Академический Документы
Профессиональный Документы
Культура Документы
General Editors
Werner Abraham Elly van Gelderen
University of Vienna / Arizona State University
Rijksuniversiteit Groningen
Volume 165
Comparative and Contrastive Studies of Information Structure
Edited by Carsten Breul and Edward Göbbel
Comparative and Contrastive
Studies of Information Structure
Edited by
Carsten Breul
Edward Göbbel
University of Wuppertal
Preface vii
List of contributors ix
List of abbreviations xi
Introduction 1
Carsten Breul & Edward Göbbel & Alexander Thiel
Contrastive topics and distributed foci as instances of sub-informativity:
A comparison of English and German 15
Volker Gast
Givenness and discourse anaphors 51
Luis López
Constraints on subject-focus mapping in French and English:
A contrastive analysis 77
Knud Lambrecht
Wh-questions in French and English:
Mapping syntax to information structure 101
Paul Boucher
A comparative perspective on intensive reflexives: English and Hebrew 139
Dana Cohen
Focus types and argument asymmetries: A cross-linguistic study
in language production 169
Stavros Skopeteas & Gisbert Fanselow
Topicality in L1-acquisition: A contrastive analysis
of null subject expressions in child French and German 199
Nicole Hauser-Grüdl
Formal and functional constraints on constituent order
and their universality 231
Peter Öhl
On the foundations of the contrastive study of information structure 277
Carsten Breul
Subject index 305
Preface
The present book contains a selection of articles based on papers presented at the
conference on “Contrastive Information Structure Analysis” organised by Carsten
Breul at the University of Wuppertal in March 2008. Two of the articles do not
originate in conference papers. The one by Luis López has been kindly contributed
upon invitation. The one by Carsten Breul reflects ideas that motivated the organi-
sation of the conference and relates them to some of the results obtained in other
contributions to this book.
We are grateful to several colleagues and friends who supported the con-
ference and without whom this book would not have come into existence. Our
thanks also go to the contributors, our assistant editor Alex Thiel, our student assis-
tants Â�Benjamin Köhnen and Ina Schlafke, and to the reviewers of the individual
articles. Each reviewer has accomplished their difficult and incongruously time-
consuming task in a very thorough and circumspect manner. Their comments and
suggestions have led to significant improvements – and to none of the shortcomings
of the present book. We are very pleased to have benefited from the expertise of
these colleagues.
Dana Cohen
UMR Structures Formelles Nicole Hauser-Grüdl
du Langage Romanistik
Université Paris 8 Bergische Universität Wuppertal
2 rue de la Liberté Gaußstr. 20
93526 Saint Denis cedex 42119 Wuppertal
France Germany
cattc@013.net nhauser@uni-wuppertal.de
The articles collected in this volume present original comparative and contrastive
research on various aspects of information structure (e.g. topic, focus, contrastivity,
givenness, anaphoricity) as well as on forms and structures whose realisation
depends on information structural factors (e.g. clefts, dislocations, reflexives, null
subjects, prosodic features, interrogatives) in a number of different languages (e.g.
Catalan, English, French, Georgian, German, Hebrew, Hungarian). Each of the
articles emphasises differences or commonalities between the languages under
investigation with respect to the realisation of information structural categories
or with respect to the information structural implications of a given form or
structure. This constitutes the specific comparative-contrastive perspective taken
in this volume, a perspective discussed from a methodological point of view in
the concluding article. The book is motivated by its editors’ and contributors’
conviction that comparative and contrastive research on information structure is
beneficial in two respects: First, it advances our knowledge of and insights into both
language specific and universal aspects of information structure; second, it raises
significant questions as to the formal representation and the functional properties
of information structural categories as well as its location in the architecture of
grammar, i.e. its relation to other components (perhaps modules) of grammar (in
the wide sense of the term). The articles collected here aim to make a substantial
contribution in these respects.
“Contrastive topics and distributed foci as instances of sub-informativity:
A comparison of English and German” is the title of Volker Gast’s contribution.
It presents a detailed elaboration of the notion ‘sub-informativity’ followed by
English/German contrastive analyses that have sub-informativity as their tertium
comparationis. The paper demonstrates clearly that the generation of insights by
contrastive analyses presupposes a well-chosen tertium comparationis. Moreover,
it is the introduction of sub-informativity as an information structural concept
capable of serving this heuristic function that constitutes one of the important
results of the paper.
Sub-informativity is an information structural notion in that it refers to
properties of declarative sentences that depend on properties of the discourse in
Carsten Breul & Edward Göbbel & Alexander Thiel
which they occur. It is defined by Gast (this volume: Section 3.2) in the following
way: “A declarative sentence S is sub-informative relative to a strategy Q
Â�(containing S) iff S does not answer all questions in Q.” The notion ‘strategy’ here
is borrowed from Büring’s (2003) model of a discourse as a tree whose nodes,
called ‘moves’, consist of explicit or implicit questions or sub-questions and their
answers. A strategy in this model is “[a]ny sub-tree […] rooted in an interrogative
move” (ib.: 518).
Sub-informativity comprises two major types, a focus-related one and a top-
ic-related one; the latter manifests itself in two sub-types, a context-preserving
one and a context-changing one. The examples in (1) (from Gast this volume:
Section 4), where each of the conjoined sentences in B’s answers is sub-informative,
give a sufficiently clear idea of what these distinctions amount to.
(1) a. A: Who read what?
B: John read the bible and Mary read the newspaper.
(focus-related sub-informativity)
b. A: What do your daughters study?
B: My older daughter studies law and my younger daughter studies
history. (topic-related sub-informativity, context-preserving)
c. A: What does your daughter study?
B: My older daughter studies law and my younger daughter studies
history. (topic-related sub-informativity, context-changing)
Gast’s English/German contrastive analyses based on these types of sub-informativity
concern selected lexical indicators (English as for vs. German was … anbetrifft),
selected syntactic indicators (English fronting vs. German movement to the
forefield) and, in greater detail, prosodic indicators. While Gast does not find any
major contrasts for these lexical and syntactic indicators of sub-informativity,1
he identifies a significant contrast regarding prosody. In English, there is an into-
national feature, the fall-rise accent, that can be used to signal incompleteness.
For Gast, incompleteness is a pragmatic notion which subsumes sub-informativity
alongside other categories. That is, in English there is no intonational contour
which is specialised for the expression of sub-informativity or any of its (sub-)
types. German, by contrast, has an intonational feature, the root contour, which is
specialised for the context-changing sub-type of topic-related sub-informativity.
Interestingly, according to Gast (this volume: Section 7.3), it is only in cases
of this latter type that the use of the fall-rise is “virtually obligatory” in English,
the other types of sub-informativity not being restricted to the fall-rise as their
.â•… As far as constituent movement to the left sentence periphery in English compared to
German is concerned, see also Breul (2007).
Introduction
�
intonational exponent. Observations like these point to interesting pathways
for further contrastive studies of the mapping between the meaning/function side
of information structural categories and their intonational exponents on the form
side (cf. Breul this volume).
The contribution by Luis López “Givenness and discourse anaphors” basically
reconsiders the traditional information structural category ‘background’. López
sets out from the acclaimed work by Vallduví (1992) and Vallduví & Engdahl
(1996), which reduces the two widely-accepted information structural divisions
topic-comment and focus-(back)ground to one single tripartite division into
link-focus-tail. López argues that Vallduví & Engdahl’s equivalence of destressing
in English and right-dislocation in Catalan, as two different grammatical repre-
sentations of tails, cannot be maintained and he shows that the grammars of the
two languages are sensitive to either givenness or discourse anaphoricity.
From a theoretical point of view, this article is a welcome contribution to the
analysis of information structure, due to its inclusion of anaphoric devices. The
study of anaphora is generally the study of pro-forms (pronouns, reflexives, sense
anaphors) and different types of ellipsis for which an antecedent is mandatory in
the same (complex) sentence or the discourse context. While the relation between
anaphora and information structural categories is generally acknowledged and
sometimes addressed (cf. Bolinger 1979; Tancredi 1992; Williams 1997; Winkler
2005), anaphora has rarely been the target of information structural studies but
has remained the realm of syntactic and semantic studies (presumably due to the
success of Binding Theory; cf. Chomsky 1981; Reinhart 1983). López however
shows that the relation between focus structure and its formal syntactic realisa-
tion (i.e. information packaging) cannot be accounted for without reference to
discourse anaphors.
The main claim is that the grammar of a language may encode either one or
the other category, while some languages may encode both. English is shown to be
sensitive to givenness (cf. Schwarzschild 1999) which is the target of deaccenting
rules. Catalan is shown to be sensitive to discourse anaphors, which are obligatorily
right-dislocated. While a focus may contain given material, discourse anaphors
are outside the focus. The distinction between the two categories and the sensitivity
of a language to one or the other category is achieved by a systematic comparison
of the realisation of the ground (Vallduví’s tail) in English and Catalan. While
English deaccents given material regardless of whether it is part of the focus or
not (hence is oblivious to focus structure), Catalan sentential stress is oblivious to
givenness. Catalan however clitic right-dislocates elements which are not part of
the focus (i.e. discourse anaphors). López also suggests that some languages, such
as German, may be sensitive to both givenness, given material being deaccented,
and discourse anaphors, which are scrambled.
Carsten Breul & Edward Göbbel & Alexander Thiel
The explanation for the different behaviour of English vs. Catalan is attributed
in part to the way sentential stress assignment works in the two languages and in part
to the timing of stress assignment, linearisation of syntactic structure and prosodic
phrasing in the mapping from syntax to PF. He argues that English stress is sensitive
to syntactic structure, while stress assignment is sensitive to linear order in Catalan.
Ordering stress assignment before linearisation and prosodic structure formation in
English accounts for this language’s sensitivity to givenness. In Catalan stress assign-
ment applies to a linearised string that has also been assigned its prosodic structure
and will therefore accent final given material that has not been right-dislocated.
This article is a fine demonstration of the role a contrastive analysis can play
in grammatical description. It demonstrates that a systematic contrastive analysis
of two languages can bring to light differences that a broad comparative approach
is likely to miss. Vallduví & Engdahl’s approach is a comparative one which tries to
identify different means of expressing the same information structural categories
cross-linguistically. López’ contrastive approach shows that a more fine-grained
analysis of the similarities and differences between two languages can lead to a
completely different conclusion.
In his contribution “Constraints on subject-focus mapping in French and
English: A contrastive analysis”, Knud Lambrecht continues his project of
showing how languages differ in the way they mark information structure by
lexico-grammatical means. In this paper he concentrates on how spoken French
differs from English in grammatically encoding focal subjects in argument-focus
and sentence-focus articulations. Lambrecht’s observations are based on the analysis
of a wide range of attested examples, many of which he presents and discusses in
the present paper.
Speakers of French show a strong tendency (more than 95% of all sentences
examined by Lambrecht) towards using a syntactic pattern that he calls “preferred
clause construction” (PCC). This has the formal make-up clitic pronoun + V +
XP+FOC and serves to avoid focal subjects in preverbal position, an effect of a
language-specific constraint in spoken French (but see also the contribution by
Skopeteas & Fanselow in this volume). One type of construction that targets the
PCC pattern is secondary predication: The logical subject of a proposition (which
would be expressed as the syntactic subject in a canonical SV(O) construction)
appears as the syntactic object of the PCC (XP+FOC in the PCC pattern) and at the
same time functions as the logical subject of the predication expressed by the rest of
the sentence (often a relative clause, a prepositional phrase or an adjective phrase).
Lambrecht’s main concern is to contrast several instances of such secondary
predication constructions of spoken French with their English translational equi�
valents under certain information structurally relevant conditions, notably those
giving rise to argument-focus on the one hand and sentence-focus on the other
Introduction
hand. The general picture emerging from these analyses is this: Spoken French
makes use of several non-canonical constructions that all conform to the PCC
pattern but are nevertheless different, while their English counterparts exhibit
canonical SV(O) constituent order throughout, making prosodic adjustments
according to the information structurally relevant conditions. Or, more specifically:
While canonical French SV(O) constructions are inappropriate in those cases
where the subject would be focused, their English canonical SV(O) counterparts
are the ‘normal’ choice, non-canonical English versions being marginal, significantly
less frequent or having a decidedly sub-standard flavour.
To give an example instantiating the argument-focus articulation: The English
version of (2) (Lambrecht’s (19)) was uttered in a restaurant when it came to
choosing an item from the wine list (square brackets indicate focal items; small
caps indicate accented items):
(2) a. I really don’t care. [You] decide.
b. Ça m’est vraiment égal. C’est [vous] qui dcidez.
The French version with canonical SV order, Vous décidez, would only be appro-
priate in a context where the addressee is already a topic under discussion. As far
as English is concerned, the clefted version of (2a), it’s you that decides, would be
marginal in the given context. Although c’est-clefts like the one above are the most
common means of expressing argument-focus articulations in spoken French,
Lambrecht also cites examples which show the avoir-cleft construction (Only [he]
understands me/Y a que [lui] qui me comprend). There are similar constructions in
English as well, but they are judged more or less sub-standard and are used far less
frequently than their French counterparts. Lambrecht mentions rather colloquial
got-constructions like I got proofs dancing in my head, syntactic amalgams like
I have a friend of mine in the history department teaches two courses per semester, or
subject-demoting constructions like We don’t last long with tablecloths in this house
(tablecloths being the logical subject). The function of all of these constructions is
to avoid an unidentifiable or inaccessible referent to be coded as a (sentence-initial)
grammatical subject.
The following example (Lambrecht’s (40)) instantiates the sentence-focus
articulation.
(3) a. A: Dis donc, Bernard!
B: Quoi?
A: [Une drôle de chose] qui m’arrive!
a′. A: #[Une drôle de chose] m’arrive!
b. A: Hey, Bernard!
B: What?
A: [A funny thing]’s happening to me!
Carsten Breul & Edward Göbbel & Alexander Thiel
The syntax of Une drôle de chose qui m’arrive! (which looks like a truncated cleft of
the type Y a une drôle de chose qui m’arrive!, but may be a construction in its own
right) conforms to the PCC pattern just like (2b) does, but is clearly different from
it. A corresponding canonical SV construction would be inappropriate in French,
but is the normal one to use in English.
Such cross-linguistic differences in grammatically encoding information
structure are viewed by Lambrecht as instantiations of a typological difference
between languages that allow for the co-mapping of subject and focus (e.g. English
and German) and languages that do not (e.g. spoken French and Sesotho).
Diachronically, modern French seems to be on its way towards grammaticalising
this co-mapping constraint.
In “Wh-questions in French and English: Mapping syntax to information
structure”, Paul Boucher addresses the question of why it is that French has so
many syntactically different types of interrogative sentences compared to the
rather sparse inventory of English. Actually, according to Boucher, the rich array
of French interrogatives can be reduced to four basic types: type 1: wh-phrase in
situ, verb in situ; type 2: wh-phrase raised, verb in situ; type 3: wh-phrase raised,
verb raised; type 4: wh-phrase raised + est-ce que, verb in situ. He then goes on to
focus on the apparent optionality of either raising French wh-phrases to the left
sentence periphery or leaving them in situ, that is, more specifically, on the differ-
ence between types 1 and 2. The important point here is that, in contrast to �English,
French wh-phrase in situ interrogatives may be employed as echo questions but
are not restricted to this function: “Type 1 questions may be used, as in English,
as ‘echo questions’ […] or as ‘phatic introductory questions’ […]. Contrary to
English, they may also be used as requests for information.” (Boucher this
volume: Section 1.2).
Making use of insights from the diachrony of French interrogatives on the
one hand and of analyses of the syntax of Romance interrogatives by Poletto &
Pollock (2004, 2009) and Munaro & Pollock (2005) on the other hand, Boucher
proposes that the two types manifest two different syntactic structures that encode
different information structural constraints. Basically, the corresponding two
syntactic structures are argued to differ in the following way: Of three operators
involved in the syntax of interrogatives – the highest ‘disjunction operator’, the
medial ‘existential operator’, the lowest ‘restrictor operator’ – either the highest or
the lowest may have a phonologically overt reflex in French interrogatives (que or
quoi respectively), with the other two having phonological null reflexes. Boucher’s
point is that the syntactic configurations corresponding to these two cases corre-
late with just those information structural constraints that he identifies for type 1
(wh-phrase in situ; verb in situ) and type 2 (wh-phrase raised; verb in situ) in
the descriptive part of the paper: “in both cases, there is a strong dependence
Introduction
Skopeteas & Fanselow challenge the first claim and provide independent
evidence for the second. By presenting statistical evaluations of semi-spontaneously
produced data, they show that speakers of the four languages select different struc-
tures under identical discourse conditions in spite of the fact that theses languages
allow in principle both marked and unmarked options of syntactic focus realisation.
The method employed, an elicitation task that establishes particular context types
by means of visual stimuli and minimal verbal contributions (e.g. questions),
guarantees that the discourse conditions are the same in the languages analysed.
The tests are restricted to narrow focus.
The results in a nutshell: Georgian allows in situ focus, preverbal focus and
clefting. The statistical evaluation of the production tests reveals that clefting,
though possible, does not occur and that there is an asymmetry both in focus type
and argument type: Identificational focus is more often realised in the preverbal
focus position and focused subjects occur more often in this position than objects.
In Hungarian, focus is always expressed ex situ regardless of focus and argument
type. In American English, identificational focus on subjects induces a low pro-
portion of cleft constructions, while objects are not clefted. In Québec French,
narrow focus on the subject induces a high proportion of cleft constructions.
The merit of the article lies in an integrated explanation of the crosslinguistic
facts. It takes into account both the possible syntactic configurations that can
express the focus structural distinctions and the role prosodic factors play in the
respective languages. The fact that the speakers of a language employ fewer options
than the language in principle allows is attributed to a minimality condition on
language production, which is analogous to Chomsky’s economy condition
on syntactic derivations (Chomsky 1992) and related ‘least effort’ principles in
language processing. The minimality condition states that, if two syntactic con-
structions can be used for the same information structural configuration, a speaker
will choose the syntactically less complex construction. For example, biclausal
constructions like clefts are not used in Georgian and Hungarian because a
syntactically less complex option (reordering) is available. The fact that Hungarian
speakers move both identificational and non-identificational foci to the structural
focus position, while Georgian speakers do so only optionally, is due to the fact
that Hungarian sentential stress is fixed, while in Georgian focus can also be
signalled in situ intonationally.
While Skopeteas & Fanselow reject a biuniqueness relation between focus
type and structural realisation, their statistical data clearly show that identifica-
tional focus and non-identificational focus at least tend to be realised differently.
For example, identificationally focused constituents are realised more frequently
in the preverbal position in Georgian and only identificationally focused subjects
invoke cleft sentences in English. Since in both languages identificational focus
Carsten Breul & Edward Göbbel & Alexander Thiel
such pronouns as topics that are permanently present in the discourse context
and therefore represent permanently available referents. This explanation reflects
the view that children have a different understanding of topic from that of adults.
The second possible explanation she offers is that children misinterpret 1st and
2nd person deictic pronouns as anaphoric expressions. This interpretation of the
facts is based on the observation that the use of deictic subject pronouns is more
difficult to learn than that of 3rd person ones. Evidence for the second explanation
comes from the fact that children sometimes use a 2nd person pronoun incor-
rectly to refer to themselves, from certain mismatches in subject-verb agreement
and from omission of 1st person subjects that are not licensed in adult speech
(i.e. in adult German a 1st person omitted subject must be anaphoric).
The contribution by Peter Öhl “Formal and functional constraints on constiÂ�
tuent order and their universality” reconsiders the traditional distinction between
subject-prominent and topic-prominent languages (Li & Thompson 1976) against
the background of recent generative research on the representation of discourse
semantic properties as well as the representation of grammatical relations in
generative grammar. The paper has a comparative-typological orientation, discuss-
ing central cases of the two types of language (e.g. English vs. Japanese, Korean,
Hungarian) with interesting consequences for SOV languages like German, which
cannot readily be classified in terms of subject/topic-prominence.
Li & Thompson’s classical distinction is now generally captured in terms of
whether a language is discourse-configurational, expressing either topic or focus
structurally (Kiss 2001), or whether the syntactic representation of grammatical
relations predominates. Since functional categories play an important role in the
expression of both types of properties, this paper critically evaluates their univer-
sality as well as their necessity for particular languages. In agreement with Kiss
(2001), Öhl argues that languages are neither purely discourse-configurational,
nor purely ‘relational-configurational’. This is due to the fact that the expression
of information structural properties and the syntactic representation of argument
structure are constrained by different principles: functional vs. purely formal ones
(e.g. the existence of a structural subject position and the EPP property typically
associated with it).
Furthermore, there is no complementarity nor any parametric option underlying
the distinction subject- vs. topic-prominence, which is part of his explanation for
the existence of a large number of languages of mixed type. Öhl argues that such
language systems can be captured if many of the functional categories proposed
in the generative literature are not universal: The more parameterised functional
projections are specified in a language, the stricter the word order in that language.
Languages which resort to adjunction have more variable word order possibilities.
Öhl argues and discusses evidence that German has no structural subject position,
Carsten Breul & Edward Göbbel & Alexander Thiel
a position also defended in work by Haider (1993, 2010) and Sternefeld (2006). He
also presents several arguments that German has no structural topic position. This
language, like many other SOV languages, has productive leftward scrambling.
Allowing free and iterative adjunction, German can encode discourse-functional
constituents without recourse to functional projections. To this end, Öhl critically
evaluates the putative universality of functional projections (cf. Cinque 1998) and
argues that there is a universal inventory of functional categories from which a
particular language selects a subset in the course of language acquisition.
The paper also contains a careful overview of the literature on and discussion
of the information structural category topic, a category notoriously difficult to
define formally. Öhl argues that properties typically associated with topics, like
givenness, specificity, definiteness, have to be considered prototypical rather than
necessary and sufficient conditions for a constituent to be chosen as a sentential
topic. He further suggests that topic and comment should not be considered prim-
itives of information structural categorisation, but should be derived from more
general cognitive principles of perspectivation, which allow speakers to present
constituents in a sequential order based on criteria like point of view, saliency,
scope, subject of predication, and also other factors which have often been claimed
to be hallmarks of topichood (e.g. given/familiar before focal/new).
As a conclusion to the volume, Carsten Breul’s “On the foundations of the
contrastive study of information structure” reflects on the underlying ideas and
the methodological motivation of a decidedly contrastive approach to information
structure and recapitulates some of the results of some of the preceding articles
in the light of his reflections. Putting emphasis on the central notion of tertium
comparationis (TC) and its being or not being consciously incorporated in the
design of a study, he first draws a tentative subdividing line between – more
generally – comparative and – more specifically – contrastive work. As an incentive
for future research in this latter direction, Breul then points out the potential
benefits of heeding rather strictly the demand for a clearly explicated TC (either
on the meaning/function side or on the form side of language) in contrastive stud-
ies in general and in contrastive information structure analysis in particular. Thus
designed, contrastive analysis not only allows for methodologically sound state-
ments about how languages A and B (similarly or differently) express the same
‘facts’ (e.g. information structural configurations), it may also lead the researcher to
critically reconsider underlying concepts she has taken for granted so far. When, for
example, a contrastive analyst chooses to establish the TC on the meaning/function
side, she is supposed to carry out the cross-linguistic comparison on the form side,
after which she may find herself in doubt as to whether a certain concept, such as
‘word’, ‘tense’, or ‘voice’ (which often serve as the starting point for comparison in
typological studies), is actually equally applicable to both languages compared.
Introduction
For Breul, the value of contrastive analysis in general lies in its potential to
supply other (theoretical and applied) linguistic disciplines (notably typology,
comparative linguistics in the vein of generative grammar, language teaching and
translation studies) with reliable, ‘pre-processed’ data clear of conceptual entangle-
ments. This also applies on a more specific level to contrastive approaches to issues
in information structure. However, the choice of the theoretical framework for
the analysis may make a difference for the aim of a non-circular establishment of
the TC. According to Breul, Halliday’s (1985/1994: 37) characterisation of ‘theme’,
for instance, conflates formal and functional aspects, while Lambrecht’s (1994)
categories such as ‘identifiability’ or ‘activation’ – being defined in terms of cognitive
states – are independent of the categories used for linguistic description. Thus they
may serve as a TC on the meaning/function side in a contrastive analysis of how
individual languages vary in formally reflecting a given (cognitive) configuration.
Breul also provides an example of how to solve the operability problem arising
from the purely cognitive nature of such categories, viz. how to detect comparable
information structural configurations.
Turning to a TC established on the form side, he states that only those forms
and structures may be so used that are equivalently present in both languages. This
could be warranted by assuming universal syntactic features and principles, which
become manifest as syntactic operations or constructions (e.g. cleft constructions)
that can be observed in both languages compared. This would yield a formal TC
against which cross-linguistic differences or commonalities in meaning/function
can be assessed, as it is exemplified with recourse to English and French cleft
constructions from Lambrecht’s contribution to the volume.
In a final, relativising step, Breul concedes that it is probably not possible for
all research questions in the domain of information structure to perfectly meet
the requirements he postulates, especially when different aspects of meaning have
to be kept apart. Yet, this is not considered to pose a threat to his central claim
that the necessity to contemplate on a suitable TC when planning and conducting
contrastive analyses will advance our understanding of the concepts involved.
References
Bolinger, D.L. 1979. Pronouns in discourse. In Discourse and Semantics, T. Givón (ed.), 289–309.
New York NY: Academic Press.
Breul, C. 2007. Focus structure, movement to spec-Foc and syntactic processing. In On
Information Structure, Meaning and Form: Generalizations Across Languages, K. Schwabe &
S. Winkler (eds), 255–274. Amsterdam: John Benjamins.
Büring, D. 2003. On D-trees, beans and B-accents. Linguistics and Philosophy 26: 511–545.
Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris.
Carsten Breul & Edward Göbbel & Alexander Thiel
Chomsky, N. 1992. A Minimalist Program for Linguistic Theory [MIT Occasional Papers in
�Linguistics 1]. Cambridge MA: The MIT Press.
Cinque, G. 1998. Adverbs and Functional Heads: A Cross-linguistic Perspective. Oxford: OUP.
Drubig, H.B. 2003. Toward a typology of focus and focus constructions. Linguistics 44(1): 1–50.
Gast, V. 2006. The Grammar of Identity: Intensifiers and Reflexives in Germanic Languages.
London: Routledge.
Haider, H. 1993. Deutsche Syntax Generativ. Tübingen: Narr.
Haider, H. 2010. The Syntax of German. Cambridge: CUP.
Halliday, M.A.K. 1985/1994. An Introduction to Functional Grammar, 2nd edn. London: Arnold.
Kiss, K.É. 1998. Identificational vs. information focus. Language 74(2): 245–273.
Kiss, K.É. 2001. Discourse configurationality. In Sprachtypologie und sprachliche Universalien: Ein
internationales Handbuch, M. Haspelmath et al. (eds), 1442–55. Berlin: Mouton de Gruyter.
König, E. 1991. The Meaning of Focus Particles: A Comparative Perspective. London: Routledge.
Lambrecht, K. 1994. Information Structure and Sentence Form: Topic, Focus and the Mental
Representations of Discourse Referents. Cambridge: CUP.
Li, C.N. & Thompson, S.A. 1976. Subject and topic: A new typology of language. In Subject and
Topic, C.N. Li (ed.), 457–89. New York: Academic Press.
Munaro N. & Pollock, J.-Y. 2005. Qu’est-ce que (qu)-est-ce que? A case study in comparative
Romance interrogative syntax. In Handbook of Comparative Syntax, G. Cinque & R. Kayne
(eds), 542–606. Oxford: OUP.
Poletto C. & Pollock, J.-Y. 2004. On the left periphery of some Romance wh-questions. In The
Structure of CP and IP, L. Rizzi (ed.), 251–296. Oxford: OUP.
Poletto C. & Pollock, J.-Y. 2009. Another look at wh-questions in Romance: The case of
Mendrisiotto and its consequences for the analysis of French in situ and embedded
interrogatives. In Romance Languages and Linguistic Theory 2006, D. Torck & W.L. Wetzels
(eds), 199–258. Amsterdam: John Benjamins.
Reinhart, T. 1983. Anaphora and Semantic Interpretation. London: Croom Helm.
Rooth, M. 1992. A theory of focus interpretation. Natural Language Semantics 1: 75–116.
Schwarzschild, R. 1999. Givenness, AvoidF and other constraints on the placement of accent.
Natural Language Semantics 7: 141–177.
Siemund, P. 2000. Intensifiers in English and German: A Comparison. London: Routledge.
Sternefeld, W. 2006. Syntax: Eine morphologisch motivierte generative Beschreibung des
Deutschen. 2 vols. Tübingen: Stauffenburg.
Tancredi, C. 1992. Deletion, Deaccenting and Presupposition. Ph.D. dissertation, MIT.
Vallduví, E. 1992. The Informational Component. New York NY: Garland.
Vallduví, E. & Engdahl, E. 1996. The linguistic realization of information packaging. Linguistics
34: 459–519.
Williams, E. 1997. Blocking and anaphora. Linguistic Inquiry 28: 577–628.
Winkler, S. 2005. Ellipsis and Focus in Generative Grammar [Studies in Generative Grammar 81].
Berlin: Mouton de Gruyter.
Contrastive topics and distributed foci
as instances of sub-informativity
A comparison of English and German
Volker Gast
Friedrich-Schiller-Universität Jena
1.â•… Introduction
Information structure has not so far been among the topics that have been
�prominently discussed in contrastive studies of English and German.1 Broadly
speaking, three types of expressive devices for the encoding of information struc-
tural categories can be distinguished: (i) lexical devices, (ii) syntactic devices and
(iii) prosodic devices. Among these, it is probably the class of ‘lexical devices’ that
has received most attention in the relevant literature. For instance, König’s (1982)
contrastive investigation of focus particles in English and German deals with an
entire lexico-grammatical sub-system relating to, and interacting with, informa-
tion structure. Breul (2008) provides a comparison of the way the information
structural category of ‘identifiability’ is lexicalized in English and German, deal-
ing with explicit topic exponents of the type as for and definite articles. As far as
the syntactic devices are concerned, some highly pertinent work has been done in
the past few years (e.g. Doherty 2005; Frey 2005, Molnár & Winkler forthcoming
on matters of topic-worthiness/topicalization and leftward movement). The third
type of expressive device – prosodic means of encoding information structure –
has, to the best of my knowledge, not so far been studied in an English-German
contrastive perspective.2 Note that the disregard for information structure in
contrastive linguistics is also mirrored in the fact that none of the relevant survey
monographs contains a section on information structure (e.g. Burgschmidt &
Götz 1974; Hawkins 1986, König & Gast 2009).3
There are several reasons for this neglect of ‘contrastive information structure
analysis’. First, the study of information structure was still in its infancy in the
1960s and 1970s, when English-German comparison flourished, at that time carried
by the wish to improve foreign language teaching. A second problem probably
concerns the conceptual basis of comparison (cf. also Breul 2008: 266–7). As any
other type of comparative investigation, contrastive studies need to be based on a
more or less clearly delimited tertium comparationis. Ideally, such a ‘third of com-
parison’ should be defined on a purely notional basis. It constitutes the invariant
in the process of language comparison, while variation is expected in the formal
means used to encode the relevant categories. In the case of information struc-
tural categories, such a tertium comparationis is very hard to define, even for basic
I would like to thank the editors of this volume, Florian Haas, Martin Schäfer, an anonymous
referee and all participants of the conference on ‘Contrastive Information Structure Analysis’,
held at the University of Wuppertal on March 18–19, 2008, for critical comments and helpful
suggestions.
.â•… For general comparisons of English and German prosody, see Sculfill (1982) and Markus
(1992, 2006).
.â•… The comparative analysis of information structure in English and German is one of the
primary objectives pursued by the project mentioned in Note 1.
Contrastive topics and distributed foci as instances of sub-informativity 
notions such as ‘topic’, ‘focus’, and ‘contrast’. The difficulties in defining the notion of
‘topic’ have been commented on, among others, by Reinhart (1981) and Polinsky
(1999) (cf. also Jacobs 2001: 643):
Although the linguistic role of the relation topic of is widely acknowledged, there
is no accepted definition for it, and not even full agreement on the intuitions of
what counts as topic (Reinhart 1981: 56).
Linguists have essentially given up on a rigorous definition of topics – almost
everyone […] mentions the aboutness condition and then moves on to more
mundane matters of topichood or topicalization (Polinsky 1999: 572).
The notion of ‘uniqueness’ in the top left cell of the table is associated with the
linguistic category of definiteness on the right (cf. also Breul 2008 and references
cited there).4 However, definiteness is not associated with ‘uniqueness’ alone but
also indicates ‘givenness’. Givenness, in turn, is usually marked by deaccentuation,
but is also associated with specific syntactic operations such as fronting in English
(cf. Birner & Ward 1998 and Section 6 below). But then, fronting also requires
contrast, which in turn is indicated by accentuation, and so on indefinitely. This
situation makes it very hard to base any notional category of information struc-
ture on formal distinctions found in natural languages, as is commonly done in
.â•… Breul (2008) solves the problem of finding a tertium comparationis by basing it on a
notional category (‘identifiability’), but using a formal one (as for-expressions) as a diagnostic
for identifying instances of identifiability.
 Volker Gast
linguistic typology (cf. Lazard 1999, 2001 for a description of such a procedure
based on ‘arbitrary conceptual frames’; see also Haspelmath 2008, Gast forthcoming
for discussion).
This study intends to make a contribution to contrastive information structure
analysis by focusing on a specific aspect of information structure in English and
German, i.e. the encoding of sub-informativity, which subsumes occurrences of
contrastive topics5 and of distributed (multiple) foci (cf. Section 4 on the term
‘distributed foci’). Contrastive topics can be illustrated with the question-answer
pair in (1). Each of the sentences in B’s answer contains a topic (my older daughter,
my younger daughter), and the two topics stand in a (paradigmatic) relationship of
contrast to each other:
(1) A: What do your daughters do?
B: [My older daughter]CT studies law, and
[my younger daughter]CT studies history.
One of the most prominent features of examples like (1) is that the answer
given by B is made up of more than one sentence, and that A’s question is only
answered by the conjunction of these sentences. We will say that elements of
such ‘conjoined answers’ or ‘answer sets’ – ‘partial answers’, as they may be called –
are sub-informative relative to the ‘question under discussion’ (What do your
daughters do?). Sub-informativity can also be observed in question-answer pairs
like the one in (2), which provides an example of a ‘distributed focus’:
(2) A: Who danced with whom?
B: Fred danced with Mary, and Bill danced with Jane.
The present study will investigate the way sub-informativity is expressed in �English
and German at different levels of the lexico-grammatical system. The discussion
starts in Section 2 with a few terminological remarks. Section 3 introduces the
concept of ‘sub-informativity’, which functions as the tertium comparationis of
this study. Section 4 distinguishes several types of sub-informativity. There are
two major types (‘focus-related’ and ‘topic-related sub-informativity’) and two
sub-types of ‘topic-related sub-informativity’ (‘context-preserving’ and ‘con-
text changing’). Sections 5 and 6 provide a comparison of the most central lexi-
cal (Section 5) and syntactic markers (Section 6) of sub-informativity in English
.â•… The notion ‘contrastive topic’ is widely used in research on information structure (see e.g.
Lambrecht 1994: 291–5; Erteschik-Shir 2007: 48ff.). The various conceptions of this notion
differ considerably, however. The concept underlying the present study will be explicated in
Section 3.1. It is mainly based on the work done by Jacobs (1982, 1996, 1997), Krifka (1994,
1998, 2007) and Büring (1994, 1997, 2003).
Contrastive topics and distributed foci as instances of sub-informativity 
and German. Only minor contrasts can be identified in these domains. In Section 7,
the most important prosodic devices for marking sub-informativity are discussed,
i.e. the fall-rise contour in English and the ‘root contour’ in German. It is shown
that they are not equivalent in terms of their information structural appropri-
ateness conditions, and that the German root contour is used only in specific
instances of sub-informativity, i.e. in ‘context-changing’ ones. Section 8 concludes
with a brief summary and some general remarks on methodological problems of
contrastive information structure analysis.
.â•… “In (X Y), X is informationally separated from Y iff the semantic processing of utterances
of (X Y) involves two steps, one for X and one for Y” (Jacobs 2001: 645).
.â•… “In (X Y), X is the semantic subject and Y the semantic predicate iff (a) X specifies a vari-
able in the semantic valency of an element in Y, and (b) there is no Z such that (i) Z specifies
a variable in the semantic valency of an element in Y and (ii) Z is hierarchically higher in
semantic form than X” (Jacobs 2001: 647).
.â•… “In (X Y), X is the address for Y iff X marks the point in the speaker–hearer knowledge
where the information carried by Y has to be stored at the moment of the utterance of (X Y)”
(Jacobs 2001: 650).
.â•… “In (X Y), X is the frame for Y iff X specifies a domain of (possible) reality to which the
proposition expressed by Y is restricted” (Jacobs 2001: 656).
Volker Gast
What restrictors and contrastive topics have in common is that they are, in a way
to be made more explicit below, ‘sub-informative’. Given that ‘sub-informativity’
will function as the tertium comparationis of this study, much of what is said will
be applicable to contrastive topics and restrictors alike, but the discussion will be
restricted to the former type of expression.
The terms ‘contrast’ and ‘focus’ are no less difficult to capture. ‘Focus’ is some-
times used for “that portion of a proposition that cannot be taken for granted at the
time of speech” (Lambrecht 1994: 207), i.e. it is regarded as a relational antonym
of ‘topic’, thus corresponding to what is called ‘comment’ in other traditions (e.g.
Hockett 1958). Other authors (most notably those in the tradition of Rooth 1985)
regard ‘being in focus’ as indicating “the presence of alternatives that are relevant
.╅ Topics that are also restrictors are regarded as contrastive topics by Krifka (2008).
Contrastive topics and distributed foci as instances of sub-informativity 
A different approach is taken by Jacobs (1982, 1996, 1997), who uses the term
‘i-topicalization’12 for (the German counterparts of) configurations like those
in (6). ‘I-topicalization’ is regarded as “a kind of contrastive topicalization in
German” (Jacobs 1997: 91). However, ‘i-topicalization’ differs from ‘contrastive
topicality’ as conceived of by Krifka in some respects. First, it does not only apply
to topics, but also to non-referential expressions like modal operators, as in (7)
(‘√’ and ‘ \’ indicate pitch movement; cf. Section 7):
(7) Man √muss das Buch \nicht mögen, aber man \kann.
one ╇ must the book ╇↜渀屮not like but one ╇↜can
‘You don’t have to like the book, but you can.’ (Jacobs 1997: 122)
.╅ In terms of Krifka (2008), contrastive topics can be regarded as constituents whose
denotations serve the function of both ‘addressation’ and ‘delimitation’: They break down a
given (referential) topic (an address) into ‘sub-topics’, so that different comments can be made
about each of these sub-topics; cf. Note 10.
.â•… ‘I’ stands for ‘intonation’, and ‘i-topicalization’ contrasts with ‘s-topicalization’ (‘s’ for
Germ. Stellung ‘position(ing)’).
Contrastive topics and distributed foci as instances of sub-informativity
discourse
question question
…
subq subq subq subq
answer answer
Each node in a d-tree is called a ‘move’, and “[a]ny sub-tree of a d-tree which is
rooted in an interrogative move is a strategy” (Büring 2003: 518). An example of a
‘strategy’ is given in Diagram 2 (from Büring 2003: 520).13
What did Fred eat? What did Mary eat? What did...
The assertion Fred ate the beans at the left terminal node serves as an answer to the
question immediately dominating it, i.e. What did Fred eat (the ‘question under
discussion’, or QUD). In the example given in Diagram 2, this sentence answers
only one of the questions in the ‘strategy’, which is rooted in the question Who
ate what?. This question dominates other ‘sub-questions’ in addition to What did
Fred eat?, e.g. What did Mary eat?. According to Büring (2003), contrastive topic
.â•… Note that ‘fred’ and ‘mary’ are regarded as contrastive topics by Büring (2003), hence the
subscript ‘ct’. In the present study they do not qualify as topics, but are regarded as ‘distributed
foci; cf. the following discussion.
Volker Gast
marking on Fred (in Fred ate the beans) indicates that this sentence does not pro-
vide a complete answer to the root question (Who ate what), i.e. there are other
sub-questions that need to be addressed. The constituent carrying a contrastive
topic accent (Fred) is the one distinguishing the relevant sub-questions from each
other (Fred, Mary). Accordingly, the sentence [Fred]CT ate the [beans]F indicates
that there is a question of the form ‘xCT ate the yF’ in the strategy to which it does not
provide an answer. In this framework, contrastive topic marking is thus regarded
as a relation between an assertion (Fred ate the beans) and the strategy containing
that assertion (i.e. the strategy rooted in the question Who ate what?).
According to the definitions adopted in this study, the examples discussed by
Büring (1997, 2003) do not contain topics but should be regarded as answers to
distributed (multiple) wh-questions (cf. Note 13 and Section 4). The question arises
whether these types of constituents should be included in a discussion of ‘con-
trastive topicality’ at all. While there are clear differences between the two types
of phenomena (cf. Jacobs 1997: Section 2.3.2 for discussion), they are nonetheless
clearly related, not least in terms of their prosodic properties, and will therefore
be taken into consideration. The introduction of the more general concept of
‘sub-informativity’ in the next section will allow us to determine both the common
denominator of distributed foci and contrastive topics, and the differences between
them: Both types of constituents are ‘sub-informative’, but they occur in different
types of discourse environments.
3.2â•… Sub-informativity
As the brief review of different conceptions of ‘contrastive topicality’ has shown,
this term can be interpreted as a property of constituents (‘constituent x is a
contrastive topic’, cf. Krifka 1998), as a property of sentences or utterances (‘this
sentence/utterance contains an illocutionary operator indicating i-topicalization/
contrastive topicalization’, cf. Jacobs 1997), or as a relation between an utterance
and the strategy containing that utterance (cf. Büring 2003). The following discus-
sion will largely be based on Büring’s (2003) analysis, even though his notion of
‘contrastive topic’ will be reinterpreted. Büring’s theory has the advantage of making
reference to higher-level entities (namely, discourse-related ones) and thus pro-
vides a reasonable basis for a comparative study, since discourse situations with
specific properties (‘d-trees’) are easily reproducible across languages. Moreover,
it allows us to abstract away from the type and locus of encoding (lexical material,
constituents, syntactic operations, prosody, etc.), and thus to determine the range
of variation found in the formal encoding of the relevant concepts in the languages
under comparison.
As has been shown by Büring (1997, 2003), the most important property of
sentences with contrastive topics (or distributed foci) is that they are, in some way,
Contrastive topics and distributed foci as instances of sub-informativity
‘sub-informative’ relative to the strategy they are contained in, insofar as they leave
open questions. The notion of ‘sub-informativity’ will be used as the basis of our
comparison. It is defined as follows:
Two major types of sub-informativity can be distinguished. The first type is found
in sentences with ‘distributed foci’. Such sentences are answers to questions con-
taining more than one wh-pronoun (‘matching questions’ in terms of Krifka 2001).
I use the term ‘distributed foci’ rather than ‘multiple foci’ because such sentences
can be analyzed as containing a single focus which is distributed over several
constituents, rather than containing several foci. Consider (9):
(10) Question:
Which pairs 〈x,y〉 are contained in the extension of the predicate read?
Answer:
The extension of the predicate read contains the pairs 〈John, bible〉
and 〈Mary, newspaper〉
Volker Gast
‘Binary foci’ such as 〈John, bible〉 are generally ‘distributed’ over (at least) two con-
stituents, hence the term ‘distributed foci’. The type of sub-informativity associated
with distributed foci as in (9) will be called ‘focus-related sub-informativity’. It is
not primarily a result of ‘imperfect common ground management’ (as in cases of
topic-related sub-informativity, cf. below), but of the fact that natural language
does not normally provide for ‘relational’ (i.e. more than unary) wh-pronouns and
corresponding foci.14 Distributed foci give rise to sub-informativity insofar as each
one of the answers given in examples like (9) above is sub-informative relative to
the superordinate question Who read what?.
In a second type of sub-informativity, more than one topic-comment rela-
tion is established. This type will be called ‘topic-related sub-informativity’.
Topic-related sub-informativity arises under one of two conditions. First, there may
be a mismatch between the background assumptions made by the interlocutors
involved; and second, the interlocutors may have the same propositional back-
ground, but the question under discussion may be phrased in such a way that it
cannot be answered in a single sentence. Instances of the first type will be called
‘context-changing’ and instances of the second type ‘context-preserving’ cases of
sub-informativity.
Let us start with the second case, where both speakers share (more or less) the
same propositional background, but one of the interlocutors phrases a question
in such a way that a single answer is not possible, i.e. s/he establishes a discourse
topic about which no single piece of information can be given. A relevant example
is given in (11) (= (1)):
(11) A: What do your daughters study?
B: My older daughter studies law and my younger daughter studies history.
Speaker A is (or at least may be) aware that B’s daughters do not study the same
subject. However, the two questions are ‘compressed’ into one. They could be
asked separately, but more ‘communicative effort’ would have to be spent. The
‘sub-informativity’ of each of the answers given by B is thus anticipated and in fact
provoked by A.
This is different in ‘context-changing’ instances of sub-informativity, where the
propositional backgrounds of the interlocutors differ, and the common ground is
actively modified by one of the speakers. This type of context is illustrated in (12):
(12) A: What does your daughter study?
B: My older daughter studies law, and my younger daughter studies history.
.╅ Note that the adverb respectively can be used in order to express distributed foci in a
single sentence (with conjoined NPs), e.g.: The cups and saucers cost £5 and £3 respectively
(Longman Dictionary of Contemporary English, s.v. respectively).
Contrastive topics and distributed foci as instances of sub-informativity
Speaker A is not aware that speaker B has two daughters. Accordingly, the infor-
mation structure of B’s utterance is not merely a reflex of the discourse structure
imposed by A; rather, B actively modifies the propositional background by intro-
ducing a new ‘sub-topic’ (my older daughter). The entire utterance establishes a new
‘super-topic’ (my [two] daughters). The difference between ‘context-preserving’ and
‘context-changing’ instances of sub-informativity can thus be described as follows:
In context-preserving cases a given set of topics is under discussion, but is sum-
marized under a single term (e.g. ‘your daughters’ = {‘your older daughter’, ‘your
younger daughter’}, cf. (11)), while in context-changing cases a new set of topics is
introduced by actively modifying the common ground (e.g. ‘my daughters’ = {‘my
older daughter’, ‘my younger daughter’} instead of ‘my daughter’, cf. (12)).
As will be seen below, the distinction between context-changing and context-
preserving topics is relevant to a comparison of English and German, as the two
languages differ in the way they encode these functions prosodically (cf. Section 7).
The three types of ‘sub-informativity’ introduced in this section are summarized
in Diagram 3:
sub-informativity
focus-related topic-related
context-preserving context-changing
We will now turn to the main topic of this study, i.e. a comparison of the lexical,
syntactic and prosodic means of indicating sub-informativity in English and
German (Sections 5–7).
English has a number of lexical devices that are commonly regarded as explicit
topic exponents, e.g. as for, speaking of and regarding. Some relevant examples are
given in (13)–(15):
(13) As for external funding, Smith has a grant application pending.
(Huddleston & Pullum 2002: 1371)
(14) Any noise it made was lost in the wind. Speaking of the wind, it was getting
stronger and I was getting colder. [BNC A6T]
(15) Regarding training, the document said that food business operators must ensure
that food handlers are fully trained or supervised … [BNC A0C]
Volker Gast
The question arises to what extent these lexical devices can be regarded as indica-
tors of sub-informativity. We will consider the most prominent relevant marker of
English, i.e. as for (cf. also Breul 2008). The main condition that must be met for a
constituent to be accompanied by as for is that it has to be ‘contextually accessible’
(cf. Lambrecht 1994: 152):
… the phrase as for NP (as well as similar phrases in other languages) can
be appropriately used only if the NP referent is already a potential topic in the
discourse at the time the phrase is used, i.e. the referent is contextually accessible.
(17) SQ1: What news are there concerning the book he was writing?
SQ2: What news are there concerning external funding?
The main topic of this passage is the ‘three goals’. This topic is broken down into
three sub-topics, i.e. each individual goal. The ‘root question’ of this strategy can
be phrased as ‘What were the three goals like?’, and it is answered in terms of three
Contrastive topics and distributed foci as instances of sub-informativity
(19) [Im Kommen sind körnige [Stoffe]CT mit leichten Strukturen, Mischgewebe
und teilweise sogar Glanzstoffe. Sie sind durchwegs leicht und weich,
angenehm anzufassen und zu tragen. Neben reiner Baumwolle kommen
Mischungen aus Cotton und Polyester, Cotton und Polyamid und vor allem
auch Leinen-Baumwoll-Mischungen. Nach wie vor überwiegen absolut
bügelfreie Hemden.]D1 [[Was [die Designs]CT anbetrifft], überwiegen bei der
Business-Line Unis, Faux-Unis und kleine Musterungen. Eine kleine
Stickerei auf der Brusttasche, der elegante Schnitt und vorwiegend klassische
Kragenformen. Kent-Kragen sind wieder im Kommen, «Button Down» ist
nach wie vor stark vertreten und vereinzelt wird das Bild abgerundet durch
den Haifischkragen.]D2
Just like the ‘third goal’ in the English example in (18) above, the topic ‘design’ is
both resumptive and contrastive. It is resumptive insofar as ‘design’ is one aspect
of the super-topic ‘trends in drapery’, and it is contrastive because it stands in a
paradigmatic relation to the other sub-topic (‘material’). Was … anbetrifft is not,
.â•… Here is an attempt at a translation: “[Grainy [fabric]CT is coming in, with light struc-
tures, blended cloth, and sometimes even artificial silk. It is light and soft throughout, pleasant
to touch and wear. Aside from pure cotton, blends of cotton and polyester, of cotton and
polyamide and, in particular, blends of linen and cotton are coming. Non-iron shirts are still
prevailing]D1. [As for [design]CT, unis, faux-unis and small patternings prevail in the busi-
ness line. A small embroidery on the breast pocket, an elegant cut and predominantly classic
collars. Kent collars are coming back, button down is still prominent and in singular cases a
shark collar rounds off the picture.]D2”
Volker Gast
A wealth of examples of this type has been assembled by Birner & Ward (1998).
Two of them are given in (22) and (23):
Contrastive topics and distributed foci as instances of sub-informativity 
(22) Humble they may be. But daft they ain’t. (Birner & Ward 1998: 46)
(23) I’ll have to introduce two principles. One I’m going to introduce now and one
I’m going to introduce later. (Birner & Ward 1998: 78)
All of the examples given above fit our definition of ‘sub-informativity’ in (8). For
instance, the sentence Mary I never saw is sub-informative insofar as it functions as
an answer to the question Did you see John and/or Mary? It thus seems that fronting
is indeed very closely associated with sub-informativity or, more specifically, con-
trastive topicalization. However, just like the lexical marker as for, fronting cannot
be regarded as a specialized expressive device for that function, as it is also used in
sentences that are not sub-informative. Pertinent examples are given in (24) and
(25), which do not require an open question in the discourse environment:
(24) I had two really good friends. Damon and Jimmy their names were.
(Huddleston & Pullum 2002: 1381)
(25) Did you want tea? Coffee I ordered. (Huddleston & Pullum 2002: 1381)
As has been shown by Birner & Ward (1998), the main conditions that must be
met for a constituent to be fronted is that this constituent has to function as a
‘link’16 between the sentence and the preceding discourse – in other words, it must
be given or at least accessible – and that there has to be a contrasting element in the
discourse environment. Note that ‘contrast’ is to be interpreted rather broadly in
this context. For instance, anything you don’t eat in (26) does not prima facie seem
to imply any type of contrast:
(26) Anything you don’t eat put back in the fridge.
