Академический Документы
Профессиональный Документы
Культура Документы
Language Change
Oxford Studies in Comparative Syntax
Richard Kayne, General Editor
246897531
Printed in the United States of America
on acid-free paper
Preface
This collection of papers developed from material presented at the First Gen-
erative Diachronic Syntax Conference, held at the University of York in April
1990. The idea for that meeting was Adrian Battye's. Since then, the Con-
ference has developed into a series, with a second meeting at the University
of Pennsylvania in November 1992 and a third one currently planned for the
Free University of Amsterdam in March 1994.
Adrian Battye passed away in March 1993. Until a short time before his death,
we collaborated in editing this collection. With the exception of one final re-
vision of the Introduction, and the choice of the running order of the papers,
the book is our joint work.
3
4 INTRODUCTION
the hierarchic structural relation between verb and object is a principle of UG,
while the variation in linear order is stated as a parameter. Schematically, we
can summarise the situation as in (1) (this mode of presentation is borrowed
from Rizzi (1988), although Rizzi uses it in a slightly different way):
(1) a. VP -> {V, NP,...} (principle of UG)
b. V precedes/follows NP (parameter of UG)
Every natural language conforms to principle (la). We have written the ma-
terial on the right of the arrow in set notation so as to indicate that the rela-
tive order of V and NP is not determined by this principle, only the hierarchical
structure of (part of) VP. The parametric choice is given in (Ib): UG allows,
in fact requires, individual languages to choose either "precedes" or "follows"
here. Each choice is a different value for the parameter. It is clear that this
choice, although extremely simple and easy to state in itself, has far-reaching
consequences for the surface orders each language allows.
Of course, the "principle" in (la) is not a real principle of UG. The state-
ment given there is an instance of a more general schema for the internal struc-
ture of syntactic constituents given by X-bar theory. One of the most important
aspects of X-bar theory is the fact that all syntactic constituents are assumed
to have the same internal structure. So, a closer approximation to the relevant
UG principle is given by (2), where X is a variable taking syntactic catego-
ries as values:
(2) a. XP -> { X', YP }
b. X' -> { X°, ZP }
Here YP is the specifier of X' and ZP is the complement of X°. Whether these
categories are present in a given representation depends on other considerations
that we cannot go into here. Still supposing the parameter to be (Ib), this
restatement of (la) leads us to expect that head-complement relations involv-
ing heads of different categories are "harmonic" in the sense that heads con-
sistently precede or consistently follow their complements. Exactly this
observation formed the basis for the typological work initiated by Greenberg
(see Hawkins (1983) for a refinement). Here we see how the principles and
parameters approach can overlap with, and in fact supersede, the earlier, more
superficial typological studies.
This outline of the basic ideas of the principles and parameters approach
to cross-linguistic variation and typology, although extremely sketchy and sim-
plified, is enough for our purposes here. We now have a notion of how cur-
rent theory accounts for syntactic differences between languages. The
principles and parameters approach was designed to account for synchronic
variation, and, as we mentioned at the outset, it has been very successful in
this. It should be immediately clear how the approach can extend from the
synchronic to the diachronic dimension: just as we try to account for syntac-
tic differences among, say, the contemporary Romance languages in terms of
differing parametric values, so we want to account for differences between
Latin and the contemporary Romance languages in terms of differing para-
6 INTRODUCTION
metric values. This naturally implies parametric change over time. The agenda
for the study of diachronic syntax is thus set: we can analyse the historical
development of a given language (or language family) in terms of differing
parametric values at different times. Syntactic change is thus seen as param-
eter change.
This approach has the immediate consequence that studying historically at-
tested (but no longer spoken) etats de langue enlarges the data base available
for comparison. Just as synchronic work on languages not previously studied
from the principles and parameters perspective can bring to light new arrays
of data which require the postulation either of new parameters or of new com-
binations of settings of existing parameters, so work on earlier etats de langue
may yield the same result. A case in point is the important work on Old French
in Adams (1987a,b). Adams showed that certain observations about Old French
word order that were made in earlier philological studies (notably in
Thurneysen (1892) and in Foulet (1919)) could be elegantly captured by say-
ing that this language is both a verb-second language and a null-subject lan-
guage, with the added restriction that null subjects are only possible in
verb-second clauses (i.e. matrix clauses and a restricted class of embedded
clauses). This kind of analysis of Old French word order is discussed, and in
some cases questioned, in several of the articles included in this volume—see
SECTION 3 for introductory comments, and the references given there. It seems
that this particular combination of parameter settings (which we could
schematize rather simplistically as +V2, +null-subject) is rare among contem-
porary languages (but cf. the remarks on Romansch at the end of Vance's
contribution—Chapter 7).
If one takes the terms "synchronic" and "diachronic" in their strict,
Saussurean sense, we have up to now been discussing synchronic historical
syntax rather than genuine diachronic syntax. "True" diachronic syntax is the
study of how one etat de langue evolves into another, while the above remarks
have been concerned largely with past etats de langue viewed synchronically.
A properly diachronic analysis of Old French, then, would take into account
the development of this +V2, +null-subject system into Middle and perhaps
Modern French, or how Old French developed from Latin.
The study of how grammatical systems change over time raises a range of
questions which do not arise in purely synchronic work: Can we construct a
typology of changes? Is there a theory of language change independently of
linguistic theory generally? Is there a tendency for languages to "drift" to-
wards particular kinds of systems (cf. Sapir (1921))? What are the relations
between language change and other factors (extralinguistic events, socio- and
psycholinguistic phenomena, language contact, etc.)? What are the relations
between language change in different components of the grammar: phonol-
ogy, syntax, etc.? These are all traditional questions of diachronic linguis-
tics, and they illustrate the extent to which the study of language change can
interact with other areas of linguistics, and possibly with other disciplines.
Any answers we may devise to the questions given above will depend upon
our answers to this question: What are the mechanisms of parametric change?
INTRODUCTION 7
The papers in this volume are primarily concerned with parametric changes,
and this is clearly the central question for diachronic syntax in terms of the
principles and parameters approach. It is probable, in fact, that more is at
stake in fully understanding the nature of parametric change; following
Lightfoot (1979, 1991) and Clark and Roberts (1993), we believe that the study
of diachronic syntax, because it gives us an insight into the mechanisms of
parameter change, can tell us something about parameter setting, i.e. about
language acquisition. This is the point that we wish to pursue in the remain-
der of this section.
It is a long-standing idea, probably due originally to Hermann Paul (see
Paul (1920)), that grammatical systems can be changed through the process
of language acquisition. To put the idea rather crudely and simplistically, a
new generation of acquirers may, as it were, "misacquire" the parental sys-
tem in the sense that they may restructure the adult system in various subtle
and presumably imperceptible ways. In the course of time, the "misacquired"
system becomes the parental system for a new generation of acquirers who in
their turn may indulge in further "misacquisition." In this way, the system
alters over time.
Framing the question of parametric change in terms of aspects of language
acquisition in this way puts diachronic linguistics in the centre of generative
theoretical endeavour, given the fundamental importance of language acqui-
sition to generative theory. In fact, the mechanisms of parameter changing
can be identified with the mechanisms of parameter setting, which are held to
play a fundamental role in the acquisition of syntax (see Hyams (1986); Pierce
(1989); Pierce and Duprez (1990); and the papers in Roeper and Williams
(1987)). This point emerges more clearly if we adopt the distinction between
I(nternal)-language and E(xternal)-language made in Chomsky (1986a: 19-24).
An E-language is a collection of actual or potential linguistic objects asso-
ciated with some population. In these terms, Bloomfield's (1933) notion of
language as a collection of speech events corresponds to an E-language, as
do the variations on the notion of "set of well-formed formulae", "set of sound-
meaning pairs" etc. that are inspired by work in analytic philosophy (see, for
example, Quine (1960); Montague (1974)). Also, most traditional work in his-
torical linguistics and philology is work on E-language. The important point
that such otherwise quite different approaches to the study of language have
in common, in Chomsky's view, is that the analysis of a language as an E-
language is quite independent in principle of any properties that may be as-
cribed to the mind/brain of native speakers of that language.
An I-language, on the other hand, is "some element of the mind of the per-
son who knows the language, acquired by the learner and used by the speaker/
hearer" (Chomsky (1986a:22)). Clearly, then, generative grammar is prima-
rily concerned with I-language. Chomsky goes on to make the following point:
The I-language L may be the one used by a speaker but not the I-language L', even
if the two generate the same class of expressions ...; L' may not even be a possible
human I-language, one attainable by the language faculty (ibid:23).
8 INTRODUCTION
have no way of explaining why acquirers would retreat from superset gram-
mars, since they would have no positive evidence that their parametric choice
did not correspond to the adult grammar. In the context of language change,
acquirers may "wrongly" choose a subset grammar, and never receive suffi-
ciently robust positive evidence that the parental grammar in fact has the
superset value.
Clark (1990) has shown that the Subset Principle is relevant for systems of
parameters, even where two parameters may give rise to an intersection (rather
than an inclusion) relation among the associated grammars. Clark refers to
this situation as a "shifting" relationship among parameters. Abstractly, we
can illustrate it as follows: suppose we have two parameters P 1 and P2 each
of which individually gives rise to an intersecting grammar. However, where
both parameters have the positive value, a grammar arises which is a superset
of either of those which emerge when only one parameter has the positive
value. The point can be illustrated by the following diagrams, where the
shaded area indicates the set of grammatical sentences:
b. P1 negative, P2 positive:
In a shifted system, the Subset Principle may lead to a change in the value of
one of the intersecting parameters—see Clark and Roberts (1993) for a case
study of exactly this kind. The conclusion is that, other things being equal,
shifted systems are disfavoured, or "marked," ultimately because of the prop-
erties of the acquirer and the process of acquisition. Of course, given the enor-
mous complexity of grammatical systems, it is frequently the case that other
things are not precisely equal. Nevertheless, the result is interesting, and in-
dicates how diachronic work can be important in developing the theory of
parameters and of parameter setting.
Our purpose in the foregoing remarks has been to sketch the conceptual
background to the papers in this volume, and to indicate what we see as the
implications and the potential importance of work of the type reported here.
It should be clear that contemporary linguistic theory is in a position to pro-
vide genuinely new insights into a range of both novel and traditional ques-
tions.
To illustrate this last assertion, consider the kinds of answers we may give
to the traditional questions about language change that we raised earlier. First,
can we construct a typology of changes? The answer to this question is that
all parametric changes must convert one possible collection of parameter val-
INTRODUCTION 11
ues into another; in this very general sense, the answer to the question is posi-
tive. However, it should also be clear that the typology of changes will be
indistinct from the typology of possible languages in general.
Is there a theory of language change independently of linguistic theory more
generally? If the core notion of the theory of language change is that of pa-
rameter change, and that notion is not distinct from the notion of parameter
setting, then we see that the answer to this question is negative. Linguistic
theory as constructed to account for the attested synchronic variation and the
facts of language acquisition gives us all the conceptual tools we need (this
point has also been argued by Lightfoot (1991)).
Is there a tendency for languages to "drift" in Sapir's sense? Parameter
theory again suggests that the answer to this question is negative. Language
change is essentially a random "walk" through the space of possible combi-
nations of parameter settings. However, at the level of individual construc-
tions there may be more to say. The kinds of reanalyses of "opaque," or
excessively effortful, input that may trigger parametric changes by altering the
nature of the trigger experience will, if Roberts (1992) is right, always sim-
plify limited classes of structures. We suggested above that this kind of change
underlies the descriptive notion of "grammaticalization," which is certainly a
very frequently attested kind of change. This kind of reanalysis is, however,
distinct from parameter change; reanalysis simplifies given structures while
parameter change alters whole systems. Note that we cannot view parameter
change as simplifying or as complicating grammatical systems; this would be
as misleading as claiming that synchronically attested languages differ in
overall complexity.
Finally, on the question of the relations between language change and other
aspects of language (whether intra- or extralinguistic), we have nothing to add
to whatever general view may be taken on how parameter setting is influenced
by these factors. Here again, questions of language change reduce to more
general questions about language acquisition. Here, as above, we see how
formulating the agenda for diachronic syntax in terms of parameter theory
holds out the prospect of understanding traditional questions of diachronic lin-
guistics in terms of general aspects of linguistic theory.
2. Verb Syntax
this approach was further developed by Chomsky (1986b) and Baker (1985b,
1988).
In SECTION 1 we introduced the X-bar schema for the internal structure of
syntactic categories. We repeat it here for convenience:
(5) a. XP -» { X', YP}
b. X' -» { X , Z P }
Originally this schema was taken to apply only to the lexical categories N, V,
A and P. It was assumed that the "clausal" categories S and S' were gener-
ated by the following PS-rules:
(6) a. S' -> COMPS
b. S -> NP Aux VP
This was clearly an undesirable state of affairs given the general programme,
initiated by Chomsky (1981) and Stowell (1981), for the elimination of inde-
pendent PS-rules from the theory. Accordingly, Chomsky (1986b) proposed
that S and S' are projections of the non-lexical categories I(nfl) and C(omp),
respectively. I corresponds roughly to the Aux node of earlier work, and con-
tains in particular, features specifying tense and agreement; this position had
originally been argued to be the head of S in Hale, Jeanne and Platero (1977).
In the system of Chomsky (1986b), the specifier of I is the subject position,
and the complement of I is VP. C is the earlier Comp position; its specifier is
the landing site for wft-movement and (in verb-second languages-see below)
fronted topics, and its complement is IP. So we have the following clause
structure:
This structure is the basis for nearly all the papers in this volume, although
some of them elaborate it in various ways.
INTRODUCTION 13
Here the verbal root first incorporates into the tense affix to give a complex
verb form which, in turn, incorporates with the agreement affix to give the
full verb form. Assuming this clause structure, i.e. the Mirror Principle and
the Head Movement Constraint (which requires heads to move cyclically
through locally governing head-positions, with no "skipping" of intermediate
positions), we derive the form of the complex verb automatically.
Chomsky (1991) elaborates Belletti's proposal further by suggesting that,
in a sense, both Belletti and Pollock are right: there is an Agr-projection above
T—this is the position of affixes specifying agreement with the subject, or Agr-
S—and there is also an Agr-projection below TP, the position for agreement
with the object, Agr-O.
In any case, whichever variant of the "split-Infl" proposal is correct, it is
now fairly clear that the basic clausal structure of many languages provides
evidence for at least two functional projections (in addition to CP). Belletti
(1990) shows that a range of what appear to be minor word-order differences
between French and Italian can be profitably analysed as the direct conse-
quence of the different scopes of verb-movement rules, rather than as rules
affecting particular lexical items.
Consider for example the different distribution of negative polarity items
with respect to the infinitive in French and Italian. While in Italian the nega-
tive polarity item invariably follows the infinitive, in French its most usual
position (and, for some items, the only position) is that preceding the
infinitive:
(15) Gianni ha deciso di non tornare mai/piu.
Gianni has decided to not return never/more.
(16) *Gianni ha deciso di non mai/piu tornare.
INTRODUCTION 17
terms, the loss of verb second is seen as the loss of this feature on C; in
Platzack's contribution to this volume he proposes that this operator has shifted
its position from C to I in the recorded history of both English and French,
thereby explaining the loss of verb second in these languages. Hulk and van
Kemenade propose a similar account based on a notion of "licenser." On the
other hand, Lemieux and Dupuis question the existence of a root/embedded
distinction in Old French verb second. Tomaselli (1989) takes a slightly dif-
ferent tack by suggesting that verb-second C contains an abstract agreement
specification; this account of verb second is taken up by Roberts (1992) in
his account of the loss of verb second in French.
Accounts vary as to the nature of the obligatory topicalization. In Roberts
(1992:71) this is taken to be forced by the following condition:
(23) A head containing Agr must have a filled specifier.
Since a verb-second C always contains Agr (unlike other Cs), this condition
requires the presence of some constituent in the specifier of those Cs. How-
ever, no further condition specifies the nature of that constituent. Roberts
suggests that the condition in (23) is related to the Extended Protection Prin-
ciple of Chomsky (1982). Another approach is to say that the topicalization
process triggers the V-fronting process. This idea was first developed by
Taraldsen (1986), and is argued for here by Lightfoot (Chapter 2). Despite a
large and growing research literature on verb second, however, the questions
concerning how this operation is triggered have not been fully answered; for
a detailed survey of all the main proposals that have been made (including
some we have not mentioned here), see Vikner (1990, ch. 2).
In this section we have given only the barest outline of the theory of verb
movement, and of the various proposals that have been made to account for
the different kinds of verb movement that have been observed. The analyses
to be found in the articles contained in this volume, which we will review in
the next section, will refine this outline somewhat.
and Dupuis argue for a particular analysis of Middle French verb second, and
Ribeiro shows that Old Portuguese had verb-second properties. Finally,
Kiparsky presents an account of the development of verb second in Germanic.
Lightfoot's paper deals with a question central to all theoretically driven
work on diachronic syntax: the precise nature of the interaction of language
acquisition and parameter setting. As we saw in SECTION 1, Universal Gram-
mar has, in recent generative research, been considered to be made up of a
set of principles and a set of parameters. Language acquisition may be thus
conceived of as the child setting the various parameters of the particular lan-
guage she is exposed to. A language will therefore change over time as
acquirers of that language come to set parameters differently (see the discus-
sion around (3) in SECTION 1). Lightfoot sets out to show, however, that the
loss of the verb-second phenomenon in the diachronic development of English,
when examined in some detail, shows that the whole question of parameter
setting is no simple matter.
The relationship between grammatical data and the triggering experience
which sets the parameters for these data may be a very indirect one, Lightfoot
maintains. He goes on to conclude "that a modern historical linguist cannot
say that some changing phenomenon is due to the new parameter setting which
accounts for it.... Rather, it manifests and provides evidence for that param-
eter setting, it is due to changes in the triggering experience which in turn
entailed the new parameter setting." This is an important conclusion for our
overall conception of language change as outlined in SECTION 1.
Beatrice Santorini's paper "The Generalisation of the Verb Second-Phenom-
enon in the History of Yiddish," also deals with the diachronic development
of verb second. Santorini shows that early Yiddish exhibits the same asym-
metry between root and embedded contexts as is to be found today in Ger-
man, while no such distinction is observable in Modern Yiddish. A quantitative
investigation of over 2,200 subordinate clauses from over 40 Yiddish texts
allowed the author to examine the generalisation of verb second from root to
subordinate clauses. This diachronic change with respect to verb second is
attributed to two other syntactic phenomena in Yiddish: a change in phrase
structure from I-final to I-medial configuration and a change in Nominative-
Case assignment which allows for SpecIP to be occupied by non-subjects. The
modification in Nominative-Case assignment is accounted for by postulating
a shift of the finiteness operator from the C-position to the I-position (see
Platzack's paper in this volume for more on the finiteness operator).
Monique Lemieux and Fernande Dupuis' contribution focusses on a "sym-
metric" verb-second system, i.e. one in which verb second is not restricted to
root clauses. Their paper, "The Locus of Verb Movement in Non-Asymmet-
ric Verb-Second Languages; the Case of Middle French" looks at the verb-
second data from the history of French and gives an analysis of them which
is at variance with the other papers in this volume dealing with verb second
in French. The fundamental difference between this approach and the others
is their claim that verb second in French can be best analysed as an IP-inter-
nal syntactic phenomenon. This analysis implies that, as in Yiddish (cf.
22 INTRODUCTION
Santorini's contribution) and Icelandic, Old and Middle French should not
display asymmetries between root and embedded environments. It is argued
that, although verb second is far less frequent in embedded contexts, it is
nevertheless possible and that extraction from verb-second embedded clauses
suggests that the exceptional "CP-recursion" analysis usually offered for them
in the context of a CP-analysis for verb second is not applicable to the French
data. The rarity of verb-second embedded clauses is considered to be ame-
nable to an analysis which has recourse to illocutionary elements in the sen-
tence.
Data from Portuguese are examined in Ilza Ribeiro's paper "Evidence for
a Verb-Second Phase in Old Portuguese." The analysis of a corpus of examples
drawn from 14th-century texts suggests that verb-second phenomena can be
identified in Old Portuguese. These data are shown to be compatible with the
sort of CP-analysis for verb-second order which has often been proposed for
Old and Middle French. Ribeiro offers an explanation of how the verb-sec-
ond configuration was lost from this language (with the exception of a cer-
tain residue). She bases her account on the proposals made in Roberts (1992),
which in turn derives from Koopmann and Sportiche (1991), that various para-
metric choices can be made with respect to the assignment of Nominative Case
by either Agr or by T. The verb-second nature of Old Portuguese is also briefly
discussed in Beninca's contribution (Chapter 12), where Portuguese is inte-
grated into a typology of verb second in Medieval Romance.
Paul Kiparsky's paper "Indo-European Origins of Germanic Syntax" at-
tempts to deal with the origins of verb second in the Germanic languages.
Kiparsky identifies the close relationship between the phrase structure of Old
English, Old High German and Old Icelandic together with their parent lan-
guage, Proto-Germanic, and other branches of Indo-European represented by
Vedic Sanskrit, Hittite, and Homeric Greek. Evidence based on data concern-
ing verb movement, clitic movement, coordination and relative-clause forma-
tion is given which argues for a Germanic clause structure which distinguishes
an adjoined topic position, a Specifier position and a head-position within CP
(it is interesting to compare this proposal with Beninca's structure for Medi-
eval Romance, discussed below).
Kiparsky argues that this structure descends directly from Indo-European,
except that Indo-European had no C-position (in the sense of having a slot
where indeclinable complementizers are inserted in embedded clauses). This
in turn is connected to the fact that finite embedded clauses are not found in
older Indo-European languages. The development of complementizers, and
consequently of a C-position capable of hosting the finite verb, is tied to the
development of finite subordination.
Part Two groups together papers which relate aspects of verb second to the
null-subject parameter. Most of the papers focus on French, although both
Platzack's paper and Hulk and van Kemenade's paper compare the develop-
ment of French with the development of English in these respects.
Barbara Vance's paper concentrates on French. As Vance points out in her
introduction, her focus is on the decline, rather than the loss, of verb second.
INTRODUCTION 23
What Vance argues is that many cases of verb-subject order in Middle French
which are superficially cases of verb second, are not to be taken as instances
of "true" verb second (which Vance takes to involve movement to C), but are
in fact cases of Italian-style "free inversion," in the sense of Rizzi (1982). The
free-inversion construction is characteristic of null-subject languages and con-
sequently available in Old and Middle French.
Vance's argument is that the availability of free inversion led to the weak-
ening and eventual loss of productive verb second at the end of the Middle
French period. She supports her arguments by showing that there is a differ-
ence in frequency in Middle French between verb-subject orders where the
subject is a pronoun and such orders where the subject is a non-pronominal
NP: the former decline through this period while the latter remain constant.
Since "free inversion" with a pronominal subject was not possible in Old
French or Middle French, inversion of a pronominal subject was an unambigu-
ous indication of a "true" verb-second structure with the verb in C, while in-
version with a non-pronominal subject was often amenable to either a
verb-second analysis (with the verb in C) or a "free-inversion" analysis (with
the verb in I). Accordingly, Vance interprets the statistical data as evidence
that "true" verb second was declining throughout Middle French. This idea
is further supported by the development of "verb-third" clauses, which is much
more frequent after the 13th century for clauses with pronominal subjects than
for those with non-pronominal subjects; again, Vance interprets this as evi-
dence that non-pronominal subjects were able to appear in a different kind of
inversion construction, and that this very fact was responsible for the decline
of verb second.
An attempt to reach a unified account of the loss of verb second in French
and English is to be found in Christer Platzack's article ("The Loss of Verb
Second in French and English"). The author's starting point is the observa-
tion that scholars agree that the loss of verb second takes place around the
14th and 15th century in these languages, but different reasons, usually lan-
guage-specific ones, are put forward for this loss.
The leading idea in this paper is that the loss of verb second is to be corre-
lated with an abstract finiteness operator which indicates the existence of a
predication at the time of the utterance. A mutual dependency is postulated
between the position of this feature and the assignment of Nominative Case.
In a verb-second language, the feature is associated with the C position and
in a non-verb-second language with I. The change of position of the feature
from C to I takes place at the same time in French and English, hence the
simultaneous loss of verb second. However the other consequences of this
change are not the same in the two languages. The main consequence of the
shift in the position of the finiteness operator for English has been the growth
of do-support and Infl-lowering while for French (at least for Middle French)
it is to be witnessed in a change in the distribution of null subjects.
The loss of verb-second configurations in the diachronic development of
Modern English and French is also the central concern of Aafke Hulk and Ans
van Kemenade's article "Verb Second, Pro-drop, Functional Projections and
24 INTRODUCTION
Language Change." The authors consider that two different kinds of abstract
structures can give rise to verb second: structures involving movement of the
verb to I and subsequent movement of V+I to C, and structures which involve
the movement of V to I. Depending on whether the verb-second language
makes use of the first or the second strategy, it is classified as a
"Complementizer Verb-Second" (CV2) or an "Inflectional Verb-Second" (IV2)
language.
With this background Hulk and van Kemenade examine in some depth the
interrelated phenomena of verb second and the distribution of referential and
expletive null subjects. While Old French and Middle French (at least ini-
tially) are considered to have an I which is morphologically rich enough to
license referential null subjects, Old English does not possess such morpho-
logical richness in I and, therefore, only expletive null subjects can be identi-
fied. Both Old French and Old English are, in terms of this analysis, to be
classified as CV2 languages. The major change which takes place around the
beginning of the 15th century is the move from CV2 to IV2 in both languages.
A consequence of this change is that SpecIP becomes the only position to
which Nominative Case can be assigned in English and, therefore, English
becomes a non-verb-second language. On the other hand, French, with a mor-
phologically rich I, even after the shift from CV2 to IV2, can still license and
Case-mark referential null subjects in either VP-internal position or in SpecIP;
a consequence of this situation is that for a short period (until this morpho-
logical richness is lost) French passes through an IV2 phase before becoming
a non-verb-second language.
Paul Hirschbiihler's contribution, "Null Subjects in Verb-First Embedded
Clauses in Philippe de Vigneulles' Cent Nouvelles Nouvelles" focusses on the
possibilities for null subjects in particular contexts in this early 16th-century
French text. The paper aims to delimit as precisely as possible the syntactic
contexts in which embedded verb-first configurations are found in this text.
The results are that it is necessary to draw up a three-way division between
the different kinds of subordinate-clause introducers according to whether they
contribute to the formal licensing of SpecIP or not. The three groupings are
shown to be those subordinators that do not license SpecIP, those that do (wh-
phrases) and those that appear with hybrid properties (allowing preverbal sub-
jects but not pro-drop): the last of these categories is instantiated by relative
and comparative que.
A similar tripartite division has to be made with respect to the person of
the verb. Pro-drop is possible with second person plural, but first person and
second person singular never allow it. Third person singular, expletive third
person and third person plural do allow pro-drop in embedded contexts intro-
duced by a w/i-item. Careful examination of the various kinds of w/i-clause
shows that pro-drop in verb-first constructions is not equally represented in
all of them. The paper is presented as strong justification of the need for more
precise and accurate empirical studies of this kind.
The last three papers all deal in various ways with the relation between
clitics and verb second. Cecilia Poletto's object of study is the status and dis-
INTRODUCTION 25
so that I became final, following its structural complement, VP. The loss of
the medial I brought in its wake the loss of the possibility for the subject clitic
to attach to I. In the development of English, however, I has remained me-
dial but the loss of the possibility of cliticisation to this node is to be attrib-
uted to the effects of deflexion (both Hulk and van Kemenade and Platzack
discuss this point further).
This collection of papers testifies to the liveliness of current work in
diachronic syntax. We see that, on the one hand, traditional questions (the
Tobler-Mussafia Law, the nature of Germanic innovations) of historical lin-
guistics are being seen in a new light, and can now be investigated using the
powerful tools of contemporary linguistic theory. On the other hand, these
papers show how diachronic data can be brought to bear in developing those
theoretical tools. In these ways, diachronic syntax is becoming fully integrated
in contemporary syntactic theory.
References
Adams, M. (1987a) "From Old French to the Theory of Pro Drop." Natural
Language and Linguistic Theory 5:1-32.
Adams, M. (1987b) Old French, Null Subjects and Verb Second Phenomena.
PhD Dissertation, UCLA.
Andersen, H. (1973) "Abductive and Deductive Change." Language
49:567-595.
Baker, M. (1985a) "The Mirror Principle and Morpho-Syntactic Derivations."
Linguistic Inquiry 16:373-416.
Baker, M. (1985b) Incorporation: A Theory of Grammatical Function
Changing. PhD Dissertation, MIT.
Baker, M. (1988) Incorporation: A Theory of Grammatical Function
Changing. Chicago: University of Chicago Press.
Belletti, A. (1990) Generalized Verb Movement: Aspects of Verb Syntax. Turin:
Rosenberg and Sellier.
Berwick, R. (1985) The Acquisition of Syntactic Knowledge. Cambridge, Mass:
MIT Press,
den Besten, H. (1983) "On the Interaction of Root Transformations and
Lexical Deletive Rules." In W. Abraham, ed., On the Formal Syntax of
the Westgermania, 47-131. Amsterdam: John Benjamins.
Bloomfield, L. (1933) Language. New York: Holt, Reinhart and Winston.
Chomsky, N. (1957) Syntactic Structures. The Hague: Mouton.
Chomsky, N. (1965) Aspects of the Theory of Syntax. Cambridge, Mass.: MIT
Press.
Chomsky, N. (1975) Reflections on Language. New York: Pantheon.
Chomsky, N. (1980) Rules and Representations. New York: Columbia.
Chomsky, N. (1981) Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, N. (1982) Some Concepts and Consequences of the Theory of
Government and Binding. Cambridge, Mass.: MIT Press.
Chomsky, N. (1986a) Knowledge of Language: Its Nature, Origin and Use.
New York: Praeger.
INTRODUCTION 27
For more than a decade generativists have viewed the linguistic genotype or
"Universal Grammar" (UG) as consisting of principles and a set of
option-points or parameters. Correspondingly, language acquisition proceeds
as children set those parameters, sometimes characterized as switches with ON
and OFF positions. And languages change over time as parameters come to
be set differently. The notion of a parameter has stolen the limelight, but
particular parameters are surprisingly evanescent. That is, for only very few
parameters can one point to a solid basis of evidence, a clear understanding
of how a particular parameter-setting interacts with other elements of various
grammars, and a plausible screen-play for how the parameter might come to
be set in the appropriate fashion. Our analytical performances do not match
the script very closely. That script is offering too simplistic a view of lan-
guage acquisition, and the discrepancy between script and performance could
bring us some bad reviews.
Like other authors in this volume, I shall discuss one of the most tantaliz-
ing topics in the current theatre of research, the verb-second phenomenon. I
hope to show that the loss of this phenomenon in the history of English sug-
gests that parameter setting is not a simple matter and that substantial ideas
are needed about how parameters come to be set. If UG consists of principles
and parameters, then one needs a distinct type of theorizing to connect more
directly with acquisition, dealing with what it takes to set parameters; this does
not follow from the nature of the parameters themselves and therefore reflects
a distinct theory of acquisition or, properly construed, a learning theory.
31
32 THE DIACHRONY OF VERB SECOND
lead the child to deduce its ungrammaticality and thus to acquire some spe-
cific device which will block its derivation. This cannot be ruled out in prin-
ciple but there are reasons to be sceptical of appeals to indirect negative data
(Lightfoot (1991, ch.l)). If one is eventually forced to appeal to negative data,
one would expect to find that this kind of deductive learning would emerge
in children somewhat differently than the usual parameter-setting, perhaps later
or after a stage of misgeneralization. Certainly an argument will be required
that children do have indirect access to negative data under, one hopes, nar-
rowly prescribed circumstances.
Nonetheless, one finds several proposals in the extensive verb-second lit-
erature postulating language-specific devices, and in such a way that they
implicitly (but never explicitly, as far as I know) presuppose indirect access
to negative data. For example, the proposals of den Besten (1983), de Haan
and Weerman (1986), Haider (1986), Koopman (1984) and others each postu-
late some element in C which must, given some principle of UG, "attract" I
(or the finite verb) to that position, but the evidence for this element is the
obligatoriness of the movement, i.e. the ungrammaticality of structures where
movement has not taken place. In the absence of an appropriate learning theory
whereby children have access to the appropriate negative data, the proposal
simply re-states the problem. Platzack's (1986) proposal suffers from the same
defect when he postulates different projections in verb-second and non-verb-
second languages. His idea is that S is a projection of I in English and other
languages which do not have verb-second properties, and a projection of C in
verb-second languages. He also allows grammars to differ in terms of whether
the subject NP is base-generated as part of the projection of I or as part of the
projection of C. Then a principle of UG, his "Case assignment rule," forces
a verb to move to C in grammars where the projection of I does not include
the subject NP. This postulates major differences in projection-types and the
evidence for the particular projection-scheme for a verb-second language is
the obligatoriness of I-to-C movement and the ungrammaticality of structures
where I fails to move to C, and again there is no discussion of how the child
would have access to such data. It is unclear how one might treat grammars
(like those of earlier forms of English) allowing verbs to move to C as an
option, and the proposal also fails to account for the necessity of topicalization
in declarative main clauses, a point to be taken up below.1
The same sort of attainability problem arises with the quite different pro-
posals of Safir (1981) and Evers (1982). Safir has UG require that verbs (and
other heads) be uniquely governed, and he makes the structure of German such
that verbs must move to C in main clauses in order for this requirement to be
satisfied. Evers keys movement of the finite verb to the assumption that the
tense element has a scope-bearing property and must c-command the clause;
in embedded clauses the verb does not need to move because the matrix verb
determines the scope of tense. In each case the child has to learn something
specific about verb-second languages: Safir's child must learn the structural
properties which entail that a verb in a base-generated main clause I position
(assuming that verbs move to I in all clause-types) fails to be uniquely gov-
WHY UG NEEDS A LEARNING THEORY 35
erned in the appropriate sense (unlike, for instance, equivalent verbs in French).
Evers' child must learn that C is the only position which c-commands the
clause, unlike in English, French, Italian, etc. In each case the evidence is
not explicitly discussed but it would appear to be the ungrammaticality of struc-
tures where movement to C has not taken place. Also, neither Safir nor Evers
discuss the relationship between movement to C and topicalization.
One solution to the problem of an appropriate triggering experience for the
proposed analysis would be to make the verb-second phenomenon unmarked.
So, if, for example, one postulates that I moves to C by attraction to some
element there, one might argue that generating that element in C reflects the
unmarked situation; English and French children, on the other hand, are ex-
posed to positive data which show that the attractive element is absent, namely
the occurrence of the finite verb in some other position. This might lead one
to expect English and French children to go through an early verb-second stage,
just as Hyams' (1986) English children go through an early pro-drop stage
reflecting the alleged unmarked status of the null-subject option in her analy-
sis. I know no evidence along these lines.
A striking feature of the basic properties of verb-second languages is the
correlation between the topicalization process and the obligatory movement
of I to C. A fundamental shortcoming of many analyses is that while they
offer some account for why the finite verb must be in some C-like position,
they have nothing to say about why some other phrasal category must occur
in initial position. Since the I-to-C movement is a common option in many
languages which are not verb-second, it is unlikely that the obligatoriness of
the movement would entail obligatory topicalization. Indeed, some
verb-subject-object languages seem to have obligatory verb fronting without
any obligatory topicalization; so, for example, Welsh seems to have underly-
ing I-subject-verb-object order and surface verb-subject-object order in all
clause-types by virtue of the verb moving obligatorily to I (Harlow (1981);
Sproat (1985)), but there is no requirement that another phrasal category be
moved forward. It is possible, of course, that the verb must not be just at the
front of the clause but actually in C to entail topicalization, but it is hard to
see why the obligatoriness of the movement should require topicalization, i.e.
why topicalization should not be required in English or French when a verb
happens to be in C.
However, the reverse relationship might be more plausible: topicalization
might entail verb movement, perhaps by a kind of predication requirement.
So Taraldsen (1986), developing work by Cinque (1982) and Kayne (1982),
argued interestingly that a topicalized argument phrase must be locally licensed
and that a verb in C effectively turns the position filled by the topic into an
argument position by yielding a predicate structure; so verb movement has the
effect of providing a local licensing environment for a displaced argument
phrase, which in turn permits its trace to be construed as a variable. He of-
fered some intriguing evidence based on the absence of verb-second effects
with bare w/z-words introducing root interrogatives in certain northern dialects
of Norwegian. Also, it is a theory of the right type in that UG forces move-
36 THE DIACHRONY OF VERB SECOND
ment to C in order to license initial argument phrases; in that case, the child
has to learn that Dutch and German sentences begin with some argument phrase
of arbitrary function. Hellan and Koch Christensen (1986, introduction) noted
that the verb-second phenomenon occurs in those languages which allow long
distance anaphors, which are sometimes argued to be sensitive to predication;
they went on to speculate that the common denominator of these languages
may be some sort of predication sensitivity. However, there are two funda-
mental problems. First, it seems unnatural to construe expletives like Dutch
er and German es as elements requiring predication, just as they cannot be
taken as topics. Second, topicalization does not entail verb movement in lan-
guages that are not verb-second (English, French, etc.). If one keys this to a
structural difference in the position of the topicalized element, then the re-
quirement of a certain kind of predication relationship becomes unclear.
Rizzi (1990a), nonetheless, builds on this idea of verb-second languages
fulfilling a predication requirement. He distinguishes "residual" verb-second
structures which occur in non-verb-second languages, like English,
subject-auxiliary inversion and French subject-clitic inversion, and he com-
pares them with the "full verb-second" phenomenon which determines the
order of constituents in all main (and some embedded) clauses in Dutch, Ger-
man and the Scandinavian languages. The distinction lies in the features of
the head of CP, [+C.-IJ in the case of a residual verb-second construction and
a kind of hybrid [+C,+I] in the case of full verb-second languages (7). Rizzi
interprets the feature +C as "prepositional" and +I as "predicational." Thus a
[+C,-I] category designates a proposition, projecting to the familiar CP of
non-verb-second languages and of non-verb-second clauses in verb-second
languages. [-C.+I] designates a predication, projecting to IP, and [+C.+I] is
the hybrid category characteristic of verb-second languages, being both prepo-
sitional and predicational. For Rizzi this is the category of verb-second clauses
in full verb-second languages. The inflected element [-C.+I] moves to the head
of the CP, necessarily in the case of a full verb-second language because of a
universal principle that the tense specification must c-command all other +1
categories in a given clause.
(7) a. Residual Verb Second
WHY UG NEEDS A LEARNING THEORY 37
The moved element properly governs a subject trace in (7b) under the
"minimality" assumptions of Rizzi (1990b), because the trace is within the
immediate projection of [+I]°, i.e. [+!]'; in (7a) "the moved head and the host
are disjoint feature bundles, hence in no sense can the moved head be said to
govern the subject trace within its (immediate) projection" (p. 385). This ac-
counts for the symmetry of subject-object extractability in full verb-second
languages, while subjects are relatively immobile in non-verb-second languages
where a subject trace would generally fail to be properly governed: this dif-
ference between verb-second and non-verb-second languages was the central
puzzle addressed by Rizzi's paper. Furthermore, different kinds of functional
heads license different kinds of specifiers: so a [+C] head licenses an opera-
tor or trace in an A-bar chain and a [+I] head licenses a subject in the speci-
fier position. So the hybrid case [+C,+I] allows its specifier to be a w/i-phrase
or trace by virtue of being +C (8) or the subject of predication by virtue of
being +1 (9).
(8) a. wer ist [e gekommen Infl]
'who has come?'
b. wer hat Johann gesagt [e ist [e gekommen Infl]]
'who did Johann say has come?'
(9) Maria ist [e gekommen Infl]
'Maria has come'.
A comparable approach, which does not involve notions of predicate for-
mation, was adopted in Lightfoot (1991, ch.3) and can be stated in terms of
constraints on phrase structure. Suppose that Dutch children learn that sen-
tences begin with an arbitrary phrasal category, NP or PP etc., which has no
fixed functional or thematic role. Since the category is utterance-initial and
is not necessarily the complement of an adjacent verb, it cannot be treated as
a complement of some head and must therefore be a specifier, since these are
the only two positions in which a phrasal category may occur in D-structure;
this assumes the familiar strong form of X-bar theory which claims that a head
38 THE DIACHRONY OF VERB SECOND
X projects to a X' which may also contain a complement, and the X' projects
to a XP which may also contain a Specifier. Lexical specifiers always project
to another phrasal category which must have a head, presumably C in this case,
projecting to CP; UG dictates that a lexical specifier requires a lexical head.
Since the initial phrasal category has no fixed thematic or functional role, the
corresponding head cannot be I or any other element associated with a par-
ticular thematic, functional or case-assigning role, and therefore will be an
empty position at D-structure, which is subsequently filled as another head
moves to that position. The only element which is local enough to move to
the empty head position and head-govern its trace, is I (with its associated
verb). Under this approach, principles of UG dictate that an initial XP be
locally licensed in a fashion which in turn requires obligatory movement of a
verb to C. Universal 11 of Greenberg (1966:83) says: "Inversion of state-
ment order so that verb precedes subject occurs only in languages where the
question word or phrase is normally initial. This same inversion occurs in
yes-no questions only if it also occurs in interrogative word questions." If
languages generally do not move verbs to the front in questions unless they
also front interrogative phrases, then it is plausible to interpret the initial phrase
as licensed by the moved verb. If children acquiring a verb-second language
learn that utterances begin with an element of arbitrary (functional or thematic)
role, then it would follow from UG that the inflected verb must occur in sec-
ond position, specifically, let us assume, in C, in order to license the initial
phrasal category. This analysis provides a different answer to Rizzi's puzzle.
(lOa) is well formed in a full verb-second language, because the subject trace
is properly governed by the inflected verb in C (details omitted); but (lOb) is
ill-formed in English because did is not sufficiently lexical to act as a proper
governor for the subject trace.
(10) a.CpLwiei bezoekt ip[eiden Haag] ]
b.cp[whoi did IP[ei visit The Hague] ]
An alternative account, adopted by Aoun, Hornstein, Lightfoot and Weinberg
(1987), would construe (lOb) as a violation of the doubly-filled Comp filter,
which requires Comp to contain no lexical material other than the head. This
filter is "unlearned" in Dutch when children are exposed to structures like
(10a).2
Although current theoretical mechanisms explain certain aspects of the verb-
second phenomenon, some things remain mysterious. Nonetheless children
attain these verb-second languages readily, more readily in fact than children
acquire the "residual" verb-second constructions in English. I have shown in
Lightfoot (1991, ch.3) that simple data from matrix Domains suffice to trig-
ger underlying object-verb order in V for Dutch and underlying verb-object
for Swedish, and children do not need rich or complex data in order to learn
that verbs move and from where they move. A crucial element in this ac-
count was that the X-bar schemata of UG require that verbs occur alongside
their complements at D-structure; therefore Dutch children have ample evi-
dence that verbs move syntactically, because they encounter plenty of utter-
WHY UG NEEDS A LEARNING THEORY 39
ances where verbs and complements are not adjacent (e.g. (2b) above). Fur-
thermore, there is ample evidence from unembedded Domains that Dutch verbs
follow their complements underlyingly; therefore Dutch children do not need
access to embedded Domains to establish the underlying position of the verb.
So the underlying position of the verb is readily attainable by a "degree-0
learner," i.e. a child who sets her parameters only on the basis of simple,
unembedded structures.
Not only is this a possible and theoretically pleasing account, but acquisi-
tional data strongly suggest that something along these lines is correct, that
there are simple, unembedded indicators which enable the child to adopt an
object-verb setting and to posit the relevant verb movement operation. Clahsen
and Smolka (1986) identify four stages in the acquisition of German verb
movement properties (11).
(11) a. stage 1 (25-29 months): no fixed order between sentence con-
stituents; all verbal elements (including verbal complexes) occur
in first/second and final position with a preference for final po-
sition.
b. stage 2 (31-33 months): verbal elements with particles occur
regularly in final position; other finite verbs occur in both first/
second and final position.
c. stage 3 (36-39 months): all and only finite verbs occur in first/
second position; verbal complexes with finite and non-finite parts
appear in discontinuous positions.
d. stage 4 (41-42 months): as soon as embedded sentences are pro-
duced, their finite verbs are in final position.
Strikingly, from the earliest relevant stage, children identify sentence-final
position as one of the possible positions for verbs, including finite verbs de-
spite the fact that they are almost never heard in this position in main clauses.
At stage 3 there is a dramatic increase in the frequency of verb-second struc-
tures: in stages 1 and 2 they are used in only 20-40% of the utterances but at
stage 3 they are used in 90%; Clahsen and Smolka (p. 149) report that this
increase takes place explosively, within a month for all the children studied.
At this stage children seem to have the object-verb D-structure order and an
operation moving a finite verb obligatorily to a C-like position. In simple
clauses there are two positions for verbs: verbal elements with the suffix -t
and modals occur in second position while infinitives and verbs with other
inflections occur sentence-finally. To this extent the adult system is in place
(Clahsen (1990)). Importantly, when they begin to use embedded structures
(stage 4), the finite verbs are invariably in final position and there seems to
be no "experimentation" or learning based on embedded clause data. Clahsen
and Smolka go further and make stronger claims: they take "move V" to
operate from the earliest stage, initially affecting verbs and verbal complexes
of all types, affecting only simple verbs at stage 2, and affecting finite simple
verbs at stage 3. This is exactly what one would expect if children are de-
gree-0 learners, and not at all what one would expect if children were sensi-
40 THE DIACHRONY OF VERB SECOND
there were two major steps in the loss of the old verb-second system and the
evolution of the "residual" system of modern English. We know that English
ceased to be a full verb-second language in Rizzi's sense by the early 15th
century. From this time verbs ceased to occur generally in C, occurring there
only when preceded by certain kinds of elements, particularly a +wh feature.
When the full verb-second properties were first lost, initial negative phrases
also ceased to trigger verb second on any regular basis (compare never did I
see such a mess, etc.). Schmidt (1980:209) claims that verb second declines
in frequency with initial negative phrases, being re-established in the 16th
century. If correct, this suggests that wh- and negative features do not have
to be treated in parallel as triggers for verb-second properties. In any case,
the first step is the restriction of verb-second forms to certain environments,
namely sentences introduced by interrogative (or negative) phrases. In those
environments verbs continued to move to C obligatorily, as in earlier English.
The restriction of verb second to a narrow class of environments seems to have
been complete by the early 15th century.
Modern English, however, shows a further restriction: only certain verbs
occur in verb-second forms, namely the modals, have and be. Since Emonds
(1978), this lexical restriction has been viewed as following from the fact that
modern English lacks the V-to-I operation, raising a verb into a position in
which it acquires various inflectional features (for tense, person, and number).
Modal verbs are base-generated in I and are therefore free to move to C, but
a verb cannot move from its D-structure position directly to C without violat-
ing the Empty Category Principle (its trace would not be properly governed).
The major evidence offered by Emonds and others for this parametric distinc-
tion is that English-type languages without the V-to-I operation have their verbs
strictly adjacent to their complements (14), whereas languages with the V-to-I
operation (like French) allow adverbs, negatives, and floating quantifiers to
intervene (15).
(14) a. *he watches always/never television on Wednesdays
b. *they watch not television
c. *they watch all television on Wednesdays
(15) a. il regarde toujours/seulement la t616vision le mercredi
b. ils regardent pas la television
c. ils regardent tous la television le mercredi
By this criterion, English used to have the V-to-I operation but lost it. One
finds sentences parallel to French (15) (examples from Kroch (1989)):
(16) a. I wende wel thys nyght to have deyed.
Caxton, The Ryall Book, lines 20-25.
'I managed almost tonight to die'.
b. ...if thay do noghte all.
Rolle, "The Bee and the Stork," lines 23-24.
'...if they don't do everything'.
44 THE DIACHRONY OF VERB SECOND
be partially correct, but it cannot be the whole story. We have seen that V-in-C
forms came to be greatly curtailed by the early 15th century but this had no
apparent effect on the V-to-I operation. After the curtailment of V-in-C to
interrogative contexts, V-to-I continued to operate: one still finds forms like
(16) robustly attested, negatives like John spoke not these words, and inter-
rogatives like how great and greuous tribulations suffered the Holy
Appostyls ...?, all of which involve V-to-I (and I-to-C for the interrogatives).
Similarly the loss of general verb second in French had no effect on the at-
tainability of V-to-I, which persists in the present day language even though
V-in-C forms are much rarer than in English (occurring only in interrogatives
with pronominal subject NPs).
This suggests that one needs to look for other changes in the primary lin-
guistic data taking place between 1400 and 1600, which might have the ef-
fect of making V-to-I.harder to attain. Two likely candidates are the demise
of inflectiorial endings on the verb and the rise of periphrastic do. The in-
flectional changes were effectively complete by 1400 and helped to distinguish
shall, may, must etc. as a distinct subclass of lexical items, instances of I rather
than of V (Lightfoot (1991, ch.6)). Therefore, the morphological changes took
place too early to affect the primary linguistic data in the relevant period, and
one cannot correlate V-to-I entirely with morphological properties (cf. Platzack
and Holmberg (1989), who make such a correlation at least for the verb-object
Germanic languages). V-to-I persisted for some time after verb morphology
had become impoverished and English had lost its rich system of subject-verb
agreement.4 Furthermore the mainland Scandinavian languages have lost their
verbal morphology but retain full verb-second properties and, therefore, a V-
to-I operation.5
Periphrastic do, on the other hand, occurred first at the beginning of the
15th century and steadily increased in frequency until it stabilized into its
modern usage by the mid-17th century. This change has been analyzed ex-
tensively and Ellegard (1953) shows that the sharpest increase came in the
period 1475-1550; for discussion and analysis, see Kroch (1989) and Lightfoot
(1991), both of whom reproduce Ellegard's important graph showing the rise
of do in different construction-types. Each insertion of a periphrastic do to
carry inflectional markers represents a case where the V-to-I operation has not
applied, so a steady increase in the distribution of do entails fewer and fewer
instances of V-to-I; the two operations are mutually incompatible.
