of spoken English and diary styles of written English, the root (i.e. topmost) CP layer
of structure can be truncated (i.e. can simply not be projected at all) in accordance with (32), with the
result that the clause projects only as far as the TP shown in (34) below:
(34) [TP It [T must] [VP [V be] on my desk]]
The root of the truncated clause structure in (34) is TP, and since it occupies the specifier position
within TP, it does indeed occupy a root specifier position, and so can receive a null spellout in
accordance with (29). Thus, Rizzi’s truncation analysis predicts that finite declarative clauses can only
have null subjects in adult English in truncated main clauses – hence in styles of adult English which
permit truncation of the topmost CP layer of structure in main clauses (e.g. informal styles of spoken
English and diary styles of written English). Rizzi refers to the relevant kinds of subjects as null root
subjects (i.e. subjects which can be null only when occupying a root specifier position).
A consequence of Rizzi’s analysis of null root subjects is that only the subject of a (truncated)
main clause can be null, not the subject of any kind of subordinate clause. Thus, in a truncated
structure such as (35) below (where CP has not been projected in the main clause):
(35) [TP I [T ø] [VP [V think] [CP [C ø] [TP it [T must] [VP [V be] on my desk]]]]]]
the main clause subject I can receive a null spellout (by virtue of being a root specifier), but not the
complement clause subject it (since this is not a root specifier). Note also that the CP layer in the
complement clause cannot be truncated, since (32) only allows CP to be truncated in a root (i.e. main)
clause.
Rizzi argues that his truncation analysis can also account for why finite wh-questions in adult
English do not allow null subjects, e.g. in sentences such as:
(36)(a) Where are you going? (b) *Where are ø going?
Since wh-expressions move to the specifier position within CP, a sentence like (36a) will be a CP with
the superficial structure shown in (37) below (simplified by not showing you originating as the subject
and where as the complement of the verb going):
(37) [CP Where [C are] [TP you [T are] going]]
Since the subject you occupies the specifier position within TP in (37), it is not a root specifier (The
root of the structure in (37) is CP, and the specifier of CP is where: hence where is the only root
specifier here). Thus, argues Rizzi, the truncation analysis (embodying the twin assumptions in (29)
and (32) above) provides a principled account of the fact that the only kind of finite clause which
generally permits a null subject in adult English is a declarative main clause 56.
Rizzi maintains that his account of null root edge constituents in (29) is grounded in a principle of
Universal Grammar which I will call the Identification Principle and which can be outlined
informally as follows:
(38) Identification Principle
A constituent can only can have a null spellout if it can be morphologically or syntactically
identified, or if it is in a (root edge) position where it cannot in principle be morphologically
or syntactically identified but can be discourse identified.
We are already familiar with the notion of morphological identification from our discussion of Italian
sentences such as (39) below in chapter 2:
(39) pro era sicura che pro avevano finito
pro was sure that pro had finished (= ‘She was sure that they had finished’)
There we saw that the person/number properties of the null pro subjects can be identified by the rich
person/number agreement inflections carried by the T constituents era ‘was’ and avevano ‘had’: as this
example shows, agreement provides a morphological mechanism by which null pronouns can be
identified.
We can illustrate syntactic identification of a null pronoun in terms of the following line from one
of Freddy Mercury’s greatest hits:
(40) I want [PRO to break free]
Here, the PRO subject of the bracketed infinitive complement clause is construed as referring to the
same individual as the I subject of the main clause: to use the relevant technical terminology, PRO is
syntactically identified by an antecedent (an expression it refers to) higher up within the overall
sentence containing it (the antecedent in this case being the pronoun I, so that I serves as a syntactic
identifier for PRO).
We can illustrate the third possibility of discourse identification in terms of the phenomenon of
Topic Drop found in German sentences such as the following (from Rizzi 1992, p.105):
(41)(a) Das habe ich gestern gekauft
This have I yesterday bought (‘This, I bought yesterday’)
(b) ø habe ich gestern gekauft
ø have I yesterday bought (‘This, I bought yesterday’)
The italicised pronoun in (41a) is a preposed topic constituent (i.e. a constituent which refers to some
entity already mentioned in the discourse). As the null symbol ø in (41b) shows, a preposed (third
person) topic pronoun can (optionally) be given a null spellout in German. Why should this be? It is
clear that the null topic pronoun ø in (41b) cannot be morphologically identified, since it doesn’t agree
with any verb carrying a rich set of agreement inflections which can identify its person/number
features. Nor can the null topic be syntactically identified in (41b), since it is the leftmost constituent
in the sentence and hence cannot in principle have an antecedent higher up (and further to the left)
within the same sentence. Accordingly, the Identification Principle (38) allows the topic pronoun to
have a null spellout only if it can be discourse identified – e.g. if it refers to some entity mentioned in
the preceding sentence in the discourse. In much the same way, the pronoun it in the truncated English
TP structure in (34) can also be discourse identified, because it occupies a root edge position (and so

56
Left out of the discussion here are imperative clauses, which allow a null counterpart of you as their subject.
Imperative null subjects appear to differ from root null subjects since they can be null even when not occupying
a root specifier position – e.g. in sentences such as Don’t (you) say anything to him! where the root projection
appears to be headed by Don’t, and yet the subject appears to be the specifier of a lower projection.
cannot in principle be syntactically identified) and refers to the expression my pen mentioned in the
preceding sentence I can’t find my pen in (30a) and so can readily be discourse-identified.

4.5 Null subjects in child English

As we noted in our discussion of the Null Subject Parameter in §2.9, two-
year-old English children frequently produce finite clauses with null subjects. We
observed, for example, that Sano and Hyams (1994) report that 56.5% of the d-
inflected verbs produced by an American boy called Adam from age 2;3 to 3;0
had null subjects, citing examples such as the following:
(42) Goed on that way. Dropped a rubber band. Slapped Becca and Rachel
Likewise, we noted that a (different) English boy called Adam produced null
subject sentences such as the following at age 2;2:
(43) Don’t know (x 14: used in reply to questions like ‘What's this?’, ‘Is it a
train?’ etc.)
Don’t paint that (= ‘I didn’t paint that’). Don’t work (x 3 = ‘It doesn't
work’)
Don’t wanna draw on this one. Does (response to ‘Yeah, it does’). Won’t
(response to ‘Does
it work?’). Can’t knock them down. Can’t get it out. Can’t stroke me now.
Can’t (x 3: reply to
questions like ‘Can you see anything, Adam?’)
How can we account for the fact that both children seem to be using null subjects
with finite verbs and finite auxiliaries in sentences such as (42/43)?
The truncation model can provide us with an answer along the following lines.
Since T is the locus of finiteness, and since the sentences in (42/43) all contain a
finite verb/auxiliary, it is clear that the children concerned must project the
clause structures in (42) and (43) beyond VP into TP. But let’s also suppose that
they don’t project them beyond TP into CP. If so, then a sentence like Can’t stroke
me now will be a TP with the structure shown in skeletal form below:
(44) [TP You [T can’t] stroke me now]
Since you is positioned at the edge of the root TP projection, it can therefore be
given a null spellout in accordance with (29). (44) is a truncated clause, in the
sense that the CP layer of structure found in canonical adult clauses is omitted.
Empirical support for the truncation analysis of null subjects in finite clauses in child English
comes from other studies. Virginia Valian’s (1991) study of 21 American children (mostly around two
years of age) showed that while they frequently produced null subjects in finite main clauses, they
never did so in finite subordinate clauses (none of the 132 finite subordinate clauses produced by the
English children she studied had a null subject). This finding follows directly from the truncation
analysis, since the truncation analysis predicts that only the subject of a finite main clause can have a
null subject (and only then if it is a truncated structure which projects no CP layer). Likewise, Roeper
and Rohbacher’s (2000) study reports that virtually none (only 5%) of the finite wh-questions
produced by the children in their study had null subjects: since wh-questions are CPs and finite clauses
only allow null subjects if their CP layer is truncated, this is exactly the situation which the truncation
analysis would predict57.

57
We might wonder why 5% of the finite wh-question in the Roeper and Rohrbacher study had null subjects,
when the truncation analysis predicts that none should have null subjects. The answer may well be that the
subject received a null spellout in such cases via a phonological process of some kind. For example the schwa
vowel represented by the final a of gonna sometimes gets dropped in colloquial English structures such as What
am I gon do? It may be that something similar happens with the schwa (representing a reduced form of the
pronoun I) in sentences like What canna do? which then gets reduced to What can do? by children via an
extension of schwa-deletion.
 Rizzi (2000) claims that children produce root null subjects in finite declarative clauses until
around the time of their third birthday, and thereafter cease to use root null subjects. He correlates this
with a change in the structure of finite declarative clauses: they are TPs for 2-year-olds, but CPs for
3-year-olds. Thus, a sentence like I want ice-cream will be a TP with the structure (45a) below for a
typical two-year-old, but will be a CP with the structure (45b) for a typical three-year-old:
(45)(a) [TP I [T ø] [VP [V want] ice-cream]]
(b) [CP [C ø] [TP I [T ø] [VP [V want] ice-cream]]]
He posits that the initial TP status of declarative clauses is the consequence of a universal Structural
Economy Principle which requires the projection of the minimum structure consistent with
grammatical requirements. The declarative complement clauses produced by two-year olds, he notes,
typically show no use of the declarative complementiser that: hence (in accordance with the Structural
Economy Principle), two-year-olds assume that declarative clauses project no further than TP. But
once they acquire declarative complement clauses headed by that (and start to produce sentences like I
don’t think that Mummy likes football), they know that declarative clauses introduced by that are CPs,
and the Categorial Uniformity Principle (6) forces them to assume that all declarative clauses
(including declarative main clauses like that in (45b) above) are CPs. What this in effect means is that
the acquisition of the complementiser that serves as a trigger which prompts the child to project
declarative clauses beyond TP into CP. This of course predicts a correlation between the acquisition of
the complementiser that and the loss of null root subjects (though Rizzi presents no empirical data in
support of such a correlation).
Although I have concentrated here on Rizzi’s analysis of null subjects in finite clauses in English
(since this is the most novel part of his work), Rizzi maintains that not all null subjects found in root
clauses can be the result of truncation. In particular, he maintains that two-year-olds frequently
produce null-subject questions such as Where go? What doing? Where gone? He argues that the null
subjects found in such sentences cannot be the kind of null root subjects found in truncated (CP-less)
clauses like (44), since preposed wh-expressions are assumed by Rizzi to move to the specifier
position within CP in both adult and child grammars alike: hence, such sentences must be CPs.
Moreover, given Rizzi’s assumption of a universal hierarchy of projections (CP>TP>(NEGP>)VP), it
follows that any CP must also contain a TP and a VP. And since Rizzi assumes that T has the same
requirement for a subject in child as in adult grammars, then the null subject must occupy the specifier
position within TP. But what kind of null subject do such clauses have? Since they contain non-finite
verb forms (like the infinitive go, the progressive participle doing and the perfect participle gone),
Rizzi concludes that the null subject is the non-finite PRO subject found in adult non-finite clauses, so
that a sentence like Where go? has the structure shown in (46) below (i.e. essentially the same
structure as the italicised complement clause in an adult sentence like ‘I wonder where to go’ – though
with the infinitive particle to having a null spellout in a main clause – as in Why worry about it?):
(46) [CP Where [C ø] [TP PRO [T ø] [VP PRO [V go] where]]]
(strikethrough indicating that PRO originates as the subject and where as the complement of go).
Rizzi suggests that root non-finite clauses cease to be used by most children around 3 years of age.
What the analysis outlined here entails is that two-year-olds often produce non-finite root/main
clauses. Since T is the locus of finiteness, this means that the head T constituent of a root clause in
child English is sometimes (e.g. in structures like (46) above) left underspecified for finiteness (i.e.
lacking the tense and/or agreement features carried by a finite T in main clauses in adult English) – an
idea which (as we shall see in the next-but-one chapter) is developed in Ken Wexler’s Optional
Infinitives model.

4.6 Evaluation of the Truncation Model

The truncation model developed by Rizzi can be seen as an interesting attempt to unify
two very different models of children’s syntax: on the one hand, the structure-building model which
claims that children’s initial clauses are VPs, and on the other hand the continuity model (outlined in
the next chapter) which claims that the earliest clauses produced by children are full CPs. In claiming
that some of the initial clauses produced by children are CPs, some are TPs, and some are VPs, Rizzi
is in effect acknowledging that both models capture important insights about the nature of child
grammars. And yet, there are a number of aspects of Rizzi’s truncation analysis which should give us
cause for reflection.
As is clear from the exposition above, Rizzi’s analysis is predicated on the assumption that UG
principles determine the types of structures produced by adults and children alike. And yet, some
aspects of his analysis would appear to violate some of the principles which he advocates. For
example, a crucial principle which he invokes is the Categorial Uniformity Principle/CUP (which
posits that all expressions of the same type belong to the same category). However, CUP would surely
entail that if Adam’s questions like (9a) What shall we shall have? and (12a) Is did part of it? are CPs,
then all Adam’s questions at the relevant stage must uniformly be CPs. This raises questions about the
plausibility of claiming that questions like (10a) They are for who? and (12b) Dis is the part? are TPs
(with the CP layer truncated), or that sentences like (11a) Use dat for what? and (12c) Go on here? are
VPs (with both the CP and TP layers truncated)58. Indeed, if we follow Chomsky (1995, p.289) in
positing that ‘the functional category C (complementizer)….determines clause type’, then it follows as
a matter of conceptual necessity that canonical (i.e. ‘ordinary’) clauses must contain a CP projection
whose head C constituent serves to mark the clause as interrogative, exclamative, imperative or
declarative in type.
More generally, such considerations raise important questions about the very nature of truncation.
The core assumption made by Rizzi is that truncated sentence structures lack some of the functional
superstructure found in the corresponding adult sentences (with either CP, or both CP and TP not being
projected). However, an alternative possibility (discussed more fully in the next chapter) is to suppose
that truncated sentences have a complete CP+TP+VP structure, but that children sometimes give
certain constituents (particularly subjects and auxiliaries) a null spellout when they occur at the very
beginning of a sentence. In this connection, it should be noted that adults (in colloquial English) can
seemingly truncate subjects and auxiliaries in main clauses, so that alongside full clauses like (47a)
below, they produce truncated clauses like (47b) and (47c):
(47)(a) Have you done any work? (b) You done any work? (c) Done any work?
While it might seem tempting at first sight to see (47a) as a full CP, (47b) as a TP (with the CP layer
truncated), and (47c) as a VP (with both CP and TP truncated), an alternative analysis is to assume that
all three sentences are full CPs, and differ only in whether have and you are given an overt or null
spellout. On this view, the three sentences in (47) would have the respective structures shown in
simplified form below (with outline font used to mark items which receive a null spellout 59:
(48)(a) [CP [C Have+Q] [TP you [T have] done any work]]
(b) [CP [C Have+Q] [TP you [T have] done any work]
(c) [CP [C Have+Q] [TP you [T have] done any work]
In (48a), both have and you are overtly spelled out; in (48b), have is given a null spellout and you an
overt spellout; in (48c) both have and you are given a null spellout. It may well be that the null
spellout of the subject and auxiliary in such cases can be accounted for in metrical terms, as being the
result of deletion of a sequence of one or more unstressed syllables at the beginning of a sentence.
Such a metrical account would generalise to sentences like Nice day, isn’t it? where the sequence It’s a
has been given a null spellout.
A null spellout analysis which takes sentences like (47b, c) to be full (untruncated) CPs in which
one or more weak syllables at the beginning of the sentence are given a null spellout offers two
distinct advantages over a syntactic truncation account which takes the CP and TP layers of structure
to be missing. The first is that it provides a principled account of why the verb DO is in the perfect
participle form in (47c) Done any work? Under the analysis of verb morphology in Adger (2003), a
non-finite verb acquires its morphological features via agreement with an auxiliary (or infinitival to):
in a structure like He was doing his duty, the verb DO agrees in aspect with the progressive-aspect
58
There may be a historical answer to this question, in that the claim about some child questions having a
truncated structure is made in his (1992) paper, whereas the Categorial Uniformity Principle appears in his later
(e.g. 2000) work; and it may be that he would no longer wish to maintain his earlier claim.
59
Q is an abstract question particle which (in main-clause questions) attracts a tensed auxiliary to move from T
to adjoin to Q. Items which receive a null spellout are shown in outline font.
auxiliary BE, and so is spelled out as the progressive participle doing; in He has done his duty, the verb
DO agrees in aspect with the perfect-aspect auxiliary HAVE, and so is spelled out as the perfect
participle done. The fact that the verb DO is spelled out as done in (47b, c) suggests that these
sentences must therefore contain a null variant of HAVE. The second advantage of the null spellout
analysis in (48) is that it provides a principled account of the occurrence of the partitive quantifier any
in such sentences. Klima (1964) noted that partitive any is a polarity item which must fall within the
scope of (and so be c-commanded by60) a negative or interrogative constituent (as we see from the
ungrammaticality of *He’s done any work). If we adopt a CP analysis along the lines of (48) above, we
can account for the occurrence of any by positing that it falls within the scope of an interrogative
question particle Q which occupies the head C position of CP: as should be obvious, if we posited that
structures like (47b, c) contain no CP layer of structure (and hence no Q particle), we would be left
with no account of the use of any in such sentences.
The conclusion which such considerations lead us to is that elliptical (i.e. abbreviated) sentences in
adult English like (47b, c) should not be analysed as the result of truncation of one of more layers of
the functional superstructure of clauses, but rather should be seen as the result of a low-level
phonological spellout operation which allows a sequence of one or more unstressed syllables at the
beginning of a sentence to have a null spellout. But if this is true of elliptical sentences produced by
adults, isn’t it plausible to suppose that the same may be true of elliptical sentences produced by
children as well? Indeed, given the well documented phonological production problems which young
children have, it would seem highly likely that ‘missing’ subjects and auxiliaries in child grammars
are the result of spellout problems rather than the result of a failure to project part of the adult
syntactic structure.
Rizzi’s attempt to capture the conditions under which an item can receive a null spellout also seem
to be descriptively inadequate – both for adult and child English. For example, he posits that only a
‘semantically empty’ auxiliary (like DO/BE/HAVE) can be given a null spellout in adult or child
grammars. And yet, if we look at examples of mother-child interaction in the Bates files on the
CHILDES corpus, we find the mothers producing structures like those below in which a (contentful)
modal is given a null spellout (items receiving a null spellout being shown in parentheses):
(49) Mommy have one? (= can). We put this here? (= shall). Like me to get your other hand?
(= Would you)
Moreover, the claim that only constituents at the edge of a root projection can be given a null spellout
proves problematic for sentences like (47c) Done any work? where both an inverted auxiliary and a
subject have been given a null spellout; while the null inverted auxiliary is in C and hence in a root
edge position, the null subject occupies the specifier position within TP, and hence is not on the edge
of a root projection (by virtue of not being the head or specifier of a CP) 61.
Moreover, Rizzi’s claim that a root edge constituent can only be null if it can be discourse
identified seems problematic in respect of null-wh structures produced by children. As the examples
below (from files on the CHILDES data-base) illustrate, alongside sentences containing an overt
wh-word, children also produce sentences in which they give the wh-word a null spellout:
(50)(a) What you making? You making? (Abe 2;5, Kuczaj corpus)
(b) This work, Mom? How this work? (Abe 2;8)
(c) Where people go? People go? (Keith 1;8, Bates corpus)
If we suppose that the relevant wh-questions have much the same syntax as in adult English and differ
only in respect of whether constituents on the edge of CP are given a null spellout or not, sentences
like those in (50a) will have the structure shown below (with outline font marking null spellout):
(51)(a) [CP What [C are] you making] (b) [CP What [C are] you making]

60
Informally, if structure is represented by a labelled bracketing, this can be thought of as meaning to the right
of.
61
It might seem to be possible to ‘rescue’ Rizzi’s analysis by positing that the highest head or the highest
specifier in a given structure can have a null spellout. But this would wrongly predict that children and adults
would produce yes-no questions like *Are gonna help me ( ‘Are you gonna help me?’)
Since are occupies the head C position of CP and what occupies the specifier position in CP, the null
spellout for what and are in such structures is consistent with Rizzi’s assumption that constituents can
have a null spellout when on the edge of a root projection. But note that Rizzi also claims that a root
edge constituent can only be null if it can be discourse identified. And yet, there seems to be no sense
in which what can be said to denote some familiar entity within the domain of discourse in such cases.
Nor is it easy to see how the null auxiliary are can be discourse identified in (51b). In short, it is not
clear that Rizzi’s attempt to identify the conditions under which constituents can have a null spellout in
syntactic terms is entirely successful: it may well be that metrical conditions of the type discussed by
Gerken (1991) determine that a sequence of one of more unstressed syllables at the beginning of a
sentence can be null.

Workbook section
4.1 Adam’s wh-questions
Discuss the syntax of the positive wh-questions italicised below (produced by Adam at age 2;4) and
show how the relevant types of question might be analysed within the truncation model.
Where dat come from? Who me tickle? What say? Where find plier?
In addition, say how negative wh-questions such as the following (produced by Adam at age 2;11, file
17) might be analysed:
Why not me break plate? Why not me drink it? Why not me sleeping? Why not
me careful?
Discuss what problems they appear to pose for Rizzi’s analysis.

4.2 Claire’s questions

The Claire files in the Appendix to this coursebook show an American girl called Claire from age 2;0
to 2;1 producing utterances like those in (a) below in contexts where adults would use a wh-question
like Where did X go? structures like (b) where adults would say What is X doing? and structures like
(c) where adults would use a yes-no question like Can X see Y?
(a) Where did the chair go? Where this go? Where go? This go? The rocking-chair goes?
(b) What squirrel doing? The bird doing? What doing?
(c) Piggie see the water in there? See baby?
Discuss how each of the relevant types of question might be analysed within the truncation model,
identifying any problems that arise. Could we account for why preposed wh-words can be null in child
English but not in adult English if we followed Rizzi (1997) in positing that CP can be split up into
two separate projections, a Focus Phrase/FocP which attracts a focused wh-word to move to specifier
position within the FocP and attracts an inverted auxiliary to move into the head Foc position of FocP,
and a Force Phrase above FocP (with a null Force head which marks the sentence as interrogative in
force/type) and that the Force Phrase layer can be truncated in child English but not adult English?
(See chapter 9 of Radford 2004a, 2004b for discussion of the split CP analysis.)

4.3 Iris’s use of nominative, accusative and null subjects

The Wells files on the CHILDES data-base show an English girl called Iris at age 3;2 (file 9)
producing nominative-subject sentences like those in the (a) examples below, accusative-subject
sentences like those in (b), and null-subject sentences like those in (c):
(a) Can I have a drink of juice? What am I going to do? I go like this. We’ll lie on the carpet. We had
a nice one. He can go up a shute. Now he going to do it.
(b) Can me put them on? Where me can sit? How me going to have a bath? How me get in that? Me
going to use this one. Me sit down. No me got one62. No me got him63.
(c) Why no gonna use that one? Won’t come. Want to go to bed?

62
Glossed in the transcription as ‘I haven’t got one’
63
Glossed in the transcript as ‘I haven’t got him’.
Say whether (under the truncation analysis) each type of clause would be finite or non-finite
and would be a CP, TP, NEGP or VP, and whether the use of a nominative/accusative/null subject
would be adequately accounted for under the proposals made in the main text.

4.4 Theoretical aspects of Rizzi’s analysis

Chomsky (1973) proposed a UG principle called the Subjacency Condition which posited that no
constituent can cross more than one bounding node in any single movement operation (TP being one
such bounding node). Say what implications such a principle has for Rizzi’s claims about how children
initially analyse the italicised declarative complement clause in adult sentences like:
1. Where do you think Daddy has gone?’
In more recent work, Chomsky (1998) has proposed a UG principle called the Phase Impenetrability
Condition which we can informally regard as saying that no constituent to the left of CP can affect any
constituent to the right of C (and concomitantly that no constituent to the right of C can move to a
position to the left of CP). Show how this condition allows the subject John of the italicised
complement clause to passivise (and receive exceptional accusative case from the transitive verb
mean) in structures like the following, if the italicised complement clause is a TP
2a. I didn’t mean John to take my remarks seriously
b. John wasn’t meant to take my remarks seriously
And show how the same condition prevent passivisaion of John in structures like the following, if the
italicised complement clause is a CP:
3a. I didn’t mean for John to take my remarks seriously
b. *John wasn’t meant for to take my remarks seriously
In the light of the Phase Impenetrability Condition, consider what Rizzi’s TP analysis of children’s
early finite declarative complement clauses would imply about the case-marking (and passivisability)
of the complement-clause subject Daddy in a sentence such as:
4. I think Daddy is working
What prediction would be made about the case-marking and passivisability of Daddy if (i) we
followed Rizzi in assuming that children’s initial finite declarative complement clauses are TPs, and
(ii) if we followed Chomsky (1995) in supposing that all finite clauses are universally CPs. What
inferences would you draw about Rizzi’s analysis from the fact that children are not reported to
produce sentences like the following?
5a. *I think him is working
b. *Daddy is thought is working

5. The Continuity Model

5.1 Introduction
In chapter 1, we saw that there are UG principles which constrain the nature of syntactic
structure and syntactic operations, and we saw that Chomsky maintains that such principles are wired
into the innate Language Faculty, and so do not have to be learned. If all such UG principles are on
line (i.e. operational) from birth (rather than e.g. different principles being programmed to come on
line at different stages of development), then it follows that the syntactic structures produced by
children at all stages of their grammatical development will conform to UG principles. If the Language
Faculty incorporates a Uniformity Principle which determines that (e.g.) all canonical clauses have a
uniform CP+TP+(NEGP+)VP structure, and if this principle is operative from birth, it also follows
that children will be born tacitly knowing that sentences are CP+TP+(NEGP+)VP structures. On this
particular view of language acquisition, there will be continuity between the structure of sentences in
adult and child English. We would therefore expect to find evidence that all children’s sentences (even
the earliest multiword utterances produced by one-year-olds) contain functional layers of structure like
CP/TP/DP. In this chapter, we look at evidence in support of the Continuity Hypothesis – i.e. the
hypothesis that all adult syntactic structure (and in particular, all functional structure) is already ‘in
place’ by the time children start to produce their earliest multiword utterances (a position argued for by
Boser et al 1992, Poeppel and Wexler 1993, and Borer and Rohrbacher 2004). Since a key assumption
within the continuity model is that children’s phrases and sentences have the same structure as their
adult counterparts and hence that children from the outset have full (adult-like) competence in syntax,
the relevant model is also known as the Full Competence Hypothesis. In this chapter, we look at
evidence that childen’s earliest multiword utterances do indeed contain functional categories like T, D
and C.

5.2 Evidence for a T constituent in children’s initial sentences

As we saw in chapter 3, part of the evidence adduced by adherents of the structure-building
model in support of positing that children’s earliest sentences contain no functional categories like T is
the absence of finite tense-marking auxiliaries (and finite forms of the copula BE) in structures like
Daddy working, Jane gone home, Teddy naughty. However, the apparent absence of a finite auxiliary
or copula in obligatory contexts (i.e. contexts where adults require their use) is not in itself conclusive
evidence of the absence of T. After all, in (adult) African American English/AAE we find similar
seemingly T-less (tenseless) sentence structures like the following (from Labov 1989, p.717):
(1) He fast in everything he do. He just feel like he getting’ cripple up from arthritis. You out the
game. We on tape
More specifically, in contexts where Standard English uses the contracted auxiliary forms ’s/’re, AAE
gives the relevant auxiliaries a null spellout. And yet, as Fasold (1980, p.29) notes, there is evidence
that AAE sentences like (1) contain a null variant of is/are. One such piece of evidence comes from
tag sentences like that below (where the part of the sentence following the comma is called a tag):
(2) He gonna be there, I know he is
The conventional wisdom about tag sentences like ‘John is working, he really is’ is that the (italicised)
auxiliary which appears in the tag is a copy of the (bold-printed) auxiliary which appears in the main
clause (to the left of the comma). If so, this means that the main clause in (2) must contain a null
counterpart of is (below symbolised as is), as shown in skeletal form below:
(3) He [T is] gonna be there, I know he [T is]
Since is can cliticise (i.e. attach itself) to the subject he in the guise of its contracted form ’s in clauses
like He’s gonna be there but not in clauses like I know he is, the relevant generalisation about AAE
would seem to be that the auxiliary has an overt spellout as is when used as a full (non-clitic,
uncontracted) form, but has a null spellout when used as a contracted clitic form. This raises the
possibility that child clauses which seemingly lack an auxiliary/copula may (just like their adult AAE
counterparts) contain an abstract T constituent which can be given a null spellout when used as a
clitic/contracted form.
In this connection, consider tag sentences like the following (from Radford 1992, p.44):
(4) I on it, aren’t I? (Sarah 2;2). I on this one, aren’t I? (Elizabeth 2;2). Those steps, those are
(Alistair 2;6). It dirty, is it? (Tony 2;6). Cos they all hided, aren’t they? (Christine 2;7)
The above are all sentences where adult Standard English speakers would require a copula form like
(a)m, (a)re or (i)s. If we make the conventional assumption that the auxiliary used in a tag is a copy of
the finite T constituent used in the main clause, then it follows that a clause like Cos they all hided
(‘Because they’re all hidden’) will contain a null counterpart of are in the head T position of TP, and
so will have the syntactic structure shown in highly simplified form below (if we follow Radford
1992, p.48 in taking coz to belong to the category C of complementiser in child English):
(5) [CP [C Coz] [TP they [T ’re] all hided]]
Since are in (5) occurs in a context where an adult could use the contracted/clitic form ’re (as in
Because they’re all hidden), it seem likely that the ‘missing’ auxiliary cliticises (i.e. attaches itself to)
its subject they and that (like adult AAE speakers) children sometimes give clitic auxiliaries a null
spellout.
But why should children give clitics a null spellout? Gleitman and Wanner (1982) argue that items
which are not acoustically salient are difficult for children to perceive, and hence to acquire. Clitics (in
their terms) are non-salient by virtue of being short word-forms which have little phonetic substance.
Leonard (1989) argues that morphemes lack phonetic substance (and are difficult to perceive and
acquire) if they contain an unstressed vowel, or a reduced vowel like schwa |ə|, and even more so if
they contain no vowel at all – especially when they lead to the creation of a cluster of two or more
adjacent consonants (since young children have problems in perceiving and producing consonant
clusters). From this perspective, it’s not surprising that (in the examples in (4) above) Sarah omits the
clitic form ’m in I’m on it, since the omitted word is a vowel-less clitic form consisting only of the
consonant segment |m|. Likewise, it’s no surprise that Tony says It dirty rather than It’s dirty, since the
omitted form ’s consists solely of the consonant segment |s| which (if not omitted) would have given
rise to the three-consonant cluster |tsd|. In the same way, it’s scarcely surprising that Alistair should
have omitted the word are in Those are steps, since in spoken English this is reduced to the vowel
schwa |ə| which is a weak vowel of very short duration. Moreover, research into the acquisition of
languages like French, Spanish and Italian which have clitic personal pronouns (e.g. clitic counterparts
of pronouns like him/it/them etc.) suggests that one-year-olds often give a null spellout to (and in this
sense, omit) clitic personal pronouns. For example, Fujino and Sano (2000) report that a Spanish child
called Maria whose progress they followed from age 1;7 to age 2;5 consistently used no clitic objects
from age 1;7 to 1;11, using null objects instead (clitic objects appearing from age 2;0 on). All of this
suggests that there is a cross-linguistic tendency for young children to use a null form in contexts
where adults use a clitic form, and adds plausibility to the claim that null auxiliary structures in early
child English contain a TP headed by a null T constituent which can be given a null spellout by
children in contexts where adults could use a contracted clitic auxiliary form.
Further evidence in support of the null T analysis (i.e. of claiming that early child clauses contain a
T constituent which can sometimes be ‘silent’) comes from the fact that many young children alternate
between producing parallel structures containing overt and null auxiliaries – as the following pairs of
examples (from Radford 1992, p.42) illustrate:
(6)(a) It not blue. It’s yellow. I teasing Mummy. I’m teasing Mummy. You big bully too. You’re a
big bully too. Mummy fix this. Mummy’ll fix this (Holly 2;0)
(b) Daddy gone. He’s gone (Neil 2;0)
(c) We been there. We’ve been there (Robert 2;2)
(d) I given those to Daddy. I’ve given those to Daddy for when he comes back (Heather 2;2)
The fact that the italicised auxiliaries are sometimes spelled out in their correct form strengthens the
case for claiming that children’s earliest sentences contain a T constituent which sometimes has a null
spellout (e.g. when corresponding to an adult clitic form) and sometimes has an overt spellout.
Theoretical considerations can be argued to point to much the same conclusion. For example, we
see from sentences like (6) that children alternate between progressive structures containing an overt
auxiliary (like I’m teasing Mummy) and progressive structures containing no overt auxiliary (like I
teasing Mummy). Since overt-auxiliary structures like I’m teasing Mummy contain a TP (headed by the
present-tense T constituent (a)m), Rizzi’s (2000) Categorial Uniformity Principle (which posits that
all expressions of the same type belong to the same category) leads us to the conclusion that all
progressive structures must likewise contain a TP, and hence that seemingly auxiliariless progressive
structures like I teasing Mummy will also contain a TP projection – one in which the auxiliary in T is
given a null spellout. Moreover, Adger (2003) posits that progressive and perfect participles like
taking/taken acquire their progressive/perfect aspect inflections -ing/-n via agreement in aspect with a
progressive/perfect aspect auxiliary. In this connection, consider adult clauses like those bracketed
below:
(7)(a) I think [CP [C that] [TP he [T is] [VP [V taking] a big risk]]]
(b) I think [CP [C that] [TP he [T has] [VP [V taken] a big risk]]]
Adger argues that the verb TAKE is spelled out as the progressive participle form taking in (7a) because
it agrees in aspect with the progressive auxiliary BE, and likewise that TAKE is spelled out as the
perfect participle form taken in (7b) because it agrees in aspect with the perfect auxiliary HAVE. On
this view, participial verb forms in sentences like (7) have their morphology determined by the choice
of auxiliary occupying the head T position in TP. If innate UG principles determine that main verbs are
universally inflected for aspect via agreement with an aspectual auxiliary, then it follows that a
seemingly T-less child structure like (8a) below must contain a TP headed by a null T constituent, as
shown in simplified form in (8b):
(8)(a) We been there (b) [ TP We [T ’ve] [VP [V been] there]]
We can then provide a principled account of why the verb BE is spelled out in the perfect participle
form been in (8a) by positing that BE agrees in aspect with a null variant of the perfect auxiliary HAVE
which occupies the head T position of TP – as claimed in (8b). Since adults would typically use the
contracted clitic form ’ve in structures like (8b), the fact that the child gives the auxiliary a null
spellout is consistent with our earlier assumption that young children often give clitics a null
spellout64.
Further evidence in support of the null T analysis comes from case-marking considerations. As we
see from the examples in (6) above, the subject pronouns used by the children in null-auxiliary
structures like I teasing Mummy/I given those to Daddy/We been there are nominative pronouns like
I/we. Since a nominative pronoun is used as the subject of a finite T constituent, this suggests that the
relevant sentence structures must contain a finite T constituent which assigns nominative case to the
subject. (Since we shall look at case-marking in some detail in the next chapter, we shall have no more
to say about it for the time being.)
A further piece of evidence which proponents of the structure-building model maintain lends
empirical support to the claim that early child clauses contain no functional categories like TP
concerns the fact that very young children use tenseless infinitive verb forms in contexts where adults
use tensed forms (i.e. where adults use verbs inflected for present/past tense), so that a child will say
Jane go home in a context where an adult will say Jane goes/went home. However, evidence from tag
sentences like (9) below calls into question the assumption that children’s apparent root infinitives
(i.e. main-clause infinitives) contain no T constituent:
(9)(a) He play with Laura, he does. Dolly fall over, Dolly does (Laura 2;6)
(b) Joan give it to you, didn’t he Ma? (Penny 2;3)
The conventional account of the use of the T/tense-marking auxiliary DO in tags is the following. The
head T constituent of the main-clause TP contains a tense/agreement affix which is copied onto the T
constituent in the tag, so forming a structure like that shown in skeletal form below (where Aff 3SG.PR
denotes a third person singular present tense affix):
(10) He [T Aff3SG.PR] play with Laura, he [T Aff3SG.PR]
If there is an adjacent main verb to the right of the affix, the affix will be lowered onto the verb by the
morphological operation of Affix Hopping: accordingly, the affix in the main clause (to the left of the
comma) in (10) is lowered onto the verb PLAY, forming the complex verb-form PLAY+Aff.SG.PR which is
ultimately spelled out in the adult grammar as plays. But when there is no adjacent verb to the right of
the tense affix for the affix to lower onto, the (meaningless) dummy/expletive auxiliary stem DO is
attached to the affix (via a morphological operation traditionally known as DO-support): consequently,
the stem DO is attached to the unsupported T-affix in the tag (following the comma) in (10), forming
the complex T constituent DO+Aff3SG.PR which is ultimately spelled out as does. Hence the syntactic
structure in (10) would be spelled out as He plays with Laura, he does in adult English.
In the light of the account of the syntax of tag sentences presented above, let’s return to consider
child sentences like (9). Given the standard T-copying analysis of auxiliaries in tags (under which the
tag contains a T constituent which is a copy of that in the main clause), the fact that the tag in these
sentences contains an (italicised) auxiliary overtly marked for tense/agreement suggests that the main
64
It would seem that adults sometimes do the same. For example, the Bates files on CHILDES show Zeke’s
mother saying to him at age 2;4 ‘You get a snack (be)cause you been a good boy’, where the clitic form ’ve of
HAVE has been given a null spellout.
clause must contain an abstract T constituent which comprises an affix carrying tense/agreement
features (precisely as in the corresponding adult structure in (10) above), with the affix in the tag being
a copy of the affix in the main clause. The only problem which then remains to be accounted for is
why the affix in the main clause which lowers onto the verb PLAY doesn’t result in PLAY being spelled
out by the relevant child as the third person singular present tense form plays (as in adult English), but
instead results in the bare form play. In this connection, it is interesting to note that Aldridge (1989,
pp.150-155) reports that many children go through a stage when they correctly inflect auxiliaries (but
not main verbs) for tense/agreement. One possibility is to suppose that such children have a
morphological deficit in that they have not yet acquired (or at any rate, not mastered the use of) the
regular third person singular present tense inflection -s or the regular past tense inflection -d, though
have acquired irregular present and past tense auxiliary forms (possibly because of their high
frequency of occurrence). At any rate, the occurrence of an overt T constituent in the tag in child
sentences like (9) provides evidence for the existence of an abstract T constituent in the main clause –
lending support to the continuity hypothesis that all child clauses (like all adult clauses) contain a T
constituent, so that there is continuity between the structure of clauses in adult and child English.
Moreover, the fact that the subject he has nominative case in (9a) He play with Laura, he does
provides further evidence in support of the assumption that the main clause He play with Laura
contains a null T constituent.
A further type of argument in support of positing that children’s initial sentences contain a TP
comes from word order facts – more particularly, from the position of subjects in child sentences. One
such piece of evidence comes from negative sentences. Although early child negative sentences often
contain no overt T (= tensed-marked auxiliary) constituent, they frequently show the word order
Subject+Negative+Verb…, as examples such as the following illustrate65:
(11)(a) Fraser not see him. Man no taste it? (Eve, 1;9-2;0)
(b) Kathryn no like celery. Man no go in there. Kathryn not go over there. Kathryn no fix this
(Kathryn 1;10-2;0)
(c) The horse not stand up. It not spill (Peter 2;6-2;7)
(d) Dis not fit. Dat one not bump dat one, no. Dat no blast off. (Adam 2;9-2;11)
If we suppose that UG principles specify that all clauses have a uniform CP+TP+(NEGP+)VP
structure, then the fact that the subject precedes the negative in sentences like (11) means that the
subject must occupy a position above NEGP; and since TP is the constituent immediately above NEGP
and subjects in English occupy the specifier position within TP, the obvious conclusion to draw is that
in a sentence like (11a) Fraser not see him, the subject Fraser must move out of its original position as
the semantic subject of the VP headed by the verb see across the intervening negative particle not to
become the syntactic subject of the TP headed by the null T constituent [ T ø], as shown by the arrow in
(12) below:
(12) [TP Fraser [T ø] [NEGP not [NEG ø] [VP Fraser [V see] him]]]

The fact that the subject Fraser ends up in front of the negative particle not provides evidence that it
occupies the specifier position in TP. More generally, negative sentences like those in (11) provide us
with empirical evidence of the existence of a TP projection in early child sentences.
A parallel kind of evidence relating to the position occupied by subjects in children’s sentences
comes from the syntax of a special class of intransitive verbs known as unaccusative verbs. These are
verbs like come/go/arrive/stay/remain/fall etc. whose superficial subjects originate as their
complements66 – and indeed, as Alison Henry (1995) notes, the (italicised) apparent subject of an
unaccusative verb like GO occupies the postverbal position typical of complements in imperative
structures like (13) below in Belfast English:

65
The examples in (11a) are from Bloom (1970); those in (11b-d) are from Harris and Wexler 1996)
66
The reason why the relevant verbs are known as unaccusatives is that an immediately following noun or
pronoun expression used as the complement of a typical transitive verb gets assigned accusative case; but the
complement of an unaccusative verb does not get assigned accusative case, but rather some other case (often
nominative).
(13) Go you to school! Leave you now! Arrive you before 6 o’clock!
Likewise, the (italicised) apparent subject of the Italian unaccusative verb ARRIVARE ‘arrive’ occupies
the postverbal position typically occupied by the complement of a verb in a sentence such as:
(14) È arrivato il treno
Is arrived the train (= ‘The train has arrived’)
As the example in (14) illustrates, unaccusative verbs also have the interesting property that (in
languages which use both HAVE and BE as perfect auxiliaries) they require the counterpart of BE as the
auxiliary marking perfect aspect. And indeed, this was also true of unaccusative verbs in earlier
varieties of English, as examples like (15) below (from various plays by Shakespeare) illustrate:
(15)(a) Mistress Page is come with me (Mrs Ford, Merry Wives of Windsor V.v)
(b) Is the duke gone? Then is your cause gone too (Duke, Measure for Measure, V.i)
(c) She is fallen into a pit of ink (Leonato, Much Ado About Nothing, IV.i)
Now if (as claimed under the structure-building model) the earliest clauses produced by one-year-olds
were simply verb phrases, then (since the apparent subjects of unaccusative verbs originate as their
complements), we should expect to find children producing verb-initial unaccusative structures like
those in (16) below, where the noun Daddy occupies the postverbal complement position associated
with the arguments67 of unaccusative verbs:
(16) [VP [V Gone] Daddy]
But if we look at the earliest unaccusative clauses produced by young children, we find that they
typically show the opposite (verb-second) order from that which the VP analysis in (16) would predict
– as the following examples (from Radford 1990, p.36) illustrate:
(17)(a) Daddy gone (Paula 1;6)
(b) Blanket gone. Geraint gone. Stick gone. (Bethan 1;8)
(c) Baby gone. Nana gone. Postman Pat gone (Jenny 1;9)
(d) Stick gone. Wayne gone Daniel 1;9)
(e) Biscuit gone. Hammer gone. Car gone (Angharad 1;10)
Since arguments of unaccusative verbs originate as their complements (in the sense of being a sister to
the verb)68, if sentences like (17) were simple verb phases (with no functional superstructure, and so
no TP) we should expect to find the verb-initial order shown in (16) above. How, then, do we account
for the verb-second word order illustrated in (17)? The Continuity Model offers us a straightforward
answer if we suppose that all child clauses (like all adult clauses) contain a TP constituent, and if (as in
adult English) the subject moves out of its original position as the complement of GO into the specifier
position within TP (in front of T), as shown by the arrow in (18) below:
(18) [TP Daddy [T ø] [VP [V gone] Daddy]]

If we further suppose that the null auxiliary [T ø] occupying the head T position of TP is the perfect
auxiliary HAVE, we can also account for the fact that the verb GO is spelled out in the perfect participle
form gone (aspectual agreement between T and V resulting in the perfect aspect feature carried by
HAVE being copied onto the verb GO, with the consequence that GO is ultimately spelled out in the
perfect participle form gone).
A final argument in support of the continuity view that all child clauses contain a T constituent
from the outset comes from children’s infinitive structures. Adherents of the structure-building model
maintain that the earliest infinitive clauses produced by young children lack infinitival to (which is
taken to be a non-finite T constituent in much recent work in syntax 69), and hence are VPs rather than

67
Argument is a cover term used to denote an expression which is the subject or complement of a verb (or, more
generally, of a predicate).
68
If we define a subject as ‘the highest argument of a verb’, we can equally say that the postverbal noun
expression in sentences like (17) is the subject of gone by virtue of being its highest (and only) argument.
TPs. However, if we look at the Eve files in the Brown corpus on the CHILDES data base, we find
that although frequently omitting infinitival to in structures like (19) below:
(19)(a) I want Mommy read (file 2, 1;6) (b) Want watch (file 2, 1;6).
(c) Sue gon read Lassie (file 7, 1;9)
Eve does occasionally use infinitival to (or forms like wanna and gonna which incorporate a
contracted form of to) in sentences like those below:
(20)(a) I don’t want to (file 3, 1;7) (b) Wanna see Eve (file 7, 1;9)
(c) Gonna write the couch (file 6, 1;9)
The occurrence of overt manifestations of infinitival to in structures such as (20) suggests that the
relevant infinitival structures in (20) contain a TP headed by infinitival to. But it then follows from
Rizzi’s (2000) Categorial Uniformity Principle that if some infinitival clauses contain a TP, all
infinitival clauses must contain a TP. This means that to-less infinitive clauses like those italicised in
(19) above must contain a TP headed by a null counterpart of infinitival to. Accordingly, the italicised
complement clause in (19a) will contain a TP with the structure shown in simplified form below:
(21) [TP Mommy [T to] read]
with strikethrough indicating that the infinitive particle to has a null spellout. Why should the
infinitive particle have a null spellout in sentences like (19)? A plausible answer is that it may have a
null spellout when it corresponds to the weak form |tə| or the clitic form |ə| - though not when it
corresponds to the uncontracted form |tu:|, as in (20a). Metrical factors (e.g. stress patterns) may play a
role here: Gerken (1991) reports that when asked to repeat sentences like He kissed her, young
children omit the subject pronoun he far more frequently than they omit the object pronoun him: this
suggests that a weak syllable immediately preceding a strong syllable is particularly vulnerable to
omission. In this connection, it is interesting to note that in the adult counterpart of the clause in (21)
Mummy to read, we find the stress pattern strong+weak+weak+strong – and so it is not surprising that
the italicised weak syllable containing to is deleted because it is immediately followed by the strong
syllable containing read. By contrast, in (20a), to follows (rather than precedes) a strong syllable
containing want.
The overall conclusion to be drawn from our discussion here is that contrary to the claim made in
the structure-building model that children’s initial clauses are simple verb phrases which contain no
functional categories like T, we have evidence from a range of different sources in support of positing
that all the sentences produced by English children (from the very onset of multiword speech) contain
a T constituent which projects into a TP. This is of course precisely as would be predicted under the
continuity view of language acquisition, if we suppose that among the UG principles wired into the
innate Language Faculty is a Uniformity Principle specifying that all sentences have a universal
CP+TP+(NEGP+)VP structure.