(Huddleston & Pullum 2002: 1372)
However, an element of contrast is also recoverable in cases like (26). The sentence
implies that some of the food will be eaten, so anything you don’t eat contrasts
with everything you do eat. (26) thus answers the (superordinate) question What
am I supposed to do with the food only partially (note that the answer to one of the
sub-questions – the one about the food that has been eaten – is of course trivial:
Digest it!).
As the preceding discussion has shown, fronting in English cannot be regarded
as a grammatical device specialized for contrastive topicalization. However, given
the two conditions of use identified by Birner & Ward (1998) – (i) that the fronted
constituent must function as a link, and (ii) that there must be an element of
contrast – it comes as no surprise that fronting is typically found in instances of
.â•… Birner & Ward’s (1998) notion of ‘link’ is not to be equated with the (more specific) term
as used by Vallduví & Engdahl (1996).
Volker Gast
sub-informativity, as links are often (sentence) topics, and contrastive topics con-
stitute a major type of sub-informativity. In fact, cases like those in (24) and (25)
(where a focus has been fronted) seem to be rather rare, and only a handful of such
examples can be found in Birner & Ward (1998).
(29) A: Was ist mit Thomas? (‘What news are there concerning Thomas?’)
B: [Er]TOP hat jetzt eine neue Freundin.
he has now a new girl.friend
‘He has a new girlfriend now.’
The traditional picture in the literature on German syntax is that the Forefield
accommodates ‘given’ material, constituents that function as a ‘link’ or otherwise
‘prominent’ matter (cf. Lötscher 1984: 118). However, as is illustrated in (27) and
(28), the Forefield can also host constituents that do not carry any specific infor-
mation structural function, for instance expletives and specific types of adverbials.
Moreover, in information structurally neutral sentences it is typically the subject
that occupies the Forefield, without there being any particular pragmatic implica-
tions. Frey (2004, 2006) has therefore argued that a distinction needs to be made
Contrastive topics and distributed foci as instances of sub-informativity
between those cases where a constituent occupies the Forefield basically because
the grammar of German requires that this position must not be empty, and those
cases where moving a given constituent to the Forefield triggers specific informa-
tion structural effects (cf. also Fanselow 2002, 2004 and references cited there for
related discussion). He calls the former type of movement ‘Formal Movement’ and
the latter (true) ‘A′-movement’. In the case of Formal Movement, it is simply the
highest element in the Middle Field that is moved to the Forefield. In most cases,
this will be a subject, which is why sentences with subjects in the Forefield tend to
be information structurally neutral. However, when some more deeply embedded
constituent is moved to the Forefield, this is normally information structurally
meaningful.17 Such constituents are necessarily prosodically prominent (cf. also
Abraham 1997; Féry 2008 on the relationship between prosody/stress and syntax
in the German sentence). Mostly, this means that they contrast with another ele-
ment (cf. Frey 2006), or that they are associated with some type of scalar implica-
ture (cf. Frey 2008). This effect can be seen most clearly when an element from a
lower clause is moved to the Forefield (‘long movement’). (31) requires a sentence
in the immediate discourse environment in which Karl is replaced with some other
referential value, i.e. den Karl is necessarily contrastive:
(31) [Den Karl]i behauptete er ti gesehen zu haben.
.â•… The idea of a syntactic movement operation in German that is specialized for ‘contrast’
can also be found in Krifka (1998), who assumes that ‘Spec-CP movement’ of a focused phrase
results in ‘contrastive topicalization’ (recall from Section 3.1 that contrastive topics are regarded
as topics that carry a focus feature and are thus associated with a set of alternative values). This
point of view is certainly compatible with Frey’s (2004, 2006) proposal.
Volker Gast
contours, and that it is, in most cases, primarily these contours that are respon-
sible for notions such as ‘contrast’, ‘topicality’, ‘focality’, etc. and, therefore, for the
encoding of ‘sub-informativity’. This takes us to the third and, in the present
context, most important set of expressive devices, i.e. the prosodic ones.
Prosody takes up a particularly central position in the present context not only
because the relevant expressive devices are ubiquitous and usually accompany
other (lexical, syntactic) means, but also because there is a relatively clear-cut con-
trast between English and German, which can be summarized as follows:
There are differences in the level of generality at which sub-informativity is
prosodically marked as such: English treats it on a par with other instances of
the more general phenomenon of ‘incompleteness’, whereas German has an
intonational contour which is specialized for one type of sub-informativity, i.e.
context-changing sub-informativity.
In what follows we will deal with the ‘fall-rise contour’ of English and the ‘root
contour’ of German, both of which have been extensively discussed in the relevant
literature. We will start with the former contour in Section 7.1 and turn to the lat-
ter in Section 7.2. The contrasts between English and German will be summarized
in Section 7.3.
.╅ A contrastive study of the fall-rise accent in English can also be found in Hetland (2008),
who compares this accent with the Korean particle nun.
Contrastive topics and distributed foci as instances of sub-informativity
(33)
Bolinger (1986), in turn, introduces a ‘profile’ that he calls ‘AC’ (as it is a combina-
tion of the ‘A-profile’ [a fall to the base line] and the ‘C-profile’ [a rise towards the
base line]) and notes that “AC becomes a pretty good theme-marker regardless of
position” (Bolinger 1986: 321). He provides the examples in (34), where Cýnthia is a
theme in both initial (cf. (34a)) and final position (cf. (34b); see also Gussenhoven
1984: 30–1 for similar examples).
(34) a. Cynthia | they a \dored
b. They a \dored | Cynthia.
Ward & Hirschberg notice that Bolinger’s ‘AC-profile’ does not correspond to the
profile shown in (36) (and, hence, not to their ‘fall-rise contour’), but rather to the
one in (37), i.e. to a contour that Bolinger (1958) calls ‘A-rise contour’:19
(37) Ward & Hirschberg’s (1985) ‘A-rise contour’
Even though Ward & Hirschberg (1985: 752) claim that there is not only a phonetic
but also a functional difference between their ‘fall-rise contour’ and the ‘A-rise
contour’, it is pointed out by Pierrehumbert & Hirschberg (1990: 285) that the
tunes share certain phonological characteristics, in particular “a L phrase accent
and a H boundary tone”, and that they, accordingly, “share also a sense that the
current utterance will be completed by a subsequent utterance”.
The following discussion will be based on the ‘fall-rise’ of the British school,
i.e. the type of pitch movement shown in (33) above. Using the ToBI annotation
conventions, it can be represented as either H*L− H% or L+H*L− H% (cf. Steedman
1991 and ex. (35) above; see also Ladd 1996: 82). I will assume that it does not
make a difference whether the contour starts off with a peak accent (H*) or a rising
peak accent (L+H*).
.â•… D. Bolinger, in turn, disagrees on this point: “Bolinger (p.c.) identifies an ‘AC contour’ he
believes is the same as our FR [fall-rise]” (Ward & Hirschberg 1985: 750).
Contrastive topics and distributed foci as instances of sub-informativity
The fall-rise can also be used as a sole sentence accent. In that case it expresses
some kind of ‘reservation’, as in (42)–(44) (cf. also Wells 2006: 27–32):
(42) It’s cheap. (reservation: ‘but that’s not the only thing that’s true about it’)
(43) It looks expensive. (reservation: ‘but is it really?’)
(44) Well … (speaker signals that information is missing)
The feeling of an ‘implication’ as stated by Tench (cf. also Wells 2006) has also been
called ‘incompleteness’, ‘up-in-the-airness’ (Bolinger) and ‘uncertainty as to the releÂ�
vance of a speaker’s contribution’ (cf. Ward & Hirschberg 1985). In the following,
I will use the term ‘incompleteness’ to characterize the type of implicature triggered
Volker Gast
by the fall-rise contour. Note that the notion of ‘uncertainty as to the relevance
of a speaker’s contribution’ (Ward & Hirschberg 1985) can also be regarded as
an instance of ‘incompleteness’, as it implies that the speaker is aware that his/
her contribution (potentially) does not qualify as ‘complete’ with respect to its
contextual implications.
The assumption that the fall-rise is used to indicate ‘incompleteness’ can explain
why it tends to be used in combination with contrastive topics. In the present study,
‘contrastive topicality’ has been defined in terms of ‘sub-informativity’: the speaker
signals that s/he is aware that there are open questions in the discourse context. In
fact, the fall-rise can be used with each one of the three types of sub-informativity
distinguished above. It is used with distributed foci (cf. (45)), with context-
preserving topics (cf. (46)) and with context-changing topics (cf. (47); ‘||’ indicates
‘major breaks’, i.e. IP boundaries in terms of Beckman & Pierrehumbert 1986):20
(45) Distributed focus sentences
A: Who ate what?
B: Fred | ate the \beans …
(46) Context-preserving topics
A: What do your parents do?
B: My father | works on a \freight ship || my mother | is a \doctor.
(47) Context-changing topics
A: What is your daughter doing?
B: My younger daughter | studies \medicine ||
my older daughter | studies \law.
It is important to note that ‘sub-informativity’ as defined in (8) above is just one
instance of the more general notion of ‘incompleteness’, with other instantiations
of this notion being the triggering of (unspoken) implications or implicatures
(cf. (42)–(44) above; see also Wells 2006: 27–29 for a number of illuminating
examples). Note furthermore that the fall-rise can also be used on foci, as in the
following example from Hedberg & Sosa (2007: 118):
(48) … and I, frankly, think this guy is pretty attractive.
I don’t find him unat tractive.
As the preceding discussion has shown, the information structural category of sub-
informativity does not have a direct prosodic correlate in English. The fall-rise con-
tour has a more general function and is often used in contexts of sub-informativity,
.â•… Note that there is an important difference between the three instances of sub-informativity:
Only in context-changing topics (as in (47)) is the fall-rise obligatory. In (45) and (46) a
different type of intonation is also possible. In particular, the nuclear accent of the first
intonation phrase could simply be a rising accent.
Contrastive topics and distributed foci as instances of sub-informativity
but it is not restricted to these contexts. As will be argued in the next section, this
is different in German, where a rather specific tune (the root contour) is used for
a rather specific information structural function (context-changing topic-related
sub-informativity).
╇
Meine jüngere Tochter studiert Medizin …
my younger daughter studies medicine
It should be mentioned that the term ‘root contour’ is also often used for the first
type of pitch movement illustrated in (49) only, which can be described either
as a low accent (L*) or a scooped accent (L*+H) followed by a H phrase accent
(H–). I will assume that the first component of the root contour has the form
L*+H H–,21 and I will call it ‘low rise’, adopting the term of the British school of
intonation (cf. Ladd 1996: 82). In examples the low rise is represented as ‘√’. The
notation used in this study is summarized in (50):
.╅ Cf. Uhmann (1991), who regards L*+H as a contrastive topic accent.
Volker Gast
Another note of caution concerns the identification of the root contour as described
above in connected speech. The pattern tends to be simplified, thus resembling a
similar, but functionally different tone pattern (cf. the quotation from Krifka 1998
above; see also Jacobs 1997: 93 and Féry 1993: 149–50). In particular, the low rise
is often realized as a simple peak accent. In this case the root contour is phonetically
similar to the sequence of two H* accents, with the pitch remaining at a high level
in between, as in (51) (from Féry 1993: 149):
The terminology chosen in this study allows us to distinguish the ‘root contour’
shown in (49) from a ‘hat contour’ as shown in (51). Note that Féry (1993) calls
both contours ‘hat contour’, the one in (51) ‘hat contour 1’ and the one in (49)
‘hat contour 2’.
.â•… Note that (53) is probably better than (52), but a boundary signal such as Engl. well
would usually be rendered with a long falling tone in German. There is probably significant
idiolectal variation, however.
Contrastive topics and distributed foci as instances of sub-informativity 
Accordingly, the low rise – or the entire root contour – cannot be regarded as
a general indicator of ‘incompleteness’ or ‘uncertainty’. Even among instances of
sub-informativity, its distribution is severely restricted. Note first that the root
contour is inappropriate in combination with distributed foci, a point also made
by Jacobs (1997). Consider (54):
Jacobs (1997: 99) points out that “the b-version is at least unusual; it is associ-
ated with additional pragmatic components of meaning” [my translation]. More
commonly, the question in (54) would be answered as shown above or also as
in (55):
We may add that the original Jackendoff example (which is an instance of ‘focus-
related sub-informativity’ as well) also sounds awkward (or ‘hyper-informative’)
when translated into German and pronounced with a root contour (cf. (56)).
Possible intonations are given in (57).
(58) Was ist mit Hans und Maria? Was haben sie gelesen?
‘What about Maria and Hans. What did they read?’
a. Ma/ ria | hat den Schatz im \Silbersee gelesen,
Maria has det treasure in.the ╇↜渀屮silver.lake read
und / Hans | den \Winnetou.
and ╇↜渀屮Hans det ╇↜渀屮Winnetou.
b. #Ma√ria | hat den Schatz im \Silbersee gelesen,
Maria has det treasure in.the ╇↜渀屮silver.lake read
und √Hans | den \Winnetou.
and ╇╛╛Hans det ╇↜Winnetou.
‘Mary read The Treasure of the Silver Lake, and John Winnetou.’
In (59), speaker B ‘inserts’ a ‘move’ into the strategy (the question What does your
younger daughter study?), thus modifying the context, and the higher-level question
(What does your daughter study?) is accordingly split up into two sub-questions,
introducing two sub-topics in the process (my older daughter, my younger daugh-
ter). This is not the only type of ‘context-changing sub-informativity’ where the
root contour can be used. It is also appropriate when a speaker refuses to provide
information about one of the sub-topics introduced by the other interlocutor, thus
‘deleting’ a move from the strategy. Consider (60):
(60) Was für Romane von Karl May haben Hans und Maria gelesen?
‘What novels by Karl May did Hans and Maria read?’
.â•… Krifka’s explanation for the infelicity of (58b) differs slightly from mine. According to
Krifka, the main reason is that the criterion of ‘disputability’ is not met, whereas in my
explanation it is the aspect of ‘context modification’ that is relevant; cf. below.
Contrastive topics and distributed foci as instances of sub-informativity
With a falling accent at the end of the sentence, the speaker indicates that s/he is
not in a position, or not willing, to provide any information about Hans. Thus, one
of the two sub-questions raised by A – What did Hans read? – is simply deleted
from the strategy. Note that the root contour is not just appropriate in this type of
context, but virtually obligatory.
So far we have considered two instances of ‘context-changing sub-informa-
tivity’: One case in which a move is inserted into a strategy (‘move insertion’), and
one case in which a move is removed (‘move deletion’). In a third type of context,
‘move insertion’ as in (59) and ‘move deletion’ as in (60) are combined, and the
sole topic of the original strategy is shifted to a new one, which is a function of
the first. This is illustrated in (61). While the question ‘Have you seen Karl?’ is
removed from the strategy, the question ‘Have you seen Karl’s wife’ is inserted by
speaker B:
Finally, we may note that the root contour can also be used in combination with
two types of topics pointed out by Büring (1997), i.e. contrastive topics that are
not sub-topics but that are simply taken from the discourse environment (e.g. the
speaker, cf. (62)), and ‘purely implicational topics’, which trigger conversational
(or perhaps conventional) implicatures (cf. (63)).
incompleteness
… sub-informativity
focus-related topic-related
context-preserving context-changing
Diagram 4.╇ Types of sub-informativity and prosodic marking in English and German
Diagram 4 also captures another generalization that has emerged in the course of
the discussion. As has been seen, ‘context-changing sub-informativity’ is the only
context in which the German root contour is appropriate. This type of context also
plays an important role in English, as it seems to be the only context where the
fall-rise is virtually obligatory. All other instances of sub-informativity, as well as
the more general notion of ‘incompleteness’, can also be indicated by other intona-
tional patterns (e.g. a simple rising accent followed by a falling focus accent). The
obligatoriness of using a fall-rise in English was illustrated in (12) above, which is
here repeated in (64). Omitting the fall-rise accent on my older daughter would be
rather unusual here, though it seems to be less compelling on my younger daughter,
where a simple rising accent could also be used.
Contrastive topics and distributed foci as instances of sub-informativity
Another important contrast between English and German concerns the phono-
logical properties of the two contours. While the fall-rise in English is not fixed
with regard to its position relative to the (falling) focus accent, the components of
the root contour cannot change places. The ‘versatility’ of the English pattern was
pointed out, among others, by Bolinger (1986) with respect to the example in (65)
(cf. (34) above). In both examples, Cynthia functions as a topic, regardless of its
position in the sentence:
(65) a. Cynthia | they a \dored.
b. They a \dored | Cynthia.
The German example in (66) does not allow such a change of position:
(66) Was macht eigentlich deine Tochter?
‘What is your younger daughter doing?’
a. Meine √jüngere Tochter | studiert Medi \zin …
my ╇ ↜younger daughter studies medicine …
b. #Medi \zin | studiert meine √jüngere Tochter …
medicine studies my ╇ ↜younger daughter
What this illustrates is that the root contour is a ‘holistic’ pattern whereas the cor-
responding English sentences are made up of two independent pitch movements,
i.e. a fall-rise and a falling focus accent, each of them associated with specific
information structural functions.
8.â•… Conclusions
References
Abraham, W. 1997. Zur Basisstruktur des deutschen Satzes unter Berücksichtigung diskursfunk-
tionaler Erwägungen. In Sprache im Fokus. Festschrift für Heinz Vater zum 65. Geburtstag,
C. Dürscheid, K.H. Ramers & M. Schwarz (eds), 59–67. Tübingen: Niemeyer.
Beckman, M.E. & Pierrehumbert, J.B. 1986. Intonational Structure in Japanese and English.
Phonology Yearbook 3: 255–309.
Contrastive topics and distributed foci as instances of sub-informativity
Birner, B.J. & Ward, G. 1998. Information Status and Noncanonical Word Order in English.
Amsterdam: John Benjamins.
Bolinger, D. 1958. A theory of pitch accent in English. Word 14: 109–149.
Bolinger, D. 1986. Intonation and its Parts: Melody in Spoken English. London: Arnold.
Breul, C. 2008. On identifiability and definiteness in English and German: An example of
contrastive information structure analysis. Languages in Contrast 8(2): 263–285.
Burgschmidt, E. & Götz, D. 1974. Kontrastive Linguistik Deutsch/Englisch. München: Hueber.
Büring, D. 1994. Topic. In Focus and Natural Language Processing, vol. 2: Semantics, P. Bosch &
R. van der Sandt (eds), 217–280. Working Papers of the Institute for Logic and Linguistics.
Heidelberg: IBM Scientific Centre.
Büring, D. 1997. The Meaning of Topic and Focus. The 59th Street Bridge Accent. London:
Routledge.
Büring, D. 2003. On D-trees, beans and B-accents. Linguistics and Philosophy 26: 511–545.
Constant, N. 2007. English rise-fall-rise: A study in the semantics and pragmatics of intonation.
Ms., University of Santa Cruz, 48 pp. Available on the Semantics Archive.
Cruttenden, A. 1986. Intonation. Cambridge: CUP.
Doherty, M. 2005. Topic-worthiness in English and German. Linguistics 43(1): 181–206.
Erteschik-Shir, N. 2007. Information Structure: The Syntax-Discourse Interface. Oxford: OUP.
Fanselow, G. 2002. Quirky subjects and other specifiers. In More than Words – A Festschrift
for Dieter Wunderlich, I. Kaufmann, & B. Stiebels (eds), 227–250. Berlin: Akademie
Verlag.
Fanselow, G. 2004. Cyclic phonology-syntax-interaction: Movement to first position in German.
In Working Papers of the SFB 632 [Interdisciplinary Studies on Information Structure 1],
S. Ishihara, M. Schmitz & S. Schwarz (eds), 1–42. Potsdam: Universitätsverlag.
Féry, C. 1993. German Intonational Patterns. Tübingen: Niemeyer.
Féry, C. 2008. The prosody of topicalization. In On Information Structure, Meaning and Form:
Generalizations across Languages, K. Schwabe & S. Winkler (eds), 69–86. Amsterdam:
John Benjamins.
Frey, W. 1993. Syntaktische Bedingungen für die semantische Interpretation. Berlin: Akademie
Verlag.
Frey, W. 2004. The grammar-pragmatics interface and the German prefield. In Sprache und
Pragmatik: Arbeitsberichte 52: 1–39. Lund: University of Lund.
Frey, W. 2005. Pragmatic properties of certain German and English left peripheral constructions.
Linguistics 43: 89–129.
Frey, W. 2006. Contrast and movement to the German prefield. In The Architecture of Focus,
V. Molnár & S. Winkler (eds), 235–264. Berlin: Mouton de Gruyter.
Frey, W. 2008. On some differences in filling the prefields of some V2-languages. Presentation
given at the conference on ‘Contrastive Information Structure Analysis’, held at the University
of Wuppertal on March 18–19, 2008.
Gast, V. forthcoming. Contrastive analysis: Theories and methods. In Dictionaries of Linguistics
and Communication Science: Linguistic Theory and Methodology, B. Kortmann & J. Kabatek
(eds), Berlin: Mouton de Gruyter.
Givón, T. 2001. Syntax, Vol. 2. Amsterdam: John Benjamins.
Gundel, J.K. 1974. The Role of Topic and Comment in Linguistic Theory. Ph.D. dissertation,
University of Texas at Austin.
Gussenhoven, C. 1984. On the Grammar and Semantics of Sentence Accents. Dordrecht.
Foris.
Volker Gast
Luis López
University of Illinois at Chicago
This chapter argues that Catalan Clitic Right Dislocation and English
deaccenting package different types of information (contra Vallduví 1992,
Vallduví & Engdahl 1996), only the former being a true discourse anaphor. This
chapter further hypothesizes the following generalization: A language in which
stress is displaced to express givenness is a language in which stress assignment
is sensitive to syntax, while in a language in which stress assignment is a purely
linear phenomenon stress cannot shift. A model of the syntax-phonology
interface is sketched in which this generalization follows from the location of
stress in the grammatical architecture.
1.â•… Introduction1
This definition requires explaining what ∃-type shifting and existential F-closure
mean. ∃-type shifting allows us to turn a DP into an expression of type 〈t〉 by having
.â•… I would like to thank Edward Göbbel, Carsten Breul, Michael Rochemont and an anony-
mous reviewer for detailed comments on an earlier draft that led to substantial improvements.
If only I had been able to incorporate more of their suggestions, this chapter would have been
much better. I would also like to thank the participants of the Tübingen workshop “Focus
and Freezing” (summer 2009), as well as Susanne Winkler for her kind invitation. Finally, it is
with pleasure that I acknowledge the long-standing support of the Alexander von Humboldt
Foundation, without which this work would not have come to fruition. All the mistakes and
shortcomings to be found in this chapter are my sole responsibility.
Luis López
unfilled arguments be bound by the existential quantifier. For instance, the phrase
green apple becomes: ∃x(green-apple(x)) (Schwarzschild 1999: 147).
The “F” in “existential F-closure” refers to Selkirk’s F feature. It is assigned
according to the following rules:
(2) F-Assignment Rules (Selkirk 1995: 555)
An accented word is F-marked
F-marking of the head of a phrase licenses F-marking of the phrase.
F-marking of an internal argument of a head licenses the F marking of the head.
For instance, a regular DP like a red apple must find a co-referent DP in the previous
discourse in order to count as given. But take the NP red apple in a context in
which red is F-marked (i.e. contrasted with, say, green apple). In this case, all
that is required for a term to count as an antecedent of red apple is that it entails
∃y(apple(y)).
Although this concept may be sufficient to account for English phenomena –
most notably, sentence stress distribution – a contrastive perspective shows that
the grammars of some languages require from us to make a distinction between
what I call accidental givenness and a mandatory antecedent-anaphor relationship.
The difference can be shown in the following pair of examples:
(4) [Context: I’m wearing a red coat. What are you wearing?]
A1 I’m wearing a blue coat.
A2 I’m wearing a blue shirt.
(5) [Context: What kind of coat are you wearing?]
A1 I’m wearing a blue coat.
A2 #I’m wearing a blue shirt.
In these and all the examples in this chapter an underlined word includes the
most prominent syllable in the sentence while italics indicate reduced prominence
or deaccenting.
In both (4A1) and (5A1) the Givenness calculus provided by Schwarzschild
gives the same result: coat is given. This has a phonetic correlate: coat is deaccented
in both examples. However, the focus structure is different: In (4), the whole direct
object is focus, since it is the answer to the wh-question. In (5), only blue qualifies as
focus because only blue answers the wh-question. The difference in focus structure
Givenness and discourse anaphors
leads to another difference. In (4), where the given coat is also part of the focus,
the fact that coat is given is accidental. As shown in (4A2), the answer could have
been about a shirt rather than a coat. In (5), where coat is not part of the focus, the
answer has to be about a coat, which makes (5A2) infelicitous.
Thus, in English, it seems that deaccenting disregards focus and reflects only
which constituents are given. Catalan works in the opposite direction: it is sensitive
to what constituents are focus or not, unfocused constituents are obligatorily
dislocated. This is exemplified in (6) and (7):
(6) [Context: I’m wearing a red coat. What are you wearing?]
A1 Porto un abric blau.
wear.1st a coat blue
‘I’m wearing a blue coat.’
A2 Porto una camisa blava.
wear.1st a shirt blue
‘I’m wearing a blue shirt.’
A3 #‘En porto un de blau, d’abric.
cl wear.1st a of blue of ’coat
‘I’m wearing a blue coat.’
(7) [Context: What kind of coat are you wearing?]
A1 En porto un de blau, d’abric.
cl wear.1st a of blue of ’coat
‘I’m wearing a blue coat.’
A2 #Porto una camisa blava.
wear.1st a shirt blue
‘I’m wearing a blue shirt.’
In Example (7), the focus is blue, as in (5). Since abric is not part of the focus, it
must be dislocated.2 The usual term for this phenomenon is Clitic Right Dislocation
(CLRD), a syntactic construction in which a constituent is moved to a right-peripheral
position and resumed by a clitic (unless the dislocated constituent is nominative, for
which there is no clitic available). In (7A1), the dislocated constituent is l’abric and it
is resumed by the clitic en, a partitive clitic.
In Example (6), abric must stay in situ because it is part of the focus. The fact
that it is given does not lead to dislocation or to any noticeable change in sentence
stress. Hence the infelicity of (6A3).
.â•… The verb and its functional projections are also outside of the focus but they cannot be
dislocated for independent reasons. In López (2009a) it is argued that the verb is outside the
scope of the rules that assign information structure features such as “anaphor” or “contrast”.
Luis López
The constituents outside the focus are discourse anaphors. I define a discourse-
anaphor as follows:
(8) Definition of Discourse-Anaphor
A discourse-anaphor obligatorily seeks an antecedent.
The first section of Vallduví and Engdahl (1996) discusses alternative theories that
aim to divide the sentence into two complementary informational segments. They
are the usual suspects: topic-comment, theme-rheme, focus-presupposition, etc.
After showing their different inadequacies, the discussion leads up to Vallduví↜’s
famous double partition: a proposition can be split into a focus and a ground. The
ground can itself be split into a link and a tail. These concepts are defined in relation
to their potential to update the information stored in the mind of the hearer. The
focus provides an update to the information stored in the hearer while the content
of the ground is already “subsumed by the input information state” (Vallduví &
Engdahl 1996: 469). Since the ground provides no new information, its role is
that of “ushering” the focus so that it is stored in the proper place in the hearer’s
memory. The file-card metaphor provides a simple method of representing these
relations (file-cards come from Heim 1982 and were adapted to the study of
information structure by Vallduví 1992 and Erteschik-Shir 1997). The ground
Givenness and discourse anaphors
directs us to the proper card, the focus adds information to the card. Let me show
it with an Example. (9a) is in English, (9b) is in Catalan:
(9) [Context: What did John do?]
a. John ate an apple.
b. El Joan es va menjar una poma.
the Joan cl past eat.inf an apple
As is common practice, I set up a context with a wh-question, which allows us to
delimit the focus/ground structure of the answer. In the sentence John ate an apple,
ate an apple is the focus, the supplier of new information. John is the ground. Using
the file card metaphor, the speaker utters the word John so that the hearer can find the
card ‘John’ in her/his memory, retrieve it and add the predicate ‘ate an apple’ to it:
(10) Card1 = John: λx. John did (x)
Update1:
Card1 = John: λx. John did (x)(eat an apple) = John ate an apple
According to Vallduví (1992), a link simply directs the focus to update the
information in the card. John/el Joan in (9) would exemplify a link. A comple-
ment can also be a link. In Catalan, a complement link involves displacement
to the left and clitic resumption; this is what is usually referred to as Clitic Left
Dislocation (CLLD). In English, a link may involve topicalization or it may stay
in situ (more on this below):
(11) [Context: Does John eat fruit?]
a. Apples he does eat.
b. Pomes, sí en menja.
apples Emph cl eats
A tail involves a more complex operation: An update introduced by a tail is meant
to dislodge an erroneous proposition in the card. The following is an example:3
(12) [Context: So, John drank the apple?]
a. No, John ate the apple.
b. No, el Joan se la va menjar, la poma.
No the Joan cl cl.acc past eat.inf the apple
Consider first (12a). In (12a), the tail is the deaccented constituent the apple. The
deaccenting of the apple tells us that the focus, ate, is displacing some other piece of
information – in this case, drank. In (12b), the tail is the CLRDed constituent la poma.
.â•… Right dislocated constituents are all deaccented, although I do not indicate it in the examples
(see Zubizarreta 1998; Frascarelli 2000; Feldhausen 2006).
Luis López
The effect of having a tail in the sentence is that the focus becomes contrastive
(although this is not how Vallduví worded it).
The Example (12) is useful to introduce an additional point of interest. On the
basis of parallel examples like this one, Vallduví and Engdahl conclude that English
deaccenting and Catalan right dislocation are informationally equivalent – they
package the same information by different means. Empirical evidence presented in
this chapter leads to a different conclusion, that English deaccenting and Catalan
dislocation are not equivalent.
Vallduví↜’s conception of link and tail has been substantially revised by later
research on information structure based on Catalan (see Villalba 2000; López
2003, 2009a). As a result of this scrutiny, Catalan dislocations are analyzed as con-
stituents that become discourse anaphors by virtue of their syntactic derivation
and they are distinguished only by the types of relationships that they have with
their antecedent. CLLDs are related to their antecedents by means of a variety of
relations (co-reference, set/subset, set/member etc.) while CLRDs can only relate
by co-reference. The following example shows that a CLRD does not necessarily
turn the focus into a constrastive focus:
(13) [Context: So, I understand you like whiskey]
a. I certainly do, like whiskey.
b. Ja ho crec, que m’agrada, el whiskey.
Already cl.neut believe.1st that cl.dat like the whiskey
‘I certainly like whiskey.’
The constituent like whiskey is a tail in Vallduví’s framework, but notice that it does
not involve contrastive focus but rather the opposite, a confirmation of the hearer’s
assumptions. Notice that the VP like whiskey is a discourse anaphor, in the way
defined in Section 1. Answers like (14a), (14b) and (14c) are infelicitous:
(14) [Context: Do you like whiskey?]
a. #I certainly do, hate whiskey.
b. #I certainly do, like brandy.
c. #No, I don’t, hate whiskey/hate brandy.
To sum up: According to Vallduví (1992), the Catalan sentence can be divided into
a focus and a ground, the latter a discourse anaphor. The two types of grounds,
links and tails, are distinguished by the type of relationship to their antecedents.4
.â•… I believe that a deeper criticism of Vallduví’s system is long overdue, but beyond the limits
of this article. For instance, recall that the function of links and tails is to usher the focus to
the right file card. However, in a typical example like (9) Juan has just been mentioned and
therefore it is uppermost in the hearer’s awareness. If so, why should the focus need to be
ushered to the most prominent card in the file? It seems to me that all functional approaches
Givenness and discourse anaphors
Lex
In other words, focus, tail and link are meant to be universal categories, derived
from the way the human mind stores and updates information. Language variation
is restricted to how information structure is linguistically realized – hence the title of
Vallduví and Engdahl’s article. In the following section I show that, whether or not
notions such as focus and ground are universal categories, language variation in
this realm is more complicated than what Vallduví and Engdahl envisioned. In par-
ticular, I show empirical evidence that English deaccenting and Catalan dislocation
do not index the same information categories.
to information structure are victims of the same circularity. It is for this reason that I take a
more descriptive-formalist approach, as will become clear in the next few pages: Constituents
in certain positions are discourse anaphoric, others are not.
Luis López
The phrases els ganivets and les forquilles are links, and as such they must be
removed to the left periphery with a resumptive clitic pronoun. The infelicity of
(16b) shows that a link cannot stay in situ in Catalan.
The following sentence exemplifies a Catalan tail:
(17) [Context: Will John drink beer?]
Clar que sí. Al Joan li agrada, la cervesa.
Of course dat.the Joan cl.dat likes the beer
‘Of course. John likes beer’ or, maybe, ‘John likes beer.’
In (14), la cervesa is a tail. As a tail, it is obligatorily dislocated to the right (although,
being a nominative argument, there is no clitic). Thus, the packaging of information
in Catalan follows a rather strict scheme:
(18) CLLD [TP …] CLRD
Recent accounts of the relationship between sentence stress and focus claim that
movement of what we are here calling tails is as a matter of fact prosodically
motivated. The claim is, in essence, that instances of displacement that we con-
nect with information structure actually take place to ensure that focus and
sentence stress converge on the same constituent (see Zubizarreta 1998 and, for
a critique, López 2009a).
Catalan dislocation is not amenable to this account. Consider (19), a ditransitive
sentence. It is an all-focus sentence, with neutral word order and stress on the last
constituent (as is the rule in Catalan, more on this below):
(19) La Joana va entregar un paquet al Pere.
the Joana past deliver a package to.the Pere
‘Joana delivered a package to Pere.’
In (20), we place the sentence in a context that forces un paquet to be a tail. Notice
that main sentence stress still falls on Pere, so prosodically there is no reason for el
paquet to be dislocated. But it is nonetheless:
(20) [Context: What happened to the package?]
La Joana el va entregar al Pere, el paquet.
the Joana cl.acc past deliver to.the Pere the package
.â•… In a discussion of (21f), an anonymous reviewer doubts that John bought can be regarded
as a focus, pointing out that it is not even a constituent. He suggests that the whole sentence
could be a focus and the books would only be given. However, I think there are good reasons
to assume the books is not part of the focus. First, with the context that I have provided, no
other constituent could make the sentence felicitous (see the discussion surrounding (4) and
(5)). Second, the Catalan equivalent would have the books dislocated.
Luis López
As shown in (23), the B accent may be associated to the lower or the higher copy.
The one that does not get the B accent gets deleted. This might seem a flight of
fancy, but let us recall that one uncontroversial property of English is its flexibility
with the place of accent and intonation contours, as shown above. Thus, this solution
simply extends this property to a new realm.6
Still following Vallduví and Engdahl’s exposition, tails in English are indexed
by means of deaccenting:
(24) [Context: Does John like beer?]
John [F loves] [T beer].
To sum up: According to Vallduví and Engdahl, the concepts focus, ground, link
and tail underlie the information structure of English and Catalan. These two lan-
guages are only different in the way these pragmatic functions are expressed: Links
and tails involve special intonations in English and dislocations in Catalan. Focus
is expressed by sentence accent in both languages.
In the rest of this article, I focus on tails.
As it turns out, the set of things that can be deaccented in English does not match
the set of things that can be CLRDed in Catalan. The crucial observation that
can be gleaned from some of the examples provided by the English literature is
that constituents contained within the focus can be deaccented – rather, they must
be deaccented if the context requires it. Therefore, English deaccenting cannot
(always) be a tail in the Vallduvian sense.
I start with a simple example:7
(25) [Context: Mary drove her blue convertible. What did John drive?]
He drove [F her blue sedan]. // He drove [F a red sedan].
.â•… However, the assumption that there is a direct correlation between an information function
like topic or focus and a prosodic contour has recently been challenged by Hedberg and Sosa
(2007), an empirical study of naturally occurring speech. My English language consultants
agree that the L+H*L–H% accent is not obligatory for links. What seems to be true is that
contrasts, topics etc. need to stand out prosodically, but it does not seem to be true that they
are linked to a particular intonation. My proposal holds the same as long as links require some
sort of melody to provide intonational relief against the rest of the utterance.
.â•… Some of the data discussed in this section are lifted from López (2009a). However, in this
work no analysis is provided.
Givenness and discourse anaphors 
The focus structure of the answer, indicated with the brackets and the subindex F
is the update in the hearer’s information store. The context tells us that the hearer
has the following cards in her archive:
(26) Card1 = Mary: Mary drove her blue convertible
Card2 = John: John λx.drove (x)
The sentence stress in this example falls on the word sedan, as is expected according
to any theory of sentence stress in English that I am aware of.
Consider now the following example (from Schwarzschild 1999):
(28) [Context: Mary drove her blue convertible. What did John drive?]
He drove [F her red convertible].
As indicated, the phrase her red convertible is the focus of the sentence, as required
by the wh-question in the preceding question. The expression her red convertible is
the update to the hearer’s file archive:
(29) Card1 = Mary: Mary drove her blue convertible
Card2 = John: John λx.drove (x)
Update2:
Card2 = John: John λx.drove (x) (her red convertible) = John drove her
red convertible
.â•… English Right Dislocation does appear to be a genuine tail with the information structure
property of being a discourse-anaphor (Birner & Ward 1998). However, English RD is,
Luis López
Consider Example (31). In this example I have set things up so that convertible
is necessarily a discourse-anaphor. Interestingly, the intonation structure is identical
to that of (28), despite the fact that the focus structure is substantially different.
A file-card style representation of the focus structure of (31) is in (32):
(31) [Context: Mary drove her blue convertible. What kind of convertible did
John drive?]
He drove a [F red] convertible.
(32) Card1 = Mary: Mary drove her blue convertible
Card2 = John: John drove λx.convertible (x)
Update2:
Card2 = John: John drove λx.convertible (x) (red) = John drove her
red convertible
Let us now turn to Catalan. As mentioned, we find that in this language the syntax
described by Vallduví does reflect information structure: A CLRDed constituent
is invariably a tail. If a constituent is part of the focus it remains in situ and is not
deaccented, even if it happens to be given.
Consider first (33). This is a direct translation of (28) above. The word descapotable
is given but it remains in situ.
(33) [Context: Mary drove her blue convertible. What did John drive?]
Va conduir [F el seu descapotable vermell].
past drive.inf the her/his convertible red
‘He drove his red convertible.’
The following, with identical focus/ground structure, also have identical syntactic
structure:
(34) [Context: Mary drove her blue convertible. What did John drive?]
a. Va conduir [F el seu sedan blau].
past drive.inf the her/his sedan blue
‘He drove his blue sedan.’
b. Va conduir [F un sedan blau].
past drive.inf a sedan blue
I believe, beyond the limits of this article. English RD, unlike Catalan CLRD, is not derived
syntactically, i.e. it does not establish a syntactic dependency with a position within the core
clause. It should be regarded as a Hanging Topic (HT), a constituent that links up to the
discourse without being a constituent of any clause in particular – an orphan, following the
terminology in Shaer and Frey (2004). For extensive discussion of HTs – albeit with a data
base drawn from French – see DeCat (2007).
Givenness and discourse anaphors
.â•… Since CLRD is generally the output of a movement rule in Catalan, I assume that (36) is
also derived by movement. I take it that de is a case marker. The fact that it is repeated should
not surprise us. Case markers are also repeated in other instances of movement that strands
DP constituents, as shown in the following stranded quantifier example:
(i) Als nens, els hi vaig donar a tots un encàrrec.
dat.the children cl cl past.1st give.inf dat all an errand
‘I gave the children all an errand.’
Luis López
We saw the same phenomenon in (20) above, where the direct object is dislocated
even if the stress falls on the indirect object in any case.
Still, to further reassure the reader that there is no connection between stress
and focus in Catalan I design an example in which the adjective precedes the noun.
This is obligatory with so-called intensional adjectives like pretès ‘alleged’:
(37) [Context: John is the real murderer. And Peter? What is Peter?]
a. El Pere es [F el pretès assassí].
the Pere is the alleged murderer
b. Peter is the alleged murderer.
Thus, assassí receives the nuclear stress even though it is given. The contrast with
the English sentence (37b) is stark.
Conclusion: Catalan CLRDs are a phenomenon of information structure,
since CLRDs are directly linked to how information is stored and updated in the
discourse model. English deaccenting is a broader phenomenon, it is not sensitive
to the role that a constituent plays in the construction of a discourse but simply
to whether a constituent is given or not. Thus, English deaccenting can affect a
segment of the focus whereas a Catalan dislocation cannot affect a segment of the
focus felicitously. The concepts of “givenness” and “discourse anaphor” need to
be separated.
Finally, I would like to present one more example that will lay out the empirical
problem in sharper lines and bring in one more piece of data to the data-pool. The
following examples combine a given constituent that is part of the focus followed
by a bona-fide tail as Vallduví would define it:
(38) [Context: Mary gave a red diamond to a politician. And Peter? What
did he give?]
a. Peter gave [F a blue diamond] to a politician.
b. El Pere li va donar [F un diamant blau], a un polític.
In both (38a) and (38b), a blue diamond/un diamant blau is the focus, while to a
politician/a un polític is the tail. In English, the sentence stress falls on blue and the
rest is deaccented, without teasing apart the given focus part from the tail. This
is of course not surprising. A given constituent is simply a repeated constituent
(modulo existential F-closure, etc). A tail is an anaphor. The set of anaphors is
properly contained within the set of given constituents. Thus, English is simply
not sensitive at the level of sentence grammar to the notion of discourse-anaphors,
rather it is sensitive to the more encompassing notion of givenness. Catalan, instead,
is sensitive to the narrower notion.
In case the data were not intricate enough, I give it another turn of the screw.
On the basis of the above examples and discussion, one could conclude that the
Givenness and discourse anaphors
notion of tail should simply be subsumed under the broader notion of givenness,
language variation would be limited to the extent that givenness is allowed to play
out. But although anaphors are, of course, given, givenness and anaphoricity should
be kept distinct. It is possible for a language to express, within sentence grammar,
both discourse anaphor and givenness. Consider the following German example, a
translation of the previous example except for the addendum of an adverb:
(39) [Context: Mary gave a red diamond to a politician. And Peter?
What did he give?]
Peter hat einem Politiker unerwartet einen blauen
Peter has a politician unexpectedly a blue
Diamanten gegeben.
diamond given
In this example, we can see that the discourse-anaphor einem Politiker is scrambled,
as shown by the position of the adverb unerwartet between the two complements.
Scrambling is customary in German for discourse anaphors – if the sentence were
pronounced without the preceding context, the indirect object would be found to
the right of the adverb. Interestingly, the given constituent, Diamanten, is deac-
cented. German and English must have a property in common that Catalan does
not have that allows them to express givenness by means of deaccenting. This
property does not conflict with the ability to express discourse anaphoricity.
So far, we have identified this difference between Catalan and English: While
Catalan grammar marks the focus/ground (discourse-anaphor) sharply, English
only expresses if something is given or not. In this section I discuss another
difference between Catalan and English: Catalan stress is linearly defined while
English stress is syntactically defined – which, I argue, reveals a minimal but
wide-ranging difference in grammatical architectures.
Taking the pioneering work of Selkirk (1984) as a basis, the following generaliz�
ations of stress in English can be assumed – and I leave aside here some irregular
cases. The following discussion assumes an out-of-the-blue context:
Generalizations 1 and 2 can also be seen in the DP. The following examples
were first noted by Bresnan (1972):
(43) a. Mary liked the proposal that John leave.
b. Mary liked the proposal that John left.
These data naturally lead to the conclusion that stress in English is assigned to
constituents according to syntactic configuration – see Selkirk (1984), (1995),
Cinque (1993), Zubizarreta (1998), Arregi (2002) for alternative analyses.11
Cinque’s (1993) seminal article unifies (40)–(46) into one compelling generalization.
Under an analysis of heads, specifiers, adjuncts and complements based on
classical X′-theory, Cinque’s main insight is that stress falls on the most deeply
embedded constituent:
.╅ I base this assertion on my consultations of native speakers as well as Selkirk (1984,
1995). Göbbel (p.c.) points out that the lesser prominence of adjuncts is not always crystal
clear for all classes of adverbs. Indeed, some examples in the literature place sentential promi-
nence on an adjunct – unfortunately, without discussion. Pending a systematic analysis of all
relevant cases, I assume the essential correctness of Selkirk’s conclusions.
.â•… But see also Féry and Samek-Lodovici (2006), who analyze English sentential stress as
falling on the rightmost lexical item of the intonational phrase, unfortunately without engaging
the data presented in earlier work and summarized in this Chapter in (40) to (46).
Givenness and discourse anaphors
(47) XP
YP X′
X′ ZP
X WP
Arregi’s NSR ensures that heads are less prominent than complements and more
prominent than anything else – assuming that, whenever there is a specifier or
adjunct, there is always an overt or covert complement (Chomsky 1995). An
additional assumption is that sentence stress in unaccusative predicates must
take the copy of the overt constituent into consideration. One advantage of
Arregi’s proposal is that it makes no use of X’-conventions, and therefore it
is easily transferable to different theories. Although Arregi’s formula does not
explain the difference between complements and specifiers, it does provide us
with a testable generalization.12
.╅ An anonymous reviewer points out that complex heads such as (i) seem to create a
problem for the SNSR:
(i) [v v+V]
In this sort of case, we have a structure γ in which both components are branching. I assume
that the rules of stress assignment are not able to look into the structure of a word (we are, as
usual, abstracting away from contrastive focus).
Luis López
Let us now go back to the connection between givenness and sentence stress. As
hinted at above, a given constituent rejects stress, even if it should get it according
to SNSR:
(49) [Context: What happened to the book?]
John burnt the book.
In order to account for these data, I propose the following simple constraint (see
Féry and Samek-Lodovici 2006):
(50) Given Non-prominence Rule (GNR)
In a structure of the form [γ α β] (order irrelevant) α is less prominent
than β if α is given.
Finally, all we have left to do is set up the system so that the SNSR acts as a default
rule that applies in contexts where the more specific GNR is trivially satisfied.13
Following current preferences, I do this by constraint ranking. Indeed, ranking
GNR “above” SNSR ensures that prominence is distributed in a descriptively ade-
quate manner. This is shown in the simple tableaux in (51) and (52). I indicate
givenness by means of a superindexed [g]:14
(51) GNR SNSR
John bought the book. √ √
John bought the book. √ *
(52) GNR SNSR
John bought [the book][g] * √
John bought [the book][g] √ *
.╅ And, I would add, contexts where rules of contrastive focus are also trivially satisfied.
Contrastive focus is not discussed in this chapter.
.╅ Notice that the candidates with a [g] feature and those without cannot be in the same
tableau or we would never have a winning candidate with a [g] feature. This is a property
of all competition-based models of focus and stress assignment (such as Szendröi 2002,
Samek-Lodovici 2005) although, as far as I know, this has never been explicitly discussed. The
question of what constitutes the input for competition is a thorny issue. For instance, Samek-
Lodovici (2005) claims that the input consists of argument structures (following Grimshaw
1997). But his analyses include inputs with the feature [focus], which is not part of argument
structure. A broader view of input is proposed in Broekhuis and Klooster (2001): Competing
candidates should have the same meaning (in a truth conditional sense). But the feature [g]
is not related to meaning. As the discussion in this chapter shows, [g] is not even related to
information structure in a clear way, but it must be present for stress assignment. At this point,
I limit myself to pointing out the problem, leaving its resolution for future research.
Givenness and discourse anaphors
Let us now turn to Catalan. In this language, the nuclear stress always falls on
the last constituent of the intonational unit, with complete disregard for syntactic
structure:
As we saw above, givenness does not affect the placement of accent in Catalan:
(55) [Context: John is the real murderer. And Peter? What is Peter?]
a. El Pere es [F el pretès assassí].
the Pere is the alleged murderer
b. Peter is the alleged murderer.
Nothing I have said so far prevents this. That is, nothing prevents a constraint like
the GNR to be operative in Catalan and alter the place of accent in (53) or (55a). As
a matter of fact, this is a question that has never, as far as I know, been asked: Why
is it that a language with syntactic stress can shift stress to avoid making a given
constituent prominent while a language with linear stress cannot? The model that I
present in the following sections provides an answer to this question, in effect, my
.â•… Zubizarreta (1998) claims that in Spanish – which is like Catalan in this respect – sentence
stress is also assigned configurationally, by means of the Structural-NSR. This assumption comes
as a consequence of a radical adoption of Kayne (1994), which forces the left to right order to
entail a c-command relation. Zubizarreta makes generous use of remnant movement in order
to maintain the tenets of Kayne’s theory.
Luis López
model will predict that if sentence stress is assigned by the LNSR givenness cannot
shift the position of the stress.
.╅ As Michael Rochemont (p.c.) points out, the model in (56) makes the construction of
prosody independent of stress placement. I take it that this is indeed the case and prosodic
boundaries are dependent exclusively on syntactic structure (as in Truckenbrodt’s (1999)
Wrap and Align constraints). Exploring the consequences of this claim go beyond the limits
of this chapter.
Givenness and discourse anaphors 
Cs in English Cs in Catalan
I claim that this difference between English and Catalan explains why the former
language is able to use stress to convey givenness while the latter is not. When Cs
applies in English, we have a full-fledged syntactic structure augmented with
features related to its connectedness to discourse. When Cs applies in Catalan, all it
has to play with is a linearized prosodic structure, from which syntactic configura-
tions and their features are gone. In the following section I present a model of the
syntax-discourse-phonology interface that expresses this difference.
In this section I would like to develop the model sketched in the previous section,
adding a discourse component to it to integrate the focus/ground structure and
givenness. I should hasten to add that this is probably not the only way that (56)
could be implemented – but it seems to me, at this stage of my research, to be a
plausible one.
I take a discourse model to be made up of Discourse Representation Structures
(DRS) (à la Kamp & Reyle 1993), which are themselves nothing but syntactic
structures annotated to express the relation of these structures to the previous
discourse.
The mapping operations follow the following steps, listed as a–e:
(57) vP
XP[+a] v′
EA v′
v VP[Foc]
V t(XP)
Call ∑[p] the syntactic structure augmented with information structure features.
c. Integrate ∑[p] into a DRS. XP[+a] links up with an antecedent.
(58) DRS
∑1 : I’m wearing a red coat. What kind of coat are you wearing?
∑2[eng] : I’m wearing a blue coat
∑2[cat] : Porto un de blau, [[+a] d’abric ]
Since English does not seem to be sensitive to the feature [+a], I assume it is
not present in the English DRS.
d. As a result of integration, some constituents will be marked as [given], simply
indicating that they are in this type of relationship with another constituent
within the same DRS:
(59) DRS
∑1 : I’m wearing a red coat. What kind of coat are you wearing?
∑2[eng] : I[g]’m[g] wearing[g] a blue coat[g]
∑2[cat] : Porto[g] un de blau, [[+a] d’abric]
.╅ Thus, being a discourse anaphor is the result of a syntactic rule while being given is not.
I think this reflects in an intuitive manner the fact that the former is mandatory while the
latter is accidental.
Givenness and discourse anaphors
Catalan: e. PL-computation
f. Stress assignment
English: e. Stress assignment
f.PL-computation
Thus, when English assigns stress, the feature [g] is present. Therefore a con-
straint like the GNR is possible and stress is shifted to avoid hitting on a given
constituent. In Catalan, stress assignment applies to a linearized intonation
unit – there is no ∑[p][g] anymore and therefore a stress constraint that refers to
any part of it cannot be formulated.