The historical facts, then, suggest that lack of strong subject-verb agree-
ment cannot be a sufficient condition for absence of V-to-I despite sugges-
tions along these lines by some authors, but it may be a necessary condition.
That is, if a language has strong verbal inflection, it will have V-to-I. Fur-
thermore, if certain individual elements of grammar show strong inflection,
then they undergo V-to-I: so English be is generally inflected richly and
undergoes V-to-I wherever possible (i.e. whenever I is not otherwise lexically
occupied).6 However, there are some forms of English which do not inflect
be, and in those cases be does not raise to I. Thus so-called Black English
uses George be president now, but the usual negative is George don't be presi-
46 THE DIACHRONY OF VERB SECOND
dent now rather than George ben't president and the uninflected be does not
invert with the subject: *be George president now! and *what be George!
(Myhill (1988)). Similarly children often use uninflected forms of be, but they
do not invert them or use them to the left of not: instead they use do-support
forms: did it be funny!, do clowns be a boy or girl!, I don't be angry.
Under this view, the possibility of V-to-I not being triggered first arose in
the history of English with the loss of rich verbal inflection; similarly in
Swedish. That possibility never arose in Dutch, French, German, etc., where
verbal inflection remained relatively robust. Despite this possibility, V-to-I
continued to be triggered and it occurred in grammars well after verbal in-
flection had been reduced to its present day level. Although V-to-I continued
to be triggered, it did not apply obligatorily as it had in Old and Middle En-
glish. There were two alternatives: the use of do as a "dummy" tense carrier
and a morphological operation lowering the tense marker on to the verb ("af-
fix hopping"). As already noted, the first option was exercised with steadily
increasing frequency from the 15th century onwards. It is impossible to know
when the second option of "affix hopping" came into use, because any effects
of affix hopping could also be produced by V-to-I.
However, it is worth noting that affix hopping or tense lowering could not
be a syntactic operation (it would leave an unbound trace) and therefore must
be a morphological operation:7 this entails automatically that it applies only
to contiguous elements (modulo Visser's (1969) transitional John not spoke
those words examples cited above) and that it does not feed other syntactic
operations (compare V-to-I, which applies across intervening adverbs and feeds
I-to-C operations). In that case it is reasonable to suppose that this morpho-
logical operation is generally available, even in grammars which also have the
V-to-I operation, and that V-to-I (in grammars which have the operation) ap-
plies only where necessary, i.e. where there would otherwise be a stranded
affix. We know that "do support" was exercised increasingly during the rel-
evant period, and it seems that it became sufficently frequent that there was
no "need" for V-to-I. That is, with the rise of periphrastic do there was no
longer anything very robust in the primary linguistic data which required
V-to-I, given that the morphological operation was always available. In par-
ticular, post-verbal adverbs and quantifiers (16) were not triggers for V-to-I
and they simply disappeared quietly. Under this analysis, the absence of
V-to-I in modern English grammars is a result of an historical convergence:
at the time that verbal inflections were simplified, verbs underwent a mitosis
whereby a subset (must, shall, etc.) came to be generated under I and there
was in addition an element do which could be analyzed as a "dummy" tense
carrier generated in I and which came to be used more and more frequently.
As a result, V-to-I ceased to be triggered.
Now that we have some idea of how the I-to-C and V-to-I parameters might
be set, we can ask why the "residual" I-to-C operation (or "subject-auxiliary
inversion") in English should be so hard for children to attain. Weinberg's.
(1990) account whereby children must learn to relax the demands of the
doubly-filled Comp filter predicts that Dutch and German children should have
WHY UG NEEDS A LEARNING THEORY 47
even more difficulty, because the filter must be relaxed still further than in
English. This seems to be contrary to the acquisitional facts discussed ear-
lier, which suggest that Dutch and German children acquire the I-to-C opera-
tion relatively early and without the kinds of errors made by English-speaking
children. The difference might arise through Rizzi's full vs residual distinc-
tion: I-to-C is harder to learn because it is restricted to a narrow class of
syntactic environments in English (i.e. to sentences introduced by +wh or a
negative phrase). However, the acquisition of French militates against this
view. Clark (1985) and others have pointed out that young French children
show a preference for verb-subject and even verb-object-subject order over
subject-verb in their early utterances. It is unclear whether such forms should
be analyzed as cases of subject-verb inversion or as right dislocation of the
subject. Whatever the analysis, this preference is surprising because other
studies show that sentences with inverted subjects are quite rare in speech ad-
dressed to children, unlike in English; Lightbown (1977) noted almost total
absence of subject-verb inversion in yes-no questions addressed to children.
Questions with inverted subjects (18a) occur much more in "text-book" French
than in colloquial forms, where people tend to indicate an interrogative by
intonation (18b) or with a wh-in-situ construction (18c) or by an est-ce que
form without inversion (18d).
(18) a. que manges-tu?
b. tu manges la poire?
c. tu manges quoi?
d. quand est-ce qu'il vient?
A common error, noted by Clark (1985) is failure to invert with subject pro-
nouns (19), but children produce forms like (20) from an early age.
(19) ou ils sont?
que ce c'est?
(20) ou est cheval?
Again, it is unclear whether (20) should be analyzed as a case of V-in-C (cf.
ou est le chevall) or as right dislocation of the subject with a missing pro-
noun (cf. ou il est, le chevalT) (Amy Pierce, p.c.). Clark notes various types
of word order errors made by French children but she does not note errors
with inversion comparable to those of English-speaking children. She con-
firms in a personal communication that she has not noted errors of copying
(2la) or of failure to invert with a fronted wh-NP (21b).
(21) a. *vient il/Jean vient a Toulon?
b. *que tu manges?
Although French subject-verb inversion is restricted to +wh contexts as in
English and, even more narrowly, to contexts where the subject NP is pro-
nominal, and although the construction is attested much less robustly in what
children hear, nonetheless French children seem to acquire the I-to-C opera-
48 THE DIACHRONY OF VERB SECOND
tion readily and without the kinds of systematic errors noted in English-
speaking children.
If I-to-C is easy for French children to learn despite being restricted to +wh
contexts with pronominal subjects, then the source of the difficulty experi-
enced by English children presumably lies elsewhere, probably in the fact that
only a small set of "auxiliary" verbs may occur in C: do, be, have and the
modals. It has been argued that these "functional" or "grammatical" items
are opaque to children and are in some sense not perceived at early ages. This
may be true but it would not explain the difficulty in moving them to C at a
time when they occur in medial position, i.e. in I. It seems that it is rela-
tively difficult to attain an operation which is manifested by a small class of
lexical items, even though that operation is attested widely and robustly in
the triggering experience. This would explain why English-speaking children
seem to acquire the I-to-C operation later than Dutch, French and German
children, and usually via systematic errors.
3. Conclusion
We have considered the loss of the verb-second phenomenon in English and
examined the conditions under which the V-to-I operation was lost and the I-
to-C operation was restricted first to +wh contexts and then to a small class
of lexical items. In doing so, we have learned something about what might
trigger these operations in grammars which have them. What emerges is that
the triggering experience may be only distantly related to the data that the
operation immediately accounts for. That is, what triggers an operation in a
child is by no means equivalent to the (positive) data that force a linguist to
postulate that operation. This should not be a surprise: if a grammatical op-
eration were triggered by precisely the positive data that the operation ac-
counted for, one would expect languages to change only by the kind of arbitrary
fluctuations of population genetics; there would be nothing very systematic
and the historical foundations of the discipline laid in the 19th century would
be shown to be weak. It means that a modern historical linguist cannot say
that some changing phenomenon "is due to" the new parameter setting which
accounts for it, although this is often said. Rather, it manifests and provides
evidence for that parameter setting; it is due to changes in the triggering ex-
perience which in turn entailed the new parameter setting. It may be true that
UG consists of principles and parameters, but an account of language acqui-
sition needs to show how those parameters are set. This raises substantive
issues and a substantive "learning theory" is required; that is, we need some
theorizing about what it takes to set a parameter one way or another. I hope
that my discussion of the triggering of the V-to-I operation has provided some
relevant suggestions. In general, I believe that one can learn much about this
from figuring out the conditions under which a parameter comes to be set
differently at some historical stage in a language's development.
WHY UG NEEDS A LEARNING THEORY 49
Notes
* Thanks to Norbert Hornstein and Peter Coopmans for comments on a prelimi-
nary version of this paper. This paper appeared in slightly different form in C.
Jones, ed. Historical Linguistics: Problems and Perspectives, London:
Longman.
7. Tense lowering might leave an unbound (and thus illicit) trace which is subse-
quently erased, for example, by a LF operation raising verbs to I (Chomsky
(1986)). Lightfoot (1991, ch.6) criticizes this approach, which fails to explain
why tense lowering affects only contiguous elements and why unstressed forms
like John did write books (which permit the most economical derivations)
entered the language along with the other periphrastic forms and then disap-
peared rapidly. Emonds (1987) also treats tense lowering as morphological.
References
Aoun, J., N. Hornstein, D. Lightfoot and A. Weinberg (1987) "Two Types of
Locality." Linguistic Inquiry 18:537-578.
Bean, M. (1983) The Development of Word Order Patterns in Old English.
London: Croom Helm.
Berwick, R. (1985) The Acquisition of Syntactic Knowledge. Cambridge,
Mass.: MIT Press.
Besten, H. den (1983) "On the Interaction of Root Transformations and Lexi-
cal Deletive Rules." In W. Abraham, ed. On the Formal Syntax of the
Westgermania, 47-131. Amsterdam: John Benjamins.
Besten, H. den and C. Moed-van Walraven (1986) "The Syntax of Verbs in
Yiddish." In Haider and Prinzhorn, eds., 111-135.
Chomsky, N. (1986) Barriers. Cambridge, Mass.: MIT Press.
Cinque, G. (1982) "Constructions with Left Peripheral Phrases, Subjacency
and ECP." Ms. University of Venice.
Clahsen, H. (1990) "Constraints on Parameter Setting: A Grammatical Analysis
of Some Acquisition Stages in German Child Language." Paper pre-
sented at the Boston University Conference on Language Development.
Clahsen, H. and K.-D. Smolka (1986) "Psycholinguistic Evidence and the
Description of V2 Phenomena in German." In Haider and Prinzhorn,
eds.
Clark, E. (1985) "The Acquisition of Romance, with Special Reference to
French." In D. Slobin, ed. The Cross-Linguistic Study of Language
Acquisition. Hillsdale, NJ: Lawrence Erlbaum Associates.
Davis, H. (1987) The Acquisition of the English Auxiliary System and Its
Relation to Linguistic Theory. PhD Dissertation, University of British
Columbia.
Ellegard, A. (1953) The Auxiliary Do: The Establishment and Regulation of
Its Use in English. Stockholm: Almqvist and Wiksell.
Emonds, J. (1978) "The Verbal Complex V'-V in French." Linguistic Inquiry
9:151-175.
Emonds, J. (1987) "Invisible Category Principle." Linguistic Inquiry 18:613-
632.
Evers, A. (1982) "Twee functionele principes voor de regel 'Verschuif het
werkwoord'." GLOT 5:11-30.
WHY UG NEEDS A LEARNING THEORY 51
Fisiak, J., ed. (1984) Historical Syntax. Trends in Linguistics: Studies and
Monographs, vol. 23. Amsterdam: Mouton.
Gerritsen, M. (1984) "Divergent Word Order Developments in Germanic Lan-
guages: A Description and a Tentative Explanation." In Fisiak, ed., 107-
135.
Greenberg, J. (1966) "Some Universals of Grammar with Particular Reference
to the Order of Meaningful Elements." In J. Greenberg, ed., Universals
of Language. Cambridge, Mass.: MIT Press.
Haan, G. de and F. Weerman (1986) "Finiteness and Verb Fronting in Frisian."
In Haider and Prinzhorn, eds., 77-110.
Haider, H. (1986) "V-Second in German." In Haider and Prinzhorn, eds., 49-
75.
Haider, H. and M. Prinzhorn, eds. (1986) Verb Second Phenomena in Ger-
manic Languages. Dordrecht: Foris.
Harlow, S. (1981) "Government and Relativization in Celtic." In F. Heny, ed.,
Binding and Filtering, 213-255. Cambridge, Mass.: MIT Press.
Hellan, L. and K. Koch Christensen, eds. (1986) Topics in Scandinavian Syn-
tax. Dordrecht: Reidel.
Hyams, N. (1986) Language Acquistion and the Theory of Parameters.
Dordrecht: Kluwer.
Jacobsson, B. (1951) Inversion in English with Special Reference to the Early
Modern English Period. Uppsala: Almqvist and Wiksell.
Jorgensen, N. (1976) Meningsbyggnaden i talad svenska. Lundastudier i
nordisk sprakvetenschap c7. Lund: Studentlitteratur.
Kayne, R. (1982) "Predicates and Arguments, Verbs and Nouns." GLOW News-
letter 8,24.
Kemenade, A. van (1987) Syntactic Case and Morphological Case in the
History of English. Dordrecht: Foris.
Koopman, H. (1984) The Syntax of Verbs: From Verb Movement Rules in the
Kru Languages to Universal Grammar. Dordrecht: Foris.
Kroch, A. (1989) "Reflexes of Grammar in Patterns of Language Use." Lan-
guage Variation and Change 1.3:199-244.
Lightbown, P.M. (1977) Consistency and Variation in the Acquisition of
French; A Study of First and Second Language Development. PhD
Dissertation, Columbia University, NY.
Lightfoot, D.W. (1991) How to Set Parameters: Arguments from Language
Change. Cambridge, Mass.: MIT Press.
Myhill, J. (1988) "The Rise of be as an Aspect Marker in BE Vernacular."
American Speech 63:304-326.
Newport, E., H. Gleitman and L. Gleitman (1977) "Mother, I'd Rather Do It
Myself: Some Effects and Non-effects of Maternal Speech Style." In
C. Snow and C. Ferguson, eds., Talking to Children: Language Input
and Acquisition. Cambridge: Cambridge University Press.
Paardekooper, P.C. (1971) Beknopte ABN-syntaxis. Den Bosch: L.C.G.
Malmberg.
52 THE DIACHRONY OF VERB SECOND
Platzack, C. (1986) "COMP, INFL and Germanic Word Order." In Hellan and
Koch Christensen, eds., 185-234.
Platzack, C. and A. Holmberg (1989) "The Role of AGR and Finiteness in
Germanic VO Languages." Working Papers in Scandinavian Syntax
43:51-76.
Richards, B.J. (1990) Language Development and Individual Differences: A
Study of Auxiliary Verb Learning. Cambridge: Cambridge University
Press.
Rizzi, L. (1990a) "Speculations on Verb Second." In J. Mascard and M. Nespor,
eds., Grammar in Progress; GLOW Essays for Henk van Riemsdijk,
375-385. Dordrecht: Foris.
Rizzi, L. (1990b) Relativized Minimality. Cambridge, Mass.: MIT Press.
Safir, K. (1981) "Inflection-Government and Inversion." The Linguistic Re-
view 1:417-467.
Schmidt, D. (1980) A History of Inversion in English. PhD Dissertation, The
Ohio State University.
Sproat, R. (1985) "Welsh Syntax and VSO Structure." Natural Language and
Linguistic Theory 3.2:173-216.
Stockwell, R. (1984) "On the History of the Verb-Second Rule in English."
In Fisiak, ed., 575-592.
Stowell, T. (1981) Origins of Phrase Structure. PhD Dissertation, MIT.
Taraldsen, K.T. (1986) "On Verb Second and the Functional Content of Syn-
tactic Categories." In Haider and Prinzhorn, eds., 7-26.
Travis, L. (1984) Parameters and Effects of Word Order Variation. PhD
Dissertation, MIT.
Vance, B. (1990) "Inversion and Pro-drop in Middle French." Paper presented
at the First Generative Diachronic Syntax Conference.
Visser, F. Th. (1969) An Historical Syntax of the English Language. Vol. IIIa.
Leiden: E.J. Brill.
Weinberg, A. (1990) "Markedness Versus Maturation: The Case of Subject-
Auxiliary Inversion." Language Acquisition 1:165-194.
3
Two Types of Verb Second
in the History of Yiddish
Beatrice Santorini
Northwestern University
As is well known, all the Germanic languages except modern English exhibit
the verb-second phenomenon, as do several (mostly medieval) varieties of
Romance. According to this word order constraint, the inflected verb appears
as the second overt constituent of a clause—regardless of whether the first
constituent is the subject. Thus, in a verb-second language, clauses in which
the first constituent is not the subject, exhibit obligatory subject-verb inver-
sion. The earliest and best-known generative analyses of the verb-second phe-
nomenon (Thiersch (1978); den Besten (1983)) were formulated to describe
languages like German and Dutch, where root clauses exhibit verb second, but
formally subordinate clauses (that is, embedded clauses introduced by an overt
complementizer or wft-phrase) do not. Since the phrase structure of these
languages is verb-final, the root/subordinate asymmetry they display with re-
gard to the position of inflected verbs is striking; it is illustrated for German
in (1) and (2) ("Vf" refers to the inflected verb).
(1) a. Root clause, subject-initial—Vf in second position:
Der Junge wird auf dem Weg eine Katze sehen.
the boy will on the way a cat see
'The boy will see a cat on the way'.
53
54 THE DIACHRONY OF VERB SECOND
1. Theoretical Assumptions
1.1 The Landing Site of Verb Movement in Verb-Second Clauses
Given their focus on asymmetric verb-second languages, early generative
studies proposed to derive the position of the inflected verb in verb-second
clauses by verb movement to Comp. Since Comp in formally subordinate
clauses is filled by a complementizer, it is not available for substitution by
the inflected verb—providing an elegant analysis of the root/subordinate asym-
metry illustrated in (1) and (2). The clause-initial constituent in a verb-sec-
ond clause occupies a non-thematic position immediately preceding the
inflected verb in Comp. Adopting current assumptions regarding X' theory
(Chomsky (1986:6)), the phrase structures for the verb-second clauses in (1)
are then as shown in (5).
56 THE DIACHRONY OF VERB SECOND
subjects in such a language need not move from their underlying position in
order to identify Agr, they are forced to move by the considerations of
minimality just described for Dutch and mainland Scandinavian. Finally,
consider the case where Agr is inherently nominative and the highest [+1] cat-
egory is Infl, as is the case in modern Yiddish (and at least for some speak-
ers, in Icelandic). This is the only case in which subjects are able to remain
in their underlying position, since they are assigned nominative case there and
do not need to raise to SpecIP in order to identify Agr. In such languages,
then, non-subjects are free to occupy SpecIP. Independent support for the
analysis just presented comes from the correlation between the loss of em-
bedded verb second and the loss of overt case-marking on noun phrases in
Old French (Cardinaletti and Roberts (1991:55f.)).10 Table 3.1 summarizes
the above discussion concerning nominative-case assignment.
Table 3.1
Distribution of Non-subjects in SpecIP in Various Languages
might reflect underlying Infl-medial phrase structure, but might just as well
reflect the combination of Infl-final phrase structure and rightward movement
processes that are independently motivated in early Yiddish (namely, verb
raising and PP extraposition).
(13) dz zi verdn bshirrnt fun irh bitrh peyn
that they become protected from their bitter pain
'that they might be protected from their bitter pain'
(Purim-shpil 876, 1697)
For each time period, dialect and style, I determined / (the number of un-
ambiguously Infl-final subordinate clauses) and m (the number of unambigu-
ously Infl-medial subordinate clauses) and calculated the relative frequency
of unambiguously Infl-medial subordinate clauses (defined as m/ (f + m)).15
Figure 3.1 shows the rise of Infl-medial phrase structure over time in West
Yiddish. It is evident that Infl-medial phrase structure is a marginal option in
West Yiddish vernacular texts. For those time periods for which we have both
vernacular and literary texts—up to the late 1600s—all but two literary texts
closely reflect the vernacular ones, with the percentage of Infl-medial subor-
dinate clauses never exceeding 5%.16 After about 1700, Infl-medial subordi-
nate clauses in West Yiddish literary texts rise in relative frequency, though
they never come to make up the majority of subordinate clauses. Given the
attested close relationship in the earlier periods of West Yiddish between lit-
erary and vernacular usage, I take this development to reflect a rise in Infl-
medial phrase structure in the vernacular.
Figure 3.2 gives the corresponding results for East Yiddish. Two aspects of
these results are noteworthy. First, in contrast to West Yiddish literary texts,
East Yiddish literary texts do not faithfully reflect vernacular usage for the
period before 1700 for which both literary and vernacular texts are available.
Infl-medial subordinate clauses are already well attested in East Yiddish ver-
nacular texts from before 1700, but they are still marginal in the literary texts
from the same period, where they occur with a frequency comparable to that
found for West Yiddish. Second, after about 1700, the great majority of struc-
turally unambiguous subordinate clauses in the literary texts have suddenly
become Infl-medial. The first fully modem texts (that is, ones without any
Infl-final subordinate clauses at all) date back to the early 1700s, though the
change from Infl-final to Infl-medial phrase structure takes roughly another
century to go to completion, with Infl-final phrase structure surviving into the
first half of the 1800s as an increasingly marginal option in literary East
Yiddish.17
The quantitative differences between West and East Yiddish evident in Fig-
ures 3.1 and 3.2 are consistent with and in turn confirm conclusions reached
by traditional philologists like Max Weinreich who have studied the history
of Yiddish phonology and morphology within its wider social context. Ac-
cording to Weinreich (1980), the Jews of western and eastern Europe belonged
to a single cultural community until the 1700s. The linguistic correlate of this
unity was the existence of a fairly uniform supraregional literary standard based
on West Yiddish, which Weinreich calls "Written Language A." While Writ-
ten Language A remained essentially impervious to Slavic influence even in
eastern Europe, vernacular Yiddish gradually developed into the two main
dialects of West and East Yiddish, the earliest reports of differences between
64 THE DIACHRONY OF VERB SECOND
the two dialects date back to the beginning of the 1600s, and by the middle
of the 1700s, they had diverged in speech to the point of causing difficulties
in mutual comprehension (Weinreich (1980:284)). This growing rift between
West and East Yiddish led to the emergence in the course of the 1700s of a
literary language based on East Yiddish vernacular usage—what Weinreich
calls "Written Language B." It is this replacement of Written Language A by
Written Language B that we see reflected in the quasi-instantaneous rise of
Infl-medial subordinate clauses in Figure 3.2.
4. Conclusion
The diachronic data discussed in this paper raise an intriguing and difficult
question—namely, whether the loss of hybrid Comp can be considered a nec-
essary consequence of the emergence of Infl-medial phrase structure. In or-
der to address this question, let us ask ourselves what evidence a child
acquiring an Infl-medial verb-second language has for positing hybrid Comp.
The first source of evidence comes from root/subordinate asymmetries. In Infl-
medial verb-second languages without verb movement to Infl.like the mod-
ern mainland Scandinavian languages, the failure of the verb to move to Infl
in subordinate clauses containing sentence negation or sentence adverbials
results in verb-third word orders that are inconsistent with verb second. The
68 THE DIACHRONY OF VERB SECOND
word order asymmetry between such verb-third subordinate clauses and verb
second root clauses therefore provides evidence that verb second in these lan-
guages involves verb movement to a hybrid Comp. By contrast, in an Infl-
medial language with verb movement to ML, like Yiddish, the root/subordinate
asymmetry resulting from verb-third subordinate clauses does not arise. It
might therefore be argued that although the first Infl-medial clauses in Yid-
dish may have been produced by a grammar with hybrid Comp, they would
provide children with no positive evidence for it and that at least some chil-
dren would therefore analyze Infl in Infl-medial clauses as the highest [+I]
category. Indeed, it might be argued further that in the absence of evidence
for verb movement to Comp in Infl-medial clauses, economy of derivation
(Chomsky (1991)) would force all children to acquire a grammar without
hybrid Comp. The appearance of lexical expletives and non-subjects in SpecIP
in embedded clauses would then follow directly, given the inherently nomi-
native character of Agr in Yiddish. The attraction of this argument is that it
relates the generalization of verb second to embedded contexts in Yiddish to
the emergence of Infl-medial phrase structure (cf. Vikner (1994)). However,
it neglects to take into account the second source of positive evidence for
hybrid Comp in Infl-medial clauses—namely, the existence of gap-Vf clauses.
Given this evidence, we would expect to find languages just like modern Yid-
dish—that is, with Infl-medial phrase structure, verb raising to Infl and overt
case morphology on full NP's—except that they do not allow lexical exple-
tives or non-subjects in SpecIP.
West Yiddish is a candidate for such a language, though the evidence is
murkier than one would like. As we saw in SECTION 2.3, West Yiddish exhib-
ited some Infl-medial phrase structure, but did not in general exhibit embed-
ded verb second. One exceptional text—the Purim-shpil of 1697—does
contain Vf-second subordinate clauses with lexical expletives in SpecIP.
However, it is not clear to what the occurrence of these clauses is attributed.
On the one hand, the scribe who copied the manuscript of the Purlm-shptl from
an earlier West Yiddish version, now lost, was from Cracow, a transition zone
between West and East Yiddish (Weinryb (1936:417)), and the occurrence of
embedded verb second in the manuscript might therefore reflect interference
from the scribe's native dialect. On the other hand, of all the West Yiddish
texts in the corpus, the Purim-shpil contains the largest absolute number of
subordinate clauses that are unambiguously Infl-medial as well as the largest
number that are consistent with Infl-medial phrase structure. Since embed-
ded verb second is not common even in languages whose grammars allow it,
the presence of embedded verb second in the Purim-shpil and its absence in
the other West Yiddish texts might simply reflect the statistically expected
values for each text. Clearly, further evidence from other late West Yiddish
texts would be needed to decide the issue. Unfortunately, however, since
written West Yiddish was being replaced by German during precisely this
period, such texts are rare, and I do not know of any besides the ones that I
have investigated. While the case of West Yiddish brings us up against the
limits of the historical record, there is synchronic evidence from insular Scan-
TWO TYPES OF VERB SECOND IN THE HISTORY OF YIDDISH 69
Appendix
Table 3.2 provides statistics on the subordinate clauses in my corpus. For
each text (or group of texts), the columns list the number of subordinate clauses
with the following properties:
1. unambiguously Infl-final (more than one constituent precedes the in-
flected verb)
2. unambiguously Infl-medial (instance of stranding or apparent verb pro-
jection raising)
3. overt subject or trace of local subject extraction in headed relative clause
in first position, Vf in second position
70 THE DIACHRONY OF VERB SECOND
Table 3.2
West Yiddish Vernacular Texts
Notes
It is a pleasure to thank the following people for their help in writing this
paper: Caroline Heycock, Anthony Kroch, Janet Pierrehumbert, Bernhard
Rohrbacher and Sten Vikner for helpful discussion, Ian Roberts for making
me work faster than I ever thought possible, and Sue Johnson, Beth Levin and
Adair Waldenburg for their refreshing senses of humor and perspective. Need-
less to say, I am responsible for all errors and shortcomings that remain.
1. For the purposes of this paper, two periods need to be distinguished in the
history of Yiddish: "Early Yiddish" (from the earliest texts dating back to the
1400s until about 1800) and "Modern Yiddish" (from about 1800 on). For a
more detailed periodization, see Weinreich (1980).
1. Yiddish verb phrases also appear to have changed from head-final to head-
initial. Further research is needed to resolve this issue, since modern Yiddish
still allows OV word order fairly productively. For discussion, see den Besten
and Moed-van Walraven (1986), GeilfuB (1991), Hall (1979), and Santorini
(1993b).
3. West Yiddish was spoken on German-speaking territory and in transition ar-
eas in contact with Slavic, and East Yiddish was originally spoken on Slavic-
speaking territory only. The earliest continuous West Yiddish texts date from
the late 1300s, and the language has been essentially extinct since 1800, at
least in written form, having given way to German. The West Yiddish sources
on which this paper is based cover a period from the early 1400s to the mid-
1700s. East Yiddish texts are more recent than West Yiddish ones, and the
oldest ones that I have examined date from the early 1500s. East Yiddish
continues to be spoken and written today both in eastern Europe and by the
descendants of eastern European Jews who migrated abroad, particularly to
North and South America. Since West Yiddish is for all practical purposes
extinct, I will generally refer to Modern East Yiddish simply as "Modern
Yiddish."
4. Arguing against Vikner (1991), Cardinaletti and Roberts (1991:6) raise the
objection that analyzing embedded verb-second clauses in symmetrical verb-
second languages as instances of CP-recursion provides "no way to avoid
unlimited recursion of C°." However, since unlimited recursion of Comp
must be avoided in any event, even in asymmetric verb-second languages
like German or residual verb-second languages like English, it is difficult to
see what would preclude extending whatever mechanism is invoked to rule it
out in those cases—say, a distinction between two types of Comp, as sug-
gested by Cardinaletti and Roberts themselves—to the case of symmetrical
verb-second languages.
5. The VP-internal subject hypothesis was originally proposed by Fillmore (1968)
and McCawley (1970), and much recent work in phrase structure theory has
been devoted to arguing in favor of it (Fukui (1986); Fukui and Speas (1986);
Kitagawa (1986); Koopman and Sportiche (1991); Kuroda (1987); Manzini
(1988); Sportiche (1988); Zagona (1988)). Several different variants of the
TWO TYPES OF VERB SECOND IN THE HISTORY OF YIDDISH 73
hypothesis have been proposed and discussed in the literature. For the pur-
poses of this paper, the precise position in which subjects originate is irrel-
evant as long as it is dominated by a maximal projection of the verb, and I
will simply refer to the position in question as the subject's "underlying"
position.
6. Kosmeijer (1991) puts forward a similar analysis, taking the finiteness opera-
tor [+F] to be a feature of Comp in German, but of Infl in Icelandic.
7. The analysis of nominative-case assignment proposed here is inspired by that
which is put forward by Platzack and Holmberg (1990), according to whom
nominative case in the verb-second languages is assigned under government
by the finiteness operator [+F]. However, there is an important difference
between their and my analysis with regard to the notion of government: ac-
cording to Platzack and Holmberg (1990:7(9)), governors m-command their
governees, whereas according to the definition in (8), they must satisfy the
stricter condition of c-command; cf. Deprez (1989:366ff.), Koopman and
Sportiche (1991:229f.) and Rizzi (1990a:30-32) for relevant discussion. As
we will see in SECTION 3.1, the distribution of empty expletives in modern
Yiddish provides empirical justification for a c-command over an m-com-
mand clause in the definition of government.
8. Platzack and Holmberg's distinction between nominative and neutral Agr has
roughly the same empirical consequences as the distinction made by Koopman
and Sportiche (1991) between nominative-case assignment under Spec-head
agreement and under government.
9. Platzack and Holmberg's idea that embedded verb second is related to rich
subject-verb agreement, incorrectly leads one to expect embedded verb sec-
ond in a language like Italian. The unacceptability of embedded verb second
in Italian is a serious difficulty for their approach and supports the case-mark-
ing approach instead.
10. Old Spanish is problematic for my analysis, since it allows embedded verb
second but fails to exhibit overt case-marking on full NP's (Josep Fontana,
p.c.).
11. Synchronic variation between Infl-final and Infl-medial phrase structure is
also attested in Old English texts (Pintzuk (1991) contra van Kemenade
(1987)).
12. The first and second numbers following each reference indicate the page,
verse or line number and the year of the text, respectively. Where no page
number is available, I indicate this by "n.p."; where the exact date is un-
known, I give a range of dates or an estimate from the secondary literature. In
contrast to modern Yiddish, vowels are not consistently represented in early
Yiddish, and I have not attempted to insert them where they are absent in the
original texts; the transliteration conventions I follow are set out in Santorini
(1989:15-17).
13. Verb projection raising is a variant of verb raising, a common and much-
studied phenomenon in West Germanic. Verb raising permutes the order of
auxiliary verbs and the heads of their infinitival complements. In verb projec-
tion raising, the post-auxiliary sequence includes arguments and adjuncts of
74 THE DIACHRONY OF VERB SECOND
the infinitive head in addition to the infinitive itself. See Evers (1975), Lotscher
(1978), Zaenen (1979), den Besten and Edmondson (1983), Haegeman and
van Riemsdijk (1986), and Kroch and Santorini (1991) for discussion.
14. See Santorini (1992:614-616) for more detailed discussion.
15. The raw numbers on which Figures 3.1 and 3.2 are based are given in the
Appendix.
16. The relative frequency of Infl-medial subordinate clauses in the two excep-
tional texts is comparable to East Yiddish vernacular usage of the same pe-
riod. Their author was from Cracow, a transition zone between West and East
Yiddish, and they are both prefaces to other works (one of these texts is by the
same author but contains hardly any Infl-medial subordinate clauses). A plau-
sible explanation for the unusually high incidence of Infl-medial phrase struc-
ture is that the usage of the author in prefaces tends to reflect his vernacular
usage.
17. Modern Yiddish still preserves some relics of Infl-final phrase structure—for
instance, the negative polarity idiom vos a/der hor vert iz 'in the slightest bit'
(literally, 'what a/the hair is worth'). Their status is comparable to that of
such relics of verb-final phrase structure in modern English as Indictments do
not a conviction make.
18. The only gap-Vf clauses that have remained acceptable in modern Yiddish
are headed relative clauses with local extraction of the subject. In such clauses,
SpecIP may exceptionally be filled by an empty category—the trace of the
subject. I will not attempt to give an analysis here of what licenses the empty
initial position in these clauses. It is worth noting, however, that the excep-
tional character of local subject extraction is not peculiar to Yiddish, as is
evidenced by the violations of the that-trace filter in precisely this context in
English; cf. Deprez (1989, ch.4) for discussion.
19. In addition to the formal licensing condition in (16), empty expletives must
also satisfy an identification requirement regarding their feature content (cf.
Rizzi 1986:520(41)). I do not wish to deny that the availability of empty
expletives in Yiddish might be related to properties like the "richness" of
subject-verb agreement; however, the details of the identification requirement
are not at issue here.
20. Without the postverbal es, the sentences in (20) are all acceptable.
21. Kosmeijer (1991) discusses parallel facts in Icelandic and German. For an
analysis of the complementary distribution of empty and lexical expletives,
see Cardinaletti (1990).
22. As in SCCTION 2.3,1 have excluded from consideration all Vf-second instances
of w/z-movement or subject postposing; the clauses under consideration are
therefore all bonafide instances of movement to SpecIP and cannot be ana-
lyzed as instances of stylistic fronting. Stylistic fronting is a process attested
in Icelandic, Faroese and Medieval Scandinavian in which a non-subject con-
stituent occupies clause-initial position, resulting in word orders that are su-
perficially consistent with verb second. Since stylistic fronting is restricted
to subject-gap clauses (Maling (1990:77-81)) and affects heads rather than
maximal projections, there is reason to distinguish it from movement of non-
TWO TYPES OF VERB SECOND IN THE HISTORY OF YIDDISH 75
References
Adams, M.P. (1.988) "Embedded pro." In J. Blevins and J. Carter, eds. Pro-
ceedings ofNELS 18, 1-21. Amherst, Mass.: GSLA.
Barnes, M. P. (1987) "Some Remarks on Subordinate-Clause Word-Order in
Faroese." Scripta Islandica 38:3-35.
Besten, H. den (1983) "On the Interaction of Root Transformations and Lexi-
cal Deletive Rules." In W. Abraham, ed. On the Formal Syntax of the
Westgermania, 47-131. Amsterdam: John Benjamins. Draft version cir-
culated in 1977.
Besten, H. den and J.A. Edmondson (1983) "The Verbal Complex in Conti-
nental West Germanic." In W. Abraham, ed. On the Formal Syntax of
the Westgermania, 155-216. Amsterdam: John Benjamins.
Besten, H. den and C. Moed-van Walraven (1986) "The Syntax of Verbs in
Yiddish." In H. Haider and M. Prinzhorn, eds. Verb Second Phenom-
ena in Germanic Languages. Publications in Language Sciences 21:111-
135. Dordrecht: Foris.
Cardinaletti, A. (1990) Pronomi nulli e pleonastici nelle lingue germaniche e
romanze: Saggio di sintassi comparata. PhD Dissertation, Universiti
di Padova e Venezia.
Cardinaletti, A. and I. Roberts (1991) "Clause Structure and X-second." To
appear in W. Chao and G. Horrocks, eds. Levels, Principles and Pro-
cesses: the Structure of Grammatical Representations. Berlin: Foris/de
Gruyter.
Chomsky, N. (1986) Barriers. Cambridge, Mass.: MIT Press.
Chomsky, N. (1991) "Economy of Derivation and Representations." In R.
Freidin, ed. Principles and Parameters in Comparative Grammar, 417-
454. Cambridge, Mass.: MIT Press.
Chung, S. and J. McCloskey (1987) "Government, Barriers and Small Clauses
in Modern Irish." Linguistic Inquiry 18:173-237.
Deprez, V. (1989) On the Typology of Syntactic Positions and the Nature of
Chains: Move a to the Specifier of Functional Projections. PhD Dis-
sertation, MIT.
Diesing, M. (1988) "Word Order and the Subject Position in Yiddish." In
J. Blevins and J. Carter, eds. Proceedings of N ELS 18, 124-140.
Amherst, Mass.: GSLA.
76 THE DIACHRONY OF VERB SECOND
Diesing, M. (1990) "Verb Second in Yiddish and the Nature of the Subject
Position." Natural Language and Linguistic Theory 8:41-79.
Dupuis, F. (1989) L'expression du sujet dans les subordonnees en ancien
francais. PhD Dissertation, Universite de Montreal.
Emonds, J.E. (1978) "The Verbal Complex V'-V in French." Linguistic Inquiry
9:151-175.
Evers, A. (1975) The Transformational Cycle in Dutch and German. PhD
Dissertation, University of Utrecht. Distributed by the Indiana Univer-
sity Linguistics Club.
Fillmore, CJ. (1968) "The Case for Case." In E. Bach and R.T. Harms, eds.
Universals in Linguistic Theory, 1-88. New York: Holt, Rinehart and
Winston.
Fontana, J.M. (1993) Phrase Structure and the Syntax of Clitics in the His-
tory of Spanish. PhD Dissertation, University of Pennsylvania.
Fukui, N. (1986) A Theory of Category Projection and its Applications. PhD
Dissertation, MIT.
Fukui, N. and M. Speas (1986) "Specifiers and Projection." In N. Fukui, T.
Rapoport and E. Sagey, eds. MIT Working Papers in Linguistics 8:128-
172.
GeilfuB, J. (1991) "Jiddisch als SOV-Sprache." In Verb- und Verbphrasen-
syntax, 3-17. Arbeitspapiere des Sonderforschungsbereichs 340:
"Sprachtheoretische Grundlagen fur die Computerlinguistik," Bericht Nr.
11. Also in Zeitschrift fur Sprachwissenschaft 9 (1990).
Haan, G. de and F. Weerman (1986) "Finiteness and Verb Fronting in Frisian."
In H. Haider and M. Prinzhorn, eds. Verb Second Phenomena in Ger-
manic Languages, Publications in Language Sciences 21:77-110.
Dordrecht: Foris.
Haegeman, L. and H. van Riemsdijk (1986) "Verb Projection Raising, Scope
and the Typology of Rules Affecting Verbs." Linguistic Inquiry 17:417—
466.
Haider, H. (1989) "0-Tracking Systems—Evidence from German." In
L. Maracz and P. Muysken, eds. Configurationality. The Typology of
Asymmetries. Studies in Generative Grammar 34:185-206. Dordrecht:
Foris.
Hall, B. L. (1979) "Accounting for Yiddish Word Order, or What's a Nice NP
Like You Doing in a Place Like This?" In J.M. Meisel and M.D. Pam,
eds. Linear Order and Generative Theory. Amsterdam Studies in the
Theory and History of Linguistic Science IV. Current Issues in Linguis-
tic Theory 7:263-287. Amsterdam: John Benjamins.
Hirschbiihler, P. and M.-O. Junker (1988) "Remarques sur les sujets nuls en
subordonnees en ancien et en moyen francais." Revue quebecoise de
linguistique theorique et appliquee 7:63-84.
Holmberg, A. (1986) Word Order and Syntactic Features in the Scandinavian
Languages and English. PhD Dissertation, University of Stockholm.
Holmberg, A. and C. Platzack (1988) "On the Role of Inflection in Scandina-
vian Syntax." Working Papers in Scandinavian Syntax 42:25-42.
TWO TYPES OF VERB SECOND IN THE HISTORY OF YIDDISH 77
latridou, S. and A.S. Kroch (1992) "The Licensing of CP-Recursion and its
Relevance to the Germanic Verb-Second Phenomenon." Working Papers
in Scandinavian Syntax 50:1-24.
Jonsson, J.G. (1991) "Stylistic Fronting in Icelandic." Working Papers in Scan-
dinavian Syntax 48:1-43.
Kemenade, A. van (1987) Syntactic Case and Morphological Case in the His-
tory of English. Dordrecht: Foris.
Kitagawa, Y. (1986) Subject in Japanese and English. PhD Dissertation, Uni-
versity of Massachusetts, Amherst.
Koopman, H. (1984) The Syntax of Verbs. From Verb Movement Rules in the
Kru Languages to Universal Grammar. Dordrecht: Foris.
Koopman, H. and D. Sportiche (1991) "The Position of Subjects." Lingua
85:211-258.
Kosmeijer, W. (1991) "The Distribution of bad: an Argument Against a Gen-
eralized COMP Analysis." In H.A. Sigur5sson, ed. Papers from the 12th
Scandinavian Conference of Linguistics, 182-193. Reykjavik: Linguis-
tic Institute, University of Iceland.
Koster, J. (1975) "Dutch as an SOV Language." Linguistic Analysis
1:111-136.
Kroch, A.S. and B. Santorini (1991) "The Derived Constituent Structure of
the West Germanic Verb Raising Construction." In R. Freidin, ed. Pro-
ceedings of the Princeton Workshop on Comparative Grammar, Current
Studies in Linguistics 20:269-338. Cambridge, Mass.: MIT Press.
Kuroda, S.-Y. (1988) "Whether We Agree or Not: A Comparative Syntax of
English and Japanese." In WJ. Poser, ed. Papers from the Second
International Workshop on Japanese Syntax, 103-143. Stanford: CSLI.
Laka, I. (1989) "Constraints on Sentence Negation: The Case of Basque." MIT
Working Papers in Linguistics 10:199-216.
Lotscher, Andreas (1978) "Zur Verbstellung im Zurichdeutschen und in anderen
Varianten des Deutschen." Zeitschrift fur Dialektologie und Linguistik
45:1-29.
Maling, J. (1980) "Inversion in Embedded Clauses in Modern Icelandic."
Islenskt mal og almenn mdlfrceSi 2:175-193. Reprinted in J. Maling and
A. Zaenen, eds. Modern Icelandic Syntax. Syntax and Semantics 24:71-
91. San Diego: Academic Press.
Manzini, M.R. (1988) "Constituent Structure and Locality." In A. Cardinaletti,
G. Cinque, and G. Giusti, eds. Constituent Structure, 157-201.
Dordrecht: Foris.
McCawley, J.D. (1970) "English as a VSO Language." Language 46:286-299.
Pintzuk, S. (1991) Phrase Structures in Competition: Variation and Change
in Old English Word Order. PhD Dissertation, University of Pennsyl-
vania.
Platzack, C. (1986) "The Position of the Finite Verb in Swedish." In H. Haider
and M. Prinzhorn, eds. Verb Second Phenomena in Germanic Lan-
guages. Publications in Language Sciences 21:27-47. Dordrecht: Foris.
78 THE DIACHRONY OF VERB SECOND
1. Introduction
In this article, we discuss the derivation of the verb-second phenomena in Old
French and Middle French (hereafter OF and MidF) in a model of grammar
where the subject is generated internal to the VP and where the Spec of IP
(or AgrP) is either an A or an A-bar position. We will provide arguments for
a V-to-I derivation of this phenomena for both main and embedded clauses
arguing that CP will be projected only when necessary (cf. Diesing (1990)).
In the spirit of Chomsky (1989), a V-to-I derivation of the verb-second con-
straint corresponds to the minimal derivation, and thus enables us to minimize
the function of transformations. We propose an analysis where verb second
has not been lost as a rule but is a consequence of the interaction of very
general principles of grammar. One important feature of this article will be
to give an explicit characterization of a small class of discursive elements trig-
gering verb second in root clauses in order to provide some new insights into
the putative asymmetry between main and embedded clauses.
Our analysis will be based on the following premises: (i) It will be shown
that MidF, like OF, exhibits the main characteristics of a verb-second language,
and that the theory that the subject is base-generated inside VP makes inter-
esting predictions for both stages of the language; (ii) contrary to what has
been proposed so far (Adams (1987a,b, 1988a,b); Vance (1988); Roberts
(1992)) we will suggest that the licensing of the subject in verb-second clauses
does not involve double verb raising to CP, that is V-to-I and V-to-C. More
importantly, we will show that the facts addressed in this paper entail theo-
retical assumptions concerning the function of some illocutionary morphemes
involved in verb second.
The discussion will proceed as follows: SECTION 2 presents the data illus-
trating the verb-second characteristics of OF and MidF. SECTION 3 focuses on
80
THE LOCUS OF VERB MOVEMENT 81
Table 4.1
Verb Second in Middle French
Texts XPV Nom. Subj. V Pron. Subj. V Subtotal % Total
The second and the third column indicate the number of preverbal subjects,
nominal or pronominal, immediately preceding the tensed verb. The residual
cases represent violations of the verb-second property. We will discuss the
importance of such data in SECTION 5.
Leaving aside for the moment the details which we will take up in SECTION
3, we will posit, along the lines of Diesing (1990) (see also Rognvaldsson and
Thr&nsson (1990)) that, in the model of grammar where the subject is gener-
ated internal to the VP and where the Spec of IP is either an A or an A-bar
position, examples (l)-(3) are derived via a single verb movement to 1° and
via the movement of some XP into the Spec of IP as shown in (4):
Yiddish
(5) a. Ikh wel avekshikn dos bukh.
I will away-send the book (Diesing (1990), example (Id))
According to Diesing, the fact that Yiddish does not show the same distribu-
tion as German is to be explained as a parametric choice for the landing site
of the verb: V-to-I rather than V-to-C. A second parameter is also respon-
sible for the dual nature of [Spec.IP]: this position is able to function either
as an A-bar position as or an A position.
3.3 The Case of Old and Middle French and the So-called
Asymmetry
While numerous analyses of the verb-second phenomenon have been pro-
vided since Thiersch (1978), most linguists in the generative framework give
an account of the root/embedded asymmetry by a double movement of the
tensed verb: to the head of IP, then to the head of CP. Adams (1987a,b) has
shown that this analysis makes interesting predictions for OF and MidF. The
THE LOCUS OF VERB MOVEMENT 87
One important consequence of this analysis is that the strict locality require-
ment necessary for subject-verb agreement to take place is not fulfilled. As
the verb raises up to amalgamate with the inflectional features, it goes further
away from the base subject. Hence the government relation for agreement
cannot be met in any way since the agreement features are outside the VP
projection. Anticipating a little at this point, let us suggest that the agreement
relation is mediated through a pro in the Spec position of TP.
Consider now the examples in (23):
(23) a. Et encore n'estoie il mie delivre desa maleurte
Berinus, I, p. 390
And still CL-NEO was he not delivered from his unhappiness
'And he still was not delivered from his unhappiness'
b. Atant regarda pro contreval la
mer, (...)
Berinus, I, p. 240
Then looked (3s SUBJ) down at the sea
'Then he looked down at the sea'.
Here we have to account for the fact that pronominal subjects manifest a quite
different pattern from nominal subjects since there is here a parallelism be-
tween the structure of MidF and the Dutch and Danish examples of (19) and
(20). Notice that the pronominal subject in (23) appears before the negation.
On the basis of general theoretical considerations having to do with Spec-Head
agreement and Case assignment, we will assume that in the above examples
the subject has been moved to the Spec of TP.
Let us start by reviewing certain issues involved in the theory of agreement.
Although there is actually no clear consensus on how the agreement relation
is to be implemented, recent proposals have provided strong arguments show-
ing that it consists of a structural relation between a Spec position and a head
in a local environment. Moreover, as noted by Rizzi (1990) (see also Koopman
and Sportiche (1991)), the theory of Agreement and the theory of Case can
be conceived of along the same lines since they both involve the same struc-
tural relation. Bearing this in mind, let us explain how the subject in TP, pro
or an overt pronominal, can trigger agreement while being assigned Nomina-
tive Case. Our analysis is based on ideas developed in Rizzi (1990) follow-
ing Tomaselli (1990).4 We assume that Agr is a governor when it has features,
hence, when it is in a tensed clause. When there is no subject in its Spec
position to agree with, Agr can transmit its features to TP. This means that
when its governing capacities are restricted to the right, Agr will govern the
pro or pronominal subject in [Spec.TP] which, in turn, being in a Spec-Head
relation with T bearing tense features, will then trigger agreement since it is
c-commanded by Agr. Let us state that this particular type of agreement is
available only because of the fully pro-drop characteristics of OF and MidF,
therefore being unavailable for partly pro-drop languages like Dutch and
Danish. So, we make the assumption that, in languages where pro can be
THE LOCUS OF VERB MOVEMENT 93
Table 4.2
Verb-Second Distribution in Root and Embedded Clauses
in Six Texts of Middle French
# # # #
si is a sentence adverbial;7
the normal position for si is at the beginning of the sentence (never at
the beginning of a text);
si is never used in subordinate clauses and never in untensed clauses;
si is never used with negative clauses;
si is strictly contiguous to the verb form.
In our analysis, si is base-generated in the Spec of AgrP, an A-bar position.
According to our model, the sentential adverbial si has the status of a clitic
element on a tensed verb. Like the clitic subjects in Modern French, si can
only be separated from the verb by clitic objects. No adverbial element can
be inserted between si and the verb; si is never found in untensed clauses, i.e.
in infinitival or participial clauses. Our claim that si occupies the [Spec,AgrP]
position, subject to some conditions on identification, explains its distinct
behavior.
