Download Structuring information in discourse: the explicit/implicit

Document related concepts

Untranslatability wikipedia , lookup

Causative wikipedia , lookup

Polish grammar wikipedia , lookup

Indexicality wikipedia , lookup

Modern Hebrew grammar wikipedia , lookup

Yiddish grammar wikipedia , lookup

Ancient Greek grammar wikipedia , lookup

Distributed morphology wikipedia , lookup

Georgian grammar wikipedia , lookup

Kannada grammar wikipedia , lookup

Musical syntax wikipedia , lookup

Universal pragmatics wikipedia , lookup

Sloppy identity wikipedia , lookup

Indeterminacy (philosophy) wikipedia , lookup

Pleonasm wikipedia , lookup

Meaning (philosophy of language) wikipedia , lookup

Japanese grammar wikipedia , lookup

Latin syntax wikipedia , lookup

Spanish grammar wikipedia , lookup

Transformational grammar wikipedia , lookup

Parsing wikipedia , lookup

Antisymmetry wikipedia , lookup

Semantic holism wikipedia , lookup

Interpretation (logic) wikipedia , lookup

Pipil grammar wikipedia , lookup

Junction Grammar wikipedia , lookup

Cognitive semantics wikipedia , lookup

Lexical semantics wikipedia , lookup

Focus (linguistics) wikipedia , lookup

Oslo Studies in Language
1 / 2009
Bergljot Behrens
& Cathrine Fabricius-Hansen (eds.)
Structuring information in discourse:
the explicit/implicit dimension
Oslo Studies in Language
General editors: Atle Grønn and Dag Haug
Editorial board
Henning Andersen, Los Angeles (historical linguistics)
Östen Dahl, Stockholm (typology)
Laura Janda, Tromsø/UNC Chapel Hill (Slavic linguistics, cognitive linguistics)
Arnim von Stechow, Tübingen (semantics and syntax)
Johanna Barðdal, Bergen (construction grammar)
Øystein Vangsnes, Tromsø (Norwegian, dialect syntax)
Cecilia Alvstad, ILOS (Spanish, translatology)
Hans Olav Enger, ILN (Norwegian, cognitive linguistics)
Ruth E. Vatvedt Fjeld, ILN (Norwegian, lexicography)
Jan Terje Faarlund, CSMN, ILN (Norwegian, syntax)
Cathrine Fabricius-Hansen, ILOS (German, contrastive linguistics)
Carsten Hansen, CSMN, IFIKK (philosophy of language)
Christoph Harbsmeier, IKOS (Chinese, lexicography)
Hilde Hasselgård, ILOS (English, corpus linguistics)
Hans Petter Helland, ILOS (French, syntax)
Kristian Emil Kristoffersen, ILN (cognitive linguistics)
Helge Lødrup, ILN (syntax)
Gunvor Mejdell, IKOS (Arabic, sociolinguistics)
Ljiljana Saric, ILOS (Slavic linguistics)
Bente Ailin Svendsen, ILN (second language acquisition)
Oslo Studies in Language
1 / 2009
Bergljot Behrens
& Cathrine Fabricius-Hansen (eds.)
Structuring information in discourse:
the explicit/implicit dimension
Oslo Studies in Language, volume 1, 2009.
Bergljot Behrens and Cathrine Fabricius-Hansen (eds.):
Structuring Information in Discourse: the Explicit/Implicit Dimension
Oslo, University of Oslo
ISSN 1890-9639
© 2009 the authors
Layout and design by Lars Bungum, Atle Grønn, Dag Haug and Karine Stjernholm
Set in LATEX fonts Gentium Book Basic and Linux Libertine by Lars Bungum
Cover design by UniPub publishing house
Printed by UniPub from camera-ready copy supplied by the editors
T  C
1 Introduction
Bergljot Behrens and Cathrine Fabricius-Hansen, University of Oslo
2 On the Functional Independence of Explicatures and Implicatures
Thorstein Fretheim, Norwegian University of Science and Technology
3 Cross-Linguistic Evidence and the Licensing of Implicit Arguments
Gergely Pethő and Eva Kardos, University of Debrecen
4 A Morpheme-based Model of Nonsentential Utterance Production
Shinji Ido, Tohoku University / University of Sydney
5 The Real, the Apparent, and What is eigentlich
Regine Eckardt, University of Göttingen
6 Self Intensification and Focus Interpretation
Kjell Johan Sæbø, University of Oslo
7 Adverbial doch and the Notion of Contrast
Elena Karagjosova, University of Oslo
8 A Formal Analysis of the French Temporal Connective alors
Myriam Bras, Anne Le Draoulec and Nicholas Asher, University of
9 Discourse Structure: Swings and Roundabouts
Bonnie Webber, University of Edinburgh and Rashmi Prasad, University
of Pennsylvania
10 Optimal Interpretation as an Alternative to Gricean Pragmatics
Henk Zeevat, University of Amsterdam
Behrens & Fabricius-Hansen (eds.) Structuring information in discourse:
the explicit/implicit dimension, Oslo Studies in Language 1(1), 2009. 1-15. (ISSN 1890-9639)
University of Oslo
   
The present thematic issue comprises a selection of nine refereed and revised
papers, most of which were presented at the SPRIK conference Explicit versus Implicit Information in Text. Information Structure across Languages in Oslo, 8-10 June
2006.1 The bipartite title of the conference reflects the main objectives of the
organizing research project SPRIK2 (‘SPRåk I Kontrast / Languages in Contrast’),
which is directed towards text-oriented, corpus-based contrastive studies (Norwegian, English, French, German) of the interplay between explicit (linguistically
encoded) information and implicit information, on the one hand, and the interaction of local, sentence-internal information structure and more global information structuring and weighting, including so-called discourse structure, on the
other. Nine explicitly contrastive and partly translation-oriented papers presented at the conference were selected for a special issue of the journal Languages
in Contrast (Behrens et al. 2007). The present issue of OSLa is more theoretically
oriented, focussing on the system(s) or procedures that can account for the structure of complex discourse. While theoretical studies and empirical studies sometimes compete in the scientific community, and researchers change from a preoccupation with theory to a preoccupation with data, the majority of the papers
in the present publication combine data and theory very closely. With a few exceptions the papers are monolingually oriented, but taken together they cover
a variety of object languages: English, German, Norwegian, French, Hungarian,
Turkish, Japanese, and Mongolian.
The contributions comprise
(i) Three keynote papers by Regine E, Kjell Johan S and Bonnie W
, and an ‘unofficial’ keynote paper by Henk Z. The papers demonstrate or argue for somewhat different theoretical approaches to the interaction of lexically encoded meaning, contextual information, information
See The SPRIK project at the Faculty of
Humanitites, University of Oslo, was funded by the Norwegian Research Council (NFR) under project
number 158447/530 (2003-2008).
  
structure and/or discourse structure: Formal (Dynamic) Semantics combined with Focus Theory à la Rooth (Eckardt, Sæbø), Optimality Theoretic
Pragmatics (Zeevat, Sæbø), and Discourse-level Lexicalized Tree Adjoining
Grammar (D-LTAG) (Webber).
(ii) Five case studies that are thematically, theoretically and/or methodologically related to one or more of the keynote papers listed in (i). These case
studies also widen the perspective somewhat: one paper presenting a unified context-dependent semantics for the conjunct adverb doch (Elena K
); one concerned with the interplay between connectives/particles
and discourse/rhetorical relations (Myriam B et al.,); one paper addressing the division of labor between syntax, semantics and pragmatics as far
as the licensing and interpretation of implicit objects is concerned (Gergely
P and Eva K); one relevance-theoretically oriented paper demonstrating lexical relations which form a very important basis for questioning
the explicature-implicature distinction (Thorstein F); and one paper outlining a model for non-sentential utterance production (Shinji I).
The general research questions addressed in this volume are highly relevant in
the present linguistic debate, centering on interfaces between different levels of
description. Where in our account of the relation between language and meaning
do the constraints on interpretation belong? In accounting for the meaning of
texts, how much and what kind of meaning do we attach to the linguistic signals
themselves? What do these signals encode that contributes to the construction
and resolution of context-dependent meaning? To put the questions differently,
how much do language users base their interpretations on world knowledge and
general pragmatic principles, such as for example the Relevance Theoretic principle of looking for optimal relevance/cognitive effects at the lowest cost, or the
Optimality Theoretic instruction to assess the relative weight of a series of basic
pragmatic principles to arrive at the best interpretation? Pure pragmatic reasoning is clearly necessary for the language user to infer intended meanings, but
while some adherents of pragmatics tend to attach most of the interpretation procedures to general pragmatic principles (see the critical discussion in Pethő and
Kardos in the present volume), several of the contributions in the present volume
demonstrate the essential role of linguistic, text-structuring clues in a compositional system of arriving at the full meaning of complex discourse.
 
This book deals with the challenges of accounting for the systems that direct the
retrieval of contextual information in the interpretation of coherent discourse.
Discourse – written text or a sequence of spoken utterances – is coherent when
OSLa volume 1(1), 2009
a single utterance or expression (sentential or non-sentential) is not only interpreted at face value, by and of itself, but makes sense as a meaningful contribution
to the linguistic and non-linguistic context it appears in.
At one level this means that we look for the factors that determine the expression’s contribution to what the discourse refers to. At another level we are after
the factors that determine its illocution, i.e. what the expression may be meant
to do in the context – whether to add, challenge or modify information, by way
of representing an opinion, underlining a viewpoint, creating a contrast etc. For
both levels of interpretation, discourse must be structured. The question for the
linguist is not only what that structure is but what the factors are that determine
the structures.
     
When we say something, we often take it for granted that what we mean by what
we say will not only be available to the listener but that what the listener takes us
to mean on the basis of what we say is in fact what we mean. That is, we expect
the linguistic code we have made use of to be sufficient to express our intended
meaning. However, as linguists, we know that a lot of what a speaker communicates by what s/he says is not explicitly stated. Well-known examples are ellipsis,
or omitted words or phrases that are somehow recoverable from general rules of
syntax, such as the implied object argument in the second conjunct in (1).
The king picked up the wine and drank.
An utterance of (1) will be understood by all reasonable listeners to mean that
the king drank (at least some of) the wine that is referred to in the first conjunct.
Although this is only a piece of  information given by the utterance, most
linguists agree that it makes up part of the truth conditions (semantic content) of
the utterance.
The same listeners will also agree that the utterance in (1) implies that the
king picked up a glass or some other container that can hold wine and that is
meant for drinking, and that the wine he drank was in that glass/bowl or bottle/flask. This is also information that is not stated explicitly. Yet the different
pieces of implicit information retrieved from the utterance of (1), one might argue, are retrieved on the basis of different interpretation principles: First of all,
the object of the verb drink may be postulated as an underspecified argument in
the syntactic structure of the second conjunct, determined by the selection restrictions for the verb. Simplifying somewhat, we may say that the underspecified
argument lacks intrinsic reference and is therefore assumed to be an anaphor,
whose reference must be resolved in context. Although everyone would agree
OSLa volume 1(1), 2009
  
that the king drank (some of) the wine, not all linguists would agree that the
above description of how that meaning comes about is a correct description. The
pragmatic principle guiding the resolution of the anaphor, furthermore, is also
a matter of debate: on a linguistically based view, a principle of non-intervening
events may be stated for events expressed in a VP conjunction, which means that
the implied object of the drinking event should find its antecedent in the first conjunct (Discourse RepresentationTheoretic (DRT) approach). On another view the
anaphor is resolved as the result of a general search for relevance (the Relevance
Theoretic (RT) approach).
The other information is retrieved on the basis of the listeners’ general knowledge about wine containers and the impossibility of picking up a liquid which is
not in a container of some sort. Typically, the exact reference of this kind of
knowledge-based information remains vague or loose without further contextual
clues or more specific knowledge about the actual state of affairs. The restricted
context given in (1) contributes to narrowing the reference of “wine” down to
“wine in some container”. A wider context might narrow it down even further.
The distinction between the types of implicit information in (1) is generally
stated as a difference between linguistically motivated information and general
world knowledge as a basis for inferences drawn as to the meaning of an utterance. This distinction is often hard to draw, and different theories of language
draw the line differently. The problem of accounting for systems of “filling in”
material “missing” in the expression to arrive at the underlying proposition is of
central concern to linguistic theory.
Grice (1975) made the important distinction between information that makes
up part of the truth conditions of an utterance and “cancellable” meanings drawn
from the utterance, which the speaker cannot be held responsible for, and correlated the division with a distinction between what is said and what is implicated.
Levinson (2000) explicates this view in saying that the “said” can be taken to be
truth-conditional content, the output of semantic interpretation (the proposition
expressed), while the “implicated” can be taken to include all the processes of
pragmatic inference.
If this is the distinction to be drawn, then information inferred from the utterance should not make up part of the truth conditions of the utterance, by definition.
Such a view, according to a broad range of linguists and language philosophers
today, is clearly mistaken. The question, therefore, is: What can be said to be ‘said’
on the basis of an utterance or piece of discourse?
Answers to this question are sought in various ways, as was illustrated by our
analysis of (1).
The first problem is to determine the grounds for which an object argument
should be postulated in the syntax of the utterance. This is a theoretical matter.
OSLa volume 1(1), 2009
The plausibility of the assumption that there is such an argument is determined
on the basis of its explanatory power. Without it one would have to postulate two
verbs drink – one transitive, the other intransitive; and one would have to account
for the relationship between the two. With an implicit object, on the other hand,
one would need just one lexical verb drink which is the same whether it is used
with or without an overt object. Without an overt object, an object position would
nevertheless be generated in the syntax, and a “zero” anaphor would be mapped
onto the syntactic position. The result is a “free” referent in the semantic representation, i.e. a referent which needs to find its descriptive content in the larger
Since the meaning of the utterance, it would be agreed, is in fact that the king
drank (some of) the wine he picked up, and nothing else, the suggested analysis
of a free, contextually resolved anaphor is highly plausible.
An alternative explanation, in which pragmatic principles play a stronger role,
is that there is no level of linguistic representation of (1) in which there is an
empty constituent slot. On this view, held by adherents to more radical pragmatics, there is no linguistic entity here at all. The material needed to complete the
proposition is supplied on purely pragmatic grounds. The RT approach, for example, aims to account for the process of recovering unarticulated material as a
process of “free enrichment”, i.e. the enriching process is “free” from linguistic
control (Carston 2004). This does not mean, though, that the enrichment does
not contribute to the truth conditions of the utterance. From a RT point of view,
pragmatic processes are not only involved in computing implicatures in a Gricean
sense, they are necessary ingredients in working out the truth conditions of utterances.
 ?
The enrichment processes taking place in the interpretation of (1) would count
as “explicatures” in the RT framework. An explicature is “an ostensively communicated assumption which is inferentially developed from one of the incomplete
conceptual representations (logical forms) encoded by the utterance” (Carston
2002, 377).
Explicatures are thus distinguished from implicatures in that the former contribute to the truth-conditions of the utterance, even though the meaning is inferentially arrived at and not linguistically encoded. The term “explicature” has
led to some confusion in the linguistic debate, since what is inferred information
is based on what is “implicit” and not “explicit”, and “what is explicit” is confused with “what is explicated”. Bach (1994) uses the term “impliciture” for a
very similar concept:
OSLa volume 1(1), 2009
  
Grice 1989
Sperber &
Carston RT
Bach 1994
“What is said”
“What is said”
“What is said”
“The coded”
. 1: Overview of pragmatics according to Levinson
In implicature one says and communicates one thing and thereby
communicates something else in addition. Impliciture, however, is
a matter of saying something but communicating something else instead, something closely related to what is said [….]. Part of what is
communicated is only implicit in what is explicitly expressed, either
because the utterance is semantically underdetermined and completion is required, or because what is being communicated is an expanded version of the proposition expressed. (Bach 1994, 126)
Bach includes ambiguity resolution and anaphor resolution as part of what is said,
and leaves impliciture to involve expansion and specification of indeterminacies,
while RT includes all of these phenomena under explicature. In his paper “Impliciture vs. explicature: What’s the difference?” Bach (2006) compares his own
concept of impliciture with the RT explicature, stating:
We agree that speakers can communicate things that are neither fully determined by the semantics of the uttered sentence nor merely conversationally implicated. We may not agree on what this involves or how it is accomplished, and
we may disagree on what constitutes sentence semantics, but clearly we agree
that there is an intermediate phenomenon. Bach (2006)
Levinson (2000, 195) gives a nice overview of the different positions in the
pragmatics literature, relevant aspects of which are in table 1. It is not our intention here to make a case for one or the other, but the terminological discussion
referred to is another indication that the topic of this book is highly motivated
in the current debate: The explicit/implicit dimension of linguistic analysis is far
from solved and the present volume contributes to furthering the discussion.
In the present volume F argues that there are serious problems with
Robyn Carston’s (1988) attempt to clarify the distinction between explicatures
and implicatures, thus keeping the debate alive and bringing it a step forward by
arguing for why the RT approach does not hinge on this distinction. P and
OSLa volume 1(1), 2009
K dismiss the RT approach to the interpretation of implicit arguments like
the missing object in (1) above, and takes a syntactic view. They demonstrate
that verbs differ not only as to whether they allow the omission of arguments,
but that the conditions for argument omission differ across languages, pointing
against free enrichment. Instead they propose a two-stage model for the licensing
of implicit arguments.
Both contributions suggest novel perspectives on the interaction between syntax, semantics and pragmatics in the account of linguistically based meaning.
I takes a new perspective on the analysis of non-sentential utterances. Such
sentence fragments have often been analyzed as elliptical sentences in the sense
of having an underlying sentence structure. Ido argues for a model of such fragments that makes do with a minimum of syntax, yet assigns them full propositional content.
“ ”:     
   ?
The explicit/implicit dimension in discourse structure is further complicated by
the fact that certain explicit markers of information do not seem to contribute to
the truth conditions of the sentences they appear in, but are explicit means to express non-truth conditional meaning: “While ‘what is said’ is largely determined
by conventional (encoded) meaning, it is not the case that all encoded meaning
goes into determining ‘what is said’ .” (Carston 2002, 107)
What Carston has in mind are  . The cases Grice
(1975) refers to are connectives such as moreover, but and therefore, later (Grice
1989) also includes other non-truth-conditional discourse connecting items such
as on the other hand and so. Although they are clearly triggers for implicit information, connectives are nonetheless explicit markers of cohesive ties in discourse,
and have become a central issue in the semantics/pragmatics debate, not least
due to Diane Blakemore’s much renowned discussion of them in the RT literature (Blakemore 1987), in which they are taken to denote “procedural” meaning,
defined as meaning which “does not contribute a concept but rather provides a
constraint on, or indication of, the way some aspect of pragmatic inference should
proceed” (Carston 2002, 379). The “procedural meaning” is investigated from different theoretical stands as applied to different connectives and discourse particles. One problem is that although the claim that connectives constrain interpretation procedures seems well founded, it also seems that the contexts may influence the interpretation of the connective. This influence may in fact be so strong
that it pushes the interpretation towards or even beyond the (fuzzy) borders of
the semantic domain covered by the connective (Fabricius-Hansen and Behrens
2001; Fabricius-Hansen 2005). The complicated interplay between what connectives allow and their contexts demand, although already given ample attention in
OSLa volume 1(1), 2009
  
the literature, is still a highly relevant issue in the ongoing discussion of the interactions between different levels of discourse interpretation; see Karagjosova’s
contribution to the present volume.
   
Some of the problems of making explicit the ways in which discourse meanings
can be accounted for are demonstrated by the discussion of (1) above: We have to
specify implicit arguments, anaphora and the denotation of lexical items in context, and we have to untangle the ways in which the implicit is made part of the
explicit (truth-conditional) meaning of the discourse. In view of the discussion of
connectives above, it should also be clear that we still have to work out the particular function(s) of non-truth-conditional expressions. If we can specify the way
the triggers work, we have important keys to an understanding of the division of
labor between (compositional) semantic and pragmatic factors in discourse interpretation.
An important goal of linguistic research is to explain how the meaning of a
discourse becomes more than a sum of the meaning of its overt elements, yet is
constrained by its overt elements and (implicit) general pragmatic principles on
interpretation. One of the great challenges is to operationalize this fine-grained
interaction between explicit and implicit means of bringing out the meaning of
utterances in context (written or spoken). Central to such operationalization is,
no doubt, the alternative semantic view of focus.
    
Authors seem to understand “information structure” differently – some take the
term to refer to sentence internal / and / structure, partly delimited by stress/accent patterns.3 Focus is a clue to identifying
contextual information that constrains interpretation – and thus to establishing
textual coherence. The concept has been made operational in the analysis of discourse through Mats Rooth’s very influential alternative semantics (Rooth 1992).
Alternative semantic analyses of focus have helped us see how not immediately
observable units of discourse operate in the structuring of textual information. Focus contributes to specifying (presupposed) information that is added to the explicitly stated information in inferential textual updating. This theory is convincingly argued for by K in the present volume. She demonstrates how
it is the syntactic position and accent pattern of the lexical item (conjunct adverb) doch that determine the factors that enter into its different interpretations,
and how accentuation triggers the focus alternatives that are needed to relate the
meaning of the lexical item precisely to the context it appears in.
OSLa volume 1(1), 2009
See Krifka (2007) for a useful overview.
The RT notion of “procedural meaning” would apply to the meaning of German doch. The analysis in the present volume is an alternative analysis which
demonstrates how a pragmatic inference procedure can be formalized if focus
presupposition is taken seriously in the analysis of discourse connecting adverbs
and particles.
The interpretive effect of focus presupposition is not limited to procedural
meaning. This is demonstrated in S’s paper, which takes up what has traditionally been studied as the “double reflexive”: the reflexive pronoun SIG (sich,
seg,… etc, depending on the language) plus the intensifier . Two interesting
observations with respect to their use are accounted for. On the one hand the 
is separated from SIG and is assigned the meaning of an intensifier – the identity function is applied to type e entitites. On the other, the verbs that require
-intensification for reflexivization are verbs like admire, accompany, give etc.
– verbs denoting “other-directedness” in the sense of activities conventionally
directed at others as opposed to conventionally self-directed activities like dress
and undress, wash etc. This lexical-pragmatic distinction is crucial to account for
the distribution of , but it is not sufficient to explain how it works. Only with
an analysis of the intensifier as a trigger of focus alternatives in the sense of Rooth
(1992) does the  provide its own contrast, which is matched in the otherdirectedness of the predicate. -intensification is used to make the contrast to
the implicit other-directedness explicit, and is needed for that very purpose. In
accordance with Bidirectional Optimality Theory (Blutner and Zeevat 2003), the
self-directedness must be marked explicitly due to the contrasting directedness
of the predicate. Enriching his focus-theoretic approach with the distinction between   and  made by Jacobs (1999), Sæbø
is also able to present a plausible explanation for the distributional differences of
-intensification across languages.
Focus and contrast also come into play in E’s account of the German
lexical itemeigentlich, which can be used as an attributive adjective and as an adverb. The expression has a cognate egentlig in Norwegian but does not have a cognate in French or English; real, really, actually being the closer English correspondences in many contexts. Eckardt proposes a compositional semantic-pragmatic
analysis based on the assumption that the adjective eigentlich contrasts the “nominal” or “named” content of a concept C with a contextually given notion of “phenomenological” evidence for C-hood, while adverbial eigentlich raises the meaning
of the adjective to the propositional level. She demonstrates that stressed adverbial eigentlich functions as a contrastive topic, i.e. it associates with a second focus in the sense of Büring (2003). Thus it gives rise to alternative propositions and
refers to a contrasting proposition q in context, conveying that q would lead one to
expect one of these (false) alternatives rather than the actual facts. In the last section of her paper, Eckardt discusses derived particle uses of eigentlich that deviate
OSLa volume 1(1), 2009
  
from the canonical adverbial eigentlich in syntax, focus sensitivity and meaning.
     –
   ?
In the invitation to our conference we used the expression “information structure” in a very wide sense – and this wide sense seems to have been captured
by the researchers who responded to our call for papers. Information structure
in a wide sense refers to the various mechanisms that are involved in distributing information and creating coherence in text units larger than the sentence.
Thus in this wide sense it entails information structure in a narrow sense – i.e.
constituted by the sentential features  and  – but it also encompasses
 , which includes coherence phenomena that are triggered by
the mere assumption that sentences in text are not put together at random. The
assumption of coherence is the basis for inferring discourse relations between the
segments of a text. Such relations may be purely rhetorical, as in Rhetorical Structure Theory (RST), or they may be at once rhetorical and semantic. Asher (1993)
distinguishes between rhetorical relations and coherence relations. Rhetorical
relations are relations between constituents of text. They segment a discourse
on the basis of the rhetorical function of particular propositions in relation to
propositions already established in the discourse, and typically contribute meaningfully to the development of the discourse. Coherence relations, on the other
hand, are relations between eventualities within discourse constituents. Unlike
rhetorical relations they directly contribute to the truth conditional content of
the constituents themselves (Asher 1993, 265).4
The various terms used in the literature to designate the relations that hold
between propositions in general reflect the uncertainty that exists among linguists as to their ontological status: discourse/semantic/logical/rhetorical relations.
Some theoretical approaches operate with a rich ontology of relations, others
restrict discourse relations to coordinating and subordinating relations without
further refinement. Topic and focus are undoubtedly involved in the establishment of discourse rhetorical structure, and this is one of the reasons why it may be
useful to operate with a concept of information or discourse structure in a wide
sense. Prosodic features marking sentence focus  with other features
in the processing of the communicated information, ranging from anaphoricity
through syntax, structural parallelism and contrast, discourse structural coordinating and subordinating structures to prosodic discourse features such as pitch
range (expanded and compressed, cf. Jasinskaja et al 2004) and pauses, in addition
to the unavoidable and very central world knowledge based inferences.
OSLa volume 1(1), 2009
In Asher and Lascarides (2003) ‘rhetorical relation’ is a more general cover term.
A very interesting aspect of some of the contributions in this volume is the
demonstration of how focus syntax and prosody contribute to distinguishing between different interpretations of context sensitive lexical items/expressions.
This is documented very clearly for the various interpretations of eigentlich in
Eckhardt’s article, as well as for German doch in Karagjosova’s paper (see above).
The discourse structural properties are shown to have an important impact on
disambiguation and determination of meaning, and these papers thus reconfirm
the importance of uncovering and categorizing the linguistic clues which contribute to constraining the contextual factors involved in the interpretation of
meaning. World knowledge is no doubt always relevant for the interpretation
of linguistic utterances, but the linguistic system is here again demonstrated to
contribute subtle, but extremely valuable clues.
What, then, is the basis for determining the discourse-rhetorical relations?
     
Recent attempts at determining the linguistic basis for discourse structure suggest that discourse coordination may be tested on the basis of syntactic coordination (Asher and Vieu 2005). One argument in favor of a correlation between the
two types of coordination/subordination is that typical cases of discourse coordination, e.g. Continuation and Narration (Asher and Vieu 2005), occur in syntactic conjunction, while a prototypical subordinating relation like Elaboration
seems blocked in coordinate syntax– at least in English (Carston 2002; Blakemore
and Carston 2005). A causal relation, which may be expressed syntactically by
subordination, is also a very likely interpretation of VP conjunction, hence a coordinating discourse relation according to the syntactic test. However, the fact
that discourse- coordinating relations may be expressed by syntactically subordinate clauses complicates the attempted correlation between discourse structure and syntax (cf. Fabricius-Hansen and Ramm 2008). Besides, there is reason
to believe that the discourse-relational potential of syntactic coordination varies
somewhat across languages (Ramm to appear). Within the rich ontologies of discourse relations the coordinate/subordinate distinction is also blurred by the fact
that one relation may be both coordinating and subordinating (Asher and Vieu
2005; Behrens and Fabricius-Hansen to appear).
Asher and Vieu (2005) define their distinction between discourse subordination and coordination in terms of Segmented Discourse Representation Theory
(SDRT), which originated with Asher (1993) and was further developed by Asher
and Lascarides (2003). Along with other pragmatic theories of discourse, it assumes a non-monotonic scheme by which the relations may be inferred, yet only
by default. It has turned out to be very hard to pin down not only the precise definition of a number of relations, but also the particular discourse properties of the
connectives used in the languages to express them. The French connective alors,
OSLa volume 1(1), 2009
  
which is the topic of B  .’s contribution, is a case in point: it is a temporal
connective (semantically), but it is also understood to contribute some discourse
structural meaning of “weak” causation between the propositions it connects. Is
this meaning truth-conditional or a matter of pragmatic, non-monotonic implicatures? Furthermore, is it part of the meaning of the connective itself, or an
element of meaning which results from combining the content of the sentences
it is used to connect?
Questions of this kind are central not only for a SDRT analysis of connectives
but for any approach to the semantics/pragmatics of connectives and conjunctions, and although the questions are not new, they are still highly relevant in
the present linguistic debate. A number of analyses of connectives have been
published during the last fifteen years, not least to give support to different theoretical approaches to connecting discourse. SDRT is central here, and competes
to some extent with the Relevance Theoretic approach, but while SDRT – as an
extension of DRT – takes the linguistic input seriously for all it is worth, Relevance Theory attaches more weight to extra-linguistic contextual knowledge in
the search for optimal relevance. The inference mechanism in the RT view is not
based on lexical structure, formal meaning postulates and default axioms, but
on the listener’s online processing of her/his knowledge of the pragmatically inferred referential meaning of the utterances.
     
The status of the discourse relations is still somewhat blurred. One interesting
issue is whether discourse relations are clearly defined at every step, or whether
there may be more than one relation at any step in the development of a text.
Baldridge et al. (2007) claim that discourse structure must be represented as graph
structures rather than as simple tree structures for the very reason that multiple relations may obtain for a discourse unit. This is not a topic taken up at any
length in the present volume, but see Bras et al. for further references; see also
Stede (2008) and relevant contributions in Benz and Kühnlein (2008). W 
P also demonstrate how discourse units are not always successively linked
linearly, i.e. to the previous discourse unit, and make it clear that discourse relations hold at different discourse levels. Their article demonstrates very clearly
how empirical data challenge existing theories of discourse structure.
Another very important complication in uncovering the nature of the relations, which also makes the discussion of discourse relations highly relevant for
the present volume, is that very often they are not marked linguistically as such,
but are left implicit. From the Gricean point of view that semantics and pragmatics are separated according to what is said and what is implicated, unexpressed
discourse relations would clearly not make up part of the semantics of language.
Yet recent pragmatic theories are more open to the possibility that what is not
OSLa volume 1(1), 2009
said may also make up part of the semantics of the utterances in the sense of
contributing to the truth conditions of the utterances, as discussed in the previous paragraphs. The way this may come about is a central issue in the ongoing linguistic debate, which is also evident from the focus of several of the papers in this volume. The fact that discourse relations are basically unexpressed
supports Z’s fundamentally novel theory of pragmatic principles that constrain interpretations. His article in the present volume breaks important ground
for revised thinking around discourse structural analysis. One ground-breaking
consequence of his theory is for example that if general pragmatic principles are
followed in the discourse, there is no need to mark them as such. The basic function of conjunctions and other connectives, therefore, is to mark deviations from
general pragmatic constraints on interpretation. His contribution is a bold attempt at grounding all of pragmatics in a wide sense, including presuppositions,
in a handful of ranked constraints, viewing explanations of linguistic events as a
special case of a general notion of explanation.
The articles in the present volume demonstrate not only that explaining connections between propositions in text is still a “hot” issue, but also that views
on how such connections come about, are very theory-sensitive. The order in
which we have presented the articles reflects a move from general relevance theoretic thinking through intricate compositionally based semantic and pragmatic
reasoning and bottom-up approaches, to a completely novel pragmatic theory of
how interpretations are constrained. It is to be hoped that the present collection
will inspire the reader’s reflection on language interpretation.
Asher, N. 1993. Reference to Abstract Objects. Dordrecht: Kluwer.
Asher, N. and Lascarides, A. 2003. Logics of Conversation. Cambridge: Cambridge
University Press.
Asher, N. and Vieu, L. 2005. Subordinating and coordinating discourse relations.
Lingua 115, 591–610.
Bach, K. 1994. Conversational impliciture. Mind and Language 9, 124–162.
Bach, K. 2006. Impliciture vs explicature:
What’s the difference?, for the Granada workshop on “Explicit Communication”, in honour of Robyn Carston.
Baldridge, J., Asher, N. and Hunter, J. 2007. Annotation for and Robust Parsing
of Discourse Structure on Unrestricted Texts. Zeitschrift for Sprachwissenschaft
26(2), 213–239.
OSLa volume 1(1), 2009
  
Behrens, B. and Fabricius-Hansen, C. to appear. The relation Accompanying Circumstance across languages. Conflict between linguistic expression and discourse subordination? In K.P. Turner (ed.), Contrasting Meaning in Languages of
the East and West., Bern: Peter Lang.
Behrens, B., Fabricius-Hansen, C., Hasselgård, H. and Johansson, S. (eds.). 2007.
Information Structuring Resources in Contrast, volume 7-2 of Languages in Contrast.
Benz, A. and Kühnlein, P. 2008. Constraints in Discourse. Amsterdam: Benjamins.
Blakemore, D. 1987. Semantic Constraints on Relevance. Oxford: Blackwell.
Blakemore, D. and Carston, R. 2005. Introduction to coordination: syntax, semantics and pragmatics. Lingua 115, 353–358.
Blutner, R. and Zeevat, H. 2003. Editors’ introduction: pragmatics in Optimality Theory. In R. Blutner and H. Zeevat (eds.), Optimality Theory and Pragmatics,
pages 1–24, New York: Palgrave.
Büring, D. 2003. On D-trees, Beans, and B-accents. Linguistics and Philosophy 26(5).
Carston, R. 1988. Implicature, explicature, and truth-theoretic semantics. In R. M.
Kempson (ed.), Mental Representations: The Interface between Language and Reality,
pages 155–182, Cambridge: Cambridge University Press.
Carston, R. 2002. Thoughts and Utterances. Oxford: Blackwell.
Carston, R. 2004. Explicature and semantics. In S. Davis and B. S. Gillon (eds.),
Semantics: A Reader, pages 817–845, Oxford: Oxford University Press.
Fabricius-Hansen, C. 2005. Elusive connectives. A case study on the explicitness of
discourse coherence. Linguistics 43, 17–48.
Fabricius-Hansen, C. and Behrens, B. 2001. Elaboration and related discourse relations viewed from an interlingual view. SPRIKreport 13.
Fabricius-Hansen, C. and Ramm, W. 2008. Editors’ introduction: Subordination and coordination from different perspectives. In C. Fabricius-Hansen and
W. Ramm (eds.), ‘Subordination’ versus ‘Coordination’ in Sentence and Text. A crosslinguistic perspective, pages 1–30, Amsterdam: Benjamins.
Grice, P. 1975. Logic and conversation. In P. Cole and J. Morgan (eds.), Syntax and
Semantics, volume 3, pages 41–58, New York: Academic Press, speech acts edition.
OSLa volume 1(1), 2009
Grice, P. 1989. Studies in the Way of Words. Cambridge, Mass.: Harvard University
Jacobs, J. 1999. Informational autonomy. In P. Bosch and R. van der Sandt (eds.), Focus: Linguistic, Cognitive, and Computational Perspectives, pages 56–81, Cambridge:
Cambridge University Press.
Krifka, M. 2007. Basic Notions of Information Structure. In C. Fery and M. Krifka
(eds.), Interdisciplinary Studies of Information Structure, volume 6, Potsdam: Potsdam Universitätsverlag.
Levinson, S. C. 2000. Presumptive Meanings: The Theory of Generalized Conversational
Implicature. Cambridge MA: The MIT Press.
Ramm, W. to appear. Discourse-structural salience from a cross-linguistic perspective: Coordination and its contribution to discourse (structure). In Grabski
et al. M. (ed.), Salience. Multidisciplinary Perspectives on its Function in Discourse,
Trends in Linguistics. Studies and Monographs, Berlin/New York: de Gruyter.
Rooth, M. 1992. A theory of focus interpretation. Natural Language Semantics 1,
Stede, M. 2008. RST revisited: Disentangling nuclearity. In C. Fabricius-Hansen
and W. Ramm (eds.), Subordination’ versus ‘Coordination’ in Sentence and Text A
cross-linguistic perspective, pages 33–58, Amsterdam: Benjamins.
  
Bergljot Behrens
University of Oslo
Dept. of Literature, Area Studies and European Languages
P.b. 1003, Blindern
N-0315 Oslo
[email protected]
Cathrine Fabricius-Hansen
University of Oslo
Dept. of Literature, Area Studies and European Languages
P.b. 1003, Blindern
N-0315 Oslo
[email protected]
OSLa volume 1(1), 2009
OSLa volume 1, 2009
Behrens & Fabricius-Hansen (eds.) Structuring information in discourse:
the explicit/implicit dimension, Oslo Studies in Language 1(1), 2009. 17-32. (ISSN 1890-9639)
    
  
Norwegian University of Science and Technology
Robyn Carston has proposed that, within Relevance Theory the proposition
expressed by an utterance U (its explicature) must play a communicative
role that is independent of the role played by any implicature of U. Thus no
implicature can entail the explicature of U, and whenever it looks as if that
sort of situation obtains, the presumed implicature should be redefined as
the explicature. This paper shows that the assumption that an explicature
may not be entailed by an implicature cannot be maintained, but the fact
that explicatures and implicatures are not always functionally independent
of one another is argued not to be a problem for Relevance Theory.
         
There is a general agreement that context and context-sensitive pragmatic processes contribute to the truth conditions of explicitly communicated propositions, but opinions diverge as soon as one starts discussing the extent to which
pragmatically derived information should be allowed to intrude into the proposition expressed by an utterance. Adherents of relevance theory (Sperber and
Wilson 1986, 1995; Carston 2002) claim that relevance-driven pragmatic inference
plays just as important a role in the recovery of the proposition expressed as in
the derivation of implicatures, and that the set of truth conditions encoded by a
given sentence hardly ever exhausts the set of truth conditions for the proposition that a speaker communicates. Carston says,
If all pragmatically derived elements are treated as implicatures
we are left with no candidate for what is said, or the explicit utterance content, other than the logical form of the linguistic expression
used, which is standardly subpropositional. But we do not communicate logical forms (though we do communicate via logical forms), and
 
we do not retain logical forms in memory; we communicate and remember assumptions or thoughts or opinions, which are fully propositional. So a distinction has to be made between pragmatic inference
that contributes to the recovery of the explicitly communicated content and pragmatic inference that eventuates in implicated assumptions. (Carston 2002, 107)
The view expressed in the final sentence of this quotation is expressed even more
strongly, and with some added implications, by Recanati (1989, 1993, 2004), because Recanati makes a distinction between ‘primary pragmatic processes’, which
are pre-propositional and outside the conscious awareness of the interpreter, and
‘secondary pragmatic processes’, which are said to be post-propositional, and
which the interpreter is consciously aware of. The former are local processes that
contribute to the hearer’s understanding of what is said and are not considered
to be inferences in the strict sense; the latter are believed to be true inferences
of a global sort. What Recanati labels the Availability Principle is based on the
assumption that two qualitatively different types of pragmatic process – primary
and secondary – are involved in utterance interpretation. Processing that is required for the determination of the truth conditions of the proposition expressed
(‘what is said’) happens at a sub-conscious level. The output of such processes
is consciously available but the processes themselves are not. True inferential
processes on the other hand are consciously available and are the ones involved
in the derivation of conversational implicatures, at least the particularized ones
(PCI’s) whose existence depends on inferences that are not of a default sort.
Recanati’s position on the difference between two kinds of pragmatic process
is generally rejected by leading relevance theorists, who hold that all pragmatic
processes are inferential processes of an unreflective nature, and that the processing of explicatures and implicatures happens in parallel, so that the latter
kind of inferential process does not depend on the hearer’s mental representation
of a pragmatically fully developed proposition, an explicature. Recanati himself
emphasizes in his most recent book that, even if what is said is logically prior to
the working out of implicatures, it is not temporally prior in the sense that recovery of what is said takes place before any implicatures can be computed (Recanati
2004, 47), but he still finds reasons to maintain his differentiation between primary and secondary pragmatic processes.
Carston (1988) is a very influential paper that deals with, and defends, the
functional autonomy of explicatures and implicatures in the pragmatic processing of utterances. Let us look at one of her examples Carston (1988, 155).
OSLa volume 1(1), 2009
How is Jane feeling after her first year at university?
She didn’t get enough units and can’t continue.
  
Carston notes that there are certain words in B’s answer that are in need of disambiguation. This is true of the verb get and the noun unit, she says – though
today she would presumably prefer to say that both get and unit are conceptually
underspecified words that are subject to ad hoc development in context, and that
no lexical ambiguity is involved here. No one would deny that on the implicit
side of the pragmatic analysis of B’s answer is the derived assumption that Jane
regrets the situation she finds herself in, but that implicature is based on an interpretation of the phrase can’t continue as ‘cannot continue with her university
studies’ and the phrase enough units as ‘enough university course units to qualify
for admission to second year study’, information which is not encoded but which
is represented in the hearer’s LoT (Language of Thought) as a result of pragmatic
enrichment of the encoded logical form. Also, the predicate get is regularly in
need of lexical narrowing, a pragmatic process that will be sensitive to the semantics of especially the direct object argument of this verb and has nothing to
do with implicature derivation.
What general criteria do we possess, which enable us to decide whether the
pragmatic developments of the grammatical phrases enough units and can’t continue in (1) belong to the proposition expressed, the output of mental processes of
‘free enrichment’ like narrowing or loosening of the encoded meaning, or are instead conceptual constituents of an implicature? Back in 1988 Carston suggested
a criterion of functional independence of the explicature of an utterance and any
implicatures of the same utterance as part of the answer to this question: the
Functional Independence Principle. Adopting this principle means that any alleged implicature that is seen to subsume all contextual effects that could be derived from the proposition expressed by an utterance cannot maintain its status
as implicature; rather it should be redefined as the explicature, a hybrid mental
representation made up of conceptual structure provided via the encoded logical
form of the utterance on the one hand and context-sensitive inferences on the
other.1 However, Carston hastens to add that “while it is instructive to consider
such criteria, they might well be seen as rather superficial, descriptive principles,
if not ad hoc”, because, she says, “they follow from a single principle directing utterance interpretation, the principle of relevance, which is itself embedded in a
general theory of human cognition and communication, the relevance theory of
Today there are certain indications (e.g. in Hall 2008) that the ‘hybrid representation’ hypothesis is less
solidly established in relevance-theoretic circles than it was just a few years ago. Is decoding of the
linguistically encoded meaning of a sentence a mental process that operates directly on a logical form
provided by the language module itself, or does all pragmatic enrichment take place in LoT because the
so-called logical form is itself a conceptual representation in LoT, distinct from any natural language
representation of decoded semantic structure (which is probably non-existent)? If the latter position is
adopted, then our recognition and understanding of the explicature of an utterance will not depend on
the output of the grammar on the one hand and inference on the other hand, because even ‘linguistic
decoding’ will then be defined as a non-linguistic thought process and as such it is hardly a process
independent of contextual input.
OSLa volume 1(1), 2009
 
Sperber & Wilson (1986)” (Carston 1988, 156). This toning down of the significance of the Functional Independence Principle was later seen to be followed up
by a stronger concession on her part (Carston 2002, 191),
However, it is not clear to me that ‘functional independence’ is
worth any kind of vigorous defence; it was in fact intended as only a
useful heuristic and should probably never have been elevated by the
label ‘principle’ at all. I was (and am still) of the view that the communicative principle of relevance itself or, more particularly, the comprehension strategy2 that follows from it, effects a sorting of pragmatic inferences into contributions to the proposition expressed (explicature) and implicatures, and so subsumes whatever correct predictions ‘functional independence’ might make. (Carston 2002, 191)
As an example of what she means by saying that the explicature of a given utterance must play a communicative role that is independent of the role of any
implicature derived via pragmatic processing, and vice versa, Carston considered
the exchange between A and B in (2) in her 1988 article.
A Have you read Susan’s book?
B I don’t read autobiographies.
implicated premise: Susan’s book is an autobiography.
implicated conclusion: B has not read Susan’s book.
(Carston 1988, 157)
In (2) the truth conditions pertaining to the implicatures ‘Susan’s book is an autobiography’ and ‘B has not read Susan’s book’ are independent of the truth conditions of the explicature ‘B does not read autobiographies’. There is no overlap in content between the explicature and the implicatures, which function independently of the explicature as premise and conclusion in a process of nondemonstrative inference (cf. Sperber and Wilson 1986).
Suppose the functional independence criterion were found not to hold water,
so that even its usefulness as a heuristic might be questioned. Would a disclosure
of the empirical shakiness of the criterion have consequences for the relevancetheoretic view of the semantics/pragmatics distinction and the distinction between explicit and implicit communication? I am going to argue that Carston’s
Functional Independence Principle should not be accorded any role in a cognitively based theory of utterance interpretation like relevance theory. There is
rather strong empirical evidence that this principle is not tenable; fortunately,
the viability of relevance theory does not in any way hinge on it.
OSLa volume 1(1), 2009
The relevance-theoretic comprehension strategy enjoins addressees to construct interpretation in order
of accessibility and to stop when their expectation of relevance is satisfied.
  
     
Carston (2002, 137-41,189-91) has observed that there can be no incompatibility
between the fact that an explicated proposition p entails a proposition q and the
assumption that the entailed proposition q is an implicature derived through the
hearer’s processing of the explicature p in a specific context. The existence of situations where an implicature of an utterance is an entailment of the explicature
of the same utterance would appear to be a counterexample to the Functional
Independence Principle. If the implicature is entailed by the explicature of the
utterance, the two communicated assumptions cannot both be autonomous, they
are not independent of one another.
Carston did not draw the conclusion that the existence of implicatures that
are entailed by the proposition expressed – as in (4) B below – is bad news for
the relevance-theoretic view of the contrast between explicated and implicated
propositions. In contrast to (3), the corresponding answer in (4), adapted from
Carston (2002, 190), implicates that B did buy vegetables, because B bought a
squash and an eggplant. In (4) the implicated conclusion is entailed by B’s explicated premise.
A Did you buy any vegetables?
B I bought some apples and pears.
implicature: B did not buy any vegetables
A Did you buy any vegetables?
B I bought a squash and an eggplant.
implicature: B bought some vegetables
Carston concludes, “In my view, the concept of ‘entailment’ and the concept of
‘implicature’ belong to different explanatory levels or different sorts of theory,
the one a static semantic theory, the other a cognitive processing pragmatic theory” (Carston 2002, 141).3
     
I am now going to show that we also have to allow for the existence of situations
in which an implicature entails the explicature of an utterance. Carston’s independence principle was meant to give pragmatists a criterion that would enable
them to tell explicatures and implicatures apart in cases of doubt. Her argument
was that, if the meaning of the explicature is fully included in the meaning of
Burton-Roberts (2005) strongly disagrees with Carston that an entailment of the explicature of an utterance can be implicated. When you explicate a proposition p, his argument goes, then you explicate
the truth-conditional content of p, which includes all entailments of p. Hence if an entailment of p is
implicated, then p is implicated. To deny this, he says, would be to construe ‘entailment’ in a novel,
idiosyncratic and unprecedented way.
OSLa volume 1(1), 2009
 
an implicature derived on the basis of the explicature and a set of contextual assumptions, then the explicature will be redundant, as no contextual effect can be
derived from the explicature which is not at the same time derivable from the
entailing implicature. The thought represented by the explicature is duplicated
in a truth-conditionally stronger implicature.
The Functional Independence Principle supports Carston’s own proposal that
all Gricean generalized conversational implicatures be eliminated, including the
reasoning from ‘p and q’ to ‘p and then q’ (‘conjunction buttressing’) and the reasoning from ‘if p, then q’ to ‘if and only if p, then q’ (‘conditional perfection’),
where the stronger readings are said to be due to generalized conversational
implicatures, Levinson’s GCI’s (Levinson 2000), which are triggered by the presence of particular linguistic elements. For relevance theorists all legitimate implicatures are of the particularized sort (PCI’s), the outcome of ‘one-off inferences’. The claim that GCI’s should be dispensed with is based on the fact that relevance theorists embrace a view of explicit communication as much more contextdependent and inference-driven than what had been assumed by Grice and the
neo-Griceans. An explicature is defined as an ostensively communicated assumption which is inferentially developed from a truth-conditionally incomplete encoded logical form (Sperber and Wilson 1986; Carston 2002). All standard examples of GCI, which a neo-Gricean like Levinson (2000) regards as a preferred or
normal interpretation based on default inference, are redefined as explicatures
in the framework of relevance theory. Like any other explicature, the ones that
Grice defined as GCI’s are conceived by relevance theorists as the result of a union
of a grammar-dependent logical form and context-based inference supported by
the communicative principle of relevance.4, 5
The reasoning that underpinned Grice’s argument that a coordinating connective like English and gives rise to what he called a generalized conversational
implicature was prompted by his intention to prove that natural language operators like conjunction and disjunction connectives and conditional connectives
work exactly like the corresponding operators of propositional logic. The fact
that we often understand a speaker’s use of the connective and as if she had actually said and then is due to factors that are extraneous to whatever information the
language code can give us. Such inferences about what the communicator means
by what she says were claimed by Grice to be due to principles of conversation,
his Co-operative Principle and his four categories of conversational maxims (Grice
OSLa volume 1(1), 2009
The communicative principle of relevance says that, “Every act of ostensive communication communicates a presumption of its own optimal relevance” (Sperber and Wilson 1995, 260)
Although it will be no major point in the present paper, I confess that I am no longer willing to take for
granted the notion of a grammatically determined semantic representation which is a subpropositional
conceptual structure. It is conceivable that logical properties are not found in natural-language representations at all and that only information represented in LoT has logical properties. See Burton-Roberts
(2007), Amfo et al. (2007), Fretheim (2008) – as well as footnote 1 of this paper.
  
