Pre-Proceedings of the SPRIK Conference 2006:
Explicit and Implicit Information in Text
Information Structure across Languages
Oslo, 8-10 June 2006
hosted by the Project SPRIK “SPRåk I Kontrast”
(Languages in Contrast)
Faculty of Humanities, University of Oslo, Norway
Editors:
C. Fabricius-Hansen, B. Behrens, M. F. Krave
Contents
Foreword
1
Code and Inference: The Meaning of Words in Context
Robyn Carston
3
Dependent on Context: eigentlich in adjectival and adverbial use
Regine Eckardt
7
Adreflexive Intensification and the Theory of Focus and Information
Kjell Johan Sæbø
13
A Lexical Perspective on Discourse Structure and Semantics
Bonnie Webber
19
A model for sentence-fragment production
Shinji Ido
25
A cross-linguistic investigation of the licensing and interpretation
of implicit object arguments
Gergely Pethõ & Eva Kardos
29
The discourse function of right-dislocation in Catalan
Laia Mayol
35
Investigating nominal coreference in originals and translations
Kerstin Kunz
41
Cohesion and explicitation in an English-German translation corpus
Silvia Hansen Schirra, Stella Neumann & Erich Steiner
45
Connectors in a cross-linguistic perspective
Henning Nølke
51
Aspectual Influence on Temporal Relations.
A Case Study of the Experiential Guo in Mandarin
Jiun-Shiung Wu
55
Tense switch in Aspect/-Non-Aspect languages
and its implications for information structure
Barbara Schmiedtová, Mary Carroll, Christiane v. Stutterheim & Natasha Sahonenko
59
Making implicit information explicit: Kortmann’s scale of informativeness
and the English V-ing free adjuncts in Catalan translations
Anna Espunya
63
Adverbial ‘doch’ and the notion of contrast
Elena Karagjosova
69
Evidence for a Scalar Analysis of Result in SDRT
from a Study of the French Temporal Connective alors
Myriam Bras, Anne Le Draoulec & Nicholas Asher
75
Contrastive Lexical Pragmatics: A relevance-theoretic approach to lexical
narrowing and broadening in English and Norwegian original texts and translations
Ingrid Lossius Falkum
81
Explaining connections in Akan discourse: the role of discourse markers
Nana Aba Appiah Amfo
85
The metarepresentational use of main clause phenomena in embedded clauses
Thorstein Fretheim
89
Discourse Structure in Optimality Theoretic Pragmatics
Henk Zeevat
95
Entailment, assertion, and textual coherence: a case study
Patrícia Matos Amaral
99
The role of discourse topic and proximity for demonstratives in German and Russian
Maria Averintseva & Manfred Consten
105
Some remarks on minimal sufficient conditions across languages
Richard Zuber
111
Foreword
The bipartite title of the conference Explicit and Implicit Information in Text. Information
Structure across Languages reflects the main objectives of the organizing research project
SPRIK (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
structuring and weighting of information, including so-called discourse structure, on the other
hand. How do presupposition-inducing and coherence-creating devices contribute to
organizing the information to be expressed in the text as a whole? To what extent and in what
ways do language-specific sentence-internal constraints – e.g. word order, syntactic
complexity, lexical options – influence local and more global structuring of information?
When different languages seem to exploit similar means differently and tend to organize
information differently in specific genres/text types, what is dictated by grammar, what is due
to competing structures available in the individual languages and what is steered by stylistic
conventions? At what level of the (written) discourse must translations deviate from originals
in order to maintain equivalence? As we see it, cross-linguistic empirical research and indepth analyses concerning these matters are still very much in need.
Our conference, then, aims to bring together researchers interested in systematizing and
explaining the variety of means exploited across languages to create cohesion and establish
dependencies and discourse hierarchies in text. We are interested in empirically based crosslinguistic studies that contribute to a better understanding of explicit and implicit information
in 'real' discourse, as well as in theoretical research modelling textual cohesion/coherence.
In our Call for Papers, we welcomed submissions from different theoretical frameworks
concerned with but not limited to the following topics:
Triggers for textual updating across languages.
Comparable clause connectives, their distribution and interpretation in different
languages.
• Discourse subordinating and discourse coordinating devices across languages.
• Implicit vs. explicit means of structuring information in text.
• Presuppositional accounts of the implicit.
• The semantic/pragmatic distinction in text interpretation (cross-linguistic research and
translation).
• The interplay between information structure (at sentence level) and discourse structure
and information flow at text level.
•
•
The present volume consists of pre-conference versions of the papers to be presented at the
workshop. We are happy to have received so many interesting contributions and wish to thank
everyone who has submitted a paper. The volume includes abstracts for the talks to be given
by the keynote speakers Robyn Carston, Regine Eckardt, Kjell Johan Sæbø, and Bonnie
Webber. The reader is referred to our home page for further information:
http://www.hf.uio.no/forskningsprosjekter/sprik/english/activities/conf.html.
1
We wish to thank Thorstein Fretheim and Kjell Johan Sæbø for the help in reviewing the
abstracts submitted for the conference.
We would also like to express our gratitude to the Norwegian Research Council for funding
the SPRIK project (2004-2006)1 in general and this conference in particular. Finally, we
would like to thank the Department of Literature, Area Studies and European Languages,
University of Oslo, for supporting the event in practical matters.
May 2006
University of Oslo, Norway
Cathrine Fabricius-Hansen, Bergljot Behrens & Maria Filiouchkina Krave
1
Grant no. 158447/530.
2
SPRIK Conference
Explicit and Implicit Information in Text - Information Structure across Languages
University of Oslo, June 8-10 2006
Code and Inference: The Meaning of Words in Context
Robyn Carston
University College London
[email protected]
According to Relevance Theory (Wilson & Sperber 2004), when we interpret utterances (whether
speech or text), we are employing the following procedure:
‘Taking a path of least processing effort, the interpreter considers possible
interpretations in their order of accessibility, stopping when he has found an
interpretation that satisfies his expectations of relevance.’
This needs some motivating. It is based on the idea that the relevance of any input to a cognitive
system (whether its source is direct perception or verbal communication) is a function of the
cognitive effects it has for the individual interpreter and the amount of effort required to derive
those effects. In the case of verbal communication, interpreters (hearers or readers) are entitled to
expect communicators to present them with an ‘optimally relevant’ input, that is, one which will
yield to them a satisfying level of cognitive effects and require from them no gratuitous
processing effort. It is this presumption that licenses them to follow the procedure above. Some
implications of this view of how linguistic communication works are that (a) the very fact that a
particular interpretation comes readily to the interpreter’s mind lends it an initial degree of
plausibility, (b) the first interpretation that satisfies the interpreter’s expectations of relevance
(that is, has sufficient effects for no undue effort) is the ‘right’ interpretation (the rationally
justified one), and (c) any utterance or text that requires a considerable degree of processing
effort from the interpreter can be expected to provide him with a correspondingly rich meaning
(range of cognitive effects).
It also follows from this view of communication that a speaker/writer should formulate her
utterance/text in such a way that her intended meaning can be grasped with a minimal
expenditure of effort by her audience. This involves choosing a linguistic stimulus that gets the
balance between decoding and pragmatic inference right, which in its turn involves the speaker
taking account of whether relevant information is or is not readily accessible to the hearer or
intended audience. So, to take a very simple case, in some contexts it will be sufficient to utter
(1a) to get across one’s intended meaning, while in another context, it will be necessary to encode
a great deal more, as in (1b), to communicate essentially the same content:
(1)
a.
b.
She felt sorry.
The woman who had caused the misunderstanding felt an unpleasant
sensation of embarrassment and regret.
3
Robyn Carston
Both the referring expression and the predicate in (1b) are much more explicit (require much
more decoding) than in (1a), and in some contexts they would be overly explicit, causing the
hearer unnecessary decoding effort with no commensurate gain in meaning (since the information
was already available).
It is widely agreed that verbal utterances quite generally require some element of pragmatic
inference in addition to decoding of the linguistic stimulus, that is, there is virtually always a gap
between linguistic meaning and speaker meaning. This is very obvious in the case of
implicatures (intended implications of an utterance), which by definition are not part of what is
encoded. However, as pragmatists in recent years have emphasised, it is also true of what we
sometimes (somewhat paradoxically) call the ‘explicit content’ of an utterance, that is, quite
generally, the sentence used does not fully encode the thought or proposition that the speaker
explicitly communicates with it. This is sometimes known as the linguistic underdeterminacy
view and it comes in two versions, the second stronger than the first. First, we are standardly a
lot less explicit than our language code allows us to be, as in the case of (1a) above, whose
explicit content could have been made much more linguistically explicit, as in (1b). Second,
however, it is not merely that we use some sentences as a convenient short-hand, relying on the
hearer/reader to make easily available inferences. Rather, it is usually just not possible to be fully
explicit. Note that while (1b) is more explicit than (1a), it is far from fully so. The strong claim,
then, is that the resources of our codes (languages) do not (cannot) encode our thoughts (a view
which I argue at length in Carston (2002)). In other words, pragmatics is not merely a
convenience but an absolute necessity in communication and probably more fundamental than
language (both evolutionarily and developmentally).
In the context of the linguistic underdeterminacy issue, much attention has been given to the
existence of ‘unarticulated constituents’, that is, inferred elements of explicit utterance which
have no reflex in the linguistic expression used (they are not linguistically motivated by
indicators such as indexicals, variables or implicit arguments). Some examples are given in (2),
where the bracketed element of meaning is both pragmatically provided and pragmatically
motivated:
(2)
a.
b.
c.
It was raining [in London]
Every student [in the semantics class at UCL in 2006] failed the course.
She stabbed him and [as a result] he died.
There is currently considerable debate about these cases and, in particular, about whether there
really is any such process of ‘free’ pragmatic enrichment that can provide entire (linguistically
unmarked) constituents of content. However, even though they are contentious, these examples
clearly fall within the first, weaker, characterisation of linguistic underdeterminacy given above.
That is, the language can easily provide a sentence which articulates the constituent in question.
It’s just not necessary in appropriate contexts and would, in fact, put the hearer to pointless effort.
So these are cases where pragmatics can be seen as providing a convenient effort-saving short-cut
to a concept that could have been linguistically encoded (and would be in contexts where it is less
readily accessible).
More recently, within relevance theory, we have been looking at another kind of free enrichment,
which exemplifies the stronger version of the underdeterminacy view, according to which the
4
Code and Inference: The Meaning of Words in Context
work done by pragmatic inference simply could not be done by choosing another linguistic
expression. In other words, the element of explicitly communicated content is not encoded
anywhere in the language system, though other concepts that are encoded can be used as a
starting point for the pragmatic process that will result in recovery of the concept the speaker
intends. This is known as ‘lexical modulation’ or ad hoc concept construction. Broadly speaking,
what goes on is that, in the process of figuring out the speaker’s intended meaning, lexically
encoded concepts are pragmatically adjusted or fine-tuned. So if a word encodes a concept BLIG,
it may be understood in context as expressing a distinct though related concept BLIG*. The use of
this starred notation indicates that the concept is not to be thought of as simply some other
encoded concept in the language but rather as a component of the thought (meaning) that the
speaker wants to communicate which may not be encodable at all in the language (or would
require a lengthy and, at best approximate, paraphrase). Of course, the lexically encoded concept
plays a crucial role in constraining the ad hoc concept that is inferred, and the two concepts will
generally at least overlap in their denotations.
Here are some examples:
(3)
a.
b.
c.
d.
e.
John is happy.
I’m not well.
His square face didn’t move.
The water is boiling.
Susan is an icicle.
Depending on context, the relevant ‘happy’ concept in (3a) could be any one of a wide number of
quite specific positive states of mind (HAPPY*, HAPPY**, etc), from mild satisfaction, or general
peace of mind, through to states of joyous exhilaration. The same goes for ‘well’ in (3b). The
outcome of the pragmatic adjustment process here is a concept with a narrower denotation than
that of the quite general concept which is lexically encoded. In the case of (3c), there is some
sort of loosening of the concept from true (right-angled) squares to the kind of square-like shape
that a human face might have. While the intended meaning of (3d) could be perfectly literal (the
water has reached actual boiling point), in different contexts, it could communicate any of a range
of broader meanings, including some for which the temperature of the water is not relevant at all,
but rather the concept refers to its whirling, bubbling, vapour-emitting appearance. This would
be a metaphorical use, as almost certainly is ‘icicle’ in (3e). On this approach, understanding
metaphorical uses falls within this general account of pragmatic modulation of word meanings in
context.
A fascinating issue here, of great interest to translators, is how the relative contributions of the
code and pragmatic inference fall out in different languages. The reasonable assumption is that
for any particular conceptual space, e.g. concepts denoting emotional states (states of being
happy, sad, angry, well, tired, etc), languages will differ in the particular concepts they encode
and that, therefore, the pragmatic adjustment processes in a given context (or cotext) will vary
accordingly. For instance, an interpretation which requires a particular kind of narrowing of a
lexical concept in English might involve less or more narrowing, or a different direction of
narrowing, or none at all in another language. I leave this kind of comparative study to those
with the requisite cross-linguistic knowledge (see the talk by Ingrid Lossius Falkum at this
conference).
5
Robyn Carston
There are two extremes in the coordination of meaning among human beings. One concerns the
stabilisation of words in a language, a relatively slow business involving collective coordination
across many individuals over longish periods of time (Sperber & Wilson 1998). The other is the
quick, spontaneous face-to-face coordination between two individuals in an ordinary
communicative exchange, in which relevance-based pragmatic inference plays a key role.
(Closer to this extreme, but requiring more time and effort on both sides, is the coordination that
happens between the author of a text and her readers.) That we have the pragmatic capacity we
do ensures that our communicative potential is not restricted to the meanings encoded in the
languages we speak.
References
Carston, Robyn. 1997. Enrichment and loosening: complementary processes in deriving the
proposition expressed? Linguistische Berichte, 8, 103-127.