It is important to notice that Spec of AgrP is not the only possible position
where the so-called sentential adverbial could arguably be generated. For
example, R&T adopt a V-to-I analysis for Icelandic, but they propose that some
root narrative facts can be explained by generating in CP an E node, a sug-
gestion inspired by Banfield (1982).8 We shall pursue this idea that there is
something distinct about root narrative clauses, but without recourse to V-to-
C movement.
In his discussion of emphatic contexts with imperative do in English, Pol-
lock (1989) suggests, as an alternative to NegP, that in such cases, do occu-
pies the head of a projection he calls "Assertion" (see also Pesetsky (1989)).
Adopting Pollock's suggestion, Laka (1991) develops a similar analysis for
Spanish in which negative polarity elements such as nadie, nada, ningun,
nunca, etc., involve movement to the Specifier of a functional projection (IP)
which is generated between CP and IP, a position also involved in emphatic
fronting. The structure Laka assigns to the sentence in (28) is depicted in (29):
(28) a nadie yo pedire perdon Laka (1991)
to nobody I will ask forgiveness
THE LOCUS OF VERB MOVEMENT 97
Table 4.3
Proportion of Discursive Elements in Prose
with Respect to XP Elements
If we compare root and embedded clauses, omitting from the root clauses the
total number of discourse elements such as si, et, or, mais, etc., that are never
found in subordinate ones, the asymmetry between the two types of clauses
radically decreases.
As is shown in Table 4.4, the three texts of poetry in our corpus (Palatinus,
Miracles and Poesies) use far fewer discursive elements than the texts in prose.
This is not a coincidence and this peculiar behavior is predicted by the pro-
posed analysis. Since poetry does not involve a narrative style, we do not
expect a large number of discursive elements.
Let us summarize the results of the discussion. This section was intended
to show that there remains some kind of asymmetry between root and
embedded clauses since verb-second structures are used more frequently in
root clauses than in embedded ones. We have claimed that this asymmetry
can be partly explained by discursive factors typically found in root contexts.
Table 4.4
Proportion of Discursive Elements in Verses
with Respect to XP Elements
Texts "et" "si" Other Discourse Total Discourse Total XPVs
XPs XPs
Palatinus 3 31 36 70 35% 201
Miracles 1 14 12 27 11% 240
Poesies 14 6 14 34 15% 226
Subtotal 18 51 62 131 20% 667
100 THE DIACHRONY OF VERB SECOND
We have proposed an analysis which accounts for the discursive particles most
frequently encountered in our data: si, et and or. The assertive level argued
for offers a new way to derive some discourse phenomena related to the verb-
second constraint within a government-binding model, without reference to
some extra mechanisms. Finally, this category gives support to our claim that
V-to-C is an undesirable analysis of the derivation of verb-second word order
in MidF.
guage. We then develop an alternative proposal which accounts for the gradual
nature of the change, as a consequence of the reanalysis of individual lexical
items. Finally, we demonstrate that a V-to-Agr analysis has nothing special
to say about most of the violations of the constraint.
Table 4.5
Verb Second (V2) and V>2 in Middle French
Texts Nominals Pronominals
V2 V>2 % Total V2 V>2 % Total
Policie 47 32 41 79 71 21 23 92
Memoires 90 18 17 108 73 21 22 94
Figure 4.5 following, shows that violations with both NP subjects and with
pronominal ones represent a stable phenomenon that is attested over two cen-
turies. Thus, the violations are surprising for two reasons: First, they occur
in contexts where they should not occur on the subject clitic hypothesis, and
second, they reflect a relatively stable property of MidF.
On the other hand, when comparing the elements responsible for the viola-
tions, there does not seem to be a clear cut difference between those that show
up with pronouns, as in (37), or with nouns, as in (38). Indeed it is easy to
find both in the same text.
(37) a. Et auxi elles s'en scevent bien mocquer entre elles
QJM, p. 23
and also they know perfectly laugh between
themselves
'and so they perfecly know to laugh about it between themselves'
b. briefment il court par la maison QJM, p. 52
quickly he runs through the house
(38) a. Et auxi toutes les aultres sont en la buee QJM, p. 53
'And also all the other ones are in the laundry'
b. briefment, le pouvre corps de lui n' avrajames repoux (...)
QJM, p. 93
in brief the poor body of him NEG will never rest
'In brief his poor body will never rest in peace'.
In our analysis if two elements precede the pronominal subject, like et and
auxi in sentences (37) and (38), the first, et, must be in 1°, and the second,
the adverb, must be adjoined to AgrP.
there are two reasons for the loss of these two characteristics of MidF. They
are the loss of directional government and a change in the accentual pattern
of the language. We agree with Adams about the importance of these two
factors, despite the differences between the analysis presented here and the
one she proposes. We have also shown that the loss of the superficial mani-
festations of the verb-second constraint in Modern French must take into ac-
count some changes in the narrative structure of the language. A small class
of elements whose original syntactic function was to satisfy the verb-second
requirement and at the same time establish a predicative relation between the
verb and its subject is no longer able to fulfill these two functions. Any analy-
sis which aims to account for syntactic change in the history of French must
take into account all three of these factors.
6. Conclusion
In conclusion, there are at least two classes of verb-second languages: Those
that show a strict asymmetry between root and embedded clauses and those
that show little or no asymmetry. In the spirit of recent work, the former in-
volve long movement of V through Agr to C, while the latter involve short
movement of V to Agr. The evidence indicates that OF and MidF are non-
asymmetric verb-second languages, like Yiddish and Icelandic. Thus, we pre
dict that apparent root-embedded asymmetries in these other non-asymmetric
languages should also be amenable to an analysis which has recourse to
illocutionary elements in the sentence.
Notes
* Research for this article was made possible by grants from SSHRC No. 410-
89-1409 and 410-89-0785. We thank all the members of our research project
for their comments and support. Special thanks to Anne Rochette, Betsy Ritter
and Isabelle Hai'k for helpful comments and suggestions.
1. The 16th-century data not considered in this paper are part of work in progress.
2. One can maintain that there is no asymmetry between root and embedded
clauses, while proposing a movement to Comp: this is indeed what is pro
posed by Haan and Weerman (1986).
3. It has been noticed that LF movement is not possible across negation, while
S-Structure raising is possible. Pesetsky's suggestion rules out the movement
at LF, arguing'that the position is filled at this level, the null version of pas
blocking movement. For a different view of the problem with far reaching
consequences, see Rizzi (1990).
4. Tomaselli (1990) has argued that the agreement in COMP is a characteristic
of all the Germanic languages, even those which do not manifest agreement
at a superficial level. This question will not be discussed in this paper.
THE LOCUS OF VERB MOVEMENT 105
5. Deprez (1988, cf. fn. 3) proposed that Stylistic Inversion licenses an exple-
tive pro in [Spec,IP]. This expletive must be replaced at LF along the lines of
Chomsky. The subject must consequently move to SpecIP at LF where Nomi-
native Case is checked.
6. We borrow from Vance (1989:88-89) the following summary of these de-
bates: "Many scholars have puzzled over the role of et in introducing VS
clauses in Old and Middle French. Some of them (for example Foulet (1963);
Nissen (1943); Franzen (1939); Lewinsky (1949); Crabb (1955); Skarup (1975)
and Adams (1987a,b)) maintain that et may in some cases be used as an ad-
verb, acting as the intial constituent in what is really an CVS structure. Oth-
ers (for example Baulier (1956)) find in their data no basis for this claim. My
data argue for the latter position." For more references and a discussion of
this question, we send the reader to Marchello-Nizia (1979:281-287). For us
et has a role in the verification of the verb-second constraint, even though it
does not have the same categorial status as si.
7. The discussion will be limited to si used as a sentence adverbial. For other
distributions of si in different contexts, see Marchello-Nizia.
8. The node E in Banfield is used as an abbreviation for "Expression." The
theory developed by this author includes in a generative framework narrative
facts that are related to the writer. Banfield elaborates these notions under the
term of "Self." Most of the assertive considerations about si, drawn from
Marchello-Nizia, could be translated into Banfield's.theory.
9. If we argue that the Spec,IP position is a possible A-bar position, si may be
generated there. This can be bound to the pecularities of the nominal subject:
an NP has to be identified by its thematic role and by its Case; both are neces-
sary to express the relation of predication between the verb with its external
argument. If this relation can be expressed elsewhere than in Spec,IP, as it is
the case in OF and MidF, this position is available for another type of predica-
tion, the one called "a second level predication" by Marchello-Nizia: as it
has been demonstrated by this author, si expresses just such a relation be-
tween the locutor and the sentence. In the case of si, this means that the
locutor takes the meaning of the sentence as true. This is why si itself cannot
be negated.
10. For more considerations on this element see Marchello-Nizia (1979, ch. 20).
11. Roberts (1992) assumes that the relevant Spec position is [Spec.CP]. How-
ever, we propose that the position in question is [Spec,AgrP].
12. As an alternative, one could argue that et occupies the head of CP. This solu-
tion would have the advantage of giving a unified analysis of et in coordina-
tion and et in discursive contexts, but in this view one does not explain why
these discursive elements like et, or, or mais are involved in verb-second
structures with null or postverbal subjects.
13. It can also be suggested that imperative sentences would be analyzed through
the category Assertion since one of their most striking characteristics at this
period of French is the presence of discursive elements like et, si, or, etc.
14. In his study, Herman (1954) discusses a few cases ofXPS V violations in OF
and says that most of them are only apparent violations. With this observa-
106 THE DIACHRONY OF VERB SECOND
tion in mind and considering the kind of violations given by Foulet (1963), it
is possible to argue that MidF did not show more violations of the verb-sec-
ond constraint in root clauses than OF. Thus expressions premierement, certes,
nonpour tant and the like as in (15) are not violations either in OF or in MidF.
(i) Certes, elle disoit verite, Berinus, I, p. 15
'Surely, she was telling the truth'
(ii) Premierement je vueil respondre... Berinus, I, p. 90
'First I want to answer...'
(iii) Non pourtant, je ne le dy pas... QJM, p. 53
'Nevertheless, I do not say so'.
In OF, violations occurred in two contexts each one with a subject NP and
with a pronominal: 1. an XP could separate the subject and the verb as in (iv):
(iv) Li reis erranment li dist
QLDR, li tiers livres, p. 223
The king promptly to him says
'The king promptly says to him'.
2. an XP precedes the subject as illustrated in (v):
(v) De ces nuveles tuit furent esfrees
QLDR, li tiers livres, p. 227
By these news all were afraid
'This news frightened everyone'.
The main difference in the corpus of MidF that we have studied is that the
pronominal subject is almost never separated from the verb. On the contrary,
NP XP V word order is still sometimes met with a nominal subject:
(vi) Li chevalier tous communement tindrent ce conseil...
Berinus I, p. 120
'The knights all collectively followed this advice...'
(vii) et la dame si fu appellee Guigamo.
Berinus I, p. 120
'and the woman was thus named Guigamo'.
15. Vance has systematically compared the violations in word order in La Queste
del Graal, a 12th-century prose text to those in Jehan de Saintre also in prose,
written around 1456 by Antoine de la Sale. The author attributes the increase
of violations in MidF to a greater liberty in adjunction to IP which has mul-
tiple consequences on word order. For Vance, this new possibility explains
the transitional period between verb second to CP and SVO.
References
Adams, M. (1987a) "From Old French to the Theory of Prodrop." Natural
Language and Linguistic Theory 5:1-32.
THE LOCUS OF VERB MOVEMENT 107
Adams, M. (1987b) Old French, Null Subjects and Verb Second Phenomena.
PhD Dissertation, UCLA.
Adams, M. (1988a) "Embedded Pro." In J. Blevins and J. Carter, eds. Pro-
ceedings ofNELS 18, 1-21. Amherst, Mass.: GLSA.
Adams, M. (1988b) "Les effets V2 en ancien et en moyen francais." Revue
quebecoise de linguistique theorique et appliquee 7:13-40.
Banfield, A. (1982) Unspeakable Sentences, Narration and Representation in
the Language of Fiction. Boston, Mass.: Routledge and Kegan Paul.
Baker, M.C. (1988) Incorporation: a Theory of Grammatical-Function Chang-
ing. Chicago: University of Chicago Press.
Baulier, F. (1956) "Contribution a 1'etude de 1'inversion du sujet apres la
conjouction 'et'." Le Francois Moderne 24:249-257.
Bayer, J. (1984) "COMP in Bavarian Syntax." The Linguistic Review 3:209-
274.
Belletti, A. (1989) Agreement and Case in Past Participial Clauses in Ital-
ian. Ms. Universite de Geneve.
Bennis, H. and L. Haegeman (1983) "On the Status of Agreement and Rela-
tive Clauses in West-Flemish." In W. de Geest and Y. Putseys, eds.
Sentential Complementation, 33-55. Dordrecht: Foris.
Besten, H. den (1983) "On the Interaction of Root Transformations and Lexi-
cal Deletive Rules." In W. Abraham, ed. On the Formal Syntax of
Westgermania, 47-131. Amsterdam: John Benjamins.
Chomsky, N. (1986a) Barriers. Cambridge, Mass.: MIT Press.
Chomsky, N. (1986b) Knowledge of Language. New York: Praeger Publish-
ers.
Chomsky, N. (1991) "Some Notes on the Economy of Derivations and Rep-
resentations." In R. Friedin, ed. Principles and Parameters in Compara-
tive Grammar, 417-454. Cambridge, Mass.: MIT Press.
Chung, S. and J. McCloskey (1987) "Government, Barriers and Small Clauses
in Modern Irish." Linguistic Inquiry 18:173-237.
Cooper R. and E. Engdahl (1989) "Null Subjects in Zurich German." In L.O.
Delsing, C. Falk and C. Platzack, eds. Comparative Germanic Syntax.
Working Papers in Scandanavian Syntax 44:31-44. Lund, Sweden.
D6prez, V. (1989) On the Typology of Syntactic Positions and the Nature of
Chains: Move a to the Specifier of Functional Projections. PhD Dis-
sertation, MIT.
Diesing, M. (1990) "Verb Second in Yiddish and the Nature of the Subject
Position." Natural Language and Linguistic Theory 8:41—79.
Dufresne, M. (1989) "Les pronoms personnels sujets en moyen francais et le
contrainte V2." Talk given at the Colloque de syntaxe historique, 57th
Conges de 1'ACFAS, Universitd du Quebec a Montreal.
Dupuis, F. (1988) "Pro-drop dans les subordonnees en ancien francais." Re-
vue quebecoise de linguistique theorique et appliquee 7:41-62.
Dupuis, F. (1989) L'expression du sujet dans les subordonnees en ancien
francais. PhD Dissertation, Universit6 de Montreal.
Foulet, L. (1963) Petite syntaxe de I 'ancien francais. Paris: Champion.
108 THE DIACHRONY OF VERB SECOND
Pollock, J.-Y. (1989) "Verb Movement, UG and the Structure of IP." Linguis-
tic Inquiry 20:365-424.
Rizzi, L. (1990) Relativized Minimality. Cambridge, Mass.: MIT Press.
Rizzi, L. and I. Roberts (1989) "Complex Inversion in French." Probus 1:1-
30.
Roberts, I. (1992) Verbs and Diachronic Syntax. Dordrecht: Kluwer.
Rognvaldsson, E. and H. Thrainsson (1990) "On Icelandic Word Order Once
More." In J. Maling and A. Zaenen, eds. Modern Icelandic Syntax. Syn-
tax and Semantics 24:3-40. San Diego: Academic Press.
Santorini, B. (1989) The Generalization of the Verb-Second Constraint in the
History of Yiddish. PhD Dissertation, University of Pennsylvania.
SigurSsson, H.A. (1990) "VI Declaratives in Icelandic: Their Syntax and Se-
mantics." In J. Maling and A. Zaenen, eds. Modern Icelandic Syntax.
Syntax and Semantics 24:41-69. San Diego: Academic Press.
Skarup, P. (1975) Les premieres zones de la proposition en ancienfrancais.
Copenhagen: Akademisk Forlag.
Thiersch, C. (1978) Topics in German Syntax. PhD Dissertation, MIT.
Thurneysen, R. (1892) "Die Stellung des Verbums im Altfranzosisehen."
Zeitschrift fur Romanische Philologie 16:289-371.
Tomaselli, A. (1990) "COMP° as a Licensing Head - An Argument Based on
Cliticization." In J. Mascaro and M. Nespor, eds. Grammar in Progress:
GLOW Essays for Henk van Riemskijk, 433-445. Dordrecht: Foris.
Vance, B.S. (1988) "L'evolution de pro-drop en franc.ais medieval." Revue
quebecoise de linguistique theorique et appliquee 7:85-109.
Vance, B.S. (1989) Null Subjects and Syntactic Change in Medieval French.
PhD Dissertation, Cornell University.
Weerman, F. (1989) The V2 Conspiracy, A Synchronic and a Diachronic Analy-
sis of Verbal Positions in Germanic Languages. Dordrecht: Foris.
Zanuttini, R. (1990) "On the Relevance of Tense for Sentential Negation." Ms.
University of Pennsylvania.
5
Evidence for a Verb-Second Phase
in Old Portuguese*
Ilza Ribeiro
State University of Sao Paulo, Campinas
The basic facts related to the verb-second phenomenon have been studied in
generative grammar based on data from Modern Germanic languages (except
Modern English) and both Old Germanic and Romance languages. There is a
consensus among researchers who have developed studies on this phenomenon
that the verb-second effects are derived from two rules of movement: a) a
rule which moves the finite verb to the second position of the sentence, the
verb fronting being obligatory for every root clause: b) a rule which moves
any XP constituent (the NP-subject or a VP element or any XP of the sen-
tence [e.g., an adverbial]) to the first position of the sentence. This movement
is obligatory for every root declarative clause.
In the string Wh V S in matrix Wh-questions, the Wh-word counts as the
first element of the sentence with respect to verb-second. In other words,
displacing the verb to a position adjacent to the Wh-word creates a verb-
second structure. Rizzi (1991) calls "residual verb-second" the Wh-construc-
tions which show subject-auxiliary inversion in English, and verb-subject clitic
inversion in French. In those languages there is no generalization of the verb-
second order to main declarative clauses. On the other hand, "full verb-sec-
ond" identifies the phenomenon that determines the order of the constituents
in all matrix clauses and in some embedded ones in languages such as Ger-
man, Dutch and Scandinavian. The distinction between the two types of phe-
nomenon lies in the fact that residues of verb second are restricted to certain
limited environments.
The examples in (la) and (Ib) are instances of the French type of verb-
second residue while (Ic) illustrates the English type:1
(1) a. A qui as-tu parle?
To whom have you spoken?
110
EVIDENCE FOR A VERB-SECOND PHASE IN OLD PORTUGUESE 111
b. Que manges-tu?
What do you eat?
c. Whom did Mary see?
Current linguistic analyses assume that these constructions instantiate Agr° to
C° movement by virtue of the fact that Spec/C' is filled in by a Wh-element
(cf. Rizzi (1990b, 1991), among others). Wh-questions with subject-verb in-
version can be found in Brazilian Portuguese (henceforth BP), as shown by
the following sentences:2
(2) a. Para quando querem os diretores esses relat6rios?
For when want the directors these reports
'When do the directors want these reports?'
b. Com quern tinha Maria pretendido sair ontem a noite?
With whom had Maria intended to go out last night?
With whom did Maria intend to go out last night?'
The respective sequences Wh - V / Aux [+fin] - Subject - Object / V [-fin]
can be analysed as evidence that the verb is in C° and the subject in Spec/
Agr', and the Wh-phrases (para quandolcom quern) in Spec/C'.
The constructions above, however, do not exhibit a homogeneous pattern
in those languages, as can be observed from the following cases:
(3) a. A qui tu as parle?
b. *Que tu manges?
c. *Who Mary saw t?
d. Who t saw Mary?
e. *Who did t see Mary?
f. Para quando os diretores querem esses relat6rios?
g. Com quern Maria tinha pretendido sair ontem a noite?
In French, verb-subject inversion is obligatory with the Wh-element que (cf.
(lb/3b)), but optional with other interrogative elements in front of the sen-
tence (cf. (la/3a)). In BP, verb-subject inversion is optional (cf. (2a/3f) and
(2b/3g)), the order with no inversion (subject-verb order) being the most com-
mon in contemporary spoken BP.3 English, on the other hand, presents a pe-
culiar behavior when compared to those two languages: subject-auxiliary
inversion is obligatory whenever a non-subject Wh-element is moved into C°
(cf. (lc/3c)), and prohibited if the moved element is a Wh-subject (cf. (3d/
3e)).
Assuming that the English and the French constructions are residues of verb
second and that the observed differences in their syntax derive from other
properties of these languages (cf. Rizzi (1991)), the first question to be posed
is: Would the interrogative constructions of BP presented above be residues
of verb second?
Studies within the variationist model have shown two things regarding the
subject-verb order in BP: on the one hand, the occurrence of the VS order is
low in interrogative structures in modern spoken BP, but, on the other hand,
112 THE DIACHRONY OF VERB SECOND
it had a very high frequency of occurrence in past periods. Duarte (1989) stud-
ies the evolution of the VS order to SV in Wh-interrogatives. She observes
that in 1734—the starting point of the time period considered—all direct Wh-
interrogatives showed VS order. The same goes for the year 1845, since only
4% of the cases displaying SV order corresponded to occurrences of only one
single sentence of a total of 28. From 1845 on, a slight increase in the occur-
rence of SV order (especially in 1882 and 1918) is observed, but a clear pref-
erence for it in the period from 1937 to 1989 is shown, and this may
characterize a change.
Thus, Duarte's data, especially the fact that in 1734 100% of the Wh-root
constructions occurred with VS order, indicate the existence of a movement
rule from Agr° to C° in 18th-century Portuguese. We may then conclude that
the high frequency of Wh+VS constructions in the 18th century must be
analysed as resulting from syntactical mechanisms related to the verb-second
phenomenon. Therefore, the second question to be posed is: If it is the case
that Old Portuguese (henceforth OP) was a verb-second language, how and
why has it lost this syntactic property?
In this work we shall limit ourselves to offering some evidence which seems
to favor a characterization of OP as a veb-second language, and to presenting
some considerations on the loss of this property.4 In SECTION 2, we shall com-
pare some analyses of the French verb-second phenomenon with data from
14th-century Portuguese. In 2.1 we shall discuss the verb-second constructions
in root and embedded clauses; in 2.2 we shall focus on verb-first clauses and
in 2.3 on verb-third ones. In 2.4 we shall deal with the clitic complement
placement and the feature [+Agr] in C°. In SECTION 3, we shall sketch a pro-
posal for analysing the loss of verb-second in a diachronic perspective, fol-
lowing Roberts' proposal of parametric possiblities for Nominative assignment
by Agr° (Roberts (1992)).
Table 5.1
The Order of Constituents in Old Portuguese6
Order SV(C) XPV(S) VSV (C) Total
Root 15 31 3 8 57
Embedded 19 12 7 18 56
Total 34 43 10 26 113
EVIDENCE FOR A VERB-SECOND PHASE IN OLD PORTUGUESE 113
indicate that enton 'then/now' is used as a syntactic recourse to meet the verb-
second restriction (example not included in Table 5.1):
(7) Enton disse San Gregorio... (1.16.37)
Then said San Gregorio
'Then San Gregorio said'.
The conclusion which may be reached at this point is that e and ca may vir-
tually appear in front of or outside of any type of matrix clause,8 since such
elements may precede SVO or CVS clause (examples not included in Table
5.1) (but cf. below: cf. also SECTION 2.3):
(8) a. E ele non Ihos quis dar (1.28.28)
and he not them+it wanted give
'And he did not want to give it to them'
b. E tan comprida era a vida (1.2.5)
and so complete was the life
'And life was so complete'.
These facts corroborate Mattos e Silva's classification of the element e (and
ca) as a conjunction rather than an adverb. Also, we agree with her that the
elements e and ca, when at the beginning of a matrix clause, are used to con-
nect a new sentence to the foregoing discourse. Thus, we conclude that they
are positioned outside of the sentential syntactic structure, i.e., outside CP.
We do not mean by these observations to exclude the possibility of e (and
ca) occurring in certain contexts with adverbial value, just as et is found in
some 13th-century French structures. A more detailed study of these construc-
tions is required before we can determine whether that is a probable property
of e. Notwithstanding, a brief and unsystematic observation of root clauses
introduced by e has provided the types of examples below (not included in
Table 5.1):
(9) a. E juntaron-se muitos homees... (1.27.3)
and gathered REFL a lot of men...
'And a lot of men gathered together...'
b. E foron- se logo muit'aglnha (1.2.45)
and went-3pp REFL soon very fast
'And soon they went out very fast'
c. E disse-o logo aos frades (4.6.7)
and told-3sp-it soon to the friars
'And he soon told it to the friars'
d. E non conta a Escritura que... (2.30.17)
and not says the Scripture that...
'And the Scripture does not say that...'.
The examples (9a-c) are all cases of the phenomenon known as Tobler/
Mussafia law in the traditional literature. This phenomenon, found in the
116 THE DIACHRONY OF VERB SECOND
Medieval Romance languages, implies that the clitic should not occur in the
sentence-initial position; so, the use of the enclitic pronoun is obligatory
whenever proclisis will place the clitic in the sentence-initial position (cf.
Cardinaletti and Roberts (1991); Salvi (1990) and Beninca (this volume) for
further details on the Tobler/Mussafia law effects).
Cardinaletti and Roberts analyse enclisis as a process involving the pres-
ence of an inflected verb in C° (the verb undergoes structure-preserving
topicalization), the Spec/C position being empty. So, the enclitic use of the
pronouns in the examples (9a-c) may be derived from the fact that the verb is
the first constituent of the sentence and the Spec/C' position is empty; thus,
the conjunction e is outside CP.9 In the example (9d) non would be consid-
ered as the first constituent of the sentence. In such a case, in this structure as
well, the conjunction e would be outside CP.
Of all the strings computed as XPV(S), the ones which are associated with
phonologically null subjects are of the type given below:
(10) a. D'alguas cousas me calarei (1.5.25)
of-some things REFL silent-will-Isp
'I shall be silent on some of the issues'
b. Tan aglyha o passaron (1.2.46)
so fast passed-3pp by it
'So fast they crossed it'.
Adams (1987) and Vance (1988, 1989) point out that null subjects are fre-
quent in matrix clauses in 13th-century French:
(11) a. Si firent grant joie la nuit (R.C1.XII)
so made (they) great joy that night
They celebrated that night' (Adams, p. 44)
b Grant piece parlerent de ceste chose (Q3)
great piece spoke-3pl of this thing
'They spoke a great deal about this thing' (Vance, p. 97)
According to Adams, in the construction types under (11), the position of the
null subject is postverbal. For her, this is the basic configuration for licensing
null subjects: pro is licensed by canonical government of Agr° in C°, in verb-
second contexts.
A strict correlation between null subject and context of inversion verb-
second context) is not kept in the 15th century, according to Vance's analysis
(cf. also Roberts (1992); Hirschbuhler and Junker (1988); Dupuis (1988)). She
also observes that in the 15th century pro (referential/expletive) also appears
in both matrix and embedded clauses in non-inversion contexts, namely, in
preverbal position. Thus, French changed from a system that only licensed null
subjects in a verb-second context (13th century) to a system that licensed null
subjects in preverbal position (15th century).
This change in the context of pro-licensing follows from the progressive
loss of the verb-second property, according to Vance, and from the possibility
EVIDENCE FOR A VERB-SECOND PHASE IN OLD PORTUGUESE 117
not the same. Pro can only be formally licensed in contexts of Case assign-
ment, i.e., in a context either under agreement or under government by Agr°.
However, when pro is licensed under agreement, rich inflectional verbal mor-
phology is required for its content-licensing or identification (cf. Roberts
(1992) and Cardinaletti and Roberts (1991)). So, the possibility of licensing
referential pro under agreement in 15th-century French texts (Nominative
Case-assignment possible) is restricted to some syntactic contexts (mainly in
2pl null subject contexts) (cf. Hirschbiihler and Junker (1988) and Vance
(1989)), since the 12th/ 13th-century OF verbal system was already undergo-
ing erosion. However, the 14th-century OP Agr-system contains person/num-
ber features sufficiently rich to recover the content of pro: thus OP verbal
morphology is able to license referential pro under agreement. This way, in
the 14th-century OP structures pro could be licensed in contexts under agree-
ment or under government by Agr° (Nominative Case assignment possible in
both contexts: cf. SECTION 3).
(13) SV(C)
a. Entendemos nos que a alma vive (4.4.55)
understand we that the soul lives
'We understand that the soul lives'
b. Dizemos que a alma recebe peas (4.27.10)
say-1 pi that the soul receives sufferings
'We say that the soul undergoes sufferings'
c. Acaeceu huu dia que sen padre e sa madre
happened one day that your father and your mother
fezeron gram jantar (1.2.6)
prepared big dinner
'It happened, one day, that your parents prepared a big dinner'.
EVIDENCE FOR A VERB-SECOND PHASE IN OLD PORTUGUESE 119
cursion and that this recursion is limited to environments where the recursive
CP is governed by a verb. Thus, there is no generalized CP-recursion (cf. also
Cardinaletti and Roberts (1991)). So, being adjunct clauses, the sentences in
(16) cannot be treated as cases of CP-recursion. We assume a double-Agr struc-
ture to derive the verb-second order in these embedded Wh-clauses.
It is worth noticing that embedded verb second in adjunct clauses is not
generalized. It seems restricted to limited clauses as those in porque 'why'
and se 'if. This situation should be compared to the one in Yiddish mentioned
in Cardinaletti and Roberts.
In his terms, the verb-first constructions in (18), as well as the ones mentioned
above (17), all present the Spec/C position filled in by a phonologically null
operator, which allows him to maintain that these are true verb-second struc-
tures. All those possible verb-first structures occur in OP; some of them were
illustrated in (17). As to the verb-first yes/no interrogative constructions, we
note that Padua (1960:94) considers that the placement of the verb in sentence-
initial position constitutes "the major recourse to mark interrogation." In what
follows, we give some examples not included in Table 5.1 of this type of con-
struction:
(19) a. Acaeceu, padre, ja algua cousa nova por que
happened-3sp, father, already some thing new why
choras mais que sooes? (1.1.8)
cry-2sp more than you are used to
'Father, has anything happened over which you cried more
than you use to?'
b. Cuidas, padre, que este homen ouve alguu meestre
care-2sp, father, that this man had some master
que o ensinasse? (1.2.17)
that him taught
'Do you think, father, that this man had a master who taught him?'
c. Queres, Pedro, que ti conte...? (1.16.38)
want-2sp, P., that you-Acc-2sp tell-lsp...?
'Do you want me, P. , to tell you...?'
Following Rizzi's (1991) proposal that yes/no questions have an empty op-
erator in Spec/C', these structures can be analysed as involving a Spec/C' po-
sition filled in by null operator and the inflected verb in C° positon. The seven
embedded VS constructions computed in Table 5.1 are of the following type
(my italics):
(20) a. Crees que ando eu sen almal (4.4.50)
believe-2sp that am I without soul
'Do you think that I have no soul?'
b. Non ti semelhas, Pedro, que deven a aver gram
not to-you appear-3sp, P., that must to have great
vergonha os homees (11.15)
shame the men
'Don t you think, P., that men must be honorable?'
c. Conta San Gregorio que veo huu principe (1.3.2)
say-3sp San Gregorio that came a prince
'San Gregorio says that a prince came'.
Although we have grouped them as VS, the constructions above comprise two
different syntactic structures: the one in (20b) is of the VCS type or Romance
inversion (free-subject inversion); those in (20a) and (20c) are both
EVIDENCE FOR A VERB-SECOND PHASE IN OLD PORTUGUESE 123
unaccusative structures, in which the subject may remain in its basic position
inside VP. In this way, there is no evidence of movement of V° to C° in
either construction.
Due to what was observed in SECTION 2.1.1, on the possibility of pro being
licensed in preverbal position, declarative verb-initial structures such as those
in (21) below occur (frequently) in the DSG text:
(21) a. Fez mui gram chanto (2.8.27)
did very great cry
'He cried a lot'
b. Acharon as sas maaos e os seus pees tan ben saos
(4.25.13)
found-3pl the his hands and the his feet also healthy
'they also found his hands and his feet recovered'.
In the examples above, referential pro occupies Spec/X' position. These
examples may be treated as cases of pro-realization in preverbal position.
Vance (1988, 1989) remarks that, in Middle French, CSV word order co-
occurs with CVS word order in matrix and embedded clauses. Verb-
subcategorized complements, which in the 13th century triggered obligatory
inversion, as well as sentential adverbs, may appear in the initial position of
structures of the form CSV in the 15th century, without triggering inversion.
The following examples illustrate this situation:
(23) a. Et lors ilz commencerent a rire (Q64) (Adams (1988:94))
And then they started laughing
b. le petit Saintre les yeux de Madame ne cessoient deregarder...
(J55) (Vance, p. 163)
the little S. the eyes of Madame not ceased
'Madame's eyes did not cease to look at little Saintre'
c. et aussi fis je de par vous (J104) (Vance p. 159)
and thus did I from by you
'and I did likewise with respect to you'.
The example in (23a) is a construction from the text La Queste del Saint Graal
(13th century). It illustrates a rare context of CSV word order in which C is
circumstantial; the construction in (23b) exemplifies a CSV structure type from
the text Jehan de Saintre (15th century) in a context which, in the 13th cen-
tury, required the CVS order. In this construction, C is the equivalent of ver-
bal complement. The case in (23c) shows that the CVS order also occurs in
the 15th century in a context similar to those found in the 13th century.
Although no occurrence of matrix CSV structure type was registered in the
data examined, some XPSV constructions did occur, in which XP is not a con-
stituent subcategorized by the verb. Certain adverbial elements did not trig-
ger obligatory subject-verb inversion as shown by the examples below (with
the exception of (24e), not included in Table 5.1):
(24) a. e assi o santo homen defendeu os seus discipulos
(1.9.13)
and thus the saintly man defended the his disciples
'and the saintly man thus protected his disciples'
b. E enton huu homen siia en sa pousada... (1.2.25)
and then a man sat down in his inn
'Then a man sat down in his inn'
c. E assi o fez o poder de Deus (1.2.47)
and thus it made the power of God
'And the power of God thus made it'
d. E pois se juntaron dois homens ou tres (2.9.3)
and then REFL gathered together two men or three
'And then two or three men gathered together'
e. E enton respondeu o abade santo e disse (1.8.33)
and then answered the abbot saintly and said
'And then the saintly abbot answered and said'.
EVIDENCE FOR A VERB-SECOND PHASE IN OLD PORTUGUESE 125
The constructions above show that adverbial elements as such as enton, pois
and assi permit two word orders, namely XPadvSV and XPadvVS. The struc-
ture of the form XPadvSV indicates that the XP-constituent is in adjunction
either to CP or to AgrP.
It is widely assumed that matrix declarative clauses are CPs in a verb-
second system and AgrPs in a non-verb-second system (but cf. note 7); and
adverbials adjoin more readily to AgrP than to CP. Consequently, the order
XPadvSV is impossible in matrix clauses in verb-second languages. How-
ever, a small number of adverbials are able to appear in that position, even in
a verb-second language like German:
(25) ...denn Johann hat gestern das Buch gelesen
so J. had yesterday the book read (Roberts (1992:186))
Vance (1989) notes that in Jehan de Saintre (15th century), a large class of
adverbial elements appear in initial position, without triggering inversion, as
seen in (26) below:
(26) Lors la royne fist Santre' appeller
then the queen made S. to-call
'Then the queen had Santre called' (Vance, p. 158).
She analyses this structure as resulting from the adjunction of the adverbial
to AgrP.
The choice between the two possibilities of analysis (adjunction to CP, as
in the case of German in (25) or adjunction to AgrP, as in the OF case in (26))
is dependent upon various factors, including the frequency of that word order
type in the DOS (and the other corpora representative of the 14th century),
and the definition of the class of elements able to appear in initial position
without triggering inversion. Only after these factors are considered can we
get the evidence needed for a decision as to whether or not matrix declarative
sentences in OP were always CPs.
In the sample analysed for Table 5.1, we registered one occurrence of the
SXPV construction type:
(27) El con sa mao deu a oferta (2.23.17)
he with his hand gave the offer
'He himself presented the offering'.
The order SXPV is also attested in embedded clauses in OF, with XP repre-
senting different complement types (cf. Vance (1988, 1989)). Below is one
example:
(28) Je vous dis que vos avec moi venez (Ql)
I to-you say that you with me come
'I tell you to come with me'.
According to Vance's analysis, the SXPV clauses in 13th-century French are
derived via topicalization to Spec/V position of an adverbial or a VP-
complement. She further adds that these structures are found only in embed-
126 THE DIACHRONY OF VERB SECOND
3. The Change
It is known that Modern BP does not manifest verb-second structures, that
the fronting of an element to clause-initial position does not result in a verb-
second string, and finally that, even in direct interrogative constructions, verb-
second residues started to disappear by the late 19th century. Assuming the
change from a verb-second language to a non-verb-second one, we will try to
present a sketchy analysis of certain facts that were brought up here in rela-
tion to the element which triggered the change.
The Principles and Parameters model considers linguistic change as a modi-
fication in the setting of certain parameters (cf. Rizzi (1988); Roberts (1992);
EVIDENCE FOR A VERB-SECOND PHASE IN OLD PORTUGUESE 129
Lightfoot (1991), among others). This theory predicts that, if a language loses
a basic property linked to a given setting of a parameter, it will also lose the
other properties related to that setting, which leads to a restructuring in the
language (i.e., a change)
According to Rizzi, restructuring is not an instantaneous process and there
is a certain amount of inertia to be taken into account when one deals with
such a complex system as language. In these terms, an accurate and realistic
characterization of the changing process must imagine that "the language in
question will enter a phase of instability which will be resolved by eliminat-
ing some of the properties related to the abandoned value, and by confining
some of them as relics on special stylistic registers or as crystallized construc-
tion markers" (Rizzi, p. 17).
The verb-second residues of English, French and Portuguese illustrate this
point, since they are relics of a general process of moving the inflected verb
to a second position in the sentence, which was very productive in these lan-
guages at former stages, but which is now confined to a few marked construc-
tions. According to theoretical assumptions children build their grammar from
what they hear (primary data) and from what they already possess innately
(Universal Grammar); also, they do not have direct access to the grammar of
the people around them (cf. Lightfoot (1979, 1981, 1991), among others).
This being so, the child's grammar (grammar 2 = G2) and the adults gram-
mar (grammar 1 = Gl) are totally discontinuous:
(33) Primary data 1 Grammar 1
Primary data 2 Grammar 2
Because the surface structures may be compatible with more than one gram-
mar and also because children do not know which analysis represents Gl,
children may opt for an incorrect hypothesis about Gl; this will result in a
grammatical change. Hence, the grammar that children build may differ from
the adult's grammar (cf. Lightfoot (1979, 1991)).
Adams (1987) argues that the reason why OF lost its verb-second effects
was that children reanalysed the derived SVO structure (= verb-second struc-
ture) as a basic SVO one. Having also lost the initial accent, typical of verb-
second languages, the surface structures did not present enough evidence for
the children to assume OF as a verb-second language, inasmuch as the SVO
order was very frequent. I take a position different from Adams, by assuming
that the sole reanalysis of a derived order as basic is not a sufficient condi-
tion for the loss of verb-second status, even if that is connected to a change
in accent pattern.
Thus, in the case of OF, Gl analyses SVO as a structure derived by the
displacing of S and V; G2 analyses SVO as a basic structure. However, chil-
dren had enough evidence of verb-second structures, produced by Gl, and these
could not be analysed as basic structures. Furthermore, by considering the time
period for the loss of verb-second (14th-16th centuries), we have to admit that
G2 also generated XPVS structures. Therefore, we believe that a single re-
analysis of a derived structure as basic and the loss of initial secondary
accent are not enough to explain the loss of verb second in OF.
130 THE DIACHRONY OF VERB SECOND
are disguised by adaptative rules. These rules are necessary only to interpret
structures produced by previous generations; these structures. consequently,
become obsolete, and this is the final stage in a change.
If adaptative rules are restricted to the interpretation of structures, and not
used to produce them, the realization of verb-second structures cannot be ex-
plained in these terms. We saw in Duarte's (1989) work that direct Wh-ques-
tions displayed verb-subject order until the 19th century. It is known that in
the Portuguese of the 18th-century and the beginning of 19th-century matrix
declarative clauses which showed movement of the auxiliary to a pre-subject
position were attested (cf. Kato and Tarallo (1986); Tarallo and Kato (1989);
and Berlinck (1989)) involve Nominative assignment under government. Con-
structions like (38) are not very much used in contemporary spoken BP; they
are still found in formal written texts:
(38) a. Tivessem eles cumprido o acordo, tudo seria resolvido
'Had they fulfilled the agreement, everything would have
been settled'
b. Maria acredita terem os meninos saido cedo
'M. believes the boys to have left early'
c. Que filme tinha Joao visto?
'Which film had John seen?'
d. Tem Joao feito o trabalho?
'Has John done the work?'
The fact that these two types of constructions (subject-verb inversion in Wh-
question and subject-auxiliary inversion like (38); cf. also the clauses in (2))
were kept in the system until the late 19th century, indicates that the change
in the Nominative assignment parameter did not happen abruptly from (35a)
to (35b).
From a letter written by Padre Leonardo do Vale in Bahia in 1561 (Serafim
Leite, 1942), we analysed the first 110 sentences and found that, out of the
thirty strings with evident subjects, eleven manifested the VS order, in con-
structions like:
(39) a. polio que vendo elle que...
for what seeing he that...
'for this reason he seeing that...'
b. primeiramente, foy a viagem muy trabalhosa
firstly, was the trip very laborious
'firstly, the trip was very hard'
c. assi que, com todos estes trabalhos, teve o Senhor
so that with all these tasks, had God
por bem de...
for the sake of...
'thus, with all these tasks, God decided...'
EVIDENCE FOR A VERB-SECOND PHASE IN OLD PORTUGUESE 133
4. Conclusion
In this work I have discussed the verb-second phenomenon found in the
DSG. I propose that this OP document reflects properties of a verb-second
language. The verb-second nature of OP in root clauses is illustrated clearly
by the examples in (5). We related the verb-second nature of OP to the fea-
ture [+Agr] in C°.
OP allows embedded verb second under limited conditions. We saw that
verb-second is possible in the complements to bridge verbs as in examples
(14). The class of bridge verbs in question is the same one which allows verb-
second completives in verb-second Germanic languages and in OF. Verb-sec-
ond is also possible in causal and conditional clauses, as in examples (16b)
and (16a), respectively. Stylistic Fronting is possible in relative clauses (ex-
amples in (15)).
In terms of current proposals, the inflected verb cannot move to an embed-
ded [+Wh] C° (cf. Rizzi (1991) and Rizzi and Roberts (1989)). So, both the
relative and the adjunct verb-second clauses motivate an analysis in terms of
a double-Agr structure. Given this analysis of relative/adjunct clauses, we do
not have to treat completive verb-second as involving movement of the verb
to C°. A double-Agr structure would also suffice to derive these constructions.
We noticed that finite verb in initial position is not restricted to yes/no ques-
tions, conditional clauses and imperatives. Verb-first declaratives are quite
common in the DSG. This may be due to the fact that movement of V° to C°
was not required for all matrix declarative clauses.
I have also discussed the loss of verb-second properties of OP, based on
the approach to Nominative-Case assignment advocated in Koopman and
Sportiche (1991) and Roberts (1992). The evidence I presented shows that in
OP Nominative Case was assigned both under government and under agree-
ment. The fundamental difference between OP and Modern Portuguese is that
the latter does not assign Nominative under government any more. Data from
the 16th century indicate that the possibility of Nominative assignment under
government was already restricted to contexts with verbs of the auxiliary type.
Notes
A first version or this paper was presented in the course "Topicos em
Linguistica Historica" given by Mary Kato and Fernando Tarallo for the PhD
Program at UNICAMP. I thank Mary Kato, Fernando Tarallo, Giampaolo
Salvi and Charlotte Galves for valuable comments and suggestions, and one
anonymous reviewer for interesting observations. Special thanks go to Ian
Roberts for his invaluable comments on the present version of this paper, and
EVIDENCE FOR A VERB-SECOND PHASE IN OLD PORTUGUESE 135
1. The French and English data in (1) and (3) are from Rizzi (1991).
2. The constructions (2) with inversion of subject-verb to verb-subject order are
not very much used in contemporary spoken BP: they are generally found in
literary and formal style. Though seldom used, these constructions are not
judged ungrammatical by all BP native speakers (cf. SECTION 3).
3. Verb-subject inversion is always obligatory in any root Wh-construction in
Modern European Portuguese, as the following examples ilustrate:
(i) (O) que estiveste tu a fazer?
'What had you been doing?'
(ii) *(O) que tu estiveste a fazer?
(iii) Onde trabalha a Maria?
'Where does Mary work?'
(iv) *Onde a Maria trabalha?
Cf. Ambar (1988) for more details about VS order in European Portuguese;
cf. also Torrego (1984) on Spanish verb-subject inversion in Wh-questions.
4. Within the limits of study, the hypotheses raised here are both qualitative and
quantitatively insufficient if we consider necessary for the investigation of
any linguistic change the analysis of individual texts, various dialects and
different epochs (Hirschbuhler and Junker (1988)).
5. Mattos e Silva's work (1971) focuses on the analysis of a Portuguese docu-
ment, named by her "A mais antiga versao portuguesa dos Quatro Livros dos
Dialogos de Sao Gregorio" (The oldest Portuguese version of the Four Books
of Saint Gregory Dialogues) and dated by internal chronology as before 1385.
The linguistic analyses of the 14th-century structures (Mattos e Silva (1989))
build upon that critical/philological reading. This is the corpus on which are
based most of my reflections regarding the verb-second phenomenon in OP.
OP examples will be cited followed by the book, chapter and line numbers,
according to Mattos e Silva's citations.
6. We name "root" the affirmative independent/matrix sentences of Mattos e
Silva's study, which include transitive, intransitive and copula constructions,
and existential structures as well. The XPV(S) and VS orders indicate that
the lexical subject is postverbal, but this is not meant to say that the subject is
always postverbal when it is phonetically null (cf. comments in 2.1.1). In the
V(C) order we disregarded the verbal complements represented by clitics,
which are in general preverbal (but see the observations in 2.4).
7. For Travis (1984) SVO verb-second matrix clauses can be AgrP structures.
For Roberts (1992) and Cardinaletti and Roberts (1991) the landing site of
the verb in verb-second structures may vary cross-linguistically; the constraints
136 THE DIACHRONY OF VERB SECOND
on verb second could be met either on the CP level or on the AgrP level, in a
double-Agr structure. For the moment, we shall analyse the matrix clauses
presented in this paper as CP-structures.
8. Vance (1989:91) proposes an analysis along the same lines to explain the
occurrence of the element et 'and' introducing sentences in Middle French.
9. Beninca (1989) analyses enclisis as a process involving a clitic+verb in C°;
since Spec/C' is empty, the verb must move to this position, resulting in an
enclitic structure. Even considering her proposal, the conjunction e is outside
CP.
10. In the DSG the SXP V order is common in both adverbial and relative clauses:
(i) as lagrimas que eu cada dia deito dos meus olhos (1.1.10)
'the tears which I drop from my eyes every day'
(ii) quando eu no moesteiro vivia (1.1.11)
'when I used to live in the monastery'.
11. In Modern BP, constructions like (30), as well as those in (31), have the clitic
adjacent to the verb.
12. Cardinaletti and Roberts (1991) analyse the "clitic-second" phenomenon in
Germanic and Romance languages. They identify the traditionally recognized
Wackernagel clitic position of Germanic languages as Agrl ° and assume Agrl °
also to be the clitic position in the Romance languages which obey the Tobler/
Mussafia law. Their proposal differentiates the clitic position from the in-
flected verb position. Agrl0 can also be the position for the inflected verb;
the possibilities of verb movement from Agr2° to Agrl 0 may vary from lan-
guage to language, and even within the same language this movement may
vary according to the sentence status: matrix or embedded. In this section I
put forward an analysis which considers clitic elements to be constituents of
C° head in embedded (and matrix) clauses. In view of Cardinaletti and Rob-
erts' proposal that Agrl 0 is the clitic position, how can these two analyses be
related as regards the 14th-century OP data? In research in progress I attempt
to argue that in OP Agrl 0 is the syntactic clitic position; however, the clitic
elements must move to C° in some embedded (and matrix) structures, such as
those (30, 31 and 32). It is the agreement (pronominal) feature in C° that
makes the clitic-movement/adjunction to this position possible. This analysis
also implies that some OP matrix clause types are instances of a double-Agr
system (cf. note 7).
References
Abraham W., ed. (1983) On the Formal Syntax of the Westgermania.
Amsterdam: John Benjamins.
Adams, M. (1987) Old French, Null Subjects, and Verb Second Phenomena.
PhD Dissertation, UCLA.
Adams, M. (1988) "Les effets V2 en ancien et en moyen francais." In
Hirschbuhler and Rochette, eds., 13-40.
EVIDENCE FOR A VERB-SECOND PHASE IN OLD PORTUGUESE 137
Rizzi, L. (1986) "Null Objects in Italian and the Theory of pro." Linguistic
Inquiry 17:501-557.
Rizzi, L. (1988) "The New Comparative Syntax: Principles and Parameters
of Universal Grammar." Ms. University of Geneva.
Rizzi, L. (1990a) Relativized Minimality. Cambridge, Mass.: MIT Press.
Rizzi, L. (1990b) "Speculations in Verb Second." In Mascaro and Nespor, eds.,
375-385.
Rizzi, L. (1991) "Residual Verb Second and the WH Criterion." Technical
Reports in Formal and Computational Linguistics, 2. University of
Geneva.
Rizzi, L. and Roberts, I. (1989) "Complex Inversion in French." Probus 21:1-
30.
Roberts, I. (1985) "Agreement Parameters and the Development of English
Modal Auxiliaries." Natural Language and Linguistic Theory 3:21-58.