1975, 1989). A given English conjunction of clauses or verb phrases conforming
to the general formula ‘p and q’ may implicate ‘p and then q’ in one context, ‘p
and therefore q’ in a different context, etc. More strongly than any other publication, Carston’s 1988 paper offered a very clearly expressed alternative to the
standard implicature analysis of the temporal and causal implications of the use
of and-conjunction.
If it turns out that we have to allow for the existence of situations where an
implicature entails the proposition expressed, the usefulness of the Functional
Independence Principle as a heuristic will be seriously diminished. On the other
hand, empirical evidence that might cause the demise of the Functional Independence Principle has been scarce. Having addressed an argument that Recanati
(1989) presented against her Functional Independence Principle, Carston (2002,
191) retorted, “A more compelling counterexample to the principle would involve
a communicated assumption which (…) is clearly an implicature, but which is an
implicated premise (rather than an implicated conclusion) and entails what is said; I
have not come across such a case.” (her emphasis). She did not tell the reader why
she feels an implicated premise that entails what is said would be more damaging
to the RT view of the saying-implicating distinction than an implicated conclusion
that entails what is said.
I have chosen not to cite Recanati’s fairly complicated counterexample in the
present paper. However, Burton-Roberts (2005), which is a review of Carston
(2002), presents data that may be a problem for Carston’s proposal about the functional autonomy of explicatures and implicatures. The kind of example on which
his argument rests is given in (5).6
There’s no milk.
The milkman’s ill.
B’s answer implicates that there is no milk because the milkman is ill. In other
words, what B implicates subsumes what B says, or explicates, which is no more
than that the milkman is ill. The implicature that there is no milk because the
milkman is ill entails the explicature of B’s utterance in (5). On Carston’s account the thought ‘There is no milk in the house because the milkman is ill’ should
preferably be explicated, not implicated, by B’s utterance of The milkman’s ill. However, the explicated assumption that he is ill is functionally independent of the
explicature of A’s utterance and independent of any inferred causal relation between B’s explicature and A’s explicature. I do not think anything would stop a
relevance theorist from saying that the logical form of B’s sentence can be developed into a more complex explicature that may be paraphrased as ‘There is no
I am grateful to Noel Burton-Roberts for pointing footnote 8 of his review article out to me
(Burton-Roberts 2005, 397).
OSLa volume 1(1), 2009
 
milk in the house because the milkman is ill’, and so it seems to me that BurtonRoberts’ milkman example might lead to a paradox for relevance theory, although
Burton-Roberts himself did not intend his counterargument to the Functional Independence Principle to be interpreted that way. He insists (p.c.) that the causal
relation between the premise p and the conclusion q in (5) cannot be explicated
even for Carston. One thing is clear: the assumption that there is no milk because
the milkman is ill cannot be both implicated and explicated by B’s utterance.
An anonymous reviewer offered me the interesting alternative of (5’) to the
data in (5). In the talk exchange of (5’) there are three participants: A, B, and C.
There’s no milk.
The milkman’s ill.
No, it’s because his company’s on strike.
What is C objecting to? The explicature of B’s utterance with a built-in causal
relation, or an implicature which communicates the causal relation between A’s
explicature concerning the lack of milk and B’s explicature concerning the milkman’s health condition? We do not have reliable tests that enable us to answer
this question in an adequate way, and from a communicative point of view the
decision does not really matter, because the relevance of C’s utterance certainly
does not depend on our professional view of how we should draw the line between free enrichment and implicature. It would appear that C, who is rejecting
Bs premise p and offering a new one, is treating the causal relation between p and
q as part of the explicit content of B’s utterance. However, there is no rule against
performing an act of denial that addresses a thought which is implicated by the
interlocutor’s previous utterance rather than explicated. The speaker’s communicated belief about the interlocutor’s belief that q was caused by p remains the
same on either analysis.
I am now going to present what I judge to be a better example of an implicature that entails the explicature of the utterance. Consider the minimally
different question-answer pairs in (6) and (7).7
What happened when that bug flew into your mouth?
I swallowed it.
What happened when that bug flew into your mouth?
I swallowed.
The linguistic intuition of native speakers of English is such that they understand
the proposition expressed by the utterance of (7) B differently than the proposition expressed by the utterance of (6) B. The version in (6) with an overt direct
OSLa volume 1(1), 2009
Anne Bezuidenhout’s paper ‘The Semantics/Pragmatics boundary’ (2005) aroused my interest in the verb
swallow and its lexical correspondents in other languages.
  
object argument is a direct way of saying that B swallowed the bug; the version
in (7) with no complement after the verb is an indirect way of conveying that
information. Non-native speakers of English generally feel the same about this
pair, because they draw on their native language competence and their own language shows exactly the same lexical difference between an intransitive verb and
a phonologically identical transitive verb that may both be glossed as ‘swallow’.
I am going to argue that there is no process of enrichment that takes you from
the encoded logical form of B’s answer in (7) to an explicature that equals the explicature communicated by B’s answer in (6). We understand the answer in (7)
to convey an explicit premise that opens for the implicated conclusion that the
bug was passed down into B’s stomach when B swallowed. The addressee of (7)
B is aware of what is said and is capable of working out the inferential connection between what is said (‘I swallowed’) and what is implicated by what is said
(‘My swallowing caused me to swallow the bug’). This is a ‘post-propositional’,
‘secondary’ pragmatic process in Recanati’s sense. It meets his availability condition, unlike the ‘primary’ pragmatic processes of mandatory saturation like the
resolution of the reference of the pronoun it in (6).
Bach (1994) introduced the notion of ‘impliciture’, standing for what is implicit in what is said, as what he considered to be a viable alternative to the relevance-theoretic ‘explicature’.8 He makes a terminological distinction between
‘conceptual incompleteness’ and ‘semantic incompleteness’. Conceptual completion, or what Bach calls ‘expansion’, is sometimes required to arrive at the ‘impliciture’ of an utterance. This is for him a case of free enrichment. Semantic
completion on the other hand is when the context supplies an argument for an
implicit argument role. In such cases Bach’s ‘impliciture’ would be arrived at via
saturation, or what Bach calls ‘completion’. When someone says I’ve finished, there
may be an implicit argument role there for the activity that was finished, and context will help in the recovery of the event which occupies the implicit argument
role. The same is presumably true of the null-complement of the verb notice in
William was very silent today.
Yes, I noticed.
A process of mandatory saturation is required for the hearer’s mental representation of what B noticed. Though there is no overt complement after the verb notice,
this is nevertheless a transitive verb whose argument structure includes an argument that refers to the object of the act of noticing. The verb finish is similarly
conceptually transitive even when it has no overt object argument.
What I’ve written in this paragraph is in large part due to personal communication with Anne Bezuidenhout.
OSLa volume 1(1), 2009
 
Could the verb swallow be analyzed in the same way as finish and notice? Does
the inferential processing of B’s answer in (7) involve mandatory saturation, Bach’s
‘completion’? Hardly, because the intransitive verb swallow encodes a different
concept than its transitive counterpart. There is no hidden or overt argument
referring to that which is swallowed when swallow is used as an intransitive verb.
Nor is it possible to let a sentence with the intransitive verb swallow undergo ‘free
enrichment’ that results in an explicature that represents an unmentioned affected object. If you wish to explicate the information that a particular solid object or liquid stuff was swallowed, you have to fill in the direct object slot, even
if the linguistic filler is just a pronoun that represents a contextually very salient
discourse entity, as in (6).
In order to convince you that intransitive and transitive uses of swallow are
semantically distinct, I suggest we consider the pair of (9)-(10).
Mark swallowed twice.
#Mark swallowed it/something twice.
An utterance of (9) causes us to infer that Mark swallowed without swallowing
anything. There may have been nothing in Mark’s mouth that could possibly be
swallowed. (9) is meaningful but (10) will not normally be, because when something has been swallowed it is hard to see how it could be swallowed once more.
The American Heritage Dictionary of the English Language lists two basic,
non-figurative meanings of the verb swallow, one of which corresponds to its meaning when it is a transitive verb and the other one to its meaning when it is used
intransitively. According to this dictionary, the transitive verb swallow means “to
cause (food, for example) to pass from the mouth via the throat and the esophagus into the stomach by muscular action”, and “ingest” is suggested as another
verb with that meaning. Under the same dictionary entry we also find an intransitive use of the verb swallow, whose meaning is defined as “to perform the act of
swallowing”. The Oxford Paperback Dictionary & Thesaurus similarly tells us that
the verb swallow has two meanings, one for its transitive and one for its intransitive use. The former is “to cause (food, drink, etc.) to pass down the throat”, and
the latter “to move the throat muscles as if doing this, especially through fear”,
where demonstrative this refers to the action denoted by the transitive verb. Finally, Shorter Oxford Dictionary distinguishes between the lexical meaning of the
transitive verb swallow, which is said to be “to take into the stomach through the
throat and gullet, as food or drink”, and the intransitive verb, whose meaning is
entered as “to perform the act of deglutition, as in an effort to suppress emotion”
(“deglutition” is a scientific word for “swallowing”). The muscular action involves
shutting of the epiglottis so that the entrance to the trachea is shut off, and the
intransitive verb swallow encodes no more information than the phrase perform an
OSLa volume 1(1), 2009
  
act of swallowing does. In contrast, the transitive verb swallow encodes the complex
deglutition concept, which necessarily includes the physical mechanism encoded
by intransitive swallow as an instrument.
Suppose we change A’s utterance in the above pair (6)-(7) slightly and replace
past tense swallowed by the present perfect tense form have swallowed.
So what happened to the bug that flew into your mouth?
I’ve swallowed it.
A So what happened to the bug that flew into your mouth?
B #I’ve swallowed.
If we accept the hypothesis that it is possible to let an utterance of the sentence
I’ve swallowed undergo free enrichment so that the contextually derived explicature will be something like ‘I have swallowed the bug that flew into my mouth’,
then we have no explanation for our feeling that B’s utterance I’ve swallowed is incoherent in the context provided by A’s question in (12), a question about the fate
of the poor bug. To the extent that (12) B makes us think of a deglutition process
of swallowing the bug, it does so in spite of the syntactic form of the sentence
used by B, which is not an acceptable way of communicating what was said in
(11). While the explicature of B’s answer in (7) above leads us straightforwardly
to the truth-conditionally more constrained implicated conclusion that B swallowed the bug, B’s answer in (12) tells us that B performed an act of swallowing
and leaves us with a feeling that the answer is either not relevant with regard to
the question asked by A, or else produced by someone whose English proficiency
is rather poor,9 which may cause the native English interlocutor A to condone the
fact that there is a missing overt object argument in B’s utterance and interpret
the utterance I’ve swallowed as if B had actually said I’ve swallowed it.
Observe that my example with the verb swallow involves no potential paradox
similar to what we experience with Burton-Robert’s example. It is not so that
the thought ‘B swallowed the bug’ arising from (7) B is implicated according to
one criterion and explicated according to a different criterion. My argument is
that there is no reason to even consider an explicature analysis to be a possible
candidate in (7), because native speakers of English do understand the proposition
expressed by B in (7) to be ostensively distinct from the proposition expressed
by B in (6), and the noted difference in acceptability between (11) B and (12) B
corroborates the evidence presented earlier.
Why do I insist that there is a genuine difference between my own example of
an implicature that entails the explicature of an utterance and Burton-Roberts’
Sperber and Wilson’s ‘presumption of optimal relevance’ (Sperber and Wilson 1995, 270) includes the following statement: The ostensive stimulus is the most relevant one compatible with the communicator’s
abilities and preferences (emphasis mine).
OSLa volume 1(1), 2009
 
mentioned example of the same phenomenon? Let us repeat his data for convenience.
There’s no milk.
The milkman’s ill.
In Burton-Roberts’ illustration rendered in (5), the causal relation between B’s
proposition (reason) and A’s proposition (consequence) is arguably expressed in
a direct manner. B could have given the same answer by using the more complex
sentence structure ‘There’s no milk because the milkman’s ill’, which requires
more decoding but less context-dependent inference. What B did instead was to
suppress repetition of information that would just be an echo of what A said in (5).
B could also have used a demonstrative pronoun to represent the consequence:
‘That’s because the milkman’s ill’. The pragmatic processing of what B says in (5)
arguably involves a pragmatic completion process, a relevance-theoretic enrichment of the free sort. At the same time there may also be fairly strong arguments
for an analysis of B’s utterance as one which gives rise to an implicature that subsumes the explicature.
My example (7), repeated here, is different. B’s ‘when’-clause in (7’), an echo
of A’s identical clause, does not change our interpretation. What we see in (7’) is
the intransitive verb swallow, grammatically and conceptually, even if the given
information expressed in the temporal clause of A’s question is echoed in B’s answer. The assumption that B swallowed the bug is no more explicated in (7’) than
in (7).
What happened when that bug flew into your mouth?
I swallowed.
What happened when that bug flew into your mouth?
I swallowed when that bug flew into my mouth..
I conclude that the assumption that speaker B swallowed the bug is conveyed in
a truly indirect way in (7), while the assumption that there is no milk because
the milkman is ill is conveyed in a less obviously indirect way in Burton-Roberts’
example (5). There was a suppressed proposition in (5) B whose recoverability
depended on the content of interlocutor A’s utterance, but there is no suppressed
proposition in (7) B that is brought into focus by the content of A’s preceding utterance. The anonymous reviewer of this paper mentioned that the answer in
(7) could be analysed as a sort of euphemism, where B is deliberately not communicating explicitly what happened when the bug flew into his mouth, maybe
because the incident was so embarrassing or so unpleasant that he does not feel
like talking about it. I agree with the reviewer’s judgment that these are possible
implicatures that would not necessarily be associated with the utterance of (6)
OSLa volume 1(1), 2009
  
B, and this corroborates my view that the utterances of (6) B and (7) B differ in
their overall relevance, even though both utterances communicate the assumption that B swallowed and the assumption that B swallowed the bug.
Neither Carston nor any other relevance theorist has thus far offered us necessary and sufficient conditions that enable us to decide where free enrichment
at the explicit level of communication stops and implicature takes over.10 A relevance theorist could, if so inclined, argue that B’s answer in (5) directly explicates
the information that there is no milk because the milkman is ill, in which case (5)
would be no threat to the Functional Independence Principle. In contrast, the explicature of B’s answer I swallowed in the exchange between A and B in (7) can be
paraphrased as ‘I performed an act of swallowing’, and this is a self-contained
communicated thought. B did not mean to swallow but it unfortunately happened, and the (implicated) consequence was the bug’s being passed down into
B’s stomach. By saying I swallowed, B did not explicate either ‘I swallowed, so I
swallowed the bug that was in my mouth’ or the truncated version ‘I swallowed
the bug that was in my mouth’.
My example involving the verb swallow, and the argument that rests on it, shows
that there are situations where an implicature of an utterance entails its explicature, but the fact that an explicature can be entailed by an implicature is not a
problem for relevance theory, it affects none of its fundamentals. We still have
what Carston (2002, 191) refers to as embedding tests, of which the most famous
one is Recanati’s Scope Principle (Recanati 1989). The Scope Principle says that a
pragmatically determined aspect of meaning is part of what is said (thus no implicature) if it falls within the scope of logical operators such as negation or a
conditional connective. When applied to conjunction data, this test reveals that
the strengthened interpretations involving a temporal sequence or a causal relation between the conjuncts come out as explicatures, or what is said in Recanati’s
Even if the explicature of B’s utterance in my example (7) is entailed by an
implicature communicated by means of that utterance, this fact does not necessarily mean that the explicature is made redundant (because it would yield no
contextual effects that are independent of the contextual effects of the entailing
implicature). It cannot be literally true that speaker B’s utterance of the sentence
I swallowed in (7) yields no contextual effect apart from what is attributable to the
implicature ‘B swallowed the bug that had flown into B’s open mouth’. The expli[10]
Lack of clarification concerning the limits of processes of free enrichment relative to derivation of
implicatures may at the end of the day turn out to be a nagging problem for relevance theory.
For some interesting thoughts on how the concept of free enrichment must be constrained, see
Nishiyama and Mineshima (2007, 2008). See also Hall (2008).
OSLa volume 1(1), 2009
 
cature of B’s utterance in (7) possibly makes it easier for the hearer to infer that
B swallowed1 (let me use subscript 1 for the concept encoded by the intransitive
verb swallow) by mistake. He should not have swallowed1 at that particular time,
because by so doing he could not help swallowing2 (let me use subscript 2 for the
concept encoded by the transitive verb swallow) the bug that had flown into his
mouth. (6) B – I swallowed it – is compatible with the assumption that B deliberately swallowed2 the bug. (7) B – I swallowed – strongly suggests that it happened
by accident.
Still, this possible pragmatic difference between (6) B and (7) B is not the main
reason why I think the kind of data that I have focused on in the present paper
should not worry relevance theorists, even though I do think the mentioned semantic properties of transitive and intransitive swallow show that the Functional
Independence Principle should be abandoned. In my view, the relevance of B’s utterance in (7) is not dependent on the inferred assumption that the swallowing1
happened by mistake. There is absolutely no reason to doubt that the explicature of the utterance of (7) B is sufficiently relevant in spite of the fact that it is
entailed by the strongly communicated implicature that B swallowed the bug. After all, Carston herself has insisted that the concept of ‘implicature’ belongs to a
cognitive theory of on-line processing of information, while the concept of ‘entailment’ belongs to a static theory of semantics. One major contextual effect of
B’s answer in (7) is the explicated information that B swallowed1 , and this gives
the hearer the opportunity to activate one or more weakly communicated implicatures that would be less accessible if the transitive verb swallow (swallow2 ) had
been used.
I would go as far as to claim, presumably against the opinion of Carston and
other relevance theorists, that for the addressee A the mental representation of
the communicated assumption that B swallowed (when the bug had flown into
his mouth) is temporally prior to A’s mental representation of the implicated conclusion that B swallowed2 the bug when he swallowed1 . In my view the temporal
sequence is no less crucial here than in an imagined situation where B actually
says the following, in a discourse where there has been no prior mention of a
bug: I swallowed – so I swallowed a bug, with a temporal break, for rhetorical effect,
between the two clauses.
I am indebted to Anne Bezuidenhout and Noel Burton-Roberts for very useful discussions, and to an anonymous reviewer for taking such an interest in my topic
and offering me some acute judgments and some further arguments. Any weaknesses are my responsibility alone.
OSLa volume 1(1), 2009
  
Amfo, N. A. A., Fretheim, T., Haugereid, L. and Nakijoba, S. 2007. The scope of
negation: How much depends on pragmatic inference? ISK Working Papers 4,
Bach, K. 1994. Conversational impliciture. Mind & Language 9, 124–162.
Bezuidenhout, A. 2005. The Semantics/Pragmatics boundary. In K. Brown (ed.),
2nd edition of The Encyclopedia of Language and Linguistics, Elsevier Publishers.
Burton-Roberts, N. 2005. Robyn Carston on semantics, pragmatics and ‘encoding’.
Journal of Linguistics 41, 389–408.
Burton-Roberts, N. 2007. Varieties of semantics and encoding: negation, narrowing/loosening and numericals. In N Burton-Roberts (ed.), Pragmatics, pages
90–114, Basingstoke: Palgrave Macmillan.
Carston, R. 1988. Implicature, explicature, and truth-theoretic semantics. In R. M.
Kempson (ed.), Mental Representations: The Interface between Language and Reality,
pages 155–182, Cambridge: Cambridge University Press.
Carston, R. 2002. Thoughts and Utterance: The Pragmatics of Explicit Communication.
Oxford: Blackwell Publishing.
Fretheim, T. 2008. Some critical remarks on the explicature–implicature distinction in relevance theory. Paper presented at the 23rd Scandinavian Conference
of Linguistics, Uppsala.
Grice, H. P. 1975. Logic and conversation. In P. Cole and J. L. Morgan (eds.), Syntax
and Semantics, volume 3, pages 41–58, New York: Academic Press.
Grice, H. P. 1989. Studies in the Way of Words. Cambridge MA: Harvard University
Hall, A. 2008. Free enrichment or hidden indexicals? Mind & Language 23, 426–456.
Levinson, S. C. 2000. Presumptive Meanings: The Theory of Generalized Conversational
Implicature. Cambridge MA: The MIT Press.
Nishiyama, Y. and Mineshima, K. 2007. Property expressions and the semanticspragmatics interface. In P. Cap and J Nijakowska (eds.), Current Trends in Pragmatics, pages 131–151, Cambridge: Cambridge Scholars Publishing.
Nishiyama, Y. and Mineshima, K. 2008. Free enrichment and the over-generation
problem. Paper presented at The conference Interpreting for Relevance: Discourse
and Translation, Kazimierz Dolny, Poland.
OSLa volume 1(1), 2009
 
Recanati, F. 1989. The pragmatics of what is said. Mind & Language 4, 295–329.
Recanati, F. 1993. Direct Reference: From Language to Thought. Oxford: Blackwell
Recanati, F. 2004. Literal Meaning. Cambridge: Cambridge University Press.
Sperber, D. and Wilson, D. 1986. Relevance: Communication and Cognition. Oxford:
Blackwell Publishing.
Sperber, D. and Wilson, D. 1995. Relevance: Communication and Cognition. 2nd edition, with a Postface. Oxford: Blackwell Publishing.
  
Thorstein Fretheim
Norwegian University of Science and Technology
Dept. of Language and Communication Studies
N-7491 Trondheim
[email protected]
OSLa volume 1(1), 2009
Behrens & Fabricius-Hansen (eds.) Structuring information in discourse:
the explicit/implicit dimension, Oslo Studies in Language 1(1), 2009. 33-61. (ISSN 1890-9639)
  
    
G E R G E L Y P E T H Ő  E V A K A R D O S
University of Debrecen
This paper discusses the occurrence and the licensing of implicit object arguments, also referred to in the literature as null complements or understood arguments. Functionalist accounts (such as those by Groefsema and
Németh T. which are couched in a relevance-theoretic framework) have repeatedly claimed that this phenomenon is fundamentally dependent on discourse-interpretational factors. In particular, it has been stated that implicit arguments can be used in Hungarian in a rather unrestricted way, and
their occurrence is only limited by considerations of interpretability. We
argue against both of these positions and try to show that cross-linguistic
data can assist in revealing the circular nature and ultimate inadequacy of
existing functional accounts of implicit argument licensing.
This paper discusses the licensing of implicit object arguments, also referred to
in the literature as null complements or understood arguments. We will argue
that theories which aim to explain this phenomenon exclusively in terms of general pragmatic principles (in particular, relevance) are insufficient. We will try
to show that some well-known results of these approaches are conceptually and
empirically problematic. Instead, a different perspective on argument omission
will be proposed, departing from Iten et al. (2005)’s suggestion according to which
more traditional lexical-syntactic explanations need to be combined with semanticpragmatic (interpretational) approaches in order to develop a satisfactory theory
of these phenomena.
The structure of this paper is as follows: Section 2 will briefly introduce the
notion of implicit arguments and summarise the most important points of Fillmore’s lexical-syntactic approach (1986), including the classic distinction between
indefinite and definite null complements, already proposed by Fodor and Fodor
(1980). Section 3 discusses Groefsema (1995), a paper that tries to explain both the
licensing and the interpretation of implicit arguments on the basis of Relevance
  
Theory. After identifying some problems connected to this position, we will conclude that Groefsema’s view of implicit argument licensing is inadequate. Section
4 first turns to the work of Németh T. (most importantly, 2001), who elaborates
on Groefsema’s theory and applies it to Hungarian. We will indicate that Németh
T.’s strongly interpretational account does not differentiate between several possible motivations for the occurrence of implicit arguments, and accordingly it is at
least not sufficiently detailed, perhaps even on the wrong track. Consequently, we
will agree with Iten et al. (2005) in that a two-stage model, consisting of a lexicalsyntactic and an interpretational-pragmatic component, is needed instead. In a
short case study, in section 5, we will further substantiate the claim that syntactic
factors are essential to this issue. This section considers superficially related phenomena like resultative constructions and verb prefixation, and aims to demonstrate why cross-linguistic data have to be taken into account in connection with
issues like the topic of this paper. Finally, section 6 concludes our discussion.
       
 
[2.1] Characteristics of implicit arguments
It is a well-known fact that there are sentences in which, judging by their surface
structure, some units of information are missing. We can distinguish between different types of such phenomena which are often referred to as implicit arguments
in a broad sense. By way of illustration, let us examine a range of examples:
My bicycle was stolen yesterday.
Come over here!
She wanted to open a shop.
Bill likes but Mary hates ice-cream.
My grandmother taught me how to knit.
I wanted to keep my new job a secret but my mother found out.
Although these examples share the property that a particular argument of a predicate is missing in each of them, they differ significantly and do not form a natural
class. They are analysed according to different criteria. Example (1) is a passive
construction in which the agent argument is not identified by an explicit phrase.
(2) is an imperative sentence that is understood with a second person singular
pronominal subject, even though this pronoun does not appear on the surface. (3)
is a so-called control structure that contains an infinitival predicate (open) that
does not have an explicit subject. Generative syntax commonly assumes that a
silent element, called PRO, occupies the subject position of this non-finite verb.
PRO, in turn, is controlled by (or is, in other words, co-referential with) the subOSLa volume 1(1), 2009
 
ject of the finite verb (i.e. she). In (4) we find what is generally considered an
elliptic sentence, in the sense that the object of the first verb likes is deleted from
the surface since it is identical with the object of hates. (5) contains (in addition
to another control structure) the transitive verb knit, which does not, however,
have an explicit object in the sentence. The sentence does not talk about a specific article of clothing being knitted, but rather how the speaker has acquired the
ability to knit. Finally, in (6) the phrasal verb found out would normally demand
a CP or a PP complement with about as its head. Although this example does not
contain either kind of argument explicitly, they can be inferred on the basis of
the information conveyed at the beginning of the sentence.
As indicated earlier, these types of construction are not treated in a uniform
way in linguistic theory. (1) and (2) are related to morpho-syntax as their description involves reference to the grammatical categories of voice and mood, respectively. The missing elements in (3) and (4) can be treated by reference to syntactic
regularities: (4) can be analysed as a result of PF deletion that is triggered by a certain syntactic configuration, whereas the absence of the agent argument in (3) is
required due to the fact that English non-finite verbal predicates cannot agree
with an explicit nominative subject. In contrast, (5) and (6) are typically not regarded as having to do primarily with syntax, but rather with lexical semantics,
argument structure, and discourse/pragmatics.
As we will discuss later, we do not believe that treating all of these phenomena
within a single unified theory of implicit arguments is viable. Instead, our main
concern will be the phenomena exemplified by (5) and (6), which can be called
null complements, understood arguments, or, alternatively, implicit arguments
in a stricter sense.
Let us briefly examine what makes implicit arguments special. Arguments
are generally defined as expressions that are combined with a predicate to form a
proposition. More specifically, each predicate is assumed to have a certain number of free positions which have to be filled by argument expressions. If we apply
this to natural languages, it means that “verbs come with a semantic frame” which
“specifies the linguistically inherent participants in the process, state, or event
that the verb denotes”, as Iten et al. (2005, 2) put it (and this characterisation can
and should be extended to other types of predicates as well, most importantly to
adjectives and nouns).
Iten et al. (2005, 3)’s example is the transitive verb lock, which ontologically
requires both a subject (agent) and an object (patient) argument in the sense that
one cannot conceive of a locking event without someone (or something) who is
doing the locking and something that is being locked. At the same time, this verb
requires both a subject and a direct object complement in a grammatical sense as
well (although there are certain syntactically defined constructions that override
this requirement, e.g. passive).
OSLa volume 1(1), 2009
  
A different type of predicate, one that can be used with an implicit argument,
is exemplified by follow. Whereas this verb also has two “linguistically inherent”
arguments, namely the follower and the thing being followed, and these are both
ontologically required for an event to qualify as a following event, the object argument can be left out under circumstances that cannot be chracterised in strictly
grammatical terms, as in:
Tom’s already gone out to Rome and his wife and children will follow shortly.
(LDOCE 4 CD-ROM:  1.)
The difference between lock and follow is that the latter verb can be used without an explicit direct object provided it becomes clear from the context who is
being followed, but the same is not possible for lock under similar conditions. In
other words, for follow there seems to be a failure of isomorphism between linguistic levels: the verb’s semantics, as per the semantic frame, specifies two linguistically inherent participants for the denoted event/action/process, but in some
fully grammatical sentences there is no (overt) constituent in the syntax for one
of them (Iten et al. 2005, 3) .
A linguistically inherent argument that is characterised by such a failure of
isomorphism in a given sentence (i.e. it is phonologically unrealised but contributes to the meaning of the whole structure) can be called an implicit argument
in a narrow sense.
Having introduced the concept of implicit arguments, let us now come to a
possible explanation for this phenomenon, which seems reasonable as a first approximation.
[2.2] Fillmore’s lexical-syntactic approach
A classic approach to implicit arguments is the so-called lexical-syntactic approach,
which assumes that the ability of a predicate to occur without some of its arguments in a sentence is encoded in the part of its lexical entry that specifies
predicate-argument relations. In his (1986) paper, Fillmore stresses that verbs are
to be distinguished on the basis of their ability to stand with null complements,
such as in the case of lock vs. follow (as mentioned above). One of his own examples was the verb-pair guarantee and promise. He states that despite the similarity
of the two verbs’ meanings (they both assume the existence of something guaranteed and something promised, respectively), the omission of the object clause
is not allowed in the case of guarantee (and similarly for the verbs pledge and vow),
whereas it results in grammatically correct sentences with promise.
OSLa volume 1(1), 2009
She guaranteed / pledged / vowed *(that she will take part in the organization of the conference).
 
She promised (that she will take part in the organization of the conference).
Fillmore explains the difference between verbs (and predicates in general) that
do allow and ones that do not allow null complements by claiming that different
argument structures are assigned to them in the lexicon. He states that complements in general are not omissible, so specific verbs that do license implicit
arguments have to be provided with a syntactic feature value marking whether a
complement can be omitted.
Fillmore further proposes a renowned classification of implicit arguments according to certain parameters. He points out that a distinction is needed between
what he calls indefinite null complements (INC) and definite null complements
(DNC).1 The members of the first group are implicit arguments, the referent of
which cannot be identified, or its identity is a matter of indifference (i.e. they are
understood as not referring to anything specific, which can be approximated by
an existentially quantified variable in semantic representations). INC verbs such
as eat, read, write, and drink can occur in structures where the object of the eating,
reading, writing or drinking event is not expressed. For instance, in (10) we can
infer the presence of a hidden object (that is a liquid substance).
John was drinking.
‘John was drinking something that is possible to drink.’
As opposed to the former group, DNC are interpreted as having definite referents
(for example, specific discourse referents) that have been mentioned in previous
utterance segments.
Peter’s parents had never got married. I wonder if he knows that.
Oh, I think he finally found out [that his parents had never got married]2 .
How come you have submitted your paper a month before the deadline of July 1st?
Well, my instructor insisted [that I submit my paper a month before
the deadline].
In (11-b) and (12-b) the DNC can be interpreted on the basis of (11-a) and (12-a)
because its referent is clearly extractable from context. Similarly to the issue
Fodor and Fodor (1980), among others, had already distinguished between two classes of verbs allowing
object omission in a similar way. They claim that the semantic representation of sentences containing a member of the eat-class of verbs involves existential quantification. John ate, for example, entails
that John ate something. In contrast, they see notice-class verbs (e.g. watch, join in, approve) as necessarily
anaphoric and entailing a definite object: John noticed entails John noticed it and not John noticed something.
Information that is understood but is not uttered explicitly is surrounded by square brackets.
OSLa volume 1(1), 2009
  
of omissibility in general, Fillmore assumes that INC versus DNC behaviour of a
complement is specified in the lexicon: “the determinants of the omissibility phenomena are lexical in the sense that individual lexical items will simply have to
be represented as having certain of their complements marked as indefinite omissible and definite omissible.” (Fillmore 1986, 98; italics in the original). He explicitly
states that he does not see a viable semantic or pragmatic explanation for the
variations in question: “There are certain semantic groupings of predicates that
allow the two kinds of complement omission, but a genuine semantic explanation does not appear to be forthcoming. In the case of DNC, no purely pragmatic
explanation will help us either” (op. cit. 98).
In other words, Fillmore believes that both the licensing of implicit arguments
and the determination of their interpretation are idiosyncratic features of individual lexical items. However, this position is not shared by everyone. A number
of authors have pointed out that there are quasi-syntactic conditions in English
that license or at least facilitate the omission of complements, like coordination,
topicality and contrastive focus among others, cf. Goldberg (2001), Cote (1997),
or Velasco and Muñoz (2002). Semantics has also been assumed to play a decisive
role in the interpretation of INC versus DNC verbs contrary to Fillmore’s position: Sæbø (1996) argues that semantic presuppositions connected to individual
verbs enable us to predict whether an implicit argument receives a definite or an
indefinite interpretation. Finally, Fillmore’s lexical-syntactic approach has also
been disputed by authors who have emphasised that pragmatics plays a significant role in licensing implicit arguments, e.g. Groefsema (1995) and Németh T.
(2000a, 2001). We will examine the views of these authors in the following two
’  
Groefsema takes a position that is radically different from Fillmore’s with respect
to both major components of the theory of implicit arguments: 1) the licensing
of omission, and 2) the choice of definite versus indefinite interpretation of an
omitted element. Firstly, she claims that “arguments can only be left implicit
if their interpretation is constrained in particular ways, because the constraints
make the interpretation of the implicit argument immediately recoverable. When
there is no constraint on the interpretation of an argument, leaving it implicit
could lead the addressee to a misinterpretation, and Relevance considerations
would rule this out” (Groefsema (1995, 139). Secondly, the difference between INC
and DNC verbs is that “INC verbs only put a selection restriction on their internal
argument as far as the type of THING that is at stake [sic], while DNC verbs specify
that an instance of a THING is at stake” (op. cit. 147). For both components of the
theory, we will first summarise what arguments Groefsema presents to support
her claims and then we will point out some empirical and conceptual problems.
OSLa volume 1(1), 2009
 
[3.1] Licensing of implicit arguments in Groefsema (1995)
Groefsema accounts for the omissibility of arguments on semantic-pragmatic grounds within the framework of Relevance Theory:
A verb can only be used with an understood argument if the interpretation of the argument is constrained in one of two ways. In the
first place, an argument can be left implicit if the verb puts a selection
restriction on the argument such that it gives us an interpretation in
accordance with the principle of relevance (Groefsema 1995, 152).
In the following lexical entry of the verb eat, it is specified that the direct object
of the verb is of the type FOOD.
EAT [Thing] ,
In case this verb is used without an explicit direct object in some context, it can
therefore be retrieved from the lexicon that it is some food that is being eaten.
Note that the selection restrictions do not mean that the predicate in question
cannot occur at all with an object that does not belong to this sort (FOOD); for example, the sentence Sandy has eaten a marble in the sense of ‘Sandy has swallowed a
marble’ is grammatical and acceptable and can be interpreted without difficulty.
It is, nevertheless, an instance of loose use in the sense of Carston (2002, Chapter
5). As Németh T. and Bibok (1999, 414) observe, implicit objects are regularly interpreted in accordance with such selection restrictions, so these restrictions can
apparently function, more generally, as some kind of default condition regarding
the content of the verb’s arguments. For example, the sentence
Sandy has eaten.
cannot be interpreted as ‘Sandy has swallowed a / the marble’ but only as ‘Sandy
has eaten some food / a meal’.
The second part of Groefsema’s condition on implicit argument licensing is
that “an argument can be left implicit if the rest of the utterance makes immediately accessible an assumption (or assumptions) which gives us an interpretation
in accordance with the principle of relevance” (op. cit. 153). Her examples are
the following:
Paul gave to Amnesty International.
? Paul gave to Ann.
I always give books on birthdays.
OSLa volume 1(1), 2009
  
In (14) the words give and Amnesty International make immediately accessible the
information (“assumption”, as Groefsema puts it) that people usually give money
to this organisation. In (16), the words give and on birthdays activate our knowledge that one usually gives presents (for instance, books) to other people on their
birthdays, and furthermore, that there is a set of people that one usually gives
presents to, namely, relatives and friends. This allows us to establish the likely
recipients of the books, even though they are not explicitly mentioned. By contrast, in (15) the words give and Ann do not enable us to identify the type of thing
that was given. Thus the phenomenon illustrated in (14) and (16) is similar to
the previous example in (13) in the sense that information that is not explicitly
spelled out in the sentence (because an argument is missing) is inherent in some
other information source. In the previous case, this source was the lexical entry
of the predicate itself. In this case, it is the immediate context of the predicate
within the utterance.
We believe that Groefsema’s position on licensing is problematic for several
reasons. In connection with simple cases like (13), her assumption is essentially
that the stricter selection restrictions apply to a certain argument (i.e. the more
information about its possible arguments the verb itself contributes to the sentence), the more likely it is that a relevant interpretation can be achieved without
any explicit mention of that argument. Consequently, if selection restrictions are
strict enough, an argument of the verb may remain implicit.
Note that Groefsema does not define a clear boundary between cases where
information provided by such restrictions is sufficient to license an implicit argument and where this is not the case. It does not become clear why e.g. eat can be
used intransitively, as opposed to lock, which requires its object to appear on the
syntactic surface, even though its relevant selection restrictions do not seem to
be any less specific: eat requires food as its object, whereas lock requires an object
that has a lock, e.g. a car or a door.
In addition to this problem, which we consider the most serious difficulty for
Groefsema’s approach, there are further issues to be noted. We do not think that
it is possible to reduce the possibility of the omission of an argument to the availability of a relevant interpretation for the utterance (on the basis of what has
been explicitly said and what is known by the language users). Consider the verb
read that can be used without an explicit direct object, in which case the implicit
argument receives an indefinite interpretation. If we add the adverb aloud to this
verb, omission of the direct object becomes far less acceptable. From the point of
view of Groefsema, this is unexpected because the set of possible candidates for
the referent of the direct object is, if anything, more restricted in the case of read
aloud than in that of read (one usually reads aloud a sentence, a passage, a poem,
or a short text, but not a novel).
OSLa volume 1(1), 2009
 
Furthermore, Iten et al. (2005) have demonstrated that Groefsema’s reasoning
in connection with (15) is stipulative and ultimately unconvincing:
If ‘being optimally relevant’ were really the issue, and things beyond selection restrictions and utterance make-up could contribute
to relevance, then any context in which the discourse situation made
clear enough that money was at issue should, it seems, be one in which
‘money’ can be omitted from (15). So, we don’t get the prediction that
(15) is unacceptable unless there is a ban on non-linguistic stuff contributing assumptions when evaluating for optimal relevance. (Iten
et al. 2005: 9)
They maintain that it makes no sense to exclude non-linguistic information from
optimally relevant interpretation, and this would indeed seem to contradict the
essence of pragmatic explanations. We will return to this issue in section 4.4 and
suggest that Groefsema’s position might be more properly seen as a result of circular reasoning instead of stipulation.
What has been said so far shows that, although there may be more than a
grain of truth in Groefsema’s picture of the licensing of implicit arguments, it is
at the very least incomplete: As it stands, the theory only allows us to predict that
if a certain predicate-argument combination has already been used successfully
in a certain context (I am eating), it will probably also be acceptable if we use it in
another similar context to similar effect. However, on the basis of what Groefsema
says in her paper, it is crucially impossible to tell in advance whether I am eating,
I have locked, Paul gave to Ann etc. per se are acceptable or not. On the other hand,
when her theory does in fact enable us to formulate a prediction, that prediction
is often wrong (as in the case of read aloud, or eat versus lock above).
Indefinite vs. definite interpretation of implicit arguments in Groefsema (1995)
Let us now turn to the second component of Groefsema’s theory of implicit arguments, namely, her position on INC versus DNC behaviour of verbs. She derives
this property from selection restrictions as well. As we have indicated, she claims
that “a verb can specify that its argument is of a particular type [or] that it is an instance of a particular type” (op. cit. 146), and that an implicit argument receives
an indefinite interpretation if it is marked in the corresponding selection restriction of the verb as a “type”, whereas definite (anaphoric) implicit arguments are
marked as “instances”. Eat as an INC verb specifies that its object has to be a type
of food, as we have seen above in section 3.1, whereas win as a DNC verb has an
object that is an instance of a competition:
OSLa volume 1(1), 2009
  
WIN [Thing] ,  Instance
However, it is unclear what Groefsema means by this distinction, since (quite trivially) being of a particular type and being an instance of a particular type are
arguably one and the same thing. Sticking by the actual characterisations that
Groefsema provides, i.e. claiming that read has a selection restriction that says
it takes objects that are types instead of instances, we would expect that read belongs to the class of predicates that do in fact take objects that are names of kinds,
very much like invent, become extinct or common:
Bell invented the telephone.
The dodo has become extinct.
The field mouse is a common animal.
Note that in these cases, the object never refers to a specific object (an instance
of a type) but to a type as such. Furthermore, these predicates cannot occur with
objects that denote specific entities at all, e.g. *Bell invented the telephone in the
corner. *My dog has become extinct. We can, therefore, conclude that these verbs do
have the exact same selection restriction that Groefsema proposes for INC verbs
like eat, drink and read, which, unfortunately for her, do not behave like this at all
(even allowing that we can regard selection restrictions as default conditions, like
in the case of eat taking an object that is food). She must have something different
in mind instead of the distinction between “type” and “instance” proper, but since
she does not provide any further explanation for her dichotomy, it is unclear how
it could be rephrased in a more coherent way.
Assuming that the above distinction does indeed make sense, it still seems
that Groefsema’s characterisation of specific verbs makes incorrect predictions
about their distributions in sentences where they do in fact occur with an explicit
argument. Although it is uncertain by what kind of complement the selection restrictions “type” and “instance” would exactly be satisfied, one could reasonably
assume that this distinction is related to some well-known linguistic semantic
dichotomy, e.g. genericity vs. specificity. Thus, read being a verb that takes, according to Groefsema, an object that is of type ‘symbolic representation’ (but not
an instance of this type), one could assume that this predicts that read may only
(or at least typically) take objects that are bare plurals or mass nouns (20) but not
NPs referring to specific entities (21).
OSLa volume 1(1), 2009
I am reading newspapers.
 
I was just reading that brown book over there.
On the other hand, win, being a verb that demands an object that is an instance of
a type, should only appear with specific NP objects (22) but not with bare plurals
She has won the race.
They play really smart ball and they often win championships.
(LDOCE 4 CD-ROM: examples bank for )
It is not only this pattern that could plausibly be assumed to be predicted by
Groefsema’s selection restrictions but fails to apply. In fact, we cannot find any
other clear distributional differences between INC and DNC verbs in general, either, when their relevant arguments are explicitly expressed. Thus it would seem
that what Groefsema has stated are not really selection restrictions at all, but
simply a description of the fact that some implicit arguments receive a definite
or indefinite interpretation, i.e. nothing that would further illuminate the issue. In other words, it seems that Groefsema’s position is in fact identical to Fillmore’s lexical theory that she attacks, with the sole difference that she (mistakenly) characterises Fillmore’s features definite and indefinite omissible as selection
restrictions and calls them, somewhat confusingly, ‘instance’ and ‘type’ (instead
of definite/indefinite or specific/non-specific).
       
[4.1] An outline of a general interpretational theory
In the previous section we summarised an approach that accounted for the licensing of implicit arguments on semantic-pragmatic grounds. We can also find
a number of further attempts to explain the phenomenon in question from a related, more or less functional perspective. For example, Goldberg (2001) concentrates on the omissibility of direct object arguments. The essential claim of her
paper is that an object argument can be omitted whenever it is not at the centre
of attention. More specifically, “omission is possible when the patient argument
is not topical (or focal) in the discourse, and the action is particularly emphasized (via repetition, strong affective stance, discourse topicality, contrastive focus, etc.)” (Goldberg 2001, 514).
Instead of trying to survey functionalist work on the issue extensively, we will
concentrate, in the following, on one author who takes a very distinctive stand
with respect to our subject. We assume that the problems that we will indicate
(and that have in part already been mentioned in the previous section) apply,
mutatis mutandis, to all approaches to implicit arguments that try to reduce licensing of omission to the aspect of interpretability.
OSLa volume 1(1), 2009
  
Németh T. (2000a, 2001, see also Németh T. and Bibok 1999, and Bibok and
Németh T. 2001) examines Hungarian verbs and claims that these “verbs do not
vary as to whether they can occur with implicit arguments, but they vary as to in
what manner and in what context they can occur with such arguments” (Németh T.
and Bibok 1999, 412). She states that there are three ways in which complements
can be left out of a sentence:
(i) if some part of the conceptual semantic representation of a verb
– including a selection restriction or a prototypical structure of a category or a lexical stereotype or a presupposition – makes it possible
to recover the implicit argument in accordance with the principle of
relevance; (ii) if the rest of the utterance in which the argument occurs makes an assumption with the typical interpretation immediately accessible in accordance with the principle of relevance; and
(iii) if extending the immediate context of the argument leads to an
interpretation consistent with the principle of relevance (Németh T.
2001, 152).
Her options (i) and (ii), taken together, are essentially the same as Groefsema’s
condition on the omissibility of arguments. (i) explains omission in cases such
as (13) above, and (ii) accounts for examples like (14). The motivation for (iii) is
exemplified in (24) (taken from Németh T., 2001: 147) and (25) below.
(The interviewee is speaking about her relationship with her younger
brother in their childhood.)
– Nem voltál rá mérges néha, hogy te most nem
mehetsz oda, ahova akarsz, mert az öcsédre kell vigyáznod?
‘Weren’t you angry with him sometimes for not being able to go where
you wanted to, because you had to take care of your younger brother?’
Dehogynem. Dehogynem. S  [ ]
yes-of-course yes-of-course often occur-PAST.INDEF.3SG
   
  [ ].
and PFX also use-to-PAST-1SG beat-INF for-it
‘Yes, of course, of course. It happened many times and I used to beat him
for this.’
In the first half of the highlighted sentence, literally ‘occurred many times’, the
argument that specifies the event that occurred is missing. This is possible because that information has already been mentioned in this context, namely in the
first sentence: ‘you were angry with him because …’. In the second half, literally
‘used to beat for this’, two arguments are left out of the sentence, namely, the first
OSLa volume 1(1), 2009
 
person singular subject, and the third person singular object. Although Németh T.
does not discuss this, it is the grammar of Hungarian that is ultimately responsible
for the possibility of omission of these arguments in the case of both predicates:
Hungarian verb inflection clearly marks number and person of the subject and
in large parts of the paradigm definiteness or indefiniteness of the direct object
by agreement morphemes. For example, the ending of előfordult ‘occurred’ shows
that the subject is third person singular.3 Context plays a role insofar as the verbal inflection (together with the rules of Hungarian pro-drop) only reveals that
both the subject of előfordul and the object of megver ‘beat’ is third person singular.
The actual reference (in a broad sense) of these missing arguments is established
by pragmatics. More specifically, what enables us to achieve a relevant interpretation is the linguistic context (preceding discourse), as explained above.
Furthermore, the elements of the situation and encyclopaedic knowledge may
also contribute to the creation of a relevant interpretation, as in the following
example (Németh T. 2000b, 201):
(A mother is strolling along with her children. One child is sitting in a
baby carriage, while the other one, Áron, is walking next to the carriage.
“I’ll step back to the nurse”, says the mother to the kids; the baby starts
The mother comforts her child:
“Ne sír-j,
Áron  [ ].”
Not cry-IMPER.2SG, Áron push.INDEF.3SG
‘Don’t cry, Áron will push you.’
In this case the subject of the verb tologat is expressed (Áron) but its object is
missing. From the inflectional suffix of the verb form, namely, the suffix that
expresses indefinite object agreement, we can only tell that the implicit object
must be either first or second person singular or indefinite third person singular (‘something’). On the basis of the situation (the speaker, the hearer, and other
people present) and world knowledge (people usually push babies in carriages and
this way they comfort them) we can infer that it is not the mother (first person
singular) or an indefinite third person singular (something) that is being pushed,
but rather the baby that is the addressee of the utterance (second person singu[3]
The Hungarian verbal paradigm consists of two sub-paradigms: one is typically used when the verb has
an indefinite direct object, whereas the other is used if there is a definite direct object. The indefinite subparadigm is “unmarked” in the sense that it is chosen not only when the verb actually has an indefinite
third person direct object, but rather as a default case. This means that all intransitive verbs (which, of
course, do not take an object at all) as well as verbs with first and second person direct objects (which
are not, strictly speaking, indefinite) appear with indefinite object agreement morphemes as well. For
example, the form előfordult in (24) is intransitive and accordingly contains the suffix characteristic of
the indefinite sub-paradigm. On the other hand, the definite sub-paradigm is only used when the direct
object is third person (singular/plural).
OSLa volume 1(1), 2009
  
Németh T. stresses that “verbs can occur with implicit arguments far more
freely in Hungarian than in English” and other languages (Németh T. 2001, 113).
By this she presumably means that if a sentence can occur with an implicit argument in a certain context in one language, it is possible that in another language
where the occurrence of implicit arguments is less free, the equivalent sentence
will be unacceptable / ungrammatical. In another paper she adds the following:
“in Hungarian, verbs do not vary as to whether they can be used with implicit
arguments or not as they do in English and in languages similar to English, but
they vary as to in what manner or in what context they can occur with such arguments” (Németh T. and Bibok 1999, 425). This claim is central to Németh T.’s view
and therefore deserves attention. Her formulation would suggest that basically
everything is possible in Hungarian with respect to argument omission, which
would make this language similar to East-Asian languages that are considered to
be extremely “free” in this respect.
However, this latter conclusion is incorrect. It only arises because one important consideration is missing from Németh T.’s account: she does not specify
what licensing factors underlie the omission of arguments but instead analyses
particular examples from the perspective of interpretation, without examining
their grammatical characteristics.
If we take into account the aspect of grammar as well, however, we get a more
complex picture, according to which Hungarian is not all that different from English. In other words, if we exclude the phenomena of ellipsis (like (4)) and prodrop4 in Hungarian, we realise that similarly to English, some verbs do whereas
others do not allow object omission. This applies both to ‘direct’ (accusative) and
to ‘indirect’ (otherwise marked) objects, as in (26) and (27), respectively:
János fogott.
‘John held/caught.’
Example (26), with an indefinite inflection form of the transitive verb fog ‘hold/
catch’, is not interpretable and is unacceptable unless a pro-drop interpretation
with a first person singular object (‘John held me’) is forced.
*János hozzáfogott.
‘*John has begun.’
Example (27) contains the verb hozzáfog which takes an indirect object with the
OSLa volume 1(1), 2009
The phenomenon of pro-drop in Hungarian is an intricate system, which we will not explain here for the
lack of space. We can only note that Hungarian object pro-drop is interconnected with the definiteness
agreement of Hungarian verbal inflection that was briefly outlined in the previous footnote. The reader
is advised to consult any good descriptive grammar of Hungarian for a more thorough description of the
 
case-ending -hoz, i.e. is not strictly speaking transitive. Object pro-drop only applies to direct (accusative-marked) objects in Hungarian, so it is no option in this
case. The verb normally requires an explicit indirect object, so the sentence is
ungrammatical unless it is parsed as an elliptic construction that requires an appropriate indirect object to be understood on the basis of the immediate context.
We can conclude that Németh T. can only maintain her claim that Hungarian
is significantly more “free” compared to English with respect to argument omission if her theory does not distinguish between different sources of omission (ellipsis, pro-drop, and others). If we do take these issues into account, the contrast
in question between these two languages has rather little to do with pragmatics,
but boils down essentially to differences between their grammars: most importantly the fact that one is a pro-drop language but the other is not; and also to divergent rules of ellipsis (which presumably follow from more abstract parametric
variation with respect to sentence structures).
To sum up: we believe that treating all kinds of argument omission as a common, unified phenomenon is not the right way to proceed. Claiming that every
Hungarian verb can occur with an implicit object is, in a way, correct. But this
would be akin to claiming that the same holds for every English verb, too, since
verbs that normally require an object, e.g. lock, like etc., are allowed to occur
without one in right node raising structures (cf. Postal 1974) like (4). Németh T.’s
strategy is to take structures where, contrary to English, Hungarian allows argument omission, and show that the way Hungarian behaves is “reasonable” from a
Relevance Theory perspective. We believe that looking for the reasons why these
cross-linguistic differences occur would lead to deeper insights. Presumably, in
many cases the outcome of this search would be a collection of language-specific
regularities that are, strictly speaking, syntactic, and that can be defined much
more precisely than the vague “freeness of argument omission” parameter that
is often referred to in the relevant literature.
An alternative: a two-stage model of implicit argument licensing
To continue the previous point, we will argue that an account of implicit argument licensing with a significantly higher explanatory power could be attained if
we chose a more differentiated theoretical framework. One such general framework was proposed by Iten et al. (2005). After pointing out some deficiencies of
Groefsema (1995), the authors conclude that an exclusively interpretational theory of implicit argument licensing is inadequate. We can call a theory of licensing
‘interpretational’ if it claims or implies, like the approaches of Groefsema (1995)
and Németh T., that argument omission is allowed if and only if the sentence containing a null complement can be interpreted (namely, by relying on word meaning, linguistic and non-linguistic context, and world knowledge).
OSLa volume 1(1), 2009
  
Iten et al. (2005) argue that an adequate explanation of licensing will have
to acknowledge the role of language-specific syntactic rules and word-specific
idiosyncrasies as factors equally important to general considerations of interpretability. They state that the question ‘what licenses null complements?’ can
be read in two ways.
It can be taken to be a question about expression types, i.e., about
what the grammar does (or does not) generate. Call this Q(A). Alternatively, it can be read as a question about which in-context utterances
are acceptable/felicitous. Call this Q(B). Put otherwise, Q(A) is about
which verbs (or verb senses) allow null complements, and of which
kind, as a matter of their context-invariant grammar (Iten et al. 2005, 10;
emphasis as in the original).
Q(B), on the other hand, enquires under what contextual conditions a speaker
may use a null complement provided that the sentence is generated by grammar.
Accordingly, we can assume a two-stage model of language use that consists of a
first stage of pure syntax, which decides whether a structure is grammatical or
ungrammatical, and a second stage of interpretation, which decides whether the
utterance makes sense (is acceptable) in a certain context. If a structure is not
generated by grammar, it might still be interpretable and even occur in everyday language use (as a performance error, which might be so frequent that it can
become lexicalised eventually as an idiomatic phrase) but it is clearly ungrammatical, e.g.: *Do you like? in the sense of ‘Do you like this?’. On the other hand,
if a structure is generated, its acceptability in a particular context will depend on
whether it is interpretable.
Let us illustrate why we need both components of this system and how the
two stages are related to each other. Firstly, we examine the English verb insist, which is a typical DNC verb: it allows its object complement to be omitted,
and if omission indeed takes place, the missing argument is interpreted as definite/anaphoric. In other words, this verb can occur with an implicit object in
contexts like (28), taken from Fillmore (1986, 98):
“Why did you marry her?”
“Because mother insisted.”
In this case, grammar generates the structure (it is grammatical) and interpretation is straightforward on the basis of the preceding sentence (it is pragmatically
acceptable). Let us contrast this with a second example, which should be interpretable but is strongly ungrammatical.
OSLa volume 1(1), 2009
“Why did you marry John?”
*“Because mother wanted.”
 