Carston, Robyn. 2002. Thoughts and Utterances. Oxford: Blackwell.
Carston, Robyn. 2004. Explicature and semantics. In: Davis, Steven & Gillon, Brendan (eds.)
Semantics: A Reader. Oxford University Press. 817-845.
Levinson, Stephen. 2000: Presumptive Meanings: The Theory of Generalised Conversational
Implicature. Cambridge, Mass.: MIT Press.
Sperber, Dan & Wilson, Deirdre. 1998: The mapping between the mental and the public lexicon.
In: Carruthers, Peter & Boucher, Jill. (eds.), Language and Thought. Cambridge:
Cambridge University Press. 184-200.
Wilson, Deirdre. 2003: Relevance Theory and lexical pragmatics. Italian Journal of Linguistics/
Rivista di Linguistica 15, 273-291. Special Issue on Pragmatics and the Lexicon.
Wilson, Deirdre. & Sperber, Dan. 2004: Relevance theory. In: Horn, Larry & Ward, Gregory
(eds.), Handbook of Pragmatics. Oxford: Blackwell. pp. 607-632.
6
SPRIK Conference
Explicit and Implicit Information in Text - Information Structure across Languages
University of Oslo, June 8-10 2006
Dependent on Context: eigentlich in adjectival and adverbial use1
Regine Eckardt (Göttingen)
(with Angelika Port, Berlin)
[email protected]
1. Data
In this talk, we will inverstigate the German word eigentlich. It can be used in a range of
contexts, with meanings, and implications that are hard to delimit. Interest of linguists was
first raised by 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.
(1)
a. Was willst Du eigentlich hier?
what want you eigentlich here
‘what do you want here after all / at all / anyway / ... thinking about it’
b. Da hast Du eigentlich recht.
there has 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 us start by recapitulating some of
the facts about eigentlich. The sentences in (2) offer some examples for adjectival eigentlich.
Adjectival eigentlich typically, but not exclusively, occurs with an accent. Small caps indicate
accents.
(2)
a. Der EIGENTLICHE Chef ist verreist.
‘the real boss is on a trip’
b. Der EIGENTLICHE Mörder war Smith.
‘the true murderer was Smith.’
c. Das EIGENTLICHE Problem ist seine Faulheit.
‘the essential problem is his lazyness’
The English translations illustrate the possible range of paraphrases where there seems to be
no single possible translation that would match all uses.
The adjective can not be used in any predicative use whatsoever. Hence, examples like in (3)
are all ungrammatical.
(3)
a. *Das Problem war eigentlich.
the problem was eigentlich
b. *Der Mörder wurde eigentlich.
the murderer became eigentlich
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.
1
The paper reflects joint work that substantially draws on Port (2006). I would like to thank
Angelika for many insightful criticisms and her continuous challenges with real data.
7
Regine Eckardt
(4)
a. *Das ist ein / der uneigentliche Garten.
this is a / the un-eigentliche garden
b. un-eigentliches Integral (= mathematics), un-eigentliches Kompositum (linguistics)
One striking fact about eigentlich is that in positive contexts, it can only be used in definite
NPs. Indefinite uses are restricted to negative contexts, like in (6) (few or rarely can not
license indefinite+eigentlich), and quantificational uses like in (7) are not allowed in Standard
High German. At the end of the paper, we will turn to GOOGLE searched data, and it will
turn out that rare positive hits of the kind in (7) offer clear evidence for an interesting dialectal
variation between Standard German and Swiss German.
(5)
a- 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
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 …’
(6)
a. Ein eigentlicher Vertrag wurde nicht abgeschlossen.
b. Niemand hatte einen eigentlichen Lösungsvorschlag.
(7)
a. *Die meisten eigentlichen Stadträte wohnen im Süden.
b. *Einige eigentliche Spieler traten am Samstag an.
In this respect, complex nouns of the form ‘eigentliche(r/s) N’ share semantic properties of
superlatives. There seems to be a notion that the individual that is the eigentliche N is a
unique single or group.
Finally, there is a strong trend to use eigentlich in the sense of true / real / actual / essential
with a stress. Stressed uses will inevitably yield a semantic contribution in one of these
senses. This sets eigentlich apart from other adjectives or adverbs which usually can also be
used in a non-stressed way. Prose discussions in earlier work usually suggest that it is
assumed that the stress on eigentlich has some kind of motivation. Authors state that some
kind of contrast is evoked to earlier utterances without, however, aiming at a concise analysis
of the prosodic facts. These should at least also include rare unstressed adjectival uses like in
(8).
(8)
Der eigentliche CHARME an der Sache ist aber, daß GM die Renovierung auch noch
zahlt.
‘the EIG TRICK about the thing is, however, that GM will even pay for the renovations’
(8), like all other examples so far, is offered without preceding discourse. In that sense, the
presentation of accenting patterns has been incomplete so far. The examples should be read
with the following footnotes:
•
•
•
It is most natural to read sentences with eigentlich in positive definite contexts with an
accent. In this accentuation, speakers are easily able to reconstruct some vague prediscourse that matches the accent pattern (whatever its contribution may be).
In negative contexts, as in (6), eigentlich needs to carry no accent.
In an example like (8), speakers seem also to be able to reconstruct some vague prediscourse if eigentlich remains unaccented (whatever the contribution of the accent
may be).
So far, we have restricted attention to adjectival eigentlich. In the adverbial use, the role of
accent becomes more prominent yet. Here, the presence or absence of the accent makes a
8
Dependent on Context: ‘eigentlich’ in adjectival and adverbial use
clear distinction between uses in a vaguely contentful sense (truly, really, in fact) and the
discourse particle use.
(9)
a. Wie heißen Sie EIGENTLICH? (= ‘what is your real name?’)
b. Wie HEISSEN 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,
we will proceed from clearer to more sophisticated cases. We will start by divising a
semantics for the uses of stressed adjectival eigentlich, including an analysis for the accent
pattern. We will propose an account for unstressed adjectival uses in positive and negative
contexts. Finally, stressed adjectival uses will be analysed in their semantic and pragmatic
dimensions. Against this background, we propose that adverbial unstressed eigentlich just
serves to annotate an utterance with a certain mental attitude which seems to have arosen as a
pale generalization over rhetoric side messages of stressed eigentlich. For this reason, we
would subsume adverbial unstressed eigentlich under the category of discourse particles, the
existence of a homophonous contentful adverb nowithstanding.
2. Semantics and pragmatics of adjectival eigentlich
We propose that eigentlich denotes an operator which maps predicates to predicates (i.e. of
type ((s,(e,t)), (s,(e,t))) ). We will use EIG to refer to this operator. Its logical type matches the
type of adjectives like former or alleged which are likewise restricted to attributive (= adnominal) uses. It is less evident how predicates P and EIG(P) relate to each other. Example
(2.c) refers to das eigentliche Problem (= ‘the EIG problem’) and suggests that other facts in
the given context could constitute ‘minor problems’. It is not clear whether these other facts
would remain ‘problems’ if the core problem were removed. Still, one could conclude that
EIG serves to map a predicate P to the stereotypical, major, important elements in its
extension.
Examples like (2.a) and (2.b) however are of a different nature. 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 Mörder. Comparing the
meaning of Mörder and eigentlicher Mörder, it turns out that the two are co-extensional.
Murderers are true murderers, and one can not more truly kill someone than by being a
murderer (eigentlich or not).
(10)
EIG( MURDERER) = MURDERER
Similar observations hold for Chef and eigentlicher Chef (= boss). Persons may be mistaken
as being the ‘boss’ but reference to EIG( CHEF ) suggests that from that point on, no such
ersatz bosses should be called Chef.
In order to come to a uniform representation of the effects of EIG, we propose that EIG
contrasts with a family of operators that will be glossed as FAKE. FAKE takes predicates P to
extensions which comprise entities that could erroneously be mistaken as P objects.
Importantly, there seems to be no uniform notion of what can count as an instantiation of
FAKE in a given context. Sometimes, FAKE is instantiated as ‘mistaken to be P’ but in other
examples, like (2.c) for instance, FAKE can be instantiated as ‘a minor nonstereotypical
instance of’. Another example discussed in Port (2006) is the following:
(11)
Der eigentliche Garten ist hinter dem Haus
‘the EIG garden is behind the house’
9
Regine Eckardt
Port clearly argues that (11) can be uttered in view of a small patch of lawn in front of the
house. While such patches sometimes already count as ‘garden’, sentence (11) signals that the
speaker will use the word ‘garden’ in its strict sense where a certain amount of square meters,
or certain kinds of plants are necessary for some piece of land to qualify as ‘garden’.
These assumptions also explain the nature of eigentlich’s context dependency. In order to use
eigentliches N, the discourse context has to provide an appropriate notion of FAKE N,
commonly instantiated by referents in earlier discourse. Given that German does not
lexicalize any uniform notion of FAKE, it can be derived that eigentlich in the described use
can not posses an antonym. Stress is licensed by the contextual contrast between FAKE and
EIG; the stress pattern can be analysed as contrastive stress in an alternative semantics.
Uniqueness does not yet fall out of the analysis. Recall that even pluralities of eigentliche N’s
need to be referred to with definites. Presently, uniqueness will be captured by the following
stipulation.
(12)
EIG(P) =λx( P(x) ∧ ¬∃y( P(y) ∧ x≤y )
EIG(P) will have singleton or empty extension. In the former case, competition between
indefinite and definite determiner will lead to definite reference in positive uses. If there are
no elements in EIG(P), however, this can be reported with an indefinite. No quantified uses
are possible on basis of (12).
3. The adverbial use
Let us now turn to adverbial uses of stressed eigentlich. We assume that EIG in such cases
combines with the proposition expressed by the sentence, and requires the hearer to evaluate
it in the domain of ‘relevant, truely important, morally preferable, etc.’ possible worlds. Like
in the adjectival case, the situation hence has to make it clear what kind of opposition between
LIBERAL and EIG cases is the intended one. Let us discuss some examples.
(14)
(15)
EIGENTLICH heiße ich ‘Luiselotte
EIG be-called I ‘Luiselotte
‘Technically speaking / in truth / in fact, my name is ‘Luiselotte.’
EIGENTLICH habe ich keine Zeit.
EIG have I no time
‘EIG, I have no time’
Closer investigation reveals that EIGENTLICH carries a topic accent and relates to a fall (focus)
accent later in the clause (most naturally on ‘Luiselotte’ and ‘keine’; but other choices are
possible, as usual). The focus indicates the kind of alternatives that the speaker wants to
address: E.g. My name is Luiselotte / Lisei / Liserl / Lizzy / … in (14), and I DO have time vs.
I DON’T have time in (15). EIG(S) states: If we evaluated the truth of S, observing legally
correct language, morally preferable evaluation of matters / … , then S is the case, and not
any of the alternatives. EIG(S) statements in the indicative always refer to the actual world
wo, it is not a counterfactual operator. Still, adverbial EIG suggests that there is evidence
against S. Reference to alternatives of S implicates that matters are, or the world looks as if,
or the speaker will behave as if other alternatives were in fact true.2 The following coherent
continuations illustrate the effect.
(14’) ... even though everyone keeps calling me ‘Lizzy’.
(15’) … but I might behave as if that were not true and join you for the movies.
2
Note that this mimicks core insights of the analysis of eigentlich in terms of default
implications which was proposed by Schmitz/Schröder (2004).
10
Dependent on Context: ‘eigentlich’ in adjectival and adverbial use
This analysis can explain why (15)—as pointed out by many authors who worked on the issue
(e.g. Kohrt, Frühwirth)—is usually understood as implying agreement to some joint activity,
and yet taking it as straight “yes” is not possible. The last turn in (16) would be incoherent.
(16)
Will you join us for dinner?
habe ich keine Zeit.
#Ok, then let’s leave now.
EIGENTLICH
4. Further issues
If time allows, we will briefly touch on two further issues. Firstly, a closer look at the
pragmatic effects of stressed adverbial EIGENTLICH reveals that the common core, roughly,
consists in ‘coming to assertion S after some thinking’. Unstressed adverbial eigentlich as a
signal for the speaker’s attitude (=discourse particle use) seems to express just this. This use
is hard to translate into English because the adverb puts less prominence on the speaker’s
attitude than full explication (‘having thought about it for some time, I hereby assert …’).
Like in many other cases, loss of accent, loss of truth conditional content, and ‘pale’ meaning
appear hand in hand.
Secondly, when we tried to verify our intuitions about possible and impossible uses of
eigentlich via GOOGLE, we were surprised by accidential 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 likely Swiss background (e.g. Swiss embassy in Berlin, etc.). 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 welldefined 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 eigentlich.
Dittmar, N. 2002. Lakmustest für funktionale Beschreibungen am Beispiel auch, eigentlich
und also. In: Fabricius-Hansen, C., O. Leirbukt and O. Letnes (eds.): Modus,
Modalverben, Modalpartikeln. Trier: 142-177.
Frühwirth, A. 1999. Syntax, Semantic und Pragmatik der deutschen Modalpartikel eigentlich.
MA Arbeit, Marburg.
Kohrt, M. 1988. ‘Eigentlich, das Eigentliche und das Nicht-Eigentliche. Zum Gebrauch einer
sogenannten Abtönungspartikel.” In: Deutsche Sprache 2: 103-130.
Port, Angelika. 2006. ‘Was heißt eigentlich eigentlich? Eine semantische und pragmatische
Analyse’. Magisterarbeit, HU Berlin.
Schmitz, H.-C. and B. Schröder. 2004. Updates with ‘eigentlich’. In: Sprache und
Datenverarbeitung 1-2.
11
SPRIK Conference
Explicit and Implicit Information in Text – Information Structure across Languages
University of Oslo, June 8-10 2006
Adreflexive Intensification and the Theory of Focus and Information
Kjell Johan Sæbø
University of Oslo
[email protected]
Introduction
Quite some languages have both “true” and “false” reflexives (Bouchard 1984), or both SELF
and SE 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:
(1)
Sangerne akkompagnerte seg #(selv) på gitar.
singers-the accompanied SEG SELF on guitar
(2)
Hun vasket seg (#selv) i en liten bekk.
she washed SEG SELF in a little creek
(3)
Gulbransson drakk [seg (#selv) full].