5.3 Evidence for a C constituent in children’s initial sentences

If all sentences do indeed have a universal CP+TP+(NEGP+)VP structure, and if this is part
of the innate knowledge of language genetically wired into the Language Faculty, then we should
expect to find evidence that from the very outset, children (tacitly) know that sentences contain a CP
layer of functional structure headed by a C constituent which marks clause type. Since CP is most
directly visible in main-clause questions in adult English (because the head C position of CP is filled
by an inverted auxiliary, and the specifier position in front of C is filled by a preposed wh-expression),
the obvious place to look for evidence of whether children’s initial sentences contain a CP constituent
is in their wh-questions. Moreover, if we assume that UG principles determine the landing-site of
moved constituents, and that spec-CP (i.e. the specifier position within CP) is universally the landing-
site for a preposed wh-expression, then we should expect to find evidence that (from the earliest
wh-questions they produce), children tacitly know that wh-questions contains an interrogative
wh-expression in spec-CP (i.e. occupying the specifier position within CP, in front of C).
The earliest wh-questions produced by one-year-old children typically include structures like those
69
Infinitival to often (though not always) marks futurity, as we see from the fact that a sentence like ‘I have
decided to give up smoking’ is paraphraseable as ‘I have decided that I will give up smoking’.
in (22) below of the form wh-word+(’s)+nominal70, with children alternating between producing and
omitting ’s, as the following examples (from Radford 1990, p.167) produced by a boy called Daniel
between ages 1;7 and 1;11 illustrate:
(22)(a) Where’s hand? Where’s chocolate? Where’s bike? Where’s Mummy? Where’s kangaroo?
Where’s hat? What’s that? What’s this?
(b) Where Carol? Where Teddy? Where Daddy? Where money? Where shoe? Where van?
Where bike? Where tractor? What this? What that?
Radford (1990) proposes a prefunctional analysis of such structures under which a child question like
Where(’s) Daddy would be a verb phrase with the simplified structure shown below:
(23) [VP Where [V ’s/ø] Daddy]
with ’s analysed as an invariable form of the copula BE which sometimes has a null spellout. However,
an analysis like that in (23) is problematic in certain respects. If we follow Radford (1990) in claiming
that where is misanalysed by the child as the subject of BE in questions like (23), we are involved in
positing an undesirable discontinuity between child and adult grammars, since where clearly functions
as the complement of BE in adult English, as we see from echo questions like Daddy is where? On the
other hand, if we posit that where is a preposed complement, then given the assumption made in the
structure-building analysis of questions like What Daddy doing? outlined in chapter 4 that a preposed
wh-complement is moved into a position to the left of the clause subject, we’d expect to end up with a
structure like that shown in simplified form below (with the arrow indicating wh-movement):
(24) [VP Where Daddy [V is/ø] where]

We would then expect that when the copula is overtly spelled out, it will be spelled out in a position
following the subject, so resulting in a sentence like Where Daddy is? (if the child knows that an
auxiliary at the end of a sentence is spelled out as a full form, not as a contracted form). But in fact
sentences like Where Daddy is? are unattested: if the copula is overtly spelled out, it is positioned
between the wh-word and the subject, as in Where’s Daddy? Thus, children’s copular wh-questions
like (22) arguably cannot be accounted for in a straightforward fashion within a structure-building
model which assumes that children’s earliest wh-questions are simple verb phrases.
By contrast, the continuity model assumes that child wh-questions have precisely the same
CP+TP+(NEGP+)VP structure as their adult counterparts. On this view, a sentence such as Where’s
Daddy would be derived as follows. The copular verb BE is merged with its complement where to form
the expression BE where, which is then merged with the subject Daddy to form the VP/verb phrase
structure shown in simplified form below:
(25) [VP Daddy [V BE] where]
The resulting VP is merged with a T constituent which contains a present-tense affix Aff. The affix
agrees with the subject Daddy and thereby comes to be marked as third person singular. The affix also
attracts the (auxiliary-like) copular verb BE to move from the head V position of VP to adjoin to the
affix in T; the affix also attracts the nominal Daddy which it agrees with to move to the specifier
position within TP (in front of T), so forming the TP shown in simplified form below:
(26) [TP Daddy [T BE+Aff3.SG.PR] [VP Daddy [V BE] where]]
The resulting TP is then merged with a C constituent containing a null question-marking particle Q (or
interrogative complementiser, to use the relevant technical term), and Q attracts the material in T to
move to C and likewise attracts the wh-word where to move to the specifier position within CP (to the
left of C), so deriving the CP shown in simplified form below (with strikethrough marking the old
positions occupied by the relevant constituents before they ended up in their new positions):
(27) [CP Where [C BE+Aff3SG.PR+Q] [TP Daddy [T BE+Aff3SG.PR] [VP Daddy [V BE] where]]]
The grammar ultimately spells out the string BE+Aff3SG.PR+Q in C as (i)s, so that the sentence
70
The term nominal is used here to mean ‘noun expression or pronoun’.
is realised as Where’s Daddy? if is cliticises onto where. Under the analysis in (26), the child is
assumed to have full adultlike competence, and the child’s question has precisely the same derivation
as its adult counterpart (with the minor difference that the inverted auxiliary may have a null spellout
in the child’s grammar, in accordance with our earlier observation that one-year-olds often give clitics
a null spellout).
Alongside copular wh-questions71 like What’s that? and Where’s Daddy? one-year-olds also
produce wh-questions such as What Daddy doing? and Where Daddy gone? Under the structure-
building analysis outlined in chapter 3, such sentences are analysed as simple verb phrases in which a
wh-word which originates as the complement of the verb is moved to the front of the overall verb
phrase (in front of the subject) to become a second specifier for the verb 72, resulting in structures such
as that below (with arrows showing how wh-movement applies):
(28)(a) [VP What Daddy [V doing] what] (b) [VP Where Daddy [V gone] where]

However, an analysis along the lines of (28) arguably violates principles of UG. If UG incorporates an
Edge Principle to the effect that a moved constituent can only move to the edge of (i.e. to the head or
specifier position in) a functional projection, the possibility of a wh-word being moved to the front of
a lexical projection such as a verb phrase is ruled out as impossible in principle 73.
By contrast, the continuity model faces no such problems in relation to sentences such as Where
Daddy gone? If we assume that arguments of unaccusative verbs originates as their complements, the
verb GO will merge with Daddy and where to form the verb phrase GO Daddy where74 (similar in
structure to a verb phrase like send Daddy there). This VP is in turn merged with a T constituent
containing the perfect-aspect T-auxiliary HAVE. Agreement in (perfect) aspect between the verb GO and
perfect aspect auxiliary HAVE results in the verb GO ultimately being spelled out in the perfect
participle form gone. The auxiliary HAVE in T agrees with Daddy (and so is spelled out as (ha)s in
adult English if used in a present-tense context), and attracts Daddy to become its syntactic subject
(thereby occupying the specifier position within TP, in front of the auxiliary has in T), so forming the
TP shown below:
(29) [TP Daddy [T has] [VP [V gone] Daddy where]]
This is then merged with an interrogative C constituent (containing a null question particle Q)
marking the overall sentence as interrogative in type. An interrogative C in a main clause attracts
whatever is in T to move to C, and likewise attracts the closest wh-word to move to spec-CP (i.e. to the
specifier position in CP, to the left of all other constituents of CP), so forming the CP shown below
(with strikethrough indicating the positions occupied by constituents before they moved to their
ultimate landing-site75):
(30) [CP Where [C has+Q] [TP Daddy [T has] [VP [V gone] Daddy where]]]
The auxiliary has can cliticise onto the wh-pronoun where, giving rise to Where’s Daddy gone? in
adult English. But since (as we noted in the previous section) very young children typically give clitic
auxiliaries a null spellout, the child’s counterpart will be Where Daddy gone?
Now, it might at first sight seem implausible to suppose that seemingly auxiliariless wh-questions
71
I.e. wh-questions containing the copular (‘linking’) verb BE.
72
The first specifier for the verb (i.e. the inner specifier closest to the verb) is the subject Daddy.
73
It might seem as if it is possible to overcome this objection by adopting the split projection analysis of verb
phrases such as that assumed by Chomsky in recent (1998, 1999, 2001) work (See Radford 2004, chapter 9 for
an outline of the split projection analysis). Under the split projection analysis, the wh-word can then be said to
move to the specifier position within a vP headed by an abstract light-verb, which (being a functional head) can
attract a wh-expression to become its specifier. A potential problem posed by this analysis, however, is that
Chomsky posits that only in transitive verb phrases can a light-verb attract a wh-expression to become its
specifier, not in intransitive verb phrases. However, Legate (2002) and Radford (2004b, chapter 10) present
evidence that preposed wh-expressions move to the left edge of the verb phrase even in intransitive clauses.
74
For a technical account of the derivation of such unaccusative sentences, see chapters 9 and 10 of Radford
(2004a,b). For present purposes, the relevant technical details are not important.
75
The landing-site for a constituent is the position it ends up in after it moves.
like Where Daddy gone? are actually CPs which involve invisible movement of an invisible auxiliary
from T to C. But there are two reasons for thinking that this is by no means implausible. For one thing,
adults allow certain clitic auxilaries to have a null spellout in rapid colloquial speech styles in
wh-questions such as the following:
(31)(a) What are you doing > What ya doin’? (b) Where have you been > Where ya been?
Moreover, alongside wh-questions with a null auxiliary, two-year-olds often produce wh-questions
with overt auxiliaries, as the following pairs of examples (from Radford 1992, p.46) illustrate:
(32)(a) What you doing? What are you doing? (Penny 2;0)
(b) Why you got your eyes shut? Why have you got it shut? (Tony 2;3)
(c) Where horse gone? Where’s Daddy gone, Mummy? (Tony 2;6)
(d) Where that go? Where does that one go? (Laura 2;8)
(e) What you say ‘Thank you’ for? What d’you say ‘Ow ow ow’ for? (Jonathan 2;9)
Alternations like those in (32) suggest that children’s wh-questions have the same CP+TP+VP
structure as their adult counterparts, but that children sometimes give weak unstressed (clitic)
auxiliaries a null spellout.
As we noted in chapter 4, Rizzi (2000) offers an alternative account of why auxiliaries have a null
spellout in structures like Where Daddy gone? in terms of the following condition:
(33) Null Spellout Condition
A constituent positioned at the edge (i.e. in the head or specifier position) of a root projection
(main clause) can receive a null spellout (in particular circumstances in particular languages)
On this view, has in a structure like (30) can receive a null spellout by virtue of occupying the head C
position in the main clause. Rizzi’s analysis can also account for why children sometimes produce
questions like those below (from Radford 1990, p.123) in which a wh-word receives a null spellout:
(34)(a) Bow-wow go? (‘Where did the dog go?’ Louise 1;3)
(b) You got? (‘What have you got?’ Harriet 1;6)
(c) Mummy doing? (‘What is Mummy doing?’ Daniel 1;9)
(d) Car going? (‘Where is the car going?’ Jem 1;9)
(e) Mouse doing? (‘What is the mouse doing?’ Paula 1;11)
If we assume continuity between the structure of child and adult wh-questions, a null-wh question like
Mummy doing? will be a CP with the structure shown in highly simplified form below:
(35) [CP What [C ’s] Mummy doing]?
Since CP is the root (i.e. topmost/leftmost) constituent in (35), and since what occupies the specifier
position within CP and (i)s occupies the head C position, both can receive a null spellout in
accordance with (33) by virtue of being positioned on the edge of a (main-clause) root projection.
The arguments presented in this section lead to the conclusion that the earliest wh-questions
produced by young children like Where Daddy? and Where Daddy gone? have precisely the same
derivation and the same CP+TP+(NEGP+)VP structure as their adult counterparts. This in turn (taken
together with the evidence in the previous section that early child clauses contain a TP constituent)
provides support for the key assumption underlying the continuity model that early child clauses
contain the same CP and TP layers of functional superstructure as their adult counterparts.

5.4 Evidence for functional categories in children’s initial nominals

As we saw in chapter 3, adherents of the structure-building model maintain that the earliest
nominals produced by children lack functional categories such as D. Consequently, they lack the
determiners used to modify nouns in adult nominals like the/a red car (instead producing bare
nominals like red car), and likewise they lack D-pronouns (e.g. personal pronouns like I/we/you/he
/she/it/they etc.) and instead produce structures containing nouns (e.g. Baby helping Mummy instead of
I’m helping you). However, the empirical basis of these claims seems questionable.
For example, if we look at the earliest of the Eve files in the Brown corpus on the CHILDES data
base, we find that (in contexts where it would be obligatory for adults to use a determiner to modify a
noun), Eve at age 1;6 alternated between producing bare (determiner-less) nouns like those italicised
in (36a) below, and determiner-modified nouns like those italicised in (36b):
(36)(a) Car coming. Where top? On wall. Doll eat celery. Spoon (reply to ‘What’s that?’)
(b) Read the puzzle. Like the farm. On the table. Drop (th)e cheese. Have a piece.
Nor does Eve seem to be untypical in this regard. Thus, on the basis of analysing a corpus of 17
English children studied longitudinally at 3-month intervals between ages 1;6 and 3;6, Abu-Akel and
Bailey (2000) – henceforth AAB – maintain that even in the earliest corpora there is evidence of the
children making sporadic use of the definite/indefinite determiners the/a. More specifically, they claim
(ibid, p.54) ‘There is no stage at which children exclusively leave out all determiners in obligatory
contexts’; instead, they maintain that ‘The use of determiners in the early stages is optional’ and that
‘when present, they are used correctly’ (ibid.). For example, AAB report a mean suppliance rate of
19% for articles in obligatory contexts by the children in their study at age 1;6 (so that 19% of the time
the children would produce nominals like e.g. the/a car and 81% of the time they would produce bare
nominals like car). If we suppose that when the children produce a nominal like the/a car they are
producing a DP of the form (37a/b) below, Rizzi’s (2000) Categorial Uniformity Principle would
lead us to suppose that when they say simply car, they are producing a DP of the form (37c), headed
by a null determiner:
(37)(a) [DP [D the] car] (b) [DP [D a] car] (c) [DP [D ø] car]
A number of factors might lead to the determiner receiving a null spellout in structures like (37c) – not
the least of these being the fact that the indefinite article a is generally realised as a weak, short
unstressed, perceptually non-salient vowel schwa |ə|, and the definite article is most frequently realised
as the form |đə| which contains the same weak schwa vowel (and an interdental fricative consonant
which children have considerable problems in articulating until around 3 years of age).
Likewise, the claim that one-year-olds do not use D pronouns would appear to be undermined by
the fact that they frequently use the demonstrative pronouns that/this and the third person singular
personal pronoun it as the following examples (from Radford 1990, p.102) illustrate:
(38)(a) Read that. That Luke, Mum (= ‘That’s Luke, Mum’). What this? (= ‘What’s this?’)
(b) Get it. Catch it. Open it (Lucy 1;8)
The use of it often seems to alternate with the use of a null pronoun, as in the following examples
(from the Bloom corpus on the CHILDES database)
(39)(a) Close it. Close. Fix it, Fix. Kick. Kick it. Open. Open it (Peter 1;9-1;10, from Bloom
corpus)
(b) Spill it. Spill (Allison 1;10, from Bloom corpus)
Rather than positing that one-year-olds have no D-pronouns, it might seem more plausible to make a
set of assumptions along the following lines. Principles of UG tell the child that all languages have D
pronouns which can be used to refer to a familiar entity mentioned in (or readily identifiable from) the
immediate discourse context. Until such time as they learn how such pronouns are spelled out,
children give D-pronouns a null spellout (by default76). Moreover, even when they acquire the relevant
D-pronouns, children may initially have problems in retrieving newly acquired pronoun forms and so
alternate between overt and null pronouns – as in (39) above 77. It may be that English D-pronouns like
that/this/it tend to be acquired early because of their relatively stable phonological form (e.g. they
76
Since a complementiser like that can have a null spellout in a sentence like I know that he is lying, a T
constituent like infinitival to can have a null spellout in a sentence like I have never known him to tell a lie, and a
D constituent like the can have a null spellout in a sentence like The Italians love the opera, it is clear that UG
principles must allow for the possibility that functional heads (like C, T and D) can be null under certain
circumstances. An interesting possibility is that functional heads receive a null spellout by default (i.e. in
structures where the conditions for using an overt constituent are not met). If so, we would precisely expect that
(e.g.) until such time as a child learns to spell out a third person singular neuter pronoun as it, the child will give
the pronoun a null spellout by default.
77
Fujino and Sano (2000) report that (Spanish-speaking) Maria used only null object pronouns from 1;7 to 1;11,
and then alternated between null and overt object pronouns from 2;0 to 2;4, before ceasing to use null object
pronouns at 2;5
don’t show nominative/accusative case alternations like irregular pronouns such as I/me, he/him,
they/them etc.). Moreover, if (as Kayne 1989 suggested) there are only two persons (first and second)
and so-called ‘third person’ expressions are actually expressions which lack person, it may be that the
absence of person in pronouns like this/that/it makes them easier to acquire than first-person pronouns
such as I/me/my/mine or second-person pronouns such as you/your/yours. After all, the referential
properties of first and second person pronouns are known to cause problems for young children, as
illustrated in the following dialogue (from Jespersen 1922, p.58):
(40) ADULT: Shall I carry you? CHILD: Carry you (Frans, 1;9)
Instead of using the first person pronoun me to refer to himself, the child in question uses the same
second person pronoun you which the adult used to refer to the child. It would seem that the child has
not yet mastered the referential properties of person (in particular, how person switches when speakers
switch) – suggesting that person is a conceptually complex feature to acquire.
A further type of functional category which supporters of the structure-building model claim to be
absent in early child grammars are case particles like genitive ’s. However, a detailed longitudinal
study of the Eve files in the Brown corpus on the CHILDES data-base by Courtney Cazden (1968)
showed that far from not using the genitive case suffix ’s at all in possessive structures, Eve between
ages 1;6 and 2;0 used genitive ’s in 13% of obligatory contexts. A reasonable inference to draw from
Cazden’s findings is that rather than genitive ’s being entirely absent from early child grammars, it is
sporadically used from the outset. If some child possessive expressions are KPs (i.e. case projections)
headed by the genitive K-particle ’s, Rizzi’s (2000) Categorial Uniformity Principle would suggest
that all child possessive structures should be analysed as KPs, with the head K constituent generally
having a null spellout because it is a phonetically insubstantial affix which (in a structure like Daddy’s
car) is realised as the (phonetically insubstantial) consonant segment |z|.

5.5 A critical perspective on the continuity model

In many ways the continuity/full competence model (which assumes that children have full
adult-like competence from the very onset of multiword speech, and that there is continuity between
child and adult grammars) is very attractive. For example, it manages to provide a (seemingly
descriptively adequate) account of the syntax of a wide range of structures produced by children while
maximising continuity with adult grammars, by assuming that children’s sentences have the same
structure as their adult counterparts, save that weak functors often have a null spellout in child
grammars (because of the perception or production problems they pose). Such an approach avoids the
learnability problems which beset any account which assumes that child grammars are radically
different in nature from adult grammars (in that we avoid the discontinuity problem of having to
explain how children initially acquire the ‘wrong’ structures, but then come to unlearn these structures
and later manage to acquire the ‘right’ structures). And yet, there are a number of searching questions
which need to be asked about the continuity approach.
From a conceptual perspective, a question which needs to be asked is whether it is intuitively
plausible to suppose (as the continuity model does) that there is no structural learning involved in the
acquisition of syntax (because all functional structure is assumed to be in place from the onset of
multiword speech). It might be argued that the structure-building model (which assumes that children
gradually build up more and more complex syntactic structures) is more intuitively plausible.
Moreover, questions arise about just how plausible it is to posit that all adult functional categories
are already in place by the onset of multiword speech, when most of them have a null spellout most of
the time (and indeed some may always have a null spellout in the very earliest stages). We can pose
this question in an extreme form in relation to an utterance like At? (the word that typically being
articulated by one-year-olds as dat or at) produced by very young children (around the age of 1;6) in a
context where adults would use a wh-question like What is that? The corresponding adult wh-question
would be derived as follows. The copula verb BE is merged with its complement what and its subject
that to form the verb phrase [VP that [V BE] what]. This in turn is merged with a present-tense T
constituent, which attracts BE to raise from V to T, and which agrees with (and triggers movement to
the specifier position within TP of) the subject that, so forming the TP [TP that [T is] what]78. The

78
This structure is simplified for expository purposes by not showing null constituents.
resulting TP is then merged with a C containing a question particle Q which attracts is to move from T
to C, and attracts what to move to the CP-specifier position CP, forming the CP [ CP What [C is] that]79.
The child’s utterance At? is assumed to have exactly the same derivation as it adult counterpart What
is that? except that the Null Spellout Condition (33) allows constituents at the edge of a root (= main-
clause) CP to have a null spellout; hence what has a null spellout by virtue of occupying a root
specifier position, and is has a null spellout by virtue of occupying a root head position. Thus, the
child’s counterpart is What is that? with the first two words being given a null spellout (indicated by
the use of outline font), and that being pronounced as at. But the crucial point is that child’s one-word
utterance At? is assumed to have the same derivation as the corresponding adult sentence.
And yet, there is a considerable amount of abstraction involved in such an analysis. After all, the
wh-word what is invisible, as is its supposed movement from VP-complement position to CP-specifier
position; likewise, the copula is also remains invisible throughout, as does its movement from V
through T into C. Equally invisible is the nominative case assigned to the subject (th)at by the
invisible T constituent containing an invisible counterpart of is…and so on. The point should be clear:
the continuity analysis entails that most of the syntactic structure which is claimed to exist (and the
syntactic operations which are claimed to take place) in sentence formation by one-year-olds is
invisible: the assumed existence of a vast panoply of insisible constituents and invisible operations
could be argued to be based on little more than wishful thinking (or methodological madness). A
simple analogy from adult English serves to reinforce the point. If a surgeon performing an operation
turns to a nurse and says ‘Scalpal’, the most straightforward account of the syntax of the surgeon’s
utterance is to say that all he produces is a bare noun, and a complex system of pragmatic inferencing
then leads the nurse to infer that he is requesting a scalpal. No syntactician (with any common sense)
would want to say that the utterance scalpal is in fact a CP of the form Would you mind passing me the
scalpal? but with all the words except the last one being given a null spellout. Why should we treat the
cryptic utterances of one-year-olds any differently?
The plausibility problem emerges in its most acute form in the case of functional categories which
always have a null spellout. In this connection, consider the suggestion made in §5.4 that UG may
‘tell’ children that all languages have D-pronouns (i.e. personal pronouns), with children giving such
pronouns a null phonetic form until such time as they learn to spell them out as overt forms like
me/you/him etc. However, we noted in §5.2 that Fujino and Sano (2000) report that a Spanish child
called Maria whose progress they followed from age 1;7 to age 2;5 consistently used no (clitic)
D-pronouns from age 1;7 to 1;11. For children like Maria (at the relevant stage of development) the
continuity analysis asks us to suppose that she has acquired D-pronouns, even though she never
produces any overt D-pronoun at the relevant stage. Positing the existence of a set of intrinsically null
functional categories which are always invisible might be said to have null plausibility.
Moreover, even in cases where the existence of null functors is claimed to be supported by
evidence that the relevant functors are sometimes overt, the empirical evidence in support of positing a
plethora of empty categories in children’s initial utterances is far from convincing. Take, for example,
the claim by Abu-Akel and Bailey (2000) that the children in their study at age 1;6 had a mean
suppliance rate of 19% for articles in obligatory contexts. What this of course means is they they
omitted articles four times as often as they used them. It seems somewhat perverse to conclude from
the fact that a tiny minority of their noun expressions at the relevant stage contained an article that all
of them contained an article, but with the article being given a null spellout in 81% of obligatory
contexts. It might be thought more plausible to suppose that the 19% of their nominals which contain
articles are D-expressions, whereas the 81% which don’t are simply N-expressions, and that the
children concerned are in a transitional phase (between an earlier prefunctional stage and a later
functional stage) in which they sometimes project N-expressions further into D-expressions, and
sometimes do not. This of course is essentially parallel to the approach to clause structure taken by
Luigi Rizzi in his truncation model (which assumes that very young children sometimes do project
functional categories, and sometimes don’t).
What undermines the continuity analysis still further is the fact that young children make frequent
morphosyntactic errors which appear hard to reconcile with the view that they have full adult-like
competence. As a case in point, consider the following two negative sentences (from the Wells corpus

79
Once again, the structure is simplified for expository purposes by not showing null constituents.
on the CHILDES data-base) produced by a girl called Iris at age 3;2:
(41) No me got one. No me got him (Iris, 3;280)
The fact that the pronoun me follows the negative particle no in both cases suggests that me is
positioned internally within VP as the subject of got. If we adopt the continuity view that child clauses
have the same CP+TP+(NEGP+)VP structure as their adult counterparts, a sentence like No me got
one will have the structure shown in (42) below:
(42) [CP [C ø] [TP [T ø] [NEGP no [NEG ø] [VP me [V got] one]]]]
But such a continuity analysis raises a number of important questions. For example, if the child’s
clause contains the same kind of finite T constituent as in adult grammars, why doesn’t this T
constituent assign nominative case to the subject me (requiring it to be spelled out as the nominative
form I), and why doesn’t T also attract I to become its syntactic subject, as in the corresponding adult
structure (43) below:
(43) [CP [C ø] [TP I [T have] [NEGP not [NEG ø] [VP [V got] one]]]]
(yielding I haven’t got one if not cliticises onto have). The fact that subject me in (42) is not assigned
nominative case and does not move out of VP into TP raises the question of whether child sentences
like (41) do indeed contain a T constituent – and if they do, why the child’s T constituent doesn’t
behave like its adult counterpart in assigning nominative case to (and attracting movement of) the
subject me. In the next chapter, we look at one way of answering this type of question.

Workbook section

5.1 Claire’s wh-questions

Below are listed the various types of wh-question structure found in the first three files from the Claire
corpus in the Appendix to this book, produced when Claire was age 2;0 (superscript numbers
indicating the number of the relevant utterance as it appears in the corpus). Discuss how the continuity
analysis might handle the syntax of the relevant questions, and any what problems (if any) arise.

1a What that?25,26,32,56,63,100,102,112,130,151,171,175,178,198,210,213,215,261,263,268
b What this?60,67,205
c What’s that?28,53
d That?185, 203 (= What’s that?)
e Animal that?211 (= What animal is that?)
2a What happen85?
b Happen weasel271?
3a What kitty doing188? What the dog doing217? What squirrel doing260? What lizard doing265?
b What doing6,30,33?
c Bunnies doing272?
4a Wheresa car75? Wheresa mommy94? Wheresa cow105? Where’s the bird119?
b Where chicken3? Where Mommy7,291? Where doggie18? 37 Where pencil37? Where people40,43,62?
Where Daddy47? Where car50? Where chicken88? Where the people156? Where Claire pencil165,166?
Where cow187? Where it208? Where two247? (= ‘Where’s the second one?’). Where chair249?
Where blinkety mole262? Where Claire horse298?
5a Where girl go34? Where pencil go35? Where cow go79,80? Where Daddy go182? Where bathtub go190?
b Where go65,168,170,171? Where tickle276?
c Chair go4? Kitty go20? Mommy gone288?
If you wish to be more comprehensive, you can take into account the full range of wh-question
structures found in the complete set of files in the Claire corpus in the Appendix, rather than just the

80
The first sentence is glossed in in the relevant file as ‘I haven’t got one’, and the second as ‘I haven’t got him’.
three files taken into consideration here.

5.2 Children’s negative sentences

Harris and Wexler (1996) claim that there is continuity between the structure of negative clauses in
child and adult English. Evaluate this claim in relation to the children’s negative sentences given
below (produced by a variety of children at a variety of ages 81):
1. No dat blast off. Dat no blast off (Adam 2;11)
2. Not Fraser read it. Fraser not see him (Eve 1;9 and 2;0)
3. It not works (Abe 2;8)
4. No Nathaniel has a microphone (Nathaniel 2;5)
5. No goes in there. But the horse not standups. That didn’t went down (Peter 2;1, 2;7 and 2;8)
6. The horse not stand up (Peter 2;6). Man no taste it? (Eve 2;0).
7. I no reach it. I no like to (Eric 2;1). I not smell (Gia 2;3). I not have some fruit (Kathryn 2;0)
8. This not fits. This one don’t fit. This one don’t fits (Kathryn 2;0)
9. No did it fall down (Cathy 2;3)
10. You go not along (Bill 2;6). He drives not in the way (Lucy 2;10)
11. He not weeing in there (Bill 2;5)
12. Her not play piano (Sophie 2;5)
13. Me no go home (Peter 2;1)
To what extent does the continuity analysis provide us with a plausible account of such sentences? And
to what extent does the notion of underspecification help us deal with problematic cases?

5.3 Data from the Adam and Iris files

Consider the extent to which data from the Adam files in exercise 4.1 (from the previous chapter) and
the Iris files in exercise 4.3 are consistent with the Continuity Model.

6. The Underspecification Model

6.1 Introduction
In the previous chapter, chapter, we looked at evidence in support of the continuity view
that the earliest syntactic structures produced by young children contain the same range of functional
categories (in particular, the same type of T, C, D and K constituents) as their adult counterparts. But,
as will have been obvious from some of the child utterances discussed in chapter 5, the relevant
children’s sentences are often very different in their superficial spellout form from their adult
counterparts, and contain what are (from an adult perspective) a number of errors. This raises the
question of how (if child utterances have the same structure as their adult counterparts), we can
account for the types of error produced by young children. In this chapter, we look at a range of types
of error produced by young children, and at attempts to account for these in terms of the assumption
that (while children’s clauses and nominals contain the same functional superstructure as their adult
counterparts), their functors (i.e. functional heads) are sometimes underspecified for (i.e. lack) one or
more of their features. Since the underspecification model posits that children sometimes omit
grammatical features on functors, it amounts to a feature-omission model.

6.2 Errors relating to finite verbs

As we saw in the previous chapter, children initially go through a stage (which typically
lasts until some point between their third and fourth birthdays) during which they alternate between
producing finite verb forms (like those italicised in ‘Daddy plays/played with me’ and bare infinitive-
like verb forms like that italicised in ‘Daddy play with me’). Ken Wexler has termed this the Optional
81
The examples in 1-6 are from Harris and Wexler 1996; those in 7-8 are from Bloom 1970; those in 9-11 are
from Anderssen 1996; that in 12 is from Ingham (1998).
Infinitive/OI stage, and (along with various collaborators) has spent many years years researching the
nature of the OI stage (see e.g. Wexler 1994, 1998, 1999; Poeppel and Wexler 1993; Bromberg and
Wexler 1995; Harris and Wexler 1996; Schütze and Wexler 1996; Schütze 1997; Wexler, Schütze and
Rice 1998). The key question which arises within a continuity framework which assumes that children
have full (adult-like) competence from the outset is how (if child clauses contain the same T
constituent as their adult counterparts) children come to produce bare verb forms in contexts where
adults require finite verb forms inflected for tense and agreement. The answer given by Wexler and his
co-researchers is that although children often correctly mark T for tense and agreement in main clauses
(so resulting in a structure containing a finite verb inflected for tense and agreement), they sometimes
omit the tense and/or agreement features of T (resulting in a structure containing a bare verb). For
obvious reasons, Wexler, Schütze and Rice (1998) refer to this as the ATOM (= Agreement and
Tense Omission) model. In this section, we look at how the ATOM model accounts for the
morphology of verbs and auxiliaries in the Optional Infinitive stage 82.
Consider first how ATOM accounts for children alternating between s-inflected main verbs like
plays and infinitive-like bare verbs such as play in sentences like Daddy plays/play with me (e.g. in
reply to a question like ‘What does Daddy do when he comes home?’). When the child produces the
finite form plays, the relevant clause is assumed to contain a TP headed by a T constituent carrying a
tense/agreement affix fully specified for both (present) tense and (third person singular) agreement, as
shown in highly simplified form below83:
(1) [TP Daddy [T Aff3SG.PRES] play with me]
The affix is lowered onto the verb PLAY by the morphological operation of Affix Hopping, forming
[V PLAY+Aff3.SG.PR]. The continuity model assumes that there is continuity between adult and child
grammars in respect of how affixes are spelled out, so that a child who has acquired all the relevant
verbal affixes found in adult English will spell out affixes on regular verbs in in the manner shown
informally in (2) below:
(2) Affix spellout
An affix on a regular verb is spelled out as84:
(i) -s if third singular present
(ii) -d if past, perfect, or passive
(iii) -ing if progressive, or gerund
(iv) -ø otherwise (by default)85
It follows that the third person singular affix on the end of the verb in [ V PLAY+Aff3SG.PR] will be spelled
out as -s in accordance with (2i), with the result that the verb is ultimately spelled out as plays. On this
view, children’s finite verb-forms arise in precisely the same way as their adult counterparts and
affixes are spelled out on regular verbs in precisely the same way as in the adult grammar (underlining
the assumption of continuity between adult and child grammars).
The T constituent in (1) is fully specified for both tense and agreement, in accordance with the
ATOM assumption that children sometimes fully specify T for tense and agreement in finite contexts
(i.e. in contexts where adults would use a T constituent fully specified for both tense and agreement).
However, a key assumption of ATOM is that (in finite contexts) children sometimes omit either the
tense feature on T as in (3a) below, or its agreement features as in (3b), or both as in (3c):
(3)(a) [TP Daddy [T Aff3SG] play with me]
(b) [TP Daddy [T AffPRES] play with me]

82
The account presented here is based on Wexler, Schütze and Rice (1998). Slightly different versions of the
account are given in Bromberg and Wexler (1995), Schütze and Wexler (1996), and Schütze (1997).
83
Here and in other structures below, I simplify exposition by not showing that the subject originates internally
within the verb phrase as the semantic subject of the main verb, by now showing the CP layer of structure above
TP, and so on.
84
The discussion throughout is simplified by ignoring the various allomorphs (i.e. variant forms) of the relevant
inflectional affixes (e.g. that present tense {s} is pronounced as |s| in helps, |z| in plays and |Iz| in watches).
85
This can be thought of either as meaning that a null affix is added to the verb stem, or that no affix is added.
 (c) [TP Daddy [T Aff] play with me]86
In each case, the affix in T will subsequently be lowered onto the verb PLAY. But in none of these
structures can the affix be spelled out as -s because (2i) tells us that -s is only used when the affix
carries both tense and agreement features, and in all three structures in (3) the affix lacks either one or
both of these features. Accordingly, the affix in all three cases will receive a null spellout by default, in
accordance with (2iv), and the verb will be spelled out as the infinitive/default form play.
Now consider how we account for children at the OI stage alternating between a past tense form
like played and a bare infinitive-like form such as play in sentences like Daddy played/play with me
yesterday. Since ATOM claims that children sometimes produce a T constituent fully specified for
tense and agreement in finite contexts, one possibility is that a child at the OI stage may produce a
structure containing a TP such as that below:
(4) [TP Daddy [T Aff3SG.PAST] play with me yesterday]
The affix in T will then be lowered onto the verb PLAY by Affix Hopping, and will be spelled out as
-d in accordance with (2ii), which says that -d is used when an affix carries a past tense feature,
irrespective of whether or not the affix carries additional (e.g. agreement) features or not. Again, things
work exactly as in the adult grammar.
But of course, the core assumption of ATOM is that children sometimes leave the affix in T
underspecified for (i.e. lacking) either tense, or agreement, or both – so that alongside TPs containing
a fully-specified T like (4) above, children at the OI stage will produce TPs like those in (5) below
containing an underspecified T:
(5)(a) [TP Daddy [T AffPAST] play with me yesterday]
(b) [TP Daddy [T Aff3SG] play with me yesterday]
(c) [TP Daddy [T Aff] play with me yesterday]
Once again, the affix will subsequently be lowered onto the verb by Affix Hopping. In a structure like
(5a) where T is marked for past tense but not agreement, the affix will be spelled out as -d in
accordance with (2ii); hence the verb will be spelled out as the past tense form played. But in
structures like (2b) and (2c) where T carries no tense feature, the affix will receive a null spellout by
default in accordance with (2iv), resulting in the bare verb form play.

6.3 Errors relating to auxiliaries

A further characteristic of the OI stage is that children alternate between giving an auxiliary
or copula in T an overt spellout (as in Daddy’s working) or a null spellout (as in Daddy working). Why
should this be? The answer provided by the ATOM model is outlined informally in (6) below:
(6) Auxiliary spellout
An auxiliary in T like BE/HAVE/DO which overtly inflects for both tense and agreement is given
an overt spellout when fully specified for both tense and agreement, but a null spellout when
underspecified (i.e. when lacking one or more of its tense/agreement features).
Let’s take a look at how (6) works in practice.
Suppose first of all that (in accordance with the ATOM assumption that children sometimes fully
specify T for tense and agreement), a child produces a sentence structure containing a TP like that
shown in simplified form below:
(7) [TP Daddy [T BE3SG.PRES] working]
Because the auxiliary BE is fully specified for (present) tense and (third person singular) agreement,
it is overtly spelled out as (i)s in accordance with (6) above.
However, ATOM maintains that children sometimes fail to mark tense and/or agreement on T.
Accordingly, alongside structures like (7) containing a fully specified T, children will also produce
86
An interesting question raised by the assumption that T can be specified for neither tense nor agreement is
what feature(s) T carries in such cases. One possibility is that T has some discourse-determined tense value, but
that the relevant tense feature is somehow is ‘invisible’ in the syntax, morphology and phonology. I return to this
question in a later section.
structures like (8) below containing a T constituent underspecified for tense, or agreement, or both:
(8)(a) [TP Daddy [T BE3SG] working]
(b) [TP Daddy [T BEPRES] working]
(c) [TP Daddy [T BE] working]
Such sentences must contain an abstract form of the progressive auxiliary BE, since the verb WORK is
in the progressive participle form working and Adger (2003) posits that a main verb acquires a
progressive participle feature via agreement with a progressive auxiliary. But if the structures in (8)
contain the progressive auxiliary BE, why does BE receive a null spellout in accordance with (6)? For
example, why should BE in (8b) receive a null spellout? After all, it contains a present-tense feature, so
why can’t the child simply go to her mental lexicon and retrieve an appropriate present tense form?
The problem is that although the child’s lexicon says that a first person singular present-tense form of
BE is spelled out as (a)m, a third person singular present-tense form as (i)s, and a second person or
plural present-tense form as (a)re, the lexicon simply contains no entry whatever for an agreementless
present-tense form of BE which is not specified for (person/number) agreement features. In cases
where the lexicon provides no means for spelling out a functor like BE, it is given a null spellout by
default. And indeed, we find the same null spellout for BE in adult English, in so-called ‘Mad
magazine’ sentences like that produced by speaker B (and italicised) below:
(9) SPEAKER A: The computer is working again! SPEAKER B: The computer working?! Great!
Hence, the same spellout mechanism which accounts for why BE receives a null spellout in adult
tenseless/ageementless sentences like that italicised in (9 B) can be invoked to account for why BE
receives a null spellout in tense- or agreement-underspecified child structures like (8).