Roberts, I. (1989) "Case-assignment Parameters and the History of French
Inversion." GLOW Newsletter, 22:55-57. Dordrecht: Foris.
Roberts, I. (1992) Verbs and Diachronic Syntax. Dordrecht: Kluwer.
Said Ali, M. (1957) Dificuldades da lingua portuguesa: estudos e observacoes.
Rio de Janeiro: Livraria Academica.
Said Ali, M. (1969) Gramatica secundaria da lingua portuguesa. Sao Paulo:
Edicoes Melhoramentos.
Salvi, G. (1990) "La sopravvivenza della legge di Wackernagel nei dialetti
occidentali della Penisola Iberica." Medioevo Romanzo 15:177-210.
Serafim Leite, S. I. (1942) Cartas dos primeiros Jesuitas do Brasil (1538-
13). Comissao do IV Centenario da Cidade de Sao Paulo.
Taraldsen, K.T. (1986) "On Verb Second and the Functional Content of Syn-
tactic Categories." In Haider and Prinzhorn, eds., 7-26.
Tarallo, F. (1990) Tempos linguisticos: itinerario historico da lingua
portuguesa. Sao Paulo: Atica.
Tarallo, F., ed. (1989) Fotografias sociolinguisticas. Campinas: Pontes Editora.
Tarallo, F. and M. Kato (1989) "Harmonia trans-sistemica: variacao intra e
inter-linguistica." Preedicao 8, UNICAMP.
Tomaselli, A. (1990) "COMP° as Licensing Head: An Argument Based on
Cliticization." In Mascaro and Nespor, eds., 433-445.
Torrego, E. (1984) "On Inversion in Spanish and Some of Its Effects." Lin-
guistic Inquiry 15:103-129.
Travis, L. (1984) Parameters and Effects of Word Order Variation. PhD Dis-
sertation, MIT.
Vance, B.S. (1988) "L'evolution de pro-drop en francais medieval." In
Hirschbuhler and Rochette, eds., 8-112.
Vance, B.S. (1989) Null Subjects and Syntactic Change in Medieval French.
PhD Dissertation, Cornell University.
6
Indo-European Origins
of Germanic Syntax
Paul Kiparsky
Stanford University
1. Introduction
In accord with the standard analysis I assume that verb-second and verb-first
order in Germanic1 is derived by movement of the finite verb to a.vacant C°
position (den Besten (1983)).2 I shall argue that in English and in the early
Germanic languages the Specifier of CP is a focus position, which hosts wh-
phrases, demonstratives, and negation. This specifier position can in turn be
preceded by a topicalized/left dislocated element adjoined to CP; if it is an
NP, it must bind a resumptive pronoun in the clause. In addition, constituents
can be topicalized by adjunction to S.
140
INDO-EUROPEAN ORIGINS OF GERMANIC SYNTAX 141
2. Germanic
2.1 V-to-C° movement
That V-to-C0 movement is obligatory is uncontroversial, though the expla-
nation for this is contested.3 Old English is important in this regard because,
alongside verb-second main clauses, it has main clauses in which the finite verb
remains inside S, including those having verb final.4 Instead of assuming that
V-to-C° movement is optional in Old English, I will argue that the category C
itself is optional, where no principle of grammar requires its presence.
On these assumptions, the word order patterns are derived in outline as fol-
lows. C° is obligatory in subordinate clauses, because they are in argument
and modifier position and sentences must be turned by complementizers into
CPs in order to function as arguments and modifiers. C° is also obligatory in
142 THE DIACHRONY OF VERB SECOND
undergo verb second or not as already discussed. But if wh-phrases must re-
side in Spec-C, sentences containing them must be CPs, therefore must con-
tain a C°, to which the finite Verb must move. Hence wh-phrases induce verb
second.11 On the assumption that topicalization and wh-movement in Old
English have the same landing site, as assumed in some analyses on the anal-
ogy of Dutch and German (e.g. Kemenade (1987)), these facts are mysteri-
ous. From a modern English perspective they are not surprising since they
persist to this day in the conditions on Subject-Aux Inversion ("residual verb
second"):12
(11) a. When will John arrive from London?
b. Tomorrow will John arrive from London.
c. Tomorrow John will arrive from London.
Maling and Zaenen (1981) noted that in Old Icelandic subordinate clauses,
topicalization but not wh-movement triggers verb second. This difference
between Topic and wh follows directly if we suppose that the complementizer
in embedded wh-clauses is lexically filled,13 and treat subordinate clauses with
topicalization as embedded root sentences with an external complementizer.14
On the assumption that adjunction is allowed to maximal projections we
would expect topicalization by adjunction to S as well. In Old English,
topicalization by adjunction to S must be assumed in main clauses with ini-
tial XP and no verb second (12a,b) and in "embedded root clauses" (12c,d):15
(12) a. & him ba Ioseph, rihtwis man, mid godcunde fultume gehealp
(Orosius 32.26)
'And him then Joseph, a righteous man, helped with holy aid.'
b. Swelcum monnum Dryhten cidde burn bone witgan (CP 27.12)
'Such men the Lord reproached through the prophet.'
c. Ic secge past behefe ic eom ge cingce & ealdormannum
(AEColl. 150)
'I say that useful I am to the king and the chiefs.'
d. Be oam is awriten oaet betera beo se geo1dlega wer oonne se
gielpna (CP 217.10)
'Therefore it is written that better is the patient man than the
boastful (one).'
the fact that non-intial conjuncts have another special word order possibility,
namely verb-initial word order, as in (20).
Given the phrase structure we proposed for Germanic, the entire range of
coordinate-clause word orders can be derived from the assumption that Spec-
C and C° may be omitted in non-initial conjuncts.
This approach gains support if we examine the conditions under which con-
joined main clauses must have verb-second word order. Sentences conjoined
with ond or oppe have obligatory verb-second order if they begin with one of
the following things: (a) a wh-phrase, (b) ba, bonne 'then' or (c) ne 'not'
(Mitchell (1985, §1723 ff. 1753, 1870)):
(22) a. Hu mceg la se blinda laedan pone blindan? (ALHom 13.124)
'O how can the blind lead the blind?'
b. ...andba wearS he oferswiSed (&CHom i. 176.15)
'and then he was overcome'
c. ...and ne mceg nan bing his willan wi5standan (JECHom i.10.1)
'and nothing can withstand his will.'
But these are the very constituents which, as we concluded in the preceding
section, are fronted to Spec-C, for they always induce V-to-C° movement, and
fail to attract pronominal subjects. So we can reduce these special cases to
the previously established generalization that sentences undergo V-to-C° move-
ment if and only if they begin with a lexically unfilled complementizer posi-
tion. Since the second conjuncts in (22) have overt wh-elements, they
necessarily have a C° head, and therefore show obligatory verb-second posi-
tion.
Now we also have the explanation for why the other Germanic verb-sec-
ond languages do not allow verb-final word order in conjoined main clauses.
In those languages C° is obligatory; it is only in Old English that it is op-
tional. We drew this conclusion from other evidence in SECTION 2.1, on the
basis of the fact that Old English is the only language which allows the verb
to remain in situ in main clauses, as in (6). The obligatoriness of C° in the
other Germanic languages would make the structure (4c) unavailable in them,
explaining why they do not have conjoined sentences with verb-final word
order as Old English does.
Since the other structures, namely (4a,b), are available in main clauses in
the other Germanic languages, non-initial conjuncts with verb-initial order of
the type (20) should occur in them. And in fact they are well attested in both
Old High German and Old Icelandic (Behaghel (1932. §1447, ff.); Heusler
(1962:175); Nygard (1905:347)):
(23) End fuorun flz sine scalcha in dea uuega" enti kasamn6tun all so
huuelthhe so sie funtun, ..., enti uuarth arfullit des bruthlauftes
kastuoli (Monsee-Vienna Fragments, Matth. 22.10)
'And his servants went out into the streets and gathered together
whoever they found, and the wedding table became filled.'
150 THE DIACHRONY OF VERB SECOND
(24) Magnus konungr f6r um haustit alt austr til landsenda, ok var hann
til konungs tekinn um alt land (Heimskringla 518, 34)
'King M. rode in the fall to the easternmost end of the land and he
was chosen king over the whole country.'
One fact which this anaysis leaves unexplained, however, is that it does not
account for the fact that verb-final yes/no questions and commands such as
(21) essentially occur only in non-initial conjuncts.
In any case, the evidence from coordinate sentences in the older Germanic
languages supports the special positional status of wh-elements and of pa,
ponne, and ne.
(26) Ac for paem he geneode swioost over pone munt oe he wiste baet
Flamineus se consul wiende paet he buton sorge mehte on baem
wintersetle gewunian (Orosius 188.30)
'but because he ventured most quickly over the mountain that he
knew the Flamineus the consul thought that he might dwell on in
winter quarters without care.'
Left-branch constraint violations (extraction of genitive modifiers from NPs)
occur in bare-complementizer relative clauses, but not in pronominal relative
clauses:
(27) se god pe bis his beacen waes
'the God that this was his sign'
sja maor, er ver segjum nu fra jartegnum, atti marga laerisveina
'the man, that we now tell about the signs had many disciples.'
NPs bearing different cases in conjoined relative clauses may be relativized
by a single bare complementizer, but not by a single pronoun + complementizer
complex:
(28) bam freondum pe ic lufige and me lufiao
'to the friends that I love and love me' (Acc. and Nom.)
pau Qtto son, er Haraldr konungr ios vatne ok gaf nafn sitt
'they had a son, that king Harald baptized and gave his name' (Acc.
and Dat.).
That-trace constraint violations occur in bare-complementizer relative
clauses but not (or very rarely) in pronominal relative clauses:
(29) ne lufige ge oisne middangeard oe ge geseoo oaet lange wunian ne
rnaeg (Alc. Th.XL. p. 614)
'do not love this world that you see (that) cannot last long.'
Relative clauses with a demonstrative pronoun adjoined tope behave much
as the relative clauses of modern English, German, French, etc.25 But, as Allen
and Maling (1976) observe, the properties of the other type of relative clause
suggest that there is no wh-movement involved. To account for it they pro-
pose a rule which deletes the relativized item in situ under identity to the head
of a relative clause. Allen considers and rejects the possibility of deriving them
by "free deletion" of a pronoun in the relative clause. However, it is possible
that the "gaps" are really null resumptive pronouns and that there is no con-
trolled deletion rule.26 One reason why this is an attractive possibility is that
overt resumptive pronouns are common in bare-pronoun relative clauses, as
in (30), but rare in pronominal relative clauses. For a proposal along these
lines, see Kemenade (1987:164).
(30) se god be bis his beacen waes
'the God that this was his sign'
sa er skilgetenn er faper hans
'he that his father is born in wedlock.'
152 THE DIACHRONY OF VERB SECOND
This agrees exactly with the description of the word orders in main and
subordinate clauses in Old English that is presented in Allen (1974). The deri-
vations are obvious. The underlying order is S O Vinf Vf. Of the two main-
clause orders, the first arises directly from V-to-C°, the second by the
additional application of Heavy NP Shift, or (if we adopt Pintzuk's proposal)
from the head-initial base structure. No further rules are applicable. In subor-
dinate clauses, V-to-C° movement is inapplicable and the underlying order is
realized directly as (32a). The type (32b) is the crucial case demonstrating
that Heavy NP Shift must be assumed in any case, even if head-initial VP/IP
is postulated. Verb raising yields (32c) and (32d), and, compounded with
Heavy NP Shift (or underlying VO order), (32e). *S ViO Vf is not derivable.
Hale uses primarily two pieces of evidence to establish that Vedic had such
a constituent structure. The first is that exactly one constituent can appear in
front of wh-elements, both in main clause questions and in relative clauses:
(34) ratham ko m'r avart ayat (RV. 10.135.5)
chariot-ACC who down rolled
'Who rolled out the chariot?
(35) sahasras'rngo vrsabho yah samudrad ud acarat, tena
thousand-horned bull which sea-ABL out rose that-INSTR
sahasyena vayam ni janan svapavamasi (RV. 7.55.7)
mighty-INSTR we in people-ACC put to sleep
'The thousand-horned bull that rose from the sea, with that mighty
one we put the people to sleep.'
The second argument, from the placement of clitics, reveals the constitu-
ency in a still more interesting way. Hale shows that clitics (all of which are
enclitics) are of two distinct types. The first type comprises emphatic clitics,
and conjunctions such as va, gha. The second type comprises pronouns.
Emphatics and conjunctions are placed after the first constituent of the entire
sentence, and clitic pronouns are placed at the beginning of S, unless there is
no fronted element, in which case they are placed after the first word of S.
See the relative clauses in (36), with va 'or' and te 'to you' positioned re-
spectively before and after the relative pronoun in Focus position:
(36) a. s"[ urau va va s'[ ye s[ antarikse madanti ] ] ]
(RV. 3.6.8)
wide-LOC or who atmosphere-LOC rejoice-3PL
'or who rejoice in the wide atmosphere'
b. s"[ idhmam s"[ yas s[ te jabharac
kindling-ACC' who you-DAT carry=PerfSubj
chasramana'h ] ] ] (RV. 4.12.2)
exerting himself
'who, exerting himself, bears the kindling to you."
Halpern's (1992) theory of cliticization accounts for the two types of clitics
in a simple way. Emphatics and conjunctions belonging to the functional sys-
tem, are syntactically affiliated with S", and pronominal clitics being argu-
ments, are syntactically affiliated with S. Both require a prosodic host on their
left, which is supplied through prosodic inversion if the base position does
not provide it.
Hale (1989) shows that the Hittite system is strikingly similar. In Hittite,
relative clauses whose head is definite, allow the topicalization of one con-
stituent out of the clause into the pre-wh slot. The proposed constituent can
be either the head, as in (37a), or some other constituent, as in (37b):
(37) a. nu DUMU-an kuin hukkiskimi
PRT child-ACC which-ACC treat-magically-lSg
'the child which I treat magically'
INDO-EUROPEAN ORIGINS OF GERMANIC SYNTAX 155
Secondly, extraposition cannot easily account for cases when the "head" is
contained in the relative clause itself, which in fact is statistically the most
frequent case for Vedic and is common in the other languages that preserve
adjoined subordinate clauses:30
(38) yam yajnam adhvaram vis'va'tah
which-ACC sacrifice-ACC service-ACC everywhere-from
paribhur asi, sa id devesu gachati (RV. 1.1.4)
surrounding are-2SG, that FOCUS gods-LOC go-3Sg
'It is the sacrifice (and) service that you surround on all sides which
reaches the gods,'
(39) ostiumquod in angiporto est horti, patefeci fores
gate which in alley-ABL is garden's open-PerflSg doors-ACC
(Plautus. Most. 1046)
'The garden gate which is in the alley I have opened (its) doors.'
"Lowering" of the head into an embedded relative clause is ruled out on gen-
eral grounds, so under the extraposition analysis some kind of reanalysis of
the head into the relative clause would have to be assumed for cases like (38).
On the adjunction account they are directly generated.
Thirdly, there can be no question of extraposition when both clauses have
separate lexical heads:
(40) a. sa gha viro na risyati, yam mdro
that EMPH man not perish-3Sg, which-ACC Indra.
brahmanaspatih somo hin6ti martyam (RV. 1.18.4)
Brahmanaspati, Soma support-3Sg mortal-ACC'
'The man who does not perish is the mortal whom Indra Brah-
manaspati Soma supports.'
b. yo martyah sisite aty aktubhir, ma
which mortal sharpen-Mid-Sg overly nights-INSTR, not
nah sa ripur TSata (RV. 1.36.16)
us-GEN that trickster dominate-Subj3Sg
'As for the mortal who makes himself too sharp by night, may
that trickster not gain power over us'
(41) nuza ANA DINGIR.MES kuit arkuwar iyami
PRT-REFLto gods which-ACC prayer-ACC make-lSg,
nukan AWATEMES ANA DINGIR.MES anda Sunni (KUBVl 46 11)
now words to gods in fill'
'As for the prayer I make to the gods, report the words to the gods.'
and when the lexical content of the "head" NP is distributed over both clauses:
INDO-EUROPEAN ORIGINS OF GERMANIC SYNTAX 157
is reanalyzed as:
4. Conclusion
I have argued that the rise of V-to-C° movement is just one of several re-
flexes of the introduction of complementizers which itself is just one of sev-
eral reflexes of the shift from adjoined to embedded clause structure. The most
celebrated feature of Germanic syntax turns out to be only the final stage of a
complex causal chain of syntactic innovations. Chronologically, too, it is the
last of the major Germanic syntactic innovations. The shift from adjunction
162 THE DIACHRONY OF VERB SECOND
to embedding, as well as the associated rise of the C° slot with its indeclin-
able subordinating complementizer, are pan-Germanic and most likely Proto-
Germanic. V-to-C° movement, on the other hand, belongs in a group of
innovations that are absent from Gothic (and probably from the earliest Scan-
dinavian of the Runic inscriptions), and which begin to spread shortly before
the historical period, with language-specific variations, throughout the rest of
the Germanic family.34
The last stage in this process is the elimination of the variation between
CP and S main clauses, preserved in the guise of apparent optionality of verb
second in Old English. This homogenization went in two opposite directions.
In English, the possibility of main clause CP was eliminated, resulting in the
loss of verb second (and, concurrently, of main clauses with overt
complementizers such as (3)). In the other Germanic languages, CP became
obligatory in main clauses, resulting in fixed verb-second word order. The
diversity of Old English word order is then an archaic trait, as is its treatment
of pronominal clitics and the structure of its relative clauses.
If this analysis is right, then the inherited distinction between the Topic and
wh landing sites is not only preserved in Old English but ultimately lies be-
hind the Modern English "residual verb second" system as well. The contrast
between Where will Max put a book? and On the table Max will put a book
has Indo-European roots. The other Germanic verb-second languages have
innovated by effectively collapsing the two positions (Wohin wird Max ein
Buch legen?, Aufden Tisch wird Max ein Buch legen.) In this respect at least,
English is syntactically the most conservative of all the modern Germanic
languages.
Notes
1. This is a revised and updated version of a paper read in 1989 at the 8th East
Coast Indo-European Conference at Harvard and at a DYANA workshop on
parametric variation in Germanic and Romance at Edinburgh, as well as in
talks at Stanford, Berkeley, and Helsinki. I would like to thank the discus-
sants on each of these occasions, especially Andrew Garrett and Elizabeth
Traugott, as well as an anonymous reviewer, for their most helpful comments.
2. The insight that subordinate-clause word order is basic goes back to Bach
(1962), Bierwisch (1963), and Koster (1975); the idea behind the V-to-C°
analysis is anticipated in the work of Diderichsen (1941) (on which see Heltoft
(1986)) and in Drach (1937). For two recent detailed theories of verb move-
ment, with full discussion of the recent literature, see Viker (1990) and Rob-
erts (1993). Lenerz (1984, 1985), and Kemenade (1987), show that the V-to-
C° analysis is also supported for the older Germanic languages (excepting
Gothic), though there are some differences in detail, as we will see below.
3. The obligatoriness of V-to-C° movement has been ascribed to the ECP (Travis
(1984)), to the feature composition of C° (den Besten (1983); Haider (1986);
Lenerz (1985); Holmberg (1986); Holmberg and Platzack (1988); Roberts
(1993)), to the need to assign Nominative Case to the subject (Platzack (1986)),
INDO-EUROPEAN ORIGINS OF GERMANIC SYNTAX 163
14. In fairness it should be noted that this solution was not available to Maling
and Zaenen since at that time they were not assuming V-to-C° movement but
rather a verb-second surface filter. In fact, they took these data as evidence
that wh-movement in Icelandic is adjunction to S to the left of C°, while
topicalization is adjunction to S, to the right of C°.
15. In Old English, as well as in Modern English and Mainland Scandinavian
languages, topicalization in embedded that-complements is allowed only af-
ter a class of non-factive verbs. Wechsler (1990) argues that this is a conse-
quence of the special illocutionary status of these complements as "asser-
tions". These clauses seem to represent the thought or words of the implicit or
explicit agent of the containing verb. In Icelandic, however, embedded pad-
clauses are reported to undergo topicalization freely.
16. This statement anticipates the treatment of relative clauses discussed in SEC-
TION 2.5.
17. See Kemenade (1987, chs.4-5) for other arguments that Old English weak
pronouns are "syntactic clitics" and for extensive discussion of preposition
stranding.
18. See Kemenade (1987:138), Mitchell (1985, §1599 ff., §3922 ff.) for the evi-
dence. Old Icelandicpa 'then' behaves in a similar way (Heusler (1962:174)).
19. For pa, though, things are a bit complicated because, in addition to being an
adverb, it can also be a subordinating conjunction, in which case the verb
must naturally be in final position. It is sometimes not possible to tell which it
is in a given sentence. Moreover, there appear to be some instances where the
adverb pa is topicalized rather than focused (Mitchell (1985, §2547 ff.)).
20. Archaic Old English seems to have had a different system in which pronouns
were proclitic (Fourquet (1974); Hock (1988)).
21. Setting aside, as before, the possibility of "embedded root clauses" under
verbs of saying and believing (fn. 15).
22. Hock (1988) suggests that the Germanic pronoun + complementizer complex
arises by rebracketing of the main clause correlative with the indeclinable
complementizer of the subordinate clause. This is an interesting possibility,
but seems to predict wrongly that the pronoun should be inflected in the case
required by the main clause, rather than in the case required by the subordi-
nate clause, and fails to explain why the pronoun + complementizer complex
is frequently accompanied by a correlative pronoun in the main clause, as
well as the obligatoriness of pied piping.
23. The two types of relative clauses are comparable to the nasalizing and leniting
relative clauses of Irish (Hale n.d.; McCloskey (1979); McCone (1980); Sells
(1984)), which have been argued to involve deletion (or, more precisely, null
resumptive pronouns) vs. wh-movement, respectively (see especially
McCloskey (1990)). Since Irish (and Celtic in general) shares also V-to-C°
movement with Germanic, there may be a deeper structural parallelism here.
24. In this section, unless otherwise noted, Old English examples are cited from
Allen and Old Icelandic examples are cited from Nygaard (1905:256-265)
and from Heusler (1962:158-163).
INDO-EUROPEAN ORIGINS OF GERMANIC SYNTAX 165
25. As Allen shows, there is a parallel distinction between relative clauses with
no overt subordinatorpe, depending on whether a demonstrative pronoun is
present or not. These relative clauses behave in exactly the same way as the
corresponding clauses with overt pe. They can be considered as having a
phonologically null (but syntactically visible) complementizer functionally
identical to pe.
26. A similar proposal was made by Ingria (1980) for Modern Greek, who showed
that it has counterparts of both Germanic types of relative clauses.
27. Let us also repeat here that Old English, like Modern English and the Scandi-
navian languages, might very well have had "root" phenomena in subordi-
nate clauses, which would have constituted another source of apparent ex-
ceptions to the basic verb-final order.
28. Particles such as nu belong in a pre-topic position (perhaps as proclitics to
S").
29. That is, the correlative pronouns could be omitted only under the same condi-
tions as pronouns in general could be omitted, namely rather freely when
nominative or accusative, and only very marginally in other case forms.
Hettrich (1988:546) notes that the conditions are different for non-restrictive
relative clauses.
30. Really this is not an autonomous generalization but a reflection of two other
tendencies: the relative clause usually precedes the main clause, and which-
ever clause comes first usually contains the "head". See Hettrich (1988:578,
682) for statistics and discussion.
31. O'Neil (1977) has claimed that relative clauses are "(almost) always at the
margins of the main clause" in Old English. Mitchell (1985, §2288 ff.) how-
ever, notes that they are "commonly found in the position usual in MnE, viz.
immediately after the antecedent or separated from it by a word or phrase
which qualifies it."
32. To say that subordinate clauses were adjoined is not to say that their syntax
was identical to main clauses. They were still subordinate clauses, linked to a
correlative element in the main clause. However, Hettrich (1988) notes that
non-restrictive relative clauses in Vedic do have essentially the syntactic prop-
erties of main clauses; in particular, the relation between relative and correla-
tive is governed essentially by the same principles that govern anaphoric re-
lations between main clauses.
33. Hock appeals to the Penthouse Principle (Ross (1973)), but this is only legiti-
mate if verb second is assumed to be V-to-C° after all, for the only genuinely
syntactic evidence for the Penthouse Principle comes precisely from the sorts
of main-clause fronting phenomena which we have been concerned with here
(including most prominently topicalization and verb second itself); there is
no evidence that anything like a Penthouse Principle constrains syntactic rules
in general. Ross also cited a number of special semantic properties character-
istic of main clauses, such as applicability of conversational postulates such
as that which allows questions to be interpreted as requests, occurrence of
pseudo-imperatives (imperatives interpreted as conditionals), and certain idi-
oms, but these are probably consequences of the status of main clauses as
166 THE DIACHRONY OF VERB SECOND
References
Allen, C. (1974) "Old English Modals." University of Massachusetts Occa-
sional Papers in Linguistics 1:89-100.
Allen, C. (1977) Topics in English Diachronic Syntax. PhD Dissertation, MIT.
Reprinted in 1981, Garland Press.
Altmann, H. (1981) Formen der "Herausstellung" im Deutschen. Tubingen:
Niemeyer.
Anderson, S. (1992) A-morphous Morphology. Cambridge: Cambridge Uni-
versity Press.
Andersson, L.-G. (1975) "Form and Function of Subordinate Clauses." Mono-
graphs in Linguistics 1.
Bach, E. (1962) "The Order of Elements in a Transformational Grammar of
German." Language 38:263-269.
Bacquet, P. (1962) La structure de la phrase verbale a I'epoque alfredienne.
Paris.
Behaghel, O. (1932) Deutsche Syntax, IV. Heidelberg: Winter.
Besten, H. den (1983) "On the Interaction of Root Transformations and Lexi-
cal Deletion Rules." In W. Abraham, ed. On the Formal Syntax of the
Westgermania, 47-131. Amsterdam: John Benjamins.
Bierwisch, M. (1963) Grammatik des Deutschen Verbs (Studia Grammatica
II). Berlin: Akademie-Verlag.
Bolinger, D. (1977) "Another Glance at Main Clause Phenomena." Language
53:511-519.
Calboli, G. (1987) "Die Syntax der altesten lateinischen Prosa." In A. G. Ramat
et al., eds. Papers from the 7th International Conference on Historical
Linguistics. Amsterdam: John Benjamins.
Chomsky, N. (1991) "Some Notes on Economy of Derivation and Represen-
tation." In R. Friedin, ed. Principles and Parameters in Comparative
Grammar, 417-454. Cambridge, Mass.: MIT Press.
Delbruck, B. (1878) Syntaktische Forschungen HI. Halle.
Diderichsen, P. (1941) Scetningsbyningen i Skaanske Lou. K0benhavn:
Munksgaard.
Drach, E. (1937) "Grundgedanken zur deutschen Satzlehre." Frankfurt.
Dressier, W. (1969) "Eine textsyntaktische Regel der idg. Wortstellung. (Zur
Anfangsstellung des Pradikatsverbums)." KZ 83:1-25.
Fourquet, J. (1974) "Genetische Betrachtungen uber den deutschen Satzbau."
In W. Besch, G. Jungbluth, G. Meissburger and E. Nellmann, eds.
INDO-EUROPEAN ORIGINS OF GERMANIC SYNTAX 167
1. Introduction
In the study presented here I examine inverted declarative clauses from sev-
eral Old and Middle French texts written between the 13th and the 15th cen-
turies and suggest, on the basis of synchronic and diachronic differences within
this class of clauses, an explanation for the decline of verb movement to Comp
over the period in question. Although it has often been claimed (cf. e.g. Martin
(1980); Adams (1987a,b), Lemieux and Dupuis (this volume)) that French
remains essentially a verb-second language until the end of the Middle French
period (14th-15th c.), my data show clearly that not all inversions are strictly
of the verb-second type, despite many surface indications that are consistent
with such an interpretation. We will look at two developments that we claim
had their origins at least as early as the 13th century, and which came to domi-
nate the grammar in Middle French: the coexistence of "Germanic" verb-
second inversions and Romance "free inversion" (SECTION 2), and the
dissociation of the initial constituent from the inverted clause, also respon-
sible for the rise of verb-third orders (SECTION 3). I will claim that the varia-
tions observable in 13th-century inversions are the seeds from which the
decline of the verb-second constraint in French developed.
It is important to emphasize that the focus of this study is the decline, rather
than the actual loss, of verb-second inversions in French. The studies of Kroch
(1989) and Roberts (1993), for example, show clearly that parameter reset-
ting takes place only after the gradual accumulation of evidence for the new
grammar results in a new generation's inability to construct the old grammar
out of the linguistic data it receives. In the case of verb-second in French,
the crucial generation seems to have lived in the 16th century, well after the
time period we will consider. I will assume for present purposes that the gram-
173
174 VERB SECOND AND THE NULL-SUBJECT PARAMETER
matical property at issue is the movement of the verb to Comp in root clauses
(V-to-C), but it should be noted that the resetting of an even more abstract
parameter—such as the Nominative-Case assignment parameter of Roberts—
is likely to be ultimately responsible for the loss of verb second. What con-
cerns us most here is simply the way in which certain inversion structures came
to be attributable to other grammatical properties than V-to-C, reducing the
amount of overall positive evidence for this movement in the late Old French/
Middle French period.
rather than C. Embedded inversion is rare in the Queste but not impossible;
ten cases exist, to my knowledge, in the entire text (but see also note 2).
Example (6) is representative:
(6) ...quant vint par la volent6 Nostre Seignor Calogrenant, uns
when came by the will (of) Our Lord Calogrenant a
chevaliers de la meson le roi Artus... (Q 190,27)
knight of the house (of) the king Arthur
An ambiguous inverted word order is thus possible in OF, variably analyzed
as the result of simple V-to-C movement or of free inversion:
(7) maintenant s'en ala la damoisele (Q 112,22)
now refl-'en' went the maiden
'now the maiden went away'.
Either this clause is directly parallel to (2), where the subject is in Spec IP, or
it is an instance of a VP-final subject, as in (5), which happens not to be sepa-
rated from the finite verb by any other VP material. If OF had obligatory V-
to-C movement, as we have assumed, then only the position of the subject is
in fact ambiguous. But since these types of clauses are not crucially depen-
dent upon V-to-C, and since we know that French has lost V-to-C movement
as a productive rule of grammar, we are led to examine (5a-d), and (7) as pos-
sible sources of diachronic change. In so doing, we extend and unify sugges-
tions made elsewhere (see Vance (1988) with respect to unaccusative and
passive declaratives and Roberts (1993) primarily on interrogatives) that the
possibility of interpreting certain CVS structures as either free inversion or
Germanic inversion can trigger grammatical reanalysis over time.
It is important to note that pronominal subjects are not possible in the in-
versions of either (5a-d) or (6). Vance, largely following Kayne (1983), ac-
counts for both facts by proposing that subject pronouns cannot occur lower
in the tree than Spec IP, and that they may in fact be enclitic on Comp when
postverbal. Roberts (section 2.2.2) invokes two additional arguments for this
condition on subject pronouns which suggest that encliticization to Comp may
be a general property of subject pronouns in OF. In inverted matrix clauses,
a pronominal subject must be immediately postverbal; in embedded clauses,
it must be preverbal, immediately to the right of the subordinator. The hypo-
thetical pronominal versions of (5a) and (6) are then (5a') and (6'), respec-
tively. These hypothetical sentences correspond exactly to the well-represented
sentence types illustrated in (1) and (4) above.
(5a) Si plorerent assez a cest departement
thus cried much at this departure
cil qui plus cuidoient avoir les cuers et durs et
those who most thought to-have the hearts both hard and
orgueillox (Q 26,19)
proud
ON THE DECLINE OF VERB MOVEMENT TO COMP 177
Table 7.1
Pronominal vs. Non-pronominal Inversion
in Old and Middle French4
Total CVSp CVSn CVXSn
Queste (1225) 419 212 (.51) 138 (.33) 69 (.16)
pp. 1-41,71-115
Joinville (1306) 74 24 (.32) 43 (.58) 7 (.10)
pp. 104-125
Saintre (1456) 44 15 (.34) 23 (.52) 6 (.14)
pp. 239-294
Commynes (1491) 93 11(.12) 50 (.54) 32 (.34)
pp. 58-97
Although the total numbers of inversions in each text are not directly com-
parable, due to the difference in the lengths of the passages, it is clear that
subject inversion, while still a prominent word order in MidF, becomes less
frequent after the 13th century. (Compare 93 inversions per 40 pages in
Commynes, the MidF text which most favors inversion, to roughly 210 in-
versions in 40 pages of the OF text.) Furthermore, within the class of inverted
word orders we find some interesting diachronic patterns. The proportion of
pronominal inversions (CVSp) to other types of inversion decreases steadily
over time. The percentage of CVXSn inversions appears to remain stable until
the late 15th century, whereas CVSn jumps to 50% of inversions by the be-
ginning of the 14th century and remains there. Recall, however, that the cat-
egory CVSn artificially groups together both simple V-to-C inversions,
identical in structure to CVSp, and ambiguous clauses such as (7), which may
be either simple V-to-C clauses or examples of CVXSn in which the subject
is VP-final but X=0. In order to interpret properly the proportion of non-pro-
nominal inversions in the data, then, we must divide this category further into
ambiguous and unambiguous clauses.
In establishing criteria for identifying ambiguous clauses, I will once again
assume that Spec IP is the normal case position for subjects. In a clause such
as (2), then, the subject is unambiguously in Spec IP, to the left of the object
in VP, NP. Also unambiguous are instances in which the subject separates the
finite verb from a non-finite verb, as in (8) (the object pronoun vos is a clitic
on the finite verb):
(8) autresint vos a Nostre Sires esleu (Q 38,16)
in the same way you has Our Lord elected.
On the other hand, the ambiguous clauses—those in which the subject may in
fact be VP-final—have a simple verb and no postverbal object. They may,
however, have extraposed PP's analogous to the one in (5d):
(9) ainz demorront li plusor en ceste Queste (Q 17,6)
rather will-remain the most in this Quest.
ON THE DECLINE OF VERB MOVEMENT TO COMP 179
Table 7.2
Simple V-to-C and VP-final Subject Clauses
in Old and Middle French
Simple V-to-C Clauses VP-final Subject Clauses
CVSn
CVSp CVSnV(-fin) CVSn# CVXSn Total VP-final
CVSnO
Q 212 (.51) 27 (.06) 111 (.26) 69 (.16) 180 (.43)
J 24 (.32) 6 (.08) 37 (.50) 7 (.10) 44 (.60)
S 15 (.34) 1 (.02) 22 (.50) 6 (.14) 28 (.64)
C 11 C12) 3 (.03) 47 (.50) 32 (.34) 79 (.85)
180 VERB SECOND AND THE NULL-SUBJECT PARAMETER
Another clue to the changing nature of inversion in late OF and MidF comes
from the high proportion of unaccusative and passive verbs in VP-final inver-
sions. Since the subjects of such verbs are underlyingly direct objects, it might
be supposed that they favor word orders in which movement out of the VP is
not necessary. Table 7.3 contrasts the percentage of unaccusatives and passives
in pronominal inversions to that in non-pronominal inversions. Although some
of the samples are too small to provide reliable percentages, the general trends
are clear. At each stage of the language, unaccusative and passive verbs occur
in a relatively small percentage of CVSp clauses which, it will be recalled,
are merely the result of V-to-C and should take place without respect to the
type of verb involved. A larger portion of CVSn clauses, and even more of
the CVXSn clauses, contain unaccusative and passive verbs. Furthermore, the
percentage of unaccusatives and passives in non-pronominal inversions in-
creases over time. These observations are consistent with the notion that such
inversions rely more and more on options other than verb-movement, e.g. the
availability of the direct object position as a location for a surface subject. It
should be pointed out that the proportion of unaccusatives and passives in
CVSp clauses rises as well. These figures are unexpected; if CVSp represents
the conservative inversion type, as we have argued, then its character should
not change over time. A possible explanation is that the overall number of
unaccusatives and passives in the texts increases, and that all categories of
inversion reflect that increase. If this account is correct, then Table 7.2 gives
evidence of synchronic differences between inversion types, as we have as-
sumed, but not of diachronic change. To verify this possibility we would need
figures on the rate of unaccusatives and passives in a general sample from each
text, to compare with the rate in inversions. Such figures are available at
present only for the Queste, where the overall percentage of unaccusatives and
passives is 40% (greater than pronominal inversions but less than non-pro-
nominal inversions). Although more work remains to be done, it seems un-
likely that the overall distribution of verb types in French should change over
time. It is more plausible that the low number of CVSp clauses in the later
texts simply makes the percentages less reliable, and therefore not directly
comparable with, that of the Queste, where CVSp accounts for fully half of
the inversions.
Table 7.3
Percentage of Unaccusative and Passive Verbs
in Old and Middle French Inversions
CVSp CVSn CVXSn
Queste 42 of 208, 20% 67 of 137, 49% 76 of 121, 63%6
pp. 1-41, 71-115
Joinville 2 of 24, 8% 24 of 43, 56% 3 of 7, 43%
pp. 104-125
Saintre 5 of 15, 33% 18 of 26, 69% 5 of 6, 83%
pp. 239-294
Commynes 5 of 11, 45% 39 of 50, 78% 25 of 32, 78%
pp. 58-97
ON THE DECLINE OF VERB MOVEMENT TO COMP 181
OF additionally allows free inversion of the Italian type. In Italian, such clauses
have the structure of (12a) (NP-movement) or (12b) (unaccusatives and
passives). (We follow here Rizzi (1982); Burzio (1986); and Pollock (1986).)
(12) a. [IP pro [r V [VP VP subject]]]
b. I[P pro [I'V [VP ...[NP subject]]
The verb-second character of OF, however, prevents surface verb-first struc-
tures like (12) from occurring; instead, the verb-second constraint and the free
inversion possibility work together to produce the hybrid constructions (13a)
and (13b), illustrated by (5) and (7) above:
(13) a. [CP XP [c' V [IP pro [I' t [VP...] subject]]
b. [CP XP [c' V [IP pro [I' t [VP... [NP subject]].
The numbers we have seen in Tables 7.1, 7.2 and 7.3 suggest that the struc-
ture in (13) is, however, unstable. Speakers of OF are already forced to posit
inversions within IP for sentences like (6):
(14) quant [IP pro [I' vint [VP t par la volente" Nostre Seignor
[NP Calogrenant]].
If Spec IP is open to non-subjects, then hybrid verb-second clauses like (13)
could be reanalyzed as matrix IP's in which free inversion, but not Germanic
inversion, has taken place. There is indeed evidence that in the 13th century
Spec IP may be an A' position. This evidence comes from a very few embed-
ded clauses, where we can be sure that an initial non-subject is in Spec IP
rather than Spec CP:
(15) quant en si haute bonte et en si haute chevalerie
when in such high goodness and in such high knighthood
seroit fichiee la bosne de son lignage (Q 221,13)
would-be fixed the end of his lineage
'when the last of his descendents would be fixed in such great good-
ness and chivalry'.
It seems plausible, then, that at least some speakers of late OF interpreted
inversions of the type (5a-d) and (7) as IP's rather than CP's. Hence (13a-b)
are simplified to (16a-b):
(16) a. I[P X P [ I ' V [yp...] subject]]
b. I[P XP [I'V [VP... [NP subject]].
Another way of putting this observation is that the data available to speakers
does not force the conclusion that V-to-C is obligatory in all matrix clauses.
182 VERB SECOND AND THE NULL-SUBJECT PARAMETER
to the left of the clause. Finally, we note the complete breakdown of the former
verb-second system in late MidF, where inversion may occur independently
of the presence of a fronted constituent. Our data in this section are slightly
more complete than in SECTION 2 and help confirm the trends noted there.
which may or may not be focussed. Furthermore, when direct objects, past
participles, or infinitives are fronted, they are more than twice as likely to occur
with postverbal Sp as with postverbal Sn. This situation hints at a contrast
between the syntactic behavior of true topics—those constituents fronted from
their normal position within the VP—and adjuncts or adverbs which happen
to occur clause-initially. The true topics occur preferentially in clauses obliga-
torily interpreted as CP's, while those clauses open to an IP interpretation host
topics more rarely. In fact, the evidence for a split between CVSp and CV(X)Sn
clauses with respect to topicalization is more significant than can be seen in
Table 7.4. As initial constituents the adverbs si and lors are very frequent:
these adverbs may have little semantic content, as example (5a) above shows
where the translation "thus" implies no particular causality. But these "mean-
ingless" adverbs are more frequent in the non-pronominal inversions. The
overall percentage of si and lors combined within the adverb class for each
of our inversion contexts is given in Table 7.5. These two adverbs are each
much more frequent than any other single adverb. Here our pronominal ex-
amples can be seen to make use of these adverbs approximately half as fre-
quently as the non-pronominal examples. This contrast is even more striking
when we consider that a large number of the instances of si in CVSp con-
structions are in fact topicalization constructions. When si triggers the inver-
sion of a subject pronoun, it usually has a contrastive function linking it to
the previous clause:9
(17) ainz vos fust avis, se vos les veissiez, qu'il en
rather to-you would-be mind, if you them would-see, that-they 'en'
fussent trop lie et si estoient il sanz faille. (Q 25,14)
would-be very happy and so were they without fail
'rather you would have thought, if you had seen them, that they were
very happy about it, and so they were indeed'.
The observations just made complement nicely the evidence that Roberts
(1993, section 2.3.1) offers for the reanalysis of matrix SVO clauses as IP's.
Pointing out that in OF the expletive subject pronoun il, which cannot be a
topic, could occur in matrix SVO constructions, Roberts (who also cites work
by Cardinaletti) hypothesizes that speakers interpret the initial position in SVO
clauses as an argument position rather than a topic position. They are then
led to assume that the subject is in Spec IP rather than Spec CP. Parallel to
Table 7.5
Si and Lors Within the Adverb Class
CVSp CVS CVXS
Table 7.6 above shows the ratio of inversion to non-inversion after an ini-
tial constituent for subject pronouns and subject NP's combined. The figures
for the Queste del Graal are slightly misleading because they include sentential
adverbs from the class we have recognized as unable to trigger inversion. If
such adverbs are not included, the figure for inversion after a non-subject
constituent is virtually 100%. The text Jehan de Saintrt appears to be slightly
anomolous in preferring non-inversion to a greater extent than the other MidF
texts. It is clear, however, that at some point between 1225 and 1306 an im-
portant change took place in French with respect to word order. Whereas in
the 13th-century verb-third orders were essentially unknown, verb-third
clauses abound in the 14th and 15th centuries.
From the numbers in Table 7.6 it would appear that, after an initial decline,
inversion stabilized over the MidF period. A slightly different situation is
suggested by Table 7.7, which breaks down the statistics in Table 7.6 accord-
ing to pronominal and non-pronominal subjects. In this table we observe a
steady decline in pronominal inversion after 1225, from nearly 100% to just
15% over a century and a half. The NP subjects, on the other hand, decline
initially and then vacillate between 50% and 73% in MidF. The fluctuation
of the percentage of CVSn clauses, I claim, reflects the fact that only a sub-
set of such clauses are directly affected by the more abstract grammatical
change in progress. Another way of looking at this is the following: because
both CSnV clauses and many of the CVSn clauses may be interpreted as IP's,
the contrast between them is in part superficial. The contrast between CVSp
and CSpV, on the other hand, is real, since if V-to-C fails to apply in the pres-
ence of an initial constituent only the latter is possible.
Table 7.7
Inversion vs. Non-inversion of Pronominal
and Non-pronominal Subjects After XP
CVSp CSpV % inv CVSn CSnV %Inv
Queste 97 3 97% 122 4 98%
1225
Joinville 24 17 59% 50 13 79%
1306
Froissart 15 26 37% 33 12 73%
c. 1375
Quinze Joies 17 28 38% 27 13 68%
1420
Jehan de Saintre 9 28 24% 30 30 50%
1456
Commynes 11 60 15% 80 29 73%
1491
ON THE DECLINE OF VERB MOVEMENT TO COMP 187
4. Conclusion
The data presented here show that a large portion of the inversion struc-
tures that occur in the MidF period can be attributed to a non-verb-second
grammar. The crucial clause-type is CV(X)Sn, a hybrid "verb second free
inversion" construction, which is consistent with both a grammar in which
obligatory V-to-C applies and with one in which it does not. My claim is that
the rise in this construction over the course of the 14th and 15th centuries
reflects the growing generalization of alternative means of producing inver-
sion structures, to the detriment of CVSp, produced strictly under V-to-C
movement. We have seen, furthermore, that the initial constituent of the clause
is gradually reanalyzed as being unrelated to inversion. Both changes in the
grammar of French had their origins at least as early as the 13th century, which
is not true of the cliticization of preverbal subject pronouns, often claimed to
be a source of the loss of verb second. This shift in the proportion of pro-
nominal and non-pronominal inversions in MidF has the effect of masking,
on the surface, the fundamental decline in verb second that is taking place.
Insofar as it views inversion structures as a source of grammatical reanaly-
sis (of CP's as IP's) in the history of French, my account is partially similar
to that of Hulk and van Kemenade (this volume). It also resembles in part
that of Lemieux and Dupuis (this volume), who see inversion structures as
IP's even in OF. Both these studies, however, consider IP inversions to be
true verb-second constructions. Under my account, the types of inverted word
orders produced in IP's and those produced in CP's are fundamentally differ-
ent. Specifically, CP inversions may—and, in the case of subject pronouns,
must—have immediately postverbal subjects. They are of course restricted
to matrix clauses, and they have an obligatory preverbal non-subject constitu-
ent. IP inversions, on the other hand, generally have VP-final subjects, strictly
exclude pronouns, and may occur in embedded contexts. Since they are es-
sentially "free inversions" and not verb-second structures, they may in fact
be verb-initial in MidF. It is the coexistence of these two underlyingly differ-
ent inversion constructions, I claim, that sets in motion the decline of verb
movement to Comp in the history of French.
The notions of language change that I have relied upon here are quite gen-
eral and should be applicable to other cases than the history of French. Spe-
cifically, my analysis makes a broad prediction which requires further
refinement: that SVO languages in which free inversion and verb-to-Comp
inversion coexist, should develop in the same direction as French, i.e. toward
the elimination of verb second characteristics. Many Northern Italian dialects
seem to have followed a similar progression (see for example Vanelli, Renzi,
and Beninca (1985) and Poletto (this volume)). Old Spanish offers another
192 VERB SECOND AND THE NULL-SUBJECT PARAMETER
Notes
* I am grateful to Deborah Arteaga, Anthony Kroch, Ian Roberts and Rex A.
Sprouse for helpful discussion and to the following contributors to this vol-
ume who shared drafts of their papers with me: Paola Beninca, Paul
Hirschbuhler, Aafke Hulk and Ans van Kemenade, Monique Lemieux and
Fernande Dupuis, David Lightfoot, and Christer Platzack. I would also like
to thank Ian Roberts for access to the manuscript of his forthcoming book,
which has now appeared as Roberts (1993).
1. For the sake of simplicity, I assume here that the base position of the subject
is Spec IP. A great deal of evidence exists, however, to place that position
within the VP or Vmax projection (see e.g. Zagona (1982) and Koopman and
Sportiche (1986, 1991)). The circumstances under which the subject moves
to the Spec of IP (or analogous position), presumably to acquire Nominative
Case, are understood differently by different authors.
If we assume, as I have here, that the subject normally occurs in Spec IP at
S-structure even if it originates in the VP, we predict that Old French does not
allow verb-second inversions in embedded clauses. If, on the other hand, OF
allows the subject to remain in a position to the left of the verb in VP, as
claimed by Adams (1987c), Dupuis (1989), and Lemieux and Dupuis (this
volume), it is predicted that OF is a symmetric verb-second language allow-
ing inversions in both main and embedded contexts. This matter is an empiri-
cal question of some importance. Although early Old French may have had a
symmetric verb-second system, I interpret prose texts from the 13th century
on to show little if any evidence of embedded clause inversion other than free
inversion (discussed below) and the type of inversion noted in note 2. On the
other hand, an analysis in which the VP-internal position of the subject is to
the right of V in OF, as in Roberts (1993), is compatible with the present
account since it predicts that embedded clause postverbal subjects will be
VP-final rather than immediately postverbal.
2. The occurrence of inverted clauses after the conjunction que is widely recog-
nized to be due to the embedding of an entire root CP under a second CP
headed by a complementizer, a phenomenon found in many verb-second lan-
guages (cf. Adams (1987a,b); Platzack (1983); Sigurosson (1990); Vance
(1988)). Such inversions are considered to be root phenomena and will not
be treated separately from matrix clauses in this paper.
ON THE DECLINE OF VERB MOVEMENT TO COMP 193
(i) (il troveroitque) au jor d'ui doit estre cist sieges aempliz (Q 4,15)
(he would-find that) today must be this seat filled.
Cist sieges is not VP-final, nor can it be in Spec IP, since it occurs after the
infinitive. It can, however, be analyzed as an underlying direct object which
precedes its past participle exactly as in (10). If this is the only interpretation
of (i) available to speakers of OF, then it seems that unaccusatives and passives
must play a syntactic role distinct from that of other verbs. This in turn lends
extra support to our analysis of (10) as ambiguous between a reading where
the subject is within the VP and one in which it occupies Spec IP like ordi-
nary subjects.