Gulbransson drank SEG SELF drunk
(4)
Huni har sitt eget band med til å akkompagnere [seg (#selv)]i.
she has her own band with to to accompany
SEG SELF
Hellan developed an ingenious theory to predict this distribution. 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 traditional treatment. It is basically syntactic.
In sharp contrast to the traditional treatment, 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 intensifier 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, it is not evident how his account can be extended to
languages like Icelandic or German, where the intensifier is on the whole less necessary than in
Mainland Scandinavian or Dutch. He does not work with a formal theory of focus (such as Rooth
1992) and intensification (such as Eckardt 2001), and I will show that if one does, the open issues
can be settled – if the theory of focus is supplemented by the theory of informational integration
(Jacobs 1999) and by optimality theoretic pragmatics (Blutner 2000).
Bergeton (2004): Severing Intensification from Binding
Bergeton’s theory rests on these assumptions:
13
Kjell Johan Sæbø
• Binding and intensification are independent.
• Intensified anaphors are parallel to intensified pronouns and DPs.
• Intensification requires contexts providing alternatives and contrast
(the “Contrastiveness condition on adnominal intensification”).
• The distribution of intensified anaphors is sensitive to predicate meaning and pragmatics
(utterance situation and common ground).
• There are three relevant classes of predicates: Anti-reflexives, presupposing non-identity
of arguments, reflexives, presupposing identity of arguments, and neutral predicates.
This theory is theoretically satisfying in two senses: It is economical, aiming at a uniform analysis of SELF whatever it modifies, and it is explanatory in that the meaning of SELF plays a role.
It also makes more precise predictions about the distribution of the forms than earlier accounts.
In particular, SIG and SIG-SELF are not predicted to be in complementary distribution, as many
of the “inherently reflexive predicates” are reassigned to a new class of neutral predicates where,
depending on the context, both can be used, and there is no requirement for SIG-SELF to be a
coargument, so it can be a SC subject, if only the context generates a set of alternatives.
(5)
Han begynte å kle av meg. Så kledde han av seg selv også.
‘He started undressing me. Then he undressed himself too.’
(6)
Når han skal sove, synger han [seg selv i søvn].
when he shall sleep sings he SEG SELF in sleep
Still, the theory leaves some questions unanswered. The most important are:
• Why is SELF necessary in many cases – like (1) and (7)?
• Why is SELF less necessary in (say) German than in (say) Dutch, cf. (8)?
(7)
Narcissos sitter og beundrer seg #(selv).
Narcissus sits and admires SEG SELF
(8)
Sie begleitet sich (selbst) auf dem Klavier.
she accompanies SIG SELF on the piano
(7) is problematic because although SELF comes with a constraint, the Contrastiveness condition,
the absence of SELF does not come with a constraint. The anti-reflexive verb presupposes that
the two arguments refer to different entities, so (7) constitutes a presupposition failure, but it is
unclear how SELF helps justify the presupposition, or remedy the failure.
As for (8), the problem is that precisely because Bergeton’s theory is compositional, the option
open to Safir (2004) is closed: One cannot well argue that SIG+SELF is a unit in Dutch but not in
German, when the cornerstone of the theory is that binding and intensification are independent.
+ Intensification Semantics (Eckardt) and Informational Integration (Jacobs)
Eckardt (2001) has devised an ingenious theory of selbst intensification in German. It is:
• The intensifier denotes the identity function on type e entities.
From this assumption, the following follows:
14
Adreflexive Intensification and the Theory of Focus and Information
• The expression denoting the type e argument, the so-called associate, is a name, variable, 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 intensification is thus to add focus, giving rise to focus presuppositions in the
focus theory of Rooth (1992), Alternative Semantics.
Neither Eckardt nor Bergeton has applied the theory to the case where the associate is a reflexive.
(9)
(En barberer barberer alle som ikke barberer seg selv og ingen andre.)
a barber shaves everyone that not shaves SEG SELF and noone else
Barberer denne barbereren [seg selv]F?
SEG SELF
shaves this barber
Here I have chosen to let the intensifier and the reflexive form a focus constituent together. Then
the focus presupposition is simpler than if the intensifier is in focus all by itself. That of (9) is:
• There is a proposition ϕ and there is an alternative y to the barber x such that ϕ = shaves(y)(x).
This is verified in the context of (9), the barber’s customers instantiating y.
Now how does the reflexive relate to focus if it is not intensified?
• Is it out of focus (as if it were old information), corresponding to (10a)? Or:
• Is it in focus with something else – say, the predicate, corresponding to (10b)?
Eckardt only considers the former focus structure, but that is probably because the associates she
considers are all anaphoric in the discourse semantic sense (names, definites, personal pronouns).
Reflexive are different. And the latter structure, with wide focus and informational integration,
offers a way to account for the necessity of SELF with “anti-reflexive” predicates, as in (1) or (7).
(10) a. Han [barberer]F seg.
he shaves
SEG
b. Han [barberer seg]F.
Jacobs (1999) has developed a theory about the grammar, semantics, and pragmatics of structures
like (10b). The less informative, not accented constituent, here seg, is informationally integrated,
or nonautonomous; predicate and argument function as one informational unit and are processed
semantically in one step. This notion of informational unity has proven difficult to make precise.
+ OT Pragmatics (Blutner 1998 etc.)
But by supplementing Rooth’s theory of focus interpretation with a pragmatic, BOT component,
we can account for what is signalled by a wide focus on a predicate and its argument.
The focus presupposition of such a structure is too weak to account for informational integration;
indeed, it is properly weaker than that of the structure with two narrow foci, one on the predicate
and the other on the argument. However, once we consider the optimal interpretation of the wide
focus version, in competition with the double focus version, it turns out that this version does not
only not communicate the stronger focus presupposition; it positively communicates the opposite,
that there are alternatives to the VP as a unit, but not to the predicate and the argument separately.
15
Kjell Johan Sæbø
It can be shown that this boils down to saying that one is sufficiently predictable from the other, –
plus that there are no local alternatives to it in the discourse.
(11) a. Han [beundrer]F [seg selv]F.
he admires
SEG SELF
b. # Han [beundrer seg]F.
Thus (10b) will signal that the reflexive argument is sufficiently predictable from the predicate,
and that the discourse does not provide local alternatives to it. For (11b) to be felicitous, the same
must be the case; but of course, the reflexive cannot be sufficiently predictable on the basis of a
verb with this meaning. Therefore, (11b) is doomed to contradict its own optimal interpretation.
This explains the necessity of SELF with so-called anti-reflexive predicates.
But what about a language like German, where selbst is rarely necessary with such predicates?
(12) a. Er [bewundert]F sich [selbst]F.
he admires
SEG SELF
b. Er [bewundert]F sich.
As long as a focus structure analogous to that in Mainland Scandinavian or Dutch is assumed,
with informational nonautonomy for the not intensified reflexive, it is of course a mystery why
intensification is superfluous with “anti-reflexive” predicates. However, it is possible to assume
that in Icelandic or German, reflexives relate to focus in the same way as personal pronouns, that
is, when not accented they are out of focus, – cf. (12a) and (12b) as compared to (11a) and (11b).
That way, the reflexive is not predicted to be predictable from the predicate; it is predicted to be a
continuing topic (and in a sense, it always is) and not a contrastive rheme or theme in the context.
Selbst can be possible but not necessary because focus presuppositions are easy to accommodate.
To account for the contrast in this way, it is necessary to assume that a German reflexive can be
out of focus while a Danish reflexive, when the internal argument of a predicate whose external
argument binds it, is always in focus, normally parasitically on the intensifier or on the predicate,
with which it integrates. To be sure, there is scarce independent evidence for this assumption –
but the same can be said of the assumption that sich selbst fails to form a unit (Safir 2004: 205).
Thus it looks as if the novel theory of complex anaphors, launched by Bergeton (2004), loosening
information theory – intensification – from syntax – binding, can be intensified.
References
Bergeton, Ulf. 2004. The Independence of Binding and Intensification. Doctoral Dissertation,
University of Southern California.
Blutner, Reinhard. 1998. Lexical Pragmatics. Journal of Semantics 15: 115–162.
Blutner, Reinhard. 2000. Some Aspects of Optimality in Natural Language Interpretation.
Journal of Semantics 17: 189–216.
Blutner, Reinhard. 2002. Lexical Semantics and Pragmatics. In: Hamm, Fritz and Zimmermann, Ede
(eds.), Semantics (= Linguistische Berichte Sonderheft 10), 27–58.
Blutner, Reinhard. 2004. Pragmatics and the lexicon. In: Horn, Larry and Ward, G (eds.),
16
Adreflexive Intensification and the Theory of Focus and Information
Handbook of Pragmatics. Oxford: Blackwell. 488–514.
Bouchard, D. 1984. On the Content of Empty Categories. Dordrecht: Foris.
Eckardt, Regine. 2001. Reanalysing selbst. Natural Language Semantics 9: 371–412.
Edmondson, J. and Frans Plank. 1978. Great Expectations: An intensive self analysis. Linguistics and
Philosophy 2: 373–413.
Gast, Volker. 2002. The Grammar of Identity: Intensifiers and Reflexives as Expressions of an Identity
Function. Doctoral Dissertation, FU Berlin.
Hellan, Lars. 1988. Anaphora in Norwegian and the Theory of Grammar. Dordrecht: Foris.
Jacobs, Joachim. 1991. Focus Ambiguities. Journal of Semantics 8: 1–36.
Jacobs, Joachim. 1999. Informational Autonomy. In: Bosch, Peter and Rob van der Sandt (eds.), Focus:
Linguistic, Cognitive, and Computational Perspectives. Cambridge: Cambridge University Press.
56–81.
König, Ekkehard and Peter Siemund. 2000. Intensifiers and reflexives: A typological perspective. In:
Frajzyngier, Z. and T. S. Curl (eds.), Reflexives. Forms and Functions. Amsterdam: John Benjamins.
41–74.
Mattausch, Jason. 2003. Optimality theoretic pragmatics and binding phenomena. In: Blutner, Reinhard
and Zeevat, Henk (eds.), Optimality Theory and Pragmatics. London: Palgrave MacMillan. 63–91.
Reinhart, Tanya and Eric Reuland. 1993. Reflexivity. Linguistic Inquiry 24: 657–720.
Rooth, Mats. 1992. A Theory of Focus Interpretation. Natural Language Semantics 1: 75–116.
Sæbø , Kjell Johan. 2005. Intensifying self in Germanic: A Reanalysis. The Stockholm Workshop on
Contrast, Information Structure and Intonation.
Sæbø, Kjell Johan. 2006. Focus Interpretation in Thetic Statements: Alternative Semantics and OT
Pragmatics. Submitted to Journal of Logic, Language and Information.
Safir, Ken. 2004. The Syntax of Anaphora. Oxford: Oxford University Press.
Siemund, P. and Volker Gast. 2006. Rethinking the relationship between self-intensifiers and reflexives.
Linguistics 44: 343–381.
Zeevat, Henk. 2003. Markedness. In: Asatiani, Rushiko et al. (eds.) Proceedings of the 5th International
Symposium on Language, Logic and Computation, Tbilisi. 183–190.
17
A Lexical Perspective on Discourse Structure and Semantics
Bonnie Webber
School of Informatics, University of Edinburgh
[email protected]
D-LTAG is a lexicalized approach to discourse relations, providing an account of how lexical
elements (including phrases) anchor discourse relations and how other parts of the text provide
arguments for those relations.
D-LTAG arose from a belief that the mechanisms for conveying discourse relations were
unlikely to be entirely different from those for conveying relations within the clause. Because
the latter can be anchored on lexical items, D-LTAG was developed as a lexicalized grammar for
discourse – specifically, a 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 and large) by discourse connectives,
whose substitution sites can be filled by the representation of anything interpretable as an abstract
object (ie, a proposition, fact or event). Elements so interpretable include discourse segments,
sentences, clauses, nominalisations and demonstrative pronouns.
As with a sentence-level LTAG, there are two types of elementary trees in D-LTAG: Initial
trees, anchored by structural connectives such as coordinating and subordinating conjunctions
and subordinators (eg, in order to, so that, etc.) as shown in Figure 1(a), and auxiliary trees,
anchored by and, an empty connective ( ), or a discourse adverbial (eg, otherwise, however as a
result, etc.) as shown in Figure 1(b).
D-LTAG structures are interpreted in terms of discourse relations between arguments (ForbesRiley et al., 2006). The process is partly compositional, partly inferential, and partly the result
of anaphor resolution. The latter is essential for intepreting discourse adverbials, which convey
a discourse relation between the abstract object (AO) interpretation of the matrix clause and an
AO interpretation in the discourse context. That discourse adverbials differ in this way from
structural connectives is demonstrated on theoretical grounds in Webber et al. (2003) and on empirical grounds in Creswell et al. (2002). An explanation for the anaphoric character of discourse
adverbials is given in Forbes (2003) and Forbes-Riley et al. (2006).
Discourse relations arising from both structural connectives and discourse adverbials can be
seen in the D-LTAG analysis of Example 1:
(1) John loves Barolo.
So he ordered three cases of the ’97.
But he had to cancel the order
because he then discovered he was broke.
The analysis is shown in Figure 2. It involves a set of elementary trees for the connectives (so,
but, because, then) and a set of leaves ( - ) corresponding to the four clauses in the example, minus the connectives. Through the operations of substitution (solid lines) and adjoining
19
(dashed lines) recorded in the derivation tree (here shown to the right of the arrow), a derived
tree is produced (here shown at the head of the arrow). More detail on both the representation of
connectives and D-LTAG derivations is given in Webber et al. (2003). A preliminary parser for
D-LTAG is described in Forbes et al. (2003) and Webber (2004).