Thus, the above model predicts that a language with a stress system based on linear
order will not express focus/background or givenness by means of stress.
Moreover, notice that this analysis does not leave German out of the loop.
Recall the example above:
(60) [Context: Mary gave a red diamond to a politician. And Peter?
What did he give?]
Peter hat einem Politiker unerwartet einen blauen Diamanten gegeben.
Peter has a politician unexpectedly a blue diamond given
.â•… Why doesn’t English scramble? That is, why is the following not grammatical?:
(i) [Context: What did you do with the blue coat?]
*I have the blue coat sold.
I do not have an answer to this question, but I may suggest an idea as to what kind of solution
is more plausible. If movement takes place freely and bad results are filtered out at the inter-
faces, as in Kučerová’s (2007) proposals, then it is not possible to explain why English cannot
scramble like German. However, a feature-based approach to movement could simply suggest
that the crucial trigger is missing in English. Arguably, positing invisible features to trigger
movement is, for the time being, unenlightening. However, the alternative is unable to begin
to address the question of language variation and thus it is in a disadvantage.
Luis López
8.â•… Conclusion
I have shown that English deaccenting does not bear the same information import
as Catalan CLRD. CLRDed constituents are discourse-anaphoric. Deaccenting is
only a reflex that a constituent has been mentioned in the previous discourse or its
contents are entailed by its F-existential closure. In German, deaccenting indexes
givenness while a scrambled constituent is discourse-anaphoric.
Catalan sentence stress is insensitive to givenness or information structure
generally. I have related this difference between English and Catalan to the fact
that stress in English is sensitive to syntax while in Catalan it is sensitive exclusively
to linear order. Finally, I have concluded that the stress difference reveals a timing
difference in the operations that take place at the syntax-PF interface.
References
Knud Lambrecht
University of Texas at Austin
There is remarkable syntactic and prosodic similarity in the way the different
languages express this focus type, at least in the given discourse context. In all
cases the initial subject expresses the topic of the sentence, about which the follow-
ing predicate expresses a comment. In the given context, the subject can be either
lexical or pronominal, or it can be null-instantiated, as in Spanish and Italian. In
spoken French, there is a clear preference for the topic NP to be left-dislocated
(hence not to be the subject). Prosodically, the common feature is the presence of
a nuclear accent at the end of the sentence and of a secondary accent on the subject
or topic NP, when it is lexical.
The second type of focus structure is the argument focus articulation, also
called ‘focus-presupposition’, ‘specificational’, ‘identificational’, or ‘contrastive’
articulation.1 Here the pragmatic assertion consists in providing the missing
argument in a pragmatically presupposed open proposition. In other words, the
assertion adds an argument to a given (incomplete) predication:
It is easy to see that in the case of the argument-focus articulation there is much
greater syntactic and prosodic diversity among the different languages than in the
predicate-focus type in (1). In English and German, the syntax of the sentence
is the same as in (1), but the nuclear accent is now on the subject instead of the
.â•… The label ‘argument focus’ is somewhat misleading as the term ‘argument’ is now almost
exclusively used to denote a complement that is required by some predicator. In the present,
somewhat old-fashioned, use ‘argument’ includes ‘adjunct’.
Constraints on subject-focus mapping in French and English
predicate, the latter being necessarily deaccented. In Spanish and Italian it is the
sequential order of the subject and the predicate that is reversed, resulting in a
case of syntactic inversion. This reversal of the two main constituents reflects
the fact that it is now the subject that represents the new or focal portion of the
proposition, while the predicate is now pragmatically presupposed. In both lan-
guages, the verb could receive a secondary sentence accent. Such an accent is
not categorial and is therefore ignored here. Finally in French the argument-focus
articulation is expressed via a type of cleft construction (the c’est-cleft, which
formally corresponds to the English it-cleft). Notice that cleft formation results in
postverbal position of the focal argument while keeping the logical subject-predicate
sequence unchanged.
The third type of focus structure is the sentence focus articulation, also referred
to as the ‘all-new’, ‘presentational’, or ‘thetic’ type. In this type, the proposition lacks
a bipartition into either topic and comment or presupposition and focus, the basic
pragmatic function being to introduce a new entity or a new situation (involving a
new entity) into the discourse. In other words, the pragmatic assertion consists in
adding both an argument and a predicate to the discourse:
2.â•… C
onstraints on the mapping from information structure
to grammatical form
and the formal structuring of sentences. These constraints restrict the possible
alignments among the following pragmatic, semantic, and syntactic parameters:
In (7a), the speaker avoids the lexical subject NP proofs whose referent is hearer-
new by resorting to a cleft-like structure headed by the verb get. In this structure
the logical predicate (are dancing in my head) is demoted to secondary predicate
status (dancing in my head), allowing the logical subject (proofs) to appear in
object position and the main subject position to be filled by the highly accessible
deictic pronoun I. Example (8a) is an instance of the ‘presentational amalgam
construction’ (Lambrecht 1988b), in which a NP with a pragmatically inaccessible
referent (here a friend of mine in the history department) functions simultaneously
as the object of the presentational verb have and as the subject of the following
main predicate, in such a way that the initial subject position can be filled by
the deictic pronoun I. (9a) is an instance of a somewhat less conventionalized
construction, in which the logical subject argument (tablecloths) is demoted to
postverbal oblique status (with tablecloths), thereby allowing the initial subject
position to be occupied by the deictic we.
While the cognitive constraints operating in (5) through (9) are universal,
their grammatical manifestation in individual languages is subject to typological
variation (Comrie 1981; Lambrecht 1994; Van Valin & LaPolla 1997; Van Valin
1999). Thus, even though English strongly favors pronominal over lexical subjects
in spontaneous discourse (Prince 1981, Francis, Gregory & Michaelis 1999), it
 Knud Lambrecht
In (12), the interrogative subject mang ‘who’ is prevented from appearing in initial
position through the use of cleft constructions. (12a) corresponds to the English
WH-cleft and (12b) to the English it-cleft. As we will see later on, a very similar
situation obtains in spoken French WH-question formation.
3.â•… M
apping constraints and preferred clause structure
in spoken French
pro+V XP
Foc [+]
pro V
Top [–]
Top [ ]
Foc [–]
Pragmatically, the PCC is of the predicate-focus type (cf. Example (1)), i.e. the PCC
normally expresses a proposition in which the initial pro element is interpreted
as having the pragmatic role of topic, about which the focal predicate expresses
a comment. In accordance with the constraints in (13), the clause-initial pro
element is non-focal and its referent is highly discourse-accessible (hence coded
pronominally). In case pro is the so-called ‘impersonal’ il ‘it’ or the generic on ‘one’,
it is neither focal nor topical, hence the empty brackets after the ‘Top’ attribute. The
postverbal XP element is focal. The verb is unmarked for the topic-focus opposition,
due to the optional nature of focus projection (Schmerling 1976; Fuchs 1976;
Höhle 1982; Selkirk 1984; Jacobs 1993).
In order to preserve the PCC as invariably as possible in discourse, spoken
French uses a number of ready-made grammatical constructions which “target”
the PCC, i.e. whose sole purpose seems to be to permit speakers to rearrange con-
stituents according to the communicative needs of the discourse without violating
the constraints in (13) on the position and morphological type of topic and focus
elements. These ‘PCC-targeted constructions’ fall into three major syntactic types:
(15) a. dislocation constructions
b. secondary predication constructions
c. inversion constructions (rare in spoken French)
The speaker’s selection among the construction types in (15) is determined by the
focus articulation of the proposition to be communicated. Generally speaking, dis-
location is used for predicate-focus, while secondary predication (and inversion)
is used for argument-focus and sentence-focus. The PCC-targeted constructions
can to some extent be combined with one another via constructional inheritance.
In this paper, I will be concerned only with the secondary predication type (15b).
Constraints on subject-focus mapping in French and English
The box diagram in (16) shows the PCC embedded in a larger sentence
structure, the secondary predication construction (labelled S2). This construc-
tion permits the generation of clefts and other PCC-targeted constructions
(Vé = past participle; RC = relative clause).
(16) Syntax and information structure of the secondary predication construction
S2 secondary predication construction
Speaker B’s reply is a striking example of the pervasive use of the c’est-cleft con-
struction in the spoken language. The canonical SVO structure in (17Ba′) would
be inappropriate in the given context. This is so because speaker A’s question ‘Do
you remember where you bought it?’ has evoked the open proposition ‘You bought
it somewhere’, as well as the desire to know the identity of the place in question.
The place of provenance of the shirt will therefore be the argument-focus element
of the answer. However since the shirt was in fact not a purchase by speaker B
but a gift from his daughter, the gift-giver, not a store, will occupy the argument-
focus position in the answer. Now since unpredictably the predicate associated
with the focus element is donner ‘give’, not acheter ‘buy’, this unpredictability must
be expressed by prosodic prominence on the verb in the RC (Lambrecht 1994).
Notice that the predicate ‘buy’, when associated with a goal argument, belongs to
the same semantic ‘giving’ frame as the verb give. Therefore the open proposition
‘x gave it to me’ is cognitively sufficiently accessible to warrant the use of the
c’est-cleft construction (see Prince 1978).
In strong contrast to French, an it-cleft construction would be clearly inappro-
priate in English in (17). Instead, English uses the canonical SVO syntax. Notice,
however, that the sentence is prosodically marked as not having the unmarked
predicate-focus or topic-comment articulation. Indeed, the sentence accent on the
subject NP Isabelle is an instance of the so-called ‘A-accent’ (Bolinger 1989) or
primary accent (Ladd 1996: 223ff.) involving a falling intonation contour (marked
‘H*L’ in the system of Pierrehumbert 1980). The referent of the subject NP is thus
formally marked as having not a topic but a focus relation to the proposition.2
Example (17) allows us to draw two tentative conclusions. The first is that the
appropriateness conditions for the use of the cleft construction in question are not
.â•… In calling the accent on Isabelle ‘primary’ or ‘A-accent’ I am leaving open the question of
the nature of the second sentence accent, on gave. I assume this is also a primary accent, even
though pragmatically it does not necessarily mark a focus element.
Constraints on subject-focus mapping in French and English
identical in French and in English, even though there may be usage overlap in other
discourse situations. The second is that it is possible in English to simply reverse
the unmarked topic-comment or theme-rheme word order without concomitant
syntactic adjustments, while this is not possible in French. Notice, however, that
the sequential order of the logical subject (here Isabelle) and the logical predicate
(here me l’a donnée/gave it to me) remains the same in the two languages: in both
languages the focal argument precedes the presupposed predicate.
Item (18) contains an attested English exchange. Speaker B has visited speaker
A and is now planning his return home. Speaker A has offered to drive B to the
airport, but B says he wants to take a cab instead:
As B’s reply shows, it is possible, and in fact quite idiomatic, in English to use the
unmarked syntactic sequence SV, even though in the given discourse situation the
predicate pays is clearly less focal than the subject company. Indeed, the semantic
‘paying’ frame has been evoked in speaker A’s utterance, while the company in
question is entirely new to the discourse, hence acts as the focus of the proposi-
tion. As a result, the subject NP company receives the focus-marking A-accent,
as in the previous example. By contrast, such a simple reversal of the topic-focus
order is unacceptable in French and a c’est-cleft construction would have to be
used, as shown in (18Bb). Nevertheless the sequential order of the logical subject
and the logical predicate is again the same in the two languages. What counts is
that in French this subject is prevented from appearing in initial subject position.
A similar situation obtains in (19). The utterance in (19a) was made in a
restaurant, in reply to another speaker’s question as to which item to choose from
the wine list:
(19) a. I really don’t care. [YOU] decide.
a′. I really don’t care. ?It’s [YOU] that decides.
b. Ça m’est vraiment égal. C’est [vous] qui decidez.
b′. Ça m’est vraiment égal. #[Vous] décidez.
The semantic structure of the utterance in (20a) contains a contrast between two
parallel predications, ‘working for us’ and ‘being paid by the government’. Indeed
in a language like Italian, the syntactic sequence would directly express the
semantic theme-rheme parallelism between the two predications, via subject-verb
inversion in the second pair. The analogous Italian sentence is shown in (20c)
(20) c. Lavorano per noi, ma paga [il governo].
which literally translates as ‘(They) work for us but pays the government’. Neither
French nor English can express the theme-rheme sequence in the same exact
parallel as in Italian (unless they were to resort to a passive construction in the
second pair, such as ‘They work for us, but they are paid by the government’). Instead
both languages resort to the chiastic structure A – B, B – A, where the theme-rheme
sequence is reversed in the second pair. However, the chiasmus is expressed with
different syntactic structures in the two languages. While English uses the cano�
nical subject-predicate sequence twice, French resorts to a c’est-cleft construction
in the second pair.
The next example shows an interesting subtle mistake made by an American
student writing in French. In (21a), the actually produced sentence, the student
describes how as a little girl she once got lost walking along a beach and how a life
guard helped her find her way back to her parents:
(21) a. #Finalement, j’ai trouvé un gardien (ou plutôt, [le gardien] m’a trouvée).
a′. Finalement, j’ai trouvé un gardien (ou plutôt, c’est [le gardien]
qui m’a trouvée).
b. Eventually I found a guard (or rather, [the guard] found me).
b′. Eventually I found a guard (or rather it’s the guard that found me).
As the comparison of the inappropriate (21a) with (21b) reveals, the student
spontaneously uses in French the canonical syntax she would have used in her
Constraints on subject-focus mapping in French and English
The stylistic effect of the English text is built on the parallel between two pairs
of contrasting items: “you tread on his foot” and “he treads on your foot”. At
the time the first pair is uttered, the fact that someone treads on someone’s
foot is new to the discourse, while in the second pair this open proposition is
now discourse-presupposed. This difference in presupposition remains formally
unexpressed in English. What is elegantly expressed is the double contrast
between agent and patient in the two pairs. In French, on the other hand, the
discourse-presupposedness of the open proposition ‘x treads on y’s foot’ is formally
expressed in the second pair via the relative clause of the c’est-cleft construction.
What remains unexpressed in French is the contrast between agent and patient.
In both pairs, the patient argument is simply expressed by an unaccented pronoun
(lui and vous).
Item (23) illustrates a little-known construction where the secondary predi-
cate represented in (16) is not a relative clause (RC) but a noun phrase (NP).
Sentence (23a), from a comic book by the French cartoonist Reiser, is uttered by
a rat sitting comfortably next to a garbage can in a Paris street. Tasting different
pieces of newspaper from the garbage can, the rat says:
(23) a. C’est [Le monde] le meilleur.
a′. #[Le Monde] est le meilleur.
b. [Le monde] is the best.
b′. *It’s [le Monde] the best.
Knud Lambrecht
The use of the structure in (23a) has the effect of avoiding the canonical structure in
(23a′), which would be appropriate in a context where the newspaper Le Monde is
already a topic under discussion, about which the predicate est le meilleur would
express a comment. However in (23a), the newspaper in question is the focus
of the proposition. What is pragmatically presupposed, given the context of the
picture, is the fact that what one eats can taste more or less good. In English, as
before, the focus element appears in the canonical subject position, followed by
the presupposed predicate, reversing the normal theme-rheme order without
concomitant syntactic changes. The structure in (23b′), which is analogous to the
French structure in (23a), is ungrammatical in English.
(24) is another instance of the secondary predication construction in (23). The
speaker is the driver of a car waiting at a red light behind three other cars. Deciding
to pass the other cars as soon as the light turns green, the speaker utters (24a):
(24) a. I’sont trois, c’est [moi] le quatrieme.
a′. #I’sont trois, [je] suis le quatrième.
b. They are three, [I]’m the fourth.
b′. *They are three, it is [me] the fourth.
As (25b′) shows, the literal translation of the English cleft is unacceptable (if
not ungrammatical) in French and a c’est-cleft has to be used instead, as in (25b).3
.â•… Note that Example (25b′), especially with left-dislocation (Le champagne, c’est ce que j’aime),
would be acceptable in the non-cleft reading, where le champagne is a topic expression.
Constraints on subject-focus mapping in French and English 
As in (25), the only possible cleft construction in this context in French is the
c’est-cleft in (26b). Sentence (26b′) is well-formed only in the reading where elle
is a topic expression, as e.g. in response to the question ‘Who was Nancy Reagan’.
Another attested Example is (27a), from a TV interview with Nelson Mandela,
in which the statesman was asked to explain how he met his second wife. After
describing the circumstances that led to the marriage, Mandela ended with
these words:
(27) a. [That]’s how I met her.4
b. C’est [comme ça] que j’ai fait sa connaissance.
c. *[Comme ça] est que j’ai fait sa connaissance.
Even though the speaker has been talking about how he met his wife, the focus
in (27a) is that element of the utterance which corresponds to the question word
‘how?’ in the original question, while the fact that the speaker met his wife at
some point is treated as presupposed. In French the focus element comme ça
‘that way’ cannot appear in preverbal position, as (27c) shows, and a c’est-cleft
must be used instead.
A revealing test case for the general claim made in this paper is the behavior of
focus-sensitive adverbs like ‘only’. French has two expressions for ‘only’: seulement
and (ne)…que. As predicted, neither expression can occur in preverbal subject
position in French:
(28) a. Only [he] understands me.
b. Y a que [lui] qui me comprend.
b′. *[lui] seulement me comprend./*Que [lui] me comprend.5
.â•… In an interesting analysis, Calude (2008) argues that the construction illustrated in (27a)
does not belong to the category Reverse WH-cleft but to a special category she refers to as
‘Demonstrative cleft’. For the purpose of the present paper, the exact categorization of the
construction is irrelevant, as long as the initial cleft constituent is a focus element.
.â•… The ungrammaticality of this sentence, as well as that of the corresponding sentence
in (29b′), is independently motivated by the fact that the que in (ne)…que can only appear
postverbally, hence is banned from initial position.
 Knud Lambrecht
While in English the focus element he is the subject, in French the focus pronoun
lui must appear in the post-copular position of the avoir-cleft construction in
(28b), of which a first example was given in (3e) above. Another Example is
shown in (29):
(29) a. Only [my parents] call me that.
b. Y a que [mes parents] qui m’appellent comme ça.
b′. *[Mes parents] seulement m’appellent comme ça.
/*Que [mes parents] m’appellent comme ça.
As in (28), the focus constituent appears in initial subject position in English, while
in French it must occur in the postverbal focus position of an avoir-cleft.
Another revealing test case for the constraint in (13a) is the behavior of
interrogative subjects in WH-questions in spoken French (see Myers 2007). As
Myers and other researchers have shown, spoken French has a baffling variety
of interrogative WH-constructions, whose existence seems to be at least in part
motivated by the constraint in (13a). Consider the data in (30), some of which are
strongly reminiscent of the Sesotho data in (11) and (12) above:
(30) a. [Who] gave you the dog?
b. [Qui] t’a donné le chien?
c. [Qui] est-ce qui t’a donné le chien?
d. [Qui] c’est qui t’a donné le chien?
e. C’est [qui] qui t’a donné le chien?
f. [Qui] qui t’a donné le chien?
(Notice that the second qui in (30c–f) is a relative pronoun, whose non-focal status
exempts it from the constraint in (13a).) The possible occurrence of (30b), with its
preverbal subject qui (‘who’), is an apparent exception to the constraint in question.
However, as Myers (2007) shows, interrogative qui is exceedingly rare as subject in
the corpora. Nevertheless, unlike Sesotho, sentences with qui in preverbal subject
position are grammatical in French. Example (30c) illustrates the use of the frozen
sequence est-ce que in question formation. This structure is acceptable because
the initial question word qui functions here predicatively rather than as a subject,
the subject being the inverted clitic ce. Although acceptable, this type of question
formation is rare in the corpora, as Myers (2007) has shown. By far the most
common WH-interrogative constructions in spoken French are those in (d) and
(e). Both are c’est-cleft constructions, with the interrogative qui either in COMP
position, as in (30d), or in situ, as in (30e). As in (30c), the interrogative word func-
tions predicatively, as the complement of the copula est. As a result, the constraint
against preverbal focus expressions is not violated. Finally (30f) can be analyzed
as a truncated form of the cleft in (30e), the sequence c’est being understood. What
counts in (f) is that interrogative qui is not in preverbal subject position.
Constraints on subject-focus mapping in French and English
Here the cleft sentence in (32a) with its inalienable possession syntax is semantically
specialized, to the point that the canonical counterpart in (32a′) would not express
the same state of affairs (i.e. dizziness of the speaker). In English, however, the literal
equivalent of (32a′) would be acceptable in the specialized meaning, even though
the alternative structure involving the adjective dizzy may be more idiomatic.
While the above-described avoir-construction involving a relative clause is no
doubt the most frequently used sentence-focus construction, it is not uncommon to
find other syntactic categories in secondary predicate position (see the diagram in
(16) above). Thus in Example (33a), the secondary predicate is an adjective phrase:
(33) a. T’as [les mains] toutes sales.
a′. #[Tes mains] sont toutes sales.
b. [Your hands] are all dirty.
context where the hands in question are not yet a topic under discussion in the
discourse. In English, however, the canonical version would be perfectly natural
in a sentence-focus context, as (33b) shows.
(34a) illustrates a peculiar French construction, in which the secondary
predicate is a past participle preceded by the word de:
(36) a. Il m’a dit qu’y avait eu une grève, qu’y avait [cent cinquante filles] dehors.
a′. #Il m’a dit qu’y avait eu une grève, que [cent cinquante filles] étaient dehors.
b. He told me that there had been a strike, that [a hundred and fifty girls]
were fired.
(37) Je ne sais pas si tu as entendu parler de la tempête qui a traversé la France mais
c’était assez terrible et du coup les fêtes ont été un peu plus réservées. Chez ma
mère, il y a eu des inondations, et chez mon père …
a. c’est [le toît] qui s’est envolé!
Constraints on subject-focus mapping in French and English
Again, the French cleft in (37a) has the effect of avoiding the canonical version in
(37a′). In English the canonical sentence is the natural choice.
The sentence-focus cleft involving être is also the construction conventionally
used to introduce the characters of jokes. An attested Example is (38):
(38) a. C’est [une cliente] qui s’approche d’un étalage de poisson et puis qui.. sur le
vieux port.. puis elle prend un rouget par la main et puis.. elle le renifle.
b. [A customer] approaches a fish stand and then.. in the old port.. then she
takes a red mullet with her hand and then.. she sniffs it.
As expected, the English version of the joke introduces the joke character in the initial
subject position of a regular canonical sentence. (39) is another attested example:
(39) a. Alors c’est [un Suisse et un Belge] qui discutent.. on mélange les deux.. c’est
[un Suisse et un Belge] qui discutent (rires) et il y a [le Suisse] qui dit euh…
b. So [a Swiss guy and a Belgian guy] are talking.. people mix up the two..
[a Swiss guy and a Belgian guy] are talking (laughs) and [the Swiss guy]
says uh…
(39a) is especially revealing because the speaker first uses the c’est-cleft to
introduce the two characters and then switches to the avoir-cleft to continue
the joke (il y a le Suisse qui dit).
The next construction I would like to discuss has a syntactic structure that
does not exactly fit the secondary predication schema in (16). Indeed it lacks the
pro and the V elements of the preferred clause construction. According to Sasse
(1987), this eventive construction is common in other languages, e.g. in Welsh and
in Egyptian Arabic. I will call this the ‘eventive (et) NP qui VP construction’. An
Example is (40a):
Whether we treat the construction illustrated by Une drôle de chose qui m’arrive
as a truncated form of the avoir-cleft Y a une drôle de chose qui m’arrive or as
a grammatical construction in its own right (as I think it is), it is clear that the
presence of the relative pronoun qui has the effect of preventing the construal
of the sentence as a canonical SV(O) sequence, thereby marking the expressed
proposition as eventive. (41) is another example, from a cartoon by Reiser.
The cartoon shows a couple of bourgeois intellectuals strolling around an
overcrowded Mediterranean beach. The man complains about working class
people wasting their hard-earned wages with stupid seaside activities, then
he says:
(41) a. A: La liberté, mais pas pour tout le monde. La liberté, ça se mérite.
B: [Ton côté fasciste] qui ressort.
B′: #[Ton côté fasciste] ressort.
b. A: Freedom, but not for everyone. Freedom has to be deserved.
B: [Your fascist side] is coming out.
As in the previous example, the presence of relative qui prevents the utterance
from taking on the canonical sentence structure while at the same time marking
the proposition as expressing an unexpected event.
When the (et) NP qui VP construction is preceded by the conjunction et it
often expresses a judgment of non-canonicity vis à vis some unusual or incon-
gruous state of affairs. An Example is (42a), from another cartoon by Reiser. The
cartoon shows a man sitting on a park bench with a dog at the end of a leash.
When a young woman passes near the bench the dog gets between her legs.
Insulted, the woman turns back to the man and says:
Here again, the special construction is used with the effect of preventing an
occurrence of the canonical SV(O) sequence as it is naturally used in English.
The last construction I would like to present here also goes beyond the
syntactic secondary predication schema in (16). It involves an interesting case
of grammaticalization via reanalysis of the subject-verb sequence je vois ‘I see’
into a kind of focus marker for an NP whose referent is not sufficiently acces-
sible in the discourse to appear directly in subject or topic position. The syntactic
structure of the construction can be represented as [je vois NP pro VP]. (43a)
is an attested example. The utterance was made in a Parisian bakery, in a dis-
cussion about a new law against fishing in the Seine. At one point, the baker’s
wife says:
Constraints on subject-focus mapping in French and English
(43) a. Les trois quarts des pêcheurs ils relâchent leurs poissons.. et même.. je vois
[mon mari].. il relâche tous ses poissons.
a′. #(…) et même.. [mon mari] relâche tous ses poissons.’
b. Three quarters of the fishermen let their fish go.. and even .. [my husband]
lets all his fish go.
In this paper I have shown that two genetically and historically closely related
languages, English and French, differ nevertheless fundamentally with respect to
the way in which pragmatic categories of information structure are paired with
syntactic categories of sentence formation. While in English the canonical sen-
tence type [NP VP] is extensively used in discourse, in spoken French this type
is subject to severe appropriateness conditions, to the point that it hardly ever
Knud Lambrecht
occurs in natural speech. In spite of its relatively rigid word order, spoken French
is a language in which focus structure contrasts are strongly realized in syntactic
structure. Through the systematic use of realignment constructions, especially
secondary predication constructions like clefts, French permits the strict syntactic
coding of focus structure distinctions of a type not seen in a language like
English. In particular, spoken French has a powerful constraint against comap-
ping of the grammatical relation subject and the pragmatic relation focus. There
is a striking similarity between spoken French and certain genetically unrelated
languages, such as the Bantu language Sesotho, which suggests a typological
division between languages in which focus constituents can be subjects and lan-
guages where they can’t. It would seem that modern French is on its way to
becoming a language like Sesotho, in which the constraint against subject-focus
mapping is grammaticalized.
References
Bolinger, D. 1989. Intonation and its Uses. Stanford CA: Stanford University Press.
Calude, A.S. 2008. Demonstrative clefts and double-cleft constructions in spontaneous spoken
English. Studia Linguistica 62(1): 78–118.
Comrie, B. 1981/1989. Language Universals and Linguistic Typology. Chicago IL: Chicago
University Press.
Creissels, D. 2008. Remarks on split intransitivity and fluid intransitivity. In Empirical Issues in
Syntax and Semantics 7, O. Bonami & P. Cabredo Hofherr (eds), 139–168. Paris: Colloque
de Syntaxe et Sémantique à Paris.
Demuth, K. 1989. Maturation and the acquisition of the Sesotho passive. Language 65: 56–84.
Francis, H.S., Gregory, M.L. & Michaelis, L.A. 1999. Are lexical subjects deviant? In
Papers from the Thirty-fifth Regional Meeting of the Chicago Linguistic Society: The Main
Session, S.J. Billings, J.P. Boyle & A.M. Griffith (eds), 85–97. Chicago IL: Chicago
Linguistic Society.
Fuchs, A. 1976. ‘Normaler’ und ‘kontrastiver’ Akzent. Lingua 38: 293–312.
Höhle, T. 1982. Explikationen für ‘normale Betonung’ und ‘normale Wortstellung’. In Satzglieder
im Deutschen, W. Abraham (ed.), 75–154. Tübingen: Gunter Narr.
Jacobs, J. 1993. Integration. In Wortstellung und Informationsstruktur, M. Reis (ed.), 63–116.
Tübingen: Niemeyer.
Katz, S.L. 1997. The Syntactic and Pragmatic Properties of the c’est-cleft Construction. dissertation,
University of Texas at Austin.
Koenig, J.-P. & Lambrecht, K. 1998. French relative clauses as secondary predicates: A case study
in Construction Theory. In Empirical Issues in Formal Syntax and Semantics 2, F. Corblin,
C. Dobrovie-Sorin, & J.-M. Marandin (eds), 191–214. The Hague: Theseus.
Ladd, D.R. 1996. Intonational Phonology. Cambridge: CUP.
Lambrecht, K. 1987. On the status of SVO sentences in French discourse. In Coherence and
Grounding in Discourse, R. Tomlin (ed.), 217–262. Amsterdam: John Benjamins.
Constraints on subject-focus mapping in French and English
Paul Boucher
Université d’Angers
1.â•… Introduction
The syntax of English wh-questions shows little or no variation. Aside from standard
requests for information, where both the wh-phrase and the inflected auxiliary must
raise to the left periphery of the root clause, only so-called echo or “reprise” ques-
tions (Bolinger 1978) (1) or what Quillard (2000) calls “phatic introductory ques-
tions” (2) allow the wh-phrase to remain in situ.
(1) A: Mary is going to visit [inaudible]
B: Mary is going to visit WHO? (rising intonation)
Elsig (2009), based on a corpus of 19th and 20th century Quebec French and Early
Modern French, lists seven types of wh-questions:
.╅ Myers (2007) also considers that there are four basic structural types of wh-questions
in French.
 Paul Boucher
French est-ce que, respectively que and é que, and Brazilian Portuguese uses the in
situ construction and both Catalan and Brazilian Portuguese avoid inversion con-
structions, no other Romance language shows anywhere near the variation noted
by Gadet for European French or Elsig for Quebec French.
In the Generative Syntax framework the apparent optionality of the syntactic
form of questions in French has attracted a certain amount of attention, especially
as concerns wh in situ questions, and a number of formal solutions have been
proposed (see for example Chang (1997); Bošković (1998); Boeckx et al. (1999);
Cheng & Rooryck (2000); Adli (2004a and b); Mathieu (2004); Poletto & Pollock
(2004, 2009), among many others.). Most of these analyses have been framed in the
context of Cheng (1997) and Chomsky (1995), according to whom, respectively,
questions are typed either by movement or by the presence of special particles,
and languages which may satisfy checking requirements without overt syntactic
movement must do so. In this light, French appears to fall somewhere between
‘true movement languages’ like English (21) and ‘true particle languages’ like
�Chinese (22).
2.1â•… Statistics
The following table gives statistics from my own study of 5 corpora for the four
question types mentioned in the introduction.
 Paul Boucher
We will now look at several corpus studies to try to pinpoint the pragmatic
functions of each question type in French before comparing these to equivalent
English questions in the following section.
Adli (2004a), in situ questions are typical of informal speech, which is characterized
by a low degree of cognitive anticipation and syntactic complexity as well as by a
strong link to the context. The statistics from the other corpora in Table 1 are not
so clear on this point however.
On the other hand, based on her study of the Barnes-Blythe corpus, Myers
(2007: 147) rejects the idea of a low sociological-stylistic register for Type 1 and
2 questions: “since the Q pro V and pro V Q structures [Myers’ terms for Types 1
and 2] were both used with great frequency in similar stylistic contexts, it must be
assumed that they share register evaluation”.
A more important factor seems to be the degree of ‘topical coherence’ of the
question relative to the preceding discourse. By this term I mean the way the ques-
tion fits into the network of presuppositions established by the preceding dis-
course, which we could call ‘contextual grounding’. In Rue, for instance, we see the
interviewers alternating between complex inversion (CI), subject clitic inversion
(SCI), est-ce que questions and in situ questions depending on how their question
fits into the on-going flow of discourse. This is precisely the sense of “alloques-
tions” I referred to in the introduction. Just what makes the speaker choose the
in situ form?
Type 1 questions may be used, as in English, as “echo questions” (23) or as
“phatic introductory questions” (24).
(24) Elle me l’a remboursé en deux fois et la deuxième fois elle est venue me
rapporter l’argent et j’ai senti que je ne la reverrais plus jamais ~ elle est restée
quoi? Cinq minutes?
‘She paid me back in two payments and the second time she came to bring me
the money and I got the feeling that I would never see her again – she stayed
what? Five minutes?’ (Quillard 2000: #717)
Myers (2007: 153) found that “The pragmatic tendencies of the pro V Q structure
are strong; the structure is often used in contexts that are highly answerable, active
and highly expected”. This corresponds to my own findings. Wh-in situ questions
in my corpora tend to occur in certain types of speech situations, such as ordering
in a restaurant, shopping, police-style interrogations and so on, where the respec-
tive roles of the interlocutors, the range of possible answers and the expectations
of the question askers are highly constrained.
(26) Ça coûte combien? (Zazie)
That costs how much
‘How much is that?’2
(27) Ce sera quoi? (Zazie)
It will be what
‘What’ll you have?’
(28) Tu es resté absent combien de temps? (Tardi)
You have stayed absent how much of time
‘How long were you gone?’
It also corroborates the notion, first introduced by Chang (1997) and subsequently
exploited by the majority of generative syntax studies that the wh-in situ question
is appropriate in a strongly presupposed context, where the speaker is asking for
details on the established topic. Chang was the first to point out that the answer to
an in situ question cannot be “nothing”.3
.â•… The translations of the examples from Zazie are taken from Barbara Wright’s English
translation, Zazie in the Metro, Penguin Classics, 2000. Those for the other corpora are my own.
.â•… As pointed out by an anonymous reviewer, this is not always true. Wh in situ questions in
French can be “out of the blue” in some cases and can have “nothing” as an answer.
Wh-questions in French and English 
Example 5.8
C. Peut-être il faut…un divan oui
Maybe we need…a sofa, yeah
M. Vous allez faire quoi ce dimanche
What are you doing on Sunday?
/inaudible/
M. J’en avais un à dîner l’autre fois euh
I had one for dinner the other time euh (ibid.)
“The proV Q structure indicates that ‘you are going to do x this Sunday’ is to be
construed as highly active, even though it is not in this context.” (ib.).
Interestingly, Myers found only one example of Type 1 questions used as echo
questions in her corpus. I got the same results. However, I did find a high percent-
age of Type 1 questions used as “requests for clarification”, or what could be called
“follow-up questions”, used “to check or secure understanding of the prior turn”
(Selting 1992: 320).
(31) – Franchement, c’est un investissement qui est parmi les plus rentables.
â•… ‘Frankly, it’s one of the best investments you can make.’
– Oui, mais, investissement, de la part de qui?
â•… ‘Best investment for who?’
In Chroniques, despite the high social status of the judge and his dominant posi-
tion in the discourse, there are several examples of wh-in situ questions. As in the
previous examples, they are all ‘follow-up’ questions, requesting more details con-
cerning the current topic.
(33) – Nous sommes arrivées au niveau d’un restaurant. J’ai vu quelqu’un tirer par
terre. J’ai dit à ma mère: “tiens, c’est bizarre”. Parce que, en Corse, la coutume,
c’est qu’on tire en l’air, quoi.
â•…‘We got to the restaurant. I saw someone shoot towards the ground. I said to
my mother: “Hey, that’s strange.” Because in Corsica, the tradition is that you
shoot in the air first.’
– En tout, vous voyez combien de personnes? demande le président
Dominique Coujard.
‘“All in all, how many people did you see?” asked Judge Dominique Coujard.’
Quillard (2000) also found that more than half of all requests for clarification in
her corpus were in situ interrogatives. She attributes this to the fact that the inter-
rogative expression represents the element to be clarified; placing it at the end of
the utterance allows it to be accentuated thereby targeting the request. Moreover,
Coveney (1989) notes that Type 1 questions cannot be rhetorical. The presence of
the question word in the focal position is apparently incompatible with a question
that is not looking for an answer.
These questions often have a challenging tone to them. Rather than requests for
information, they seem to call into question the validity of the information pro-
vided in the previous speech turn. In some cases these questions can be analysed
as a colloquial version of so-called “What the devil” questions.
(39) Pourquoi diable voulez-vous l’appeler? (Formal Register)
‘Why the devil do you want to call him?’
We note that in German or English such ‘surprise’ questions are characterized, not
by syntactic restructuring, but by prosody. They correspond, for instance, to Selting’s
(1992: 335) ‘astonished’ questions, “signaling a problem of expectation”.
(41) WAS. has du gemacht?
‘What did you do?’
calls into question the presuppositions attached to the clause itself. This is opposed
to what was observed in Type 1 questions, in which the presuppositions repre-
sented by the contextual ground are maintained and an information gap is indi-
cated by the wh-word. We will pursue this point in Section 5.
Type 2 questions may also be used as simple requests for information, especially
with the question word où (where).
(44) Où tu les amènes dîner? (Zazie)
Where you they take dine
‘Where are you taking them for dinner?’
Quillard finds a strong tie between Type 2 questions and rhetorical questions.
I found a certain number of examples of this function in my corpus, with or with-
out the complementizer que after the question word.
(46) Alors moi, de quoi j’aurais l’air? (Zazie)
So me of what I would have the look
‘Well how would I look?’
(50) Pourquoi n’avez-vous jamais voulu dire qui était ce quelqu’un? lui demande
le président, Dominique Coujard.
‘Why haven’t you ever been willing to say who this ‘someone’ was? Judge
Coujard asked her.’
In the Tardi corpus, where there is a lot of third degree interrogation, the detec-
tive interrogating the suspect often uses the inversion structure, while the suspect
never does so.
Myers (2007) notes that inversion can be used to mark a change in register, such
as in “teacherese”, while Quillard (2000) suggests that inversion can be used to put
ironic distance between the speaker and his/her question.
Interestingly, in the Julliard corpus, one of the sole uses of inversion corres�
ponds to a play-acting context, where the students are clearly making fun of
themselves. This ironic, mocking use of inversion is also reported in Myers
(2007: 170–171).
Dewaele (1999) found that inversion is used to introduce a new topic or a new
theme within a topic. In the Rue corpus, made up of transcriptions of a radio talk
show, this is especially clear. At the beginning of the program, and each time a
change in topic is formally introduced and a new discussion is begun, the speaker
uses the complex inversion construction.5
(56) Les centres d’excellence attirant les meilleurs chercheurs, les meilleurs
professeurs et les meilleurs étudiants, la France peut-elle retrouver une place
digne de ses ambitions?
‘The best universities attract the best researchers, the best teachers and the best
students. (Given this fact) can France attain the rank she deserves?’
(57) Université-entreprises: le loup est-il dans la bergerie? C’est le dossier aujourd’hui
de Rue des Entrepreneurs.
‘Businesses on campus: is the wolf in the sheep-shed? This is today’s topic on
“Rue des Entrepreneurs”.’
(58) Les universités françaises vont-elles retrouver leur place dans ces
classements mondiaux?
‘Will French universities be able to regain their international ranking?’
.â•… As the reader can see, these are in fact examples of Yes-No complex inversion questions.
I did not find examples of wh-inversion questions with this particular function in my corpus.
Wh-questions in French and English 
C: Mais c’est dingue, comment est-ce que euh est-ce qu’on peut nous donner (It’s
crazy. How can they give us…)
M: Ce que je comprends pas (What I don’t understand…)
C: la responsabilité de, de de, de corriger des copies (…the responsability for
correcting papers…)
M: quand on a juste une connaissance, enfin de (When we barely know…)
C: sur un cours, qu’on on (…for a course that we…)
M: Un cours où les gens ils payent euh quand même cinquante dollars le crédit, je
sais pas combien, et et on a aucun/inaudible/c’est un cours, un cours trois mille
hein, ça veut dire qu’ils vont bientôt avoir leur license!
(a course that people are paying fifty dollars a credit for, or something like that,
and you have…it’s a course, a course three thousand euh that means that they’ll
be getting their B.A.!)
The nature of these questions clearly necessitates overt marking of the illocutionary
question-asking feature, which explains the presence of est-ce que.
In my corpora, I found that Type 4 questions are most often “unrestricted
‘open’ conversational questions”, in the terms of Selting (1992), that is, explicit
requests for information rather than requests for explanation or for clarification.
They may be used as rhetorical questions but never as echo questions. They are fre-
quently used to indicate a move to a new topic or a new aspect of a current topic,
without the stylistic connotations of the inversion construction. When combined
with a cleft structure (c’est) they often have a ‘challenging’ tone to them, that is,
something we could paraphrase as “Is it really true that X?” or “Do you really think
that X?” This last value is especially clear in the Rue corpus.6
(59) Mais, Jean Tirol, est-ce que c’est le manque d’argent, dans une interview récent
du Nouvel Observateur, Bernard Belloc, l’ancien président de l’université de
Toulouse I, disait ce n’est pas le manque d’argent qui fait des projets mais les
projets qui font venir l’argent.
‘Listen, Jean Tirol, is it really the lack of money? In a recent interview in the
Nouvel Observateur, Bernard Belloc, the former president of the University of
Toulouse 1, said that it’s not a lack of money that creates projects but it’s projects
that create money.’
(60) Est-ce que c’est la bonne approche aujourd’hui?
‘(but) Is that (really) the best way to go about it today?’
.â•… Again, I have had to use Yes-No rather than wh est-ce que questions to make my point. In
this particular case, my choice is due to the fact that 99% of my est-ce questions were “Qu’est-ce
que” questions. Since que cannot be fronted without either verb movement or est-ce que inser-
tion, it is impossible to test its pragmatic function.
 Paul Boucher
Myers (2007: 179) notes that Type 4 questions often occur when the speaker
already knows the answer to his/her question. This may explain the “challenging”
or rhetorical nature of many of these questions. Since the information contained
in the answer to the question is already known to both speaker and hearer, the
est-ce que question serves to comment on it in some way.
2.6â•… Conclusion
In this section, after looking at some statistics, we have discussed the pragmatic
use of the four basic question types in contemporary spoken French. We can
summarize our findings as follows:
Type 1:
–â•fi Syntax: Wh in situ, SVO order;
–â•fi Pragmatics: appropriate in a strongly presupposed context, which often takes
the form of a highly constrained social situation (ordering, shopping, interro-
gation, etc.); frequently used as requests for clarification, as echo or as phatic
questions; cannot be used rhetorically.
Type 2:
–â•fi Syntax: Wh-pronoun raises to the left edge of the clause, SVO order;
–â•fi Pragmatics: generally correspond to a request for explanation; may also express
surprise, disapproval or incredulity, possibly corresponding to a colloquial
version of ‘What the hell/que diable’ questions; may also be used as requests
for information or as rhetorical questions.
Type 3:
–â•fi Syntax: Wh-pronoun raises to the left edge of the clause, “S–V inversion”;
–â•fi Pragmatics: true requests for information but with a strong connotation of
social distance; can be used to change topics.
Type 4:
–â•fi Syntax: Wh-pronoun raises to the left edge of the clause, est-ce que is generated
to the left of the in situ subject;
–â•fi Pragmatics: “unrestricted ‘open’ conversational questions” which optionally
may signal a change of topic or call into question the validity of a presupposi-
tion linked to the question.
Table 3.╇
Concerning their pragmatic functions, wh in situ questions occur in all of the four
main pragmatic categories, making up 7% of “external” questions, 2% of “talk”
questions, 1% of “relational” questions and 1% of “expressive” questions. While
36% of standard wh-questions were used for “external” or information seeking
questions, 43% of them were used to express information, including rhetorical
questions, the rest, respectively 9 and 12%, were used either as “talk” questions or
as “relational” questions.
Looking at the problem the other way around, that is, from function to form,
we find that 65% of “questions seeking information about the immediate conver-
sational context”, which we called “requests for clarification” or “follow-up ques-
tions” above, were Yes-No intonation questions, that is questions which left the
verb in situ and simply signaled the question form with rising intonation at the
end of the sentence. Freed also found that tag questions tended to be used as “rela-
tional” questions, especially as “phatic information” questions.
In other words, while most of the functions observed in Section 2 for French
questions are also found in the English language corpus, though not necessarily
in the same order or with the same name, the mapping from function to form
gives totally different results, aside from a clear tendency in both languages to use
intonation rather than movement for requests for clarification of information in
the preceding discourse. Whatever light future studies may be able to shed on this
question, it seems clear that English does not use syntactic variation in the French
sense to express pragmatic functions.
This is confirmed by the English translations of the four basic French question
forms found in Zazie dans le Métro. Only two syntactic forms are observed, the
standard construction and the in situ construction used for echo questions. Below
are some representative examples.
I. Wh in situ + V in situ
(66) (Faut pas egzagérer, dit le type.) – Egzagérer quoi? demande Charles.
Egzaggerate what? (echo)
(67) Ce sera quoi? (Du brie?)
What would you like? Brie?
(68) Je vous dois combien?
How much do I owe you?
(69) Vous, dit le type, vous vous appelez comment?
What about you, what’s your name?
(70) (Je vois.) Elle voit quoi?
What does she see?
(71) (Tu verras de tes propres yeux.) Je verrai quoi?
What’ll I see?
(72) (C’est la seule façon.) La seule façon de quoi?
The only way to what? (echo)
II. Wh moved, V in situ
(73) Alors moi, de quoi j’aurais l’air?
And what’d I look like then?
(74) Non mais de quoi je me mêle?
No but what am I letting myself in for?
(75) En quoi ça consiste?
And what does that consist of?
(76) Quelle gueule il fait maintenant?
How does he look now?
(77) Où vous avez pris ça, s’il vous plait?
Where’d you get that from please?
III. Wh moved, V moved
(78) Et petite, où vas-tu comme ça?
Hey, Zazie, where do you think you’re going?
(79) Pourquoi aurais-tu témoigné à huis clos?
Why should you give evidence in camera?
(80) Et où habites-tu?
And where do you live?
(81) Et pourquoi pleurais-tu tout à l’heure sur le banc?
And why were you crying just now on the bank?
(82) Alors, tonton, comment trouves-tu mes bloudjinnzes?
Well, uncle Gabriel, how d’you like my blewgenes?
 Paul Boucher
(88) Et vous? Dans quoi est-ce que vous vous mettez pour qu’on vous admire?
And what about you? What do you put yourself in so’s people can admire you?
In the following section we will try to sketch out some of the reasons for this situation.
This in turn will lay the foundations for a formal analysis of the two languages.
Old French (9th to 12th centuries) was a verb second (V2) language that
required inversion of the inflected verb and the subject in both declaratives and
interrogatives.
She gives the following examples of declarative (89) and interrogative inversion (90).
“By the beginning of the Modern French period in the 17th century, all types of
inversion had been lost in declaratives, while in interrogatives, only subject-clitic
inversion, as in [(90c)] and [(90d)] above, remained.” O’Connor found the following
statistics for the 16th century:
–â•fi 92% of all questions, both Yes/No and wh-questions, were formed by subject-
clitic inversion;
–â•fi est-ce que questions or SVO ‘intonation questions’ accounted for only 1% each
of the corpus;
–â•fi questions with complex inversion (Jean, va-t-il en ville?) made up approxi-
mately 3% of the total.
further argues that changes in the prosodic system of French may have played a
role in this evolution, by contributing to the grammaticalisation of est-ce que and
reinforcing the wh-in situ construction:
Enfin nous avons cru constater que la cause essentielle de la formation et de la
‘grammaticalisation’ des formules interrogatives du français doit être cherchée
dans l’accentuation finale spéciale du français, rythme qui, d’un côté, invite la
langue à placer le mot à accentuer sous l’accent final, et qui, d’autre part, rend
la place initiale tellement faible, que cette faiblesse force presque la langue à
renforcer le mot interrogatif, lorsque celui-ci se trouve au début de la phrase.
(Rouquier 2002: 102)7
4.3â•… Wh-pronouns
Old French inherited a paradigm of strong, or ‘tonic’ interrogative words from
Latin, which I will call ‘kw-words’ (quoi [kwa] (< Latin quid, quod, quia, quam),
and cui [kwi] (< Latin qui), and had a complementary paradigm of ‘weak’ or
‘atonic’ wh-pronouns, which I will call ‘ke-words’ (que [kε], qui [ki]) (Foulet(1974),
Rey (2006), Brunot and Bruneau (1969), Druetta (2002)). Like all ‘strong’
forms (Cardinaletti and Starke (1999)), the former are stressed, can be objects
of prepositions, have richer morpho-semantic features and remain in situ,
while the latter, like other ‘weak’ or ‘clitic’ forms, are unstressed, cannot be
objects of prepositions, have underspecified semantic content and must leave
their base position and raise to some higher functional position in the left
periphery of the clause.
While the two ‘who’ pronouns – cui, [kwi] and qui [ki] – eventually merged
into a single form, for reasons that we need not go into here, the two ‘what’ pro-
nouns – quoi [kwa] and que [kε]] – survived in Modern French. The clitic form que
continues to raise to its high left periphery target, either by attaching itself to the
‘spurious cleft’ construction, est-ce que, in the terms of Poletto and Pollock (2004),
or by getting a ‘free ride’ through subject-clitic inversion, while the tonic form quoi
is maintained in the base position in all SVO questions. The co-existence of two
complementary positions for question words led to a form of ‘pragmatic specialiÂ�
zation’, with a difference not only in register, but also in function.
.â•… My translation: ‘It would seem that the main cause of the formation and grammaticalisa-
tion of French question forms is to be found in the phrase-final stress marking which is peculiar
to French. This type of stress placement not only encourages placement of the word to be
stressed in the final position, but also so weakens the initial position that an interrogative
pronoun placed there needs reinforcement.’
Wh-questions in French and English 
However, Poletto & Pollock (2006) find all of the above constructions perfectly
acceptable and Adli (2004a) found that the vast majority of the native French
speakers he interviewed accepted such constructions as grammatical. This differ-
ence of acceptability corresponds in my opinion to different interpretations of the
scope of the question. Given the strong contextualization attached to Type 1 ques-
tions, they can always be interpreted as “echo questions” or as “requests for clarifi-
cation or confirmation” and in this case become perfectly acceptable.
These two examples are in no way intended to represent the entire range of pro-
posals in the literature on French wh-questions. They leave aside a number of
important papers, including recent work by Eric Mathieu which goes in the direc-
tion indicated here, that is, that the syntactic variants discussed in Section 2 are
pragmatically motivated.8 However, they do serve to underscore the necessity of
studying French interrogatives in terms of their specific pragmatic functions in
authentic contexts. In the following sub-sections I will suggest a more fruitful line
of investigation, which will rely on two important notions to distinguish English
and French, namely contextual grounding and pronoun strength. I’ll begin with
the latter.
5.2â•… Pronouns
The investigation of French pronouns, both interrogative and personal, leads us in
two interesting directions. First of all, several recent studies, summarized in Elsig
(2009), suggest that French subject pronouns are no longer independent syntactic
arguments but are evolving towards, or perhaps have already reached the status of
.â•… Thanks to an anonymous reviewer for pointing this out to me, referring to Mathieu (2009).
Wh-questions in French and English 
agreement markers, that is, of functional heads. This has important consequences
for the analysis of Type 3 questions, which Elsig claims are no longer productive
in French. Crucially, he argues that so-called “subject-verb inversion” has effec-
tively disappeared from Modern French and that wh-raising in French does not
target a high left periphery target such as Spec,CP, but rather a clause-internal
position, Spec,TP.