On the other hand, this example may reflect the sort of movement of non-
finite verbs to Spec VP proposed by D6prez (1989), where cist sieges would
be base-generated in the VP of aemplir and where estre moves to the specifier
of that VP. Such movement would not depend on the type of the verb. Al-
though Ddprez's analysis appears in most instances to introduce unnecessary
complexities into the grammar, there are two examples in the Queste of a
con-struction that resists most other interpretations. One of them is given
in (ii):
(ii) bien nos a tenu ceste damoisele convenant (Q 275,20)
well to-us has kept this maiden covenant
194 VERB SECOND AND THE NULL-SUBJECT PARAMETER
13. It should be pointed out here that there is a slight preference in Middle French
for CSpV over CSnV when the initial constituent is a direct object. Propo-
nents of the cliticization account interpret this fact as evidence for their posi-
tion, supposing that true VP complements, as opposed to adverbials, must
appear in Spec CP and thus can only co-occur with preverbal subjects if the
latter are clitics. The evidence in question, however, is not available prior to
the decline observed in Table 7.7 and so cannot be used to argue that
cliticization is a source of that decline. The discussion in Vance (1988) to the
effect that preverbal subject pronouns could be phonological but not syntac-
tic clitics in Middle French is consistent with the data examined here as well
as with the main thrust of the account in Kroch (1989), where rhythmic fac-
tors are held to be ultimately responsible for the loss of verb second. Further-
more, Kroch's account of the rise of the "reprise" construction in French makes
the importance of the behavior of objects in CSV constructions questionable;
both OVS and OSV are being replaced by a new structure that makes use of
left-dislocation and a resumptive pronoun.
14. In fact, SVO clauses may have been IP's in Old French as well as in Middle
French. If the grammar of the texts is representative, there are no data avail-
able to speakers of Old French that would force a CP interpretation of SVO
clauses. Under these circumstances, given the principle of economy of
Chomsky (1991) and Roberts' (1993, section 2.3.2) formal interpretation of
it, we would expect that the least complex possible structure, here IP, would
be posited from the start. I will not explore here the consequences of such an
account, which also forms part of the analysis suggested in Lemieux and
Dupuis (this volume), but the reader is referred to Travis (1984) and Zwart
(1991) for broader theoretical discussions of this view of verb second.
15. The preliminary results of this research, which require verification, indicate
that in at least one dialect of Rhaeto-Romance the status of free inversion is
different from that of French and Italian. From the earliest texts available
(16th century) to the present day, the Surselvan dialect appears to have VP-
final inversion only with unaccusative and passive verbs; even French-style
stylistic inversion with other intransitive verbs is prohibited. Such inversion,
however, has been observed in the Engadine dialects.
References
Primary
Adams, M. (1987a) "From Old French to the Theory of Prodrop." Natural
Language and Linguistic Theory 5:1-32.
Adams, M. (1987b) Old French, Null Subjects, and Verb Second Phenomena.
PhD Dissertation, UCLA.
Adams, M. (1988) "Embedded pro." In J. Blevins and J. Carter, eds. Proceed-
ings ofNELS 18, 1-21. Amherst, Mass.: GLSA, University of Massa-
chusetts.
196 VERB SECOND AND THE NULL-SUBJECT PARAMETER
Auger, Julie (1992) Pronominal Clitics vs. Agreement Affixes: On the Verb
Morphology of Colloquial Quebec French. Dissertation proposal, Uni-
versity of Pennsylvania.
Besten, H. den (1983) "On the Interaction of Root Transformations and Lexi-
cal Deletive Rules." In W. Abraham, ed. On the Formal Syntax of the
Westgermania, 47-131. Amsterdam: John Benjamins.
Besten, H. den (1985) "The Ergative Hypothesis and Free Word Order in Dutch
and German." In J. Toman, ed. Studies in German Grammar, 23-64.
Dordrecht: Foris.
Burzio, L. (1986) Italian Syntax: A GB Approach. Dordrecht: Reidel.
Calabrese, A. (1992) "Some Remarks on Focus and Logical Structures in Ital-
ian." Harvard Working Papers in Linguistics 91-127.
Chomsky, N. (1991) "Some Notes on Economy of Derivations and Represen-
tations." In R. Friedin, ed. Principles and Parameters of Comparative
Grammar, 417-454. Cambridge, Mass.: MIT Press.
Contreras, H. (1987) "Small Clauses in Spanish and English." Natural Lan-
guage and Linguistic Theory 5:225-243.
Deprez, V. (1989) "Sujets inverses dans Vigneulles." Ms. MIT and Universite
d'Ottawa.
Dupuis, F. (1989) L'expression du sujet dans les propositions subordonnees
en ancien francais. PhD Dissertation, Universite de Montreal.
Fontaine, C. (1985) Application de methodes quantitatives en diachronie:
I'inversion du sujet enfrancais. MA Thesis, UQAM.
Fontana, J.-M. (1993) Phrase Structure and the Syntax of Clitics in the His-
tory of Spanish. PhD Dissertation, University of Pennsylvania.
Foulet, L. (1930) Petite syntaxe de l'ancien francais. 3rd ed. Paris: Honore
Champion.
Giorgi, A. and G. Longobardi (1991) The Syntax of Noun Phrases: Configu-
ration, Parameters and Empty Categories. Cambridge: Cambridge
University Press.
Groos, A. and R. Bok-Bennema (1986) "The Structure of the Sentence in
Spanish." In I. Bordelois, ed. Generative Studies in Spanish Syntax, 67-
80. Dordrecht: Foris.
Hirschbuhler, P. (this volume) "Null Subjects in Verb-First Embedded Clauses
in Phillippe de Vigneulles' Cent Nouvelles Nouvelles."
Hulk, A. and A. van Kemenade (this volume) "Verb Second, Pro-drop, Func-
tional Projections and Language Change."
Kayne, R. (1983) "Chains, Categories External to S, and French Complex
Inversion." Natural Language and Linguistic Theory 1:107-139.
Kayne, R. and J.-Y. Pollock (1978) "Stylistic Inversion, Successive Cyclicity,
and Move NP in French." Linguistic Inquiry 9:595-621.
Koopman, H. and D. Sportiche (1986) "A Note on Long Extraction in Vata
and the ECP." Natural Language and Linguistic Theory 4:357-373.
Koopman, H. and D. Sportiche (1991) "The Position of Subjects." Lingua
85:211-258.
ON THE DECLINE OF VERB MOVEMENT TO COMP 197
Editions
La Queste del Saint Graal. Ed. A. Pauphilet. Classiques Francais du Moyen
Age. Paris: Champion, 1984. (c. 1225)
Joinville, Jehan de. La Vie de Saint Louis. Ed. Noel Corbett. Sherbrooke:
Editions Naaman, 1977. (c. 1306)
ON THE DECLINE OF VERB MOVEMENT TO COMP 199
1. Introduction
In this paper I will attempt to give a unified explanation of the loss of verb
second in English and French during the Middle Ages. Prior to the loss of
verb second, both languages had developed a grammar with verb second SVO,
and subject clitics; in such a grammar, there are very few signs which un-
ambiguously indicate verb second. The hypothesis I will defend in this paper
is that the existence of such a stage triggered the loss of verb second, and that
certain differences between Modern French and Modern English resulted from
the loss of verb second in combination with the fact that Infl was "weak" in
English but not in French at the time of the word order change.
Scholars studying the diachronic development of English and French seem
to agree that the medieval stages of these languages displayed verb-second
phenomena which were lost in the 14th and 15th centuries, whereas they may
have different opinions concerning the reason why verb second was lost. Con-
sider e.g. van Kemenade (1987) and Weerman (1989) for English, and Adams
(1987) and Vance (1989) for French. To my knowledge there is no attempt to
search for a common reason why both English and French lost verb second.1
Naturally there does not have to be a common reason; however, since verb
second seems to be quite a stable phenomenon once it has entered a language
(as evidenced by the bulk of the Germanic languages, both in the OV and the
VO variants); it appears to be a good research strategy to try to find a unified
explanation before attempting language particular solutions.
My discussion of the loss of verb second in English and French will be
conducted in the light of ideas which have developed in close contact between
me and Anders Holmberg during the last couple of years. Different versions
of this concept have been presented in Holmberg and Platzack (1988), Platzack
200
THE LOSS OF VERB SECOND IN ENGLISH AND FRENCH 201
and Holmberg (1989) and Holmberg and Platzack (forthcoming); unless there
is specific reason to refer to the earlier papers, I will confine myself to refer
to the last mentioned work, which contains the most recent and explicit pre-
sentation of our ideas.
From the point of view of this paper, the main hypothesis developed in the
works referred to in the last paragraph is that verb-second languages are lan-
guages where the tense affix and the finiteness feature [+F] are in different
functional heads (I° and C°, respectively), whereas non-verb-second languages
have tense and [+F] in the same head, I°. In terms of this description, the
loss of verb second in English and French is a change of position for [+F],
from C° to I°. Assuming that the head where the finiteness feature is realized
must be lexicalized in order to license Nominative Case, this description ac-
counts for the need to move the tensed verb to C° in main clauses of verb-
second languages, producing the verb-second effect, and the need to lexicalize
I° in non-verb-second languages (cf. Pollock (1989)). In subordinate clauses
of verb-second languages, movement of the tensed verb to C° is blocked by
the presence of a complementizer, leading to a word order difference between
main clauses and subordinate clauses.2 No such difference is predicted for
non-verb-second languages, where the presence of a complementizer in C° does
not interfere with the lexicalization of the head carrying [+F].3
My paper is organized in the following way. In SECTION 2 some of the lead-
ing ideas of Holmberg and Platzack are presented, followed by a survey of
recent attempts to explain the loss of verb second in English and French.
SECTION 3 is devoted to a presentation of the present description, whereas
SECTION 4 contains a discussion of some consequences of the proposed expla-
nation. SECTION 5 is the summary.
2. Background
Like Holmberg and Platzack, I will express the category finiteness as a fea-
ture [F], realized in some functional category. There are two values of this
feature: [+F], which is used to express finite categories, and [-F], which is
used to express non-finite categories. A unitary verb-second parameter may
now be formalized as in (1), expressing the hypothesis that the feature [+F] is
in C° in verb-second languages, in I° in non-verb-second languages:
(1) The Verb-Second Parameter
±([+F] is located in C°)
According to the principle of full interpretation, every element in a structure
must be licensed. Since the finiteness feature has the function of indicating
presence of a predication, it is conceivable that it should be interpreted as an
element relating a subject to its predicate; a way to implement this idea is to
assume that [+F] has to govern Nominative Case to be licit, according to the
condition in (2):
(2) Licensing Condition for the Finiteness Feature [+F]
An occurrence of the feature [+F] is licit if and only if the head host-
ing it governs a phonetically realized element bearing Nominative
Case, or the trace of such an element.
Furthermore, following Holmberg and Platzack, I will assume that subject-
verb agreement is represented by the feature Agr, realized in I°, and that Agr
may be either a nominal element with certain syntactic consequences, or just
a form of the tense affix. A similar idea is developed in Zagona (1988, e.g.
p. 168) to account for differences between English and Spanish. In Platzack
and Holmberg (1989) it is argued that both Old French and Middle English
(henceforth OF and ME, respectively) belong to a group of verb-second lan-
guages where Agr is a nominal element; I will adopt this idea in the present
paper. When Agr is nominal, it must bear Case: Holmberg and Platzack as-
sume that nominal Agr is inherently Nominative. As such it must be licensed
according to the same rules as other Nominative categories. The licensing
conditions for structural Nominative Case, assumed by Holmberg and Platzack
are given in (3):4
(3) Nominative is licit iff
a. it is head governed by a head with lexical features hosting [+F]
(direct licensing), or
b. it is governed by a member of a chain, the head of which is li-
censed by virtue of (3a) (indirect licensing).
Together, (2) and (3) express a mutual dependency relation between [+F] and
Nominative Case: just as every instance of Nominative Case must be licensed,
directly or indirectly, by the finiteness feature, every example of this feature
must have an instance of Nominative Case to license. Since (2) and (3) are
formulated in terms of government, we must indicate how government is de-
fined. Following Rizzi (1990) we assume two kinds of government: Head
Government and Antecedent Government;5 the definitions given in (4) and (5)
are taken from Rizzi (1990:25) unless otherwise stated:
THE LOSS OF VERB SECOND IN ENGLISH AND FRENCH 203
compatible with the fact that the Scandinavian languages, which are SVO, have
managed to uphold verb second. Secondly, as Weerman (1989:186) remarks,
main clauses with SVO order are at best ambiguous between the verb-second
interpretation and the interpretation that SVO is the basic word order. Hence,
even if the language learners of English and French at the time of the loss of
verb second experienced a majority of SVO main clauses, the structural am-
biguity of these clauses must have made them inappropriate as indicators for
or against a verb-second grammar. At the same time, the language learners
must also have met examples which explicitly signified the presence of verb
second, e.g. all main clauses where a constituent other than the subject is
fronted and the subject is in third position, following the fronted element and
the tensed verb. We thus have a situation where the language learner must
have experienced a certain number of sentences11 which unambiguously indi-
cated the presence of verb second, and a bulk of sentences which were struc-
turally ambiguous between a verb-second interpretation and a basic SVO
interpretation. It is unclear why the language learners should ignore these
unambiguous cases in favour of a particular interpretation of the ambiguous
ones. The conclusion must be that although Adams and van Kemenade may
be on the right track, their explanation has to be supplemented to account for
the loss of verb second in French and English.
Consider next the account of the loss of verb second in French proposed
by Vance (1989). Following Travis (1986), Vance assumes that verb-second
languages are languages which do not tolerate adjunction to IP of fronted el-
ements: hence, the loss of verb second in French is described as the intro-
duction of the possibility to adjoin fronted elements to IP. Since the loss of
is not in focus for Vance, she just states that there has been a resetting of the
parameter of adjunction to IP for adverbs and VP complements from negative
to positive. There is no attempt in Vance (1989) to explain how this resetting
was triggered.
Turning to English, we find a serious attempt to explain the loss in English
by Weerman (1989:235~240),12 who claims that the loss of verb second is due
to a change in his proposed system of S-identification, triggered by the emer-
gence of blockades (mainly the negation not) which prevented the finite fea-
tures of V from reaching Vmax. According to Weerman (1989:78ff.), both
verbal projections and nominal projections must be identified at S-structure:
for nominal projections, S-identification is syntactic case, for verbal projec-
tions, S-identification is conjugation (i.e., tense and mood). In the same
manner as for Case, conjugation can be assigned either inherently or structur-
ally (via C°, governing a finite Vmax; it is to be noticed that there is no I-pro-
jection in Weerman's system). The contrast between structurally and inherently
assigned conjugation is the basis for the contrast between verb-second
languages and non-verb-second languages, according to Weerman: in verb-sec-
ond languages, C must be lexicalized in order to S-identify Vmax, whereas in
non-verb-second languages, S-identification can take place within Vmax, e.g.
THE LOSS OF VERB SECOND IN ENGLISH AND FRENCH 207
As pointed out in the previous section, the presence of sentences with am-
biguous word orders, as illustrated in (8), cannot be the sole explanation for
the loss of verb second. However, at the time when verb second was lost in
English and French, there was also another type of ambiguous sentence present:
THE LOSS OF VERB SECOND IN ENGLISH AND FRENCH 209
a sentence with a weak subject pronoun between the topic and the tensed verb.
An English and a French example are given in (9): in these examples the weak
pronoun is in bold face:
(9) a. Certis pei ben opyn foolis, and... (Wycliffe, late 14C15)
Certainly they are open fools
b. En verite, il a este et est bon valeton (Jehan de Saintre",
in truth he has been and is a good valet-DIM mid 15C16)
'In truth he has been and is a good little valet'.
There are two different structural interpretations of sentences like (9): the
pronoun may be in Spec-IP, as ordinary subjects, or it may be cliticized to the
tensed verb. These two options are illustrated in (10), where '+' indicates the
clitic attachment of the pronoun to the tensed verb in (l0b):
(10) a. Certis [IP pei i [ r benj ] [VP ei [v ej ] opyn foolis]]
b. Certis [c peii+benj]k [IP e [I ek] [VP ei [v ej ] opyn foolis]]
Consider first (10b), where the subject pronoun is cliticized to the tensed verb.
I have not indicated whether this cliticization takes place in VP or in IP; for
the sake of my argument this does not matter. The important thing to notice
is that the pronoun is cliticized to the verb prior to the movement from I° to
C°. Therefore (lOb) counts as a verb-second structure, and can be produced
by a person with a verb-second grammar. Consider next the structure out-
lined in (l0a), where the pronoun is not interpreted as a clitic element. This
structure violates the verb-second requirement that the tensed verb must be in
second position, hence such a structure cannot be produced by a person with
a verb-second grammar. Obviously, sentences like (9) are structurally ambigu-
ous; thus, together with the SVO sentences, they constitute possible triggers
for the loss of verb second.17
Summarizing, at the time of the loss of verb second in English and French,
we seem to have a situation which can be described as follows. Most utter-
ances which the children met had a word order which was ambiguous and thus
did not enable them to determine the position of the finiteness feature: these
sentences could be produced by parents having either the plus or the minus
value of the Verb-Second Parameter (2). Only sentences with inversion (di-
rect questions, wh-questions and cases with topicalization) unambiguously in-
dicated that the finiteness feature was in C°. The possibility of using examples
like (9) with a weak pronoun between the topic and the tensed verb consti-
tuted a further ambiguity. Together this led to a situation where true verb-
second indicators were few in number. In spoken language, and especially in
the language spoken to children, the number of sentences unambiguously in-
dicating a verb-second grammar must have been quite small: we know from
studies of modern verb-second languages that the number of sentences with
non-pronominal subjects is low,18 and we can infer that the number of sen-
tences with inverted non-pronominal subjects must have been even lower. In
such a situation it is conceivable that many children selected the minus value
210 VERB SECOND AND THE NULL-SUBJECT PARAMETER
lands, c. 1380) have optional inversion with nominal subjects, i.e. there are
cases in the writings of these two authors where not only a subject pronoun
but a full subject NP is found in the position between the topic and the tensed
verb. Rolle has optional inversion also with pronominal subjects, indicating
that there are no subject clitics in his language (van Kemenade 1987:220),
whereas Wycliffe always has subject pronouns in the clitic position between
the topic and the tensed verb. The language of Chaucer (E. Midland/London,
c. 1390), on the other hand, is consistently verb second both with nominal and
pronominal subjects; there is no trace of an ongoing change. Hence, although
the text of Chaucer is the youngest one of the three texts investigated by van
Kemenade, it is the most conservative with respect to verb second.
Discussing the texts of Rolle and Wycliffe, van Kemenade actually proposes
that the use of subject clitics may be a reason for the loss of verb second,
arguing more or less as I have done above. However, taking into consider-
ation the regular verb second found in Chaucer, she concludes that "the be-
havior of pronominal subjects cannot have provided the basis for a
reinterpretation" (p. 222), settling for the insufficient explanation of the loss
of verb second discussed in the previous section. In my view this is a mis-
take. The alternative seems to be to treat the language/dialect of Chaucer as
representing a blind alley in the development of English syntax.
There are certain properties of the syntax found in Chaucer's prose that
might support the assumption that Chaucer's language is not representative
for the development of English at the end of the 14th century. For example,
there are very few examples of periphrastic do in the texts of Chaucer, who
seems to prefer gin, which was lost in the early part of the Modern English
period (Mustanoja (1960:614)), although one of the first cases of do in a ques-
tion occurs in Chaucer's poetry (Mustanoja (1960:607)). Another case where
Chaucer's language is not representative for the development of English is
found in his use of overt subjects with infinitives, as in (12) (cf. Einenkel
(1887:80ff.)):
(12) And I to ben youre veray humble trewe
and I to be your very humble faithful.
Summing up, provided it is correct to consider Chaucer's language somewhat
exceptional with respect to verb second,19 there seems to be a period at the
end of the 14th century when the use of subject clitics in ME, together with
the general SVO word order, paved the way for the loss of verb second. We
have a situation where the language learners could interpret an overwhelming
majority of the utterances they met as indicating a non-verb-second grammar,
whereas their parents produced these utterances with the help of a verb-sec-
ond grammar.
3.4 Conclusion
Concluding this section, we have found that different developments in
English and French led to a situation where both languages had subject clitics,
were verb-second languages and had an underlying SVO word order. In such
languages there are very few indications of how to set the verb-second pa-
rameter. As I have argued, language learners of both English and French re-
acted in the same way to this situation: they gave up verb second. In my
terms, this means that they changed the position of the finiteness feature, from
C° to I°. In the next section I will consider certain syntactic consequences of
this change.
that this requirement is the reason for the verb to move to C° in verb-second
languages. When verb second is lost in French and English, and [+F] is in I°,
there is a similar demand for I° to be lexicalized. French and English have
reacted differently to this lexicalization requirement, due to the different sta-
tus of Agr in I° at the time of the loss of verb second.
Since Agr in I° is nominal in nature and hence marked with the feature [+N]
in both OF/MF and ME, I° is a head containing [+N] and consequently a head
governor in both languages; cf. the definition of Head Government in (4)
above. When [+F] is placed in I°, the governor status of I° becomes indepen-
dent of the categorial status of Agr. In present day English and French, Agr
has lost its categorial status, as is evident from the impossibility of using empty
subjects in the modern versions of these languages (with respect to the corre-
lation between nominal/Nominative Agr and empty subjects, cf. SECTION 2
above).24 However, whereas this loss of nominal/Nominative Agr seems to
take place more or less at the same time as the loss of verb second in English,
the loss of nominal/Nominative Agr is much later in French.
The two factors mentioned above, i.e. the change of position of [+F] and
the different developments of Agr, are responsible for several syntactic dif-
ferences between French and English which begin to be visible in the first half
of the 15th century. In this section I will discuss these differences, starting in
SECTION 4.1 with a discussion of the differences with respect to how the two
languages meet the demand of lexicalizing [+F], following this with a discus-
sion of null subjects in SECTION 4.2.
phrastic auxiliary do came into existence in the late 13th century. One of the
factors playing an important role for its regulation was actually the loss of
verb second, according to Ellegard (1953:209).28
It is also possible to trace the emergence of Infl-lowering in English to the
time when verb second was lost. At this time there occur in English instances
which seem to indicate some kind of tense copying or tense agreement within
the single clause, cases where tense seems to be realized both in I° and on the
verb in VP. Such examples provide evidence for the presence of Infl-lower-
ing, since there is no other way for tense to be realized within VP. As no-
ticed by Ellegard (1953:123), discussing the use of semi-auxiliary do +
"explanatory" verb, before 1400 "the explanatory verb was generally in the
same form as do: finite if do was finite, infinite if do was infinite." An ex-
ample is given in (18):
(18) Thalestris ... did wroot to kyng Alexandre
T. did wrote to king A.
In (18) there are two tense affixes: on the auxiliary do, and on the verb in VP.
The empty auxiliary 0, suggested by Pollock, may be seen as a remnant of
the overt do in cases like (18).
In conclusion, we have found that the loss of verb second in English oc-
curs at the end of the 14th century, more or less simultaneously with the
emergence of auxiliary do, Infl-lowering, and the weakening of Agr. The
hypothesis that these changes are related to the change of position for [+F]
thus seems to be supported.29
ond leads to a change with respect to the licenser of small pro: prior to the
change, the licenser of pro is different from the head hosting Agr, after the
change, the licenser of pro is identical to the head hosting Agr.
Following Rizzi (1986:520) Holmberg and Platzack assume that the differ-
ent interpretations of pro depend on which (p-features pro is associated with,
according to the specification in (20); see Rizzi (1986:543):31
(20) a, referential pro: [+number, +person]
b. quasi-argumental pro: [+number, -person]
c. true expletive pro: [-number, -person]
It follows from (19) that pro has the grammatical specification (i.e. '(o-features)
of its licensing head. Hence a change of position for [+F] does not only change
the licensing head of small pro, it also changes the way in which pro is iden-
tified. Whereas it is clear that I° contains o-features due to the presence of
Agr in I°, it is not evident that there are any o-features in C°.
To understand how pro is identified in verb-second languages, i.e. in cases
where the licensing head (C°) is not automatically associated with any cp-fea-
ture, we must consider how pro might be interpreted in such languages.32
Obviously verb-second languages with syntactically active Agr differ in their
abilities to represent (p-features in C° and hence to identify small pro. Mod-
ern German seems to lack (p-features in C° altogether: the only possible iden-
tification of pro in such a case is that pro is a true expletive. Thus, an example
like (2la) with expletive pro is well formed in modern German, whereas ex-
amples like (21b,c) are not well formed, since C° without any (p-features can-
not identify either referential pro or quasi-argumental pro:
(21) a. Gestern wurde pro getanzt.
yesterday was danced
b. (Sein Buch ist sehr interessant.) *Leider hat pro sehr schlechte
Bilder.
his book is very interesting. Unfortunately has very bad pictures
c. * Gestern hat pro geregnet.
yesterday has rained
On the other hand, modern Icelandic differs from German in allowing quasi-
argumental pro: consider the well-formed example in (22):
(22) Rigndi pro i gaer?
rained yesterday
Since C° is the identifying head of pro in Icelandic as in German, the con-
trast illustrated by (21c) and (22) indicates that C° may be marked for [+num-
ber] in Icelandic, but not in German. Consider (20) above. There is no
morphological support for this difference: in both (21c) and (22), the tensed
verb in C° is in its 3sg form. Furthermore, the facts are the same in embed-
ded clauses: although neither German nor Icelandic complementizers show
agreement, quasi-argumental pro is possible in Icelandic embedded clauses,
but not in German embedded clauses:
218 VERB SECOND AND THE NULL-SUBJECT PARAMETER
(23) a. Hann sagoi ao hefoi rignt mjog allan daginn i gaer. (Ice.)
he said that had rained much whole day yesterday
b. * Er sagte daB gestern hatte den ganzen Tag vieles geregnet.
(Ger.)
he said that yesterday had the whole day much rained.
To account for the difference between Icelandic and German with respect to
the interpretation of pro, Platzack (1992) proposes that Icelandic C° differs
from German C° in respect to being marked with the feature [+number].
As indicated by the examples in (22) and (23a), the ability of C° to license
quasi-argumental pro seems to be the same in main clauses as in subordinate
clauses, notwithstanding the fact that the agreement morpheme realized on the
tensed verb is actually occurring in C° in main clauses, but not in subordinate
clauses. This observation supports the assumption that Agr is not necessarily
to be identified with the verbal inflection: Agr is a property of I°, which may
or may not be interpreted as a property of the inflected verb. We conclude
that languages seem to differ with respect to whether or not (p-features may
appear in C°.33
Whereas modern Icelandic, as we have seen, does not allow referential pro,
and hence does not allow the (p-feature [-t-person] of Agr to be realized in C°
even when the inflected verb is moved to C, the situation is different in Old
French: as, for example, Adams (1987:2) points out, empty referential sub-
jects typically occur only in main clauses with inversion in OF. Hence small
pro can be identified as referential in OF in case the p-feature [+person] is
allowed to follow the verb to C°.34
Now, when verb second is lost, small pro is identified by I°, not by C°, as
mentioned above. In a situation where Agr has the same status before and
after [+F] changed its position, we expect this change to lead to an extension
of the distribution of pro: after the change, the interpretation of pro should
no longer be dependent on verb movement to C°.35 This is exactly what we
find in French: in 15th-century texts, i.e. texts produced after the loss of verb
second, null subjects are found not only in main clauses, but also in subordi-
nate clauses. Consider the following example, taken from Vance (1989:3):
(24) Puet bien estre que n'en avez point (Jehan de Saintre)
may (3sg) well be that neg-en have neg
'It may well be that you have none'.
Since the increased distribution of small pro in French appears after the loss
of verb second we have perfect timing: the change of position of [+F], which
is the theoretical consequence of the loss of verb second, has visible effects
on the grammatical output.36
Turning to English, null-subject data do not give us a similar support for
the change of position of [+F]. At the time of the loss of verb second, En-
glish and French differed considerably with respect to subject-verb agreement:
whereas both OF and MF have rich verbal inflection (although Vance
(1989:267) notices that this inflection is eroding in MF), the ME system of
THE LOSS OF VERB SECOND IN ENGLISH AND FRENCH 219
5. Conclusion
In this paper I have tried to show that the loss of verb second in French
and English during the 14th and 15th centuries could be given a common ex-
planation: at the time of the loss of verb second, both languages had a pre-
dominant use of SVO word order, and in addition both languages had
developed subject clitics. The SVO word order can be produced both by a
grammar with verb second and by a non-verb-second grammar where SVO is
the underlying word order. The same is true of sentences with a pronominal
subject between the topic and the tensed verb. Hence for both French and
English there is a period when an overwhelming majority of the sentences
uttered could be given two different structural interpretations. Such a situa-
tion is a necessary prerequisite for a syntactic change to take place.
According to Holmberg and Platzack, who provide the theoretical frame-
work for the present study, the loss of verb second is described as a change in
position of the finiteness feature [+F]: in verb-second languages, this feature
is in C°, in non-verb-second languages it is in I°. This change in position has
different consequences for English and French. For English the consequences
involve the development of do-insertion and Infl-lowering, for French a change
in the distribution of null subjects. The difference between English and French
in this respect has to do with the status of Agr: whereas Agr is nominal and
Nominative in French both prior to and after the loss of verb second, Agr is
losing its nominal status in English more or less at the same time as verb
second is lost.
Notes
A preliminary version of this paper was read at the First Generative Diachronic
Syntax Conference in York, England, April 1990, whereby I got several valu-
able comments by the audience. Thanks also to Lars-Olof Delsing, Cecilia
Falk, Anders Holmberg, Ans van Kemenade, Beatrice Santorini and Halld6r
Sigur5sson for their reactions to the preliminary version, and to Ian Roberts
for letting me read an early draft of Roberts (1992), a book which has been of
220 VERB SECOND AND THE NULL-SUBJECT PARAMETER
subject element is fronted have the tensed verb in C°. Recently, Vikner and
Schwartz (1991) have given empirical support to the assumption that the tensed
verb is always outside of IP in verb-second clauses.
3. This account of the verb-second effects has much in common with the stan-
dard V-to-Comp analysis which was first suggested by den Besten in an un-
published paper from 1977, later printed as den Besten (1983) (see also den
Besten (1989:94)).
4. Compare also Borer (1986).
5. Consider especially Rizzi's (1990:76-80) arguments against the existence of
Theta Government.
6. Rizzi (1990:25) formulates this condition as follows: X = { [±N], [±V], Agr,
T}. The difference between the formulation in (4ii) and Rizzi's formulation
is not dramatic: for most non-verb-second languages, [+F] is situated in the
same position as T, hence for these languages the difference is just notational.
With respect to Agr, Holmberg and Platzack claim that only Agr which is
nominal, i.e. containing the feature [+N], qualifies as a head governor. Hence
most cases where Rizzi explicitly refers to Agr are handled by the presence of
the feature [+N] in (4). The only case where the definition in (4) and the
definition in Rizzi (1990) differ is for languages where Agr is not nominal
and not in the same position as [+F], i.e. verb-second languages with non-
nominal Agr. Such languages should have no V-to-I in spite of having sub-
ject-verb agreement. There are some MSc. dialects of this kind, e.g. the dia-
lect of Hallingdalen in Norway (see Trosterud (1989)).
7. Rizzi uses this formulation in order to specify the three subcases of anteced-
ent government: chains formed by NP-movement, wh movement, and head
movement.
8. Rizzi (1990:92) reformulates this condition as: (i) X and Y are non-distinct.
9. Richard Rolle of Hampole, The Bee and the Stork. From English Writings of
Richard Rolle Hermit of Hampole, ed. by Hope Emily Allen, Oxford 1931.
10. The Peterborough Chronicle, ed. by C. Clark. 2nd Edition, Clarendon Press,
Oxford 1970.
11. The number of sentences which unambiguously indicated verb second was
probably quite low, however. For Chaucer's Treatise on the Astrolabe,
Macleish (1969) has recorded 83% SV order in main clauses, leaving just
17% for S-V inversion. For the first 135 lines of the OF prose text La Queste
del Saint Graal, Vance (1989:35) reports a frequency of 75% SV order in
main clauses (only clauses with overt subject are taken into consideration);
the frequency of S-V inversion of full NPs is 13%.
12. This is a simplification of Weerman's position: according to Weerman (p.
234), English never was verb-second language in all respects.
13. Naturally this is a problem only if you try to find a common explanation for
the loss verb second in French and English. It should be pointed out that
Weerman is only concerned with the loss of verb second in English.
14. As I have mentioned several times by now, Roberts (1992) might be an ex-
ception. In the part of his book that I have had access to, Roberts gives a
detailed and very interesting account of the loss of verb second in French,
222 VERB SECOND AND THE NULL-SUBJECT PARAMETER
which in its essence seems to be quite close to the account I will present in
SECTION 3 below. In particular, it can be noticed that Roberts derives the loss
of verb second from a change in the way Nominative Case is assigned. In my
description, verb second is a consequence of the position of the finiteness
feature, which in its turn has an important role to play for the licensing of
Nominative Case. However, since I have not had access to the final version
of his book, I will abstain from making a detailed comparison of our ap-
proaches.
15. The example is found in van Kemenade (1987:200).
16. The example is found in Vance (1989:200); notice that Vance does not con-
sider il in this example to be a true syntactic clitic.
17. A third factor might have been the word order of subordinate clauses. As
noticed in note 2 above, both OF and ME belong to a type of verb-second
languages with few differences between main clauses and embedded clauses.
18. Einarsson (1978:143) reports that 45% of the subjects in written Swedish are
pronouns (an average of different genres), whereas 82% of the subjects in
modern spoken Swedish are pronouns (once again this figure is an average).
In this material there are unfortunately no frequency figures for the amount of
subject pronouns in the language of adults talking to young children.
19. Van Gelderen (1989) notices that verb second is optionally used in Chaucer's
Prologue to The Wife of Bath, from which she quotes the following examples:
(i) a. An housbande I wol have...
a husband I will have
b. In wyfhod I wol use myn instrument
In wifehood I will use my instrument.
However, since the Prologue is poetry, not prose, examples like these should
most probably not be taken into consideration.
20. Both Vance (1989:183ff.) and Roberts (1992) argue that there is no syntactic
cliticization in French at this time, only phonological cliticization: hence,
according to Vance, the subject pronoun is not syntactically adjoined to the
tensed verb, meaning that cases where the subject pronoun occurs between
the topic and the tensed verb show that MF is not a verb-second language.
21. Another type of main clause lacking verb second found in all Scandinavian
languages, is introduced by adverbs corresponding to the English maybe, for
instance Swedish kanske, Icelandic kannski. Illustrating examples are given
in (i):
(i) a. Kanske jag kommer.
b. Kannski eg komi.
maybe I come.
The position of the adverb in examples like (i) has been debated. Platzack
(1986) notices that Swedish kanske can appear in the positions which are
normally reserved for the tensed verb, i.e. in C° in main clauses. It is, for
example, found in second position in declarative clauses, as shown in (ii):
(ii) a. Allan kanske redan har fatt jobbet.
Alan maybe already has got the job
THE LOSS OF VERB SECOND IN ENGLISH AND FRENCH 223
33. Naturally these data might also be interpreted to show that Rizzi (1986) is
wrong when he claims that pro is both licensed and identified by the same
head. If that restriction is lifted, we could say that pro is identified by Agr in
either C° or I°. In the lack of any independent reasons to perform such a
weakening of the theory, I will stick to the more restricted version given in
(19). See Platzack (1992) for a discussion of the consequences of such an
approach.
34. Roberts (1992) extensively discusses cases with null subjects in embedded
clauses in OF, showing that null subjects are found only in cases where em-
bedded verb second is possible.
35. Independent evidence for this hypothesis is provided by the development of
Romance varieties in Northen Italy and Southern France, where null subjects
were generalized from verb-second contexts to all contexts with the loss of
verb second. See references in Roberts (1992).
36. Consider Vance (1989) and Hirschbuhler (this volume) for detailed investiga-
tions of the exact contexts inwhich null subjects are possible in MF texts.
References
Abraham, W. and E. Reuland, eds. (1991) Issues in Germanic Syntax. Ber-
lin/New York: Mouton de Gruyter.
Adams, M. (1987) "From Old French to the Theory of Prodrop." Natural
Language and Linguistic Theory 5:1-32.
Besten, H. den (1983) "On the Interaction of Root Transformations and Lexi-
cal Deletive Rules." In W. Abraham, ed. On the Formal Syntax of the
Westgermania, 147-131. Amsterdam: John Benjamins.
Besten, H. den (1989) Studies in West Germanic Syntax. Amsterdam/Atlanta:
Rodopi.
Borer, H. (1986) "I-Subjects." Linguistic Inquiry 17:375-416.
Cardinaletti, A. (1990) Impersonal Constructions and Sentential Arguments in
German. Padova: Unipress.
Chomsky, N. (1986) Barriers. Cambridge, Mass.: MIT Press.
Diesing, M. (1990) "Verb-Second in Yiddish and the Nature of Subject Posi-
tion." Natural Language and Linguistic Theory 8:41-79.
Einarsson, J. (1978) Talad och skriven svenska. Lund: Walter Ekstrand.
Einenkel, E. (1887) Streifuge durch die mittelenglische Syntax unter besonderer
Berucksichtigung der Sprache Chaucer's. Munster: Verlag von Heinrich
Schoning.
Ekerot, L.-J. (1990) "The Swedish sa-Construction in a Functional and Con-
trastive Perspective." Paper presented at the Seventh International Con-
ference of Nordic and General Linguistics, Torshavn 1989.
Ellegard, A. (1953) The Auxiliary Do. The Establishment and Regulation of
its Use in English. Stockholm: Almqvist and Wiksell.
Falk, C. (1990) "Pro-drop in Early Modern Swedish." Paper read at the llth
Groningen Grammar Talks, November 1990.
THE LOSS OF VERB SECOND IN ENGLISH AND FRENCH 225
Gelderen, E. van (1989) "The Historical Rationale Behind Split Infinitives and
Kindred Constructions." Archivfur das Studium der neueren Sprachen
und Literaturen 26:1 -18.
Hirschbuhler, P. (this volume) "Null Subjects in Verb-First Embedded Clauses
in Philippe de Vigneulles' Cent Nouvelles Nouvelles."
Hjartard6ttir, Th. (1987) Getio i eyournar. Um cyour fyrir frumlog og audlog
i eldri islensku. MA Thesis, University of Iceland.
Holmberg, A. and C. Platzack (1988) "On the Role of Inflection in Scandina-
vian Syntax." Working Papers in Scandinavian Syntax 42:25-42.
Holmberg, A. and C. Platzack (1988) On the Role of Inflection in the Syntax
of the Scandinavian Languages. Oxford: Oxford University Press.
Kemenade, A. van (1987) Syntactic Case and Morphological Case in the His-
tory of English. Dordrecht: Foris.
Macleish, A. (1969) The Middle English Subject-Verb Cluster. The Hague:
Mouton.
Mustanoja, T. (1960) A Middle English Syntax. Part 1: Parts of Speech.
Helsinki: Societe Neophilologique.
Platzack, C. (1986) "COMP, INFL, and Germanic Word Order." In L. Hellan
and K. Koch Christensen, eds. Topics in Scandinavian Syntax, 185-234.
Dordrecht: Reidel..
Platzack, C. (1989) "Review of Ans van Kemenade: Syntactic Case and Mor-
phological Case in the History of English." Linguistics 27:948-953.
Platzack, C. (1990) "Review of Fred Weerman: The V2 Conspiracy. A
Synchronic and a Diachronic Analysis." Nordic Journal of Linguistics
13:77-88.
Platzack, C. (1992) "Complementizer Agreement and Argument Clitics."
Working Papers in Scandinavian Syntax 50:25-54.
Platzack, C. and A. Holmberg (1989) "The Role of AGR and Finiteness in
Germanic VO Languages." Working Papers in Scandinavian Syntax
43:1-76.
Pollock, J.-Y. (1989) "Verb Movement, UG and the Structure of IP." Linguis-
tic Inquiry 20:365-424.
Rizzi, L. (1986) "Null Objects in Italian and the Theory of pro." Linguistic
Inquiry 17:501-557.
Rizzi, L. (1990) Relativized Minimality. Cambridge: MIT Press.
Rizzi, L. and I. Roberts (1989) "Complex Inversion in French." Probus 1:1-
30.
Roberts, I. (1989) "Case Assignment Parameters and the History of French
Inversion." Glow Newsletter 22:55-57.
Roberts, I. (forthcoming) "Object Movement and Verb-Movement in Early
Modern English." To appear in H. Haider, S. Olsen and S. Vikner, eds.
Proceedings of the Seventh Comparative Germanic Syntax Conference.
Dordrecht: Kluwer.
Roberts, I. (1992) Verbs and Diachronic Syntax. Dordrecht: Kluwer.
Rognvaldsson, E. and H. Thrainsson (1990) "On Icelandic Word Order Once
More." In J. Maling and A. Zaenen, eds. Modern Icelandic Syntax.
226 VERB SECOND AND THE NULL-SUBJECT PARAMETER
In this article we present an account of the history of verb second and pro-
drop in French and English. We shall identify two types of verb second lan-
guages, which we call C-oriented or CV2 languages, and I-oriented or IV2
languages, and the specific properties correlating with this distinction. Next,
we shall see how the distinction as drawn, in conjunction with certain assump-
tions concerning the licensing of functional properties, allows a very articu-
late account of the history of French, in terms of a development from CV2 to
IV2 to non-V2, and sound theoretical motivation why English developed from
CV2 to I-non-V2 and did not move through the IV2 stage. The similarities
and differences between the diachrony of verb-second and pro-drop in French
and English are thus analyzed in a surprisingly uniform way.
The paper is organized as follows: SECTION 1 presents the theoretical back-
ground to our account of the history of verb second and pro-drop in English
and French; 1.1 contains a discussion of verb second and C/I orientation; 1.2
presents the outline of a theory of licensing that we develop in more detail
elsewhere (Hulk and van Kemenade (1993)); in 1.3 the role of functional pro-
jections in language change is considered. In SECTION 2 an analysis is given
of the history of verb second and pro-drop in French and English. In SECTION
3 we summarize our conclusions.
1. Theoretical Background
227
228 VERB SECOND AND THE NULL-SUBJECT PARAMETER
of the basic sentence structure, and fronts the finite verb (Vf) to presentential
position in root clauses. We analyze this as in the structure (1), adapted on
the basis of Chomsky (1986) and Sportiche (1988).
One of the core characteristics of verb second in this sense is the asymme-
try between root and non-root clauses with respect to verb-fronting. Fronting
of Vf is found in root clauses only, cf. early work on verb-second in den Besten
(1983). This is the main reason why verb second is widely analyzed as move-
ment of the finite verb/I to Comp, which is blocked in non-root clauses by
the presence of a base-generated complementizer. Thus, in verb-second lan-
guages, movement of Vf in non-root clauses can take place to I, but never to
C. Evidence for this root/non-root asymmetry is the fact that apparent em-
bedded verb-second with topicalization takes place only after bridge verbs,
as exemplified for modern Dutch in (2):
(2) a. Hij zei hij heeft hem gisteren gezien
he said he has him yesterday seen
b. Hij zei gisteren heeft hij hem gezien
he said yesterday has he him seen
c. Hij zei dat hij hem gisteren gezien heeft
he said that he him yesterday seen has
d. Hij zei dat gisteren heeft hij hem gezien
he said that yesterday has he him seen
VERB SECOND, PRO-DROP, FUNCTIONAL PROJECTIONS 229
Similar examples can be adduced from Middle Dutch, Modern German as well
as older stages of German and Modern Icelandic, but we will not do this here
for reasons of space. We now discuss how the various types of pro-drop should
be analyzed.
Our analysis of pro-drop essentially follows up the classical analysis of full
pro-drop as in Italian (Taraldsen (1978); Chomsky (1982); Rizzi (1982, 1986))
and of expletive pro-drop as in the Scandinavian languages analyzed by
Platzack (1987). The option of having an empty subject at all depends on the
possibility of identifying it. Following Rizzi (1982, 1986) we assume that a
pro-subject must satisfy two conditions: it must be formally licensed, and its
content (i.e. (p-features) must be identified. Rizzi (1986) defines these condi-
tions as follows:
(12) pro-licensing:
a. pro must be formally licensed under government by a designated
Case-assigning head
b. the content of pro ((p-features) must be identified.
Notice that (12a) in fact imposes two requirements on the governing head:
first of all that it be a "designated" head (in Rizzi (1982) the head is desig-
nated by virtue of a feature specification [+ pronominal]); secondly that the
head be a Case-assigner. The term "designated" indicates that such a head is
marked for pro-licensing language-specifically, i.e. presumably each language
has an inventory of heads that are termed designated; it is even quite possible
that languages lack such an inventory entirely. The latter seems to be the case
in Modern English. Rizzi (1986) considers that the licensing conditions for
pro (12a-b) should be satisfied by one and the same head.
We propose that the designated Case-assigning head in (12a) coincides cru-
cially with the dominant head for C/I orientation in terms of our theory. In C-
oriented languages, C is the relevant governor/Case marker, and can be listed
as the designated head; in CV2 languages where C is a designated head, it
formally licenses pro in Spec,IP. Modern Dutch is an example of such a lan-
guage. In I-oriented languages, I, if listed as a designated head, is the relevant
governor and can license pro in Spec.VP. Modern Spanish and Modern Ital-
ian are cases in point.
We furthermore recall the definition for Nominative Case ((6) above), which
states that Nominative is assigned under government by the dominant head C
or I. Finally, we propose that content-licensing of pro (12b) is by the (p-fea-
tures of I, if these are morphologically realized.
With these theoretical preliminaries in mind we can consider the question
of what it is that determines whether a language has full or expletive pro-drop
characteristics. This is related to the fact that a pronominal I, but not a pro-
nominal C, may contain (p-features: [±person, ±number, ±gender]. We follow
Rizzi (1982:42) in assuming that it is the specification [+person] that is nec-
essary to license referential pro-drop. If a pronominal I does not have the
[+person] specification, only expletive pro-drop is possible. In I-oriented lan-
guages, a pronominal I may optionally be specified [+person]. Therefore, we
234 VERB SECOND AND THE NULL-SUBJECT PARAMETER
find both referential pro-drop (13) and expletive pro-drop (14) in these lan-
guages:
(13) hanno parlato
have spoken
'they have spoken'
(14) Sembra che gli amici sono arrivati
seems that the friends have arrived
"It seems that the friends have arrived'.
When I or C is a designated head, and there are no p-features under I, we find
expletive pro-drop. Pro can be formally licensed, but there are no (p-features
that must be identified, hence formal licensing is sufficient. C-oriented lan-
guages for which this is relevant, are Modern Dutch and earlier stages of En-
glish, as exemplified above. We will discuss the latter in more detail below.
The possibilities for pro-drop discussed are summarized in (15), with men-
tion of some of the core instantiations of the various options.11
(15) V2 Comp Infl
MoDu/OE/ME + +DH -phi expl.pro-drop
Cont. Scand. + -DH - Phi no pro-drop
Ital/Span - - +DH+phi full pro-drop
MoFr/MoE - - -DH-phi no pro-drop
Icelandic + -DH +DH-phi expl.pro-drop
The table in (15) is not exhaustive. We shall see below that Old French, which
cannot be accommodated in (15) in any obvious way, is an interesting mix-
ture of the COMP and INFL properties listed. We discuss this in SECTION 2.1.
these two types of verb-second language. We shall argue that French changed
from a CV2 language with full pro-drop in root contexts through an IV2 stage
with postverbal subjects and pro-drop in both root and non-root contexts, to
non-verb second, i.e. Modern French. This scenario involves a number of phe-
nomena that will be difficult to account for in a framework with only one
functional projection CP. One may even want to carry this further and hypoth-
esize that functional projections may yield an interesting account of synchronic
variation.
The theoretical framework we assume also gives the motivation for why
English did not go through this intermediate stage and changed from CV2 to
non-verb second. Thus, we will see that the analysis we present in the fol-
lowing sections provides crucial support for distinguishing at least two func-
tional projections above VP.
under government by C, but its p-features are licensed by I. This may present
an interesting starting point for the developments seen in Middle French
(MidF), which we deal with in the next subsection.
(27) a. quand elle veit qu' aultre reponse ne pouvait avoir de luy
(Vigneulles099112)
when she saw that other answer not could have from him
'when she realized that she would not get any other answer
from him'
b. promist que ainsi feroit- ilz (Vigneulles 077041)
promised that thus would-do they
'promised that they would act like that'.
Recall that these patterns already existed in OF, though with an unambiguous
CP interpretation. The extension of verb second to clauses embedded under
non-bridge verbs should be interpreted as a shift to IV2, we would argue. As
noted above, the causal factor responsible for the increase was the cliticization
of subject pronouns, resulting in ambiguity between a CP and an IP interpre-
tation; as an IP interpretation for the verb-second phenomenon became avail-
able, its increase in non-root clauses is expected. Note that this constitutes a
reinterpretation of the infrequent CP pattern of OF as an IP pattern. At the
same time, the increase of verb second in non-root clauses further weakens
the evidence for a root/non-root asymmetry with respect to verb second. Dur-
ing the same period an increasing number of verb-first sentences are reported
(Vance (1988); Dupuis (1988)); some examples:
(28) a. et se retira le roy a Corbeil
and himself withdrew the king to Corbeil
'and the king retired to Corbeil' (Commynes 62, 2)
b. et luy fut adoubee sa playe qu'il avoit au col
and to-him was dressed his wound that-he had in the neck
'and his neck wound was dressed'. (Commynes 63, 20)
These two patterns—verb-first structures and the loss of the root/non-root
asymmetry—are characteristic of languages which are typically analyzed as
IV2, see for instance Santorini's (1989) analysis of the history of Yiddish.
What we propose then is that while OF was a CV2 language with a root/
non-root asymmetry with respect to verb second and pro-drop, the cliticization
of subject pronouns, resulting in ambiguity between a CP and IP interpreta-
tion, triggered a shift to IV2. Verb-second structures in MidF are then ana-
lyzed as follows:
(29) IP
[ XP [I Vf [VP NP [v, Vj ]]]]
Vf moves to I; XP moves to Spec,IP; the subject remains in its base-gener-
ated position. Evidence for this VP-internal position is given by Lemieux and
Dupuis (this volume) who cite examples in which a VP-adjoined adverb in-
tervenes between Vf and the postverbal subject:
242 VERB SECOND AND THE NULL-SUBJECT PARAMETER
In fact, this appears to be borne out: Hirschbuhler and Junker (1988) give
evidence that the position of pro in, for example, Vigneulles is postverbal, as
it alternates only with postverbal lexical subjects.