Compositional interpretation of the derivation tree produces the discourse relations associated
with because, so and but, while anaphor resolution produces the other argument to the discourse
relation 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 D-LTAG’s syntacticsemantic interface are given in Forbes-Riley et al. (2006).
Note that, although D-LTAG produces only analyses in the form of trees, it is acknowledged
in Webber et al. (2003) – as noted earlier by Bateman (1999) and Gardent (1997) – that a discourse unit must be allowed to participate in one constituent structure with left-adjacent material
and another with right-adjacent material, as in Figure 3. This would require relaxing substitution
constraints in D-LTAG that the same tree only substitutes into a single site.
Empirical data on the predicate-argument structure of discourse connectives are now available in Release 1.0 of the annotated Penn Discourse TreeBank1 (Dinesh et al., 2005; Miltsakaki
et al., 2004a,b; Prasad et al., 2004; PDTB-Group, 2006; Webber, 2005). In the PDTB, discourse
connectives have been manually annotated, along with the text spans that give rise to their two
arguments. The text that has been annotated is the same 1-million word Wall Street Journal
corpus that has been annotated for syntactic structure in the Penn TreeBank 2 and for verbs and
their arguments in PropBank.3 Annotated in Release 1.0 are tokens of the following explicit
connectives and their arguments:
31 subordinating conjunctions (eg. when, because, as soon as, now that, etc.) and their
modifiers (only, just, even, mainly, etc.)
7 coordinating conjunctions (and, but, or, either/or, nor, neither/nor, so
62 discourse adverbials
This constitutes 18505 tokens, one for each instance of an explicit connective and its arguments.
Also annotated in Release 1.0 are implicit connectives in three sections of the corpus, comprising
2003 tokens. An early effort to use data in the PDTB to develop a procedure for resolving the
anaphoric argument of the discourse adverbial instead is described in Miltsakaki et al. (2003),
and the effect of Information Structure on the preferred argument of the discourse adverbial
otherwise is described in Kruijff-Korbayová and Webber (2001). Although English is currently
the only language for which there is a Discourse TreeBank, I hope that the idea of a lexicalized
discourse-level grammar and a discourse treebank would be of interest and of value for other
languages as well.
1
http://www.seas.upenn.edu/˜pdtb
http://www.ldc.upenn.edu/Catalog/CatalogEntry.jsp?catalogId=LDC99T42
3
http://www.ldc.upenn.edu/Catalog/CatalogEntry.jsp?catalogId=LDC2004T14
2
20
References
Bateman, J. (1999). The dynamics of ‘surfacing’: An initial exploration. In Proceedings of
International Workshop on Levels of Representation in Discourse (LORID’99), pages 127–
133, Edinburgh.
Creswell, C., Forbes, K., Miltsakaki, E., Prasad, R., Joshi, A., and Webber, B. (2002). The
discourse anaphoric properties of connectives. In Proceedings of the Discourse Anaphora and
Anaphor Resolution Colloquium, Lisbon, Portugal.
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, R., Sarkar, A., 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.
Gardent, C. (1997). Discourse tree adjoining grammars. Claus report nr.89, University of the
Saarland, Saarbrücken.
Kruijff-Korbayová, I. and Webber, B. (2001). Information structure and the semantics of “otherwise”. In ESSLLI’2001 Workshop on Information Structure, Discourse Structure and Discourse Semantics, pages 61–78, Helsinki, Finland.
Miltsakaki, E., Creswell, C., Forbes, K., Prasad, R., Joshi, A., and Webber, B. (2003).
Anaphoric arguments of discourse connectives: Semantic properties of antecedents versus
non-antecedents. In EACL Workshop on Computational Treatment of Anaphora, Budapest,
Hungary.
Miltsakaki, E., Prasad, R., Joshi, A., and Webber, B. (2004a). Annotating discourse connectives
and their arguments. In NAACL/HLT Workshop on Frontiers in Corpus Annotation, Boston
MA.
Miltsakaki, E., Prasad, R., Joshi, A., and Webber, B. (2004b). The Penn Discourse Treebank. In
LREC, Lisbon, Portugal.
PDTB-Group, T. (2006). The Penn Discourse TreeBank 1.0 annotation manual. Technical Report
IRCS 06-01, University of Pennsylvania.
21
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, Barcelona, Spain.
Schabes, Y. (1990). Mathematical and Computational Aspects of Lexicalized Grammars. Ph.D.
thesis, Department of Computer and Information Science, University of Pennsylvania.
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., Stone, M., Joshi, A., and Knott, A. (2003). Anaphora and discourse structure.
Computational Linguistics, 29, 545–587.
22
α:or
α:so
α :while−init
or
so
while
(a) Initial trees for ’or’, ’so’ and clause−initial ’while’
β :and
β :otherwise
β:Φ
and
*
*
Φ
otherwise
*
(b) Auxiliary trees for ’and’, the empty connective and ’otherwise’
Figure 1: Initial and Auxiliary Trees in D-LTAG
α: so
β: but
*
then
*
so
but
because
α: so
1
T1 love
α: because_mid
β:then
0
3
β: but
T2
T1
T2 order
3
α: because_mid
T3 cancel
1
T4 discover
T3
3
T4
0
β:then
but
because
so
T1 love
T2 order T3 cancel
then
T4
Figure 2: Derivation of Example 1
sequence
manner
Figure 3: Simple multi-parent structure
23
SPRIK Conference
Explicit and Implicit Information in Text - Information Structure across Languages
University of Oslo, June 8-10 2006
A model for sentence-fragment production
Shinji Ido
Tohoku University / University of Sydney
[email protected]
1. The issue
In various areas of study within the discipline of linguistics, sentence-fragments (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),
Malmkjaer (2002: 543)). Accounts of sentence-fragment production where sentencefragments (hereafter SFs) are identified as sentences with missing constituents are necessarily
based on the assumption that every SF has a corresponding ‘full sentence’ into which native
speakers can ‘reconstruct’ it with (near-)unanimous agreement. However, despite its
popularity, one can find without much difficulty examples that contradict this assumption. For
example, in as early as 1974, Gunter (1974: 12-13) devised the term ‘telegraphic ellipses’ to
refer to SFs (‘elliptical sentences’ in his terminology) of which ‘informants do not agree …
on the proper expansion’ (ibid: 13). The existence of such SFs raises the question of whether
it is reasonable to assume the existence of a ‘full sentence’ for every SF.
The present study is an attempt to account for SF production without assuming the existence
of a ‘full sentence’ for every SF. Although this study is at a preliminary stage, the model for
SF production that derives from it has the following three advantages over the popular
‘constituent-omission’ model explained above: 1) it accounts for the production of the sort of
SFs that Gunter calls ‘telegraphic ellipses’, 2) it predicts what constituents have to be present
in a given SF, and, perhaps more significantly, 3) it explains why in certain contexts pro-drop
cannot occur in languages that have subject-verb agreement morphology. I describe below
how this model, which I tentatively call the composite model, accounts for SF production.
2. The model
The principle on which the composite model is based is simple: ‘in informative
communication, foci (which are, as will be explained below, morphemes in this model) and
morphemes that are grammatically required to accompany them necessarily occur’. I present
below a simple schematic representation of the model in which a Turkish dialogue taken from
Enç (1986: 195) is used as an example. fs and ms in the following chart represent foci and
morphemes that are grammatically required to accompany them, respectively.
25
Shinji Ido
Mary’s utterance:
(1)
Herkes
Ali-’yle
everyone
Ali-COM
‘Everyone met Ali.’
tanış-tı.
meet-PAST.3SG
Mary thinks that:
Everyone met Ali.
(2)
∀x∃e(Meeting(e) & Past(e) & Agent(e,x) & Theme(e,a))
John wants to contradict Mary because:
John did not meet Ali.
(3)
¬∃e(Meeting(e) & Past(e) & Agent(e,j) & Theme(e,a))
negative morpheme -ma (NEG)
{f: f is a focus}
{m: m is a morpheme grammatically required to accompany f}
Sentence-fragment
1st person singular pronoun ben (I)
ben
benf
-ma
tanış-dı
tanış-m -maf -dım
-m
-mm
John’s response to (1):
(4)
Ben tanış-ma-dı-m.
I
meet-NEG-PAST-1SG
‘I didn’t.’
(Note that (4) is an SF which would, in the ‘constituent-omission’ model, be identified as the
following ‘full sentence’ from which Ali’yle has been omitted: Ben Ali-’yle tanış-ma-dı-m. (I
Ali-COMITATIVE meet-NEGATIVE-PAST-1SG) ‘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).)
At the time (1) is uttered, John’s proposition (3) differs from (2) by having one connective
and one individual constant, namely ¬ and j, which map onto the negative morpheme -ma and
the first person singular pronoun ben, respectively. These morphemes, which are called foci in
this model, need to be present in (4) because the principle mentioned above calls for their
occurrence. This accounts for the obligatory occurrence of the pronominal subject ben ‘I’, i.e.
the non-occurrence of pro-drop, in (4). The other of the two foci, namely -ma, cannot occur in
isolation and calls for the accompaniments of three morphemes, namely tanış-, -dı, and -m,
the occurrence of which is also required by the principle. (See Ido (2003: 50-51) for a detailed
explanation of why these morphemes specifically need to occur with -ma.) Thus the
composite model determines which constituents must be present in John’s response to (1).
In summary, the production of an SF is represented in the composite model as a process in
which foci and their accompaniments are ‘put together’ rather than as a process where
constituents are omitted from a ‘full’ sentence. This model thus 1) explains SF production
26
A model for sentence-fragment production
without assuming a ‘full sentence’ for every SF, 2) identifies constituents that must occur in
an SF, and 3) explains why pro-drop cannot occur in certain contexts.
3. Cross-linguistic validity
The validity of the 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 that is observed in Turkish (4). (The information
structural property of each morpheme is shown in subscript.) All of these examples obey the
principle stated in § 1: ‘foci (fs) and morphemes that are grammatically required to
accompany them (ms) necessarily occur’.
Mary’s utterance:
(1)
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):
(4) in Turkish
benf
I
tanışm-maf-dım-mm
meet-NEG-PAST-1SG
(4) in Mongolian
bif
I
uulzm-aam-güyf1
meet-IMPERFECTIVE-NEG
(4) in Bukharan Tajik
manf na-fšudm-amm
I
NEG-became-1SG
(4) in Japanese
watashif-wam awm-anaf-kattam
I-TOPIC
meet-NEG-PAST
‘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.
4. Summary
The composite model comprises three mappings, namely the mapping of parts of a
proposition onto morphemes, which are then identified as fs (mapping 1), the mapping
between fs and ms (mapping 2), and the mapping of fs and ms onto a linear line of time
(mapping 3)2. The point at which this model departs most radically from the constituentomission model is the role of syntax in SF production. In the constituent-omission model,
1
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 which is not used here because of the modality of regretfulness that it encodes.
2
Mapping 3, which is not discussed in the present paper, is omitted from the chart in § 1. (3) in fact yields
three sets of morphemes, namely the sets of fs, ms, and ns. Mapping 3 linearly aligns fs, ms, and ns in
accordance with syntax and/or morphology. {n: n is a morpheme whose occurrence or non-occurrence does
not alter the proposition expressed by fs and ms} will be explained in detail in a follow-up paper. Unlike fs,
ms are not constant across varieties and dialects and are susceptible to the differences between, say, the
grammar of the informal style and that of the formal style — in general, the grammar of the formal style
requires more morphemes to accompany fs, turning morphemes that would be ns in other styles into ms. A
‘full’ sentence often consists of fs, ms, and ns, whereas a ‘fragment’ often consists only of fs with or
without ms. The following may serve as an example of such a ‘fragment’ in English: Good short-term, bad
long-term (a fifteen-year-old Australian’s response to you have a good memory!).
27
Shinji Ido
meaning is mapped onto a syntactic structure, following which ellipsis takes place within that
structure (see analyses in Merchant (2004)). On the other hand, in the composite model,
mappings 1 and 2 restrict the role of syntax in SF production to often simple linear alignment
of fs and ms. This order in which the three mappings are executed means that fs and ms can
often be aligned without an elaborate syntactic structure.3 This model thus postulates that
there may not always be ‘syntactically full’ sentences in SF production.4
References
Bavin, E. L. 2000. Introduction: a functional approach to ellipsis. Linguistics, 38, 3. 449-455.
Brown, K. / Miller, J. 1991. Syntax 2nd edition. London: Harper Collins Academic.
Enç, M. 1986. Topic switching and pronominal subjects in Turkish. IN Slobin, D. I. /
Zimmer, K. eds. 1986. Studies in Turkish linguistics. John Benjamins. 195-208.
Gunter, R. 1974. Sentences in dialog. Columbia: Hornbeam Press.
Ido, S. 2003. Agglutinative information. Wiesbaden: Harrassowitz.
Lyons, J. 1977. Semantics volume 2. Cambridge University Press.
Malmkjaer, K. ed. 2002. The linguistics encyclopedia 2nd edition. Routledge.
Matthews, P. H. 1997. The concise Oxford dictionary of linguistics. Oxford University Press.
Merchant, J. 2004. Fragments and ellipsis. Linguistics and philosophy, 27. 661-738.
Napoli, D. Jo. 1996. Linguistics: An introduction. Oxford University Press.
3
In some cases, linear alignment of fs and ms does not call for the involvement of any syntax. For example,
in the following utterance that appears in a Japanese interview transcription: -na-i. (-NEG-NONPAST), which
is the interviewee’s answer to … zutto ita wake de wa nai? ‘… so you weren’t there all the time?’
(http://www.hotexpress.co.jp/interview/kyoko/ (2005/09/16)), the order in which -na and -i occur can be
determined by the simple rule of ‘the negative suffix precedes the tense suffix’.
4
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 nor an n. This analysis is also consistent with the fact that (c), which consists of one f, is an
acceptable answer to the question.