6.4 Case-marking errors

In addition to making spellout errors on verbs and auxiliaries, children at the OI stage also
make errors relating to the case-marking of subjects. More specifically, children at the OI stage are
reported to alternate between using nominative subjects (as in I want one) and accusative subjects (as
in Me want one). Within the ATOM model, case errors on subjects are said to be the result of T being
underspecified for agreement. Let’s see how this works.
Suppose that the Case Assignment Conditions found in adult English are along the lines
presented in simplified form in below:
(10) A nominal (i.e. noun or pronoun expression) is assigned
(i) nominative case if in the domain87 of an agreeing T constituent
(ii) genitive case if in the domain of a genitive-case-assigning head 88
(iii) accusative case if in the domain of an accusative-case-assigning head (e.g. a transitive
V, P or C), or if not in the domain of any case-assigning head 89

We can illustrate how these case conditions work in relation to the following dialogue:
(11) SPEAKER A: I have heard that he is not taking the exams
SPEAKER B: Him not taking them? Wow!

In (11A), the subject pronouns I/he are assigned nominative case via (10i) by the present-tense T
constituents have/is which agree with them in person and number. In (11B) the object them is assigned
accusative case via (10iii) by virtue of falling within the domain of the transitive verb taking. The
subject him in (11B) is assigned default (accusative) case va (10iii) by virtue of not falling within the
domain of any case assigner: although the subject him in (11B) occupies the specifier position within
TP (as we see from the fact that it precedes the negative particle not) the relevant T constituent is

87
More fully, ‘in the local c-command domain of’ a (i.e. c-commanded by a nearby) head of the relevant kind
88
Abney (1987) argued that possessors are assigned genitive case by a null determiner ‘higher up’ in the nominal
containing them. So, in a structure like [DP [D ø] [PossP John [Poss ø] car]], John is assigned genitive case by a null
determiner above (and to the left of) it.
89
The case assigned to a noun or pronoun expression which does not fall within the domain of any case assigner
is termed the default case: the default case is accusative in English, but nominative in German (Schütze 2001).
unspecified for agreement (which is why it has a null spellout in accordance with (6) rather than being
spelled out as is), and this means that T cannot assign nominative case to the subject; and since him
does not immediately follow a transitive verb either, it is in a position where it does not fall within the
domain of any case assigner, and so receives default accusative case by (10iii).
If (as the ATOM model assumes) there is continuity between the set of case conditions operating in
child and adult grammars (so that the adult case conditions (10) are operative in child grammars from
a very early stage), then it follows that we should expect to find a correlation between the case
properties of subjects and the agreement properties of T. More specifically, in contexts where T is
specified for agreement, its subject will be assigned nominative case in accordance with (10i); and in
contexts where T is underspecified for agreement (i.e. its agreement features are omitted), its subject
will be assigned default accusative case by (10iii). What all of his predicts (in relation to the various
types of child clause discussed in the previous section) is the following.
In contexts where adults would use a present-tense main verb like plays, what we expect to find is
the following. In an adult-like structure such as (1) [ TP Daddy [T Aff3SG.PRES] play with me], the subject
will be assigned nominative case by the agreeing (3rd person singular) T constituent in accordance with
(10i), predicting that children should produce adult-like structures such as He plays with me. In (3b)
[TP Daddy [T Aff3SG] play with me], the subject will again be assigned nominative case by the agreeing T
constituent, predicting that children will also produce structures like He play with me. But in (3a)
[TP Daddy [T AffPRES] play with me] and (3c) [TP Daddy [T Aff] play with me], the subject will be
assigned accusative case by default because T lacks the agreement features required for the assignment
of nominative case. What all of this predicts is that in contexts where adults say He plays with me,
children at the OI stage will alternate between saying He plays with me, He play with me and Him play
with me.
Now consider what happens in contexts where adults use a past tense main verb like played. In a
structure like (4) [TP Daddy [T Aff3SG.PAST] play with me] the subject Daddy will be assigned nominative
case by the agreeing (3rd person singular) T constituent, predicting that if Daddy is replaced by a
pronoun, a child will say He played with me. In (5b) [TP Daddy [T Aff3SG] play with me], the subject will
again be assigned nominative case by the agreeing T constituent, predicting that children will also
produce structures like He play with me. In (5a) [TP Daddy [T AffPAST] play with me], the subject will be
assigned accusative case by default because T lacks the agreement features required for the assignment
of nominative case, so predicting that children will say Him played with me. Finally, in (5c) [TP Daddy
[T Aff] play with me], the subject will again be assigned accusative case by default (because T lacks
agreement features), and the verb will surface as a bare form (because T also lacks tense features),
predicting that children will say Him play with me. Thus, in contexts where adults would say He
played with me, the ATOM model predicts that children at the OI stage will alternate between saying
He played with me, He play with me, Him played with me and Him play with me.
Next consider the case-marking of the subject in contexts where adults would use an auxiliary like
(i)s – e.g. in Daddy’s working. In a structure like (7) [TP Daddy [T BE3SG.PRES] working], the subject will
be assigned nominative case by the agreeing T constituent, so that we expect to find children
producing sentences such as He’s working. In (8a) [TP Daddy [T BE3SG] working], the subject is again
assigned nominative case by the agreeing T, but BE is given a null spellout in accordance with (6) by
virtue of lacking tense, so resulting in sentences such as He working. In structures like (8b)
[TP Daddy [T BEPRES] working] and (8c) [TP Daddy [T BE] working] the subject is assigned accusative
case by default because T carries no agreement features, and the auxiliary is given a null spellout by
(6) because it is underspecified, so resulting in sentences such as Him working. Thus, in contexts
where adults produce sentences such as He’s working, the ATOM model predicts that children will
alternate between producing structures such as He’s working, He working and Him working.
Rather different in respect of the ATOM model’s predictions about how they case-mark their
subjects are modal auxiliaries like can/could, will/would, may/might etc. These are forms which
visibly inflect for tense (can being a present-tense form, and could being its past tense counterpart),
but which don’t overtly inflect for agreement. Let’s therefore suppose that the lexical entry for a modal
auxiliary like CAN (in adult and child grammars alike) says something along the following lines:
(12) CAN is spelled out as can if specified as present-tense, and as could if specified as past tense
(irrespective of whether or not CAN also carries agreement features)
In this connection, consider how we would expect a child to case-mark the subject in a modal sentence
such as Daddy can swim. Suppose that a child fully specifies the T constituent containing CAN for both
tense and agreement (as would be required in adult English, where a finite T always carries both tense
and agreement), as shown in simplified form below:
(13) [TP Daddy [T can3SG.PRES] swim]
In such a case, the subject would be assigned nominative case by the agreeing (third person singular) T
auxiliary CAN, and the auxiliary would be spelled out as can in accordance with (12) by virtue of
carrying a present-tense feature: if the subject Daddy were replaced by a pronoun, we’d therefore
expect to find a nominative pronoun such as he, giving rise to He can swim. But now suppose that (in
accordance with the central claim of the ATOM model), children sometimes leave T underspecified for
agreement, resulting in a structure such as (14) below:
(14) [TP Daddy [T canPRES] swim]
T in (14) carries no agreement features and hence its subject cannot be assigned nominative case and
so will be assigned accusative case by default. The auxiliary will still be spelled out as can in
accordance with (12), by virtue of its present-tense feature. Hence, if Daddy is replaced by a personal
pronoun in a structure like (14), we expect to find Him can swim. So, what the ATOM model predicts
in relation to child structures containing an overt modal auxiliary like can is that children will alternate
between using nominative and accusative subjects (e.g. saying He can swim and Him can swim).

6.5 Errors with negation

A frequent type of error made by young children in negative sentences is auxiliary
omission90, as sentences those below illustrate:
(15)(a) I not swimming. (Adam 2;10) (b) I not big enough (Adam 2;10)
(c) I not your daughter (Nina 2;5) (d) I no taste dem (Sarah 2;11)
(15a) will be derived as follows. The verb SWIM will be merged with its subject I to form a verb phrase
which we can represent informally as [VP I [V swim]]. A null negative head [NEG ø] is then merged with
this VP and with not, forming the Negative Phrase [NEGP not [NEG ø] [VP I [V swim]]. This is in turn
merged with a T constituent containing the progressive auxiliary BE, deriving the TP shown in
simplified form below:
(16) [TP [T BE] [NEGP not [NEG ø] [VP I [V swim]]]
The verb SWIM agrees in progressive aspect with the auxiliary BE, marking swim as a progressive
participle (which is ultimately spelled out as swimming). BE also agrees with, and assigns nominative
case to the subject I. Given Chomsky’s assumption that movement is contingent on agreement (so that
T attracts the closest constituent it agrees with to becomes its syntactic subject), BE will also attract I to
move from being the semantic subject of swim to becoming the syntactic subject of BE, so deriving the
structure shown informally below:
(17) [TP I [T BE1SG] [NEGP not [NEG ø] [VP I [V swimming]]]
If BE is specified for (present) tense as well as agreement, it will ultimately be spelled out as (a)m. But
suppose that BE is specified for agreement but not tense (as in (17) above): in such a case, BE will
receive a null spellout in accordance with (6) by virtue of being underspecified, and the child will say
I not swimming. If (as claimed by Chomsky in recent work) all sentences contain a C constituent
which marks whether the sentence is declarative (used to make a statement), interrogative (used to
ask a question), exclamative (used to exclaim e.g. surprise or delight) or imperative (used to issue an
order) in type, the TP in (17) will subsequently be merged with a null declarative (statement-marking)
C constituent to form the CP shown below:
(18) [CP [C ø] [TP I [T ø] [NEGP not [NEG ø] [VP I [V swimming]]]]

90
For present purposes, the copular verb BE functions as an auxiliary, since (when finite) it raises from V into the
T position occupied by (uninverted) finite auxiliaries.
On this view, there is continuity between adult and child negative sentences, in that both have the
same CP+TP+(NEGP+)VP structure, and both involve T attracting a subject which it agrees with and
assigns nominative case to. The only difference between the child and adult structures is that T is
underspecified for tense in the child’s structure, and hence the child gives the auxiliary BE a null
spellout.
However, alongside medial negatives like (15), young children sometimes also produce initial
negatives like the following (from the Iris files in the Wells corpus on the CHILDES database):
(19) No me got one. No me got him (Iris, 3;291)
In order to try and undertand what’s going on here, let’s look at the derivation of No me got one. This
proceeds as follows. The verb GET merges with its complement one and its subject me to form the verb
phrase [VP me [V get] one]. A null NEG constituent then merges with this VP and with the negative
particle no to form the NEGP [NEG no [NEG ø] [VP me [V get] one]]. The resulting NEGP is then merged
with a T constituent containing the perfect-aspect auxiliary HAVE. If HAVE is specified for (present)
tense but not agreement, merging HAVE with the relevant NEGP will form the TP shown in simplified
form below:
(20) [TP [T HAVEPRES] [NEGP no [NEG ø] [VP me [V get] one]]]
GET then agrees in perfect aspect with the perfect auxiliary HAVE, and GET is thereby marked as a
perfect participle form and so is ultimately spelled out as got. Let’s suppose (as in (20) above) that the
T constituent containing HAVE is specified for present tense but not for agreement. This means that
HAVE does not agree with me, and so cannot assign nominative case to me (because this is contingent
on agreement as we see from (10i) above), and likewise cannot attract me to become its subject (since
under Chomsky’s agreement-based account of movement, T can only attract a nominal which it agrees
with to become its subject). Absence of agreement between HAVE and me means that me receives
default accusative case by (10iii), and remains in situ in the specifier position within VP. The resulting
TP is merged with a null C constituent marking the sentence as declarative in force, forming the CP in
(21) below:
(21) [CP [C ø] [TP [T HAVEPRES] [NEGP no [NEG ø] [VP me [V got] one]]]]
Because HAVE is underspecified for agreement, it receives a null spellout in accordance with (6).
Accordingly, the structure in (21) is ultimately spelled out as No me got one. While at first sight it may
look as if there is considerable discontinuity between a child negative structure like (21) and its adult
counterpart I haven’t got one (e.g. in respect of the case-marking and position of the subject and the
null spellout of the auxiliary), all three differences are ultimately reducible to a single property – viz.
absence of agreement features on the head T constituent of TP. Thus, the underspecification analysis
provides a principled account of both medial child negatives like I not swimming and initial child
negatives like No me got one.

6.6 Errors with questions

Alongside adult-like wh-questions such as (22a) below, children also produce wh-questions
such as (22b-d):
(22)(a) What’s he doing? (b) What he’s doing? (c) What he doing? (d) What him doing?
In order to get a better understanding of what’s going on here, let’s look at the derivation of the adult
sentence (22a). The verb DO merges with its complement what and its semantic subject he to form the
verb phrase [VP he [V do] what]. This in turn is merged with a T constituent containing the progressive
auxiliary BE to form the TP shown below:
(23) [TP [T BE] [VP he [V do] what]]
The verb DO agrees in progressive aspect with the auxiliary BE, and so is marked as a progressive
participle form (ultimately being spelled out as doing). If BE is fully specified for (e.g. present) tense
and agreement, it will agree with, assign nominative case to, and attract he to become its syntactic
subject, and will ultimately be spelled out as is – so deriving the TP shown informally below (with
91
The first sentence is glossed in in the relevant file as ‘I haven’t got one’, and the second as ‘I haven’t got him’.
relevant items shown in their spellout form, to make discussion less abstract):
(24) [TP He [T is] [VP he [V doing] what]]
The resulting TP is then merged with a C constituent containing a null question affix Q carrying a
T-feature (i.e. a tense feature) which attracts the closest T constituent to attach to it, and a wh-feature
which attracts the closest wh-expression to move to the front of the overall CP. Merger of C with the
TP in (24) above derives the CP shown in simplified form below:
(25) [CP [C QT, WH] [TP He [T is] doing what]]
The T-feature on Q then attracts is (which carries a present-tense feature) to adjoin to Q, and the
wh-feature on Q attracts where to move to the specifier position within CP, so deriving the structure
shown in highly simplified form below (simplified inter alia by showing only overt constituents):
(26) [CP What [C is] he doing]
Cliticisation of is onto what will ultimately derive What’s he doing?
Now consider the nature of the error in child sentences such as (22b) What he’s doing? Suppose
the derivation proceeds as before until we have formed the TP in (24) above, repeated as (27) below:
(27) [TP He [T is] [VP he [V doing] what]]
The TP in (27) will then be merged with a C constituent containing a null question particle Q (serving
to mark the sentence as interrogative in type). This Q constituent (as before) will carry a wh-feature
which attracts the closest wh-expression to move into the specifier position within CP. In adult main-
clause questions, Q also carries a T-feature, attracting the closest T-constituent to move into C. But
suppose that (just like T) C can be unspecified for tense, and sometimes (as here) carries no T-feature.
The result of merging a (wh-specified but tense-unspecified) C constituent with the TP in (27) will be
to form the CP shown in simplified form below:
(28) [CP [C QWH] [TP He [T is] doing what]]
The wh-feature of Q attracts the wh-word what to move to the specifier position in front of C. But
because Q is underspecified for tense and contains no T-feature, it cannot attract the present-tense
auxiliary is to move to C, with the result that is remains in situ in T, and we derive the structure shown
in simplified form below:
(29) [CP What [C ø] [TP he [T is] doing]]
Such an analysis is entirely in keeping with the continuity hypothesis, since it assumes that child
questions have the same CP+TP+VP structure as their adult counterparts – the only difference between
the two being that children sometimes leave functional heads like T and C underspecified for tense.
In a structure like (29), T is specified for tense, but C is not.
Now consider what happens in the case of questions like (22c) What he doing? Once again, let’s
suppose that DO merges with its complement what and its semantic subject he to form a verb phrase,
and that the resulting VP is merged with a T constituent containing the progressive auxiliary BE, so
forming the TP structure (23) above, repeated below:
(30) [TP [T BE] [VP he [V DO] what]]
The verb DO will agree in aspect with the progressive auxiliary BE, and hence ultimately be spelled out
as the progressive participle form doing. Let’s further suppose that BE is specified for agreement but
unspecified for tense. By virtue of being specified for agreement, BE will agree with, assign
nominative case to and attract the subject he. Accordingly, he moves to become the syntactic subject of
BE, forming the TP shown in simplified form below:

(31) [TP he [T BE3SG] [VP he [V doing] what]]

The resulting TP is then merged with a C containing a null question particle Q which (like T) is
underspecified for tense, and hence carries a wh-feature but no T-feature, so forming the CP shown in
simplified form below:
(32) [CP [C QWH] [TP he [T BE3SG] [VP he [V doing] what]]]
The wh-feature of C attracts the wh-word what to move into the specifier position within CP to the left
of C, so forming the CP shown in skeletal form below:
(33) [CP What [C QWH] [TP he [T BE3SG] [VP he [V doing] what]]]
The question particle Q receives a null spellout by virtue of being inherently null. The progressive
auxiliary BE receives a null spellout in accordance with (6) by virtue of being underspecified (i.e.
lacking tense). The structure in (23) is thus spelled out as What he doing? In such structures, both T
and C are underspecified for tense.
Finally, consider the derivation of (22d) What him doing? Suppose that things work as before until
we reach the stage of derivation shown in (23) above, repeated below:
(34) [TP [T BE] [VP he [V do] what]]
As before, the main verb DO agrees in aspect with the progressive auxiliary BE and is thereby marked
as a progressive participle (ultimately being spelled out as doing). In addition, suppose that BE is
specified for (present) tense but unspecified for agreement. Its lack of agreement properties means that
BE is unable to agree with, assign nominative case to or attract he. This means that he is assigned
accusative case by default (and so is ultimately spelled out as him) and remains in situ within the verb
phrase. The resulting TP is merged with a C containing a null question particle Q which carries a
wh-feature, so forming the CP shown below:
(35) [CP [C QWH] [TP [T BEPRES] [VP him [V doing] what]]]
The wh-feature of Q will attract the wh-word what to move into the specifier position in front of C, so
forming the CP shown in simplified form below:
(36) [CP What [C QWH] [TP [T BEPRES] [VP him [V doing] what]]]
Since BE is underspecified for agreement, it will receive a null spellout in accordance with (6), so that
(36) is spelled out as What him doing?
Note, however, that we tacitly assumed in (35) that Q is underspecified by virtue of carrying a
wh-feature but no T-feature. But what if Q were fully specified for both wh- and T-features, as in (37)
below?
(37) [CP [C QT,WH] [TP [T BEPRES] [VP him [V doing] what]]]
In this case, the T-feature of Q would attract the tense-specified auxiliary BE to move to C, forming
the structure shown below (simplified by omitting the features carried by Q and BE, in order to avoid
excessive visual clutter):
(38) [CP What [C BE+Q] [TP [T BE] [VP him [V doing] what]]]
However, since the auxiliary BE is underspecified (by virtue of carrying tense but not agreement
features), it will again receive a null spellout in accordance with (6), so that structure in (38) will also
be spelled out as What him doing?
An interesting consequence of the two92 alternative derivations sketched out here for What him
doing? is that in both cases the default-case subject him remains in situ in the specifier position within
VP as the subject of the verb doing and at no point raises to the specifier position within TP to become
the subject of the null auxiliary BE. Since negative particles are positioned above/to the left of VP but
below/to the right of TP, this makes the interesting prediction that the negative counterpart of
accusative-subject sentences like What him doing? should show the negative particle positioned to the
left of the accusative subject (him). Some evidence that this is indeed the case comes from negative
wh-questions such as the following (produced by Adam at age 2;11 – the relevant data being from the
Brown corpus on the CHILDES data-base):
(39) Why not me break plate? Why not me careful? Why not me drink it? Why not me hurt?
Why not me sleeping?
Given the assumption of structural continuity between adult and child sentences, a negative question
92
A third possibility which ATOM allows for (the consequences of which I leave you to work out for yourself) is
that T could be unspecified for both tense and agreement.
like Why not me break blate? will be a CP+TP+(NEGP+)VP structure, and have the structure shown
in simplified form in (40) below93:
(40) [CP Why [C ø] [TP [T ø] [NEGP not [NEG ø] [VP me [V break] plate]]]]
The fact that the subject me is positioned after the negative particle not provides empirical evidence
that accusative subjects remain internally within VP – unlike nominative subjects which raise to the
specifier position within TP (and hence end up positioned in front of the negative particle).
A final type of wh-question error which I shall touch on briefly here are auxiliariless questions
containing a finite main verb – such as the following (produced by Adam at 3;0):
(41) Why me bent dat game?
Since the past tense form bent is used here, it would seem that T must carry a (past) tense feature in
such sentences; but since the subject me has default accusative case, it is equally clear that T is not
specified for agreement. Furthermore, it seems likely that C is underspecified for tense in structures
such as the above, with the result that C does not attract the affix in T to move to C but rather the affix
remains in T, so that the overall sentence has the syntactic structure shown in simplified form below 94:
(42) [CP Why [C ø] [TP [T AffPAST] [VP me [V bend] that game]]]
In the morphological component, the past-tense affix will be lowered from T onto the adjacent verb
BEND by Affix Hopping, and the resulting inflected word BEND+AffPAST will be spelled out as bent. If,
by contrast, the affix in (42) had carried agreement features as well as tense, the subject would have
been assigned nominative case and would have been attracted by T to move into the specifier position
within TP, with the resulting structure ultimately being spelled out as Why I bent that game?
The most important point to note in relation to our discussion in §6.2-§6.6 is that the continuity
model can provide a principled account of a wide range of un-adultlike sentences produced by young
children on the basis of the assumption that children’s sentences have the same CP+TP+(NEGP+)VP
structure as their adult counterparts, that children have essentially the same system of verb
morphology and case-marking as adults, and that the principal difference between the two is that
children sometimes leave functional categories like T and C underspecified for tense/agreement.

6.7 Determiner errors

If we follow Longobardi (1994) in positing that principles of UG require nominal
arguments (i.e. noun-expressions which function as subjects or complements) to be DPs/determiner
phrases in adult grammars, and if we espouse the hypothesis that there is structural continuity between
child grammars and their adult counterparts, then it follows that children’s subject and object nominals
likewise have the syntactic status of DPs/Determiner Phrases from the outset. Since young children
alternate between nominals like the car containing an overt determiner and seemingly determiner-less
nominals like car, it is clear that the continuity model is going to have to posit that children sometimes
give determiners an overt spellout, and sometimes give them a null spellout. This in turn means that
the model meeds to develop a principled account of how determiners come to have a null spellout in
child grammars. In a series of interesting papers (Hyams 1996, Hoekstra and Hyams 1998, 1999, and
Hoekstra, Hyams and Becker 1998), Nina Hyams and her late husband Teun Hoekstra argued that in
the same way as auxiliaries like BE/HAVE/DO are overtly spelled out when fully specified for all the
features they carry in adult English and have a null spellout when underspecified for one or more of
the relevant features, so too determiners have an overt spellout when fully specified and a null spellout
when underspecified.
Reasoning along these lines, Hoekstra and Hyams argue that just as children at the Optional
Infinitive stage alternate between producing finite and non-finite clauses, so too they alternate between
producing finite and non-finite nominals (i.e. noun expressions). A finite nominal (Hoekstra and
Hyams maintain) is one which is specified for (singular/plural) number, whereas a non-finite nominal
is one which is unspecified for number. Since mass nouns in English (i.e. nouns like butter or furniture
which denote an undifferentiated mass) give no visible indication of being marked for number,

93
A question set aside here is where exactly why originates with the sentence.
94
Simplified, inter alia, by ignoring the question of where why originates.
Hoekstra and Hyams restrict their attention to nominals containing count nouns (i.e. nouns like dog
which can be counted, in that e.g. we can say one dog, two dogs, three dogs etc.). They maintain that a
plural count noun such as dogs is self-evidently specified for (plural) number, and that an expression
like the dog or a dog is likewise specified for (singular) number, since a singular count noun has to be
modified by a determiner in English, so using a determiner in front of the noun in such structures
provides a clear indication that the noun is specified as singular in number. By contrast, they maintain,
a bare count noun like dog is unspecified for number: their reasoning is that if the noun were plural in
number it would carry the plural -s suffix; and if it were singular in number, it would be modified by
an appropriate determiner like the/a, so that the absence of any determiner when a child produces a
bare count noun like dog (i.e. one which is neither marked as plural nor modified by a determiner)
indicates that the noun is unspecified for number. So, for Hoekstra and Hyams, nominals like the dog,
the dogs, and dogs are specified for number (and so finite), whereas bare count nouns like dog are
unspecified for number, and so non-finite95.
Hoestra and Hyams claim that there is a correlation in child grammars between the finiteness
features of T and those of its subject. More particularly, they posit a constraint which we can outline
informally in the following terms:
(43) Finiteness Constraint
A finite T requires a finite subject, whereas a non-finite T allows a finite or non-finite subject
Hoekstra and Hyams (1998) maintain that there is strong empirical evidence in support of the
finiteness constraint: for example, they claim that 99% of the relevant utterances they looked at in the
Adam files in the Brown corpus on the CHILDES data-based show the expected finiteness correlation.
They argue that their findings support the view that there is continuity between the structure of child
and adult clauses and nominals, but that just as children at the Optional Infinitives stage sometimes
leave clauses underspecified for finiteness, so too they sometimes leave nominals underspecified for
finiteness: however, the fact that children’s finite clauses invariably have finite subjects underlines the
fact that they have adult-like competence, and hence ‘know’ that a finite T constituent needs a finite
subject to agree with.
But how can we explain the finiteness constraint? Arguably, the constraint is a reflex of
person/number agreement between T and its subject. A fully specified finite T carries uninterpretable
person/number features which need to be checked (i.e. valued and deleted) via agreement with a
subject carrying interpretable person/number features. By contrast, if T is underspecified for
agreement, it will allow either a subject specified for number (because the number features of the
subjects are interpretable and so do not need to be checked), or a numberless subject. Now, it seems
reasonable to suppose that in a determiner+noun structure such as a dog, the determiner a enters the
derivation marked for singular number (because a is inherently singular and has no plural form),
whereas the noun dog enters the derivation with an unvalued number feature which is valued as
singular via Concord with the determiner. Likewise, it seems reasonable to suppose that determiners
carry the person properties of nominals (with determiners like the being third person, so that the
president is a third person expression whereas you presidents is a second person expression)96. If so, it
follows that a bare noun expression like dog will be unvalued for number or person, whereas a DP like
a/the dog will be a third person singular expression. Given these assumptions, let’s look in rather more
concrete terms at what the agreement account of the Finiteness Constraint predicts.
If a child uses an (auxiliary or main) verb like is/are/likes which is overtly specified for agreement,
it will require an agreeing subject specified for number and person, so that we expect to find that

95
It might at first sight seem as if a noun like dog must be singular, since it lacks the plural -s suffix found in
dogs. But if we suppose that the spellout rule for regular nouns says that that they carry the plural suffix -s when
specified as plural in number but a zero suffix otherwise, it follows that the inflected form dogs will only occur
when the noun is marked as plural in number, but that the bare form dog will be used either where the noun is
specified as singular in number, or where it is unspecified for number (e.g. in compounds like dog handler where
the meaning is clearly ‘person who handles dogs’, and yet the noun dog is unspecified for number).
96
In the case of a bare plural noun like dogs, we can assume that this is modified by a null plural partitive or
generic quantifier which (via Concord) values the (initially unvalued) number feature of the noun DOG as plural.
See Radford (2004, chapter 4) for arguments in support of positing null determiners and quantifiers.
children at the OI/Optional Infinitive stage producing sentences like the (a) examples in (44-46)
below, but not sentences like the (b) examples:
(44)(a) The dog’s barking (b) *Dog’s barking
(45)(a) (The) dogs are barking (b) *Dog are barking
(46)(a) The dog bites people (b) *Dog bites people
Since the verb forms (i)s/are/bites are overtly specified for agreement, it follows that they will require
a matching number-and-person-specified subject like the dog/(the)dogs and will not allow a
numberless and personless subject like dog.
By contrast, other types of structure like those below (on the agreement account outlined here) are
predicted to allow both agreeing (i.e. person-and-number-specified) and agreementless (i.e. personless
and numberless) subjects:
(47)(a) The dog can bark (b) (The) dogs can bark (c) Dog can bark
(48)(a) The dog barked (b) (The) dogs barked (c) Dog barked
(49)(a) The dog bark (b) (The) dogs bark (c) Dog bark
(50)(a) The dog barking (b) (The) dogs barking (c) Dog barking
On Schütze and Wexler’s assumptions, can is specified for (present tense), but (for children at the OI
stage) may or may not be specified for agreement: if specified for agreement, can will require an
agreeing (third person singular) subject like the dog in (47a) or a (third person plural) subject like (the)
dogs in (47b); if unspecified for agreement, can will also allow an agreementless subject like dog in
(47c).
Much the same can be said about past tense verbs like barked. If the verb barked is specified for
agreement, it will require a third person singular subject like the dog in (48a) or a third person plural
subject like like (the) dogs in (48b); but if unspecified for agreement, barked will also allow a
personless and numberless subject like dog in (48c).
If the sentences in (49) occur in a past tense context, it is clear that the T constituent associated
with the verb bark is unspecified for tense (since it does not carry the past tense affix -ed). If T is
specified for agreement, it will require a third person singular subject like the dog in (49a) or a third
person plural subject like (the) dogs ikn (49b); but if T is underspecified for agreement, it will also
allow a personless and numberless subject like dog in (49c).
However, if the sentences in (49) occur in a present-tense context, the picture becomes rather more
complex. The use of bark rather than barks in (49a) indicates that the T constituent associated with the
verb bark is not specified for agreement (though it’s impossible to tell whether it is or isn’t marked for
present tense); and an agreementless T constituent allows either an agreeing subject like the dog in
(49a) or an agreementless subject like dog in (49c). In (49b), it is impossible to tell whether the verb
bark is specified for tense and/or agreement: but since a third person plural subject like (the)dogs can
occur both as the subject of a T specified for agreement and as the subject of an agreementless T, we
expect children at the OI stage to produce sentences like (47b) (The) dogs bark in a present-tense
context. In (49c), since the subject dog is numberless, the T constituent associated with the verb bark
must be underspecified for agreement – in which case we expect it to allow both a numberless subject
like dog and a number-specified subject like the dog or (the) dogs.
In (50), the absence of the auxiliary (i)s/are indicates (on Schütze and Wexler’s assumptions) that
T must be underspecified. (50a) has a third person singular subject the dog and (50b) has a third
person plural subject (the) dogs and these could come about if T were unspecified for tense, or
agreement, or both97. (50c) has a personless and numberless subject, and so T must be unspecified for
agreement here (though may or may not be specified for tense).
Hoekstra and Hyams’ underspecification analysis of children’s nominals (allied to Chomsky’s
claim about movement being contingent on agreement) makes an interesting prediction about what
would be expected to happen in negative clauses. Consider, first of all, what is expected to happen
with a non-finite subject like man. If T can only attract a nominal it agrees with in person and number,
97
Note that a T underspecified for agreement will allow either a number-specified subject or a numberless
subject, for technical reasons (relating to the number features on T being uninterpretable and needing to be
checked by an agreeing subject, but those on the subject being interpretable and so not needing to be checked).
T clearly cannot attract a non-finite subject which is unspecified for person or number, with the result
that the subject will have to remain in situ within VP and hence will follow the negative particle no(t)
which occupies a position above/to the left of VP. These assumptions lead us to expect that children
will produce utterances like (51a), but not (51b) (an asterisk indicating a type of structure predicted
not to be produced by children):
(51)(a) No boy like cabbage (b) *Boy no like cabbage
But what if a negative sentence has a finite subject like the man? If T is specified for agreement, we
expect the subject the man to be raised out of VP (across the intervening negative) into TP, so resulting
in (51a) below; but if T is underspecified for agreement, we expect the subject to remain in situ within
VP, so resulting in (51b):
(51)(a) The boy no like cabbage (b) No the boy like cabbage
While the predictions of the number-underspecification analysis about the position of subjects seem
clear, as far as I know they remain empirically untested.
If Hoekstra and Hyams are right in their claim that children go through a stage when they
optionally mark finiteness in nominals (and likewise optionally mark finiteness in clauses), then this
takes us towards a more general characterisation of the relevant stage in children’s development as an
Optional Finiteness stage during which children’s clauses and nominals have the same structure as
their adult counterparts, but the functional categories (like T, C and D) which they contain are
sometimes underspecified for finiteness (i.e. features such as tense and person/number).

6.8 A critical perspective on the Underspecification Model

From some perspectives, the underspecification model is conceptually attractive. After all, it
enables us to maintain that children’s sentences which superficially appear to be radically different in
structure from their adult counterparts have exactly the same syntactic structure as the corresponding
adult structures (and hence that there is continuity between the structure of adult and child grammars)
and differ only in that one or more of the functors in the children’s structures is is featurally
underspecified. So, for example, a child question like (39) Why not me sleeping? has the same
CP+TP+NEGP+VP structure as the corresponding adult question Why am I not sleeping? but differs in
that T is underspecified for agreement and so cannot attract (or assign nominative case to) the subject
me. Underspecification of the T-auxiliary BE in turn means that it has a null spellout.
And yet, the crucial mechanism of underspecification which is used to account for why children
produce structures that are often substantially different in form from their adult counterparts deserves
close critical scrutiny. One kind of question which underspecification poses is a conceptual one
relating to the notion of continuity. After all, if T in finite contexts in adult English is specified for both
tense and agreement, in what sense is a child’s T constituent which is underspecified for tense and/or
agreement the same type of constituent as its adult counterpart? In other words, if a child’s T is
sometimes different in feature composition from an adult T, in what sense is there continuity between
adult and child T constituents?
Moreover, the notion of underspecification is potentially problematic from a theoretical point of
view. Chomsky (1995) argues that all heads (i.e. word-level categories) in a syntactic structure must be
interpretable, in the sense that syntactic structures must not contain heads to which the semantic
component of the grammar cannot assign any semantic interpretation. He further maintains the tense
feature carried by T is interpretable (in that it contributes to determining part of the meaning of a
sentence, by providing information about the time at which an event takes place), whereas the
person/number agreement features carried by T are uninterpretable (in that they are just a copy of the
corresponding features carried by the subject). From this perspective, there is no theoretical problem in
claiming that T can be underspecified in respect of its agreement features (because these do not
contribute to determining meaning): but by contrast, there are problems in claiming that T can be
underspecified for tense (and so carry no tense feature whatever), since this will mean that T cannot be
assigned any semantic interpretation, so causing the derivation to crash (i.e. fail).
One way of seemingly overcoming the interpretability problem posed by the assumption that T can
be underspecified for tense is to suppose that T carries some additional interpretable feature which is
always present when T is underspecified for tense. So, for example, Carson Schütze (1997) maintains
that T carries an interpretable finiteness feature in addition to tense and agreement; and in a similar
vein, Nina Hyams (2001) suggests that T carries an interpretable (realis/irrealis) mood feature
whenever underspecified for tense. However, these apparent solutions to the interpretability issue face
the problem that theoretical linguists such as Luigi Rizzi (1997) and Memo Cinque (1999) have argued
that finiteness and mood features are associated with entirely different heads from T. An alternative
possibility (which does not presuppose that T carries some feature in addition to tense and agreement)
is that when T is underspecified for tense, it has a discourse-determined tense value which is visible to
the semantic component, but is invisible in the syntax, morphology and phonology. This amounts to
saying that the tense feature carried by T in such cases is an intrinsically semantic feature rather than a
grammatical feature, and hence plays no role in syntax, morphology or phonology. While many
linguists have assumed that there are inherently semantic features which play no role in grammar (e.g.
a feature such as [±ADULT] differenting man from boy, or horse from foal, or cow from calf, or cat
from kitten, for example), it is less clear whether there is any theoretical basis for positing that a
feature like tense which plays a central role in grammar in many languages could be taken to be purely
semantic in child grammars. The problem is compounded by the fact that because tense is sometimes
morphologically marked at the Optional Infinitive stage, it would be necessary to assume that tense is
given a dual status by the child, sometimes being treated as semantic, sometimes as grammatical. At
any rate, the issue of what it means for T to be underspecified for tense (and the theoretical
consequences of this) need to be addressed.
Moreover, the claim that T can be underspecified for agreement also poses potential theoretical
problems. In English, T agrees in person and number with a particular noun or pronoun expression.
Person and number are separate features, and yet within the ATOM model, they are in effect treated as
a single composite feature: consequently, T is assumed to be specified either for both person and
number, or for neither person nor number. But since person and number are separate features, we
would expect (given the ATOM assumption that a functional head like T may be underspecified in
respect of one or more of the features which its adult counterpart carries) that in finite contexts T may
sometimes be specified for both person and number, sometimes specified for person but not number,
sometimes specified for number but not person, and sometimes specified for neither number nor
person.
In fact, there seems to be some empirical evidence that it is necessary to recognise that T can be
underspecified for one of its two agreement features. In this connection, consider the following
negative sentence (from Anderssen 1996):
(52) Me haven’t seen a duck (Bill 2;5)
Given the continuity assumption that children’s sentences have the same CP+TP+(NEGP+)VP
structure as their adult counterparts, (52) will be derived as follows. The verb SEE will be merged with
its complement a duck and its subject me to form the verb phrase [VP me [V see] a duck]. A null NEG
head will then merge with this VP and with the negative particle n(o)t to form the NEGP structure [NEGP
not [NEG ø] [VP me [V see] a duck]]. This will then be merged with a T constituent containing the perfect
auxiliary HAVE, to form the TP shown below:
(53) [TP [T have] [NEGP not [NEG ø] [VP me [V see] a duck]]]
The verb SEE will agree in aspect with the perfect auxiliary HAVE and so be marked as a perfect
participle (and ultimately be spelled out as seen). Since HAVE is spelled out as the present tense form
have in (52) rather than as the past tense form had, we can suppose that T is specified for (present)
tense in (53). Under the ATOM account, T will additionally either be fully specified for agreement or
entirely unspecified for agreement. If [ T HAVE] is fully specified for agreement, it will be expected to
agree in person and number with, assign nominative case to, and attract the subject ME, so ultimately
deriving I haven’t seen a duck (if the negative particle n(o)t cliticises onto have). By contrast, if
[T HAVE] is entirely unspecified for agreement, it will not agree with, not assign nominative case to and
not attract the subject me, with the result that the subject is assigned default accusative case by (10iii)
and remains in situ within VP, so deriving *Haven’t me seen a duck. Since neither of these ATOM-
based analyses derives the target sentence Me haven’t seen a duck, it would seem that some other
analysis is called for.
One possibility along these lines is to posit that T can sometimes be underspecified for one of its
agreement features – say number (since Hoekstra and Hyams claim that number is underspecified in
child nominals sometimes). Suppose, then, that T in (53) is specified for tense and person (and hence
is first person by agreement with the first person subject me), but not for number – as in (54) below:
(54) [TP [T have1PERS.PRES] [NEGP not [NEG ø] [VP me [V see] a duck]]]
Suppose that T attracts (to become its syntactic subject) the closest expression which it agrees in
person with, but that T only assigns nominative case to an expression it agrees in person and number
with. Given two these assumptions (which are made in Chomsky’s recent theoretical work on adult
grammars), it will follow that T can attract me to become its subject (by virtue of agreeing in person
with ME), but that T cannot assign nominative case to ME (by virtue of not agreeing in number with
ME), and so the subject is assigned default accusative case by (10iii) and is ultimately spelled out as
me. This results in the structure (55) below:
(55) [TP Me [T have1PERS.PRES] [NEGP not [NEG ø] [VP me [V seen] a duck]]]
Merging the resulting TP with a C constituent marking the sentence as declarative in type results in a
structure which is ultimately spelled out as Me haven’t seen a duck (if the negative particle cliticises
onto have). Note that a crucial assumption underlying the analysis in (55) is that T may sometimes be
underspecified for (say) just number alone, rather than being underspecified for both number and
person.
An alternative approach to sentences like (52) Me haven’t seen a duck is to maintain that they are
the result of a spellout error. Reasoning along these lines, let’s suppose that T in such cases is fully
specified for both person and number, and so agrees with and assigns nominative case to the subject
ME, but that (for some reason) the nominative case feature on the pronoun is wrongly spelled out as
me. How might such a case spellout error come about? One answer might be as a result of retrieval
failure. Suppose that the child’s grammar specifies that a first person singular pronoun is spelled out
as follows:
(56) 1SG = I if nominative, my if a weak genitive, mine if a strong genitive98, and me otherwise.
Suppose also that retrieving the right form of a pronoun requires considerable practice (with hundreds
of practice attempts being required before perfect retrieval results), and that failure to retrieve the
correct form typically results in retrieval of the default (‘otherwise’) form me. We would then expect
to find that a child who has not yet had enough practice in retrieving a nominative form like I to ensure
perfect retrieval will occasionally mis-retrieve the default form me instead (resulting in structures like
Me am naughty), and that a child who has not yet had enough practice at retrieving the genitive form
my will occasionally retrieve the default form me instead (resulting in structures like Daddy eat me
biscuit).
A related spellout error account would be to assume that spellout forms in the child’s lexicon are
sometimes underspecified for features which they carry in the adult grammar. For example, suppose
that a two-year-old child has acquired the third person feminine singular pronoun her, but not she. The
child’s entry for her will then simply specify that any 3FSg pronoun (whatever case – if any – it
carries) is spelled out as her (so that the child will say e.g. Her is naughty, Help her! and Find her
dress!). Suppose that sometime later, the child acquires the pronoun she and knows that this is used to
spell out a third person feminine singular nominative pronoun, but continues to assume that her can
spell out any (case-marked or caseless) third person feminine singular pronoun, so that the child has
the following spellout for the relevant items:
(57)(a) she = 3.F.Sg.Nom (b) her = 3.F.Sg
This would mean that in an accusative context (e.g. Help her!) or genitive context (e.g. Find her
dress!) the child would use only her, but in a nominative context (e.g. She/Her is naughty), the child
would alternate between using she (because this can spell out a third person feminine nominative
pronoun) and her (because this can spell out any third person feminine singular pronoun). This would
provide one way of accounting for children who are reported to alternate between nominative and
98
A weak genitive form is one used before a nominal expression (e.g. my in my new car), whereas a strong
genitive form is one used when not followed by a nominal (e.g. mine in ‘This is mine’ or ‘Mine is bigger than
yours’).
accusative subjects with agreeing verbs (See the EXERCISES at the end of the chapter for some
illustrative material).
An alterative Paradigm Building Hypothesis developed by Matt Rispoli (1994, 1995, 1998a,b,
1999a,b, 2000, 2002, 2005)99 posits that case errors are the result of children building a pronoun
paradigm which is slightly different the adult paradigm, resulting in them using pronouns in rather
different ways from adults, and in using rather different retrieval processes, so that (e.g.) a child
unable to retrieve the 1SG nominative form I may instead retrieve the rhyming 1SG form my and so
use it in a nominative context, e.g. in a sentence such as My am hungry.
A final point to be made about the ATOM model’s treatment of the case-marking of subjects is that
its implicit assumption that underspecification is the only mechanism we need to invoke in order to
account for the use of default subjects in child grammars appears to be problematic in respect of
examples such as the following:
(58)(a) Where’s me? (Abe 2;5, covering his head with a piece of paper)
(b) Me’s a big one. Me’s a big one, Mom (Abe 2;7)
Here, T is clearly fully specified for (present) tense, (third) person, and (singular) number, and yet its
subject surfaces in the default form me rather than the nominative form I found in the corresponding
adult sentences Where am I? and I’m a big one. What’s going on here? The answer would seem to be
along the following lines. The T-constituent containing BE carries person and number features which
need to be assigned a value. The relevant person/number feature-valuation operation can be sketched
informally as in (59) below:
(59)(i) T is valued as first person if agreeing with a first person expression, second person if
agreeing with a second person expression, and third person otherwise
(ii) T is valued as plural if agreeing with a plural expression, and singular otherwise
In adult English, the relevant T constituent agrees in person and number with first person singular
subject, and so is ultimately spelled out in the first person singular form am; conversely, because it
agrees in person and number with the relevant T constituent, the first person singular subject is
assigned nominative case, and so is spelled out as I. But suppose that in the child’s grammar, the
present-tense T constituent is fully specified for person and number, but that the agreement operation
which simultaneously values the person/number features of T and assigns nominative case to the
subject fails to apply. In consequence, T is assigned the italicised default/otherwise (third-person,
singular-number) values in (59) and so is spelled out as (i)s, and its subject is spelled out in the default
case form me. On this view, the type of error found in (58) involves agreement failure and not
underspecification.
The notion of underspecification also raises important developmental questions. For example, if T
can initially be underspecified for either tense or agreement (or both), which of the two types of
feature carried by T (its tense feature or its person/number agreement features) is mastered before the
other? In other words, do children learn to mark tense obligatorily in finite contexts before they learn
to mark agreement – or is the converse the case? Richard Ingham (1998) provides evidence from a
longitudinal study of a single child that tense is acquired before agreement. If indeed this is so, it raises
the question of why. One possibility may be that tense features are interpretable (i.e. contribute to
determining meaning) and agreement features uninterpretable, and that (all things being equal)
children acquire interpretable features (because they are essential to conveying meaning) before
uninterpretable features. If we assume that the number features of nominals are interpretable, this
would predict that number-marking would be acquired earlier on nouns than on verbs (since on
nominals it is an interpretable feature, but on verbs it is an uninterpretable agreement feature
associated with T).
Mention of the number properties of nominals leads us on to the research by Hoekstra and Hyams
into the correlation between finiteness on subjects and on auxilaries/main verbs. One potentially
problematic assumption that they make is that an expression like a dog is specified as singular in
number whereas a bare nominal like dog is unspecified for number. This raises the following question.
Suppose that an expression like a dog is a DP/determiner phrase formed by merging the singular