Recall, however, the exceptional cases of referential pro-drop in non-root
clauses in OF. In OF pro is formally licensed by a lexical C, and we sug-
gested that OF had a marked option of (p-identifying pro under Spec-head
agreement in IP. Since this option was already marginally available in OF, it
seems plausible that it remained available in MidF, the more so because in
MidF I had become the formal licenser. However, licensing of p-features un-
der Spec-head agreement could never be the only option, as postverbal sub-
jects in Spec.VP were possible too.17 So in fact null subjects in MidF could
be found in two positions: in Spec.IP and in Spec.VP. In both these positions
pro is formally licensed by I under government;18 in Spec.IP the p-features
are identified under Spec-head agreement; in Spec.VP the p-features are iden-
tified under government.19
Summarizing, our analysis for OF and MidF is as follows:
(32) Old French: C-oriented; Verb movement: V to I, V/I to C;
Nominative Case: government by C
Fro-licensing: formal: government by C
content: government by V/I/C in roots
Middle French: I-oriented; Verb movement: V to I
Nominative case: government by I
Fro-licensing: formal: government by I
content: government by I/Spec-Head Agreement
in IP
nario makes crucial use of two functional projections CP and IP, which both
allow verb-second facts, given certain Infl properties. This distinction explains
the contrast between Old and Middle French with respect to verb second and
pro-drop: the root/non-root asymmetry in OF versus the development towards
symmetry in MidF . Our assumptions about the licensing of p-features explain
why the erosion of agreement morphology on V/I led to the loss of pro-drop
and the obligatoriness of NP subject in Spec,IP. Our analysis thus argues for
at least two functional projections.
show that we find expletive pro-drop until the early 15th century; this is borne
out by such data studies as van der Gaaf (1904) and Elmer (1981);24 some
ME data are given in (39):
(39) a. benne scheomep me (Obj) berwib
then shames me with that
'then I am ashamed of that' (St. Marh, 34, 30)
b. him (Obj) wile sone longe barafter
him will soon long after that
'he will soon long for that'. (Trin., 148,19)
With respect to constructions like those in (39), it is important to realize that
the emerging obligatorinesss of a Nominative subject is only part of the story.
While in OE the range of complements and cases available to subjectless verbs
was a lot wider than, for example, in the 13th century, due to the loss f
morphological case, these verbs could appear with complements, as (3 )
shows: in both sentences we find an object pronoun and a PP complement.
The most important observation with respect to the loss of expletive pro-drop
however, is the obligatory presence or not of a Nominative subject. The Nomi-
native subject became obligatory in the early 15th century.
From the middle of the 14th century, verb second declines. As to the pri-
mary cause for this, we have to go back to a peculiarity of verb second in
earlier times. While in OE verb second, as analyzed here, is fairly regular given
that patterns like XP-Vf-subject NP-(...)V are the norm, this pattern is not fol-
lowed if the subject is a personal pronoun. In that case the subject is found at
the left of the finite V. Van Kemenade (1987) analyzes this as a process of
cliticization of the pronominal subject to V/I. We do not wish to go into the
nature of this process here. It suffices for our purposes that there is a system-
atic discrepancy between the position of the nominal subject and the pronomi-
nal subject with respect to the position of the finite verb in root clauses. Some
examples are given in (40):
(40) a. For5on we sceolan mid ealle mod and masgene to Gode
therefore we shall with all mind and power to God
gecyrran
turn
'therefore we must turn to God with all our mind and power'
(Blickling 97)
b. Be oaem we magon suioe swutule oncnawan oaet
by that we may very clearly perceive that...
'by that we can perceive very clearly that...' (CP, 181,16)
This discrepancy gradually vanishes in the course of the second half of the
14th century and first quarter of the 15th century. This, we propose, is the main
trigger for the loss of verb second. Pronominal subjects regularly appeared
on the left of verb second in root clauses, while nominal subjects regularly
"inverted" (cf. the examples (38) with (40)). As this discrepancy between nomi-
nal and pronominal subjects got lost, nominal subjects adopted the pro-
248 VERB SECOND AND THE NULL-SUBJECT PARAMETER
nominal pattern.25 We can show this with some examples from the Wycliffite
writings:
(41) a. wip newe wenchis is Crist now weddid and
with new wenches is Christ now wedded and
on newe maner he kepte his furste matrimonye
in new manner he kept his first matrimony
(The Wycliffite sermons, ed. Hudson, p. 361, 1.42)
b. And by pis same skyle hope and sorwe schulle iugen us
and by this same skill hope and sorrow shall judge us
(The Wycliffite sermons, ed. Hudson, p. 372, 1.97)
c. for more ioyse pei myhte not haue
for more joy they could not have
(The Wycliffite sermons, ed. Hudson, p. 382, 1.106)
We regularly find these three patterns: inversion of a subject noun (41a); non-
inversion of a subject noun (41b); and non-inversion of a subject pronoun
(41c). There are very few examples of inverted subject pronouns. Inversion
of a subject pronoun is very infrequent, except with an operator type of first
constituent like a w/z-element. What these examples indicate is that in this
dialect, which there is some reason to assume played a role in the rise of the
standard language, the nominal subjects tended to adopt the pattern of pro-
nominal subjects. The result of this was that fronting of XP to Spec.CP no
longer went hand-in-hand with V/I to C. Hence verb second declined and
English changed from a C-oriented language to an I-oriented language. Let
us consider the relationship between the loss of verb second and the loss of
expletive pro-drop more closely.
At the stage when English was still verb-second with expletive pro-drop,
say c. 1250, the basic sentence structure is as in (42):
VERB SECOND, PRO-DROP, FUNCTIONAL PROJECTIONS 249
4. Conclusions
Let us conclude by comparing the scenario for the history of English with
that for the history of French. OF and OE are both CV2 languages. The dif-
ference between OF and OE is that in OF I has inherent p-features, whereas
I in OE does not. As a result OF has referential pro-drop in root contexts and
expletive pro-drop in both root and non-root clauses. OE on the other hand
allows expletive pro-drop everywhere; since I has no p-features, a referential
subject is obligatorily in Spec-head agreement relation in IP. Since a referen-
tial subject could never be dropped, English, when it lost its CV2 characteris-
tic, could not become an IV2 language, since Spec.IP was a subject position,
not an XP position. In contrast, when French lost its CV2 characteristic, I still
had inherent p-features and a subject, whether lexical or null, could be licensed
in its base-generated position in Spec.VP in the structure (1). As a result
Spec.IP can be an XP position and the IV2 characteristic obtains. The final
step in the development of French was that inherent p-features were lost. As
a result, I came to be licensed by Spec-head agreement with NP in Spec.IP.
250 VERB SECOND AND THE NULL-SUBJECT PARAMETER
The CV2 stage that we argu d for in Old French, and the IV2 stage that
we identified in Middle French argue for at least these two functional pro-
jections.
Notes
1. This paper was presented at the First Diachronic Generative Syntax Confer-
ence in York, April 1990. We thank the audience, the editors of this volume
and an anonymous reviewer for comments. We would like to thank in particu-
lar Barbara Vance for her constructive criticisms of various versions of the
material on French. The first draft of this paper was circulated in 1990. Since
then, the theoretical work on which this paper is partly based, and the histori-
cal work reported here, have led partially separate lives. Although the
theoretical preliminaries of this paper are closely related to Hulk and van
Kemenade (1993), the theory in the latter has evolved much since this work
was written up.
2. Sometimes root clauses have base-generated complementizers, hence no verb
fronting, as in, for example, the Modern Dutch exclamation:
(i) dat -ie dat zo gedaan heeft!
that he that so done has!
'I am surprised he has done it is way'.
3. For instance, Evers (1982) proposes at tense is an operator in Comp and
must be lexical; Koopman (1984) proposes that lexicalization is triggered
for Nominative Case assignment; de Haan and Weerman (1986) suggest that
Comp (containing Agr) has categorical features and must therefore be lexi-
cal; Weerman (1989) assumes that C assigns a modal role to the proposition,
and that C needs to be lexical to be visible as an assigner of modal role.
4. For the moment we do not wish to go into the details of the relation between
V and I in these languages. At least for Modern English there is evidence, for
example, from Neg placement facts (see Pollock (1989); Roberts (1985)) that
V does not move to I. These facts also underlie the pie-Barriers idea that in
Modern English I is lowered to V, see Lasnik (1981). We will come back to
this below.
5. We will come back below to the theoretical status of this assumption.
6. Consider for instance the following Southern Dutch sentences (from den
Besten (1979)). For related facts in German, see Tomaselli (1990).
(i) a. date Jan en Rees morgen zullen komen
that+pl John and Bill tomorrow shall +pl come
'that John and Bill will come tomorrow'
b. dat(*e) Jan morgen zal komen
that(*+pl) John tomorrow shall -pl come
'that John will come tomorrow'.
These examples show that C has number agreement with Vf.
VERB SECOND, PRO-DROP, FUNCTIONAL PROJECTIONS 251
infinitival verb, given further movement rules operative in OE; in fact all van
Kemenade's examples are sentences with two verbs). A thorough look at such
constructions bears out van Kemenade's system, with two provisos. First, V-
fronting is not completely obligatory; and second, the position preceding the
finite V in root clauses is sometimes occupied by two constituents. This, how-
ever, is not counterevidence to V-fronting, as it can still be shown that there is
fronting.
Allen's criticisms seem to be intended as refutation. As such, they mislead
the reader about the status of the counterevidence, and appear to stem from a
basic misappreciation of the grammatical status of the phenomenon.
22. These examples are suggestive of a fairly homogeneous word order in OE,
but in fact this is far from being the case. The exceptions are explained by van
Kemenade (1987) as (heavy) NP shift to postverbal position; a relatively free
process of verb-raising as attested in various West Germanic languages, and a
clitic analysis of personal pronouns. Pintzuk (1991) analyzes OE as variable
between I-medial and I-final word order. This does not bear on the issue
whether OE verb-second is movement to I or to C.
23. Only a pattern with a fronted XP is valid as evidence here, as examples with
Subject-Vf-(XP)-V-(XP) occur quite frequently in non-root clauses. The latter
may be analyzed either as an instance of a variant pattern with I-medial order
in OE, or as an instance of verb-projection raising.
24. Apart from these two data studies, Allen (1986) adduces interesting evidence
that in constructions with like, Nominative subjects are favoured as of the
15th century.
25. Actually this is subject to some dialectal variation (cf. van Kemenade (1987)).
It seems that in the prose of Chaucer, pronouns adopted the nominal subject
pattern. For instance, in one of Chaucer's prose works, the Treatise on the
Astrolabe, verb-second inversion is completely regular. This is in marked
contrast to the dialect of Wycliffe, who is a near contemporary of Chaucer.
Interestingly, there is good reason to assume that the Wycliffite sermons rep-
resent an important literary standard, see Warner (1982). Allen (1990) criti-
cizes van Kemenade's (1987) observations on the dialects of Wycliffe's and
Chaucer's prose as being highly inaccurate. It is unclear how she arrived at
this remark. Detailed further investigations have shown that these observa-
tions are pretty much correct (see van Kemenade (1990); Kroch (1989)).
26. At a later stage, even V to I is lost. This change does not concern us at present,
but should be dated as later than the general loss of verb second. Roberts
(1985) dates the loss of V to I in the 16th century.
References
Abraham, W. ed. (1983) On the Formal Syntax of the Westgermania.
Amsterdam: John Benjamins.
Adams. M. (1987) "From Old French to the Theory of Prodrop." Natural
Lanugage and Linguistic Theory 5:1-32.
254 VERB SECOND AND THE NULL-SUBJECT PARAMETER
Sportiche, D. (1988) "A Theory of Floating Qualifiers and its Corollaries for
Constituent Structure." Linguistic Inquiry 19:425-449.
Taraldsen, K.T. (1978) "On the NIC, Vacuous Application and the That-trace
Filter." Unpublished Ms, MIT.
Thrainsson, H. (1986) "VI, V2, V3 in Icelandic." In Haider and Prinzhorn,
eds., 169-194.
Tomaselli, A. (1990) "COMP as a Licensing Head: An Argument Based on
Cliticization." In Mascaro and Nespor, eds., 433—445.
Vance, B. (1988) "Uevolution de pro-drop en frangais me'di6val." Revue
quebecoise de linguistique thlorique et appliquee 7:85-109.
Vance, B. (1989) Null Subjects and Syntactic Change in Medieval French. PhD
Dissertation, Cornell University.
Warner, A. (1982) Complementation in Middle English and the Methodology
of Historical Syntax. Pennsylvania State University Press.
Weerraan, F. (1989) The V2 Conspiracy: A Synchronic and a Diachronic
Analysis of Verbal Positions in Germanic Languages. Dordrecht: Foris.
10
Null Subjects in Verb-First Embedded
Clauses in Philippe de Vigneu es'
Cent Nouvelles Nouvelles
Paul Hirschbiihler
Universitt d'Ottawa
"C'est un syst&me bien dangereux, monsieur Fred, bien dangereux, que celui qui
consiste a partir de Fid6e qu'on se fait de 1'assassin pour arriver aux preuves dont
on a besoin!" (Rouletabille)
The focus of the present study concerns the licensing of null subjects in em-
bedded verb-first constructions in a late Middle French text, the Cent Nouvelles
Nouvelles (from now on, the CNNV) by Philippes de Vigneulles.1 More
precisely, our goal is to characterize as precisely as possible the conditions
under which these null subjects are found, hoping that some of the ideas dis-
cussed will find their way in a more principled account than the one that is
offered here. One point that should be clear from the start is that I do not claim
that all the variables that appear to be required in characterizing the data ex-
amined here are necessarily relevant for other Middle French texts. Some of
the facts are unexpected given previous literature. It may therefore be tempt-
ing to consider them as accidental patterns rather than considering them as
significant. The second position will be adopted here, in the hope that close
study of other texts will reveal similar patterns.2
This paper is organized as follows: I first very briefly review the main points
of the grammar of Middle French with respect to main clauses that are rel-
evant for the discussion of embedded clauses; then I present the main results
of my analysis of null subjects in verb-first embedded clauses; finally I turn
to the data and discuss it in detail.
257
258 VERB SECOND AND THE NULL-SUBJECT PARAMETER
I will se it to you for an amount that will be said, and [we] will
put the ines at a good price and so [we] will drink from the
trade.
g. .... j'en prendra la chose sur moy, et 05 direz que je le vous ay
fait faire. 015089
I will take it upon myself, and [you] will say that I had you
doit.
We now turn to our gene l claims concerning null subjects in verb-first em-
bedded clauses in the CNNV.
3. Weak Agr. 1st and 2nd singular. With some exceptions that will be
discussed, they are not found in a verb-first construction in any of
the embedded clause types. One way to describe their behaviour
would be to say that their ability to identify the content of pro in a
configuration of Spec-Head agreement is parasitic on whether or not
they also actually formally license the subject pro position. Given
that in embedded clauses any element in C or SpecCP, when adja-
cent to SpecIP, is a potential formal licenser, weak Agr is never in a
position to license pro in that position.
2. Data
Following a standard view, the null subject (Spo) is analyzed as pro. In what
follows, its distribution is compared to that of the lexical pronominal subject
(Spr). Both types of pronominal subjects together are designated by Sp.
Spr and Spo have been divided in several groups to determine the possible
role of various factors in the licensing of Spo. The following potential factors
have been examined:
the type of subordinate clause as well as the specific subordinator
used
the verb position: does it immediately follow the subordinator or not?
the person and number of the verb
the particular verb used
The distribution of Spo in subordinate clauses of the type "Sub (XP) V..."
is summarized in Table 10.2 and Table 10.15 below. Table 10.3 and Table
10.14 give us the same information for subordinate clauses with a preverbal
lexical pronominal subject Spr.
The tables are divided into three parts, the two main categories correspond-
ing in part to the divisions discussed in previous studies: Hirschbiihler and
Junker (1988), Vance (1988, 1989), Adams (1987a,b, 1988) and Dupuis (1988,
1989). In part A, we have grouped together the conjunctive clauses introduced
by que or by a conjunctive phrase ending with que, as well as wh-clauses
introduced by que: cleft sentences, restrictive relative, comparative and su-
perlative clauses, and indirect questions. Part B contains temporal and con-
ditional clauses, as there were reasons to expect that they might pattern with
category C as well as A, given some of the existing literature (Hirschbiihler
and Junker (1988); Adams (1988)). Finally, part C of the tables contains wh-
clauses (indirect questions, relatives, comparatives), except for those intro-
duced by que. Before turning to these tables, we shall give a brief overview
of the distribution of null and pronominal subjects in the CNNV.
The total number of null subjects in the text is 2403. This number excludes
(1) direct questions and (2) the second (third, etc.) of a series of coordinate
clauses whose null subject is interpreted as identical to the subject of the pre-
ceding coordinated clause. The null subjects can be further subdivided as in
Table 10.1. As can be seen, the 513 examples of null subjects in subordinate
clauses represent 21.35% of the total number of null subjects in the text.
Table 10.1
Distribution of Null Subjects (Spo):
Person 12 2 3 31 7 ce 4 5 6 Total %
Main Clause 127 12 927 415 28 28 53 300 1890 78.61
Subord. Clause 7 6 124 207 4 7 127 31 514 21.38
Total 134 18 1051 622 32 35 180 331 2404 100.00
264 VERB SECOND AND THE NULL-SUBJECT PARAMETER
3. Sub 0V
We first examine the cases where no constituent intervenes between the
subordinator and SpecIP, leaving to SECTION 4 the study of examples with an
intervening XP, Table 10.2 summarizes the facts for null subjects in embed-
ded verb-first constructions. There are 179 such Spo (second column). The
column "Total 0" gives the sum of the embedded Spo in verb-first (i.e. Sub 0
VX, Table 10.2) with those in verb-second (i.e. Sub XP VX, Table 10.15)
constructions.
The 179 Spo examples in Table 10.2 are to be compared to the 3884 Spr
examples of Table 10.3; they represent 4.40% of the total Sp in verb-first con-
structions (4063). Null subjects are excluded in category A (except for the
Strong Agr fifth person and, to some extent the fourth). If only categories B
and C are considered, Spo represents 7.41% of the total of Spo + Spr (6.51%
for category B and 8.06% for category C). The embedded verb-first construc-
tion with Spo is therefore not that frequent compared to the embedded verb
NULL SUBJECTS IN CENT NOUVELLES NOUVELLES 265
second with preverbal Spr, but it cannot be ignored, since its distribution
appears to be subject to restrictions related to the type of subordinate clause,
the person of the verb, and the personal vs impersonal distinction, as is
discussed below.
Table 10.2
Subordinator 0 V X (179/514 = 34.82% of Embedded Spo)
Table 10.3
Subordinator Spr V X (3884)
were to behave very much like person 5. But there are only three examples of
a Spo subject against 53 cases of a Spr subject for person 4, i.e. only 5.35%
of Spo subjects. If a total of 56 cases is enough to feel confident that these
5.35% represent a regular syntactic phenomenon, but of low frequency, fac-
tors separate from morphological distinctness will have to be found to explain
the difference between person 4 and person 5.1 have nothing to contribute on
this aspect of the data at the present time.
A comparison of Table 10.2 with the parallel section of Table 10.3 shows
that the absence of null subjects from a person distinct from 5 or 4 is signifi-
cant beyond any doubt, as there is only one null subject in that case against
2309 examples with a Spr subject. This unique example is of person 3 (against
1365 Spr examples of person 3). Thus the subordinators of category A do
not formally license null subjects in SpecIP and prevent Infl from doing it.
(3) ...et se print a luitter encontre celle tendre bergiere par tel maniere
et par tel force que la ruyst au bas.... 069015
...and [he] started to fight with this shepherdess in such a way and
with such force that he threw her down...
One more comment needs to be made regarding category A of Table 10.2.
Given their behaviour with respect to null subjects, comparatives-superlatives,
restrictive relatives and clefts (grouped under R.Rel. que), as well as indirect
questions headed by que (absent from Table 10.2 from lack of examples) have
been grouped with conjunctive clauses introduced by que rather than with the
wh-constructions found under category C. When we take into account only
persons 3, 31 (expletive) and 6 (excluding persons 1 and 2, given that they
are never represented by null subjects in category C, as well as person 4 and
5 because they could allow a null subject in SpecIP whatever the type of the
embedded clause), there are 486 examples of wh-clauses introduced by que
with a pronominal subject against none with a null subject. By comparison,
the percentage of Spo as opposed to Spr is 34.54% for person 5, and 8.33%
for person 4 (but again, there are only two examples of a null subject in this
last case). Thus, these wh-clauses headed by que behave just as regular non-
wh-que clauses. This is important since in other cases they have been assumed
or shown to behave like wh-clauses in general (see for example, Adams (1988);
Hirscbiihler and Junker (1988); Vance (1988); Roberts (1992)), the distinction
between the two types of clauses headed by que being attributed then to the
presence of a +wh feature in the que relatives. In the present case, one is led
to conclude that in the CNNV, a wh-feature associated with que is not enough
to formally license a SpecIP position that dominates pro. As we suggested in
SECTION 1.2.4, the situation just described could be described by saying that
when Infl/Agr is grammatically active with respect to SpecIP as the identifier
of its content, it counts as a potential formal licenser for it, but not as an ac-
tual one, because of the filled embedded C or SpecCP. Competition with an-
other potential formal licenser prevents +wh-que from formally identifying
SpecIP.
This concludes our survey of category A with verb-first embedded clauses.
We now turn to categories B and C which will be discussed at some length.
268 VERB SECOND AND THE NULL-SUBJECT PARAMETER
3.3 Category C
We will go into the details of the various clauses of part C of Tables 10.2
and 10.3. But first, let us have a brief overview of the main facts regarding
clauses of this category. These are summarized in Table 10.5 (a subpart of
Table 10.4).
We may first note that there is not a single example of a null subject of
person 1 or 2, while there are examples of null subjects for all the other per-
sons. The facts are clear for person 1, as there are 122 examples with a pro-
nominal subject of this person in clauses of type C; they are perhaps less clear
for person 2, given that there are only thirteen examples with a pronominal
subject.
The absence of null subjects of persons 1 and 2 in clauses of type C is what
led us to suggest in SECTION 1.2.2 that for persons 1 and 2, the capacity to
identify the content of SpecIP via Spec-Head agreement is available only if
Infl formally licenses the same position simultaneously. As in subordinate
clauses the power to formally license SpecIP is normally reserved to the C
system, realization of persons 1 and 2 by a pro subject will be excluded in
the environment discussed. The only examples which go against the general
ban against null subject of person 1 or 2 in a verb-first embedded clause are
(5) and (6b), discussed above. There are two possible analyses for these. Ei-
ther the complementizer licenses SpecIP, in which case identification by per-
son 1 or 2 is exceptionally not "parasitic," or Infl acts exceptionally both as a
licenser and as an identifier for SpecIP in a verb-first embedded context.
In contrast to persons 1 and 2, the impersonal and person 5 are frequently
null, although person 5 clearly not as much. One unexpected fact is that per-
son 5 is found more frequently in category A of Table 10.2 than in category
C (40.81% vs. 12.80%). For the other persons, there are only between 4%
and 6% of null subjects.
NULL SUBJECTS IN CENT NOUVELLES NOUVELLES 273
Table 10.5
Spo vs Spr in Subordinate Clauses of Type C
Person 1 2 ce 3 31 4 56 6 Total
Spo 0 0 0 20 31 1 16 5 73
Spr 122 13 19 403 68 19 109 79 832
% Spo 0 0 0 4.72 31.31 5 12.80 5.95 8.,06
One factor that we did not discuss in SECTION 1 is the opposition between
impersonal and non-impersonal verbs or uses of verbs in clauses of type C.
In these clauses, the null subject is more frequent with impersonal uses than
with person 3. The contrast may be attributable to discourse conditions or it
may simply reflect a weaker requirement on local identification for the im-
personal, i.e. person and number agreement could be a process independent
from transmission of a referential index. We have no definite idea on this
aspect of the data presently.
These general considerations being made, let us take a closer look at the
data of category C. This will provide more justification of what was said about
wh-que in SECTION 1.2.4 as well as allow additional observations. In particu-
lar, we will see that there is no clear indication that null subjects are regu-
larly allowed in indirect questions or in clauses introduced by comme. This
would leave only the introducer of relative clauses as a regular formal licenser
of SpecIP. If other texts display the same characteristics, it would show ei-
ther that the licenser of the empty category in subject position is not simply
the wh-word introducing the clause, or that other factors, still to be discov-
ered, play a role. This detailed examination, some aspects of which perhaps
may not be of immediate theoretical import for the analysis presented in this
paper, will allow us to get a clearer picture of the grammar of the CNNV and
may become important if new facts or new theoretical frameworks raise new
questions.
Table 10.6
Spo vs Spr in Indirect Questions (excluding que)
Category C Sub Sp V 1 2 ce 3 31 4 5 6
Q.I.: si Spr 66 2 2 2 42 6 0 3 9
Q.I.: si Spo 3 0 0 0 2 0 0 0 1
% Spo 4.34 4.54 — 10
Table 10.7
I.Q. Introduced by que with Spr Preverbal Subject
Category A Sub Sp V 1 2 ce 3 31 4 5 6
I.Q.: wh-que 129 12 2 24 37 27 5 7 15
Indirect questions are the clearest case where que introducing an embed-
ded clause is a wh-word; we consider it to occupy SpecCP, just as it would in
a main clause. The absence of null subject is not only an indication that wh-
que does not formally license a null SpecIP, it also shows that the verb does
not move to C in the context under consideration, as otherwise we would
expect a number of postverbal null subjects. Particularly strong support for
the view that indirect question que is not a licenser comes from impersonal
verbs, which represent the category most easily found with a null subject in
relative clauses. Their total absence with a null expletive subject in que indi-
rect questions contrasts with the high number (i.e. 27) of impersonal verbs
with a pronominal expletive subject.
(13) a. Et luy, qui..., ne sentit rien et ne savoit qu'i luy fut advenu,
091288
And he, who..., didn't feel anything and didn't know what it
happened to him
b. ...je te dires qu 'il te fault faire. 094055
...I will tell you what it is necessary for you to do
c. ...et veit bien quelle heure qu'il estoit.10 091111
...and [he] saw very well what time it was
The variety of the verbs involved (advenir (15), falloir/failloir (7), sembler d
(2), y avoir (2), etre (1)) reduces the possibility of an accidental gap in the data.
3.5.2 Relatives
The relative clauses have been divided into appositive relatives with a nomi-
nal antecedent, appositive relatives with a CP antecedent, and restrictive rela-
tives. If the three types of relative clauses are put together, it can be seen
that the proportion of null subjects in this construction compared to that of
Spr subjects is relatively important for some persons and that the number of
examples may be high enough for the percentages to be reliable, in particular
for persons 31 and 3, and perhaps 6.
Table 10.8
Relatives not introduced by que
Type Total 1 2 3 31 4 5 6 Total
Relatives Spo 50 0 0 16 28 1 1 4 22
Relatives Spr 363 65 2 216 23 17 9 31 341
% Spo 11.89 0 0 6.89 54.90 5.55 10 11.42 6.06%
276 VERB SECOND AND THE NULL-SUBJECT PARAMETER
Table 10.9
Spo vs Spr in Appositive Relatives
Category C Sub (Sp) V 1 ce 3 31 4 5 6
R.A. Spo 36 0 0 0 11 22 1 0 2
R.A. Spr 87 15 1 0 55 4 3 4 5
%Spo 29.26 0 0 0 16.66 84.61 25 0 28.57
Table 10.8 clearly shows that relatives not introduced by que allow null sub-
jects, especially in the case of person 31, and to a lesser extent 3 and 6. Per-
sons 1 and 2 seem to be categorically excluded. Some of the mechanisms
involved have been discussed in SECTION 1.2. The discussion there is however
incomplete as we refrain from being concrete as to what the property of a
subset of relatives and indirect questions (those not introduced by que) is that
leads to formal licensing of SpecIP. But see Santorini's approach in note 5.
Appositive relatives have been divided in two groups, those with NP ante-
cedent and those with CP antecedent. They are examined in Tables 10.9 and
10.12 respectively.
NP appositive relatives have a large proportion of null subjects: 29.26%
Spo against 70.73% of Spr. (To the examples of Spr, one may add three ex-
amples with on which have not been included in the table, as on has no null
variant.) It is important to note the weight of the impersonal: of 36 null sub-
jects, there are 22 examples of expletive ones; this is to be compared to only
four examples of impersonal verbs with an expressed pronominal subject. Null
subjects appear to be the rule in the case of the impersonal. The proportio
of null subjects is high for person 3. It is higher for person 6, but here th
total number of occurrences of Sp (7) is not high enough to make definitive
claims regarding that person. Attention will therefore be focused on persons
3 and 31.
First of all, a closer look at the examples seems to justify a tentative divi-
sion of NP appositive relatives in two classes. The first one, with 13 Spo ex-
amples, is characterized by the fact that the relative is introduced by a wh-word
acting as a determiner of a noun or by a wh-word starting a new sentence after
a full stop (or both).
Table 10.10
Appositive Relatives of Type 1
In this category, null subjects are not limited to person 3 and 31. There is
also a large variety of verbs: se pouvoir emerveiller 'to be able to be amazed',
souvenir (2) 'to remember', y avoir (6) 'to be', etre compte 'to be counted',
prendre devotion 'to take devotion', se doubter 'to be sure', penser 'to think',
tarder 'to be late', etre congneu 'to be known', prendre 'to take', venir 'to
come', donner 'to give', pouvoir aller 'to be able to go', s'accompagner 'to
be accompanied', parler 'to speak', avoir 'to have', se trouver 'to be situated'
(once for each verb). The fact that null subjects are well distributed among
persons and that it is not restricted to a few particular verbs indicates that the
verb-first construction with Spo reflects a productive grammatical phenom-
enon in appositive relatives. With 80% of Spo and the bulk of the cases, the
impersonal is responsible for the high proportion of null subjects in the ap-
positive relatives of type 2. These 80% have to be qualified by the fact that
half of the cases of Spo involve y avoir. Even if we take out the examples
involving this expression, we still have 66.66% of expletive Spo. The examples
involve souvenir a quelqu 'un 'to remember' twice, tarder a quelqu'un 'to be
delayed', prendre devotion a quelqu'um 'to honour someone', venir quelqu'un
'to come' and etre conte de quelqu'un par quelqu'un 'to be talked about by
someone to someone', i.e. they involve obligatory impersonal verbs,
unaccusative ones, and one impersonal passive. With 17.18% from a total of
64 Sp, person 3 confirms that null subjects are not an accident in NP apposi-
tive relatives. A few examples are given in (16).
(16) a. ...une grainge joindant de leur maison, en laquelle 03 pourroit
aller facillement et secrettement de nuyct... 047012
...a barn joining their house, in which [she] could easily and se-
cretly go at night...
b. .. .il demanderoit 1'ausmone, laquelle 03 ne pensoit jamais qu'elle
luy deust refuser... 018030
...he would ask for charity, which [he] never though that she
would refuse him...
c. Symonnat, a qui 031 ne souvenoit plus de 1'an passes, vint a
confesse. 002036
Symonnat, to whom [there] was no recollection of the previous
year, went to confession...
d. ...il arivait en 1'abayee aucun gentil homme de Lorraine, auquel
031 fut comptez des fais de messire Jehan Pare... 005067
...there came to the abbey a gentleman from Lorraine, to whom
[it] was told of the deeds of master Jehan Pare...
e. ...je vueil icy compter ceste presente adventure laquelle 031 n'a
guiere qu'elle est advenue a Mets. 041002
...I want to tell here this present event which [it] is not long ago
that it happened in Mets. [with resumptive pronoun]
NULL SUBJECTS IN CENTNOUVELLES NOUVEULES 279
As we saw in Table 10.8, when the three types of relative clauses are put
together, the proportion of null subjects in this construction compared to that
of Spr subject is relatively important for some persons. The number of ex-
amples may perhaps be high enough for the percentages to be significant, in
particular for person 31 (54.90% on 51 examples), 3 (6.89% on 232 examples)
and perhaps 6 (11.42% on 35 examples). However, looking more closely, it
is clear that the vast majority of examples with Spo come from NP and CP
appositive relatives, and that restrictive relatives not introduced by que rarely
contribute an Spo example. It is therefore clear that some factor (either gram-
matical, discourse or other) should allow us to distinguish between the two
types of relatives. One hypothesis that we briefly considered was that in NP
appositive relatives: the same hypothesis could have been made for CP ap-
positive relatives, the fronted wh-phrase was taken as a topicalized constitu-
ent triggering V-to-C, exactly in the way that any non-wh-phrase fronted to
SpecCP results in V-to-C. We do not adopt that approach, given the 27 cases
of Spr in person 1, vs none with Spo, in NP and CP appositive relatives. As
made clear earlier, we have not identified the particular factor responsible for
formal licensing of a null subject in SpecIP in some wh-clauses. As a result,
it is useless to speculate at this point on what the relevant difference between
NP and CP appositive relatives and restrictive relatives—and for that matter,
wh-indirect questions perhaps—is. Clearly, this area is open for further re-
search.
3.3.3 Comme
In order to be complete, one last type of clause must be considered, i.e.
clauses introduced by comme 'as'. In this case, apart from person 5, where
null subject is licit independently of the complementizer, there are only two
examples of a null subject, both impersonal.
In the fifteen cases of person 5, we find the same verb in the same formula
fourteen times (comme aves ouy 'as you have heard' thirteen times and comme
aves cy devant ouy 'as you have here before heard' once), and in a variant of
the formula once (comme aves tousjours dit 'as you have always said'), which
results in the percentage for Spo being very low if the one expression com-
monly used with Spo is discarded. If, as should be the case, person 5 is not
taken into account, facts are very similar to what was seen in restrictive rela-
tives not introduced by que, i.e. the remaining null subjects are limited to the
impersonal.
Table 10.13
comme
Category C Sub Sp V 1 2 ce 3 31 4 5 6
comme Spr 275 38 7 0 78 39 1 90 22
comme Spo 17 0 0 0 0 2 0 15 0
% of Spo 5.82 0 0 — 0 4.87 0 14.28 0
282 VERB SECOND AND THE NULL-SUBJECT PARAMETER
4. Sub XP (Subject) V
One position adopted at the beginning of this paper, following insights from
Roberts, is that, in embedded clauses, whether or not SpecIP is formally li-
censed depends on properties of C, and not on Infl. This was discussed in
relation to the embedded verb-first construction. In the analysis defended here,
the verb-first construction is an [lPproVX] construction. The SVX construc-
tion may be part of a larger construction where a first constituent (or a series
of such constituents) is adjoined to IP, as in the following examples:
(20) a. ...mais d'une chose je vous prie... 002045
...but one thing I ask you...
b. Et ce fait, il bouta ce maistre tuppin dedans le mur... 011025
And this being done, he put this large pot in the wall.
284 VERB SECOND AND THE NULL-SUBJECT PARAMETER
We saw that this construction allows for null subjects, in other words, Infl or
Agr licenses and identifies a pro subject on its left in SpecIP, as in (21), which
is similar to (Id) above. As the participial phrase is never found with a
postverbal pronominal subject, it is clear that the null subject is to the left of
the tensed verb:
(21) Et cela faict, marchanda audit poinctre de luy faire ung sepulchre
And this done, [he] made a deal with the painter that he would paint
a sepulchre. 011032
The same type of constructions are found in embedded clauses, i.e. with Spr
or Spo, as clearly shown in (1c) above where Spr and Spo alternate. The
distribution of each construction is given in Table 10.14 and Table 10.15 with
examples following each table:
Table 10.14
Sub XP Spr V (72+9=81)
A: Que Sub XPSp V 1 2 3 31 ce 4 5 6 on
purpose 1 1
cause 5 1 2 1 1
complement 27 2 1 12 3 2 3 4
misc 1 1
concessive
result 19 13 1 1 1 1 2
optative
comp.-superl. 1 1
Rest.ReL: que 2 2
Total A 56 6 2 28 4 1 4 5 6
C: Wh # que Sub XP Sp V 1 2 3 31 ce 4 5 6 on
Q.I.: si I 1
Q.I.: wh # que 0
NP Appos. Rel. 6 1 4 1
CP Appos. Rel. 5 4 1
R.Rel. # que 6 2 3 1
comme 'as' 2 1 1
Total C 20 4 11 1 1 3
Total A, B, C 81 10 2 42 4 1 4 2 7 9
NULL SUBJECTS IN CENTNOUVELLES NOUVELLES 285
Table 10.15
Sub XP V X (with null subject)11
(335/514 = 65.17% Embedded Spo )
A: Que Total Sub X V 0 1 2 3 31 4 5 6
purpose 7 2 1 1
cause 16 10 1 5 4
complement 114 92 1 36 43 1 4 7
misc. 13 8 1 4 2 1
concessive 7 4 2 1 1
result 47 45 1 28 8 1 7
optative 2 2 1 1
comp.-superl 6 4 2 2
Rest.Rel.: que 28 9 1 1 6 1
Total A 240 176 4 4 83 61 1 5 18
Along the lines of Adams (1988), examples of that type can be analyzed as
embedded CPs, with the first phrase after the subordinator in SpecCP and the
verb in C. These examples, with the verb governing SpecIP, allow for a null
subject, but as I indicated just above, examples like those in (23) are not of
the verb-second type, since they are found in all types of embedded clauses,
in particular clauses in which a postverbal Spr is excluded. In addition, they
are often introduced by phrases that are not found with a postverbal pronomi-
nal subject, i.e. that do not trigger V-to-C and are attested in conjunction with
a preverbal pronominal subject, as can be seen by comparing some of the
examples in (22) and (23).
The array of facts discussed suggests that null subjects in examples of the
type in (23) must be accounted for exactly like the parallel cases in main
clauses, i.e. Infl is responsible for both formal licensing and identification,
despite the presence of a subordinate clause marker. I assume that here Infl is
able to act as the formal licenser of the embedded SpecIP because it is the
only potential formal licenser for SpecIP. The first C up is not a potential
formal licenser of SpecIP because it does not govern it, as the result of the
intervention of the XP adjoined to IP.
Given that Infl is the formal licenser of SpecIP in the construction discussed,
we expect null subjects of person 1 and 2 to be found. There are six examples
in the text of person 1 (25) and five of person 2 (26) (the fifth example is in
(lc)):
(25) a. je me repens fort que jamais 0l vous feis venir ceans. 018137
I regret very much that ever [I] made you come here
b. nonobstant qu'en ma vie plus chandeilles 01 n'acheta 019128
despite [the fact] that in whole my life I had not bought more
candles
c. ...de la mesure que ainsi 0l as oublies 029023
...of the measure that so [you] forgot
d. va a tous les deable, que jamais plus 01 ne te voie. 066046
go to the devils, that never more [I] see you
e. par quoy a ceste heure 01 vous en vueil dire aucune chose
080001
as a result of which at this hour [I] I want to tell you something
about it
f. Par quoy, en acquitant ma promesse, 0l vous dir6s de celuy
chareton 082069
By which, by accomplishing my promise, [I] will tell you of this
carter
(26) a. Ne quel affaire a tu eu que au moins 02 ne t'es venu confesser
013019
And what happened to you that at least [you] didn't come to
confess yourself
288 VERB SECOND AND THE NULL-SUBJECT PARAMETER
5. Conclusion
Most of the main points having been stressed several times in the text, this
conclusion will be brief.
First of all, we saw that it is necessary to establish a tripartition between
the various types of subordinate clause introducers according to whether they
contributed to the formal licensing of SpecIP, or not. The division is not sim-
ply between elements that do and elements that do not formally license SpecIP,
since we have seen that wh-que did not license SpecIP when it was connected
to Infl (in the sense that this category was the source of the potential identi-
fication of pro in SpecIP), while it did in the Stylistic Inversion construction
(where we took the position that SpecIP is not associated with Infl).
Second, we saw that it is necessary to establish a tripartition between the
various persons. Person 5 is always allowed, persons 1 and 2 never (when
adjacent to a complementizer), and persons 3, 31 and 6 are allowed in wh-
clauses.
Moreover, a detailed examination of the various types of wh-clauses has
shown that a null subject in a verb-first construction is not equally well rep-
resented in all of them. At this point, we have no clear idea as to why this is
so, in particular as to whether this is to be attributed to the way formal li-
censing works, or whether it follows from different considerations, having to
do with discourse, for example, or with the author's stylistic preferences.
As a final note, I hope that the observations that have been made here have
shown that the discovery of some regularities—as well as the discovery that
some facts which appear at first sight to go in one direction are, upon close
examination, cases of unproductive formulas—requires much more detailed
descriptions than those which are usually found in the literature, however te-
dious this may be at times.
Notes
* This research has been made possible thanks to grants 410-87-0332 and 410-
89-1131 from the Conseil de Recherches en Sciences Humaines du Canada.
1. Philippe de Vigneulles was a draper from Metz. He wrote his CNNs between
1505 and 1515. He has been described as a semi-literary author.
NULL SUBJECTS IN CENTNOUVELLES NOUVELLES 289
2. For a comparison between the CNNV and the anonymous Cent Nouvelles
Nouvelles (c. 1462), see Hirschbiihler (1992).
3. For the sake of clarity, we will tend to use IP to refer to the highest maximal
projection immediately dominated by CP.
4. There is no example of null subjects of the 2nd person singular in a verb-first
construction in main clauses, but we take this not as the reflection of a prohi-
bition against that construction with a 2nd person sing., but rather as being
related to the fact that there are only a handful of examples (i.e. 12) of null
subjects of the 2nd person sing. in assertive main clauses in this text.
5. The interlinear translations are literal translation for the most relevant parts,
and more idiomatic English glosses for the rest. The first three numbers after
an example correspond to the number of the short story, the next three to the
line in that story. So, '002042' means story number 2, line 42. A null subject
will be indicated by "0"; when present, a numeric subscript indicates the
person.
6. Santorini (1992) develops an analysis of Yiddish based on Platzack and
Holmberg (1989). She suggests that a finiteness operator [+F] is located in
Comp (for verb-second languages) or Infl (for non-verb-second languages).
[+F] licenses Nominative Case under government (strict c-command) and/or
agreement configurations, depending on the language. In addition, she de-
fends the idea that empty expletives must be (i) licensed by being head-gov-
erned by a case assigner, and (ii) identified by being co-indexed with Agr.
The analysis sketched in the text could be reformulated by saying that in wh-
clauses not introduced by que, C° contains the [+F] feature, while this feature
would be absent from the other embedded clauses. It could then be said that
licensing in Vigneulles is restricted to head government by a case-assigner
when there is a configuration of head-government at hand. The "when-clause"
is to ensure that licensing under Spec-Head agreement is possible in Vigneulles
when there is no potential head governor, i.e. in verb-first assertive main
clauses of the SVO type (whether they be analyzed as CPs or IPs). Consider-
ing embedded clauses, an empty SpecIP could not be licensed by a [+F] fea-
ture in Infl given that there is a higher position (Comp) where a [+F] feature
might have appeared. One attractive aspect of the analysis is that it identifies
more precisely than in the approach adopted here the particular grammatical
property responsible for the licensing of SpecIP under government. I have
not yet committed myself to this analysis because I do not see why in embed-
ded clauses the finiteness feature would be located in C0 only in those wh-
clauses not introduced by que (as I consider that the interrogative que is a wh-
word in SpecCP, it could not be that the [+F] feature is present in C only when
C is devoid of lexical material); I also have difficulty conceiving a finiteness
operator distinct of Infl (or Tense).
7. '31' represents impersonal subjects.
8. According to Santorini (1992, n. 10), 1% is the level at which a phenomenon
does not reflect a regular feature of the language: "On the basis of detailed
quantitative work of my own and others, it appears that it is common for well-
established generalizations in a language to be violated in naturally-occur-
290 VERB SECOND AND THE NULL-SUBJECT PARAMETER
ring usage at a low, relatively constant rate of about 1%. For instance, the
relative frequency of resumptive pronouns in English in non-island environ-
ments is around 1% (Anthony Kroch, p.c.)."
9. The null subject is parallel to ce rather than il, and the omission of ce could
therefore be due to avoidance of strings like se ce. See the discussion in the
text to follow.
(i) ...et vous dis bien, ..., que se ce ne fut de peur de gaster mon lit,...
041066
... and I tell you, ..., that if it were not for fear of spoiling my bed,...
10. Wh-phrase quelle heure is followed by the complementizer que, the example
has been classified with the I.Q. introduced by que.
11. Some of the subjects are preverbal, others postverbal. See Vance (1988) for a
discussion on how preverbal null subjects may have arisen in Middle French
as a result of the coexistence of "XP Spr V," "XP V Spr" and "XP V" with
null subject.
12. These examples are also distinct from those with a null subjects in that the
types of initial XPs are much more restricted. The XP preceding the verb may
be an adverb or an adverbial phrase: si (twice), ainsi (twice), aussi (twice), a
peine (5 times), voirement (3 times), tout ainsi, en mal an, au moins, encore
au debout de quinze jours; an object as se (= ce) (2 fois) and moult de telles
femmes; or a combination of several phrases: aufait de bien mentir, a cela,
par Dieu, voirement ands'il se courroucait, encor.
References
Adams, M. (1987a) "From Old French to the Theory of Prodrop." Natural
Language and Linguistic Theory 5:1-32.
Adams, M.(1987b) Old French, Null Subjects, and Verb Second Phenomena.
PhD Dissertation, UCLA.
Adams, M. (1988) "Embedded Pro." In J. Blevins and J. Carter, eds. NELS
18, 1-21. Amherst: University of Massachusetts, GLSA.
Deprez, V. (1988) "Stylistic Inversion and the Structure of COMP." Escol
Proceedings.
Dupuis, F. (1988) "Pro-drop dans les subordonnfes en ancien francais." Re-
vue quebecoise de linguistique theorique et appliquee 7:41-62.
Dupuis, F. (1989) L'expression du sujet dans les propositions subordonnfes
en ancien francais. These Doctorale, UniversitS de Montreal.
Hirschbiihler, P. (1990) "La legitimation de la construction Via sujet nul en
subordonn6e dans la prose et le vers en ancien frangais." Revue
quebicoise de linguistique theorique et appliqute 19:32-55.
Hirschbiihler, P. (1992) "Uomission du sujet dans les subordonnees V1: Les
Cent Nouvelles Nouvelles de Vigneulles et les Cent Nouvelles Nouvelles
anonymes." Travaux de Linguistique 25:25-46.
NULL SUBJECTS IN CENTNOUVELLES NOUVELLES 291
Hirschbiihler, P. and M.O. Junker (1988) "Remarques sur les sujets nuls en
subordonnee en ancien et en moyen francais." Revue quebecoise de
linguistique theorique et applique'e 7:63-84.
Martin, R. and M. Wilmet (1980) Syntaxe du moyen francais. Bordeaux:
Sobodi.
Platzack, C. and A. Holmberg (1989) "The Role of AGR and Finiteness in
Germanic VO Languages." Working Papers in Scandinavian Syntax
43:51-76.
Rizzi, L. (1986) "Null Objects in Italian and the Theory of pro." Linguistic
Inquiry 17:501-557.
Roberts, I. (1992) Verbs and Diachronic Syntax. Dordrecht: Kluwer.
Santorini, B. (1992) "Variation and Change in Yiddish Subordinate Clause
Word Order." Natural Language and Linguistic Theory 10:595-640.
Vance, B. (1988) Null Subjects and Syntactic Change in Medieval French. PhD
Dissertation, Cornell University.
Vance, B. (1989) "The Evolution of Prodrop in Medieval French." In J. de
Cesaris and C. Kirschner, Studies in Romance Linguistics, 413-441.
Amsterdam: John Benjamins. (Appeared also as "L'evolution de Pro-
drop en francais medieval," Revue quebecoise de linguistique theorique
et appliquee 1 (1988):85-109.
Zink, G. (1987) "«Quant ce vint au congietprendre». De ce anaphorique a ce
auto-referentiel en ancien frangais." In Etudes de linguistique generale
et de linguistique latine offertes en hommage a Guy Serbat. Bibliotheque
de 1'Information Grammaticale, 417-426. Paris.
This page intentionally left blank
Part Three
Clitics and Verb Second
This page intentionally left blank
11
The Diachronic Development of Subject
Clitics in North Eastern Italian Dialects*
Cecilia Poletto
Universities of Venice and Padua
Introduction
The aim of this paper is to provide some insight into the evolution f subject
clitics of Northern Italian Dialects from the Renaissance period to the present
stage. It will be shown that subject clitics are strictly related to the head of
the Agreement projection. In particular it will be argued that subject clitics
have been reanalyzed as heads that take on functions normally related to the
Agr head in Standard Italian.
From a relatively homogeneous stage Northern Italian Dialects have devel-
oped different systems in which subject clitics have specialized as the pro-
drop licenser head, or the Nominative Case-assigning element, or can even
occupy an additional Agr head. In SECTION 2 and 2.1 the Renaissance Veneto
dialects will be shown to be exactly parallel to Renaissance French, both in
the treatment of subject clitics as well as subject DPs and in the pro-drop
system which is activated by the presence of a "strong" Agr or C head (where
strong is defined as carrying a particular grammatical feature). SECTION 3 deals
with the Veneto dialect of the 16th century which presents the system of a
full pro-drop language in which subject clitics are specified as pro licenser
heads. Subject clitics can specialize not only as pro-drop licensers, but as
Nominative Case-assigners, also, as is the Case for some modern varieties (cf.
Poletto (1993a)).
In the last two sections it will be shown that the spectrum of the functions
played by subject clitics can be even wider: a special series of subject clitics
will be shown to appear only with auxiliaries, in order to lexicalize a higher
Agreement projection available only to auxiliaries as verbs that do not assign
theta-roles. The most advanced variety, namely Friulano, shows a very wide-
spread use of subject clitics which signal the presence of another Agreement
head that attracts clitics.
295
296 CLITICS AND VERB SECOND
In (1) the head of AgrP assigns Nominative Case to the subject, which is placed
in its specifier position. This is the position in which a null subject is licensed
in Standard Italian. Northern Italian Dialects (from now on NIDs) are pro-drop
languages just as Standard Italian is. In (1) in fact, a pro is licensed in the
SpecAgr position.