Korean
Q:
Nu-ka
bought
a.
A:
b.
A:
c.
A:
ku
who-NOM
this
‘Who bought this book?’
Yongsu-ka.
Yongsu-NOM
*Yongsu-rul
Yongsu-ACC
Yongsu.
Yongsu
28
chaek-ul
sa-ss-ni?
book-ACC
SPRIK Conference
Explicit and Implicit Information in Text - Information Structure across Languages
University of Oslo, June 8-10 2006
A cross-linguistic investigation of the licensing and interpretation
of implicit object arguments
Gergely Pethő, Éva Kardos
University of Debrecen, Hungary
[email protected]
1. Introduction
This paper discusses the occurrence and the licensing of implicit object arguments, also referred
to in the literature as null complements or understood arguments. We will argue that approaches
which aim to explain this phenomenon in terms of general pragmatic principles (in particular,
relevance) are not sufficient. Furthermore, we will try to show that particular well-known results
of these approaches are conceptually and empirically problematic.
Whereas arguments in general are both syntactically and ontologically required by their
predicates (for example, a locking event necessarily involves an agent and a patient argument,
both of which must normally be expressed syntactically), it is possible under certain
circumstances to produce grammatical utterances that involve phonologically unrealised
arguments, cf. the often cited example of the intransitive use of the verb eat, as in:
(1)
John is eating.
In (1), a patient of the eating event is still ontologically necessary, i.e. an eating is not
conceivable without there being something eaten, but it is not realised syntactically. Such
mismatches between the presence of participants in the situation expressed by a predicate on the
one hand and the presence of syntactic complements that name these participants on the other
hand are rather numerous, as is well known. Some of these can be characterised as properties of
the verbal predicate itself, i.e. the verb has both a transitive and an intransitive version in the
lexicon, like eat, drink, cook, read. Others seem to be relatively independent of the verb, but have
to be associated instead with certain (grammatically and semantically characterisable)
constructions, e.g. habituality (2) and coordination (3):
(2)
(3)
I like to knit.
He will steal, rob and murder.
It has been repeatedly suggested, for example by Groefsema (1995) and Németh T. (2001), that
the omission of complements is primarily driven by pragmatic and discourse factors. More
exactly, according to this view it is lexical and encyclopedic knowledge or previous contextual
information that helps the hearer recover the reference of the unrealised element and thus reach
an interpretation of the utterance that conforms to the pragmatic principle of relevance. This
approach claims that implicit arguments are always ultimately licensed by the interpretability of
the utterance.
In this paper, we present arguments against this claim. Firstly, a semantic-pragmatic account of
argument omission, as elaborated in Groefsema (1995), is problematic in several respects.
Secondly, certain well-known phenomena of the syntax-semantics interface (e.g. resultative
constructions, prefixed verbs) seem to exhibit regularities with respect to argument omission
29
Gergely Pethõ & Éva Kardos
which are independent of the factors of interpretability or relevance. We will explain these two
issues in more detail in the following sections.
2. The role of selection restrictions in the licensing of implicit argument
Groefsema (1995) outlined a theory of the lexical licensing of argument omission, according to
which this process is driven by selection restrictions of individual verbs. Essentially, her
assumption is 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. For example, the verb drink puts a specific restriction on its direct object
argument: what is drunk is liquid (Groefsema 1995: 146), and similarly for verbs like eat and
read. In such cases, the patient is not required to be expressed. 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).
Groefsema also aims to account for a classic distinction between two types of implicit argument:
so-called indefinite null complements (INC) and definite null complements (DNC). This
distinction was introduced by Fillmore (1986), and aims to capture a semantic difference between
two types of verbs. Indefinite null complements (i.e. implicit objects) of verbs receive an
“existentially quantified” interpretation, e.g. I am eating approximately means ‘I am eating
something’, but not ‘I am eating it.’. On the other hand, a definite null complement is interpreted
anaphorically and must therefore have an appropriate antecedent in context to make sense, e.g.
(4)
“Why did you marry her?”
“Because mother insisted.”
Here, the object of insist is missing and can be recovered on the basis of the previous discourse
segment.
Fillmore’s original lexical-syntactic account of this observation was that individual verbs are
assigned a syntactic feature value (for whether they allow argument omission or not) in the
lexicon. Groefsema tried to incorporate this distinction into her theory and explain INC and DNC
behaviour of verbs in terms of selection restrictions as well. She distinguishes two ontological
categories of entities (things), namely, types and instances. She assumes that an implicit
argument receives an indefinite interpretation if this argument is marked in the corresponding
selection restriction of the verb as a type, whereas definite (anaphoric) implicit arguments are
marked as instances. Drink as an INC verb specifies that its object has to be a type of liquid,
whereas win as a DNC verb has an object that is an instance of a competition.
Although it is without doubt an attractive goal to derive both the possibility of argument omission
and the definite or indefinite interpretation of an implicit argument from a single lexical semantic
property (selection restrictions), there are some problems with this theory which render it
somewhat unconvincing.
Firstly, it is not clear what it means that “a verb can specify that its argument is of a particular
type [or] that it is an instance of a particular type” (Groefsema 1995: 146), since quite trivially,
30
A cross-linguistic investigation of the licensing and interpretation of implicit object arguments
being of a particular type and being an instance of a particular type seems to us to be one and a
same thing. Unfortunately, Groefsema does not provide any further explanation for this
distinction, and therefore it is unclear to us how it could be rephrased in a more coherent way.
Secondly, assuming that this distinction does indeed make sense, it seems that Groefsema’s
characterisation of specific verbs makes incorrect predictions about their distributions. Although
it is not completely clear 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 more 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 take objects that are bare plurals or mass nouns (I am reading newspapers.), but not
NPs referring to specific entities (I was just reading that brown book over there...). On the other
hand, win, being a verb that demands its object to be an instance of a type, should only appear
with specific NP objects (She has won the race.), but never with bare plurals (They play really
smart ball and they often win championships.). It is not only this plausibly predicted pattern that
fails to apply, but 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 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 that fact at all.
3. Cross-linguistic considerations and implicit arguments
It has been pointed out by Iten et al. (2004) that the interpretational (semantic-pragmatic) account
of implicit argument licensing raises some unanswered questions that become especially clear
when several languages are compared to each other. In particular, this approach does not seem to
be able to explain cross-linguistic differences in argument structure variations, because there is
no independent reason to assume that a relevant interpretation can be reached in one of the
examined languages, but not in the other. For example, it has been noted (Sæbø 1996) that the
possibility of using verbs like insist with a definitely interpreted implicit object in English, e.g.
Because she insisted., does not generalise to other languages. The equivalent structures in
Hungarian or German, for instance, would obligatorily contain a pronominal representing the
state of affairs that is demanded by the subject of the sentence. What appears to license the
omission in English as opposed to Hungarian or German does not seem to have anything to do
with whether a relevant interpretation for the utterance can (or cannot, respectively) be derived,
but rather with whether the appropriate syntactic construction is available in a language (as in
English) or not (as in Hungarian or German).
Furthermore, it is not only cross-linguistic differences, but also similarities that support the view
that argument omission is in certain cases more a matter of grammar and syntax rather than one
of pragmatics. More specifically, it has been noted (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, e.g. John painted *(the fence) red. 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, e.g. Hun.
(5)
János pirosra
festett *(egy kerítést).
John red-result.
painted a fence-acc.
‘John painted *(a fence) red.’
31
Gergely Pethõ & Éva Kardos
A syntactic-semantic explanation could be attempted for this characteristic property of resultative
constructions, 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.
(6)
*János egész délután a fürdőszobában volt és pirosra festett.
*‘John spent the whole afternoon in the bathroom and painted red.’
Similar differences between different occurences 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. Ger. and Hun.
(7)
Sie näht für ihre Kinder.
she sews for her children
‘She sews for her children.’
(8)
Wir nähen
*(die kleinen Stoffstücke) zusammen.
we sew
the small pieces-of-fabric together.
‘We sew *(the small pieces of fabric) together.’
(9)
Kati írt
délután.
Kati wrote
afternoon
‘Kati wrote in the afternoon.’
(10)
Kati kiírt
*(három kérdést)
a könyvből.
Kati out-wrote
three questions-acc the book-from
‘Kati wrote *(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. Tenny 1994). 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 syntax-semantics 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), 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. Ger.
(11)
Er brüllte
seine Freude
heraus.
he screamed
his happiness out
‘He cried out with joy.’
Similarly to the cases above (8, 10), the omission of the object leads to ungrammaticality.
Considering the examples mentioned so far, it would 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
32
A cross-linguistic investigation of the licensing and interpretation of implicit object arguments
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 predicate.
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.
(12)
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 friends.’
(13)
Der
Schienenbus befährt die Strecke am Sonntag dreimal in beide Richtungen.
the rail-bus
pfx-drives the distance on-the Sunday three-times in both
directions
‘The train travels this distance every Sunday three times in both directions.’
Similarly, in Hungarian:
(14)
János eszik (egy almát).
John eats an apple
‘John is eating an apple.’
vs.
János megeszik *(egy almát).
John pfx-eats an apple
‘John is eating an apple.’
In the case, the simplex verb is lexically imperfective (that is, of the event type process), the
prefixed verb is lexically perfective (an accomplishment).
Both of these functions of verbal prefixes, i.e. transitivisation (along with argument structure
modification) and perfectivisation, appear in large numbers in both of these languages that have
an extensive productive system of verb formation by prefixation.
The same observation can be made in connection with English as well. It was Fillmore in his
classic paper (1986) who cited the famous pair of wait and await, two verbs which seem to be
closely synonymous, but differ in their ability to occur with an implicit object: I am waiting. vs. I
am awaiting *(John). He used this example to argue that what we call the pragmatic-semantic
approach to implicit argument licensing is wrong, since that approach should predict that two
synonymous predicates should indeed behave identically with respect to argument omission.
Note that there is also a clear difference between the argument structures of these verbs (wait
takes an indirect object, whereas await takes a direct object) in parallel with the possibility of
intransitive use. This correspondence between transitivisation and the impossibility of argument
omission appears quite systematically in English as well as in the languages mentioned above.
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.
(15)
(16)
For years, parents and teachers have bemoaned the fact that we do not have a
national childcare policy.
I'm fed up with hearing you moaning the whole time!
33
Gergely Pethõ & Éva Kardos
4. Conclusions
Although we have concentrated only on its problematic aspects, the semantic-pragmatic approach
to implicit argument licensing has shown quite convincingly (e.g. Németh T. 2001) that
pragmatics does play an important role in this process. World knowledge, information contained
in the linguistic and situational context, and the consideration whether, based on these types of
knowledge, a consistent interpretation for the utterance can be achieved do seem to play a very
significant role in determining whether certain arguments may be or are even preferred to be
omitted in certain contexts. On the other hand, we hope to have shown in section 3 that certain
syntactic factors (which are connected to semantic factors like event and argument structure)
prohibit argument omission in many cases. Therefore, we believe that an adequate framework in
which the licensing and interpretation of implicit arguments may be satisfyingly explained has to
involve a division of labour, in accordance with Iten et al. (2004)’s suggestion. On the one hand,
syntax (including the syntax-semantic interface and morphosyntax) has to license a structure
which involves an omitted argument of a predicate, which may fail to apply in certain
constructions (like resultatives or transitivisation by prefixation), thus blocking the generation of
the relevant structures. On the other hand, once a structure is successfully generated by syntax
(i.e. is grammatical in the given language), the speaker has to consider (usually unconsciously)
whether the expression can be interpreted if uttered in the given context. Making this distinction
between grammaticality on the level of syntax and interpretability (which is equated with
acceptability by the pragmatic approach) on the level of semantics-pragmatics should improve
the plausibility and power of pragmatic explanations of argument omission significantly.
References
Fillmore, Charles J. 1986. Pragmatically controlled zero anaphora. Proceedings of the Berkeley
Linguistics Society 12. 95-107.
Groefsema, Marjolein. 1995. Understood arguments: A semantic/pragmatic approach. Lingua 96:
139-161.
Iten, Corinne, Marie-Odile Junker, Aryn Pike, Robert Stainton, Catherine Wearing. 2004. Null
complements: licensed by syntax or by semantics-pragmatics? In: M. O. Junker, M.
McGinnis & Y.Roberge (eds.), Proceedings of the 2004 Canadian Linguistics Association. 115.
Németh T., Enikő. 2001. Implicit arguments in Hungarian: manners of their occurrence and
possibilities of their identification. In: I. Kenesei (ed.), Argument structure in Hungarian.
Budapest: Akadémiai. 113-156.
Sæbø, Kjell Johann. 1996. Anaphoric presuppositions and zero anaphora. Linguistics and
Philosophy 19: 187-209.
Tenny, Carol. 1994. Aspectual roles and the syntax-semantics interface. Dordrecht: Kluwer.
34
SPRIK Conference
Explicit and Implicit Information in Text - Information Structure across Languages
University of Oslo, June 8-10 2006
The discourse function of right-dislocation in Catalan1
Laia Mayol
University of Pennsylvania
[email protected]
1. Introduction
All languages have the means to express the same propositional content in many different ways.
Although truth-conditionally equivalent, these variants are often not felicitous in the same
contexts. Instead, they impose different constraints on the kind of context they may appear in and
encode some non-truth-conditional meaning, some “informational component” (Vallduví, 1994).
By structuring the information in a particular way, the speaker is instructing the hearer about how
to process and interpret the utterance, especially in relation to the givenness status of the entities.
This paper presents a corpus study of right dislocation (RD), the non-canonical construction in
which a clitic pronoun in the main clause is coreferential with a final peripheral noun phrase, and
is aimed at describing its main discourse functions. I argue that RD in Catalan is clearly a means
to structure information in a coherent way by displacing old information from the main clause.
2. Previous approaches to right dislocation
RDs have not received as much attention in the literature as other non-canonical constructions
since they have been regarded by some authors as performance errors, as “afterthoughts”.