99
For an ATOM response to these, see Schütze (1999, 2001).
indefinite determiner a with the noun dog (the noun dog being marked as singular via concord with the
singular determiner a). Given the core continuity assumption that functional categories in child
grammars can be underspecified for one or more of the features they carry, and that a functional
category has a null spellout when underspecified, we would expect a child to produce the following
three different types of DP. If D is specified for both definiteness (e.g. as indefinite) and number (e.g.
as singular), we expect the child to say a dog. If the child specifies the determiner for definiteness but
not for number, then we expect the child to say simply dog (a numberless expression) if an
underspecified determiner has a null spellout. But suppose the child specifies the determiner for
(singular) number but not for definiteness: in such a case, the determiner will again be null (because it
is underspecified for definiteness) but the noun dog will be singular (via concord with the null singular
determiner). On this scenario, we would expect it to be possible for a child to use a bare noun like dog
as a singular form which is unspecified for definiteness. This in turn would mean that we seemingly
have no theoretical basis for predicting that a child will not say Dog is barking (using a singular
subject unspecified for definiteness with a singular verb). Perhaps one way of getting round this
problem would be to suppose that the determiner the/a contributes the third person feature which is
essential for agreement in number and person between a finite T and its subject. So, while dog could
be singular, it would be unspecified for person, and so unable to serve as the subject of a finite T
constituent (which requires as its subject an expression which it can agree with in number and person).
In this connection, it is interesting to note that Lyons (1999) argues that person and definiteness are
inter-related properties.
Two additional factors make the finiteness correlation appear even more problematic. One is that
even adults seem to omit articles with singular count nouns on occasions – as illustrated by the
following examples produced by mothers talking to their children (from the Bates corpus on the
CHILDES data-base):

(60)(a) Dog’s gonna eat first (Mother talking to Amy at 1;8)

(b) Bird’s happy because he has a house now (Mother talking to Zeke at 2;4)
(c) Pizza’s in the refridgerator? (Mother talking to Rick at 2;4)
It may be that we can account for the first two examples by supposing that (at least when talking to
children) adults sometimes treat animal-denoting nouns as proper names (so that dog is treated like the
name Fido, and hence occurs without an article) – and if so, it would scarcely be surprising if children
do the same. However, this account clearly does not generalise to the use of the bare noun pizza in the
third example100.
An alternative possibility to consider is that metrical factors may lead children to omit articles –
and indeed adults in sentences like (60). In this connection, we should note that Gerken (1991)
suggested that young children tend to omit unstressed syllables in iambic feet (i.e. metrical feet which
consisting of a weak+strong syllable) more frequently than in trochaic (strong+weak) feet. More
specifically, she claimed that young children tend to omit function words (including determiners like
a/the) when they are unstressed and occur in iambic feet. The percentages in (61) below show the rate
of omission of unstressed function words in two target sentences on a sentence repetition task given to
a group of English children with a mean age of 2;3 (with square brackets indicating metrical feet, and
italics marking unstressed words):
(61)(a) [She39% kissed] [the28% dog] (b) [The39% dog] [kissed her0%]
Thus, there may be a number of factors which lead children to omit articles in obligatory contexts.
A final point to note here concerns the suggestion made in the main text about extending the
underspecification analysis from T to C: recall that in §6.6 we argued that absence of auxiliary
inversion in wh-questions follows from C being underspecified for tense (with the result that T cannot
attract a tensed auxiliary to move to C). Given the key claim in the ATOM model that children
sometimes leave functional categories underspecified for agreement as well as tense, we should expect
100
Although a noun like pizza can be used as a bare mass noun in sentences like ‘Pizza is fattening’, this is
generally only the case in generic statements (used to make timeless generalisations), not in sentences like this
one which describe the location of an entity at a specific point of time.
to find that children sometimes leave C underspecified for agreement. Let’s think through what this
would mean. Chomsky (1998, 1999, 2001) argues that syntactic movement operations involve a
functional head attracting the closest constituent which it agrees with: for example, A-movement (e.g.
passivisation) involves T attracting the closest expression which it agrees in person with, and A-bar
movement (e.g. wh-movement) involves C attracting the closest expression which it agrees with in
some peripheral feature (e.g. a topic, focus or wh-feature). What this means is that wh-movement
involves C attracting an expression which is agrees in wh-ness with (wh-agreement being indicated
informally in our earlier discussion by adding a WH subscript to C). But if functional heads in child
grammars can be underspecified for agreement features, this would lead us to expect that C will
sometimes be underspecified for the wh-agreement feature which makes it attract a wh-expression. If
C is underspecified for wh-agreement but specified for tense, we expect to find children producing
structures like:
(62) [CP [C Has+QT] [TP Daddy [T has] gone where]] ‘Has Daddy gone where?’
If C is underspecified for both tense and agreement, we expect to find children producing structures
like:
(63) [CP [C Q] [TP Daddy [T has] gone where]] ‘Daddy has gone where?’
More generally, we would expect to find children sometimes failing to prepose wh-expressions in
questions. And yet (as we noted in chapter 1), a robust finding which comes out of Guasti’s (2000)
study of data from the Adam, Eve, Nina and Sarah files on the CHILDES data-base is that although
English children sometimes fail to invert auxiliaries in wh-questions, they never fail to apply
wh-movement. The range of wh-question structures reported by Guasti are shown below (the numbers
in parentheses giving the raw number and frequency of the relevant type of construction in the four
sets of files which Guasti examines):
(64)(a) Where (i)s Daddy going? (1440/52%) (b) Where Daddy go? (652/24%)
(c) Where Daddy going? (354/13%) (d) Where Daddy (i)s going? (84/3%)
(e) Where Daddy goes/went? (237/9%)
If (as Chomsky claims) wh-movement is contingent on wh-agreement and if (as the ATOM model
claims) children sometimes leave functional heads underspecified for tense and agreement, we would
expect to find both auxiliary-inversion errors (resulting from C being underspecified for tense), and
wh-movement errors (resulting from C being underspecified for wh-agreement). It would therefore
seem that underspecification in itself does not provide us with an explanation for the pattern in (64).
This being so, how can we account for the fact that children frequently make errors with auxiliary
inversion, but not with wh-movement?
Chomsky (1999) argues that wh-movement and Auxiliary Inversion are two entirely different
kinds of linguistic operation. Wh-movement is a syntactic operation, and syntactic operations have the
twin properties that (i) that they can target (i.e. do things to) phrases rather than just words alone101,
and (ii) they have an effect on semantic interpretation. By contrast, Auxiliary Inversion is a
morphological operation, and morphological operations have the twin properties that (i) they can only
affect words (not phrases)102, and (ii) they have no effect on semantic interpretation.
We can illustrate the properties of these two different types of movement operation in the
following terms. As (65) below illustrates, in a question containing a wh-phrase, wh-movement has to
prepose the whole wh-phrase, not just the wh-word alone:
(65)(a) Which two films did you like best? (b) *Which did you like two films best?
This illustrates that wh-movement is a phrasal movement operation, and hence (since syntax is
concerned with the distribution of phrases) intrinsically syntactic in nature. Moreover, wh-movement
is an operation which has an effect on the semantic interpretion (i.e. meaning) of sentences, in that:
(66) A CP is interpreted as a (non-echoic) question if it has an interrogative specifier, as an
exclamation if it has an exclamative specifier, and as a relative clause if it has a relative

101
More precisely, they target maximal projections – but this is a technicality which I set aside here.
102
More precisely, they are operations on minimal projections – but again, I set aside technical details.
 pronoun as its specifier
We can illustrate how (66) works in terms of the bracketed clauses/CPs in the following sentences:
(67)(a) I wonder [CP which dress [C ø] she prefers]
(b) I was surprised at [CP what a lot of food [C ø] she ate]
(c) She is someone [CP who [C ø] I really like]
In accordance with (66), the bracketed clause in (67a) is interpreted as interrogative because its
specifier is the interrogative expression which dress; that in (67b) is interpreted as exclamative
because its specifier is the exclamative expression what a lot of food; and that in (67c) is interpreted as
a relative clause because its specifier is the relative pronoun who. Note that since there are no
auxiliaries in any of the bracketed clauses, auxiliaries cannot in principle play any role in determining
how the clauses are interpreted. So, our discussion suggests that wh-movement is a syntactic operation
with an effect on semantic interpretation.
By contrast, Auxiliary Inversion is very different in nature. For one thing, it is a purely word-based
operation, in that it moves a particular kind of word (an auxiliary) from one word position to another
(more technically, from the T position to the C position in a clause). It appears to have no effect on
semantic interpretation. For example, we could not claim that inverting an auxiliary (i.e. moving it
from T to C) results in a sentence being interpreted as interrogative, for the obvious reason that a
sentence like:
(68) [CP Never before [C have] [TP I [T have] seen him looking so happy]
is declarative in force, not interrogative (i.e. it is a statement rather than a question). Because they
affect only words (not phrases), and because they have no effect on semantic interpretation (Chomsky
reasons), operations like Auxiliary Inversion are purely morphological in nature 103. One
implementation of this idea is to suppose that in structures like (68), C contains a null affix which
attracts a tensed auxiliary to attach to it (and thereby move to C).
If wh-movement is a syntactic operation which has an effect on semantic interpretation, but
auxiliary inversion is a morphological operation with no interpretive effect, we can then hypothesise
that children acquire operations (like wh-movement) which have an effect on semantic interpretation
before they acquire operations like auxiliary-inversion which do not. The reason would be that
children are more concerned to acquire operations which are essential to communication (i.e. to
conveying an intended meaning) then operations which are intrinsically grammatical in nature.

Workbook section
6.1 Children’s my subjects
A matter not discussed in the main text is that (as noted by Anne Vainikka 1994) one- and two-year-old
children sometimes produce my subjects. Schütze and Wexler (1996, p.679) suggested that my subjects
are assigned genitive case by a (tenseless and agreementless) T constituent which is underspecified for
both tense and agreement. By contrast, Radford (1998) argued that my subjects are child nominatives
(and so, given ATOM assumptions, would be expected to occur as the subject of a T specified for
agreement), and that on the basis that many pronouns have full and contracted forms (like he and ’e,
them and ’em) some children mistakenly analyse nominative I (phonetically |aI|) as a contracted form
of my (phonetically |maI|), with the initial |m| segment of my subjects being the first person singular
stem consonant found in forms like me/my/mine. Discuss which of these two (genitive/nominative)
competing analyses of my subjects better accounts for the full range of my subjects found in child
sentences such as those below.
1. My get my car (Nina 1;11, from Vainikka 1994).
103
It might seem as if Auxiliary Inversion has an effect on semantic interpretation in Yes-No Questions such as Is
it raining? But these are assumed by many linguists to be interpreted as interrogative because they have an
abstract yes-no question particle (which you can think of as a null counterpart of whether) in spec-CP (as in
Whether is it raining?). In Skakespeare’s time, yes-no questions could be introduced by whether – cf. Whether
dost thou profess thyself a knave or a fool?
2. My see that (Adam 2;3, from Vainikka 1994)
3. Look what my got (Nina 2;3, from Vainikka 1994)
4. My seen tractors. My seen Terrence the digger (Bill 2;5, from Anderssen 1996)
5. My been the sweeties shop (Kenny 2;8, from Anderssen 1996)
6. My driving this car (Kenny 2;8, from Anderssen 1996)
7. My moving the legs. My going in. (Nina 2;0/2;3, from Powers 1996)
8. My taked it off. My cracked the eggs. My blew the candles out (Jeffrey 2;6, from Budwig 1995)
9. Oh, my can’t open it by myself (Child 3, 2;6, from Rispoli 1995)
10. My will do it again (Child 7, 2;4, from Rispoli 1995)
11. My would be lonely, won’t I? (Douglas, 3;2 from Huxley 1970)
12. My did get my leg dry (Betty 2;6, from Radford 1998)
13. My am coming up to play in there (Child 6, 2;5 from Rispoli 1995)
14. My am mad (Child 9, 2;10 from Rispoli 1995)

6.2 Case-marking of subjects by Douglas

Renira Huxley (1970) reported the following subject-pronoun structures produced by a boy called
Douglas at the ages specified below in parentheses. (Here, I exemplify only three types of subject
pronoun produced by Douglas: the full range of subject pronouns he produces can be found in the
Appendix to Huxley 1970.)
1ST PERSON PRONOUNS
1. I’ve got out garage (2;7) 2. How me get them out? (2;8) 3. Look where I’m running (2;8).
4. Me finish that up (2;9) 5. Me filling it up now (2;9) 6. I am taking my shoes off (2;9)
7. It is pretty when me put my socks on (2;10) 8. My been at their house already (2;10)
9. Me able to nip onto that one (2;11) 10. How me put it under? (2;11)
11. Know what me keep for you? (3;0) 12. Douglas see if me can do it (3;0)
13. Douglas leaves on my shirt, haven’t I?104 (3;0) 14. Douglas eat all my sweets, didn’t I? (3;0)
15. I’d forgotten Douglas call him a boy (3;1) 16. I coming for a drink (3;1)
17. I bited my own thumb (3;1) 18. I want a sweet (3;1)
19. My would be lonely, won’t I? (3;2) 20. When me big, I will go to playgroup (3;2)
21. I can put that in again (3;2) 22. I must bang this jolly hard (3;2)
23. I am Douglas Scott Brown (3;2) 24. I got…my making a spiral ladder (3;2)
(Thereafter, only ‘I’ was used as a subject pronoun in finite contexts)

3RD PERSON SINGULAR MASCULINE PRONOUNS

25. He saying jingle bell (2;8) 26. Him jumping out (2;11) 27. Him is driver (2;11)
28. Him only in the picture (2;11) 29. Him is bear (2;11) 30. He’s got a mother (2;11)
31. He killed it (2;11) 32. Him was at Granny’s house, too (2;11)
33. Now him happy (2;11) 34. He cut it with scissors (3;1)
35. Him did get stung, didn’t he? (3;2) 36. Him pulled out the telephone (3;2)
37. Then the postman comes to get it, then outs it, then puts it into big piles, doesn’t him? (3;3)
38. He jump down (3;3) 39. Him hits it with it (3;3)
40. Him is getting some petrol (3;4) 41. He climbed up the ladder (3;4)
42. He’s a clever pilot, he can fly upside-down (3;4) 43. Driver can peer out, couldn’t him? (3;4)
(Thereafter only ‘he’ was used as a subject pronoun in finite contexts)

3RD PERSON PLURAL PRONOUNS

44. They fallen down (2;4) 45. Them up (2;4) 46. Douglas can’t say them big (2;8)
47. Yes, them go round like that (2;10) 48. Them match (2;11)
49. Them putting their shoes into water (3;0) 50. Them are, them are 105 (3;0)
51. They’ve got two little dogs (3;0) 52. Taked off all the wheels, so they are not old (3;1)

104
The adult counterpart would be ‘I’ve got leaves on my shirt, haven’t I?’
105
Produced in reply to the question ‘Are they potatoes?’
53. Them able to go round on their back wheel (3;1) 54. Them are only little, aren’t they? (3;2)
55. Buses don’t take ladders, don’t they not? (3;2) 56. Them are things like sticks (3;3)
57. So they’re all yours (3;3) 58. Them can’t go (3;3)
59. See them are new ones, aren’t they? (3;3) 60. And then them drove away (3;4)
61. They don’t have windows (3;4) 62. Them got names (3;4)
63. They’re going to kill animals for food (3;4) 64. When them have gone away (3;5)
(Thereafter, only ‘they’ was used as a subject pronoun in finite contexts)
Discuss which of the above sentences do and don’t appear to be compatible with the ATOM account of
subject case-marking (saying why), and see if you can suggest ways of accounting for the problematic
examples106. See exercise 6.1 for suggested analyses of my subjects.

6.3 Abe’s s-inflected verb-forms

The Kuczaj files in the CHILDES data-base show that a boy called Abe (in addition to produce a large
number of correct uses of s-inflected verb-forms with third person singular nominative subjects) also
produced the following structures containing s-forms (at the ages indicated in paretheses). Say what
apparent problems utterances like those below pose for the ATOM model, and see if you can suggest a
possible way round the problems.
1. Where’s me? (2;5.7, covering head with piece of paper) 2. Me’s a big one (2;7.15).
3. What’s those? (2;7.26). 4. What’s those books doing up here? (2;9.11)
5. I wants taste one (2;10.12) 6. You don’t write in my books when them is old ones? (2;10.22)
7. Is those mine? (2;11.18) 8. Her is in the closet, looking for her baby (2;11.25)
9. Sometimes her’s not friendly, sometimes her barks nice, sometimes her don’ts. (2;11.30)
10. And what is they called? (3;0.29)

6.4 Claire’s nominal subjects

Below are listed sentences with count noun subjects (or, in the case of one a count N-pronoun)
produced by Claire at age 2;1 (from files 6-9 of the Claire corpus in the Appendix to this book). To
what extent do they provide evidence of the finiteness effect observed by Hoekstra and Hyams? What
problems (if any) arise in attempting to answer this question, and what solution to them (if any) can
you envisage? (The numbers given below are the numbers of the relevant utterances in the Claire
corpus. Of course, you are welcome to include additional data from utterances 1-304 in the Claire
corpus, if you wish to provide a more comprehensive answer.)
305 Where chair? 320 Where table? 326 Where the girl? 331 Where another fence? 338 People
calling dinner time. 345 What the cow say? 349 Horsie swimming pool 107. 350 Horsie swimming in
the pool. 352 Bunny standing up. 353 Bunny stand up. 361. Where girl? 362 Where another girl?
363 There baby. 364 Where baby? 394 Where the hat? 396 There’s the hat. 401 There’s the tail.
402 Where the horse go? 410 Where the puzzle go? 413 Where that puzzle? 417 There’s the owl.
442 What frog doing? 451 There the daddy, right there. 458 There Claire daddy bed. 460 Baby fit that?
464 There the daddy chair. 487 The rocking chair goes? 488 The rocking chair go? 489 Car go? 493
There the baby. 507 Where worm? 509 There’s a spider. 51 There pear. 515 Bird flying. 516 Bird big
flying. 517 There the bird. 524 Baby drink the coffee. 527 Where the baby go? 531 There baby. 538
Baby crying. 544 There mousie. 546 Where the horsie? 551. Where Claire pencil? 552. Where the
little pencil? 553 Where the little one? 554 Where the big pencil? 555. Where the cow go?
562+567 There the daddy. 585 Where baby chair? 586 Baby sitting. 587 Baby eating dinner.
588 Baby eating juice. 597 Where did the chair go? 598 Where the daddy? 599 Where the people?
600 Where the stairs go? 601 There the man people. 602 Where the man? 603 Girl sleeping. 606 What

106
You might want to bear in mind that in some varieties of English, them can function as a distal demonstrative
used in much the same was as those – hence occurring both in accusative contexts like I don’t wanna wear them
(shoes) and in nominative contexts like Them (shoes) are dirty.
107
The adult counterpart for 349 and 350 would be ‘The horsie is swimming in the pool’.
dolly name? 615 What the daddy doing? 623 The pixie doing? 624 Pixie barking. 625 There horsie.
627 Horsie upstairs. 645 Bear did it. 644+648 Bear do it. 650 What’s the snake saying? 655 Pig say
oink. 657 Where the baby? 670 That one goes another one. 674 There goes another one. 685 There’s
another one, Jane. 726 The Dumpty doing? 728 The bird doing? 736 What the boot doing? 738 The
bunny say? 739 There’s more108 bunny. 740 There’s a bunny. 741 What the mousie doing? 744 There’s
the kitty. 746 Kitty say? 755 Porcupine lie down. 757 Porcupine sleeping. 762 Where the house go?
763 The porcupine doing? 768 What the porcupine doing? 791 Piggy see the water in there? 792 The
bird doing? 793+794 Monkey go xxx (makes monkey noise). 799 Giraffe under bed. 801. The giraffe
doing? 804 There Claire slipper 807 Where the doggy? 818 Peoples talking. 819 Where’s the mommy
talk? 822.Oh, there’s the hammer.
Note that in structures corresponding to adult sentences like ‘Where’s the chair?’ ‘There’s the baby’,
‘There goes the train’, the italicised postverbal expression is the clause subject.

6.5 Abe’s nominal subjects

The Kuczaj files in the CHILDES data-base contain the following exchanges between a boy called
Abe (at age 2;5) and his father (= FAT), with his (linguist!) father attempting to get him to repeat a
number of model sentences whose subjects contain a count noun. To what extent do the relevant data
provide empirical support for the finitness effect discussed in §6.7?
1. FAT: How does the hippopotamus feel? ABE: Hippopotamus is growing
2. FAT: Say this: the dog went home ABE: Dog went home
3. FAT: Here’s another one: a cow ate the carrot ABE: Cow ate the carrot
4. FAT: Now do this one: the boy fell down the stairs ABE: Boy fell down the stairs
5. FAT: Here’s another one: a alligator kissed the donkey ABE: Alligator kissed the donkey
6. FAT: The cow ate a carrot ABE: Cow ate a carrot
7. FAT: The little boy is happy ABE: Little boy’s happy
8. FAT: The red wagon is fun ABE: Red wagon is fun
9. FAT: The turtle will go home ABE: Turtle will go home
10. FAT: The goat did eat his cake ABE: Goat eat cake
11. FAT: The fire will burn the tree ABE: Fire will burn tree
12. FAT: The car can take three people ABE: Car can take three people
13. FAT: The mouse did go to the country ABE: Mouse go to the country
14. FAT: The cow did not eat his dinner ABE: Cow didn’t eat his dinner
15. FAT: The elephant cannot go home ABE: Elephant can’t go home
16. FAT: The boy is glad that the grass is good ABE: The boy is glad that the grass is good
17. FAT: The girl is like her mother ABE: Girl is like his mother
18. FAT: The hippo wanted to kiss the lion ABE: Hippo want to kiss the lion
19. FAT: The baby cries when his mother goes away ABE: Baby cries when mother’s 109 go away
20. FAT: The boy watches Sesame Street while he eats dinner
ABE: The boy watch Sesame Street he eat dinner

6.6 Iris’s declarative and interrogative sentences

Below are a selection of the declarative and interrogative sentences produced by Iris at age 3;2 (the
data being from one of the Iris files on the Wells corpus on the CHILDES data base). Say how these
sentences might be analysed in terms of the continuity model, commenting on points of special
interest, and saying what problems (if any) arise.
Iris’s declarative sentences
Me put my baby in here. Me put Fraser on me bike. Me sit down. Me sit down here. Me keep me baby
xxx. Me try. Me going to use this one. No me got one (glossed as ‘I haven’t got one’). No me got him

108
The adult counterpart of more would be another.
109
Although this is how the form is transcribed, it could equally have been the plural noun mothers.
(glossed as ‘I haven’t got him’). I go like this. Now I’ve got a windy (= window). I can’t xxx. When we
go up Gran’s we do that. We went to xxx. We had a nice one. And you got sit on there. You’s 110 can lie
on here (talking to mother). Now he going to do it. He can go up a shute. There’s tiger. That is Iris’.
That’s a funny bed for you lie on. This is Iris’ windy (= window). This is my play house. It’ll dry up.
No Mummy have to go in. And Fraser want more. Won’t come. Going to have a go on the swing. Want
sit there.

Iris’s interrogative sentences

Can you? Can me put them on? Can I have a drink of juice? Teddy done it? Me get in bath? Want to go
to bed? What’s that? Who doing that? Who going to bath Iris? What am I going to do? Where me can
sit? How me going to get on? How me going to have a bath? How me get in that? How Fraser have his
dinner? Where Fraser put his plate? Where Fraser go? (x3) How going put it on? Why no gonna use
that one?

7 The syntactic development of bilingual children

7.1 Introduction
Bilingualism takes a variety of forms and can arise in a variety of settings, and these may
influence the course of language development in bilingual children (See de Houwer 1995, Bhatia and
Ritchie 1999 and Genesee 2002 for overviews of the linguistic development of bilingual children).
One environmental factor which plays an important role is the nature of the speech input children
receive from their parents – in particular, whether they receive separate input (with each parent
speaking to the child in a different language – sometimes referred to as a one parent, one language
environment) or mixed input (with both parents talking to the child in both languages). A second issue
which needs to be borne in mind is whether the child acquires both languages from birth (so that there
is simultaneous acquisition of the two languages – a phenomenon referred to as bilingual first
language acquisition), or whether one of the languages is acquired from birth and the other at a later
stage (e.g. from age 3 years) – a phenomenon referred to as bilingual second language acquisition. A
third question which it is important to ask is whether a bilingual child receives roughly equal amounts
of input (and produces roughly equal amounts of output) in both languages and so is a balanced
bilingual, or whether the child receives substantially more input (and produces substantially more
output) in one language than the other, with the result that one language is dominant – either overall,
or in particular settings111.
A question which has been central to research into the linguistic development of bilingual children
since its inception is whether children growing up with two languages from birth have one or two
linguistic systems in the early stages of acquisition. Leopold (1939-1949/1970) was the first to suggest
that bilingual children initially develop a single (fused/mixed) language system which eventually
separates (somewhere between two and three years of age) into two independent systems – an idea
widely referred to as the Single System Hypothesis. This position was predominant until the the mid
1980s, and was argued for in Imedadze (1967), Oksaar 1976a/b, Swain (1977), Volterra and Taeschner
(1978), Redlinger & Park (1980), Saunders (1982), Taeschner (1983), and Vihman (1982, 1985).
However, a alternative strand of research from the mid 1970s on argued for a diametrically opposed
110
In some varieties of English, the pronoun you has the plural form you’s/youse.
111
Note that dominance may change over time (e.g. if a child moves abroad).
Separate Systems Hypothesis (or Separate Development Hypothesis) whereby bilingual children
develop separate grammars of each language from the point when they start producing their earliest
multiword utterances, and bilingual language development is said to follow the same course as
monolingual first language acquisition (See e.g. Padilla and Liebman 1975, Bergman 1976, Lindholm
and Padilla 1978, Meisel and Mahlau 1988; Meisel 1989; Genesee 1989; de Houwer 1990; Parodi
1990; Pfaff 1990, 1992; Schlyter 1990; Goodz 1994; Meisel 1994; Paradis and Genesee 1996, and
Lanza 1997).
In what follows below, I begin by outlining Volterra and Taeschner’s (1978) arguments for the
Single System Hypothesis in §7.2, before turning in §7.3 to consider possible implications of their
findings for parameter-setting models of acquisition. In §7.4, I outline Genesee and Paradis’ (1996)
paper arguing for the Separate Systems Hypothesis. And in §7.5 I pose a number of questions (and
give you some exercise material) designed to get you to evaluate the two models discussed here.

7.2 Volterra and Taeschner (1978)

Volterra and Taeschner (henceforth VT) base their study primarily on data from two
sources. The first is data from Leopold’s (1970) study of his (English-German) bilingual daughter
Hildegard, who was brought up in a one parent/one language environment (her mother speaking to her
in English, and her father in German). The second are 30-minute recordings made once a month by VT
of the spontaneous speech production of two German/Italian bilingual sisters (Lisa and Giulia) who
were also brought up in a one parent/one language environment (their father talking to them in Italian
and their mother in German). On the basis of analysing these two sets of data, VT propose a three-
stage model of bilingual acquisition (briefly outlined below).
In VT’s stage 1, the bilingual child is said to have a single lexicon and a single syntax shared in
common between the two languages. In relation to lexical development at this stage, VT claim (p.312)
that ‘A word in one language almost always does not have a corresponding word with the same
meaning in the other language’ (so that e.g. a child may know the word for shoe in one language and
the word for sock in the other, but will not have a synonym/translation equivalent for either word in
the other language). In relation to syntactic development, VT maintain that ‘Words from both
languages frequently occur together in two- to three-word constructions’ (p.312), so that a bilingual
child at stage 1 will produce code-mixed sentences which contain a mixture of words from both
languages. In stage II, the child develops two distinct lexicons, so that ‘for almost any word in one
language, the child has a corresponding word in the other language’, but continues to have a common
grammar for the two languages, and so ‘applies the same syntactic rules to both languages’ (ibid). It is
only at stage III that the child develops separate grammars for the two languages, and reaches the
point where ‘The child speaks two languages differentiated both in lexicon and syntax’.
We can summarise the claims made by VT about the three key developmental stages that bilingual
children go through in schematic terms as follows (where L1 and L2 are their two languages):
(1) Volterra and Taeschner’s three stages in bilingual acquisition
Stage Lexicon Syntax
Stage I common (L1+L2) lexicon common (L1+L2) syntax
Stage II L1 lexicon L2 lexicon common (L1+L2) syntax
Stage III L1 lexicon L2 lexicon L1 syntax L2 syntax
Thus, at stage I, the children have a common lexicon and a common syntax; at stage II, they have
separate lexicons but a common syntax; at stage III they have separate lexicons and separate syntactic
systems. Let’s now examine some of the evidence which VT present in support of their three-stage
model – starting with a look at their claims about lexical development.
In support of their claim that bilingual children at stage 1 have a common lexicon, VT adduce
evidence from Leopold’s (1970) study of his daughter Hildegard. VT classify Hildegard’s early
words into three types: (i) those that are English; (ii) those that are German; and (iii) and those
that are common to both languages. Examples of VT’s classification are given in (2) below:

(2) Volterra and Taeschner’s classification of Hildegard’s lexicon at stage I

English words German words Common words
pretty Bild ‘picture’ Mamma
down Ball ‘ball’ Carolyn
hot Blumen ‘flowers’ Rita
naughty mehr ‘more’ Bett/bed
night auf ‘on’ Marion
there da ‘there’ baby
thank you danke ‘thanks’ Mann/man
yes ja ‘yes’ quak-quak

VT (p.317) maintain that at stage I ‘All the words of the child’s speech appear to form one lexical
system’ and that ‘The use of one language or the other depends on what the child wants to say and not
so much on the language spoken to him’. In other words, if asked a question like ‘What are those?’ in
English, a bilingual English-German child with a lexicon like that in (2) will reply with the German
word Blumen ‘flowers’ because this is the only word which she has in her common lexicon for the
concept of FLOWERS at this stage.
VT’s single-lexicon claim that words in one language at stage I have no synonyms in the other
might at first sight appear to be compromised by related pairs like the last three (italicised) entries in
the list in (2) above. However, VT counter that ‘The children often do not appear to consider such
words as exactly corresponding to each other’ (p.314): e.g. Lisa at stage I uses Italian là ‘there’ ‘for
things that are not visible at the time of speaking’ and German da ‘there’ ‘for things that are present
and visible to her’ (p.315).
As we see from (1) above, when bilingual children reach stage II, they are claimed to have
developed two separate lexicons. VT remark (p.317) that ‘In the second stage, the child reaches
the point where he can be said to possess two lexical systems, in the sense that the same object
or event is indicated with two different words pertaining to the two languages’ (p.317). In the
early part of stage II, the child may associate synonyms in the two languages with different
objects: e.g. Lisa at age 2;5 says occhialiI ‘glasses’ when she sees her Italian-speaking father’s
glasses, but BrillenG ‘glasses’ when looking at a picture of someone else’s glasses (a subscript
I/G being used to indicate that the relevant word is Italian/German). Later, she alternates
between the two words – e.g. after drawing glasses, she says to her mother BrillenG…occhialiI.
VT claim that during stage II, the child ‘learns to distinguish between the two lexical systems.
He recognises that a group of words belongs to the same lexical system and uses only such
words in a sentence. When this happens, most of the child’s sentences are in one or the other
language, the choice depending on the person with whom he is speaking.’ So e.g. Lisa (at 3;3)
when talking to an Italian friends says Dov’è Kitty? (‘Where’s Kitty?’) and then turns round to
her German-speaking mother and says Wo ist Kitty? (‘Where is Kitty?’).
Now consider the evidence which VT adduce in support of their claims about syntactic
development in bilingual children. As noted above, in both stage 1 and stage II bilingual children are
claimed to have a common syntax and to produce code-mixing sentences which are a mixture of words
and syntactic structures from both languages. Implicit in the claim that they have a common syntax at
stages I and II, however, is the idea that (for any given construction) they will use a fixed structure,
and not two different structures. So, for example, VT maintain that up until the age of two-and-a-half
years, Lisa has ‘two lexicons but only one syntax’ and so produces only possessive structures like
those in (3) below with German (POSSESSOR+NOUN) word order (as in Sonjas Auto ‘Sonja’s car’), not
possessives with Italian (NOUN+POSSESSOR) order (as in la machina di Sonja ‘the car of Sonja’):
(3)(a) GiuliaC BuchG (b) GiuliaC giammaI (G = German, I = Italian, C = common)
Giulia book Giulia pyjama
‘Giulia’s book’ ‘Giulia’s pyjamas’
Likewise, Lisa’s adnominal adjective structures (i.e. structures containing an adjective used to modify
a noun) at stages I and II (according to VT) always show the ADJECTIVE+NOUN order found in
German, not the NOUN+ADJECTIVE order which is predominant in Italian: cf.
(4)(a) schöne Blume ‘pretty flower’ (b) kleine ladio ‘little radio’
And Lisa negates sentences in both German and Italian by use of the clause-final negative particle no,
as the following examples illustrate:
(5)(a) Lisa haia haia machen no ‘Lisa haia haia do no’ [German]
(b) Lisa va da là no ‘Lisa goes over there no’ [Italian]
What is curious about the pattern of negation found in (5) is that it is found neither in German nor
Italian, but instead seems to be Lisa’s own creation. A similar pattern of sentence-final negation is
reported in monolingual English children by Klima and Bellugi (1966).
However, once Lisa enters stage III (at around two-and-a-half years of age) we see her developing
language-specific syntactic structures (e.g. separate German and Italian possessive and
adnominal adjective structures), initially overgeneralising both types of structure to both
languages. So, at 2;8 she produces both POSSESSOR+NOUN and NOUN+POSSESSOR
structures like the following in both languages:
(6) Lisa’s possessive structures at stage III
Language Possessor+noun (German pattern) Noun+possessor (Italian pattern)
German Lisa Arbeit Shappen Lisa
Lisa work ‘Lisa’s work’ Slippers Lisa ‘Lisa’s slippers’
Italian Lisa gomma a penna di Lisa
Lisa rubber ‘Lisa’s rubber’ the pen of Lisa ‘Lisa’s pen’

And at age 2;9 she produces the following italicised possessive blend (i.e. mixed-language possessive
structure) with Italian words and Italian case-marking (via use of the preposition di to
mark genitive case) but German POSSESSOR+NOUN word order:
(7) Quetto è di Giulia libro
This is of Giulia book (= ‘This is Giulia’s book’)
Moreover, in the adnominal adjective structures she produces at stage III, she initially alternates (in
both German and Italian) between the ADJECTIVE+NOUN order found in German and the
NOUN+ADJECTIVE order which predominates in Italian – as the following structures she produces at
3;6 illustrate:

(8) Lisa’s adnominal adjective structures at stage III

Language Adjective+noun (German) Noun+adjective (Italian)
German der kleine Badezimmer Schuhe dunkelbraun
the small bathroom Shoe brown ‘brown shoe’
Italian quel bianco pecora un sole rosso
that white sheep a sun red ‘a red sun’

And in her negative sentences at stage III, she correctly positions finite verbs after the negative
particle non ‘not’ in Italian, but positions finite verbs before (correctly) and after (incorrectly) the
negative particle nicht ‘not’ in German – as the following examples illustrate:
(9)(a) Così anche non va bene (Italian, 3;6)
Thus also not goes well (‘It isn’t OK like this either’)
(b) Giulia will nicht weiter schlafen (German 3;3)
Giulia wants not longer sleep’ (‘Giulia doesn’t want to sleep any longer’)
(c) Onkel Karlos nicht versteht Italienisch (German 3;3)
Uncle Carlos not understands Italian (‘Uncle Carlos doesn’t understand Italian’)
It would seem that she has correctly identified that the Italian negative particle non is a proclitic (a
clitic particle which attaches to the front of an auxiliary/verb), and wrongly assumes in
structures like (9c) that the German negative particle nicht can also be a proclitic. At a later
phase in stage III, she learns to differentiate between structures in the two languages, reserving
Italian structures for Italian sentences, and German structures for German sentences.
 7.3 A parameter-setting reinterpretation of Volterra and Taeschner’s claims
Volterra and Taeschner’s paper was written three years before Chomsky (1981) developed
Principles and Parameters Theory/PPT. However, their main findings about the acquisition of syntax
can be reinterpreted in PPT terms in the following fashion.When bilingual children set parameters,
they have to determine the setting of each parameter for each language. In other words, for a bilingual
child parameter-setting involves fixing both a structural value (e.g. head-initial/head-final) for the
parameter and a language value (e.g. Italian). Where a parameter P has different settings for each
language (setting S1 for language L1 and setting S2 for language L2), one of the two structural values
(S1) is generally acquired before the other. This parameter setting is initially unvalued for language
(i.e. the language value for the parameter has not been set) and hence used in both languages.
Likewise, when the second setting (structural value) of the parameter (S2) is acquired, it is initially
unvalued for language (hence used for both languages). For a while, both structural settings (S1 and
S2) may be used for both languages, but gradually the two settings are differentiated (valued for
language), so that S1 comes to be associated with L1 and S2 with L2. There may be a similar pattern
in the way the child acquires the lexicon, in that words (like structures) may initially be unvalued for
language (hence used in both languages), and only at a later stage valued for language (hence
thereafter used e.g. in German sentences only). This would amount to setting a language parameter
for each word acquired (e.g. identifying it as an English or German word). In stage I, the relevant
lexical parameters have not been set, so words are acquired without being tagged for language – and
hence can be used in both languages.
We can illustrate the general idea in terms of the position occupied by (adjectival and possessive)
modifiers with respect to the noun they modify in Italian and German. Simplifying what is in reality a
considerably more complex situation, let’s say that the two languages differ in respect of a Modifier
Position Parameter in that German normally positions (s-possessive and adjectival) modifiers before
any noun they modify (and so has modifier+noun word order), whereas Italian typically positions
di-possessive and (most) adjectival modifiers after the noun and so has noun+modifier order. We can
thus say that German has a modifier-first (i.e. modifier+noun) setting for the Modifier Position
Parameter, whereas Italian has a modifier-last (i.e. noun+modifier) setting112.
Given the (highly simplified) analysis presented above of word-order variation in modified noun
structures, we can recast VT’s account of the development of Lisa’s noun expressions in the following
terms. In stages I and II (when Lisa produces only ADJECTIVE+NOUN and POSSESSOR+NOUN
structures), Lisa has set the Modifier Position Parameter at the modifier-first value appropriate for
German, but has not set the language value for the relevant parameter setting and so generalises this
setting from German to Italian – hence producing ADJECTIVE+NOUN and POSSESSOR+NOUN word order
in both languages. At the beginning of stage III, Lisa learns that the parameter has two distinct
structural settings, but does not initially learn the language setting associated with each of these
individual parameter settings, and so (as we saw in (8) above) alternates in both languages between
ADJECTIVE+NOUN and NOUN+ADJECTIVE structures on the one hand, and between POSSESSOR+NOUN
and NOUN+POSSESSOR structures on the other. At the end of stage III, Lisa learns that Italian requires a
modifier-last setting for the Modifier Position Parameter and German a modifier-first setting, and so
comes to use ADJECTIVE+NOUN and POSSESSOR+NOUN order in German and restricts
NOUN+ADJECTIVE and NOUN+POSSESSOR order to Italian.
As we have seen in this section, Volterra and Taeschner’s findings have potentally interesting
implications for how parameters are set by bilingual children. Unfortunately, however, there are
severe methodological and empirical limitations to VT’s work (discussed in de Houwer 1995, pp.231-
5). For one thing, VT’s key claims are based on a very limited corpus (mainly a handful of short
recordings of the speech output of Lisa and Giulia at monthly intervals), and it is therefore impossible
to be sure about the precise range of structures produced by either of the sisters in either language.
Moreover, the evidence VT present in support of their findings is anecdotal rather than quantitative
in nature, in the sense that it consists of presenting example sentences in support of the claim that a
112
I am simplifying here in numerous respects. In Italian, most adjectives follow any noun they modify, though a
few precede the relevant noun. Cinque (1994) argues that the different positions occupied by nouns in Italian and
German noun expressions reflect the fact that nouns in German remain in situ whereas nouns in Italian raise to a
higher functional head position ‘above’ an adjectival or possessive modifier. I set aside technical details (and
numerous descriptive complications) here, in order to simplify exposition.
given child at a given stage produced such-and-such a type of structure, without any attempt to
quantify the raw number of occurrences (and relative frequency) of the structure in question. And in
some cases, there seems to be no evidence in support of a specific key claim: for example, the claim
that in the early stages, the two Italian-German bilingual children they studied generalise the word
order ADJECTIVE+NOUN from German to Italian is undermined by the seeming lack of any structures
containing an adective-modified noun produced by the children in Italian at the relevant stage. In other
cases, what little evidence VT do present seems potentially to contradict their claims (as we see from
the fact that the examples in (1) appear to show that Hildegard has a number of translation equivalents
– e.g.yes/ja). Moreover, much of the evidence VT bring to bear in support of their claim that children
start out with a common syntax for their two languages is based to a large extent on code-mixing
(more precisely, on the claim that the children frequently produce mixed-language utterances). And yet
code-mixing is also found in adult bilinguals (who are generally assumed to have separate grammars
for each of their languages), and so is far from providing conclusive evidence in support of the Single
System Hypothesis. In short, there are considerable methodological, empirical and theoretical
shortcomings in VT’s research.