Nevertheless the Agreement structure of NIDs is more complex: a subject
clitic appears adjoined to the head of AgrP, where the verb has moved from
THE DIACHRONIC DEVELOPMENT OF SUBJECT CLITICS 297
the V position through T in order to incorporate the Tense and Agreement mor-
phemes placed respectively under T and Agr.1
We will briefly review the arguments used by Rizzi (1986) and reported by
Brandi and Cordin (1989) in order to show that subject clitics of the NIDs
are heads, because these tests will be important for the following discussion
about their development from the Renaissance period to their present status.
Subject clitics are considered to be heads because they appear to the right of
the preverbal negative marker, while subject DPs and French subject clitics
appear on the left:
(2) To mama no vien Venetian
Your mother not comes
(3) Elle ne vient pas French
She not comes not
(4) No la vien Venetian
Not she comes
As the position of NID subject clitics is to the right of the preverbal negation
marker, while subject DPs (and French subject clitics) appear on its left, (cf.
(2)-(4)), we cannot assume that NID subject clitics occupy the same position
that DPs fill at S-structure.2 Another test that reveals the status of NID sub-
ject clitics as heads adjoined to Agr is Agr' coordination. It is a fact that NID
subject clitics have to be repeated in coordinate structures, while subject DPs
and French subject clitics can be omitted in the second conjunct of the coor-
dination:
(5) Nane lese el giornale e fuma un toscan Venetian
John reads the newspaper and smokes a cigar
(6) II lit le journal et fume un cigare French
He reads the newspaper and smokes a cigar
(7) El lese el giornal e *(el) fuma un toscan Venetian
He reads the newspaper and *(he) smokes a cigar
In (5) the subject DP Nane can be omitted in the second member of the coor-
dination, the same is possible for French subject clitics as (6) shows, but in
NIDs this is excluded. In (7), in fact, the sentence is grammatical only if the
subject clitic is repeated. This contrast can be explained only accepting that
the subject clitics in the NIDs are structurally closer to the inflected verb than
a normal subject DP, and precisely that subject clitics occupy a position un-
der Agr', while subject DPs occupy the SpecAgr position. On the contrary,
French subject clitics occupy a DP position, namely SpecAgr, and cliticize to
the inflected verb only at PF (cf. Kayne (1975)).
In purely structural terms, it can be assumed that subject clitics in NIDs
are adjoined to the head of Agr as in (1). According to Belletti (1990), the
inflected verb moves up to the Agr head position in order to incorporate the
Agreement morpheme.
298 CLITICS AND VERB SECOND
As (I) shows, the subject clitic is adjoined to this head and this explains
why subject clitics cannot be separated from the inflected verb by any other
element than other clitics.
In the dialects studied by Rizzi (1986a) and Brandi and Cordin (1989) the
subject clitic is always obligatorily expressed, even if a subject DP is present:
(8) La Maria la magna Trentino
The Mary she eats
(9) La Maria magna
The Mary eats
Also the contrast between (8) and (9) suggests that the subject clitic is not a
true subject but a sort of morphological specification that is always expressed
on the head of Agreement, independently of the element that is realized in
SpecAgr which can be a null subject or a phonetically realized DP.
In order to avoid the possibility of interpreting (8) as an instance of left
dislocation of the subject DP, Rizzi observes that the subject clitic is obliga-
tory even when the subject DP is a Quantifier phrase, which cannot be left
dislocated:3
(10) Tut *(l)'e capita de not Trentino
Everything it is happened by night
(11) Tout (*il) s'est passe' dans la nuit French
While the Trentino data in (10), show that the subject clitic has to co-occur
with a quantifier subject, this is not possible in the French example (11).
Subject clitics of the NIDs are thus a part of the Agreement morphology
and not true subject pronouns. NIDs correlate typologically with French, be-
cause they have subject clitics, but their structure is similar to Standard Ital-
ian because they are pro-drop languages. This assumption also explains why
the series of subject clitics is not complete for all persons in most NIDs, while
it is complete in French, where subject clitics behave as subject DPs with re-
spect to the tests presented here.
A closer examination of the distribution of subject clitics in other NIDs
shows that not all subject clitics have the distribution described by Rizzi
(1986a) and by Brandi and Cordin (1989).
In particular, the tests in (4) and (7) are valid also for the subject clitics of
Veneto that we will examine here, suggesting that they are all heads. On the
contrary, the distribution of subject clitics can vary with respect to subject DP.
Not all subject clitics can appear when there is a phonetically realized subject
DP in the sentence. As proposed in Poletto (1991), I will assume that subject
clitics in NIDs can be distinguished on the basis of a movement versus base-
generation analysis.
As proposed by many authors (cf. in particular Koopman and Sportiche
(1991)), I will assume that the subject is generated inside the VP, and pre-
cisely in the SpecVP position, where it gets its theta-role assigned and it is
raised successively to SpecAgr in order to get Nominative Case. I will refer
THE DIACHRONIC DEVELOPMENT OF SUBJECT CLITICS 299
to this subject position inside the VP as the basic argumental subject position.
When a subject clitic is generated in the basic argumental position inside the
VP, it gets the subject theta-role, which is assigned in that position, and then
moves to Agr. No other subject can occur in this structure because the basic
subject position is occupied by the trace of the subject clitic.
If the subject clitic on the contrary is base-generated in its surface position
in Agr, it is an expletive, deprived of the subject theta-role. The subject theta-
role is in fact assigned into the lower position in the sentence structure inside
the VP. As the basic subject position inside the VP is empty, it can be filled
by another DP which absorbs the subject theta-role.
So, subject clitics that are generated inside the VP and then moved to Agr
can receive the subject theta-role and are argumental clitics, while subject
clitics base generated in Agr are expletive elements, as they do not have any
theta-role.
The tests that permit us to distinguish between argumental and expletive
clitics are the following:
(12) a. 1'ha paria qualcheduni
cl has spoken somebody
b. *E1 parla qualcheduni
cl speaks somebody
In (12a) the subject clitic can co-occur with a subject DP which is realized in
the postverbal subject position, while the clitic in (12b) cannot.4 So the sub-
ject clitics described in (10) for Trentino can be assimilated to the expletive
clitic in (12a) because they are compatible with a subject DP in argumental
position.
Subject clitics of the type of el cannot appear if the subject DP has been
moved through wh-movement (as for instance restrictive relatives,
topicalization or clefting), while subject clitics of the type of l can:5
(13) a. El puteo che (*el) vien vanti... Veneto
The boy that (*he) comes along
b. Ti che *(te) vien vanti
You that *(you) come along
(14) a. NANE, che (*el) vien vanti...
JOHN, that (*he) comes along
b. TI, che *(te) vien vanti
YOU, that *(you) come along
(15) a. Ze Nane, che (*el) vien vanti
Is John, that (*he) comes along
b. Te si TI, che *(te) vien vanti
You is YOU, that *(you) come along.
300 CLITICS AND VERB SECOND
(13), (14) and (15) represent respectively cases of restrictive relative clause,
topicalization and clefting. In all these cases the third person subject clitic
cannot co-occur with the variable trace, while the second person singular sub-
ject clitic can (indeed it must).
The explanation for the contrasts in (12), (13), (14) and (15) is that, as men-
tioned above, argumental subject clitics leave a trace in the basic subject po-
sition through which the subject theta-role is transmitted. Hence they cannot
co-occur with another subject, which would occupy the position of the trace.
Non-argumental subject clitics on the contrary are base-generated in their
superficial position, leaving the basic position free for another subject, which
is the OP nisuni in (12) and the variable trace of wh-movement in (13b), (14b)
and (15b).
Some dialects have both expletive and argumental clitics; the Veneto vari-
ety that we used for the examples above is precisely this kind. Other varieties
realize only one of the two possibilities.
From a diachronic point of view, it is interesting to investigate how sub-
ject clitics of the NIDs have developed to reach their present status. Have they
always been heads like today, or were they similar to French in some previ-
ous stages of evolution? Renzi (1989) has shown that Fiorentino of the 18th
century was like Modem Standard French with respect to the distribution of
subject clitics. If this is true, the same could be valid for North Eastern Ital-
ian dialects too, in particular for Veneto (cf. Vanelli (1987)). In the following
section the tests presented here will be applied to Veneto of the Renaissance
in order to determine which syntactic status subject clitics have in this period.
not count for C to be a pro licenser. The intuition is that C counts for the pro-
drop theory only if it is "visible" in some sense to be defined.
On the other hand, the distribution of argumental null subjects of first per-
son singular and plural and second person plural does not seem to be depen-
dent on any feature in C. There are examples of null subjects of first person
and second person plural both in main and embedded sentences:
(27) a. Ve suplico (Calmo p. 72) RVe
(I) pray you
b. Havemo buo notita che.. (Calmo p. 129)
(We) have had news that...
c. Dire a Ser Zuan che... (Ruz. p. 107)
(You+plur.) will say to Sir John that...
(28) a. Co avesse ben dissenao (Calmo p. 1ll)
When (I) had well dined
b. Quando aspetemo suto, ... (Calmo p. 73)
When (we) await dry weather,...
c. Si vole scambiar tuto... (Calmo p. 94)
If (you+plur) want to exchange everything...
(27) shows that a first person singular and plural and a second plural null
subject is possible in a main clause. Hence, a particular type of Comp (such
as a +wh or a +subjunctive one) is not relevant for the licensing of the null
subject. The relevant head that licenses and identifies the contentive features
of the null subject must then be the head of the Agreement projection.
At this point we have two classes of null subjects. True expletives and first
person singular and plural and second person plural null subjects can be li-
censed both in a main and in an embedded context. On the other hand, exple-
tives coindexed with an argumental subject position, second person singular,
and third person singular and plural null subjects are sensitive to the type of
element which is realized in the Comp position: only a +wh or a +subjunctive
Comp can license this type of pro. The situation is summarized by the fol-
lowing schema:
(29) MAIN CL EMBEDDED CL. EMBEDDED CL
-wh /subjunct. +wh /+subjunct.
expletive + + +
pro -theta
l.person + + +
sing.pro
l.person + + +
plur.pro
2.person + + +
plur.pro
THE DIACHRONIC DEVELOPMENT OF SUBJECT CLITICS 305
2.person - - +
sing.pro
3.person - - +
sing.pro
3.person - - +
plur.pro
expletive - - +
pro+postv.DP
Null subjects can thus be divided into two groups. We will refer to the first
group of null subjects which are not sensitive to the type of Comp as "ex-
tended pro-drop." The second group of null subjects which can only be li-
censed if the Comp projection has a particular type of feature (+wh or
+subjunctive) will be termed restricted pro-drop.
Looking at the distribution of expletive subjects and argumental subjects
in RVe, it is evident that the pro-drop conditions in RVe are strongly reminis-
cent of the situation in Renaissance French (henceorth RFr) type of pro-drop
studied in Roberts (1992) (see also references quoted there).
In RFr the distribution of the null subjects as described by Roberts (1992)
can be resumed as follows: expletive subjects, first person plural and second
person plural null subjects can be found in both main and embedded clauses.
On the other hand, first person singular, second person si gular, and third
person singular and plural can only be licensed in embedded contexts and only
if there is a +wh-item in the Comp projection of the sentence.
The distribution of null subject in RFr is thus the following:
(30) MAIN CL. EMBEDDED CL. EMBEDDED CL.
-wh/-subjunct. +wh/+subjunct.
expletive + + +
pro
l.person + + +
plur.pro
2. person + + +
plur.pro
l.person - - +
sing.pro
2.person - - +
sing.pro
3.person - - +
sing.pro
3.person - - +
plur.pro
The table in (30) shows exactly the same partitioning of table (29) between
extended and restricted pro-drop.
306 CLITICS AND VERB SECOND
not think that verbal morphology is of such a great importance in this matter
that we can attribute the different evolution of these two languages only to
the difference in the number of the morphological distinctions on the inflected
verb. In other words, the richer morphological paradigm of the Veneto vari-
ety is not the only factor that has determined the evolution of this language
into a pro-drop language.
I would like to connect the different evolution of French and Veneto not
only with the number of morphological specifications on the verbal head, but
with the relative balance between verbal morphology and the paradigm of the
subject clitics. The different evolution of RFr and RVe is a particular case of
a generalization formulated by Renzi and Vanelli (1983), which states that the
subject person and number features must always be phonetically expressed by
Agreement or by the subject itself. In other words there must always be at
least one element, verbal morphology or the subject pronoun itself, that ex-
presses the number and person features of the subject. This seems to be true
for all Romance dialects examined by Renzi and Vanelli.
Both RFr and RVe have a restricted system of pro-drop and six subject pro-
nouns which appear in the SpecAgr position. But, in RVe the subject clitics
of first person singular, plural and second person plural have the same form a
or e depending on the variety (cf. (16)). In RFr the series of subject clitics
has a distinct element for all persons of the verb. Hence even in RVe, not only
in Modern Veneto, the morphologically realized inflectional features are the
only elements able to identify the number and the person of the subject. Even
if the subject clitic is in SpecAgr, it has no features that could convey infor-
mation about the subject.
As in RVe subject clitics are not always distinguished for person and num-
ber, so RVe has no other choice than to maintain the person and number fea-
tures on Agreement, reinterpreting subject clitics as heads for the persons that
are not fully specified by verbal morphology. French on the other hand, hav-
ing a full discrete series of subject clitics, has been able to keep them as true
DPs, further limiting the role of Agreement as pro-drop licenser.
Hence, the factor that has determined the split between French and the
Veneto variety (and probably other Northern Italian dialects as well) is not
only the different number of morphological specifications on the verb. It is
the relation between the number of morphological specifications on the verb
and the number of morphological specifications on the subject clitics.
It is interesting to note, however, that both languages have evolved in a way
that respects Renzi and Vanelli's generalization: the person and number fea-
tures of the subject are realized at least once in both languages. The necessity
of expressing these features can thus be considered not only as a synchronic
property of Romance dialects in general, but also a diachronic tendency to
maintain a sort of balance between the features expressed in Agr and in its
Spec position.
I will now examine a problem which is closely connected with the pro-drop
system and the distribution of subject clitics, namely postverbal subjects.
There is another quite interesting problem that is connected with the facts
discussed up to now, namely the free inversion cases in RVe noted by Vanelli
(1987) with a subject clitic in preverbal position.
308 CLITICS AND VERB SECOND
Vanelli observes that examples like (33) constitute a puzzle for Case theory,
given the hypothesis that subject clitics are true subjects in RVe (cf. SEC-
TION 2):
(33) a. El viene quel so fraelo (Ruz. p. 94) RVe
Cl comes that his brother
b. L'e sta suspeso le prediche al Sior Geronimo (Calmo p. 15)
Cl is been suspended the sermons to Mr. Geronimo
(33a) presents a case of postverbal definite subject with an ergative verb and
a subject clitic which is realized in preverbal position. On the basis of the
discussion about the position of subject clitics it is clear that they cannot be
considered as morphological affixes at this stage of evolution. They are true
DPs which absorb the Case of the subject. The problem for the theory is pre-
sented by the fact that the definite subject in the postverbal position needs a
Case, too. It is generally assumed that two phonetically realized elements
cannot be assigned the same Case (cf. Kayne (1983)). So, in this structure
we need two distinct Cases, one for the subject clitic and one for the postverbal
subject DP.10 Looking at verbal Agreement it seems that the Nominative Case
is assigned to the subject because the verb agrees with the clitic and not with
the subject DP. In (33b) the postverbal DP is feminine plural, but the verb is
marked as masculine singular on the past participle and as singular on the
auxiliary.
We will thus assume that the subject clitic in preverbal position absorbs
the Nominative Case, as the verbal morphology indicates. What about the
postverbal DP? The Case assigned to the postverbal DP cannot be Accusa-
tive, because the verb is an ergative one. It cannot be the Partitive Case pos-
tulated in Belletti (1988) either, because Partitive is assigned only to indefinite
DPs and the DPs in (33a,b) are both definite. So the Case assigned to the
postverbal DP can be neither Nominative through Spec-head agreement with
the head of AgrP nor Partitive. In order to solve this problem, we have to
consider how Nominative Case is assigned. I will assume Roberts' (1992) idea
that Nominative Case can be assigned in two different configurations: Spec-
head Agreement with the head of AgrP and government by the head of TP.
The possibilities of Nominative Case-assignment correspond thus to (34)
(cf. Roberts (1992:29 ff.)):
(34) a. Agr assigns Case through Spec-head agreement,
b. T assigns Case through government.
Such a parameter of Nominative Case-assignment has been proposed by Rob-
erts in order to explain the difference between languages such as French and
Welsh. In French the subject appears in the preverbal subject position and it
triggers morphological agreement of person and number with the verb. Fol-
lowing Roberts' hypothesis, French exploits the possibility expressed by (34a).
Hence the subject DP moves from its base position inside the VP to the
SpecAgr position, where it is assigned Case and it triggers morphological
agreement of number and person.
THE DIACHRONIC DEVELOPMENT OF SUBJECT CLITICS 309
On the other hand, in Welsh the subject appears after the inflected verb and
it does not trigger morphological agreement in person and number. This means
that Welsh adopts (34b): the subject DP does not need to move to SpecAgr,
on the contrary it must remain in situ, in order to get Nominative Case as-
signed by the head of TP. Given that that there is no Spec-head agreement
relation between the subject DP and the head of AgrP, there is no morpho-
logical agreement of number and person.
Roberts further assumes that in the Romance languages the subject can be
in the postverbal position because both options in (34) can be selected: Nomi-
native Case can be assigned both by Spec-head agreement with the head of
AgrP or by government from the head of TP. Nevertheless, languages like
Standard Italian always show morphological agreement of person and num-
ber between the subject DP and the verb, while Welsh never does. Roberts
explains this difference on the basis of the observation that in Welsh AgrP is
never active in Nominative Case-assignment, while it is in Romance. On the
basis of this difference, a rule of cosuperscripting between the heads of AgrP
and TP applies in Romance, but not in Welsh.
(35) Coindex Agr and T
A rule like (35) will thus be active in the Romance languages because both
Agr and T are able to assign Nominative, but it will fail to apply in Welsh,
because Agr in Welsh is inert with respect to Nominative Case-assignment.
This cosuperscripting determines the passage of morphological agreement
features of person and number so that the verb and the postverbal subject agree
in person and number in Romance.
Let us now consider the structure of sentences like (33):
310 CLITICS AND VERB SECOND
In (36) the subject clitic el is realized in SpecAgr, while the postverbal sub-
ject is inside the V governed by the head of TP. It is possible to think that in
RVe, as in other Romance languages, both mechanisms of Case assignment
can be exploited, namely that the head T can assign Case to the postverbal
subject DP through government and the head of AgrP can assign Nominative
through Spec-head agreement. It is interesting to note, however, that in RVe
(as in Modern NIDs) no agreement of person and number between the verb
and the subject DP appears to be active. In other words, RVe is more similar
to Welsh than to Standard Italian and other Romance languages.
We have to assume that the rule of cosuperscripting postulated in (35) for
Romance languages does not apply here, but why? Also in RVe there are
preverbal subjects that trigger morphological agreement of person and num-
ber with the verb. Hence also in RVe the AgrP projection is active for Nomi-
native Case-assignment exactly as in other Romance languages. Is the fact that
the rule of cosuperscripting fail to apply a mere coincidence or not? And, if it
is not, is it connected with other selectional choices that the grammar of the
dialect in question makes? It seems plausible to think that the fact that rule
(35) does not apply in RVe is somehow connected with the particular type of
postverbal subjects observed in this dialect.
In other words, the fact that there is no cosuperscripting must be related to
the problem of Case assignment to the postverbal subject in a structure like
(36). We already excluded the possibility that the postverbal subject DP re-
ceives Partitive Case, because it is a definite DP. It cannot receive Nomina-
tive Case through a chain with the expletive, because the expletive is a
phonetically realized element, and it needs a Case of its own.
Now, Kayne (1984) proposes that two phonetically realized elements can-
not be in the same chain and share the same Case, while an overt and a silent
element can. Considering the Nominative Case-assignment possibilities ex-
pressed in (34), we can make the hypothesis that Case is assigned to the post-
verbal subject DP by the head of TP, while the expletive in preverbal position
receives Nominative through Spec-head agreement with the head of AgrP.
At first sight, it might seem strange to assume that two Nominative Cases
are assigned at the same time, even in different structural configurations and
by different structural configurations and by different heads. Note however,
that a system of Nominative Case-assignment like (34) does not specify any-
thing about the possibility that both heads assign Case at the same time. In a
language that selects both heads Agr and T as Nominative Case-assigners it
could be the Case that the two heads are both active, and that two different
DPs get Nominative Case, one through Spec-head agreement with the head
Agr and one through government by the head T.
This double mechanism of Case assignment is restricted by theta theory that
admits only one DP for each thematic role assigned by the verb. So, even if
there are two possible Nominative Cases available, only one of the two will
be realized, because there is only one subject theta-role. If both Nominative
Cases are assigned to two distinct DPs, one of the two will be left without a
thematic role, violating the theta criterion. There is only one situation in which
a DP can be left without a thematic role, namely where DP is an expletive.
THE DIACHRONIC DEVELOPMENT OF SUBJECT CLITICS 311
A structure with double Case assignment is thus possible only when one of
the two elements is an expletive. Furthermore, the expletive element must be
the higher one, because SpecAgr is the non-thematic position. If the exple-
tive were realized in the postverbal position and the subject DP in the preverbal
one, it would be impossible for the subject DP to receive the subject theta-
role, which is assigned inside the VP.
The only situation in which the two Nominatives can be assigned thus cor-
responds to a structure like (36) which does not violate the theta criterion. The
subject clitic in SpecAgr is in fact an expletive, and as such it does not ab-
sorb the subject theta-role, while the postverbal subject DP does.
Hence, RVe has the possibility of exploiting both options expressed in (34)
at the same time. Moreover, it must do so, otherwise one of the two elements
would remain without a Case. This kind of analysis does not seem necessary
for languages such as Standard Italian or Standard French. In Standard Ital-
ian in fact there is a null element in preverbal position, and not a phonetically
realized one. In this case Kayne's restriction about the presence of two ele-
ments sharing the same Case does not apply, because one of the two is silent.
Hence, Standard Italian does not need to exploit both options of Nominative
Case-assignment described in (34) at the same time. The same is true for
French postverbal subjects in the case of Stylistic Inversion: in the preverbal
position a pro is licensed probably by a +wh C (see Kayne and Pollock (1978)),
and the subject DP receives Case directly from the head of TP.
Once we have seen how the mechanism of Case assignment works in a struc-
ture like (36), we can go back to the hypothesis that it may be connected to
the difference that we noted before with respect to morphological agreement
of person and number. In Standard Italian and French postverbal subjects
trigger morphological agreement with the verb, while in RVe this is not so. In
order to explain this fact, we assumed Roberts' cosuperscripting rule between
the heads of AgrP and TP to be active in Romance but not in RVe. It seems
that when the cosuperscripting rule applies, the two heads of AgrP and TP are
treated as one, both with respect to the morphological features of person and
number and with respect to Case assignment.
We can thus assume that the rule of cosuperscripting blocks independent
Case assignment by the two heads that are able to assign it. Hence the double
head constituted by Agr+T can only assign Case once: through government
or through Spec-head Agreement. On the contrary, when the cosuperscripting
does not apply, the two heads are considered as distinct elements by the gram-
mar: they do not share morphological agreement features and can both assign
Case independently.
In RVe the rule of cosuperscripting cannot apply, otherwise one of the two
Nominatives would get lost and the lexical expletive or the postverbal sub-
ject DP would remain without a Case. Moreover, a structure like (36) is the
one in which the possibility of a double Case assignment is realized, because
it is the only situation which is not blocked by the theta criterion.
If the rule in (35) does not apply, no sharing of the morphological features
between Agr and T is possible: hence the verb must agree with the preverbal
expletive clitic and not with the postverbal DP. In particular, we expect that
312 CLITICS AND VERB SECOND
century (henceforth SVe) some subject clitics appear at the right of the
preverbal negative marker:
(38) Perche no la pole (Oda p. 441) SVe
Because not she can
(39) No i te fa male (Oda p. 443)
Not they to-you do harm
In a sample of 145 sentences there are no cases of coordinated structures, so
the second test cannot apply. We are thus compelled to base our analysis only
on the fact that subject clitics appear at the right of the preverbal negative
marker, and for this reason they are to be considered heads at S-structure. As
discussed in SECTION 1, the fact that a subject clitic appears after the negative
marker shows that subject clitics and subject DPs do not occupy the same
position in the syntax: subject DPs in fact can only appear at the left and never
at the right of the negative marker. Hence, we can conclude that SVe subject
clitics are analogous to their modern counterparts.
So, it seems that subject clitics have been reanalyzed, during the period
between the 16th and the 17th century, as part of the inflectional head of AgrP.
As already discussed in SECTION 1, this does not mean that subject clitics at
this point of their evolution are not arguments in the sense that they do not
absorb the subject theta-role. Even if they are heads, they can start out from
a thematic position inside the VP and adjoin to the head of Agr blocking the
insertion of another subject, because the thematic position is filled by the trace
of the subject clitic.
Object clitics in Romance have normally the distribution of argumental
heads: when the object clitics are inserted, no object DP can be phonetically
realized and no variable can occupy the object position.11
If we apply the tests already discussed in SECTION 1, we are compelled to
admit that subject clitics of this period are bound to an argumental position.
A subject clitic is not required when a subject DP is present, as in (40):
(40) I toroere vale pi che no valse qui de Hisperite (Oda p. 442) SVe
Your oak woods are more precious that not those of Hesperide.
Subject QPs always appear without a subject clitic and there is no subject clitic
when the subject is marked +wh and moved outside the sentence:
(41) Agno pomaro fea pumi indore (Oda p. 441) SVe
Every apple tree made golden apples
(42) Agnun che bita dentro i tredese comun (Oda p. 443)
Everyone that lives in the thirteen villages
(43) Chi po far retirare el mare si ingordo? (Oda p. 443)
Who can let retreat the sea (which is) so greedy.
(40), (41), (42) and (43) show that subject clitics in SVe are parallel to object
clitics: they absorb the subject theta-role and are incompatible with other sub-
jects in argumental position.
314 CLITICS AND VERB SECOND
Therefore, the structure of a sentence with a subject clitic will be (1) (here
repeated as (44)):12
THE DIACHRONIC DEVELOPMENT OF SUBJECT CLITICS 315
In (44) the subject clitic starts in the VP internal subject position as the trace
t! under DP1 indicates, and moves up to Agr. It ends up in an adjoined posi-
tion to the head of AgrP where the inflected verb is placed after having incor-
porated the affixes of Tense and Agreement. This kind of adjoined position is
the same as that postulated for Modern NIDs (cf. (1)). On the other hand, (45)
describes the situation that we found in RVe, in which subject clitics are still
equivalent to maximal projections in the syntax and are clitics only at PR The
difference between (44) and (45) can be interpreted as a modification of the
subject clitic, which changes its categorical status. It is no longer analyzed as
an XP that does not branch, as it does not have a Specifier and a Complement
position, but as a simple head. As the structure preservation principle states
that all XPs must move to an XP position and all X must move to head posi-
tions (cf. Chomsky (1986)), the subject clitic can no longer move to the
SpecAgr position, which is an XP position, it can only move up to the head
of this projection from the basic subject position inside the VP. Hence, the
reanalysis of subject clitics as heads implies that they move to a head posi-
tion.
As (44) illustrates, subject clitics move to the head of AgrP. We can imag-
ine different motivations that induce subject clitics to move just into this head:
first of all no head containing a trace can host the subject clitic. T and V are
both occupied by the trace of the verb which has moved to Agr. If the clitic
adjoined to T or to V, it would induce a minimality effect between the trace
and the inflected verb in Agr, yielding a structure like (46):
A configuration like (46) is excluded by Baker (1988). In fact the subject clitic
would be a closer potential governor for the trace in T and it would prevent
the correct relation between the verb in Agr and its trace in T.
Hence the subject clitic must adjoin to a head which is not filled by a trace,
but by a phonetically realized element, and only Agr is such a head.
Second, the movement of the subject clitic to a left adjoined position to
Agr recreates the same configuration at the X level that subject DPs have with
Agr at the XP level. Adjunction of the subject clitic is structurally similar to
a relation of Spec-head Agreement, but at a lower level.13
Third, if we consider Kayne's (1989) proposal that all syntactic clitics move
to the head of AgrP in Romance, then also subject clitics, being syntactic
clitics, will be attracted by this head.
At this point we can ask if the reanalysis of subject clitics from purely
phonological clitics as in RVe to syntactic clitics in SVe exerts some influ-
ence on other fields of the grammar. In particular we expect that the processes
connected with the AgrP projection are influenced by this readjustment of the
structure of Agr. Let us for instance, take into consideration the pro-drop sys-
tem. We saw that the pro-drop system of RVe is fairly complex. Two heads
316 CLITICS AND VERB SECOND
are marked as possible pro licensers, namely C and Agr. But only in the case
where they are filled by a particular feature are they visible for the pro-drop
licensing condition.
Looking at the data, it may seem strange to postulate a pro-drop system
for SVe, because in this dialect, there seem to be no cases of null subjects at
all. In fact, a subject clitic or a subject DP is always phonetically realized.
(47) a. Quand'a me tacco a cantare (Oda p. 440) SVe
When I me begin to sing
b. Te si ti solo " "
You are YOU alone
c. La mormolla de ti " "
She murmurs of you
d. A sagion darme... " "
We know to give
e. O golusi slecaizzi ch'a si " "
Oh, greedy that you are
f. Quel ch'j dise " "
What that they say
At a superficial glance, it seems that pro-drop has completely disappeared from
the language. In fact, there is a subject clitic which is obligatory for all the
persons of the verb, a phonetically realized subject DP, or a variable in the
case of wh-movement of the subject.
One may assume that the pro-drop character of RVe has been completely
lost during this century and that SVe is a non-pro-drop language. Things do
not appear to be so simple if we consider that subject clitics are no longer
true subjects in SVe, but heads, as indicated by the tests in (38) and (39) and
by structure (44). At this point three questions arise:
(a) If subject clitics are heads, what kind of element fills the SpecAgr
position?
(b) Why are subject clitics obligatory, when there is no other phoneti-
cally realized subject DP?
(c) Why have pro-drop phenomena disappeared?
On the basis of the Extended Projection Principle, we must assume that
SpecAgr is filled by some element, because the preverbal subject position
cannot be left totally empty in any language. Hence, a null category must fill
it: this category cannot be a variable, because it is not bound by any operator,
it cannot be an DP-trace or a pro, because it is a Case-marked position. The
only category that can occupy the SpecAgr position is a pro. This element,
like all null categories, has to be licensed by a head which in RVe was Agr or
C. We have seen that in SVe subject clitics are obligatory when there is no
subject DP, but that they do not co-occur with subject DPs. If subject clitics
THE DIACHRONIC DEVELOPMENT OF SUBJECT CLITICS 317
appear only when a pro, and no subject DP occupies the SpecAgr position,
we can make the hypotheses that the head that licenses the null subject is
neither Agr nor C, but the subject clitic. The pro-drop conditions of SVe are
expressed in (48):
(48) pro is licensed by a clitic head in Agr through Spec-head agreement.
The null subject is co-indexed with the subject clitic which licenses it through
Spec-head Agreement.
We can thus answer question (b): subject clitics are always obligatory when
there is no phonetically realized subject DP because they license pro. If the
subject clitic is omitted there is no head that can license pro and the sentence
is ungrammatical.
At this point the answer to the third question is quite simple. Pro-drop
phenomena have not disappeared from the language at all. On the contrary,
they are more widespread in the language than before. The change regards
only the type of head that licenses the null subject. This head is neither C nor
Agr as it was in RVe, but the subject clitic adjoined to Agr. The obligatory
presence of a subject clitic simulates the requirement of a non-pro-drop lan-
guage, in which a subject pronoun must always be present. The subject pro-
noun of SVe is nevertheless not a true subject DP, but a syntactic clitic in Agr.
Sve is thus a pro-drop language like Standard Italian is, but it differs from
Standard Italian because the head that licenses pro is not Agr itself, but a
subject clitic adjoined to Agr. The structural configuration is the same in the
two languages, namely Spec-head Agreement, but the head that licenses the
null element is different.
SVe has lost both strategies of pro-licensing that we found in RVe: neither
C, nor Agr are possible pro-drop licensers. It has developed in the direction
of a simpler system, in which only one head can license pro and only through
a unique structural configuration. We see that the evolution of French and
the Veneto variety are parallel. Modern French has also completely lost the
possibility of pro-drop licensing through Spec-head agreement with the head
Agr: no first or second plural person null subjects are admitted in Modern
French as is the case in RFr. French has maintained pro-drop licensing from
C through government, when C is marked +wh or +subjunctive, even if only
for expletive subjects (cf. Kayne and Pollock (1978)):
(49) a. Quand pro viendra Jean? French
When will come John?
b. J'aimerais que pro sorte Paul
I wish that goes out Paul
SVe has lost pro-drop licensing from both C and Agr, but it has developed a
new system, in which another head has this function.
Both SVe and French have developed into systems in which Agr is not a
possible pro-drop licenser. This is the reason why they both have maintained
subject clitics. As has often been noted in the literature, the languages that
have developed subject clitics are precisely those that, in their medieval stage,
318 CLITICS AND VERB SECOND
could only license a pro through government by Agr, which had moved to C
in accordance with the verb-second constraint. Agr was not able to license a
pro through the configuration of Spec-head agreement. The similarity between
French and SVe is to be found in the fact that in both languages Agr was not
able to take up the function of pro licenser through Spec-head Agreement as
was the case in Southern Italian Dialects and other Romance languages like
Spanish.
This weakness of Agr (which we assume to be syntactic and not only
morphological) has brought about the development of an alternative system
in SVe: a subject clitic licenses pro because Agr is not strong enough to do it
in the relevant configuration of Spec-head Agreement. This system is still
adopted by some conservative varieties in isolated areas. This fact is very
important because it permits us to study the licensing conditions of a dead
language such as SVe more deeply and to check our predictions by construct-
ing ungrammatical sentences. One such variety is Rovignese spoken in
Slovenia in the town of Rovigno. The subject clitic series of Rovignese is
complete for all persons (cf. TekavCk; (1986)):
(50) 1. 2. 3. l.plur. 2.plur 3.plur
i ti el/la i i i/le
When a subject DP is not realized, a subject clitic is obligatory:
(51) a. Sa *(ti) me dive la paca Rovignese
If you to-me give a hit
b. * (A) ta par
It to-you seems
The subject clitic is not obligatory when there is a phonetically realized sub-
ject DP:
(52) a. Se Paron Giacomo gira furbo... Rovignese
If Mr. Giacomo was clever
b. La Francia gaviva tuchisto tira veia suldadi de 1'Istria
The France had had to take away soldiers from Istria
c. La feila spativa
The girl waited
Indeed, subject clitics and subject DPs in SpecAgr must be incompatible. In
fact if we substitute the definite subject DP with a QP, which cannot be left
dislocated and can only occupy the SpecAgr position, the subject clitic can-
not appear:
(53) a. Qualunque pol meti la man sul fogo Rovignese
Everyone can put the hand on the fire
b. * Qualunque el pol meti la man sul fogo
Everyone he can put the hand on the fire
THE DIACHRONIC DEVELOPMENT OF SUBJECT CLITICS 319
4. Conclusion
The status and the distribution of subject clitics in Northern Italian Dia-
lects is connected to, at least, three components of the grammar: the pro-drop
parameter, the Case-assignment conditions, and the visibility of empty Agree-
ment heads. It is possible to summarize the entire discussion about the de-
velopment of subject clitics making a quite simple hypothesis regarding the
relation between verbal morphology and subject clitics.
In Standard French, subject clitics have remained true subjects, which ap-
pear in the SpecAgr position like other subject DPs. On the contrary, in all
Northern Italian Dialects, subject clitics have been reinterpreted as a possible
candidate for substituting agreement in various syntactic mechanisms. In all
these cases the function of agreement, both intended as a syntactic position
and as morphological specification, is to identify the subject of a predicate.
Subject clitics, starting as true subjects, have slowly been reanalyzed as heads
that interfere in the strict relation between the head and SpecAgr. In SVe, for
instance, subject clitics mimic the relation of Spec-head Agreement that Agr
has with the subject adjoining to the head and licensing a pro in the SpecAgr
position. The subject clitic thus constitutes a new type of Agreement morphol-
ogy following Renzi and Vanelli's (1983) generalization that the subject fea-
tures must be encoded at least on one of the two elements, but can also be
encoded on both.
Notes
Thanks are due to A. Battye, A. Belletti, G. Cinque, T. Guasti, I. Roberts, L.
Rizzi, A. Tomaselli, R. Zanuttini, and in particular P. Beninca and L. Vanelli
for comments and discussion. All errors are naturally my own.
2. Some varieties allow both the order negation-subject clitic and the order sub-
ject clitic-negation:
(i) Tu un vieni Fiorentino
You not come
(ii) Un tu vieni
Note however, that the order in (ii) is never admitted when the subject is a full
DP. Moreover, there are reasons to believe that in (i) the subject clitic is a
head adjoined to the negative marker (see Poletto (1993a)).
3. Sentences with compound tenses have a different series of subject clitics from
sentences with simple tenses.
4. The Veneto variety used for the examples is the dialect of Oderzo.
5. We use here the second person singular subject clitic, which behaves like the
/ clitic.
6. We will use examples from plays by Ruzante for the Paduan variety and from
a letter collection by Calmo for the Venetian variety. There are only some
minor morphological distinctions between the two. In (16) the first form is
the Paduan, the second corresponds to Venetian.
7. Ca seems to be a specialized form for the comparative complementizer which
is found only in Venetian texts.
8. Subject clitics are still arguments also in Modern Veneto. They have lost the
status of XPs, and are heads just like object clitics, but they never co-occur
with subject QPs or subject variables either in preverbal or in postverbal sub-
ject position.
9. The pro-drop system of Renaissance French and Veneto is different from the
Medieval system. In their Medieval stage, these languages were verb-second.
Pro-drop was licensed by the verb in C, hence possible only in matrix verb-
second clauses. In the Renaissance period, French and NIDs lost verb second,
but the licensing of pro still comes from the C head. As the verb does not
move any more into C, this must be marked with a particular feature in order
to be visible. Agr can only take on the function of pro-licenser if it is morpho-
logically strong.
10. The theory that we propose here cannot be applied to Modem NIDs as it is
formulated. NID subject clitics are in fact heads, and it is not obvious that
they need to be independently Case-marked. We will not discuss the phenom-
enon of Quirky Agreement (cf. Battye (1990)) in Modern NIDs here.
11. We are not considering here cases of clitic doubling, which are quite frequent
in NIDs, but only with indirect object clitics.
12. From the diachronic point of view, it seems quite reasonable to admit that the
change in the structure must happen by means of ambiguous strings of words
(cf. Lightfoot (1979) and Roberts (1992)) that give rise to a possibility of
"misunderstanding" the structure of the sentence. This is surely not the only
reason for diachronic change, because there must be some parametric choices
that "push" a language in a given direction. In any case, the structures pre-
sented in (44) and (45) present just the case of ambiguity that seems to be
THE DIACHRONIC DEVELOPMENT OF SUBJECT CLITICS 321
implied in the reanalysis of a structure. For instance a sentence like (i) can be
interpreted as having the structure (44) or 5):
(i) El vien
He comes
This ambiguity must have been the "bridge" which permitted the reanalysis
from (45) to (44).
13. It is interesting to note that there seems to exist a relation of mutual exclusion
between Nominative Case-assignment through government from Agr and
Nominative Case-assignment through government from T. English, for in-
stance, is a language that does not permit free inversion of the subject. Hence,
following the parameter in (34) it does not select T as a possible Nominative
Case-assigner. Nevertheless, in main interrogative sentences, Agr can assign
Nominative to the subject DP in SpecAgr, as in (i):
(i) What has John done?
Romance languages, on the other hand, select T as possible Case-assigner,
but do not permit Nominative Case assigned by Agr in a sentence like (i):
(ii) *Qui a Jean vu?
Who has John seen?
So, we can observe that a language can exploit a Nominative Case-assign-
ment configuration only once: if the subject gets Nominative from T, it can-
not get it from Agr under the same type of configuration. This could be valid
not only for government, but also for Spec-head agreement.
References
Baker, M. (1988) Incorporation: A Theory of Grammatical Function Chang-
ing. The University of Chicago Press,
Battye, A. (1990) "Quirky Agreement in Genovese." Ms. University of York.
Belletti, A. (1988) "Unaccusatives as Case Assigners." Linguistic Inquiry 19:1-
34.
Belletti, A. (1990) Generalized Verb Movement: Aspects of Verb Syntax. Torino:
Rosenberg and Sellier.
Beninca, P. (1983a) "II clitico a nel dialetto padovano." Scritti linguistici in
onore di Giovan Battista Pellegrini, 25-35. Pisa: Pacini.
Beninca, P. (1983b) "Osservazioni sulla sintassi dei testi di Lio Mazor." In C.
Angelet et al., eds. Langue, Dialecte, Literature. Etudes romanes a la
memoire de Hugo Plomteux, 187-197. Leuven.
Beninca, P. (1985a) "II mutamento sintattico: di un aspetto della sintassi
ladina considerate di origine tedesca." Quaderni Patavini di Linguistica
5:3-15.
Beninca, P. (1985b) "Uso dell'ausiliare e accordo verbale nei dialetti veneti e
friulani." Rivista Italiana di Dialettologia 8:178-194.
322 CLITICS AND VERB SECOND
Pollock, J.-Y. (1989) "Verb Movement, UG, and the Structure of IP." Linguistic
Inquiry 20:365-424.
Renzi, L. (1992) "I pronomi soggetto di due varieta substandard: fiorentino e
francais avanc6." Zeitschrift fur romanische Philologie 108:72-98.
Renzi L. and L. Vanelli (1983) "I pronomi soggetto in alcune varieta romanze."
Scritti linguistici in onore di G.B. Pellegrini, 121-145. Pisa: Pacini.
Rizzi, L. (1982) Issues in Italian Syntax. Foris: Dordrecht.
Rizzi, L. (1986a) "On the Status of Subject Clitics in Romance." In O. Jaeggli
and C. Silva-Corvalan, eds. Studies in Romance Linguistics, 391-419.
Dordrecht: Foris.
Rizzi, L. (1986b) "Null Objects in Italian and the Theory of Pro." Linguistic
Inquiry 17:501-557.
Rizzi, L. (1990) Relativized Minimality. Cambridge, Mass.: MIT Press,
Rizzi, L. (1991) "Residual Verb Second and the Wh Criterion." Technical
Reports in Formal and Computational Linguistics, 2. University of
Geneva.
Rizzi, L. and I. Roberts (1989) "Complex Inversion in French." Probus 1:1-
30.
Roberts, I. (1992) Verbs and Diachronic Syntax. Dordrecht: Kluwer.
Tekavcic, P. (1986) Grammatica storica dell'Italiano. Bologna: II Mulino.
Tomaselli, A. (1989) La sintassi del verbofinito nelle lingue germaniche. PhD
Dissertation, University of Pavia.
Tuttle E. (1983) "L'Oda rusticale di Nicol6 Zotti." Scritti Linguistici in Onore
di G.B. Pellegrini, 431-464. Pisa: Pacini.
Vanelli, L. (1987) "I pronomi soggetto nei dialetti italiani settentrionali dal
Medio Evo ad oggi." Medioevo Romanio XIII173-211.
Vikner, S. (1990) Verb Movement and the Licensing of NP-Positions in the
Germanic Languages. PhD Dissertation, University of Geneva.
Zorzi, L. (1967) Ruzante: Teatro. Torino: Einaudi.
12
Complement Clitics in Medieval
Romance: the Tobler-Mussafia Law*
Paola Beninca
State University of Milan
325
326 CLITICS AND VERB SECOND
As is clear from the figure, I will not take into account possible splittings
of the Infl node, and, as far as CP is concerned, I will only consider Top as a
functional projection higher than CP. It appears from recent work on Romance
languages that both IP and CP are presumably a cover term for a series of
functional projections. For the time being, it does not seem possible to me to
explore and test these assumptions on MRLs. Nor will I be concerned with
the structure under IP, taking for granted that the subject and the main verb,
at a certain point of the derivation, happen to be in the SpecI and I° position
respectively; the main verb then moves to the head of CP in main clauses,
and a constituent moves to the SpecC, producing the typical configuration of
verb-second languages. Top is here considered an unanalyzed projection, the
COMPLEMENT CLITICS IN MEDIEVAL ROMANCE 327
position in which left dislocated and topicalized constituents are found in the
Modern Romance languages.
Top is probably a cover term as well, concealing more functional projec-
tions; it has specialized functions, in particular, it is not able (for principled
reasons?) to have a constituent in its Spec which enters a Quantifier-variable
relation with an argument position in the sentence. I will not pursue this sub-
ject further, and I will call this upper projection Top; I will assume also that,
as in Modern Romance languages, it can be indefinitely iterated.
Using unanalyzed IP I also mean to avoid more precise hypotheses regard-
ing the exact way the verb and its inflection join each other and incorporate
complement clitics: if a complex head results, in main clauses the whole
complex head moves to C.6 The aim of the present paper is to show that this
happens if and only if the Spec of C is filled. If the sequence clitic-verb is to
be seen as an incorporation, in order to get the order verb-clitic we are led to
admit the possibility of excorporating the verb, whatever the process of
encliticization may be. In the present paper I will keep these descriptive and
interpretive problems—which still have a controversial status—out of the scope
of the phenomenon considered.
I shall compare three types of MRLs, exemplified here by:
a) French, limited to the 12th and 13th centuries;
b) Northern Italian dialects (NIDs) of the 12th to the early 14th centu-
ries, represented here by three varieties sharing the relevant features;
c) Old Portuguese and Southern Italian, as instances of pro-drop MRLs.
Old Portuguese will allow us to take into consideration a Modern
Romance language, namely Modern Portuguese, which shares with its
old variant the essential syntactic features with regard to the phenom-
ena we are considering. This will give us the possibility to elicit and
test the relevant cases in a living language.
(3) a. Nouveles vos sei dire del tornoiement de Karahe's (Artu 87, 11)
'News you-(I) can tell of the joust of Karahes'
b. A ceste paroles respont la reine (59, 84)
'to these words responded the queen'
c. Un peu apres eure de prime fu Mador venuz a cort (81, 6)
'Soon after the first hour was Mador come at court'
d. Or voiz tu bien (72, 7)
'Now see you well'
e. Si en est li rois moult a malese (79, 3)
'So of-it is the king very troubled'
f. Non est ce la premiere foiz que vos 1'avez quis (Artu 24, 11)
'Not is this the first time that you him-have looked for'
(= that you have looked for him).
From this comparison some superficial features emerge which may be con-
sidered as diagnostic of the underlying structure. The sentences in (4) show
the so-called "asymmetry of pro-drop," i.e., the inflected verb in main clauses
can have a pro as subject in SpecI position, while in dependent clauses the
verb must have a phonetically realized subject (a pronoun or a Noun Phrase).
This property has been connected to the fact that in main clauses the verb is
normally in IP (see Beninca (1983-84); Vanelli, Renzi and Beninca (1985);
Adams (1987)). (4c) is an example of the same fact with a verb having a non-
argumental subject:
(4) a. Si errerent-tant en tele maniere qu'il vindrent en la praerie de
Wincestre (Artu 16, 66)
'so wandered (3.pl)- so much in such a way that they came in
the prairie of Winchester'
b. Atant en lessent -la parole ester et chevauchent a petites
jornees tant que il vindrent a Kamaalot (Artu 31, 1)
'Then make -the speech stop and ride by easy stages so that
they came to Camelot'
c. Cel jour fist -moult lait tans, car il plut et espart (Berthe au
grand pied, 76-77)
That day was -very nasty weather, because it rained and light-
eninged'.
A second important feature is shown in (5): a direct object in first position
is never accompanied by a resumptive pronoun.
(5) a. Itiex paroles disoit la reine a soi meisme (Artu 32, 30)
'Such words said the queen to herself
b. Mes Lancelot ne connut il mie, car trop estoit enbrons (Artu 11, 13)
'But Lancelot (obj.) not recognized he (subj.), because too much was
sullen'
COMPLEMENT CLITICS IN MEDIEVAL ROMANCE 329
quent in the older texts), though rare, are possible, mainly for reasons that I
will not investigate here, with non-argumental subjects.
The facts are roughly as follows. When a verb in sentence-initial position
is accompanied by a clitic pronoun, the latter cannot take the usual position
before the verb: either the pronoun takes the "full" form (which means that
it is no longer a clitic), as in (7a), or it must follow the verb, becoming enclitic,
as in (7b).
(7) a. Moi semble que.... (Floire et Blanchifleur, 1538)
'Me seems that... (= it seems to me)
b. Sire, ai le ge bien fait (Perceval, 1471)
'have-it I well done' (= I have done it well).
In what follows, I will try to provide a description that connects the enclisis
with a precise syntactic condition.
It could be argued that in the period considered the T-M generalization is
no longer completely valid in French, because it seems to be contradicted by
a number of facts. Occasions for enclisis are uncommon, as I said, because
the verb is rarely in first position. Any constituent can be moved into first
position without any particular requirements; otherwise, filler-words, such as
si 'so' and or 'now', then appear in first position. Thus there are fewer occa-
sions for the verb to be in first position, but this does not imply that the con-
straint against clitics appearing in first position is itself no longer valid. We
can say that it begins to be "obscured" by being avoided with various devices.
Two further cases apparently contradicting the T-M generalization deserve
more careful consideration.
There are, in fact, precise contexts in which clitics are found in first posi-
tion: one case is in yes-no questions, where, at the time we are considering,
clitics are allowed to appear before the inflected verb in first position, as in
(8a,b). This was not the case before the 12th century, when only sentences
like (8c) were possible; this fact is often interpreted in the literature as an
indication that, whatever the reason for the prohibition against clitics in first
position, this prohibition was getting weaker in the 12th century.