However, more recently, most accounts agree that RDs genuinely contribute to the packaging of
the information. According to Vallduví (1994), in Catalan, RDs are tails, pieces of old
information not maximally salient at the moment of utterance that establish how the new
information must be updated. RDs are different from left-dislocations in that only the latter
encode links, which may be contrastive and which indicate where the information must be
updated. RD constructions are different from sentences with pure pronominalization in that, in
the latter, there is no mention of how the information must be updated.
For English, Ziv and Grosz (1993) note that the use of RDs is proscribed to refer to an entity
which has just been mentioned unless it adds some attributive meaning, as (1) shows. According
to them, the RD is felicitous in English when referring (i) to an entity present in the discourse
situation but not mentioned or (ii) to entities textually evoked only when (a) they have been
mentioned in discourse, but not recently or (b) the NP adds some attributive meaning (as in 1b).
(1) a. I took my dog to the vet yesterday. # He is getting unaffordable, my dog
b.
He is getting unaffordable, the old beast
3. Data
My corpus consists of 93 naturally occurring instances containing RD in Catalan: 67 instances
are transcriptions of a famous comedian’s shows and the rest were taken from radio, television,
1
I wish to thank Enric Vallduví for advice and encouragement and to the Penn’s Semantics Reading Group for many
useful comments.
35
Laia Mayol
press, novels or conversations2. The data has been analyzed according to the following
parameters (the percentage of instances falling in each category is indicated in parenthesis):
1. Status of the detached NP
a. Mentioned in the previous sentence (60.2%)
b. Previously mentioned, but not in the previous sentence (23.6%)
c. Non-textually mentioned, although may be situationally implicit (16.2%)
2. Consequences of eliminating the right-dislocated constituent of the sentence:
a. ok the sentence is still fully acceptable (23.7%)
b. * the sentence is not acceptable (47.3%)
c. ? the sentence is worse than the original, but still acceptable (29%)
3. Consequences of restoring the canonical order:
a. ok the sentence is still fully acceptable (10.7%)
b. * the sentence is not acceptable (47.3%)
c. ? the sentence is worse than the original, but still acceptable (42%)
Table 1 shows the data broken down according to the first parameter (boldface indicates the
highest value in each category).
Status of detached NP
Eliminating RD
Canonical order
%
%
%
1a
1b
1c
60.2
23.6
16.2
ok
32.1
ok
1.8
*
25
*
60.7
?
42.9
?
37.5
ok
18.2
ok
13.6
*
72.7
*
36.4
?
9.1
?
50
ok
0
ok
40
*
93.3
*
13.3
?
6.7
?
46.7
4. Data analysis: Status of the dislocated information
The data resulting from parameter 1 backs up the claim that right-dislocated elements encode
discourse-old information: in 83.6% (60.2% from 1a plus 23.6% from 1b) of occurrences of RDs,
the detached element had been mentioned in the previous discourse. RD is a good strategy to
mention old information and also to displace it from the main clause. However, it seems that the
constraints Ziv and Grosz identify for the use of RD in English do not hold for Catalan. Recall
that they claim that an NP cannot be right-dislocated if the entity to which it refers has been
mentioned in the previous utterance, unless it adds some attributive meaning. However, 60.2% of
2
The full corpus is available at http://www.ling.upenn.edu/~laia/papers/dd.pdf
36
The discourse function of right-dislocation in Catalan
RDs in Catalan belong to this category (1a in Table 1). In the following sections, I analyze some
examples that fall under each of the categories of parameter 1.
4.1 Detached NP refers to a discourse-old but not recently mentioned entity
In 23.6% of our examples (1b type), the RD refers to discourse-old information which had not
been introduced in the previous sentence, but at some earlier point. This fits very well with
Vallduví’s definition of tail and with function (iia) of Grosz and Ziv. (2) is an example of this
situation (the sentence with the RD is glossed in 2c) 3:
(2) a. Per fer règim, s’ha de tenir una especial voluntat. No com el Gallardo, que me’l trobo
l’altre dia i li dic: “Com estàs?”. I em diu: “Fa tres setmanes que faig règim”. Dic: “Ah,
sí? I quan has perdut?”. Diu: “Tres setmanes”. I té raó. Jo tampoc en tinc, de voluntat.
Quan faig règim, ho passo fatal.
b. If you are on a diet, you must have special willpower. Not like Gallardo. I met him the
other day and I asked him: “How are you doing?” And he answered: “I’ve been on a diet
for three weeks”. And I said: “Really? How much have you lost”. And he said: “Three
weeks”. And he’s right. I don’t have the willpower either. When I am on a diet, I have a
very bad time.
c. Jo tampoc en
tinc, de voluntat
´I neither part-pr have, of willpower’
In this example, the RD is used to refer to an entity which had been introduced at some previous
utterance. Therefore, it is a way of referring to an entity which was not accessible anymore and
make it highly salient. In these cases, the RD cannot be eliminated in 72.7% of the instances,
since it would not be clear to which entity the pronoun is referring. If the RD were not there in
the last example, the pronoun would probably be interpreted as referring back to setmanes
(weeks). Also, if we restore the canonical order, 50% of the instances are still acceptable, but
their degree of acceptability decreases. That is, since the entity had not been introduced in the
previous utterance, it is possible to refer to it with a full NP in its canonical position. However,
the sentence is less felicitous since it does not make explicit that the speaker intends to refer to a
discourse-old entity. In the last example, it would not be clear that the speaker, after telling us
about his conversation with a friend, intends to go back to the topic of having willpower to be on
a diet. Thus, the RD accomplishes two functions: unambiguously picks the entity to which the
speaker wants to refer, while marking it as a discourse-old entity.
4.2 Detached NP had not been explicitly mentioned
This category of RDs (1c type) would fulfil function (i) of Ziv and Grosz (1993): that is, either
the referent of the NP was situationally present but never mentioned or it was inferable from
another mentioned entity. (3) is an example of such a case4.
(3)
L’oració Diga trenta y tres mereix una reflexió.
‘The sentence Say thirty-three deserves some thought.’
Realment la fan servir tant, els metges?
‘Really do-pr do use much, the doctors?’
3
I use the following abbreviations: part-pr (partitive clitic pronoun), do-pr (direct object clitic pronoun), io-pr
(indirect object clitic pronoun).
4
There is no clitic in the main clause because Catalan has no subject clitics. However, (3) has a clearly different
intonation pattern than a sentence with a postverbal subject.
37
Laia Mayol
“Do doctors really use it that much?”
After mentioning the sentence “Say thirty-three” (which is stereotypically used by doctors when
they want to examine their patients throats), the entity the doctors is directly inferable and,
therefore, detachable. Since the entity to which the RD refers had not been previously been
mentioned, if the RD is eliminated, the sentence is no longer acceptable (this is the case for
93.3% of the instances in our corpus). The canonical order is, in these cases, either acceptable
(40%) or dubious (46.7%). However, the canonical order fails to make explicit the link between
the displaced entity and the situation or the other entity from which the RD is inferable.
4.3 Detached NP had been mentioned in the previous sentence
In most instances of our corpus (60.7%), the detached NP refers to an entity which had been
mentioned in the previous utterance of the discourse (1a type). According to Ziv and Grosz
(1993), this use is proscribed in English, unless the RD adds some attributive meaning. There are
four examples in our corpus in which the RD seems to fulfil this condition or some similar
condition (rather than adding attributive meaning, the RD is used to refer to the same entity using
a different noun, therefore characterizing it in a different way):
(4) (during a power outage) Les famílies de sobte s’han de parlar. Mires els teus germans i
tenen una altra cara
“Suddenly, families must talk. You look at your brothers and you almost don’t recognize
them”
Jo què li
dic, a aquest tio?
‘I what io-pr say, to this guy?’
“What can I say to this guy?”
In this case, the RD is a good strategy to refer back to a discourse-old entity (your brothers) with
another NP, which can provide additional meaning. In this case, it provides a humorous way to
refer back to the same entity. However, this explanation does not account for the great majority
of cases, in which the displaced entity had just been mentioned, as the example in (5) shows.
(5) A: A mi el Phil Collins em sembla una mica hortera
“I find Phil Collins a little bit tacky”
B: Doncs a mi m’agrada, el Phil Collins
well to me please, the Phil Collins
“Well, I like him/Phil Collins”
In this instance, the dislocated NP (Phil Collins) refers to an entity just mentioned in the previous
sentence. The detached NP does not add any attributive meaning to the referred entity and, still, is
perfectly natural. Therefore, either the attributive meaning restriction does not hold for Catalan or
the notion of “attributive meaning” is different for both languages and must be redefined. Also,
these RDs do not fit very well into Vallduví’s notion of tail, since the RD conveys information
which is salient at the moment of utterance. Also, this is the type of RD which seems less
necessary: if the RD is removed, the sentence is still felicitous (32.1%) or dubious (42.9%).
However, in all examples involving a RD that refers to an entity just mentioned, the RD does
indeed add some meaning, some “emotional content”. For instance, in the previous examples, it
helps to make explicit the contrast between A and B’s opinion regarding Phil Collins. The
sentence would also have been felicitous only with the pronoun, but the contrast would not have
been as forcefully stated. It is also worth noting that this kind of RD appears frequently in
38
The discourse function of right-dislocation in Catalan
interrogative (23.2%) and exclamative sentences (12.5%), in which the speaker tends to express a
higher degree of subjectivity. Lambrecht (1981) also claims that some examples of RD in French
transmit a certain feeling of “camaraderie”: a stronger link between speaker and hearer is
conveyed, since the speaker is making it clear that there is some information that both know. If
the RD is omitted, this link disappears.
Therefore, in Catalan, an element mentioned in the previous utterance is a candidate to be rightdislocated, not only if the RD adds some attributive meaning, but also if the speaker wants to
convey some additional emotive content: emphasis, opposition or camaraderie. Unlike the other
two types, this type of RD is not strictly necessary to signal how the information must be
updated. However, by doing so, the speaker signals to the hearer that he wishes to communicate
some extra meaning. We could think of this process as a type of implicature: the speaker could
have used a simpler structure (i.e. could have omitted the RD, since the referent of the pronoun is
perfectly clear), but has chosen not to do so, because he wants to convey an extra meaning5.
5. Conclusion
To sum up, I argue that RD in Catalan is clearly a means to structure information in a coherent
way by displacing old information from the main clause. It is not as constrained as in English and
it can appear in broader contexts than those described by Ziv and Grosz (1993). We can find
three main kinds of NPs in the detached phrase, which have different discourse functions:
(1) NPs referring to an entity mentioned some utterances back. In this case, the RD is a necessary
strategy to accomplish two tasks: to activate an entity which was not accessible anymore and
make it highly salient, while still marking its discourse-old status.
(2) NPs referring to entities never textually mentioned, but which were implicitly present in the
situation. As in (1), the RD also has a twofold function: it makes explicit the implicit referent and
it places it in a discourse-old information position.
(3) NPs referring to entities mentioned in the previous sentence. In Catalan, most cases of RD fall
under this category and only a small part add some attributive meaning. However, such RDs do
convey an additional meaning, some “emotional content”, having to do mainly with expression of
opposition or emphasis.
References
Gundel, Jeanette, Nancy Hedberg and Ron Zacharski (1989). “Givenness, implicature, and
demonstrative expressions in English discourse”. Chicago Linguistics Society. 89-103.
Lambrecht, Knud. (1981). Topic, antitopic and verb agreement in non-standard French.
Amsterdam/Philadelphia: Benjamins.
Vallduví, Enric. (1992). “Detachment in Catalan and information packaging”.
Pragmatics 22, 573-601.
Journal of
Ziv, Yael and Barbara Grosz (1993). “Right Dislocation and Attentional State”. In: Buchalla,
Rhobna and Anita Mittwoch (ed.). The Proceedings of the Ninth Annual Conference Ben
Gurion University of the Negev 1993 Jerusalem: Akademon. P. 184-199.
5
Gundel et. al. (1990) also invoke implicatures to explain unexpected uses of referring expressions: i.e. definite noun
phrases used in situations where the referent is in focus and a pronoun would have been sufficient.
39
SPRIK Conference
Explicit and Implicit Information in Text - Information Structure across Languages
University of Oslo, June 8-10 2006
Investigating nominal coreference in originals and translations
Kerstin Kunz
Saarland University
[email protected]
The present paper deals with the elaboration of a method for analysing various types of nominal
coreference in originals and translations. We focus on designing a model for the manual
annotation of coreferential noun phrases to capture differences on the lexicogrammatical level as
well as on a more semantic and conceptual level. Note that, in this paper, we concentrate on the
analysis of German and English texts only.
(1) A relation of coherence and cohesion
Coreference is a strategy of text creation and text reception linking pieces of text in a coherent
way. It goes beyond the text surface as it is not only a relation between linguistic elements but
also a relation of meaning - a cognitive connection between mental concepts. The recipient of a
text establishes this connection not only on the basis of linguistic expressions in the text, but also
by inferring knowledge stored in cognitive systems other than the language system, e.g. world
knowledge. The relation of coreference consists in identity of reference as the same cognitive
concept is activated and reactivated. However, this mental relation is evoked by linguistic
expressions and cohesive devices which create coreference on the text surface in a more or less
explicit way. Nominal coreference therefore constitutes both, a textexternal and a textinternal
relation:
•
textexternal with two or more noun phrases in a text referring to the same cognitive or
extralinguistic concept
•
textinternal with a directional pointer being set by a cohesive marker from one coreferring
noun phrase to another
(2) Types of nominal coreference
Consider the following example of succeeding noun phrases as the may appear in a text:
A blue Porsche .... The car ..... it.... The Porsche
In the example, an extralinguistic concept or referent is activated for the first time by the noun
phrase A blue Porsche - the antecedent. The same referent is reactivated by several other noun
phrases ('The car, it, The Porsche) in the succeeding text – the anaphors. Personal pronouns, such
as it can be stated as a typical and very common cohesive device to create identity of reference.