7.4 Paradis and Genesee (1996)

Paradis and Genesee (henceforth PG) argue that bilingual children develop separate
grammars for their two languages from the outset, and that these grammars remain completely
autonomous of each other at all stages of development, and at no stage are in any way
interdependent (where ‘We define interdependence as being the systemic influence of the grammar of
one language on the other’, p.3). They identify two potential manifestations of interdependence. One
is transfer, which they characterise as involving ‘the incorporation of a grammatical property into one
language from the other’ (p.3). A second is the rate of acquisition (in the sense of one language
influencing the speed with which a particular property is acquired in the other): one such effect might
be acceleration so that ‘a certain property emerges in the grammar earlier than would be the norm in
monolingual acquisition’ (p.3); the converse effect would be delay so that ‘The burden of acquiring
two languages could slow down the acquisition process in bilinguals, causing them to be behind
monolinguals in their overall progress in grammatical development’ (p.3). PG undertake an original
empirical study which sets out to demonstrate the complete absence of any interdependence between
the grammars of bilingual children.
PG base their study on longitudinal spontaneous-speech data which they collected from three
(French-English) bilingual Canadian children at 6-monthly intervals (roughly at ages 2;0, 2;6, and
3;0113). The children in their study were all from Quebec and had an English-speaking mother and a
French-speaking father: in each case, the father spoke to the children in French, the mother in English.
PG used data from Amy Pierce’s (1992) study of the linguistic development of monolingual English
and monolingual French children in order to make comparisons between monolingual and bilingual
development. PG focused on three main areas of grammatical development, viz the acquisition of (i)
finite verb morphology, (ii) negation, and (iii) subject pronouns. Let’s look at each of these in turn.
In her (1992) study, Amy Pierce reported that (during the Optional Infinitives stage when children
alternate between producing finite and non-finite verbs in contexts where adults use finite verbs)
monolingual French children acquire inflected forms of finite verbs earlier (and use them more
frequently) than monolingual English children. To see whether this is true of the three bilingual
children in their study, PG calculate the percentage of the verb-forms they produced which were finite
and produce the following table (p.13):

(10) Mean % of the bilingual childen’s verbs which were finite at each recording time
Language Time 1 (around 2;0) Time 2 (around 2;6) Time 3 (around 3;0)
English 10% 24% 44%
French 51% 74% 85%

PG observe that the bilingual children show much the same pattern of development in this respect as
Pierce reports for monolingual French and English children, and conclude that the bilinguals’
grammars of English and French develop independently. (Although Pierce 1992 gives no figures for
113
More precisely, two of the children were aged 1;11 at the start of the study, and the third 2;2.
the English children she studies, she reports that 51% of Grégoire’s verbs at 2;0 were finite – a figure
identical to that reported by Paradis and Genesee for the French verbs produced by the bilingual
children.) What the findings in (10) mean in PG’s terms is that there is no evidence for acceleration
(i.e. no evidence that the bilinguals acquire English finite verb morphology faster than monolingual
English children because of the influence of French) or delay (i.e. no evidence that bilinguals acquire
French morphology more slowly than monolingual French children because of the influence of
English).
A second topic investigated by PG is the acquisition of negation. Before considering the results of
their research, let’s briefly look at key differences between negative sentences in adult English and
French. T in English is always weak, with the result that the T position in a negative clause like that in
(11a) below can only be filled by merging (= directly positioning) an auxiliary like DO in T. By
contrast, T is strong in finite clauses in French and so can attract a verb to move from V to T (across
an intervening negative particle like pas ‘not’) in the manner showed by the arrow in (11b):
(11)(a) John [T does] not [V like] the police (b) Jean [T aimelikes] pasnot [V aime] lesthe flicscops

As a result of V-to-T movement in French structures like (11b), a finite main verb like aime ‘likes’
comes to be positioned in front of the negative particle pas ‘not’. However, in non-finite negative
clauses in both languages, verbs remain in situ within the verb phrase and hence follow the negative
particle not/pas, as we see from the bracketed non-finite clauses in the examples below:
(12)(a) I advise you [not to say a single word]
(b) Je te conseille [de pas prononcer un seul mot]
I you advise of not pronounce a sole word (= same meaning as 12a)
The different patterns of negation in the two languages reflect different settings for the Head Strength
Parameter for T (= the T-Strength Parameter): T is strong in finite clauses in French but weak in
non-finite clauses, and so only a finite T can attract a verb to move from V to T; by contrast, T is weak
in all kinds of clause (finite and non-finite alike) in English, and so does not have the power to attract
a verb to move from V to T. Using rather different (but equivalent) terminology, PG argue that French
and English differ in respect to a Verb Movement Parameter, with main verbs move from V to T in
finite (though not non-finite) clauses in French, but always remaining in situ in English.
Having briefly looked at the syntax of negation in adult English and French, let’s now return to
acquisition. Amy Pierce (1992) reported that (like their English counterparts) monolingual French
children go through an Optional Infinitives stage at which they alternate between producing finite
and nonfinite verbs in finite contexts (e.g. main clauses). Pierce claims that monolingual French
children only raise verbs to T (across negative pas) in finite negative clauses like (13a) below, but (like
adults) leave the verb in situ (following pas) in nonfinite negative clauses like (13b):
(13)(a) Ça tourne pas (b) Pas chercher les voitures
It turns not Not seek INF the cars
‘It won’t turn’ ‘(Let’s) not look for the cars’
Pierce also reported that monolingual English children always correctly leave verbs in situ in negative
clauses (so that the verb is positioned after negative no/not), and never raise the verb to T across no(t).
This suggests that (from the outset) monolingual French and English children set the Verb Movement
Parameter (= T-Strength Parameter) at the value appropriate to the language they are acquiring.
However, Pierce noted that clause subjects in early child English can either remain in situ in the
specifier position within VP (so giving rise to sentence-initial negative structures containing a TP like
that in (14a) below), or raise to occupy the specifier position in TP (so giving rise to sentence-medial
negative structures containing a TP like that in (14b) below):
(14)(a) [TP [T ø] [NEGP no [NEG ø] [VP Leila [V have] a turn]]] (Nina 2;1)
(b) [TP Me [T ø] [NEGP no [NEG ø] [VP [V go] home]]] (Peter 2;1)
Déprez and Pierce (1993) reported that for three English children they studied aged 1;10-2;4, the mean
percentage of sentence-medial negatives like (14b) (with subject+negative+verb word order) was
48%.
 Having briefly surveyed the acquisition of negation by monolingual children, let’s now look at the
types of negative sentence produced by bilingual children. PG claim that the three bilingual
children in their study correctly raise finite verbs to T (across the negative particle pas) in
French, and correctly leave nonfinite verbs in situ. They give the following examples in which
an (italicised) finite verb has raised to T across pas:
(15)(a) Va pas là (b) Je veux pas parler à Papa
Goes not there I want not talk to Daddy
‘Doesn’t go there’ ‘I don’t want to talk to Daddy’
They report (but do not cite) one example of a nonfinite negative clause in which a nonfinite verb
remains in situ and is correctly positioned after pas.
PG also report that in their English structures, the bilingual children always correctly leave verbs
in situ (following no/not), but that they may either leave the subject in situ within VP (giving
the word order Negative+Subject+Verb) or raise the subject to spec-TP (giving
Subject+Negative+Verb). The frequency of these two types of negation pattern at the
relevant recording times is shown below:

(16) Number/percentage of English negative structures produced by the bilingual children

Word order Times 1 & 2 (around 2;0 & 2;6) Time 3 (around 3;0)
Negative+Subject+Verb 12 (60%) 5 (26%)
Subject+Negative+Verb 8 (40%) 14 (74%)

PG claim that their results are comparable with those of Déprez and Pierce (1993), who report a mean
of 48% medial negation (Subject+Negative+Verb structures) in the English of 3 monolinguals aged
1;10-2;4. The overall conclusion PG arrive at is that the acquisition of negation in English and French
by the bilingual children in their study is parallel to that of the monolingual English and French
children in Pierce’s study, and provides clearcut evidence that bilingual children develop separate
grammars for each of the languages they are acquiring from the outset.
The third topic which PG look at is the acquisition of subject pronouns. Like English, French has a
set of nominative pronouns (such as je ‘I’, tu ‘you’, il ‘he’ etc.). However, there is an important
difference between English nominative pronouns and their French counterparts, as can be illustrated
seen in relation to the sentences below:
(17)(a) Il parle rarement à ses parents (b) *Il rarement parle à ses parents
He speaks rarely to his parents He rarely speaks to his parents
While a nominative pronoun like he in English can freely be separated from the finite (auxiliary or
main) verb of which it is the subject, this is not true of its French counterpart il ‘he’ – as we see from
the fact that we cannot position an adverb like rarement ‘rarely’ between the nominative pronoun il
‘he’ and the verb parle ‘speaks’ in (17b). Why should this be? The answer is that nominative pronouns
in French are clitics and must attach to an appropriate host – the appropriate host for a nominative
pronoun in French being a T constituent containing a finite auxiliary or (raised) main verb. Hence
(17a) is grammatical because the pronoun il ‘he’ can procliticise (= attach to the beginning of) the
finite verb parle ‘speaks’ in T, but (17b) is ungrammatical because the intervening adverb rarement
‘rarely’ prevents il ‘he’ from procliticising onto the finite verb parle ‘speaks’. By contrast, nominative
pronouns are not clitics in English, and so do not have to be immediately adjacent to a finite verb – as
we see from the fact that the nominative pronoun he is separated from the finite verb speaks in the
English gloss for (17b) by the intervening adverb rarely.
 Amy Pierce (1992) reports that English monolingual children (at the Optional Infinitives stage) use
nominative pronouns as subjects of both finite (auxiliary or main) verbs as in He can help
me/He is helping me and non-finite verbs as in He help me/He helping me. However, she notes
that monolingual French children use nominative pronouns only as subjects of finite verbs
(and hence say e.g. Il dort ‘He sleeps’ using il ‘he’ as the subject of the finite verb dort
‘sleeps’) and not as subjects of non-finite verbs (and so do not say e.g. Il dormir ‘He sleep’,
using il ‘he’ as the subject of the infinitive form dormir ‘sleep’). Paradis and Genesee report
that the three bilingual children in their study used nominative pronouns in English with finite
and nonfinite verbs alike (in roughly equal proportions) – hence alternating between
structures like He went home and He going home. By contrast, PG note that 99% of the French
nominative pronouns produced by their bilingual subjects were used with finite verbs. From
this they conclude that the bilingual children in their study know that French nominative
pronouns are clitics which require a finite verb host to attach to, and likewise know that
English subject pronouns are not clitics (and so do not have to occur immediately in front of
as finite verb)and hence can occur both as the subject of a finite or nonfinite verb (e.g. He
teased/teasing me).
PG argue that further evidence in support of their claim that the bilingual children in their study
know that nominative pronouns are clitics in French but not in English comes from code-mixing. The
note that when the bilingual children code-switch (change language) between subject and verb, they
respect the clitic/non-clitic pronoun distinction. So, for example, they use English (non-clitic)
nominative pronouns as subjects of French finite and nonfinite verbs alike – as the examples below
illustrate:
(18)(a) He a eyes (b) They manger bonbon
‘He hasFINITE eyes’ ‘They eatINFINITIVE sweet’
Conversely (PG claim) the children use French (clitic) nominative pronouns only as subjects of
English finite verbs, not as subjects of English non-finite verbs. The overall conclusion which
their research leads them to is that the bilingual children in their study develop entirely distinct
pronoun systems for the two languages they are acquiring from the outset, lending futher
empirical support to the separate development hypothesis.
The more general conclusion which PG draw from their research is the following:
‘The acquisition of finiteness, negation and pronominal subjects in these bilingual children
follows the same pattern as those of monolinguals. The large gap between French and English
in the use of finite utterances and the absence of English utterances with postverbal negatives
indicate that the children are not transferring the Verb Movement Parameter 114 from French
into their English grammar, nor is the presence of French accelerating their acquisition of
English syntax. Similarly, the distribution of pronominal subjects in each language shows that
the children have correctly classified French weak pronouns as clitics and French strong
pronouns and English pronouns as NPs. We conclude that our bilingual children were
acquiring French and English separately and autonomously.’ (p.19) ‘In addition to showing the
same patterns of acquisition as monolinguals, the bilingual children in our study seemed to be
acquiring these aspects of French and English syntax at a rate similar to that of monolinguals.’
(p.20)
They thus claim that their findings provide empirical support for the separate systems hypothesis.
While PG’s claims might at first sight appear to be persuasive, there are nonetheless pitfalls in
their work (as should become clear to you when you look at some of the material in the next section).
For example, their study is based on a very small corpus (just three recordings each of three children),
so questions arise about (i) whether they really have enough empirical evidence to support their
findings for the three children in their study, and (ii) whether their research findings would be
replicated by a larger-scale studies of other bilingual children. A further weakness in their research is
that they fail to acknowledge the extensive literature on transfer effects in bilingual children (such as
those reported by Volterra and Taeschner). To illustrate from just one such study: Döpke (2000)

114
They mean ‘are not transferring the setting of the Verb Movement Parameter…’
reports that although German shows verb-final word order in subordinate clauses (and so positions the
italicised verb after its complement in a subordinate clause like that bracketed in (19a) below), a group
of bilingual Australian children she studied who acquired English and German simultaneously
produced complement clauses like that italicised in (19b) with verb-initial word order (i.e. with the
italicised verb positioned in front of its complement):
(19)(a) Ich möchte [dich tragen]
I must [you carry] ‘I must carry you’
(b) Ich möchte [tragen dich]
I must [carry you] ‘I must carry you’
The fact that English shows verb-initial order in complement clauses (and indeed in main clauses)
suggests that structures like (19b) result from transferring the relevant setting of the Head Position
Paramater from English to German. If so, we have clear evidence that the grammars of the two
languages are not treated as entirely independent by the bilingual children concerned.
Main clauses in German show verb-medial word order (as in Ich sehe dich ‘I see you’), whereas
subordinate clauses show verb-final order (as in Er kann dich sehen ‘He can you see’). Consider, for
example, their claim that the negatives sentences produced by the bilingual children in their study
show that they have developed separate grammars for the two languages, and know verbs raise to T in
finite clauses in French, but not in English. This might lead us to the conclusion that PG simply don’t
have enough data from enough children to support the sweeping theoretical claims that they make.

Workbook section

7.1 Equivalent words

Quay (1995) reports that 36% of the English words produced by a one-year-old Spanish-English
bilingual girl called Manuela had Spanish equivalents, and conversely that 41% of her Spanish words
had English equivalents. Pearson, Fernández and Oller (1995) report in relation to a study of 27
Spanish-English bilingual children recorded between 0;8 and 2;6 that translation equivalents were
found for all but one of the children, with an average of 30% of all words having equivalents (both at
earlier and at later stages). Taeschner’s (1983) study shows that in the first six months that they were
studied (at which time they were said to be at stage I), 8% of Lisa’s Italian words and 12% of her
German ones had equivalents in the other language, while 18% of Giulia’s Italian words and 20% of
her German ones also had equivalents. Discuss the potential significance of these findings for models
of bilingual acquisition. To what extent might the fact that most of the Taeschner data were collected
by the children’s German-speaking mother have affected the number of translation equivalents found?

7.2 Genesee et al’s (1995, 1996) studies

Genesee, Nicoladis & Paradis (1995) report on a study of a group of French-English bilingual children
ages between 1;10 and 2;2. They observed that the children predominantly used French when talking
to a French-speaking parent, and predominantly used English when talking to an English-speaking
parent. In a follow-up study, Genesee, Boivin & Nicoladis (1996) reported that 3 of the 4 bilingual
children they studied used a higher proportion of English when talking to an English-speaking
monolingual stranger than to an English-speaking parent (and conversely used a a higher proportion of
French when talking to a French-speaking monolingual stranger than to their French-speaking parent).
What are the potential implications of these studies for models of bilingual acquisition?

7.3 Frequency of code-mixing

Genesee, Nicoladis & Paradis (1995) estimate that between 1% and 7% of the sentences produced by
bilingual children show code-mixing. Vihman (1982) reports a mixing rate of 34% in her
Estonian/English bilingual subject at 1;8, 22% at 1;9, 20% at 1;10, 11% at 1;11 and 4% at 2;0.
Discuss the potential significance of these observations for the single system/separate systems models
of bilingual acquisition.

7.4 Bilingual children’s negative sentences

What is the potential significance of the remarks by Paradis and Genesee (1996, p. 15) that ‘The
children in this study did not produce a large number of negative utterances in either language’, and
that ‘There is only one example of a nonfinite negative utterance’. Consider also the potential
significance of the fact that one of the three bilingual children in PG’s study (Gene) produced the
following sentences in French:
(a) Il pas joue dehors (Gene 1;11) (b) Pas il va là (Gene 2;7)
He not plays outside Not he goes there
How convinced are you by PG’s suggestion (p.15) that such sentences are ‘most likely performance
errors’?

7.5 Joshua’s English

Below is a small selection of the English sentences produced by a two-year-old English-German
bilingual boy called Joshua (at age 2;7.13), brought up in a one-parent, one-language environment by a
German mother and English father. To what extent do the relevant data (collected and kindly made
available to me by Bettina Knipschild as part of her ongoing PhD dissertation) lend support for Paradis
and Genesee’s view that the two grammars of bilingual children develop entirely independently?
1. I want it. 2 I want that one. 3. I cut it. 4. I like not that 5. I like it not. 6. Is very good (= It is..)
7. I did make it nice and clean. 8. I did fall it over. 9. I did throw it. 10. Don’t like it (= I don’t…)
11. This is very hot 12. This is cool down. 13. Is warm. 14. I going to bang that one.
15. I going to repair it. 16. I have to bang this. 17. Let’s read Postman Pat and the Woolly Jumpers.
18. He’s pulling the cart. 19. He’s gonna go mith115 the dog. 20. They’re going to play.
21. They’re going there sleep. 22. Going to pick the flowers. 23. I want to put them here.
24. They’re pretty 25. I did pick them. 26. I did put them in here. 27. I did pick not one
28. I going that to bang. 29. I going to bang the railway track. 30. I did drink it. 31 I drink the coffee.
32. We’re gonna go to bed. 33. I want to change the nappy. 34. Juice is that. 35. Milk is that.
36. It is just raining 37. Is just raining. 38. I did walk. 39. I did go in the puddles.
40. I want not tea. 41. I have orange juice. 42. I want to drink it. 43. I want Wurst.
44. Sausages essen116. 45. I want some sausage. 46. I want some cheese. 47. Cheese is that.
48. Up here is some 49. Up there is sandwich. 50. Where’s the hammer? 51. You sit on it.
52. Is gone. 53. I think the locomotive is gonna come. 54. Is going to push.
55. Is going to push them not. 56. No, it’s going to push them. 57. That very clever.
58. That’s very clever. 59. It’s gonna go fast. 60. There’s gonna crash the locomotive as well.
61. There117 was too fast. 62. Es war118 too fast 63. The locomotive is just fell in the ditch119.
64. It was too fast. 65. It is just get out of the ditch. 66. Is going to go again in the ditch
67. There fell in the ditch the locomotive. 68. What is it time? (= ‘What time is it?’)
69. I want take that one. 70. I want to read it not. 71. I want to not read it. 72. I want not 120

115
Mith appears to be a blend of German mit and English with.
116
Essen is the infinitive form of the German verb meaning EAT.
117
The German pronoun er means ‘he/it’.
118
Es occurs in sentences like Es regnet ‘It rains’ and Es kam niemand ‘There came nobody’.
119
Presumably meaning ‘has just fallen’.
120
In reply to his mother asking ‘Do you want to go to the toilet?’
73. Mummy do it not.
Background information
There are several characteristics of German which you should bear in mind when analysing these
examples. (For an introduction to the syntax of clauses in German, see Radford et al. 1999, pp. 349-
56.) Some of these can be illustrated by the sentence below:

i) Ich weiss, dass er nach Berlin gefahren ist

I know that he to Berlin gone is (‘I know that he has gone to Berlin’)
Heads generally precede complements in German (for example, the complementiser dass ‘that’
precedes its complement er nach Berlin gefahren ist ‘he to Berlin gone is’), but there are two
exceptions to this: (i) V occupies head-final position within VP (so that the verb gefahren ‘gone’
follows its complement nach Berlin ‘to Berlin’); and (ii) T occupies head-final position within TP (so
that the present-tense T-auxiliary ist ‘is’ follows its verb phrase complement nach Berlin gefahren ‘to
Berlin gone’). All finite clauses are CPs in German, and both T and C are always strong in finite
clauses in German: accordingly an auxiliary can raise from T to C in a finite main clause, and a main
verb can raise from V through T into C in an auxiliariless finite main clause – and indeed this is what
happens to the verb weiss ‘know’ in the above example. The specifier position within CP in a main
clause gets filled by movement of some expression from (e.g. a subject or complement position)
within TP into CP. Hence we find a range of different types of superficial CP structures like:
ii) [CP Er [C ist] nach Berlin gefahren] iii) [ CP Nach Berlin] [C ist] er gefahren]
He is to Berlin gone To Berlin is he gone
iv) [CP Er [C fährt] nicht nach Berlin] (v) [ CP Nach Berlin [C fährt] er nicht]
He goes not to Berlin To Berlin goes he not
In (ii) and (iii), the T-auxiliary ist (which would otherwise be expected to be positioned at the end of
the clause, as in (i) above) moves from T to C, and the specifier position of CP is filled by preposing
the subject er in (ii) and the complement nach Berlin in (iii). In (iv) and (v), the verb fährt moves from
V through T into C, and the specifier position within CP is filled by movement of the subject er in (iv)
and of the complement nach Berlin in (v). German uses the auxiliary BE with intransitive verbs of
motion121 as in (i-iii), and HAVE with other verbs. German also makes frequent use of expletive-
pronoun structures like Es kam niemand ‘There came nobody’. It also has a phenomenon known as
object shift whereby a definite direct object (like diese Bücher ‘this book’ in the example below) is
moved out of the verb phrase into some position higher up in the clause structure (above/preceding the
negative particle nicht ‘not’):
(vi) Harald darf diese Bücher nicht lesen
Harald must this book not read (‘Harald must not read this book’)
German is also a topic drop language which allows a third-person subject or object pronoun which
occupies the specifier position within CP to have a null spellout if it is a topic (i.e. if it refers to some
previously mentioned or salient entity within the immediate discourse setting). Hence, the preposed

121
More technically, with unaccusative verbs.
object pronoun das ‘this’ can optionally have a null spellout in the following example (from Rizzi
1992, p.105) e.g. if it denotes an object which has already been mentioned in the discourse:
(vii) Das habe ich gestern gekauft
This have I yesterday bought (‘This, I bought yesterday’)

8 Language mixing in bilingual children

8.1 Introduction
 In this chapter we look at language-mixing in bilingual children, focusing on code-mixing
(where children switch between one language and another internally within a sentence).
This phenomenon of intrasentential/sentence-internal code-switching 122 can be illustrated
by the following sentence produced by a bilingual English-Italian child called Lucy at
age 2;8123, with Italian words italicised):
(1) And then ho visto little baby sheep
And then have seen little baby sheep (‘And then I saw some little baby sheep’)
Lucy switches from English to Italian after the adverb then, and switches back into English after the
verb visto ‘seen’. In the previous chapter, we noted Volterra and Taeschner’s claim (1978, p.312) that
in the early stages of bilingual acquisition ‘Words from both languages frequently occur together in
two- to three-word constructions’, and that this is part of the evidence they adduce in support of their
view that bilingual children initially develop a single mixed grammar. However, a number of
quantitative studies have suggested that code-switching is relatively infrequent: reviewing evidence
from a number of studies, Genesee, Nicoladis & Paradis (1995) report that code-mixing is found in
only 1% -7% of bilingual childen’s utterances, and conclude that this very low rate of code-mixing is
inconsistent with Volterra and Taeschner’s Single System Hypothesis/SSH (a key argument given by
Volterra and Taeschner in support of SSH being that bilingual children code-mix frequently at stage I).
Genesee (1989) suggests that mixing may be more common in bilingual children whose parents code-
mix in talking to their children than in children whose parents obey the one person, one language
principle.
Two questions about code-mixing which have been extensively debated in the bilingual acquisition
literature are: (i) Why does code-mixing occur? and (ii) Where does code-mixing take place (i.e. at
what points within the structure of a phase/sentence). I’ll look briefly at the question of why children
code-mix before turning to consider in some detail the question of where code-switching does (and
does not) occur.

8.2 Why bilingual children code-mix

Two main answers have traditionally been given in the research literature to the question of
why children code-mix internally within sentences. Lindholm and Padilla (1978) formulated a Lexical
Gap Hypothesis to the effect that bilingual children code-mix when they have a lexical gap (i.e. no
word for a particular concept in one language, but a word for the concept in the other language): more
specifically, they suggested (1978, p.332) that ‘Children mix when they do not know the
corresponding word in the other language124’. A related Dominance Hypothesis was put forward by
Petersen (1988) who maintained that bilingual children tend to switch into their dominant/stronger
language when using their weaker language because many of the words/structures they need are
lacking in the weaker language: this predicts that switching takes place from the the weaker to the
stronger language far more often than from the stronger to the weaker. See also Kielhöfer (1987).
A study which presents empirical evidence in support of the Lexical Gap Hypothesis is Gawlitzek-
Maiwald and Tracy (1994). They analyse the speech of a two-year-old bilingual German-English child
(Hannah), and report that she went through a stage at which she had acquired auxiliaries in German
but not in English. So, for example, when her father chases her and says I’ve got you using the present-
122
The term code-mixing is used to ‘refer to the co-occurrence of elements from two or more languages in a
single utterance’ (Genesee 1989, p.162), whereas the more general term code-switching is used to describe ‘the
mixing of languages within and across utterances or sentences’ (Lanza 1992, p.636). The further term code-
changing is employed by McClure and McClure (1975) and Wentz and McClure (1977), but I shall not use it
here.
123
The Lucy examples given in this chapter are from a private corpus collected by Gabriele Azzaro which he has
kindly made available to me.
124
However, they note an apparent counterexample in their data, in that the Spanish-English bilingual children
they studied alternated between English and and Spanish y ‘and’ in English co-ordinate structures like ‘This one
y this one y this one’. They maintain that this is because Spanish y ‘and’ was the dominant coordinating
conjunction for the children (i.e. the one they used most frequently). They seem to be implying that children may
switch from one language to another when they use a particular word more frequently in one language than its
counterpart in the other.
tense T constituent (ha)ve, Hannah repeats his sentence as I got you, omitting the auxiliary have. By
contrast, she has acquired a number of auxiliaries in German, and so in sentences like (2) below where
she wants to use an auxiliary, she will use a German auxiliary like haben ‘have’ and then switch to
English afterwards (German words being italicised):
(2) Sie haben gone away ‘They have gone away’
Gawlitzek-Maiwald and Tracy term this kind of mixing bilingual bootstrapping (the idea being that
borrowing auxiliaries from German helps Hannah build clauses in English) and argue that Hannah
uses German auxiliaries to compensate for a lexical gap (viz. the absence of auxiliaries in her English).
Nicoladis and Secco (1998) set out to test the Lexical Gap and Dominance Hypotheses by
undertaking a longitudinal study (from age from 1;0 to age 1;6) of spontaneous speech samples from a
bilingual Portuguese-English boy whose father spoke Brazilian Portuguese and whose mother spoke
English. English was the boy’s dominant language (measured in terms of vocabulary size). They found
that 97% (88/91 cases) of the code-switching by the child occurred when the child had no translation
equivalent, suggesting that ‘the child’s code-switching...can be accounted for by gaps in the productive
lexicon’ (p.582) and that ‘Bilingual children code-switch to fill lexical gaps in their vocabulary’. They
also reported that the child showed an average rate of code-mixing of 26% when speaking in his
weaker language (Portuguese) to his father and 7% when speaking in his dominant language (English)
to his mother. This suggests that there is also a strong dominance effect (which could perhaps be
accounted for in terms of the Lexical Gap Hypothesis analysis, if the child has more lexical gaps in the
weaker language).

8.3 Where bilingual children code-mix: Poplack’s constraints

There have been a number of attempts to argue that principles of Universal Grammar
determine where code-switching can (and cannot) take place within sentences. Shana Poplack (1980)
argued that UG principles impose both morphological and syntactic constraints on Code-Switching in
adult grammars. A morphological constraint which she proposed (1980, p.585) is the Free Morpheme
Constraint, which can be paraphrased informally as follows:
(3) Free Morpheme Constraint/FMC
Code-switching is barred between a bound morpheme (i.e. an inflectional or derivational affix)
and another morpheme.
FMC predicts (e.g.) that bilinguals do not attach English affixes to Spanish word stems and hence do
not produce blends like canting (attaching the English progressive affix -ing to the Spanish verb-stem
cant- ‘sing-’) instead of Spanish cantando or English singing; nor do they attach a Spanish affix to an
English stem (and so they don’t produce forms like singando, with the Spanish progressive suffix
-ando added to the English verb stem sing-). In a study of 1,835 code-switches by adult Puerto-Rican
bilinguals in New York, Poplack found only 5 violations of FMC. A number of other studies have
claimed empirical support for the constraint, including Bentahila and Davies (1983) for Arabic-French
bilinguals, Berk-Selligson (1986) for Spanish-Hebrew bilinguals, and Clyne (1987) for
German/Dutch-English bilinguals.
However, the descriptive adequacy of FMC has been questioned in a number of subsequent studies
of adult bilinguals, including Nartey (1982), Bokamba (1989), Kamwangamalu (1989), Stenson
(1990), Choi (1991), Mahootian (1993), Myers-Scotton (1993), Timm (1994), Halmari (1997) and
Chan (1999). Many of the reported counterexamples involve switching between a stem in one
language and an affix in another: for example, Gardner-Chloros and Edwards (2004, p.109) report
English-French adult bilinguals producing mixes like Je sunbathais ‘I was sunbathing’, in which the
French past tense affix -ais is attached to the English verb stem sunbath-. It is therefore
conceivable that (in the words of Sankoff and Poplack 1981, p.5) that in such
cases, the stem has been ‘integrated into the language of the bound morpheme’,
so that they involve borrowing rather than switching (as suggested by Poplack,
Wheeler and Westwood 1989, Sankoff, Poplack and Vanniarajan 1990, and Poplack and Meechan
1995). This would mean that a stem is being borrowed from one language L1 into a
second language L2, and then being treated as an L2 stem and so being inflected
with L2 affixes. (See Pfaff 1979, Muysken 1995, and Boumans 2001 for discussion
of the concepts of borrowing and switching and Poplack 1980, Nartey 1982, and
Halmari 1997 on the idea that phonological integration may be a criterion for
borrowing.)125
Violations of FMC have also been reported in studies of bilingual children. Wentz and
McClure (1975, p.7) report that only English verb stems which have been adapted
to Spanish carry Spanish affixes: hence the children in their study produced forms
like tinqué for ‘I thought’ but not thinké – suggesting that borrowing may be a
possible account. Petersen (1988) reports a Danish-English bilingual child attaching English
affixes to Danish word-stems (as in bors ‘lives’ and vasking ‘washing’, where English present-tense
and progressive-aspect affixes have been added to Danish verb-stems) but not the converse pattern of
attaching Danish affixes to English stems (so that there are no mixed forms like washer ‘washes’
where the Danish third person singular present tense affix -er has been added to the English stem
wash-). Lanza (1992) reports the converse pattern of an English-Norwegian bilingual child attaching
Norwegian affixes to English stems (as in looker ‘looks’ where the Norwegian third person singular
present tense affix -er is attached to the English stem look), but not the converse pattern of attaching
an English affix to a Norwegian stem (hence no occurrences of forms like husks ‘swings’, where the
English third person singular present tense -s affix has been attached to the Norwegian stem husk-).
Petersen and Lanza maintain that the relevant data show that affixes from a child’s dominant language
can be attached to stems from the child’s non-dominant language (but an affix from a non-dominant
language cannot be attached to a stem from a dominant language). Since affixes are functional
morphemes, we can generalise this condition as follows:
(4) Dominance Constraint
Functional morphemes in code-mixed sentences are from the dominant language
Formulated in these terms, such a constraint has implications for syntax, in that it implies that the
functional superstructure of code-mixed utterances (i.e. the functional categories they contain, and the
position which they occupy) will be from the dominant language – and indeed claims of this ilk have
been made in the guise of the Matrix Code Principle of Kamwangamalu (1989), the Dual Structure
Principle of Sridhar and Sridhar (1980), and the The Matrix Language Frame Model of Myers-Scotton
(1992, 1993, 1995). However, the Dominance Constraint does not seem to hold for all children: the
English-Italian bilingual child (Lucy) in the Azarro corpus has English as her dominant language, but
produced both mixed nouns combining an Italian stem with an English affix (like muccas ‘cows’) and
mixed nouns comprising an Italian suffix added to an English noun stem (like bono ‘bone’, and
butterflaia ‘butterfly’).
Poplack (1980, p.586) also proposed an Equivalence Constraint on code-mixing which amounts to
positing that mixing is only possible where two languages have equivalent word orders. Sankoff and
Poplack (1981, p.4) formulate the constraint in the following terms:
(5) Equivalence Constraint/EC
‘The order of sentence constituents immediately adjacent to and on both sides of the switch
must be grammatical with respect to both languages involved.’
EC predicts that an French-English bilingual could say I’ve lost les clefs ‘I’ve lost the keys’ because
there is superficial equivalence between the word order in the English sentence I’ve lost the keys and
its French counterpart J’ai perdu les clefs; but an adult French-English bilingual would not switch in a
sentence like I’ve lost les (les = ‘them’) because les in French is a clitic pronoun which must
procliticise (= attach to the front of) the finite auxiliary/verb in a finite clause, so that the absence of
equivalence in word order between the English sentence I’ve lost them and its French counterpart Je
les ai perdues (literally ‘I them have lost’) bars code-switching between an English finite verb and a
125
An alternative account of switching between affix and stem would be to suppose that switching is contingent
on lexical properties of individual affixes (rather than UG principles). For example, just as English some
selective affixes which attach only to particular types of stem (e.g. -ity attaches only to adjectives of Graeco-
Roman origin), so too English also has unselective affixes which attach to any type of stem (e.g. -ness attaches to
any kind of adjective). Whether or nor switching is possible from an affix in one language to a stem in another
may therefore dependent on whether the affix is a selective or unselective one.
French clitic pronoun. Poplack reports only 11 violations of EC in 1,835 switches produced by her
subjects.
Poplack argues that empirical support for her Equivalence Constraint comes from work by Gringàs
(1974), who constructed model Spanish-English code-switching sentences and asked adult Chicano
bilinguals to judge whether or not code-switching was acceptable at particular points in each sentence.
One of Gringàs’ findings was that all 26 informants rejected code-switching between the verb want
and its italicised clausal complement in the following sentence:
(6) El man que came ayer wants John comprar a car nuevo
The man that came yesterday wants John to.buy a car new
‘The man that came yesterday wants John to buy a new car’
Why should they reject switching between want and its italicised infinitival complement? Poplack’s
answer is that this is because verbs of the WANT class are not followed by the same type of
complement in the two languages: the English verb want is followed by an infinitive complement in
sentences like (6) (where the complement clause has a subject distinct from that of the main clause),
whereas its Spanish counterpart querer is followed by a subjunctive complement introduced by que
‘that’ (so that Spanish requires a structure paraphrasable as ‘He wants that John buy a new car’, where
the verb buy is subjunctive in mood). Because want and querer are not followed by the same types of
complement, the Equivalence Constraint predicts that switching will be blocked in sentences like (6).
By contrast, switching is predicted to occur between the italicised auxiliary and the bold-printed
main verb in sentences such as:
(7) Siempre está promising cosas
Always is promising things (‘He/She is always promising things’)
This is because both the Spanish progressive auxiliary ESTAR ‘be’ and its English counterpart BE are
followed by a complement headed by a verb in the progressive participle form (ending in -ing in
English, and -ndo in Spanish). Because ESTAR and BE are followed by equivalent types of complement
in the two languages, the Equivalence Constraint predicts switching to be possible in structures like (7)
– and indeed Poplack (1980, p.596) reports adult Puerto-Rican bilinguals in New York producing such
structures. More generally, she reports (loc. cit.) that 99% of code-switching in her study conformed to
the Equivalence Constraint.
Although Poplack’s constraints were formulated in relation to code-switching in adult bilinguals,
Fred Genesee (2002) claims that Poplack’s constraints also operate in child grammars, reporting
(2002, p. 186) that 99% (425/429) of the code-mixed utterances produced by a group of 10
English/French bilingual children that he studied conformed to the Free Morpheme Constraint and the
Equivalence Constraint.
However, violations of the Equivalence Constraint have been reported in a number of studies of
adult bilinguals. One such case relates to head-complement structures. More specifically, examples
have been reported of switching between languages in structures where one language has head-initial
HEAD+COMPLEMENT order, and the other has head-final COMPLEMENT+HEAD order. In this regard,
consider the position of the bold-printed complementiser with regard to its italicised clausal
complement in the following instances of English-Korean code-switching (from Nishimura and Yoon
1998):
(8)(a) I am out of town ira malhaseyo
I am out of town that tell (= Tell (him) that I am out of town)
(b) I realized that nae hangukmal ajik yakhanko
I realized that my Korean (is) still weak
English positions heads before (and Korean after) their complements, so the differing word-orders in
the two languages mean that the Equivalence Constraint would wrongly predict that it should not be
possible to switch between a head in one language and a complement in the other. Yet in (8a) we find
switching between a Korean complementiser and its English complement, and in (8b) between an
English complementiser and its Korean complement. Numerous other studies have reported a similar
pattern of switching in a range of other structures: see Chan (1999, chapters 3 and 4) for further
exemplification and discussion.
An interesting property of the structures in (8) is that the Korean complementiser ira in (8a)
occupies the head-final position which it occupies in Korean, and conversely the English
complementiser that occupies the head-initial position which it occupies in English. This suggests the
possibility of an alternative linearity constraint along the following lines:
(9) Head Position Constraint
The position of a head with respect to its complement(s) is determined by
the language of the
head (i.e. conforms to the position of the relevant type of head in the
language of the head).
A proposal within the spirit of (9) is made (in rather different frameworks) by
Mahootian and collaborators (Mahootian 1993, 1996; Santorini and Mahootian
1995; Mahootian and Santorini 1996), Nishimura (1997), and Chan (1999).
However, it seems doubtful that (9) could be of more general validity, as the
following examples of English-Japanese code-switching produced by a Japanese
L2 learner of English (from Radford and Yokota 2005) illustrate:
(10)(a) What color do you think to Hanako like?
(b) Which flower do you think to Hanako like?
Although (as we have seen) Japanese is a head-final language in which
complementisers are positioned at the end of the clause they introduce, the bold-
printed Japanese complementiser to ‘that’ in each of the examples in (10) is
positioned in front its italicised complement, and hence shows the head-initial
word order characteristic of English. What seems to be going on in (10) is that the
sentence has an English functional superstructure (with all heads positioned
before their complements), but a Japanese function word is inserted in the head C
position of CP.
Something similar seems to be going on in relation to noun-adjective switches in
nominal expressions like those italicised in the sentences below:
(11)(a) Tenían platas flacas, pechos flat (Spanish-English, Poplack 1980)
Had legs skinny, chests flat (= ‘They had skinny legs, flat chests’)
(b) He presented a paper exceptionnel (English-French, Bokamba 1989: 282, (16a))
He presented a paper exceptional (= ‘He presented an exceptional paper’)
The Equivalence Constraint wrongly predicts that the italicised strings in (11) should not occur, since
(for the relevant class of adjectives) Spanish and French show NOUN+ADJECTIVE word-order, whereas
English shows the converse ADJECTIVE+NOUN order. And since adjectives are modifiers of nouns (and
not heads of the relevant nominal expression, the Head Position Constraint (9) does not account for the
data either. If we adopt the proposal by Picallo (1991), Ritter (1991) and Cinque
(1994) that nouns in French and Spanish raise to an intermediate functional head
(= Num) between D and N, we can suppose that the italicised nominals have a
French/Spanish functional superstructure containing a Num head which triggers
raising of the nouns pechos/paper to Num.
Other structures produced by adult bilinguals which would seem to violate the Equivalence
Constraint are reported by Nartey (1982), Bentahila and Davies (1983), Berk-Seligson (1986),
Bokamba (1989), Kamwangamalu (1989), Stenson (1990), Chan (1993), Myers-Scotton (1993), Timm
(1994), Halmari (1997) and Chan (1999). Poplack and her colleagues (Poplack 1988, Poplack, Sankoff
and Miller 1988, Poplack, Wheeler and Westwood 1989, Sankoff, Poplack and Vanniarajan 1990,
Poplack and Meechan 1995) have attempted to argue that such cases may involve borrowing (but see
Myers-Scotton 1993, Santorini and Mahootian 1995, and Chan 1999 for a critique
of this idea).
A theoretical problem posed by the Equivalence Constraint lies in Poplack’s assumption that
equivalent word order (rather than equivalent hierarchical structure) is required for code-switching to
be possible. This poses the problem that ‘Principles of grammar are formulated in terms of hierarchical
relations rather than linear order’, so that constraints on code-mixing would be expected to be
‘structural rather than linear’ (Di Sciullo, Muysken and Singh 1986, p.3). The problem is
compounded in theories (like the Minimalist model of syntax developed in Chomsky 1995, 1998,
1999, 2001, 2002, 2005a,b) which sees syntactic structures as inherently unordered (word-order being
added to unordered syntactic structures via linearization conditions in the PF component 126).