This example shows why it is necessary to have a grammatical component in our
theory of licensing: if context (28) allows for the object of insist to be omitted, the
same should be possible in (28)’s minimal pair (29), which only differs from the
former with respect to the choice of the verb. Since the context is identical, there
are only two possible explanations for the difference of acceptability between the
two examples (we exclude interpretational options along the lines of Groefsema
(1995), as we should on the grounds of what has been said in section 3): Either the
lexical entries of the two verbs specify that the first is a DNC verb, whereas the
second must not omit its object, or the syntax of English tells us in general that
finite clauses and indirect objects (like the object clause or the on-PP complement
of insist) may, but non-finite clauses and direct objects (which occur with want)
must not be omitted. At this point, we will not address the question which of
these options is the correct one, but we can at least state that one of them seems
logically necessary.
Finally, let us see an example which shows that we cannot do without a pragmatic component of our theory of licensing, either.
“Are you coming to the party today?”
# “No, mother insisted.”
Since (28) is perfect, we can infer that the sentence ‘Mother insisted.’ is generated
as such by grammar. However, its occurrence is restricted to contexts where it
becomes clear what mother insists on. Since this is not the case in (30), the answer
is pragmatically unacceptable, even though it is grammatical.
To conclude our introduction of the two-stage model, let us briefly consider
the status of its notions of grammaticality and acceptability. If language use were
“ideal”, we would only produce utterances that are both grammatical and acceptable. However, we know that this is not always the case. As a result of what is
generally known as performance errors, we do use sentences that should not be
generated by grammar. On the other hand, as a result of an inadequate estimation
of the context we also produce utterances that are unintelligible for the interlocutors and are therefore pragmatically unacceptable. Nevertheless, we have no
choice but to assume that language use is mostly acceptable and grammatical and
that we can at least consciously judge whether a given utterance is problematic
in some respect.
[4.3] Signs of circularity in interpretational theories
We believe that the problems noted by Iten et al. (2005) in connection with Groefsema (1995) can be stated on a higher level of abstraction as well. Interpretational (or we could also call them functional) explanations of implicit argument
licensing seem very plausible. After all, we are indeed able to interpret acceptable sentences with implicit arguments and do indeed have trouble making sense
OSLa volume 1(1), 2009
  
of unacceptable ones. As a consequence, it is a straightforward choice to regard
interpretability or, put more generally, ease of interpretation as a decisive condition on omissibility. However, this begs the question whether this condition is
just necessary or indeed sufficient as well. Although the advocates of the interpretational approach do not address this issue, they obviously assume that interpretability is indeed a sufficient condition of omissibility. (Note that this may be
regarded by some as a general problem for functionalist approaches in linguistics
as a whole.) This assumption, i.e. that we only have to consider a single factor
when deciding whether an argument is omissible, ultimately gives rise to circular reasoning, as is observable in the papers we have cited. Let us illustrate this
by three typical examples.
Example 1: Groefsema (1995, 140) mentions that, according to Fillmore, argument omission cannot be given a semantic explanation because “semantically
related groups of verbs do not display the same behaviour regarding whether or
not they allow understood arguments.” Cf. She found out versus *She discovered; I
am waiting versus *I am awaiting; I tried versus *I attempted; insist versus require etc.
Groefsema does not accept Fillmore’s position because “he does not show in what
way these verbs are semantically related” (Groefsema 1995, 144) and furthermore
she assumes, without explicitly arguing for her decision, that the meanings of
these verbs are indeed significantly different from one another. Obviously, it is
no wonder that Fillmore does not go out of his way to prove the semantic similarity of these verbs because they are, in fact, relatively closely synonymous:
they can be freely exchanged within appropriate contexts, contribute the same
truth conditions to sentences uttered in the same context, can be paraphrased in
roughly the same way, etc. The only conceivable reason for claiming that they are
semantically not related would be the consideration that they absolutely have to
be semantically different in case they behave differently with regard to argument
omission. In other words, the reasoning goes: “Why is it that find out occurs with
a DNC and discover does not? Because they are semantically unrelated. How do we
know that they are semantically unrelated? Because find out occurs with a DNC
but discover does not.”
Example 2: This is mentioned in Iten et al. (2005, 9) as Groefsema’s Dictum. It
can be illustrated by our examples (14) and (15), repeated below:
Paul gave to Amnesty International.
?Paul gave to Ann.
The question is: following Groefsema’s relevance theoretic view, why is (15) less
acceptable than (14), even in contexts that make it clear what Paul was going to
give to Ann. Note further that Paul gave it to Ann would be the preferred way of
conveying the message in question, which only differs from (15) in that it conOSLa volume 1(1), 2009
 
tains a semantically quite empty but explicit pronominal it. Groefsema is forced
to answer: “Relying on non-linguistic context inevitably leads to unnecessary processing
effort, so that the resulting utterance will not be optimally relevant” (Iten et al. 2005, 9;
emphasis as in the original). Now how do we know that it leads to unnecessary
processing effort? Because (15) is less acceptable than (14) or the same sentence
with an explicit pronominal. Again, the argumentation is viciously circular.
Example 3: We have mentioned above that Groefsema (1995, 146-147) tries to
explain the difference between INC and DNC verbs by claiming that different selection restrictions are assigned to their relevant arguments. But how do we know
that the selection restrictions assigned to these arguments are different? As we
have pointed out, this question would normally be answered by enumerating distributional limitations of the verbs in question (i.e. what kinds of argument they
can co-occur with). However, Groefsema does not mention such limitations and it
is questionable whether there are any at all. Therefore, we can assume that Groefsema’s answer would be: they must be different because some of the arguments
are DNC and others are INC, and there must be a lexical-semantic explanation for
We believe that the appearance of such examples of circular reasoning in
Groefsema (1995) is only an indication of a more fundamental conceptual problem: exactly because of their apparent plausibility, purely functional explanations for implicit argument licensing are prone to be circular, or, in other words,
be no more than pseudo-explanations. The best way to avoid this is to compare,
in several languages, otherwise equivalent structures that differ with respect to
the possibility of argument omission.
[4.4] Cross-linguistic differences and similarities
It has been stated several times in the literature (as early as Fillmore 1986) that
a purely functional approach to implicit arguments is not viable, especially when
we take into account differences between languages. Let us see a few examples to
illustrate this point.
Groefsema (1995, 155) claims that (15) above is bad because it is not optimally
relevant without an explicit pronoun. However, relevance is a pragmatic notion
that is essentially not language specific, but a universal cognitive category. Therefore, if a structure without a pronoun violates the principle of relevance in one
language, it follows that it must be equally problematic in another language. Arguably, this is not the case. For example, the following sentence (31) in Hungarian
would be grammatical, even though it does not contain an explicit pronoun.
Pál Annának adta (?#AZt).
Paul Ann-to gave (it-ACC).
‘Paul gave it to Ann.’
OSLa volume 1(1), 2009
  
In Hungarian the anaphoric pronoun azt is grammatical in a sentence like (31),
but in most contexts e.g. an answer to a question like ‘What did Paul do with the
money?’ it would be completely unacceptable, and even in narrative texts, where
the pronoun is allowed to appear, its omission is mostly preferred. However, this
apparent difference between Hungarian and English is in some sense only superficial. Sentence (31) contains a definite object agreement verb form adta, and this
inflection is basically identical in function to the explicit anaphoric pronominal
it. Therefore, we could claim with Groefsema that in whatever way the pronoun
contributes to optimal relevance in an English sentence, verbal inflection plays
the same role in its Hungarian equivalent. So there is no necessary theoretically
relevant difference between (15) and (31). The same cannot be said, however,
about examples from other languages such as Southern Cone Spanish mentioned
in (Iten et al. 2005), where the morpho-syntactic features of the sentence do not
help the hearer recover the missing object, like in (32) (taken from op. cit. 15):
para vos.
Already put-1ST.SING for you
‘I already put for you.’
We can see that Groefsema’s theory predicts that (32) should be just as unacceptable as (15), which is wrong. We can again conclude that a universal pragmatic
theory is not the correct way to account for the licensing of implicit arguments,
but we need language specific grammars to achieve this goal.
There is another way as well in which cross-linguistic evidence can play an important role in rejecting a purely functional view. There are certain similarities
between languages with respect to implicit arguments for which an interpretational explanation does not suggest itself.
This line of reasoning goes as follows: Let us assume that there is, within the
same language, a contrast regarding argument omissibility between two structures, which seems to be their particular idiosyncratic property. When arguing
for his lexical-syntactic approach, Fillmore (1986) mentions quite a few such contrasts, for example, I am waiting versus *I am awaiting. It would seem that the only
way to account for this and similar cases would be to state in the lexicon that
the oblique object of wait is omissible, whereas the direct object of await is required. In other words, there is no obvious semantic difference between the two
verbs and no obvious grammatical rule, either, by which this contrast could be
deduced. Consider an apparently unrelated phenomenon as well: There is a difference between the behaviour of a particular verb in a resultative construction,
on the one hand, and in various other syntactic environments, on the other hand,
in that verbs never allow the omission of a direct object in a resultative structure
John painted *(the door) white, even if they can be used with an implicit INC object
in a non-resultative sentence, e.g. wash, paint. Again, it would not be entirely
OSLa volume 1(1), 2009
 
trivial to find an interpretational explanation for this behaviour of resultatives
and it would at the very least be logically possible that it is an idiosyncratic property of this particular English construction that direct objects must appear in it.
However, we cannot exclude the possibility that these phenomena only appear
idiosyncratic because we are looking at a single language and do not have a basis
of comparison. We believe that this is exactly the case. If we consider other languages in addition to English, we will find that a) prefixed verbs in general tend
to require their direct object to be explicit and b) resultative constructions are always obligatorily transitive in any language. Assuming that these generalizations
are correct and that they cannot be explained by reference to the relative interpretability of the structures in question, they must have some kind of grammatical explanation. In other words, a theory of implicit argument licensing should
not be exclusively based on functional principles, but must have a grammatical
component that is in part universal. We will explore this idea in more detail in
the next section.
  :  ,  ,
   
We have mentioned in the previous section the well-known generalisation (stated
for example in Tenny (1994)) that transitive verbs that would otherwise allow an
intransitive use (e.g. paint) cannot omit their internal argument in a resultative
construction. A similar restriction seems to apply to resultatives and purportedly similar structures (e.g. prefixed and particle verbs) cross-linguistically, in
unrelated languages like Hungarian as well.
János pirosra
*(egy kerítést)5 .
John red-RESULT. painted-INDEF. a
‘John painted *(a fence) red.’
A syntactic-semantic explanation could be attempted for this characteristic property of resultatives, namely, by stipulating a principle that secondary predication,
like the one present in resultatives, requires the argument predicated about to
appear explicitly. By contrast, an interpretational (semantic-pragmatic) explanation does not seem to be feasible; note e.g. that the resultative construction’s
inability to appear with an implicit object is not influenced by context (especially
salience of an eligible patient in context) at all, e.g.:
This sentence contains the indefinite object agreement form of the verb fest ‘paint’, which would be fine
in a non-resultative structure with an omitted INC object like János festett ‘John painted’. Note that inserting the definite object agreement form of the same verb instead, festette, would render (33) perfectly
grammatical, since that would be a pro-drop structure equivalent to the English sentence ‘John painted
it red.’
OSLa volume 1(1), 2009
  
*János egész délután a fürdőszobában volt, és pirosra festett.
6= ‘John spent the whole afternoon in the bathroom and painted [it /
something] red.’
Similar differences between different occurrences of the same verb can be observed in other cases that can be argued to relate to secondary predication as well,
namely, particle verbs in certain languages. The particle, which may predicate a
result state of the internal argument, forces that argument to occur explicitly,
even if it could remain implicit otherwise, e.g. German and Hungarian:
Sie näht für ihre Kinder. (DUW 4:  1.)
she sews for her children
‘She sews for her children.’
Wir nähen *(die kleinen Stoffstücke)
we sew the small pieces-of-fabric together.
‘We sew *(the small pieces of fabric) together.’ (based on COSMAS)
Kati írt
Kati wrote afternoon
‘Kati wrote in the afternoon.’
Kati kiírt
*(három kérdést)
a könyvből.
Kati out-wrote three
questions-acc the book-from
‘Kati copied *(three questions) out of the book.’
As is well-known, these changes in argument structure correlate with a shift in
the aspectual properties of the verbal predicate, i.e. whereas the relevant simple
verbs are of the event type ‘process’ (‘activity’), the resultatives and verbs complemented by similar secondary predicates are ‘accomplishments’/’achievements’
(cf. Rothstein 2004, Chapter 5). In other words, whereas no clear relationship
between the role of context and the possibility of argument omission could be
observed in these cases, there is in fact a connection between event structure and
argument omission, which indicates that both phenomena belong to the syntaxsemantics interface.
The assignment of accusative case by the verb (i.e. a typical grammatical argument structure phenomenon) is another property which is connected to resultatives and prefixed verbs. In the case of resultatives, an otherwise intransitive verb can become transitive (more specifically, reflexive; compare e.g. Müller
(2002, 213-215)), e.g. She would cry herself to sleep at night. The same is true for
many particle verbs which can also be argued to involve secondary predication,
e.g. German herausschreien ‘cry out’.
OSLa volume 1(1), 2009
 
Er schrie
seine Freude
he screamed his happiness out
‘He cried out with joy.’
Similarly to the cases (36) and (38) above, the omission of the object leads to ungrammaticality.
Considering the examples mentioned so far, it may seem plausible to suggest,
along the lines of the semantic-pragmatic approach, that what necessitates the
appearance of explicit arguments with secondary predication and related phenomena does in fact have something to do with the possibility of reaching a relevant interpretation. One could claim, for example that in the absence of an explicit argument expression, we would not be able to make sense of a secondary
However, on further scrutiny this reasoning does not seem to be correct. Note
that prefixation of verbs leads to the observed effect, i.e. transitivisation and the
impossibility of argument omission, even in cases where the prefix quite clearly
does not predicate anything about the internal argument in question. This observation can be made with respect to several languages as well. In the following
examples (German prefixed verbs with be-), the simplex verb has both a transitive and an intransitive use (in other words, an optional, omissible argument),
whereas its prefixed version may only be used transitively (both taken from COSMAS).
Heute bekocht er nur noch seine Familie und Freunde.
today PFX-cooks he only still his family and friends
‘Today, he does not cook for anybody anymore except his family and
Der Schienenbus befährt
die Strecke am
Sonntag dreimal
the rail-bus
PFX-drives the distance on-the Sunday three-times in
beide Richtungen.
both directions
‘The train travels this distance every Sunday three times in both directions.’
Similarly, in Hungarian:
János eszik
(egy almát). vs. János megeszik
*(egy almát).
John eats.INDEF an apple vs. John PFX-eats-INDEF an apple
‘John is eating an apple.’
‘John is eating an apple.’
In these cases, the simplex verb is lexically imperfective (that is, of the event type
‘process’), the prefixed one is lexically perfective (an accomplishment). In neither
OSLa volume 1(1), 2009
  
of these examples can we account for the required appearance of the object by
the tentative interpretational explanation we have proposed above, since these
prefixes do not express secondary predication in a strict sense6 . Instead the effect
of the verbal prefix in such cases is simply transitivisation (along with argument
structure modification) and perfectivisation.
Thus far, we have considered verbs from languages that have an extensive
productive system of verb formation by prefixation. However, the same observation can be made in connection with English as well. We have mentioned earlier the fact that despite being closely synonymous, wait and await differ in their
ability to occur with an implicit object. Such examples are problematic for the
interpretational approach since that would predict that two synonymous predicates should behave identically with respect to argument omission. We have also
mentioned that these differences appear to be idiosyncratic. However, this can be
slightly revised here considering the preceding discussion. For note that there is
also a clear difference between the argument structures of these verbs (wait takes
an oblique object, whereas await takes a direct object) in parallel with the possibility of intransitive use. Similarly to the cross-linguistic tendency we have already
referred to, this correspondence between transitivisation and the impossibility of
argument omission appears to a lesser degree in English as well. English prefixed
verbs with a-, be- and en- are regularly obligatorily transitive, regardless of the
lexical category of the base word of the prefixation (i.e. noun, verb or adjective),
cf. typical examples derived from verbs: belabour, bescribble, besmear, entrust, arch.
bepaint, berob etc.and their simplex versions.
For years, parents and teachers have bemoaned the fact that we do not have
a national childcare policy.
(LDOCE 4 CD-ROM: examples bank for )
I’m fed up with hearing you moaning the whole time.
(LDOCE 4 CD-ROM: examples bank for )
We have to stress that this phenomenon is not productive in present-day English.
Instead, these forms are the remnants of a formerly existing rule of English grammar, namely verb transitivisation by prefixation. We also need to add that in
German and in Hungarian, where the morphological process is indeed productive, there is no clear dichotomy between prefixed and non-prefixed verbs in the
sense that all and only those verbs that are prefixed should be transitive, perfective, and unable to occur with an implicit object, whereas non-prefixed verbs
OSLa volume 1(1), 2009
Note that whereas this claim seems highly plausible to us, there are authors (notably, Kiss 2006) who
would claim, on theoretical grounds, that even prefixes like the apparently semantically empty meg or
verb modifiers in idiomatic complex verbs are secondary predicates (e.g. the formal object in the idiom
csütörtököt mond ‘fail’, literally ‘say Thursday’; or the goal adverbial in fűbe harap ‘die’, literally ‘bite into
 
should be intransitive, imperfective, and object-omitting. Still, there is a large
number of verb-pairs that exhibit this behaviour, which demands a grammatical
explanation. Because there is no obvious functional grammatical explanation, we
can assume that we need to account for these phenomena in the syntax-semantics
interface of a formalist, universal grammar.
   
We hope to have demonstrated some of the shortcomings of interpretational (semantic-pragmatic) theories of implicit arguments. In particular, we tried to make
the case that the licensing of such arguments cannot be explained by a theory
which exclusively relies on the notion of ease of interpretability (or, in other
words, optimal relevance), because such an approach runs into serious conceptual and empirical problems, especially when cross-linguistic evidence is considered. Whereas a semantic-pragmatic approach apparently makes very reasonable claims about implicit arguments, certain differences and similarities between
data from different languages indicate that under closer scrutiny, it is grammatical (especially syntactic) regularities rather than interpretability or pragmatics
that probably constitute the most important factor of the licensing process. Accordingly, we claimed that a two-stage model of licensing (consisting of a grammar and a pragmatics/interpretation component) promises more adequate explanations for the phenomena in question.
At the same time, we have to admit that what we have done in this paper is
no more than pointing out some shortcomings of existing approaches and giving a very vague outline of an alternative position. In particular, it should be
clear that the two-stage model of licensing that we have sketched is not, by any
measure, a proper theory of implicit arguments, but rather a research strategy
or a general attitude toward the description and explanation of the relevant phenomena. In order to turn it into a theory that can be debated, it should be decided exactly what kind of grammar underlies the first stage, and what the pragmatic/interpretational regularities are that make up the second stage. This will
have to be filled in by future research.
As a closing thought, let us mention some of our reservations about the picture that Iten et al. (2005) suggest about the relationship of the two subsystems responsible for licensing. We have mentioned that they assume that interpretability
decides about acceptability if the structure in question is generated by grammar.
However, consider examples like (45), which is very similar to (28) in section 4.2:
“Are you coming to the party today?”
#“No, because I have work to do and mother insisted.”
Example (28) is fine, i.e. mother insisted has to be a well-formed structure, and
OSLa volume 1(1), 2009
  
there is nothing else in the answer above that one would expect to render the
utterance unacceptable. Furthermore, there is a very straightforward candidate
in the directly preceding context that could be identified with the DNC of insist,
namely: mother insisted that I do the work. On the basis of these premises and
what Iten et al. say about the licensing process, we would expect the answer in
(45) to be acceptable. However, it is not. To see what causes the problem in this
example, compare the following question-answer pair:
“Are you coming to the party today?”
“Yes, because I have to meet Jack and, besides, mother insisted.”
It seems that the difference between the unacceptable answer in (45) and the acceptable one in (46) is, indeed, that the latter makes sense whereas the former
does not. Apparently, the DNC in structures like (28) and (46) is only allowed to
pick up an antecedent from a preceding question, but not from the speaker’s own
answer. This way we correctly get the interpretation ‘mother insisted that I come
to the party’, which makes sense. On the other hand, by the same rule we have
to get the very same interpretation for the DNC in (45) as well, which would be
incoherent: if mother insisted that I come to the party, this cannot be straightforwardly interpreted as an explanation of why the speaker is not coming to the
party, as in the answer in (45).
In other words, something (presumably, the grammar of DNC in English) forces
an incoherent interpretation for the answer in (45) and even blocks the selection
of an appropriate antecedent for the DNC that would allow a coherent interpretation. Such cases show that the relationship between the grammar and the pragmatics component of the licensing system is more complex than what Iten et al.’s
suggestion would imply. The grammar component apparently does not just generate a structure like mother insisted and leave the interpretation of the DNC to the
pragmatic component. Instead, grammar can rather strongly restrict this interpretation by reference to context, thereby further limiting the role of pragmatics.
Because of the complex nature of this interaction between context and syntax,
it is unclear whether all theories of grammar are equally able to formulate the
relevant generalisations. Arguably a strongly “context sensitive” syntactic theory
like dynamic syntax (Cann et al. 2005) would be a more promising framework to
describe such phenomena than a more standard generative model.
To turn to a different issue, it seems to us that even though pragmatics is not
capable of explaining all aspects of implicit arguments, it does play an extremely
important role in the process of an obligatorily transitive verb becoming an optionally intransitive verb. We are referring to cases like certain intransitive uses
of the verb draw, used (in special contexts) to express things like ‘draw a card’,
‘draw a gun’, etc. It seems that when a verb very frequently and predictably takes
a certain kind of argument in a certain kind of situation (like draw taking a card
OSLa volume 1(1), 2009
 
as an object when the interlocutors are playing cards, or taking a gun as object in
a Western movie context), it can occasionally be used intransitively. This intransitive use can eventually become lexicalised, as a sort of “technical term” used in
appropriate contexts. Although this seems to be a very common and systematic
phenomenon, it has been largely ignored in the literature. Still, we can safely
assume that intransitivisation in such cases is intimately connected to specific
situations of communication and considerations of economy, and can therefore
be regarded as belonging to pragmatics. However, this conjecture will have to
be confirmed by research, and the actual extent of the phenomenon in everyday
language use also remains to be established.
    
(Examples for which we have not specified, as a source, either one of the following
or one of the literature references below, are our own.)
• COSMAS: COSMAS II (Corpus Search, Management and Analysis System),
Mannheim: Institut für Deutsche Sprache.
• DUW 4: Duden Universalwörterbuch. 4th Edition. Mannheim: Bibliographisches Institut and Brockhaus, 2001.
• LDOCE 4 CD-ROM: Longman’s Dictionary of Contemporary English. 4th Edition, CD-ROM edition. Harlow: Pearson Education, 2003.
Bibok, K. and Németh T., E. 2001. How the lexicon and context interact in the
meaning construction of utterances. In E. Németh T. and K. Bibok (eds.), Pragmatics and the flexibility of word meaning, pages 289–320, Amsterdam: Elsevier.
Cann, R., Kempson, R. and Marten, L. 2005. The dynamics of language: An introduction. Syntax and Semantics 35, Amsterdam: Elsevier.
Carston, R. 2002. Thoughts and utterances. Oxford: Blackwell.
Cote, S. 1997. Grammatical and discourse properties of null arguments in English.
Ph. D.thesis, University of Pennsylvania.
Fillmore, C. 1986. Pragmatically controlled zero anaphora. Proceedings of the Berkeley Linguistics Society 12, 95–107.
Fodor, J. and Fodor, J. D. 1980. Functional structure, quantifiers, and meaning postulates. Linguistic Inquiry 11, 759–769.
OSLa volume 1(1), 2009
  
Goldberg, A. 2001. Patient arguments of causative verbs can be omitted: the role of
information structure in argument distribution. Language Sciences 23, 503–524.
Groefsema, M. 1995. Understood arguments: A semantic/pragmatic approach.
Lingua 96, 139–161.
Iten, C., Junker, M.-O., Pyke, A., Stainton, R. and Wearing, C. 2005. Null Complements: Licensed by Syntax or by Semantics-Pragmatics? In M.-O. Junker,
M. McGinnis and Y. Roberge (eds.), Proceedings of the 2004 Canadian Linguistics
Association Annual Conference, ACL.
Kiss, K. 2006. Focussing as predication. In V. Molnár and S. Winkler (eds.), The
Architecture of Focus, pages 169–196, Berlin: Mouton de Gruyter.
Müller, S. 2002. Complex predicates: Verbal complexes, resultative constructions, and
particle verbs in German. Stanford: CSLI.
Németh T., E. 2000a. Occurrence and identification of implicit arguments in Hungarian. Journal of Pragmatics 32, 1657–1683.
Németh T., E. 2000b. Implicit argumentumok a magyarban: előfordulásuk módjai és azonosításuk lehetőségei (Implicit arguments in Hungarian: Modes of their
occurrence and possibilities of their identification). In I. Kenesei (ed.), Igei vonzatszerkezetek a magyarban (Verb complement structures in Hungarian), pages 197–252,
Budapest: Osiris.
Németh T., E. 2001. Implicit arguments in Hungarian: manners of their occurrence and possibilities of their identification. In I. Kenesei (ed.), The syntax and
semantics of argument structure, pages 113–156, Budapest: Akadémiai.
Németh T., E. and Bibok, K. 1999. Implicit arguments in Hungarian. In J. Verschueren (ed.), Pragmatics in 1998. Selected papers from the 6th International Pragmatics Conference, volume 2, pages 412–426, Antwerp: IPrA.
Postal, P. 1974. On Raising. Cambridge: MIT Press.
Rothstein, S. 2004. Structuring events: A study in the semantics of lexical aspect. Oxford: Blackwell.
Sæbø, K. J. 1996. Anaphoric Presuppositions and Zero Anaphora. Linguistics and
Philosophy 19, 187–209.
Tenny, C. 1994. Aspectual roles and the syntax-semantics interface. Dordrecht:
OSLa volume 1(1), 2009
 
Velasco, D. and Muñoz, C. 2002. Understood objects in Functional Grammar.
Working Papers in Functional Grammar 76. Amsterdam: University of Amsterdam.
  
Gergely Pethő
University of Debrecen
Institute of German Studies
H-4010 Debrecen, Pf. 47
[email protected]
Eva Kardos
University of Debrecen
Institute of English and American Studies
H-4010 Debrecen, Pf. 73
[email protected]
OSLa volume 1(1), 2009
OSLa volume 1, 2009
Behrens & Fabricius-Hansen (eds.) Structuring information in discourse:
the explicit/implicit dimension, Oslo Studies in Language 1(1), 2009. 63-75. (ISSN 1890-9639)
   
  
Tohoku University / University of Sydney
The present study is an attempt to account for non-sentential utterance
(NSU) production without assuming the existence of a ‘syntactically full sentence’ for every NSU. The model for NSU production that derives from this
study has the following four advantages over the popular ‘constituent-omission’ model: It 1) accounts for the production of NSUs that native speakers
variably ‘reconstruct’, 2) explains why in certain contexts pro-drop cannot
occur in languages that have morphologically marked subject-verb agreement 3) models the production of NSUs without devising separate production processes for ‘ellipses’ and ‘fragments’, and 4) predicts what constituents
have to be present in a given NSU. It also keeps the involvement of syntax
in NSU production to a minimum.
 
In various areas of study within the discipline of linguistics, nonsentential utterances (including many of those that are referred to as elliptical sentences) are
identified as ‘full sentences’ from which syntactic or grammatical constituents
are omitted (see, e.g., Lyons (1977, 589), Brown and Miller (1991, 144-146), Napoli
(1996, 200), Matthews (1997, 111), Bavin (2000), Malmkjær (2002, 543), Merchant
(2004)). Accounts of non-sentential utterance (hereafter NSU) production where
NSUs are identified as sentences with missing constituents are necessarily based
on the assumption that every NSU has a corresponding ‘full sentence’ which native speakers ‘reconstruct’ with (near-)unanimous agreement. However, despite
its popularity, one can find without much difficulty examples that contradict this
assumption. For example, as early as 1974, Gunter (1974, 12-13) devised the term
‘telegraphic ellipses’ to refer to NSUs (‘elliptical sentences’ in his terminology) of
which ‘informants do not agree … on the proper expansion’ (ibid: 13).1 The exis[*]
Financial support for this research came in part from MEXT Grant-in-Aid for Young Scientists (B)
(20720101) and the Matsushita International Foundation. I would like to thank Nayanbaatar Amarjargal,
Naoya Arakawa, Joungmin Kim, Shuhrat Rahmatov, Sarvar Rahmatullaev, Kei Takahashi, Kei Yoshimoto,
and the participants of the SPRIK conference 2006 for their comments, help, and criticism.
For an argument concerning the indeterminacy of the contents of NSUs, see Clapp (2005, 120-121). See
also Barton (2006, 19). The model presented in this paper postulates that an utterance may be in any
form provided that f ’s (foci) and m’s (morphemes grammatically required to accompany f ’s) (if there are
any) are present in it and it does not contradict the proposition of the utterer.
 
tence of such NSUs questions the assumption that a ‘full sentence’ underlies every
Recent years have seen a few separate attempts to analyze NSUs without assuming underlying full sentences. Progovac (2006) argues that full sentences are
underlyingly small clauses (rather than the other way around) and that many
NSUs are in fact small clauses that are not transformed into full sentences. Progovac thus argues for the feasibility of identifying NSUs as small clauses which
would become full sentences through transformation. Culicover and Jackendoff
(2005, chapter 7) introduce a process they call indirect licensing, through which
certain phrases in NSUs are assigned semantic roles and syntactic features. This
process allows NSUs to have the semantics of full sentences without actually being full sentences. Culicover and Jackendoff’s analysis of NSUs hence genuinely
dispenses with the notion of the underlying full sentence. These two studies aim
at accounting for NSU interpretation and do not attempt an explanation of NSU
The present study is an attempt to account for NSU production without assuming the existence of a ‘(syntactically) full sentence’ for every NSU. Although
this study is at a somewhat preliminary stage, the model for NSU production that
derives from it has the following four advantages over the popular ‘constituentomission’ model: 1) it accounts for the production of the sort of NSUs that Gunter
calls ‘telegraphic ellipses’, 2) it explains why in certain contexts pro-drop cannot
occur in languages that have subject-verb agreement morphology, 3) it models
the production of NSUs without devising separate production processes for ‘fragments’ and ‘ellipses’, and, perhaps more significantly, 4) it predicts what constituents have to be present in a given NSU. Another characteristic of the model,
which may also be an advantage over the ‘constituent-omission’ model, is its ability to produce NSUs with a minimum of syntax.2
The present study is also distinct from many other studies of NSUs in that it
presents a model for NSU production rather than for NSU resolution, which, incidentally, has attracted enough attention among specialists to become a focus of
inquiry in linguistics. Since this study models NSU production, given an appropriate set of data, it can make certain predictions about the forms of NSUs, which
differentiates it from the majority of NSU studies.
The following section outlines how this model, which I tentatively call the
composite model, accounts for NSU production. As will be apparent from the outline that will be presented in the following section, the composite model is capable
of accounting for the production of utterances that are not syntactic constituents.
(Indeed, section 4 will present an NSU that makes perfect sense despite the fact
that it consists solely of bound morphemes.) In this respect too, the composite
OSLa volume 1(1), 2009
Progovac (2006) and Culicover and Jackendoff (2005) share this characteristic with the model presented
 
model diverges radically from the majority of studies on NSUs.3
Before we proceed to the outline of the composite model, a note on the referent of the term NSU is in order. NSUs as they are identified in the present paper
include not only ‘elliptic utterances’ but also utterances that make good sense
despite their divergence from prescriptive sentence grammar, some examples of
which will be dealt with in section 4.
 
The principle on which the composite model is based is simple:
if an NSU contains foci (which are, as will be explained below, morphemes
in this model) and morphemes that are grammatically required to accompany them, that NSU is sensical (i.e. makes sense).
In the remainder of this paper, I will refer to the first half of this principle, namely
‘if an NSU makes sense, it necessarily contains one or more foci’ as P 1 and
the second half, i.e. ‘if an NSU makes sense, it necessarily contains morphemes
that are grammatically required to accompany the focus or foci’ as P 2.
Naturally, this second principle applies only if the grammar according to which
an NSU is produced requires the focus or foci in that NSU to be accompanied by
other morphemes.
The definition of ‘focus’ in the present study differs from current definitions
in the linguistic literature. ‘Focus’ in the present study is, roughly speaking, the
morphemic representation of the difference between discourse participants’ propositions. Foci in the composite model are defined as ‘morphemes that correspond with the propositional components by which a proposition differs from
another proposition in the current context’. This definition, which as it is probably does not make much sense, will be explicated below with an example.
In figure 1 I present a simple schematic representation of the model in which
a Turkish dialogue taken from Enç (1986, 195) with modification is used as an
example. f ’s and m’s in the following chart represent foci and morphemes that
are grammatically required to accompany them, respectively.
The composite model may hence have implications for the discussion about the categorical distinction (or
lack thereof) between ‘elliptical sentences’ and ‘sub-sentences’, exemplified by Ellugardo and Stainton
(2005). These implications are not unintentional but are outside the scope of the present paper.
OSLa volume 1(1), 2009
 
m -3
m m
 1: Schematic representation of the model
OSLa volume 1(1), 2009
 
Notice that (4) in figure 1 on the facing page is ‘elliptic’ – it is an NSU which
would, in the ‘constituent-omission’ model, be identified as the following ‘full sentence’ from which Ali’yle has been omitted:
Ben Ali-’yle
‘I did not meet Ali’.
Note also that the occurrence of the pronoun ben ‘I’ is obligatory in (4) despite
the presence of the subject-verb agreement suffix -m (1SG) — ## Tanışmadım is
awkward as a response to (1).
As was stated above, in the composite model, ‘foci’ are defined as ‘morphemes
that correspond with the ‘propositional components’ by which a proposition differs from another proposition’. A more precise definition would be ‘the set of
morphemes onto which the difference of two sets of propositional components
map’. F in the following chart is the set of foci.
For example, if we consider (2) and (3) in figure 1 to be two sets of ‘propositional components’ A and B, respectively, the difference of B and A, or B−A, the
elements of which are ¬ and j, corresponds to the morphemes -ma and ben. -ma
and ben are then identified as foci. Foci are thus the morphemes that correspond
with the elements of the relative complement of A in B. (The use of the logical form
as a set of ‘propositional components’ in this section is a temporary expedient; see
section 3.)
At the time (1) is uttered, John’s proposition4 (3) differs from Mary’s (2) by
having one connective and one individual constant, namely ¬ and j, which correspond respectively to the negative morpheme -ma and the first person singular
pronoun ben. These morphemes, which are called foci in this model, need to be
present in (4) because Principle 1 calls for their occurrence. This explains the
obligatory occurrence in (4) of the pronominal subject ben ‘I’, whose occurrence
is otherwise not obligatory. ‘Pro-drop’ cannot occur in (4) since Principle 1 does
not allow a covert element to constitute a focus.
On the other hand, the other of the two foci, namely -ma, cannot occur in
isolation and calls for the accompaniment of three morphemes, namely tanış-, -dı,
See the next section for an explanation of the use of the term ‘proposition’ in this paper.
OSLa volume 1(1), 2009
 
and -m, the occurrence of which is required by Principle 2. I explain below why
specifically tanış-, -dı, and -m are m’s. Principle 1 calls for the occurrence of f ’s.
Accordingly, -ma has to occur:
In Turkish one cannot say -ma in isolation, as it is a bound morpheme. Accordingly, -ma needs to appear with some other morphemes to be grammatically acceptable. Above all, it needs a verb stem:
tanışm -maf
This means ‘do not meet’ in the imperative mood, which contradicts (3). Another
m needs to occur to render the mood of John’s utterance indicative.
tanışm -maf -dım
This means ‘s/he did not meet’, which also contradicts (3). Accordingly, another
m, namely -m (1SG) needs to accompany the above. Thus, if one adds morphemes
to -ma until it becomes grammatically acceptable without contradicting (3), the
result will be as follows.5
tanışm -maf -dım -mm
Ben and tanışmadım are then linearly aligned in accordance with the minimum
syntax which the alignment calls for. In the case of the production of (4), the
syntactic constraint that the alignment necessitates can be as simple as ‘Turkish
is a verb-final language’. (If an NSU contains only one free morpheme, syntax
would have no role in its production.) Syntactic constituents that Principle 1 and
Principle 2 produce are often very small in number, which minimizes syntax in
NSU production (likely enhancing the efficiency of utterance production).
In summary, the production of an NSU is represented in the composite model
as a process in which foci (f s) and their required accompaniments (ms) are ‘put
together’ rather than as a process where constituents are omitted from a ‘full’
sentence. This model thus 1) accounts for NSU production without assuming the
existence of a ‘full sentence’ for every NSU, 2) identifies constituents that must
occur in a given NSU, 3) explains why pro-drop cannot occur in certain contexts,
OSLa volume 1(1), 2009
The morphemes added to the f should be within the range of a function that has the intersection of A
and B as its domain (see section 2).
 
and 4) minimizes the role of syntax in NSU production.
[2.1] Cross-linguistic validity
The validity of the composite model is not limited to Turkish. For example, the
Mongolian, Bukharan Tajik, and Japanese equivalents of (4) exhibit the same obligatory occurrence of a first person singular pronoun and a negative morpheme as
was observed in Turkish (4) in the previous section. (The information structural
property of each morpheme is shown in subscript.) All of these examples obey
both Principle 1 and Principle 2: ‘foci (f ’s) and morphemes that are grammatically required to accompany them (m’s) necessarily occur’.
Mary’s utterance:
Herkes Ali’yle tanıştı. (Turkish)
Bügd Jontoy uulzsan (biz dee). (Mongolian)
Hamma Ali kati šinos šud. (Bukharan Tajik)
Minna Arini atta (ne). (Japanese)
‘Everyone met Ali’.
John’s response to (1):
In Turkish
benf tanışm -maf -dım -mm
In Mongolian6
bif uulzm -aam -güyf
In Bukharan Tajik
manf na-f šudm -amm
In Japanese
watashif -wam awm -anaf -kattam
‘I didn’t’
Note that the first person singular pronoun has to occur in (4) in all of these languages, regardless of whether they utilize subject-verb agreement morphology or
(4) contains several m’s in all of the four languages. There are, however, also
languages in which (4) contains no m’s, such as German, in which (4) consists only
of f ’s:
The meaning of this sentence is closer to ‘I haven’t met Ali’ than it is to ‘I didn’t meet Ali’. The Mongolian
translation of the latter is not used here because of the modality of regretfulness that it encodes.
OSLa volume 1(1), 2009
 
In German
Ichf nichtf
I not
This German response to (1) exemplifies NSU production where there are no m’s,
which, unlike f ’s, do not occur unless the grammar according to which the utterance is produced calls for their occurrence. (The occurrence of f ’s, on the other
hand, is obligatory in any language.)7
Finally, let us observe the English utterance I didn’t in the light of Principle 1
and Principle 2. The information structure of I didn’t is analogous to that of (4) in
German; I didn’t has both I and not as f ’s. However, unlike German, not requires
the accompaniment of at least one morpheme, namely the auxiliary did, thus: If
didm -n’tf / If didm notf .8
     
 
This section explicates the representation of propositions used in the chart presented in the previous section. As may have been apparent from the discussion
in the previous section, propositions in this model are compositional and can be
decomposed into components, which are, in the outline in section 2, tentatively
shown with symbols such as ¬ and j. The logical form used in (2) and (3) in figure
1 should therefore be considered representations of structured, decomposable semantic entities, for which ‘propositional structures’ may be a more suitable label.
In this paper, however, I will continue using the term ‘proposition’ for the sake of
The logical form is in fact far from adequate for the purpose of representing
such propositional structures. My use of logical form in representing propositional structures is essentially an expedient, though it has an advantage, too. The
logical form is appropriate for our purposes here because, unlike a number of
other formalisms, such as LFG, it is impervious to syntax and grammatical relations, both of which have little place in the production of (4).9 Notwithstanding
this advantage, ideally another formalism should be devised that integrates with
the composite model seamlessly10 , because, as an anonymous reviewer of this pa[7]
OSLa volume 1(1), 2009
Unless, of course, the absence of a morpheme or morphemes corresponds with an element in B−A (see
section 2).
Note that Principle 2 does not restrict m’s to bound morphemes. The auxiliary did in this example is an
m though it is arguably a free morpheme.
In section 2, I present Mary and John’s propositions as (2) and (3) because for the examples that we look
at in the present paper, the ‘components’ of the logical form link with morphemes more or less directly.
A formalism where components of a semantic representation can be mapped into a set of morphemes
is particularly desirable because the logical form used in section 2 is in fact incapable of representing
some types of propositions in such a way that their components can be mapped readily into a set of
 
per points out, the logical form does not fare well with other discourse contexts
in which NSUs are particularly commonplace; namely discourse contexts where
questions and answers take place, to which we turn now.11
In the following analysis of a question-answer pair, I tentatively use Jackendoff’s
Lexical Conceptual Structure 1990 as an alternative to the logical form. LCS is less
of a lingua franca than first-order logic, but now that the crux of the model has
been introduced, I think that the use of a notation that is of limited popularity
will not render the argument here intractable. Observe the following questionanswer pair.
What did you eat?
The following are propositional structures of the speakers of (5a) and (5b) interpreted into LCSs.12
[Event CAUSE ([Thing B]α A , [GO ([Thing x]‹A› , [TO [IN [MOUTH-OF [α]]]])])]
[Event CAUSE ([Thing B]α A ,
[GO ([Thing PILAF]‹A› , [TO [IN [MOUTH-OF [α]]]])])]
(6a) represents, roughly, ‘(there exists) an event where B causes a thing x13 to
go to the interior of the mouth of B’. (6b) is identical with (6a) except that it has
PILAF where (6a) has an open variable x. Obviously, the propositional component
by which (6b) differs from (6a) is PILAF. This component then finds its morphemic
representation in the morpheme pilaf , which then becomes an f . In accordance
with Principle 1, pilaf f has to occur in speaker B’s response to (5a). Colloquial
English grammar does not require any ms to accompany pilaf f and accordingly
Principle 2 does not trigger any addition of ms to the pilaf f
Turkish pilavf ‘pilaf’ as a response to ne ye-di-n? (what eat-PAST-2SG) ‘what
did you eat?’ can be produced in the same way as English pilaf f . So can Japanese
I limit the scope of the present paper to NSUs that occur in some specific ‘discourse relations’ in which
NSUs appear to occur routinely (e.g. correction and question-answer pair), effectively excluding discourseinitial NSUs from my discussion. The production of discourse-initial NSUs is presumably not different
from that of discourse-internal NSUs except that in the former the utterer speculates on the hearer’s
proposition using non-linguistic clues. For instance if the hearer is a taxi driver, the utterer (B), whose
proposition is [Event GO ([Thing B], [Path TO ([Place NEWTOWN])] (roughly ‘B goes to Newtown’) may calculate
that the taxi driver’s proposition is something like [Event GO ([Thing B], [Path TO ([Place x])] (roughly ‘B goes
to x’) and say Newtown.
These LCSs are based on the tentative analysis of the semantics of eat in Jackendoff (1990, 253). Details
(tense etc.) that are not relevant to the argument here are omitted from these simplified LCSs.
In her augmentation of LCS, Dorr (1993, 169) introduces such primitives as WH-MANNER, WH-THING,
and WH-TIME to equip LCS with the capability to handle wh-questions (x-interrogatives). The x here
corresponds to Dorr’s WH-THING.
OSLa volume 1(1), 2009
 
pirafuf and Uzbek palovf .14
This section recapitulates the composite model of NSU production and discusses
the implications that it has for linguistic theory.
As is clear from the chart presented in section 2, the composite model comprises three main functions/mappings, namely the function that assigns certain
components of a propositional structure to morphemes, which are then identified
as f ’s (correspondence 1), the mapping between f ’s and m’s, i.e. word-formation
(correspondence 2), and the mapping of f ’s and m’s onto a linear structure (correspondence 3). In figure 1 in section 2, these correspondences are shown with
c1, c2, and c3 in squares. The point at which the composite model departs most
radically from the constituent-omission model is thus the role of syntax in NSU
production. In the constituent-omission model, meaning (proposition) is mapped
onto a syntactic structure, following which ellipsis takes place within that structure (see analyses in Merchant 2004). In the composite model, on the other hand,
correspondences c1 and c2 restrict the role of syntax in NSU production to often
very simple linear alignment of words consisting of f ’s and m’s. The differing processes of utterance production in the constituent-omission model and that in the
composite model can be schematised roughly as follows:
Constituent-omission proposition > syntax > morphemes
Composite proposition > morphemes (> word-formation > syntax)
This order in which the three mappings take place means that the linear alignment of f ’s and m’s often does not necessitate the involvement of an elaborate
syntactic structure. In the case of the production of (4) in Turkish, Japanese, Mongolian, and Bukharan Tajik, there needs to be only one constraint, namely ‘verbfinal’, for (4) to be produced. In highly polysynthetic languages, correspondences
c1 and c2 may suffice for the production of (4) because, in the composite model,
syntax does not participate in NSU production unless f ’s and m’s constitute multiple syntactic constituents. However, I would like to stress that I have no intention
of defending the idea that the role of syntax is always highly restricted in NSU production. After all, the production of many of the example utterances in this paper
(particularly (4) in German or English) clearly calls for the involvement of syntax.
One consequence which the morpheme-based nature of the model brings about is
that the model allows for the production of NSUs that do not constitute syntactic
constituents. One example of such NSUs appears in the following dialogue taken
OSLa volume 1(1), 2009
Though the grammars of a number of other languages also do not require any ms to accompany the
focus, this does not mean that all grammars of natural languages require no ms for (6b). (Indeed, the
prescriptive sentence grammars of all of these languages require the occurrence of some m’s for (6b).)
 
from a Japanese cartoon.
Jinsei yarinaos-e-mas-u-ka-ne.
‘Can I start (my) life all over again?’
–masm -enf
‘(You can)not.’
(8b)15 consists of two bound morphemes.16 While it is admittedly a rare occurrence for NSUs to consist exclusively of bound morphemes, such NSUs can occur
and make perfect sense. Their occurrence/production hence demands a theoretical explanation. The composite model provides one such explanation: the
grammar of highly colloquial Japanese requires only one m, namely -masm to accompany the focus -enf .17
The abbreviations used in this section are Q: interrogative, SFP: sentence-final particle, INTENTION: intentional mood, AHON: addressee honorifics, PROG: progressive aspect.
This is by no means an isolated example of utterances consisting only of bound morphemes. An Uzbek
informant provided the following constructed dialogue in which B’s utterance is an NSU that consists of
one bound morpheme:
s/he leave-PROG-3SG-Q
‘is s/he leaving?’
‘(s/he) intends (to)’.
The grammar of less colloquial Japanese would call for the occurrence of more m’s, which would result
in a longer (8b), e.g.
yarinaosm -em -masm -enf
‘(You) cannot start (your life) all over again’.
OSLa volume 1(1), 2009
 
Thus, one theoretical consequence that Principle 1 and Principle 2 bring about
is the non-existence of ‘syntactically full’ sentences in the process of NSU production.18
Such a process in which NSUs are produced appears to fit the psychological
reality of a number of speakers more readily than a process where components
of utterances are ‘omitted’ or ‘inaudible’, though admittedly the speakers’ intuitions are, in the absence of any empirical evidence, open to dispute. In summary, the composite model is based on the simple principle of the necessary minimum, namely ‘for an utterance to be sensical, (only) foci and morphemes that are
grammatically required to accompany them necessarily occur’. This simple principle 1) accounts for the production of NSUs that native speakers variably ‘reconstruct’, 2) explains why in certain contexts pro-drop cannot occur in languages
that have subject-verb agreement morphology, 3) models the production of NSUs
without devising separate production processes for ‘ellipses’ and ‘fragments/subsentences’, and 4) predicts what constituents have to be present in a given NSU.
The composite model keeps the involvement of syntax in NSU production to a
minimum, which in turn may reduce the (mental) cost of producing NSUs.
Barton, E. 2006. Toward a nonsentential analysis in generative grammar. In L. Progovac, K. Paesani, E. Casielles and E. Barton (eds.), The syntax of nonsententials,
pages 11–31, Amsterdam: John Benjamins.
Bavin, E. 2000. Introduction: a functional approach to ellipsis. Linguistics 38(3),
Brown, K. and Miller, J. 1991. Syntax 2nd edition. London: Harper Collins Academic.
Clapp, L. 2005. On the interpretation and performance of non-sentential asser[18]
For example, the ungrammaticality of (b) in the following example, which Merchant (2004, 676-705)
claims to be ‘expected under the ellipsis analysis, since the distribution of case morphology on DPs will
be regulated by the same mechanism in both elliptical and non-elliptical contexts’ (Ibid.: 679), is also
expected under the present analysis, not because (b) is elliptic but because, while Yongsu is an f , -rul is
neither an f nor an m. This analysis is also consistent with the fact that (c), which consists of one f , is an
acceptable answer to the question.
(i) Korean Q:
OSLa volume 1(1), 2009
ku chaek-ul sa-ss-ni?
who-NOM this book-ACC bought
 
tions. In R Ellugardo and R. J. Stainton (eds.), Ellipsis and nonsentential speech,
pages 109–129, Dordrecht: Springer.
Culicover, P. W. and Jackendoff, R. 2005. Simpler syntax. Oxford University Press.
Dorr, B. J. 1993. The use of lexical semantics in interlingual machine translation.
Machine Translation 7, 135–193.
Ellugardo, R. and Stainton, R. J. 2005. Introduction. In Ellipsis and nonsentential
speech, pages 1–26, Dordrecht: Springer.
Enç, M. 1986. Topic switching and pronominal subjects in Turkish. In D. I Slobin
and K Zimmer (eds.), Studies in Turkish linguistics, pages 195–208, John Benjamins.
Gunter, R. 1974. Sentences in dialog. Columbia: Hornbeam Press.
Jackendoff, R. 1990. Semantic structures. Cambridge, Massachusetts: The MIT Press.
Lyons, J. 1977. Semantics volume 2. Cambridge University Press.
Malmkjær, K. 2002. The linguistics encyclopedia 2nd edition. Routledge.
Matthews, P. H. 1997. The concise Oxford dictionary of linguistics. Oxford University
Merchant, J. 2004. Fragments and ellipsis. Linguistics and philosophy 27, 661–738.
Napoli, D. J. 1996. Linguistics: An Introduction. Oxford University Press.
Progovac, L. 2006. The syntax of nonsententials: Small clauses and phrases at the
root. In L. Progovac, K. Paesani, E. Casielles and E. Barton (eds.), The syntax of
nonsententials, pages 33–71, Amsterdam: John Benjamins.
  
Shinji Ido
Tohoku University
Research Center for Language, Brain and Cognition
41 Kawauchi, Aoba-ku, Sendai
[email protected]
OSLa volume 1(1), 2009
OSLa volume 1, 2009
Behrens & Fabricius-Hansen (eds.) Structuring information in discourse:
the explicit/implicit dimension, Oslo Studies in Language 1(1), 2009. 77-108. (ISSN 1890-9639)
 ,  ,   
University of Göttingen
In this paper, I will propose a uniform analysis for adjectival and certain
instances of adverbial eigentlich. The analysis rests on the assumption that
eigentlich contrasts the nominal content of a concept C with a contextually
given notion of phenomenological evidence for C-hood. eigentlich offers the
semantic frame to refer to two ways of C-hood which are usually supplied
by context. In the adjectival use, the discourse content should provide an
N exemplar in the true sense (an eigentliches N) along with an apparent N,
thereby proving the two notions to be different. In the adverbial use, the
message conveyed is usually that the actual world nominally satisfies some
proposition p (eigentlich, p) while the actual world looks as if a contrasting
proposition q were the case. The analysis improves on earlier accounts in
German descriptive linguistics in that it offers a fully compositional account
of the semantic and pragmatic contribution of eigentlich in a wide variety of
constructions, including focus, contrastive topic and questions. The analysis
proposes a delineation of eigentlich as an emotive marker which differs from
the content use in prosody, syntax, focus sensitivity and meaning.
The German word eigentlich can be used in a range of contexts, with meanings
and implications that are hard to delimit. Linguists’ interest was first drawn to
eigentlich as a discourse particle, like in (1) (where eigentlich is to be read without
accent). Its contribution is hard to translate into English in such examples, and
it is equally difficult for native speakers of German to define or paraphrase the
meaning of the word.
Was willst Du eigentlich hier?
what want you eigentlich here
This paper is a substantially revised version of Eckardt (2006), presented at the Oslo SPRIK Colloquium in
2006. Due to valuable comments of two anonymous reviewers, the analysis in the present paper differs
substantially from the one proposed there. My eigentlich project rests on joint work with Angelika Port
whom I need to thank for many insightful criticisms and her continuous challenge with real data. All
unclarities and faults in the analysis are in my responsibility.
 
‘what do you want here after all / at all / anyway / ... thinking about
Da hast Du eigentlich recht.
there have you eigentlich right
‘You are right after all / thinking about it / to tell the truth’
We will approach these uses indirectly, via the meaning of the intuitively more
contentful, more graspable stressed adjectival and adverbial use. Let me start by
listing some of the facts about the use of eigentlich. The sentences in (2) offer some
examples for adjectival eigentlich, which typically, but not exclusively, occurs with
an accent. Small caps indicate accents.
Der  Chef ist verreist.
‘the real boss is on a trip’
Der  Mörder war Smith.
‘the actual murderer was Smith’
Das  Problem ist seine Faulheit.
‘the essential problem is his laziness’
The English translations illustrate the possible range of paraphrases. Given that
there is no single translation that would match all uses, a precise semantic analysis might even be of practical interest. I will use eig to stand in for eigentlich in
translations and glosses. Another fact about eigentlich is that the adjective cannot
occur in predicative use. Hence, examples like in (3) are all ungrammatical.
a. *Das Problem war eigentlich.
*the problem was eig
b. *Der Mörder wurde eigentlich.
*the murderer became eig
Moreover, there is no antonym to eigentlich that can be morphologically derived
in a transparent way. Isolated exceptions can be found in expert languages where
eigentlich has a specific, theory-internal meaning. (4-b) lists examples.
a. *Das ist ein / der uneigentliche Garten.
this is a / the un-eig
b. un-eigentliches Integral (mathematics),
un-eigentliches Kompositum (linguistics)
One striking observation about eigentlich is a definiteness effect. In positive contexts, it can only be used in definite NPs. Indefinite uses are restricted to negative
contexts, like in (6) (few or rarely cannot license indefinite + eigentlich), and quanOSLa volume 1(1), 2009
  eigentli
tificational uses like in (7) are not allowed in Standard High German.1
Der eigentliche Chef kommt nur dienstags.
‘the eig boss only comes on Tuesdays’
b. Die eigentlichen Bewohner sind gerade verreist.
‘the eig inhabitants are just away’
‘The real inhabitants are presently on a trip’
c. *Ein eigentlicher Chef kommt nur dienstags.
‘an eig boss …’ (even if there are several bosses.)
d. *Ein eigentlicher Bewohner hat gerade das Haus verlassen.
‘an eig inhabitant …’
Ein eigentlicher Vertrag wurde nicht abgeschlossen.
‘there was not made any eig contract’
Niemand hatte einen eigentlichen Lösungsvorschlag.
‘no one had any eig proposal how to solve the problem’
a. *Die meisten eigentlichen Stadträte wohnen im Süden.
‘most eig senators live in the South’
b. *Einige eigentliche Spieler traten am Samstag an.
‘some eig players came on Saturday’
Complex nouns of the form ‘eigentliche(r/s) N’ share definiteness effects of superlatives. There seems to be a notion that the individual that is the eigentliche N is a
unique (single or plural) individual. In section 3, I will take a closer look at the
semantic rationale that could motivate this fact about the use of eigentlich.
Finally, there is a strong tendency to use eigentlich in the sense of true / real
/ actual / essential with a stress. Stressed uses will invariably express a semantic
contribution in one of these senses. This sets eigentlich apart from ordinary adjectives or adverbs where pitch accent and meaning are usually independent. Earlier
authors have suggested that the stress on eigentlich has some kind of motivation
(see Frühwirth 1999). They propose that some kind of contrast is evoked without,
however, offering any concise analysis of the prosodic facts. An analysis should
also allow for rare unstressed adjectival uses like in (8).
Der eigentliche C an der Sache ist aber, daß GM die Renovierung
auch noch zahlt.
‘the eig  about the thing is, however, that GM will even pay for the
GOOGLE searched data show convincingly that rare positive hits of the kind in (7) offer clear evidence
for an interesting dialectal variation between Standard German and Swiss German.
OSLa volume 1(1), 2009
 
The example in (8) (a real online quotation) did not contrast the true trick with
other, minor advantages of some plan. Most examples, however, and certainly
those that I have listed so far, will occur most naturally in a discourse where
eigentliches N contrasts with objects or persons that are not real Ns even though
you might think so. The above examples should be read with the following trends
and generalizations:
• If eigentlich occurs in positive definite contexts, it is most naturally read
with a pitch accent. Presented with this accent pattern, hearers/readers
will reconstruct a suitable preceding discourse where the eigentliche N is
contrasted with something else. We will take a closer look at the nature of
this contrast in section 3.
• In negative contexts, as in (6), eigentlich needs no accent.
• In an example like (8), eigentlich carries no stress. Nevertheless, the use of
eigentlich suggests a contrast to other things – in the example, other positive aspects about the renovations – even though these may not have been
explicitely listed in the actual text. (If you remove eigentlich in (8), the resulting sentence suggests that the payments are the one and only charming
aspect of the renovation plan.)
So far, we have restricted attention to adjectival eigentlich. In the adverbial use,
the role of accent is even more important. The presence or absence of the accent
makes a clear distinction between uses in a sense that closely corresponds to truly,
really, in fact, and the discourse particle use. This latter use is intuitively different,
as the paraphrases in (9) illustrate.
Wie heißen Sie ?
‘what is your real name?’
Wie  Sie eigentlich?
‘what’s your name, by the way?’
It is all the more necessary to understand the role of accenting, literal contribution and pragmatic effects of adverbial eigentlich. In delimitating the semantic
analysis of eigentlich, I will proceed from clearer to more sophisticated cases. I
will start by devising a semantics for the uses of adjectival eigentlich, including an
analysis for the accent pattern. I will propose an account for unstressed adjectival uses in positive and negative contexts. Finally, stressed adjectival uses will be
analyzed in their semantic and pragmatic dimensions. Against this background,
I will turn to adverbial unstressed eigentlich, which serves to annotate an utterance with a certain speaker attitude. It can be hypothesized how this use arose
as a generalization of side messages of stressed eigentlich. In view of a very nicely
OSLa volume 1(1), 2009
  eigentli
working semantic analysis of stressable eigentlich, I will propose that adverbial
unstressed eigentlich belongs to the category of discourse particles, and should be
distinguished from stressable eigentlich.
 