However, as the preceding example displays, the system of a language provides other cohesive
devices like hyperonymy, repetition, etc. The occurrence and use of these various types in
specific texts may not only be instantiations of the language system but may also depend on the
accessibility of the respective referents: different types of nominal coreference may indicate
differences in information status (e.g. discourse old or discourse new) and attentional state (e.g.
topic or focus). Referent accessibility not only may affect form and semantics of the coreferring
41
Kerstin Kunz
noun phrases but also their position and function in the sentences as well as in the text. In
addition, the realization of reference identity via nominal coreference may also be a question of
register.
(3) Nominal coreference in translation
Creating identity of reference constitutes a fundamental strategy to establish meaning in a given
text as it provides thematic continuity. We consider the creation of coreference relations to be an
essential part of the translation process as they help to create connectivity between textual
elements in original and translation, and at the same time, retain the meaning of an original in its
translation. Relations such as identity of reference on the mental level, which, in the original are
expressed by cohesive devices in the source text language have to be preserved in its translation
expressing them by adequate cohesive devices in the target text language. More precisely, the
translator has to:
•
identify noun phrases as referring expressions in the source text and identify cohesive
devices indicating relations between these noun phrases
•
identify reference markers as well as position and function of coreferring noun phrases
signalling the accessibility of referents
•
interprete the relation between the referring expressions to be identity, e.g. establish a
mental relation of identity of cognitive concepts
•
indicate referent accessibility by form, function and position of coreferring noun phrases
in the target text
•
transform the relation of sense into wording, e.g. express identity of reference by
coreferring noun phrases in the target text language to create thematic continuity
(4) Analysing coreference in translations and originals
In order to analyse nominal coreference in originals and translations with corpuslinguistic
methods we have to concentrate on how the same cognitive relation of identity is realized by
noun phrases and cohesive markers within these noun phrases in the original and translation,
respectively. Furthermore, we have to investigate if and how referent accessibility is expressed
differently on the surface of source and target text. The sentences below, extracted from a
German original and its English translation serve as an example. As the example displays, in the
German source text the coreference relation is expressed by several noun phrases, which form a
coreferential chain (underlined noun phrases). In the English translation, a coreferential chain is
set up for the same extralinguistic concept but differs in number of anaphoric noun phrases and
realization via lexicogrammatical elements within these noun phrases. In few cases, the
coreferring noun phrases in original and translation also differ in their syntactic function and
position. The example below can hardly reflect the network of different coreferential chains
being set up in a complete text, nor does it display all possible varieties in form, semantics,
position and syntactic function of coreferring noun phrases in source text and target text
language.
42
Investigating nominal coreference in originals and translations
Die Sojabohne zählt zu den bekanntesten
Nutzpflanzen der Erde. Kaum eine andere
Pflanze ist so reich an Wirkstoffen wie sie. Die
Pflanze gehört als Hülsenfrucht zur Familie der
Schmetterlingsblütler und stammt ursprünglich
aus dem nordöstlichen Asien. In China und
Japan haben Anbau und Verarbeitung der
Sojabohne seit Jahrtausenden Tradition. Sie
wird dort als eines von fünf "heiligen
Getreiden" gepriesen und ist die Basis für eine
Vielzahl unterschiedlicher Lebensmittel. Soja
genießt in Asien in etwa den gleichen
Stellenwert wie Fleisch und Kuhmilch in
Europa.
The soybean is one of the most well-known
agricultural products on earth with more
beneficial components than virtually any other
plant. Soybeans are legumes belonging to the
pea family and originally came from northeast
Asia, where they have traditionally been
cultivated and processed for thousands of years
in countries like China and Japan. Here they
are praised as one of the five "holy grains" and
serve as the basis for a variety of different
foods. In Asia, soybeans enjoy approximately
the same status as meat and dairy products in
Europe.
(4) Model for an annotation scheme
In order to track down as many differences and commonalities in nominal coreference as possible
our annotation scheme, which is designed for the manual annotation of corpora, is very fine
grained.
We set up the following categories and subcategories in order to differ types of anaphora:
•
recurrence: lexical and orthographical identity between anaphor and antecedent (total or
partial recurrence)
•
pronominal relations: the anaphor is a personal, possessive, demonstrative or relative
pronoun
•
is-a relations: the anaphor constitutes a synonym, a hyperonym, or a hyponym of its
antecedent
With the following category, we intend to differ items indicating the type of reference
•
reference marker: definite, indefinite and zero article, pronoun, possessive determiner,
demonstrative determiner
We analyse the noun phrase by establishing the following categories:
•
modification: to differ pre, post and zero modification
•
noun phrase head: to indicate the level of embedding of a subordinate anaphoric noun
phrase
•
number: to differ plural and singular of the noun phrase
•
noun phrase type: to differ noun phrases referring to generic or specific referents
•
The category function looks into the syntactic function of coreferring noun phrase
differing subject, different types of objects, complement adverbial, predicator, etc.
•
The categories positionGER and positionEN analyse the position of coreferring noun
phrases in the sentence, considering differences in German and English constituent
structure.
43
Kerstin Kunz
Apart from these classifications, we number all coreferring noun phrases and assign each
coreferring noun phrase to its respective coreferential chain in the text. Finally, we align a
coreferring noun phrase in the translation to the corresponding coreferring noun phrase in the
original.
(5) Conclusions and outlook
The classifications described above serve as a scheme for the manual annotation of nominal
coreference e.g. with an XML editor. They help us to trace and compare coreferential chains
referring to the same concept in originals and translations. For instance, the number of noun
phrases in a coreferential chain may vary in original and translation. In addition, we may detect
differences resulting from various modifications on the lexicogrammatical level e.g. different
cohesive markers, different semantics of the head noun, different types of modification in the
coreferential noun phrases, or differences in position and syntactic function of the coreferring
noun phrases. We may also capture changes due to the more or less explicit marking of
coreference on the text surface. These shifts in nominal coreference on the text level can cause
shifts of coreference on the semantic and cognitive level: They may affect referent accessibility
or cause the creation of new or different coreferential chains. Eventually, this may result in
changes in the coherence and thematic continuity of the translation in comparison to its original.
References
Blum-Kulka, S. 1986. Shifts of Cohesion and Coherence in Translation. In: J. House & S. BlumKulka (eds.), Interlingual and Intercultural Communication. Tübingen: Narr.17-35
Gundel, J., N. Hedberg & R. Zacharski. 1993. Cognitive Status and the Form of Referring
Expressions in Discourse. Language, 69/2: 274-307
Halliday, M.A.K. & R. Hasan. 1976. Cohesion in English. London: Longman
House, J. 1997. Translation Quality Assessment: A Model Revisited. Tübingen: Narr.
Martin, J.R. 1992. English Text: System and Structure. Amsterdam: Benjamins
Steiner, E. 2004. Translated Texts: Properties, Variants, Evaluations. Frankfurt/M. etc.: Lang
Strube, M. & U. Hahn. 1999. Functional Centering: Grounding Referential Coherence in
Information Structure. Computational Linguistics 25/3: 309-344
Grosz, B. J. , A. K. Joshi & S. Weinstein. 1995. Centering: A framework for modeling the local
coherence of discourse. Computational Linguistics, 21: 203-225.
44
SPRIK Conference
Explicit and Implicit Information in Text - Information Structure across Languages
University of Oslo, June 8-10 2006
Cohesion and explicitation in an English-German translation corpus1
Silvia Hansen-Schirra, Stella Neumann, Erich Steiner
Saarland University, Saarbrücken, Germany
[email protected], [email protected], [email protected]
In translation studies, cohesive features as indicators for explicitation have been analysed either
in an example-based way (Blum-Kulka 1986) or as concordances in monolingually comparable
corpora of raw text (cf. several contributions in Laviosa (ed.) 1998, Olohan & Baker 2000). In
spite of the insight gained from this line of research, we argue that where explicitation is
investigated without taking into account the source texts, the interpretation of results remains
restricted and problematic. Work on translations against a more linguistic background has
addressed some of these restrictions and problems (cf. relevant work as in Johansson & Oksefjell
(eds.) 1998, Fabricius-Hansen 1999); the focus of these research interests and methodologies is
however different from, and partly complementary to, ours with respect to corpus architecture,
querying techniques and underlying linguistic modelling (for which cf. Hansen 2003, Neumann
2003, Steiner 2001, 2005a,b,c, Teich 2003 ).
The basic assumption for the analysis of explicitation in the present paper is that the element
explicitated in the target text has to be present implicitly in a linguistically traceable way in the
source text and vice versa for implicitated elements. Explicitation is thus defined as a relationship
and a process between instantiated and aligned pieces of translated texts. Furthermore, we stratify
the notion of explicitation according to the linguistic levels of lexicogrammar (not in focus in this
paper) and cohesion. As this stratification is still too abstract to be directly quantifiable on
linguistic data in an electronic corpus, a series of further micro-level operationalisations is
undertaken which are meant to bring the relevant phenomena down to an empirically accessible
level.
Our investigation of explicitation and implicitation of cohesion markers in translations is based
on a cross-linguistic corpus containing statistically meaningful and representative samples (cf.
Biber 1993) of German and English registerially parallel texts from 8 registers annotated with
parts of speech, morphology, phrase structure and grammatical functions. In addition to these two
sub-corpora, two further sub-copora have been compiled consisting of translations of the samples
from the first two sub-corpora into the respective other language, yielding 4 sub-copora. The
overall corpus comprises 1 million words (approx. 250 000 for each of the four sub-corpora) plus
68,000 words in register-neutral (cross-register) reference corpora in both languages. A
characteristic feature of our corpus is the alignment of source and target texts on different
linguistically motivated layers: we not only align sentences (which is state of the art in
Translation Memories; e.g. Johansson et al. 1996) and words (which is state of the art in Machine
Translation; cf. Och & Ney 2003) but also clauses.
1
The authors would like to thank the reviewers for helpful comments on an earlier version. The research reported on
here is sponsored by the German Research Foundation (DFG) as project no. STE 840/5-1. For current information cf.
http://fr46.uni-saarland.de/croco/
45
Silvia Hansen-Schirra, Stella Neumann & Erich Steiner
One of the methodological principles for the compilation of the resource is the distinction
between lexico-grammatical/ cohesive annotation of source and target language texts (including
the alignment) on the one hand, and the interpretation of the data in view of more abstract
concepts like “explicitation” on the other. This distinction allows us to pose queries on
(combinations of) lower level linguistic features assumed to be indicators of the more abstract
concept. One technical precondition for the comprehensive analysis of the corpus is the use of
XML stand-off mark-up as representation format for annotation and alignment. This is necessary
because we annotate the corpus on different layers, thus keeping the annotation and alignment of
overlapping and/ or discontinuous units in separate files. Thus it becomes possible to view the
annotation in aligned segments and to pose queries (using XSLT and XQuery) combining
different layers (cf. Neumann & Hansen-Schirra 2005, Hansen-Schirra et al. 2006). The resource
thus permits the analysis of a wealth of linguistic information on each level helping us to
understand the interplay of the different levels and the relationship of lower level features to more
abstract concepts such as explicitation.
In what follows we will exemplify the queries possible on the basis of the annotation and
alignment for the cohesion markers described by Halliday & Hasan (1976) and their equivalents
for German. It is, for instance, a straightforward step to retrieve (co-)reference markers separately
from the source and target language corpora. The part-of-speech information contained in the two
corpora permits precise queries. Specific queries into these reference markers in the target texts
which have no equivalent in the source texts are more complex. However, they address more
linguistically meaningful levels of encoding than merely string-based queries which are unable to
retrieve information encoded on higher linguistic levels.
1. to 7. below are examples of hypotheses about cohesion to be tested on the data: For either a
given pair of non-aligned text segments globally, or else for a given aligned source – target
fragment of two texts in a translation relationship, we expect differences along the following
parameters:
1. the proportion of explicit to implicit referents;
2. the proportion of phoric to fully lexical (auto-semantic) phrases;
3. the number of newly introduced discourse referents per discourse segment;
4. the amount of cohesive ellipsis and substitution;
5. the strength of lexical chains as measured by various ratios between content and function
words, and as measured by type-token relationships;
6. the strength (internal connectivity) of lexical chains as measured by average number of items
per lexical chains;
7. the ratio between explicit and implicit encoding of conjunctive relations.
Observe that in comparing any text fragments which are not in a unit-of-translation-relationship,
as in our registerially parallel sub-corpora of originals, we are testing for the global property of
(relative) explicitness. However, whenever we are comparing a specific aligned and instantiated
source-target (translation) unit, we are testing “explicitation” (or its opposite, implicitation).
An indicator for the first hypothesis mentioned above could be the proportion of explicit
(pronominal) referents vs. implicit ones in comparing German relative clauses with their nonfinite English correspondences. The relevant evidence is reflected in the annotation and
alignment at word level. Relative pronouns receive the part-of-speech tag prels and if they occur
in both languages, they are linked to each other. Is there, however, a relative pronoun in the
German translation which cannot be found in the English original text, the German relative
46
Cohesion and Explicitation in an English-German Translation Corpus
pronoun is not aligned at all - it receives a so-called empty link. Figure 1 shows the XML
representation of the tokenised corpus, the part-of-speech tagging and the alignment on word
level. In the token index file each token is assigned an index number. The part-of-speech
annotation of tokens refers back to this index file via xlinks specifying the index number of the
respective token. The word alignment then refers to the index numbers of both source and target
token in turn. In cases of empty links, they receive the value undefined.