8.4 Structural constraints on code-mixing

In work subsequent to Poplack’s study, a number of linguists suggested syntactic constraints
on code-switching in adult grammars which were purely structural in nature (in the sense that they
refer to hierarchical syntactic structure alone and ignore word-order differences between languages).
An early constraint of this kind was the Government Constraint proposed in a (1986) paper by Di
Sciullo, Muysken and Singh (=DMS), which we can outline informally as follows:
(12) Government Constraint
No code-mixing can take place between a governor and an expression which it governs
Since DMS take contentives (i.e. content words such as verbs, adjectives, nouns and prepositions) to
be governors and assume a particular definition of government (based on the syntactic relation
c-command) which takes heads to govern their complements but not their specifiers, we can
informally regard their constraint as stating that mixing cannot take place between a content word and
its complement. Consequently, DMS’s Government Constraint allows code-mixing between a
specifier and a head (e.g. between a subject and an auxiliary/verb), and between an adjunct and the
expression it is attached to (e.g. an adverb and an expression which it modifies); it also allows
switching between a functor (i.e. a function word like a determiner or auxiliary) and its complement 127.
DMS present empirical evidence in support of their constraint from studies they undertook of French-
Italian and English-Italian code-mixing by adults in Montreal, and of Hindi-English code-mixing by
adults in urban North India.
Illustrative sentences showing typical patterns of code-switching which DMS found in their study
are given below (where italicised words are French, and other words are Italian):
(13)(a) Ha ricevuto il diplôme
Has received the diploma (‘He/She received the diploma’)

(b) Basta che marche

Suffices that works (‘It suffices that it works’)
(c) Parce que hanno donné des cours
For.it that have given some courses (‘Because they gave some courses’)
(d) La plupart des Canadiens scrivono ‘c’
The majority of.the Canadians write ‘c’
In examples (13a-c), switching takes place between a function word in one language and a
complement in another: in (13a) between the Italian determiner il ‘the’ and its complement (the French
noun diplôme ‘diploma’); in (13b) between the Italian complementiser che ‘that’ and its complement

126
The PF component is the component of the grammar which converts syntactic structures into PF-
representations (i.e. representations of their Phonetic Form). One such linearization (i.e. word-order) condition
can be paraphrased informally by saying that ‘In English, specifiers precede heads, and complements follow
heads.’
127
I have tried to give a general sense of the spirit of the constraint proposed by Di
Sciullo, Muysken and Singh without entering into specific details, because government has not played a
role in syntax since the end of the 1980s, a baffling variety of different definitions of government were proposed
in work in the 1980s, and some of the assumptions made by Di Sciullo, Muysken and Singh about syntactic
structure are somewhat different from those which would be made nowadays. Attempts to refine the Government
Constraint can be found in Meisel (1994) and Halmari (1997): an evaluation of the latter can be found in Chan
(1999) and Boumans (2001).
(the French verb marche ‘works’); in (13c) between the French complementiser que ‘that’ and a
complement headed by the Italian auxiliary hanno ‘have’, and between the Italian auxiliary and its
French verb phrase complement donné des cours ‘given some lectures’. In (13d), switching takes
place between a specifier and a head – i.e. between the French subject expression la plupart des
Canadiens ‘the majority of Canadiens’ and the associated Italian verbal head scrivono ‘write’128. Since
none of these cases involve switching between a content word and its complement, they all conform to
DMS’s Government Constraint.
Belazi, Rubin and Toribio (henceforth BRT) argued that DMS’s Government Constraint ‘fails in
that it is too restrictive’ (1994, p.224). In BRT’s own study of code-mixing in adult Spanish-English
adult bilinguals, they found patterns such as the following (with Spanish words italicised):
(14)(a) The professor said que el estudiante había recibido una A
‘The professor said that the student had received an A’
(b) El professor dijo that the student had received an A
‘The professor said that the student had received an A’
(c) *The professor said that el estudiante había recibido una A
‘The professor said that the student had received an A’
(d) *El professor dijo que the student had received an A
‘The professor said that the student had received an A’
BRT argue that the Government Constraint wrongly predicts that the verb and the complement clause
following it must be in the same language – and yet this claim is clearly falsified by sentences like
(10a) and (10b). Conversely, it also wrongly predicts that switching should be possible between a
complementiser like that/che and the complement expression (= TP) following it, and yet BRT claim
that structures like (10c) and (10d) were not produced by the Spanish-English bilingual adults in their
study129. A number of other studies (including Clyne 1987, Klavans 1985, Romaine 1989, Nortier
1990, Muysken 1990, Stenson 1990, Chan 1999 and Gardner-Chloros and Edwards 2004) have
pointed out that the Government Constraint wrongly predicts that switching should not occur between
verb and object – and yet verb-object switches are reported by the relevant authors to be relatively
frequent.
BRT agree with DMS that constraints on code-switching should be expressed in purely structural
terms, but unlike DMS, BRT maintain that the relevant constraint is that switching is barred between a
functor and its complement (not between a contentive and its complement). We can outline BRT’s
constraint informally in the following terms:
(15) Functional Head Constraint/FHC
Code switching is barred between a functional head (like C, T, Neg, D or Q) and its
complement, because the two must carry the same language feature
BRT argue that FHC makes correct predictions about where adult Spanish-English bilinguals code-
switch in sentences such as the following (with Spanish words italicised):
(16)(a) The police officers have seen un ladrón (= ‘The police officers have seen a thief’)
(b) Los policías han visto a thief (= ‘The police officers have seen a thief’)
(c) *The police officers have visto un ladrón (= ‘The police officers have seen a thief’)
(d) *Los policías han seen a thief (= ‘The police officers have seen a thief’)
More specifically, the Functional Head Constraint correctly predicts that switching can take place
between a verb like seen/visto and the complement following it (because main verbs are content
128
Since finite verbs in Italian (as in French) raise from V to T, switching in (13d) takes place between the
specifier and head of TP.
129
However, the empirical basis of BRT’s claim would seem to be undermined by the observations made by
Halmari (1997) and MacSwan (1997) that Spanish-English bilinguals they consulted disagreed with the
judgments in Belazi et al. (1994) and (e.g.) accepted sentences in which code-switching takes place between a
Spanish complementiser and an English clause.
words), but not between an auxiliary like have/han and the complement following it (because
auxiliaries are function words, and FHC bars switching between a function word and its complement).
BRT further note that if inflectional morphemes are analysed as functional heads, FHC will provide a
principled account of Poplack’s (1980) Free Morpheme Constraint/FMC which bars switching
between a head and an inflectional morpheme.

Toribio and Brown (1995) claim that FHC holds in child as well as adult grammars. Although they
present no developmental evidence in support of this, we can illustrate the kind of code-mixing
structures predicted to occur by FHC in terms of the following utterances reported by Lanza (1992) to
have been produced by a two-year-old bilingual Norwegian-English girl called Siri (with Norwegian
words being italicised):
(17)(a) Mama stay ut ‘Mama stay out’ (b) Bygge cow ‘build cow’
(c) I et lite house ‘In a little house’ (d) Jeg lean over ‘I lean over’
In (17a), switching takes place between an English lexical verb stay and its Norwegian complement ut
‘out’; conversely in (17b), switching occurs between a Norwegian lexical verb bygge ‘build’ and its
English complement cow. In (17c) switching occurs between an adjective lite ‘little’ and the noun
house which it modifies; and in (17d) switching takes place between a Norwegian subject jeg ‘I’ and
its associated English verb lean. Since in none of these cases does switching take place between a
functor and its complement, data such as (17) are consistent with the Functional Head Constraint.
However, if we look at some of the other code-mixed utterances produced by Siri (like those in
(18) below), we find that they don’t appear to obey FHC:
(18)(a) Ikke rain now ‘Not rain now’ (b) Mer paper ‘more paper’
In (18a), switching appears to take place between a NEG constituent (the negative particle ikke ‘not’)
and its complement rain now; and in (18b), switching occurs between a Q constituent (the quantifier
mer ‘more’) and its complement. Since Belazi, Rubin and Toribio (1994) claim that NEG and Q are
functional categories, sentences such as (18a,b) seemingly challenge the descriptive adequacy of the
claim by Toribio and Brown (1995) that the Functional Head Constraint holds in child grammars.
Moreover, Vihman (1985) reports that the majority of her son’s mixed utterances involve the use of an
English function word with an Estonian noun – in clear violation of the Functional Head Constraint 130.
Similarly, Petersen (1988) reports a bilingual Danish-English child combining English determiners
with Danish nouns to form structures such as the dukke ‘the doll’, a bog ‘a book’, and a flaske ‘a
bottle’. Additionally, examples like (13a-c) above suggest that bilingiual adults also produce structures
which violate FHC, in that we find switching between the Italian determiner il ‘the’ and the French
noun diplôme ‘diploma’ in (13a), between the Italian complementiser che ‘that’ and the French verb
marche ‘works’ in (13b), between the French complementiser que ‘that’ and a TP complement headed
by the Italian T-auxiliary hanno ‘have’ in (13c), and between the Italian T-auxiliary hanno ‘have’ and a
VP complement headed by the French verb donné ‘given’ in (13c). Such structures call into question
the descriptive adequacy of the FHC. Moreover, given BRT’s claims that FHC subsumes Poplack’s
Free Morpheme Constraint/FMC, structures which are problematic for FMC are also problematic for
FHC.
In addition to questions surrounding its descriptive adequacy, there are theoretical questions raised
by the Functional Head Constraint. One such relates to the difficulty of determining whether a given
function word is or is not a head in a given type of structure. In this connection, take another look at
the structure of (18a). If we suppose that the negative particle ikke is the head NEG constituent of
NEGP, then this will be a sentence containing the NEGP constituent shown in skeletal form below, in
which ikke is the head of NEGP and rain now is its complement:
(19) [NEGP [NEG ikke] rain now]
And since the Functional Head Constraint bars switching language between a functional head and its
complement, FHC would wrongly predict that children should not produce structures like (19), since

130
However, as Meisel (1994) notes, not all the items which Vihman takes to be functors would be considered to
be such in more recent work.
they involve switching after a functional NEG head containing ikke. However, work dating back to
Pollock (1989) has argued that negative particles like English not and French pas actually occupy the
specifier position within a NEGP headed by an (often null) NEG constituent. If so, the NEGP
constituent in (19) will have the structure shown in (20) below:
(20) [NEGP ikke [NEG ø] rain now]
And since the Functional Head Constraint/FHC does not bar switching between a functional head and
its specifier (e.g. between the null NEG head in (20) and its specifier ikke), FHC would then make the
contrary prediction that children would indeed code-switch in sentences like (18a). Parallel
indeterminacy problems arise with determiner and quantifiers, which are taken by some linguists to
occupy the head position of DP or QP, but by others to occupy the specifier position within DP/QP.
Problems also arise in relation to determining whether a particular type of switch involves a
function word or not. In this connection, consider adjective-noun switches in nominals like et lite
house ‘a little house’ in (17c) above. At first sight, such structures might simply seem to involve
switching between two substantive categories. However, Cinque (1994) argues that attributive
adjectives are specifiers of an abstract functional head (below denoted as F). If this is so, the relevant
nominal will have a structure along the lines shown below:
(21) [DP [D et] [FP lite [F ø] [N house]]]
The Functional Head Constraint will then predict that it should be possible to switch language between
the adjective lite and the functional head F, but not between F and its complement. However, where (as
here) the functional head is null, it is impossible to be certain where the switch takes place. The more
general point being made here is that structural indeterminacy makes it difficult to test FHC
empirically.
A further problem posed by FHC is that it is often difficult to determine whether a given word in a
language is a function word or not: for example, modals like will/would, can/could etc. are generally
taken to be auxiliaries (hence function words) in English because they only have finite forms (hence
can only occupy the head T position of TP). But their counterparts in many other languages (e.g.
French, Italian, Spanish, German etc.) have non-finite (infinitive and participle) forms and in this
respect seem to behave more like typical main verbs (and hence to resemble content words more than
functors). Moreover, if (as in Roberts 1998) English narrow scope modals like need do not originate in
T (but rather in a position below T), it is again far from clear whether they are true functors (and hence
whether FHC would predict that switching can take place between need and its complement).
Furthermore, even the functional status of T constituents is far from clear, given Chomsky’s (1999)
suggestion that T may not be a functional category at all, but a substantive category.
An additional indeterminacy surrounding FHC concerns the question of the level at which the
constraint applies. Meisel (1994, p.419) claims that ‘Grammatical constraints on code-switching apply
to surface structure properties of the languages involved’, and Gardner-Chloros and Edwards (2004,
p.125) echo this view with their claim that ‘the processes involved in CS are “surface” processes’.
However, if (as Belazi, Rubin and Toribio assume) switching is permitted only when a functional head
and its complement agree in respect of the language feature which they carry, and if this type of
language-agreement is checked in the same way as other (e.g. person/number) agreement features,
theoretical considerations suggest that this cannot be so. To see why, consider what the Functional
Head Constraint would say about code-mixing in a sentence like the following, produced by an
English-German bilingual boy called Danny at age 2;2 (from Redlinger and Park 1980):
(22) Die Mädchen’s going night-night (‘The girl’s going night-night’)
Let us assume that (22) has a structure which includes the TP shown in simplified form below:
(23) [TP Die Mädchen [T (i)s] going night-night]
If we suppose that FHC holds of superficial syntactic structures, a structure like (23) will seemingly
satisfy FHC, because mixing takes place between a functional head (= the present-tense auxiliary is)
and its subject/specifier in (23), and FHC only bars code-switching between a functor and its
complement. But when we look rather more carefully at the derivation of the sentence, the picture
changes somewhat.
 Under the VP-Internal Subject Hypothesis widely assumed in work in syntax since the mid 1980s,
the syntactic subject die Mädchen ‘the girl’ of the T-constituent (present-tense auxiliary) (i)s originates
as the subject of the verb go. Accordingly, the sentence is derived by merging the verb go with its
complement night-night and its subject die Mädchen to form the verb phrase [VP die Mädchen go
night-night]131. The resulting verb phrase is then merged with a present-tense T constituent containing
the progressive-aspect auxiliary be, forming the TP shown in simplified form below:
(24) [TP [T be] [VP die Mädchen [V go] night-night]]
Pesetsky’s (1995) Earliness Principle requires that feature checking/valuation should take place as
early as possible in the derivation: this means that all feature-checking operations affecting T must
apply at this point, on the TP cycle. Accordingly, the T-auxiliary be agrees in aspect with the verb go
and assigns go a progressive aspect feature (as in Adger 2003), so that go is ultimately spelled out as
the progressive participle form going. Likewise, be also agrees in person and number with (and
assigns nominative case to) die Mädchen at this point. Given that FHC requires a functional head to
carry the same language feature as a constituent which it enters into a checking relation with and sees
this as part of a more general feature-checking operation, FHC will require T and the subject it agrees
with to be in the same language. However, this same-language condition is not met in (24), since the
head T constituent of TP contains the English auxiliary be and agrees with the German nominal
expression die Mädchen. Thus, FHC wrongly predicts that switching will not be possible in such
structures – and yet we see from sentences like (22) that it nonetheless occurs.
More importantly, the Functional Head Constraint seems to be implausible from a conceptual point
of view, in implicitly treating language as a grammatical feature. After all, the reason why switching is
not possible between a functional head and its complement (according to BRT) is that a functional
head and its complement must match in the language feature they carry (e.g. a functional head and its
complement must both be English words). But this amounts to treating language as a grammatical
feature, and saying e.g. that just as an aspectual auxiliary like perfect HAVE (in a sentence like They
have gone) requires a perfect participle complement like gone which matches the [perfect-aspect]
feature of HAVE, so too HAVE requires a complement carrying a language feature which matches the
[English-language] feature of HAVE. However, being an English word is arguably a very different kind
of property from being perfect in aspect, and while it is in the nature of grammatical operations that
they are sensitive to grammatical features like tense, aspect, mood, person, number, gender and so on,
it seems far less plausible to suppose that they are sensitive to a language feature of this kind.
Chan (1999) proposes to deal with problems besetting FHC by revising it in the manner sketched
informally below:
(25) Head Selection Constraint/HSC
The selectional requirements of heads must be met 132
(25) offers the obvious advantage over FHC that it does away with the need to posit that syntactic
computations are sensitive to language features. It is also a constraint which is not specific to bilingual
grammars, since monolingual grammars are subject to the same requirement. It also avoids singling
out functional heads as subject to special constraints which lexical heads are not subject to. FHC also
offers the empirical advantage that it can handle many switches which proved problematic for FHC to
account for – as we can illustrate in relation to (13c) above, repeated as (26) below (with the word
hanno ‘have’ being Italian and the italicised words being French):
(26) Parce que hanno donné des cours
131
Syntactic structures are simplified throughout by showing only heads and phrases, not intermediate
projections (and ignoring the possibility that verb phrases may have a more abstract shell structure incorporating
a light verb). Note that there is code-switching internally within the verb phrase here, between the subject die
Mädchen and the verb expression going night-night, but this type of switching would be predicted to occur by
FHC.
132
As Chan notes, there are parallels with the Subcategorization Constraint proposed by Bentahila and Davis
(1983, p.329) to the effect that ‘All items must be used in such a way as to satisfy the (language-particular)
subcategorization restrictions imposed upon them.’ A technical complication which I gloss over here is that Chan
claims that functional heads impose both c- and s-selection restrictions (i.e. categorial and semantic restrictions)
on their complements, whereas lexical heads impose only s-selection restrictions.
 For.it that have given some courses (‘Because they gave some courses’)
The requirement for the French complementiser que ‘that’ (as used here) to select a complement
headed by an indicative T-constituent is met in (26) by virtue that the head word of its complement is
the indicative T-auxiliary hanno ‘have’. Likewise the requirement for the Italian T-auxiliary hanno
‘have’ to select a complement headed by a perfect participle is met in (26) in that the head word of its
complement is the French perfect participle donné ‘given’. Thus, Chan’s Head Selection Constraint
would appear to be descriptively superior to the Functional Head Constraint 133.
And yet, a theoretical question which arises from HSC is why the constraint should single out the
selectional properties of heads, and thereby exclude other grammatical properties (e.g. their agreement
properties, case-marking properties, and so on). After all, in a sentence such as (13a) above (repeated
below):
(27) Ha ricevuto il diplôme
Has received the diploma (‘He/She received the diploma’)
switching between the Italian determiner il ‘the’ and the French noun diplôme ‘diploma’ not only has
to satisfy the selectional requirement for D to have an N complement, but also has to satisfy the
Concord requirement for the determiner and noun to agree in number and gender (both being
masculine singular forms). In much the same way, in a structure such as (24) above, the English
T-auxiliary is has to agree with and assign nominative case to the German subject die Mädchen ‘the
girl’ – suggesting that case and agreement properties need to be taken into account as well as
selectional properties. Such considerations suggest that we need a much more general constraint than
FHSC. In the next section, we look at the form that such a generalised constraint might take.

8.5 An alternative convergence account of code-switching

An intuition which a number of studies have sought to capture is that code-mixing is only
possible between one constituent and another if there is congruence between the two (Weinreich 1964,
Myers-Scotton 1993, Myers-Scotton and Jake 1995, Jake and Myers-Scotton 1997, Sebba 1998), or
structural consistency (Lindholm and Padilla 1978), or absence of structural conflict (Pfaff 1979), or
equivalence (Poplack 1980), or structural integrity (Sankoff and Poplack 1981), or correspondence
(Joshi 1985), or syntagmatic coherence (Di Sciullo, Muysken and Singh 1986), or grammatical
cohesiveness (Meisel 1994), or feature-matching (MacSwann 1997). What I shall argue here is that we
can provide a coherent characterisation of the intuition underlying these claims in terms of the notion
of convergence which is a core concept of the Minimalist theory of syntax developed by Chomsky
(1995, 1998, 1999, 2001, 2002, 2005a,b).
Chomsky posits that syntactic structures are derived by taking a set of items out of the lexicon
(these constituting the lexical array out of which a given sentence structure is to be built), and then
combining these into successively larger expressions via a series of merger, movement and agreement
operations. On this view, mixing may come about when the items in the lexical array are drawn from
two different lexicons. Once a particular structure has been formed, it is handed over to the semantic
component of the grammar to be assigned a semantic representation, and to the PF component to be
assigned a representation of its Phonetic Form. In order to avoid a derivation crashing (i.e. failing), the
syntactic structures which the grammar generates must be convergent at the PF and semantics
interfaces – i.e. they must be of such a form that they can be assigned appropriate semantic and
phonetic representations. Below, I shall suggest that the same condition holds for code-mixing –
namely that code-mixing is only possible where it leads to a convergent derivation. In what follows, I
shall concentrate on the role of agreement in ensuring that a given syntactic structure is convergent at
the PF interface (i.e. contains all the information required in order for each of the constituents it
contains to be given an appropriate phonetic spellout) 134.
133
A technical complication which I gloss over here is that Chan claims that functional heads impose both c- and
s-selection restriciotns on their complements, whereas lexical heads impose only s-selection restrictions.
134
Chomsky (2005b) maintains that in order for derivations to be convergent at the conceptual-intentional
interface (C-I), uninterpretable grammatical features are deleted at the end of each phase as part of the transfer
operation which input syntactic structures into the semantic component. Undeleted uninterpretable features
would cause a crash at the C-I interface because they cannot be assigned any semantic interpretation. I do not
 A key assumption made by Chomsky is that some constituents enter the syntax already valued for
particular features, whereas others are initially unvalued for certain features, with unvalued features
being assigned a value in the syntax via agreement with constituents carrying matching valued
features. We can make our discussion more concrete by considering how Concord (i.e. number and
gender agreement) works in a French nominal expression like une grande tasse ‘a big cup’. I assume
that a French noun like TASSE ‘cup’ enters the syntax carrying a feminine-gender feature [F-Gen], but
an initially unvalued number feature [u-Num]: this reflects the fact that TASSE is inherently feminine in
gender, but can be either singular or plural in number 135. Conversely, I assume that the indefinite article
UN ‘a’ enters the derivation with an already-valued singular-number feature [Sg-Num], but an
unvalued gender feature [u-Gen], reflecting the fact that UN is inherently singular in number but can be
either masculine or feminine in gender. Finally, I assume that an adjective like GRAND ‘big’ enters the
derivation with an initially unvalued number feature [u-Num] and an unvalued gender feature [u-Gen],
because adjectives can be either singular or plural, and either masculine or feminine. This means that
in a nominal like une grande tasse ‘a big cup’, the relevant items enter the syntax carrying the C-
features (i.e. number/gender Concord features) shown below (with lexical items CAPITALISED and
shown in their dictionary citation form):
(28) UNa GRANDbig TASSEcup
[Sg-Num] [u-Num] [u-Num]
[u-Gen] [u-Gen] [F-Gen]
The unvalued number features on the adjective and noun will subsequently be valued as singular in
number by agreement with the singular-number feature on the article, and the unvalued gender
features on the determiner and adjective will subsequently be valued as feminine in gender via
agreement with the feminine-gender feature on the noun 136, with the outcome shown below:
(29) UNa GRANDbig TASSEcup
[Sg-Num] [Sg-Num] [Sg-Num]
[F-Gen] [F-Gen] [F-Gen]
The resulting structure will be convergent at the PF interface, since all the relevant features have been
valued, and so each of the words is spelled out as a feminine singular form, yielding the string une
grande tasse ‘a big cup’.
Now consider the derivation of the corresponding English nominal a big cup. In keeping with the
claim by Chomsky (1999, p.2) that parametric variation is restricted to inflectional properties which
are ‘easily detectable in data available for language acquisition’, I shall assume that one typological
difference between English and French relates to the C-features carried by relevant types of
constituent. I shall make the traditional assumption that C-features in English include number but not
gender, so that both determiners and nouns carry number in English (easily detectable in contrasts like
this book/these books) but not gender (since there are no nouns, adjectives or determiners which inflect
for gender in English, making gender undetectable on the relevant items). On the grounds that they
inflect for neither number nor gender, I shall also assume that adjectives are numberless and
genderless in present-day English. Given these assumptions, when the items A, BIG and CUP are

discuss the mechanics of feature-deletion here, since nothing discussed here is contingent on adopting any
particular account of feature-deletion.
135
Chomsky (1999) posits that inherent features (i.e. those which are already valued when they enter the
derivation) are interpretable; however, the intrinsic (but arbitrary) feminine gender feature on a French noun like
tasse ‘cup’ seems to be uninterpretable, so calling the generality of Chomsky’s observation into question.
136
For ease of exposition, I set aside a number of technical questions here, such as whether Concord is an
independent operation from Agreement (which in part depends on whether DPs are phases as argued in
Svenonius 2003 and Adger and Ramchand 2004), and whether Number Concord and Gender Concord are
independent operations (as suggested in psycholinguistic research by Antón-Méndez, Nicol and Garrett 2002). I
also set aside the question of what abstract case feature (if any) each of the constituents carries, and how this is
valued. Finally, I also set aside the proposal by Picallo (1991), Ritter (1991) and Cinque (1994) that nouns in
French raise to an intermediate functional head position (= Num) between D and N.
combined by successive merger operations into a nominal structure, they will initially carry the C-
features shown below:

(30) A BIG CUP

[Sg-Num] [u-Num]
The unvalued number feature on the noun will be valued as singular via agreement with the singular
article, yielding the outcome shown below:
(31) A BIG CUP
[Sg-Num] [Sg-Num]
Since all the relevant features have been valued, the structure converges at the PF interface and the
resulting string is ultimately spelled out as a big cup.
Given the assumptions made above about the different C-features carried by relevant constituents
in English and French, consider now what the expected outcome would be if an English-French
bilingual used an English article like A and an English adjective like BIG to modify a French noun like
TASSE ‘cup’. On the assumptions made here, the relevant items would enter the syntax carrying the C-
features shown below:
(32) A BIG TASSEcup
[Sg-Num] [u-Num]
[F-Gen]
The unvalued number feature on the noun tasse would be valued by agreement with the singular-
number feature on the article, with the outcome shown below:
(33) A BIG TASSEcup
[Sg-Num] [Sg-Num]
[F-Gen]
Consequently, the article will be spelled out as the singular form a, the adjective will be spelled out in
the default form big, and the noun in the feminine singular form tasse. Since each item can be given an
appropriate spellout (i.e. can be spelled out as an appropriate Phonetic Form), the derivation converges
at PF (i.e. yields a successful PF outcome).
In empirical terms, what our discussion means is that we’d expect a code-mixing bilingual to be
able to use an English genderless determiner or adjective to modify a gendered noun in another
language. Interestingly, structures containing an English genderless determiner modifying a gendered
noun in another language are frequently reported to be produced by bilingual children, as illustrated by
the examples below (where subscripts are used to indicate the gender of the relevant nouns):
(34)(a) the vacaF ‘cow’, the aguaF ‘water’; a perroM ‘dog’ (Spanish: Lindholm & Padilla 1978)
(b) the dukke ‘doll’; a bog ‘book’; a flaske ‘bottle’ (Danish: Petersen 1988)
(c) I can’t give you any KussM [= ‘kiss’] because I have a SchmutznaseF [= ‘dirty nose’]
(German: Leopold 1970)
Moreover, as the expression example pechos flat ‘chests flat’ in example (11a) above illustrates, we
also find cases of a genderless and numberless English adjective modifying a gender- and number-
marked noun in another language. More generally, the convergence account of code-mixing outlined
here would lead us to expect that code-switching is possible except where it leads to one or more
features remaining unvalued.

8.6 Resolving potential feature mismatches under the convergence account

In the previous section, we saw that mixing is possible in cases where unvalued features on
a constituent in one language can be valued by corresponding valued features on a constituent in
another language. In this section, we look at what happens when mixing gives rise to a potential
feature-mismatch in which a constituent in one language has unvalued features which have no valued
counterpart in the other language. We can illustrate this problem by considering what might happen in
a situation where a bilingual speaker uses a French article like UN ‘a’ and a French adjective like
GRAND ‘big’ to modify an English noun like CUP. On the assumptions made in the previous section,
the relevant items would enter the derivation carrying the C-features shown below:
(35) UNa GRANDbig CUP
[Sg-Num] [u-Num] [u-Num]
[u-Gen] [u-Gen]
The unvalued number features on the adjective and noun can be valued as singular via agreement with
the singular-number feature on the article, so that we would expect number concord to be correct (and
unsurprisingly, Lindholm and Padilla 1978 report correct number agreement in children’s code-mixed
nominals containing a Spanish article or demonstrative and an English noun). However, the unvalued
gender features on the determiner and adjective cannot be valued by the genderless English noun cup.
Unless something is done to repair the potential feature-mismatch here, the gender features on the
determiner and adjective will remain unvalued. If we suppose that unvalued features are illegible at the
PF interface (i.e. that the PF component simply cannot process unvalued features), then we would
expect that such a derivation would not be convergent at the PF interface and so would automatically
crash (i.e. fail). This would lead to the prediction that bilinguals will never use a gendered determiner
or adjective in one language to modify a genderless noun in another. In this context, it is interesting to
note that Petersen (1988) reports that the Danish-English bilingual child she studied (Thea) used
English determiners to modify Danish nouns (as in that mad ‘that food’) but never used Danish
determiners to modify English nouns (hence did not produce structures like dette house ‘that house’).
Since Danish determiners (like their French counterparts) carry gender, it seems reasonable to suppose
that the absence of gender on English nouns means that the gender feature on Danish determiners
cannot be valued by genderless English nouns, and that this is why the child did not produce such
structures.
However, an alternative possibility is that where bilinguals use a gendered modifier (e.g. a
determiner or adjective) to modify a genderless noun, they will assign the modifier the default gender.
For example, suppose that gender concord works along the lines suggested informally below in
French:
(36) Gender Concord Condition
A modifier is assigned feminine gender if modifying a feminine noun, but masculine gender
otherwise
A condition like (36) would account for the pattern of Concord which we find in (37) below:
(37)(a) uneFSG filleFSG intelligenteFSG ‘a girl intelligent’ (= an intelligent girl)
(b) unMSG garçonMSG intelligentMSG ‘a boy intelligent’ (= ‘an intelligent man’)
(c) unMSG garcon MSG et uneFSG filleFSG intelligentsMPL ‘a boy and a girl intelligent’ (= an
intelligent boy and girl)
This, we find the feminine form of the adjective INTELLIGENT used to modify a feminine noun in (37a)
above (-e being an orthographic marker of feminine gender in these examples), and the masculine
form used to modify either a masculine noun as in (37b) or a mixed-gender expression as in (37c) –
the use of a masculine form to modify a mixed-gender expression coming about because masculine is
the default (‘otherwise’) gender in French. On this account, what we might expect a French-English
bilingual to do in a structure like (35) is assign the determiner and adjective the default (masculine)
gender, and so say un grand cup (with un ‘a’ and grand ‘big’ both being masculine singular forms).
The default gender analysis outlined here could account for why Pfaff (1979, p.305) reports that the
bilinguals in her study used masculine modifiers in Spanish to modify English nouns in structures like
unos traditions ‘some traditions’ (where unos is a masculine plural quantifier).
However, it would seem that some bilingual speakers of languages like Spanish where nouns carry
grammatical gender assign gender to English nouns in order to ensure that a gender feature on an
article used to modify an English noun can nonetheless be assigned a value. To understand how this
works, let us briefly look at how nouns are assigned gender in a language like Spanish. On the grounds
of economy (i.e. eliminating redundancy in order to reduce the memory burden placed on language
learners), we don’t want to list the gender of every noun in the language in the Lexicon if the gender of
most nouns is predictable from the meaning or form of the noun. Instead, lexical entries for words will
specify the gender of irregular nouns with unpredictable gender, but regular nouns with preditable
gender will be unspecified for gender in the Lexicon and will be assigned gender by Gender
Assignment Conditions such as those sketched informally in (38ii, iii) below:
(38) Gender Assignment Conditions
A noun carries the gender specified in its lexical entry if it has unpredictable gender, but if
unspecified for gender in the lexicon will be assigned:
(i) masculine gender if it denotes a male being137, feminine if denoting a female being
(ii) masculine gender if ending in -o138, feminine if ending in -a
(iii) masculine gender otherwise (by default)
We can illustrate how gender assignment in Spanish works in the following terms. A noun like mujer
‘woman’ will be unspecified for gender in its lexical entry, and will be assigned feminine gender by
virtue of its meaning (i.e. by virtue of denoting a female) in accordance with the gender assignment
condition (38i). In the same way, a noun like hombre ‘man’ will be unspecified for gender in its lexical
entry, and again be assigned masculine gender in virtue of its meaning by (38i). The noun libro ‘book’
will also be unspecified for gender in its lexical entry, and (by virtue of its form – i.e. the fact that it
ends in -o) will be assigned masculine gender by (38ii). Similarly, the noun tienda ‘shop’ will be
unspecified for gender in its lexical entry, and will be assigned feminine gender by virtue of ending in
-a via (38ii). In contrast, the noun catedral ‘cathedral’ has unpredictable gender, and so has to be
marked in its lexical entry as feminine; although the irregular masculine noun hospital also has
unpredictable gender, it can be left unspecified for gender in the lexicon and will then be assigned
masculine gender by default via condition (38iii)139. The irregular noun mano ‘hand will be listed in
the lexicon as feminine it does not have the masculine gender of regular nouns ending in -o; and
likewise, the irregular gender noun día ‘day’ will be listed in the lexicon as masculine in gender
because it does not have the feminine gender of regular nouns ending in -a.
The assumption that nouns which are assigned no gender in their lexical entry can nonetheless be
assigned gender via Gender Assignment Conditions like those in (38) has interesting implications for
how bilinguals treat nouns in English, since English nouns do not carry grammatical gender (as argued
by Namai 2000). In particular, it could be that (e.g.) Spanish-English bilinguals assign English nouns
gender by applying the Case Assignment Conditions which operate in Spanish to English nouns. This
is what Carol Pfaff (1979, p.305) suggests in relation to a study of English-Spanish code-switching in
adult Mexican Americans. For example, she notes that her subject treated English nouns which denote
males as masculine (saying e.g. el trainer ‘the trainer’ and un boxer ‘a boxer’ where el/un are
masculine singular forms), and English nouns which denote females as feminine (saying e.g. la maid
‘the maid’ and esa girl ‘that girl’, where la/esa are feminine singular forms). However, Pfaff also
noted that an alternative strategy used by her subjects for assigning gender to English nouns was a
transfer-based gender assignment condition, which we can outline informally as follows:
(39) Gender Transfer Condition
Assign an English noun the same gender as a Spanish counterpart which resembles it
(i) in meaning (i.e. which is synonymous with it), or
(ii) in form (e.g. which carries a similar ending)
So, for example, she reported her subjects treating an English noun like coast as feminine (and hence
saying la coast ‘the coast’, where la ‘the’ is a feminine singular article) because its Spanish
counterpart costa is feminine. Similarly, she reported her bilingual subjects treating nouns like liability
as feminine (hence modifying it by the feminine article la ‘the’ in a structure like la liability), and

137
I set aside here the important descriptive detail of what is meant by the word being in this context (e.g.
whether it refers to human beings, or humans and certain types of animals, etc.)
138
That is, if its default (singular) form ends in -o.
139
We could arguably simplify conditions (i) and (ii) still further by simply saying that regular nouns denoting
females or ending in -a are feminine in gender, because it would then follow condition (iii) that nouns denoting
males or ending in -o are assigned default masculine gender. We set aside the issue of the optimal formulation of
case assignment conditions here, since this is not our concern.
claimed that this was because they equate the English derivational suffix -ity with its feminine Spanish
counterpart -idad (found in feminine nouns such as comunidad ‘community’).
The observation by Lindholm and Padilla (1978, p.331) that in some cases, the Spanish-English
bilingual children in their study (aged from 2;10 to 6;2) assigned English nouns gender ‘in accordance
with their natural gender (e.g. la lady, un boy)’ suggests that bilingual children may use Spanish
gender assignment conditions like (38) for assigning gender to English nouns 140. However, the data
reported in their study also show that (with other types of noun) the children alternated between using
masculine and feminine Spanish determiners to modify genderless English nouns. So, for example,
they report the children using both the masculine singular Spanish determiner este ‘this’ and the
feminine singular indefinite article una ‘a’ to modify the noun bird (cf. este bird/una bird). It is
conceivable that uncertainly about the sex of the bird leads children to alternate between treating it as
masculine or feminine in gender in accordance with (38i).
However, such an account would not generalise to cases where children altermate betweeing
masculine and feminine modifiers with sexless nouns. For example, Lindholm and Padilla (1978)
report Spanish-English bilingual children using both the masculine plural determiner los ‘the’ and the
feminine singular determiner esa ‘that’ to modify the noun window (cf. los windows/esa window).
Likewise, Cornejo (1975) reports Spanish-English bilingual children using both the masculine singular
article un ‘a’ and the feminine singular article la ‘the’ to modify the noun bottle (cf. un bottle/la
bottle). What may be going on in such cases is that the child alternates between using the transfer
condition (39) and treating the nouns window and bottle as feminine because their Spanish
counterparts ventana and botella are feminine, or assigning the nouns masculine gender via the default
condition (38iii)141.
Still, even this account would not work for a case like that reported by Cornejo (1975) where
bilingual children use both the Spanish masculine singular article un ‘a’ and its feminine singular
counterpart una to modify the noun book (cf. un book/una book): while the use of a masculine article
un could be argued to reflect default gender assignment, the use of the feminine form una would
appear not to be the result of transfer, since its Spanish counterpart libro is masculine in gender. What
is going on here?
It might seem as it one possibility would be to assume that the children simply assign gender to
English nouns which don’t have ‘natural’ gender on a random basis. But this raises the obvious
problem that the proposal to assign random feature-values to items is otherwise unprecedented, and
hence would seem to be unprincipled. A more principled way of dealing with the dual behaviour of
nouns like window is to assume that they remain genderless throughout the derivation, and that any
Spanish article modifying them will consequently remain unvalued, so resulting in a structure such as
the following (once the noun has undergone number concord with the plural article):
(40) UNA BOOK
[Sg-Num] [Sg-Num]
[u-Gen]
Singular forms of the indefinite article in Spanish are spelled out as follows:
(41)(i) una if feminine singular (ii) un if masculine singular
Let us further suppose that a given spellout form can be used to spell out a given syntactic target form
if the two match in features. If we follow Chomsky (1999, p.4) in assuming that ‘Match is not strictly
speaking identity but nondistinctness’, we can suppose that a target form and a spellout form match in
respect of a given set of features unless they have distinct values for one or more of the features (e.g.
one is singular and the other plural, or one is masculine and the other feminine). Given this
assumption, let us consider whether the spellout form una in (41i) matches the target form of the
article required in (40). The two carry the same singular-number feature, and so clearly match in
number. The spellout form una in (41i) also carries a feminine-gender feature, while the target form
140
The fact that in sentence (22), Danny uses the German feminine determiner die ‘the’ to modify the irregular
German neuter noun Mädchen ‘girl’ suggests that he uses a condition like (38i) to assign the noun feminine
gender.
141
A further possibility would be that they treat the nouns as genderless, and assign their modifiers default
masculine gender via condition (36).
UN in (40) carries an unvalued-gender feature: however, since these two do not have distinct gender
values (i.e. it is not the case that one is masculine and the other feminine), they match in gender (in the
sense of having non-distinct gender values). Thus, the spellout form una in (41i) matches the target
form UN in (40) in both number and gender, and so is a possible spellout for the article in (40),
resulting in una book. But by parity of reasoning, the masculine singular spellout form un in (41ii)
provides an equally optimal match in number and gender with the target form UN in (40), so that the
structure in (40) could also be spelled out as un book. On this view, the crucial factor in code-mixing is
whether there are one or more spellout forms which provide an optimal match for the target form 142.
In this section and the last, I have outlined an alternative convergence (feature-matching) account
of code-mixing which posits that code-mixing is permitted only if it leads to a convergent structure.
Although the supporting evidence has come from code-mixing internally within nominal structures,
such an approach can in principle be extended to other types of structure – for example to code-mixing
in clauses such as the following (where Italian words are italicised):
(42) Io voglio hold the trapano (Lucy, 3;2)
I want hold the drill (‘I want to hold the drill’)
It is a property of the verb VOLERE ‘want’ in Italian that it selects an infinitive complement. This
selectional requirement is met in the code-mixed structure in (42) by virtue of the fact that the verb
hold is in the infinitive form – resulting in a PF-convergent structure (i.e. one which enables all the
relevant items to be given an appropriate phonetic spellout).
An important methodological point which needs to be made in relation to the convergence account
is the following. At first sight, there would appear to be cases of child code-mixing which seemingly
violate the convergence (feature-maching) requirement. For example, Caroline Koehn (1994) reports
that a French-German bilingual boy called Ivar produced nominal expressions such as the following:
(43) dasneuter bateaumasculine [Ivar 2;0]
thatGerman boatFrench
Since das ‘that’ is neuter in gender in German, and bateau ‘boat’ is masculine in gender in French,
there would at first sight appear to be a gender-mismatch here, and the feature-matching account of
code-switching might seem to be falsified by examples like (38). But Koehn (1994, p.38) notes that
until the age of 2;5, ‘no gender distinctions are made’ by Ivar. If he has not yet acquired the functional
feature gender and his nouns and determiners are underspecified for gender at this stage, there will
self-evidently be no gender-mismatch between determiner and noun in (38), and hence no empirical
challenge to the convergence account. If determiners and nouns carry number but not gender in Ivar’s
grammar at the relevant stage, then the fact that both are singular in (38) would suggest that the
nominal is convergent143. The more general point being made here is that claims about whether child
code-mixing in a particular utterance does (or does not) result in a convergent structure have to be
evaluated in terms of the features carried by the relevant items in the child’s grammar – not by the
features carried by the corresponding items in the adult grammar.