(8) a. S"est il donques corrouciez a nos? (Artu 66, 29)
'Himself-is then vexed with us'
b. ...me fetes vos droit de doner a la reine si lonc respit? (68, 30)
'(to) me-give you the right...'
c. ...haez le vos donques si durement? (59,7)
'Hate-him you then so harshly?'
Another case is found in coordinate sentences: a coordinate sentence can
begin with a clitic or a clitic cluster, flying in the face of Tobler and Mussafia
and in fact of any theory considering clitics as being elements that, due to their
very semantics or prosody, cannot begin a sentence.
COMPLEMENT CLITICS IN MEDIEVAL ROMANCE 331
less, they share with French other features which indicate that we are still
dealing with a language with V movement to C and restrictions on the pos-
sible preverbal constituent. In this sense I continue to consider them to be
verb-second languages, though the verb is not constrained to surface in sec-
ond position. The behaviour of these languages, being more richly varied,
gives us the possibility to check some predictions in a wider range of con-
texts.
I shall give examples drawn from three distinct languages of Medieval Italy:
Piedmontese, Venetian and Florentine. With respect to the facts considered,
these languages are to be considered interchangeable, sharing as they do the
relevant features. To begin with, the following sentences show the asymme-
try between main and dependent clauses with regard to pro-drop:7
Piedmontese: Sermones Subalpini (ed. Babilas, 1968. 12th-13th century)
(10) a. Done-li terme per tal convent que, si el al terme non aves paia
quest aver, qu-el serea pendu (sermo VII)
'(he) gave-him due-date...that, if he at the due-date would not
pay..., that he will be hanged'
b. Torne-sen, se ane a 1'autre so ami, si lie ai coita so desasi (VII)
'(he) came back, so (he) went to the other friend, so (he) told
him his disease'
c. Quar eu no savea que tu fuses tal hom com tu eres (II)
'Because I didn't know that you were such a man as you are'.
Venetian of the Lagoon (Lio Mazor, ed. Levi, 1904. early 14th century)
(11) a. Et en questa lo Saracho dis "Pouse, cunpare!", et leva lo rem et
mena-me 50 per lo brago si ch'e/ me lo scavega (3t, 48)
'at that point the Saracho said "Stop, my fellow!," and raised the
sweep and stroke-me on the arm so that he broke it'
b. el dis ch'el me pagarave quando el vorave (2t, 3)
'he said that he would pay me when he wanted'.
Old Florentine (ed. Schiaffini (1954) 13th-14th centuries)
(12) a. La formica € piu savia di te e ongni altro animale, inper6 che ella
raguna la state ond'ella vive di verno (p. 24: 89 v, 5)
'The ant is wiser than you and any other animal because she gath-
ers together in the summer what she lives on in the winter'
b. E cosi ne prov6 de piu cari ch'elli avea (p. 74: 90 r, 30)
'And so tested some of the dearest that he had'
The emphasized pronouns in the examples above would be ungrammatical
in Modern Italian, or they would produce disjoint reference effects. In other
Modern Northern Italian dialects, on the contrary, ungrammaticality would
stem from the absence of subject pronouns.
COMPLEMENT CLITICS IN MEDIEVAL ROMANCE 333
Apart from the asymmetry of pro-drop, the examples above show that, un-
like Old French, Italian varieties do not appear to be verb-second languages:
in main clauses we also find the verb in first, third, fourth, etc. position.
Nevertheless, I argue that the verb is in the head of CP in main clauses, and
therefore that NIDs and Old French share the property of verb movement to
C in main clauses.
I interpret the difference between Old French and NIDs with respect to the
possibility of verb third etc. by supposing that SpecC in NIDs is not the only
position available for preposing a constituent. A preposed constituent can be
in Top as well, and, moreover, as for Modern Romance languages, we assume
a Top position which can be rewritten. This is an even more natural assump-
tion, if we consider Top, as suggested above, as a recursion of Comp with
specialized (reduced) properties.
In Italian dialects, then, a constituent which appears on the surface before
the verb, is not necessarily in SpecC. On the basis of NIDs we can try to fix
a number of properties that distinguish constituents located in SpecC from
other preposed constituents, supposedly in Top. Some items are lexically speci-
fied as "fillers" of SpecC: (co)si is such an item, as it is in Old French. In
main clauses it is strictly preverbal and causes obligatory subject inversion:
(13) et cosi lo mis-e co (Lio Mazor: 17, r 17)
'and so him put-I down'.
The same is true of preverbal objects, which always lack a resumptive pro-
noun:
(14) a. 90 dis-el plusor fiade (Lio Mazor: 1t, 61)
'that said-he many times'
b. una fertra fei lo reis Salomon. (...) Las colones fei d'argent e
1'apoail fei d'or; li degrai per unt hom i montava covrf de purpra
(Serm. Sub., V)
'a sedan-chair made the king Salomon (subj.)... The columns
made (3.sg.) of silver and the support made of gold; the steps by
which one there-climbed, covered (3.sg.) with purple'.
Again, a direct object adjacent to the verb has no resumptive pronoun. Since
the Italian varieties are freer than Old French with regard to the accessibility
of structural positions in the left portion of the sentence, we can observe that
the items that we suppose to occupy SpecC cannot appear together. We can
have more than one preverbal constituent, but we never find, for example, si
and a preverbal object with no resumptive pronoun together. We find instead
a subject followed by si or by an object. Let us say that the subject in this
case is left dislocated in Top.
(15) L'autre ami si est la moillier (Serm., VII)
'The other friend so is the wife'
More evidence in favour of this assumption will be given below.
334 CLITICS AND VERB SECOND
Let us summarize now the generalization that Mussafia drew from Medi-
eval Italian texts with regard to clitic pronouns. In contrast with the French
type, the Italian type, while prohibiting clitics from beginning a sentence, al-
lows clitics to follow a verb in sentence-internal position. In Mussafia's view,
the sequence "verb-clitic" is never ungrammatical, only the sequence "clitic-
verb" is, and precisely in sentence-initial position.
I am going to argue that this is not true. For the moment we will have to
be content with negative evidence and consider the absence of relevant cases
as evidence of ungrammticality; comparison with a living language will sup-
port the assumption.
Compared to French, Italian generally shows more frequent occurrences of
the sequence "verb-clitic." A coordinate sentence is the typical context for
verb-clitic in Italian, but not in French (I do not take into consideration Old
Provengal, which behaves like Italian in this respect): sentences (16) contrasted
with sentences (17) illustrate this difference.
(16) a. e si la lave e forbf e retorne"-la en sen loc (Serm., X)
'and so her-washed (3.sg.) and wiped and put-her again in its
place'
b. et he li tras la fosina de man et branchai-lo per li caveli (4r, 36)
'and I him-tore the harpoon from his hand and caught-him by the
hair'
c. Allora quelli il si mise in casa (...) e cominciollo a confortare
(Schiaffini (1954:76))
'Then he him-put in his house and began-him to console*
(17) a. Mes Nouvele (...) ala (...) a la Joieuse Garde et I'i porta uns
valets (Artu 105, 15)
'But News went to the Joieuse Garde and it-there-brought a
valet (subj.)'
b. Je m' en irai arriere a mon signour et li conterai (Artu 110, 57)
'I will go after my lord and him-tell'.
We could conclude from this feature that the constraint against first posi-
tion is stronger in Italian than in French. But if we look at yes-no questions,
the conclusion is the opposite: in this context, Italian varieties very consis-
tently have the clitics in first position:
(18) a. Me voj tu dar la taverna? (Lio Mazor: 8t, 28)
'Me-want you give the tavern? (= will you give me the tavern?)
b. Se vastarave lo pes... ? (Lio Mazor: 20t, 46)
'Self-would taint the fish?' (= will the fish go bad?)
I said that in French the clitic in first position in yes-no questions becomes
more frequent in relatively later texts, and this is interpreted as evidence of
the weakening of the constraint against clitics in first position. From this point
of view, the Italian pattern, with clitics in first position in yes-no questions
COMPLEMENT CLITICS IN MEDIEVAL ROMANCE 335
and clitics after the verb in coordinate sentences, is inconsistent, the former
indicating a strengthening, the latter a weakening of the constraint. It seems,
then, necessary to interpret these facts as only indirectly having to do with
the constraint itself; they depend instead, I would suggest, on the different
structural descriptions assigned to these types of sentence. What is interpreted
as "first position" (i.e., empty SpecC) in Old Italian, is not "seen" the same
way in Old French syntax, but the constraint remains unaffected.
Another difference, first noticed by Mussafia himself, is the fact that, as I
mentioned before, in Italian varieties the enclisis of the clitic, while being
obligatory in first position, is widely found in sentence-internal position as
well. Some work has been devoted defining the type of contexts prohibiting,
or else allowing, enclisis.8
As Schiaffini (1954) already pointed out in his comments on the collection
of 13th-century Florentine texts he edited, we never have enclisis in a depen-
dent clause introduced by a complementizer or a wh-pronoun, even in the case
where the verb is clause-initial. This syntactic type is illustrated by the fol-
lowing examples, which are good evidence against any proposal to consider
enclisis as a process connected with the "given" character of the clitics in
contrast with the verb. This property is expected not to change—and so to be
relevant—in a complement clause: but this is not the case, since we never
have enclisis in a clause introduced by a complementizer, in any Romance
variety of the Middle Ages:
(19) a. (...) che vi si po a riporre, (...) che vi si possa sedere
(Schiaffini (1954:68))
'that there-one-can put..., that there-one-can sit...'
b. (...) queli che ve dis ste parole (Lio Mazor: Levi 1904, 1 r, 21)
'those that to-you-tell these words'.
The fact that enclisis is never found in a sentence introduced by a
complementizer (which is a designated occupier of the head of C), suggests
that verb movement to C feeds the enclisis.
We can add that the occupiers of SpecC that we isolated before (si, an in
Venetian, the direct object) also block enclisis; we never have enclitic pro-
nouns if the verb is preceded by these elements: in (20a) we have preposed
direct objects (quello che tu vorrai, gli altri), in (20b) an, in (20c) si.
(20) a. Quello che tu .vorrai mi renderai e gli altri ti terrai (Florentine:
Novellino, IX)
'What you want to-me-will give (2.sq) back and the others for-
you-will keep' (= you will give back to me what you want and
will keep the others)
b. An lo dies-tu ben! (Lio Mazor: a3 t, 68)
'Indeed it-said you well' (= indeed, you said it!)
c. si la lave...e retornela... (cfr. 16a)
'so her-washed and put-her'.
336 CLITICS AND VERB SECOND
If the grammar considers the Spec to be empty, we have the situation found
in Old French, where there is enclisis. If the Spec is considered not to be
empty, since it contains an abstract operator, we have the situation found in
Italian varieties, i.e. we always have proclisis: the Spec being filled, the clitics
appear in sentence-initial position.
In this interpretation, the different behaviour of Italian and French in coor-
dinate sentences depends on the level where coordination occurs: if the at-
tachment is at CP level, we will have an empty Spec, which will trigger
enclisis,11 this is the situation of Italian varieties. If the coordination is at the
level of C', we do not have an empty Spec in the second clause; this is the
case of French.12
As I said above, it is possible to make some simple and straightforward
predictions stemming from the generalization that when we have an empty
SpecC in a clause, we have enclisis of clitic pronouns, while when we have a
constituent that must be in the Spec of CP, we never expect to have enclisis.
Furthermore, verb movement is to be considered as feeding the further move-
ment of the verb. Then if verb movement to C is blocked, we predict that
"verb-clitic" order does not occur.
As expected, we never have "verb-clitic" in relative clauses or in wh-ques-
tions. In relative clauses we do not have an accessible head of CP for the
verb to move into, in a wh-question there is no empty SpecC, as the wh-word
is obligatorily there.
I also assumed that a direct object in first position with no resumptive pro-
noun is obligatorily in SpecC. If the verb has other clitics corresponding to
different arguments, these are always expected to be proclitics. This is also
confirmed by the data and will be checked later in slightly different languages,
Old Portuguese and Old Southern Italian, which are pro-drop languages.
If, on the other hand, we could have a constituent in first position, which
cannot be in the SpecC but has to be in Top, we could build up a complemen-
tary argument.
In my theory, a preposed object with a resumptive pronoun has to be in Top,
even if in the surface it appears immediately adjacent to the verb. If between
a preposed object and the governing verb there is no intervening constituent
and the object has a resumptive clitic, this clitic will always be enclitic. For
some reason, a "heavy" constituent—in particular a NP containing a relative
clause—is likely to be left dislocated (i.e. in Top), as shown in Vanelli (1986).
So we have in the following sentence (from Schiaffini (1954:282, n.9) an
appropriate example, with two preverbal constituents that are both in Top:
(23) [A voi], [le mie poche parole ch'avete intese], olle dette con grande
fede.
'To you, my few words that you have heard, (I) have-them said with
great faith'.
A Hanging Topic could also be a constituent that has to be in Top, as far as
we know. If there is no other constituent between the Hanging Topic and the
verb, I expect a clitic—if there is one—to always be enclitic, as is in fact the
338 CLITICS AND VERB SECOND
case. See the following example from Old Florentine: we can argue that the
first constituent is a Hanging Topic from the fact that it does not have the
preposition a required by the subc tegorizing verb (see Cinque (1983)).
(24) [Quelgli il quale ndasse per Firenze...in die di lavorare], debbialgli
essere soddisfatto... (Schiaffini (1954:54, 1.13)
'He who should go through Florence...in a working day, be-to him
paid compensation...'
In the following sentence (25) (from Schiaffini (1954: 46, 11.32-33)) we
have apparently a counter example: the first constituent is again a Hanging
Topic—it lacks the preposition a—and we would expect enclisis of gli:
(25) Et chi facesse contra la prima volta, gli sia imposta penitenca, et la
seconda sia cacciato.
'And he who should act against the first time, to-him be imposed a
penance, and the second be pulled away'.
But we can easily accommodate this sentence both for the theory and for
the sense itself; as indicated in (25') with a different punctuation, the Hang-
ing Topic is only chi facesse contra, while la prima volta is a separate con-
stituent, which we can suppose to be in SpecC:
(25') Et chi facesse contra, la prima volta gli sia imposta penitenca, et la
seconda sia cacciato.
'And he who should act against, the first time to-him be imposed a
penance, and the second be pulled away'.
Notice that the emendation permits a more natural contrast between the—
now isolated and symmetrical—constituents la prima volta...la seconda volta.
a language which does not show second position of the verb in main clauses,
nor the asymmetry of pro-drop, can nevertheless be a verb-second language.
The broad type, then, apart from pro-drop asymmetry which is independently
related to verb second, differs from the strict type only owing to the avail-
ability of the Top position.
In a dependent clause introduced by a wh-pronoun or a complementizer we
always have the order clitic-verb, as in the following Old Portuguese examples
(from Huber (1933:160 ff.)):
(26) a. As tuas doces palavras per que me prometias
'your sweet words with which you promised me'
b. O rato respondeo que Ihe prazia
'The mouse answered that to-him-pleased'.
The behaviour of a preposed direct object is exemplified in (27-29):
(27) a. Sao ko kelle terre, per kelle fini que ki contene, trenta anni le
possette parte Santi Benedicti (the oldest Italian text, 960, from
Capua)
'I know that those lands, in those boundaries that here contains,
thirty years them-owned the party of St. Benedict'
(28) a. La salamandra audivi... (Jacopo da Lentini, Sicily)
'The salamander (I) heard...'
b. La mia gran pena e lo gravoso affanno..., madonna lo m'ha in
gioia ritornato (Guido delle Colonne, Sicily)
'My great sorrow and grievous pain...my Lady it to-me has into
joy turned'
(29) a. Tal service Ihe pode fazer hun homen pequenho (OPortug:
Huber)
'Such service to-him can do a little man (subj.)'
b. O trigo que eu como, guanco-0 per meu trabalho
'The wheat that I eat, (I) earn-if with my work'.
Example (29a) has a preposed object with no resumptive pronoun and
proclisis of the indirect object pronoun; (29b) has a preposed direct object with
a resumptive pronoun: the clitic is enclitic. This pattern is consistently ob-
served in Old Portuguese.13
I conclude that the preposed object is in SpecCP when it has no resumptive
pronoun. Sentences like (29a,b) are evidence in favour of this hypothesis. If
we have a direct object in SpecC, a resumptive pronoun is not expected. If
instead we have a clitic corresponding to a preposed object, the object has to
be in Top and not in SpecC. If SpecC is empty, this will trigger enclisis of the
clitic (as in 29b). Another symmetrical prediction is the following: we will
never have enclisis of a clitic if the preposed direct object has no resumptive
pronoun. The object has to be in the Spec of C in this case and thus no fur-
340 CLITICS AND VERB SECOND
ther movement of the verb is possible. These predictions are completely con-
firmed by the available data.
But including pro-drop Medieval Romance in the discussion, and in par-
ticular Old Portuguese, has a useful consequence: in fact, with respect to these
patterns, Old Portuguese differs very little from Modern Portuguese. We can
test our prediction then in a wider range of contexts, and, what is more im-
portant, we can elicit judgments about ungrammatical sentences.
In Modern Portuguese (of Portugal) the subject is regularly in Top, as we
detect from the fact that clitics normally follow the verb when there is a
preverbal subject (in Old Portuguese, this was very common but not obliga-
tory).14
(30) a. O Joao disse-nos
b. *O Joao nos disse
'John told us'.
But some subjects are never followed by a verb with enclitic pronouns. I
would predict that, if my hypothesis is correct, the subjects that block enclisis
are the subjects that must be in SpecC, as the following contrasts show:
(31) a. ...quern me chamou?
'Who called me'
b. *...quern chemou-me?
(32) a. Ninguem nos viu /*viu-nos
'Nobody saw us'
b. Todos se lembran /*lembran-se
'Everybody remembers'.
Quantified subjects are considered to occupy SpecC obligatorily. Thus, the
verb must be in the head of CP and enclisis of the clitics is impossible.
The same happens with the negation:
(33) a. Nao os comprendo
b. *Nao comprendo-os
'(I) them understand'.
There is here a difference with medieval varieties, where the negation could
either satisfy the verb-second requirement, i.e. occupy SpecC, or be proclitic
of the verb (see Vanelli, Renzi and Beninca (1985)).
With the support of the proposed tests the Tobler-Mussafia law can be re-
stated as follows:
In Medieval Romance (and in Modern Portuguese) complement clitics
occur after an inflected verb if and only if the governing verb is in C and
the Spec of CP is empty.
COMPLEMENT CLITICS IN MEDIEVAL ROMANCE 341
Notes
* Previous versions of this paper have been presented at the Linguistics semi-
nar in Venice in 1989 and at the First Generative Diachronic Syntax Confer-
ence in York, April 1990. I am greatly indebted to Giulio Lepschy and Paolo
Salvi for discussing both the form and content of this paper.
I also owe a debt of gratitude to Adrian Battye: he made precise observa-
tions on the analysis presented in first version of this paper; his organization
of the York conference gave that meeting an especially pleasant atmosphere.
To a deeply missed friend and colleague, I dedicate these pages.
1. There is evidence that the relation between a verb and an enclitic pronoun in
Romance involves a process in some sense going "deeper" into morphology
than is the case with proclitics. For some evidence of this close relation be-
tween an enclitic pronoun and a verb, see Beninca and Cinque (1993).
Mussafia began his note of 1886 with these introductory lines: "I pronomi
personali oblique atoni mi, ti, si, ecc., e le particelle pronominali atone ci, vi,
ne o precedono il verbo di forma finita (proclisi) o gli tengono dietro, formando
con esso una parola sola (enclisi)." ("The oblique atonic personal pronouns
me, ti, si, etc., and the atonic pronominal particles mi, vi, ne, either precede
the finite verb [proclisis] or follow it forming a single word with it [enclisis]"—
editors.) It seems to me that we still have no means to account for intuitions
of this kind.
2. See Tobler (1875), Mussafia (1886).
3. In infinitival and gerundive clauses probably other factors complicate the pic-
ture, but it seems that the main lines of the description I will provide hold in
the same way. This means that clitics were not constrained always to be
proclitics with these verb moods, as in Modern French for example, nor al-
ways to be enclitics, as in Modern Italian (excepted some cases, in fact), but
again the position of complement clitics relative to gerund and infinitive de-
pended on a broader syntactic context. A similar, though more restricted,
possibility is still found in Modern Neapolitan and in other Italian dialects.
4. For example, Mussafia (1886), Meyer-Lubke (1897), Foulet (1924), Adams
(1987), Cardinaletti and Roberts (1991). Marcantonio (1976,1980), Antinucci
and Marcantonio (1980) proposed an analysis connecting enclisis to syntax,
into a generative semantics framework.
5. See Benacchio and Renzi (1987), Renzi (1992).
6. In Old Spanish, as in Old Portuguese and Galego-Portuguese, a different pos-
sibility is met, with complement clitics attached to the complementizer or a
Wh-element in dependent clauses, and separated from the verb by the subject
and/or a restricted set of adverbs: the following example is from Ramsden
(1963:137):
(i) quando la el rey dixo
'when-it the king said'.
See also Salvi (1990, 1991).
342 CLITICS AND VERB SECOND
References
Adams, M. (1987) Old French, Null Subjects and Verb Second Phenomena.
PhD Dissertation, UCLA.
Adams, M. (1988) "Parametric Change: Empty Subjects in Old French." In
D. Birdsong and J.P. Montreuil, eds. Advances in Romance Linguistics,
1-16. Dordrecht: Foris.
Antinucci, F. and A. Marcantonio (1980) "I meccanismi del mutamento
diacronico: il cambiamento d'ordine dei clitici in italiano." Rivista di
Grammatica Generativa 5:3-50.
Babilas, W., ed. (1968) Untersuchungen zu den Sermones Sabalpini. Miinchen:
Hiiber.
Benacchio, R. and L. Renzi (1987) Clitici slavi e ramanzi. Padova: Clesp.
Beninca, P. (1983-84) "Un'ipotesi sulla sintassi delle lingue romanze
medievali." Quaderni Patamini di Linguistica 4:3-19.
Beninca, P. and G. Cinque (1993) "Su alcune differenze fra enclisi e proclisi."
In Omaggio a Gianfranco Folena. Padova: Programma.
Cardinaletti, A., and I. Roberts (1991) "Clause Structure and X-second." In
W. Chao and G. Horrocks, eds. Levels, Principles and Processes: the
Structure of Grammatical Representation. Berlin: Foris/de Gruyter.
Cinque, G. (1983) "Topic Constructions in Some European Language and
Connectedness." In I. Ehlich and H. van Riemskijk, eds. Connectedness
in Sentence, Discourse and Text. Tilburg.
Foulet, L. (1924) "L, accent tonique et 1'ordre des mots." Romania 50:54-93.
Huber, J. (1933) Altportugiesisches Elementarbuch. Heidelberg.
Kayne, R. (1972) "Subject Inversion in French Interrogatives." In J.
Casagrande and B. Saciuk, eds. Generative Studies in the Romance
Languages, 70-126. Rowley, Mass.: Newbury House.
Kayne, R. (1990) "Romance Clitics and PRO." Ms. CUNY (Published as
Kayne, R. "Romance Clitics, Verb Movement and PRO." Linguistic
Inquiry (1991) 22:647-686.)
Kayne, R. and J.-Y. Pollock (1978) "Stylistic Inversion, Successive Cyclicity
and Move NP in French." Linguistic Inquiry 9:595-621.
Lema, J. and M.L. Rivero (1990) "Types of Verbal Movement in Old Span-
ish: Modal, Futures and Perfects." Ms. University of Ottawa.
Levi, U. (1904) / Monumenti del dialetto di Lio Mazor. Venezia: Visentini.
Marcantonio, A. (1976) "Un aspetto dell'ordine delle parole nell'italiano del
Due-Trecento." Rivista di Grammatica Generativa 1,2:57-77.
Marcantonio, A. (1980) "Alcune considerazioni sulla legge Tobler-Mussafia."
In P. Gerrettoni, eds. Problemi di analisi linguistica. Perugia.
Melander, J. (1929) "L'origine de 1'italien me ne, me lo, te la, etc." Studia
Neophilologica 2:169-203.
Meyer-Lubke, W. (1897) "Zur Stellung der tonlosen Objekts-pronomina."
Zeitschrift fur Romanishe Philologie 21:313-334.
Mussafia, A. (1886) "Una particolaritd sintattica della lingua italiana dei primi
secoli." In Miscellanea di filologia e linguistica, dedicata alla memoria
344 CLITICS AND VERB SECOND
This paper is organized into five sections. In the first section, I introduce and
discuss the particular construction which this investigation will concentrate
on: the cases of verb-third order in matrix clauses in Old High German (OHG
henceforth). In SECTION 2 OHG data will be compared with Old English (OE)
data (cf. Kemenade (1987)). The third section is devoted to the presentation
and discussion of two alternative analyses of these verb-third word order pat-
terns already proposed in the literature: that in Lenerz (1985), and that in
Kemenade (1987). Finally, in the last two sections, I will propose a partially
new analysis, which will lead us to discuss two important issues: the clitic-
head relation and the w/t-construction.
patterns. Under this label one could in fact subsume all the cases in which
the finite verb occupies a position between the second and the final position
of the sentence: verb-third, verb-fourth, etc.. Inside this "basket" of excep-
tions to verb second our attention will concentrate on what certainly seems
more interesting with respect to the general problem concerning the "degree
of realization" of the constraint in OHG syntax: the cases of verb third.
First of all it should be stressed that even an apparently simple label like
"verb-third" does not unambiguously refer to a unique word order pattern. It
is well known that OHG prose consists of translations from Latin. This fact
helps us to differentiate between two different types of verb third. Let us first
consider the following examples:
(1) Isidors Schrlft contra ludaeos (8th/9th century)
a. dhaz ir chichundida
Obj Subj Vfnt
that he showed
'He showed that' Braune and Ebbinghaus (1979:19,135)
On the other hand, examples (la-c), taken from the OHG translation of De
fide catholica contra Judaeos by Isidorus Hispalensis, exemplify a different
type of construction which could be considered peculiar to Germanic syntax:
the verb-second constraint is violated by the presence of the subject pronoun
which intervenes between the fronted constituent and the finite verb.2
It is clear that examples (la) and (Ib) could be considered cases of verb-
third as well as cases of verb last (and, in fact, a verb-last analysis has been
proposed by Lenerz (1985)). However, I prefer to consider these examples
as cases of verb third given that they clearly exemplify the same kind of con-
struction together with example (1c), where the third position of the finite verb
does not correspond to the final position of the sentence.3
What has been said till now is well known to people familiar with Germanic
philology. There is, in fact, general agreement among Germanists that the fol-
lowing properties hold for OHG:
a) OHG prose was characterized by the verb-second constraint
(subject-verb inversion in the main declarative clause);
b) the finite verb may shift to third position given the presence of a pro-
nominal element in second position (cf., among others Lippert
(1974:15)).
In what follows I will concentrate on exactly this kind of construction.
a) this construction is well attested in Old English prose (on the con-
trary, in OHG the construction under consideration is attested only
in the translation of Isidor and partially in the Monsee-Wiener
Fragmenten);
b) while in OHG the pronoun which intervenes between the fronted
element and the finite verb generally corresponds to the subject pro-
noun, in OE it can also be the object of the verb or the object of a
preposition.
Note that this difference between OE and OHG seems, in fact, to confirm the
traditional hypothesis that OHG was a more consistently verb-second language
than Old English (cf. Fourquet (1938)).
Despite these differences, Old English and OHG syntax show two impor-
tant similarities: both allow the relative order pronoun-Vfnt (cf. SECTION 2.1);
both have the same range of positions of the finite verb in the subordinate
clause (cf. SECTION 2.2).
w
ne - Vfnt+pronoun - / * WH/ne/ha - pronoun+Vfnt
Subordinate Clause:
c. Comp.+pronoun - . . . . . / * pronoun+Comp. .....
Note that, as far as the main clause is concerned, the pronoun precedes the
finite verb only when the fronted element corresponds to a non-negative de-
clarative constituent (XP) (pattern (6a)— cf. (la-c), (3), (4) and (5)). On the
contrary, when the first constituent of the main clause corresponds either to a
wh-constituent or to the negative clitic or to the adverb ba 'then' the pronoun
follows the finite verb in third position. This word order pattern (cf. (6b))
will be analyzed later (cf. SECTION 5) where the syntactic nature of the adverb
ba will also be discussed.
In the subordinate clause (cf. (6c)), the pronoun immediately follows the
lexical complementizer:
Old English (examples from Kemenade (1987:59)):
CASES OF VERB THIRD IN OLD HIGH GERMAN 349
Old English + + + + +
OHG
Isidor + + + + +
Muspilli + + (+) -
Williram + + + +
Memento Mori + •f - +
West Flemish + + + +
Zurich Swiss German + + + +
German + + _ +
Dutch + + - +
This table shows that there is a correlation between the position of the Vfnt in
the main clause and the position of the Vfnt in the embedded clase. The Comp
NPsubj Vfnt NPobj V order is found in the subordinate clause whenever the XP-
pronoun-Vfnt order occurs in the main clause. Prima facie, the hypothesis that
these two word order patterns should both be related to the medial position of
I° seems to be confirmed. Before exploring this hypothesis (cf. SECTION 4),
some points should be stressed concerning (26):
1) Diachronically, the order XP-pronoun-Vfnt (...) is lost before VPR in
the subordinate clause.
354 CLITICS AND VERB SECOND
2) The word order pattern (lle) (i.e. Comp NP(Subj) Vfnt NP(Obj) V) could
refer to two different syntactic constructions (cf. note 5). This fact
obviously plays an important role in establishing the correlation un-
der discussion, which does not seem to hold the other way round in
Williram (OHG) and more obviously in two Modern Germanic lan-
guages like West Flemish and Zurich Swiss German. In these lan-
guages, it is clear that: i) the possibility of having instances of bracket
structure in the subordinate clause does not correlate with verb-third
cases in the main clause order pattern Comp NP(Subj) Vfnt NP(Obj) V
must be attributed to VPR (see, among others, Haegeman and
Riemsdijk 1986); iii) the presence of VPR is strictly dependent on
the presence of "simple" VR (den Besten (1986)).7 On the other
hand, in OHG and OE, at least some instances of bracket structure
in the subordinate clause could be analyzed, in principle, as the re-
sult of V movement to a medial I0.8 Here in fact the correlation be-
tween the two word order patterns under consideration (verb-third
cases in the main declarative clause and bracket structure in the sub-
ordinate clause) seems to hold in both directions. The precise na-
ture of this correlation will be explored in SECTION 4.
The preposed verb in (52b) [= (28b)] may then have been reanalysed as S-
initial Infl in base structure. Since there were also structures with Comp in
OHG, S-initial Infl became identified with Comp (Confl).
In other words, according to Lenerz (who bases his analysis of verb second
on Platzack (1983)), the rise of verb second in OHG is crucially dependent
on two different processes of reanalysis:
356 CLITICS AND VERB SECOND
b.
- Vfnt+pronoun-....../* WH/ne/pa-pronoun+Vfnt
Subordinate Clause:
c. Comp+pronoun - / * pronoun+Comp
Kemenade assumes that pronominal elements in OE could have the status of
"syntactic" clitics. In these terms, the word order pattern XP-pronoun-Vfnt-
(...) (cf. (31a)) is attributed to a process of cliticization on the left of C°. More
precisely, the S-structure representation of the word order pattern (3la) is as
in (32) (where C°= Confl):
ond only holds for the history of English (cf. Kemenade (1987) and
Hulk and Kemenade (this volume)). Both phenomena could in fact
be reasonably traced back to the general process of deflexion which
this language underwent. On the other hand, however, in the his-
tory of German the loss of cliticization corresponds to the strength-
ening of verb second. This can be explained if we assume that the
loss of cliticization, in this case, is not to be linked to a change of
the morpho-syntactic characterization of I° but rather to the differ-
ent position occupied by I° at D-structure (cf. (36)).
After this excursus through some of the most interesting consequences of
the hypothesis that both OE and (what may be more problematical) OHG were
characterized by a head-medial IP, let us briefly consider some problems which
this hypothesis involves.
Why can the (clitic) pronoun not intervene between the element in the speci-
fier of CP ([Spec, CP]) and the Vfnt in C°? The answer provided by Kemenade
(1987) consists, as we have already noted above (cf. SECTION 4), of two inter-
related assumptions:
i) the wh-element, the clitic of negation (ne) and pa are operators;
ii) the relation between an operator in [Spec, CP] and the Vfnt in C° is
such that nothing can interrupt it.14
The following objection can be raised regarding the first of these assumptions:
while we can certainly assume that both wh-elements and negation are opera-
tors, the syntactic status of the adverb pa is uncertain. Stockwell (1977), in
fact, assumes (on the basis of unpublished work by W. Rybarkiewicz; cf.
Stockwell (1977:311, n.2)) that transitional adverbs like pa, ponne, paer, are
to be considered conjunction elements together with the coordinating conjunc-
tion. Following this hypothesis, sentences introduced by pa should be more
appropriately analyzed as verb-first sentences, like yes/no questions (cf. also
Fourquet (1938)).
Regarding the assumption in (ii), the idea that the relation between the wh-
element and the Vfnt is "special" in a certain way is immediately captured by
the wh-criterion recently proposed by Rizzi (1990a). Following, in essence,
May (1985) and updating his proposal in terms of Chomsky's (1986) theory
of clausal projections, Rizzi assumes that the occurrence and position of wh-
elements is determined by the following principles (cf. Rizzi (1990a:378):
(38) Wh-Criterion:
Principle A: Each [+wh] X° must be in a Specifier-Head relation
with a wh-phrase
Principle B: Each wh-phrase must be in a Specifier-Head relation
with a [+wh] X°
One of the main purposes which underlies the formulation of the wh-criterion
consists in the possibility of accounting for what Rizzi calls "residual verb
second." This label refers to the constructions which involve the mechanics
of verb second (i.e. V° to I° to C°) in non-verb-second languages, in particu-
lar subject-aux inversion (SAI) in English and subject-clitic inversion in French
(SCI). These two syntactic phenomena are exemplified in (39) and (40)
below:
362 CLITICS AND VERB SECOND
(39) SAI
a. Who i didj [Mary tj see ti ]?
b. *WhOi [Mary I° saw ti ]?
(40) SCI
a. Que manges-tu?
b. *Que tu manges?
Note that, while the adjacency requirement between the wh-word and the Vfnt
is accounted for in terms of V° (to I°) to C° movement,15 on the other hand,
the explanation for why the Vfnt must move to C° in a non-verb-second lan-
guage crucially relies on the wh-criterion.
In fact, if we assume (following Rizzi (1990a:378-379)) that the feature
[+wh] may be generated:
i) in C° in the subordinate clause (through selection by the matrix verb),
ii) in I° in the main clause,16
then it follows that I° (i.e. the Vfnt) must move to C in the main wh-clanse in
order to satisfy principle B of the wh-criterion.
The hypothesis that the movement of the Vtnt to C° in the main wh-clause
is forced by the wh-criterion receives independent evidence from the diachronic
perspective. If we consider the historical evolution of verb second, in fact,
we do not find any violation of the verb-second constraint in wh-constructions.
First of all it is well known that in OHG one possible exception to the verb-
second constraint in the main declarative clause consisted in the (strongly)
limited occurrence of verb-last constructions.17 It conies as no surprise that
no verb-last construction is attested in the main interrogative clause in OHG.
Secondly, it is clear that the distributional facts presented by Kemenade (1987)
with respect to the relative order pronoun-Vfnt in OE (here extended to OHG
data) go exactly in this direction. The fact that the verb-second constraint
seems to be "stronger" in the wh-construction should be, in fact, simply at-
tributed to the wh-criterion, independently from whatever explanation one
could provide for "full verb second."
Turning now to our original problem, if what has been said so far proves
to be reasonable, then in order to explain the word order pattern in (37) we
have simply to assume the following:
(41) The complex head which is derived from a process of cliticization
(= [p clitic [r Vfnt]]) is unable to satisfy principle B of the wh-crite-
rion.
At this point it is important to note that:
a) The wh-criterion together with (41) cover not only the OE (and OHG)
data under discussion (cf. (37)) but can be immediately extended to
the phenomenon of subject-clitic inversion in French (cf. (40));
b) An important distinction must be drawn between subject clitics and
object clitics. In fact, while the generalization captured in (35) does
CASES OF VERB THIRD IN OLD HIGH GERMAN 363
not imply any asymmetry between subject clitics and object clitics
(as it seems to be the case in OE), in OHG and, more obviously, in
modern French, the same generalization holds only for subject clitics.
Compare (40) with the following two French examples:
(42) Qui 1'a mangee?
(Who it-has eaten = Who ate it?)
(43) Quand 1'as-tu mangee?
(When it-have-you eaten = When did you eat it?)
As we can see, while the subject clitic must occur to the right of the
finite verb (cf. tu in (43) and (40)), the object clitic (cf. /' in (42)
and (43)) regularly occurs to the left of it.18
c) the constraint (41) crucially implies that the process of (subject-)
cliticization modifies the status of the head on which it applies.
From this last assumption at least two important related issues arise:
I) How does cliticization modify the status of the head?
II) Which principle underlies the constraint proposed in (41)?
Concerning II), note that cliticization has been generally analyzed as a pro-
cess of adjunction of a head (the clitic itself) to another head (I° or, more
controversially, C°).19 Given this assumption, the interference caused by the
clitic could be reduced to the notion of intervention as stated, for example, in
Rizzi (1990b). In fact, the clitic adjoined to C° (either directly as proposed
by Kemenade (1987), or indirectly through I° to C° movement as proposed
here) interferes in the relation between the wh-element in [Spec, CP] and the
Vfnt in C° in a way which is similar to the interference caused by a preposi-
tion intervening between a verb and its complements in a Case-assignment
relation. This situation is illustrated in the following schema:
Given this distinction, it is clear that only cliticization through adjunction (cf.
(45)) would be pertinent as far as the constraint (41) is concerned. On the
contrary, the complex head resulting after cliticization to an X internal level
(cf. (46)), should not cause any violation of principle B of the wh-criterion.
Note that this distinction provides us with a principled explanation for the
different (morpho-)syntactic behaviour between:
i) cliticization phenomena in Romance languages versus cliticization
phenomena in Germanic languages;
ii) cliticization of the subject pronoun versus cliticization of object pro-
nouns.
A deeper investigation of the consequences of such speculative assumptions
goes further beyond the purposes of the present paper and is therefore left to
future work.
Notes
* This paper is a revised and expanded version of a talk presented at the First
Generative Diachronic Syntax Conference in York (April 1990). A prelimi-
nary version of this work was submitted to the University of Geneva for the
attainment of the Certificat de Specialisation en Linguistique (Theorie de la
syntaxe et syntaxe comparative), academic year 1989/90, and appeared in
GAGL 33 (1991). In addition to instructive discussion from the participants
at the conference in York (particularly from Ans van Kemenade, Tony Kroch,
Cecilia Poletto, Ian Roberts, Beatrice Santorini, Sten Vikner and Fred
Weerman), comments and other help came from the following people: Werner
Abraham, Harald Clahsen, Denis Delfitto, Giorgio Graffi, Maria Teresa Guasti,
Lidia Lonzi, Andrea Moro, Luigi Rizzi and Raffaella Zanuttini. Of course,
responsibility for errors is just mine.
CASES OF VERB THIRD IN OLD HIGH GERMAN 365
This work was partially financed by the research project 40% "Variazione e
conservazione delle strutture sintattiche," headed by Prof. Riccardo Ambrosisi
(Universita di Pisa) and Prof. Giorgio Graffi (local coordinator, Universita di
Pavia).
1. For the relevance of the influence of Latin syntax in Tatian cf. the detailed
study by Lippert (1974).
2. Note that in (la-c) the fronted constituent always corresponds to the object
NP. This is due to chance; other XPs can appear here.
3. As Beatrice Santorini (p.c.) pointed out to me, (lc) is not the clearest example
one could provide in order to differentiate this particular kind of verb-third
construction from verb last, which was also possible in OHG. Since one has
to assume, independently from the facts under consideration, that OHG syn-
tax is characterized by a process of extraposition, then (1c) could be ana-
lyzed, in principle, as one of the possible variants which could be derived
exactly from a verb-last construction through extraposition. A sentence with
a complex verbal form (auxiliary/modal - past participle/infinitive) or a par
ticle verb (e.g. liebhaben) would certainly provide more direct and convinc-
ing evidence in order to draw a precise distinction between (la-c) and the
verb-last construction.
4. The so-called Monsee-Wiener Fragmenten present an ambiguous situation
with respect to the position of the subject pronoun in the main declarative
clause. If, on one hand, the subject pronoun generally follows the Vfnt as in
Williram, on the other hand, some cases of verb third (XP-pronoun-Vfnt) are
also attested.
5. As Ans van Kemenade (p.c.) pointed out to me, the word order pattern (9b),
here repeated for the sake of simplicity:
a) CompNP (subj) V fnt NP (0bj) V
could ambiguously refer to two rather different syntactic constructions, which
should be more precisely characterized by two distinct word order patterns:
b) CompNP(Subj)XPVfntNP(0bj)V
c) Comp NP(Subj) *(XP) Vfnt NP(0bj) V
Note that only the word order pattern in b), where a maximal projection XP
intervenes between the NP(subj) and the Vfnt, can be unambiguously analyzed
as an instance of VPR. On the other hand, the word order pattern in c) could
be much more adequately analyzed as the result of V° movement to I° given
the hypothesis of a deep structure word order of the type:
d) [ I p NP ( S u b j ) [I°][ V P NP ( 0 b j ) V°]]
where IP is a head-medial maximal projection. It is precisely this idea that
we are going to explore in SECTION 4.
6. For a detailed investigation of the syntax of verbs in OHG, cf. Tomaseli
(1989, ch. 2).
7. Muspilli shows a quite interesting situation. In fact we find an instance of
VPR (just one example amongst 65 subordinate clauses) in a dialect which
366 CLITICS AND VERB SECOND
does not have simple VR. This is particularly interesting as far as the nature
of VPR in OHG is concerned (cf. note 8). On the other hand, in Modern West
Germanic languages the presence of VPR is strictly dependent on the pres-
ence of "simple" VR (cf. den Besten (1986)).
8. This hypothesis receives independent evidence from recent research by Ans
van Kemenade on Old English (cf. Kemenade (1990)).
9. This ungrammatical word order pattern is, in fact, attested in at least one
OHG text. The following example, taken from Muspilli, is the only example
of complementizer-Vfnt adjacency usually cited in the historical grammars of
German (cf. Erdmann (1985); Jolivet and Mosse" (1972)):
a) daz sculi der antichristo mit Eliase pagan
CompVfnt NP(subj) PP V
that must the antichrist with Elias fight
Note that in the relative clause the possible adjacency of the relative pronoun
and the Vfnt has a clearly different syntactic value. The following examples
are taken from the Monsee-Wiener Fragmenten (M) and from Tatian (T):
b) (M): der framtregit fona siemo horte niuuui ioh firni
REL. Vfnt PP NP(0bj)
b') (T): thie tha frambringit fon snemo treseuue nivvu inti altiu
(Latin: qui profert de thesauro suo nova et vetera)
who derives from his treasure the new and the old
c) (M): Enti sohuuer soquuidit le uuort uuidar mannes sune
Coor. REL. Vfnt O SP
c') (T): Inti souuer soquidit uuort uuidar then mannes sun
(Latin: Et quicumque dixerit verbum contra filium hominis)
and whoever says a word against the son of the man
Leaving aside the influence of the Latin syntax, it is clear that the adjacency
of relative pronoun-Vfnt could not be stated in structural terms. Assuming
that the relative pronoun occupies the specifier of CP, both the head of CP
(C°) and the trace of the relative pronoun in subject position ([Spec.IP]) inter-
vene between the relative pronoun and the finite verb.
10. Cf. the review by Allen (1990), where this objection against Kemenade (1987)
is also raised.
11. According to Steele et al. (1981:285 ff.), the evolution of English went through
the following stages:
a. S O V Aux (Old English)
b. S Aux O V (Early Modern English)
c. S Aux V O (Modern English)
Note that while, on one hand, there is a certain agreement an the relative
chronology of these three stages (cf., amongst others, Roberts (1985) and, for
independent evidence coming from a different field of research, Schwartz
CASES OF VERB THIRD IN OLD HIGH GERMAN 367
and Tomaselli (1988)), on the other hand, what is at stake here concerns the
absolute chronology. In fact it is crucial for the purposes of our analysis to
assume that OE was already characterized by a head-medial IP (cf. b).
12. Unless one wants to assume with Travis (1984) that Modern German is char-
acterized by a head-medial IP as well. The fact that in recent works it has
been convincingly shown that the system proposed by Travis is not an
adequate description for Modern German syntax (cf. den Besten (1986);
Tomaselli (1989); Schwartz and Vikner (1989)) does not compromise, how-
ever, the validity of her analysis for an older stage of the language (for related
ideas concerning the historical development of French, cf. Roberts (1992)).
Note, by the way, that the hypothesis that the history of German was charac-
terized by an SIOV stage finds an interesting parallelism within the typologi-
cal framework (cf. Lehmann (1971)).
13. The fact that I cannot represent the final landing site for verb movement (but
just an intermediate step in V to C° movement) must be independently as-
sumed for Modern Scandinavian languages (with the exception of Icelandic)
(cf., amongst others, Platzack (1986)).
14. More precisely, Kemenade (1987:139-140) assumes the following: "if an
operator with an index (wh-elements, Pa and ne) moves to Comp (= Spec,
CP[AT]), it transmits this index to the head Infl (= C°, in a system where both
CP and IP are distinct maximal projections [AT]) as in (44) (cf. a below [AT]):
a. [infip Comp Oi Infl] > [InflP Comp Oi Infli]
with respect to cliticization, (44) has the following effect: Comp and
Infl behave as one constituent, so that cliticization is on the Infl projection
rather than on Infl0." Apart from obvious terminological differences,
Kemenade's basic idea is the following: when the specifier of CP is occupied
by an operator, the specifier and the head of CP behave as one constituent
preventing anything from intervening between them.
15. For a first formulation of SAI and SCI in terms of movement of the Vfnt to
Comp, cf. den Besten (1983).
16. For an exhaustive and detailed explanation, cf. Rizzi (1990a).
17. This construction could be exemplified by the following word order pattern:
XP YP (Z) Vfnt (Y)
where both XP and YP are maximal projections. For a first analysis of this
construction in OHG, cf. Lenerz (1985).
18. The situation represented by some Northern Italian Dialects (like Basso
Polesano (cf. Poletto (1990a)) and Trentino (cf. Brandi and Cordin (1981)
and Rizzi (1987)) seems to provide immediate support to this claim. A dis-
cussion of these data goes beyond the purposes of this paper.
19. Cf. Roberts (1991) and the literature cited there.
20. As for the postulation of X internal levels, cf. Selkirk (1982) and Roberts
(1991).
368 CLITICS AND VERB SECOND
References
Allen, C. (1990) "Review of A. van Kemenade (1987)." Language 66:146-
152.
Behaghel, O. (1923-32) Deutsche Syntax. Eine geschichtliche Darstellung.
Heidelberg: Carl Winter.
Besten, H. den (1983) "On the Interaction of Root Transformations and Lexi-
cal Deletive Rules." In W. Abraham, ed. On the Formal Syntax of
Westgermania, 47-131. Amsterdam: John Benjamins.
Besten, H. den (1986) "Decidability in the Syntax of Verbs of (not necessar-
ily) West-Germanic Languages." Groninger Arbeiten zur
Germanistischen Linguistik 28:232-256.
Brandi, L. and P. Cordin (1981) "Dialetti e italiano: un confronto sul parametro
del soggetto nullo." Rivista di Grammatica Generativa 6:33-87.
Braune, W and E.A. Ebbinghaus (1979) Althochdeutsches Lesebuch. Tubingen:
Niemeyer.
Chomsky, N. (1986) Barriers. Cambridge: MIT Press.
Erdmann, O. (1985) Grundziige der deutschen Syntax. Ed. by W.A. Koch.
Hildesheim: Olms.
Fourquet, J. (1938) L'ordre des elements de la phrase en germanique ancien.
Paris: Les Belles Lettres.
Haegeman, L. (1991) Generative Syntax: Theory and Description. A Case
Study from West Flemish. Oxford: Blackwell.
Haegeman, L. and H. van Riemsdijk (1986) "Verb Projection Raising, Scope
and the Typology of Rules Affecting Verbs." Linguistic Inquiry 17:417-
466.
Hopper, P. (1975) The Syntax of the Simple Sentence in Proto-Germanic. The
Hague/Paris: Mouton.
Hulk, A. and A. van Kemenade (this volume) "Verb Second, Pro-drop, Func-
tional Projections and Language Change."
Jolivet, A. and F. Mosse (1972) Manuel de I'allemand du Moyen Age. Vol. 1.
Paris: Aubier Montaigne.
Kemenade, A. van (1987) Syntactic Case and Morphological Case in the His-
tory of English. Dordrecht: Foris.
Kemenade, A. van (1990) "The Loss of Subject-Verb Inversion in Late Middle
English." Paper given at Manchester University, April 1990.
Lehmann, W.P. (1971) "On the Rise of SOV Patterns in New High German."
In K.G. Schweisthal, ed. Grammatik, Kybernetik, Kommunikation.
Festschrift fur A. Hoppe, 19-24. Bonn, Ferd: Dummlers Verlag.
Lenerz, J. (1984) Syntaktischer Wandel und Grammatiktheorie. Eine
Untersuchung an Beispielen aus der Sprachgeschichte des Deutschen.
Tubingen: Niemeyer.
Lenerz, J. (1985) "Diachronic Syntax: Verb Position and COMP in German."
In J. Toman, ed. Studies in German Grammar, 103-132. Dordrecht:
Foris.
Lippert, J. (1974) Beitrage zu Technik und Syntax althochdeutscher
Ubersertzungen. Miinchen: Wilhelm Fink.
CASES OF VERB THIRD IN OLD HIGH GERMAN 369