A wide range of authors have approached the meaning and use of eigentlich from
a descriptive perspective in German linguistics. While I will not review all these
in detail, there are several main issues that delineate different positions in this
(i) Are there one or two homophonous adverbial eigentlich?
(ii) Does its effect lie in weakening or strengthening the assertion?
(iii) What is the role of its contrasting function?
(iv) In questions: Does it always convey casualness, friendliness?
The most important question is whether we should distinguish adverbial eigentlich
from a homophonous discourse marker. Intuitively, it seems clear that two opposed uses should be distinguished, differing in grammatical as well as semantic
properties. On the one hand, we find discourse particle uses of eigentlich where
nothing except a certain “flavor” is added to the assertion. Diagnoses vary as to
what the exact nature of this flavor should be, but it seems hard to give any clear
paraphrase (or English translation). (10) offers an example.
Ich mag ihn eigentlich.
I like him eig
‘thinking about it a bit, I’d say that I like him in fact’.
Such uses usually occur sentence internal or final. Eigentlich is not stressed. On
the other hand, we also find eigentlich in stressed uses, and in topic positions, two
features that discourse markers do not normally allow (see Meibauer 1994). Such
uses moreover convey a more graspable message, something like “in truth”, “in
actual fact”, “really”.
 heisse ich Max.
be-called I Max
‘It’s true, my name is Max’
Even though no author so far could define a clear demarcation between uses like
in (10) and uses like in (11), the poles of the continuum of usages are distinct
enough to warrant an ambiguity debate. However, only those few who are interested in a formal semantic analysis (notably Schmitz and Schröder 2004; Schmitz
OSLa volume 1(1), 2009
 
2008) adopt the assumption of a “discourse particle” at one end and the conceptually richer adverb at the other
A majority of authors (Weydt 1977; Thurmair 1989; Frühwirth 1999) however
assume that only one entry eigentlich is sufficient. They point out that the proposed grammatical properties of discourse markers (in Meibauer 1994) should
be seen as prototypical features rather than necessary properties. According to
these authors, eigentlich with a pitch accent, and in topic position, like in (11),
could still be a discourse marker in the sense of Meibauer’s classification. Their
main point in favor of a single lexical item eigentlich is parsimony in semantic
modeling. They point out that all uses of eigentlich rest on one common underlying semantic core (without, however, specifying that core clearly). In addition,
they stress that no morpho-phonological dissociations can be traced that would
support two different lexical entries.
Let us address the speculations about the semantic core of eigentlich in some
more detail. Most traditional approaches characterize the contribution of eigentlich
in terms of “weakening” or “strengthening” the assertion. This is puzzling not
only in that it remains unclear in what sense an assertion can be strengthened or
weakened. It is also puzzling in that different authors (and sometimes even one
and the same) diagnose eigentlich as both weakening and strengthening. I will
offer two examples, both quoted after Frühwirth (1999), which have been categorized as a “weakening” and a “strengthening” use respectively, without further
attempts at spelling out the intuition.
 habe ich keine Zeit. (“weakening”, discussed in Kohrt 1988)
have I no time
Gehst Du heute Abend mit ins Training? – Ich habe eigentlich keine Zeit.
‘Will you join me to go to training tonight? – I have eig no time’
(“strengthening”, discussed in Weydt and Hentschel 1983)
These apparently conflicting diagnoses in terms of traditional grammar do not
add to our understanding of the meaning of the term, and urgently require elucidation.
A further observation concerns the relation between an eigentlich sentence
and its local discourse context. Frequently, the sentence is contrasted with a preceding or following utterance. Frühwirth offers examples like the following.
Ich hasse Oliven. Naja, eigentlich mag ich sie nur nicht besonders gerne.
‘I hate olives. Oh well, in fact I just don’t like them very much.’
Ja, aber eigentlich kam es mir nicht nur so vor! Ich war es tatsächlich!
‘Yeah, but in fact it not just seemed to be like that. I really was it.’
Obwohl es sehr belastend war, hat mich die Wiedervereinigung eigentlich
OSLa volume 1(1), 2009
  eigentli
Although it was very depressing, the reunification eig made me happy
The contrasting function of eigentlich is unanimously acknowledged by authors.
It is at the basis of the only formal approach to the contribution of eigentlich that
seems to exist to date, the analysis by Schmitz and Schröder (2004). They assume
that eigentlich serves to block some default inference that the listener would otherwise derive from the sentence. The sentence in (17) the prime example in their
Eigentlich geht Ostwind. Aber es regnet.
‘Eigentlich is an easterly wind. But it’s raining.’
Schmitz and Schröder observe that the assertion “there is an easterly wind” will
give rise to the default expectation (in the meteorological context of Germany)
that the weather is dry. In (17) ‘there is an easterly wind’ warrants the default
inference: ‘the weather is sunny’ (in a central German climate). The function of
eigentlich lies in signaling that this non-monotonic inference should be blocked.
The following sentence asserts a proposition that is in conflict with this default
inference. The authors offer a very elegant implementation of this idea in terms
of update semantics.
Schmitz and Schröder in fact capture an essential insight about the use of
eigentlich. The analysis leads to very reasonable predictions in many examples.
However, for the sake of a unified semantic analysis the authors disregard both
the adjectival use as well as the question whether an extra discourse marking
function must be acknowledged. The latter seems the more serious omission. Examples like (18) show that not all uses of eigentlich block default inferences:
Sollen wir zum Frühstück einen Sekt
Shall we for breakfast a
Champagne open?
b. (hesitating)Ach – heute ist ja
eigentlich Sonntag! Gut,
Ach – today is PARTICLE eig
Sunday. Well,
machen wir das.
do so.
‘Should we have Champagne for breakfast? – I don’t know … well, yes,
after all, it’s Sunday. Let’s have some.’
Du hast eigentlich recht.
You have eig
‘You are right after all.’
In (18), speaker B first contemplates the negative consequences of having champagne so early in the morning – first, its fun, but afterwards one is tired, one
OSLa volume 1(1), 2009
 
wants to take a nap, or one can not work properly. Yet, he then comes across a
fact that might still support A’s suggestion: After all, it’s Sunday. So, let us behave accordingly. In this example, no default inference of “it is Sunday” seems
to be withdrawn. On the contrary, if any default plays a role here, it is “Sundays are lazy days” which is supported, not blocked, in this context. Likewise, an
utterance like (19) is used normally to convey agreement-after-some-reflection.
The speaker states that the hearer is right. And that one should act, behave, or
decide accordingly. I will in the following label such examples as the “let’s act
accordingly”, or pensive use of eigentlich. In later sections, it will be argued that
this distinction corresponds to the following:
adv. eigentlich: in association with focus; operator on properties; stressed;
related to adjectival eigentlich;
in questions: operates on questioned property.
mp eigentlich: unstressed; speech act signal; no relation to focus;
no transparent relation to adjectival eigentlich;
let’s act accordingly uses, pensiveness;
in questions: adds the ‘after some reflection’ flavor
At the present point, the data simply show that Schmitz and Schröder’s uniform
analysis of adverbial eigentlich does not as yet cover all data, and that the analysis
moreover does not offer any clue as to how the prosody and word order facts of
examples can be turned into a prediction about the nature of the examples in
question. Do they represent an eigentlich – but use (i.e. in line with their analysis)
or an eigentlich – let’s act accordingly use (and hence unpredicted). Importantly,
Schmitz and Schröder acknowledge an extra discourse marking use in questions
like (20)
Wie heißen
Sie eigentlich?
How are-named you eig?
‘what’s your name, by the way?’
They briefly characterize the function of eigentlich as marking a casual, friendly
question. I will come back to this claim when we deal with the discourse marking
function of eigentlich.
                                  eigentli
An appropriate analysis of adjectival eigentlich should explain the definiteness effects, the fact that the adjective is non-intersective, its preference for pitch accent
as well as the contrastive function that has been repeatedly observed in grammars,
dictionaries and the literature. Let us take a closer look at this function, which
OSLa volume 1(1), 2009
  eigentli
constitutes the core contribution of the word.2 In all cases, there is a feeling that
the speaker wants to contrast the “real stuff” to something which, even though it
might look similar, is not the “real stuff”. So, example (2.c) refers to das eigentliche
Problem (= ‘the eig problem’) and suggests that other facts in the given context
could constitute ‘minor problems’.
Examples like (2.a) and (2.b) are similar, though slightly different. If one calls
someone the eigentliche Mörder (‘eig murderer’), one rarely refers to circumstances
where more persons hurt the victim in minor ways. Usually, the speaker refers to
hypothetical or apparent murderers, persons which were for some time hypothesized to be murderers but turned out not to be. The opposite of eigentlicher Mörder
hence is scheinbarer (apparent, seeming) Mörder. Comparing the meaning of Mörder
and eigentlicher Mörder, it turns out that the two are co-extensional. Murderers are
true murderers, and one cannot more truly kill someone than by being a murderer
(eigentlich or not).
〚eigentlich + murderer〛= 〚murderer〛
Similar observations hold for Chef and eigentlicher Chef (= boss). Persons may be
mistaken as being the ‘boss’ but reference to eigentlicher Chef suggests that from
that point on, no such “substitute” bosses should be called Chef.
The following classical quote in Adelung (1774, taken from Schmitz and Schröder
2004, 12), in turn, suggests that sometimes the speaker wants to distinguish between a “nominal” extension of the noun in contrast to the “true” extension of
the noun, which is a subset of the former.
Das eigentliche Griechenland, derjenige Theil Griechenlands,
the true/real Greece,
part-of Greece
welchem dieser Name der schärfsten Wahrheit nach
to-which this name the sharpest truth
acoording-to belongs
‘the true/real Greece, that part of Greece that most truly deserves that
If we imagine the geographical map of Greece, we can delineate those parts that
count as the “true, real” extension of Greece. Everything outside is “Greeceunder-a weaker-standard”, “parts that were erroneously taken to belong to Greece”,
or similar hedges. The resulting picture reflects nicely what we assume to be the
semantic contribution of eigentlich. It is used to cut a property’s extension down
to the true core.
I propose that eigentlich generally leads from a property concept C to two further semantic objects: the ‘nominal’ notion of ‘being a C’ as contrasted to the
Throughout this section, eigentlich will always be the adjective. I omitt this qualification to ease readability.
OSLa volume 1(1), 2009
 
phenomenological notion of ‘being a C’. More specifically, the phenomenological
notion usually is tied to a set of typical properties that would allow subjects to
identify an object as ‘being a C’. I will use P(C) and N(C) to refer to the two
sets of properties.
Let N be a noun and C be the property denoted by that noun. Moreover,
assume that N is used in an utterance situation s.
N(C) = {N1 , N2 …} the set of property concepts that define the extension of N. In the normal case, this can be just C; in some cases the
speaker intends to cut down C to a subset of C.
P(C) = {P1 , P2 …} a set of property concepts that the speaker in
s presents as the typical phenomenological evidence to classify an
object as having C. Good paraphrases could be “what a C object usually looks/is/behaves like”. Note that there need not be a unique set
P(C) tied to C. What is perceived as typical for C can vary from
utterance situation to utterance situation.
These two sets of properties are at the basis of adjectival eigentlich and will later
serve as the basis to reconstruct Schmitz and Schröder’s observation about blocked
default inferences in terms of a truth conditional setup. Before turning to the
semantic analysis of the complex phrase 〚eigentlich N〛, we will have to check the
contribution of eigentlich in referring to a set of objects related to N. Interestingly,
eigentliche N can be used in two senses: In one kind of use, the speaker wants to
refer to the nominal core of N. In this case, the overall referent of the NP ‘Det
eigentlich(e) N’ is characterized as being N in the ‘nominal’ sense, and not just looking like an N. In another kind of use, however, the speaker wants to present the
referent of ‘Det eigentlich(e) N’ as exhibiting all typical features of N even though
it is not in the actual extension of N in the ‘nominal’ sense.3 The two following
examples illustrate the two usages.
Frau Meier leitet die Geschäfte von Tag zu Tag. Die 
Mrs. Meier leads the business from day to day. The eig
Chefin ist Frau Schmitz.
boss is Mrs. Schmitz.
‘Mrs. Meier makes the day-to-day business decisions. The true boss is
Mrs. Schmitz.’
Frau Schmitz steht der Firma
offiziell vor.
Die 
Mrs. Schmitz stands the company officially before. The eig
OSLa volume 1(1), 2009
I want to thank an anonymous reviewer for pointing out this really crucial observation on the two kinds
of usages of eigentlich.
  eigentli
Chefin ist aber
unsere Sekretärin, Frau Meier.
boss is however our
secretary, Mrs. Meier.
‘Mrs. Schmitz is the official leader of the company. The true boss, however, is our secretary Mrs. Meier.’
Let me go through the rationale behind these two examples. Sentence (24) states
• Mrs. M fulfills functions which are phenomenologically tied to the notion
be the boss like taking decisions, leading business, etc.
• Mrs. S is the boss in the ‘nominal’ sense (for instance, in legal respects).
• Mrs. S is the eigentliche Chef.
Sentence (25) contrasts the same two ways in which people can be boss using
eigentlich in the opposite direction:
• Mrs. S is the boss in the ‘nominal’ sense (for instance, in legal respects).
• Mrs. M fulfills functions which are phenomenologically tied to the notion
be the boss like taking decisions, leading business etc.
• Mrs. M is the eigentliche Chef.
This twin pair of examples can be systematically multiplied. It shows that adjectival eigentlich can not be analyzed as always leading from look-like N to actual
N. Neither does it always serve to lead from N-by-name to N-by-typical-property.
However, it systematically serves to point out that this difference exists and that
N-by-nameand N-by-typical-property have different extensions. This is likewise
exemplified in (24)/(25) The little sample discourses contain referents which show
the difference between N-by-name and N-by-typical-property. In our example,
the extensions of Chefin-by-name and Chefin-by-function are singletons, and the
speaker states that the two singleton sets differ. In other cases, it can be sufficient to state that some referent is in the extension of one property, but not the
other. This option is exemplified in the adverbial uses (see below).
The definition in (23) introduced two sets of properties that are tied to a concept C. In the discussion of (24)/(25), I have repeatedly referred to properties simpliciter that are exhibited by some individual. I will now turn to a definition that
leads from a set of properties to the mereological sum of these. This sum will be
a unique mereological object (or, intension thereof) which will explain the definiteness effects that were stated at the outset of the paper. The definition is
stated in terms of an analysis of properties in two-sorted type theory. Moreover,
we assume that properties are attributed to singular and plural objects that are
OSLa volume 1(1), 2009
 
mereologically ordered by a relation ≤. For details, the reader is referred to Link
(1983) or other introductions to plural ontologies. It should, however, be possible
to follow the subsequent discussion on basis of an intuitive understanding of the
formal details.
Let {P1 , P2 … Pn be a set of property concepts of type (s,(e,t)). Let the domain of objects be mereologically ordered ≤. Let S(X) be the supremum
object of a subset X of the domain of entities. The sum of the properties
is defined as
{P1 , P2 … Pn } := λwλZ[Z = S({P1 (w), P2 (w) … Pn (w)})]
Note that in each possible world w, {P1 , P2 … Pn }(w) is either the bottom element of the mereology or else a unique entity. This entity can be a true group
object (i.e. a plurality of atomic objects) or an atomic object (i.e. what one would
normally consider an individual entity). The underlying function of this object
seems to be, that this object (in context) uniquely represents the conjunction of
exactly the listed properties. If the sum object is a plurality of atomic objects,
importantly, none of the atomic objects in isolation will represent the sum, even
though of course they all have the conjoined properties. This underlying rationale
leads to the definiteness effects for eigentlich. Before moving on, I want to provide
a prose paraphrase for the sum operation in (26). { P1 , P2 … Pn } can be read as
“that unique property concept that, for each possible world, yields the maximal
group of entities who have properties P1 and P2 and … and Pn .”
The next step in the analysis of adjectival eigentlich will consist in offering a
semantic value for ‘eigentlich(e) N’ on the basis of the ingredients that have been
introduced so far.
[[eigentlich N]] ∈ {(N[[N ]]), P([[N ]])}
Support by discourse context: should entail (N[[N ]]), 6= P([[N ]])
for example by naming an x < (N[[N ]]) such that ¬(x < (P[[N ]]))
or by naming
a = (N[[N ]]), b = (N[[N ]]) where a 6= b or by entailments that
follow from what the speaker is currently talking about.
In prose, this analysis consists of the following steps:
(i) The noun N has a reliable interpretation, and can be interpreted in a literal
sense. (Hence, this analysis does not rest on the idea of ‘different ways to
understand the word N’.)
(ii) The literal sense of N is tied to typical properties that could be used to identify N objects (P(〚N〛)).
OSLa volume 1(1), 2009
  eigentli
(iii) The literal sense of N is also tied to N(〚N〛). In practically all cases, N(〚N〛)
is just the singleton set of 〚N〛.
(iv) Summation: In each world, (P[[N ]]) is the singleton set that contains
the plurality that exhibits all and exactly the properties in P(〚N〛). If an
object a is in the extension of N but not part of this plurality, then a is an
object that fails to show some typical property of N-hood. If an object is part
of the plurality but not an N-object, it could be characterized as a “fake” or
“seeming” or “apparent” N.
(v) Summation: In each world,
are N.
(N[[N ]]) is the single plurality of things that
(vi) The complex nominal eigentlich(e,er,es) N can denote either one of the two
If the plurality is not the bottom element, then it is unique and requires definite
reference. Let us see some examples.
(…) Die  Schüler kommen im Oktober.
‘the eig pupils arrive in October ’
*Ein paar eigentliche Schüler stehen vor der Tür.
some eig pupils
stand before the door
‘there are some eig pupils at the door’
In (28-a), the noun Schüler leads to contrast the plurality of ‘nominal’ pupils to
individuals that could seem to be pupils. Most naturally, the preceding context
will have introduced the seeming pupils, i.e. individuals in P(〚Schüler〛),
whereas the sentence refers to N(〚Schüler〛), the plurality of pupils in the
literal sense. This is a unique plurality, hence the failure in (28-b) to use an indefinite. Note that a partitive construction will be perfectly acceptable: Einige der
eigentlichen Schüler stehen vor der Tür ‘Some of the eig pupils stand before the door’
can be used to refer to a subgroup of the maximality “the true pupils”.
Applied to (25)/(26), the analysis allows the NP “die eigentliche Chefin” to refer
to the unique individual which shows all typical properties of bosses, and contrast
it with the ‘nominal’ boss. However, it also allows the NP “die eigentliche Chefin”
to refer to the unique individual who is the ‘nominal’ boss, and contrast it with
another person who shows typical properties of the boss. Both applications share
the entailment that in some sense “the boss is not as she ought to be”. This is intuitively a valid prediction.
What does the proposed analysis have to say about the pitch accent on eigentlich?
Crucially, the proposed analysis is a choice analysis in that ‘eigentlich(e/r/s) N’ refers
to one of two possible interpretations (N[[N ]]), (P[[N ]]). Given that the
OSLa volume 1(1), 2009
 
choice is left to context, eigentlich does not have a constant meaning that could
be contrasted with other meanings. In context, however, eigentlich will either be
interpreted as the function fNOM that maps a property C to N(C), or else it
will denote the function fPHÄN that maps properties to P(C). In either case,
we can assume that 〚eigentlichF 〛f = {λC. N(C), λC. P(C)}. That is, the contrast between two different ways of determining the extension of C is the constant
element in the meaning of eigentlich. This view would warrant the pitch accent
on eigentlich as a contrastive focus accent. However, some refinements seem to be
While the above analysis accounts for the pitch accented examples in a natural
way, it does not account for the non-accented use of the adjectival eigentlich This
is not only so in negated examples but also in examples like (8). I therefore think
that the accented cases in fact come about by an interaction of conspiring factors.
The use of eigentlich typically takes place in discourse when the notion N is under
debate and hence discourse-given. Under such circumstances, speakers evoke a
contrast between eigentliche N and other kinds of N, as captured in the semantic
analysis. Contrastive focus, hence, lies on the entire NP but in view of the fact that
N is discourse-given, the AF principle relegates the accent on the adjective
(Schwarzschild 1999). This leads to the impression that the adjective carries a
focus accent but in fact, it is the full NP that is in contrastive focus. This analysis
works better for examples where no NP referents are contrasted:
Hans hatte einige Zuhörer, aber eigentliche  waren nicht
Hans had some listeners, but eig
fans were not
‘John had some listeners but there weren’t any true fans present’
Der eigentliche  der Sache ist, daß …
‘The eig appeal of the matter is that …’
An example like (29-b) in the given intonation could be uttered as a final argument in favor of “the matter” without other charming aspects having been explicitly mentioned before. After all, the use of charm for non-human referents is
already non-literal. However, the utterance cleverly evokes the impression that
other seemingly charming aspects are at the back of the speaker’s mind, and that
the positive aspect to be mentioned now just adds the true positive aspect of the
matter as a kind of climax. Hence, the utterance in (29-b) effectively conveys that
the matter has a lot of positive properties without actually listing them (as revealed by the accent pattern). In conclusion, the analysis offers a good basis for
the prosodic behavior of adjectival eigentlich NPs. I will leave the details to be
filled in, and return to semantic matters.
OSLa volume 1(1), 2009
  eigentli
The proposed analysis is open for a wide variety of instantiations of P(〚N〛)
and N(〚N 〛). At first sight, this appears a disadvantage because the analysis is
not very specific. However, in view of the wide variety of possible uses in real discourse, this flexibility might be an advantage. Consider a case like (30), discussed
in Port (2006):
Der  Garten ist hinter dem Haus.
‘the eig garden is behind the house’
Port argues that (30) can be uttered in view of a small patch of lawn in front of the
house. While such patches could already count as ‘garden’, sentence (30) signals
that the speaker will use the word ‘garden’ in a stricter sense. In this example,
it is hard to determine the status of different notions of garden that are at stake.
One could assume that eigentlicher Garten (‘true garden’) is used in the sense “garden with all relevant properties” and contrasted with what ‘nominally’ could be
classed as garden. In this analysis, the nominal extension of Garten can comprise
little patches of green as well as more extended gardens, and the eig garden is a
restricted form of ‘all gardens’. Alternatively, however, the speaker could be intending to say “in a certain phenomenological sense, this little patch could fulfill
the criteria to be a garden – a fenced piece of lawn in front of a house. Yet, the
nominal sense of Garten includes a functional element and refers only to that piece
of lawn that is functionally assigned to the house as being its garden. This sense
of Garten does not apply here”. However, both analyses have in common that the
speaker points out that we could use Garten in two senses, and demonstrates the
difference on the basis of discourse referents. This is exactly the intuition underlying examples like (30). Moreover, the analysis predicts that the speaker believes
that the word Garten does have a literal sense, and that sentences like (30) are not
used to point out a language error. Again, this is a valid prediction. The extra
level of complexity that remains in example (30) is the fact that Garten, unlike
pupil or boss, does not have a fixed legal or institutional definition.
To conclude the discussion of adjectival eigentlich, the analysis predicts that
it does not make much sense to negate eigentlich. The only reasonable denota∑
tion for un-eigentliches N could be the opposed semantic value in { (N[[N ]]),
(P[[N ]])}. Given that eigentlich has the choice to pick up either of the two,
uneigentlich simply provides a morphologically more complex way to express the
same thing. Likewise, it does not make much sense to use un-eigentlich to name
the “second meaning” after the first one has been chosen. If used in this way, ‘un-’
would not have a constant interpretation but serve to express something like the
other operator. This is not an acceptable form of negating a concept (see e.g. Horn
2002). Only in those rare cases where the adjective is used context-independently
to map a property N to a sub-property can un- apply. In this case, the prefixed adjective constantly refers to the complementing sub-property. Examples comprise
OSLa volume 1(1), 2009
 
cases like eigentliches/uneigentliches Integral, and a few more as mentioned above.
 
While most other studies on eigentlich address only its adverbial use, the present
study deliberately spent time on a close investigation of the adjectival case. This
effort will pay off in the present section. I will argue that the adjectival use can
be adapted to the adverbial case with minor changes. At the end of the section, I
will briefly point out why other possible analyses, not inspired by the adjectival
case, do less justice to the data.
[4.1] Stress = contrastive topic accent
The first important observation concerns the prosodic facts of the adverbial use.
It was mentioned in section 2 that adverbial eigentlich can occur stressed or unstressed, and that authors tend to correlate this difference with two possible uses
(which will be confirmed in the present analysis). It has, to my knowledge, never
been pointed out that the stress of stressed eigentlich (as in (31)) can not be a simple focus accent. In fact, (31) with a single accent on eigentlich is prosodically and
pragmatically ill formed (see (32)). The hearer perceives but cannot interpret the
single accent. A more appropriate prosody is given in (33) where a second accent
occurs on keine (‘no’); on the interpretation of accent contours see Büring (2007)
and references therein.
Eigentlich habe ich keine Zeit.
have I no time (no accent pattern)
single HL* accent is illformed:
*E habe ich keine Zeit.
contrastive topic L*H accent is acceptable:
/ habe ich \ Zeit.
This suggests that adverbial eigentlich can associate with a second focus, and that
an analysis in terms of contrastive topic would be more faithful to the data than
one in terms of simple focus. I will completely disregard the option of an uninterpreted “lexical” accent here. Let us investigate how different loci of the second
accent influence our understanding of the overall message conveyed.
[4.2] eigentlich in association with focus
Sentences like (33), as well as (34) below, strongly suggest certain kinds of context
of use. They echo situations that we all know too well, and we would strongly tend
to read them in the prosodic pattern that seems most natural in these situations.
Hence, the literature contains no discussion of the fact that other accents would
be possible in other situations.
OSLa volume 1(1), 2009
  eigentli
Eigentlich heiße
ich TF . (But everyone calls me “Ede”)
am-named I Thomas
Let us take the minimal pair in (35)/(36) as starting point to investigate accentuation.
Eigentlich F
ich gerade.
take-a-shower I now
Eigentlich dusche
F gerade.
take-a-shower I
Both sentences have a ring of protest to them and suggest possible continuations
which spell out this undertone. However, different continuations are natural for
either example.
Eigentlich F ich gerade… but there are certain suggestions p around
that seem to be based on the assumption that I am not taking a shower.
E.g. p = You ask me to answer the phone.
Eigentlich dusche F gerade… but there are certain propositions p that are
more coherent with someone else taking a shower. E.g. p = Tom occupies
the shower.
The observations in (35) and (36) are hard coherency facts. Violations are as bad
as coherency violations can ever be. A cross-change of the continuations as in
(35) and (36) yields clearly incoherent discourses.
# Eigentlich F ich gerade. And/but Tom occupies the shower.
# Eigentlich dusche F gerade. And/but you ask me to answer the phone.
An adequate analysis of the meaning of eigentlich needs to be able to predict these
differences in coherency, and hence has to take focus into account. Note that
the default inference blocking analysis by Schmitz and Schröder (2004) correctly
predicts that the acceptable cases are acceptable (default inferences of S are contradicted by next sentence). Yet, it cannot explain why the nature of possible
contrasting propositions is influenced by focus structure.
[4.3] Adverbial eigentlich: What holds and does not hold true in wo ?
I will propose an analysis for adverbial eigentlich that raises the meaning of the
adjective to the propositional level. Adverbial eigentlich takes the proposition p
expressed by the sentence as its argument. Once again, it refers to two propositions that can be derived from p:
• N(p) = those worlds where p (the proposition expressed by the sentence)
OSLa volume 1(1), 2009
 
holds true
• P(p) = conjunction of propositions {Q1 , Q2 … Qn } which are normally indicative for p-worlds.
The definitions are slightly simpler than those in section 3, refraining from sum
formation. This is due to the fact that adverbial eigentlich shows no definiteness
effects. The semantic contribution of eigentlich is to state that the two sets of
worlds N(p) and P(p) differ. This difference is evidenced by the real world
wo which is an element of one, but not the other set.
Eigentlich heiße
ich Thomas. Aber jeder
nennt mich Ede.
named-am I Thomas. But everybody calls me Ede.
‘My name is Thomas really. But everybody calls me Ede’
The first sentence in the discourse in (37) makes two statements, actually. First
it states that wo is in the proposition p = ‘my name is Thomas’. Second, it states
that wo is not in the proposition P(p). Hence we know that at least one of
the typical indicative circumstances that normally come along with ‘my name is
Thomas’ does not hold true in wo . Typically, a following aber (‘but’) sentence will
make it clearer which of the Qi is false in wo . In the present example, Qi = ‘everyone
calls me Thomas’ as a typical side effect of a person’s name being Thomas. The
content of the ‘but’ sentence entails the negation of Qi . We have hence effectively
recast the analysis of Schmitz and Schröder (2004) in a truth conditional setting.
Note that eigentlich denies the truth of evidence for p, not default inferences
simpliciter. This is made clear by the following incoherent discourse:
Eigentlich ist Dani kahlköpfig. #Aber Dani ist eine Frau.
is Dani bald.
But Dani is a
‘(*In fact, Dani is bald. But Dani is a woman.)’
It is certainly true that bald persons are typically male persons. Hence, the second sentence contradicts a default inference of the first sentence, and can serve to
block default inferences (Schmitz and Schröder do not specify a particular framework of default logic; the model theoretic reconstruction of non-monotonic reasoning in Eckardt (1999) reveals some of the assumptions about normality and
defaults). The example in (38) should be well-formed, yet this prediction is not
borne out. In terms of the truth conditional analysis presented here, we can state
that baldness is not one of the phenomenological features that are used to identify
male persons. The eigentlich in the first sentence tells us that one in a set of characterizing or defining propositions Qi does not hold true. The second but clause
is supposed to spell out which one. And ‘Dani is female’ is just not suited to do this,
because ‘Dani is male’ is not a good phenomenological clue to single out persons
OSLa volume 1(1), 2009
  eigentli
who have no hair.
I want to end this subsection by pointing out a possible variant of the analysis. In principle, the analysis would allow the same choice of uses of eigentlich for
propositions like the adjectival case.
wo ∈ N(p) and wo 6∈ P(p)
wo ∈ P(p) and wo 6∈ N(p)
I am not sure, however, how prominent the second option is in actual practice.
Attempts to create examples of this kind sound somewhat artificial and forced,
which is not the case for the adverbial. I will therefore leave it open for the
moment whether this second use really exists, and will confine myself to sideremarks when such a second option might occasionally do better justice to the
[4.4] The pragmatics of prosody
In section 4.2 it was argued that the prosody of the sentence can serve to further
narrow down the expectations as to what kinds of typical circumstance Qi might
be violated. Moreover, we observe that uses of eigentlich often come along with
a L*H accent that is typically used, in German, to indicate contrastive topics. In
this section I will show that these prosodic patterns are indeed meaningful and
receive their normal pragmatic interpretation. My analysis will rest on Büring
(2003) where an analysis of complex pitch accent patterns in terms of contrastive
topic and focus is proposed. We will use the German variant of this analysis, which
differs from the English version in that the contrastive topic accent is a L*H (simpler than the English counterpart) and always has to precede the focused constituent. The analysis reveals that contrastive topicalization presents an utterance as part of a strategy where a topical question is answered by addressing several sub-questions in turn. Hence, an utterance like die WEIBlichen (L*H) Popstars
trugen Kaftans (HL*) (‘the female pop stars wore caftans’) answers the sub-question
‘What did the female pop stars wear?’ and relates to the contrasting sub-question
‘What did the male pop stars wear?’ Both utterances together serve to settle the
discourse topic ‘What did the pop stars wear?’. The information structure of the
utterance is as follows: Die weiblichenCT Popstars trugen KaftansF where CT and F
both receive a systematic semantic/pragmatic interpretation. I will use the interpretations proposed in Büring (2003); the reader is referred to the article for
The most important step consists in understanding the semantic components
of a sentence with adverbial eigentlich, a CT accent on eigentlich and a focus accent
elsewhere in the clause. It will turn out that the CT construction creates a clearly
defined logical tie between the eigentlich-clause and the following sentence, usually referred to as the “contrasting sentence”, and also the sentence which, in
OSLa volume 1(1), 2009
 
terms of Schmitz and Schröder (2004), offers indications as to which default inferences are blocked.
According to the general pragmatics of CT+F constructions, we should expect
that (i) the meaning of eigentlich is contrasted with a focus alternative A(‘eigentlich’), that (ii) the sentence addresses a local question under discussion about
eigentlich(S), and (iii) the sentence is part of a strategy that will next address a
contrasting question under discussion about A(‘eigentlich’) (S’). The first important player that we need to find is A(‘eigentlich’). This player is difficult to
paraphrase because, like in the adjectival case, there is no good lexicalization of
the focus alternative of eigentlich. Jocular “un-eigentlich” in quotation marks is the
best approximation that one can find in actual conversational situations.4 However, in terms of the semantic analysis of eigentlich above, the semantic contribution of the focus alternative can be stated quite clearly:
〚eigentlich〛= λpλw.(N(p)(w) ∧ ¬P(p)(w))
A(〚eigentlich〛) = λqλw.(¬N(q)(w) ∧ P(q)(w) )
In a loose prose paraphrase, eigentlich serves to state that, for some given proposition p, the current world is such that p nominally holds true, and the current
world does not look as if p were true. The alternative to eigentlich states that for
some proposition q, the current world does not make q true even though it might
look as if q were true. In the contrastive topic construction, the proposition p will
be the propositional content of the eigentlich sentence. The contrastive proposition should be one that differs from p in exactly that second position that is
indicated by focus. What is non-standard about the contrastive topic construction is that the second part of the strategy can never fully exhibit the appropriate CT+F structure, due to the fact that there is no good word that expresses
A(〚eigentlich〛). In order to show how contrast, focus and discourse conspire, I
will discuss an example in detail.
ECT heisse ich TomF .
a. 〚S〛o = λw.(N(λw’.N(M, T,w’))(w)
∧ ¬P(λw’.N(M, T, w’))(w))
Ordinary meaning of the sentence: ‘We are in a world w in which my name
is Tom, but w fails to show one of the typical features of worlds in which
my name is Tom’
For instance, you may overhear conversations like the following.
OSLa volume 1(1), 2009
Eigentlich bin ich schon satt
‘I am eig no longer hungry.’
Na, und “un-eigentlich“ ? Was möchtest Du noch essen?
‘Well, and un-eig? Which dish can I offer you?’
  eigentli
〚S〛f = { λw.(N(λw’.N(M, T,w’))(w) ∧
λw.(N(λw’.N(M, E,w’))(w) ∧
¬P(λw’.N(M,E,w’))(w)), … }
= 〚‘what is my name, eigentlich?’〛o
Focus semantic value of the sentence: Set of propositions of the form “eig,
my name is X”. Semantically identical to the question ‘what is my name,
eig ?’.
c. 〚S〛ct
= {‘what is my name, eigentlich?’, ‘what is my name, Aeigentlich?’}
= { {λw.( N(‘my name is T’)(w) ∧ ¬P(‘my name is T’)(w) ),
λw.( N(‘my name is E’)(w) ∧ ¬P(‘my name is E’)(w) ), …}
λw.( ¬N(‘my name is T’)(w) ∧ P(‘my name is T’)(w) ),
λw.(¬N(‘my name is E’)(w) ∧ P(‘my name is E’)(w) ), …} }
Propositions in 〚S〛ct in English paraphrase:
{ {‘my name is T, but matters don’t look that way’,
‘my name is E, but matters don’t look that way’, … },
{‘It looks as if my name were T, but that’s not true’,
‘It looks as if my name were E, but that’s not true’, …} }
The step-by-step computation of ordinary semantic value of S, focus semantic
value of S and contrastive-topic value of S reveals that the single components
of the sentence contribute in a fully systematic fashion to reveal the pragmatic
skeleton of eigentlich discourse. The paraphrases in d. show that the indicated
strategy matches the observed structure of eigentlich-sentences in discourse. A
first utterance, containing ‘eigentlich’, serves to state what truly is even though
matters do not look that way. In a second utterance the speaker can go on and
point out what our world looks like, even though the respective proposition is
false in our world. Let us see how a typical continuation will fit into this strategy.5
S1: ECT heisse ich TomF .
S2: Aber jeder nennt mich Ede.
The second sentence is supposed to contribute to the following implicit question:
‘For which alternative name X does the world look as if my name were X,
yet it is not actually so?’.
Admittedly, the eigentliche eigentliche name of the person called ‘Ede’ is Thomas. Due to its length, however, this eigentliche eigentliche name would have caused awkward line breaks in formulae.
OSLa volume 1(1), 2009
 
The second sentence contributes a proposition that would typically lead us to
expect that ‘speaker’s name is Ede’ is true. That is, we are in a world which phenomenologically looks like a ‘my name is Ede’ world. However, our world is not
a ‘my name is Ede’ world in the technical sense (N), because sentence 1 of the
discourse has just asserted that the name of the speaker in the technical sense
is ‘Tom’. Against the knowledge background supplied by sentence 1, hence, sentence 2 offers a full answer to the remaining implicit question of the strategy. The
second sentence does even more in that it names a specific proposition that would
be typical for a ‘my name is Ede’ world. A literal answer to the remaining implicit
question could confine itself to ‘it might look as if my name were Ede’ without any
information as to what precisely creates the impression that the speaker’s name
might be Ede.6
Let us proceed to one of the examples that have been described as “weakening
of the proposition”. In the following two-sentence discourse, the speaker weighs
reasons to reject or accept a proposal.
(A: Do you want to join me to the movies tonight?)
B: S1 Eigentlich muss ich arbeiten. S2 Aber ich komme trotzdem mit.
‘Eig I have to work. But I will join you nevertheless.’
I assume that S1 has the following information structure:
ECT [muss ich arbeiten]f
The full sentence content is in focus, and the underlying strategy should consist
of the following two questions, which I have numbered for ease of reference.
〚ECT [muss ich arbeiten]f 〛ct
= {Q1: ‘what is the world eigentlich like, in spite of contradicting evidence?’,
Q2: ‘what does the world look like, even though it is not true?’ }
I will refrain from offering the full representation in a formal format. However,
note that sentence-wide focus allows the speaker to address a maximally varied
set of alternative propositions q as focus alternatives. S1 answers Q1 in the strategy: The world is such that the speaker has to work. Such obligations are typically
evidenced by the speaker’s being busy, the speaker not leaving her desk for days,
the speaker just taking breaks to have a cup of tea, etc. However, the eigentlich
statement already asserts that one of these typical pieces of evidence for ‘speaker
has to work’ fails to hold. Not surprisingly, sentence S2 reveals which one. Once
again, S2 contributes to question Q2 in an indirect manner. It names a typical
OSLa volume 1(1), 2009
In this sense, a fully matching answer like ‘uneigentlich heisse ich Ede’ (‘un-eig, my name is Ede’) is underinformative. If you consider the ‘jocular literal answer’, it is easy to see that indirect answers are better
answers in this case.
  eigentli
feature of worlds where the speaker does not have to work: ‘I agree to go to the
movies tonight’. The current world phenomenologically shares this feature of ‘I do
not have to work’ worlds even though – as asserted in S1 – the proposition ‘I do not
have to work’ is not true. We can indirectly reconstruct the focus alternative of
[muss ich arbeiten]f (‘I have to work’):
〚[muss ich arbeiten]f 〛f = {‘I have to work’, ‘I don’t have to work’}
Under this analysis, the short discourse in (42) once again exhibits a fully coherent example of a strategy. It also becomes clear why the eigentlich sentence
alone is not sufficient for the addressee as an agreement to the proposal. Sentence (43) in the indicated information structure only announces that something
in the present world is not as it should be in an ‘I have to work’ world. It does not
commit the speaker to anything – specifically as the announced counterevidence
is subject to the speaker’s decisions. This turns eigentlich sentences into ideal discourse moves, allowing the speaker some time to think while leaving all options
Let me summarize the building blocks of the present analysis of adverbial
eigentlich in cases discussed so far.
• eigentlich relates the proposition p expressed by the sentence to two sets of
worlds N(p) and P(p).
• The overall sentence denotes those sets of worlds w that are in N(p) but
not in P(p). Hence, the function expressed by eigentlich can be stated as
〚eigentlich〛 = λpλw.(N(p)(w) ∧ ¬P(p)(w))
• eigentlich can be focused, and will give rise to the focus alternative
A(〚eigentlich〛) = λqλw.(P(q)(w) ∧ ¬N(q)(w))
• eigentlich can be in contrastive focus. In this case, it “associates” with a
second focus in the sentence in the way all contrastive foci do. The second focus can be on a lower constituent, or on the entire sentence except
• Contrastive topic on eigentlich offers the typical starting point for an ‘eigentlich
S1. But S2.’ discourse. The discourse follows a strategy with two underlying
questions. S1 answers the question what the world really is like. S2 answers
the question what the world looks like, by contributing a proposition which
is typically linked with some second state of affairs which contradicts the
content of S1.
OSLa volume 1(1), 2009
 
The present analysis derives the ‘blocking of default inferences’ in an indirect
manner. All eigentlich – but discourses contrast a state of affairs p against evidence
φ for a conflicting state of affairs q. This evidence φ will contradict some other
piece of evidence ψ that is normally typically linked with p worlds. The default
inference would hence be the following.
Usually if p then ψ.
If the underlying logic of an eigentlich-but discourse in fact consisted in blocking
a default inference, one might wonder why the blocked inference is not simply
and straightforwardly named. Would it not be most cooperative to say “eigentlich,
p is the case, even though ¬ψ”? According to an analysis in terms of blocked
default inferences, such uses should be the best and most cooperative way to
use eigentlich. In actual fact, however, linguistic descriptions rest on a shared
intuition that eigentlich-but data constitute the prototypical core data, whereas
eigentlich-even though examples are at best mentioned tangentially in formal analyses.7 We will turn to these in the next section, under the heading of “derived
uses”. The present analysis predicts that the eigentlich-but use addresses the underlying strategy in an optimal manner. Notably, the analysis assumes that the
main function of an eigentlich statement is to contrast the world as it is with the
world as it looks. Blocked default inferences are a side effect of this pragmatic
function, but not its primary aim.
[4.5] Extension to Questions
Note that this analysis naturally extends to stressed eigentlich in wh-questions. We
have, in fact, used these questions in the contrastive topic analyses above.
Wie heißt
Du ?
How are-called you eigentlich?
E heisse ich T. (But everyone calls me “Ede”)
Wann hast Du  Sprechstunde?
When have you eigentlich conference
E habe ich D Sprechstunde.
Eigentlich have I on-Thursday conference
(But I offer advice to interested students at any time of the week.)
Focus in questions standardly serves to contrast them with other questions. These
alternative questions generally look like the original question, except that the focused constituent is replaced by one of the focus alternatives. In the case of a focus
on eigentlich, then, the alternative question will have the same structure as the
OSLa volume 1(1), 2009
Notably, Schmitz and Schröder (2004) build their entire analysis on basis of eigentlich-but examples.
  eigentli
original one, except that eigentlich is semantically replaced by A(〚eigentlich〛).
In the present example, this semantic alternative could be expressed as seemingly,
apparently. The speaker implicitly evokes a question “ When do you seem to have
conference hours?”, or more precisely, “What time looks as if it could be the time
of your conference hour?” The speaker alludes to facts in the context that answer
this question, such as for example “It is Monday morning at 8, and I am already
supervising students, so you might think that this is my conference hour.” Questions like (47) and (48) are standardly used if the answer q to the question about
what seems to be the case is salient in the context. The respective propositions
are given in brackets above. The contribution of the question in (47) could be
paraphrased as follows:
Contextually salient: ‘What name x is such that a contextually given fact
q suggests that x is your name?’
Given fact q: everyone calls you ‘Ede’.
Suggested answer: ‘Ede’ is the name such that ... Ede might be your name.
Explicit question: What name is eigentlich your name?
Explicit answer: Eigentlich, my name is ‘Thomas’
Use of eigentlich licensed by alternative q: ‘Your name is Ede’.
The semantics and pragmatics of the answer is exactly the same as the one that
was computed in the previous section. I refrain from a second spell-out here.
Note that an analysis which cuts orthogonally through standard semantic analyses (Schmitz and Schröder 2004 would certainly count as one amongst these) is
problematic at this point. Specifically, an analysis in terms of blocked default inferences will make the claim that the eigentlich question asks for a fact that the
speaker does not yet know, but she already is supposed to know which default
entailment of this unknown fact she wants to cancel. While this might be a technically feasible claim, it does not plausibly model actual question-answer discourse.
  
In this section, I want to outline usages of eigentlich which deviate more and more
from the analysis in section 4. The analysis in 4 is tailored for cases where eigentlich
associates with focus, brings alternative propositions to the fore, refers to a contrasting proposition q in context and conveys that q would lead one to expect
one of these (false) alternatives rather than the actual facts. Deviant cases vary
from those that differ only slightly from those discussed in the previous section
to those where eigentlich arguably does not contribute in the sense of section 4.
The first constellation is still closely related to the systematic use that has
been analyzed in section 4, and in fact has been mentioned briefly at the end of
the previous section: Continuations with obwohl (‘even though’) instead of aber
OSLa volume 1(1), 2009
 
Ich mag Peter eigentlich, obwohl er manchmal frech wird.
I like Peter eig
although he sometimes nasty becomes
‘As a matter of fact, I like Peter although he’s sometimes nasty.’
Schematically, the rationale behind (49) is this.
S, obwohl P.
P normally would lead you to expect not S. Here, however, S holds true.
This type of use is fully in line with a ‘blocked inference’ analysis. Note that (49)
with a pitch accent on eigentlich is deviant. A literal analysis like in section 4 will
look as follows.
〚eigentlich [ Ich mag Peter ]〛o
= λw.(N(‘I like Peter’)(w) ∧¬P(‘I like Peter’)(w))
According to this analysis, sentence (49) should state that in a technical sense
I like Peter, although it might look as if I didn’t. Intuitively, this is not at all the
message that is conveyed by the discourse in (49).(49) states that the speaker likes
Peter, and is prepared to act according to this sympathy (so the world, in all respects, will also look like a world where the speaker likes Peter), and that she is
moreover aware of the fact that Peter shows certain types of behavior that would
normally lead people to not like him. In terms of its paraphrase, then, the content
of (49) does not match earlier paraphrases that were used to spell out the contrast
between what the world is and what the world looks like.
It is extremely hard to pinpoint the origins of the global difference between
eigentlich-but examples and eigentlich - although examples. One major difference
between eigentlich – but examples and eigentlich – although examples, intuitively,
consists in the role that various conflicting propositions will play in further discourse. In an eigentlich – S1 – but S2 construction, the speaker signals that S2 is
more salient and will be considered a leading fact in the sequel. In an eigentlich
S1 – although S2 example, the speaker is willing to continue the discourse with a
focus on S1 which is the most salient fact, in spite of S2. This can be illustrated
with the following minimal pair.
OSLa volume 1(1), 2009
Eigentlich ist dein Zimmer schön aufgeräumt, obwohl im Eck noch etwas
Krempel liegt.
eig is your room nicely cleaned, even though in the corner still some rubbish lies
‘Your room is eig nice and clean, even though it is still a little messy in
one corner.’
  eigentli
The speaker is willing to acknowledge the room as being in the state of “cleanliness”, disregarding minor faults. This is definitely not so in (53).
Eigentlich ist Dein Zimmer schön aufgeräumt. Aber im
is your room nicely cleaned.
But in-the corner lies
noch etwas Krempel.
still some rubbish
‘Your room is nicely cleaned, but it is still a little messy in one corner’
In this case, the speaker most likely will insist that the mess in the corner has to
be removed before the state of “cleanliness” is fully acknowledged.8
In many cases, the differences between eigentlich – but and eigentlich – although
variants are subtle. However, the eigentlich – although use opens up a road that is
definitely blocked for eigentlich – but uses. Pushing the although use even further,
the speaker can use eigentlich without any obligation to specify any additional
proposition which would contradict S, or lead one to expect Not S. I will call this the
“pensive use of eigentlich” and show examples presently. Uses of this kind do not
suggest blocking of default inferences, or phenomenologically conflicting propositions. The utterances in (54) – (56) show some cases, and the English translations
are faithful in that they at best convey a moment of reflection before making the
assertion. Formally, the examples require a de-accented use of eigentlich.
Peter ist eigentlich ein netter Kerl.
‘Thinking about it, Peter is actually a nice guy.’
Wir schlafen eigentlich nur.
‘Thinking about it, we are sleeping all the time.’
Da hast Du eigentlich recht.
‘Thinking about it, you are right.’
In using sentence (54), the speaker asserts that ‘Peter is a nice guy’, that she has
come to this conclusion after some thinking about Peter’s character, and that
there is currently nothing that would cast doubt on this fact. Using (56), the
speaker signals friendly agreement, after a moment’s reflection or doubt. (The
doubts will never have been uttered explicitly: Sentence (56) would never serve
to end a longer monologue where the speaker lists the pros and cons to a position
uttered by the second interlocutor.) And so on.
If we were to speculate on the development of the “pensive use of eigentlich”
Examples like this could suggest that adverbial eigentlich, like its adjectival counterpart, could require
a choice analysis. In the present case, it is doubtful whether we want to license the inference from
eigentlich, the room is clean to ‘the room is clean’ holds true in the nominal sense of the sentence. If needed,
the analysis of section 3 can easily be extended to adverbial eigentlich. I will not burden the discussion
with this extra move.
OSLa volume 1(1), 2009
 
from the conceptually richer adverb uses, a metalinguistic intermediate stage
could be hypothesized. The speaker would make an although statement, signaling that the content of the sentence will serve as the basis for further discussion:
let’s act accordingly. The content of the although phrase could be about the speaker
trying to think about counterarguments to S.
Wir schlafen eigentlich nur.
Eigentlich: we sleep all the time. Even though: I tried hard to think of other
interesting things that I could tell you we’re doing.
Da hast Du eigentlich recht.
Eigentlich: you are right. Even though: I tried hard to think of counterarguments to your position.
At present, I am unable to offer a decent analysis of such a – speculative – intermediate use in terms of metalinguistic eigentlich in any serious sense. It seems
clear to me that the literal sense of eigentlich in terms of the preceding sections
will not be straightforwardly applicable, even if we raise it to the level of speech
acts. The mental and semantic processes that lead from a straightforward CF+F
structure via a related eigentlich-although sequence to an undertone like the one
paraphrased in (57) are still beyond formal analysis. Curiously, the pensive undertone carries over to unstressed eigentlich in questions. These suggest that the
question came to mind after some thinking/interaction. This interaction can be
friendly (‘casual question’) or aggressive.
Sind Sie eigentlich wahnsinnig?
(Having interacted with you for some time I feel pressed to ask:)
‘Are you mad?’
Wie heissen Sie eigentlich?9
(Having talked to you for some time, I realize that I could ask:)
‘What’s your name?’
Wollen Sie eigentlich noch Kuchen?
(Let me interrupt our friendly ongoing interaction to ask you:)
‘Would you like some more cake?’
The unstressed eigentlich in questions is definitely bad if the speaker starts an interaction with the hearer with the intention of asking exactly this question, or if
the question is an essential part of a professional interaction between speaker and
hearer. The following exchanges are all marked, the effects ranging from “funny”
OSLa volume 1(1), 2009
Also warranted in an aggressive interaction like in the preceding example.
  eigentli
to “rude/offensive”10 .
Waiter when approaching customers:
Was wollen Sie (#eigentlich) essen?
what want you eig
≈ ‘What do you want to eat anyway?’
Dentist asks whimpering patient:
Haben Sie (#eigentlich) Schmerzen?
have you eig
≈ ‘By the way, are you in pain?’
Police officer checking you after you passed a red traffic light:
Wie ist (#eigentlich) Ihr Name?
what is eig
your name?
‘What’s your name anyway?’
[5.1] Postlude: A dialectal twist
When my colleague and I tried to verify our intuitions about possible and impossible uses of adjectival eigentlich via GOOGLE, we were surprised by accidental
matches for patterns that the analysis, as well as our intuitions, would not support. Among these were hits for ein eigentlicher/-s in positive contexts, quantified uses like die meisten eigentlichen … and uneigentliche in non-expert language.
Closer investigation revealed that all such matches came from Swiss sites, or were
quotes from Swiss authors or newspapers, or were on sites / by authors with a
very likely Swiss background (e.g. Swiss embassy in Berlin). Further explorations
on such sites suggest that Swiss German uses eigentlich more or less as a synonym
of wirklich, echt (‘true’). Specifically, Swiss eigentlich needs no contextual licensing, can hence be used in quantificational NPs, and has a well-defined antonym.
We could not so far establish the prosodic patterns of adjectival eigentlich in Swiss
German but would expect that accenting is much freer than in German. In summary, Swiss eigentlich offers a minimally contrasting ‘normal’ eigentlich variant
and hence highlights the context dependency and discourse function of German
In the present paper, I proposed a truth conditional analysis of adjectival and adverbial eigentlich. I assume that the item serves to relate a property or proposition C to two derived cognitive objects: N(C), the nominal sense of C and
P(C), the conjunction of properties (propositions) that together would offer
Hence, learners of German need to be cautioned against the idea that eigentlich invariably makes a question sound more polite!
OSLa volume 1(1), 2009
 
phenomenological evidence for a case of C-hood. These two derived concepts underlie the use of both, adjectival and adverbial eigentlich, but the two cases differ in
the details. Adjectival eigentlich shows definiteness effects, which arise due to the
fact that the phrase eigentlich(e/r/s) N refers to the unique plural object that exemplifies the sum of properties in N(C) or P(C) respectively. Eigentliche/r/s
N is typically used in situations where the discourse content entails that the two
concepts are different. However, statements about empty domains (es gibt hier
keinen eigentlichen Chef = ‘there is no true boss here’) or other uses without explicit
contrast are possible, and in the range of the proposed analysis. Finally, the adverbial use is characterized by the fact that eigentlich has a choice semantics: The
full phrase eigentliche/r/s N can refer either to N(〚N〛) or P(〚C〛), depending on the speaker’s intentions. This also explains why eigentlich can hardly be
Adverbial eigentlich is mereologically simpler. It maps propositions to propositions and does not create pluralities of worlds. Likewise, a majority of cases can
be analyzed on basis of the assumption that eigentlich S states that the world is
in N(〚S〛), but not in P(〚S〛): 〚eigentlich〛= λpλw.(N(p)(w) ∧ ¬P(p)(w)).
eigentlich has a systematically available focus alternative, and can be used in focus
constructions and contrastive topic constructions in both assertions and questions. I demonstrated that the predicted semantic representations are faithful to
the intuitive meanings of examples in many cases; and specifically in those cases
where prosody indicates a clear information structure.
The proposed analysis ties in nicely with an earlier formal analysis proposed
by Schmitz and Schröder (2004), treating eigentlich as a blocker of default inferences. The present proposal goes beyond its predecessor in several respects. First,
it offers a uniform analysis for adjectival and adverbial use. Second, the analysis makes crucial use of the information structure of the utterance and offers
a fully compositional account in terms of focus semantic value, literal meaning
and contrast. Third, the analysis of stressable eigentlich carries over to the question case without any additional stipulations. The approach likewise allows us to
speculate on the demarcation line between truth conditional and emotive uses
of eigentlich and offers straightforward reasons to distinguish truth conditional
uses at least from “pensive” uses at the end of the spectrum of emotive usages.
Finally, the present approach could be helpful in clarifying the notions of “normally behaved” and “normally looking” exemplars of a kind, or worlds in a proposition. This could help to cut short some overgenerations inherent in Schmitz and
Schröder’s account.
Büring, D. 2003. On D-trees, Beans, and B-accents. Linguistics and Philosophy 26(5),
OSLa volume 1(1), 2009
  eigentli
Büring, D. 2007. Semantics, Intonation, and Information Structure. In G. Ramchand and C. Reiss (eds.), The Oxford Handbook of Linguistic Interfaces, pages
445–474, Oxford.
Eckardt, R. 1999. Normal Objects, Normal Worlds, and the Meaning of Generic
Sentences. Journal of Semantics 16(3).
Eckardt, R. 2006. Dependent on context: eigentlich in adjectival and adverbial use.
In C. Fabricius-Hansen, B. Behrens and M.F. Krave (eds.), Pre-Proceedings of the
SPRIK Conference 2006, pages 7–11, Oslo.
Frühwirth, A. 1999. Syntax, Semantik und Pragmatik der deutschen Modalpartikel
eigentlich. Marburg, mA Arbeit.
Horn, L. 2002. Uncovering the ‘un’-word: A study in lexical pragmatics. Sophia
Linguistica 45, 1–64.
Kohrt, M. 1988. Deutsche Sprache 2, Chapter Eigentlich, das Eigentliche und das
Nicht-Eigentliche. Zum Gebrauch einer sogenannten Abtönungspartikel, pages
Link, G. 1983. The Logical Analysis of Plural and Mass Terms: A Lattice-Theoretical
Approach. In R. Bäuerle, Ch. Schwarze and A. von Stechow (eds.), Meaning, Use
and Interpretation of Language, pages 302–323., Berlin: Walter de Gruyter.
Meibauer, J. 1994. Modaler Kontrast und konzeptuelle Verschiebung. Studien zur Syntax
und Semantik deutscher Modalpartikeln. Tübingen: Niemeyer Verlag.
Port, A. 2006. Was heißt eigentlich eigentlich? Eine semantische und pragmatische Analyse. Masters Thesis, HU, Berlin, magisterarbeit.
Schmitz, H.-C. 2008. Eigentlich again. In A. Grønn (ed.), Proceedings of SuB 12, pages
567–581, Oslo.
Schmitz, H.-C. and Schröder, B. 2004. Sprache und Datenverarbeitung, Chapter Updates with ‘eigentlich. 1-2.
Schwarzschild, R. 1999. Givenness, AvoidF and Other Constraints on the Placement of Accent. Natural Language Semantics 7(2), 141–177.
Thurmair, M. 1989. Modalpartikeln und ihre Kombinationen. Tübingen: Niemeyer.
Weydt, H. 1977. Abtönungspartikel. Die deutschen Modalwörter und ihre französischen
Entsprechungen. Bad Homburg vor der Höhe: Gehlen.
Weydt, H. and Hentschel, E.l. 1983. Kleines Abtönungswörterbuch. In Partikeln und
Interaktion, volume 3, Tübingen.
OSLa volume 1(1), 2009
 
  
Regine Eckardt
University of Göttingen
English Dept.
Käte-Hamburger-Weg 3
D-37073 Göttingen
[email protected]
OSLa volume 1(1), 2009
Behrens & Fabricius-Hansen (eds.) Structuring information in discourse:
the explicit/implicit dimension, Oslo Studies in Language 1(1), 2009. 109-129. (ISSN 1890-9639)
  
 
University of Oslo
Standardly (Safir 2004), the “complex reflexive”  +  in Dutch or Scandinavian is treated as a special species of anaphora, stronger than  alone.
This approach has a number of disadvantages, descriptive and theoretical.
Theoretically, it is desirable to treat  the same as when it modifies another element. Bergeton (2004) argues that a uniform analysis of  as
an intensifier is feasible and that the descriptive shortcomings of standard
treatments can be overcome if intensification is severed from binding ().
However, his account is incomplete in a few regards. Building on a formal
theory of focus (Rooth 1992), I show that the distribution of simple and complex reflexives – almost complementary in Dutch and Scandinavian, freer in
German – can be more fully explained on the basis of a theory of intensification (Eckardt 2001) supplemented by Bidirectional OT (Blutner 1998, 2000,
2002, 2004, 2006).
A number of languages have both “true” and “false” reflexives (Bouchard 1984),
or both  and  anaphors (Reinhart and Reuland 1993), serving different functions; specifically, there are predicates where only the former can occur. Norwegian exhibits the complex reflexive seg selv beside the simple reflexive seg. Hellan
(1988) argued that basically, these two are in complementary distribution, as suggested by the following examples:
Sangerne akkompagnerte seg #(selv) på gitar.
singers-the accompanied  #() on guitar
Christina Aguilera har sagt ja til å kle
av seg (# selv).
Christina Aguilera has said yes to to dress off  (# )
Gulbransson drakk [ seg (# selv) full ] .
Gulbransson drank  (# ) drunk
Huni har sitt eget band med til å akkompagnere [seg (# selv)]i .
shei has her own band with to to accompany [ (# )]i
  
Hellan developed an ingenious theory to predict this distribution (see Section 2).
However, there are problems with this account, both theoretical and descriptive.
These problems persist in more recent analyses of this and similar phenomena
(Reinhart and Reuland 1993; Safir 2004). I will refer to these analyses collectively
as the  , and I will treat them and their problems, and the
facts that (1)–(4) indicate, more extensively in Section 2. In sharp contrast to the
traditional treatment, which is basically syntactic, Bergeton (2004) proposes to
derive the different distribution of (Danish) sig and sig selv from the meaning of
selv. This word is used to modify other individual denoting words than reflexive
pronouns, for instance, personal pronouns, and a uniform description as an 
 is theoretically attractive. Intensifiers are assumed to supply focus, and
focus has to do with alternatives and contrast. Bergeton can answer why sig is
preferred over sig selv in cases like (2) and (3) but sig selv is good in cases like (1),
but not really why sig selv is preferred over sig – that is, why the intensifier is necessary – in cases like (1). Besides, he does not work with a formal theory of focus
(such as Rooth 1992) and intensification (such as Eckardt 2001), so his account is
not as precise as it could be. Finally, it is not evident how it can be extended to
languages like German, where the intensifier is on the whole less necessary than
in Scandinavian or Dutch.
     