Secondly, the fact that French nominal wh-pronouns, as we saw in the preced-
ing section, fall into two different morphological categories: strong or “tonic” on
the one hand, clitic or “atonic” on the other, not only explains the presence of in
situ questions from the very beginning of French, but it also provides us with the
basis for a formal binding mechanism for analyzing them and for differentiating
them from their English counterparts.
If this analysis is on the right track then we can truly say that verb movement has
ceased to be an option in contemporary French. This has direct consequences for
our study of French wh-questions. If verb raising has ceased to be an option in
Modern French, then it follows that wh-raising cannot target the high left periphery
of the clause, since in that case the wh-criterion could not be satisfied at the CP
level. In order for the wh-phrase and the inflected verb to be in a local Spec-Head
relationship, the former must raise to the Spec,TP position. That French satisfies
the wh-criterion at the TP level, while English does so at the XP level as we will
claim below (Section 5.2.5), turns out to be a crucial part of our answer to the
question of how to map the syntax of French questions to their pragmatic func-
tions. We will now look at the nature of French wh-pronouns, comparing them
respectively to their equivalents in NIDs and in English.
Poletto & Pollock’s analysis of the NID wh-clitic doubling construction of the type
found in (100) assumes that such wh-words form ‘pairs’, of the type found when
clitic pronouns double DPs, so-called ‘DP doubling’ constructions, originally dis-
cussed in Kayne (1972) and further studied in Uriagereka (1995). They propose that
this analysis “should carry over to wh-doubling: the wh-clitic and the wh-phrase,
start out as a single complex item and then split and move to different projections,
Wh-questions in French and English 
due to the distinct features they have to check.” (Poletto & Pollock 2009: 2) In
order to host these doubled wh-words, they posit two distinct operator positions
in the interrogative clause, WhP1 and WhP2.
(104) [WHP1 Wh01 [FORCEP F0 [GROUNDP G0 [TOPP [WHP2 Wh02 [IP …]]]]]]
They further claim that the members of this ‘wh-pair’ “instantiate a (set of)
feature(s) in the complex functional structure of wh-items which parallels the dif-
ferent layers of the CP projections” (p.11). Since in some NIDs three sorts of wh-
words are found: clitic forms, like se, weak forms (Cardinaletti 1991) like cusa,
and strong forms like cusé, with three different distributions, respectively “high”,
“middle” and “low”, the authors reason that the clitic member is a disjunction
operator instantiating the higher ‘disjunction feature’, the ‘weak’ member is an
existential operator instantiating the ‘existential feature’, while the ‘strong’ member
is a restrictor operator instantiating the ‘restrictor feature’.
(105) [DISJUNCTION OPERATOR se [EXISTENTIAL OPERATOR cusa [RESTRICTOR cusè ]]]
In Poletto & Pollock’s ‘remnant movement’ analysis, the lower wh-word is not actu-
ally in situ, that is, in its base position. It has, in fact, risen to WhP2. Then the rest
of the clause moves to GroundP (or alternatively the VP raises to GroundP and
the subject to TopP; see Poletto & Pollock 2004: 262). The left-edge clitic wh-word
then raises to WhP1, from which position it has scope over the entire structure
below it. I will not discuss this aspect of their analysis but concentrate on the clitic
doubling aspect only.
In their analysis of French, the clitic que lexicalises the ‘disjunction feature’, while
its null associate has both existential and restrictor features; the non-clitic quoi lexi-
calises the ‘restrictor feature’, while its non-lexical associate has both disjunction
and existential features. If we assume, like Poletto & Pollock, that French interro�
gatives involve covert wh-doubling, that is, “that wh-in situ and wh-doubling are
two sides of the same coin”,10 the natural conclusion would then seem to be that
French in situ expressions in fact function like the lower member of a wh-operator
pair, with the null ‘disjunction operator’ acting as the higher member of the pair.
(108) øTu as fait quoi?
You have done what
‘What did you do?’
Let us pursue this line of reasoning and now try to tie together the descriptive data
discussed in Section 2 with the formal analysis.
We saw that Type 1 in situ questions in French are always strongly grounded
in their contexts, being used in most cases either in highly constrained situations
where a limited number of possible answers are available, or as “follow up” ques-
tions or “requests for clarification” of the previous conversational turn. I would like
to propose that French wh-in situ questions in fact exploit the intermediate “exis-
tential” position by inserting a covert existential operator in just such contexts.
(109) [DISJUNCTION OPERATOR Ø/que [EXISTENTIAL OPERATOR Ø/∃ [RESTRICTOR quoi/Ø ]]]
D-linked wh-expressions, on the other hand, do not show such effects and are
therefore assumed not to move at LF but rather to have their scope assigned by
some other mechanism such as unselective binding (Pesetsky 1987) or choice
function (Tsai 1994; Reinhart 1987). In Pesetsky’s analysis the higher wh-word
unselectively binds both its own trace and the in situ word, which is therefore
considered to function like an indefinite in the sense of Heim (1982).
which allows the higher indefinite to function like a quantifier (≅ every man) and
thus to bind the lower indefinite, which in turn can bind the pronoun it.11
(113) A man who has a donkey (always) beats it.
.â•… Kratzer (2004: 2): “Irene Heim (1982) and Hans Kamp (1981) […] argued that the quan-
tificational behavior of indefinites was an illusion. It derived from overt or non-overt opera-
tors present in semantic representations, or alternatively, from the mechanics of the semantic
interpretation procedure itself. According to Heim and Kamp, indefinites introduced mere
variables with conditions attached to them into semantic representations.” Partee (1995: 567):
“What appears to account for the quantificational interpretation of such sentences is the
existence of various ‘default’ or unmarked operators with interpretations such as ‘universal’,
‘modalized universal’, and ‘generic’, alongside other possibilities such as implicit existential
quantification.”
 Paul Boucher
This fact seems all the more puzzling since, argues Tsai, the in situ solution should
always be preferred by UG on the grounds of economy. In the terms of the mini-
malist approach,
Tsai’s answer to this is based on the notion of pronoun strength: “given that opera-
tor features such as [+wh] are strong in English, procrastination of wh-movement
is not allowed” (ib.: 18). In fact, he pursues, English does take advantage of the
economical solution in (8a), but it does so at the XP level, that is, within the wh-word
itself. His reasoning is based on a proposal that goes back to Katz & Postal (1964)
in which English interrogative and demonstrative pronouns are analyzed in terms
of their sub-morphemes:
Wh words pronominals
wh-o wh-en th-ey th-en
wh-om wh-ere th-em th-ere
wh-at th-at
Given the free relative construals of wh-words (whoever, etc.), which are impos-
sible with th-words (*thatever, etc.), Tsai (1994: 20) concludes “wh does not block
binding from the suffix -ever, which contributes universal force to the indefi-
nites” (whatever ≅ anything, whoever ≅ anyone, etc.). In contrast, pronominals
cannot be suffixed by -ever since th- blocks the binding construal between -ever
and the indefinite.
(116) a. N0
N0 -everx
wh- ind(x)
b. *N0
N0 -everx
[…] Since the [wh] feature is strong in English, singulary substitution must apply
before SPELLOUT to make sure that what is in the matrix CP Spec for feature-
checking, as illustrated by (18b). (Ib.: 22)
Thus the in situ interrogative (114) is ruled out for failing to check the strong fea-
ture of what in overt syntax.
As we have seen no such rule applies in French and the reason seems to be the
difference in pronoun strength between the two languages. French interrogative
pronouns have a weak [wh] feature, which allows them to remain in situ. As argued
above, the French pronoun system overall is evolving towards a state similar to
that observed for NIDs. Covert wh-doubling ensures that there will always be a
higher disjunctive operator to bind the in situ wh-phrase.
6.â•… Conclusion
My goal in this paper has been to find a way to map the syntax of French
�wh-questions to their pragmatic functions and to distinguish them formally from
their English counterparts. This is not an easy task and it would be presumptuous
to assume that I have attained my goal and that everything now fits together seam-
lessly. Obviously there are still many loose strings to tie up. Let us review the various
aspects of the problem discussed in Sections 2 and 2 before concluding.
The central thesis here, following Lambrecht (1994), is that syntactic variation
in French questions is grounded in pragmatics. The same idea underlies several
of the studies discussed above, especially Quillard (2000), Myers (2007) and Elsig
(2009). What emerges from these studies, all of which are based on corpora of
authentic data, is that the fronting (Type 2) and the in situ construction (Type 1)
have become the default interrogative constructions in Modern French and that
verb movement has effectively disappeared from Modern French. Therefore, con-
trary to English, French appears to respect economy and to resort to binary substi-
tution whenever possible, possibly using covert operators to ensure proper binding
 Paul Boucher
References
Adli, A. 2004a. French wh-in situ and syntactic optionality: Evidence from three data types.
Zeitschrift für Sprachwissenschaft 25(2): 163–203.
Adli, A. 2004b. Y a-t-il des morphemes intonatifs impliqués dans la syntaxe interrogative
en français? Le cas du qu- in situ. In Nouveaux départs en phonologie: Les conceptions sub-
et supra-segmentales, T. Meisenberg & M. Selig (eds), 199–215. Tübingen: Narr.
Wh-questions in French and English 
Boeckx, C., Stateva, P. & Stepanov, A. 1999. Presupposition projection and the semantics of
French wh in situ. Ms, University of Connecticut.
Bolinger, D. 1978. Asking more than one thing at a time. In Questions, H. Hiz (ed.), 107–150.
Dordrecht: Reidel.
Boškovic, Ž. 1998. LF movement and the minimalist program. Proceedings of NELS 28, vol. 1,
P.N. Tamanji & K. Kusumoto (eds), 43–58.
Brunot, F. & Bruneau, C. 1969. Précis de grammaire historique de la langue française. Paris:
Masson.
Cardinaletti, A. 1991. On pronoun movement: the Italian dative loro. Probus 3(2): 127–153.
Cardinaletti, A. & Starke, M. 1999. The typology of structural deficiency: A case study of the
three classes of pronouns. In Clitics in the Languages of Europe, H. van Riemsdijk (ed.),
145–233. Berlin: Mouton de Gruyter.
Chang, L. 1997. Wh-in-situ Phenomena in French. MA thesis, University of British
Columbia.
Cheng, L.S. 1997. On the Typology of Wh-questions. New York NY: Garland.
Cheng, L.S. & Rooryck, J. 2000. Licensing wh-in-situ. Syntax 3(1): 1–19.
Chomsky, N. 1995. The Minimalist Program. Cambridge MA: The MIT Press.
Coveney, A. 1989. Pragmatic constraints on interrogatives in spoken French. York Papers in
Linguistics 13: 89–99.
Coveney, A. 1996. Variability in Spoken French: A Sociolinguistic Study of Interrogation and
Negation. Exeter: Elm Bank.
Dewaele, J.-M. 1999. Word Order Variation in French Interrogative Structures. Leuven: Peeters.
Di Cristo, A. 1998. Intonation in French. In Intonation System: A Survey of 20 Languages,
D. Hirst & A. Di Cristo (eds), 95–218. Cambridge: CUP.
Druetta, R. 2002. Qu’est-ce que tu fais? Etat d’avancement de la grammaticalisation de est-ce que,
première partie. Lingua 2: 67–88.
Elsig, M. 2009. Grammatical Variation across Space and Time: The French Interrogative System.
Amsterdam: John Benjamins.
Engdahl, E. 2006. Information packaging in questions. In Empirical Issues in Syntax and
Semantics 6, O. Bonami & P. Cabredo Hofherr (eds), 93–111.
Féry, C. 2001. Focus and phrasing in French. In Audiatur Vox Sapientiae. A Festschrift for Arnim
von Stechow, C. Féry & W. Sternefeld (eds), 153–181. Berlin: Akademie-Verlag.
Freed, A. 1994. The form and function of questions in informal dyadic conversation. Journal of
Pragmatics 21: 621–644.
Friedemann, M.-A. 1997. Sujets syntaxiques: Positions, inversions et pro. Bern: Lang.
Foulet, L. 1974. Petite syntaxe de l’ancien français. Paris: Honoré Champion.
Gadet, F. 1989. Le français ordinaire. Paris: Colin.
Heap, D. & Roberge, Y. 2001. Cliticisation et théorie syntaxique 1971–2001. Revue québécoise de
linguistique 30(1): 63–90.
Heim, I. 1982. The Semantics of Definite and Indefinite Noun Phrases. Ph.D. dissertation,
University of Massachusetts.
Kayne, R. 1972. Subject inversion in French interrogatives. In Generative Studies in
Romance Languages, J. Casagrande & B. Saciuk (eds), 70–126. Rowley MA: Newbury
House.
Katz, J. & Postal, P. 1964. An Integrated Theory of Linguistic Descriptions. Cambridge MA: The
MIT Press.
 Paul Boucher
Kratzer, A. 2004. Indefinites and the operators they depend on: From Japanese to Salish. Ms,
University of Massachusetts, Amherst.
Lambrecht, K. 1994. Information Structure and Sentence Form: Topic, Focus, and the Mental
Representations of Discourse Referents. Cambridge: CUP.
Mathieu, E. 2004. The mapping of form and interpretation: The case of optional wh-movement
in French. Lingua 114(9–10): 1090–1132.
Mathieu, E. 2009. Les questions en français: Micro- et macro-variation. In Le français d’ici:
Études linguistiques et sociolinguistiques sur la variation du français au Québec et en Ontario,
F. Martineau et al. (eds), 37–66. Toronto: Éditions du GREF.
Moran, J. 1992. How to ask: Question formation in written representations of spoken French.
Georgetown University Round Table on Language & Linguistics 1992: Language, Com-
munication and Social Meaning, J.E. Alatis (ed.), 135–146. Washington DC: Georgetown
University Press.
Munaro N. & Pollock, J.-Y. 2005. Qu’est-ce que (qu)-est-ce que? A case study in comparative
Romance interrogative syntax. In Handbook of Comparative Syntax, G. Cinque & R. Kayne
(eds), 542–606. Oxford: OUP.
Myers, L. 2007. Wh-Interrogatives in Spoken French: A Corpus-Based Analysis of their Form
and Function. Ph.D. dissertation, University of Texas at Austin.
Noonan, M. 1989. Operator licencing and the case of French interrogatives. In Proceedings of the
8th West Coast Conference on Formal Linguistics, J. Fee & K. Hunt (eds), 315–330. Stanford
CA: CSLI.
O’Connor, K. 2001. Parametric change and the development of SVO interrogatives in the his-
tory of French. Handout for the talk given at the 31st meeting of the LSRL, Chicago IL,
April 19–21, 2001.
Partee, B.H. 1995. Quantificational structures and compositionality. In Quantification in Natural
Language, E. Bach et al. (eds), Vol. 2, 541–601. Dordrecht: Kluwer.
Pesetsky, D. 1987. Wh-in-situ: movement and unselective binding. In The Representa-
tion of (In)definites. E. Reuland & A. ter Meulen (eds), 98–129. Cambridge MA: The
MIT Press.
Poletto C. & Pollock, J.-Y. 2004. On the left periphery of some Romance wh-questions. In The
Structure of CP and IP, L. Rizzi (ed.), 251–296. Oxford: OUP.
Poletto, C. & Pollock, J.-Y. 2006. Wh-questions in Romance: the case of Mendrisiotto and some
of its consequences for the analysis of French wh- in situ. Talk delivered at the 30th Going
Romance Conference, Amsterdam, December 7–9, 2006.
Poletto C. & Pollock, J.-Y. 2009. Another look at wh-questions in Romance: The case of Mendri-
siotto and its consequences for the analysis of French in situ and embedded interrogatives.
In Romance Languages and Linguistic Theory 2006, D. Torck & W.L. Wetzels (eds), 199–258.
Amsterdam: John Benjamins.
Quillard, V. 2000. Interroger en français parlé: Études syntaxique, pragmatique et sociolinguis-
tique. Ph.D. dissertation, Université de Tours.
Rey, A. 2006. Dictionnaire du français historique. Paris: Robert.
Reinhart, T. 1987. Specifier and operator binding. In The Representation of (In)
definites. E. Reuland & A. ter Meulen (eds), 130–167. Cambridge MA: The MIT
Press.
Rouquier, M. 2002. Les interrogatives en ‘qui/qu’est-ce qui/que’ en ancien français et en moyen
français. Cahiers de Grammaire 27: 97–120.
Wh-questions in French and English 
Schegloff, E. 1977. Identification and recognition in interactional openings. In The Social Impact
of the Telephone, I. de Sola Pool (ed.), 415–450. Cambridge MA: The MIT Press.
Selting, M. 1992. Prosody in conversational questions. Journal of Pragmatics 17: 315–345.
Tsai, W.-T.D. 1994. On Economizing the Theory of A-bar Dependencies. Ph.D. dissertation, MIT.
Uriagereka, J. 1995. Aspects of the syntax of clitic placement in Western Romance. Linguistic
Inquiry 26: 79–123.
Weinrich, H. 1989. Grammaire textuelle du français. Paris: Didier/Hatier.
A comparative perspective
on intensive reflexives
English and Hebrew
Dana Cohen
Université Paris 8
1.â•… Introduction1
for the constructions under discussion; the term ‘intensifier’ also refers to adverbs like very
(Louw 2005; Anderson 2006), while the term ‘emphatic reflexive’ also indicates phonologi-
cally stressed anaphors or point-of-view reflexives (Ferro 1992; Reinhart & Reuland 1993;
Kemmer 1995).
A comparative perspective on intensive reflexives 
2.â•… Background
English IRs appear in three positions: the post-nominal IR (PNself, 1a, b) that
immediately follows the antecedent, the post-verbal IR (VPself, 1d) that follows
the verb and any complements, but can precede or follow other adjuncts, and
the post-auxiliary IR (PAUXself, 1c) that follows (at least one element of) the
auxiliary set. The same positions are available for the Hebrew IR (2a–d), but its
distribution is even freer, as it can, for example, intercede between the V and DO
(2e). In both languages, IRs that are not adjacent to their antecedent must refer to
the subject while the PNself can accompany an antecedent in all syntactic positions
(1a, b, e, 2a, b).
.â•… For the purposes of this paper, this is intended simply as a descriptive characterisation,
without any theoretical implications as to the relations between the two functions.
.â•… Hebrew displays a rich agreement system, in which number, gender and person marking
appears on nouns, verbs, adjectives (not person) and occasionally on prepositions. Whenever
possible, I leave most agreement information out of the gloss, to enhance expository simplicity.
 Dana Cohen
.â•… Hebrew is a pro-drop language. Only a b- IR can refer to the null subject.
.â•… The confusing range of labels and categories drawn in the various analyses underscores
the absence of a clear correlation. A case in point is the classification of the PAUXself. This
position is identified as a distinct IR in Edmondson & Plank (1978) and Verheijen (1986), the
first identifying its function as ‘reversal of semantic roles’. Gast (2006: 82) does not distinguish
the PAUXself, but includes it with some uses of the VPself, as necessarily inclusive. Conversely,
Siemund (2000: 85–87, 112–114) argues that tokens of PAUXself can fall into all three IR types
that he identifies (adverbial-exclusive, adverbial-inclusive and even adnominal).
.â•… The influence of contextual factors on IR interpretation has been noted in some polysemy/
ambiguity analyses (see, König 1991: 93; Siemund 2000; Gast 2006).
 Dana Cohen
The analysis outlined here aims to account for IR data in both English and Hebrew.
The primary goal of this analysis is to account for the fluidity of interpretation
evident in IRs. The present analysis was originally developed for English and its
successful adaptation to Hebrew raises language-specific questions. An additional
goal of the proposed treatment, therefore, is to pinpoint the contribution of the
preposition that accompanies the Hebrew IR and contrast it with the English data.
The analysis outlined here gives a significant role to parallels between IRs and
markers like even, only and too. Markers of this type contribute a contrastive/com-
parative aspect to the meaning of a sentence, which depends on the identification
of two components, frequently termed focus and scope.8 Such a marker focuses
on a constituent and relates it to a set of contextually dependent alternatives of
the same conceptual or semantic type (König 1991; Kratzer 1991; Rooth 1992;
inter alia). The focused entity is typically related to the alternatives in terms of
exclusion, inclusion and scalar ranking.9 Broadly speaking, exclusion indicates that
the alternatives are not substitutes for the focused entity (as in only); inclusion indi-
cates that relevant information on the focused entity also applies to the alternatives
(as in also); scalarity indicates that the alternatives are hierarchically ordered on
some scale and the focused entity is at one extreme of that scale (as in even).
.â•… This is reflected in the terms used for these markers, including ‘focal particles’ (Atlas 1991),
‘focus particles’ (Hoeksema & Zwarts 1991; König 1991; Rullmann 2003), ‘focusing adverbs’
(Taglicht 1984; Rooth 1992), ‘focus-sensitive operators’ (Vallduví & Zacharski 1994; Krifka
2006), ‘scopal adverbs’ (Brugman 1986; McCawley 1996), and ‘scopal operators’ (Kay 1990).
.â•… The status of these aspects of meaning in set markers is debated; they have been classi-
fied as presuppositions (Horn 1969; König 1991; Lycan 1991; Rooth 1992; Kadmon 2001),
conventional implicatures (Karttunen & Peters 1979; Bennett 1982; Francescotti 1995) and
generalized conversational implicatures (see, Atlas 1996).
A comparative perspective on intensive reflexives 
.╅ The main distinction in information structure studies divides the conceptual/semantic
representation of a sentence into relationally given and relationally new information (Gundel &
Fretheim 2004), frequently termed topic/ground/theme and focus/rheme. The differences
between researchers are not only terminological but also conceptual. Consequently, the same
phenomenon can be classified as new by some definition and given by another (see, Hedberg
2006). Various theories recognise that a binary distinction is not sufficient to account for the
range of phenomena sensitive to information structure. Prince (1992), for instance, proposes
a combination of factors: discourse-old/-new, hearer-old/-new and presupposition/focus (see
also, Birner & Ward 1998).
.╅ There are two major differences between set markers and IRs (in English and Hebrew):
set markers are typically uninflected while IRs in both languages are inflected for number,
person and gender; set markers can typically focus on constituents of any category while IRs
can only focus on nominal constituents. These factors are briefly addressed below.
 Dana Cohen
and against this categorisation, see, König 1991: 87ff.; Siemund 2000: 13ff.; Gast
2006: 62ff.; Cohen in prep. §2.3).
Following König (1991), I assume that IRs function like set markers, and show
that the operative elements in the analysis of set markers can be employed in a sys-
tematic monosemous analysis of IRs. This paper outlines the general principles of
the proposed treatment from a pragmatic perspective, linking various interpreta-
tions through conversational implicature. No specific syntactic or semantic analy-
sis is addressed in this work.12 The following section illustrates how the range of
meanings associated with IRs can be incorporated into an overall generalisation.
As noted above, set markers relate the focused entity to a set of alternatives
in terms of exclusion, inclusion and scalar ranking. These relations, in isolation
and in combinations, can be extended systematically over the various IRs in both
languages. The interpretations provided are approximations intended to articulate
the senses relevant for discussion and do not illustrate the full range of positions
or possible interpretations (especially given the effect of contextual factors on this
construction, discussed below).
(6) a. Derek (himself) wrote the letter (himself).
b. Yoni (acmo) katav (beacmo) et ha.mixtav (beacmo).
Yoni (himself) wrote (p.himself) om df.letter (p.himself)
Exclusive: Nobody else (in the relevant set) wrote the letter.
(7) a. Derek (himself) signed the petition (himself).
b. Yoni (acmo) xatam (beacmo) al ha.acuma (beacmo).
Yoni (himself) signed (p.himself) p(on) df.petition (p.himself)
Inclusive: Someone besides Derek/Yoni signed the petition.
(8) a. Derek (himself) invited me (himself).
b. Yoni (acmo) hizmin oti (beacmo).
Yoni (himself) invited me (p.himself)
Scalar: Derek/Yoni is a particularly prominent member of the set to invite me.
The examples above show the three interpretations in isolation. In the absence of
further context, (8a, b) do not necessarily display either the inclusive or the exclu-
sive interpretation – the sentences do not imply that anyone else did or did not
invite the speaker. Similarly, (6–7) do not necessarily display scalarity, providing
.╅ There are various proposals for the semantic derivation of focus interpretations (Kratzer
1991; Rooth 1992; Kadmon 2001 for overview and references). Various syntactic analyses of
adverbs have also been put forward (as in Alexiadou 1997; Cinque 1999; Ernst 2002; Göbbel
2007). The choice between these proposals is outside the scope of this paper.
A comparative perspective on intensive reflexives 
no necessary requirement that the focused entities rank higher or lower relative
to the alternatives. This option is not ruled out, however. The scalar implicature
produced by the IR can be combined with exclusion and inclusion, as it does with
prototypical set markers. This option is illustrated below. The examples in (9) illus-
trate the scalarâ•‚exclusive combination.
(9) (Context: Since the plumber never showed up)
a. Chloe (herself) fixed the tap (herself).
b. Miri (acma) tikna (beacma) et ha.berez (beacma).
Miri (herself) fixed (p.herself) om df.tap (p.herself)
Exclusive: Nobody but Chloe/Miri fixed the tap.
Scalar: Chloe/Miri is a particularly prominent member of the set to fix the tap.
In (9), the referent is interpreted as a prominent member of the set to fix the tap
(in this case, the relevant scale can be unexpectedness), and the activity is associated
with her exclusively. (10) exemplifies the scalar-inclusive combination, typically
associated with even.
(10) a. Bill Gates (himself) uses Firefox (himself).
b. Bill Gates (acmo) miStameS (beacmo) b.firefox (beacmo).
Bill Gates (himself) uses (p.himself) p(in).firefox (p.himself)
Inclusive: Someone besides Gates uses Firefox.
Scalar: Gates is a particularly prominent member of the set to use Firefox.
.╅ The referential properties of English IRs as a reflexive pronoun are discussed in Leskosky
(1972) and Moravcsik (1972).
 Dana Cohen
A sentence like (11/12a), out of context, does not indicate the nature of the rela-
tion between the entity in the IR focus (the sniper) and its alternatives and implies
neither the exclusion nor inclusion of the alternatives. Out of context, the contri-
bution of the IR is only that there are others besides the sniper who are relevant.
Since the choice of relation to the alternatives is left open, the sentence can display
different implicatures in different contexts.
(11) b. The DA prosecuted the heads of the organisation. The sniper himself
was released.
c. The DA couldn’t convict the heads of the organisation. The sniper himself
was released.
.╅ Underspecified meaning and the pragmatic inferential processes required to comple-
ment it are discussed extensively within Relevance Theory (see, Sperber & Wilson 1986/1995;
Carston 2002; Wedgwood 2003, 2007). Relevance theorists emphasise the importance of extra-
grammatical inferencing processes in the construction of meaning. It is argued that the
meaning of any utterance must combine encoded and inferential elements. The processes of
inferencing and context selection are based on a search for relevance, guided by principles of
informativity and processing effort.
A comparative perspective on intensive reflexives 
The context given in (11/12b) raises the exclusive implicature – nobody (in the
set) but the sniper was released, whereas the context in (11/12c) raises the inclusive
implicature – others besides the sniper were also released. A similar effect is evi-
dent in (13/14).
(13) a. The mayor attends the talks himself.
b. The mayor attends the talks himself and his deputy is never seen there.
c. The mayor attends the talks himself and his deputy is there all the time.
Out of context, (13/14a) implies neither exclusion nor inclusion. The context of
(13/14b) raises the exclusive implicature – no one (in the set) but the mayor attends
the talks, whereas the context in (13/14c) raises the inclusive implicature – others
also attend the talks. The same effect applies to the scalar implicature. Consider (15).
(15) I can’t look after the children tonight. I am going out myself.
The context clearly raises the inclusive implicature, but the scalar implicature may
or may not arise in (15) depending on additional contextual factors relevant for
the interlocutors. If the addressee assumes that the speaker does not go out (as the
regular babysitter perhaps), the scalar implicature may be triggered as well. If, how-
ever, no such assumption exists in the relevant context, scalarity need not arise.
These examples illustrate that these aspects of interpretation are not specifically
encoded in the IR. The results of comparison remain context-dependent aspects
 Dana Cohen
In the appropriate (and all too familiar) context in which a computer malfunctions
regularly and a specific application is causing additional problems, (16) can easily
trigger both inclusion and exclusion implicatures. This theoretically contradictory
combination is possible because the two relations stem from somewhat different
aspects of the situation. In this case, both operate on the same set of alternatives,
{causes for instability in the system}, but the inferencing process involves a divi-
sion of this set based on the relevance of distinct time frames. Thus, the result-
ing inclusion aspect indicates that at various points other elements besides the
program display instability, while exclusion indicates that in certain time frames
instability is directly caused by this program rather than by elements outside it.
A comparative perspective on intensive reflexives 
The equivalent sentence without an IR does not trigger any set of alternatives and
neither of these interpretation arises. Their appearance can therefore be attributed
to the IR itself.15 The existence of the inclusive-exclusive combination in the range
of interpretations available to the IR emphasises the fluidity of IR interpretations.
As argued above, the results of comparison are conversational implicatures,
not an obligatory element encoded in the IR. In this light, a basic, core function of
the IR emerges, the element common to all its instantiations – triggering a set of
alternatives against which the focused entity is evaluated.16 This core function is
sometimes evident in isolation, as illustrated in (17).
(17) a. His [John Clay] grandfather was a Royal Duke, and he himself has been to
Eton and Oxford. [Conan Doyle: 142]
b. At his hip he [Arren] wore a sword in a sheath of new leather figured with
inlay of red and gold; but the sword itself was plain, with a worn cross hilt
of silvered bronze. [Le Guin, Shore: 7]
c. Context: a report of a tour for prospective parents at the maternity facilities
of a hospital.
exad me.ha.avot Se.hiStatfu ba.siyur haya
One p(from).df.fathers that.participated dfp(in).tour was
beacmo pag Se.nolad ba.makom.
p.himself preemie that.born dfp(in).place
‘One of the fathers that participated in the tour was himself a preemie born
in the hospital.’
[http://www↜.shnorkel.co.il/gal.asp?gal_id=443]
d. im pa’am nod’a ta’avato le.Seonim […] harey hayom
if once known desireposs p(for).watches […] indeed today
hu mit’ave la.zman acmo
he desires dfp(for).time himself
‘While he was once known for his love of watches […], now he longs for
time itself.’
[http:↜//www↜.haaretz.co.il/hasite/pages/ShArt.jhtml?more=1&itemNo=
865579&contrassID=2&subContrassID=4&sbSubContrassID=0]
.╅ While this combination requires some division, it is not, of course, limited to the temporal
set-subset configuration in this example.
.â•… This aspect is most clearly evident in cases where explicit comparison markers like ‘as x
as’ supplement the function of the IR.
 Dana Cohen
The contribution of the IR in these examples does not lie in the exclusive or inclu-
sive implicature, and there is certainly no scalar implicature involved. The import
of the IRs in (17) seems to be in relating the focused entity to a relevant alternative
in the context. In other words, the IR can be used to contribute nothing more than
triggering a comparison between the focused entity and its alternatives. The vari-
ous options illustrated above are summarised in Table 1.
Inclusive Inclusive-Scalar
Comparison Exclusive Exclusive-Scalar
Scalar Inclusive-Exclusive
.╅ Definitions of contrast abound in the linguistic literature (see, Bolinger 1961; Chafe 1976;
Prince 1998; Culicover & Rochemont 1983; Taglicht 1984; Lambrecht 1994; Erteschik-Shir
1997; de Hoop & de Swart 2004); characterisations of the concept often imply or explicitly call
for opposition, frequently requiring a preference between the contrasted items. For discussion
and comparison of various uses of the term contrast in linguistics and the relation between
the underlying concepts, see, Molnár (2002) and Umbach (2004).
.╅ The term contrast has been raised in several treatments of IRs, typically in relation to the
PNself, although the underlying concept differs (see, Edmondson & Plank 1978; Baker 1995;
Kemmer 1995; Golde 1999; Creswell 2002). Creswell explicitly suggests a non-oppositional
notion of contrast in order to include relations of similarity and identity.
A comparative perspective on intensive reflexives 
incorporating under a single notion the range of relations signalled between the
focused entity and its alternatives.
The second clause of (18a) means, to quote Moravcsik: “’but I didn’t sign it myself;
rather, I had someone else do it.’ [(18b)] means ‘but I didn’t sign it.’ Do includes
the sentence-final intensifier [=VPself] but not the head-bound one [=PNself] in
its scope” (1972: 274–5). The interpretation of (18b) indicates that the PNself is
outside the scope of do and suggests that it is not a part of the predicate in the first
clause. Hebrew data, as in (19) shows parallel effects.
In (19a), as in (18a), the ellipsis in the second clause of the VPself example includes
the IR, so the clause can mean that someone else read the letter to the speaker.
This interpretation is not available in (19b), in which the second clause means
 Dana Cohen
‘but I didn’t read it.’ The interpretation thus indicates that the PNself is not
in the ellipted material and suggests that it is not a part of the predicate in the
first clause.
IR interaction with negation in both languages supports this conclusion (as
shown in (20–21)).
In the PNself examples in (20a, 21a), set construction is based on the negated open
proposition x not erased the tapes. The PNself is therefore outside the scope of
negation. The negated predicate then applies to the focused entity (the president),
producing the set of {individuals who did not erase the tapes} and the focused
entity is interpreted as a member in it. In (20b, 21b), negation affects the entire
clause including the VPself. The comparison set is {individuals who erased the
tapes}, formed prior to the application of negation. Negation applies to this set,
and negates the inclusion of the focused entity (the president) in this set.
The interaction of IRs with scope-related phenomena suggests that scope
plays a significant role in accounting for differences between IRs. Specifically, it
is argued that IR scope affects the determination of the relevant context for set
construction and for comparison. The importance of surface position to the estab-
lishment of the relevant set is illustrated in cases such as (22–24) where the VPself
is not predicate-final and the set is based on the part of the predicate preceding it,
without including the following material. Hebrew allows a wider range of options
here, allowing an IR between verb and DO.
(22) He had a friend who lost his money in a bank, and another friend who was
ruined by an absconding solicitor, and he lost some money himself in a
fraudulent company. [Christie, FC: 39]
As shown above, the English PNself is outside the scope of negation and not
included in VP ellipsis; the English VPself is in the scope of negation and included
in VP ellipsis, while the set it triggers does not include parts of the predicate that
follow the VPself. Examination of similar data for various IR positions indicates
that the English IR takes scope over the element that immediately precedes it. The
integration of these factors into the analysis leads to the conclusion that the scope
of the IR restricts the type of comparison set triggered, directing the addressee
in selecting the relevant context for IR interpretation. The VPself has scope over
the part of the VP that precedes it; therefore, the focused entity is established
as a member of the set triggered by the predication and evaluated against other
members of the same set. The PNself takes scope only over its nominal antece�
dent, indicating that the choice of alternatives is not restricted by the predicate
(although this is the most immediate interpretation, especially in the absence of
additional context). The set is chosen based on any relevant context (linguistic,
previous discourse or encyclopaedic knowledge) based on some property of the
focused entity that is contextually relevant. The PAUXself takes scope over the
preceding auxiliary. Therefore, the set it triggers necessarily involves the predi-
cate, and is based on the elements of the auxiliary set that precede the IR, but not
on the ones following it. The examples in (25) illustrate these distinctions (square
brackets indicate IR scope).
In the simplest case, Derek is evaluated against a set of {dancers with Harriet}.
This is the only possibility for (25b) and no other set is under consideration.
In (25a), the set is not necessarily related to the predicate. Set construction is
inferred based on properties of the referent in any way that seems relevant for
comparison in the context. The predicate is a relevant piece of information in the
context, so (25a) can have the same interpretation as (25b), but it is also appro-
priate in a context where the predicate itself is irrelevant for comparison and
triggers other sets. This is evident in the context in (26) where (25a) is possible
but (25b) is inappropriate.
 Dana Cohen
(26) Derek’s party will be as boring as the last one! Jack and Fred will play cards,
and everyone else will just watch TV.
a. Derek himself will dance with Harriet (and won’t notice anyone).
b. #Derek will dance with Harriet himself (and won’t notice anyone).
The PAUXself in (25c) produces a more subtle effect. The IR in (25b) takes scope
over the VP which triggers the set of {dancers with Harriet}, and the utterance
indicates that Derek will be included in that set at some point in the future. The
PAUXself takes scope over the auxiliary itself, so its contribution, temporal, aspec-
tual etc., is the governing factor in the set construction. The set triggered in (25c)
is the set of {future dancers with Harriet}. The difference is highlighted in (27).
(27) a. Harriet will waltz with Jack and Fred. Derek will dance with her himself
later tonight.
b. Harriet will waltz with Jack and Fred. Derek will himself dance with her
later tonight.
c. Harriet waltzed with Jack and Fred. Derek will dance with her himself
later tonight.
d. Harriet waltzed with Jack and Fred. # Derek will himself dance with her
later tonight.
When the context includes future dancers, as in (27a–b), both positions are pos-
sible. However, when the context includes past dancers with Harriet, as in (27c–d),
the set of future dancers is irrelevant and the PAUXself version that triggers it is
dispreferred. Returning to (25c), the predicate is signalled as relevant for set con-
struction, as it is in (25b). Consequently, the PAUXself is also dispreferred in the
list setting in (26), where the predicate is not relevant for the set. The crucial aspect
is that a set based on the predicate is only one of many possible options for the
PNself, whereas it is the only possibility for the VPself and the PAUXself.
As illustrated in (19–23) above, the Hebrew IR displays similar scope effects
to its English counterpart. As in English, the scope of the Hebrew IR directs the
addressee in selecting the context relevant for the set, as shown in the following
examples.
(28) a. Derek acmo cilem et ha.zira.
Derek himself photographed om df.crime-scene
b. Derek cilem et ha.zira beacmo.
Derek photographed om df.crime-scene p.himself
While IR scope effects are evident in both languages, they differ somewhat in the
specific linguistic marking of this scope. As in English, some Hebrew IRs mark
their scope by linear position. Thus, the bare PNself and the b- marked VPself
take scope backwards over the preceding segment (28–29 above). However, two
important differences must be considered: the wider range of positions open to the
Hebrew IR and its occurrence with a preposition. Unlike English, Hebrew requires
prepositional marking with some IRs. As noted in Section 2, bare IRs can occur
with any nominal antecedent and must immediately follow it, and so are identi-
fied as PNself. In contrast, b- IRs require subject antecedents. The latter point is
illustrated in the contrast between (31a) and (31b).
(31) a. ha.ed (beacmo) hu’aSam (beacmo) be.bizyon (beacmo).
df.witness (p.himself) accusedpass (p.himself) p(in).contempt (p.himself)
‘The witness was accused of contempt (of court) himself.’
b. dibarnu im ha.yeled acmo/?? beacmo.
talked1pl p(with) df.boy himself/?? p.himself
‘We talked with the boy himself.’
The data suggests that the b- IRs, regardless of their positions, take scope over the
predicate (or some part of it) and mark it as relevant for set construction. This is
 Dana Cohen
most evident in post-subject IRs, where both bare and b- forms can occur in the
same linear position. In (32) the b- IR is only appropriate in the context in (32a)
that allows a set based on the information in the predicate, {individuals accused},
suggesting that it takes scope forward over the predicate, while the bare IR is com-
patible with other contexts as well, as in (32b), and allows other relevant sets.
Thus, a b- IR in this position behaves like its VPself counterpart in taking scope
over the predicate. The bare PNself in the same linear position takes scope over the
antecedent, and its set is not restricted to the predicate. The combination of these
factors indicates that IR scope in Hebrew is not exclusively signalled by position
but also marked by the preposition b-.19
To summarise, this section illustrates how IR scope constrains the selection
of the comparison set. In English, scope is signalled by the linear position of the
IR, while in Hebrew, both position and prepositional marking play a role. In both
languages, the PNself takes scope over the antecedent, indicating that the choice
of alternatives for comparison is not restricted to the predicate. The resulting
comparison set may be based on any relevant property of the focused entity. The
focused entity is thus evaluated on a contextually relevant parameter against a
contextually relevant set. The VPself takes scope over the preceding part of the VP.
The focused entity is therefore established as a member of the comparison set trig-
gered by the predication in question and compared to other members of the same
set. The PAUXself takes scope over the auxiliary preceding it. The triggered set
.╅ In some varieties of colloquial spoken Hebrew, the b- IR alternates with the bare IR in
PNself positions and for some speakers even replaces the bare version completely. In these
cases, the preposition no longer marks scope but seems to be reinterpreted as a frozen part
of the IR form.
A comparative perspective on intensive reflexives 
requires the predicate but is centred on the contribution of the specific auxiliary.
This pattern, in English, leads to the generalisation that the IR systematically takes
scope over the segment that immediately precedes it. Hebrew also utilises prepo-
sitional marking, whereby the prepositional variant of the IR marks the predicate
as necessary for set construction. In both languages, IR scope affects the set of
alternatives and restricts the search for relevant context.
thereby marking the set based on it as the anchor entry. The PAUXself takes scope
over the informationally poor auxiliary. In this case, the IR signals that both the
predicate and the referent are discourse-old and already activated, thereby mark-
ing them as anchor entries, while highlighting the connection between them as the
new information in the discourse.
The linking relations signalled by the IR involve subtle distinctions in status.
These subtleties are highlighted when contrasted in the same context, as in (34),
based on the attested (33).
(33) Context: an article discussing criticism of ethnic jokes recounts a joke about
Catholic schools.
My mind goes back to a teenager I met in America in the 1960s, who told me a
story about Christianity which I have never forgotten. […] the kid who told me
the story was himself Catholic. Catholics often tell the best anti-Catholic jokes.
[Independent]
(34) a. The kid himself was Catholic.
b. The kid was Catholic himself.
c. The kid was himself Catholic.
As is evident from the full original context, both antecedent and predicate refer to
discourse-old entities, already activated in the prior context. It is only the relation
between them that is discourse-new. The scope of the various positions signals
different linking patterns to the context. In (34a), the PNself marks the referent
as the anchor to which the information in the predicate is related – Catholicism
is a property ascribed to the kid; in (34b) the VPself marks the predicate as the
anchor – the kid is linked to the previously established set of {Catholics}; in (34c)
both referent and predicate serve as anchors, and the relation between them is the
element being highlighted as new – a link is established between the previously
evoked kid and Catholicism.
Subtle differences in anchoring to the prior context are evident in Hebrew
IRs as well, as illustrated below. The alternations in (36) show different anchoring
variations, based on the attested context in (35).
(35) Context: Article details a debate over the pension that is due to a specific retiring
judge. The issue was resolved in arbitration.
ha.poe’nta hi Se.ha.borer haya beacmo Sofet.
df.point is that.df.arbitrator was p.himself judge
[http:↜//www↜.faz.co.il/story_2052]
(36) a. ha.borer acmo haya Sofet.
b. ha.borer haya Sofet beacmo.
c. ha.borer haya beacmo Sofet.
A comparative perspective on intensive reflexives 
In the original context, both antecedent (the arbitrator) and predicate (being a
judge) refer to discourse-old entities and it is only the relation between them that
is discourse-new. In (36a), the PNself marks the referent as the anchor to which
the information in the predicate is related – being a judge is a property ascribed
to the arbitrator; in (36b) the VPself marks the predicate as the anchor – the arbi-
trator is linked to the previously established set of {judges}; in (36c) both referent
and predicate serve as anchor, and the relation between them is the element being
highlighted as new – a link is established between the previously evoked arbitrator
and judge.
The different linking signalled by the choice of IRs account for the so-called
changes in emphasis associated with IRs. The delicate shifts in linking created by
this choice are nicely illustrated in the attested (37), in which two IR examples are
juxtaposed.
The prior context in this example ascribes a limited, often binary discreteness to
structural coding. The referent of both the PNself and the PAUXself is the semantic
or pragmatic dimension, and the alternative is the structural code. In both cases,
the IR triggers the inclusive interpretation, highlighting the similarity between the
referent and the alternative. The PNself indicates that the properties of binarity and
discreteness should be attached to the referent marked as anchor (all this embed-
ded under illusion). The following PAUXself signals that the current informational
contribution is the connection created between the referent (the aforementioned
dimension) and the predicate (the non-discrete continuum), both of which serve
as anchors.20 The cues of discourse linking and anchoring signalled by the IRs thus
enable subtle shifts in the status of the referent and predicate.
.â•… The position of the IR in both cases complements additional linguistic cues, such as the
use of however and even. Nevertheless, the effects of the IRs remain intact if these additional
linguistic instructions are removed.
 Dana Cohen
5.â•… Conclusion
This paper presents a unified analysis of IRs in English and in Hebrew. The pro-
posed treatment adopts mechanisms and concepts that are independently moti-
vated and applied in the analysis of other linguistic expressions. It has been shown
that in both languages a similar range of interpretations is available to the IR in its
various positions, a fluidity of interpretation that is more easily and economically
accounted for by a monosemous analysis. According to the proposed analysis the
core function of the IR is to signal a comparison of its focused entity against a
contextually determined set of alternatives. The choice of appropriate set and the
relevant parameters for comparison are determined by context-driven, relevance-
oriented inferencing processes.
Scope differences affect context selection and mark information structure,
indicating a subtle change in anchoring to the prior context. English scope is sig-
nalled by position and Hebrew scope is marked by position and prepositional
marking. In both languages, the VPself has scope over the preceding part of the
VP, so the predicate is necessary for the choice of the set. The predicate is marked
as the anchor, and the referent is identified as a member of the set based on it. The
PNself takes scope only over its antecedent, so the set of alternatives is based on
properties of the referent that are relevant in the context and is not restricted to
the predicate. The referent is marked as the anchor in this case. The PAUXself takes
scope over the auxiliary preceding it, and incorporates the predicate due to the
auxiliary-predicate relationship. Consequently, the set triggered by the PAUXself
requires the predicate but is centred on the contribution of the specific auxiliary.
Here both referent and predicate serve as anchors, and the relation between them
is the element being highlighted. These factors are summarised in Table 2.
It is argued that the behaviour of the IR in both languages stems from the com-
bination of its properties as a pronominal/referential element and its properties as
a set marker. IRs differ from prototypical set markers in two respects: (a) easily
identifiable set-focus and scope (via position in English and a combination of
A comparative perspective on intensive reflexives 
possessive form of ‘bone’, and already displaying a similar range of uses in Mishnaic
Hebrew. The properties of the IR construction in the two languages are almost
completely parallel, as shown above, and both function similarly as reflexive ana-
phors. Given the diachronic differences, the strong synchronic parallels suggest
that these properties are not language-specific and are therefore unlikely to be the
result of unique historical processes. Rather, these parallels point to a more cogni-
tive, synchronically-based approach to the combination of functions as offered
here with respect to the IR function alone.
[Christie FC] Christie, A. 1980. Miss Marple’s Final Cases. London: Fontana/ Collins.
[Conan Doyle] Conan Doyle, A. 1891. The red-headed league. In The Adventures of Sherlock
Holmes [Wordsworth Classics edition (1992)], p.142. London: Wordsworth.
[Independent] Kingston, M. 2005. Go on, have a laugh “while it’s still legal”. In The Independent
(London), June 24, 2005.
[Le Guin Shore] Le Guin, U.K. 1975. The Farthest Shore. Toronto: Bantam Books.
References
Cohen, D. 2004. Intensive Reflexives: from Sentence to Discourse. Ph.D. dissertation, The Hebrew
University of Jerusalem.
Cohen, D. In prep. Self-Focusing: The English Intensive Reflexive. University of Paris 8.
Creswell, C. 2002. The use of emphatic reflexives with NPs in English. In Information Sharing:
Reference and Presupposition in Language Generation and Interpretation, K. van Deemter &
R. Kibble (eds), 137–166. Stanford CA: CSLI.
Culicover, P. & Rochemont, M. 1983. Stress and focus in English. Language 59(1): 123–165.
de Hoop, H. & de Swart, P. 2004. Contrast in discourse. Journal of Semantics 21: 87–93.
Edmondson J. & Plank, F. 1978. Great expectations: An intensive self analysis. Linguistics &
Philosophy 2: 373–413.
Ernst, T. 2002. The Syntax of Adjuncts. Cambridge: CUP.
Erteschik-Shir, N. 1997. The Dynamics of Focus Structure. Cambridge: CUP.
Ferro, L. 1992. On self as a focus marker. In Proceedings of the Eastern States Conference on
Linguistics (ESCOL), M. Bernsten (ed.), 68–79. Ithaca NY: Cornell University.
Francescotti, R. 1995. Even: The conventional implicature approach reconsidered. Linguistics &
Philosophy 18(2): 153–173.
Gast, V. 2006. The Grammar of Identity: Intensifiers and Reflexives in Germanic Languages.
London: Routledge.
Givón, T. 1987. Beyond foreground and background. In Coherence and Grounding inÂ�
Discourse [Typological Studies in Language 11], R. Tomlin (ed.), 175–188. Amsterdam:
John Benjamins.
Göbbel, E. 2007. Focus and marked positions for VP adverbs. In On Information Structure,
Meaning and Form: Generalizations across Languages, K. Schwabe & S. Winkler (eds), 275–300.
Amsterdam: John Benjamins.
Golde, K. 1999. Evidence for two types of English intensive NPs. Chicago Linguistic Society
35: 99–108.
Gundel, J.K. 1978. Stress, pronominalization and the given-new distinction. University of Hawaii
Working Papers in Linguistics 10(2): 1–13.
Gundel, J.K. 1985. Shared knowledge and topicality. Journal of Pragmatics 9: 83–107.
Gundel, J.K. & Fretheim, T. 2004. Topic and focus. In Handbook of Pragmatics, L. Horn &
G. Ward (eds), 175–196. Oxford: Blackwell.
Hedberg, N. 2006. Topic-focus controversies. In The Architecture of Focus, V. Molnár & S. Winkler
(eds), 373–397. Berlin: Mouton de Gruyter.
Hirschberg, J. 1991. A Theory of Scalar Implicature. New York NY: Garland.
Hoeksema, J. & Zwarts, F. 1991. Some remarks on focus adverbs. Journal of Semantics 8(1–2):
51–70.
Horn, L. 1969. A presuppositional analysis of only and even. Chicago Linguistic Society 5: 98–107.
Israel, M. 2004. The pragmatics of polarity. In Handbook of Pragmatics, L. Horn & G. Ward (eds),
701–723. Oxford: Blackwell.
Kadmon, N. 2001. Formal Pragmatics. Malden MA: Blackwell.
Karttunen, L. & Peters, S. 1979. Conventional implicatures. In Syntax and Semantics, vol. 11:
Presupposition, C. Oh & D.A. Dinneen (eds), 1–52. New York NY: Academic Press.
Kay, P. 1990. ‘Even’. Linguistics & Philosophy 13: 59–111.
Keenan, E. 2002. Explaining the creation of reflexive pronouns in English. In Studies in the
History of English: A millennial Perspective, D. Minkova & R. Stockwell (eds), 325–355.
Berlin: Mouton de Gruyter.
 Dana Cohen
Kemmer, S. 1995. Emphatic and reflexive -self: expectations, viewpoint, and subjectivity. In
Subjectivity and Subjectivisation: Linguistic Perspectives, D. Stein, S. Wright & E. Finegan
(eds), 55–82. Cambridge: CUP.
König, E. 1991. The Meaning of Focus Particles: A Comparative Perspective. London: Routledge.
König, E. & Siemund, P. 2000. Intensifiers and reflexives: A typological perspective. In Reflex-
ives: Forms and Functions, Z. Frajzyngier & T.S. Curl (eds), 41–74. Amsterdam: John
Benjamins.