142
A key theoretical assumption being made here is that a feature which remains unvalued at the end of the
syntactic derivation does not automatically cause the derivation to crash in the PF component: rather, a crash
only occurs when it leads to spellout failure – i.e. when the PF component has no form which can spell out the
features on the target form.
143
Redlinger and Park (1980) report a French-German bilingual child (Henrik) using the German masculine
article der ‘the’ to modify the French feminine noun mouche ‘fly’, and the German feminine article die to modify
the French masculine nouns livre ‘book’ and bateau ‘boat’. This may be a similar case – though we obviously
cannot draw any conclusions from a handful of examples.
 Workbook section
8.1 Testing the Lexical Gap account of why children code-switch
The following material is designed to get you to test the Lexical Gap Hypothesis (that children only
switch language when they lack a word for a given concept in a particular language). The data given
below are from Deuchar and Quay (1999). They report on a longitudinal study of a one-year old
Spanish-English bilingual, Manuela. They list (p.473) all the two-word utterances produced by
Manuela at ages 1;7 and 1;8, and below are all the code-mixed utterances she produced at the relevant
stage (with a subscript S being used to mark Spanish words):
(1) MásS banana ‘More banana’ [1;7.26] (2) More galletaS ‘More biscuit’ [1;7.26]
(3) Oh-dear camaS ‘Oh-dear bed’ [1;7.29] (4) Dos S ball ‘Two ball’ [1;8.12]
(5) Daddy dosS ‘Daddy two’ [1;8.16] (6) MásS paper ‘More paper’ 1;8.16
The list of Manuela’s translation equivalents given in Quay (1995) suggests that (at the time the
utterances were produced) she had no translation equivalent for banana, galleta, oh-dear, cama, dos,
ball, or paper, though she did have equivalents/synonyms for más/more and for daddy/papá. The
question which you should ask yourself is: Are the relevant data consistent with Lexical Gap account
or not? To what extent does this depend on whether we take a top-down (right-to-left) or bottom-up
(left-to-right) view of how sentence structures are formed? 144

8.2 English-Italian code-mixing by Lucy

The following code-mixed utterances (kindly provided by Gabriele Azzaro) were produced by a
bilingual English-Italian girl called Lucy between ages 2;4 and 3;4 (English being her dominant
language): English glosses are provided in single inverted commas. Discuss the extent to which they
are compatible with a number of different accounts of code-mixing.
1. Make the bua better (‘Make the sore better’, in reply to ‘What did the doctor do?’)
2. Eating a bone outside, eating a bono (reply to ‘What’s the doggy doing?’)
3. I want the ciuccio (‘I want the dummy/pacifier’)
4. In the acqua (‘In the water’)
5. And then the mucca said “Moo” (‘And then the cow said “Moo”’)
6. I want to go to the muccas (‘I want to go to the cows’)
7. Io ho fatto lo shopping (‘I have done the shopping’)
8. Questo è un gooso (‘This is a goose’)
9. Io vo(gl)io fare la puzzle (‘I want to do the puzzle’)
10. È un puzzle piccolo (‘Is a puzzle tiny’ = ‘It’s a tiny puzzle’)
11. Con la butterfly (‘With the butterfly’)
12. Come questa butterflaia (‘Like this butterfly’)
13. La lion non vuol far la cacca ‘The lion not wants do the poo’ (= ‘The lion doesn’t want to poo’)
14. On Lucy’s testa (‘On Lucy’s head’)
15. I’m doing cavallino (‘I’m doing horsey’)
16. Io faccio drawings? (‘I do drawings?’ = ‘Shall I do drawings?’)
17. No, io stay (‘No, I stay’, in reply to “Vieni fuori?” ‘Are you coming outside?’)
18. Questo è brown, questo è white (‘This is brown, this is white’)
19. Questo is pink (‘This is pink’)
20. It’s rotto (‘It’s broken’)

144
Chomsky takes a bottom-up view of syntax in which sentence structures are built up one layer at a time, from
bottom to top (with lower layers of structure being formed before higher layers). An alternative top-down model
is presented in Phillips (2003) under which sentence structures are built up from top to bottom (with higher tiers
of structure being formed before lower tiers). Under the top-down model, a quantifier phrase like more galleta
‘more biscuit’ would be formed by (i) first selecting the quantifier more, (ii) then selecting the noun galleta, and
(iii) then merging the quantifier with the noun to form the quantifier phrase more galleta. Under the bottom-up
model, the phrase would be formed by (i) first selecting the noun galleta, (ii) then selecting the the quantifier
more, and (iii) then merging the two to form the quantifier phrase more galleta.
Note that -o is a masculine singular affix on Italian nouns/pronouns/articles/adjectives/participles, and
-a a feminine singular affix, -i a masculine plural affix, and -e a feminine plural one (so that e.g. the
plural of mucca ‘cow’ is mucche ‘cows’). Note also that un ‘a’ is a masculine singular indefinite article
(its feminine counterpart being una). The Italian counterparts of key English nouns are listed below,
with their masculine/feminine gender indicated by a subscript M/F:
bone = ossoM; goose = ocaF; puzzle = ; butterfly = farfallaF; lion = leoneM; puzzle = puzzleM
Note that in possessive structures where English says ‘Gianni’s car’, Italian says (the equivalent of)
‘the car of Gianni’.

8.3 English-French code-mixing by Nicholas

Joseph Galasso (1999) reports his English-French bilingual son Nicholas (whose dominant language is
English) producing code-mixed structures such as the following (where French words are italicised).
Discuss the extent to which they are compatible with a number of different accounts of code-mixing
discussed in the main text.
1. Baby pousseD me ‘Baby push me’
2. MoucheF want out ‘Fly want out’
3. I want glaceF ‘I want ice-cream’
4. Me want bonbonM ‘Me want sweety/candy’
5. I want mangeD ‘I want eat’
6. I got a bonbonM for you ‘I’ve got a sweety/candy for you’
7. Juice in a pailleF ‘Juice in a straw’
8. All wet the chaussureF ‘All wet the sock145’
9. A big bateauM ‘A big boat’
10. Where Daddy chaussureF?
11. I want me chaussureF ‘I want my shoe’
12. MaF car ‘My car’
13. I got beaucoup money ‘I got much money’
14. I want encore juice ‘I want more juice’
15. I want encore bonbon ‘I want more sweety/candy’
16. I want run encore ‘I want run again’
Note that a subscript M/F indicates that the relevant French words are masculine/feminine in gender
(and are also singular in number). In relation to 12, you might note that possessives like ma ‘my’ agree
in person and number with the possessum (i.e. with the possessed noun following them) in French, so
that we find the masculine singular form mon in mon père ‘my father’ and the feminine singular form
ma in ma mère ‘my mother’: you might also bear in mind that the French counterparts of the English
word car are the feminine nouns voiture/auto. A subscript D indicates that the relevant French verb-
form is the default form of the item for Nicholas (by virtue of being the only form of the verb which he
uses at the relevant stage, albeit distinct from the adult infinitive forms pousser/manger). In relation to
13, note that beaucoup ‘much’ in adult French (rather like lots in English) requires a complement
headed by de ‘of’ (as in beaucoup de fromage ‘lots of cheese’). In relation to 14-16, note that encore is
an adverb corresponding to English ‘again’ and can be used to modify an indefinite article or quantifier
(so that in a context where an English waiter might ask ‘More coffee?’, his French counterpart would
say ‘Encore un café?’ (literally ‘Again a coffee’) or ‘Encore du café?’ (literally ‘Again some coffee?’).

8.4 Other examples of child code-mixing

Below are listed examples of sentences produced by young children in which they code-mix between
145
Chaussure in adult French means ‘shoe’, but Nicholas here uses it to refer to a sock.
English and another language (the names and ages – where known – of the children being specified in
parentheses). Say whether (and if so how) each of the examples is consistent with a number of
different accounts of code-mixing. Subscripts mark non-English words (F = French, G = German,
I = Italian, N = Norwegian).
(1)
KlappeN hand ‘Clap hand’ (Siri)
(2)
DenN full ‘It (is) full’ (Siri)
(3)
OgN nyN diaper ‘And new diaper’ (Siri)
DerG pushtG derG kleineG JosefG ‘He pushes the little Josef’ (Danny 2;2)146
(4)
(5)
EinG big cow ‘A big cow’ (Danny)
(6)
MehrG books ‘More books’ (Danny)
(7)
And the froggy’s getting nassG ‘And the froggy’s getting wet’ (Danny)
(8)
Daddy’s red AutoG fallen ‘Daddy’s red car [has] fallen’ (Danny)
(9)
IchG can’t see it ‘I can’t see it’ (Danny)
(10)
DerG monkey willG beissenG ‘The monkey wants to bite’ (Henrik)
(11)
KannstG duG move a bit? ‘Can you move a bit?’ (Hannah 2;4)
(12)
IchG habG geGclimbed up ‘I have ge-climbed up’ (Hannah 2;4)
(13)
DieG mamaG helfG mirG strap it in ‘The mummy help me strap it in’ (Hannah 2;3)
(14)
You mettezF honey? ‘You put honey?’ (Gene 3;1)
(15)
He aF eyes ‘He has eyes’ (Gene 3;1)
(16)
I peuxF pasF wash the couF, me (Gene 3;1)
I can not wash the neck, me (‘I can’t wash my neck, me’)
(17) I ask him quéS yoS voyS aS casaS ‘I ask him that I go to home’ (unnamed
child)
(18) Shoté ese ‘Shot that’ (= ‘I shot that’: -é is a Spanish 1st person singular
past tense affix)
(19) HizoS unaS birthday ‘(He) had a birthday’ (una is feminine singular)
Examples (1-3) are from Lanza (1992), (4-10) from Redlinger and Park (1980), (11-12) from
Gawlitzek-Maiwald and Tracy (1994), (14-15) from Genesee and Paradis (1996), and (17-19) from
Lindholm and Padilla (1978).

146
In pusht, the German 3.SG present-tense affix -t appears to have been added to the English verb-stem push.
 9 References
Abels K (2001) ‘The predicate cleft construction in Russian’, in S. Franks (ed.) Proceedings of FASL 9
Abney SP (1987) The English Noun Phase in its Sentential Aspect, unpublished PhD diss., MIT
Abu-Akel A & Bailey AL (2000) ‘Acquisition of use of a and the in English by young children’, in SC
Howell et al (eds) BUCLD 24 Proceedings, Somerville Mass, pp. 45-57
Adger D (2003) Core Syntax, Oxford University Press.
Adone MCD (1993) ‘IP and its Development in Mauritian Creole’, draft manuscript,
University of
Hamburg
Akmajian, A. & Heny, F. (1975) An Introduction to the Principles of
Transformational Syntax, MIT
Press, Cambridge Mass.
Aldridge M (1989) The Acquisition of INFL, Indiana University Linguistics Club
Publications.
Aldridge M, Borsley RD & Clack S (1995) ‘The Acquisition of Welsh Clause
Structure’ in D
MacLaughlin and S McEwen (eds) Proceedings of the 19th Annual Boston
University Conference
on Language Development, Cascadilla Press, Somerville Mass., pp.37-47.
Anderrsen M (1996) The Acquisition of Functional Categories, Tromsø Studies in Linguistics 17,
Novus,Oslo
Antony, L. and Hornstein, N. (2002) Chomsky and His Critics, Blackwell, Oxford.
Atkinson M (1992) Children’s Syntax, Blackwell, Oxford.
Baker CL (1970) ‘Notes on the description of English questions: The role of an abstract question
morpheme’, Foundations of Language 6: 197-219.
Belazi H, Rubin EJ and Toribio JA (1994) ‘Code switching and X-bar theory: The Functional Head
Constraint’, Linguistic Inquiry 25: 221-237.
Bentahila A & Davies EE (1983) ‘The syntax of Arabic-French code-switching’, Lingua 59: 301-330.
Bentahila A & Davies E (1992) ‘Code-switching and dominance’ in RJ Harris (ed) Cognitive
Processes in Bilinguals, Elsevier, Amsterdam, pp. 443-458.
Bergman CR (1976) ‘Interference vs independent development in infant bilingualism’, in Keller G
(ed) Bilingualism in the Bicentennial and Beyond, Bilimngual Press, New York.
Berg-Seligson S (1986) ‘Linguistic constraints on intrasentential code-switching: A study of
Spanish/Hebrew bilingualism’, Language in Society 15: 313-48.
Bhatia TK & Ritchie WC (1999) ‘The bilingual child: Some issues and perspectives’ in WC Ritchie
& TK Bhatia (eds) Handbook of Child Language Acquisition, Academic Press pp.569-643.
Bloom L (1970) Language Development, MIT Press, Cambridge Mass.
Bloom L et al. (1978) ‘Pronominal-nominal variation in child language’, in L. Bloom (ed) Readings
in Language Development, Wiley, New York, pp. 231-238
Bloom P (1990) ‘Subjectless sentences in child language’, Linguistic Inquiry 21: 491-504.
Bokamba, E. G. (1989) ‘Are there syntactic constraints on code-mixing?’ World Englishes 8: 277–292.
Borer H & Rohrbacher B (2004) ‘Minding the Absent: Arguments for the Full
Competence
Hypothesis’, paper to appear in Language Acquisition.
Bowerman M (1988) ‘The “no negative evidence” problem: How do children avoid
an overly general
grammar?’, in J Hawkins (ed.) Explaining Language Universals, Blackwell,
Oxford, pp.73-101.
Boser K, Lust B, Santelmann L & Whitman J (1992) ‘The syntax of CP and V2 in
Early Child
German: The strong continuity hypothesis’, Proceedings of NELS 22: 51-65.
Bošković Z (2001) On the Nature of the Syntax-Phonology Interface, Elsevier
Bošković Z (2002) ‘On multiple wh-fronting’, Linguistic Inquiry 33: 351-383.
Boumans L (2001) ‘Searching for syntactic explanations of code switching’,
Linguistics 39: 437-453.
Braine MDS (1971) ‘Three suggestions regarding grammatical analyses of
children’s language’ in
CA Ferguson & DI Slobin (eds) Studies of Child Language Development, Holt
Rinehart and
Winston, New York, pp.421-429.
Bromberg H & Wexler K (1995) ‘Null subjects in child wh-questions’, MIT Working Papers in
Linguistics 26: 221-247.
Brown R, Cazden C & Bellugi U (1968) ‘The child’s grammar from I to III’, in JP Hill
(ed.)
Minnesota Symposium on Child Development vol 2, pp. 28-73.

Brown R & Hanlon C (1970) ‘Derivational complexity and order of acquisition in

child speech’, in
JR Hayes (ed.) Cognition and the Development of Language, Wiley, New York,
pp. 11-53.
Budwig N (1995) A Developental-Functionalist Approach to Child Language,
Erlbaum, Mahwah NJ.
Cazden C B (1968) ‘The acquisition of noun and verb inflections’, Child
Development 39: 433-48.
Chan B (1999) Aspects of the syntax, production and pragmatics of code-
switching with special
reference to Cantonese-English, PhD dissertation, University College London.
Cheng, L. (1997) On the Typology of Wh-Questions, Garland, New York.
Cheng L & Rooryck J (2000) ‘Licensing Wh-in-situ’, Syntax 3: 1-19.
Choi J (1991) ‘Korean-English code-switching: Switch-alpha and linguistic constraints’, Linguistics
29: 877–902.
Chomsky, N. (1965) Aspects of the Theory of Syntax, MIT Press, Cambridge Mass.
Chomsky, N. (1968) Interview with S. Hamshire in The Listener, May 1968.
Chomsky, N. (1972) Language and Mind (enlarged edition), Harcourt Brace
Jovanovich, New York.
Chomsky N (1981) Lectures on Government and Binding, Foris, Dordrecht.
Chomsky N (1986), Barriers, MIT Press, Cambridge Mass.
Chomsky N (1989) ‘Some notes on Economy of Derivation and Representation’,
MIT Working Papers
in Linguistics 10: 43-74.
Chomsky N (1995) The Minimalist Program, MIT Press, Cambridge Mass.
Chomsky N (1998) Minimalist Inquiries: The Framework, MIT Occasional Papers in
Linguistics, no.
15 (also published in R Martin, D Michaels and J Uriagereka (eds) Step by Step:
Essays on
Minimalism in Honor of Howard Lasnik, MIT Press, Cambridge Mass., pp. 89-
155).
Chomsky N (1999) Derivation by Phase, MIT Occasional Papers in Linguistics, no.
18 (also
in M Kenstowicz (ed) (2001) Ken Hale: A Life in Language, MIT Press, Cambridge
Mass., pp.1-52).
Chomsky N (2001) Beyond Explanatory Adequacy, unpublished manuscript, MIT.
Chomsky N (2002) On Nature and Language, Cambridge University Press
Chomsky, N. (2005a) ‘Three factors in language design’, Linguistic Inquiry 36: 1-
22.
Chomsky, N.(2005b) ‘On Phases’, unpublished paper, MIT (to appear in C. P. Otero
et. al. eds.,
 Foundational Issues in Linguistic Theory, MIT Press, Cambridge Mass.).
Cinque G (1988) ‘Parameter-setting in "instantaneous" and "real-time"
acquisition’, The Behavioural
and Brain Sciences 12: 336-337.
Cinque G (1994) ‘Evidence for partial movement in the Romance DP’ in G Cinque et al (eds) Towards
Universal Grammar, Georgetown University Press, Washington.
Cinque G (1999) Adverbs and Functional Projections, Oxford University Press.
Clahsen H, Penke M & Parodi T (1994) ‘Functional Categories in Early Child
German’, Language
Acquisition 3.395-429.
Clyne M (1987) ‘Constraints on code-switching: How universal are they?’,
Linguistics 25: 739-64.
Cornejo R (1975) ‘The acquisition of the lexicon in the speech of bilingual
children’, in P. Turner (ed)
Bilingualism in the Southwest, University of Arizona Press, Tucson, Arizona.
Crain, S. & Pietroski, P. (2002) ‘Why language acquisition is a snap’, The
Linguistic Review 19: 163-
183.
Crisma P (1992) ‘On the acquisition of wh-questions in French’, Geneva Generative Papers vol. 0,
pp.115-22.
Curtiss, S. (1977) Genie: A Psycholinguistic Study of a Modern Day “Wild Child”,
Academic Press,
London.
de Houwer A (1990) The Acquisition of Two Languages from Birth, Cambridge University Press.
de Houver A (1995) ‘Bilingual Acquisition’ in P Fletcher & B MacWhinney (eds) Handbook of Child
Language, Blackwell, Oxford, pp.219-250
Denham K (2000) ‘Optional wh-movement in Babine-Witsuwit’en’, Natural
Language and
Linguistic Theory 18: 199-251.
Déprez V & Pierce A (1993) ‘Negation and functional projections in early grammar’, Linguistic
Inquiry 24: 25-67.
Déprez V & Pierce A (1994) ‘Crosslinguistic evidence for functional projections in early child
grammar’ in T Hoekstra & B D Schwartz (eds) Language Acquisition Studies in Generative
Grammar, Benjamins, Amsterdam, pp. 57-84.
Deuchar M & Quay S (1999) ‘Language choice in the earliest utterances: a case study with
methodological implications’ Journal of Child Language 26: 461-75.
Deuchar M & Quay S (2000) Bilingual Acquisition: Theoretical Implications of a Case Study,
Oxford University Press.
de Villiers JG & PA (1973) ‘A cross-sectional study of the acquisition of grammatical morphemes in
child speech’, Journal of Psycholinguistic Research vol.2 no.3.
Di Sciullo A-M, Muysken P & Sngh R (1986) ‘Government and code-mixing’, Journal of Linguistics
22: 1-24.
Döpke S (1992) One Parent - One Language, Benjamins, Amsterdam
Döpke S (2000) ‘Generation of and retraction from cross-linguistically motivated structures in
bilingual first language acquisition’, in Bilingualism 3: 209-226.
Drozd, K.F (1995) ‘Child English pre-sentential negation as metalinguistic exclamatory sentence
negation’, Journal of Child Language 22: 583-610.
Fanselow G. & Mahajan. A. (2000) ‘Towards a minimalist theory of wh-expletives,
wh-copying and
successive cyclicity’, in Lutz, Müller & von Stechow (eds), pp.195-230.
Fasold R (1980) ‘The relation between Black and White speech in the south’,
unpublished ms. School
of Languages and Linguistics, Georgetown University, Washington DC
Felser C (2004) ‘Wh-copying, phases and successive cyclicity’, Lingua 114: 543-574.
Fodor JD (1998) ‘Parsing to Learn’, Journal of Psycholinguistic Research 27: 339-374.
Fodor, JD & Crowther C (2002) ‘Understanding stimulus poverty arguments’, The
Linguistic
Review 19: 105-145.
Friedemann M-A & Rizzi L (eds) The Acquisition of Syntax, Longman, London, pp.269-292.
Fujino H & Sano T (2000) ‘Some notes on the null object phenomenon in child Spanish’ in
Proceedings of the 24th Annual Boston University Conference on Language Development, Cascadilla
Press, Somerville Mass, pp.308-318.
Garman M et al (2004) ‘The emergence of tense and agreement’, unpublished
paper, University of
Reading.
Gawlitzek-Maiwald I & Tracy R (1996) ‘Bilingual bootstrapping’, Linguistics 34: 901-926.
Galasso J (1998) On the Acquisition of Functional Categories: A Case Study, PhD diss, University of
Essex.
Gardner-Chloros P & Edwards M (2004) ‘Assumptions behind grammatical approaches to code-
switching: When the blueprint is a red herring’, Transactions of the Philological Society
102: 103-129.
Gavruseva L & Thornton R (2001) ‘Getting it right: Acquisition of whose-questions
in child
English’, Language Acquisition 9: 229-67.
Gawlitzek-Maiwald I & Tracy R (1996) ‘Bilingual bootstrapping’, Linguistics 34: 901-926.
Genesee F (1989) ‘Early language development: one language or two?’ Journal of Child
Language 16: 161-179
Genesee F (2002) ‘Portrait of the bilingual child’, in VJ Cook (ed) Portraits of the L2 User,
Multilingual Matters, Clevedon, pp.170-96.
Genesee F, Boivin I & Nicoladis E (1996) ‘Talking with strangers: A study of bilingual children’s
competence’, Applied Psycholinguistics 17: 427-42.
Genesee F, Nicoladis E & Paradis J (1995) ‘Language differentiation in early bilingual
development’, Journal of Child Language 22: 611-631
Gerken L (1991) ‘The metrical basis for children’s subjectless sentences’, Journal of Memory and
Language 30: 431-51.
Gleitman L & Wanner E (1982) ‘Language acquisition: The state of the state of the art’, in E Wanner
& L Gleitman (eds) Language Acquisition: The State of the Art, Cambridge University Press,
London, pp.3-48.
Goodz N (1994) ‘Interactions between parents and children in bilingual families’ in F Genesee (ed)
Educating Second Language Children, Cambridge University Press, London, pp. 61-81.
Greenfield P (1991) ‘Language, tools and the brain’, Behavioural and Brain
Sciences 14: 531-95.
Gringàs R (1974) ‘Problems in the description of Spanish-English intra-sentential
code-switching’, in
Bills GA (ed.) Southwest Areal Linguistics, Institute for Cultural Pluralism, San
Diego.
Guasti MT (1992) ‘Verb Syntax in Italian Child Grammar’ Geneva Generative
Papers 2: 145-162.
Guasti MT (1996) ‘Acquisition of Italian interrogatives’ in H Clahsen (ed) Generative Perspectives on
Language Acquisition, Benjamins, Amsterdam, pp. 241-269.
Guasti MT (2000) ‘An excursion into interrogatives in Early English and Italian’ in M-A Friedemann
& L Rizzi (eds) (2000), The Acquisition of Syntax, Longman, pp.105-128
Guasti MT (2002) Language Acquisition, MIT Press, Cambridge Mass.
Guasti MT & Rizzi L (1996) ‘Null AUX and the acquisition of residual V2’ in A Stringfellow et al
(eds) BUCLD 20 Proceedings Cascadilla Press, Somerville Mass, pp. 284-295
Guasti, M.T., Thornton R. & Wexler K. (1995) ‘Negation in children’s questions:
The case of
 English’, in B. MacLaughlin & S. McEwen (eds) Proceedings of the 19th Annual
Boston University
Conference on Language Development, Cascadilla Press, Somerville Mass,
pp.228-239.
Guilfoyle E (1994) ‘The acquisition of tense and the emergence of lexical subjects in child
grammars’, McGill Working Papers in Linguistics 2.1: 20-30
Guilfoyle E and Noonan M (1988) ‘Functional categories and language acquisition’, manuscript,
McGill University (revised version published in Canadian Journal of Linguistics, 1992).
Hale, K. & Keyser, S. J. (1991) On the Syntax of Argument Structure, Lexicon
Project Working
Papers, MIT, Center for Cognitive Science, Cambridge Mass.
Hale, K. & Keyser, S. J. (1993) ‘On argument structure and the lexical expression
of semantic
relations’, in K. Hale & S. J. Keyser (eds) The View from Building 20, MIT Press,
Cambridge Mass.,
pp. 53-109.
Hale, K. & Keyser, S. J. (1994) ‘Constraints on argument structure’, in B. Lust, M.
Suñer & J.
Whitman (eds) Heads, Projections and Learnability, Erlbaum, Hillsdale New
Jersey, vol 1, pp.
53-71.
Halmari H (1997) Government and Codeswitching: Explaining American Finnish, Benjamins,
Amsterdam.
Harris T & Wexler K (1996) ‘The optional-infinitive stage in child English: Evidence from negation’,
in H Clahsen (ed) Generative Perspectives in Language Acquisition, Benjamins, Amsterdam,
pp.1-42.
Henry A (1995) Belfast English, Oxford University Press
Hiraira K (2005) ‘Dimensions of agreement’, in C. Boeckx (ed.) Complex Agreement Systems, Oxford
University Press, Oxford.
Hoekstra T & Hyams N (1998) ‘Aspects of root infinitives’, Lingua 106: 81-112
Hoekstra T & Hyams N (1999) ‘The eventivity constraint and modal reference effects in root
infinitives’ in A Greenhill et al (eds) BULCD 23 Proceedings, Cascadilla Press, Somerville Mass,
pp. 360-373.
Hoekstra T, Hyams N & Becker M (1998) ‘Agreement and finiteness of V2: Evidence from child
language’. In A Greenhill et al (eds) BULCD 22 Proceedings, Cascadilla Press, Somerville Mass,
pp. 360-373
Hoekstra T & Schwartz BD (eds) Language Acquisition Studies in Generative
Grammar, Benjamins,
Amsterdam
Hong S-H (2004) Aspects of the Syntax of Questions in English and Korean, PhD
dissertation,
University of Essex.
Hurford, J. (1991) ‘The evolution of the critical period for language acquisition’,
Cognition 40: 159-201.
Huxley R (1970) ‘The development of the correct use of subject pronouns in two children’ in GB
Flores d’Arcais and WJ Levelt (eds) Advances in Psycholinguistics, North-Holland, Amsterdam, pp.
141-165.
Hyams N (1986) Language Acquisition and the Theory of Paramaters, Reidel, Dordrecht
Hyams N (1987) ‘The theory of parameters and syntactic development’ in T Roeper & E Williams
(eds) Parameter Setting, Reidel, Dordrecht, pp. 1-22.
Hyams N (1989) ‘The null subject parameter in language acquisition’, in O Jaeggli & KJ Safir (eds)
The Null Subject Parameter, Kluwer Academic Publishers, Dordrecht, pp. 215-238.
Hyams N (1992) ‘A reanalysis of null subjects in child language’ in J Weissenborn et al (eds)
Theoretical Issues in Language Acquisition, pp. 249-267.
Hyams, N (1994) ‘V2, Null Arguments and COMP projections’ in T Hoekstra & BD
Schwartz (eds)
Language Acquisition Studies in Generative Grammar, Benjamins,Amsterdam
pp.21-55.
Hyams N (1996) ‘The underspecification of functional categories in early grammar’ in H Clahsen (ed)
Generative Perspectives on Language Acquisition, Benjamins, Amsterdam, pp.91-127
Hyams N (2001) ‘Now you hear it, now you don’t: The nature of optionality in child grammars’, in
AH-J Do, L Domínguez and A Johansen (eds) Proceedings of the 25th Annual Boston University
Conference on Language Development, Cascadilla Press, Somerville Mass., pp 34-58
Hyams, N. & K. Wexler (1993) ‘On the grammatical basis of null subjects in child language’,
Linguistic Inquiry 24:421-459.
Imedadze NV (1967) ‘On the psychological nature of child speech formation
under condition of
exposure to two languages’, International Journal of Psychology 2: 129-132.
Ingham R (1992) ‘The optional subject phenomenon in young children’s English:
A case study’,
Journal of Child Language 19: 133-151.
Ingham R (1998) ‘Tense without agreement in early clause structure’, Language Acquisition 7: 51-81.
Jaeggli O & Safir K (1989) The Null Subject Parameter, Kluwer, Dordrecht.
Jake J & Myers-Scotton C (1997) ‘Codeswitching and compromise strategies: Implications for lexical
structure’, International Journal of Bilingualism 1: 25–39.
Jespersen O (1922) Language: Its Nature, Development and Origin, MacMillan, New York
Joshi AK (1985) ‘Processing of sentences with intrasentential code-switching’, in Dowty D,
Kartunnen L, and Zwicky A (eds) Natural Language Processing: Psychological, Computational
and Theoretical Perspectives, Cambridge University Press, Cambridge.
Kamwangamalu NM (1989) ‘The morphosyntactic aspects of French/English-Bantu code-mixing:
Evidence for universal constraints’, In CLS 25, Parasession on Language in Context, Chicago
Linguistic Society, Chicago, pp. 157–170..
Kayne RS (1989) ‘Notes on English agreement’, CIEFL Bulletin 1: 40-67.
Kazman R (1988) ‘Null arguments and the acquisition of case and INFL’, text of paper presented at
the Boston University Conference on Language Development.
Kielhöfer B (1987) ‘Le “bon” changement de langue et le “mauvais” mélange de langues’, in Lüdi G
(ed) Devenir Bilingue – Parler Bilingue, Niemeyer, Tübingen, pp.135-155.
Klavans JL (1985) ‘The syntax of code-switching: Spanish and English’ in King LD & Maley CA
(eds) Selected Papers from the XVIIIth Linguistic Symposium on Romance Languages, Benjamins,
Amsterdam, pp.213-231.
Klima ES and Bellugi U (1966) ‘Syntactic regularities in the speech of children’, in J Lyons and
R Wales (eds) Psycholinguistic Papers, Edinburgh University Press, Edinburgh, pp.183-207.
Koehn C (1994) ‘The acquisition of gender and number morphology within NP’ in JM Meisel (ed)
Bilingual First Language Acquisition, Benjamins, Amsterdam, pp.29-51.
Köppe R & Meisel JM (1995) ‘Code-switching in bilingual first language acquisition’ in L Milroy & P
Muysken (eds) One Speaker, Two Languages, Cambridge University Press, pp.276-301.
Kwan-Terry A (1992) ‘Code-switching and code-mixing: The case of a child learning English
and Chinese simulaneously’, Journal of Multicultural and Multilingual Development 13:
243-259.
Labov W (1969) ‘Contraction, Deletion, and Inherent Variability of the English
Copula’, Language
45: 715-762.
Labov W & Labov T (1978) ‘Learning the syntax of questions’, in RN Campbell & PT Smith (eds)
Recent Advances in the Psychology of Language, Plenum Press, New York, pp. 1-44.
Lanza E (1992) ‘Can bilingual two-year-olds code-switch?’ Journal of Child Language 19: 633-658.
Lanza E (1997) Language Mixing in Infant Bilingualism: A Sociolinguistic Perspective, Oxford
University Press, New York.
Lanza E (1997) ‘Language contact in bilingual two-year old and code-switching: language
 encounters of a different kind?’, International Journal of Bilingualism 2: 135-162.
Larson, R. (1988) ‘On the double object construction’, Linguistic Inquiry 19: 335-
91.
Larson, R. (1990) ‘Double objects revisited: reply to Jackendoff’, Linguistic Inquiry
21: 589-632.
Lasnik, H. & Uriagereka, J. (2002) ‘On the poverty of the challenge’, The Linguistic
Review 19: 147-
150.
Lebeaux D (1988) Language Acquisition and the Form of the Grammar, PhD diss., U.Mass