Below, I review three theories of  +  which I take to be representative of
the syntactic tradition: Hellan (1988), Reinhart and Reuland (1993), Safir (2004).
I first present their key elements, then I discuss what I see as their problems.
[2.1] Three Theories
According to Hellan, the near complementary distribution of seg and seg selv, as
witnessed by (1)–(4), results from a division of labour between seg and selv: The
former indicates binding (sloppily, coreference with some accessible subject), while
the latter encodes .
Thus the fact that seg selv is impossible as a ‘long-distance anaphor’, as shown
by (4), follows from the constraint that binder and bindee be coarguments, as does
the alleged fact that seg selv is excluded as a small clause subject, as suggested by
(3). Because conversely, the absence of selv bars coargumenthood, the fact that
seg selv is necessary in cases like (1) follows as well. The preference for seg alone
in cases like (2) is attributed to the hypothesis that with such verbs, the reflexive
does not really code an argument. Sloppily, what you do when you wash is not
what you do when you wash somebody – it is a different predicate.
The reflexivity theory developed by Reinhart and Reuland (1993) predicts more
or less the same facts concerning the Dutch pair zich and zichzelf (and 1st and 2nd
person forms). Only  anaphors, e.g. zichzelf , reflexivize predicates, cf. (5);
OSLa volume 1(1), 2009
  
predicates with  anaphors are intrinsically reflexive, rendering  anaphors
redundant, cf. (6).
Bart bewondert zich*(zelf).
Bart admires *()
Petra waste zich(??zelf).
Petra washed (??)
Since to reflexivize a predicate means to indicate that its object corefers with its
subject,  small clause subjects and long-distance anaphors are ruled out:
Het publiek danste [ zich(#zelf) warm ] .
the audience danced (#) warm
Zij smeekte mij *zichzelf te helpen.
she beseeched me * to help
Safir (2004) understands the near complementary distribution of - and 
in Germanic as a reflex of a competition. He assumes a scale of forms where the
former is more dependent than the latter, and a principle requiring the most dependent form available to be used. Together with availability constraints, specifying, for example, that - depends on a local subject and that it could “represent a referential value distinct from what it depends on”, these assumptions
are to account for all the facts about the distribution of different anaphoric forms,
in Germanic and in other languages.
[2.2] Problems
There are a number of problems, though, shared by the three theories sketched;
some of which are addressed by the scholars themselves, some of which are not.
Two problems are descriptive in nature, one has a more theoretical status.
The Cross-Linguistic Challenge: How can intrinsic reflexivity vary?
Cross-linguistic variation offers a challenge to any theory of complex reflexives.
German, too, has phrases built from the string  +  – sich selbst – but here the
 element is optional in many cases where, mutatis mutandis, it is obligatory
in Dutch or Scandinavian; “the addition of the  form is either disambiguating
or emphatic” (Safir 2004, 99). According to Mattausch (2003), the fundamental
challenge facing Reinhart’s and Reuland’s account is that it is not obviously extendable to other languages; in particular, in German or Icelandic,  marking is
not always mandatory for non-intrinsically reflexive predicates. The same challenge faces Hellan’s and Safir’s accounts.
Safir counters this challenge by hypothesizing that in German, selbst does not
OSLa volume 1(1), 2009
  
form a morphological unit with sich; not entering the numeration by a single selection, sich selbst is not in competition with sich and sich is not obviated from the
contexts where - is mandatory in Dutch or Mainland Scandinavian (Safir
2004, 100; 205). This may seem a reasonable enough solution, but there is no independent evidence to support it.
The claim that verbs like vaske ‘wash’ are intrinsically or inherently reflexive,
meaning that  is a nonargument or that the object cannot represent anything
distinct from the subject, is intuitively rather weak. Naively, one is tempted to
say that something else can absolutely be substituted for . Indeed, as conceded
by most scholars, the complementarity between  and - is not complete:
- forms are possible with intrinsically reflexive verbs provided that the
discourse justifies them, as illustrated below:
Han begynte å kle av meg. Så kledde han av seg selv også.
he began to dress off me then dressed he off   too
‘He started undressing me. Then he undressed himself too.’
Here, the context provides an alternative – meg ‘me’ – to the reflexive referent,
and this seems to require that (seg be stressed or) seg selv replace seg.
To this, a syntactician may respond that it is only apparently possible to substitute something else (i.a., -) for , because when we do so, we change the
verb, going from one variant to another.1  and - are not interchangeable – in fact, when we substitute the latter, we select the other variant of the
verb, the one that can have an object distinct from the subject. Thus , though
it may be an argument, is informationally redundant after all.
And to be sure, there do seem to be two different actions described in the two
halves of (9). Or, with reference to (2), if Christina Aguilera undresses, the action is
more or less automatic and only semiconscious, whereas if I were to undress her, I
should feel at a loss over how to go about it. But the same can be said of predicates
where the  element is obligatory, such as Norwegian undersøke ‘examine’; if
a doctor examines herself, she performs a different, this time much less routine,
action than if she examines someone else.2
It is probably true that  and - are never freely interchangeable when
bound by a local subject – there will always be a change in the contextual conditions. However, it is very difficult to argue, without risking circularity, that we invariably alter the meaning of the predicate when we substitute one for the other.
OSLa volume 1(1), 2009
See e.g. Safir (2004, 130) on “choices of homophonic predicates”.
This case is also a counterargument to the theory of Lødrup (2007), who argues that simple and complex
reflexives are both used in the local domain but that the simple reflexive is used when the physical aspect
of the referent of the binder is in focus.
  
In fact, on close inspection, there seem to be some clear cases where we do alter
the meaning – cf. (10), some intermediate cases, and some clear cases where we
do not alter the meaning, cf. (9) and (11): In a situation where it is appropriate to
use just  and in one where it is appropriate to use -, you are describing
exactly the same action. This, to my mind, is a strong sign that we need more
precise tools than those provided in the syntactic tradition.
Det er mange anekdoter om folk som har skadd seg
it is many anecdotes about people that have injured 
alvorlig i ekstreme situasjoner uten
å merke det.
seriously in extreme situations without to notice it
Filmens utgangspunkt er 1. verdenskrig, der fem soldater
film-the’s departurepoint is 1st worldwar where five soldiers
som har skadd seg selv, skal henrettes for forræderi.
that have injured   shall execute-s for treason
Trangen til å vaske seg ofte er den vanligste nevrosen.
urge-the to to wash  often is the commonest neurosis
Noen nevrotikere må vaske seg selv eller huset veldig ofte.
some neurotics must wash   or house very often
In the same vein, it is easy to show that the alleged ban on - forms as small
clause subjects, in resultatives or perceptives (cf. (3) and (7)), often fails, and that
(in Norwegian) seg and seg selv can overlap in the same type of context without
reflecting any change in interpretation, however subtle:
I toårsalderen sang hun seg i søvn i stedet for å gråte.
in twoyearsage sang she  in sleep in stead of to cry
Når han skal sove, synger han seg selv i søvn.
when he shall sleep sings he   in sleep
What can be observed is that the frequency of - forms vs.  forms as subjects in resultative small clauses varies with the frequency of referentially distinct
DPs in the same context: The more likely the “superordinate” event is to cause a
distinct referent to undergo the “subordinate” event, the more likely is a nondistinct referent to be articulated by a  form (cf. Section 4.3).
Intensification and Compositionality
The traditional, syntactically based accounts of  +  anaphors do not answer
(or even ask) the question why  forms (or, in Romance,  forms) are chosen to augment  forms. These forms have a use outside the reflexive domain,
as intensifiers, modifying other (individual-denoting, type e) DPs, for instance,
personal pronouns. Some scholars (Moravcsik 1972, Edmondson and Plank 1978,
OSLa volume 1(1), 2009
  
Gast 2006, Siemund and Gast 2006, Zribi-Hertz 1995) have argued that the reflexive usage is historically derivable from, if not synchronically identical to, the intensive (emphatic) usage. In fact, a uniform, synchronic treatment might well be
feasible, and if it is, it is clearly desirable. Bergeton (2004) has made a serious
attempt at such a treatment (Section 3).
There is a core of common sense in the contention that verbs like vaske ‘wash’
are intrinsically or inherently reflexive: Washing is typically something you do
to yourself, it is a predominantly self-directed action. In most or all cases of
locally bound , a case can be made that the predicate denotes such an action. We have also seen that when - figures as a small clause subject, it
is not uncommon for somebody else to occupy that position in the same context. König and Siemund (2000) have hypothesized a cross-linguistic correlation
between the type of reflexivizing strategy and the self- vs. other-directedness of
the predicate: The more marked, or other-directed, a reflexivizing situation, the
more marked, or complex, a reflexivizing strategy will be used to encode it.
A marked-form – marked-content correlation such as this is in itself plausible,
not least in a perspective of OT pragmatics (cf. Zeevat 2004). Mattausch 2003 has
shown how it can have developed as a result of bidirectional learning. What remains to be shown is that - is not only more complex and marked than ,
having two syllables instead of just one, but that it is ideally suited to the purpose
because  makes a specific and constructive contribution to the marking of
other-directedness – in other words, that the use of  (or ) to augment 
is not accidental, but that the meaning of  +  is a function of the meanings
of its parts.
      
Following König and Siemund (2000) in assuming that, across many languages,
 anaphors are combinations of a  anaphor and an intensifier whose overall
meaning is a function of that of the two components, Bergeton (2004) develops
a semantic-pragmatic analysis of - as the intensified version of . This
work represents a radical reinterpretation of the relevant facts and has decisive
advantages over most previous work, descriptively and theoretically. However, it
is still in need of improvement in some respects. In this section, I will first outline
the analysis and point out its strengths; then I will discuss what I perceive as its
[3.1] Adreflexive Intensification
Bergeton’s theory rests on these assumptions:
• Binding and intensification are independent.
• Intensification requires contexts providing alternatives and contrast.
OSLa volume 1(1), 2009
  
• Intensification of anaphors is sensitive to predicate meaning and pragmatics (utterance situation and common ground).
• There are four relevant classes of predicates (Bergeton 2004, 160):
(i) Reflexive predicates, presupposing identity of arguments,
(ii) “anti-reflexives”, presupposing non-identity of arguments,
(iii) neutral predicates, presupposing nothing of the kind, and
(iv) “hidden” neutrals, coming close to anti-reflexives; presupposing nothing, but evoking expectations of non-identity of arguments.
This theory is theoretically satisfying in a double sense: It is economical, aiming
at a uniform analysis of  whatever it modifies, and it is explanatory in the
sense that the meaning of  is assigned a role.
The theory also makes more precise predictions about the distribution of the
forms than do earlier accounts (see 2.2). In particular,  and - are not predicted to be, in the strict, lexical sense, in complementary distribution, as a large
subclass of the “inherently reflexive predicates” are reassigned to a new class of
neutral predicates where, depending on the context, both are possible. Likewise,
the theory correctly predicts that - forms can be SC subjects; there is no requirement for them to represent coarguments, and again, the choice will depend
on whether the context generates a contrast set of alternatives.
Finally, the theory predicts that  +  forms can be ‘long-distance’ bindees;
in fact, Bergeton contests the traditional notion that when a language has both
simplex and complex reflexives and both local and non-local binding, it is the
former that are non-locally bound – “Pica’s generalization” (Pica 1985), cf. (13),
maintaining that intensification is independent of binding and that the distribution of  follows basically the same semantic-pragmatic pattern in non-local
as in local environments: when the context offers alternatives with which the
bindee is explicitly contrasted, as in (14) or (15), sig selv is possible:
Sjeherasadi ba
Dunjasad hjelpe [seg (# selv)]i .
Sheherazadei asked Dunyazade help [ (# )]i
McArthur, an extremely tough general, feels that the lightly wounded soldiers ought to put up with the pain in order to save painkillers for the truly
needing. When he himself got a large piece of shrapnel in his thigh he stubbornly refused to take any kind of painkillers. But in the end the pain became too much for him. So far his principles had dictated him to ask the
nurses to give the painkillers to the other soldiers in his ward.
OSLa volume 1(1), 2009
  
b. ?Men igår,
sent på natten bad McArthuri mig endeligt
But yesterday, late on night-the asked McArthuri me finally
give sigi (selv) en morfinindsprøjtning.
give i () a morphineinjection
Hani vil ha meg til å forsørge både segi selv og foreldrene.
‘He wants me to provide for both himself and his parents.’
[3.2] Loose Ends in Bergeton’s Account
Although Bergeton’s work represents a significant step in a promising direction,
it also leaves some questions unanswered. Among these are:
(i) Why is  necessary in many cases – like (1) and (5)?
(ii) Why is  less necessary in (say) German than in (say) Dutch?
The Necessity of Intensification
It is evident that the theory is intended to account for the necessity of intensification by  in connection with “anti-reflexive” or “hidden neutral” predicates
like beundre, cf. (16). These predicates are assumed to induce the presupposition
or at least the expectation that their arguments refer to different entities; people do not normally admire themselves. According to Bergeton’s “Contrastiveness Condition”, intensification requires contexts providing contrasting alternatives. Since this condition is, as it were, lexically satisfied in connection with
anti-reflexive or hidden neutral predicates, it is clear that  is possible here, –
but not, strictly, that it is necessary. It would seem that intensification is needed
to overcome anti-reflexivity or hidden neutrality, but this is not spelt out, and it
does not follow from the analysis; it remains unclear what is wrong with  forms
in connection with anti-reflexives or hidden neutrals – or what sets it right once
 is augmented by .
Narcissos sitter og beundrer seg *(selv).
Narcissus sits and admires  *()
The Cross-Linguistic Challenge
The fact that  +  is clearly less frequent in German (or Icelandic) than in
Dutch or Mainland Scandinavian, noted in Section 2.2, is just as problematic on
Bergeton’s account as on earlier accounts. With many predicates that correspond
to hidden neutral predicates in, say, Danish, sich is possible on its own – even
though it is not accented – and selbst intensification is just possible, not necessary.
Yet hidden neutrality would seem a universal property – one would not expect it
to vary cross-linguistically.
OSLa volume 1(1), 2009
  
Sie begleitet
sich (selbst) auf dem Klavier.
she accompanies  () on the piano
So the challenge is to motivate some linguistic contrast between Dutch and MSc
on the one hand and German and Icelandic on the other, to explain how antireflexivity or hidden neutrality, semantic properties of predicates, may relate to
intensification differently from one language to another.
Note that precisely because Bergeton’s theory is compositional, the option
open to Safir (Safir 2004) (see Section 2.2) is here closed: One cannot very well
argue that  +  is a unit in Dutch but not in German, when the cornerstone
of the theory is that binding and intensification are independent of each other.
The reason that these issues remain open in Bergeton’s analysis may be that
he does not work with a formal theory of intensification, such as Eckardt (2001),
set in a formal theory of focus and alternatives, such as Rooth (1992). Such a basis
would sharpen the theory, also with respect to the facts about which it appears
to make valid predictions. Thus the theory seems to explain why  is sometimes appropriate and sometimes inappropriate with neutral predicates; but to
ascertain these predictions, it is really necessary to seek a firm footing in a formal framework.
     
I believe that the problems identified in the last section can be solved – or, at
any rate, that solutions can be developed and assessed – if Bergeton’s theory is
supplemented by a theory of intensification based on a formal theory of focus
interpretation. Such a theory of intensification has been proposed by Eckardt
(2001), building on the focus theory of Rooth (1992), Alternative Semantics. In
addition, there will be a need to augment Alternative Semantics by Bidirectional
Optimality Theory, developed by Blutner (Blutner 1998 and later work).
Eckardt’s Theory of Intensification
Eckardt’s theory is primarily about the German intensifier selbst (≈ English intensifying -self , -selves) as it appears in (18), adnominally, or (19), adverbally.
Ich möchte spätestens mit 17 ein Kind haben, weil
ich es am besten
I wantto latest
at 17 a child have because I it at best
finde, früh Kinder zu bekommen, um mein Kind,
wenn es
find early children to get
for my child
when it
selbst Teenager ist, zu verstehen.
self teenager is to understand
OSLa volume 1(1), 2009
  
Ihr Kind spürt, dass Sie mit ihm nicht zufrieden sind – und es ist
your child senses that you with it not content are – and it is
selbst nicht zufrieden.
itself not content
The theory is very simple. It says:
   E
The intensifier denotes the identity function on type e entities.
From this assumption, the following follows:
• The expression denoting the type e argument, the so-called , is a
name, variable, (personal or reflexive) pronoun, or definite description.
• The associate is out of focus (or in focus together with the intensifier), and
the intensifier is in focus (alone or together with the associate) – or else
intensification would be redundant.
    
The net effect of selbst is to add focus, giving rise to focus presuppositions in Alternative Semantics, the focus theory of Rooth (1992).
Let us work our way through (20).
As Elizabeth Brinker cares for her mother, she knows she herself F
is [ at risk of inheriting ]F Alzheimer’s disease.
We have the adnominal case: She and herself form a constituent, she herself . Suppose that the ordinary semantic value of she is x, bound in a presupposition. Then
the ordinary semantic value of she herself is the same, x, and so is also the focus
semantic value of she. But the focus semantic value of she herself is the set of values of alternatives to the identity function at x (the focus semantic value of herself
being the set of alternatives to that function):
     she herself
{ z | there is an alternative f to ID<e,e> such that z = f (x)}
Alternatives to the identity function on individuals are operations on individuals
that do not map them onto themselves but onto others. At sentence level, the two
foci are interpreted in terms of a focus presupposition:
OSLa volume 1(1), 2009
  
    (20)
There is a proposition φ such that there is a function f ≈ ID<e,e> and
there is a relation R ≈ at risk of inheriting such that
φ = R(alzheimer’s)(f (x))
This presupposition is verified in the context: f = one’s mother, R = afflicted with.
To be sure, there is more to say about, in particular, adverbal intensification in
Eckardt’s theory (see also Sæbø 2005). However, since intensification of reflexives
is adnominal intensification, we can concentrate on this.
[4.2] Adreflexive Intensification in Eckardt’s Theory
Eckardt does not explicitly consider selbst intensification of reflexive pronouns –
all her examples have subject associates – but the theory can accommodate it. As
a step towards it, consider intensification of a Norwegian object pronoun:
Glahn elsker drømmen om Edvarda mer enn han elsker henne selv.
Glahn loves dream-the of Edvarda more than he loves her 
De gode gjerningene hennes har reelt sett gagnet
henne selv.
the good deeds
hers have really seen benefitted her 
As the focus semantic value of henne selv, we can maintain the focus semantic value
of she herself defined above, with x as the focus and ordinary semantic value of
henne. The focus presupposition in terms of which focus is interpreted at sentence
level might here be (assuming y to represent the good deeds):
    (22)
There is a proposition φ such that there is a function f ≈ ID<e,e> and
there is an operation O ≈ really such that
φ = O(benefit(y)(f (x)))
Again, this is verified in the context: f = one’s beneficiaries, O = supposedly.
From here, there is a short step to the case where the associate is a reflexive.
Consider first the intensification of a reflexive object of a “neutral predicate” (in
Bergeton’s term), where its appropriateness in general depends on the context:
(En barberer barberer alle
som ikke barberer seg selv
a barber shaves everyone that not shaves  
ingen andre.) Barberer denne barbereren seg selv?
noone else
shaves this barber
 
Applying the same method to this case as to (20) and (22), we can compute this
focus presupposition (assuming x to represent the barber):
OSLa volume 1(1), 2009
  
    (23)
There is a proposition φ such that there is a function f ≈ ID<e,e>
such that
φ = shaves(f (x))(x)
This is verified in the context, for one’s barbershop customers instantiating f .
Now there is an alternative to this method, modifying it slightly as regards the
focus structure of the intensified phrase. Eckardt assumes that the intensifier
is in focus all by itself, forming a separate focus domain. The associate is out
of focus. This is reasonable as long as – and this is the case she considers – the
associate is a familiar referent, a name, personal pronoun, or definite description.
But even then, it is imaginable that the intensifier, inherently accented, forms a
focus domain together with the associate; cf. the following alternatives:
She [ herself ]F is [ at risk of inheriting ]F Alzheimer’s disease.
[ She herself ]F is [ at risk of inheriting ]F Alzheimer’s disease.
Barberer denne barbereren seg [  ]F ?
Barberer denne barbereren [ seg  ]F ?
And when the associate is a reflexive, the option that it and the intensifier are
in focus together, corresponding to (25b), may appear to be particularly viable.
Then, the focus presupposition is simplified, since the focus semantic value of the
intensified pronoun or anaphor is simply the set of alternatives to x:3
    (23) ()
There is a proposition φ such that there is a y ≈ x such that
φ = shaves(y)(x)
The central question is how the associate relates to focus if it is not intensified:
(i) Is it out of focus (old information)? Or:
(ii) Is it in focus with something else – the predicate?
The facts about the distribution of  () in connection with anti-reflexive
predicates (cf. Section 4.4) seem to point in the direction of 2. That is, there is
reason to assume (26b), with wide focus and so-called informational integration
(Jacobs 1999), instead of (26a), where the reflexive is out of focus:
OSLa volume 1(1), 2009
Hun [ vasker ]F seg.
she washes 
Eckardt derives so-called centrality effects from alternatives to the identity function. It is debatable
whether there is a real difference between the two options in this regard, or whether centrality should
be explained otherwise. Anyway, centrality effects do not seem to play a central role in connection with
intensified anaphors.
  
Hun [ vasker seg ]F .
she washes 
And in that case, the ceteris paribus situation with intensification is surely not
the one where the associate is out of focus by itself, corresponding to (25a) and
(27a), but the one where the associate and the intensifier are in focus together,
corresponding to (25b) and (27b). This is the structure I will henceforth assume.
Hun [ vasker ]F seg [ selv ]F .
she washes  
Hun [ vasker ]F [ seg selv ]F .
she washes  
[4.3] Predicate Meaning and Local vs. Global Focus Justification
On the theory of focus interpretation and intensification as applied to , a sentence with [   ] presupposes that there is a proposition corresponding to
the sentence but for the substitution of some alternative for the focus or foci, in
particular, for the  referent. Generally, the existence of such a proposition can
be given contextually, in the discourse, or more globally, through our lexical and
encyclopaedic knowledge.
    a p  
There is a proposition φ such that there is a y ≈ a such that
φ = P (y)(a)
It has often been noted that the predicates that Bergeton calls neutral (at any rate
those that are traditionally called intrinsically reflexive) require “discourse justification” for their reflexive to be intensified (e.g., Reinhart and Reuland 1993,
fn. 15). This means that the focus presupposition arising from the intensifier
must be justified contextually; say, (2) with seg selv is not felicitous out of the
blue. On the other hand, the “anti-reflexive” and “hidden neutral” predicates are
felicitous with seg selv without contextual support, so the focus presupposition
is evidently justified lexically. In other words: By virtue of their meaning, their
other-directedness, the anti-reflexives and the hidden neutrals provide propositions involving alternative referents, verifying the presupposition induced by
the intensifier “globally”, as opposed to the neutrals, where it must be verified
In this perspective, resultative constructions where simple and intensified reflexives coexist as small clause subjects, with a more or less pronounced bias for
one or the other from case to case, form an intermediate stage; and Bergeton’s
theory is very well adapted to account for this gradient variation. Often enough,
intensification does not need to be justified contextually, it depends on whether
OSLa volume 1(1), 2009
  
it would be sufficiently natural for another referent to occupy the relevant role in
relation to the caused event as a result of the causing event. So it does not depend
on one predicate but on two predicates in a resultative relation.
Note, for example, that the frequency relation between gråte seg selv i søvn (‘cry
-self to sleep’) and gråte seg i søvn (by a Google search) is at about 1 : 100, while that
between synge seg selv i søvn (‘sing -self to sleep’) and synge seg i søvn is at about 1 : 1.
I toårsalderen sang hun seg i søvn i stedet for å gråte.
in twoyearsage sang she  in sleep in stead of to cry
Når han skal sove, synger han seg selv i søvn.
when he shall sleep, sings he   in sleep
This can be taken to reflect that the focus presupposition coming from selv is much
more readily justified when singing is the causing activity – we normally sing
others to sleep – but that presupposition can still be accommodated when crying
is the causing activity.
Note, by the way, that there are also neutral predicates where there is no clear
preference for simple reflexives and complex ones do not require discourse justification, and where, as in the resultative constructions, this coexistence is not
connected to a polysemy or a “doppelgänger” or “proxy” interpretation effect.
In Norwegian, such truly neutral predicates include redde ‘save’, ofre ‘sacrifice’,
forsvare ‘defend’, beskytte ‘protect’, and ditransitive nekte ‘deny’.
[4.4] The Necessity of Intensification and Bidirectional Focus Theory
As stated in Section 3.2, Bergeton’s theory as it stands does not really answer the
question why intensification is necessary when a reflexive object is bound by the
subject of an anti-reflexive or a hidden neutral predicate. On the grounds of the
presupposition or expectation coming with such predicates – that the arguments
refer to different entities – it seems clear that sentences with reflexives constitute presupposition failures or expectation denials. However, it is not clear how
intensification would help; adding focus adds a(nother) presupposition, one verified through the predicate, as we have seen. But that does not undo the first
presupposition failure or cause the expectation denial not to be grave anymore.
The key to a solution to this problem lies in considering what is signalled by
the focus structure of the sentence with the simple reflexive, without intensifier.
This depends on what that focus structure is. As suggested in Section 4.2, there
is in general a choice between assuming that the reflexive is out of focus, cf. (29),
and assuming that it is in focus together with the predicate, cf. (30):
OSLa volume 1(1), 2009
#Han [ beundrer ]F seg.
he admires
  
#Hun [ beundrer seg ]F .
she admires 
Eckardt (2001) appears to only consider the structure corresponding to (29), but
that is probably only because the associates she considers are all anaphoric in
the discourse semantic sense (names, definite descriptions, personal pronouns).
Reflexives are different. And if we choose the structure corresponding to (30), we
have a way of explaining the infelicity of simple reflexives in these contexts.
Jacobs (1999) has developed a theory about the grammar, semantics, and pragmatics of “wide focus” constructions such as those where a predicate and an argument (usually a theme argument) form one focus domain with one accent (on
the more informative constituent; usually the argument, but the predicate if the
argument is e.g. an indefinite pronoun). The less informative constituent is said
to be informationally integrated, or nonautonomous; the predicate and the argument function as one informational unit and must be processed semantically in
one step (Jacobs 1991, 18) and (1999, 68). Here are two typical examples:
They have [ discovered  ]F .
They have [ disered something ]F .
This notion of informational unity is difficult to make precise; in particular, the
semantic and pragmatic – contextual and lexical – constraints on integration have
proven elusive. The basic problem is this: Focus theory has a story to tell about the
constraints on one or more narrow foci versus wide focus, these being relatively
strong focus presuppositions, but in the opposite direction, the focus presupposition is weaker, so a preference for, say, two narrow foci, one for the predicate
and another for the argument, is prima facie mysterious.
Thus (33) will carry the focus presupposition that there is an alternative a1
to stolen and there is an alternative a2 to a ring such that there is a proposition,
sloppily, of the form he has a1 a2 , while (34) will just carry the presupposition that
there is an alternative a3 to stolen a ring such that there is a proposition, sloppily,
of the form he has a3 . The former subsumes the latter.
He has [ en ]F [ a  ]F .
He has [ stolen a  ]F .
However, recent developments in pragmatics, Bidirectional OT (Blutner 1998 and
later work), can provide a bridge between Focus Theory and Informational Integration. The basic idea is that expressions compete with each other for interpretations (and vice versa), and when one expression loses to another for an interpretation, an alternative interpretation wins as the optimal interpretation for
that expression. Thus a sentence with a relatively nonspecific focus interpretaOSLa volume 1(1), 2009
  
tion, for example, (34), will not only not communicate the more specific focus
interpretation of (33); it will actually communicate the opposite, namely, there
are not alternatives to stolen and to a ring – to the VP as a whole there are alternatives, but not to the predicate and the argument separately.
As shown in (Sæbø 2007), this optimal interpretation – an implicature which,
though basically conversational, can become conventional (Blutner 2006) – boils
down to saying that the predicate is sufficiently predictable from the argument a
ring, plus that there are no local alternatives to the predicate in the discourse.
This reasoning carries over to the two competitors (30) and (35), only that
here, the predictability requirement for wide focus is reversed – for (30) to be
felicitous, the argument seg must be sufficiently predictable from the predicate,
plus that the discourse must not overtly provide alternatives to it.
Han [ beundrer ]F [ seg selv ]F .
 
he admires
But of course, the reflexive is not, and cannot be, sufficiently predictable on the
basis of a verb with this meaning. This holds true regardless of discourse relations
otherwise known to facilitate relative predictability.
Generally, complementing the focus presupposition of a P F [ sig  ]F , we
will (in languages like Norwegian), have a (presupposition and an) implicature
for a [ P sig ]F , along the following lines:
    a P F [ sig  ]F
There are propositions φ s. t. there is an R ≈ P and a y ≈ a such that
φ = R(y)(a)
      a [ P sig ]F
There are propositions φ s. t. there is a Q ≈ P (a) such that φ = Q(a),
  such that there is a R ≈ P and a y ≈ a such that φ = R(y)(a)
For P a neutral predicate, when the sentence is uttered relatively out of the blue,
so that there are no overt alternatives to predicate and argument separately, the
implicature will as a rule be appropriate. But for P an anti-reflexive predicate
or a hidden neutral predicate, even when there is no local motivation for one
focus for the predicate and one for the argument, there will always be a global
motivation, as long as there are alternatives to the predicate-argument merge;
neither is sufficiently predictable from the other to justify wide focus. The focus
presupposition of two separate foci will be verified on a lexical-conceptual basis,
and we can assume that this accounts for the necessity of the intensifier.
OSLa volume 1(1), 2009
  
The Role of Self across Languages: A Suggestion
Recall from 2.2. and 3.2 that while in Dutch and Mainland Scandinavian, there are
many predicates where a locally bound  argument must be intensified, in German or Icelandic,  intensification is much more rarely mandatory. A syntactician can maintain that  +  somehow has a tighter structure in the languages
where complex reflexives are often obligatory (Safir 2004, 205). However, if binding is independent of intensification, that strategy is unavailable. The relative
freedom of German , illustrated in (38), presents a problem.
Anstatt sich (selbst) anzuklagen, lernt man sich (selbst) zu akzeptieren
instead  () accuse
learns one  () to accept
und anzunehmen.
and embrace
All three verbs correspond to “hidden neutral” verbs in Mainland Scandinavian,
necessitating ; and it is easy to agree that they are all “other-directed” in the
sense that it is much more common to accuse, accept and embrace somebody else
than to accuse, accept and embrace oneself. As we have seen, this is enough to
force intensification in Mainland Scandinavian (but not in German).
The explanation given in Section 4.4 (based on Bergeton’s analysis) is that
if we do not intensify, we implicate that there are no salient alternatives to the
bindee, globally (lexically) or locally (contextually). Other-directedness in a predicate means that there are salient alternatives, lexically; thus the implicature we
make when we do not intensify is systematically contradicted by the predicate we
use. The same conclusion would seem inescapable for German.
However, reconsider the focus structure assumed for the not intensified cases
in MSc, inappropriate for hidden neutral and anti-reflexive predicates, in 4.4.
Both (29) and (30) were candidates for the not intensified case, alternating with
(35), the intensified case; both were considered but (30) was selected. The difference is this: whereas in (29) the reflexive is out of focus when it is not intensified,
in (30) it is in focus together with the predicate, forming one informational unit.
The main motive for choosing this latter option is that exactly because predicate
and reflexive argument form (in the sense of Jacobs 1999) one informational unit,
the not intensified case is predicted to obey semantic and pragmatic constraints
parallel to those that informational integration between predicate and argument
obey generally: One must be sufficiently predictable from the other; otherwise,
the two must be in focus separately.
When in German, however, the not intensified case is evidently not subject
to the same constraints, there may be reason to reconsider the focus structure
corresponding to (29), cf. (37) and (38):
OSLa volume 1(1), 2009
  
Er [ bewundert ]F sich.
he admires
Anstatt sich ([ selbst ]F ) anzuklagenF , lernt man sich ([ selbst ]F ) zu
instead  ()
learns one  ()
akzeptierenF und anzunehmenF .
and embrace
Assuming that this is what is appropriate for German means that the reflexive
behaves like a personal pronoun as far as focus is concerned. Out of focus means
familiarity, or given information; and in a sense, a reflexive can always be viewed
as given information, binding being what unites “anaphors” and “pronouns”. The
essential clue is the parallelism between the conditions for personal pronoun intensification and those for reflexive intensification in a language like German.
Both are governed by discourse and information structure in the following sense:
Intensification is possible when the resulting focus presupposition is justifiable
globally (lexically) or locally (contextually); intensification is necessary just in
case that presupposition is verified .4 This is the case whenever beside
the discourse function as a continuous topic, the reflexive or personal pronoun
referent acts as a contrastive topic or focus.
For this to account for the contrast, we must assume that a German reflexive
can be out of focus while a Dutch or MSc reflexive, as the object of a predicate
whose subject binds it, is always in focus, usually together with the predicate,
with which it integrates, or with the intensifier.5
It is important to limit the assumption that  is never out of focus to those
cases where it is the internal argument of a predicate whose external argument
binds it. Thus both long-distance  and SC subject  can be out of focus. Here,
the distribution of simple  is much freer than elsewhere and similar to the general pattern in German. This indicates that it is governed more by contextual than
by lexical conditions, as is to be expected if it is not necessary for the reflexive referent to be the expected referent, only that it is topical enough to count as given
information and not overtly contrasted with alternatives.
I have contrasted two widely different approaches to (non)intensive reflexives,
the traditional, syntactically based, line and a novel, semantically based line. For
the former, (Safir 2004) is a good representative. The latter, following up on sug[4]
OSLa volume 1(1), 2009
Crudely put, “sich [ selbst ]F ” is necessary in fewer cases than “[ seg selv ]F ” because there are more alternatives, possible in more cases. But see (König and Siemund 2000) for predicates where the former 
To be sure, there is scarce independent evidence for this – but the same holds for the assumption that
sich selbst fails to form a morphological unit (Safir 2004, 205).
  
gestions by several scholars, has been developed into a coherent theory by Bergeton (2004). In many respects, that work speaks for itself, arguing that a uniform
treatment of intensifiers wherever they occur is theoretically attractive, avoiding
duplication and stipulation, and that it solves the descriptive problems facing the
traditional approach; in particular, it is no longer necessary to defend, in ways so
often running counter to intuition, the postulate of a complementary distribution
between the simple and the complex anaphor, predicting things to be impossible
that common sense says are possible.
On the other hand, the novel approach risks predicting too much flexibility.
What is clear is that the intensifier, introducing a (focus) presupposition (the Contrastiveness Condition in Bergeton’s framework), will impose constraints on the
context or the predicate or both. It is less clear what makes it necessary (and
sufficient) in many contexts and – in some, but not all, relevant languages – with
many predicates. Bergeton ascribes to these predicates a presupposition or expectation to the effect that the subject and object refer to different things, a move
which does not yet answer the open issues.
I have tried to show that the way to complete Bergeton’s account and retain
a viable theory in the novel, pragmatic-semantic line is to embed it in a formal
theory of focus and intensification, exploiting the options offered in this theory,
and to complement it with Bidirectional Optimality Theory. Not only does this
strategy seem to yield the right results regarding the discourse-triggered necessity of adreflexive intensification generally and its predicate-triggered necessity
in a language like Norwegian; it offers a suggestion about the absence of this latter necessity in a language like German. Thus the novel theory, dividing the labor
of complex anaphors between syntax and information structure, is intensified.
Bergeton, U. 2004. The Independence of Binding and Intensification. Ph. D.thesis, University of Southern California.
Blutner, R. 1998. Lexical Pragmatics. Journal of Semantics 15, 115–162.
Blutner, R. 2000. Some Aspects of Optimality in Natural Language Interpretation.
Journal of Semantics 17, 189–216.
Blutner, R. 2002. Lexical Semantics and Pragmatics. In F. Hamm and E. Zimmermann (eds.), Linguistische Berichte Sonderheft, volume 10, pages 27–58, Hamburg:
Helmut Buske Verlag.
Blutner, R. 2004. Pragmatics and the lexicon. In L. R. Horn and G. Ward (eds.), The
Handbook of Pragmatics (Blackwell Handbooks in Linguistics), pages 488–514, Oxford: Blackwell Publishers.
OSLa volume 1(1), 2009
  
Blutner, R. 2006. Embedded Implicatures and Optimality Theoretic Pragmatics. In
T. Solstad, A. Grønn and D. Haug (eds.), A Festschrift for Kjell Johan Sæbø, pages
11–29, University of Oslo.
Bouchard, D. 1984. On the Content of Empty Categories. Dordrecht: Foris.
Eckardt, R. 2001. Reanalysing selbst. Natural Language Semantics 9, 371–412.
Edmondson, J. and Plank, F. 1978. Great Expectations: An intensive self analysis.
Linguistics and Philosophy 2, 373–413.
Gast, V. 2006. The Grammar of Identity: Intensifiers and Reflexives in Germanic Languages. Routledge Studies in Germanic Linguistics, Address: Routledge (Taylor
and Francis).
Hellan, L. 1988. Anaphora in Norwegian and the Theory of Grammar. Dordrecht: Foris.
Jacobs, J. 1991. Focus Ambiguities. Journal of Semantics 8(1-2), 1–36.
Jacobs, J. 1999. Informational Autonomy. In P. Bosch and R. van der Sandt (eds.),
Focus: Linguistic, Cognitive and Computational Perspectives, Cambridge University
Press, 56-81.
König, E. and Siemund, P. 2000. Reflexives: Forms and Functions. Philadelphia: John
Benjamins, 41-74.
Lødrup, H. 2007. A New Account of Simple and Complex Reflexives. Journal of Comparative Germanic Linguistics 10, 183–201.
Mattausch, J. 2003. Optimality theoretic pragmatics and binding phenomena. In
R. Blutner and H. Zeevat (eds.), Optimality Theory and Pragmatics, pages 63–91,
Basingstoke: Palgrave Macmillan.
Moravcsik, E. 1972. Some cross-linguistic generalizations about intensifier constructions. CLS 8, 217–277.
Pica, P. 1985. Subject, tense and truth: Towards a modular approach to binding.
In H. Obernauer and J.-Y. Pollock (eds.), Grammatocal Representation, Dordrecht:
Foris, 259-291.
Reinhart, T. and Reuland, E. 1993. Reflexivity. Linguistic Inquiry 24, 657–720.
Rooth, M. 1992. A Theory of Focus Interpretation. Natural Language Semantics 1,
Safir, K. 2004. The Syntax of Anaphora. Oxford Studies in Comparative Syntax, New
York: Oxford University Press.
OSLa volume 1(1), 2009
  
Siemund, P. and Gast, V. 2006. Rethinking the relationship between selfintensifiers and reflexives. Linguistics 44, 343–381.
Sæbø, K. J. 2005. Intensifying self in Germanic: A Reanalysis. In The Stockholm Workshop on Contrast, Information Structure and Intonation.
Sæbø, K. J. 2007. Focus Interpretation in Thetic Statements: Alternative Semantics
and OT Pragmatics. Journal of Logic, Language and Information 16(1), 15–33.
Zeevat, H. 2004. Markedness. In Asatiani R., R. K. Balogh, G. Chikoidze, P. Dekker
and D. De Jongh (eds.), Proceedings of the 5th International Symposium on Language,
Logic and Computation, Tbilisi, 183-190.
Zribi-Hertz, A. 1995. Emphatic or reflexive? On the endophoric character
of French lui-même and similar complex pronouns. Journal of Linguistics 31,
  
Kjell Johan Sæbø
University of Oslo
Dept. of Literature, Area Studies and European Languages
P.b. 1003, Blindern
N-0315 Oslo
[email protected]
OSLa volume 1(1), 2009
OSLa volume 1, 2009
Behrens & Fabricius-Hansen (eds.) Structuring information in discourse:
the explicit/implicit dimension, Oslo Studies in Language 1(1), 2009. 131-148. (ISSN 1890-9639)
 do 
   
University of Oslo
The paper provides a first analysis of one use of the German discourse marker
doch, namely doch as conjunct adverb (CA). As CA, doch has two properties,
concessivity and anaphoricity, that distinguish it from the conjunction doch,
which is the other clause connecting use of the word. As conjunction, doch
is not anaphoric and it may mark a greater variety of (contrastive) discourse
relations besides concession. I argue that CA doch acquires anaphoricity and
concessivity due to a combination of factors related to the fact that it is
accented. More specifically, I claim that (i) accent on CA doch leads to an
asymmetric focus-background structure of the doch-conjunct which in turn
leads to the concessive interpretation of the discourse relation marked by
doch; (ii) accent on CA doch evokes as a focus alternative the negation of the
conjunct to which doch is attached and (iii) focus on CA doch is contrastive,
which means that the alternative CA doch evokes is anaphorically linked to
an antecedent in the preceding discourse. I also sketch a formal account of
the concessivity and anaphoricity of CA doch.
The goal of this paper is to provide an analysis of one use of the German discourse
marker doch, namely of doch as conjunct adverb (CA). An example of CA doch is
Draußen vor dem Zelt stand Artax, sein Pferd. Es war gefleckt und klein wie
ein Wildpferd, [seine Beine waren stämmig und kurz]C1 , und  [war es
This work is supported by the SPRIK project (Språk i Kontrast [Languages in Contrast], NFR 158447/530).
For valuable comments, I would like to thank the members of the SPRIK group and especially Cathrine
Fabricius-Hansen and Torgrim Solstad, the audience of the conference SPRIK 06, especially Regine
Eckardt, Carla Umbach and Henk Zeevat, as well as an anonymous reviewer.
This and the remaining examples in this section, including their English translations, are from the Oslo
Multilingual Corpus (, a parallel multilingual (Norwegian, English,
German, French and Russian) corpus of mainly fictional original texts and their translations, developed
at the University of Oslo within the project SPRIK (
sprik/). The indices are mine. The small capital letters denote (nuclear) accent and are also mine.
 
der schnellste und ausdauerndste Renner weit und breit]C2 .
‘His horse, Artax, was standing outside the tent. He was small and spotted
like a wild horse. His legs were short and stocky, but he was the fastest,
most tireless runner far and wide.’
As it becomes obvious from the example, this use of doch has the following properties:
(i) First, doch occupies the initial field (Vorfeld) of the German sentence, which
is the typical position of conjunct adverbs.
(ii) Second, it functions as a sentence connector linking the two conjuncts C1
and C2 to each other.
(iii) Third, it has a concessive meaning synonymous with trotzdem (‘despite that’).
This means roughly that doch specifies the relation between C1 and C2 as
being such that normally one would not expect C2 given that C1 is the case,
i.e. if a horse has short legs, then we do not expect it to be the fastest runner
far and wide.
(iv) Fourth, doch establishes an anaphoric link to the preceding clause C1 , which
is a general property of conjuct adverbs (Duden 2005).
(v) And finally, it is accented.
The above properties, taken together, distinguish CA doch clearly from the other
grammatical uses of the word: the conjunction doch (as in (2)), the unaccented
modal particle doch (3), the response particle doch (4) and the accented adverbial
doch (5):
Abstraktes und Abstraktionen lagen ihr zwar nicht gänzlich fern, doch die
“Hackschrift”, wie sie Stenografie nannte, mochte sie nicht lernen.
‘It was not that abstract thinking and abstractions were entirely without
interest for her, but “chopped-up writing,” her name for shorthand, was
something she had no wish to learn.’
Ich sage: “Der Mensch ist doch kein Flußbett.”
‘ “A person isn’t just a riverbed, is he?” I say.’
Ich sagte schnell: “Doch, doch, ich weiß.”
‘I quickly said: “Oh, of course I do.” ’
Ich weiß nicht, warum ich schließlich doch geheiratet habe.
‘I don’t know why I did marry in the end.’
OSLa volume 1(1), 2009
 do  
Note that CA doch is always accompanied by und (‘and’), thus being part of a bipartite connector und doch, in which we can observe the following division of labour:
The coordinating conjunction und provides for the syntactic link between C1 and
the clause C2 in which CA doch is syntactically integrated, whereas CA doch contributes a semantically richer relation between C1 and C2 . This division of labor
is necessary because adverbial connectors are syntactically (one-place) modifiers,
but take semantically two arguments, cf. Pasch et al. (2003).
The main question that the paper addresses is why doch as conjunct adverb
has the properties under (iii) and (iv). This question is motivated by a comparison between doch as conjunct adverb and doch as ordinary conjunction, which
shows that the conjunction doch is not anaphoric and it may mark a greater variety of (contrastive) discourse relations besides concession. More specifically,
the question that I attempt to answer is: Given that the discourse marker doch is
not always anaphoric, nor always concessive, what is it that makes this particular
use of doch capable of, and at the same time restricts it to, rendering the same
interpretation as anaphoric concessive CAs like trotzdem?
In spite of recent revived interest in the semantics and pragmatics of connectors in general and adversative connectors in particular, CA doch has to my
knowledge not been given any closer attention before. It has even been completely ignored in a comprehensive work such as Pasch et al. (2003). The analysis
I propose is in this sense a first attempt to understand the properties of CA doch
that distinguish it from other uses of the word.
In a nutshell, I will argue that CA doch acquires anaphoricity and concessivity due to a combination of factors related to the fact that it is accented. More
specifically, I claim that
• accent on CA doch leads to an asymmetric focus-background structure of
the doch-conjunct which in turn leads to the concessive interpretation of
the discourse relation marked by doch.
• accent on CA doch evokes as a focus alternative the negation of the conjunct
to which doch is attached.
• focus on CA doch is contrastive, which means that the alternative CA doch
evokes is anaphorically linked to an antecedent in the preceding discourse.
The paper is structured as follows: First I will elaborate on the differences between the conjunct adverb and the conjunction doch with respect to the discourse
relation they mark (section 2). Then, I will provide my explanation of the concessivity and anaphoricity of CA doch (sections 3 and 4). In section 5 I sketch a formal
account of the concessivity and anaphoricity of CA doch. Finally, section 6 provides summary and conclusions.
OSLa volume 1(1), 2009
 