Kratzer, A. 1991. The Representation of Focus. Berlin: Walter de Gruyter.
Krifka, M. 1995. The semantics and pragmatics of polarity items. Linguistic Analysis 25: 209–257.
Krifka, M. 2006. Association with focus phrases. In The Architecture of Focus, V. Molnár &
S. Winkler (eds), 105–136. Berlin: Mouton de Gruyter.
Lambrecht, K. 1994. Information Structure and Sentence Form: Topic, Focus, and the Mental
Representations of Discourse Referents. Cambridge: CUP.
Leskosky, R. 1972. Intensive reflexives. In Studies in the Linguistic Science. vol. 2.1, G. Green
(ed.), 42–65. Urbana IL: University of Illinois.
Louw, H. 2005. Really too very much: adverbial intensifiers in Black South African English. In
Proceedings from The Corpus Linguistics Conference Series, Vol. 1(1). 〈http:↜//www↜.corpus.
bham.ac.uk/pclc/〉.
Lycan, W. 1991. Even and even if. Linguistics & Philosophy 14(2): 115–150.
McKay, T. 1991. He himself: undiscovering an anaphor. Linguistic Inquiry 22(2): 368–373.
McCawley, J.D. 1996. The focus and scope of only. In Discourse and Meaning, B. Partee & P. Sgall
(eds), 171–93. Amsterdam: John Benjamins.
Molnár, V. 2002. Contrast – from a contrastive perspective. In Information Structure in a Cross-
Linguistic Perspective, H. Hasselgård et al. (eds), 147–161. Amsterdam: Rodopi.
Moravcsik, E. 1972. Some crosslinguistic generalizations about intensifier constructions.
Chicago Linguistic Society 8: 271–277.
Moyne, J. 1971. Reflexive and emphatic. Language 47(1): 141–163.
Prince, E. 1981. Toward a taxonomy of given–new information. In Radical Pragmatics, P. Cole
(ed.), 223–255. New York NY: Academic Press.
Prince, E. 1992. The ZPG letter: subjects, definiteness, and information-status. In Discourse
Description: Diverse Analyses of a Fund Raising Text, W. Mann & S. Thompson (eds), 295–325.
Amsterdam: John Benjamins.
Prince, E. 1998. On the limits of syntax, with reference to topicalization and left-dislocation.
In Syntax and Semantics, Vol. 29: The Limits of Syntax, P. Culicover & L. McNally (eds),
281–302. New York NY: Academic Press.
Reinhart, T. & Reuland, E. 1993. Reflexivity. Linguistic Inquiry 24(4): 657–720.
Rooth, M. 1992. A theory of focus interpretation. Natural Language Semantics 1: 75–116.
Rullmann, H. 2003. Additive particles and polarity. Journal of Semantics 20(4): 329–401.
Siemund, P. 2000. Intensifiers in English and German: a Comparison. London: Routledge.
Sperber, D. & Wilson, D. 1986/1995. Relevance: Communication and Cognition. Cambridge MA:
Harvard University Press.
Taglicht, J. 1984. Message and Emphasis: On Focus and Scope in English. New York NY: Longman.
Umbach, C. 2004. On the notion of contrast in information structure and discourse structure.
Journal of Semantics 21: 155–175.
Vallduví, E. 1990/1992. The Informational Component. New York NY: Garland.
Vallduví, E. & Vilkuna, M. 1998. On rheme and kontrast. In Syntax and Semantics, Vol 29: The
Limits of Syntax, P. Culicover & L. McNally (eds), 79–108. New York NY: Academic Press.
A comparative perspective on intensive reflexives 
Vallduví, E. & Zacharski, R. 1994. Accenting phenomena, association with focus, and the
recursiveness of focus-ground. In Proceedings of the Ninth Amsterdam Colloquium,
P. Dekker & M. Stokhof (eds), 638–702. Amsterdam: The Institute for Logic, Language and
Computation.
Van Gelderen, E. 2000. A History of English Reflexive Pronouns: Person, Self and Interpretability.
Amsterdam: John Benjamins.
Verheijen, R. 1986. A phrase structure syntax for emphatic self-forms. Linguistics 24: 681–695.
Ward, G. & Birner, B. 2004. Information structure. In Handbook of Pragmatics, L. Horn &
G. Ward (eds), 153–174. Oxford: Blackwell.
Ward, G. & Prince, E. 1991. On the topicalization of indefinite NPs. Journal of Pragmatics
16: 167–77.
Wedgwood, D. 2003. Predication and Information Structure: A Dynamic Account of Hungarian
Pre-Verbal Syntax. Ph.D. dissertation, University of Edinburgh.
Wedgwood, D. 2007. Identifying inferences in focus. In On Information Structure, Meaning
and Form: Generalizations across Languages Linguistik Aktuell/Linguistics Today 100],
K. Schwabe & S. Winkler (eds), 207–227. Amsterdam: John Benjamins.
Focus types and argument asymmetries
A cross-linguistic study in language production
The effects of focus on syntax differ across languages: some languages encode
focus in situ, while in other languages focus induces an array of constructions that
deviate from the canonical configuration, such as non-canonical orders or clefts.
This article presents semi-spontaneously produced data from American English,
Québec French, Hungarian, and Georgian which shows that speakers of these
languages select different structures under identical discourse conditions. The
observed cross-linguistic differences are accounted for by means of grammatical
properties of the object languages that hold independently of information
structure. This account leads to the conclusion that a non-compositional mapping
between information structural concepts and structural configurations is an
unnecessary complication of the grammatical model.
1.â•… Preliminaries1
.╅ The present article evolved within the project D2 Typology of Information Structure, which
is part of the SFB 632 Information Structure at the University of Potsdam/Humboldt University
Berlin (financed by the German Research Foundation). We would like to thank Carsten Breul,
Caroline Féry, Edward Göbbel, Sam Hellmuth, Manfred Krifka, and Malte Zimmermann for
their comments on the interpretation of the experimental data and on previous versions of
this article. Special thanks are due to Rusudan Asatiani, Alain Thériault, Elizabeth Medvedovsky,
and Krisztián Tronka, who contributed to the data collection and the analysis of the data sets
of the individual languages. This article was presented at the conference Contrastive Information
Structure Analysis (Wuppertal, 18 March 2008).
 Stavros Skopeteas & Gisbert Fanselow
Gussenhoven 2007, and Krifka 2007 for some detailed classifications), there is a
major division between those instances of focus that simply express non-presupposed
information and those that come with an additional function that operates on
the relation between the focused constituent and its antecedent(s) in discourse.
Following Kiss (1998: 262), we use the term ‘identificational focus’ for the latter
variety and we assume that this type of focus involves a quantificational operation
over a set of referents, in particular an operation excluding some (contrastive) or all
(exhaustive) relevant alternative referents to the focused element in discourse. We
use the term ‘non-identificational focus’ for the former instances of focus that do
not bear any quantificational properties (also called ‘information focus’, see Kiss
1998). Cross-linguistically, it has been claimed that these focus types differ in their
structural realization. In general, deviations from the canonical syntactic configu-
ration are more likely to be induced by the identificational instances of focus than
by the non-identificational ones. Some syntactic models capture this asymmetry
by assuming that non-canonical syntactic configurations arise through the appli-
cation of some syntactic operation that is associated with identificational focus
(or a subtype of it) (see Kiss 1998, 2009; Drubig 2003). The asymmetry of focus
types is summarized in the implicative relation in (1) which should be read as
follows: “If a non-canonical structure occurs with the non-identificational
instances of focus, it is expected to occur with identificational instances of focus”.
The predictive power of (1) is that it excludes a grammar in which non-canonical
structures occur with non-identificational instances of focus while identificational
instances of focus are expressed through canonical structures. We conceive the
asymmetry in (1) as an observational generalization. As we are going to show
in the discussion of our empirical data, this asymmetry may be derived by the
interaction of contextual conditions with particular structural properties of the
grammars at issue.
(1) Asymmetry of focus types
Identificational focus ← Non-identificational focus
The second asymmetry that is discussed in this article relates to the argument
hierarchy. It has already been observed for some languages that focus on subjects
obligatorily induces a non-canonical structure while focus on non-subjects only
optionally does so. Evidence for subject/non-subject asymmetries has been pro-
vided for several languages including French (Lambrecht 2001), Spanish (Büring &
Gutiérrez-Bravo 2001), Hausa (Hartmann & Zimmermann 2007), West Chadic
languages (Zimmermann 2008), several Kwa and Gur languages (Fiedler & Schwarz
2005), Northern Sotho (Zerbian 2007), etc. This asymmetry is summarized in the
implicative relation in (2) which should be read as follows: “If a non-canonical
structure occurs with focus on non-subjects, it is expected to occur with focus on
Focus types and argument asymmetries 
subjects too”. This implicative relation reflects the observation that nonâ•‚canonical
structures for the expression of focus occur either (a) equally for subjects and non-
subjects, or (b) for subjects but not for non-subjects, or (c) for neither structural
category. The argument asymmetry in (2) excludes a language type in which a
non-canonical construction is used for focusing non-subjects and a canonical one
for focusing subjects.
(2) Asymmetry of focused arguments
Subject ← Non-subject
verbal contributions (e.g. several questions).2 The use of the same elicitation
procedure in all object languages yields a data set of semi-spontaneous expressions
that is ideal for the testing of cross-linguistic hypotheses. This elicitation task is
presented in Section 3 and the empirical results are reported in Section 4.
The theoretical question of this article is whether the cross-linguistic differences
that are captured by the observational generalizations in (1) and (2) reflect:
(a) non-further-decomposable differences of the individual grammars with respect
to the association of information structural concepts with structural operations
or (b) the interaction of universal information structural principles with structural
differences of the grammars at issue. From a conceptual viewpoint, an account
of the latter type has the theoretical advantage of being less stipulative, since it
explains discourse-related phenomena on the basis of structural rules that inde-
pendently hold. To the extent that a compositional account of this type is possible,
it will give further support to the view that the correlation between information
structural concepts and structural operations is not the result of a non-further
decomposable ‘discourse:syntax’ association but rather the product of the
interaction of discourse-related principles with the output of syntactic rules, i.e.
particular linearizations and prosodic possibilities (see Wedgwood 2003; Fanselow
2006, 2007; Fanselow & Lenertová (to appear); Zimmermann 2007). Nevertheless,
the possibility of an account of this type is an empirical question that is dis-
cussed in Section 5.
.â•… The task presented in this paper is part of a longer elicitation agenda, namely the
Questionnaire on Information Structure (QUIS), which is the collaborative product of the
project Typology of Information Structure at the University of Potsdam/Humboldt University
Berlin (see Skopeteas et al. 2006).
Focus types and argument asymmetries 
2.2â•… Reordering
A further source of cross-linguistic variation with respect to the expression of
information structure relates to the structural possibility of a grammar to allow for
alternative linearizations of the same constituents. Two structural operations are at
.â•… The analysis of the prosodic properties of Georgian is a matter of ongoing research by
Caroline Féry in association with Rusudan Asatiani and Stavros Skopeteas that we do not
anticipate in this paper.
 Stavros Skopeteas & Gisbert Fanselow
issue: (a) instances of movement to A-bar positions that are headed by functional
projections outside the lexical domain, and (b) instances of scrambling within the
lexical domain of the hierarchical clause structure.
Hungarian is a language with VSO canonical order. The occurrence of a
constituent in a preverbal position is licensed by restricted contextual conditions.
Two configurations involving preverbal realization of constituents have to be
distinguished, see (3a) and (3b). Example (3a) could occur in a context with a subject
topic (e.g. as an answer to the question ‘What did Mary do?’), while Example (3b)
could occur in a context that licenses narrow focus on the subject (e.g. as an answer
to the question ‘Who called up Peter?’). In both cases, the subject constituent surfaces
in a position that precedes the predicate. However, the preverb fel ‘up’ surfaces in
its default position in (3a), while in (3b) it surfaces postverbally.
(3) Hungarian (Kiss 1998: 256)
a. Mari fel hívta Pétert.
Mary up called Peter.acc
‘Mary called up Peter.’
b. Mari hívta fel Pétert.
Mary called up Peter.acc
‘It was Mary that called up Peter.’
The phenomenon illustrated through (3a–b) is the basic evidence for distinguishing
two preverbal positions in Hungarian. Topics are realized in a sentence-initial
position which is identified by the fact that it precedes the landing site of focused
constituents (see examples in Kiss 1998). Focused constituents undergo movement
to the specifier position of another functional projection, whose head attracts the
V to the effect that the latter precedes the preverb in the linear order (see Kiss
1998: 256). Both preverbal positions are not argument positions, i.e. they are A-bar
positions above the predicate phrase. The range of contexts that induce the operation
exemplified in (3b) is a matter of debate. Some accounts assume that this position
is associated with a quantificational operator encoding exhaustive identification
of the moved constituent (see Kiss 1998), while other accounts assume that this
position is semantically underspecified (see Wedgwood 2003, 2007).
Georgian is a verb final language (SOV) allowing for considerable word order
freedom determined by information structure (see Apridonidze 1986: 136–143;
Vogt 1971: 222; Skopeteas & Fanselow 2009, 2010; Skopeteas et al. 2009). As
expected for V-final languages (see Haider & Rosengren 2003), Georgian allows
for word order changes of the scrambling type. This means that movement targets
argument positions, which is empirically supported by the fact that the non-
canonical orders establish new binding relations (see evidence and discussion
in McGinnis 1999; Skopeteas & Fanselow 2009). The interaction of scrambling
Focus types and argument asymmetries 
with the focus set of the utterance is exemplified in (4): (4a) is a canonical SOV
sentence that could be an answer to the question ‘What happened?’ (all focus)
or ‘What did a/the man do?’ (VP focus) or ‘What did a/the man push?’ (object
focus). (4b) illustrates a sentence in which the object is scrambled over the subject
constituent. This order is contextually restricted, i.e. it could be the answer to the
question ‘Who pushed the chair?’ (subject focus) (see experimental evidence as
well as competence data on Georgian word order in Skopeteas & Fanselow 2010).
Speakers’ intuitions indicate that the SOV linearization in (4a) is not felicitous
in subject focus contexts and that the OSV linearization in (4b) is not felicitous
in object focus contexts (see Skopeteas et al. 2009; Skopeteas & Fanselow 2010).
Hence, the generalization in the Georgian data is that a preverbal constituent
in narrow focus has to be realized adjacent to the verb (see Kim 1988 about the
existence of a preverbal focus position in V-final languages).
(4) Georgian
a. k’ac-i sk’am-s a–c’v-eb-a.
man-nom chair-dat pv(io.3)-push-thm-s.3.sg
‘A/the man pushes a/the chair.’
b. sk’am-s k’ac-i a–c’v-eb-a.
chair-dat man-nom pv(io.3)-push-thm-s.3.sg
‘A/the man pushes a/the chair.’
A complication in the Georgian data results from the fact that this language
involves an operation of optional V-fronting. Hence, the SVO order in (5) may
occur in subject focus contexts, in which case it can be accounted for through
the assumption that the focused subject occupies the specifier of a functional
projection whose head attracts the finite verb (see account on the Hungarian
data above). Crucially, the SVO linearization in (5) may also occur out of the
blue as well as in object focus contexts, a fact that motivated previous accounts
that the order of V projection in this language is unspecified (see Anderson
1984: 186). Based on evidence that the V-final order is the basic configura-
tion, Skopeteas & Fanselow (2010) conclude that the SVO order results from
an operation of optional V-fronting (to the position projected by the head of
the tense phrase). The notion of an ‘optional’ structural operation means that
V-fronting is not associated with a restricted information structural trigger, but
it does not imply that it is a random choice. The choice between a VO and an OV
order corresponds to alternative linear and prosodic options whose occurrence
can be motivated by discourse-related phenomena but cannot be captured by
an operation of matching a discrete semantic or pragmatic feature (see detailed
discussion about the consequences for constituent strucure and evidence from
interpretation in Skopeteas & Fanselow 2010).
 Stavros Skopeteas & Gisbert Fanselow
(5) Georgian
k’ac-i a–c’v-eb-a sk’am-s.
man-nom pv(io.3)-push-thm-s.3.sg chair-dat
‘A/the man pushes a/the chair.’
The two further languages in our sample, namely Hungarian and Georgian, also
have the structural possibility to form cleft constructions (the corresponding
constructions in these languages are reversed pseudo-clefts). However, these
constructions occur only rarely in discourse and native speakers’ intuitions
suggest that they are restricted to specific registers (“written styles”).
3.â•… Method
The aim of the elicitation task that is presented in this section is to create a
semi-naturalistic data set that allows us to observe the effects of the asymmetries
presented in Section 1. This elicitation task is part of the Questionnaire on Infor-
mation Structure (see Section 1, Footnote 2). The experimental procedure is based
on the elicitation of spontaneous answers to several question types. The speaker
is presented four pictures and is instructed to look at the presented scenes. When
(s)he is ready, the pictures are taken away and the instructor asks four questions
 Stavros Skopeteas & Gisbert Fanselow
.â•… The data presented in this paper is part of a larger data set that contains two further
question types (selection and confirmation) and has been carried out in 15 languages. A full
account of the obtained data is under preparation.
Focus types and argument asymmetries 
4.â•… Results
The effects of focus on the clause structure may be tested in the subset of answers
that (i) realize the intended contextual conditions and (ii) involve a lexically realized
verb. Answers that do not meet these requirements were coded as ‘non-valid’ and
are discarded in the further analysis (which means that they are natural answers
in the examined discourse condition, but irrelevant for the hypotheses at issue).
(9a) illustrates an answer in the English data set that does not meet requirement (i)
and (9b) an answer in the Georgian data set that does not meet requirement (ii). The
observations made in the following sections are based on the remaining answers that
were decoded as ‘valid’.
(9) a. {In the scene with cloudy sky, who is looking at the girl?}
Who is looking at the g…? The man is looking at the girl? (Condition n/sbj)
b. {In front of the well, who is pushing a/the man?}
bič’-i.
boy-nom
‘A/the boy.’ (Condition n/sbj)
4.1â•… Georgian5
In the set of valid data, we encountered two types of realization of the focused
constituent. The first type consists of sentences in which the focused constituent
(either subject or object) is placed in the immediately preverbal position, which is
the case in the orders SOFV, OFVS, OSFV, SFVO, OFV, and SFV (see Table 1). The
crucial observation is that while the (X)YFV pattern occurs in several configura-
tions, the XFYV pattern is not attested at all. This contrast provides evidence for
the generalization that a preverbal constituent in narrow focus has to be realized
adjacent to the verb, see 2.2. The following examples illustrate two deviations from
the canonical SOV order (see Skopeteas & Fanselow 2010 for further examples
.â•… A first dataset with 4 speakers was recorded and transcribed by Rusudan Asatiani
(January-June 2005). A second dataset containing 16 further speakers was collected by
S. Skopeteas and transcribed by Sh. Bartaia and N. Tsereteli (September 2005). All participants
are native speakers of Georgian and residents of Tbilisi (11 women, 9 men, age range: 18–26,
average: 21.9).
 Stavros Skopeteas & Gisbert Fanselow
and discussion of this data set): the focused subject in (10a) is realized adjacent to
the verb in an OSFV order; in (10b), the focused object is left adjacent to the verb,
while the given argument is realized postverbally.
(10) a. OSFV
{In the scene with the blue sky, is a/the man hitting a/the man?}
ara, k’ats-s kal-i u-rt’q’-am-s.
no man-dat woman-nom pv(io.3)-hit-thm-s.3.sg
‘No, a/the woman is hitting a/the man.’ (Condition i/sbj)
b. OFVS
{In the scene in the room, what is a/the man hitting?}
sk’am-s u-rt’q’-am-s igi.
chair-dat pv(io.3)-hit-thm-s.3.sg that:nom
‘He is hitting a/the chair.’ (Condition n/obj)
total 40 40 40 40
â•… non-valid 16 19 14 15
â•… valid 24 100.0 21 100.0 26 100.0 25 100.0
â•…â•… SVO 12 50.0 11 52.4 7 26.9 20 80.0
â•…â•… SOV 6 25.0 – – 13 50.0 – –
â•…â•… OVS 3 12.5 6 28.6 – – 1 4.0
â•…â•… OSV – – 3 14.3 – – 2 8.0
â•…â•… OV 3 12.5 – – 6 23.1 – –
â•…â•… SV – – 1 4.8 – – 2 8.0
(11) a. SVOF
{In the scene in front of the fence, what is a/the girl hitting?}
gogo u-rt’q’-am-s mankana-s.
girl(nom) pv(io.3)-hit-thm-s.3.sg car-dat
‘A/the girl is hitting a/the car.’ (Condition n/obj)
Focus types and argument asymmetries 
b. OVSF
{In the scene with the blue sky, who is hitting a/the man?}
k’ac-s u-rt’q’-am-s kal-i.
man-dat pv(io.3)-hit-thm-s.3.sg woman-nom
‘A/the woman is hitting a/the man.’ (Condition n/obj)
100
80
% of n valid answers
60
Non-identificational
Identificational
40
20
0
Object Subject
4.2â•… Hungarian6
We have seen in Section 2.2 that the surface placement of Hungarian preverbs
provides evidence for the distinction between a sentence initial position and an
.â•… The Hungarian data was collected and transcribed by Krisztián Tronka (Piliscsaba,
Hungary, 2006–2007). Four native speakers participated in the experiments, all residents of
Piliscsaba and students.
 Stavros Skopeteas & Gisbert Fanselow
immediately preverbal position. The Examples in (12) illustrate the word orders
in our data set (SVO, SOV, and OVS) in which the verb precedes the preverb, indi-
cating thus that the preverbal constituent occupies the type of preverbal position
in Hungarian that invokes Vâ•‚attraction (compare with the preverb-verb order in
(14)).
(12) a. SF Vp O
{Is a man hitting the man?}
Nem, egy nő üti meg a férfit.
no indef woman hit:3.sg prf def man:acc
‘No, a woman is hitting the man.’ (Condition i/sbj)
b. S OF Vp
{What is the man kicking?}
A férfi a széket rúgja meg.
def man def chair:acc kick:3.sg prf
‘The man is kicking the chair.’ (Condition n/obj)
c. OF Vp S
{Whom is the man kicking?}
Egy másik férfit rúg meg a férfi.
indef other man:acc kick:3.sg prf def man
‘The man is kicking another man.’ (Condition n/obj)
In a further subset of our data, the speakers selected verbs without preverbs (see
(13)). In these utterances, the only evidence for the properties of the position at
issue is the adjacency to the verb. The distribution of these sentences in Table 2
shows that SVO sentences only occur with subject focus, while SOV/OVS sentences
only occur with object focus.
(13) a. SF V O
{Who is carrying the pot?}
Egy férfi cipeli a cserepet.
indef man carry:3.sg def pot-acc
‘A man is carrying the pot.’ (Condition n/sbj)
b. OF V S
{Whom is the man carrying?}
Egy nőt cipel a férfi.
indef woman-acc carry:3.sg def man
‘The man is carrying a woman.’ (Condition n/obj)
Example (14) is the only utterance in our data set, in which the focused constituent
(object) is realized in situ. The preverbal realization of the preverb adds evidence
Focus types and argument asymmetries 
that the given subject in the left periphery is not in the position that invokes
V-attraction. Following discourse-configurational accounts on Hungarian syntax,
the postverbal argument may only bear new information focus, which means
that the answer in (14) is not contextually congruent, since the context involves
correction (see Kiss 1998). However, Example (14) displays a heavy object con-
stituent, indicating that movement to the position that hosts focused constituents
in Hungarian interacts with non-pragmatic preferences on the linearization (such
as the preference for heavy constituents to be realized late in the utterance, that is
known to influence Hungarian word order, see Kiss 2008: 445–447).
(14) Identificational focus in situ
{Is the woman hitting a flower?}
Nem, a nő ki-tépi az utolsó fát a környéken.
no def woman out-pull:3.sg def last tree-acc def neighborhood-sup
‘No, the woman is pulling out the last tree in the neighborhood.’
(Condition i/obj)
The Hungarian data set reveals a categorical pattern as shown in Figure 2. The
focused constituent is realized immediately in front of the verb and this holds for
both focus types and both focused arguments examined in this elicitation task.
Whenever a preverb is available, then this preverb appears postverbally which
supports the view that the constituent that occurs left adjacent to the verb occupies
the specifier position of a functional projection whose head attracts the verb. The
only exception to this pattern is a single example in the condition of identification-
ally focused objects. However, we argued that there is no reason to assume that
 Stavros Skopeteas & Gisbert Fanselow
100
80
% of n valid answers
60 Non-identificational
Identificational
40
20
0
Object Subject
this �difference depends on the examined condition, since the utterance at issue
contains a heavy object constituent that is probably realized in situ for reasons that
do not relate to information structure.
The presentational constructions in (15c) are not bi-clausal, since the predicate
is not expressed through a relative clause. These constructions may occur in two
.â•… The data was collected and transcribed by Elizabeth Medvedovsky (Chicago, December
2005). 20 native speakers (age range 20–26), all inhabitants of Chicago participated in the
elicitation task.
Focus types and argument asymmetries 
total 40 40 40 40
â•… non-valid 12 12 6 12
â•… valid 28 28 34 28
â•…â•… simple clause 28 100.0 24 85.7 34 100.0 20 71.4
â•…â•… it-cleft – – – 6 21.4
â•…â•… presentational – 4 14.3 – 2 7.1
Figure 3 presents the percentages of it-clefts in the data set and shows that the
only context in which this type of cleft occurs in our data set is the condition of
identificational focus on subjects.
100
80
% of n valid answers
60
Non-identificational
Identificational
40
20
0
Object Subject
Figure 3.╇ Percentage of it-clefts in English (averages of speakers’ means)
 Stavros Skopeteas & Gisbert Fanselow
.â•… The data was collected and transcribed by Alain Thériault in Montreal (August–December
2007). 10 speakers (4 men, 6 women; age range: 25–49; average: 34.6) participated in the
experiment, all residents of Montreal, native speakers of Québec French and bilingual in
English. Each speaker has been presented the entire set of questions (hence gave 8 tokens for
each experimental condition), which resulted in a larger data set (total: 320 answers).
Focus types and argument asymmetries 
total 80 80 80 80
â•… non-valid 11 5 10 7
â•… valid 69 100.0 75 100.0 70 100.0 73 100.0
â•…â•… canonical 68 98.6 34 45.3 68 97.1 19 26.0
â•…â•… cleft 1 1.4 41 54.7 2 2.9 54 74.0
â•…â•…â•… c’est – 32 78.0 2 100.0 49 90.7
â•…â•…â•… y a 1 100.0 9 22.0 – 5 9.3
Figure 4 presents the proportions of the data in which the respective focused
constituent is clefted. The data pattern is different from the English one in Figure 3.
First, clefting the focused constituent frequently occurs in both conditions of subject
focus and only in these (recall that the three cleft constructions in the object-focus
conditions involve clefted subjects, see Example (17)). Second, the proportions
of cleft constructions in these conditions are higher than the corresponding
 Stavros Skopeteas & Gisbert Fanselow
100
80
% of n valid answers
60
Non-identificational
Identificational
40
20
0
Object Subject
Figure 4.╇ Clefted focus constituent in Québec French (averages of speakers’ means)
5.â•… Discussion
.â•… Recall that also the clefts that were encountered in the object-focus contexts involved a
clefted subject constituent, see Section 4.4.
 Stavros Skopeteas & Gisbert Fanselow
shown to account for several properties of language processing (see Bornkessel &
Schlesewsky 2006 and references therein); it has already been observed that the
asymmetry between optimal and suboptimal structures has an even stronger effect
in production data since optimal candidates always win the competition to their
alternatives in discourse (see Featherston 2005). In this spirit, we formulate the
minimality condition for the production data as follows.
The minimality condition accounts for a further subset of the empirically attested
differences. It explains why French speakers did not use cleft constructions in the
object focus conditions, as well as why cleft constructions are not attested at all in
the Georgian and Hungarian data sets (though they are possible structural con-
figurations in the grammar). That is, with the assumptions made so far, we may
completely account for the data pattern in Hungarian and Québec French, but not
yet for the subject/object asymmetry and the identificational/non-identificational
asymmetry in the Georgian and American English data.
(21) Georgian10
{Maria, Nino, Kote, and Lela are sitting in the room.}
a. KOT’E u-cem-i-a maria-s.
Kote(nom) pv(io.3)-hit-pf-s.3.sg Maria-dat
‘Maria has hit KOTE.’ (→ not Nino and Lela)
b. maria-s u-cem-i-a KOT’E.
Maria-dat pv(io.3)-hit-pf-s.3.sg Kote(nom)
‘Maria has hit KOTE.’ (→ not Nino and Lela)
The same holds for the English counterparts in (22). The effect of excluding
possible alternatives in discourse does not only hold for cleft constructions, such
as in (22a) (see Kiss 1998: 268), but also for in situ focus in (22b).
(22) {Mary, Paul, John, and Tom are sitting in the room.}
a. It’s Mary that hit John. (→ Paul and Tom did not)
b. MARY hit John. (→ Paul and Tom did not)
The interpretational properties show that both the canonical and the non-canonical
options of expressing focus in English and Georgian allow for the inference of
exhaustive identification. Hence, English and Georgian differ from Hungarian, in
which postverbal constituents do not exhibit exhaustive readings (see Kiss 1998,
2009). The interpretational evidence supports the view that identificational focus
is not a sufficient condition for the licensing of the non-canonical structures in
these languages, as already suggested by the data pattern of our elicitation task.
For these reasons, we assume that the empirically attested asymmetry of focus
types in Georgian and English does not reflect the non-compositional associa-
tion between the feature [+identificational] and particular syntactic operations,
but a contextual asymmetry resulting in a probabilistic correlation with certain
types of answers. Wh-questions (conditions n/sbj, n/obj) introduce a variable and
a presupposition. Answers that only assert the referent that instantiates the variable
are highly expected, i.e. their information structure is fully predictable by the
context, even if it is not signaled by grammatical means. The conditions i/sbj and
i/obj, on the other hand, involve rejection of a part of the presuppositions of the
speaker, hence involving a focus feature that is not predictable by the question.
By means of this asymmetry we may reasonably assume that the latter context
is more likely than the former to induce a structure that articulates the focus
domain at issue. The empirical confirmation of this prediction is the significant
.╅ Note that these examples involve case inversion which is licensed by the perfect tense,
i.e. the agent constituent bears dative case and the patient constituent nominative case (see
Harris 1981).
 Stavros Skopeteas & Gisbert Fanselow
main effect of ‘focus type’ in Georgian and American English. During the produc-
tion process, this asymmetry interacts with markedness constraints resulting in
a data pattern that contains a larger proportion of violations of the minimality
condition in the identificational contexts. The empirical proof of this expectation
is the significant interaction effect between ‘focus type’ and ‘focused argument’ in
American English.
6.â•… Conclusions
The semi-spontaneous data presented in this article shows that the asymmetry of
focus types and the asymmetry of focused arguments have cross-linguistically dif-
ferent effects on the choice of syntactic structure. We accounted for the obtained
differences by means of grammatical differences between the languages at issue,
notably the possibility of expressing narrow focus in situ and the availability of
operations that allow for the expression of focus through simple manipulation
of the linear order. By assuming a minimality condition in language production
we were able to predict the preference for structurally less complex operations
whenever they compete with more complex alternatives in particular contexts. By
assuming a difference between identificational and non-identificational contexts,
we predicted that the former are more likely than the latter to license violations of
minimality in the speakers’ choices.
In sum, we were able to explain the properties of the behavioral data set on
the basis of structural differences between the observed languages and without
recourse to the assumption of associations between certain information struc-
tural concepts and particular syntactic operations. The obtained data provides
evidence against a cross-linguistic 1:1 mapping between types of focus and struc-
tural operations. Hence, while French clefts occur whenever the subject is part
of whatever focus domain, English clefts only occur in contexts that license an
identificationally focused subject. The empirical data shows that English clefts
occur in different contextual conditions than focus movement in Hungarian,
which is counterevidence to the assumption that both structures are licensed by
the same feature of exhaustive identification (see Kiss 1998). The difference in
our data is in line with the conclusion of Wedgwood et al. (2006) that the range
of interpretations and corpus occurrences of focus movement in Hungarian has
a significantly underspecified semantics in comparison to English clefting. In
our view, this difference is accounted for by the fact that English has an in situ
alternative for signaling narrow focus, while Hungarian does not, and furthermore
by the fact that the choice between an in situ alternative and a cleft construction
interacts with structural factors.
Focus types and argument asymmetries 
The argumentation in this article advocates the line of thought that a substantial
portion of the attested cross-linguistic differences on the effects of information
structure on syntax is explained if we take into account the structural possibilities
of the grammars at issue and their interaction with communicative intentions
in discourse. To the extent that these effects are predictable through structural
generalizations, a non-compositional mapping between information structural
concepts and structural operations leads to an unnecessary contamination of the
constituent structure with pragmatic concepts.
References
Féry, C. 2001. Focus and phrasing in French. In Audiatur Vox Sapientiae: A Festschrift for Arnim
von Stechow, C. Féry & W. Sternefeld (eds), 153–181. Berlin: Akademie.
Féry, C. 2010. Information structure and prosody. To appear in The Expression of Informa-
tion Structure: The Interaction of Syntax and Phonology in Cross-Linguistic Perspective,
S. Â�Skopeteas, S. Hellmuth, G. Fanselow & C. Féry (eds). Berlin: de Gruyter.
Fiedler, I. & Schwarz, A. 2005. Out-of-focus encoding in Gur and Kwa. In Approaches and Findings
in Oral, Written and Gestural Language (Interdisciplinary Studies on Information Structure
3), S. Ishihara, M. Schmitz & A. Schwarz (eds), 111–142. Potsdam: Universitätsverlag.
Groenendijk, J. & Stokhof M. 1984. Studies on the Semantics of Questions and the Pragmatics
of Answers. Ph.D. dissertation, Universiteit van Amsterdam.
Gussenhoven, C. 2007. Types of focus in English. In Topic and Focus: Cross-Linguistic Perspectives on
Meaning and Intonation, C. Lee, M. Gordon, D. Büring (eds), 83–100. Heidelberg: Springer.
Haider, H. & Rosengren, I. 2003. Scrambling: Nontriggered chain formation in OV languages.
Journal of Germanic Linguistics 15(3), 203–267.
Harris, A.C. 1981. Georgian syntax: A study in Relational Grammar. Cambridge: CUP.
Hartmann, K. & Zimmermann M. 2007. In place – out of place? Focus strategies in Hausa. In
On Information Structure, Meaning and Form: Generalizations across Languages [Linguistik
Aktuell/Linguistics Today 100], K. Schwabe & S. Winkler (eds), 365–403. Amsterdam:
John Benjamins.
Hedberg, N. 2000. The referential status of clefts. Language 76(4), 891–920.
Kim, A.H.O. 1988. Preverbal focusing and type XXIII languages. In Studies in Syntactic
Typology [Typologicial studies in Language 17], M. Hammond, E. Moravcsik & J. Wirth
(eds), 147–72. Amsterdam: Benjamins.
Kiss, K.É. 1998. Identificational vs. information focus. Language 74(2): 245–273.
Kiss, K.É. 2008. Free word order, (non)configurationality, and phases. Linguistic Inquiry
39(3): 441–475.
Kiss, K.É. 2009. Structural focus and exhaustivity. In Information Structure: Theoretical, Typological,
and Experimental Perspectives, M. Zimmerman & C. Féry (eds), 64–88. Oxford: OUP.
Krifka, M. 2007. Basic notions of information structure. In The Notions of Information Structure
[Interdisciplinary Studies on Information Structure 6], C. Féry, G. Fanselow & M. Krifka
(eds), 13–55. Potsdam: Universitätsverlag.
Lambrecht, K. 2001. A framework for the analysis of cleft constructions. Linguistics 39(3): 463–516.
McGinnis, M. 1999. A-scrambling exists! In Proceedings of the 23rd Annual Penn Linguistics
Colloquium [Penn Working Papers in Linguistics 6.1], M. Minnick & N.-R. Han (eds), 283–
297. Philadelphia PA: University of Pennsylvania.
Rochemont, M.S. 1986. Focus in Generative Grammar. Amsterdam: John Benjamins.
Siewierska, A. 1991. Functional Grammar. London: Routledge.
Skopeteas, S. & Fanselow, G. 2009. Effects of givenness and constraints on free word
order. In Information Structure: Theoretical, Typological, and Experimental Perspectives,
M. Zimmerman & C. Féry (eds), 307–331. Oxford: OUP.
Skopeteas, S. & Fanselow, G. 2010. Focus in Georgian and the expression of contrast. In Lingua
120(6): 1370–1391.
Skopeteas, S., Féry, C. & Asatiani, R. 2009. Word order and intonation in Georgian. Lingua
119(1): 102–127.
Skopeteas, S., Fiedler, I., Hellmuth, S., Schwarz, A., Stoel, R., Fanselow, G., Féry, C. &
Krifka, M. 2006. Questionnaire on Information Structure: Reference Manual [Special issue of
Interdisciplinary Studies on Information Structure 4]. Potsdam: Universitätsverlag.
Focus types and argument asymmetries 
Szendrői, K. 2001. Focus and the Syntax-Phonology Interface. Ph.D. dissertation, University
College London.
Szendrői, K. 2003. A stress based approach on the syntax of Hungarian focus. The Linguistic
Review 20: 37–78.
Thériault, A., Hellmuth, S. & Skopeteas, S. 2008. The Subject Focus Constraint and cleft
constructions in Québec French. Ms, Universität Potsdam.
Van Valin, R.D. 1999. A typology of the interaction of focus structure and syntax. In Typology and
the Theory of Language: From Description to Explanation, E. Raxilina & J. Testelec (eds).
Moscow: Jazyki Russkoi Kultury Publishers.
Vogt, H. 1971. Grammaire de la langue géorgienne. Oslo: Universitetsforlaget.
Wedgwood, D. 2003. Predication and Information Structure: A Dynamic Account of Hungarian
Pre-verbal Syntax. Ph.D. dissertation, University of Edinburgh.
Wedgwood, D. 2007. Identifying inferences in focus. In On Information Structure, Meaning
and Form: Generalizations across Languages [Linguistik Aktuell/Linguistics Today 100],
K. Schwabe & S. Winkler (eds), 207–227. Amsterdam: John Benjamins.
Wedgwood, D., Pethő, G., & Cann, R. 2006. Hungarian ‘focus’ position and English it-clefts: The
semantic underspecification of focus readings. Ms, University of Edinburgh.
Zerbian, S. 2007. The subject/object asymmetry in Northern Sotho. In On Information Structure,
Meaning and Form: Generalizations across Languages [Linguistik Aktuell/Linguistics Today
100], K. Schwabe & S. Winkler (eds), 323–345. Amsterdam: John Benjamins.
Zimmermann, M. 2007. Contrastive focus. In The Notions of Information Structure [Interdis-
ciplinary Studies on Information Structure 6], C. Féry, G. Fanselow & M. Krifka (eds),
147–159. Potsdam: Universitätsverlag.
Zimmermann, M. 2008. Focus realization in West Chadic. Ms, University of Potsdam.
Topicality in L1-acquisition
A contrastive analysis of null subject expressions
in child French and German
Nicole Hauser-Grüdl
Bergische Universität Wuppertal
Although the study of human information structure has become a major field
of linguistic interest, little research has examined topicality in L1-acquisition.
Those studies dealing with topicality in child speech almost all focus on the
question if children do or do not possess the adult ability to correctly identify
and encode topics. The child’s performance of encoding topical information is
thus classified as “correct” if it corresponds to the adult norm or as “deficient”
in case it doesn’t. Such a way of interpreting the data does not consider that
children’s topic-marking may follow its own rules, which are based on a
different understanding of children of what a topic is. This latter possibility
will be discussed in this article for the domain of null subjects in child French
and German.
1.â•… Introduction
Up to now, the notion of topicality has almost exclusively been examined on the
basis of adult speech data. Contrastive analyses in this field usually aim at describ-
ing what is common and what is different about two or more languages with respect
to how topic expressions are marked syntactically, morphologically and phoneti-
cally (see e.g. Cinque 1983; Molnár 1998 and Frascarelli & Hinterhölzl 2007).
Though contrastive in nature, this study looks at topicality from a different per-
spective – namely the child’s. By analyzing and contrasting the person and num-
ber feature specification of null subjects in early French and German of three
bilingual and two monolingual children, it is shown that – despite considerable
differences in the quantity of the children’s null subject use in French and German –
it is in both languages preferably 1st and, in around 40% of the cases, also 2nd �person
referents that are expressed by a null subject in early L1-acquisition. Since 1st and
2nd person null subjects in child French and German typically represent topic
expressions (see Section 4), this result of the study allows two conclusions as to the
child’s understanding of topicality.
Nicole Hauser-Grüdl
First, it could be argued that the preferred null-realization of 1st and in many
cases 2nd person subject referents reflects a child’s understanding of topicality
which is different from that of adult speakers in that in children it is predomi-
nantly permanently available referents, i.e. entities specified for either [+speaker]
or [+hearer], that are interpreted as topics and expressed by means of a null subject.
In adults, on the contrary, subject topic drop mainly affects discourse-active 3rd
person referents. Schmitz, Patuto & Müller (forthcoming) show for German, for
instance, that null subjects are specified for 3rd person at a rate of more than 90%.
An alternative interpretation of the results of the study would be to argue that
the predominant use of 1st and in some children also of 2nd person null subjects
is due to the particular difficulty of learning the syntax and pragmatics of deictic
subject pronouns. Analyses of the different contexts in which 1st person subject
omissions occur suggest that, in the early years of L1-acquisition, children may
treat 1st person subject expressions in analogy to 3rd person anaphoric expres-
sions, i.e. as referring back to a preceding discourse topic. If this were the case,
children would not necessarily have an understanding of topicality different from
adult speakers. Their overuse of 1st and partly 2nd person null subjects would
rather be due to a misinterpretation of deictic subject expressions. Both ways of
interpreting the results of the study will be discussed and evaluated in detail.
The rising linguistic interest in information structure has, over the past decades,
produced an extensive amount of literature in this field with the result that numer-
ous definitions of the term “topic” have been put forward. Generally, definitions can
be grouped into syntactic ones, where the topic is defined as a particular constitu-
ent located in the left periphery of the main clause, into semantic ones, where the
topic is understood to be what is being spoken about in a sentence, and into infor-
mational ones where the topic is seen to represent “old” or “given” information, i.e.
information already known to the interlocutor.1 For L1-acquisition research, the
problem with all these definitions is that they have been elaborated on the basis of
.â•… For an overview of syntactic approaches see Gómez-González (2001: 49ff.), for a semantic
approach see Lambrecht (1994). A detailed discussion of various informational approaches as
well as of their sometimes differing interpretation of what exactly “givenness” means can be
found in Prince (1981: 225ff.).
Topicality in L1-acquisition 
adult speech data, which makes them rather unsuitable to be applied to early child
speech, i.e. to adequately determine what a topic is for young L1-learners.
To give an example, even very elaborate approaches to adult information struc-
ture, such as the one by Lambrecht (1994), encounter difficulties if they are used
to identify topic expressions in child speech. This is due to the fact that although
Lambrecht (1994: 118) defines “topic” semantically as “[…] the thing which the
proposition expressed by the sentence is ABOUT”, in his approach the identifi-
cation of topic constituents draws heavily upon specific prosodic and syntactic
means used by adults to explicitly mark topicality. In the early stages of language
acquisition, however, children typically do not yet master the full range of for-
mal devices needed for adult-like topic marking in their L1(s). Therefore, to try
to identify topic expressions in child speech on the basis of adult-typical formal
marking of topic constituents will, in many cases, prove impossible.
What further complicates the study of topicality in child language is that at
least some of the factors considered crucial in determining topicality in adults
do not seem to hold true for young children. For adult speakers it is commonly
assumed, for example, that a referent, in order to be pragmatically acceptable as
a topic, needs to be discourse-active and thus cognitively accessible to the inter-
locutor. In other words, the topical status of an entity in adult speech depends
on either its prior mention in the linguistic context (in which case topicality is
established anaphorically) or its saliency in the extralinguistic context. In the latter
case, topicality is established deictically, i.e. the referent of a given discourse entity
is unequivocally determined by gestures or eye gaze. This connection between an
entity’s discourse activation, its cognitive accessibility and its pragmatic accept-
ability as a topic is illustrated by Lambrecht (1994: 165) by means of his so-called
“topic acceptability scale” (see Table 1):
As can be seen from the scale, it is those referents which are active in the
discourse that are most easily cognitively accessible and, at the same time, most
acceptable as topics from a pragmatic point of view. For such active, easily accessible
topics it has been shown that they are typically expressed by either an unaccented
Nicole Hauser-Grüdl
(1) (context: Kerstin and adult are looking at models in a fashion magazine)
(child): gucke mal / neue hos(e), ge(ll) /
Ø hat s(ch)oene schuhe an /
Ø hat auch s(ch)uhe an /
‘Look! a new pair of trousers, isn’t it?’
‘(She) is wearing nice shoes.’
‘(She), too, is wearing shoes.’ Kerstin 2;7,23
Since the entity represented by the omitted subject expressions (i.e. the respec-
tive model the child has in mind when speaking) is clearly “the thing which the
proposition expressed by the sentence is ABOUT” (Lambrecht 1994: 118), it is
reasonable to assume the dropped subjects are regarded as topical by Kerstin. This
topical status, however, does not seem to be linked to the subject referent’s prior
discourse-activation and thereby its unequivocal contextual identification, as is
the case with adults. For adult speakers, the subject referents in (1) clearly would
be less acceptable topics (if they were classified as topics at all), since the inter-
locutor has to try to infer them from a rather ambiguous extralinguistic context.
Kerstin’s subject drop, therefore, is illicit in adult speech.
The question which thus arises from child speech samples such as the one
above is on what basis young L1-learners determine topicality. The fact that subject
expressions obviously are omitted regardless of whether or not their referent has
previously been mentioned or, alternatively, been activated deictically, suggests
that children may have a different (possibly broader) understanding of “topic”
than adult speakers – at least in the early stages of language acquisition.
Topicality in L1-acquisition
finding an answer to the question if children lack or if they possess the ability to
correctly identify and encode topics from early on. What is focused on instead
is if the differences observable in children’s and adults’ encoding of information
may be due to a different understanding by children and adults of what a topic is.
To be more precise, it will be suggested that children do not consider the topical
status of an entity dependent on the entity’s discourse activation in the adult sense
(which is through prior mention or explicit deictic reference, see page 201 above),
but primarily on the psychological status of this entity as a permanently available
referent, i.e. its specification as either [+speaker] or [+hearer]. According to this
view, the higher frequency of subject drop in child speech, for the most part, is the
result of children regarding other factors relevant for an entity to be topical than
adults. This does not exclude, however, that there are, as de Cat has observed, cer-
tain language-specific linguistic structures used to mark topicality by both adults
and children, since there may be a certain overlap in what children and adults
identify as a topic.
In order to test the hypothesis that children, at least in the early years of lan-
guage acquisition, may have a different understanding of “topic” than adult speak-
ers, a detailed analysis of child null subject expressions has been carried out in this
study. The results of this analysis, which focuses on examining the person speci-
fication of the omitted child subjects but also considers the frequencies of child
subject drop, will be presented and discussed in detail in Sections 5 and 6. Before
this is achieved, however, subject omissions in adult and child French and German
will be investigated more thoroughly in order to identify the differences between
the two languages and between both adult and child speakers.
4.â•… Subjects and subject omission in adult and child French and German
In languages like French and German, there is a close relationship between sub-
ject and topic, which makes the subject domain an interesting field for studying
topicality. This close relationship between subject and topic is due to the fact that
in such languages “[…] the topic-comment articulation is the ‘unmarked prag-
matic sentence articulation’” (Lambrecht 1994: 132), which means that, if there
is no syntactic or prosodic evidence to the contrary, the subject of a sentence
is automatically interpreted as a topic expression. In isolation, i.e. without con-
textual or prosodic clues, a German sentence such as Die Kinder lieben Fußball
(‘The children love football’) is thus construed by the language user as a topic-
comment structure where the subject die Kinder is identified with the pragmatic
role of topic and lieben Fußball as a property attributed to the children.
Topicality in L1-acquisition
There are, however, a few exceptions to this rule which concern sentences
with a topicalized non-subject constituent as well as so-called event-reporting and
presentational sentences. In the case of topicalization, as for instance in German
Den Mann dort hat sie geküsst (‘That man over there she kissed’), both the topi-
calized constituent (i.e. den Mann dort) and the subject (i.e. sie) can be regarded
as topical expressions. This is because the sentence is intended to not only con-
vey information about the referent of sie, but also about the man (cf. Lambrecht
1994: 147, who suggests that in cases like this the subject may be called the “primary
topic”, the topicalized object the “secondary topic”).
As for event-reporting sentences (i.e. sentences such as Es regnet ‘It is rain-
ing’), which, instead of ascribing a certain property to the subject, merely inform
the addressee of an event that has happened) and presentational sentences (i.e.
sentences such as Es gibt hunderte verschiedener Blumen ‘There are hundreds of
different flowers’), it is commonly assumed that these are entirely focal, since they
represent new information only. Unlike in most other cases, the subject, therefore,
does not have topical status (see e.g. Lambrecht 1994: 133). This is important to
know since in adult French subject omission often affects presentational sentences.
This is illustrated in the following.
Although neither French nor German is a null-subject language, subject omis-
sion is target-like in both languages in certain well-defined contexts (cf. Pillunat
et al. 2006: 9). These contexts differ, however, depending on which of the two lan-
guages is being spoken. In French, subject omission is lexically licensed. The drop
of the subject only occurs with a very restricted number of verbs such as falloir
and avoir and almost always affects 3rd person expletive il as can be seen from the
examples in (2). (2b) illustrates the drop of a subject in a presentational sentence.
Since expletive il only has morpho-syntactic function but neither meaning nor ref-
erence, it does not have topical status. In adult French, therefore, subject omission
is expletive drop, no topic drop.
In German this is different. Here, subject omission is licensed syntactically. As
a V2-language, German allows sentence-initial topic-drop, which means that con�
s�tituents immediately preceding the inflected verb can systematically be omitted in
sentence-initial position only if their referent is unambiguously identifiable from the
discourse. This typically requires that the referent has been mentioned at least once
before in the immediately preceding speech context (see e.g. Hamann 1996: 170;
Nicole Hauser-Grüdl
Pillunat et al. 2006: 9). Subject omission in adult German hence is the omission of
discourse-active, topical information (see (3)).
As can be seen from (3b), the drop may not just affect 3rd person subjects (although
this is the most frequent case (see page 200 above)), but also 1st and 2nd person
ones, provided their referent represents the previously established discourse topic.
Child subject drop differs from adult subject drop in many respects. The rea-
son for this is that children, at least in the early stages of L1-acquisition, omit
subjects irrespective of the restrictions there are for such drop in their respec-
tive L1(s). In French, for instance, children omit topical subjects although only
expletive drop is licit (see examples (4a) and (4b)). In German, children omit
subjects regardless of the syntactic restriction that the drop is only allowed with
constituents immediately preceding the inflected verb in sentence-initial posi-
tion (see (4c)). Furthermore, they leave out subjects which, against the rules of
German subject topic drop, are not identical to the precedingly established dis-
course topic, but which cause a shift in topic instead (see (4d)). This results in a
considerable number of target-deviant, i.e. ungrammatical, subject omissions in
early L1-acquisition.
(4) a. (context: Amélie and adult are looking at a picture showing a boy
dressing himself)
(adult): i met son pyjama /
‘He puts on his pyjama.’