Lee I (1999) A principles-and-parameters approach to the acquisition of the

morphosyntax of IP in
Korean, PhD diss., University of Essex
Legate, J.A. (2003) ‘Some interface properties of the phase’, Linguistic Inquiry 34:
506-16.
Legate, J.A. & Yang, C.D (2002) ‘Empirical re-assessment of stimulus poverty
arguments’, The
Linguistic Review 19: 151-162.
Lenneberg, E. (1967) Biological Foundations of Language, Wiley, New York.
Leonard LB (1989) ‘Language learnability and specific language impairment in children’, Applied
Psycholinguistics 10: 179-202
Leopold W (1939-1949) Speech Development of a Bilingual Child: A Linguist’s Record (4 volumes),
Northwestern University Press, Evanston, Il. (Republished as Leopold 1970)
Leopold W 1970 Speech Development of a Bilingual Child (4 vols), AMS Press, New York
Lindholm K & Padilla A (1978) ‘Language mixing in bilingual children’ Journal of Child
Language, 5: 327-335.
Longobardi G (1994) ‘Reference and proper names’, Linguistic Inquiry 25: 609-66.
Lutz, U., Müller, G. and von Stechow A. (eds) (2000) Wh-Scope Marking,
Benjamins, Amsterdam.
Lyons C (1999) Definiteness, Cambridge University Press.
MacSwan J (1997) A Minimalist Approach to Intra-sentential Code-switching: Spanish-Nahuatl
Bilingualism in Central Mexico, PhD Dissertation, University of California, Los Angeles.
Mahootian S (1993) A Null Theory of Code-switching, PhD dissertation, Northwestern University.
Marcus, GF (1993) ‘Negative evidence in language acquisition’, Cognition 46: 53-
85.
McClure E & McClure M (1975) ‘Ethnoreconstruction’, Mid-America Linguitsics
Conference Papers,
University of Kansas, Lawrence, pp. 327-37.
McClure E & Wentz J (1975) ‘Functions of code-switching among Mexican-
American children’,
Papers from the Parasession on Funcionalism, Chicago Linguistic Society,
Chicago, pp.421-32.
McNeill D (1966) ‘Developmental Psycholinguistics’ in F. Smith & G.A. Miller (eds)
The Genesis of
Language, MIT Press, Cambridge Mass., pp.15-84.
McDaniel D (1986) Conditions on Wh-Chains, doctoral dissertation, City University
of New York.
McDaniel D (1989) ‘Partial and multiple wh-movement’, Natural Language and Linguistic Theory
7: 565-604.
Meisel JM (1989) ‘Early differentiation of languages in bilingual children’, in K Hyltestam &
L Obler (eds) Bilingualism Across the Lifespan, Cambridge University Press, pp. 13-40
Meisel JM (1994) ‘Code-switching in young bilingual children: The acquisition of grammatical
constraints’, Studies in Second Language Acquisition 16: 413-439.
Meisel JM and Mahlau A (1988) ‘La acquisición simultanea de dos primeras lenguas’, Actos del II
 Congreso Mundial Vasco, Vitoria, Servicio de publicaciones del Gobierno Vasco.
Morgan JL and Travis L (1989) ‘Limits on negative information in language input’,
Journal of
Child Language 16: 531-552.
Muysken P (1990) ‘Concepts, methodology and data in language research: Ten remarks from the
perspective of grammatical theory’, Papers for the Workshop on Concepts, Methodology and Data,
European Science Foundation Network on Code-Switching.
Muysken P (1995) ‘Code-switching and grammatical theory’, in Milroy L & Muysken P (eds) One
Speaker, Two Languages:Cross-disiplinary Perspectives on Code-Switching, Cambridge University
Press, Cambridge
Myers-Scotton C (1992) ‘Constructing the frame in intrasentential codeswitching’ Multilingua 11(1),
101–127.
Myers-Scotton C (1993) Duelling Languages: Grammatical Structure in Codeswitching,
Clarendon Press, Oxford.
Myers-Scotton C (1995) ‘A lexically based model of code-switching’ in L Milroy & P
Muysken (eds) One Speaker, Two Languages, Cambridge University Press, pp. 233-256.
Myers-Scotton C & Jake J L (1995) ‘Matching lemmas in bilingual competence-performance
model: Evidence from intrasentential code-switching’, Linguistics 33: 981-1002.
Myers-Scotton C & Jake J L (1997) ‘Code-switching and compromise strategies: Implications
for lexical structure’, International Journal of Bilingualism 1: 25-39.
Namai K (2000) ‘Gender features in English’, Linguistics 38: 771-779.
Nartey J S (1982) ‘Code-switching, Interference or Faddism? Language use among educated
Ghanaians’, Anthropological Linguistics 24, 183–192.
Nicoladis E & Secco G (1998) ‘The role of translation equivalents in a bilingual family’s code
-mixing’, In A Greenhill et al (eds) BUCLD 22 Proceedings 576-585.
Nishimura M (1997) Japanese/English Code-switching – Syntax and Pragmatics, Peter Lang, New
York.
Nishimura M & Yoon K-K (1998) ‘Head directionality and intrasentential code-switching: A study of
Japanese Canadians and Korean Americans’ bilingual speech’, in D Silva (ed.) Japanese/Korean
Linguistics 8: 121–130, CSLI Publications.
Nortier J (1990) Dutch-Moroccan Arabic Code Switching, Foris, Dordrecht.
O’Grady W, Peters A & Masterson D (1989) ‘The transition from optional to required subjects’,
Journal of Child Language 16: 513-29.
O’Grady W (1997) Syntactic Development , University of Chicago Press, London.
Ohara, H (2001) The nature of Children’s Initial Syntactic Structures: A Case Study, MA dissertation,
University of Essex
Oksaar E (1976a) ‘Code-switching as an interactional strategy for developing bilingual competence’,
Word 27
Oksaar E (1976b) ‘Implications of language contact for bilingual acquisition’, in McCormack W &
Wurm S (eds) Language and Man: Anthropological Issues, Mouton, The Hague
Padilla AM and Liebman E (1975) ‘Language acquisition in the bilingual child’, The Bilingual Review
Paradis J & Genesee F (1996) ‘Syntactic acquisition in bilingual children: autonomous or
interdependent?’ Studies in Second Language Acquisition 18: 1-25.
Parodi T (1990) ‘The acquisition of word order regularities and case morphology’, in JM Meisel (ed)
Two First Languages: Early Grammatical Development in Bilingual Children, Foris, Dordrecht,
pp. 157-92.
Pearson BZ, Fernández S & Oller DK (1995) ‘Cross-language synonyms in the lexicons of bilingual
infants: One language or two?’, Journal of Child Language 22: 34-68.
Petersen J (1988) ‘Word-internal code-switching constraints on a bilingual child’s grammar’,
Linguistics 26: 479-93.
Pfaff CW (1979) ‘Constraints on language mixing: Intrasentential codeswitching and
borrowing in Spanish/English’, Language 55: 291-318.
Pfaff C (1990) ‘Mixing and linguistic convergence in migrant speech communities: Linguistic
constraints, social conditions and models of acquisition’, in Papers for the Workshop on Constraints,
 Conditions and Models, ESF Network on Code-Switchnig and Language Contact, London,
pp.119-53.
Pfaff C (1992) ‘The issue of grammaticalization in early German second language’, Studies in Second
Language Acquisition, 14: 273-96.
Phillips C (2003) ‘Linear order and constituency’, Linguistic Inquiry 34: 37-90.
Pierce A (1989) On the Emergence of Syntax: A Crosslinguistic Study, PhD
dissertation, MIT
(published as Pierce 1992).
Pierce, A (1992) Language Acquisition and Syntactic Theory, Kluwer, Dordrecht.
Poeppel D & Wexler K (1993) ‘The full competence hypothesis of clause structure
in early German’,
Language 69: 365-424.
Pollock J-Y (1989) ‘Verb Movement, Universal Grammar and the Structure of IP’, Linguistic Inquiry
20: 365-424
Poplack S (1980) ‘Sometimes I’ll start a sentence in Spanish y yo termino en español: Towards a
typology of code-switching’, Linguistics 18: 581-618.
Poplack S (1988) ‘Contrasting patterns of code-switching in two communities, in M Heller (ed.)
Codeswitching – Anthropological and Sociolinguistic Perspectives, Mouton de Gruyter, Berlin, pp.
215–244.
Poplack S, Sankoff D & Miller C (1988) ‘The social correlates and linguistic processes of lexical
borrowing and assimilation’, Linguistics 26: 47–104.
Poplack S, Wheeler S & Westwood A (1989) ‘Distinguishing language contact phenomena: Evidence
from Finnish-English bilingualism’, World Englishes 8: 389–406.
Poplack S & Meechan M (1995) ‘Patterns of language mixture: Nominal structure in Wolof-French
and Fongbe-French bilingual discourse’, In L Milroy and P Muysken (eds.) One Speaker, Two
Languages: Cross-disciplinary Perspectives, Cambridge University Press, Cambridge, pp. 199–
232.
Powers SM (1996) The Growth of the Phrase Marker: Evidence from Subjects, PhD diss., University
of Maryland.
Pullum, G.K. & Scholz, B.C. (2002) ‘Empirical assessment of stimulus poverty
arguments’, The
Linguistic Review 19: 9-50.
Quay S (1995) ‘The bilingual lexicon: Implications for studies of language choice’, Journal of Child
Language 22: 369-387.
Radford A (1986) ‘Small children's small clauses’, Research Papers in Linguistics 1: 1-38, University
College of North Wales, Bangor.
Radford A (1988) ‘Small children's small clauses’, Transactions of the Philological Society vol. 86,
pp. 1-46.
Radford A (1990) Syntactic Theory and the Acquisition of English Syntax, Blackwell, Oxford
Radford A (1992) ‘The acquisition of the morphosyntax of finite verbs in English’, in JM Meisel (ed)
The Acquisition of Verb Placement, Kluwer, Dordrecht, pp.23-62.
Radford A (1995) ‘Children: Architects or Brickies?’ in D MacLaughlin & S. McEwen (eds)
Proceedings of the 19th Annual Boston University Conference on Language Development,
Cascadilla Press, Somerville Mass, vol. 1, pp. 1-19.
Radford A (1998) ‘Genitive subjects in child English’, Lingua 106: 113-131.
Radford, A. (2004a) English Syntax: An Introduction, Cambridge University Press.
Radford A (2004b) Minimalist Syntax: Exploring the Structure of English, Cambridge University
Press.
Radford A et al. (eds) (1999)Linguistics: An Introduction, Cambridge University Press.
Radford A and Yokota H (2005) ‘Principles, Parameters and Transfer in Second Language
Acquisition: Evidence from Long-Distance wh-questions produced by elementary Japanese Learners
of English’, unpublished paper, University of Essex.
Redlinger WE & Park T (1980) ‘Language mixing in young bilinguals’, Journal of Child Language
 7: 337-352.
Redlinger WE & Park T (1980) ‘Language mixing in young bilinguals’, Journal of Child Language
7: 337-352
Reidegeld A (1992) Grammatische Beschränkungen beim Code-switching: Eine kritische Literatur-
discussion, unpublished Master’s thesis, University of Hamburg.
Rispoli M (1994) ‘Pronoun case overextensions and paradigm bjuilding’, Journal of Child Language
21: 157-72.
Rispoli M (1995) ‘Mechanisms of pronoun case error: biased retrieval not
syntactic incompetence’,
ms, Northern Arizona University.
Rispoli M (1998a) ‘Patterns of pronoun case error’, Journal of Child Language 25:
533-84.
Rispoli M (1998b) ‘Me or my: Two different patterns of pronoun case errors’,
Journal of Speech and
Hearing Research 41: 385-93.
Rispoli M (1999a) ‘Case and agreement in English language development’,
Journbal of Child
Language 26: 357-72.
Rispoli M (1999b) ‘A developmental psycholinguistic approach to pronoun case
error: A reply to
Schütze’, Journal of Speech, Language and Hearing Research, 42: 1020-2.
Rispoli M (2000) ‘Towards a more precise model of pronoun case error: A
response to Schütze’,
Journal of Child Language 27: 707-14.
Rispoli M (2002) ‘Theory and methods in the study of the development of case and agreement: A
response to Schütze’, Journal of Child Language 29: 151-9.
Rispoli M (2005) ‘When children reach beyond their grasp: Why some children make pronoun case
errors and others don’t’, Journal of Child Language 32: 93-116
Rizzi L (1992) ‘Early Null Subjects & Root Null Subjects’, Geneva Generative Papers vol 0, pp.102-
114.
Rizzi L (1994a) ‘Some notes on linguistic theory and language development: The case of root
infinitives’, Language Acquisition 3: 371-393
Rizzi L (1994b) ‘Early null subjects and root null subjects’ in T Hoekstra & B Schwartz (eds)
Language Acquisition Studies in Generative Grammar, Benjamins, Amsterdam, pp. 151-176.
Rizzi L (1997) ‘The fine structure of the left periphery’ in L Haegeman (ed.) Elements of Grammar,
Kluwer, Dordrecht, pp. 281-337.
Rizzi L (2000) ‘Remarks on early null subjects’ in M-A Friedemann & L Rizzi (eds) The Acquisition
of Syntax, Longman, London, pp.269-292.
Roeper T & Rohrbacher B (2000) ‘True pro-drop in child Enlgish and the principle of economy of
projection’, in C Hamann & S Powers (eds) The Acquisition of Scrambling and Cliticisation,
Kluwer, Dordrecht.
Romaine S (1989) Bilingualism, Blackwell, Oxford.
Rymer, R. (1993) Genie: A Scientific Tragedy, Harper Perennial, New York.
Sampson, G. (2002) ‘Exploring the richness of the stimulus’, The Linguistic
Review 19: 73-104.
Sankoff D, Poplack S & Vanniarajan S (1990) ‘The case of the nonce loan in Tamil’, Language
Variation and Change 2: 71–101.
Sano T & Hyams N (1994) ‘Agreement, finiteness and the development of null arguments’,
Proceedings of NELS 24: 543-58
Santorini B & Mahootian S (1995) ‘Codeswitching and the syntactic status of adnominal adjectives’,
Lingua 96: 1–27.
Saunders (1982) Bilingual Children: Guidance for the Family, Multilingual Matters, Clevedon.
Swain MK (1977) ‘Bilingualism, monolingualism and code acquisition’, in Mackey W & Andersson T
(eds) Bilingualism in Early Childhood, Newbury House, Rowley Mass.
Schlyter (1990) ‘The acquisition of tense and aspect’ in in JM Meisel (ed) Two First Languages: Early
Grammatical Development in Bilingual Children, Foris, Dordrecht, pp. 87-122.
Scholz, B.C. & Pullum, G.K. (2002) ‘Searching for arguments to support linguistic
nativism’, The
Linguistic Review 19: 185-223.
Schütze C (1997) INFL in Child and Adult Language: Agreement, Case and Licensing, PhD diss.,
MIT.
Schütze C (1999) ‘Different rates of pronoun case error: Comments on Rispoli (1998), Journal of
Child Language 26: 749-55.
Schütze C (2001) ‘Productive inventory and case/agreement contingencies: A methodological note on
Rispoli (1999)’, Jounal of Child Language 28: 507-15.
Schütze C & Wexler K (1996) ‘Subject case licensing and English root infinitives’, in A. Stringfellow
et al (eds) Proceedings of BUCLD 20 vol 2, Cascadilla Press, Somerville Mass., pp. 670-681.
Sebba M (1998) ‘A congruence approach to the syntax of code-switching’, International Journal of
Bilingualism 2: 1-19.
Simonds R & Scheibel AD (1989) ‘The postnatal development of the motor
speech area: a preliminary
study’, Brain and Language 37: 42-58.
Sinka I & Schelleter C (1998) ‘Language mixing in young bilinguals’ The International
Journal of Bilingualism 2: 301-326.
Smith, N. (1998) ‘Jackdaws, sex and language acquisition’, Glot International 3, 7:
7.
Smith, N. (1999) Chomsky: Ideas and Ideals, Cambridge University Press,
Cambridge.
Sportiche D (1988) ‘A theory of floating quantifiers and its corollaries for constituent structure’,
Linguistic Inquiry 19: 425-49.
Sridhar SN & Sridhar KK (1980) ‘The syntax and psycholinguistics of bilingual code-switching’,
Canadian Journal of Psychology 34, 407–416.
Stenson N (1990) ‘Phrase structure, Congruence, Government and Irish-English code-switching’. In R
Hendrick (ed.), The Syntax of Modern Celtic Languages 169–197, Acadamic Press, San Diego.
Stromswold K (1990) Learnability and the Acquisition of Auxiliaries, PhD diss.,
MIT.
Stromswold K (1996) ‘Does the VP-internal subject stage really exist?’, unpublished paper presented
to the Boston University Conference on Language Development, Nov. 3 1996.
Taeschner T (1983) The Sun is Feminine: A Study on Language Acquisition in Bilingual Children,
Springer, Berlin.
Thomas, M. (2002) ‘Development of the concept of “the poverty of stimulus”’,
The Linguistic Review
19: 51-71.
Thornton R (1990) Adventures in Long-Distance Moving: The Acquisition of Complex Wh-questions,
PhD diss., University of Connecticut, Storrs.
Thornton R (1995) ‘Referentiality and Wh-movement in Child English: Juvenile D-Linkuency’,
Language Acquisition 4: 139-175.
Thornton R & Crain S (1994) ‘Successful cyclic movement’, in Hoekstra & Schwartz (eds), pp. 215-
252.
Timm L (1994) ‘The limits of code-switching constraints, with some evidence from Breton-French
code-switching’, Journal of Celtic Languages 3: 95–134.
Toribio, JA & Brown BL (1995) ‘Language contact and differentiation in child bilingualism:
a syntactic analysis’ in MacLaughlin D & McEwen S (eds) BUCLD 19 Proceedings pp.629
-642.
Tsimpli I-M (1992) Functional Categories and Maturation: The Prefunctional Stage of Language
Acquisition, PhD diss, University College London [published in 1996 as The Prefunctional Stage of
First Language Acquisition, Garland Publishing Inc, London]
Vainikka A (1994) ‘Case in the development of English Syntax’, Language
Acquisition 3: 257-325.
Vainikka A & Levy Y (1999) ‘Empty subjects in Finnish and Hebrew’, Natural
Language and
Linguistic Theory 17: 613-671.
Valian V (1991) ‘Syntactic subjects in the early speech of American and Italian children’, Cognition
40: 21-81.
Vihman M (1982) ‘The acquisition of morphology by a bilingual child’ Journal of Child Language
12: 297-324
Vihman M (1985) ‘Language differentiation by the bilingual infant’, Journal of Child Language 12:
297-324
Vihman M (1998) ‘A developmental perspective on codeswitching: conversations between a pair of
bilingual siblings’, The International Journal of Bilingualism 2: 45-84.
Volterra V & Taeschner T (1978) ‘The acquisition and development of language by bilingual
children’, Journal of Child Language 5: 311-326.
Wakefield J & Wilcox JM (1995) ‘Brain maturation and language acquisition: A theoretical model
and preliminary investigation’, in D MacLaughlin & S McEwen (eds) Proceedings of the 19th
Annual Boston University Conference on Language Development, Cascadilla Press, Somerville
Mass, vol. 2, pp. 643-654.
Wang Q et al. (1992) ‘Null subject versus null object: Some evidence from the
acquisition of Chinese
and English’ Language Acquisition 2: 221-254.
Weinreich U (1964) Languages in Contact, Mouton, The Hague
Weissenborn J (1992) ‘Null subjects in early grammars: Implications for parameter-setting theories’ in
J Weissenborn, H Goodluck & T Roeper (eds) Theoretical Issues in Language Acquisition, Erlbaum,
New York
Wexler K (1994) ‘Optional infinitives, head movement and the economy of derivation’, in D
Lightfoot and N Hornstein (eds) Verb Movement, Cambridge University Press, pp. 305-350.
Wexler K (1998) ‘Very early parameter-setting and the unique checking constraint: A new explanation
of the optional infinitive stage’, Lingua 106: 23-79.
Wexler K. (1999) ‘Maturation and growth of grammar’ in W Ritchie & TK Bhatia (eds) Handbook of
Child Language Acquisition, Academic Press, London
Wexler K, Schütze C & Rice M (1998) ‘Subject case in children with SLI and unaffected controls:
Evidence for the Agr/Tns Omission Model’, Language Acquisition 7: 317-344.
Wijnen F et al (2001) ‘Root infinitives in Dutch early child language’, Journal of Child Language
629-660.
Woolford E (1993) ‘Bilingual code-switching and syntactic theory’, Linguistic Inquiry,
pp.520-536.

10 Suggested assignment and dissertation topics

Assignment topics
1. Discuss evidence for the operation of UG principles in child grammars, basing your answer (at least
in part) on the material in the exercises in chapter 1. You will find more detailed discussion of how the
relevant principles operate in adult grammars in Radford (2004a, 2004b).

2. Discuss the nature of parameter-setting in child grammars, and include an evaluation of Hyams’
research on the acquisition of the Null Subject Parameter (whether the parameter-resetting account
outlined in Hyams 1986/1987/1989 or the morphological uniformity account in Hyams 1992). You
will find textbook accounts of Hyams’ work in the relevant chapters of Atkinson (1992), O’Grady
(1997) and Guasti (2002). For an important critique of Hyams’ early work, see Valian (1991) and other
works cited in chapter 2. You should also include discussion of some of the relevant exercise material
from chapter 2.

3. Provide a critical evaluation of two or more alternative accounts of null subjects in child English.
For example, you could compare one or more of Hyams’ syntactic accounts (referred to in topic 3)
with Radford’s (1990) null noun analysis discussed in chapter 3, or Paul Bloom’s (1990) processing
account (See Hyams and Wexler 1993 for a critique of this), or Gerken’s (1991) metrical account, or
Rizzi’s (2000) truncation account, and so on. You could use the data from the Claire files in the
Appendix to this book to look at the use which Claire makes of null subjects (and indeed null objects,
since a number of accounts of child null subjects predict that English children use null subjects but not
null objects).

4. Provide a critical evaluation of the structure-building model of the acquisition of English, outlined
in Radford (1990, 1995) and Vainikka (1994), making use of exercise material in chapter 3 as part of
your evaluation.

5. Provide a critical evaluation of Rizzi’s truncation model, making use of exercise material in chapter
4 as part of your evaluation. If you wish to look at how the truncation model deals with the acquisition
of languages other than English, look at the collection of papers (on child French, German, Italian etc.)
in Friedeman and Rizzi (2000).

6. Make a critical assessment of the ATOM account of subject case-marking (slightly different
versions of which are found in Bromber & Wexler 1995, Schütze and Wexler 1996, Schütze 1997, and
Wexler, Schütze and Rice 1998), comparing it with one or more alternative accounts like those in
Vainikka (1994), or Radford (1998) or one or more of Rispoli’s papers (1994, 1995, 1998, 1999,
2000). It would be instructive to use the data from exercises 6.1 and 6.2 to test the alternative
accounts.

7. Outline and evaluate the research by Hoekstra and Hyams (1998, 1999) and Hoekstra, Hyams and
Becker (1998) into the finiteness effect in child grammars, using the data from exercises 6.3 and/or
6.4 to determine whether the effect holds for Abe and Claire.

8. Evaluate Volterra & Taeschner’s claim that bilingual children initially start out with a single (mixed-
language) lexicon which has no synonyms/translation equivalents. In addition to Volterra and
Taeschner’s (1978) paper, other works you might want to look at include Redlinger & Park (1980),
Saunders (1982), Taeschner (1983), Vihman (1982, 1985), Quay (1995), Pearson, Fernández and Oller
(1995), and Deuchar and Quay (1999, 2000). For an overview, see the survey articles by de Houwer
(1995) and Bhatia and Ritchie (1999). (This is probably the least technical topic.)

9. Discuss the nature of parameter-setting by bilingual children. Under the single system model, we’d
expect bilingual children initially to have a single setting for any given parameter P which is
generalised to both languages, whereas under the separate systems model we’d expect them to have
completely separate settings for P for each language (where P has distinct settings in each of the adult
languages). Discuss how the Joshua data in exercise 7.6 call both these models into question, and
instead suggest that bilingual children initially have partially overlapping systems (such that Joshua’s
English may show the English setting for the Head Position Parameter, the German setting for the
Head Strength Parameter, and both German and English settings for the Object Shift Parameter).
Discuss what this implies for the notion of dominance in bilingual children.

10. Discuss one or more accounts of code-mixing in child grammars, and the extent to which each
manages to provide a principled account of the data in a selection of the exercise material at the end of
chapter 8.
Dissertation topics
1. The syntax of wh-questions in child English: Stromswold (1990), Thornton (1990), Thornton and
Crain (1994), Thornton (1995), Guasti (2000) and Gavruseva and Thornton (2001) are seminal studies
to get you started – though some are based on dated theoretical frameworks. You could examine a set
of files on the CHILDES data-base, and look at the various types of wh-question structure produced
by young children, and at whether/how a range of different models can account for them.

2. The syntax of negative sentences in child English (which brings into question issues about the
position occupied by bare/determiner-modified nominal subjects like man/the man and
nominative/accusative pronoun subjects like I/me). Again you could collate examples of child negative
sentences from the CHILDES data-base – though since it’s rare for children to produce negative
sentences with an overt subject, you’re going to have to do a lot of file-searching to find enough data
to lead you to an empirically well informed view of children’s early negative sentences.

3. Aspects of the morphosyntax of subjects in child English – where they are positioned in various
types of clause structure, whether subject nominals are bare or modified by a determiner, what case is
assigned to subject pronouns, and so on. You will be able to find plenty of relevant data from
CHILDES, though you have to be careful to select (e.g. two-year-old) children who use a reasonably
high proportion of bare nouns and non-nominative pronouns as subjects.

4. Aspects of the morphosyntactic development of a particular child. You could analyse a set of
transcripts of the speech output of one particular child and use them to test various claims made within
particular models of acquisition. You do, however, need to make sure you choose a corpus which
contains plenty of errors (e.g. null determiners, null pronouns, case errors, tense and agreement errors,
in-situ subjects and so on). The Iris files from the Wells corpus on the CHILDES data-base include a
wide range of errors; and the Claire files included in the Appendix to this book contain lots of null
determiners, null pronouns, as well as plenty of wh-questions (many with null subjects and null wh-
words), so provides a suitable corpus for testing specific aspects of the continuity+underspecification
model; but it isn’t suitable for working on the case-marking of subjects, for example. You could of
course try and collect a corpus of your own – but recording and transcribing child speech is a hugely
time-consuming task, and there is ultimately no guarantee that the little monster you decide to record
will produce the kind of errors and structures which provide you with the sort of evidence you need if
you are to arrive at a reasoned, empirically substantiated evaluation of the models you are testing.

5. The lexical development of a one-year-old bilingual child. If you have access to a bilingual child,
you can collect examples of her vocabulary (by spontaneous speech recordings, and/or by getting one
of her parents to do a diary study, recording each new word as it is produced), and use your corpus to
test claims about whether one-year-olds do (or don’t) have a single fused/mixed lexicon with no
translation equivalents. However, this is probably a topic in which it is hard to produce genuinely
insightful work (e.g. you could end up collecting a lot of data which ultimately show very little).

6. The nature of syntax in bilingual children. You can look at data from one or more bilingual children
around two years of age to see if they have completely overlapping, partially overlapping or non-
overlapping grammars. You can either collect data of you own from a bilingual child you have access
to (though this can be time-consuming), or use data from the CHILDES data-base or some other
source (I have a large corpus of data on English-French, English-Spanish, and English-Italian bilingual
children).

7. The nature of code-switching in bilingual children. Since children code-switch relatively

infrequently, you are unlikely to be able to get enough data from simply looking at transcripts of the
speech production of one particular child. Rather, you’ll have to collect published examples of code-
switching by reading a wide range of articles on bilingual children’s development, and looking at
which of the various models of code-switching that you discuss can deal with the widest range of
examples of children’s code-switching. Alternatively, you can use data from existing corpora (though
once again, note that even a large corpus may contain few examples of code-switching).

11. Appendix: The Claire files

The data below are transcripts of 45-minute samples of the spontaneous speech production of a girl
called Claire recorded over a 9-week period, starting at age 24 months 6 days. The recordings were
made by Jane Anne Collins Hill, and the transcripts form an appendix to her 1983 Indiana University
PhD dissertation A Computational Model of Language Acquisition in the Two-Year-Old. Collins
(p.126) reports that recordings made in weeks 4 and 5 were ‘lost due to a malfunctioning tape
recorder’. I have set out the transcripts exactly as Collins does, except that (for ease of reference) I
have numbered each of Claire’s utterances (in the column headed No): where two words are
transcribed as a single word (e.g. whats, looka, right-here) this indicates that Collins thinks that the
two items may form a single indivisible word for the child (though whether this is really so is perhaps
doubtful). The adult participant in the conversations was Collins herself (referred to as Jane by
Claire); I have italicised what the adult says.

SESSION 1: CLAIRE AT AGE 24.0 (24 MONTHS 0 WEEKS)

ADULT CHILD No
Oh Claire, is your house shut up?........................ house shut up 1
Would you like some help?................................... help 2
where chicken? 3
chair go? 4
kitty cat? 5
What’s the kitty cat doing? ................................. what doing? 6
Is he riding a bicycle? ......................................... riding bicycle 7
I’m getting my glasses so I can read the book .... read book 8
looka that 9
thatsa trash 10
Is it a garbage pail? ............................................ pail 11
tail 12
uh oh 13
where Mommy? 14
Big cats ............................................................... big cats 15
Small cats ............................................................ small cats 16
Short cats ............................................................ short cats 17
Where’s the doggie? ............................................ where doggie? 18
kitty 19
kitty go? 20
Is Pixie outside?................................................... Pixie outside 21
Pix outside 22
cow 23
Shall we read more books? ................................. read book 24
what that? 25
A big bad wolf .................................................... what that? 26
That’s the wolf’s eye ........................................... eye 27
whats that? 28
outside 29
What’s he doing? ................................................. what doing? 30
There’s the big yellow cat ....................................yellow cat 31
what that? 32
what doing? 33
That’s a little girl ................................................. where girl go? 34
where pencil go? 35
downstairs 36
ADULT CHILD No
where pencil? 37
Is Claire writing? ......................................... writing house 38
pencil 39
where people? 40
The people are right here .............................. people right-here 41
uh oh 42
where people? 43
uh oh 44
notherone 45
uh oh 46
 where daddy? 47
uh oh 48
notherone 49
where car? 50
car right-here 51
notherone chair 52
whats that? 53
uh oh 54
daddy 55
what that? 56
uh oh broken 57
bunny broken foot 58
uh oh 59
What’s this? .................................................... what this? 60
notherone 61
where people? 62
what that? 63
Upstairs ......................................................... upstairs 64
where go? 65
there it is 66
What’s this? .................................................... whats this? 67
uh oh 68
looka that 69
dumpit out 70
daddy in-there 71
mommy in-there 72
night-night peoples 73
uh oh 74
wheresa car? 75
moo 76
Two mommies ................................................ two mommies 77
put right-here 78
where cow go? 79
where cow go? 80
Look at this, Claire ........................................ looka this 81
Here’s the rooster .......................................... rooster 82
drinka water 83
Here’s the water trough ................................. drinka water 84
what happen? 85
notherone 86
again 87
where chicken? 88
here it is 89
ADULT CHILD No
notherone 90
notherone 91
chicken 92
notherone daddy 93
wheresa mommy? 94
notherone 95
right-here 96
cat 97
kitty 98
sit down mommy 99
 what that? 100
cow fall down 101
what that? 102
ring doorbell 103
more doorbell 104
wheresa cow? 105
open it 106
open mommy 107
open it 108
________________________________________________________
SESSION 2: CLAIRE AT AGE 24.1 (24 MONTHS 1 WEEK)

ADULT CHILD No
Let’s get a chair ................................................. a chair mommy 109
right here 110
Right here? ......................................................... right-here 111
Alright, here’s a chair for Mommy ..................... what that? 112
Mommy’s chair ................................................... get chair this one 113
This mommy needs a chair ................................. this mommy chair 114
Mommy’s eating supper ..................................... the chicken supper 115
More chicken ...................................................... sit up 116
close the door 117
chicken 118
wheres the bird? 119
bird 120
right-here 121
Cup of tea
Do you think this Mommy wants a cup of tea?.... mommy cup of tea 122
Do you think this chicken wants a cup of tea?.... chicken cup of tea 123
open it 124
Now close it ......................................................... open it 125
close 126
open it 127
close 128
open it 129
what that? 130
Shut .................................................................... shut 131
uh oh 132
That’s such a big chair ...................................... such a big chair 133
open it 134
ADULT CHILD No
pick up 135
pick it up 136
Jane do it 137
Jane do it 138
uh oh 139
open it 140
uh oh 141
open it 142
uh oh 143
pick up 144
 chicken 145
fall down 146
look it 147
nother mommy 148
nother cow 149
sleeping 150
daddy sleeping 151
what that? 152
baby 153
daddy sleeping bed 154
another one 155
All the people right-here 156
where the people? 157
moo 158
uh oh 159
bathtub 160
Everybody goes to sleep in the bathtub .............. chicken bathtub 161
baby 162
Here’s the bathtub .............................................. Claire bathtub 163
teddy bear 164
Claire 165
Where’s Claire’s pencil? ................................... where Claire pencil? 166
where Claire pencil? 167
Claire do it 168
where go? 169
There it is ............................................................ there it is 170
where go? 171
There it is ............................................................ where go? 172
what that? 173
broken 174
look it 175
what that? 176
broken 177
look it 178
what that? 179
open it cat 180
(asking to open book about cats)
what that? 181
is it? 182
where daddy go? 183
mamma rocking 184
looka that 185
that? 186
ADULT CHILD No
looka that 187
where cow? 188
What’s this kitty doing? ...................................... what kitty doing? 189
bathtub 190
where bathtub go? 191
another one boy 192
mommy shoe 193
nother one 194
cow 195
Now we’ll look at the cow book .......................... moo 196
 look it 197
cow 198
what that? 199
moon 200
meow kitty 201
big kitty 202
little one 203
that? 204
uh oh 205
What’s this? ........................................................ what this? 206
A boy falling down .............................................. boy 207
kitty picture 208
where it? 209
owl 210
what that? 211
animal that? 212
uh oh 213
what that? 214
A cow .................................................................. cow 215
what that? 216
look at that 217
What’s this doggy doing? ................................... what the dog doing? 218
He’s standing on his head ................................... standing 219
open 220
shut open 221
put in 222
knock it down 223
Claire picture 224
Claire eye 225
another Claire 226
one Claire 227
more 228
more Claire 229
mouth 230
put the mouth 231
right-here 232
looka Claire 233
more Claire 234
notherone 235
________________________________________________________

SESSION 3 CLAIRE AT AGE 24.2 (24 MONTHS 2 WEEKS)

ADULT CHILD No
fit it in there 236
more one daddy one 237
more daddy one 238
daddy, put it in in there 239
another one fit it 240
this mommy put it in 241
Jane 242
 little kitty 243
little kitty one 244
turn around 245
little one mommy bear 246
big porcupine 247
where two? 248
(meaning ‘Where’s the second fence?)
more one Claire 249
Where’s the chair? ............................................. where chair? 250
Is that a hanger?................................................. no hanger 251
Is it a hook? ....................................................... two hook 252
door 253
There’s a porcupine. Want to read it?................ read it 254
mommy bear 255
another bear mommy bear 256
There’s the mommy bear ................................... another mommy bear 257
little bear baby bear 258
There’s baby bear ............................................. there 259
look it that 260
what squirrel doing? 261
what that? 262
where blinkety mole? 263
what that? 264
there lizard 265
what lizard doing? 266
See the raccoon ................................................... raccoon 267
porcupine 268
what that? 269
That’s a weasel ................................................... that weasel 270
get up weasel 271
happen weasel? 272
Look at the bunnies ............................................ bunnies doing? 273
oh 274
bunnies up 275
Tickle .................................................................. tickle 276
where tickle? 277
no tickle him 278
Do you think he’ll wake up?............................... wake up 279
the prickly porcupine nose 280
nose 281
He went back to sleep ........................................ sleep porcupine 282
there fox 283
there the fox 284
ADULT CHILD No
There’s the nice box ............................................ the fox 285
put in there 286
theres the moms chair 287
theres a new one mommy chair 288
mommy gone? 289
people chair 290
notherone 291
where Mommy? 292
Downstairs. Shall we go find her?...................... go find her 293
no this one 294
 want cottage cheese 295
over there 296
Claire full 297
daddy ice cream 298
where Claire horse? 299
two Daddy fork 300
make another mommy 301
two mommies 302
lots of mommies 303
__________________________________________________________

SESSION 6 CLAIRE AT AGE 25.1 (25 MONTHS 1 WEEK)

(sessions 4 & 5 lost owing to a malfunctioning tape-recorder)

ADULT CHILD No
in the house 304
another chair 305
where chair? 306
where daddy? 307
Daddy’s in the chair............................................ daddy in 308
mommy 309
daddy chair 310
this daddy 311
two mommy 312
Where’s Daddy? ................................................ right-there 313
daddy thirsty 314
uh oh 315
uh oh 316
daddy 317
Who’s this? ................................ Pixie 318
girl 319
want dinner 320
where table? 321
daddy 322
uh oh 323
Pixie eat dinner 324
here Pixie 325
eat dinner 326
where the girl? 327
this Mommy 328
this puppy 329
ADULT CHILD No
uh oh 330
bunny 331
where another fence? 332
Where’s another fence?.............. put another fence 333
Is there another fence? .............. uh oh 334
horsie 335
horsie 336
put the people dinner time 337
dinner time (calling) 338
people calling dinner time 339
 dinner 340
no Pixie 341
daddy table too 342
where Daddy? 343
Jane help dinner time 344
moo 345
what the cow say? 346
(animal noises)
chair too (two?) 347
daddy chair 348
horsie 349
horsie swimming pool 350
horsie swimming in the pool 351
moo 352
bunny standing up 353
bunny stand up 354
what that? 355
A frisbie ............................................................... frisbie 356
what that? 357
A black spot on the frisbie .................................. Jane what is it? 358
David did it 359
where go? 360
house 361
where girl? 362
where another girl? 363
there baby 364
where baby? 365
What’s this?......................................................... open it 366
close 367
nother one 368
little one 369
Mommy won’t fit in the refridgerator ................. mommy fit refridgerator 370
open it 371
Shall I open the little one?................................... open little 372
Shall I open the big one?..................................... big one 373
shut 374
open door 375
What room is this?.............................................. what room this? 376
What room is this?.............................................. bed 377
What room is this?.............................................. kitchen 378
What room is this?.............................................. table 379
That must be the dining room ............................ dining room 380
chair 381
ADULT CHILD No
ring doorbell 382
Claire close it 383
mommy at church 384
where go? 385
ring bell 386
wheres Daddy? 387
in the house 388
Daddy’s in the chimney....................................... daddy in the hot chimney 389
chair 390
play this 391
 over here 392
play this one 393
put in the lap 394
where the hat? 395
the hat 396
theres the hat 397
this go? 398
where this go? 399
turn 400
dump it out 401
theres the tail 402
where the horse go? 403
here is 404
what do? 405
put in lap 406
Eyes...................................................................... eyes 407
woo (wolf) 408
mommy sheep 409
big hat 410
where the puzzle go? 411
Where did it go? It’s gone ................................... it all gone 412
gone back 413
Oh, behind your back, I see it ............................. where that puzzle? 414
uh oh 415
pick up 416
put it 417
theres the owl 418
eyebrow 419
chin 420
dirty 421
(pointing to clothes hamper)
jama dirty too 422
giraffe 423
baby 424
Is that a horsie? .................................................. no, giraffe 425
giraffe airplane 426
in the sky 427
ball 428
Play ball ............................................................. play ball 429
Play giraffe? ...................................................... play ball 430
Play ball ............................................................ play giraffe 431
Can you catch a giraffe? ................................... little flower 432
little birdie 433
ADULT CHILD No
horsie 434
play baby 435
baby in the box 436
uh oh 437
Claire fell down 438
get up 439
again 440
Where’s Claire? ................................................. Claire 441
ribbit (frog sound) 442
what frog doing? 443
 read book 444

SESSION 7 CLAIRE AT AGE 25.2 (25 MONTHS 2 WEEKS)

ADULT CHILD No
Oh, let’s play with the people ..... play 445
Is that what they came in? .......... box 446
play the carriage 447
Oh, I see the carriage ................. put that mommy in the carriage 448
mommy carriage 449
here daddy 450
here daddy 451
Is that daddy? ............................. there the daddy right-there 452
that the horsie 453
that the girl 454
Claire 455
put in bed 456
go nightie night 457
another daddy in 458
there Claire daddy bed 459
Daddy’s too big for the baby’s bed.
It’s the baby’s bed .............. baby bed 460
baby fit that? 461
girl 462
what this doing, Jane? 463
There’s the baby in the baby’s chair ............ put in there 464
there the daddy chair 465
another one chair 466
baby 467
that baby 468
see baby? 469
uh oh 470
Pick up daddy ............................. put the baby chair 471
num num (meaning ‘yum yum’) 472
table 473
like that 474
Claire 475
Claire baby 476
rock 477
rock in the chair 478
Rock? Is that a rocking chair?.... thata rocking chair 479
rock 480
ADULT CHILD No
what that? 481
that Claire 482
two Claire 483
puzzle 484
there go 485
that go there 486
that horse 487
the rocking chair goes? 488
the rocking chair go? 489
car go? 490
 right-here 491
chair go there 492
All right ...................................... put this, this Jane 493
there the baby (new sister) 494
Is the baby crying? Claire flower 495
see flower 496
Judy can play peekeboo with Paul ..... peekaboo 497
You can play peekaboo with Paul
Peekaboo Paul ................................... look it 498
handkerchief 499
You can smell the flowers .................. smell the flowers 500
You can look in the mirror ................ Claire in the mirror 501
Feel daddy’s whiskers ....................... whiskers 502
You can read Judy’s book ................. where Judy go? 503
open 504
You can put your finger through
Mommy’s ring ............................... finger 505
what daddy doing? 506
another bunny 507
Do you want to read about ‘Where
is Willy Worm?’ ............................ where worm? 508
Wow, juicy pears, oh boy ................. juicy pears 509
theresa spider 510
where daddy? 511
Ripe apples, said Willie .................... there pear 512
No, that’s an apple ........................... that tail 513
worm 514
what that? 515
bird flying 516
Home just in time ............................... bird big flying 517
there the bird 518
owl 519
worm 520
where that go? 521
push 522
It fits without the people .................... oh yes 523
There’s the baby ................................ baby dinner 524
baby drink the coffee 525
look at that 526
Claire do puzzle 527
where the baby go? 528
in the lap 529
Nice and smooth ................................. smooth 530
ADULT CHILD No
two 531
there baby 532
open eye 533
Raggedy Ann crying 534
down 535
Raggedy Ann sit down too 536
There are diapers for the baby ............ baby diapers 537
baby crying 538
baby finger nail 539
that jamas 540
Striped jamas ...................................... what that? 541
horse jamas 542
Or is that a giraffe? raffe 543
there mousie 544
lookit that 545
where the horsie? 546
in the box 547
taking horsie 548
Did you fall in the box again, Claire?.. where Claire? 549
where the Claire? 550
where Claire pencil? 551
I’ll find your pencil ............................. where the little pencil? 552
where the little one? 553
where the big pencil? 554
where the cow go? 555
more Claire (drawing) 556
What next? .......................................... eye 557
sleeping Claire 558
eyebrow 559
mouth 560
nose 561
there the daddy 562
more Claire 563
eye 564
sleeping Claire more 565
Claire woke up 566
there the daddy 567
Claire 568
more Claire 569
more Claire eyebrow 570
more daddy eyebrow 571
Does he need ears? ............................
Claire mouth 572
put more daddy 573
more big daddy 574
eye 575
Do you want him to wake up? ............ put daddy wake up 576
wake up 577
more wake up 578
more big daddy 579
more big daddy wake up 580
what that? 581

ADULT CHILD No
mommy 582
daddy sit in chair 583
where baby chair? 584
baby sitting 585
baby eating dinner 586
baby eating juice 587
there daddy in the chair 588
there daddy in the chair two 589
everybody sleeping 590
another chair 591
Is there a chair in the house? ..............
go night night 592
daddy sitting in the chair 593
eat dinner 594
who’s this? 595
where did the chair go? 596
where the daddy? 597
where the people? 598
where the stairs go? 599
there the man people 600
(fireman doll)
where the man? 601
__________________________________________________________

SESSION 8 CLAIRE AT AGE 25.3 (25 MONTHS 3 WEEKS)

ADULT CHILD No
girl sleeping 602
its a baby 603
its a dolly 604
what dolly name? 605
its a girl 606
doll name 607
big truck 608
uh oh 609
theresa one 610
theres two 611
thatsa one 612
that the man 613
Is that the fireman? ............................. what the daddy doing? 614
there daddy 615
where that go? 616
what daddy brought? 617
fire 618
truck 619
back it in 620
pick it up 621
the Pixie doing? 622
Pixie barking 623
there horsie 624
Pix downstairs 625
horsie upstairs 626
ADULT CHILD No
block 627
play blocks 628
Claire’s making a long line of blocks ……. push 629
Pixie go to sleep now 630
Is this a street? .................................... this a street 631
Pixie street 632
Pixie hiding 633
there 634
peekaboo 635
 big bear 636
uh oh 637
snap 638
night night 639
here 640
Shall I do it? I’ll do the snap (on toy
snake). Claire do it ............................ Claire did it 641
I wonder if bear could do it ............... want bear do it 642
Bear do it ........................................... bear do it 643
Bear did it .......................................... bear did it 644
snap again 645
Come on, bear push hard .................. push it 646
You did it, bear .................................. bear do it 647
the snake 648
whats the snake saying? 649
Going ssssss ....................................... the pig 650
where is it? 651
Little pig? ........................................... little 652
Is this a little pig or is this a big pig?.. little pig 653
What does the pig say? pig say oink 654
baby hat 655
where the baby? 656
duck 657
myna bird 658
truck 659
duck, gramma 660
giraffe 661
What’s this, Claire? ........................... blocks 662
whats this? 663
put the truck 664
Dump it out ......................................... dump 665
dump it out 666
more 667
Can you put it in one? ........................ that goes little one 668
that one goes another one 669
Goes in there ...................................... goes in there 670
letters on it 671
ready 672
there goes another one 673
where that go Jane? 674
goes there 675
goes here 676
that one fit 677
It goes here ......................................... goes here 678
ADULT CHILD No
goes here 679
dump it out 680
ohhhh 681
goes here 682
goes there 683
theres another one Jane 684
six 685
two 686
right-here 687
 ready 688
I’m ready too 689
one fits 690
fit here 691
theres more 692
two 693
three 694
six 695
goes here 696
dump it out 697
goes there 698
Make a tower ..................................... uh oh 699
do it 700
one 701
six 702
seven 703
two 704
two 705
two 706
Where did it go? ................................. Claire get it 707
Claire threw it .................................... Claire did it 708
water in it 709
play the water 710
The water ........................................... the water in 711
in there 712
From the bathtub? ............................. from in the bathtub 713
There’s no water in this one .............. that water this one 714
theres water in it 715
in it 716
the bathtub 717
Jane go home? 718
(Jane is putting toys away)
No, find a book ................................... color book 719
color 720
play with it 721
Dumpty Dumpty 722
thats Dumpty Dumpty 723
thats Dumpty Dumpty, Jane 724
the Dumpty doing? 725
bird 726
the bird doing? 727
what that? 728
the daddy hat 729
Daddy’s hat? ...................................... doggie 730
ADULT CHILD No
doggie hat 731
what that? 732
That’s a loaf of bread ......................... big doggie hat 733
whats that? 734
That is a boot ...................................... what the boot doing? 735
bunny 736
the bunny say? 737
theres more bunny 738
theresa bunny 739
That’s a mouse .................................... what the mousie doing? 740
moo 741
kitty 742
theres the kitty 743
meow meow 744
kitty say? 745
Meow meow ........................................ woof woof 746
now read porcupine 747
Jane read porcupine 748
mommy 749
oh 750
big 751
porpupine 752
water 753
porcupine lie down 754
what Jane doing? 755
I’m listening to Claire read the book .. porcupine sleeping 756
oh 757
that 758
blinkety mole 759
blinkety the mole 760
where the house go? 761
the porcupine doing? 762
Sweeping his house. The porcupine
is sweeping the porcupine house .... the porcupine the house 763
house 764
blinkety mole 765
whats that? 766
what the porcupine doing? 767
blanket 768
Claire’s blanket 769
Raggedy Ann lie down 770
dolly too 771
put on the bed 772
Raggedy Anne stay here 773
put the boy in 774
the little too 775
everybody sleeping 776
put the boy in too 777
put there 778
Raggedy Anne there 779
pick up 780
little blanket for dolly 781
Raggedy Ann 782
ADULT CHILD No
put blanket Raggedy Ann 783
everybody resting 784
blanket 785
Raggedy Ann too 786
wake up 787
peekaboo 788
Listen, I hear the water ...................... the water, Raggedy Ann? 789
piggie see the water in there? 790
the bird doing? 791
 SESSION 9 CLAIRE AT AGE 25.4 (25 MONTHS 4 WEEKS)

ADULT CHILD No
Oh, a monkey. What’s the monkey do? monkey go (wrinkles nose) 792
monkey go (wrinkles nose) 793
whats that? 794
giraffe 795
right-here 796
under bed 797
giraffe under bed 798
see the giraffe 799
I don’t see a giraffe ............................ the giraffe doing? 800
over there 801
Claire go get him 802
I see Claire’s slipper .......................... there Claire slipper 803
what that? 804
Oh, that’s part of the puzzle ............... no do puzzle 805
where the doggie? 806
doggie puzzle? 807
Here’s the giraffe ............................... now read the monkey one 808
read the monkey 809
Thank you, here’s the monkey ........... now Jane the monkey 810
put it on the giraffe 811
There’s the monkey on the giraffe ..... high (holding in air) 812
look at that 813
its mommys 814
Look at all the people in the truck mommy too 815
see the peoples talking 816
peoples talking 817
wheres the mommy talk? 818
mommy sleeping 819
mommy night night in the truck 820
oh theres the hammer 821
there it is 822
Claire walking 823
Claire found the rest of the
kangaroo puzzle ............................... where that go? 824
uh oh 825
here 826

ADULT CHILD No
You put the beads together. Claire
went bump ....................................... again 827
Jane 828
Where are you going Claire? See that under there? 829
What? I don’t see anything ................ the hammer 830
Play with the hammer ........................ play the hammer 831
here 832
Thank you. Here ................................ push him out 833
Put it back ......................................... put him back 834
 where that go? 835
Goes there ........................................... goes here 836
that go? 837
happened the hammer? 838
uh oh 839
all fall down 840
do that one 841
Do it again .......................................... again 842
push 843
Push hard ........................................... do it 844
Do it again ......................................... do it again 845
Do green one ..................................... do yellow one 846
that orange 847
no that orange 848
that yellow 849
that pink 850
that one 851
that, do it 852
its out 853
here 854
do the green one 855
do the red 856
Which one? ......................................... that one 857
take out now 858
I took it out ......................................... too hard 859
to get 860
do to hammer 861
I’ll do the hammer ............................. now take out 862
no 863
take out 864
put another one 865
take out 866
no take out 867
I did it again 868
here 869
do the hammer 870
take it out 871
pick it up 872
another one 873
take it out 874
take it out one 875
hold it 876
there 877
Claire did it 878
ADULT CHILD No
can’t do it 879
take out 880
hard 881
can’t do it 882
here 883
there 884
Claire did it 885
take it out 886
Claire do it 887
 theres Jane jacket 888
Jane see mommy 889
Claire standing on it 890
all fall down box 891
Claire turn it on 892
go get it 893
read the toys 894
look I found 895
the cup 896
look I found 897
theres two 898
theres three 899
theres one 900
________________________________________________________