   
There are basically three kinds of contrastive relations assumed in the literature.
Here I will use the term “concession” for what is also called “denial of expectation”, e.g., König (1991), Lagerwerf (1998). Following Kruijff-Korbayová and Webber (2005), I reserve Spooren’s (1989) “concessive opposition” for the argumentative interpretation of contrast observed in Anscombre and Ducrot (1977) and
Dascal and Katriel (1977). The latter is induced by a context in which a claim is
given for which the aber/doch/etc.-construction provides both arguments pro and
It seems that CA doch marks a contrastive discourse relation that can only
be interpreted in terms of concession, unlike the conjunction doch that can also
mark semantic opposition and concessive opposition. Let’s turn to some examples
which illustrate this point.
S . An example of semantic opposition is (6), where two
subjects with two different properties are contrasted:
Hans ist reich, doch Peter ist arm.
‘Hans is rich but Peter is poor.’
Replacing the conjunction doch with CA doch renders the sentence rather awkward, since CA doch forces a concessive reading that is a bit counterintuitive, or
at least requires a very specific context, i.e. one in which Hans is the father of
(7) ??Hans ist reich, und  ist Peter arm.
‘Hans is rich, and yet Peter is poor.’
C . An example for this contrastive relation is (8) where the
first conjunct provides an argument for and the second an argument against taking the room in question:
(Wollen wir das Zimmer nehmen?)
(‘Shall we take this room?’)
Es hat einen tollen Ausblick, doch der Preis ist zu hoch.
‘It has a beautiful view, but it is very expensive.’
Here as well, the CA doch is not an appropriate connector, cf. (9):
(Wollen wir das Zimmer nehmen?)
??Man hat einen tollen Ausblick, und  ist der Preis zu hoch.
‘It has a beautiful view, and yet it is very expensive.’
The reason for that is the same as in (7): the hearer is forced to reconstruct a
OSLa volume 1(1), 2009
 do  
relation between the first and the second conjunct which is such that if the first
conjunct is true, then normally the second conjunct is not true. The utterance is
awkward, since the suggested connection between the two states of affairs is not
plausible, i.e. rooms with a nice view are rather expected to be expensive.
In order to confirm the intuition that CA doch is specialised in marking concession, I carried out a small scale corpus study involving an analysis of all occurrences of CA doch in the Oslo Multilingual Corpus. From 21 tokens of CA doch in a
clause connecting function2 out of 375 doch corpus matches, all had a concessive
My diagnostics for determinig the discourse relation that CA doch marks in the
corpus examples is based on Umbach’s (2001a) distinction between four structural
cases of but-conjunctions:
(i) two different predicates are being contrasted with respect to the same subject,
(ii) two different subjects are being contrasted with respect to the same predicate,
(iii) different but comparable subjects and predicates are being contrasted, and
(iv) the subjects and predicates are not comparable, i.e. the entire propositions
are contrasted.
Cases (i)-(iii) cover the cases known as semantic opposition. Case (iv) covers concession and concessive opposition, where the latter is dependent on a contextually given tertium comparationis.
I also looked at the English and Norwegian translations of the 21 CA doch corpus matches. In the English translations, the majority (17 of 21) of the occurrences
of CA doch were translated with concessive markers (16 with yet, one with still),
cf. table 1.3
In the Norwegian translations, the majority (16 of 21) of the instances of CA
doch were also translated with concessive markers. Of them, 15 were translated
with (al)likevel (‘yet’, ‘nevertheless’, ‘although’), with or without other co-occurring
connectors, and one with selv om (‘even if’), cf. table 2.4
CA doch can also connect phrases.
The term adversative used in tables 1 and 2 is used as a general, cover term for all three relations of
contrast discussed above.
The translation with forresten (‘by the way’) is a bit strage and could be explained by the fact that the
first conjunct of the relation is somewhat obscure in the original. In the English translation, however,
the concessive and yet is used. One sentence has no Norwegian translation in the corpus.
OSLa volume 1(1), 2009
 
. 1: English translations of conjunct adverb doch
and 
. 2: Norwegian translations of conjunct adverb doch
og 
og 
 
men 
men ... 
OSLa volume 1(1), 2009
 do  
do               
My first claim is that the accent on CA doch plays a role in specifying the interpretation of the relation between the conjuncts as concession.
It has been argued (Umbach 2001a, 2004; Lang 2004; Lang and Adamíková 2005)
that concession can be seen as contrast between two propositions that leads to
specifying the conjuncts linked by adversative connectors like aber and but as focus alternatives with respect to each other.
Moreover, Lang and Adamíková (2005) argue that the focus-background structure of the sentence (FBS) specifies which entities are contrasted, which in turn
determines the interpretation of the contrast relation as one of either concession
or semantic opposition. They observe the following correlation between the FBS
of the conjuncts and the interpretation of the relation between the conjuncts: if
the FBS is parallel, then the relation is interpreted as semantic opposition, if it is
asymmetric, then the relation is concessive.
Consider (10), where the FBS is parallel and involves two topics and two foci:5
[Mein Vater]T [ist ernsthaft krank]F , doch [meine Mutter]T [geht arbeiten]F .
‘My father is seriously ill, but my mother goes out to work.’
Such a parallel FBS renders a semantic opposition interpretation of contrast. In
(11), we have two foci comprising the entire conjuncts:
[Mein Vater ist ernsthaft krank]F , doch [meine Mutter geht arbeiten]F .
Here, the FBS is asymmetric which renders a concessive interpretation of the contrast relation.
A similar correlation is observed by Umbach (2001a): “in a but-conjunction,
there are two corresponding foci (in the first and in the second conjunct, respectively) which establish alternatives with respect to each other”. The corresponding foci may encompass the entire conjuncts, i.e. the entire propositions
are alternatives to each other (cf. case (iv) of Umbach’s structural cases of butconjunctions on page 135).
Returning to CA doch, the fact that it is accented seems to have consequences
for the FBS of the entire construction. Thus, it does not seem to allow a parallel FBS, cf. (12-a), contrary to the unaccented conjunction doch in (12-b), which
allows for both kinds of FBS (as illustrated by (10) and (11):
a. #[H]T [ist ]F , und  [ist P]T []F .
b. [H ]T [ist ]F , doch [P]T [ist ]F .
This suggests that given that the asymmetric FBS of the sentence determines a
The example is taken from Lang (2004) where the synonymous connector aber is used instead of doch.
OSLa volume 1(1), 2009
 
concessive reading, the accent on doch is crucial for this interpretation, since it
determines the asymmetry of the FBS in the first place.
do               
According to standard theories of focus, an accented constituent is a focussed constituent evoking focus alternatives. In the alternative semantics of Rooth (1992),
a focused expression is accounted for by assuming that it adds an additional sef
mantic value [[.]] of the sentence next to its ordinary semantic value [[.]] . The
focus semantic value represents a set of alternatives, i.e. a set of propositions
which contrast with the ordinary semantic value. The ordinary semantic value is
always an element of the focus semantic value. The set of alternatives is salient
but not necessarily explicitly mentioned and contains only alternatives which are
type-identical with the focussed expression.
Since CA doch is accented and hence focussed, the obvious question to ask is
what alternatives it evokes. Now, the ordinary semantic value of a sentence of
the form ‘ C2 ’ is [[p]] , where p is the proposition expressed by C2 , since
discourse markers like doch do not influence the truth conditions of the sentence.
Intuitively, the alternative that accented doch evokes is the sentence negation
nicht. E.g., the alternative evoked in (13-a) is (13-b).
Es war gefleckt und klein wie ein Wildpferd, [seine Beine waren stämmig und kurz] C1 , und [ war es der schnellste und ausdauerndste Renner weit und breit]C2 .
it is not the case that he was the fastest, most tireless runner far and
In other words, the focus alternative evoked by a sentence of the form ‘ C2 ’
seems to be the logical complement of the proposition p expressed of the dochconjunct. The focus semantic value of such a sentence is then the set containing
its ordinary semantic value [[p]] and its sole focus alternative ¬p:
[[[doch]F p]] = {p, ¬p}
I suggest that the anaphoricity of CA doch can be accounted for by treating the
focus it carries as contrastive focus. According to Rooth (1992), in the case of
contrastive focus, accent signals that the focussed expression contrasts with a
previously uttered member of the focus set of alternatives, i.e. a phrase α is cono
trasting with a phrase β, if [[β]] ∈ [[α]] . In other words, contrastive focus signals
that the focus alternative evoked by a focussed expression is anaphorically linked
to the preceding context.
The classical Roothian example of contrastive focus is (15), where “Canadian
farmer” evokes a set of focus alternatives containing farmers of different nationOSLa volume 1(1), 2009
 do  
alities and where one member of this set, namely “American farmer” is provided
by the preceding context:
An AmericanF farmer was talking to a CanadianF farmer.
In the case of CA doch, the relevant member of the focus set of alternatives is the
negation ¬C2 of the doch-conjunct C2 . Contrastive focus signals that the evoked
alternative ¬C2 should be linked to the preceding context, i.e. that a matching
antecedent should be available. The alternative ¬C2 is not explicitly provided by
the preceding context, but is suggested as a default consequence of the state of
affairs expressed by the first conjunct (due to the concessive interpretation of the
contrast relation). In other words, the anaphoric antecedent to which the focus
alternative ¬C2 is linked is implicit.
Rooth (1992) notes a similar case of contrastive focus involving an implicit
anaphoric antecedent. He observes that in (16), the entailment relationship can
mediate contrastive focus:
he1 called her2 a Republican and then 2,F insulted 1,F .
In (16), contrastive focus on the pronouns she and him in the sentence she insulted
him leads to evoking as a focus alternative the sentence he insulted her, i.e.
[[she insulted him]] = {she insulted him, he insulted her}.
However, the contrasting expression he insulted her is not explicitly mentioned
but must be derived from he called her a Republican, i.e. a presupposed axiom
has to be assumed that to call someone a Republican is to insult him, such as
∀x∀y[call-a-Republican(x, y) → insult(x, y)]. Rooth suggests therefore that the
definition of contrastive focus must be modified to the following: a phrase α is
contrasting with a phrase β, if [[β]] entails some element of [[α]] , i.e. [[β]] |=
[[γ]] such that [[γ]] ∈ [[α]] . Then the contrast between she insulted him and
he called her a Republican is indirect and consists in a contrast between the contrastively focussed she insulted him and its focus alternative he insulted her, where
the focus alternative is an entailment of he called her a Republican.
Applying the modified definition of contrastive focus to CA doch, we need a
similar presupposed axiom for the derivation of ¬C2 from C1 . This axiom is provided by the concessive interpretation of the contrast relation, i.e. C1 > ¬C2 .
Here, the relation is however not material implication but rather a defeasible implication (‘>’ (Asher and Morreau 1991)), as is usually assumed in the literature
on concession, cf. e.g. (Lagerwerf 1998). Then, C2 is contrasting with C1
since ¬C2 ∈ [[[doch]F C2 ]] and C1 > ¬C2 .
OSLa volume 1(1), 2009
 
                         do
So far I argued that accent on CA doch is responsible for rendering the concessive
interpretation of the contrast relation that doch marks. I also argued that focus
on doch is contrastive and that the anaphoricity of contrastive focus can be made
responsible for the anaphoric properties of CA doch.
In other words, I claim that concessivity and anaphoricity are not some inherent properties of the lexical item doch, as also the comparison to the conjunction
doch in section 3 showed.
I also argued that in the case of CA doch, the contrast between the doch-conjunct
and the first conjunct consists in linking the focus alternative that doch evokes to
a defeasible implication of the first conjunct, i.e. to an implicit anaphoric antecedent.
Consequently, a formal account of CA doch must address the following questions:
• How can the meaning of doch be specified?
• How can the anaphoric resolution to an implicit antecedent be modelled?
[5.1] The meaning of adversative conjunctions
Lang (Lang 2004; Lang and Adamíková 2005) provides a framework for the interpretation of adversative constructions with aber that accounts both for the meaning of the conjuncts, including their FBS, and the meaning of the connectors.
I will adopt this framework to account for the lexical meaning of the adversative connector doch.
Lang formalizes the (non truth-conditional) meaning of adversative connectors in terms of a presupposition variable q these connectors introduce which is
such that it is a member of the (focus) set of alternatives of the conjunct they are
attached to and meets a preference condition (where ‘≫’ is a preference relation
“is preferred to” and p1 and p2 are the propositions expressed by the first and the
second conjunct respectively):
∃q[q ∈ ALT (p2) : [[p1 & q] ≫ [p1 & p2]]
The preference condition is formulated such that it covers all possible interpretations of contrast, esp. semantic opposition and concession (Lang does not mention concessive opposition). The variable q is to be instantiated by means of contextual information: (i) structural, (ii) discourse or (iii) world knowledge (where
q 6= p1 and q 6= p2). The search for a suitable instance proceeds from (i) to (ii) to
In the case of concession, the asymmetric FBS indicates that some proposition
p has to be inferred from outside the construction. More closely, the asymmetric
OSLa volume 1(1), 2009
 do  
FBS triggers a world knowledge based inference process. This inference process
leads to inferral of some proposition p which instantiates the variable q introduced by the adversative connector.6 The inference process is restricted by a
selection principle, a structural specification for restricting q to the most obvious
element ¬p of ALT (p2) for a given p.
Lang does not mention doch but his formalisation is intended to cover adversative connectors and constructions in general. Applying his system to our example
with doch as a conjunction, the concessive interpretation is derived as in (19-h).
In the derivation below, the propositional variables p1 and p2 denote the meaning
of the conjuncts and variables of type /pn / denote inferred propositions. I use the
defeasible implication sign ‘>’ instead of the original relation ’⇒’ (where ’p ⇒ q’
denotes “given our everyday experience, p is expected to have the consequence
q”). Step b. represents the meaning of doch, step c. the selection principle, d. is
the result of the application of the selection principle on the meaning of doch, and
steps e.-h. represent the inference process based on world knowledge as a result
of which the variable q becomes instantiated.
[Seine Beine waren kurz]F , [doch es war der schnellste Renner weit
und breit]F
∃q[q ∈ ALT (p2) : [[p1 & q] ≫ [p1 & p2]] (non truth-functional
meaning of doch)
∀q∀p[q ∈ ALT (p) ≡ q → ¬p] (selection principle)
∃q[q → ¬p2 : [p1 & q] ≫ [p1 & p2]] (selection principle applied)
/p1/ = ∀x ∈ {creatures’}[short-legged’(x) > slow’(x)]
/p2/ = ∀x ∈ {creatures’}[slow’(x) > ¬fast’(x)]
/p3/ = ∀x ∈ {creatures’}[short-legged’(x) > ¬fast’(x)]
/p4/ = ¬he was the fastest runner (=q), i.e. /p1 > q/ from e.–g.,
controlled by q → ¬p2
There are however strong arguments against basing the concessive interpretation on a world-knowledge inference: there are cases in which we do not know
that a default rule exists, but we learn it from the speaker, as pointed out in
Umbach and Stede (1999) who give the following example:
Es war Juli, aber wir haben keine Safranschirmlinge gefunden.
‘It was July, but we did not find any shaggy parasols.’
Even if the hearer has not the faintest idea of what a “Safranschirmling” is (it is a
kind of mushroom), he understands that normally one can find them in July.
Therefore, a different explanation for the concessive interpretation is needed
Lang and Adamíková (2005) claim that contrastive focus accents in the absence of adversative connectors
may also introduce a variable q like the one introduced by adversative connectors.
OSLa volume 1(1), 2009
 
that we cannot provide here. One possibility is already hinted at by Lang when he
claims that prosody is the decisive factor in determining the FBS of the construction and hence the reading of the connector. If this is so, then we do not need
to access world knowledge at all. We would know from prosody that q has to be
instantiated by some p which is such that q ∈ ALT (p) ∧ q → ¬p.
In the case of CA doch, access to world knowledge is clearly gratuitous, as is
the derivation of some inference p constrained by a selection principle. With CA
doch, the variable q is directly instantiated by the focus alternative evoked by
accented doch, namely ¬p2. According to Lang’s specification of the meaning of
adversative connectors, q must be a proposition which is a member of the set of
alternatives evoked by the focussed element, i.e. ∃q[q ∈ ALT (doch(p2))]. Since
q cannot be instantiated by p1 according to the definition (q 6= p1), a second (and
only) possibility is that q is instantiated by the other element of the focus set of
alternatives of the construction ‘ C2 ’, namely ¬p2. The preference relation,
which is part of the meaning of doch, tells us that p1&¬p2 should come closer to
what the speaker expects, rather than what obtains, namely p1 & p2. This can be
translated as p1 > ¬p2.
[5.2] Anaphoric resolution with implicit antecedents
In this section I adopt a solution to the question of how the anaphoric resolution
to an implicit antecedent that seems to take place in the case of CA doch can be
modelled. My proposal is based on Bos et al. (1995) who extend van der Sandt’s
theory of presupposition with the notion of bridging anaphora. An example of a
bridging anaphor is (21):
When I go to a bar, the barkeeper always throws me out.
Here, the NP the barkeeper triggers an existential presupposition. The preceding sentence however does not provide an overt antecedent, i.e. a barkeeper,
to which the presupposed material can be linked. But the anaphor goes through
because of our world knowledge telling us that a bar has a barkeeper. This knowledge is modelled in Bos et al. (1995) as lexical information in terms of Pustejovsky’s
(1991) qualia structure, which is a set of lexical entailments modelling the partial
meanings (polysemy) of lexical items. The qualia structure is made available by
means of accommodation when necessary, like in the case of bridging anaphora
where it plays the role of antecedent. Here is how a DRS for (21) looks like:
OSLa volume 1(1), 2009
 do  
When I go to a bar, the barkeeper always throws me out.
α : barkeeper(y)
Q : barkeeper(z)
The representation captures the information that there is a bar the speaker goes
to; the Q-box contains the qualia information that a bar has a barkeeper, and the
definite NP the barkeeper is a presupposition trigger that introduces the anaphoric
information α. The qualia information is accommodated, i.e. it enters the main
DRS, and the presupposed barkeeper is linked to the barkeeper that is part of the
qualia, i.e. bridging yields the resolved DRS where the presupposed barkeeper y
is linked to the qualia-barkeeper z:
When I go to a bar, the barkeeper always throws me out.
A similar solution is possible for the anaphoric resolution of the focus alternative
evoked by CA doch. Here we need the axiom p1 > ¬p2 to become available as the
antecedent of the focus alternative ¬p2 evoked by focussed CA doch. The axiom
becomes available due to the concessive interpretation of the discourse relation,
OSLa volume 1(1), 2009
 
as we saw in section 5.1. The presupposed material is accommodated, i.e. added
to the main DRS and thus made available for anaphoric resolution.
Applied to our example, the technique described above renders the following
Seine Beine waren kurz, und doch war es der schnellste Renner.
¬ fastest-runner(x)
The DRS reflects the information that there is a horse which has short legs and is
the fastest runner. The presupposed information is that normally, if the/a horse
is short-legged, then it is not the fastest runner.7
After accommodating the default rule, it becomes part of the main DRS and
its consequent is available as the antecedent for the alternative evoked by doch:
OSLa volume 1(1), 2009
Standard DRT does not include defeasible implication, so a more elaborated formalisation should be concerned with integrating it into the theory. This is however outside the scope of the present proposal.
 do  
¬ fastest-runner(x)
The defeasible implication prevents by its definition the inference of information
that contradicts the information accumulated in the hitherto context, here that
the horse is not the fastest runner. Otherwise, the accommodation of the axiom
would have been prohibited by van Sandt’s (1992) consistency constraint.
The anaphoric resolution happens when the focus alternative ¬fastest-runner(x) enters the DRS. Focus alternatives are generally treated as presuppositional (cf. e.g. Umbach 2001b; Bende-Farkas et al. 2003). I.e. we have the representation below:
¬ fastest-runner(x)
α: ¬
OSLa volume 1(1), 2009
 
The presupposed focus alternative above is locally bound by the defeasible rule,
i.e., the final representation will be the one in (25) above.8 The defeasible relation
in the concessive axiom makes sure that the global DRS stays consistent, since it
prevents the inference of information that would contradict information already
contained in the context. This accounts for the difference between CA doch and
cases of corrections with accented doch, such as (27), where the context must be
updated with contradicting information and where a context revision is required
in order to keep the DRS consistent, cf. Maier and van der Sandt (2003).
A: Karl war nicht auf meiner Party. ¬p
‘Karl was not at my party.’
B: Karl war  auf der Party. p
‘He was indeed.’
  
I suggested an analysis of what I claimed was the anaphoricity and concessivity of
German CA doch. I argued that these properties are acquired by the CA doch rather
than being inherent to the lexical item doch, and that both properties can be ascribed to the effects of the prosodic accent that CA doch carries. I also sketched
a formal account of the concessivity and anaphoricity of CA doch. My account
relates CA doch to its use as conjunction as well as to the remaining uses of doch
where it receives a correction interpretation.
Further work must be carried out to spell out more precisely the details of the
analysis, as well as to elaborate on the relation between the different uses of the
discourse marker doch.
Anscombre, J.-C. and Ducrot, O. 1977. Deux mais en français? Lingua 43, 23–40.
Asher, N. and Morreau, M. 1991. Commonsense entailment: a modal theory of
nonmonotonic reasoning. In IJCAI’91, Proceedings of the Ninth International Joint
Conference on Artificial Intelligence.
Bende-Farkas, Á, von Genabith, J. and Kamp, H. 2003. DRT: An updated survey, ms.
Bos, J., Buitelaar, P. and Mineur, A.-M. 1995. Bridging as Coercive Accommodation.
In Computational Logic for Natural Language Processing–(Workshop Proceedings).
OSLa volume 1(1), 2009
A problem with this solution that was pointed out to me by an anonymous reviwer is that in standard
DRT, material on the right-hand side of conditionals is not available for presuppositions triggered in the
main DRS. The question whether this restriction can be relaxed for cases involving defeasible implication
must be however postponed for further work.
 do  
Dascal, M. and Katriel, T. 1977. Between Semantics and Pragmatics: The two Types
of “but” - Hebrew “aval” and “ela”. Theoretical Linguistics 4, 143–172.
Duden. 2005. Duden. Die Grammatik. Mannheim: Dudenverlag.
Kruijff-Korbayová, I. and Webber, B. L. 2005. Interpreting Concession Statements
in Light of Information Structure. In H. Bunt and R. Muskens (eds.), Computing
Meaning, volume 3, pages 145–172, Kluwer.
König, E. 1991. Konzessive Konjunktionen. In A. von Stechow and D. Wunderlich
(eds.), Handbuch der Semantik, Berlin, New York: Walter de Gruyter.
Lagerwerf, L. 1998. Causal Connectives have Presuppositions. Ph. D.thesis, Tilburg.
Lang, E. 2004. Schnittstellen bei der Konnektoren-Beschreibung. In H. Blühdorn,
E. Breindl and U. H. Waßner (eds.), Brücken schlagen, Walter de Gruyter.
Lang, E. and Adamíková, M. 2005. The lexical content of connectors and its interplay with intonation. An interim balance on sentential connection in discourse.
In A. Späth (ed.), Interfaces and Interface Conditions, Berlin-New York: Walter de
Maier, E. and van der Sandt, R. 2003. Denial and correction in layered DRT. In
Ivana Kruijff-Korbayová and Claudia Kosny (eds.), Proceedings of DiaBruck, (The
7th Workshop on the Semantics and Pragmatics of Dialogue), Saarbrücken.
Pasch, R., Brauße, U., Breindl, E. and Waßner, U. H. (eds.). 2003. Handbuch der
deutschen Konnektoren. Schriften des Instituts für deutsche Sprache, Berlin - New
York: Walter de Gruyter.
Pustejovsky, J. 1991. The generative lexicon. Computational Linguistics 17(4).
Rooth, M. 1992. A Theory of Focus Interpretation. Natural Language Semantics (1),
Spooren, W. P. M. S. 1989. Some Aspects of the Form and Interpretation of Global Contrastive Coherence Relations. Ph. D.thesis, University of Nijmegen.
Umbach, C. 2001a. Contrast and Contrastive Topic. In Proceedings of the ESSLLI 2001
Workshop on Information Structure, Discourse Structure and Discourse Semantics.
Umbach, C. 2001b. Restriktion der Alternativen. Linguistische Arbeitsberichte (77),
Umbach, C. 2004. On the Notion of Contrast in Information Structure and Discourse Structure. Journal of Semantics 21(2), 155–175.
OSLa volume 1(1), 2009
 
Umbach, C. and Stede, M. 1999. Kohärenzrelationen: Ein Vergleich von Kontrast
und Konzession. KIT-Report 148, Technische Universität Berlin, Berlin.
van Sandt, R. 1992. Presupposition Projection as Anaphora Resolution. Journal of
Semantics (9), 333–377.
  
Elena Karagjosova
University of Oslo
Dept. of Literature, Area Studies and European Languages
P.b. 1003, Blindern
N-0315 Oslo
[email protected]
OSLa volume 1(1), 2009
Behrens & Fabricius-Hansen (eds.) Structuring information in discourse:
the explicit/implicit dimension, Oslo Studies in Language 1(1), 2009. 149-170. (ISSN 1890-9639)
     
  alors
University of Toulouse
This paper presents an analysis of the French adverbial alors (then, at that
time, so). Among the linguistic markers that establish a temporal relation
between the eventualities introduced by two clauses, we define as temporal connectives those that introduce at the same time some sort of discourse
relation. We argue that alors can be such a connective. Within the framework of Segmented Discourse Representation Theory, our investigation concerning the discourse information conveyed by sentence initial alors reveals at
least three sorts of relations close to the Result relation of SDRT. Alors alone
conveys a weak causal relation, which we formalize using Lewis’ counterfactual, and encode in the discourse relation Weak-Result. We distinguish it
from Strong-Result, which is inferred when lexical or other contextual information suggests a causal, discursive link. We also show that alors can,
when Weak-Result is blocked, suggest an inferential relation, which we express using the weak conditional already present in SDRT.
Et qu’est-ce que l’autre lui raconte ! Qu’il devrait ajouter un bouton ! Oui ! Un bouton à son pardessus ! à son pardessus ! alors. alors
l’autobus est arrivé. Alors j’ai monté dedans. Alors j’ai vu un citoyen
qui m’a saisi l’oeil. Alors j’ai vu son long cou et j’ai vu la tresse qu’il
y avait autour de son chapeau. Alors il s’est mis à pester contre son
voisin […] R Q, Exercices de style, 1947)
This paper presents an analysis of the French temporal connective alors (generally translated in English by then, at that time, so). It is part of a broader project
aimed to provide a systematic analysis of French temporal connectives within the
formal framework of Segmented Discourse Representation Theory (Asher 1993; Asher
and Lascarides 2003). We define as a temporal connective any adverb that expresses
,    
a temporal (or aspectuo-temporal) relation between two eventualities and, at the
same time, implies a logico-pragmatic relation between the two utterances in
which the eventualities are described (Le Draoulec and Bras 2006). In other words,
temporal connectives are adverbs that play a role at the discourse level in introducing discourse relations.
In previous work, we showed that puis is just such a connective. Its role in
SDRT is to impose a relation of Narration (whose semantics includes a temporal relation of succession) and to block causal relations like Result (cf. Bras et al.
2001, Borillo et al. 2004). We argue here that alors can also be such a temporal
connective under specific conditions.
Much work has been done on alors (cf. inter alia Jayez 1981, 1988b,a, Franckel
1987, Gerecht 1987, Hybertie 1996, Reyle 1998). Our starting point is the work
of Hybertie (1996), partly based on the analyses of Jayez, Franckel and Gerecht.
Hybertie distinguishes three major uses of alors in assertions1 :
 :
with only a temporal link between the eventualities:
(1) J’ai rencontré Pierre en 1987. J’étais alors une jeune étudiante.2
‘I met Pierre in 1987. I was alors a young student.’
with a temporal link and a dependency link (that we’ll explain below):
Je suis allée jusqu’à la place du village, alors je l’ai vu arriver.
‘I walked up to the village square. Alors I saw him arrive.’
 :
between eventualities:
J’étais pressé, alors j’ai pris le sens interdit.
‘I was in a hurry. Alors I took the one way street the wrong way.’
Olivier a fait tomber la carafe. Alors elle s’est cassée.3
‘Oliver dropped the carafe. Alors it broke.’
between two cognitive states in an inferential process:
OSLa volume 1(1), 2009
Les volets sont fermés, alors ils sont partis.’
‘The shutters are closed, alors they have left.’
(I see that the shutters are closed, so I infer that they must have
The cases in which alors is associated with an interrogative or an exclamative intonation are left aside.
All the examples below but (4) and (6) are taken from (Hybertie 1996). In the English glosses, we prefer
not to choose a translation of alors, so as not to blur the problem.
In (4), contrary to (2), alors could be replaced by donc.
  alors
Ce nombre est égal à 4. Alors il est pair. (Jayez 1988b, 158)
‘This number is equal to 4. Alors it is even.’
and   where alors is a kind of “structuration” marker:
with only a temporal link between the eventualities:
Oui bonsoir j’habite en moselle, alors actuellement il existe une
loi sur le travail à mi-temps pour les femmes […]
‘Yes, good evening. I live in Moselle county, alors at present
there is a law concerning part time work for women […]’
Le Draoulec and Bras (2007) analyze the uses of alors that Hybertie classifies as
“temporal uses”. Restricting their analysis to uses of alors that involve only events,
they showed that the temporal link conveyed by alors is associated with a rhetorical relation between constituents only when alors is in clause initial position, as in
example (2). This rhetorical relation expresses a dependency link that has been
described by Hybertie as follows: the event expressed by the first constituent is a
“condition for the realization” of the event described by the second constituent.
In these cases only, alors meets the requirements of a temporal connective in the
sense defined above.
When alors is in a clause internal or final position, its role is merely that of
a temporal anaphoric adverb conveying a temporal relation (with only possible
semantic effects of consequentiality when it is in internal position), as in example
Moreover, as far as the temporal relation between the events is concerned, the
temporal value itself depends on the sentential position: clause initial alors gives
rise to a relation of temporal succession between the events, as in (2); clause internal or final alors denotes a temporal relation of concomitance or coincidence
as in (1).
In this paper we consider the uses of alors when it is a temporal connective, leaving aside its temporal adverbial uses (examples such as (1) are thus excluded). We
provide a formal analysis of the “temporal uses”, and we extend the analysis to
the “consequential uses” in the classification of Hybertie above.
Some of these consequential cases – those linking eventualities – have a temporal value that leads us to group them together with the temporal cases. In this
enlarged set of temporal uses of alors, we have developed a new classification on
the basis of the discursive link at issue: the link either provides a dependency link
like the one described by Hybertie, or a cause-consequence relation. Alors as a temporal connective will be analysed in section 3, where we investigate which discourse
relations are able to account for the different types of dependency/causality associated with alors.
OSLa volume 1(1), 2009
,    
The restriction to clauses describing events is still in force in this analysis –
states are left for future research. Thus, we will exclude examples such as (3). For
the consequential cases listed above involving cognitive states such as (5) and (6),
we will only sketch lines of analyses for cases where events are involved. These
cases will form a new class and will be analysed in section 4 where we deal with
alors as a “logical connective”.
Finally, the “structuration marker” cases such as (7) won’t be taken into account.
  : 
[2.1] Overwiew of SDRT
SDRT is a formal theory of the semantics-pragmatics interface. It is an extension of Kamp’s Discourse Representation Theory (Kamp and Reyle 1993) that takes
discourse relations into account. We will give a brief outline of SDRT here. The
reader is invited to refer to (Asher and Lascarides 2003) for a thorough presentation. In SDRT, a discourse is represented by an SDRS (Segmented Discourse Representation Structure). It is a recursive structure consisting of labelled elementary
DRSs (i.e., Discourse Representation Structures, as described by DRT) representing a single clause and labelled sub-SDRSs linked together by Discourse Relations,
such as Narration, Elaboration, Background, Continuation, Result, Contrast, Explanation…
These elementary DRSs and the sub-SDRSs corresponding to complex discourse
segments are the constituents of the SDRS representing the discourse. The elementary constituents describe eventualities, i.e. events or states.
Labels are discourse referents. They are used to distinguish different occurrences of constituents, since each occurrence of a constituent in a discourse structure will be affected differently, at least in principle, by the discourse context. The
labels for the constituents are essential because a given proposition or semantic
content may have many different uses in different discourse contexts4 . To keep
the whole theory within a first order setting, discourse relations take labels as
arguments. In our formulation of the axioms for inferring and interpreting the
various discourse relations linking constituents, we will use Greek letters (α, β,
…) as variables to represent labels, Kα to represent the constituent labelled with
α, and eα to represent the main eventuality (event or state) described by Kα .
To construct an SDRS for a discourse, we need to add considerably to the construction algorithm posited for DRSs. SDRT defines a “Glue Logic” and an “Update
Function” that together determine a new SDRS for a given SDRS K 0 representing
the context (the discourse already processed) and a new constituent K α representing the information to be integrated into that context. The Glue Logic ex[4]
OSLa volume 1(1), 2009
A label can be seen as a kind of speech act discourse referent.
  alors
ploits the framework of “Commonsense Entailment” (Asher and Morreau 1991), a
logic that exploits both monotonic (→) and non-monotonic (>) conditionals. In
Commonsense Entailment (CE), ϕ > ψ means “if ϕ then normally ψ”. From ϕ >
ψ and ϕ, CE entails ψ “by default”, that is, defeasibly, in the absence of further
information regarding the truth value of ψ 5 .
Asher and Lascarides (2003) analyze the Glue Logic as a description logic that
makes extensive use of underspecification. We will adopt the convention used in
the Glue Logic of adding a third argument to two place predicates like discourse
relation predicates: a formula of the form R(α, β, λ) is to be understood as a constraint on the third argument λ, which is the label of the smallest SDRS containing
the formula linking the label α to the label β via the discourse relation R.
As far as the inferential tasks are concerned, the Glue Logic contains:
(i) definitions characterizing which constituents in the contextually given SDRS
are open for attaching the current constituent β,
(ii) axioms detailing what discourse relations may be inferred, on the basis of
a variety of linguistic and common knowledge clues, in order to actualize
the attachment of β to some open constituent α of the contextually given
SDRT also contains axioms specifying the semantic effects of the discourse relations, which can be considered as meaning postulates. We will give examples of
such axioms in the next section.
The Update Function is in charge of integrating the discourse relations the
Glue Logic infers and its choice of attachment sites within the contextually given
SDRS representing the structure of the discourse processed so far. This integration may also involve the resolution of underspecifications (e.g., anaphora and
SDRT distinguishes coordinating relations from subordinating ones. Explanation and Elaboration are examples of subordinating relations, while Narration is
a coordinating relation. Subordinating and coordinating relations affect the Update Function differently. Only subordinating relations may introduce complex
SDRSs, in other words, the Update Function may gather several SDRSs into a new
complex SDRS only if these constituents are attached to the same site with the
same subordinating relation. Asher and Vieu (2005) provide criteria within SDRT
for coordinating and subordinating relations. They also postulate that some relations like Result, though in most cases a coordinating discourse relation, may in
certain contexts be subordinating instead.
From ϕ > ψ, ϕ and ¬ψ, CE no longer entails ψ, but ¬ψ. From ϕ > ψ, ζ > ¬ψ, ϕ → ζ, ϕ and ζ, ψ (and not
¬ψ) is inferred (Penguin principle). From ϕ > ψ, ζ > ¬ψ, ϕ and ζ, if ϕ and ζ are logically independent,
CE cannot conclude ψ nor ¬ψ (Nixon Diamond).
OSLa volume 1(1), 2009
,    
In SDRT several discourse relations may simultaneously link the same two
constituents, thus distinguishing SDRT from other discourse approaches, notably
RST (Mann and Thompson 1988). SDRT also allows for multiple “superordinate’’
parents of a constituent, which means that SDRSs must be graphs and cannot be
represented completely faithfully as trees.6
In the following, we only present the discourse relations needed to describe
the discourse contribution of temporal connectives such as puis and alors: the SDRT
relations Narration and Result.
[2.2] The Discourse Relation of Narration
Narration is a relation which is based on the Gricean pragmatic maxim of manner
“be orderly”. When two clauses are linked by Narration, they describe in sequence
two successive events “of the same story”.
Narration can be non-monotonically inferred if the two clauses to be related
contain clues indicating that their main eventualities are of types that may belong
to ‘the same story’. This “condition” on the types of the eventualities described
by the clauses is formalized by the predicate Occasion, which serves as a triggering
condition to the non-monotonical inference of Narration in the following axiom,
where λ represents the smallest constituent that will end up containing the formula that links α and β once the SDRS for the discourse is constructed and fully
A N
(? (α, β, λ) ∧ Occasion(α, β)) > Narration(α, β, λ)
Occasion is a predicate of SDRT’s Glue Logic whose semantics involves those linguistic elements available in the logical forms of the discourse constituents that
are relevant to inferring Narration. It exploits lexical semantics and shared knowledge in terms of scripts connecting certain event types in sequences in which one
event ‘naturally’ leads to the next, though it is not the cause thereof. For instance,
(8) is an example of Narration where Occasion holds, since there is clearly in the
shared knowledge a script in which, before entering, people knock at the door7 .
This kind of knowledge is represented by A O 1.
Paul frappa à la porte. Il entra.
In effect, however, this added expressive power of the theory does not come at great computational cost:
Baldridge et al. (2007) show that relational parsers for SDRS graphs have no greater complexity than
ordinary tree parsers.
Cases where Occasion does not hold may be illustrated by an example such as
OSLa volume 1(1), 2009
Paul frappa à la porte. Il se mit à chanter.
‘Paul knocked at the door. He began to sing.’
  alors
‘Paul knocked at the door. He entered.’
A O 1
(? (α, β, λ) ∧ [frapper-à-la-porte(eα , x) ∧ [entrer(eβ , x)](β) > Occasion(α,
We now turn to another way of inferring Narration, with a monotonic inference
this time. This is when a specific discourse marker is present. In (Bras et al. 2001),
it was shown that puis is such a marker – which endows it with the status of temporal connective defined in section 1. Its role in SDRT is described by the following
A N 8
(? (α, β, λ) ∧ [puis](β)) → Narration (α, β, λ)
Last, Narration can be inferred between α and β from information relevant to
subsequent constituents, one of which is linked to β. An example motivating this
sort of rule occurs with a simple discourse like (9):
Nicholas a travaillé sur son papier, il a dîné avec Myriam et Anne, puis il
est reparti pour Teilhet.
‘Nicholas worked on the paper, had a meal with Myriam and Anne and then
went home to Teilhet.’
The last two clauses are forcibly linked by Narration because of the discourse connective puis. But this discursive link leads us to interpret the link between the
first and the second clause as one of Narration. It is this sort of situation that the
following axiom on Subsequent Relations, is designed to address:
A S R
(?(α, β, λ) ∧ R(β, γ, λ)’) > R(α, β, λ)
where R is any discourse relation used in SDRT.
Let us now examine the two major semantics effects of Narration on discourse
content. The first one aims at capturing the fact that narratives must cohere in
the sense that the events linked by Narration must fit consistently and without
significant spatio-temporal gaps as expressed in (Asher et al. 1995) and (Asher
1996) and observed in earlier work on temporal order in narratives (see for example Caenepeel 1989). This doesn’t mean that there should be no interval of time
between the two events eα and eβ , but rather that no relevant event, i.e. no event
interfering with eα or eβ , can occur during this interval. This constraint is for[8]
We remark that our axiom does not of course entail that every instance of Narration can be marked by
puis. In our approach, there is not equivalence between the discourse relation and the presence of the
OSLa volume 1(1), 2009
,    
malized in (Bras et al. 2001) by the following axiom rewritten in the notation of
(Asher and Lascarides 2003):9
A N T 
Narration (α, β) ⇒ eα ⊃⊂ (post(eα ) ∩ pre(eβ )) ⊃⊂ eβ
This axiom expresses a meaning constraint on Narration. It uses not the language
of Glue Logic itself but the base logic in which the semantics of SDRSs is given. This
language is much richer than the Glue Logic and can appeal to notions that are
used traditionally in semantics, including the relation of abutment (⊃⊂) used in
DRT, which is equivalent to the “meets” relation as used in Allen’s theory (Allen
1984), a function ∩ taking pairs of events into the maximum interval during which
both events obtain; and the functions post-state and pre-state – the post-state of
eα , post(eα ), is a state that begins at the end of eα and persists by default indefinitely into the future, while the pre-state of eα is the state that terminates at the
beginning of eα and extends indefinitely far back into the past. The right hand
side of the axiom expresses that eα abuts post(eα ) ∩ pre(eβ ) which in turn abuts
eβ . It entails in addition that there is no intervening event whose propositional
content interferes with either that of post(eα ) or that of pre(eβ ), i.e., no event that
ends post(eα ) before eβ starts or prevents pre(eβ ) from holding right after eα has
ended10 . For illustrations, see (Bras et al. 2001), (Borillo et al. 2004).
Narration has a second semantic effect. It is motivated by the intuition that
the elements of a Narration must belong to the “same story”, i.e., they must have
some common subject matter. To this effect, axiom A N T expresses that the constituents held together by Narration must have a common
Topic. A topic is a simple constituent which is contingent (i.e., not vacuous, not
contradictory, not tautological), and subsumes the constituents of a sub-SDRS, in
this case, the constituents linked by Narration. If not already present in the context, it has to be added to the SDRS during the update. A N T
and the rules of Glue Logic actually imply that Narration can be non-monotonically
inferred only if such a topic exists or can be built.
A N T
Narration(α, β ⇒ ¬¤(Kα ⊓ Kβ )
where ⊓ is a merge operation defined in Asher (1993) for defining topics. The
axiom states that the topic of α and β cannot be vacuous, which we formalize in
OSLa volume 1(1), 2009
We will forego the use ϕR(α,β) of (Asher and Lascarides 2003) and use R(α, β) instead for ease of reading.
From A N T  and uncontroversial ordering assumptions on events and
their pre- and post-states (Event(e) → pre(e) ⊃⊂ e ⊃⊂ post(e)), we can deduce a relation of temporal precedence between the events eα and eβ : eα ≺ eβ , which was the original temporal effect of Narration in
(Lascarides and Asher 1993).
  alors
SDRT as a necessary truth.
To illustrate these structural effects, let us consider the simple example (8)
again. This small text clearly tells the story of Paul arriving at someone’s home
or office. Such a “topic” is inferred only because the two sentences are textually
linked together and in this order.
[2.3] The Discourse Relation of Result
Result can be non-monotonically inferred on the basis of lexical semantics or of
some shared knowledge on the types of eventualities in α and β as in (10):
Tarzan poussa Jane. Elle tomba.
‘Tarzan pushed Jane. She fell.’
For this example, some information that generalizes on the event types of pushing and falling suggests to the reader a causal link that the narrator most likely
wants to express. The presence of such clues indicating a possible causal link is
expressed by the predicate CauseD where CauseD (α, β) means that α describes an
eventuality which is a possible cause for β’s main eventuality11 . For instance, the
following axiom encodes a plausible piece of shared knowledge:
A CD 1
(? (α, β, λ) ∧ [change_of_force(eα , x, y)](α) ∧ [change_of_position(eβ , y)]
(β)) → CauseD (α, β)
SDRT relies on a lexicon to infer that pousser(eα , x, y) is an instance of change_of_force(eα , x, y) and that tomber(eβ , y) is an instance of change_of_position(eβ , y) in order to deduce that CauseD (α, β) holds in this particular case. SDRT then allows us
to deduce the discourse relation Result when one can infer the predicate CauseD
(we simplify the axiom of Asher and Lascarides (2003) as we will not consider different aspectual classes):
A R
(? (α, β, λ ) ∧ CauseD (α, β)) > Result (α, β, λ)
A Result relation between constituents α and β represents the narrator’s intention to signify that β is a result of α (or α results in β). It has the semantic effect
of implying a causal link between the main eventualities of the constituents it
relates, which we write using the formalism of Asher and Lascarides (2003) as follows:
The causal link expressed by CauseD is neither entailing nor being entailed by Occasion. We remind the
reader that Occasion and CauseD are not Discourse Relations but predicates specifying information from
a variety of knowledge sources leading to the inference of the Discourse Relation at stake: Narration or
OSLa volume 1(1), 2009
,    
A R SE
Result(α, β) ⇒ Cause(eα , eβ )
The predicate Cause(eα , eβ ), to be read eα causes eβ , implies, among other things,
that if eα and eβ are events, the first temporally precedes the second eα ≺ eβ .
Asher and Lascarides (2003) don’t give any explicit marker of Result, as we
did above for Narration with puis. We mention that puis, together with the explicit marking of Narration, is a kind of “negative clue” for Result, as it blocks the
inference of this relation (Bras et al. 2001)12 :
A R 
(? (α, β, λ) ∧ [puis](β)) > ¬ Result (α, β, λ)
It was also suggested in (Bras et al. 2001) that Result may be monotonically inferred on the basis of the presence in β of an explicit marker of causation, such
as the conjunct donc (therefore):
Max a eu son bac. Donc il a pu entrer à l’université.
‘Max passed his A-levels. Therefore he could go to the university.’
But donc has other meanings, which do not express a “material” consequence, but
rather a logical or argumentative consequence, which contradicts the temporal
counterpart of Cause, as in:
Il a réussi à faire l’exercice, donc il a bien compris le cours.
‘He succeeded in solving the problem, therefore he understood the class
As a matter of fact, in order to define an axiom such as A N , we
need a detailed study of the lexical marker. Such an analysis for alors is spelled
out in the next section.
alors                     
[3.1] Formalization in extant SDRT
In this section we specify which extant discourse relations of SDRT are able to
express the consequential value necessarily involved by alors when it plays the
role of a temporal connective.
For instance, the reading we get in an example such as
L’acide tomba dans le liquide. Puis une explosion se produisit.
‘The acid fell into the liquid. Puis an explosion happened.’
is one in which the author does not commit himself regarding a possible causal relationship between the
events and tries to objectively tell the story of what happened during the chemical experiment.
OSLa volume 1(1), 2009
  alors
Let us first examine example (4) repeated below:
Olivier a fait tomber la carafe. Alors elle s’est cassée.
According to the A R given above, SDRT predicts a discourse relation of
Result between the two constituents described by example (4): the information
needed to infer CauseD is readily available – tomber(x) is a permissible cause of
se casser(x). Although SDRT does not specify sentence initial alors as a discourse
marker of Result to draw this inference, sentence initial alors in this case would
not block the inference but rather lead to the same conclusion. However, in real
texts, it is difficult to find examples similar to (4), and in any case, (4) itself is not
an example of perfectly idiomatic French. Alors doesn’t appear to fit in contexts
where the CauseD predicate expresses an “objective” causal link, such as the one
conveyed by (4). Actually, the kind of attested examples with alors where Result
can be inferred thanks to CauseD are cases where the causal link is more “subjective”, as in (13) and (14) below13 :
Mme Couze a dit d’abord qu’on ne te voyait plus jamais et que ces messieurs
du restaurant devaient te regretter. Alors Mme Londe a répondu: Une de
perdue, dix de retrouvées. (Julien Green, Léviathan).
‘Mme Couze first said that we never saw you any more and that the gentlemen at the restaurant must miss you. Alors Mme Londe responded: one
lost, but ten found.’
Il fit ses comptes et s’aperçut qu’il avait tant donné de traites au porteur,
qu’il ne lui restait plus que cinquante mille francs. Alors il se fit en lui
une réaction étrange: lui qui venait d’abandonner cinq millions, il essaya
de sauver les cinquante mille francs qui lui restaient. (Alexandre Dumas
père, Le Comte de Monte-Christo)
‘He did the books and then noticed that he had given so much to the porter
that he only had 50 000 francs left. Alors he had a strange reaction: he who
had just lost five million now tried to save the 50 000 francs that he still
Examples (13) and (14) are clear cases of the Response relation, along the lines of
Sandström (1993, 63): “Response is the relation between an event e1 and an action
e2 which it evokes in a sentient agent”. In SDRT these cases are subsumed under
Result because of their causal nature (Bras et al. 2001, 137). So here again, in the
extant SDRT framework, Result can be inferred thanks to CauseD without the help
of alors.
In order to understand the role of alors, it is necessary to consider examples
where alors is responsible for the causal link at stake, such as in (2), repeated
All our attested examples are taken from the textual database Frantext:
OSLa volume 1(1), 2009
,    
below, or the constructed example (15). Without the connective, (2’) and (15’)
lack any causal link of that sort and are rather marginal examples of Narration,
given a null context.
Je suis allée jusqu’à la place du village, alors je l’ai vu arriver.
Il m’a rejointe. Alors je me suis souvenue que j’avais oublié mes clés.
‘He joined me. Alors I remembered that I had forgotten my keys.’
Je suis allée jusqu’à la place du village, je l’ai vu arriver.
Il m’a rejointe. Je me suis souvenue que j’avais oublié mes clés.
In the examples (16)-(18) extracted from real texts, we find evidence of this causal
Elle ne vit pas le télégramme. Un peu plus tard elle ressortit avec un jeu de
boules de pétanque en matière plastique contenu dans une espèce d’étui
à claire-voie. Alors elle vit le télégramme, le ramassa et le lut et courut
vers la plage. (Jean-Patrick Manchette, Trois hommes à abattre)
‘She didn’t see the telegram. A little later she came out with a pair of
plastic bacci balls contained in a sort of see through carrying case. Alors
she saw the telegram, picked it up, read it, and ran towards the beach.’
Puis les allées commencèrent de monter assez raide et, bientôt, Joseph
se trouva sur le plateau. Alors Joseph leva la tête, vit le ciel et poussa un
long soupir. Jamais il ne songeait à regarder le ciel. (Georges Duhamel,
Chronique des Pasquier)
‘At that point the streets started to climb pretty rapidly and soon Joseph
found himself on top of the plateau. Alors Joseph raised his head, looked
at the sky and let out a long sigh. Never did he think to look at the sky.’
Je l’ai emporté dans ma chambre. Avant de l’ouvrir, je me suis assise et
je l’ai posé sur mes genoux. Alors je me suis souvenue que je l’avais eu il
y avait très longtemps, et que, lorsque je l’avais regardé pour la première
fois, je n’étais pas dans la maison de M. Drapeur. (André Dhôtel, Le Pays
où l’on n’arrive jamais)
‘I brought it into my room. Before opening it, I sat down and placed it on
my knees Alors I remembered that I had possessed it long long ago and
that when I had gazed on it for the first time, I was not in the house of M.
From a strict SDRT point of view, the requisite information needed to infer CauseD
for (2) and (15)-(18) is lacking. For instance in (2), to infer Result we would have
to assume that an event of the type “going to the village square” is normally a
plausible possible cause of an event of the type “seeing someone arrive”, which is
OSLa volume 1(1), 2009
  alors
not the case. So, with the extant SDRT axioms, we cannot infer Result.
It also seems improbable that Occasion, the relevant predicate on eventuality
types needed to infer Narration, holds between the two constituents in (2) and
(15)-(18), because the types of events exemplified in this text do not specify a
stereotypical sequence. So the appropriate axiom A N can not be
used to infer Narration. As SDRT does not yet account for the role of alors, it would
predict Narration thanks to the A S R. However, this inference does not match our description of the role of alors: in these examples, as
shown in section 1, alors triggers a discourse relation requiring that ‘the event
expressed by the first constituent is a “condition of realization” for the event described by the second constituent’. This relation differs from the extant, similar
SDRT relations of Result and Narration.
[3.2] Formalization within an enriched set of discourse relations
Asher and Lascarides (2003) don’t give a complete definition of Result, but they
take Result(α, β) to imply Cause(eα , eβ ), which means that the main eventuality
in α “is the cause of” the main eventuality in β. Cause is not to be confused with
CauseD , it is a predicate that holds between eventualities and not eventuality types
as CauseD does. Although Cause is not formally defined in (Asher 1993), nor in
(Lascarides and Asher 1993) and (Asher and Lascarides 2003), we take it to correspond to a strong causal link between eα and eβ , which we will presently define.
With this informal definition of Cause as a “is the cause of” relation, we can account for (4) and (13)-(14) correctly. For (2) and (15)-(18), a strict cause-consequence
link is too strong: so we do not want to introduce an axiom stipulating that alors
would imply Result (after the model of A N ). Alors does not
trigger an inference to the discourse relation Narration either. Rather, it expresses another kind of causal link, that we described above, following Hybertie,
as a “condition of realization”, which we feel to be somewhere between Narration
and Result.
It thus seems necessary to introduce a new relation, that should have weaker
causal implications (and for which alors would be a trigger). To capture this weaker
causal link, we turn to Lewis (1973) who defines several causal relations. In particular, his “causal dependency” relation is close to what we need to express, except
that we will restrict our causal links to actually occurring events. Following Lewis,
we define the relation Weakly_causally_depends_on:
eβ Weakly_causally_depends_on eα if and only if
(i) if eα had not occurred, eβ wouldn’t have occurred either, in all the
worlds closest to α’s world, and
(ii) (it is true that) eα occurred and (it is true that) eβ occurred.
We would like to add the requirement that eα precedes eβ . Formally, we offer the
OSLa volume 1(1), 2009
,    
following definition in SDRT, where we appeal to Lewis’ similarity relation that
he uses to define counterfactuals and where the causal relation is expressed as
holding between event types or facts:
An event eβ associated with a description Kβ (in a discourse constituent α) Weakly_causally_depends_on an event eα associated with a
description Kα (in a discourse constituent β) if and only if (¬Kα ¤ →
¬Kβ ) ∧ (Kα ∧ Kβ ) ∧ eα ≺ eβ ), where A¤ → B is true in a world w if
and only if in every world closest to w where A is true, B is true too.
We emphasize that “in every world closest to w” means that the implication has
to be considered “other things being equal”, and hence defeasible. For instance,
for (2), we cannot consider an event of “her climbing up on a wall”. Of course
that event could have been a cause for the event “her seeing him arrive”, but this
event occurs in a world that doesn’t belong to the closest worlds to w.
Having defined the relation Weakly_causally_depends_on, we can formally define the stronger causal relation Causally_depends_on:
eβ Causally_depends_on eβ if and only if eβ Weakly_causally_depends_on
eα , and Kα > Kβ
Causally_depends_on is stronger than Weakly_causally_depends_on in a strict sense,
thanks to the formula Kα > Kβ 14 .
Weak causal dependency and causal dependency can hold between events,
but they can also hold between event descriptions or facts, elements described by
We have expressed the semantics for our two causal relations in the full SDRT
language, not the description language in which the Glue Logic axioms for inferring discourse relations is stated. Thus, since we are speaking about the causal
relations themselves and not about a description of the occurrence of this relation symbol in some SDRS, as we do in the Glue Logic, we have expressed our definitions using intuitive two place relations, rather than the description language
three place formulas.
Our two causal relations permit us to make some distinctions at the discursive
level. The first is that we introduce a new discourse relation, which we call WeakResult. As is usual, we constrain the semantics of this relation via axioms about
its semantics effects:
OSLa volume 1(1), 2009
Our notion of causal dependency differs from that of (Lewis 1973) only in that we replace the counterfactual (Kα ¤ → Kβ ) in his definition with a normality conditional and we insist that the relation
hold between actually occurring events. This replacement is essential if we wish, as we do, to restrict
our relations to actually occurring events, since in Lewis’ logic of counterfactuals (Kα ∧ Kβ ) implies
(Kα ¤ → Kβ ). Thus using Lewis’ definition in our case would reduce our relation of causal dependency
to the relation of weak causal dependency.
  alors
A WR S
Weak-Result (α, β) ⇒ Weakly_causally_depends_on (Kα , Kβ )
This formulation of Weak-Result encodes that it entails the Weakly_causally_depends_on relation between the event descriptions in the constituents labelled with
α and β. We use this relation to account for non purely logical cases where causes
precede consequences (the logical cases will be analysed further).
Alors, as we have seen, is responsible for introducing a relation whose content
is just that for Weak-Result. This means that we should introduce an axiom in
the Glue Logic and its description language, expressing that alors is a trigger for
A W R 
(? (α, β, λ) ∧ [alors](β)) → Weak-Result (α, β, λ)
This holds only for clause initial alors, according to the description in section 1.
When alors is not initial, it does not play a role at the discourse level but at the
sentence level. Therefore its contribution will be taken into account with compositional semantic rules (i.e. within the constituent representation) and not with
discourse rules such as A WR.
Let us see how we can now account for our examples with the definition and
axiom on Weak-Result given above. The relation of Weak-Result as defined is
appropriate for linking the clauses in (15)-(18). For example, in (15) we have: “if
he hadn’t rejoined me, I wouldn’t have remembered”; further, it is both true that
“he rejoined me” and that “I remembered”; and finally, the event of his rejoining
me precedes the event of my remembering.
On the other hand, Weak-Result is insufficient to describe the discourse link in
(4) and (13)-(14): the causal link at stake is stronger than the weak causal dependency: it corresponds to the relation of causal dependency defined above. This
leads us to the conclusion that Result is a scalar relation: along with Weak-Result,
there is also a relation that we call Strong-Result, which can be defeasibly inferred
from CauseD , and which we define as reflecting the relation of causal dependency
and the relation Causally_depends_on.
A SR S
Strong-Result (α, β) ⇒ Causally_depends_on (Kα , Kβ )
A SR 
(? (α, β, λ ) ∧ CauseD (α, β)) > Strong-Result (α, β, λ)
For examples (4) and (13)-(14) both A WR and A SR
will apply: both Weak-Result and Strong-Result will be inferred. For the same examples without alors, Strong-Result would be inferred too. However, unlike the
OSLa volume 1(1), 2009
,    
naturally occurring examples (13)-(14), (4), as we noted, is not very good, for reasons which remain to be explored. Our intuition is that the objective versus the
subjective nature of the causality should be taken into account (some hints in this
sense are also given by Hybertie 1996).
We note that our constraints are also compatible with what we have said about
puis. Puis signals that we should normally not infer Result. A R 
given above should be replaced by the two following axioms:
A WR 
(? (α, β, λ) ∧ [puis](β)) > ¬Weak-Result (α, β, λ)
A SR 
(? (α, β, λ) ∧ [puis](β)) > ¬Strong-Result (α, β, λ)
This will be consistent with the fact that events in the world be linked via causal
dependency or weak causal dependency, as long as these relations cannot be inferred from the information present in the discourse and the context.
We finally examine cases when alors combines with Occasion, as in (19):
Pierre est tombé. Alors je l’ai aidé à se relever.
‘Peter fell. Alors I helped him up.’
In such a case, the extant axioms lead to infer Narration and Weak-Result. The
conjunction of these two relations strengthens the discourse connection between
the two constituents. (19) is a constructed example, but there are many examples
in real texts with a similar interpretation:
L’empereur se tourna à demi, avec un léger hochement de tête, pour autoriser le ministre de l’intérieur à continuer. Alors Rougon entra dans des
détails préliminaires. (Émile Zola, Son Excellence Eugène Rougon)
‘The emperor turned around halfway with a slight nod of the head, signalling to the interior minister to continue. Alors Rougon began to lay out
the preliminary details.’
La grille l’attira. Il serra les barreaux rouillés très fort, comme un prisonnier. Il y appuya son front pour sentir le froid du métal. La porte s’ouvrit.
Alors, il entra. (Robert Sabatier, Les Noisettes sauvages)
‘The ironwork attracted him. He clasped the rusted bars very strongly,
like a prisoner. He put his forehead to the metal to feel its cold. The door
opened. Alors he entered.’
The combination of Occasion and alors would always seem to yield a coherent discourse in SDRT’s terms. That is, they suffice to infer a discourse relation between
two constituents with this information. However, this is not the case. Let us for
OSLa volume 1(1), 2009
  alors
instance compare (19) with (22):
?Pierre est tombé. Alors il s’est relevé.
‘?Peter fell. Alors he stood up.’
Our intuition is that alors better combines with Occasion when there is a change
of subject as in (19), which is similar, on that point, to (20) and (21). The contrast
between (23) and (24) reinforces this hypothesis. Finally the contrast between (24)
and (25) illustrates the fact that puis is particularly felicitous in the configuration
of Occasion with a same subject (Bras et al. 2001, 128-129).
Ils se mirent à table. Alors Pierre souhaita bon appétit à tout le monde.
‘They sat down to table. Alors Pierre wished ‘bon appetit’ to everyone.’
?Ils se mirent à table. Alors ils se souhaitèrent bon appétit.
‘They sat down to table. Alors they wished ‘bon appetit’ to everyone.’
Ils se mirent à table. Puis ils se souhaitèrent bon appétit.
‘They sat down to table. Puis they wished ‘bon appetit’ to everyone.’
We won’t try to axiomatize here the prediction of the compatibility between alors
and Occasion, as this would require a more fine-grained description of Occasion,
taking the change of subject into account15 . We leave this matter at this stage.
alors                    
In the cases we have examined so far, alors is a temporal connective, in the sense
defined in section 1. We also want to account for cases in which alors plays the
role of a logical consequence connective. This logical role of alors is more frequent in
discourses describing states, as in (5) and (6) repeated below:
Les volets sont fermés, alors ils sont partis.
Ce nombre est égal à 4. Alors il est pair.
As said above, in this paper, we restrict ourselves to events. But in fact we can
also have events in discourses where alors expresses a logical consequence:
Taking the change of subject into account would help make more precise the frontier between Occasion
and CauseD , as Occasion with a change of subject is often very close to CauseD . If we take the example of
Occasion mentioned in section 3, (8), we notice that adding alors does not yield a felicitous discourse:
?Pierre frappa à la porte. Alors il entra.
Now if we modify (8’) by introducing a change of subject:
Pierre frappa à la porte. Alors Marie lui ouvrit.
we have difficulties in classifying the event type predicate at stake between Occasion and CauseD .
OSLa volume 1(1), 2009
,    
Toutes les filles sont arrivées à l’heure, alors Marie est arrivée à l’heure.
‘All the girls arrived on time, alors Marie arrived on time.’
In order to account for these cases, we introduce a discourse relation, Inferential
result, expressing the logical link at stake:
A IR SE
Inferential-Result (α, β) ⇒ (Kα ∧ Kβ ∧ (Kα > Kβ ))
Actually, in cases like (6) and (26) we have ¤(Kα → Kβ ). Kα > Kβ is meant
for cases like (5). Insofar as ¤(Kα → Kβ ) implies Kα > Kβ both cases fit our
definition. Further research will have to include the analysis of inferential result
involving states and will probably lead to a refined formalization.
We see now that alors can trigger both Weak-Result and Inferential-Result.
Since Inferential-Result is triggered relative to information that entails the information used to infer Weak-Result, we rewrite our A WR, changing
the monotonic axiom for a non-monotonic one:
A WR 2
(? (α, β, λ) ∧ [alors](β)) > Weak-Result (α, β, λ)
This allows us to introduce a more specific axiom to trigger Inferential-Result.
A IR T
(? (α, β, λ) ∧ [alors](β) ∧ ¬ Weakly_causally_depends_on (Kα , Kβ )) >
Inferential-Result (α, β, λ)
Let us illustrate the application of these axioms for (26). With A W
R 2 we infer Weak-Result(π 1 , π 2 ). But Weak-Result(π 1 , π 2 ) does not hold because the temporal constraint e1 ≺ e2 is wrong. The temporal relation is e2 ⊆ e1, as
obtained by the following reasoning: the proper noun Marie in K2 triggers the presupposition of existence of a girl named Marie. We can bind this presupposition if
we can infer that this referent is part of the plural referent in the universe of K1
thanks to the semantics of the quantifier toutes les filles. As we have the same event
types for e1 and e2 , and the subject referent of e2 being a part of the subject referent of e1 , we infer that the event of Marie arriving on time (e2 ) is part of the event
of all the girls arriving on time (e1 ). As the binding of a presupposition is preferred
to accommodation in standard theories of presupposition (Van der Sandt 1992) as
well as in SDRT (Asher and Lascarides 2003), we are forced to this treatment of the
presupposition and to the inference concerning the temporal relation between e1
and e2 . This means that Weak-Result cannot apply, and by A I
OSLa volume 1(1), 2009
  alors
R T16 , we conclude Inferential-Result.
Our investigation concerning the discourse information conveyed by sentence
initial alors has revealed at least three sorts of relations close to the informal
gloss given by Asher (1993) and Asher and Lascarides (2003) for Result. We have
seen that alors alone suggests a weak causal relation, which we have formalized
using Lewis’ counterfactual. We encoded this information in the discourse relation Weak-Result, which we distinguished from Strong-Result. Strong-Result is
inferred when lexical or other contextual information triggers a causal, discursive link. Our definitions immediately imply that both Strong-Result and WeakResult are veridical relations in the sense of Asher and Lascarides (2003). We also
saw that alors can, when Weak-Result is blocked, suggest an inferential relation,
which we expressed using the weak conditional > already present in the SDRT
Our analysis of the discursive uses of alors in initial position with clauses that
involve events paints a uniform but complex picture of this discourse connective.
In future work we intend to extend this study to treat uses of alors that involve
reference to states. Our proposal to analyse Result as a scalar relation should now
be put to the test and refined through a systematic comparison of alors with other
discourse connectives also related to causality such as donc, du coup or de ce fait (cf.
Rossari and Jayez 1996, 2000).
Inferential-Result differs from the SDRT relation called Defeasible-Consequence, used to express conditionals, which is not veridical. It would be interesting, as one of our reviewers suggests, to compare our
treatment of Inferential-Result as triggered by alors with the relation triggered by donc: donc does not
embed within conditionals, while alors does. While we have not yet studied the role of alors within conditionals, and we have not studied donc either, we can offer some suggestions as to how these connectives
might differ. The semantics for alors that we have given, as well as the general architecture of SDRT,
does not preclude the embedding of alors within a conditional; when this occurs, this means that two
relations hold between the antecedent and the consequent of the conditional, the one specified by alors
and the one specified by the conditional. There is, however, a clash between these two relations: the
conditional does not imply a veridical relation between antecedent and consequent, whereas alors does.
But notice that alors is naturally understood as occurring within the scope of the conditional; this means
that we should interpret alors as contributing material to the consequent of the conditional, specifying,
as it were, the relation between antecedent and consequent. Interpreted in this way, there is no clash;
the consequent is to be interpreted on the assumption that the antecedent is true, and so the veridicality of the discourse relation contributed by alors is no longer problematic. None of this tells us why
donc cannot be used within the scope of a conditional. One hypothesis worth exploring, however, is that
donc introduces the same relation as the conditional, plus the constraint that its terms be true. This is,
donc would have a similar analysis to alors, except that instead of allowing us to infer Weak-Result or
Inferential-Result, it leads to the inference of a consequence relation, exactly the relation introduced by
the conditional. Some preliminary evidence that this is so comes from the observation that the deduction theorem of classical logic seems to hold with donc : A, donc B seems to be equivalent to Si A, B. This
does not seem to be the case with alors. If this analysis is on the right track, then we would have at least
a pragmatic explanation for why si A donc B is not very good: donc at this point is redundant.
OSLa volume 1(1), 2009
,    
Allen, J. A. 1984. Towards a General Theory of Action and Time. Artificial Intelligence
23, 123–154.
Asher, N. 1993. Reference to Abstract Objects in Discourse. Dordrecht: Kluwer.
Asher, N. 1996. Mathematical Treatments of Discourse Contexts. In Proceedings of
the 10th Amsterdam Conference on Formal Semantics, volume 1, pages 21–40, ILLC
Publications, Amsterdam.
Asher, N., Aurnague, M., Bras, M., Sablayrolles, P. and Vieu, L. 1995. De l’espacetemps dans l’analyse du discours. Sémiotiques 9, 11–62.
Asher, N. and Lascarides, A. 2003. Logics of Conversation. Cambridge: Cambridge
University Press.
Asher, N. and Morreau, M. 1991. Commonsense Entailment: A Modal Theory of
Nonmonotonic Reasoning. In J. Mylopoulos and R. Reiter (eds.), Proceedings of
the Twelfth IJCAI, pages 387–392, Los Altos, CA.: Morgan Kaufman.
Asher, N. and Vieu, L. 2005. Subordinating and Coordinating Discourse Relations.
Lingua 115(4), 591–610.
Baldridge, J., Asher, N. and Hunter, J. 2007. Annotation for and Robust Parsing
of Discourse Structure on Unrestricted Texts. Zeitschrift für Sprachwissenschaft
26(2), 213–239.
Borillo, A., Bras, M., De Swart, H., Le Draoulec, A., Molendijk, A., Verkuyl, H., Vet,
C., Vetters, C. and Vieu, L. 2004. Tense and Aspect. In F. Corblin and H. de Swart
(eds.), Handbook of French Semantics, pages 233–348, Stanford: CSLI Publications.
Bras, M., Le Draoulec, A. and Vieu, L. 2001. French Adverbial Puis between Temporal Structure and Discourse Structure. In M. Bras and L. Vieu (eds.), Semantic
and Pragmatic Issues in Discourse and Dialogue: Experimenting with Current Theories,
CRiSPI series,, volume 9, pages 109–146, Amsterdam: Elsevier.
Caenepeel, M. 1989. Aspect, Temporal Ordering and Perspective in Narrative Fiction.
Ph. D.thesis, University of Edinburgh, Edinburgh.
Franckel, J. J. 1987. Alors, alors que. Bulag 13, 17–49.
Gerecht, M.-J. 1987. Alors: opérateur temporel, connecteur argumentatif et marqueur de discours. Cahiers de linguistique française 8, 69–79.
Hybertie, C. 1996. La conséquence en français. Paris-Gap: Ophrys.
OSLa volume 1(1), 2009
  alors
Jayez, J. 1981. Etude des rapports entre l’argumentation et certains adverbes français.
Masters Thesis, Aix-Marseille 2.
Jayez, J. 1988a. L’inférence en langue naturelle. Paris: Hermès.
Jayez, J. 1988b. Alors, descriptions et paramètres. Cahiers de Linguistique Française
9, 135–175.
Kamp, H. and Reyle, U. 1993. From Discourse to Logic. Dordrecht: Kluwer.
Lascarides, A. and Asher, N. 1993. Temporal Interpretation, Discourse Relations,
and Commonsense Entailment. Linguistics and Philosophy 16(5), 437–493.
Le Draoulec, A. and Bras, M. 2006. Quelques candidats au statut de connecteur
temporel. Cahiers de Grammaire 30, 219–237.
Le Draoulec, A. and Bras, M. 2007. Alors as a possible Temporal Connective in Discourse. Cahiers Chronos 17, 81–94.
Lewis, D. 1973. Causation. Journal of Philosophy 70, 556–567, reprinted in Lewis,
1986, Philosophical Papers II. Oxford: Oxford University Press. 159-213.
Mann, W. C. and Thompson, S. A. 1988. Rhetorical Structure Theory: A Theory of
Text Organization. Text 8:3 8(3), 243–281.
Reyle, U. 1998. A note on enumerations and the semantics of puis and alors. Cahiers
de Grammaire 23, 67–79.
Rossari, C. and Jayez, J. 1996. Donc et les consécutifs: des systèmes de contraintes
différentiels. Lingvisticae investigationes 20(1), 117–143.
Rossari, C. and Jayez, J. 2000. Du coup et les connecteurs de conséquence dans une
perspective dynamique. Lingvisticae investigationes 23(2), 303–326.
Sandström, G. 1993. When-clauses and the temporal interpretation of narrative discourse. Ph. D.thesis, University of Umeå.
Van der Sandt, R. A. 1992. Presupposition projection as anaphor resolution. Journal
of Semantics 9(4), 333–378.
OSLa volume 1(1), 2009
,    
  