(child): oui / Ø va faire dodo maint’nant /
‘Yes / (He) is going to sleep now.’ Amélie 2;7,6
b. (context: Amélie pretends she is making a cup of coffee)
(adult): alors montre moi comment tu prépares /
‘Okay show me how you do it!’
(child): euh / Ø prends ça /
‘Euh. (I) take this.’ Amélie 2;6,11
Topicality in L1-acquisition
c. (context: Kerstin has said something the adult has not properly understood)
(adult): was macht die mama jetzt nicht mehr? /
‘What has Mum stopped doing?’
(child, not quite answering the question, but referring to what her mother
is doing at the moment):
bett mach Ø jetzt /
‘(She) is making the bed now’ Kerstin 2;9,11
d. (context: Kerstin and adult are looking at a picture showing a child sitting
on a tricycle)
(adult): gucke mal, was macht denn das kind hier? /
‘Look, what is the child doing here?’
(child, turning round to leave the room):
Ø auch auto holen /
‘(I), too, will get my car.’ Kerstin 2;1,2
Characteristically, and this, too, can be seen from the examples in (4), target-deviant
child subject drop is the omission of a topic expression. This means that the sub-
jects left out by the children, in the case of target-deviant drop, typically repre-
sent the entity the child wants to say something about in the sentence. In (4a), for
instance, it is the boy from the picture book about whom Amélie wants to provide
information, namely that he “is going to sleep now”. In (4b), the topic is Amé-
lie herself and the information conveyed by the speaker about herself is how she
proceeds in making a cup a coffee. Target-like child subject drop, however, is not
necessarily the omission of a topical entity, since 3rd person expletives may also be
dropped (see above).
In the analysis of the bilingual and monolingual French and German child
speech data presented in the following, both target-like and target-deviant subject
omissions have been investigated. However, since it is primarily the target-deviant
omissions that represent the drop of a topic expression, it is these which are of
particular interest to this study.
between the languages result from inter-child performance differences (such as,
for instance, different capacities of children’s working memories), as may be the
case in a contrastive study of monolingual French and German L1-acquisition.
The data examined in this study are longitudinal speech data of three bilingual
French-German children as well as one monolingual French child and one mono�
lingual German child – the latter two being included in the study to test if the results
obtained for the bilinguals’ subject omissions can be generalized to monolingual
Â�children, too. The bilingual data were collected in the research project “Bilingualism in
early childhood: Comparing Italian/German and French/German”,2 the monolingual
data have been taken from the CHILDES database (McWhinney & Snow 1985).
As for the bilinguals, all three children examined (Alexander, Amélie and
Céline) lived in Hamburg, Germany, at the time the research was carried out. They
were raised by one German and one French parent who spoke to them in their
respective mother tongue (so-called “one parent – one language principle”, see
Romaine 1995: 181ff.). The children were videotaped during spontaneous interac-
tion with either an adult French or German native speaker in regular two-week
intervals for a period of approximately five years. Each recording was around
30 minutes long for each of the children’s languages.
The monolinguals’ data are based on either audio recordings (in the case of
Kerstin, Miller corpus) or both audio recordings and videotapes (in the case of
Max, York corpus). All recordings were conducted either at the children’s home
or at a location which was at least familiar to the children. Intervals between the
recordings varied depending on the respective corpus and researcher and so did the
period of time for which the two children were studied. Child Max (monolingual
French) was recorded from age 1;9 to 3;2, child Kerstin (monolingual German)
from 1;3 to 3;4.3
To make the data of the bilingual and monolingual children comparable,
approximately one recording per month has been analyzed for each child and each
of its languages from age 2;0 to 3;5 in this study. For some children no recordings
were made at certain age points so that the number of transcripts analyzed slightly
.â•… This project was financed by a grant given to Natascha Müller by the Deutsche ForschungsÂ�
gemeinschaft and was run at Hamburg University from 1999 till 2005. The data gathered in
the project are now being analyzed in the Wuppertal research project “The architecture of
the early bilingual language faculty: a comparison of Italian/German and French/German
children in Italy, Germany and France” financed, too, by the DFG.
.â•… Max is a Canadian child who, at the time of the recordings, lived in Montréal and thus
acquired Québec French. Since no major differences are known between subject omission in
European and Canadian French, Max’s Montréal origin should not prove problematic as to
what is tested in this study.
Topicality in L1-acquisition
varies from child to child.4 On average, 14 transcripts per child and language have
been analyzed, and a total of 112 transcripts are included in this study of child
subject drop.
–â•fi Since I was interested in the relative frequency with which subject omissions
marked for 1st, 2nd and 3rd person occur in proportion to subject realizations
bearing the same feature specification, both omitted and overtly realized sub-
jects were counted and analyzed for their person and number feature.
–â•fi Omissions were divided into target-like and target-deviant omissions, i.e.
those resulting in grammatically acceptable utterances and those resulting in
ungrammatical ones. Since target-deviant omissions in particular reveal pos-
sible differences in the children’s and adults’ understanding of topic, they were
primarily focused on in this study.
.â•… Alexander, for instance, was first recorded at age 2;2 and Max’s recordings stop at age 3;2
already. For Kerstin, only very few recordings are available from age 2;10 onwards.
.â•… The idea that the person feature may play a significant role in the realization and omission
of arguments in early child language goes back, amongst others, to Clancy (1993), Serratrice &
Sorace (2003), Serratrice, Sorace & Paoli (2004), Serratrice (2005) and Allen (2008).
.â•… Evidence that topic expressions may be dropped comes both from German sentence-
initial subject and object topic drop (cf. Fries 1988; Cardinaletti 1990 and Rizzi 1992) as
well as from French object drop in, for instance, imperative constructions (cf. Cummins &
Roberge 2005).
 Nicole Hauser-Grüdl
–â•fi Only those utterances were included in the analysis that contained either a
finite or a non-finite verb form. Others, such as ich da in German or moi là in
French, were not taken into account.
–â•fi Imperatives as well as verb forms which, in child speech, could possibly repre-
sent an imperative were excluded from the analysis, since the syntax of impera-
tives differs considerably from that of declaratives, especially in French.
–â•fi Root infinitive constructions, such as essen das in German or manger ça in
French, were interpreted as incidents of target-deviant subject omission. If it
was clear from the situational context who was meant to be the subject referent
of essen (‘eat’), the person feature of the omitted subject was noted accordingly.
In all other cases, the feature specification was listed as being unknown.
–â•fi Incomplete child utterances, imitations and repetitions were discarded.
–â•fi In total, 9006 subject expressions were analyzed, 3907 French and 5099 German
ones. The number of subjects analyzed per child, language and age period can
be seen in Table 2.
French German
those obtained for German), it can easily be seen that the rate of target-deviant
subject omission is clearly higher in German than it is in French. This is true for
each of the three stages of language development examined (i.e. age 2;0 to 2;5, age
2;5 to 3;0 and age 3;0 to 3;5), but particularly for the first stage, where in German
the rate of target-deviant omissions amounts to almost 40%, whereas in French it
amounts to only 13%. With increasing age of Alexander, the rate of target-deviant
omissions decreases in both languages but, as mentioned earlier, this decrease
takes longer in German than it does in French. Whereas in French, non-target-like
omissions are reduced to 4% at age 3;0–3;5, in German the percentage of target-
deviant omissions still amounts to 10% at the same age period.
100%
90%
80%
70%
60% Target-deviant omissions
50% Target-like omissions
40% Realizations
30%
20%
10%
0%
Age 2;0–2;5 Age 2;5–3;0 Age 3;0–3;5
100%
90%
80%
70%
60% Target-deviant omissions
50% Target-like omissions
40% Realizations
30%
20%
10%
0%
Age 2;0–2,5 Age 2;5–3;0 Age 3;0–3;5
What has been said for Alexander also holds true for Amélie, whose devel-
opment of subject omissions and realizations is illustrated in Figures 3 and 4 for
French and German respectively. Like Alexander, Amélie, too, shows higher rates
of target-deviant omissions in German than in French and she, too, takes longer to
eliminate those non-target-like null subjects in her German L1. In fact, the develop�
ment of the rate of subject omission is almost identical in both children. In German,
they both start with a relatively high percentage of target-deviant omissions of
around 40% at age 2;0–2;5, which they then reduce to 10% or less (7% in the case
of Amélie) at age 3;0–3;5. In French, target-deviant subject drop is comparably low
in both children at all three age periods. The initial rate of 5% target-deviant omis-
sions at age 2;0–2;5 in Amélie and of 13% in Alexander decreases in both children to
around 3% at age period 3;0 to 3;5. Compared to the children’s average rate of 8.5%
target-deviant subject drop in German at this stage, this rate is considerably lower.
100%
90%
80%
70%
60% Target-deviant omissions
50% Target-like omissions
40% Realizations
30%
20%
10%
0%
Age 2;0–2;5 Age 2;5–3;0 Age 3;0–3;5
100%
90%
80%
70%
60% Target-deviant omissions
50% Target-like omissions
40% Realizations
30%
20%
10%
0%
Age 2;0–2,5 Age 2;5–3;0 Age 3;0–3;5
Figure 4.╇ Subject omission and realization in German, Amélie
Topicality in L1-acquisition 
As for Céline, her results are more complicated. Whereas her development
of subject omissions and realizations in German (see Figure 6) is almost the
same as in the other two bilinguals (target-deviant omissions amount to 38% at
age 2;0–2;5 and to 10% at age 3;0–3;5), the results obtained for her French (see
Figure 5) seem, at first glance, to contradict those obtained for Alexander and
Amélie. In contrast to the latter, Céline shows rates of target-deviant French
subject drop which are substantially higher than those of the other two children
and higher, too, than the percentages obtained for target-deviant subject drop in
her German. As can be seen from Figure 5, a rate of 34% non-target-like French
subject drop has been calculated for age 2;5–3;0 and one of 12% for age 3;0–3;5
100%
90%
80%
70%
60% Target-deviant omissions
50% Target-like omissions
40% Realizations
30%
20%
10%
0%
Age 2;0–2,5 Age 2;5–3;0 Age 3;0–3;5
100%
90%
80%
70%
60% Target-deviant omissions
50% Target-like omissions
40% Realizations
30%
20%
10%
0%
Age 2;0–2,5 Age 2;5–3;0 Age 3;0–3;5
(percentages in German are 13% and 10% respectively). For age 2;0–2;5 no sub-
ject omission rate has been determined, since Céline’s French data, at this age
period, only contain 3 utterances with a verb (cf. Table 2, page 210 above), which
is too small a sample to be meaningful.
Céline’s unexpectedly high rate of target-deviant subject omissions in French
can easily be explained, however. It is a likely consequence of Céline being a lan-
guage unbalanced bilingual child, which means that she develops her French more
slowly than her German and more slowly also than the French of Alexander and
Amélie, who both have been shown to be language balanced.7 Céline’s rate of 34%
target-deviant French subject omissions at age 2;5–3;0 hence can be seen to corresÂ�
pond to similar rates of target-deviant subject drop in the other children’s French
at an earlier stage of language development some time before the age of 2. Further-
more, when interpreting Céline’s data, it has to be considered that at a very young
age Céline shows a strong preference for using her German L1, which means
that she hardly ever speaks any French at the early stages of language acquisition.
Therefore, it is not surprising, in fact, that at age 2;0–2;5 only 3 utterances contain-
ing a verb can be found in Céline’s French.8 However, since Céline’s development
in French is clearly delayed, her strikingly high rate of French subject omission
should by no means be considered as counter-evidence to what has been said ear-
lier, namely, that omission rates, in general, are clearly higher in German than they
are in French.
This, too, is confirmed by the results of the analysis of the monolingual children’s
subject omissions, as can be seen from Figures 7 and 8.9 Figure 7 shows for the
monolingual French speaking child Max that target-deviant subject drop amounts
to 11% at age 2;0–2;5 – a rate which equals that of child Alexander at the same
age period. From age 2;5–3;0 onwards, target-deviant omissions are almost non-
existent in Max’s French.
100%
90%
80%
70%
60% Target-deviant omissions
50% Target-like omissions
40% Realizations
30%
20%
10%
0%
Age 2;0–2;5 Age 2;5–3;0 Age 3;0–3;2,23
100%
90%
80%
70%
60% Target-deviant omissions
50% Target-like omissions
40% Realizations
30%
20%
10%
0%
Age 2;0–2;5 Age 2;5–3;0 Age 3;0–3;4,3
.╅ Plural subject expressions, whether omitted or overtly realized, amount to only about
5% in this study.
Topicality in L1-acquisition 
Let us look again at each child separately. As can be seen from Figures 9 and 10
for Alexander, omission rates clearly are highest for 1st person singular subjects,
followed in 2nd place by 2nd person singular subject omissions and finally by omis-
sion of 3rd person subjects.11 Although the actual percentages of the dropped 1st,
2nd and 3rd person subjects vary for French and German (rates of target-deviant
drop are consistently higher in German than in French no matter which person fea-
ture the subjects are specified for), this is the same for both the child’s languages.
100%
90%
80%
70%
60% Realizations
50% Target-like omissions
40% Target-deviant omissions
30%
20%
10%
0%
1. sg. 2. sg. 3. sg.
Figure 9.╇ Relation of 1st, 2nd and 3rd person singular subject omission and realization;
Alexander, French
100%
90%
80%
70%
60% Realizations
50% Target-like omissions
40% Target-deviant omissions
30%
20%
10%
0%
1. sg. 2. sg. 3. sg.
Figure 10.╇ Relation of 1st, 2nd and 3rd person singular subject omission and realization;
Alexander, German
.╅ Only the target-deviant omissions are of interest here, since target-like subject drop, at
least in French, is the drop of a non-topic expression (see Section 4).
 Nicole Hauser-Grüdl
The analysis of Céline’s subject realizations and omissions (see Figures 11 and
12) shows similar results. Again, of all singular subject expressions used by the child,
it is in both languages 1st person subjects which are most frequently omitted. With
omission rates of 1st, 2nd and 3rd person singular subjects of 19%, 13% and again
13% in German, Céline very much performs like Alexander, who drops 19% of his
1st, 12% of his 2nd and 9% of his 3rd person singular subjects. The only difference
is that, whereas Céline shows equal rates of 2nd and 3rd person subject drop, Alex-
ander omits slightly more of all subjects marked for 2nd person than he does of all
subjects marked for 3rd person. In French, subjects of any feature specification are
100%
90%
80%
70%
60% Realizations
50% Target-like omissions
40% Target-deviant omissions
30%
20%
10%
0%
1. sg. 2. sg. 3. sg.
Figure 11.╇ Relation of 1st, 2nd and 3rd person singular subject omission and realization;
Céline, French
100%
90%
80%
70%
60% Realizations
50% Target-like omissions
40% Target-deviant omissions
30%
20%
10%
0%
1. sg. 2. sg. 3. sg.
Figure 12.╇ Relation of 1st, 2nd and 3rd person singular subject omission and realization;
Céline, German
Topicality in L1-acquisition 
more often omitted by Céline than by Alexander, which is likely to be due to her
slower development of French (see above). However, the sequence of the omitted
subjects as to which expressions are left out most frequently is the same in both
children, namely 1st person singular omissions > 2nd person singular omissions >
3rd person omissions.
The results obtained for the third child examined, Amélie, do not quite fit
the observation, based on the other bilinguals’ speech data, that 1st person sub-
jects are most prone to be omitted in both French and German. Whereas Amélie’s
French data, illustrated in Figure 13, show a relatively clear tendency for Amélie to
preferably drop 1st person singular subjects – thereby supporting the observation
mentioned above – her German data, illustrated in Figure 14, do not. Of all German
1st person singular subjects, a rate of 9% is dropped by Amélie, whereas of all 3rd
person singular subjects, 11% are omitted (target-like ones not being taken into
account). In other words, in the case of Amélie’s German, the rate of omitted 3rd
person subjects slightly exceeds that of omitted 1st person ones. Furthermore, of
all the bilinguals examined, Amélie is the only child who, in one of her languages,
namely in German, shows lower rates for 2nd than for 3rd person singular subject
drop. In all other children as well as in Amélie’s French, the percentage of 2nd
person omissions is either higher than or at least equally high as that of 3rd person
subject omissions.12
100%
90%
80%
70%
60% Realizations
50% Target-like omissions
40% Target-deviant omissions
30%
20%
10%
0%
1. sg. 2. sg. 3. sg.
Figure 13.╇ Relation of 1st, 2nd and 3rd person singular subject omission and realization;
Amélie, French
.â•… In Amélie’s French, the relation of 2nd and 3rd person singular subject realizations and
target-deviant omissions is identical (omissions amount to 1% in each case).
Nicole Hauser-Grüdl
100%
90%
80%
70%
60% Realizations
50% Target-like omissions
40% Target-deviant omissions
30%
20%
10%
0%
1. sg. 2. sg. 3. sg.
Figure 14.╇ Relation of 1st, 2nd and 3rd person singular subject omission and realization;
Amélie, German
To summarize the results of the person feature analysis of the bilingual chil-
dren’s subject omissions, one can say that all three children preferably drop 1st
person subjects and that, except for Amélie in German, they do so in both their
languages. As for 2nd person subject drop, of all six child/language combina-
tions examined, there are three cases where 2nd person drop is more frequent
than 3rd person drop, two cases where the relation of 2nd and 3rd person subject
omissions and realizations is identical and one case where 2nd person subjects
are the ones that are omitted the least. In what follows, the results of the bilingual
children’s data will be cross-checked against the results obtained for the mono-
lingual children to determine if they hold true also for monolingual learners of
French and German.
As can be seen from Figures 15 and 16, which illustrate the person specifi-
cation of the monolingual children’s subject realizations and omissions, it is in
both Max’s and Kerstin’s language 1st person singular subjects which are most fre-
quently omitted. The observation, based on the bilingual speech data analysis, that
1st person subjects are the most preferably dropped ones, is thus supported by the
monolingual data. No matter if a child is raised with either French or German or
with both languages simultaneously, 1st person subject omissions seem to be the
ones which occur most often. Amélie’s slightly diverging German data, therefore,
is seen in this study as some kind of individual variation from the norm. Whereas
it is always questionable to try and explain diverging results by means of inter- or
intra-individual variation, it is argued that in the case of Amélie’s German such
an interpretation is licit, since, although omission of 3rd person subjects is more
frequent than that of 1st person subjects, both rates are almost equally high (11% vs.
9%, see Figure 14).
Topicality in L1-acquisition 
100%
90%
80%
70%
60% Realizations
50% Target-like omissions
40% Target-deviant omissions
30%
20%
10%
0%
1. sg. 2. sg. 3. sg.
Figure 15.╇ Relation of 1st, 2nd and 3rd person singular subject omission and realization;
Max, monol. French
100%
90%
80%
70%
60% Realizations
50% Target-like omissions
40% Target-deviant omissions
30%
20%
10%
0%
1. sg. 2. sg. 3. sg.
Figure 16.╇ Relation of 1st, 2nd and 3rd person singular subject omission and realization;
Kerstin, monol. German
As for 2nd person subject omissions, Max and Kerstin slightly diverge from
the majority of the bilingual children in that, like Amélie in German, they omit
fewer 2nd person singular subjects than both 1st and 3rd person ones (in the case
of Max, 2nd person singular subject drop is, in fact, non-existent). What seems
to be a general tendency for most bilingual children, namely that the rate of 2nd
person omissions is either higher than or at least equally high as that of 3rd per-
son subject omissions, hence is not confirmed by the monolingual data. On the
whole, there are three cases in which the rate of omitted 2nd person subjects
exceeds that of 3rd person ones (2nd > 3rd), three cases in which exactly the
opposite (3rd > 2nd) holds true and two cases in which the rate of 2nd and 3rd
person omission is equally high (2nd= 3rd).
Nicole Hauser-Grüdl
6.â•… Discussion
The results of the analysis of the children’s subject omissions as described above
raise two major questions which will be discussed in the following. (1) How can
the quantitative differences observable in both bilingual and monolingual children
between subject omissions in German on the one hand and French on the other
hand be explained? (2) What exactly does the children’s preference to omit 1st
(and in some cases also 2nd) person subjects tell us about the children’s under-
standing of “topic”?
As to question (1) it is argued that the differences in the rate of subject omission
in French and German are due to the differing grammatical systems of the two
languages. This claim is based on observations which show that the higher rate of
subject omission in German than in French, particularly in the early stages of lan-
guage acquisition, correlates with a far more frequent use in child German of root
infinitive constructions. In such constructions, for which examples are given in (5),
the vast majority of subjects are omitted by the children (as shown in (5b)).
(5) a. (context: Alexander is trying to climb onto a cupboard)
(adult): du willst auf den schrank klettern? /
‘You want to climb onto the cupboard?’
(child): ich krabbeln /
‘I crawl’ (where crawl is the infinitive) Alexander 2;2,20
b. (context: Adult and child are talking about what would happen if the child
hurt its knee)
(adult): dann weinst du. /
und was soll dann die mama machen ? /
‘Then you cry.’
‘And what is mum supposed to do then?’
(child): pflaster drauftun /
‘put on plaster’ (where put on is the infinitive and no subject
is mentioned) Kerstin 3;2,8
The above mentioned facts about child root infinitives are illustrated in Figures 17
and 18 for bilingual Alexander. Figure 17 shows the difference in Alexander’s use
of root infinitives in French and German, Figure 18 demonstrates that root infini-
tive constructions are extremely prone to subject drop.
As can be seen from Figure 17, Alexander’s use of root infinitives at age 2;0–2;5
amounts to 29% in German, but only to 4% in French.13 More than four fifths,
.╅ These percentages are consistent with those of Lasser (1997) and Jakubowicz & Rigaut
(1997). The latter find 32% root infinitives in the monolingual German child Kerstin at age 2;1
and 3.9% root infinitives in French children with average age 2;4,10.
Topicality in L1-acquisition
namely 81%, of his root infinitives in German are used without a subject (see
Figure 18). This naturally makes the rate of subject omissions increase in Alexander’s
German and thereby helps to account for the discrepancy there is between the
quantity of his subject omissions in German and in French at age 2;0–2;5 (39% vs.
13% respectively, see page 211 above). The higher rate of subject omissions in
German than in French in later periods of language acquisition (child age 2;5–3;0
and 3;0–3;5) seems to reflect that the French adult system, in which subject omis-
sions are licensed lexically, is easier to acquire for children than the German adult
system, in which subject omission represents a more complex grammatical phe-
nomenon requiring both specific syntactic and pragmatic knowledge.
100%
90%
80%
70%
60%
50% Finite verb forms
Root infinitives
40%
30%
20%
10%
0%
German French
19%
RI + subject omission
RI + subject realization
81%
Figure 18.╇ Subject omission and realization in root infinitive contexts in German;
Alexander (age 2;0–2;5)
How can it be explained, however, – and this brings us to question (2) – that
despite the differing target systems and despite the differences there are between
subject omissions in French and German both languages are alike in that subject
drop mostly affects 1st person entities? Or put differently: What are the possible
reasons that make children preferably omit 1st (and some children also 2nd) person
Nicole Hauser-Grüdl
subjects both in French and in German14 and what may be concluded from this as to
a child’s understanding of “topic”?15 It is argued that two conclusions are possible.
Following Hauser-Grüdl (2007), it could first be argued that the children’s
preferred omission of 1st (and sometimes also 2nd) person singular subjects is
due to the fact that children in the early stages of language acquisition primarily
interpret those entities as topics which are permanently physically present in a
given discourse context and which, therefore, are constantly active in the children’s
memory. This is equivalent to the claim that, whereas linguistic discourse activa-
tion is a decisive factor for topicality and thus subject topic drop in adult speakers,
in children it is rather the psychological status of an entity as a permanently avail-
able referent, i.e. its specification as being either [+speaker] or [+hearer], which
forces its topic interpretation and which makes children drop 1st and 2nd person
subject pronouns.16 Such an explanation of the data supports the hypothesis pre-
sented in Section 3 that children have a different understanding of what a topic is
than adult speakers.
A second, alternative interpretation of the results of the study is as follows. It
could be argued that rather than having a different understanding of topicality, young
children misinterpret 1st and 2nd person deictic subjects as anaphoric expres-
sions, i.e. as expressions in line with 3rd person subjects that refer back to a previ-
ously mentioned discourse-topic (see Schmitz, Patuto & Müller Â�(forthcoming)).
Such an interpretation is based on both the observation that the use of deictic
subject pronouns is more difficult for children to learn than that of 3rd person
ones and the assumption that this is due to a more complex kind of interaction
between pragmatics and syntax in the former. Since there is plenty of evidence
that young children tend to avoid more complex grammatical structures by using
less complex ones instead,17 the interpretation of 1st and 2nd person subjects as
anaphoric could be claimed to be such a case where a more complex analysis is
abandoned in favor of a less complex one.
Evidence for the assumption that the use of deictic subjects poses a problem
for young children comes from child speech samples like the ones in (6), which
illustrate children’s difficulties in assigning correct reference to deictic subject pro-
nouns. In the examples given, the child incorrectly uses a 2nd person singular
subject pronoun to refer to herself instead of using a 1st person one.
(6) a. (context: Child is pointing at her head)
(child): ei- vi- vie- viele vögl aufm kopf hat /
‘One – man- man- many birds has on the head.’
(adult): wer ? /
‘Who?’
(child): du / (referring to herself)
‘You.’
(adult): du hast viel vögl auf dem kopf talli ? /
‘You have many birds on your head, Talli?’
(child): ja /
‘Yes.’ Chantal 2;7,0
b. (context: Child is sucking a coconut sweet; she wants to say she likes the
coconut flavour)
(child): du mag nüsse /
‘You like nuts.’
(where you (German du) refers to the child herself and like
(mag) is either 1st or 3rd person singular)
(adult): du magst nüsse ? /
‘You like nuts?’ (in surprise)
(child): ja /
‘Yes.’
(adult): aber du isst gar keine nuss / […] / da is’n
bisschen kokosnuss vielleicht drin […] /
‘But you are not eating a nut. […]. There may be a bit of coconut
in there […].’ Chantal 2;10,0
.╅ Cross-linguistic phenomena in bilingual children are a very good example here. As
shown for example by Müller et al. (2002), it is always the language which, in a particular
grammatical domain, is the less complex one of a child’s two L1s which exerts influence on
the child’s other L1.
Nicole Hauser-Grüdl
.╅ According to one reviewer of this article, there is also a different explanation for why
children use deictic subject pronouns with verbs of 3rd person inflection, namely that children
treat 3rd person singular verb forms as some sort of unmarked default. There are, however,
examples of cases where children combine a 2nd person subject pronoun with a 1st person
verb form, which cannot be explained by means of a default hypothesis, but rather supports
the analysis put forward in this article. A couple of examples of children’s use of 2nd person
subject pronouns with a finite verb with 1st person inflection are shown in (i) below:
(i) a. (context): Adult pretends he is removing fleas from a toy monkey
(child): was du hole da ? /
‘What are you picking out of there?’
(where German hole (‘pick’) is 1st person singular) Kerstin 2;7,23
b. (context): Child is sliding forwards and backwards with her feet, pretending
she is skiing; the referent she is talking about is herself
(child): du mach gleich [z]agat / (= Spagat)
‘You’ll soon do the splits.’
(where German mach (‘do’), again, is 1st person singular)
(adult) ja / du machst gleich spagat /
‘Yes. You’ll soon do the splits.’ Chantal 2;8,2
.╅ German 1st and 2nd person subject topic drop only occurs in very restricted contexts
such as, for instance, diary, e-mail or telegram writing where the 1st person subject referent
can unambiguously be identified since only one person is involved in the specific situational
context (cf. Fries 1988: 27). The drop of 1st and 2nd person deictic subjects furthermore is licit
in specific question-answer contexts like the one in (ii), where speaker B is explicitly being
asked to provide information about himself, so that, again, the referent of the omitted 1st
person subject is unambiguously identifiable in the given situational context.
Topicality in L1-acquisition
As child speech samples such as the one in (7) suggest, however, this necessity for
subject drop of an unambiguously identifiable subject referent may not be obvi-
ous to child language learners during the early stages of L1- acquisition. In (7), for
instance, which is an example of early child German speech, the child leaves out
a 1st person deictic subject where no such drop is permitted. It is reasonable to
assume that the child interprets the omitted 1st person subject as somehow linked
to the preceding ich-expression in the adult utterance and that this is what triggers
the drop.20
(7) (context: Adult and child are fishing small plastic fish)
(adult): huh, huh, ich hab einen /
‘Huh, huh, I’ve got one.’
(child): ha, Ø habe einen auch /
‘Ha, (I) have one, too.’ Kerstin 2;4,16
The child would thus treat the 1st person deictic subject as if it were anaphoric, i.e.
as if it referred to a previously mentioned, and therefore droppable, discourse topic.
Evidence for such anaphoric topic drop comes from 3rd person subject and object
drop in German as well as from 3rd person object drop in French (cf. Note 6). What
the child does not yet seem to have understood then is that 1st person subjects
cannot be anaphoric in nature since, rather than having one fixed referent, they
may refer to different people depending on who is speaking.
Why adult speakers use 1st and 2nd person subject drop rather infrequently even in examples
in which, like in (ii), the drop seems licit, is still unknown. One possible explanation, suggested
to me by Joachim Jacobs, is that 1st and 2nd person subject drop is no anaphoric drop such
as is typical of German topic-drop constructions. The omitted 1st person ich-expression in
(ii), for instance, is not anaphorically linked to the preceding 2nd person du pronoun but
indirectly via its deictic interpretation. This obviously makes the drop less likely to happen.
.â•… Another, possibly even more convincing example is the one in (iii), which I witnessed
when visiting one of my godchildren who, at that time, was 2;6 years old. While we were
playing Memory, his mother asked me to come and help her with something in the kitchen.
The utterances in (iii) are to be understood as both my own and Tim’s reaction to
her demand:
(iii) (adult): ich komme gleich /
‘I’ll come in a minute.’
(child): Ø komme auch /
‘(I) come, too.’
Nicole Hauser-Grüdl
7.â•… Conclusion
This article has dealt with the question if children have, at least in the early years
of language acquisition, a different understanding than adult speakers of what
a topic is. Since in the two languages examined, French and German, there is a
strong correlation between the (aboutness-)topic and subject, the subject domain
was chosen to investigate children’s understanding of topicality. In consistency with
the concept of topicality as an information structural, language-universal pheno�m��
enon, the study can show that in all children and both languages examined it is
preferably 1st (and sometimes also 2nd) person referents which are omitted.
The discussion of the results has illustrated that two ways of interpreting
the data are possible and that, depending on which of the two interpretations is
adopted, the hypothesis that young children have a different understanding of topic
than adults may or may not be found confirmed. Which of the two interpretations
is more adequate to account for the data – the psychological interpretation suggesÂ�
ted first or the linguistically oriented second one – needs to be further investigated
in future studies, ideally examining child speech data of languages other than
French and German.
Since the two interpretations suggested do not exclude each other, a third
explanation of the results of the study seems reasonable, too, namely that psycho-
logical as well as linguistic factors play a major role in the preferred omission of
1st and partly also 2nd person subjects. This is equivalent to the claim that both
interpretations discussed in this article are correct and that the extensive use
of 1st and 2nd person subject drop is a phenomenon linked both to a different
understanding of children of what a topic is and to difficulties in acquiring the
syntax and pragmatics of deictic subject pronouns. What this study shows, in
any case, is that since no evidence disproving the hypothesis of a different under-
standing of topicality by children as opposed to adults has been found, further
research should always consider the possibility that children, at least in the early
stages of L1-acquisition, may deem other factors relevant for an entity to be topical
than adults.
References
Allen, S.E.M. 2008. Interacting pragmatic influences on children’s argument realization. In Cross-
linguistic Perspectives on Argument Structure: Implications for Learnability, M. Bowerman &
P. Brown (eds), 191–210. New York NY: Lawrence Erlbaum.
Baker, N.D. & Greenfield, P.M. 1988. The development of new and old information in young
children’s early language. Language Sciences 10(1): 3–34.
Bromberg, H. & Wexler, K. 1995. Null subjects in child wh-questions. MIT Working Papers in
Linguistics 26: 221–247.
Topicality in L1-acquisition
Cantone, K., Kupisch, T., Müller, N. & Schmitz, K. 2008. Rethinking language dominance in
bilingual children. Linguistische Berichte 215: 307–343.
Cardinaletti, A. 1990. Subject/object asymmetries in German null-topic constructions and the
status of SpecCP. In Grammar in Progress: Glow Essays for Henk van Riemsdijk, J. Mascaró &
M. Nespor (eds), 75–84. Dordrecht: Foris.
Cinque, G. 1983. ‘Topic’ constructions in some European languages and connectedness. In
Connectedness in Sentence, Discourse and Text, K. Ehlich & H. van Riemsdijk (eds), 7–41.
Tilburg: Katholieke Hogeschool.
Clancy, P. 1993. Preferred argument structure in Korean acquisition. In The Proceedings of the
25th Annual Child Language Research Forum, E. Clark (ed.), 307–314. Standford CA: CSLI.
Cummins, S. & Roberge, Y. 2005. A modular account of null objects in French. Syntax 8: 44–64.
De Cat, C. 2002. Early ‘pragmatic’ competence and the null subject phenomenon. In Romance
Languages and Linguistic Theory 2002: Selected Papers from Going Romance 2002,
R. Bok-Bennema, B. Hollebrandse, B. Kampers-Manhe & P. Sleeman (eds), 17–32. Amsterdam:
John Benjamins.
De Cat, C. 2005. Eliciting topics in child French. Handout presented at the workshop “Language
Acquisition between Sentence and Discourse”, Nijmegen, May 13, 2005.
De Cat, C. 2008. Experimental evidence for preschoolers’ mastery of ‘topic’. In Leeds Working
Papers of Linguistics 13, B. Heselwood & C. De Cat (eds), 76–84. Leeds: University of Leeds.
Emslie, H. & Stevenson, R. 1981. Pre-school children’s use of the articles in definite and indefinite
referring expressions. Journal of Child Language 8(2): 313–328.
Erteschik-Shir, N. 2007. Information Structure: The Syntax-Discourse Interface. Oxford: OUP.
Frascarelli, M. & Hinterhölzl, R. 2007. Types of topics in German and Italian. In On Information
Structure, Meaning and Form: Generalizations across Languages [Linguistik Aktuell/Linguistics
Today 100], K. Schwabe & S. Winkler (eds), 87–116. Amsterdam: John Benjamins.
Fries, N. 1988. Über das Null-Topik im Deutschen. In Sprache und Pragmatik: Arbeitsberichte
3, 19–49. Lund: University of Lund.
Gómez-González, M.A. 2001. The Theme Topic Interface: Evidence from English [Pragmatics &
Beyond New Series 71]. Amsterdam: John Benjamins.
Gundel, J.K., Hedberg, N. & Zacharski, R. 1993. Cognitive status and the form of referring
expressions. Language 69: 274–307.
Hamann, C. 1996. Null arguments in German child language. Language Acquisition 5(3): 155–208.
Hauser-Grüdl, N. 2007. “Topik” im kindlichen Spracherwerb: Eine Analyse des Subjektbereichs.
Ms, Universität Wuppertal.
Hauser-Grüdl, N. 2008. Topikalität im L1-Erwerb: Eine quantitative und qualitative Analyse
von Nullsubjekten bilingual französisch-deutscher sowie monolingual deutscher und
französischer Kinder. Ms, Universität Wuppertal.
Jakubowicz, C. & Rigaut, C. 1997. L’↜acquisition des clitiques nominatifs en français. In Les
pronoms: Morphologie, syntaxe et typologie, A. Zribi-Hertz (ed.), 57–101. Paris: Presses
Universitaires de Vincennes.
Karmiloff-Smith, A. 1981. The grammatical marking of thematic structure in the development
of language production. In The Child’s Construction of Language, W. Deutsch (ed.), 121–147.
London: Academic Press.
Lambrecht, K. 1994. Information Structure and Sentence Form: Topic, Focus, and the Mental
Representations of Discourse Referents. Cambridge: CUP.
Lasser, I. 1997. The interpretation of root infinitive constructions in adult and child German. In
The Interpretation of Root Infinitives and Bare Nouns in Child Language [MIT Occasional
Papers in Linguistics 12], J.C. Schaeffer (ed.), 26–65. Cambridge MA: The MIT Press.
Nicole Hauser-Grüdl
McWhinney, B. & Snow, C. 1985. The Child Language Data Exchange System. Journal of Child
Language 12: 271–296.
Molnár, V. 1998. Topic in focus: On the syntax, phonology, semantics and pragmatics of the
so-called ‘contrastive topic’ in Hungarian and German. Acta Linguistica Hungarica 45(1–2):
89–166.
Müller, N., Cantone, K., Kupisch, T. & Schmitz, K. 2002. Zum Spracheneinfluss im bilingualen
Erstspracherwerb: Italienisch – Deutsch. Linguistische Berichte 190: 157–206.
Pillunat, A., Schmitz, K. & Müller, N. 2006. Die Schnittstelle Syntax-Pragmatik: Subjektauslas-
sungen bei bilingual deutsch-französisch aufwachsenden Kindern. Zeitschrift für Literatur-
wissenschaft und Linguistik 143: 7–24.
Prince, E.F. 1981. Toward a taxonomy of given–new information. In Radical Pragmatics, P. Cole
(ed.), 223–55. New York NY: Academic Press.
Rizzi, L. 1992. Early null subjects & root null subjects. In Geneva Generative Papers, M. Starke
(ed.), 102–14. Geneva: University of Geneva.
Romaine, S. 1995. Bilingualism. Oxford: Blackwell.
Schaeffer, J., Gordishevsky, G., Hadar, G. & Hacohen, A. 2002. Subjects in English-speaking
children with Specific Language Impairment. In Proceedings of the GALA 2001 Conference,
J. Costa & M.J. Freitas (eds), 223–231. Lisbon: Associação Portuguesa de Linguística.
Schmitz, K., Patuto, M. & Müller, N. Forthcoming. The null-subject paraÂ�meter at the interface
between syntax and pragmatics. In Interfaces in Child Language Acquisition [Special Issue
of First Language], P. Guijarro-Fuentes & J. Rothman (eds).
Schwabe, K. & Winkler, S. 2007. On Information Structure, Meaning and Form: Generalizations
across Languages [Linguistik Aktuell/Linguistics Today 100]. Amsterdam: John Benjamins.
Serratrice, L. 2005. The role of discourse pragmatics in the acquisition of subjects in Italian.
Applied Psycholinguistics 26: 437–62.
Serratrice, L. & Sorace, A. 2003. Overt and null subjects in monolingual and bilingual Italian
acquisition. In BUCLD 27 Proceedings, B. Beachley et al. (eds), 739–750. Somerville MA:
Cascadilla Press.
Serratrice, L., Sorace, A. & Paoli, S. 2004. Crosslinguistic influence at the syntax-pragmatics
interface: Subjects and objects in Italian-English bilingual and monolingual acquisition.
Bilingualism: Language and Cognition 7: 183–205.
Wexler, K. 1998. Very early parameter setting and the unique checking constraint: A new expla-
nation of the optional infinitive stage. Lingua 106: 23–79.
Formal and functional constraints
on constituent order and their universality
Peter Öhl
Ludwig-Maximilians-Universität München
This paper addresses the competition between syntactic relations and the
structural encoding of discourse semantic functions, and, more specifically,
the question of how much the distributional restriction of syntactic linearisation
can be constrained by the inventory of functional features of Generative Syntax.
The variation between typologically distinct languages is explained by the
potential of a generative syntactic system to fix constituent order by means
of functional phrases, something not implemented to the same degree in all
languages. Moreover, the various information structural properties a sentence
topic can have imply that notions like topic and comment are not primitives as
such, but that it is more primitive features of perspectivation which determine
the choice of constituents to act as sentence topics.
1.â•… Introduction
The competition between syntactic relations and the structural encoding of discourse
semantic functions, which, according to a commonly held view, yields two different
syntactic systems of natural language, has often been discussed in the literature:
Languages have been said to be either subject-prominent or topic-prominent, or,
following a more concise division, some languages are discourse configurational,
whereas for other languages, the canonical order1 is primarily constrained by
syntactic relations and/or argument structure.
The more recent the research on this topic (cf. Kiss 2001), the more common
the opinion that there is no real sharp division between these two classes. Almost
every language can be said to have either property to at least some degree. The
only statement that remains unarguably true is that languages may differ in their
options for marking discourse functions structurally, as they may differ in their
formal restrictions on linearising the sentence constituents. Whether there is a
.╅ By canonical we understand the normal case, which is the most unmarked one according
to structural regularities.
Peter Öhl
correlation between these two domains of syntactic organisation is not yet clear.
Thus, our challenge is to identify two different classes of universal principles of
ordering and their parameterisation and to elaborate on the features relevant for
ordering. Therefore, we discuss data from and different accounts of languages
where the order of constituents seems to be fixed through various criteria (English,
Hungarian, Italian) and compare them to languages where the order is freer
(German, Japanese, Korean).
The paper is structured as follows: after introducing our basic assumptions
in Section 2, we discuss earlier approaches to discourse configurationality and con-
stituent order (Section 3.1), the identification of subject and topic positions (3.2),
and the classification of languages according to parameters in systems with (3.3)
and without (3.4) functional phrases (henceforth FPs). In Section 4, we develop a
model of interacting constraints on constituent order. We take a closer look at the
notion of topicality and discuss further options of information structuring and
the evidence this gives for and against FPs. Finally, we propose a model of features
constraining the linearisation in different kinds of syntactic systems.
.â•… Of course we do not claim that this is the only semantic aspect of linearisation, it is just the
most basic one. Scope relations in quantification, the definiteness or specificity of referents, and
other similar properties are all semantic factors applying in addition to (though not necessarily
after) the conceptual order. More on this follows in Section 4.
Formal and functional constraints on constituent order and their universality
(1) a. [IP The police (AGENT; nom pl) have (3pl) [VP put a linguist
(PATIENT; acc sg) into jail].
b. [IP No linguist (PATIENT; nom sg) has (3sg) [VP been put into jail
[PP by the police (AGENT)]].
Thirdly, there appears to be a cognitive need for fitting information into the hearer’s
knowledge store. There are pragmatic rules concerning the systematic selection and
structuring of information and its integration in a (linguistic) context, determined
by the need to make the hearer comprehend the message communicated. Thus the
constituents may be ordered by being packed in a syntactic information structure.
Vallduví (1992: 15; 53ff.) defines information packaging in terms of instructions to
the hearer:
.â•… We ignore here other relational systems, such as the ergative-absolutive one, which we
regard as variants of the universal formal condition of linking conceptually hierarchical
semantic roles to arguments ordered according to a language specific case system.
Peter Öhl
Vallduví 1992: 43ff.) and a comment-like part.4 We also adopt the well-founded view
that constituents in their basic order and with normal accentuation are unmarked
with respect to information structure, which means that they potentially represent
maximal focus (i.e. sentence focus in the terms of Lambrecht 1994: 223).5 Discourse
semantic features such as familiarity, salience, point of view etc. apply in addition
to the basic syntactic rules and may change both the order and the accentuation
of defocused elements (cf. Cinque 1993; Höhle 1982; Abraham 2007 and Molnárfi
2007 for German). We suggest subsuming them under the term perspectivation,
which we borrow from functionalist work such as Graumann & Kallmeyer (2002;
see Section 4.2). How they apply will be as much a point of our discussion as the
question of how many of them are necessary or sufficient features of topicality.
We want to emphasise that we consider syntactic relations primarily a formal
aspect of sentence structure, as we do argument structure and other semantic
factors of linearisation. The discourse configuration is primarily a functional
aspect of structure building. Taking the proper distinction of form and function
seriously, we strongly reject any account attempting to explain the notion of gram-
matical subject functionally, as proto-topic (Lambrecht 1994: 131) or mediator of
topicality (Sasse 1995: 1065):
Subjects are essentially topics that have become integrated into the case frame
of a verb. (Sasse 1995: 1067)
In a relational system (…), the primary grammatical relation (PGR) indicating
the topic has obvious implications of semantic roles and can therefore be used to
denote them.6 (Sasse 1982: 276)
.â•… Vallduví (1992: 53ff.) argues that the term ‘comment’ can be abandoned, if the packages of
information are divided into focus and (back)ground. We agree that the term as used in the lit-
erature is imprecise and misleading if it is not made clear whether defocused elements belong
to the comment or not. In Vallduví’s (1992: 53ff.) system, they form the ground together with
the link, i.e. an aboutness topic. We will continue to use the term comment in a rather more
traditional and theory-neutral sense, as something that is said about one (or more) topic(s),
ignoring the differentiation of focused and defocused constituents it may contain.
.â•… Lambrecht (1994) distinguishes the sentence focus of thetic sentences from the predicate
focus of categorical sentences and from argument focus, i.e. focus on just one constituent.
.â•… Original: “In einem relational gerichteten System, gleichgültig ob es ergativisch oder
akkusativisch funktioniert, enthält die TOPIC-anzeigende primäre grammatische Relation
(PGR) eindeutige Implikationen semantischer Rollen und kann daher als Mittel zu ihrer
Bezeichnung eingesetzt werden.”
Formal and functional constraints on constituent order and their universality
We provide some evidence showing that neither topicality nor a semantic role can
define a ‘grammatical subject’. Data from languages like English or German clearly
show that the presence of a subject is independent from any discourse function.
(3) a. [IP The police (nom pl) are (3pl) [VP coming]]. (thetic sentence)
b. Die Polizei (nom sg) kommt (3sg).
(4) a. [IP There has (3sg) [VP happened an accident (nom sg)]].
(presentational sentence)
b. Es geschah (3sg) ein Unfall (nom sg).7
Both thetic (Jacobs 2001: 674; Frey 2004a: 11) and presentational sentences (Kuno
1972: 299; Frey 2004a: 11) belong to the so-called ‘anti-topic constructions’ (Jacobs
2001: 674). This does not at all touch the subject condition, however. Moreover,
subjects can even be fully unsuitable as topics, as shown by the following examples
from German, where the prepositional object in (5a) and the temporal adverbial
in (5b), respectively, are the sentence topics.
(5) a. Damit hatte niemand gerechnet. (negative quantifier)
that-with had nobody calculatetd
‘Nobody expected that.’
b. Damals wurde ein Knabe geboren.
then was a boy born
(unspecific indefinite subject; Kiss 1996: 120ff.)
Both temporal and locative adverbials are generally quite suitable as topics:
(6) a. In Wuppertal leben etwa 360.000 Menschen.
in W. live about 360,000 people
b. Im Jahr 2007 lebten in Wuppertal etwa 360.000 Menschen.
in-the year 2007 lived in W. about 360,000 people
The sentence initial referents in (6) are as topical as they are in (7).
(7) a. Wuppertal hat etwa 360.000 Einwohner.
W. has about 360,000 inhabitants
b. Das Jahr 2007 hatte 365 Tage.
the year 2007 had 365 days
.â•… Here and below we use italics to highlight constituents referred to in the text, in this
case the subjects. In order to indicate pitch accented syllables, we follow the convention of
using small capitals. Occasionally, in later examples, rising and falling tones are indicated
by ‘/’ and ‘\’, respectively.
Peter Öhl
The functional overlap of subject and topic is empirically much smaller than is
often assumed. In our view, there is no direct relation between topicality and any
one syntactic function. Therefore, subjects cannot be something like a ‘grammati-
calised topic’. Moreover, they cannot be identified with a salient semantic role (like
AGENT) either. It is well known that subjects can even be patients or carry no
semantic role at all:
We will defend the view that the problem can be solved by a small number of
assumptions: firstly, there is no universal hierarchy of FPs. Instead, languages
differ parametrically by the number of phrases hosting specific kinds (or classes)
of constituents. The fewer FPs there are in the syntax of a language, the more
options it has for the linearisation of constituents. Secondly, instead of a universal
hierarchy of phrases, there is a global hierarchy of structural, conceptual and
discourse-functional features determining the parameterisation of FPs on the one
hand, and the linearisation of constituents in the so-called ‘free word order lan-
guages’ that are characterised by only few FPs, on the other hand. The hierarchy
of these features is not absolute but relative and rankable, the options depending
on the class a specific feature belongs to. The reason why word order is more or
less fixed in different kinds of languages is that different kinds of FPs are fixed by
language acquisition, blocking the structural variation in performance.
All of these properties except (c), which can be attributed to properties such as the
high productivity of scrambling in SOV languages (cf. Section 4.2 below), can be
exemplified by the sentences in (10).
(10) a. sakana wa tai ga oisii
fish top Tai nom delicious
‘Speaking of fish, Tai is delicious.’ (Japanese; Li & Thompson 1976: 468)
b. pihengki nun 747 ka khu-ta.
airplane top 747 nom big-stative
‘Speaking of airplanes, the 747 is big.’ (Korean; ibid.)
Subject-prominent languages, on the other hand, are said to have the following
characteristics (among others; abstracted from Li & Thompson 1976: 466f.):
(11) a. Subject-prominent languages have a canonical subject position.
b. Subject-prominent languages have a regular active/passive diathesis.
c. Subject-prominent languages have formal subjects without any semantic role.
More recent Generative accounts (e.g. Kiss 1995, 2001) broaden this spectrum to
a more general division into discourse configurationality and the prominence of
grammatical relations, both of them assumed to be bound to specific positions.
Formal and functional constraints on constituent order and their universality
This means that the syntactic configuration is criterial for typological classifica-
tion rather than morphological marking. Languages like Japanese, which have a
canonical position for topics but not for subjects, are then to be regarded as typical
discourse configurational languages.8
Kiss (1994, 2002) compares mainly English and Hungarian as prototypes of
their classes. She makes generalisations about the treatment of subjects and topics
on the basis of data as follows (Kiss 1995: 7f.):
(13) a. Fido is [VP chewing a bone]. (categorical sentence)
b. A dog [VP came into the room]. (thetic sentence)
Kiss (1995: 7f.) assumes that the two languages basically differ by one property: the
constraint forcing a constituent to leave the VP. In English, the grammatical subject
is always fronted. In Hungarian, it has to be topical. Moreover, the following data
show that non-subjects are also fronted if they are topics:
(15) a. *Egy kutya [VP van a szobában]
b. A szobában [VP van egy kutya]
the room-into â•… came a dog
According to Kiss, this suggests that English and Hungarian differ by the inventory
of canonical positions in at least one respect: Whereas English has one for the
subject, Hungarian has one for the topic. Since, unfortunately, Example (15b) from
.â•… Note that Kiss (1995: 6) calls Japanese one of the best known examples of discourse
configurational languages.
Peter Öhl
Kiss (1995) resembles cases of locative inversion in English where the corresponding
order is also grammatical, some brief discussion is needed.
(16) [[PP into the room] [VP came a dog]]
Thus, there is no canonical topic position in English but there is a canonical subject
position. In Hungarian, however, background subjects always stay in situ, whereas
topical elements are fronted independently of their grammatical function:
(20) A szintaxist jól [VP tudják a diákok]
the syntax-acc well â•… know the students-nom
.â•… Since locative inversion is not possible with transitive verbs, Alexiadou & Agnastopoulou
(2001) argue that the ‘subject’ in situ does not have to move on grounds of the option to leave
one case feature in the VP unchecked.
Formal and functional constraints on constituent order and their universality 
that there is a position SPEC/IP for an argument agreeing with the finite verb to be
assigned nominative case.
The EPP is another principle regulating syntactic structure […]: sentences must
have subject positions, [Spec/IP] positions, at all syntactic levels. It is important
to point out here that the EPP imposes that the [Spec/IP] position be generated.
(Haegeman 1994: 339f.)