Myriam Bras
University of Toulouse
Maison de la Recherche &
Université Toulouse Le Mirail
5 allées Antonio Machado
F-31058 Toulouse Cedex 09
[email protected]
Anne Le Draoulec
University of Toulouse
Maison de la Recherche &
Université Toulouse Le Mirail
5 allées Antonio Machado
F-31058 Toulouse Cedex 09
[email protected]
Nicolas Asher
University of Toulouse
IRIT UMR 5505 &
Université Paul Sabatier
118 Route de Narbonne
31062 Toulouse Cedex 09
[email protected]
OSLa volume 1(1), 2009
Behrens & Fabricius-Hansen (eds.) Structuring information in discourse:
the explicit/implicit dimension, Oslo Studies in Language 1(1), 2009. 171-190. (ISSN 1890-9639)
 :
  
University of Edinburgh
University of Pennsylvania
The goal of understanding how discourse is more than a sequence of sentences has engaged researchers for many years. Researchers in the 1970’s attempted to gain such understanding by identifying and classifying the phenomena involved in discourse. This was followed by attempts in the 1980s
and early 1990s to explain discourse phenomena in terms of theories of abstract structure. Recent efforts to develop large-scale annotated discourse
corpora, along with more lexically grounded theories of discourse are now
beginning to reveal interesting patterns and show where and how early theories might be revised to better account for discourse data.
In the sciences, theory and data often compete for the hearts and minds of researchers. In linguistics, this has been as true of research on discourse structure
as of research on syntax. Researchers’ changing engagement with data versus
theory in discourse structure – and the hope for more progress by engaging with
both of them – is the subject of this brief paper.
 70:    
The 1970’s saw a focus on data, with Cohesion in English (Halliday and Hasan 1976)
an important milestone. This volume catalogued linguistic features in English
that impart cohesion to a text, by which Halliday and Hasan meant the network
of lexical, referential, and conjunctive relations which link together its different
parts. These relations contribute to creating a text from disparate sentences by
requiring words and expressions in one sentence be interpreted by reference to
words and expressions in the surrounding sentences and paragraphs.
  
Of particular interest here are conjunctive elements, which Halliday and Hasan
took to signal how an upcoming sentence is related to what has been said before. Such conjunctive elements include both co-ordinating and sub-ordinating conjunctions and what they call conjunctive adjuncts (eg, adverbs such as but, so, next,
accordingly, actually, instead, besides, etc.; and prepositional phrases such as as a
result, in addition, in spite of that, in that case, etc.).
More specifically, a conjunctive element is taken to convey a cohesive relation between its matrix sentence and that part of the surrounding discourse that
supports its effective decoding – resolving its reference, identifying its sense, or
recovering missing material needed for its interpretation. Three examples of conjunctive elements can be found in this extract given from Meeting Wilfred Pickles,
by Frank Haley.
Then we moved into the country, to a lovely little village called Warley.
It is about three miles from Halifax.
There are quite a few about.
There is a Warley in Worcester and one in Essex.
But the one not far out of Halifax had had a maypole, and a fountain.
By this time the maypole has gone, but the pub is still there called the
Halliday and Hasan labelled the adverb then in (1-a) a   
 (like next), which is a type of   ,
which is itself a type of  . It is decoded as conveying a simple sequential temporal relation to something in the preceding text (not provided
here). But in (1-e) was labelled a    
(like and), which is a type of a   , itself a
type of  . But is decoded here as conveying a simple contrastive adversative relation to (1-d). The final conjunctive element in this text,
by this time, Halliday and Hasan labelled a    
 (like until then), which is a type of   , which
is another type of  . By this time is decoded as conveying a
complex temporal relation to (1-a).
Although Halliday and Hasan provided a very elaborate taxonomy of conjunctive elements in terms of the hierarchy of detailed labels for conjunctive relations
illustrated above, they did not embed it in any kind of theoretical framework that
would explain, for example, how meaning is projected systematically from a conjunctive relation with a given label or what “surrounding text” a link can be made
to. Without such a theoretical framework, it was difficult to make use of their data
analysis in the systematic way required for computational applications.
OSLa volume 1(1), 2009
 
 80  90:    
Providing a theoretic framework for what can be termed lexically grounded discourse relations was not, however, what researchers in the 1980s and 1990s were
concerned with. Rather, they aimed to provide a complete theoretical account
of a text in terms of abstract discourse relations. Such accounts included Rhetorical Structure Theory (hereafter, RST) developed by Mann and Thompson (1988);
a theory (hereafter, GS) developed by Grosz and Sidner (1986) that posited three
separate but isomorphic discourse structures – an intentional structure, a linguistic structure and an attentional structure; the Linguistic Discourse Model (hereafter, LDM) developed by Polanyi and her colleagues (Polanyi 1988; Polanyi and
van den Berg 1996); Relational Discourse Analysis (hereafter RDA) developed by
Moser and Moore (1996) as a way of reconciling RST and ; and Structured Discourse Representation Theory (hereafter ) developed by Asher and Lascarides
(2003) as an extension to DRT (Kamp and Reyle 1993) to account for how discourse
relations arise from what the authors call commonsense entailment (Lascarides and
Asher 1993). The abstract discourse relations used in these theories included both
semantic relations between the facts, beliefs, situations, eventualities, etc. described in a text (more often called informational relations) and pragmatic relations
between what a speaker is trying to accomplish with one part of a text with respect to another (more often called intentional relations). Together, these are often
simply called discourse relations.
Unlike the lexically-grounded discourse relations of Halliday and Hasan (1976),
theories of abstract discourse relations see comprehensive structure (based on
discourse relations) underlying a text, just as theories of syntax see a comprehensive syntactic structure underlying a sentence. In particular, theories of abstract
discourse relations assume, as in formal grammar, a set of terminal elements
(called either elementary or basic discourse units). A discourse relation holding
between adjacent elements joins them recursively into a larger unit, with a text
being recursively analysable down to its terminal elements. Such an analysis covers the entire text, just as in syntax a single parse tree or dependency analysis
covers an entire sentence. Also as in syntax, the analysis is essentially a tree
structure since, for the most part, it is also assumed that no discourse element
is part of more than one larger element. Where these theories differ is in (1) the
specific types of relations they take to hold between units; (2) the amount of attention they give to, eg, how a hearer establishes the relation that holds between
two units; (3) whether they take there to be separate but related informational and
intentional discourse structures (GS, ) or a single structure (RST, , );
and (4) how to provide a compositional syntactic-semantic interface that would
systematically interpret a discourse unit in terms of the interpretations assigned
to its component parts.
These theories were not without practical application, and came to underOSLa volume 1(1), 2009
  
pin work in Natural Language Generation (Marcu 1996; Mellish et al. 1998; Moore
1995) and document summarization (Bosma 2004; Marcu 1998, 2000). However,
these theories were based on very little data, and problems began to be noticed
early on. For example, Moore and Pollack (1992) noticed that the same piece of
text could simultaneously be given different  analyses, each with a different
structure. These were not alternative analyses that could be disambiguated: all
of them seemed simultaneously appropriate. But if this was the case, what was
the consequence for associating a text with a single discourse structure (or even
structurally isomorphic intentional and informational structures)?
Elsewhere, Scott and de Souza (1990) and later Carlson et al. (2003) pointed
out that the meaning conveyed by a sequence of sentences or a complex sentence
could also be conveyed by a single clause. But if this was the case, did it make
sense to posit an independent existence for elementary discourse units?
And questions about discourse having an underlying recursive tree-like structure were raised by Wiebe (1993) based on examples like
The car was finally coming toward him.
He [Chee] finished his diagnostic tests,
feeling relief.
But then the car started to turn right.
The problem she noted was that the discourse connectives but and then appear to
link clause (2-d) to two different things: then to clause (2-b) in a  relation – i.e., the car starting to turn right being the next relevant event after Chee’s
finishing his tests – and but to a grouping of clauses (2-a) and (2-c) – i.e., reporting
a contrast between, on the one hand, Chee’s attitude towards the car coming towards him and his feeling of relief and, on the other hand, his seeing the car turning right. But a structure with one subtree over the non-adjacent units (2-b) and
(2-d) and another over the units (2-a), (2-c) and (2-d) again is not itself a tree. Did
such examples really raise a problem for postulating an underlying comprehensive tree-like structure for discourse, and if they did, what kind of comprehensive
structure, if any, did discourse have?
  :   
While the problems noted above weren’t enough to refute theories of abstract
discourse relations, two distinct currents gaining momentum in sentence-level
syntax through the 1990s have led discourse research to re-focus on data again
in the new century. These were the emergence of (1) part-of-speech and syntactically annotated corpora such as the Penn TreeBank (Marcus et al. 1993); and
(2) lexicalized grammars, in which syntactic contexts were associated with words
directly rather than indirectly through phrase structure rules (eg, syntactic conOSLa volume 1(1), 2009
 
texts in the form of tree fragments in Lexicalized Tree-Adjoining Grammar (LTAG)
(Schabes 1990; XTAG-Group 2001) and in the form of complex categories in Combinatory Categorial Grammar (CCG) (Steedman 1996, 2000)).
In Section 3.1, we will make some general remarks about annotated corpora,
followed by a brief description in Section 3.2 of a lexicalized grammar for discourse, and then finally, in Section 3.3, a bit about a particular annotated discourse corpus – the Penn Discourse TreeBank – that was stimulated by this work
to return to a focus on lexically-grounded discourse relations. We will close with
some predictions about the future.
[3.1] Annotated Discourse Corpora
While the automatically generated, manually corrected part-of-speech and syntactic annotation that makes up the Penn TreeBank (PTB) Wall Street Journal Corpus was developed as a community-accepted “gold standard” on which parsers
and parsing techniques could be evaluated (Marcus et al. 1993), the PTB soon became a basis for inducing parsers using statistical and machine learning techniques that were much more successful in wide-coverage parsing than any ones
previously developed. Similar techniques based on appropriately annotated corpora enabled the development of other language technology, including part-ofspeech taggers, reference resolution procedures, semantic role labellers, etc., which
are beginning to be used to improve performance of applications in information
retrieval, automated question answering, statistical machine translation, etc.
Eventually, researchers turned to consider whether annotated discourse corpora can yield similar benefits by supporting the development of technology that
required sensitivity to discourse structure, such as in extractive summarization,
where one identifies and includes in a summary of one or more source texts, only
their most important sentences, along with, perhaps, other sentences needed to
make sense of them. This was one of the motivations behind the creation of the
 Corpus (Carlson et al. 2003), comprising 385 documents from the Penn TreeBank corpus that have been manually segmented into elementary discourse units,
linked into a hierarchy of larger and larger units, and annotated with  relations
taken to hold between linked units. It was also the acknowledged reason for developing an -annotated corpus (Polanyi et al. 2004). Another corpus was annotated according to  (Moser and Moore 1996), with an aim of improving the
quality of Natural Language Generation (NLG) – in particular, to identify which of
several syntactic variants is most natural in a given context – information that can
then be incorporated into the sentence planning phase of NLG (Di Eugenio et al.
1997). (Prasad et al. (2005) discusses how the Penn Discourse TreeBank, to be discussed in Section 3.3, can also be used for this purpose.)
While the promise of benefits for Language Technology have helped attract
funding for annotated discourse corpora, such corpora serve other objectives as
OSLa volume 1(1), 2009
  
well. For example, researchers want to use them to test hypotheses about the
effect (or co-dependency) of discourse structure on other aspects of language
such as argumentation (Stede 2004; Stede et al. 2007) or reference resolution (cf.
work done at the University of Texas on a corpus annotated according to 
(Stede et al. 2007)).
But a fundamental reason for developing annotated discourse corpora is to
enable us to advance towards a theoretically well-founded understanding of discourse relations – ie, of how they arise from text, or of constraints on their possible arguments, or of how the resulting structures pattern – that is well-grounded
in empirical data. This was a main goal behind the development of the Discourse
GraphBank (Wolf and Gibson 2005).1 And already there has been a real advance in
discourse theory based on empirical data (Stede 2008): Problems with annotating
 relations in the Potsdam Commentary Corpus (Stede 2004; Stede et al. 2007)
led Stede to examine in detail the notion of nuclearity that has been fundamental
to  as a theory but is problematic in practice. The result is an argument for
nuclearity as primitive only for intentional discourse relations. For informational relations, nuclearity is best discarded in favor of independently motivated notions of
discourse salience or prominence (associated with entities) and discourse topic. Before
turning to a brief description of the Penn Discourse TreeBank (Miltsakaki et al.
2004; Prasad et al. 2004; Webber 2005) and some things we have already learned
from the process of creating it, we will briefly describe the lexicalized approach
to discourse that led to the creation of this lexically-grounded corpus.
[3.2] Discourse Lexicalized Tree-Adjoining Grammar (D-LTAG)
D-LTAG is a lexicalized approach to discourse relations, which aims to provide an
account of how lexical elements (including phrases) anchor discourse relations
and how other parts of the text provide arguments for those relations (Webber et al.
2003; Webber 2004). D-LTAG arose from a belief that language has only a limited
number of ways to convey relations between things. For example, within a clause,
• a verb or preposition can convey a relation holding between its arguments
(eg, the cat ate the cheese; the cat in the hat), as can some nouns and adjectives (eg, the love of man for his family; an easy mountain to climb);
• adjacency of two or more elements can convey implicit relations holding
between them, as in soup pot, soup pot cover, aluminum soup pot cover adjustment screw, etc;
• an anaphoric expression conveys a relation between all or part of its denotation and some element of the surrounding discourse – an  rela[1]
OSLa volume 1(1), 2009
(Webber 2006) argues against the strong claims (Wolf and Gibson 2005) make about discourse structure
based on this annotation, and many of them have subsequently been withdrawn (Kraemer and Gibson
 
tion in the case of coreference, other relations in the case of comparative
anaphora such as the smaller boys;
• intonation and information structure can convey relations of  between discourse elements and/or their different roles with respect to information structure (Steedman 2007).
If one assumes that many of the same means are operative outside the clause as
within it, then it makes sense to adopt a similar approach to discourse analysis
as to syntactic analysis. Since lexicalized grammars seemed to provide a clearer,
more direct handle on relations and their arguments at the clause-level, D-LTAG
adopted a lexicalized approach to discourse based on lexicalized Tree Adjoining
Grammar (Schabes 1990). A lexicalized TAG (LTAG) differs from a basic TAG in taking each lexical entry to be associated with the set of elementary tree structures
that specify its local syntactic configurations. These structures can be combined
via either substitution or adjoining, to produce a complete sentential analysis.
The elementary trees of D-LTAG are anchored by discourse connectives whose
substitution sites correspond to their arguments. These can be filled by anything
interpretable as an abstract object (ie, as a proposition, fact, eventuality, situation,
etc.). Elements so interpretable include discourse segments, sentences, clauses,
nominalisations and demonstrative pronouns. Adjacency in D-LTAG is handled
by an elementary tree anchored by an empty connective.
As with a sentence-level LTAG, there are two types of elementary trees in DLTAG: Initial trees, anchored by structural connectives such as subordinating conjunctions and subordinators (eg, in order to, so that, etc.) illustrated by the trees
labeled α:so and α:because_mid in Figure 1(a), and auxiliary trees, anchored by a
coordinating conjunction or an empty connective or by a discourse adverbial, illustrated by the trees labeled β:but and β:then in Figure 1(a).
Both the initial trees of structural connectives and the auxiliary trees of coordinating conjunctions and the empty connective reflect the fact that both their
arguments are provided through structure – in the case of initial trees, the two
substitution arguments labelled with ↓ in α:so and α:because_mid, and in the case
of auxiliary trees, the one substitution argument labelled with ↓ and the adjunction argument labelled with *, as in β:but. In contrast, a discourse adverbial is
anaphoric – with the discourse relation, part of its semantics, but with one of its
arguments coming from the discourse context by a process of anaphor resolution.
Its other argument is provided structurally, in the form of its matrix clause or
sentence. That discourse adverbials such as instead, afterwards, as a result, etc. are
anaphoric, differing from structural connectives in getting their second argument
from the discourse context, is argued on theoretical grounds in (Webber et al.
2003) and on empirical grounds in (Creswell et al. 2004). It also echoes in part
the claim of Halliday and Hasan (1976), noted in Section 1, that all conjunctive elOSLa volume 1(1), 2009
  
α: so
β: but
α: because_mid
T1 love
T2 order
T3 cancel
T4 discover
T2 order T3 cancel
T1 love
(a) Derived Tree for Example (3).
α: so
β: but
α: because_mid
(b) Derivation Tree for Example (3)
 1: Tree analyses of Example (3)
OSLa volume 1(1), 2009
 
ements were interpreted in this way. Justification of the anaphoric character of
discourse adverbials is given in (Forbes 2003) and (Forbes-Riley et al. 2006).
Discourse relations arising from both structural and anaphoric connectives
can be seen in the D-LTAG analysis of Example (3) below.
John loves Barolo.
S he ordered three cases of the ’97.
B he had to cancel the order
 he  discovered he was broke.
Figure 1(a) illustrates both the starting point of the analysis — a set of elementary
trees for the connectives (so, but, because, then) and a set of leaves (T 1-T 4) for the
four clauses in Example (3) without the connectives — and its ending point, the
derived tree that results from substituting at the nodes in α:because_mid, α:so and
β:but marked ↓ and adjoining the trees β:but and β:then at their nodes marked ∗.
These operations of substitution and adjoining are shown as solid and dashed lines
respectively in Figure 1(b), in what is called a derivation tree. The numbers on the
arcs of the derivation tree refer to the node of the tree that an operation has been
performed on. For example, the label 1 on the solid line from α:so to T1 means
that T1 has substituted at the leftmost node of the tree α:so. The label 3 refers to
the node that is third from the left. The label 0 on the dashed line from α:so to
β:but means that β:but has adjoined at the root of α:so.
The structural arguments of a connective can come about through either substitution or adjoining. The derived tree in Figure 1(a) shows the two structural
arguments of so, but and because as their left and right sisters. The derivation tree
in Figure 1(a) shows both arguments to so and because coming from substitution,
with one structural argument to but coming from substitution and the other coming through adjoining. Finally, then only has one structural argument, shown in
Figure 1(a) as its right sister. Figure 1(b) shows it coming from adjoining. The dotted line in Figure 1(a) shows then linked anaphorically to the clause that gives rise
to its first argument. More detail on both the representation of connectives and
D-LTAG derivations is given in (Webber et al. 2003). A preliminary parser producing such derivations is described in (Forbes et al. 2003) and (Webber 2004).
Compositional interpretation of the derivation tree produces the discourse
relation intepretations associated with because, so and but, while anaphor resolution produces the second argument to the discourse relation interpretation associated with then (ie, the ordering event), just as it would if then were paraphrased
as soon after that, with the pronoun that resolved anaphorically. Details on this
syntactic-semantic interface are given in (Forbes-Riley et al. 2006).
Although D-LTAG produces only analyses in the form of trees, Webber et al.
(2003) recognized that occasionally the same discourse unit participates in one
relation with its left-adjacent material and another distinct relation with its rightOSLa volume 1(1), 2009
  
adjacent material, as in
A   the tremor passed, many people spontaneously arose and cheered,
  it had been a novel kind of pre-game show.2
W Ms. Evans took her job, several important divisions that had reported
to her predecessor weren’t included   she didn’t wish to be a
full administrator.
In Example (4), the main clause “many people spontaneously arose and cheered”
serves as ARG1 to both the subordinating conjunction    on its left and
the subordinating conjunction   on its right. Example (5) shows a similar pattern. Such examples, however, would require relaxing substitution constraints in D-LTAG that the same tree only substitutes into a single site. This has
not yet been done. It should be clear that D-LTAG is only a conservative extension
of the theories mentioned in Section 2 in that it shares their assumptions that
a text can be divided into discourse units corresponding to clauses, with a discourse analysis covering those units – hence the work on developing a discourse
parser for D-LTAG (Forbes et al. 2003; Webber 2004). The main way in which DLTAG diverges from these other theories is in anchoring discourse relations in, on
the one hand, structural connectives and adjacency, and on the other, anaphoric
connectives. The latter provide additional relations between material that is not
necessarily adjacent, but not in a way that changes the complexity of discourse
structure: D-LTAG analyses are still trees.
[3.3] The Penn Discourse TreeBank
The Penn Discourse TreeBank (PDTB) annotates discourse relations in 2304 articles
of the Wall Street Journal corpus (Marcus et al. 1993) in terms of discourse connectives, the minimal text spans that give rise to their arguments, and the attribution of both the connectives and their arguments (Dinesh et al. 2005; Prasad et al.
2007). For example, in Example (6)
Factory orders and construction outlays were largely flat in December 
purchasing agents said manufacturing shrank further in October.
both ARG1 and the connective while are attributed to the writer, while ARG2 is
attributed to someone else via the attributive phrase “purchasing agents said”,
which is not included in that argument.
Primarily two types of connectives have been annotated in the PDTB: explicit
connectives and implicit connectives, the latter being inserted between adjacent
OSLa volume 1(1), 2009
Both these examples come from the Penn Discourse TreeBank (Section 3.3) and reflect the actual annotation of these connectives. In all PDTB examples, ARG1 is in italics and ARG2 in boldface, and the
connective is underlined.
 
S-medial so
S-medial but
S-initial So
S-initial But
 2: S-medial vs. S-initial Connectives
paragraph-internal sentences not related by an explicit connective. As in D-LTAG
(Section 3.3), explicit connectives include coordinating conjunctions, subordinating conjunctions and discourse adverbials. The argument associated syntactically
with the discourse connective is conventionally referred to as ARG2 (eg, the subordinate clause of a subordinating conjunction) and the other argument as ARG1.
Because annotators were asked to annotate only the minimal span associated with
an argument, they were also allowed (but not required) to indicate spans adjacent
to ARG1 and ARG2 that were relevant to them but still supplementary, using the
tags SUP1 and SUP2.
A preliminary version of the PDTB containing the annotation of 18505 explicit
connectives was released in April 2006 (PDTB-Group 2006), and received over 120
downloads. The completed PDTB (Version 2.0) was released by the Linguistic Data
Consortium (LDC) in February 2008, and includes annotation of all implicit connectives as well, along with a hierarchical semantic annotation of both explicit
and implicit connectives (Miltsakaki et al. 2008). More information on the PDTB
can be found at its homepage (
Although annotation has sometimes been a challenge, it has nevertheless begun to yield some useful observations. One such observation is that the position
of ARG1 can differ significantly, depending on whether its associated connective
occurs sentence-medially (S-medial) or sentence-initially (S-initial). Figure 2 contrasts the instances of S-medial and S-initial so and but in the corpus3 with respect
to whether their ARG1 spans the immediately preceding clause(s) or sentence(s)
(Adjacent) or not (Non-adjacent). This difference in patterning between medial
instances of but and so and sentence-initial instances, is statistically significant
The difference will certainly be relevant to those researchers interested in
developing the technology to automatically recognize the arguments to discourse
With But, figures are based on the first 213 of 1188 S-medial instances in the corpus and the first 410 of
2124 S-initial instances.
OSLa volume 1(1), 2009
  
connectives (cf. Wellner and Pustejovsky 2007). But it is not just the distance of
ARG1 from its connective: it is also a difference in function. As with S-initial but
and so, A1 of S-initial discourse adverbials like instead can also be found at a
distance – eg,
On a level site you can provide a cross pitch to the entire slab by raising
one side of the form, but for a 20-foot-wide drive this results in an awkward
5-inch slant across the drive’s width. I, make the drive higher at
the center.
(Reader’s Digest New Complete Do-it-yourself Manual, p. 154)
If government or private watchdogs insist, however, on introducing greater friction between the markets (limits on price moves, two-tiered execution, higher margin requirements, taxation, etc.), the end loser will be the markets themselves.
I, we ought to be inviting more liquidity with cheaper ways to
trade and transfer capital among all participants.
But of the 48 occurrences of S-initial Instead in the corpus, ARG1 spans text other
than the main clause in 26 (54.2%), independent of its position with respect to
the connective. In Example (8), ARG1 spans the subordinate clause, and in Example (9), the gerund complement of an appositive NP:
The tension was evident on Wednesday evening during Mr. Nixon’s final
banquet toast, normally an opportunity for reciting platitudes about eternal
friendship. I, Mr. Nixon reminded his host, Chinese President
Yang Shangkun, that Americans haven’t forgiven China’s leaders for
the military assault of June 3-4 that killed hundreds, and perhaps
thousands, of demonstrators.
This suggests that ARG1 of a discourse adverbial like instead comprises a span with
a specific semantic character — here, one that can be interpreted as something
for which an alternative exists (Webber 2004). The span can be anywhere in the
sentence and serve any role. In contrast, ARG1 of S-initial So or But comprises
a span at the same level of discourse embedding as ARG2, with any intervening
material appearing to serve a supporting role to the span identified as ARG1, as
OSLa volume 1(1), 2009
The 40-year-old Mr. Murakami is a publishing sensation in Japan. A more recent novel, “Norwegian Wood” (every Japanese under 40 seems to be fluent in Beatles lyrics), has sold more than four million copies since Kodansha published it in 1987. B he is just one of several youthful writers
– Tokyo’s brat pack – who are dominating the best-seller charts in
 
It is difficult, if not impossible, for anyone who has not pored over the thousands
of pages of court pleadings and transcripts to have a worthwhile opinion on the
underlying merits of the controversy. Certainly I do not. S we must look
elsewhere for an explanation of the unusual power this case has exerted over the minds of many, not just in Washington but elsewhere
in the country and even the world.
Additional (albeit incomplete) evidence for this role of the intervening text comes
from the frequency with which some or all of the intervening material has been
annotated SUP1 in the PDTB, even though the annotators were not required to be
systematic or rigorous in their use of this label.
This pragmatic/intentional sense of a supporting role (or a difference in focality or discourse salience) appears to be what Mann and Thompson (1988) had in
mind with presentational rhetorical relations in which one argument (the satellite)
supported the other argument (the nucleus) – and what Grosz and Sidner (1986)
had in mind when they posited a dominance relation in which one discourse segment supported the discourse purpose of another. It also appears related to what
Blühdorn (2007) and Ramm and Fabricius-Hansen (2005) refer to as subordination
in discourse. It has not been directly annotated in the PDTB, which has focussed
on the arguments to explicit and implicit connectives. On the other hand, we
should be able to gather additional evidence related to this issue once the annotation of implicit connectives in the PDTB has been adjudicated, at which point
we can survey what relations annotators have taken to hold between ARG1 and
material intervening between it and its associated connective. Such empirical
evidence of something like intentional structure would be very exciting to find.
A second observation relates to other linguistic devices that convey discourse
relations, besides the discourse connectives annotated in the PDTB. In annotating
implicit connectives, we noticed different cases whose paraphrase in terms of an
explicit connective sounded redundant. A closer look revealed systematic nonlexical indicators of discourse relations, including:
• cases of S-initial PPs and adjuncts with anaphoric or deictic NPs such as at
the other end of the spectrum, adding to that speculation, which convey a relation (eg,  and   in the above two cases)
in which the immediately preceding sentence provides a referent for the
anaphoric or deictic expression, which thereby makes it available as ARG1.
• cases of equative be in which the immediately preceding sentence provides
a referent for an internal anaphoric argument of a relation-containing NP
in subject position – eg, the  relation conveyed by the effect is
(ie, the effect [of that] is) in
OSLa volume 1(1), 2009
  
The New York court also upheld a state law, passed in 1986, extending for
one year the statute of limitations on filing DES lawsuits. T  
 lawsuits that might have been barred because they were
filed too late could proceed because of the one-year extension.
Identification of these alternative ways of conveying discourse relations (labelled
 in the PDTB) will allow for a more complete annotation of discourse relations in other corpora, including corpora in languages other than English.4
A third observation relates to the distinction that Blühdorn (2007) has drawn
between discourse hierarchy associated with the pragmatic/intentional concept
of focality versus discourse hierarchy associated with the syntactic/semantic concept of constituency. It is clear that both some sort of intentional structure and some
sort of informational structure are needed in discourse, just as was suggested in
the theories of Grosz and Sidner (1986), Moore and Pollack (1992), and Moser and
Moore (1996). It is just that the data seem to tell a somewhat different story of
the properties of these structures than those assumed in these earlier theories of
abstract discourse relations.
Our previous observation about the status of material intervening between
ARG1 and S-initial But and So related to a discourse hierarchy associated with focality, with the intervening material playing a supplemental role in the discourse.
Our observation here relates to constituency and is a consequence of the procedure
used in annotating the PDTB. This differed from the procedure used in annotating
the  corpus (Carlson et al. 2003), the  corpus (Polanyi et al. 2004) and the
GraphBank corpus (Wolf and Gibson 2005), where annotators first marked up the
text with a sequence of elementary discourse units that exhaustively covered it (just
as a sequence of words and punctuation exhaustively covers a sentence) and then
identified which of the units served as arguments to each relation. Instead, as
already noted, PDTB annotators were instructed to select the minimal clausal text
span needed to interpret each of the two arguments to each explicit and implicit
discourse connective.
We reported in (Dinesh et al. 2005) cases where the so-annotated arguments
to discourse connectives diverged from syntactic constituents within the Penn
TreeBank (PTB) — in particular, where instances of attribution (eg, “purchasing
Independently of the PDTB, we also noticed that discourse relations could be conveyed by marked syntax
– for example, the expression of a  relation in the marked syntax of
Had I known the Queen would be here, I would have dressed better.
or of a  in the marked syntax of
The more food you eat, the more weight you’ll gain.
Taking account of these examples will allow a more complete annotation of other corpora.
OSLa volume 1(1), 2009
 
agents said” in Example (6) earlier and “analysts said” in Example (13)), headed
non-restrictive relatives (eg, “which couldn’t be confirmed” in Example (13)), and
headless non-restrictive relatives (eg, “led by Rep. Jack Brooks (D., Texas)” in
Example (14)), all of which are constituents within the PTB’s syntactic analyses,
are nevertheless not part of the discourse arguments of which they are syntactic
A traders rushed to buy futures contracts, many remained skeptical about the Brazilian development, which couldn’t be confirmed, analysts
Some Democrats, led by Rep. Jack Brooks (D., Texas), unsuccessfully opposed
the measure because they fear that the fees may not fully make up for the budget
B Justice Department and FTC officials said they expect the filing fees
to make up for the budget reductions and possibly exceed them.
This suggests another way in which syntax and discourse may diverge, in addition to the lack of correspondence between coordination and subordination in
syntax and in discourse noted by Blühdorn (2007): Material that is part of a constituent analysis in syntax may not be part of a constituent analysis in discourse.
Also relevant here are examples such as (4) above, where constituency structure
in syntax is standardly taken to be a tree, while in discourse, it appears to be a
simple DAG (with the main clause a constituent of two distinct discourse connectives). Lee et al. (2006) presents an array of fairly complex constituency patterns
of spans within and across sentences that serve as arguments to different connectives, as well as parts of sentences that don’t appear within the span of any
connective, explicit or implicit. The result is that the PDTB provides only a partial but complexly-patterned cover of the corpus. Understanding what’s going on
and what it implies for discourse structure (and possibly syntactic structure as
well) is a challenge we’re currently trying to address.
There is renewed interest in discourse structure, as more and larger annotated
discourse corpora are becoming available for analysis. While there is still much
to be gained from trying to extract as much as possible through machine learning
methods based on superficial features of discourse that we believe we understand,
there is also much to be gained from a deeper analysis of discourse structure that
suggests new features that we are only now beginning to discover.
OSLa volume 1(1), 2009
  
We would like to thank Nikhil Dinesh, Aravind Joshi, Mark Steedman and Bergljot
Behrens for their comments on an earlier draft of this paper. The result of their
suggestions is a much more coherent and, to our minds, more interesting paper.
Asher, N. and Lascarides, A. 2003. Logics of Conversation. Cambridge UK: Cambridge
University Press.
Blühdorn, H. 2007. Subordination and Coordination in Syntax, Semantics and Discourse: Evidence from the study of Connectives. In C. Fabricius-Hansen and
W. Ramm (eds.), ’Subordination’ versus ’Coordination’ in Sentence and Text, pages
59–89, John Benjamins.
Bosma, W. 2004. Query-Based Summarization Using Rhetorical Structure Theory.
In Proceedings of 15th Annual Meeting of Computational Linguistics in the Netherlands
(CLIN), pages 29–44.
Carlson, L., Marcu, D. and Okurowski, M. E. 2003. Building a Discourse-Tagged
Corpus in the framework of Rhetorical Structure Theory. In J. van Kuppevelt &
R. Smith (ed.), Current Directions in Discourse and Dialogue, New York: Kluwer.
Creswell, C., Forbes, K., Miltsakaki, E., Prasad, R., Joshi, A. and Webber, B. 2004. The
Predicate-Argument Structure of Discourse Connectives: A corpus-based study.
In Tony McEnery Antonio Branco and Ruslan Mitkov (eds.), Anaphora Processing:
Linguistic, Cognitive and Computational Modeling, John Benjamins.
Di Eugenio, B., Moore, J. and Paolucci, M. 1997. Learning Features that Predict Cue
Usage. In Proceedings of the 35th Annual Meeting of the Association for Computational
Linguistics (ACL97/EACL97), pages 80–87, Madrid, Spain.
Dinesh, N., Lee, A., Miltsakaki, E., Prasad, R., Joshi, A. and Webber, B. 2005. Attribution and the (Non-)Alignment of Syntactic and Discourse Arguments of Connectives. In ACL Workshop on Frontiers in Corpus Annotation, Ann Arbor MI.
Forbes, K. 2003. Discourse Semantics of S--Modifying Adverbials. Ph. D.thesis, Department of Linguistics, University of Pennsylvania.
Forbes, K., Miltsakaki, E., Prasad, Rashmi, Sarkar, Anoop, Joshi, A. and Webber,
B. 2003. D-LTAG System: Discourse Parsing with a Lexicalized Tree-Adjoining
Grammar. Journal of Logic, Language and Information 12.
Forbes-Riley, K., Webber, B. and Joshi, A. 2006. Computing Discourse Semantics:
The Predicate-Argument Semantics of Discourse Connectives in D-LTAG. Journal
of Semantics 23, 55–106.
OSLa volume 1(1), 2009
 
Grosz, B. and Sidner, C. 1986. Attention, Intention and the Structure of Discourse.
Computational Linguistics 12(3), 175–204.
Halliday, M. and Hasan, R. 1976. Cohesion in English. Longman.
Kamp, H. and Reyle, U. 1993. From Discourse to Logic. Dordrecht NL: Kluwer.
Kraemer, J. and Gibson, T. 2007. Ordering Constraints on Discourse Relations. In
Proceedings of the 20th Annual CUNY Conference on Human Sentence Processing, page
116, La Jolla CA.
Lascarides, Alex and Asher, Nicholas. 1993. Temporal Interpretation, Discourse Relations and Commonsense Entailment. Linguistics and Philosophy 16(5),
Lee, A., Prasad, R., Joshi, A., Dinesh, N. and Webber, B. 2006. Complexity of Dependencies in Discourse: Are dependencies in discourse more complex than in
syntax? In Proc. 5th Workshop on Treebanks and Linguistic Theory (TLT’06), Prague
Mann, William and Thompson, Sandra. 1988. Rhetorical Structure Theory: Toward a Functional Theory of Text Organization. Text 8(3), 243–281.
Marcu, D. 1996. Building up Rhetorical Structure Trees. In Proceedings of AAAI--96,
pages 1069–1074, Portland OR.
Marcu, D. 1998. Improving summarization through rhetorical parsing tuning.
In Proceedings of 6th Workshop on Very Large Corpora, pages 206–215, Montreal,
Marcu, D. 2000. The theory and practice of discourse parsing and summarization. MIT
Marcus, M., Santorini, B. and Marcinkiewicz, M. A. 1993. Building a Large Scale
Annotated Corpus of English: The Penn TreeBank. Computational Linguistics 19,
Mellish, C., Knott, A., Oberlander, J. and O’Donnell, M. 1998. Experiments using
stochastic search for text planning. In Proceedings of the Ninth International Workshop on Natural Language Generation, pages 98–107, Niagara-on-the-Lake, Canada.
Miltsakaki, E., Prasad, R., Joshi, A. and Webber, B. 2004. Annotating Discourse Connectives and Their Arguments. In NAACL/HLT Workshop on Frontiers in Corpus Annotation, Boston.
OSLa volume 1(1), 2009
  
Miltsakaki, E., Robaldo, Livio, Lee, A. and Joshi, A. 2008. Sense Annotation in the
Penn Discourse Treebank. In Computational Linguistics and Intelligent Text Processing, pages 275–286, Springer.
Moore, J. 1995. Participating in Explanatory Dialogues. Cambridge MA: MIT Press.
Moore, J. and Pollack, M. 1992. A problem for RST: The need for multi-level discouse analysis. Computational Linguistics 18(4), 537–544.
Moser, M. and Moore, J. 1996. Toward a Synthesis of Two Accounts of Discourse
Structure. Computational Linguistics 22(3), 409–419.
PDTB-Group. 2006. The Penn Discourse TreeBank 1.0 Annotation
Manual. Technical Report IRCS 06-01, University of Pennsylvania,
Polanyi, L. 1988. A Formal Model of the Structure of Discourse. Journal of Pragmatics
12, 601–638.
Polanyi, L., Culy, C., van den Berg, M. H., Thione, G. L. and Ahn, D. 2004. Sentential
Structure and Discourse Parsing. In Proceedings of the ACL Workshop on Discourse
Annotation, pages 80–87, Barcelona, Spain.
Polanyi, L. and van den Berg, M. H. 1996. Discourse Structure and Discourse Interpretation. In P. Dekker and M. Stokhof (eds.), Proceedings of the Tenth Amsterdam
Colloquium, pages 113–131, University of Amsterdam.
Prasad, R., Dinesh, N., Lee, A., Joshi, A. and Webber, B. 2007. Attribution and its
Annotation in the Penn Discourse TreeBank. TAL (Traitement Automatique des
Langues) .
Prasad, R., Joshi, A., Dinesh, N., Lee, A., Miltsakaki, E. and Webber, B.
2005. The Penn Discourse TreeBank as a Resource for Natural Language
Generation. In Proceedings of the Corpus Linguistics Workshop on Using Corpora for Natural Language Generation, Birmingham UK, slides available at
Prasad, R., Miltsakaki, E., Joshi, A. and Webber, B. 2004. Annotation and Data Mining of the Penn Discourse TreeBank. In ACL Workshop on Discourse Annotation,
pages 88–95, Barcelona, Spain.
Ramm, W. and Fabricius-Hansen, C. 2005. Coordination and Discourse-structural
Salience from a Cross-linguistic Perspective. In Salience in Discourse: Multidisciplinary Approaches to Discourse, pages 119–128, Munster, Germany: Stichting/Nodus.
OSLa volume 1(1), 2009
 
Schabes, Y. 1990. Mathematical and Computational Aspects of Lexicalized Grammars.
Ph. D.thesis, Department of Computer and Information Science, University of
Scott, D. and de Souza, C. S. 1990. Getting the Message Across in RST--based Text
Generation. In Robert Dale, C. Mellish and Michael Zock (eds.), Current Research
in Natural Language Generation, pages 47–73, London, England: Academic Press.
Stede, M. 2004. The Potsdam Commentary Corpus. In ACL Workshop on Discourse
Annotation, Barcelona, Spain.
Stede, M. 2008. RST revisted: Disentangling nuclearity. In C. Fabricius-Hansen and
W. Ramm (eds.), ’Subordination’ versus ’Coordination’ in Sentence and Text, pages
33–59, Amsterdam: John Benjamins.
Stede, M., Wiebe, J., Hajičová, E., Reese, B., Webber, B. and Wilson, T. 2007. Panel
Session on Discourse Annotation. In ACL Workshop on Language Annotation Workshop, Prague.
Steedman, M. 1996. Surface Structure and Interpretation. Cambridge MA: Linguistic
Inquiry Monograph 30, MIT Press.
Steedman, M. 2000. The Syntactic Process. Cambridge MA: MIT Press.
Steedman, Mark. 2007. Surface Compositional Semantics of Intonation, submitted.
Webber, B. 2004. D-LTAG: Extending Lexicalized TAG to Discourse. Cognitive Science
28, 751–779.
Webber, B. 2005. A Short Introduction to the Penn Discourse TreeBank. In Copenhagen Working Papers in Language and Speech Processing.
Webber, B. 2006. Accounting for Discourse Relations: Constituency and Dependency. In M. Dalrymple M. Butt and T. King (eds.), Intelligent Linguistic Architectures, pages 339–360, CSLI Publications.
Webber, B., Stone, Matthew, Joshi, A. and Knott, Alistair. 2003. Anaphora and Discourse Structure. Computational Linguistics 29, 545–587.
Wellner, B. and Pustejovsky, J. 2007. Automatically Identifying the Arguments to
Discourse Connectives. In Proceedings of the 2007 Conference on Empirical Methods
in Natural Language Processing (EMNLP), Prague CZ.
Wiebe, J. 1993. Issues in Linguistic Segmentation. In Workshop on Intentionality
and Structure in Discourse Relations, Association for Computational Linguistics, pages
148–151, Ohio StateUniversity.
OSLa volume 1(1), 2009
  
Wolf, F. and Gibson, E. 2005. Representing Discourse Coherence: A Corpus-based
Study. Computational Linguistics 31, 249–287.
XTAG-Group. 2001. A Lexicalized Tree Adjoining Grammar for English. Technical Report IRCS 01-03, University of Pennsylvania, see
  
Bonnie Webber
University of Edinburgh
2 Buccleuch Place
Edinburgh EH8 9LW
United Kingdom
[email protected]
Rashmi Prasad
University of Pennsylvania
Institute for Research in Cognitive Science
3401 Walnut Street, Suite 400A
Philadelphia PA 19104-6228
[email protected]
OSLa volume 1(1), 2009
Behrens & Fabricius-Hansen (eds.) Structuring information in discourse:
the explicit/implicit dimension, Oslo Studies in Language 1(1), 2009. 191-216. (ISSN 1890-9639)
   
   
University of Amsterdam
The paper introduces an optimality theoretic notion of optimal interpretation based on presupposition theory and shows that it is a viable alternative
for Gricean pragmatics. It moreover is directly applicable to presupposition
and rhetorical structure and improves the insights in those areas. The last
sections are concerned with provisionally making this point.
Gricean pragmatics is the theory of implicatures formulated by Grice (1975) based
on the principle of cooperativity. Though there has been a lot of well-founded
criticism of this conception of pragmatics (Cohen 1971; Sperber and Wilson 1984),
it is still the dominant concept of pragmatics among semanticists and in larger
sections of the linguistic community, where typically, Gricean explanations are
sought when semantical explanations do not seem to work. Interestingly, Gricean
pragmatics1 falls short of one goal it sets itself. It is meant as an account of everything that happens over and above truth-conditional meaning and a clear example of something that goes beyond truth-conditions is presupposition. In (Grice
1981), Grice discusses presupposition but ends up with using a syntactic device
to deal with the more problematic cases. Apart from the correctness of what
Grice proposes there (there are a number of predictions there that few recent
authors on presupposition would adhere to) the syntactic device itself seems unmotivated. So I claim that Grice in fact does not manage to reduce presupposition
to the cooperativity principle.
This lacuna has been inherited by all the successors to Grice – to the best of my knowledge. There is no
Neo-Gricean or Relevance Theoretic account of presupposition and accounts of presupposition normally
only interact with other aspects of pragmatics.
 
This paper starts from the common core of the theories of Van der Sandt (1992)
and Heim (1983) on presupposition2 and develops it to be a theory of interpretation in the style of Optimality Theory (Prince and Smolensky 1993). The theory
describes what is the best interpretation of a linguistic utterance in a context by
maximising a number of properties. The rest of the paper shows that it can do everything that Gricean pragmatics can do but also presupposition and rhetorical
This is important because it brings unity to pragmatics, a field that has been
described as the waste-basket of semantics. Grice was aiming for a structured
general pragmatics based on the cooperativity principle. The alternative system
proposed here has the same aim and can be seen as a general theory of explanation
for natural events specialised to communicative events. As such, the pragmatic
system would pre-date human language and can be plausibly argued to be the
motor of language change but also a driving force behind the genesis of human
language from its precursors. The case for the system is however primarily its
application to the descriptive problems that pragmatics has to deal with.
Without a unified approach, pragmatics consists of different areas: anaphora,
deixis, presupposition, rhetorical structure, implicature, speech acts, explicature
and others and one is tempted to invent principles governing each of those areas. In a unified theory, this freedom is considerably reduced. This means that
observations in one area start having a bearing on what happens elsewhere and
that explanations must be made uniform. The claim is that this leads to important
progress in presupposition and rhetorical structure and to a better understanding
of implicature and anaphora. The paper is structured as follows. The system of
constraints is derived from the explicit and implicit assumptions behind the presupposition accounts of Van der Sandt and Heim. It is then related to explanation
and abduction (Hobbs et al. 1990). The application to the four areas is briefly discussed in the last sections.
     
Since Karttunen (1973) the aim of presupposition theory has been to deal with
the projection problem and influential and lasting contributions have been made
by Karttunen himself, Gazdar (1979), Soames (1982) and others. Though there is
definitely still controversy and quite a lot of problems, the formulations of Heim
and Van der Sandt from the 80’s have not been superseded by better and more
comprehensive theories3 . The projection problem is the problem of predicting
OSLa volume 1(1), 2009
It builds on the unpublished (Blutner and Jäger 1999) which reconstructs Van der Sandt’s account in
optimality theory.
E.g. the recent attempts of Schlenker and his students at providing different foundations for presupposition theory all seem to be aimed at reconstructing Heim’s predictions, even in the cases where these
are problematic.
 
which presuppositions emerge as implicatures or entailments of the whole sentence in complex sentences that contain presupposition triggers. The formulation pre-dates the dynamic views of semantics from the 80’s and uses notions like
implicature or entailment that come from the logical tradition. Within dynamic
semantics, one could restate it as follows: If a sentence with a presupposition trigger is used in a context, under what conditions on the context and the sentence
will the presupposition be true in the context that incorporates the interpretation of the sentence. In this formulation, the presupposition can be in the new
context because it was already in the old context or because it is added by the
interpretation of the utterance. A better alternative, which is needed to incorporate local accommodation and intermediate accommodation in the style of Heim
and Van der Sandt would be:
If a sentence with a presupposition trigger is used in a context, under what
conditions on the context and the sentence will the presupposition be true in
which subcontexts of the context that contains the interpretation of the sentence.
The typical facts that need to be explained can be illustrated by the following
John thinks that if Mary left, Harry will be glad that she left.
John thinks that if Mary was angry, Harry will be glad that she left.
The second but not the first example will force the new context to contain the
information that Mary left. In Van der Sandt’s theory or in Heim’s theory, this is
because the trigger be glad has access to Mary left in the condition of the embedded
conditional in the first example (resolution), but not in the second. The second
sentence can thereby be used only in contexts that either have the information
that Mary left or where that information can be added to the context without
making it inconsistent (accommodation). For a fuller discussion of Heim’s and
Van der Sandt’s theory see Beaver (1997) and Geurts (1999). The following is an
abstract version of these theories.
(i) Presupposition triggers have a presupposition p that must hold at the site
of the trigger
(ii) p should be resolved to an accessible part of the context of the trigger
(iii) If this is not possible p should be accommodated (added to some context of
the trigger)
(iv) p should preferably be added to the outermost context of the trigger if it is
consistent there.
Prinicples (i) and (ii) are in essence due to Karttunen. (iii) and (iv) are more problematic. The problem with (iii) is that not all triggers accommodate and with
OSLa volume 1(1), 2009
 
(iv) that accommodations do not always go to the outermost context. (iii) is not
hard to motivate: it follows from the fact that presuppositions are not superfluous (they can be conditions on definedness of concepts or mark properties of the
contexts in which they are used) and so are related to the acceptability of the
utterance. The surrounding contexts are the ones that are relevant for resolution. So any accommodation that meets (iii) restores the situation that would be
needed for resolution. But (iv) is not easy to motivate and the lack of motivation is an embarassment to the theory. Van der Sandt (but not Heim) advocated
a stronger version (iv’) of (iv).
(iv’) p should preferably be added to the outermost context of the trigger in
which it is consistent.
This version has been shown by Beaver to make some wrong predictions, especially in cases like (2).
Most German housewives wash their Porsche on Sundays.
Principle (iv’) predicts that this should mean (3) (which might well be true).
Most German housewives who have a Porsche wash it on Sundays.
But nearly everybody understands it as meaning (4) which is clearly not true4
(local accommodation).
Most German housewives have a Porsche and wash it on Sundays.
The four principles can be generalised to four general constraints. What principle
(i) says is that the local5 truth of the presupposition for the speaker is required for
the use of the presupposition trigger: if the local truth fails, the presupposition
trigger cannot be used. From the hearer perspective, it means that her reconstruction of what the speaker is intending to say must take the presupposition
trigger into account and the local truth must be checked or created within her
interpretation of what the speaker is saying. Why? Because otherwise the interpretation would not be such that the hearer could have used the same words to
achieve the same intention, putting herself in the speaker’s position. This last formulation can be underpinned by Grice’s theory of meaningNN (Grice 1957) which
requires the recognition of the intention of the speaker by the hearer to be intended by the speaker as part of her plan to realise that intention. Recognition
OSLa volume 1(1), 2009
In fact, most people read it as implicating that all German housewives have Porsches, a reading not
predicted by Van der Sandt.
This formulation ignores a complication with referring expressions and particles, where the presupposition can fail to be true in the local context, provided it is true in an accessible context. This complication
can be ignored, as discussed later, because such triggers do not accommodate.
 
of the intention by the hearer would involve at least the truth of the statement
that the hearer can imagine that had she been the speaker, she could have used
the same words. And this is the general constraint on interpretations that we will
1 6 : the hearer could have used the utterance herself to express the interpretation – if she would have been the speaker.
The last proviso is for dealing with perspectival shifts, for limitations on the speaker’s command of the language, for errors etc.
For presupposition triggers,  captures the generalisation that triggers
can be used only where the local truth of their presupposition is assumed. (This
could be taken as a definition of what it means to be presupposition trigger.) But
presupposition is only a special case: the  constraint is about all restrictions on how to express oneself in a language which includes all lexical, syntactic,
semantic, morphological and phonological rules and facts. The lexical, phonological and semantic facts that some words, intonation patterns and constructions
are presupposition triggers is the fact needed for presupposition.  can be
articulated in many ways. One could look at the language and speech generation
systems and the principles on which these are based. Or one can look at optimality theoretic syntax and phonology. But in principle any theory that defines the
relation between linguistic form and interpretation can be used, including the recent probabilistic approaches. In this paper, OT syntax will be assumed and – at
one point – max -constraints in OT syntax7 . Since expressive constraints should be
part of any theory of grammar, this again does not rule out other models. 
is not an absolute constraint. It allows for imperfect use of language or speech
in which case the hearer has to correct for the imperfections. This aspect will be
ignored in the rest of the paper.
One aspect of principle (iv) – see above – is a well-known principle of interpretation: the fact that internally inconsistent interpretations, and interpretations
that clash with the context give way to consistent interpretations. In probabilistic
approaches to language one implicitly makes the further assumption that interpretations that are less plausible than others given the context also give way to
more plausible ones. This can be formulated simply as:
2 : maximise plausibility (an interpretation is bad if there is a more
plausible interpretation)
In OT, faithfulness constraints are the constraints on the relation between the input and the output.
This allows a precise formulation in terms of OT syntax and phonology. Suppose a complete OT system
is given deciding which candidate pronunciation is optimal for a certain semantical input in a context.
An interpretation meets  if there is no alternative interpretation for which the utterance would be
better in the context from which the speaker made the utterance. The notion of “better” can be spelt
out in terms of the given constraint system.
OSLa volume 1(1), 2009
 
 applied without restriction would make any interpretation maximally
plausible, i.e. trivial.  is a necessary restriction: only more plausible interpretations for which the hearer, stepping in the speaker’s shoes, could have used
the same expression will come in place of less plausible ones. So  does
not rule out surprising interpretations or even corrections (implausible in any
context, because they are inconsistent with it). The rules that govern language
production allow people to say interesting things, because it is not possible to
ignore them in interpretation. Without them, language users are condemned to
exchanging unsurprising messages. For presupposition accommodation, 
 just rules out accommodation sites if more plausible ones are available.
 is the central principle for disambiguation. The rest of (iv) will be
dealt with later on. Principles (ii) and (iii) are repeated below:
(ii) p should be resolved to an accessible part of the context of the trigger.
(iii) If this is not possible p should be accommodated (added to some context of
the trigger).
These principles together prefer resolution over accommodation and this can
again be seen as the outcome of a more general constraint, that of preferring
not to make unnecessary extra assumptions in interpretations, especially the assumption of new discourse referents of all kinds8 . A similar constraint has been
adopted in another optimality theoretic treatment, Hendriks and de Hoop (2001):
    .
3 *: old referents are preferred over connected referents which are preferred over new referents.
Again, without limitations this principle condemns the hearer to interpret anything as old news. But it is meant to apply after  and  have ruled
out other interpretations.
It is below , because new interpretations can be more plausible than
ones in which the information expressed is identified with old information and
where referents are identified or connected with old discourse referents. *
is not the reason why presupposition triggers or pronouns are resolved to given
information or given identity. The assumption is that triggers – as lexical, phonological or syntactic items – mark their referents and presuppositions as given (or
more accurately, they mark for properties that imply givenness). So  is responsible for the obligation to resolve. * is responsible for optional anaphora
OSLa volume 1(1), 2009
Discourse referents go back to Karttunen (1976) and are incorporated into all schemes of dynamic interpretation. Nothing in this paper conflicts with the discourse representation theory of Kamp and Reyle
(1993). See also the appendix.
 
only. For example, the Slavic bare singular lexical NP can be both definite and
indefinite. It is therefore not a presupposition trigger even though it has all the
anaphoric and bridging possibilities of a definite NP.
For presupposition, resolution of the presupposed material to existing material accessible in a context of the trigger means that the demands of  are
satisfied. There is then no need for accommodation and accommodation will be
blocked by *.
* is a powerful principle. It prefers resolving interpretations of Russian
bare NPs over bridging ones and bridging ones over indefinite ones. And it gives
the same preferences for English definite NPs and prefers bridging indefinites
over unconnected ones. Moreover, as will be discussed later, in combination with
, it is also responsible for the defaults in rhetorical discourse structure.
(iv) p should preferably be added to the outermost context of the trigger if it is
consistent there
This preference will be a consequence of .
4 : let the interpretation decide any of the activated questions it
seems to address.
In this formulation of , “the activated questions” should be read
as presupposing that there are such questions. In particular, there must be one
activated question that captures the point of the utterance9 . Questions can be
activated by conversation participants raising them explicitly, by being evoked
by previous contributions and other features of the context and finally by being
evoked by the utterance itself.
Unrestricted  would not work:  comes after the other
constraints. The reason is simple: any arbitrary way of settling all activated questions would always be the most relevant interpretation. But such an interpretation is random and would not be what the speaker intended. Therefore it should
be below . But also below *, because otherwise global accommodation
would be allowed when local resolution is possible. It should be below 
, because what is implausible cannot be implicated. Similar principles have
been defended before by Grice (1975), by Sperber and Wilson (1984), by Van Rooy
(2003) and others. It can be derived from cooperativity: activated questions are
goals of the conversation, whether they are already given or activated by the utterance itself. Cooperativity can also be used to argue that the utterance itself
addresses an activated question or one activated by itself.
Without this assumption, it would not be possible to recapture the insight of Stalnaker (1979) that assertions cannot just bring information that is already in the common ground. The question answered by
the clause as such can be identified with its clausal topic.
OSLa volume 1(1), 2009
 
For presupposition,  works as follows. Assume that a trigger presupposes p and that p cannot be resolved. This activates the question “whether
p”. The speaker indicates by her use of the trigger that she is assuming a positive
answer and thereby seems to address the question, unless the context already has
conflicting information. The interpretation that answers “whether p” positively
adds the information that p to the outermost context.
The constraints are meant to apply in the order in which they were given.
 can force implausible interpretations and that possibility is necessary for
telling interesting things and for carrying out corrections.  also may force
the introduction of new discourse referents and prevent the settling of activated
question.  is a filter on how items can be resolved and on whether they
can be resolved, thus sometimes overriding *.  must be below *
because global accommodation can be more relevant than local resolution (which
appears to be always preferred). The constraint system of these four constraints
in the given order can be understood as a ordered optimality theoretic system
of contraints, an OT pragmatics. The constraints allow violation and the ordering
fully decides the conflicts that can arise. But there are also substantial differences
with optimality theoretic systems as employed in phonology and syntax.
First of all, there is no indication that any kind of variation is possible with
respect to the ordering. Second, it is not clear that finite scoring of errors is the
most appropriate. A candidate interpretation loses if another interpretation is
better with respect to the constraint where better may be interpreted by finite
as well as continuous scoring techniques. Both of these features suggest that this
OT pragmatics should not be part of Optimality Theoretic linguistics (on a narrow
view, OT linguistics is just the account of ) but is rather just another optimisation problem, like robot ethics, choosing a party dress, making an investment
decision in a situation of partial information and others that have been noted in
the literature. The special problem solved by pragmatics is explaining linguistic behaviour as a special case of explaining human behaviour and of explaining
natural events.
,    
Grice (1957) starts by giving some examples of natural meaning (5-b).
These spots mean that the child has the measles.
Smoke means fire.
These are statements in which some natural phenomena – the spots and the smoke
– are explained. The statements explain them by giving their cause: the measles
and the fire. A natural explanation is just supplying another natural phenomenon
(temporally and spatially connected to the explanandum) for which there is a
OSLa volume 1(1), 2009
 
causal principle that can derive the explanandum from the explanation. The correctness question about the explanation is whether the explanation really caused
the explanandum. It is not clear that one can ever be entirely sure. To account
for the powers of explanation of natural phenomena that one finds in human and
animals, it seems useful to think of explanation in terms of optimisation. The
quality of explanation can then be thought of as something that biological and
cultural evolution maximised by providing an ever sounder grasp on the component factors. The quality of explanation is clearly a factor in survival and sexual
selection as well as in social success. So there is evolutionary pressure for better explanation abilties both in biological and in cultural evolution. The overall
concept that is optimised can be articulated as a constraint system much like the
interpretation system of the last section.
1 The explanation should be a possible cause.
Presumably, earlier successful explanations are remembered which leads to associative links that provide a range of possible causes. This is analogous to :
only interpretations that can be possible causes of the utterances are considered.
The difference is that an utterance is behaviour that the interpreter could also
carry out herself to realise the communicative intention that is the cause. This
provides a strong test on the causal link. In the explanation of natural phenomena
this must be filled by representations of causal knowledge.
2 If there is more than one possible cause one should compare them for probability in the context. The most likely should be chosen: this maximises the
chance of being right.
3 If there is more than one most probable cause, they can be inspected for
newness: is part of the cause already known? What part is completely new?
What new states of affairs do we need to postulate? Minimising the new
hypotheses is best.
In the philosophy of science, [3] is Ockham’s razor. [1] and [2] can be related
to techniques like Bayesian nets in AI.
If the analogy is correct, the pragmatic system is just the ability to form successful explanations of utterances. The only new element is  which
does not have an analogue in the explanation of natural phenomena and would
indeed be expressive of cooperation and communication. If questions have been
raised or are being raised and information is offered that pertains to them, the
hearer has the right to expect that the interlocutor is cooperative and that the
information given answers those questions. Language use is voluntary behaviour
and a natural explanation for why the behaviour is wanted and produced is given
by precisely the activated questions.
OSLa volume 1(1), 2009
 
The analogy with the philosophy of science can then be carried further. Philology would go beyond natural science in asking for understanding of the texts and
artefacts that it studies, a reconstruction of the intentions behind those objects.
The difference with natural science stems from the fact that the aim of a communicative event is communication and that the human interpreter can – as a
potential producer – try to reconstruct the intentions and goals behind the communicative act. And the same situation holds for the objects studied by philology:
they have a purpose and as a fellow human being, one can try to reconstruct the
purposes that led to their creation.
There is continuity in this account between interpretations of natural phenomena, behaviour, involuntary vocalisations, non-linguistic communication and
linguistic communication. It is the same system relying on the same kind of resources (connections between causes and effects in experience) and producing explanations of experiences. Where the explanation of behaviour starts to include
the intentions of the person who carried out the behaviour, these intentions may
themselves be explained by the currently activated questions that are common
ground between speaker and hearer.
As an example, consider the following situation. John stands by the road waving his jacket at me. I should be asking myself first when I would be standing by
the road waving my jacket at me. This is : it requires me to consider possible
explanations. The possible answers should be weighed by plausibility and better
answers should be selected in preference over less plausible ones (). I
should then be wondering about the new elements in my explanation, can they
be eliminated, can I connect them to already known things (*). And if there
are activated questions can John be settling them by his waving ()? If
one supposes that we were looking for a lost cow, the proper explanation may be
that John has seen it and that he is indicating with his waving where it is. In this
case, the pragmatic system makes a by itself not very meaningful gesture into a
statement: I have found the cow! Here it is!
The theory cannot be a general pragmatics unless it is able to recapture the effects
of Gricean pragmatics, in particular conversational implicatures. This section discusses a number of examples only and tries to make it clear how the implicatures
should follow. Grice explains particularised conversational implicatures as deliberate floutings of his maxims: the speaker is overtly uncooperative. It is then the
task of the hearer to work out reasons for why the speaker does not follow the
maxim she is flouting.
The following extra assumption is necessary for these implicatures. It belongs
to linguistic competence that special expression strategies can be followed: sometimes things can be expressed better by a joke, irony, understatement, hyperbole,
OSLa volume 1(1), 2009
 
metaphor and marked forms (forms whose meaning is normally expressed in another way). It can be more effective to do so, more amusing or more polite.
The speaker and a hearer share knowledge of these strategies and it would
be part of production optimality theory to say more about when certain intentions prefer expression by these special strategies and the rules that define the
strategies. This area is largely unexplored, but nothing indicates that an account
is impossible. An account of this kind is necessary to make these cases pass .
Here, it is only attempted to show that the hearer can infer the intention behind
the utterances in question.
For a group of implicatures, a conflict arises with . Irony, metaphor,
understatement and hyperbole have literal interpretations that cannot be true,
because they conflict with knowledge about the speaker, general knowledge or
conceptual knowledge. Assuming that the speaker is being ironical, understating,
metaphoric or hyperbolic leads to a different hypothesis about what the speaker
is trying to express and drastically improves plausibility. Saying that you heard
a thousand feet on the corridor relies on the hearer recognising that this cannot be true and on the hearer realising that this is only a way of expressing the
impression the event made on the speaker. Saying that John is a cunning fox similarly conflicts with plausibility and the interpretation relies on figuring out how
it could still be saying something that is true of John. In a marked expression
 does not map the straightforward interpretation to the chosen form as in
Mrs T. produced a series of sounds closely resembling the score of “Home
Sweet Home”
The literal interpretation would be mapped to: Mrs. T sang “Home Sweet Home”.
But with the extra innuendo that there was something wrong with the singing,
the hearer can imagine producing the form herself (if she were as clever as Grice
or the reviewer he is quoting).
In another example (7) of Grice, the speaker addresses a different topic from
the one given: the qualifications of John as a lecturer.
John is an excellent cyclist.
The statement is on topic again (and thus meets * and  better) if the
refusal to address the given topic is recognised including the reasons that might
have led the speaker to do so.
 gives rise to generalised conversational implicatures: exhaustivity
implicatures including scalar implicatures.
John has 3 sheep and five cows.
OSLa volume 1(1), 2009
 
Questions it seems to address:
What livestock does John have? How many sheep?
How many cows?
John does not have other lifestock except sheep and cows.
John does not have n sheep with n > 3
John does not have m cows with m > 2.
These implicatures do not arise in situations where the questions are not activated.
 would be responsible for the general assumption of normality (a
normal car, a normal killing, a normal drink) since it prefers the most probable
interpretations. For a case like: drink (normally: alcoholic drink) this would be
a question of figuring out what is typically meant with drink, something that is
correlated with the frequency with which drink is used for alcoholic drinks. The
context (e.g. the person who is supposed to drink it is 6 years old) or modifiers
(soft) can influence this decision.
For clausal implicatures, expressive constraints in syntax need to be assumed.
i.e. they are in the province of . For clausal implicatures, the following two
are needed: max () and max () which force the formal expression of
the speaker’s adherence to the proposition expressed or the formal expression
of the doubts that the speaker has with respect to the clause expressed. The expressive constraints can only be satisfied if the morphology or lexical material is
available in the language for the expression of adherence or doubt. This is the case
for English indicative conditionals that stand in opposition with both the causal
and the counterfactual constructions.
If John came here, he would have seen the problem.
Because John came here, he has seen the problem.
If John had come here, he would have seen the problem.
This means that the speaker cannot take the condition or the consequence of an
indicative conditional to be true or subject to doubt without being forced into
an alternative construction. And thereby as implicating that both condition and
consequence are open issues, if the indicative conditional is chosen. The clausal
implicatures of disjuncts of a disjunction follow in the same way from the competition of just saying one of the disjuncts in isolation. Other non-entailed clauses
(e.g. complements of belief attributions) do not give rise to the clausal implicatures because expression alternatives are not available.
OSLa volume 1(1), 2009
 
 ,    
The constraints except  (which merely determines the space of possible interpretations) promote certain interpretations over others.  militates in
favour of probable interpretations. * tries to keep referents old. 
prefers settling questions over leaving them open. What to do if the speaker wants
to express an implausible reading, or draw the attention to new objects, or does
not want to settle an activated question or a question that is activated by his utterance? Such a case automatically puts the speaker under an expressive obligation,
since doing nothing means that the speaker will be misunderstood. If the misunderstanding is a question of the value of a semantic feature, it is likely that there
is an expressive constraint to express the feature. And there are often grammaticalized devices for expressing such semantic features. The following are some
markers that can associated with the task of putting the constraints out of action.
• : mirative markers (even, only, just, yet, already), correction markers (no, but, however), intonational contrastive marking. Even lexical words
can be thought of as devices that mark against .
• *: indefinite marking, presentational constructions, contrastive markers and contrastive intonation, tense change (tense morphemes, now, then),
mood change, additive marking (too, also, and)
• : sentence final rising intonation, “wave” intonation, words like
“some” and “certain” expressing uncertainty, downtoners, mood.
This is as predicted. Language evolution has created markers to act against the
tendencies promoted by pragmatics.
The real problem is that one can also find many grammaticalised markers for
tendencies that are being enforced by the constraints, in particular old markers,
while old interpretations are already enforced by *.
The following are such “old markers”: particles like indeed, immers (Dutch)/ja
(German), deaccented toch/doch (Dutch/German), finiteness, indexicals, pronouns,
demonstratives, (short) definite descriptions.
The problem can be solved. If  were not restricted by , i.e. by
absolute rules, it would make it nearly impossible to say anything interesting. E.g.
I could not correct your opinion by saying (10).
He is  in Spain.
It would be much better for you to interpret he as referring to somebody else than
the person you have your opinion about: some person who – also according to you
– is not in Spain. And if no such person presents himself, by inventing a person
OSLa volume 1(1), 2009
 
like that. This would violate *, but would also make the interpretation more
plausible. Consider another example (11).
John believes that that is precious.
(pointing at a plastic coffeemug or without a pointing)
Given that John is reasonable, any other referent for that for which it is more
plausible that it is precious is preferable and such objects can be created by accommodation. But accommodations or resolutions to less activated antecedents
are excluded for pronouns and demonstratives and that is exactly what allows
corrections and the attribution of implausible opinions using these devices. The
typical old markers are marking for certain properties: he marks its referent as
a highly activated male person, that marks its referent as an object in the utterance situation that is indicated by the speaker. The hearer cannot reconstruct
the marked feature of these items unless the referents are really highly activated
or really in the utterance situation and indicated. If they are not, the interpreter
has to give up or enter into real reparation strategies. So the conclusion must be
that old marking of this kind is lexically coded and belongs to  in order to
override .
It is not an accident that the DRT development algorithm (Kamp and Reyle
1993) includes absolute pronoun resolution: all pronouns must be resolved. This
indicates that proper anaphora belongs to : otherwise, there would be exceptions due to . The property that licenses their use is a feature of the
context as such and they are typically ways of satisfying expressive constraints.
Expressive constraints would also be responsible for preference based on the
referential hierarchy assumed in Gundel et al. (1993) or on the similar hierarchies
in natural language generation (Reiter and Dale 2000). The following hierarchy
could be part of a generation system for Dutch.
first and second person pronouns > reflexives > 3rd person pronoun >
demonstrative pronouns > anaphoric and bridging definites and short
names > full demonstrative NPs > full descriptions and full names > indefinites
The higher type of NP is preferred if the condition for its use is met. This gives
a corresponding hierarchy of max () constraints where F varies over conversational participant, c-commanded, high activation, indicated, activation10 visible
and uniqueness.
OSLa volume 1(1), 2009
This would be the common denominator of bridging and anaphoric uses of short definites. The high
activation of the antecedent for bridging spreads to the referent of the bridging NP, earlier mention
activated the antecedent of the anaphoric definite NP.
 
Most of these marking items can be thought of as presupposition triggers
and the features they express as the associated presupposition, but they lack
many of the properties of normal triggers like bindability and accommodation
(see Beaver and Zeevat (2006)) for more discussion).
  
Proper presupposition triggers can be distinguished from marking devices and
particles because there is no associated expressive constraint. This restricts this
class to lexical triggers, i.e. to predicates that can be applied to an arrangement
of objects only if that arrangement already meets some criterion, the presupposition. This leaves us with factives, quantifiers, start and finish and a class of nouns
and verbs (bachelor, buy)11 . Lexical presupposition seems important for disambiguation. Derived uses of words like strong or run are safe because conflicts between presuppositions force the assumption of different readings. In (13), one
cannot assume a meaning with much muscular power or jumping locomotion, because
that presupposes animacy (of the subject).
strong (of drinks): with much alcohol
run (of engines): functioning
The emergence of lexical presupposition cannot be just due to the needs of disambigation however. The simplest way to think of how lexical presupposition arises
is by general considerations about classification. Once one has a class of objects
or situations, the members of that class can be classified further. Membership of
the class is a lexical presupposition of the words that are used to divide the class
in subclasses.
It is a consequence of being a lexical presupposition that the speaker must
take the presupposition to hold in the local context of its use (see Zeevat (1992)
for a fuller discussion). If it is not common ground that the presupposition holds
in the local context there are open and activated questions “whether p?” for any
context C of the trigger (searching for p in a context C is conceptually hard to
distinguish from raising the question “whether p” in that context). To be more
precise, positive contexts determine an operator X from which the natural question “whether X p? ” can be formed. E.g. the context representing what is attributed as a propositional attitude to John corresponds with the operator “John
believes that” and the question “whether John believes that p?”. The context C
corresponding to the consequence of a conditional to the operator “if the condition is true” and the question “whether if the condition is true, then p?”. The
This leaves a number of triggers in an unclear position. E.g. clefts and pseudo-clefts do not clearly have
the properties of lexical triggers, yet also seem to lack expressive constraints. I will defer discussion to
another occasion.
OSLa volume 1(1), 2009
 
operator corresponding to the scope of a proper quantifier to the operator: “if an
arbitrary x satisfies the restrictor” and the question to: “whether p(x) if x satisfies the restrictor”. Non-positive contexts (monotone decreasing contexts) do
not have associated operators.
For any of these questions, the speaker has suggested that they have a positive
answer by using the trigger in them. By  the speaker should be taken
as settling these questions positively unless this leads to violations of .
Zeevat (ms) applies a formal version of these ideas to a range of outstanding
problems of presupposition projection. Here I just mention three problems that
seem to go away at once. First, Heim has always insisted on a proper rational
motivation for the preference for global accommodation, while admitting that
there was a strong empirical case for assuming it. The  constraint is
such a motivation, since it can be rationally motivated and it gives a preference
for global accommodation. The question with respect to the outermost context is
always raised – if resolution is impossible – and positively settled unless there is
a inconsistency with the outermost context.
Second, an old puzzle first noted by Karttunen (1973) is illustrated in (14).
Mary hopes her admirer will visit her tonight.
Karttunen notes that (by projection tests) two presuppositions are projected by
“her admirer”, the ones in (15).
Mary has an admirer.
Mary believes she has an admirer.
The present account immediately predicts this (the question associated with the
context of Mary’s hopes is: Does Mary believe that she has an admirer) while the
utterance has a marked reading with only local accommodation. Heim’s and Vander Sandt’s account have no explanation. Third, the present account gives the
correct explanation of partial accommodation where an incomplete antecedent
of the presupposition is enriched to a fully matching one as in (16).
John was ill last week and now he is glad he took care of his fever.
* will prefer adding fewer new discourse referents over adding more and will
so prefer partial resolution over full accommodation and will thereby implicate
that John was ill and had a fever last week. The main objection to an account of
this kind – accommodation in non-positive contexts – is discussed in Zeevat (ms).
 
It came as a surprise that presupposition theory would be relevant to rhetorical
structure. To the degree that the following points are convincing, they form the
OSLa volume 1(1), 2009
 
strongest argument for the view on pragmatics presented in this paper: the application to implicature brings nothing new (except the programme of giving a
generative account of non-literal use of language), the application to presupposition is important, but since presupposition was the stating point of the account,
it is unsurprising that it should be applicable.
The application to rhetorical structure however came for free and a proper
foundation in pragmatics was much needed. The current account does not lead
far away from the extant accounts such as SDRT, Abduction12 , Discourse Grammar and Rhetorical Structure Theory. From the perspective of the many views on
rhetorical structure that there are, it is surprising that there is a purely pragmatic
foundation for a number of central phenomena. The phenomena themselves are
well-known and have been described by additional axioms or mechanisms.
Defaults on rhetorical relations and the Right Frontier Constraint can be underpinned by *
There are two versions of developing this idea. The first starts from the assumption that topics are just another discourse referent, which is then subject to *.
This version will be adopted here13 . The topic should be old, otherwise a part of
the given topic and otherwise bridge. Being totally new is the worst case. There
should also be the assumption that there is a preference for higher activation
when selecting the old topic, the topic of which the current topic is part or to
which it bridges. All of this should be part of *.
Now adopt the following two common assumptions: (a) Lists, Contrastive Pairs
and Narrations have a discourse topic that is different from the topics of the
clauses that make them up, and (b) clauses have their own topic. Topics should
here be taken as the characterisation of the point of the clause in the larger discourse, in terms of a question. E.g. the second clauses in (17-h) would be dealing
with the indicated topics.
John fell. Bill pushed him.
Why did John fall? (Explanation)
John fell. I saw it.
How do you know that John fell? (Justification)
This is perhaps the closest relative to the account in this paper. Abduction is explanation and it is Hobbs’
view that pragmatics is explanation. From the perspective of this paper, abduction is a combination of
 and , but lacks * and . This section presents the case for the importance
of these principles. Abduction could be a promising way of implementing the proposal in this paper, if
one finds the right way of integrating the two additional principles.
The other starts from the main temporal discourse referent of the causes and applies * to those. The
version with topics needs an argument to show that topics are discourse referents in order to motivate
the application of *. The version with temporalities is slightly less straigthforward and used in Zeevat
OSLa volume 1(1), 2009
 
John fell. Then Bill pushed him.
What happened when John fell? (Narration)
John fell. So Mary had to continue by herself.
What was the result of John’s falling. (Result)
John fell. He slipped on a banana skin.
What happened? (Restatement)
John fell. He stumbled,
How did John fall? (Elaboration)
John fell. Bill managed to regain his balance.
Did Bill fall? (Contrast)
John fell. Bill fell too.
Did Bill fall? (List)
It is in virtue of those topics that one can assign rhetorical relations to the second
clause, indicated between the brackets after the question.
If the pivot as in the example is the last clause, * gives the following preferences:
pivot topic
part of pivot topic
bridging to pivot topic
part of shared discourse topic
complement of discourse topic
These correspond to:
Explanation, Background and Justification
List and Narration
Contrast and Concession
The last clause always gives the most activated topic. But when it cannot be the
pivot (no plausible identity, part of, bridging, overarching topic or complement
relation applies) the most activated topic takes its place.
That the defaults between discourse relations are in fact as indicated can be
seen by considering ambiguous pairs. C.f (Jasinskaja 2007). Concessions are almost obligatory marked, contrast less, while restatements and elaboration almost
never and lists and narrations have a middle position (see Taboada (2006) for a systematic investigation). Non-transparent markers like “and”, “but” and “though”
occur at the bottom end which points to a much greater need for these markers
and has presumably caused their grammaticalisation. If one finds markers higher
up, they tend to be transparant, i.e. lexical. The exception are some markers for
OSLa volume 1(1), 2009
 
Explanation (e.g. English “for”, Dutch “immers” and “want” and German “ja” and
If one assumes that Lists, Narrations, Contrasts and Concessions deactivate
the topic of their left daughters, while Elaborations, Backgrounds, and Justifications merely lower the activation level of the topic of the pivot, the Right Frontier
Constraint also follows from * and moreover motivates a default for the closest element on the right frontier .
Both assumptions are very reasonable: the top three levels can be seen as the
epistemic and causal grounding of the pivot and the pivot topic should be kept
active while this happens. Moving to the next element in a List, Narration or
Contrast is a sign that the last element (the pivot)’s topic has been dealt with to
the satisfaction of the speaker and can be deactivated.
Optional marking can be seen as an interplay between , * and 
Marking of rhetorical relations (especially in the higher levels of (19) is optional.
The semantic/pragmatic effect of the marked and the unmarked clause seems
the same in a context that prefers the relation marked by the marker. This sort
of situation is difficult to capture in an account of the relation between meaning
and form. In Montague grammar, one would have to construct multiple syntactic analyses (with hidden operators?) to account for the unmarked clause. In the
more modern design of DRT and SDRT, one needs a separate process of discourse
resolution that adds the extra operator after syntactic and semantic processing.
This improves on Montague grammar by avoiding syntactically unmotivated syntactic ambiguities and by solving the problem of selecting the right reading in
the ambiguous situation, but it offers no account of the decision problem for the
speaker: when should the marker be added, when is it superfluous?
The following example, due to Jason Mattausch, illustrates this problem.
John fell.
John fell.
John fell.
John fell.
Bill pushed him.s
Then Bill pushed him.
Mary smiled at him.
Because Mary smiled at him.
The markers “then” and “because” are necessary because if they were omitted,
the normal interpretation would be a different one (Explanation rather than Narration or the other way round). If the marker is inserted when the markerless
version gives the same result, the markers seem redundant and, while not grammatically incorrect, it is definitely not good style to use them in that case.
John fell. Because Bill pushed him.
John fell. Then Mary smiled at him.
OSLa volume 1(1), 2009
 
 should be able to disregard a feature Explanation or Narration in the input,
but also be able to insert “because” or “then” to express it, i.e. the version with
and without the marker should be equally optimal. That is the grammatical side,
but for an account of speaking, the form should be monitored by pragmatics. This
means that the speaker (while planning, speaking or even afterwards) should try
to interpret what she is producing. If the interpretation does not match what she
is trying to say – the input to the selection of the optimal form – she should see if
there is an equally optimal form which would lead to the intended interpretation.
Early detection would lead to an improved form, later detection to self-correction.
In this way, the markerless optimal form can lose out to an other optimal form
that marks for the distinction. The selection of the form should be simultaneous
with the interpretation system selecting the best interpretation of the form in the
Monitoring should be conceived as another optimisation problem, one that
checks a set of ranked features14 of the input for their intended values in the interpretation prefered by the system  > * > . Together
with an economy constraint this gives a notion of pragmatic production optimality:  >  >  which would be able to handle optional marking
and expressive constraints.
This comes down to saying that rhetorical relation is not grammatically marked
(unlike number, definiteness, tense and aspect in English) but still has to be expressed in the sense that wrong preferred interpretations need to be marked
against. The adequate characterisation of the process requires full-scale interpretation and has been one of the arguments for bidirectional optimality theory.
Many proposals for bidirectional OT have been found wanting in Beaver and Lee
(2003). Assuming interpretational monitoring is however enough and there is
good psychological evidence that monitoring indeed takes place.
The reason for optional marking of rhetorical relations lies in the other constraints. In the example, it is  that prefers to let Mary smile after John’s
falling, overriding *. In the case of Bill’s pushing, both  and *
would prefer Explanation.  should be unimportant, since the three
stronger constraints suffice for a decision. The need to mark explains why lexical
items have been recruited for marking rhetorical relations. The defaults generated by * explain why grammaticalisation is stronger at the lower end of the
Finally,  captures the effect that the strongest rhetorical relation
compatible with the utterance in the context is always the one that is prefered by
OSLa volume 1(1), 2009
In case of several features competing for a means of expression the most important feature should win.
The famous case is competition for priority in word order between subjects and topics. If there is a
conflict which would leave the subject unmarked, the subject wins this competition. For more discussion,
see Zeevat (2006).
 
interpreters. This can strengthen a List to Narration and a Narration to Result,
a Background to an Explanation, an Elaboration to a Restatement and a Restatement to a Conclusion. It is natural to assume that reporting a new event activates
the questions like “what then?”, “what resulted?”, “why?”, “so?”, etc. When these
questions are not answered yet and the new statement can be taken as addressing
them, the interpreter must assume that the statement settles them.
This brief section tried to show that the interpretational constraint system
derived from presupposition theory can explain the defaults in discourse interpretation and can in fact refine the picture. It would however be wrong to see it
as a revolution: most of the insights in the area are preserved.
The pragmatic constraint system used in this paper comes from studies of the
interpretational effect of presupposition triggers. If one tries to understand the
particular assumptions in presupposition theory as instances of plausible general
interpretational principles – as is done in this paper – and makes the further assumption that this is all there is in interpretation, one obtains a mono-directional
pragmatic system like Relevance Theory or Hobbs’s abduction framework. There
is a divergence here with attempts to do pragmatics in the other direction as Grice
does or the intriguing intermediate proposals by Horn (1984); Levinson (2000);
Blutner (2000) to develop a bidirectional pragmatics.
Grice’s insight that meaningNN crucially involves the intention that the intention behind the utterance is recognised seems to lead directly to the attempt to
better understand the intentions that speakers have when they produce something with meaningNN : the intention is to be cooperative given the goals of the
conversation. But to stop there misses the point that meaningNN is also quite continuous with natural meaning. It is at least a partial explanation of John’s saying
“I am hungry” that John is hungry. And this partial explanation is enough for
reaching the further effects that John is hoping to achieve with his utterance. Or
for trying the same trick oneself in order to achieve the same effects.
Cooperativity or relevance could well be something that emerges after an earlier stage in which communicative acts are just interpreted as natural events, with
the emergence of cooperativity mainly due to engaging in such acts oneself. This
makes the interpretation system a specialisation of the explanation system for
natural events to communicative acts. The development of complex natural languages is conditioned by this interpretation system and the continuous monitoring that the system makes possible.
Put rather crudely, explanation can be refined to include relevance or cooperativity. But cooperativity cannot be refined to include explanation. It follows
that there is a limit to how much a cooperativity based pragmatics can achieve.
The main case for the system proposed in this paper is however the broadness
OSLa volume 1(1), 2009
 
of its scope without losing applicability ot detailed problems. The aim of this paper was to make this point in a sketchy and programmatic way. It surprised me
that the “general nonsense” of presupposition theory can be used to do Gricean
pragamtics and to explain rhetorical structure. But this is exactly what general
pragmatics should be: insights on politeness marking should be applicable to the
marking of discourse relations, obscure insights in pronoun resolution on conversational implicatures, and so on.
The idea that optimality theoretic pragmatics can be done as a generalisation of
presupposition theory dates from January 2000 and appears in Zeevat (2001). Helen de Hoop and David Beaver managed to throw me in despair however by accurate questions about the relation of pronouns and *. In retrospect, I can only
be grateful to these excellent colleagues. I owe a debt to Masha Averintseva who
asked me to teach a course in Berlin about rhetorical structure, during which I
discovered the relevance of the system for rhetorical structure. A first full version of this paper was presented at the Ustaoset Sprik workshop and I am very
grateful to the organisers of that stimulating experience. I owe important feedback to audiences in Amsterdam, Oslo, Austin and Baltimore. Special thanks go
to Katrin Erk who is responsible for a reformulation of the  constraint.
Many thanks also to Katja Jasinskaja who managed to shake up my thinking about
rhetorical relations in a profound way.
This appendix tries to give a DRT-based approximation to the system of four constraints introduced in the paper. My aim is to show that an approximation to
an implementation is quite feasible, though much more work is needed in three
areas: estimating plausibility that is not consistency, dealing with incorrect utterances and dealing with non-literal speech.
Question evocation for presupposition triggers can be handled, a full treatment of other cases needs further study.
I: an utterance U and an old DRS K0
C: the set of all legal DRSs K. Evaluation of candidate interpretations
K of U in K0 :
: The hearer, imagining to be the speaker, should be able to use U to effect
the change from K0 to K.
Approximation: an adaptation of the DRS development algorithm geared
to give all possible developments including corrections. The development
OSLa volume 1(1), 2009
 
algorithm must obey the following question regime: All questions activated in K0 and all questions evoked by U of which K does not entail that
they have no true answer are in K in one of the following three ways:
as an activated question of K
if K but not K0 entails that the question has a positive answer, the
unification of the positive answer with the exhaustive answer to the
question is in K
if the question is evoked by a presupposition trigger, its positive answer is in K
: There is no more plausible alternative DRS given K0 .
prove the consistency of K. Corrections lose from extensions of K0 . Smaller
corrections win over larger corrections; (approximation) use a sort-based
plausibility estimate over all atomic predications in the new K \ (K0 ∩ K)
and multiply.
Winners are the maximally plausible extensions allowing for some uncertainty.
*: There is no alternative interpretation in which a discourse referent is more
activated or, if not activated, more connected than in the candidate.
Count the number of new DRs. Subtract 2/3 for each new DR that is part of
another DR, 1/3 for each new DR that is bridging to another DR. Candidates
with lowest scores go through.
: All activated questions that may be settled are settled.
There is no other candidate K with fewer activated questions.
Beaver, D. 1997. Presupposition. In J. van Benthem and A. ter Meulen (eds.), The
Handbook of Logic and Language, pages 939–1008, Amsterdam: Elsevier.
Beaver, D. and Lee, H. 2003. Input-output mismatches in OT. In R. Blutner and
H. Zeevat (eds.), Pragmatics and Optimality Theory, Palgrave.
Beaver, D. and Zeevat, H. 2006. Accommodation. In G. Ramchand (ed.), Oxford
Handbook of Linguistic Interfaces, Oxford: Oxford University Press.
OSLa volume 1(1), 2009
 
Blutner, R. and Jäger, G. 1999. Competition and interpretation: The German adverbs of repetition. Presented at the conference “Approaching the Grammar of
Blutner, Reinhard. 2000. Some aspects of optimality in natural language interpretation. Journal of Semantics 17, 189–216.
Cohen, J. 1971. Some remarks on Grice’s views about the logical particles of natural language. In Y. Bar-Hillel (ed.), Pragmatics of Natural Languages, pages 50–68,
Dordrecht: Reidel.
Gazdar, G. 1979. Pragmatics: Implicature, Presupposition and Logical Form. New York:
Academic Press.
Geurts, B. 1999. Presuppositions and Pronouns. Amsterdam: Elsevier.
Grice, P. 1957. Meaning. Philosophical Review 67, 377–388.
Grice, P. 1975. Logic and conversation. In P. Cole and J. Morgan (eds.), Syntax and
Semantics 3: Speech Acts, pages 41–58, New York: Academic Press.
Grice, P. 1981. Presupposition and Implicature. In P. Cole (ed.), Radical Pragmatics,
pages 183–199, New York: Academic Press, reprinted in Paul Grice, 1989, Studies
in the Way of Words Harvard University Press, Cambridge, Massachusetts, pp.
Gundel, J., Hedberg, N. and Zacharski, R. 1993. Cognitive status and the form of
referring expressions in discourse. Language 69(2), 274–307.
Heim, I. 1983. On the projection problem for presuppositions. In M. Barlow,
D. Flickinger and M. Westcoat (eds.), Second Annual West Coast Conference on Formal Linguistics, pages 114–126, Stanford University.
Hendriks, P. and de Hoop, H. 2001. Optimality theoretic semantics. Linguistics and
Philosophy 24, 1–32.
Hobbs, J., Stickel, M., Appelt, D., and Martin, P. 1990. Interpretation as abduction.
Technical Report 499, SRI International, Menlo Park, California.
Horn, L. 1984. Towards a new taxonomy of pragmatic inference: Q-based and
R-based implicature. In Meaning, form, and use in context: Linguistic applications,
pages 11–42, Washington: Georgetown University Press.
Jasinskaja, E. 2007. Pragmatics and Prosody of Implicit Discourse Relations: The Case of
Restatement.. Ph. D.thesis, University of Tübingen.
OSLa volume 1(1), 2009
 
Kamp, H. and Reyle, U. 1993. From Discourse to Logic. Kluwer, Dordrecht.
Karttunen, L. 1973. Presuppositions of Compound Sentences. Linguistic Inquiry 4,
Karttunen, L. 1976. Syntax and Semantics 2: Notes From the Linguistic Underground,
Chapter “Discourse referents”, pages 363–385. New York: Academic Press.
Levinson, Stephen. 2000. Presumptive meaning: The theory of generalized conversational implicature. Cambridge, Mass: MIT Press.
Prince, A. and Smolensky, P. 1993. Optimality Theory: Constraint Interaction in
Generative Grammar. Technical Report, Center for Cognitive Science, Rutgers
Reiter, E. and Dale, R. 2000. Building Natural-Language Generation Systems. Cambridge: Cambridge University Press.
Soames, S. 1982. How presuppositions are inherited: a solution to the projection
problem. Linguistic Inquiry 13, 483–545.
Sperber, D. and Wilson, D. 1984. Relevance: Communication and Cognition. Oxford:
Basil Blackwell.
Stalnaker, R. 1979. Assertion. In P. Cole (ed.), Syntax and Semantics, volume 9, London: Academic Press.
Taboada, M. 2006. Discourse Markers as Signals (or not). Journal of Pragmatics 38,
Van der Sandt, R. 1992. Presupposition projection as anaphora resolution. Journal
of Semantics 9, 333–377.
Van Rooy, R. 2003. Relevance and bidirectional OT. In R. Blutner and H. Zeevat
(eds.), Pragmatics and Optimality Theory, pages 173–210, Palgrave.
Zeevat, H. 1992. Presupposition and accommodation in update semantics. Journal
of Semantics 9, 379–412.
Zeevat, H. 2001. The asymmetry of optimality theoretic syntax and semantics.
Journal of Semantics 17(3), 243–262.
Zeevat, H. 2006. Freezing and marking. Linguistics 44(5), 1097–1111.
Zeevat, H. ms. A full solution to the projection problem. University of Amsterdam.
OSLa volume 1(1), 2009
 
  
Henk Zeevat University of Amsterdam
c/o Philosophy
Nieuwe Doelenstraat 15
NL-1012 CP Amsterdam
The Netherlands
[email protected]
OSLa volume 1(1), 2009
Oslo Studies in Language (OSLa) is a peer-reviewed open access series. The focus of OSLa is mainly on thematically related collections of papers from any area
of linguistics, and edited conference proceedings from internationally oriented
linguistic conferences in Norway.
Ideas for the OSLa series should be directed to the general editors:
[email protected] or [email protected]
In 2008 the OSLa initiative received financial support from the Department of Literature, Area Studies and European Languages and the Department of Linguistics
and Scandinavian Studies – both at the Faculty of Humanities, University of Oslo.