In the feature based phrase structure model of the Minimalist Program, the
concept of the EPP was saved but implemented by a formal feature that is checked
irrespectively of case and agreement,10 which allows accounting for subjects in situ
agreeing with the verb (see 16 and 21a above):
I suggest that the strong feature in this instance is an “EPP feature” residing in
Agr, hence the same feature that drives overt subject raising, the modern technical
implementation of the EPP. (Lasnik 2001: 81)
Note that this formal way of deriving obligatory SPEC positions is not bound to a
specific phrase like IP. Thus, it can also be applied to other functional projections
and it can be considered a matter of parameterisation whether a SPEC position
has to be filled or not (the generalised EPP from Chomsky 2000: 109). Language
specific properties like V2 in German can be explained in an elegant way by the
application of the EPP to a higher FP (e.g. CP) plus an independent condition
making the finite verb move to C0 (see also Roberts & Roussou 2002):
(22) [CP Linguisten [C’ sind [IP immer wieder Fehler unterlaufen]]]
╅ linguists-dat ╅╛↜渀屮are ╅╛↜渀屮always again errors-nom happen-past-ptcp
‘Errors were made by linguists again and again.’
Very much in this sense, Kiss (1995: 6f., 14) distinguishes between the canonical
position of grammatical subjects and notional subjects (i.e. the subjects of predication).
In subsequent work (Kiss 2002), she suggests that a subject-prominent language
like English is characterised by the EPP parametrically yielding the obligatory
.╅ It is not clear to us why only some arguments but no adjuncts can check this feature.
This is another technical problem that cannot be discussed here.
Peter Öhl
filling of SPEC/IP. Languages like Hungarian, on the other hand, do not have to fill
SPEC/IP. A non-notional subject stays inside the VP:
(23) a. [IP a guest [VP has telephoned]
b. [IP[VP telefonált [DP egy vendég]]]
â•…â•… telefoned â•… â•⁄ a guest (Kiss 2002: 109)
In the following subsection, we take a closer look at the two kinds of canonical
positions and how they are identified in the Generative model of phrase structure.
Assuming the orders in (25) to be canonical, authors have made several proposals to
derive the data in (26). Firstly, Italian being a pro-drop-language, data like (26a)
.╅ Note that earlier GB accounts like Rothstein (1985; also quoted by Chomsky 1986: 4, fn. 5)
derived the EPP from a condition of predicate linking.
Formal and functional constraints on constituent order and their universality
.╅ Note that there is also independent distributional evidence for the movement of non-finite
verb forms to a rather high position in Italian (Giusti 1995: 1352).
.╅ Note that left dislocation also indicates topicality in languages like German (Jacobs
2001: 659).
Peter Öhl
The right peripheral topic in (28b) is doubled by a clitic, exactly like the left
peripheral ones in (27). Subjects are never doubled by a clitic in pro-drop Italian.
Thus, defocused postverbal subjects can be treated on a par with topics.
Moreover, the options of ordering multiple topical elements are equivalent
on both sides of the sentence (Frascarelli 2000: 139). Furthermore, Frascarelli
(139; 161ff.) provides several arguments for the position of both kinds of topics
being identical and proposes that the whole FocP moves to the left in cases of right
topicalisation. We do not want to go into the details of this analysis but would
also like to refer the reader to Vallduví’s (1992: 85; 101) analysis of right periph-
eral topics in Catalan. He proposes a model of mirrored adjunct positions for
topics in the same c-command position, which could also be translated into left
and right specification of a TopP. The difference is not really crucial from our
point of view. The advantage of such a model would be the replacement of the
notion of precedence by a notion of c-command. Either option, both movement
of the FocP and mirroring, provides the possibility of a canonical topic position
in phrase structural terms also for languages like Italian. Assuming a canonical
subject position on the grounds of these observations the following hierarchy
is derived:
(29) TopP > FocP > IP > VP
Kiss (2002) proposes a different hierarchy of FPs for Hungarian. Firstly, she assumes
no IP. Secondly, there is evidence for a position specific to strong quantifiers.
(30) TopP > QP > FocP > VP
Kiss (2002) states that quantified and focalised expressions precede the verb but
follow sentence adverbials which express the speaker’s attitude (31a). Adverbials
are often taken as indicating the borders between specific positions for different
classes of constituents (cf. Cinque 1998). She generalises elements preceding them
as topics and, thus, as identifying a topic phrase (31b).
(31) a. Szerintem [QP minden diák [FocP a szintaxist
in-my-opinion â•… every student â•…â•… the syntax
[VP szereti legjobban]]]
â•… likes best
b. [TopP A diádok [szerintem [VP jól [VP tudják a szintaxist]]]]
╅╅↜the students in-my-opinion ╅ well ╅ know the syntax
topicality (cf. also Reinhart 2004: 296), Frey (2007) concludes that the subject
preceding the sentence adverbial in sentences like those below is a topic:
(32) a. Weil er müde war, hat ein Student leider während
because he tired was has a student unfortunately during
der Vorlesung geschlafen.
the lecture slept (Frey 2007: 333)
b. *Weil er müde war, hat leider ein Student während der Vorlesung geschlafen.
On the basis of data like the above, (Frey 2004a+b; 2007) makes a strong assumption
about a canonical (or designated) topic position in German:
(34) Designated Topic Position
In the middle field of the German clause, directly above the base position
of sentential adverbials (…), there is a designated position for topics (in the
aboutness sense): all topics occuring in the middle field, and only them,
occur in this position. (Frey 2007: 232)
Frey does not use a model with a fixed hierarchy of FPs. Instead, he assumes
that FPs vary according to the class of elements they can host. His topic position
is below the FinP hosting the finite verb. Contrasted elements are located in a
ContrP dominating FinP (35a). Non-contrastive topical elements move to the
C-domain only in the absence of contrastive ones, and in one of two scenarios:
firstly, when there is formal movement of the highest element from the so-called
middle field to SPEC/FinP (35b) due to an EPP-like condition that one specifier in
the C-domain has to be occupied; secondly, left dislocation, which he also regards
as a case of topicalisation, bringing topics to SPEC/CP. This is possible both in
root clauses where SPEC/FinP is occupied by a co-indexed pronoun (35c) and in
subordinate clauses with a complementiser (35d), where the co-indexed pronoun
is in the actual topic position.
(35) a. [ContrP Mit dem Hammer [FinP[Fin’ hat [TopP Otto [das
╅╅╇╛↜with the hammer ╅╅ ╅ ↜渀屮has ╅╅ Otto ╗↜the
Fenster eingeschlagen]]]]]
window smashed
b. [FinP Otto1 [Fin’ hat [TopP t1 [das Fenster eingeschlagen]]]]
╅ Otto ╅ ╇ ↜渀屮↜has ╅╅╅ the window smashed
c. [CP Den Otto1 [FinP den1 [Fin’ mag [TopP t’1 [jeder t1]]]]]
╅ the Otto ╅ ╇ him ╅╇↜likes ╅╅ ╅ ╛everyone
Peter Öhl
d. Jeder glaubt, [CP den Hans1 [C’ dass [FinP [TopP den1 [jeder t1 mag]]]]]14
everybody thinks ╅ ╇ the John ╅ that ╅╅╅ ╅ him everyone likes
He concludes that the canonical topic position is not in the so called prefield but
topmost in the middle field, which yields the following hierarchy for German:
(36) CP > ContrP > FinP > TopP (Frey 2004b: 29)
yes e.g. English, Italian, German (?) e.g. Hungarian, Italian, German (?)
no e.g. Tagalog, Hungarian, German (?) e.g. Tagalog, English, German (?)
.╅ Note that this example taken from Frey (2004b) is considered ungrammatical by speakers
of northern German varieties. Sentences like these are quite common in the south of Germany,
though, especially in Bavaria.
Formal and functional constraints on constituent order and their universality
the relevant features apply more freely, and that languages like German (and
also Japanese and Korean) differ from most languages in this table with regard to
this property.
There has been much discussion of the status of German as being either
topic- or subject-prominent from the typological point of view (e.g. Lötscher 1992).
Even though German has some of the features typical for ‘subject-prominence’ in Li
& Thompson’s (1976) classification, it also has properties of topic-prominence. So
called ‘double subject constructions’ and free nominatives (‘nominativus pendens’)
as hanging topics occur with high frequency at least in the spoken language:
(37) a. %Bäume stehen dort nur (noch) Tannen. (double subjects)
trees stand there only ↜yet firs
‘As far as trees are concerned, there are only firs left there.’
b. %Rotwein schmeckt mir (eigentlich) nur Bordeaux.
red-wine tastes to-me ↜actually only B.
‘As far as red wine is concerned, I actually like nothing but Bordeaux.’
It is also significant that there are verbs which can fully dispense with an
overt subject:
(39) a. Mir graut vor aller Theorie.
pron-(dat) cause-shudder by all theory
‘I shudder to think of any theory.’
b. Mich dürstet nach Wissen.
pron-(acc) thirst(verb)-3sg after knowledge
‘I am thirsty for knowledge.’
That the order DAT-NOM is not derived by scrambling but is actually the
basic one is also shown by the fact that it is the only grammatical order found in a
fronted VP (Haider 1993: 132ff.):
(43) a. [Syntaktikern Fehler unterlaufen]i sind immer wieder ti
syntacticians-dat mistakes-nom happen-past-ptcp are always again
b. *[Fehler Syntaktikern unterlaufen]i sind immer wieder ti
Examples like these serve Haider (1993, 1997b; 2000, 2010) with a major empiri-
cal argument against a canonical subject position in German that the fronted VP
can contain the subject with ergative and passivised transitive verbs shows that
there is no English-like syntactic requirement of ‘externalisation’:
(44) a. [VP Bäume ausgerissen] wurden hier heute noch nicht.
â•… trees pulled-out were here today still not
(passivised transitive verb)
b. [VP Zuhörer eingeschlafen] sind uns aber, Gott sei Dank,
â•… listeners slept-in are us-dat but god-dat be thank
auch nicht. (unaccusative verb)
also not
‘But we didn’t have any listeners falling asleep either, thank God.’
That it is the VP and not the IP which is fronted is indicated by the fact that the
phrase preceding Vfin in C0 may not contain its trace (Haider 1993: 151):
(45) a. [VP ein Zug angekommen (*tk)]i istk hier noch nie ti
â•… a train arrived is here still never
b. *[IP ein Zug an tk]i kamk hier noch nie ti
Obviously, German subjects do not move to SPEC/IP, but stay in situ in the
unmarked case (whereas cases of subject in situ, as in locative inversion, are clearly
the marked case in languages such as English). The two potential explanations
in the Generative model are that either the EPP does not hold for the IP in
German, or that German does not have an IP at all – and thus no canonical subject
position (Haider 1993: 142ff.; 1997a+b; 2000, 2010: 45–85; detailed discussion of
arguments against an IP in German can also be found in Sternefeld 2006: 507ff.
and in Öhl 2003: 104–134). Haider (2010: 271) puts it like this: “the word order is
not determined by case licensing requirements but by the ranked lexical argument
structure that determines the order of projection/merger”.
There are also reasons to doubt whether the criteria presented in Section 3.2
for the identification of a TopP in German are sufficient. Firstly, it is obvious that
more than one sentence adverbial can occur in a sentence. If the positions of these
adverbials were fixed, one would expect that nothing could intervene between them.
Peter Öhl
Moreover, all topics, but nothing else, should precede them. Neither prediction
seems to hold, however. In the following examples, there are no topics given by
the context. That each of the three answers to the question in (46) is thetic is
also indicated by the initial expletive that cannot occur in sentences with topical
subjects (Frey 2004a: 11).
(46) You are looking so glad – what do you expect?
According to our native intuition, the main difference between the sentences in
(47a) vs. (b) and (c) is that the subject in (47a) is non-specific. This can be easily
confirmed by a discourse continuation like “I am going to introduce him to you
soon.”, which would be possible only in (47b+c). Thus, the specific subject can occur
in two positions. It seems that in German, these adverbials may precede, follow,
or frame other constituents. Whether all elements preceding sentence adverbials
in languages like Hungarian are in fact topical is difficult for us to test. Assuming
them to be universal indicators of topicality, Kiss (1996: 128ff.) proposes two
different ‘subject positions’ in English, one being associated with topic features.
(48) a. In most cases, boys will be born.
b. *Boys will in most cases be born.
c. Boys will in most cases know the novels of Karl May.
However, if the subject in (c) above is in a higher position than in (a), why should the
modal will also be forced to move there? It is not in other cases of topicalisation:
(49) *The novels of Karl May will in most cases boys know.
In our view, data like these only show that an adverbial taking scope over the
proposition cannot be adjoined below a subject within sentence focus. Sentence
adverbials precede the discourse semantically unmarked constituents, but they can
follow defocused constituents on the grounds of factors which will be discussed
more deeply in Section 4. For the time being, we just state that the data in (47) do
not support the assumption of a fixed position for sentence adverbials in German.
Moreover, constituent order also varies when there are several elements that
are topical according to the assumption that cataphors are a reliable indicator of
topicality (cf. Frey 2007; Reinhart 2004).
Formal and functional constraints on constituent order and their universality 
(50) a. Weil sie ihn interessierte, hat ein Student diese Vorlesung
because it him interested has a student this lecture
erfreulicherweise ganz aufgezeichnet.
to-my-pleasure totally recorded
b. Weil sie ihn interessierte, hat ein Student erfreulicherweise diese Vorlesung
ganz aufgezeichnet.
c. Weil sie ihn interessierte, hat diese Vorlesung erfreulicherweise ein Student
ganz aufgezeichnet.
d. ?Weil sie ihn interessierte, hat erfreulicherweise ein Student diese Vorlesung
ganz aufgezeichnet.
e. ??Weil sie ihn interessierte, hat erfreulicherweise diese Vorlesung ein Student
ganz aufgezeichnet.
f. Weil sie ihn interessierte, hat diese Vorlesung ein Student erfreulicherweise
ganz aufgezeichnet.
The orders in (50b) and (c) are as natural as those in (50a), and not even the order
in (50d) is fully bad. The order in (50e) is rather marked because of scrambling
in the basic focal domain, whereas (50f) shows that different orders above these
adverbials are possible. This also indicates that the order may be determined by
various factors. We shall return to this point in Section 4 as well, where we investi-
gate the distributional options of non-focal elements in more detail. Again, we just
state that the existence of a canonical topic position for German is debatable.15
There is also Italian evidence for several potential topic positions. It is for this
reason that the cartographic split-CP-approach of Rizzi (1997) has to build on the
assumption of at least two TopPs, one between the positions of complementisers
and focused elements, and one between focused elements and Rizzi’s FinP.
(51) Credo [ForceP che [TopP a Gianni [FocP QUESTO [TopP domani
think-1sg ╅╅ that ╅╇ ╛↜渀屮to G. ╅╅ THIS ╅╇ tomorrow
[FinP [IP gli dovremmo dire]…]
╅╅╅↜渀屮objcl must-fut-1pl say
‘THIS, I think we have to tell to John tomorrow.’ (Rizzi 1997: 295)
.╅ Among others, Abraham (1997) also argues for such a canonical topic position on the
grounds of functional projection. In order to avoid misunderstanding, it might be necessary
to draw the attention to some of our assumptions defended in the course of this paper. We
would like to emphasise that we are not arguing against discourse configurational properties of
German in general. However, we assume two factors supporting discourse configurationality:
first, the existence of specific functional phrases like TopP; second, the absence of functional
phrases restricting word order, which facilitates free serialisation according to information
structural features by means of adjunction. See also the arguments against functional phrases
between the German VP and CP in Haider (2010: 45–85).
Peter Öhl
Rizzi (1997) obviously takes different kinds of thematic elements to be of the same
class. Since more than one of them can occur in both partitions of the split CP, he
simply states that the TopPs in his system are iterable.
(52) [ForceP [TopP Il libro [TopP a Gianni [TopP domani
╅╅╅╅ the book ╅╅ to Gianni ╅╇ tomorrow
[FinP[IP glielo darò senz’altro]…]
╅╅╇╛╛╛indirObjCl.dirObjCl give-fut.1sg surely (Rizzi 1997: 290)
Rizzi (2001) found evidence for even another potential Italian topic position. The
sentences in (53) show that the Italian subordination marker che cannot follow a
topic (53a), whereas the interrogative complementiser se can (53b). He concludes
that there is an Interrogative Phrase (IntP) below ForceP, with another potential
topic position above it:
(53) a. *Credo, a Gianni, che avrebbero dovuto
think-1sg dat G. that aux-pastperf-subj-3pl must-pastpart
dirgli la verità.
say-inf det truth
‘I think that they should have told the truth to John.’
(Italian; Rizzi 2001: 289)
b. Non so, [ForceP [TopP a Gianni [IntP se [[IP avrebbero
neg know-1sg ╅╅╅╅╇ dat G. ╅╇╛↜if aux-pastperf-subj-3pl
potuto dirgli la verità]…]
can-pastpart say-inf det truth
‘I do not know if they could have told the truth to John.’ (Rizzi 2001: 289)
Similar accounts have been proposed by Roussou (2000: 79ff.) for Greek and by
Öhl (2004: 165) for Persian and Bengali. The options of positioning topics are
very clearly not bound to a single canonical position in these languages. Moreover,
Benincà & Poletto (2004) observe that the distribution of different kinds of topics
in Italian is, in fact, restricted to different layers of the C-domain. Instead of
iterable topic phrases, they propose a more differentiated hierarchical order of FPs
which they relate to specific discourse semantic functions:
(54) Sublayers of the C-Domain (Benincà & Poletto 2004: 73)
[ForceP [hanging topic [scene setter [left dislocation [list interpr.
frame theme
focus
Formal and functional constraints on constituent order and their universality
Thus, the following structure should be ungrammatical, too. The given example,
however, is not.
(57) [VP (*tk) Fehler unterlaufen]i sind [TopP Syntaktikernk [immer wieder ti]]
Thus it is improbable that topics and comparable elements are moved from the
VP to structurally higher positions in FPs. Since it is implausible to assume that
V’ can be fronted to SPEC/CP, the fronted phrase must be a complete VP. This is
possible if we assume that the fronted phrase is a segment of VP, i.e. if the whole
VP is analysed as an adjunction structure. We follow Haider (1997a+b; 2000) in
assuming that head final VPs are not constituted by shells that are projected by
separate heads like v0, but rather are iterated or extended by adjuncts licensed to
the left by V0.
(58) a. [VP Kindern Märchen erzählt] haben [VP Großeltern [VP schon
╅╇ to-children fairy-tales told have ╅ ╛↜grandparents ╅ ╛↜yet
immer [VP gerne [VP ti]]]]
always ╅ ╛↜gladly
b. [VP Märchen erzählt] haben [VP Großeltern [VP Kindern [VP schon immer
[VP gerne [VP ti]]]]]
Peter Öhl
Thus, only arguments which do not have arguments above them in the Θ-hierarchy
are found outside of a fronted VP.
(60) a. Syntaktikerni sind ti wahrscheinlich schon oft [VP solche Fehler unterlaufen].
b. [VP Solche Fehler unterlaufen]k sind Syntaktikerni wahrscheinlich schon oft tk.
Abraham (2007: 194ff.) also argues against movement of elements out of VP:
‘Themata are “born” outside, i.e. to the left of VP, whereas rhemata are “born”
inside VP’ (Abraham 2007: 197). We agree with Abraham (2007: 197) that what he
calls thematic elements (we subsume under them topics, discourse old information,
etc.) are base generated in layers above VP interacting with discourse semantics.
Formal and functional constraints on constituent order and their universality
Furthermore, as will be shown in the next section, the order of the elements in this
partition is not as strict as a cartographic approach would predict. Thus we also
agree with accounts like Abraham (2007) and Molnárfi (2007) which state that
linearisation in languages like German is not triggered by ‘formalised pragmatic
features’ (Molnárfi 2007: 176).
Two needs are made obvious by the observations presented in the previous section.
Firstly, if there are, in fact, numerous potential positions for topical elements
and if we still want to assume canonical positions for them as we do for subjects,
we have to find a proper means of identifying these positions. Secondly, it does
not really seem clear how topical elements should be classified. What we need
first of all is a more differentiated model of the semantic and pragmatic criteria
for linearisation. In past research, discourse configurational properties of many
languages have been stated on the basis of empirical observations. Besides those
mentioned above, there are data from Catalan, Romanian, Bulgarian, Russian,
Greek, Nepali, Hindi, Finnish, Arabic and many more (cf. Kiss 2001). Besides
diverging accounts of focus-prominence, we find different uses of the term topic
(as well as of terms such as thematicy and familiarity) that are even more prob-
lematic. Quite often, they are confused with other discourse functional features
and abused to subsume them (cf. Vallduví 1992: 28ff.). Therefore, we would like
to continue by discussing the notion of topicality.
concerned. Just as there can be non-topical old information included in the focus,
there are newly introduced ‘shifted’ topics (as discussed by Frascarelli & Hinterhölzl
2007: 88, 109), probably restricted by an accessibility condition for their being used
as a link for information storage (Vallduví & Engdahl 1996: 498). The following
sentence shows that a topic can be formally licensed although not given contextually
nor by inference:
(62) When she was five years old, a child of my acquaintance announced a theory
that she was inhabited by rabbits. (Reinhart 2004: 296)
Why should a propositional utterance not be ‘about’ more than one discourse
referent, however? Lambrecht (1994: 150) suspects that Reinhart’s (1981) observa-
tions concern the “pragmatic salience of the various topic referents at given points
of the discourse [rather than] the difference between topics and non-topics”. We
also think that Reinhart’s (1981, 2004) restrictions limit the range of the term
topic too much. Firstly, she analyses only NP topics. Reinhart (1981: 56) states that
her analysis could be extended to other topic expressions, but she does not
show this. Secondly, as far as we can see, all of her examples concern the subject
of predication. Reinhart (1981: 54) explicitly identifies aboutness with predication;
even though Reinhart (1981: 58) calls for a pragmatic definition of aboutness
instead of a semantic one, we miss the discussion of non-arguments as topics in
both Reinhart (1981) and (2004).
If the sentence topic were identified with the subject of predication only, there
could be only one. Semantic predication can, however, be only one of the factors
determining topicality. Therefore, it is just one of the dimensions of topic/comment
in work like Jacobs (2001), where it is defined as follows:
(63) Semantic Predication (Jacobs 2001: 647)
In P = (X … Y), X is the semantic subject and Y the semantic predicate iff
a. X specifies a variable in the semantic valency of Y
b. there is no Z such that (i) Z specifies a variable in the semantic valency of
an element in Y and (ii) Z is hierarchically higher in semantic form than X
He gives an example from German where the indirect object is the fronted
topic and subject of predication. This is formalised in the semantic form given
in (b) below.
(64) a. Der Polizei misstraut er. (Jacobs 2001: 648)
the police-(dat) mistrusts he
b. [THE-POLICE(y) & [HE(x) & MISTRUST(x,y)]
valency of the predicate. Breul (2004, 2007) develops an account whereby exactly
one topic can move to a canonical SPEC position, which is triggered by a specific
formal feature [–foc]. If we are interpreting his analysis correctly, this movement
applies either to the subjects of categorical sentences, or to elements that are
fronted due to the assignment of a feature turning them into subjects of predica-
tion in the sense of Jacobs (2001: 657). Thus, the singular topic in accounts like
these seems to be restricted to the dimension of semantic subjecthood, which does
not, in our view, exclude further topical elements from being licensed by discourse
semantic functions.
As indicated in Section 1 (and also in the view defended by Reinhart 1981,
2004), a topic should be defined in terms of pragmatic aboutness. Sentence topics
are used as links (Vallduví 1992: 43) respectively cataloguing addresses (Reinhart
1981: 24), i.e. as instructions to the hearer as to where to store the information. A
definition of addressation can also be found in Jacobs (2001: 650).
(65) Addressation
In (X Y), X is the address for Y iff X marks the point in the speaker-hearer
knowledge where the information carried by Y has to be stored at the moment
of the utterance of (X Y).
.â•… Original interlinear translation: “The broccoli the boss (they) gave it to him (for free).”
Like Vallduví (1992) and Frascarelli (2000) we take clitic doubling as indicating topicality. Left
dislocation can be considered a strong indicator of topicality in languages like German as well
(Jacobs 2001: 658).
Peter Öhl
The elements Reinhart considers topics have to fulfil several formal properties,
including having the highest accessibility among potential antecedents for discourse
anaphora (Reinhart 2004: 299).
(68) Max was walking down from school, pondering about the meaning of life. Soon
he ran into Felix and then he suggested that they stop at the bar.
However, if the subject is intended as the sole element with topic potential, this
example seems questionable to us. If it is true that topics are the most accessible
antecedent, this test should also be valid for the exclusion of multiple topics. This
does not seem to be the case, as shown by the following example:
(69) Soon he ran into Felix. Max did not actually want to meet Felix, but then he
suggested stopping at the bar.
In the second sentence above, there is no absolute preference for how to interpret
the discourse anaphor, which should mean that both referents, Max and Felix,
have the potential of being the addressation topic.
Focusing on the function of topics as an instruction for information storage
brings us instantly to another ‘dimension of topic-comment’ from Jacobs (2001),
which is frame setting.
(70) Frame Setting
In (X Y), X is the frame for Y iff X specifies a domain of (possible) reality to
which the proposition expressed by Y is restricted. (Jacobs 2001: 656)
Frame setting expressions are the part of the sentence that specifies the spatial or
temporal framework for the event reported in the sentence, or a particular state of
affairs in which the sentence asserts something (Reinhart 1980: 173). It has often
been noted that languages with topic morphology like Japanese and Korean use
the same markers with frame setters and sentence topics. 17
(71) a. Kinoo wa Hurankuhuruto wa tenki ga yokat ta. (Japanese)
yesterday top Frankfurt top weather nom good past
‘The weather was fine in Frankfurt, yesterday.’
.╅ For data and discussion we are indebted to Yuko-Shige Tamura and Jiro Inaba (Japanese),
Ki-Hyun Yoon (Korean).
Formal and functional constraints on constituent order and their universality
We do not think that, in cases like these, there are two subjects of predication (as
the assumptions made by Jacobs 2001: 15; 660f. would imply). It is rather that the
frame setter restricts the domain for which the predication is valid. This assumption
is supported by an interesting argument by Strawson (1964; Reinhart 2004: 279;
we also refer to earlier discussions in Strawson 1950: 327ff.). He assumes that sen-
tences like the king of France is bald provoke a conflict in assigning a truth value
because, in an extensional context, the subject of predication lacks an extension. If
an element lacking an extension is part of an extensional predication, however, the
proposition must be false.
(74) The exhibition was visited yesterday by the king of France. (→ f!)
Peter Öhl
We found that this is also true if the subject of predication follows a frame setter
or a contrastive topic:
(75) a. On Friday morning, the king of France was sad. (→ f!)
b. In London, the king of France is adored. (→ f!)
c. For Mary, the king of France would do everything. (→ f!)
There are three topical elements present in the left periphery. That the object
demonstrative is a topic is indicated by clitic doubling; the subject pronoun has a
contrastive value, since the speaker “wants to stress that, as for him, he is not going
to tell anything to his students“ (Frascarelli & Hinterhölzl 2007: 88). They label the
constituent ai ragazzi as familiar or background topic (ibid.), since it is contextually
given and moved to the left. Questo is a topic in the aboutness sense, signalling a
shift in the conversation to the addressee. Interestingly, framing can be analysed
in a parallel way (ibid.: 89):
(77) Gestern hat der/Hans die Maria ge\troffen. (ibid.)
Similar observations have been made by Choi (1997: 550ff.) for Korean (see also
Hetland 2007). She states that Korean nún is not a topic marker but a marker of
contrastiveness. Contrastive elements can occur in several positions:
(78) a. Mary-ka ecey Boston-ey-nun ka-ss-ta.
Mary-nom yesterday Boston-to-contr went
b. Mary-ka Boston-ey-nun ecey ka-ss-ta.
c. Boston-ey-nun Mary-ka ecey ka-ss-ta. (Choi 1997: 550)
.╅ Cf. also McNay (2009: 199), who, looking at a range of different languages, uses a recursive
phase edge feature, [+Link], which carries different semantic import at each phase level, namely
contrastivity at the vP level and aboutness at the TP.
Formal and functional constraints on constituent order and their universality 
continuing topic Mary in (78a+b) can also be the aboutness topic, used as an address
for information storage. Boston in (78a) has contrastive focus simply because it is
marked by nún and has focus in situ, whereas it is a contrastive topic in Frascarelli &
Hinterhölzl’s (2007: 88, 109) sense in (78b). In (78c) it is an aboutness topic, giving
an addressing instruction to the hearer. Similar observations are described by
Vermeulen (2007: 187ff.; cf. also Kuno 1972; Deguchi 2008) for Japanese.
(79) a. sono inu-wa kinoo kooen-de John-o kande-simatta.
That dog-contr yesterday park-at John-acc bite-closed
b. John-o sono inu-wa kinoo kooen-de kande-simatta.
John-acc that dog-contr yesterday park-at bite-closed
‘The dog bit John in the park yesterday.’ (adapted from Vermeulen 2007: 184)
In both Japanese (83) and Korean (84), non-focused contrastive elements can
occur in a lower position and more than one contrastively focused element can
occur per sentence.
(82) A: Of course, mistakes can occur to everyone.
B: Yes, but such a mistake should not happen to such a man.
All these examples suggest that contrastivity is a property both topical and focal
constituents can have as an additional discourse semantic function. We propose
that contrastivity raises a topical element in the hierarchy by making it a delimiter in
Krifka’s (2007: 49f.) sense, which would mean shifting it to the ‘aboutness’-position
in Frascarelli & Hinterhölzl’s (2007: 95ff., 109ff.) model.
Given the notion of a familiar or continuing topic in the accounts just discussed,
we would like to return to the notion of givenness. Whereas Reinhart (1981) strictly
rejects the familiarity account, many authors rely on its constituting effect on topicality
(Vallduví 1992: 20ff.). However, note that the condition of familiarity can be decisively
weakened by replacing contextual givenness with accessibility (cf. Vallduví & Engdahl
1996: 498, who borrow the term from Ariel 1988; cf. also Reinhart 2004: 298ff.). It is
well known that topics can also be inferred from the background.
(85) a. Gustav hat die ganze Nacht nicht geschlafen. Studenten sind
G. has the whole night not slept students are
ja ununterbrochen am arbeiten.
ptc uninterruptedly at work-inf
‘Gustav hasn’t slept all night. Students are uninterruptedly at work,
as you know.’
b. Gustav geht gleich an die Uni. In der Mensa gibt es
G. goes shortly to the university in the refectory gives it
heute glhwein.
today mulled-wine
‘Gustav is going to the university, soon. In the refectory, they serve mulled
wine today.’
The second sentence in (a) uses a generic expression which can be the topic if it is
part of the common ground that the pre-mentioned referent belongs to this class of
individuals. Similarly, the scene setter in (b) is a suitable topic because speaker and
hearer share the knowledge that universities have refectories. But the conditions
on accessibility are even weaker. It may even be sufficient to know of the existence
or the properties of a referent. Thus, frame setters are always highly accessible.
Formal and functional constraints on constituent order and their universality
They are presupposed, like e.g. the existence of a yesterday or the expectation of
a tomorrow.
(86) Yesterday it was raining. Tomorrow it hopefully won’t.
Topics such as those in Reinhart’s Example (62) (When she was five years old, a
child of my acquaintance announced a theory that she was inhabited by rabbits)
are accessible since it is presupposed that people can have acquaintances and that
they can have children. Note that the sentence becomes significantly worse if the
subject is replaced by a bare indefinite that cannot be presupposed, e.g. a person.
However, as also argued by Reinhart (2004: 275ff.), this does not mean that
presupposition implies topichood. Note that the subjects of thetic sentences are
also often presupposed:
(87) The police are coming.
The reason for the often observed interference of topicality with properties such
as givenness, specificity, definiteness etc. must be that they imply high accessibility,
which is a precondition for discourse cohesion (Vallduví 1992: 20), which again is
a reason for choosing constituents as topics. This is why we would like to suggest
regarding these properties as prototypical rather than necessary and sufficient
features of topics.
4.2â•… ‘Perspectivation’
“The speaker may choose very different ‘perspectives’ under which the entire
information to be verbalised is put into sequential order” (Stutterheim & Klein
2002: 66). Since several of the features influencing constituent order seem to exist
independently of topicality, the discussion may be reduced to two major questions.
Firstly, what are the primitives of what we call perspectivation? Secondly, how do
they, in fact, interact with topicality? In the context of the term perspectivation,
the notions of salience and of ‘point of view’ (called empathy by Chafe 1976)
frequently occur in the literature besides ‘topicality’. The options a speaker has for
marking what he finds relevant (or assumes the hearer to), or for illustrating his
point of view, are quite different across the languages of the world due to different
structural means and different formal restrictions.
Since English has a syntactic system with very few options for permutation,
the role of passivisation as a means of changing the word order is important
for marking the point of view. At the same time, the speaker can also choose this
construction type as an instruction to the hearer regarding what to take as an
address for storing the information. e.g.:
(88) a. Caesar conquered Gaul in 52 BC.
b. Gaul was conquered (by Caesar) in 52 BC.
Peter Öhl
Even though it changes the point of view along with the subject, passivisation
cannot be a means specific to topicalisation. Passivisation does not always result
in the creation of an address for information storing. This is shown by thetic
passive sentences:
(89) Numerous peoples were defeated by the Romans.
Elements marking the point of view relative to other elements tend to precede
them in the sentence. Being the ‘point of view’ may be a feature that is typical
of topics, but it is neither a necessary nor a sufficient condition. There are
several factors causing elements to precede others. It seems obvious that the more
discourse-prominent features, such as point of view or familiarity, an element has,
the less acceptable the precedence of other elements in the sentence is. The crucial
question to ask now is: When do these features determine the choice of a topic?
Firstly, we would like to discuss some more properties of elements which can
trigger their fronting. Specificity is another feature that is characteristic of topics
but is not a sufficient condition for topicality. Assume the following context:
(90) A: What did he say?
B: That there is probably exactly one country in the world where everything is
better than here.
The following sentences with quantified expressions show that specific indefinites
tend to precede the sentence adverbial in German, whereas non-specific ones
follow it. In (91a), the phrase in genau einem Land (‘in exactly one country’)
follows the sentence adverbial wahrscheinlich (‘probably’). In this case, it must have
a non-specific reading. If it precedes the adverbial, it has a specific but implicit
reference (e.g. Switzerland in 91b).
(91) a. dass auf der ganzen Welt wahrscheinlich in genau einem Land
that in the whole world probably in exactly one land
alles besser ist, als hier.
everything better is than here
b. dass auf der ganzen Welt in genau einem Land wahrscheinlich alles besser
ist, als hier.
It has often been observed that non-specific indefinites must not be moved higher
than specific expressions. The reason is that they always take narrow scope, whereas
specific indefinites may take wide scope (Pafel 1997: 31ff.).
(92) A: Wem hast du ein Buch geschickt? (Lenerz 2000: 266)
‘Who did you send a book?’
B: Ich habe (*ein Buch) dem Verlag *(ein Buch) geschickt.
I have a-acc book the-dat publishers sent
Formal and functional constraints on constituent order and their universality
The example in (B) with *… ein Buch dem Verlag geschickt improves immediately if
the indefinite is stressed, or if it occurs with a restrictive attribute (such as ‘delayed
far too long’), giving a quantified object a topical (Endriss & Hinterwimmer
2007: 85; 88) or a specific reading:
This means defocused phrases must be in a position outside of the VP. One often
observed result is the definiteness effect raising discourse semantically marked
definites. Note, however, that definiteness itself does not imply defocusing of a dis-
course referent (Lambrecht 1994: 108; Molnárfi 2007: 176ff.). Salient or singular
referents from the common ground can be inside the focus even if definite
(cf. Molnárfi 2007: 178 about referents like the president or the cat). We consider
definiteness to be primarily a quantificational feature restricting a reference set in
relation to the discourse domain. That is why definites have neither to leave the
‘focus domain’ VP nor be refocused in situ, as proposed by Abraham (2007: 199f.).
Peter Öhl
Neither definiteness nor specificity crucially forces elements out of the VP (96a);
there must be additional properties, such as being the point of view (96b) or
contrastivity (96c).
(95) What did he say?
(96) a. …dass [VP schon zweimal [VP die Olympiade an Wuppertal vergeben
worden] ist
b. …dass die Olympiade [VP schon zweimal [VP an Wuppertal vergeben
worden] ist
c. …dass die Winterolympiade [VP schon zweimal [VP an Wuppertal vergeben
worden] ist
What occurs in the fronted VP is also in the focus. Therefore, VP-fronting may be
taken as a test for predicate focus or sentence focus.
(97) a. [VP schon zweimal [VP eine Olympiade an Wuppertal vergeben worden]]
ist seiner Ansicht nach
b. [VP eine Olympiade an Wuppertal vergeben worden]] ist seiner Ansicht
nach schon zwei Mal
c. [VP schon zwei Mal [VP an Wuppertal vergeben worden] ist die (Winter)
olympiade seiner Ansicht nach
d. *[VP die Winterolympiade an Wuppertal vergeben worden] ist seiner
Ansicht nach schon zwei Mal
These examples show that the German middle field can roughly be divided into two
partitions, the lower one containing the sentence or predicate focus, the higher
one containing discourse semantically marked elements, among them also addres-
sation topics. There are options of ‘perspectivation’ in both partitions. In the ‘focus
domain’ VP, however, scrambling is restricted:
(100) What did he say?
(101) a. …dass [VP bald [VP eine Olympiade in Wuppertal stattfinden]] soll
b. ?…dass [VP bald [VP in Wuppertal eine Olympiade stattfinden]] soll
Thus, languages like German are discourse configurational because they lack
specific positions for different kinds of constituents. The order is constrained by
the features themselves, and this obviously often allows for different options. Similar
analyses should be possible for languages like Japanese or Korean. As mentioned
above, earlier research on information structure relied on the assumption that, in
these languages, topics are marked both by specific particles and by fronting to
a specific clause initial position. We have also shown evidence from more recent
research, however, that this view needs more differentiation by considering con-
trastiveness, suggesting that the occurrence of these particles and perspectivation
(especially topicalisation) are independent. This assumption can be supported by
further phenomena of perspectivation as shown by contexts like the one below, pre-
supposing nothing but the frame setter today.
(102) A: Do you know what will happen today?
B: Fortunately, some student will probably record the whole lecture today.
.╅ For these judgements, we thank again Jiro Inaba (Japanese) and Ki-Hyun Yoon (Korean).
Peter Öhl
These examples, which parallel the orders in the German examples in (50) above,
show that there is no clear ‘base position’ for sentence adverbials in Japanese and
Korean either. Besides topicalisation, there are more discourse semantic movements
changing the order between more or less prominent elements in the sentence – based
on the fact that SOV languages like Japanese, Korean and German have scrambling,
offering options of perspectivation beyond those of the canonical positions provided
by functional projections in languages such as English. We would like to conclude
this subsection by referring to an earlier account by Fukui (1995) who proposed
that, besides head finalness, the syntax of Japanese differs from the syntax of English
mainly by the lack of any functional projection dominating the VP.
5.â•… C
onclusion: Towards a model of interacting constraints
on linearisation
Secondly, if a language can be identified as having one or the other kind of canonical
position relating to functional projections, their hierarchy is fixed not only by the
order chosen during language acquisition, but also by universal conceptual factors
(cf. Parodi 1998; Rizzi 2000). This does not only apply to functional features: The
order of constituents within the domain of V (including the vP of more recent
Generative accounts) is very clearly constrained by a lexical conceptual hierarchy
(e.g. the Θ-hierarchy of the LCS).
(108) Hierarchy of canonical positions
– C-domain > I-domain
– I-domain > V-domain
– C-domain: TopP > FocP (languages like Italian; cf. Rizzi 1997)
– I-domain: Agr > T
– V-domain: Θ-hierarchy
It seems obvious from the discussions in this paper that languages may lack
canonical positions for topics as they may for subjects. The syntax of these lan-
guages is more liberal with respect to what we called perspectivation. We assume
that there is no universal hierarchy of FPs, but that features, especially those
interfacing with pragmatics (such as the features of illocutionary force, clause
mood and perspectivation), are acquired by the first language learner as he
extends his conceptual knowledge (in fact, this is a proposal we made earlier in a
model of grammar change; cf. Öhl 2009). It depends on the parameterisation of
the inventory of functional features whether the child acquires phrases that then
provide more or less fixed positions in the language specific syntax.
We think that the options for parametrising FPs are subject to more global
hierarchies of features that are made evident by several general tendencies which
Peter Öhl
.â•… We concede that this list is incomplete and neglects some phenomena like floating
quantifiers and others. We also concede that we did not consider the whole range of
literature that might be relevant to that topic. We did not discuss backgrounding and
foregrounding in detail. Much more could be said about it and the topic would deserve a
much more elaborate model to account for it. Nevertheless we hope we could outline the
main idea that we developed on the basis of the contrastive phenomena we compared and
analysed here. This might be the right place and the right time to use the familiar formula
promising future research …
Formal and functional constraints on constituent order and their universality 
However, in our view, it is not at all clear whether all of these properties are primi-
tives. Some may instead be consequences of the interaction of more primitive fea-
tures. An adverbial scoping over the whole proposition, and thus marking the point
of view, serves, at the same time, as delimitation for the information processing
(Krifka 2007: 49f.; see above, p. 259). Thus it can serve as a frame setter. A specific
and defocused subject of predication marking the point of view can serve as an
address in Reinhart’s (1981) sense (see above, p. 257). These potential functions
may also be supported by the degree of cognitive accessibility of an element.
From this point of view, the fact that topics and frame setters tend to precede
all other constituents may just follow from the fact that they have several of the
properties in (109). Or, put differently, the more of these properties an element
has, the more probable it is that it is chosen as a frame setter or as a topic. A
closer look at the relevant features and the ways in which they interact should be
possible through intensive empirical research, providing more contrastive data
from syntactic systems with a greater range of variation.
References
Abraham, W. 1997. The base structure of the German clause under discourse functional weight:
contentful functional categories vs. derivative functional categories. In German: Syntactic
Problems – Problematic Syntax, W. Abraham & E. van Gelderen (eds), 11–42. Tübingen:
Niemeyer.
Abraham, W. 2007. Topic, focus, and default vs. contrastive accent: Typological differences
with respect to discourse prominence. In On Information Structure, Meaning and Form:
Generalizations across Languages [Linguistik Aktuell/Linguistics Today 100], K. Schwabe &
S. Winkler (eds), 183–203. Amsterdam: John Benjamins.
Alexiadou, A. & Agnastopoulou, E. 2001. The subject-in-situ generalization and the role of case
in driving computations. Linguistic Inquiry 32: 193–232.
Ariel, M. 1988. Referring and accessibility. Journal of Linguistics 24: 67–87.
Benincà, P. & Poletto, C. (2004). Topic, focus and V2: Defining the CP-sublayers. In The Structure of
CP and IP. The Cartography of Syntactic Structures, Vol. 2, L. Rizzi (ed.), 52–75. Oxford: OUP.
Peter Öhl
Bresnan, J. 1994. Locative inversion and the architecture of Universal Grammar. Language
70: 71–131.
Breul, C. 2004. Focus Structure in Generative Grammar: An Integrated Syntactic, Semantic
and Intonational Approach [Linguistik Aktuell/Linguistics Today 68]. Amsterdam: John
Benjamins.
Breul, C. 2007. Focus structure, movement to spec-Foc and syntactic processing. In On
Information Structure, Meaning and Form: Generalizations across Languages [Linguistik
Aktuell/Linguistics Today 100] K. Schwabe & S. Winkler (eds), 255–74. Amsterdam:
John Benjamins.
Chafe, W.L. 1976. Givenness, contrastiveness, definiteness, subjects, topics and point of view. In
Subject and Topic, C.N. Li (ed.), 25–56. New York NY: Academic Press.
Choi, H.-W. 1997. Topic and focus in Korean: The information partition by phrase structure and
morphology. In Japanese/Korean Linguistics, Vol. VI, H.-M. Sohn & J. Haig (eds), 545–561.
Stanford CA: CSLI.
Chomsky, N. 1986. Barriers. Cambridge MA: The MIT Press.
Chomsky, N. 1995. The Minimalist Program. Cambridge MA: The MIT Press.
Chomsky, N. 2000. Minimalist inquiries: The framework. In Step by Step: Essays on Minimalist
Syntax in Honor of Howard Lasnik, R. Martin, D. Michaels & J. Uriagereka (eds), 89–155.
Cambridge MA: The MIT Press.
Chomsky, N. 2008. On phases. In Foundational Issues in Linguistic Theory, R. Freidin, C.P. Otero
& M.L. Zubizarreta (eds), 133–166. Cambridge MA: The MIT Press.
Cinque, G. 1993. A null theory of phrase and compound stress. Linguistic Inquiry 24: 239–297.
Cinque, G. 1998. Adverbs and Functional Heads: A Cross-linguistic Perspective. Oxford: OUP.
Culicover, P. & Levine, R.D. 2001. Stylistic inversion in English: A reconsideration. Natural
Language & Linguistic Theory 19: 283–310.
Deguchi, M. 2008. Topicalization and contrastive readings: Insights from Japanese wa. Talk
given at the workshop on ‘Topicality’, DGfS annual meeting, Bamberg.
Drubig, H.B. 2007. Phases and the typology of focus constructions. In On Information Structure,
Meaning and Form: Generalizations across Languages [Linguistik Aktuell/Linguistics Today
100], K. Schwabe & S. Winkler (eds), 33–66. Amsterdam: John Benjamins.
Endriss, C. & Hinterwimmer, S. 2007. Direct and indirect aboutness topics. In Interdisciplinary
Studies of Information Structure [Working Papers of the SFB 632 No. 6], C. Féry et al. (eds),
83–96. Potsdam: Universitätsverlag.
Frascarelli, M. & Hinterhölzl, R. 2007. Types of topics in German and Italian. In On Infor-
mation Structure, Meaning and Form: Generalizations across Languages [Linguistik
Aktuell/Linguistics Today 100], K. Schwabe & S. Winkler (eds), 87–116. Amsterdam:
John Benjamins.
Frascarelli, M. 2000. The Syntax Phonology Interface in Focus and Topic Constructions in Italian
[Studies in Natural Language and Linguistic Theory 50]. Dordrecht: Kluwer.
Frey, W. 2000. Über die syntaktische Position der Satztopiks im Deutschen. In Issues on Topics
[ZAS Papers in Linguistics 20], K. Schwabe, A. Meinunger & D. Gasde (eds), 137–172.
Berlin: ZAS.
Frey, W. 2004a. A medial topic position for German. Linguistische Berichte 198: 153–190.
Frey, W. 2004b. The grammar-pragmatics interface and the German prefield. Sprache &
Pragmatik 52: 1–39.
Frey, W. 2007. Some contextual effects of aboutness topics in German. In Interfaces and Interface
Conditions, A. Späth (ed.), 329–348. Berlin: Walter de Gruyter.
Formal and functional constraints on constituent order and their universality
Lambrecht, K.1994. Information Structure and Sentence Form: Topic, Focus, and the Mental
Representations of Discourse Referents. Cambridge: CUP.
Lasnik, H. 2001. Derivation and representation in modern transformational syntax. In The
Handbook of Contemporary Syntactic Theory, M. Baltin & C. Collins (eds), 62–88. Oxford:
Blackwell.
Lenerz, J. 2000. Word order variation: Competition or co-operation. In Competition in Syntax,
G. Müller & W. Sternefeld (eds), 249–281. Berlin: de Gruyter.
Li, C.N. & Thompson, S.A. 1976. Subject and topic: A new typology of language. In Subject and
Topic, C.N. Li (ed.), 457–89. New York NY: Academic Press.
Lötscher, A. 1992. The relativity of subject/topic prominence in German, English and Russian.
Folia Linguistica 26: 95–105.
McNay, A. 2009. Information structural recursion at the phase level. In Selected Papers from the
2006. Cyprus Syntaxfest, K.K. Grohmann & E.P. Panagiotidis (eds), 195–236. Cambridge:
Scholars Publishing.
Molnárfi, L. 2007. On the discourse configurationality of West Germanic. In On Information
Structure, Meaning and Form: Generalizations across Languages [Linguistik Aktuell/Linguistics
Today 100], K. Schwabe & S. Winkler (eds), 155–181. Amsterdam: John Benjamins.
Öhl, P. 2003. Economical Computation of Structural Descriptions in Natural Language. Ph.D.
dissertation, University of Stuttgart.
Öhl, P. 2004. Satztypen und Hypotaxe im typologischen Vergleich. In Beiträge zu Sprache &
Sprachen 4: Vorträge der Bochumer Linguistik-Tage, K. Pittner, R.J. Pittner & J.C. Schütte
(eds), 159–170. Munich: Lincom.
Öhl, P. 2009. Sprachwandel und kognitive Ökonomie: Zur Grammatikalisierung und Substitution
von Satzkonnektoren. Linguistische Berichte 220, 393–438.
Pafel, J. 1997. Skopus und logische Struktur. Studien zum Quantorenskopus im Deutschen.
Habilitation thesis, Universität Tübingen.
Parodi, T. 1998. Der Erwerb funktionaler Kategorien im Deutschen: Eine Untersuchung.
Tübingen: Narr.
Reinhart, T. 1980. Conditions for text coherence. Poetics Today 1(4): 161–180.
Reinhart, T. 1981. Pragmatics and linguistics: An analysis of sentence topics. Philosophica 27: 53–94.
Reinhart, T. 2004. Topics and the conceptual interface. In Context Dependence in the Analysis of
Linguistic Meaning, H. Kamp & B. Partee (eds), 275–305. Amsterdam: Elsevier.
Rizzi, L. 1997. The fine structure of the left periphery. In Elements of Grammar: Handbook in
Generative Syntax, L. Haegeman (ed.), 281–337. Dordrecht: Kluwer.
Rizzi, L. 2000. Comparative Syntax and Language Acquisition. London: Routledge.
Rizzi, L. 2001. On the position “Int(errogative)” in the left periphery of the clause. In Current
Studies in Italian Syntax: Essays Offered to Lorenzo Renzi, G. Cinque & C. Salvi (eds),
287–96. Amsterdam: North Holland.
Rizzi, L. (ed.). 2004. The Structure of CP and IP. The Cartography of Syntactic Structures, Vol. 2.
Oxford: OUP.
Roberts, I. & Roussou, A. 2002. The extended projection principle as a condition on the
tense dependency. In Subjects, Expletives, and the EPP, P. Svenonius (ed.), 123–154.
Oxford: OUP.
Roberts, I. & Roussou, A. 2003. Syntactic Change: A Minimalist Approach to Grammaticalization.
Cambridge: CUP.
Rothstein, S.D. 1985. The Syntactic Forms of Predication. Bloomington IN: Indiana Linguistic Club.
Formal and functional constraints on constituent order and their universality
Roussou, A. 2000. On the left periphery: Modal particles and complementisers. Journal of Greek
Linguistics 1: 65–94.
Sasse, H.-J. 1982. Subjektprominenz. In Fakten und Theorien. Festschrift für Helmut Stimm zum
65. Geburtstag, S. Heinz & U. Wandruszka (eds), 267–286. Tübingen: Narr.
Sasse, H.-J. 1987. The thetic/categorical distinction revisited. Linguistics 25: 511–580.
Sasse, H.-J. 1995. Prominence typology. In Syntax: An International Handbook of Contemporary
Research, Vol. 2, J. Jacobs, A. von Stechow, W. Sternefeld & T. Vennemann (eds), 1065–
1075. Berlin: de Gruyter.
Schachter, P. 1976. The subject in Philippine languages: Topic, actor, actor topic, or none of the
above. In Subject and Topic, C.N. Li (ed.), 491–518. New York NY: Academic Press.
Sternefeld, W. 2006. Syntax: Eine morphologisch motivierte generative Beschreibung des Deutschen.
Tübingen: Stauffenburg.
Strawson, P.F. 1950. On referring. Mind 59: 320–344.
Strawson, P.F. 1964. Identifying reference and truth-value. Theoria 30: 96–118.
Stutterheim, C. v. & Klein, W. 2002. Quaestio and L-perspectivation. In Perspective and
Perspectivation in Discourse [Human Cognitive Processing 9], C.F. Graumann & W. Kallmeyer
(eds), 59–88. Amsterdam: John Benjamins.
Vallduví, E. 1992. The Informational Component. New York NY: Garland.
Vallduví, E. & Engdahl, E. 1996. The linguistic realization of information packaging. Linguistics
34: 459–519.
Vermeulen, R. 2007. Japanese wa-phrases that aren’t topics. In UCL Working Papers in Linguistics
19, R. Breheny & N. Velegrakis (eds), 183–201. London: University College.
On the foundations of the contrastive study
of information structure
Carsten Breul
Bergische Universität Wuppertal
The paper presents some ideas on the delineation of a more specific contrastive
approach to information structure analysis within the larger and more loosely
circumscribed comparative area. It will be argued that this delineation can
be effected by emphasising the methodological role of the notion tertium
comparationis. Ontological and methodological aspects of contrastive information
structure analysis will be discussed, and benefits of taking a specifically
contrastive approach to information structure will be pointed out. Finally, some
lines of argumentation and observations from the preceding chapters of the
present volume that can be construed as instances of contrastive information
structure analysis will be briefly recapitulated.
1.â•… Introduction
The articles in this volume are concerned with comparative approaches to issues
of information structure. This chapter puts forward some ideas on the delineation
of a more specific contrastive approach within the larger and more loosely circum-
scribed comparative area.1 I will be arguing that this delineation can be effected by
emphasising the methodological role of the notion tertium comparationis, which
has been prominent and the focus of much discussion in the history of contras-
tive linguistics and contrastive analysis (see e.g. Krzeszowski 1990: Chapter 1, 2,
Chesterman 1998: Chapter 1 and the references given there). Note that I will
be using the terms contrastive linguistics and contrastive analysis not as referring
to a certain theoretical framework alongside those such as structuralist linguis-
tics, (systemic-)functional linguistics, generative linguistics, and, respectively, to
a style of analysis typical of these frameworks. Rather, I will be using them to refer
to kinds of linguistic research where the concern with the question of the tertium
.â•… I am grateful to Edward Göbbel, Alex Thiel and an anonymous reviewer for corrections
and valuable suggestions. All remaining errors are mine.
Carsten Breul
.â•… Restricting myself to rather explicitly contrastive research on the language pair English/
German where the interplay of syntactic and information structural aspects plays an important
role (i.e. the field of my own main research interests), I may mention, among others, Breul 2007,
2008a, 2008b; Doherty 1996, 1999, 2002; Erdmann 1990, 1993; Esser 1995; Fabricius-Hansen
1999; Firbas 1959, 1964; Kirkwood 1969, 1970; 1978; Klein 1988; Legenhausen & Rohdenburg
1995; Weinert 1995; Zimmermann 1972. In terms of theoretical backgrounds, approaches and
objectives, these works form a highly heterogeneous, but thereby also a rather representative
set for the research in the said domain for the past 50 years.
On the foundations of the contrastive study of information structure
.â•… The various papers in the present volume provide examples and many references to
comparative research on information structure within functional, generative, and typological
frameworks. For a quite recent overview of research on information structure, making ref-
erence to various languages, see Erteschik-Shir (2007). The publications of the Collabora-
tive Research Centre (Sonderforschungsbereich, SFB) “Information Structure: The Linguistic
Means for Structuring Utterances, Sentences and Texts“ (see http:↜//www↜.sfb632.uni-potsdam.
de/main.html) and the references contained in these publications provide a wealth of relevant
bibliographical material. So do the articles collected in Schwabe & Winkler (eds) (2007).
.â•… According to Kortmann (1998: 138f.), the decline of contrastive linguistics was much more
pronounced in the USA than in Europe. Disappointment generated by exaggerated expectations
concerning the benefits of contrastive linguistics (CL) for the purposes of language teaching
and learning affected the more pedagogically oriented CL researchers in the USA more strongly
than the more theoretically and descriptively oriented CL researchers in Europe.
.â•… As to the relevance of a comparative approach that focuses on differences for the study of
universals as conceived of in generative grammar, I may quote Kayne (1996/2000: 3):
The study of differences among languages must obviously proceed in tandem with
the study of what they have in common, that is, with the study of the principles of
Universal Grammar (UG) that interact with language specific parameters to yield
observed variation. Similarly there is every reason to believe that the search for
universal syntactic principles cannot proceed without close attention being paid to
syntactic variation.
Carsten Breul
.â•… See the overview in Pan & Tham (2007: Chapter 4) and the works referred to there.
Pan & Tham’s (2007: 208) own proposal for a “definition” of contrastive linguistics is
as follows:
The blind spot in this type of approach is the question which lexico-grammatical
forms from the different languages in focus one is to select for inclusion in the
study of a given principle or parameter in the first place. (My two examples
involving (1)–(4) below will illustrate this problem.) This question is usually not
raised, and it is usually answered implicitly by taking structures filled with lexi-
cal items into account that are considered to be more or less cross-linguistically
the same in terms of meaning. But this methodological step is not motivated by
the theoretical framework. It may be thought that it is motivated by the addi-
tional assumption that the universal syntactic features and principles are associ-
ated with semantic content from the different languages in such a way that, say,
the same features and the same syntactic operations conditioned by the same
syntactic principles will result in the same meaning. But this, in fact, is rather an
implication of the methodological decision to compare just those structures that
seem to have (roughly) the same meaning; it is not an assumption which could,
in principle, be falsified and which could be taken to motivate this methodologi-
cal decision as long as it is not falsified. Now, what would it mean if it should
turn out that those structures compared are not actually the same in meaning?
Or what does it mean to say that those structures are only roughly the same in
meaning? I will mention two cases that raise these questions in a more concrete
and illustrative way.7
What, for instance, does the availability of know and the unavailability of kennen
in English and German passives mean for generative accounts of passives?
.â•… The examples involve (lexico-)syntactic topics – rather than information structural
ones – in order for me to be able to connect them to the mainstream generative view of the
relation between language specificity, universality and parametricity as laid down in the quo-
tations from Kayne (1996/2000) and Haegeman & Guéron (1999) given above. There is no
mainstream generative view of the elements and principles of information structure, nor of its
place and role in – or in relation to – the architecture of grammar.
On the foundations of the contrastive study of information structure
Does it mean that the syntactic principles involved in passive formation are
essentially different in English and German? Does it mean that the syntactic prin-
ciples are the same, but in different ways sensitive to syntactically relevant lexical
characteristics of know and kennen, due to different parameter settings? Or does it
mean that the syntactic principles as well as potentially involved parameter settings
are the same, while syntactically relevant lexical features of know and kennen
are different in just those respects that are relevant for passive formation? For a
convincing analysis set in a universalist (notably generative) framework to be
possible, questions such as these have to be raised and plausibly answered. In
addition to triggering such questions, a contrastive syntactic approach provides
the data on the basis of which answers to such questions are to be assessed.
To give another example: Imagine a generative syntactician investigating the
ordering options among phrases that premodify the nominal head in German noun
phrases. Confronted with data such as those in (2), she may start by hypothesising
that the order of adjectival phrase and numerical phrase in pre-head position is
reversible in German.
Bringing English into play, the syntactician may then observe that this language
only allows the order numerical phrase > adjectival phrase.
At this point the syntactician may conclude her analysis by proposing a parametrical
difference between German and English in this respect. However, she may also
continue the investigation by taking a more contrastively oriented perspective,
observing that nominal phrases with the German pre-head order adjectival phrase >
numerical phrase actually seem to have English nominal phrases as translational
and semanto-syntactic equivalents (see Krzeszowski 1990) that are slightly, but
significantly, different in structure, as suggested by the examples in (4).
Carsten Breul
(4) a. And while pollution incidents for 1991–92 reached an all-time high of 29,
524 […], there were a meagre 536 prosecutions. (British National Corpus
(BNC), document CH6)
a′. … magere 536 Anklagen (not: 536 magere Anklagen ‘536 meagre
prosecutions’)
b. Terry, a biscuit kiln fireman, shaved a remarkable 12 minutes off his best
ever time to finish in three hours five minutes. (BNC HBE)
b′. … bemerkenswerte 12 Minuten (not: 12 bemerkenswerte Minuten ‘12
remarkable minutes’)
c. Situated in a former ex-Great Western Railway coach, No. 1160
liveried in chocolate and cream at platform one, the exhibition boasts a
staggering 200 visitors per day during the operating season of the SVR.
(BNC CKK)
c′. … erstaunliche 200 Besucher (not: 200 erstaunliche Besucher ‘200
staggering visitors’)
These data suggest that the pre-head adjectival phrase > numerical phrase order
in German nominal phrases corresponds rather to the somewhat peculiar English
nominal phrases displayed in (4), also in cases such as those in (2b), correspond-
ing to We spent a pleasant three weeks there. (These nominal phrases are peculiar,
of course, in that they apparently feature an indefinite article in construction with
a plural noun, which ‘usually’ leads to ungrammaticality; cf. *a 536 prosecutions;
*a 12 minutes etc.)8 This contrastive observation, then, may lead the hypothetical
generative syntactician to a significantly different analysis than the one mentioned
above. In this alternative analysis based on a contrastive observation the German pre-
head adjectival phrase > numerical phrase order is structurally brought in line with
the syntactic structure of the respective nominal phrases in (4) – whatever their struc-
ture may be – rather than that in (3a). An analysis along these lines would certainly
be more explanatorily adequate than simply postulating a parametrical difference as
in the analytical approach concentrating only on data such as (2) and (3). The second
line of analysis is contrastive in that it takes relevant translation equivalents as tertia
comparationis into account (schöne drei Wochen/a pleasant three weeks) that may have
significant implications for the syntactic analyis.9
.â•… I have found no reference to (analyses of) the peculiar type of construction exemplified
by a pleasant three weeks in the recent comprehensive survey volume Noun Phrase in
the Generative Perspective (Alexiadou & Haegeman & Stavrou 2007). In the Cambridge
Grammar of the English Language (Huddleston & Pullum 2002), this construction type is
described as involving a recategorisation or respecification of a plural measure phrase as
singular (see ib.: 346, 353f.).
.â•… By “relevant translation equivalents” I mean translation equivalents that are semanto-syntactic
equivalents in Krzeszowski’s (1990) sense at the same time.
On the foundations of the contrastive study of information structure
Of course there are tertia comparationis in typological studies too. These are
certain grammatical categories or concepts, such as ‘word’, ‘word order’, ‘voice’,
‘tense’, which are mostly taken from or rooted in traditional grammar of the western,
Greek and Latin based, school. Typological studies aim at classifying languages
according to the manifestation of such categories, potentially with the further aim
of uncovering universals. (The notion ‘classifying’ in this context comprises the
assignment of languages to certain positions on continuous classificatory scales,
or to certain positions in spheres around a prototype, in those cases where the
researcher rejects the existence of discrete classes.) Here as well, researchers often
take a more or less implicit recourse to equivalence in meaning or function when
they explore which classes to set up in the first place. This may put the typological
endeavour to a certain risk. On the one hand, a thorough analysis of the ways in
which language B expresses what language A expresses by operating with or on
a form that has been analysed as the manifestation of a certain grammatical cat-
egory or concept makes this analysis look very much like a contrastive analysis. This
may mean nothing more, but also nothing less, than that typology is dependent on
contrastive analysis. On the other hand, the preliminary contrastive analysis may
reveal that the putative grammatical category or concept that the typologist sets out
to explore is so heterogeneous in nature cross-linguistically that it is questionable
whether it can survive as a typological tertium comparationis.10 This is to say that
contrastive analysis has a complementing function vis-à-vis typology as well, a point
that has been emphasised in a series of articles by E. König (1990, 1992, 1993, 1996).
Note also that contrastive analysis cannot be subsumed under typology, at least
.╅ For instance, it may be debated whether the impressive amount of knowledge that has been
accumulated by typologists concerned with ‘grammatical voice’ have left anything coherent of
this notion that may qualify it for the status of grammatical category and thus as typological
tertium comparationis. See e.g. Klaiman (1991), who insightfully surveys and classifies different
voice system, concluding:
Plausibly, then, what is common to different types of voice systems may be that
wherever voice alternations occur, they encode alternative assignments of arguments
to positions which have superior ranking at some grammatically significant level of
organization, be that of relational structure, information structure or some other level.
(One alternative level which will be taken up momentarily is ontological structure.)
(Ib.: 262f.)
The author hopes that, for all its tentativeness, the present study might provide a
fruitful basis for an enhanced understanding on an intriguing grammatical category
whose nature has long seemed obscure. (Ib.: 271)
It is not obvious that the generalisation over quite heterogeneous “grammatically significant
level[s] of organization” suggested in the first quotation may serve as a definition or
characterisation of a grammatical category.
Carsten Breul
2.2â•… Th
e distinctive features applied to contrastive information
structure analysis
What do the distinctive features (a) and (b) of contrastive linguistics mentioned at the
beginning of Section 2.1 mean in more concrete terms for contrastive information
structure analysis?
The meaning/function side is manifested by the categories of information structure
as identified by a given theory of information structure. By this I mean categories
such as theme and rheme, given and new information, topic and comment, focus
and background – as long as they have been characterised or defined indepen-
dently of formal criteria (cf. further below) – or the three categories identifiability,
activation and focus structure as thoroughly discussed especially by Lambrecht
(1994). As my own conception of information structure is coined by Lambrecht’s,
I may quote his definition:11
Information structure: That component of sentence grammar in which
propositions as conceptual representations of states of affairs are paired
with lexicogrammatical structures in accordance with the mental states of
interlocutors who use and interpret these structures as units of information in
given discourse contexts. (Lambrecht 1994: 5)
This definition allows one to see why, for Lambrecht (and the present author as
well), identifiability and activation also figure under the comprehensive label
‘information structure’, alongside the more prominently discussed phenomena
that he subsumes under the label ‘focus structure’. To give a simple example: The
lexico-grammatical alternatives displayed by (5a, b) below may, in an appropriate
context, be due to nothing but a difference as to whether or not the producer of
(5a) assumes that the policeman is identifiable by her addressee, and as to whether
or not the producer of (5b) assumes that the referent of John/he is active in the
mind of the addressee at the given point in the discourse.
.╅ My reference to Lambrecht (1994) here to the exclusion of other theories of information
structure reflects a personal view to the effect that this work is (still) outstanding by its pre-
cision, clarity, homogeneity and comprehensiveness in explaining what information structure
is about in general. There are other works that are highly insightful and analytically deep on
specific topics of information structure, many of which mentioned in the present volume. But
they do not fit my “metalevel” (as an anonymous reviewer puts it) purposes in this paper.
On the foundations of the contrastive study of information structure
(5) a. There is {a/the} policeman standing at the corner. (acceptable with the if
there is only one policeman available in the discourse world)
b. {John/He} is ill.
However, as is well known, theories of information structure have not yet developed
to a stage where there is broad terminological and conceptual consensus and
where the open theoretical issues or questions could be formulated in a manner
that would make sense to every researcher. This is only to be expected in a fairly
young domain of investigation, and contrastive information structure analysis
may be one of the factors that help information structure theory to reach the next
stage in its development.
The form side is manifested by the forms, categories, operations, principles,
constructions identified on the levels of phonology, including intonational pho-
nology, morphology, syntax and the lexicon that have been associated with catego-
ries of information structure by researchers with diverse theoretical backgrounds.
Here as well there is no consensus between representatives of different theoretical
schools about even the most elementary aspects of the nature of such areas as
phonology, mor�phology, syntax, or the lexicon. However, on a more shallow, or
‘surface’, level, certain phonological, morphological, syntactic and lexical phenom-
ena seem to be unanimously associated with information structural categories in
at least some languages. I am thinking of certain intonational phenomena (such
as falling pitch accents or tunes), bound morphemes (such as Japanese -wa and
Korean -nun), constituent order phenomena (such as phrase movement to clause
initial position), constructions (such as cleft-constructions) and lexical items (such
as pronouns, definite articles, focus and topic particles) about whose information
structural relevance in some way in certain languages there is no disagreement –
even if disagreement may begin as soon as one delves a little deeper into ques-
tions of the description and explanation of these phenomena as such and of their
relation to information structural categories.
A variety of theoretical backgrounds is no problem in principle for contrastive
information structure analysis. What seems necessary, though, is that the theoretical
background chosen allow the researcher, first, to identify and demarcate their
tertium comparationis and, second, to motivate the assumption that it applies to
both languages under investigation. As to the second point, we may recall the
dilemma of those in the early days of contrastive linguistics who adhered to the
structuralist (of a certain brand) conviction that each language is unique and has
to be described in its own terms and who were interested in doing contrastive
analysis at the same time (see James 1980: 166f.). That is, in order to avoid such
a dilemma, it is necessary that one’s theoretical background not exclude the possi-
bility that the tertium comparationis does in fact apply to both languages involved.
It would make the task of the contrastive analyst even easier if their theoretical
Carsten Breul
background not only allows for, but suggests that the respective tertium compa-
rationis is cross-linguistically, perhaps even universally, applicable. It has been
pointed out more than once that cross-linguistic or universal applicability of a
tertium comparationis is by no means guaranteed by the mere fact that the same
grammatical term has traditionally been used for the description of a linguistic
phenomenon in language A and of a linguistic phenomenon in language B.12
.â•… Among others, this point was made by Lattey (1982: 133) and König (1993: 290f.).
According to the latter:
This conflation – form: first vs. second position; meaning/function: expressing “that
with which the clause is concerned” vs. expressing that by which “the message is
developed” – renders these notions unsuitable for serving as tertia comparationis in
contrastive information structure analysis. They already presuppose what may actu-
ally be a result of a contrastive information structure analysis, namely that there may
be a systematic relation between the linear position of a formal unit in a linguistic
expression on the one hand and its information structural value (meaning/func-
tion) on the other hand. Actually, the fact that this presupposition does not hold
in a language such as English, as in cases like (6B) (from Lambrecht 1994: 223),
where small capitals signal the word that carries the primary sentence accent, is
then accounted for in terms of markedness in the Hallidayan framework.
(6) A: I heard your motorcycle broke down.
B: My car broke down.
Here my car cannot be said to be “that with which the clause is concerned” (see
quotation from Halliday above) and where broke down cannot be said to be “the
part in which the message is developed” (see ib.) – rather the reversed characteri-
sation is correct. Such cases are conceived of as being ‘marked’ in the Hallidayan
framework (see e.g. Halliday 1985/1994: 59). Markedness and unmarkedness in
fact reintroduce the separation between the formal and the meaning/function
sides, as they provide for the situation that the initial element in a clause is not
“that with which the clause is concerned” (the marked case for themes) and the rest
apart from the initial element is not “the part in which the message is developed”
(correspondingly the marked case for rhemes). This allows one to state, for
instance, that while (6B) is marked, its Italian counterpart in (7) is unmarked in
the Hallidayan sense.
(7) Si è rotta la mia macchina.
However, this simply translates back as saying that in the English case the initial
clausal element (in the Hallidayan sense), my car, is not “that with which the clause
is concerned” whereas in the Italian case it (si è rotta) is. The argumentation has
come back to square one in a full circle. We could have spared ourselves this circle
and started right away with a syntactic analysis of the interesting observation that,
whatever information structural category is equivalently expressed by my car on the
one hand and la mia macchina on the other hand – the category focus for Lambrecht
(1994) – it seems to correlate with a syntactic difference manifested by the different
positions of these phrases in surface structure. And, as Lambrecht (1994) has
shown, this leads to the further interesting observation that the same lexical
material arranged in the same predicate-argument configuration as in sentences
(6B) and (7) requires a different syntactic configuration in Italian while disallowing
Carsten Breul
.╅ In previous work (Breul 2004, 2007), I argued that there is an underlying syntactic difference
between such cases as (6B) and (8a):
(i) [FocP [My car]1[+foc] [Foc′ Foc[+foc] [IP t1 broke down]]] (6B)
(ii) [FocP [My car]1[–foc] [Foc′ Foc[–foc] [IP t1 broke down]]] (8a)
In both cases the phrase my car moves to the specifier position of a functional phrase above
IP, called FocP. The movement is triggered by the checking requirement of a syntactic feature,
[+foc] in the first case, and [–foc] in the second case. [+foc] is paired with a semantic focus
feature (my car is an identificational focus expression); [–foc] is paired with a semantic topic
feature (my car is a topic expression).
On the foundations of the contrastive study of information structure 
(10) My encounters with girls were destined always to end in rejection until I’d left
my teens behind me. I caught up a bit during the ‘60s when I became the oldest
teenager in town – in fact I was in my early thirties. As for love, I fell easily
and often. (CH8 1762)
The comparison of attested English DPs in as for DP expressions with their trans-
lationally equivalent German counterparts in corresponding was DP angeht
expressions yields interesting observations concerning contrasts in definite article
Carsten Breul
.╅ Of course, there are contrastive aspects involved in the syntactic structures of the sentences
in (11) (such as the differences in the structures of the verb phrases, gave it to me versus
me l‘a donnée) whose investigation belongs to the domain of contrastive syntactic analysis. The
implication that these differences are irrelevant may be challenged. Challenges of presupposed
or implicated assumptions of this kind will potentially advance our understanding of the
phenomena involved.
On the foundations of the contrastive study of information structure
derived from universal features and principles – an assumption that forms our
tertium comparationis.
(11) a. It’s Isabelle that gave it to me five years ago.
b. C’est Isabelle qui me l’a donnée il y a cinq ans.
Lambrecht’s observation is that the French cleft sentence provided with a suitable
intonation constitutes a perfectly acceptable reply to a context utterance like (the
French counterpart of) I like your shirt, do you remember where you bought it?
whereas the English cleft sentence would be inappropriate in this context with
any intonation. This, then, is a contrastive information structural observation that
involves the information structural category focus structure on the meaning/func-
tion side. The context utterance puts certain constraints on the focus structure of
the reply; the English cleft sentence does not comply with these constraints, while
the French one does.
There is the problem that, most generative approaches apart, some grammatical
theories have built into their very structure the assumption that the formal side,
syntax, for instance, cannot be conceived of independently from the meaning/
function side, information structural categories, for instance. In a strictly func-
tionalist grammatical framework, an item or pattern on the form side is only
identified if it is associated with a certain meaning or function, and the formal
characteristics of this item or pattern are assumed to be determined by this mean-
ing or function. According to Nichols (1984: 97), “[f]unctionalists maintain that
the communicative situation motivates, constrains, explains, or otherwise deter-
mines grammatical structure, and that a structural or formal approach is not
merely limited to an artificially restricted data base, but is inadequate even as a
structural account.” For a contrastive information structure analysis that is based
on such a grammatical theory it does not make sense to state its tertium compa-
rationis on the form side. This is because a meaning or function must have been
identified before an item or pattern on the form side can be said to have been
identified (for otherwise there would be non-functional criteria for identifying
formal entities; the existence, though, of such non-functional criteria is denied).
Consequently, in such a framework it does not make sense to speak of a formal
entity that is equivalent in two languages and whose potentially different charac-
teristics on the meaning/function side are to be investigated. Thus, a contrastive
information structure analysis which is based on a linguistic theory which denies
the independence of the form from the meaning/function side is only possible if
the tertium comparationis is stated on the meaning/function side. That is, in such
a framework contrastive information structure analysis is restricted to the basic
question what the formal means are of expressing a given information structural
category in language A as opposed to language B.
Carsten Breul
derives directly from Popper’s (e.g. 1972) philosophy of science. Popper argues
that all growth of objective knowledge proceeds as a form of problem-solving,
in which hypotheses (tentative theories) are suggested, tested and refuted,
giving rise to revised hypotheses which are in turn tested and revised, and so
on in an endless process of conjecture and refutation. On this view, the goal of
cross-linguistic comparisons is to propose and test falsifiable hypotheses, and
their theoretical value lies precisely in this function (see also Janicki 1990).
(Chesterman 1998: 53)
However, there does not seem to be too deep a rift between Chesterman’s view
and the one sketched in Section 2 above. Of course, the trigger for the detection
of an interesting topic for a contrastive information structure analysis is a per-
ception of similarity or dissimilarity as described by Chesterman. I consider the
reflection on the tertium comparationis and equivalence issues to be the crucial
part of the stage where this perception is “refined and operationally defined as a
similarity constraint” (Chesterman quotation above).15
There is another respect in which the ideal that requires a clear demarcation
of the tertium comparationis on the meaning/function side may have to be rela-
tivised. The point is that a category of information structure is never manifested
independently of other aspects of meaning. For example, identifiability and acti-
vation have to do with denotations or referents of expressions; focus structure,
or the dimensions of topic-comment, of focus-background, and of theme-rheme
have to do with the way in which propositional information is expressed. ‘Denota-
tion’, ‘referent’ and ‘proposition’ essentially being semantic notions, there is no way
of cutting off these meaning aspects from the respective category of information
structure. Consequently, any consideration of the ways in which such a category
is formally expressed in two languages can only proceed against the background
assumption that these necessarily involved meaning aspects do not interfere with
the contrastive analysis. And this presupposes a reliance on being able to keep
these meaning aspects equivalent, with nothing more at our disposal than the
.╅ As far as English/German contrastive information structure analyses are concerned, I may
point to work by Doherty (see Note 2) where this methodological step – refining perceptions
of (dis)similarity and defining them as a similarity constraint – is carried out with utmost
expertise, leading to highly insightful and interesting results.
Carsten Breul
3.â•… E
xamples of contrastive information structure analyses
in the present volume
This section illustrates how the ontological and methodological points about contrastive
information structure analysis presented above are at work in some of the contribu-
tions to the present volume. In addition to the reference to Lambrecht’s work (this
volume) in Section 2.2.2 above, I will briefly refer to four more examples.
The main point of Cohen’s contribution with respect to contrastive information
structure analysis – ‘main’ in my construal for the purposes in the present paper,
not necessarily in Cohen’s – is based on a tertium comparationis on the meaning/
function side. It is actually the building up of the tertium comparationis which
provides the bulk of Cohen’s article. Against the gist of several previous studies,
the author presents a unified, monosemous, account of the function of intensive
reflexives (IRs) in English. She explains how her account relates to information
structure, an important aspect of the explanation being as follows:
[T]he scope of the PNself [i.e. post-nominal IR] marks the referent as an anchor
to which the newer information should be linked, the entry under which new
information is inserted. The VPself [i.e. post-verbal IR] marks the predicate
similarly, thereby marking the set based on it as the anchor entry. The
PAUXself [i.e. post-auxiliary IR] takes scope over the informationally poor
auxiliary. In this case, the IR signals that both the predicate and the referent are
discourse-old and already activated, thereby marking them as anchor entries,
while highlighting the connection between them as the new information
in the discourse. (Cohen this volume: Section 4.2)
Cohen argues that her theory of the function of IRs in English applies equally well to
Hebrew IRs. This step in the line of argumentation might be considered a contras-
tive observation in its own right, based on a tertium comparationis on the form side,
i.e. on the identification of the lexico-grammatical category IR in both languages, and
revealing a cross-linguistic commonality rather than a difference. However, my con-
strual of Cohen’s main contrastive observation affects precisely formal aspects of IRs:
While IR scope effects are evident in both languages, they differ somewhat in
the specific linguistic marking of this scope. As in English, some Hebrew IRs
mark their scope by linear position. Thus, the bare PNself and the b- marked
VPself take scope backwards over the preceding segment […]. However, two
important differences must be considered: the wider range of positions open to
the Hebrew IR and its occurrence with a preposition. Unlike English, Hebrew
requires prepositional marking with some IRs. As noted in Section 2, bare
IRs can occur with any nominal antecedent and must immediately follow it,
and so are identified as PNself. In contrast, b- IRs require subject antecedents.
(Cohen this volume: Section 4.1)
Thus, strictly speaking, we cannot take the step of argumentation just mentioned to
be a legitimate contrastive observation in its own right, embedded in the main one.
The form side of the IRs in Hebrew in comparison to English is at issue in the main
contrastive observation, revealing a difference. Consequently, it cannot be made the
tertium comparationis of another contrastive analysis aiming at the meaning/function
side. The challenge that Cohen’s main contrastive observation sets, a challenge inviting
attempts at falsification and thus being of significant scientific value, is this: Can
the claim that the differences between English and Hebrew IRs on the form side
do not correspond to differences on the information structural meaning/function
side be maintained in the light of further evidence?
The contribution by López contains several lines of argumentation that con-
stitute instances of contrastive information structure analysis. I will pick one of
them for illustration: On the meaning/function side, López makes a distinction
between ‘givenness’ and ‘discourse anaphoricity’.
(12) a. Context: I’m wearing a red coat. What are you wearing?
i. I’m wearing a blue coat.
ii. I’m wearing a blue shirt.
b. Context: What kind of coat are you wearing?
i. I’m wearing a blue coat.
ii. #I’m wearing a blue shirt.
For López (this volume: Section 1), the expression coat in (12ai) is ‘given’ – by virtue of
being only ‘accidentally’ occurring again, after having been mentioned in the
Carsten Breul
preceding question, as part of the larger focus expression; cf. the appropriateness
of (12aii) as a reply to (12a). In (12bi) coat is ‘discourse anaphoric’ – by virtue of
being mandatorily coreferential with an antecedent in the previous discourse and not
being a part of the focus expression; cf. the inappropriateness of an alternative like
(12bii) as a reply to (12a). These two information structural categories, ‘givenness’
and ‘discourse anaphoricity’, can be applied to both English and Catalan, that is,
each of them may serve as a tertium comparationis on the meaning/function side.
English and Catalan show interesting non-equivalences on the form side in the
manifestation of these information structure categories. For example, a Catalan
‘discourse anaphoric’ constituent, in contrast to an English one, is obligatorily
clitic right-dislocated, as shown by (13) (see ib., and p.c.).
(14) A: Mary drove her blue convertible. What did John drive?
B: Va conduir el seu sedan blau.
past drive.inf the her/his sedan blue
‘He drove her/his blue sedan.’
both English and German have contours that are used in contexts of
‘sub-informativity’, but they are used at different levels of generality: the English
fall-rise is a general marker of ‘incompleteness’, and therefore covers ‘sub-
informativity’ […] as one of its functions, whereas the German root contour is a
rather specific marker of ‘context-changing sub-informativity’. (Ib.: Section 7.3)
This remark reminds me of Hawkins’s (1986, 1988) well-known claim about the
tighter fit in German than in English between syntactic surface structures on
the one hand and semantic relations between predicates and their arguments on
the other. It seems worthwhile investigating whether there is also a tighter fit in
German than in English between intonational patterns on the one hand and the
information structural categories they are associated with on the other. Gast’s
observation concerning the English fall-rise and the German root contour points
in this direction.
4.â•… Conclusion
References
Chafe, W. 1976. Givenness, contrastiveness, definiteness, subjects, topics and point of view. In
Subject and Topic, C. Li (ed.), 25–56. New York NY: Academic Press.
Chafe, W. 1987. Cognitive constraints on information flow. In Coherence and Grounding in
Discourse [Typological Studies in Language 11], R. Tomlin (ed.), 21–52. Amsterdam:
John Benjamins.
Chesterman, A. 1998. Contrastive Functional Analysis [Pragmatics & Beyond New Series 47].
Amsterdam: John Benjamins.
Collaborative Research Centre (Sonderforschungsbereich, SFB) ‘Information Structure: The
Linguistic Means for Structuring Utterances, Sentences and Texts’. 〈http:↜//www↜.sfb632.
uni-potsdam.de/main.html〉.
Comrie, B. 1986. Contrastive linguistics and language typology. In Linguistics across Historical and
Geographical Boundaries: In Honour of Jacek Fisiak on the Occasion of his Fiftieth Birthday,
Vol. 2: Descriptive, Contrastive and Applied Linguistics, D. Kastovsky & A. Szwedek (eds),
1155–1163. Berlin: Mouton de Gruyter.
Doherty, M. 1996. Passive perspectives; different preferences in English and German: A result of
parameterized processing. Linguistics 34(3): 591–643.
Doherty, M. 1999. Position and explicitness: Language specific conditions for the use of adverbial
clauses in translations between German and English. In Sprachspezifische Aspekte der
Informationsverteilung, M. Doherty (ed.), 112–148. Berlin: Akademie Verlag.
Doherty, M. 2002. Language Processing in Discourse: A Key to Felicitous Translation. London:
Routledge.
Erdmann, P. 1990. Fokuskonstruktionen im Deutschen und Englischen. In Kontrastive Linguistik,
C. Gnutzmann (ed.), 69–83. Frankfurt: Peter Lang.
Erdmann, P. 1993. Subjekt- und Topikprominenz im Deutschen, Englischen und Koreanischen.
In Nachbarschaften: Festschrift für Max Mangold, M. Bonner & E. Braun & H. Fix (eds),
319–336. Saarbrücken: Saarbrücker.
Erteschik-Shir, N. 2007. Information Structure: The Syntax-Discourse Interface. Oxford: OUP.
Esser, J. 1995. Einzelaspekt: Wortstellung und Fokusstrukturen. In Handbuch Englisch als
Fremdsprache, R. Ahrens & W.-D. Bald & W. Hüllen (eds), 151–154. Berlin: Erich Schmidt.
Fabricius-Hansen, C. 1999. Information packaging and translation: Aspects of translational sentence
splitting (German – English/Norwegian). In Sprachspezifische Aspekte der Informations-
verteilung, M. Doherty (ed.), 175–214. Berlin: Akademie Verlag.
Firbas, J. 1959. Thoughts on the communicative function of the verb in English, German and
Czech. Brno Studies in English 1: 39–63.
Firbas, J. 1964. From comparative word-order studies. Brno Studies in English 4: 111–128.
Haegeman, L. & Guéron, J. 1999. English Grammar: A Generative Perspective. Oxford: Blackwell.
Halliday, M.A.K. 1985/1994. An Introduction to Functional Grammar, 2nd edn. London: Arnold.
Hawkins, J.A. 1986. A Comparative Typology of English and German: Unifying the Contrasts.
London: Croom Helm
Hawkins, J.A. 1988. The unity of English/German contrasts: Inferring a typological parameter.
In On Language: Rhetorica, Phonologica, Syntactica: A Festschrift for Robert P. Stockwell
from his Friends and Colleagues, C. Duncan-Rose & T. Vennemann (eds), 361–380.
London: Routledge.
Huddleston, R. & Pullum, G.K. 2002. The Cambridge Grammar of the English Language.
Cambridge: CUP.
James, C. 1980. Contrastive Analysis. London: Longman.
Carsten Breul
James, C. (ed). 1996. Cross-linguistic Approaches to Language Awareness. Special issue of Language
Awareness 5(3–4).
Janicki, K. 1990. A brief falsificationist look at contrastive sociolinguistics. Papers and Studies in
Contrastive Linguistics 26: 5–10.
Kayne, R.S. 1996/2000. Parameters and Universals. Oxford: OUP.
Kirkwood, H.W. 1969. Aspects of word order and its communicative function in English and
German. Journal of Linguistics 5: 85–107.
Kirkwood, H.W. 1970. Some systemic means of ‘functional sentence perspective’ in English and
German. International Review of Applied Linguistics in Language Teaching (IRAL) 8(2):
103–114.
Kirkwood, H.W. 1978. Options and constraints in the surface ordering of noun phrases in English
and German. Journal of Pragmatics 2: 225–245.
Klaiman, M.H. 1991. Grammatical Voice. Cambridge: CUP.
Klein, E. 1988. A contrastive analysis of focus phenomena in English and German on a functional
basis and some implications for a didactic grammar. Die Neueren Sprachen 87(4): 371–386.
König, E. 1990. Kontrastive Linguistik als Komplement zur Typologie. In Kontrastive Linguistik,
C. Gnutzmann (ed.), 117–131. Frankfurt: Peter Lang.
König, E. 1992. Contrastive linguistics and language typology. In New Departures in Contrastive
Linguistics: Proceedings of the Conference Held at the Leopold-Franzens-University of Inns-
bruck, Austria, 10–12 May 1991, Vol. 1, C. Mair & M. Markus (eds), 137–154. Innsbruck:
Institut für Anglistik, Universität Innsbruck.
König, E. 1993. Contrastive linguistics: Language comparison or language pedagogy? In Ang-
listentag 1992 Stuttgart: Proceedings, H.U. Seeber & W. Göbel (eds), 289–302. Tübingen:
Niemeyer.
König, E. 1996. Kontrastive Grammatik und Typologie. In Deutsch – typologisch, E. Lang &
G. Zifonun (eds), 31–54. Berlin: Walter de Gruyter.
Kortmann, B. 1998. Kontrastive Linguistik und Fremdsprachenunterricht. In: Kontrast und
Äquivalenz: Beiträge zu Sprachvergleich und Übersetzung, W. Börner & K. Vogel (eds),
136–167. Tübingen: Narr.
Krzeszowski, T. 1990. Contrasting Languages: The Scope of Contrastive Linguistics. Berlin: Mouton
de Gruyter.
Lambrecht, K. 1994. Information Structure and Sentence Form: Topic, Focus and the Mental
Representations of Discourse Referents. Cambridge: CUP.
Lattey, E. 1982. What is the ‘same thing’ in interlinguistic comparison? In The Contrastive
Grammar of English and German, W.F.W. Lohnes & E.A. Hopkins (eds), 131–144. Ann
Arbor MI: Karoma.
Legenhausen, L. & Rohdenburg, G. 1995. Kontrastivierung ausgewählter Strukturen im Englischen
und Deutschen. In Handbuch Englisch als Fremdsprache, R. Ahrens & W.-D. Bald &
W. Hüllen (eds), 133–139. Berlin: Erich Schmidt.
Nichols, J. 1984. Functional theories of grammar. Annual Review of Anthropology 13: 97–117.
Pan, W. & Tham, W.M. 2007. Contrastive Linguistics: History, Philosophy and Methodology.
London: Continuum.
Popper, K. 1972. Objective Knowledge. An Evolutionary Approach. Oxford: Clarendon Press.
Schwabe, K. & Winkler, S. (eds). 2007. On Information Structure, Meaning and Form:
Generali�zations across Languages, [Linguistik Aktuell/Linguistics Today 100]. Amsterdam:
John Benjamins.
On the foundations of the contrastive study of information structure
Sheen, R. 1996. The advantage of exploiting contrastive analysis in teaching and learning a foreign
language. International Review of Applied Linguistics in Language Teaching 34: 183–197.
Weinert, R. 1995. Focusing constructions in spoken language. Clefts, Y-movement, thematization
and deixis in English and German. Linguistische Berichte 159: 341–369.
Zimmermann, R. 1972. Themenfrontierung, Wortstellung und Intonation im Deutschen und
Englischen. Die Neueren Sprachen 71: 15–28.
Subject index
A E I
accessibility╇ 201, 256ff. echo question╇ 107ff. identifiability╇ 16f., 286, 290f.
AC-profile╇ 35f. economy╇ 190 implicature╇ 33, 37f., 43,
activation╇ 201ff., 286 English╇ 15ff., 51ff., 77ff., 101ff., 146ff., 163
addressation╇ 19, 257f. 139ff., 169ff., 231ff., 277ff. inclusion╇ 142, 144, 146ff.
alloquestion╇ 104ff. equivalence╇ 283ff., 288ff., 292, incompleteness╇ 34, 37f., 41, 44
alternatives (set of ~)╇ 20f., 42, 294ff. inference╇ 148, 150, 162f., 178
144ff., 158f., 162f., 193 exclusion╇ 142, 144, 146ff. Italian╇ 231ff.
Extended Projection i-topicalisation╇ 22
B Principle╇ 240ff.,
B-accent╇ 34f., 59 246ff. J
Japanese╇ 231ff.
C F
canonical / non-canonical fall-rise contour╇ 34ff. K
(constituent order, familiarity╇ 255, 262ff. Korean╇ 231ff.
position, etc.)╇ 83, 86, feature specification of omitted
88, 90, 93ff., 170f., 174ff., subjects╇ 209f., 216f. L
184ff., 190ff., 231, 233, focus structure╇ 61ff., 77ff., linearisation╇ 70ff., 171ff., 231ff.,
238ff., 269 286, 293 253ff.
Catalan╇ 51ff. focus╇ pass. link╇ 31f., 54ff., 233f.
cleft construction╇ 176ff., 186ff., – (non-)identificational
pass. focus╇ 170f. M
– avoir cleft╇ 93ff. – in situ focus╇ 172f. minimality╇ 190f.
– c’est cleft╇ 86ff. forefield╇ 32ff. movement pass.
clitic left dislocation╇ 55ff. French╇ 77ff., 101ff., 169ff., – Aʹ-movement╇ 33
clitic pronoun╇ 83f., 125ff. 199ff. – Formal Movement╇ 33
clitic right dislocation╇ 53ff., fronting; see also
62, 64 ‘preposing’╇ 30ff., 175, N
comparison (set); see also 253f., 266f. Northern Italian dialects╇ 128f.
‘alternatives’╇ 152, 154f.,
158f., 162f. G P
contrast, contrastiveness, Georgian╇ 169ff. perspectivation╇ 234, 263ff.
contrastivity╇ 15ff., 89, 142, German╇ 15ff., 199ff., 231ff., point of view╇ 263ff.
145, 152, 260ff. 277ff. precedence╇ 269f.
cross-linguistic influence╇ 216 givenness╇ 17, 33, 51ff., 262f. preferred clause
ground╇ 54ff. construction╇ 83ff.
D preposing; see also
definiteness╇ 17, 93f., 263, 265ff. H ‘fronting’╇ 30, 83, 176
discourse hat contour╇ 39f. preposition b- (Hebrew)╇ 142ff.
configurationality╇ 234, Hebrew╇ 139ff. presentational
237ff. hierarchy╇ 232ff., 242ff., 253ff., construction╇ 184ff.
D-linking╇ 130f. 269f. preverb╇ 174
d-tree╇ 22ff. Hungarian╇ 169ff., 231ff. pronoun strength╇ 126, 132f.
Subject index
169 Sánchez, Liliana: The Morphology and Syntax of Topic and Focus. Minimalist inquiries in the Quechua
periphery. Expected November 2010
168 Feldhausen, Ingo: Sentential Form and Prosodic Structure of Catalan. xiii, 280 + index. Expected
October 2010
167 Mercado, Raphael, Eric Potsdam and Lisa deMena Travis (eds.): Austronesian and Theoretical
Linguistics. vii, 374 pp. + index. Expected October 2010
166 Brandt, Patrick and Marco García García (eds.): Transitivity. Form, Meaning, Acquisition, and
Processing. vi, 300 pp. + index. Expected October 2010
165 Breul, Carsten and Edward Göbbel (eds.): Comparative and Contrastive Studies of Information
Structure. 2010. xii, 306 pp.
164 Zwart, Jan-Wouter and Mark de Vries (eds.): Structure Preserved. Studies in syntax for Jan Koster.
2010. xxiii, 395 pp.
163 Kiziak, Tanja: Extraction Asymmetries. Experimental evidence from German. 2010. xvi, 273 pp.
162 Bott, Oliver: The Processing of Events. xix, 379 pp. + index. Expected September 2010
161 Haan, Germen J. de: Studies in West Frisian Grammar. Edited by Jarich Hoekstra, Willem Visser and
Goffe Jensma. 2010. x, 384 pp.
160 Mavrogiorgos, Marios: Clitics in Greek. A minimalist account of proclisis and enclisis. 2010.
x, 294 pp.
159 Breitbarth, Anne, Christopher Lucas, Sheila Watts and David Willis (eds.): Continuity and
Change in Grammar. 2010. viii, 359 pp.
158 Duguine, Maia, Susana Huidobro and Nerea Madariaga (eds.): Argument Structure and
Syntactic Relations. A cross-linguistic perspective. 2010. vi, 348 pp.
157 Fischer, Susann: Word-Order Change as a Source of Grammaticalisation. 2010. ix, 200 pp.
156 Di Sciullo, Anna Maria and Virginia Hill (eds.): Edges, Heads, and Projections. Interface properties.
2010. vii, 265 pp.
155 Sato, Yosuke: Minimalist Interfaces. Evidence from Indonesian and Javanese. 2010. xiii, 159 pp.
154 Hornstein, Norbert and Maria Polinsky (eds.): Movement Theory of Control. 2010. vii, 330 pp.
153 Cabredo Hofherr, Patricia and Ora Matushansky (eds.): Adjectives. Formal analyses in syntax
and semantics. 2010. vii, 335 pp.
152 Gallego, Ængel J.: Phase Theory. 2010. xii, 365 pp.
151 Sudhoff, Stefan: Focus Particles in German. Syntax, prosody, and information structure. 2010.
xiii, 335 pp.
150 Everaert, Martin, Tom Lentz, Hannah de Mulder, Øystein Nilsen and Arjen Zondervan
(eds.): The Linguistics Enterprise. From knowledge of language to knowledge in linguistics. 2010. ix, 379 pp.
149 Aelbrecht, Lobke: The Syntactic Licensing of Ellipsis. 2010. xii, 230 pp.
148 Hogeweg, Lotte, Helen de Hoop and Andrej Malchukov (eds.): Cross-linguistic Semantics of
Tense, Aspect, and Modality. 2009. vii, 406 pp.
147 Ghomeshi, Jila, Ileana Paul and Martina Wiltschko (eds.): Determiners. Universals and variation.
2009. vii, 247 pp.
146 Gelderen, Elly van (ed.): Cyclical Change. 2009. viii, 329 pp.
145 Westergaard, Marit: The Acquisition of Word Order. Micro-cues, information structure, and
economy. 2009. xii, 245 pp.
144 Putnam, Michael T. (ed.): Towards a Derivational Syntax. Survive-minimalism. 2009. x, 269 pp.
143 Rothmayr, Antonia: The Structure of Stative Verbs. 2009. xv, 216 pp.
142 Nunes, Jairo (ed.): Minimalist Essays on Brazilian Portuguese Syntax. 2009. vi, 243 pp.
141 Alexiadou, Artemis, Jorge Hankamer, Thomas McFadden, Justin Nuger and Florian
Schäfer (eds.): Advances in Comparative Germanic Syntax. 2009. xv, 395 pp.
140 Roehrs, Dorian: Demonstratives and Definite Articles as Nominal Auxiliaries. 2009. xii, 196 pp.
139 Hicks, Glyn: The Derivation of Anaphoric Relations. 2009. xii, 309 pp.
138 Siddiqi, Daniel: Syntax within the Word. Economy, allomorphy, and argument selection in Distributed
Morphology. 2009. xii, 138 pp.
137 Pfau, Roland: Grammar as Processor. A Distributed Morphology account of spontaneous speech errors.
2009. xiii, 372 pp.
136 Kandybowicz, Jason: The Grammar of Repetition. Nupe grammar at the syntax–phonology interface.
2008. xiii, 168 pp.
135 Lewis, William D., Simin Karimi, Heidi Harley and Scott O. Farrar (eds.): Time and Again.
Theoretical perspectives on formal linguistics. In honor of D. Terence Langendoen. 2009. xiv, 265 pp.
134 Armon-Lotem, Sharon, Gabi Danon and Susan D. Rothstein (eds.): Current Issues in
Generative Hebrew Linguistics. 2008. vii, 393 pp.
133 MacDonald, Jonathan E.: The Syntactic Nature of Inner Aspect. A minimalist perspective. 2008.
xv, 241 pp.
132 Biberauer, Theresa (ed.): The Limits of Syntactic Variation. 2008. vii, 521 pp.
131 De Cat, Cécile and Katherine Demuth (eds.): The Bantu–Romance Connection. A comparative
investigation of verbal agreement, DPs, and information structure. 2008. xix, 355 pp.
130 Kallulli, Dalina and Liliane Tasmowski (eds.): Clitic Doubling in the Balkan Languages. 2008.
ix, 442 pp.
129 Sturgeon, Anne: The Left Periphery. The interaction of syntax, pragmatics and prosody in Czech. 2008.
xi, 143 pp.
128 Taleghani, Azita H.: Modality, Aspect and Negation in Persian. 2008. ix, 183 pp.
127 Durrleman-Tame, Stephanie: The Syntax of Jamaican Creole. A cartographic perspective. 2008.
xii, 190 pp.
126 Schäfer, Florian: The Syntax of (Anti-)Causatives. External arguments in change-of-state contexts. 2008.
xi, 324 pp.
125 Rothstein, Björn: The Perfect Time Span. On the present perfect in German, Swedish and English.
2008. xi, 171 pp.
124 Ihsane, Tabea: The Layered DP. Form and meaning of French indefinites. 2008. ix, 260 pp.
123 Stoyanova, Marina: Unique Focus. Languages without multiple wh-questions. 2008. xi, 184 pp.
122 Oosterhof, Albert M.: The Semantics of Generics in Dutch and Related Languages. 2008. xviii, 286 pp.
121 Tungseth, Mai Ellin: Verbal Prepositions and Argument Structure. Path, place and possession in
Norwegian. 2008. ix, 187 pp.
120 Asbury, Anna, Jakub Dotlačil, Berit Gehrke and Rick Nouwen (eds.): Syntax and Semantics of
Spatial P. 2008. vi, 416 pp.
119 Fortuny, Jordi: The Emergence of Order in Syntax. 2008. viii, 211 pp.
118 Jäger, Agnes: History of German Negation. 2008. ix, 350 pp.
117 Haugen, Jason D.: Morphology at the Interfaces. Reduplication and Noun Incorporation in Uto-Aztecan.
2008. xv, 257 pp.
116 Endo, Yoshio: Locality and Information Structure. A cartographic approach to Japanese. 2007. x, 235 pp.
115 Putnam, Michael T.: Scrambling and the Survive Principle. 2007. x, 216 pp.
114 Lee-Schoenfeld, Vera: Beyond Coherence. The syntax of opacity in German. 2007. viii, 206 pp.
113 Eythórsson, Thórhallur (ed.): Grammatical Change and Linguistic Theory. The Rosendal papers. 2008.
vi, 441 pp.
112 Axel, Katrin: Studies on Old High German Syntax. Left sentence periphery, verb placement and verb-
second. 2007. xii, 364 pp.
111 Eguren, Luis and Olga Fernández-Soriano (eds.): Coreference, Modality, and Focus. Studies on
the syntax–semantics interface. 2007. xii, 239 pp.
110 Rothstein, Susan D. (ed.): Theoretical and Crosslinguistic Approaches to the Semantics of Aspect. 2008.
viii, 453 pp.
109 Chocano, Gema: Narrow Syntax and Phonological Form. Scrambling in the Germanic languages. 2007.
x, 333 pp.
108 Reuland, Eric, Tanmoy Bhattacharya and Giorgos Spathas (eds.): Argument Structure. 2007.
xviii, 243 pp.
107 Corver, Norbert and Jairo Nunes (eds.): The Copy Theory of Movement. 2007. vi, 388 pp.
106 Dehé, Nicole and Yordanka Kavalova (eds.): Parentheticals. 2007. xii, 314 pp.
105 Haumann, Dagmar: Adverb Licensing and Clause Structure in English. 2007. ix, 438 pp.
104 Jeong, Youngmi: Applicatives. Structure and interpretation from a minimalist perspective. 2007.
vii, 144 pp.