English original:
German translation:
... a palmist, inferring the future out of his own lined flesh
... ein Handleser, der seine Zukunft aus den eigenen Linien ableitete
( a palmist who his future out-of the own
lines inferred )
token index file
<token id="t64" strg="ein"/>
<token id="t65" strg="Handleser"/>
<token id="t66" strg=","/>
<token id="t67„ strg="der"/>
<token id="t68" strg="seine"/>
<token id="t69" strg="Zukunft"/>
part-of-speech annotation
<token pos="art" xlink:href="#t64"/>
<token pos="nn" xlink:href="#t65"/>
<token pos="yc" xlink:href="#t66"/>
<token pos="prels" xlink:href="#t67"/>
<token pos="pposat" xlink:href="#68"/>
<token pos="nn" xlink:href="#t69"/>
word alignment
<token> <align xlink:href="#t55"/>
<align xlink:href="#t66"/> </token>
<token> <align xlink:href="#t56"/>
<align xlink:href="#undefined"/> </token>
<token> <align xlink:href="#tundefined"/>
<align xlink:href="#t67"/> </token>
Figure 1: XML corpus annotation and alignment on word level including empty links
For the investigation of explicit pronominal referents in German relative clauses vs. implicitly
encoded English referents, all German tokens with the part-of-speech tag prels (for relative
pronoun) have to be extracted which are not aligned on the word level (since the pronominal
reference is encoded in the English participle). The respective XQuery is shown in Figure 2.
for $k in $doc//tokens/token
let $fileName := $doc//translations/translation[@n='1']/@trans.loc
let $fileNameNew := replace($fileName,"tok","tag" )
where ($k/align[1][@xlink:href != "#undefined"] and $k/align[2]
[@xlink:href = "#undefined"] and doc($fileNameNew)//token
[@xlink:href eq $k/align[1]/@xlink:href][@pos eq "prels"])
Figure 2: XQuery for relative pronouns with empty links
The output of this query are sentences like the ones displayed in Figure 1. This example (taken
from the fiction sub-corpus) is interpreted as explicitation since participant role (and thus the
reactivation of the referent), tense and mood are explicitly realised in the finite relative clause of
the German translation, whereas they are implicit in the English original.
<result no="13"><ori_en>Baker Hughes Business Support Services has assumed accounting, payroll, benefits and IT
support duties for many of the company's U.S. operations, eliminating duplicate efforts by division personnel. </ori_en>
<trans_ge>Baker Hughes Business Support Services hat die Buchführung, Gehalts- und Sozialleistungen sowie ITAufgaben für viele Niederlassungen des Unternehmens in den Vereinigten Staaten übernommen, wodurch doppelte Arbeit
durch das Personal in den Tochterunternehmen vermieden werden konnte. </trans_ge></result>
<result no="14"><ori_en>In this environment, Baker Hughes revenue declined 22% to $4.5 billion for 1999, compared to
$5.8 billion in 1998. </ori_en>
<trans_ge>Vor diesem Hintergrund sanken die Umsatzerlöse von Baker Hughes im Jahre 1999 um 22% auf 4,5 Mrd.
Dollar, während sie 1998 noch 5,8 Mrd. Dollar betragen hatten. </trans_ge></result>
Figure 3: Results for conjunctions with empty links
A similar query could be posed for conjunctive relations. Here, all German tokens with the partof-speech tag kous (for conjunction) are to be extracted which are not aligned on word level
(since the conjunctive relation is encoded implicitly, for instance through a participle clause). The
results for this query displayed in Figure 3 are taken from the sub-corpus of shareholder
information. Here, all examples show explicitation in the German translations, since the implicit
47
Silvia Hansen-Schirra, Stella Neumann & Erich Steiner
conjunctive relation encoded in the English participles (marked in bold face) are translated
explicitly with German conjunctions.
Another interesting phenomenon is ellipsis in translations, since it is a direct indicator for
implicitation. The alignment of our corpus on different layers enables us to find ellipsis through
empty alignment links on all alignment levels. On the basis of these empty links, we can
furthermore investigate in which syntactic functions ellipsis occurs, whether it prefers finite or
non-finite clauses or how it is dealt with in different translation directions.
The long-term aim of the present study is to identify, count and interpret cohesion markers and
lexicogrammatical markers (not addressed in this paper) and their quantitative (co-)occurrences
and patterns found in translations and their source texts as indicators of explicitation/
implicitation. This will be interpreted against the background of three sources of explanation:
language typology, text typology and the translator’s language processing (cf. Steiner 2001,
Hansen 2003, Neumann 2003, Teich 2003). Our methodologically motivated separation of
linguistic annotation/ alignment from their interpretation in pursuing a research question (in our
case explicitation) makes the corpus resource flexible enough to allow research into other
phenomena of interest in connection with translation, such as simplification, normalisation,
levelling-out (Baker 1996), culturally determined preferences (House 2002), shining through (cf.
Teich 2003), density and directness (cf. Steiner 2005 a,b,c). Beyond properties of translation the
resource opens up new research perspectives on basic questions of translation studies like the
translation unit.
In an overall perspective, we are working towards constructing and making available our corpus
as a resource, which is theory-neutral and should be usable as an empirical basis for research
informed by different theories and models. At the same time, research in our own group is
corpus-based, but not corpus-driven, in the sense that our research questions and hypotheses do
not “emerge” out of the data, but are derived from a range of theories and models about language,
texts, and translations. On the basis of the research design presented here we hope to be able to
report on some initial empirical findings relating to our hypotheses by the time of the talk.
Returning, finally, to a comparison of our approach with earlier investigations of “explicitation”
in translation studies (as in work by Blum-Kulka 1986, Baker 1996, Laviosa 1998, Olohan and
Baker 2000, Englund-Dimitrova 2005), we are aiming at significant progress towards closing the
methodological gap between high-level notions, such as “explicitness” or “explicitation” on the
one hand and levels of linguistic encoding in our data on the other, while not resorting to heavily
interpretative example-based examinations of individual cases.
References
Baker, M. 1996. Corpus-based Translation Studies: The challenges that lie ahead. In: Somers, H.
(ed.), Terminology, LSP and Translation Studies in Language Engineering. Amsterdam/
Philadelphia: John Benjamins. 175-186.
Biber, D. 1993. Representativeness in Corpus Design. Literary and Linguistic Computing 8/4:
243-257.
Blum-Kulka, S. 1986. Shifts of cohesion and coherence in Translation. In: House, J. & BlumKulka, S. (eds.), Interlingual and Intercultural Communication: Discourse and Cognition in
Translation and Second Language Acquisition Studies. Tübingen: Narr. 17-35.
Englund-Dimitrova, B. 2005. Expertise and Explicitation in the Translation Process.
Amsterdam: John Benjamins.
48
Cohesion and Explicitation in an English-German Translation Corpus
Fabricius-Hansen, C. 1999. Information packaging and translation: Aspects of translational
sentence splitting (German – English/ Norwegian). In: Doherty, M. (ed.), Sprachspezifische
Aspekte der Informationsverteilung. Berlin: Akademie-Verlag. 175-214.
Halliday, M.A.K. & Hasan, R. 1976. Cohesion in English. London: Longman.
Hansen, S. 2003. The Nature of Translated Text. An interdisciplinary methodology for the
investigation of the specific properties of translations. Saarbrücken: Saarbrücken
Dissertations in Computational Linguistics and Language Technology. vol. 13.
Hansen-Schirra, S.; Neumann, S. & Vela, M. 2006. Multi-dimensional Annotation and
Alignment in an English-German Translation Corpus. In: Proceedings of the 5th Workshop on
Multi-dimensional markup in NLP. Trento. 35-42.
Hasselgard, H.; Johansson, S.; Behrens, B. & Fabricius-Hansen, C. (eds.) 2002. Information
Structure in a Cross-Linguistic Perspective. Amsterdam, New York: Rodopi.
House, J. 2002. Maintenance and convergence in translation – some methods for corpus-based
investigations. In: Hasselgard et al. (eds.) 2002. 199-212.
Johansson, S.; Ebeling, J. & Hofland, K. 1996. Coding and Aligning the English-Norwegian
Parallel Corpus. In: Aijmer, K.; Altenberg, B. & Johansson, M. (eds.), Papers from
Symposium on Text-based Cross-linguistic Studies. Lund. 87-112.
Johansson, S. & Oksefjell, S. (eds) 1998. Corpora and Cross-linguistic Research. Amsterdam:
Rodopi.
Laviosa, S. (ed.). 1998. Meta Translators Journal. vol. 43 no.4.
Neumann, S. (2003) Die Beschreibung von Textsorten und ihre Nutzung beim Übersetzen.
Frankfurt etc.: Peter Lang Verlag.
Neumann, S. & Hansen-Schirra, S. 2005. The CroCo Project. Cross-linguistic corpora for the
investigation of explicitation in translations. In: Proceedings from the Corpus Linguistics
Conference Series, vol. 1, no. 1, ISSN 1747-9398.
Och, F. J. & Ney, H. 2003. A Systematic Comparison of Various Statistical Alignment Models.
Computational Linguistics, vol. 29, no. 1: 19-51.
Olohan, M. & Baker, M. 2000. Reporting that in Translated English. Evidence for Subconscious
Processes of Explicitation? Across Languages and Cultures 1(2): 141-158.
Steiner, E. 2001. Translations English - German: investigating the relative importance of
systemic contrasts and of the text-type "translation". SPRIK-Reports no.7. Oslo.
Steiner, E. 2005a. Some properties of texts in terms of ‘information distribution across
languages’. Languages in Contrast 5:1 (2004-2005): 49-72.
Steiner, E. 2005b. Some properties of lexicogrammatical encoding and their implications for
situations of language contact and multilinguality. In: Franceschini, R. (ed.), Zeitschrift für
Literaturwissenschaft und Linguistik Jahrgang 35, Heft 139. Stuttgart: Metzler Verlag. 54-75.
Steiner, E. 2005c. Explicitation, its lexiocogrammatical realization, and its determining
(independent) variables – towards an empirical and corpus-based methodology. SPRIKreports no. 36. Oslo.
Teich, E. 2003. Cross-linguistic variation in system and text: a methodology for the investigation
of translations and comparable texts. Berlin: Mouton de Gruyter.
49
SPRIK Conference
Explicit and Implicit Information in Text - Information Structure across Languages
University of Oslo, June 8-10 2006
Connectors in a cross-linguistic perspective
Henning Nølke
University of Aarhus
[email protected]/[email protected]
This paper will take up the following main questions:
•
Can connectors be properly translated?
•
How should connectors be described in a bilingual dictionary?
•
Can Contrastive Network Analysis (Nølke 1995) be used both as a heuristic tool and for
language-specific analyses?
I shall begin by briefly introducing connector grammar (Nølke 2001). Connectors are defined as a
functional category. They are characterised by having a structural function mainly concerned
with scope properties and a logical or semantic function that works upon semantic units called
connector arguments. Both functions are strongly context sensitive. Prototypical connectors have
two scopes: one at the left and one at the right. Each connector yields specific instructions about
the type of scope and how to assign it. Figure 1 gives an illustration of this process:
Figure 1
X
p
CON
Y
q
X and Y symbolise the left scope and the right scope, respectively, and p and q the arguments that
result from semantic interpretations of X and Y, respectively. P and q are used by the semantic
function, which works through specialisation and combination, forming a new and more complex
meaning from the argument meanings. To a certain extent, these basic functions also depend on
the grammatical properties of the connectors. Items from different word classes may indeed
function as connectors. For example, connector adverbs behave differently from conjunctions.
This means that we should develop a methodological framework making it possible to duly
account for this complex interaction, which is why I adopt a modular approach (Nølke 1999).
In a cross-linguistic perspective we may note that completely analogous connectors barely exist
across languages. Therefore, their complex semantic nature, and in particular their context
sensitivity, makes translation, or even simple comprehension, extremely difficult. Connectors
thus constitute a huge challenge to contrastive linguistics. Applying the descriptive tools offered
by connector grammar, first I shall describe the basic properties of the two French connectors
donc and or and illustrate the analysis with some examples taken from Flaubert’s novel Madame
Bovary. I have chosen this text partly because it contains rather complex examples allowing us to
appreciate the complexity of the connector functions and partly because it has been translated
into many different languages. I shall then discuss some of the proposed translations in an
attempt to pinpoint what has been changed – lost or added – during the translation process. I will
take a closer look at two different translations into Danish that show very clearly how the
translator often has to make a choice because she cannot transfer all the subtle nuances conveyed
by the French connector.
51
Henning Nølke
This leads us to address the first two questions mentioned above. Why is it that translations into
the same language can differ so much and still all be considered acceptable? I shall argue that
connectors bring bundles of complex instructions concerning their functions. These bundles may
be described as function domains. The translator then looks for connectors in the target language,
L2, that create function domains as close as possible to the ones she finds in the L1 text. This is
why translation is possible, but as is well known, it is never completely equivalent. But if this is
so, how can dictionaries or linguistic analysis be an aid? Traditional dictionaries typically give a
list of possible corresponding connectors and a set of examples. Due to the high context
sensitivity of connectors, this list can never be exhaustive and therefore this procedure demands a
lot of the user’s competence and imagination. In fact, the advanced translator would rather avoid
the use of a dictionary. Hence the question is whether linguistic analysis and description help.
My hypothesis, which I have developed in past work, is that one should look for some basic
abstract instructions attached to each connector which anticipate the emergence of the function
domains by anticipating context influence. The function domains are thus conceived of as
pragmatic effects (effets de sens). In other words, the function domains are highly dependent
upon the actual context to which the connector instructions are applied. Since the context
contribution also varies between languages, there may be examples where the target context can
be judged to contain in itself most of the instructions attached to the source connector. These
considerations may explain why experienced translators sometimes choose simply to omit the
connector.
Though the linguist’s ultimate objective is to formulate abstract instructions supposed to be part
of the language system, a first step has to be to circumscribe the possible function domains
attached to each connector. To this end, we need some analytical tools. I shall argue that
Contrastive Network Analysis may help. Combined with a study of authentic translations, it may
comprise an excellent starting point for proposing genuine explanatory hypotheses anticipating
the emergence of the observed function domains.
The Contrastive Network (CN) is a heuristic device that is supposed to give us hints as to how to
find L2 equivalents of L1 items. It is a “to-and-fro” translation system. First we single out the L2
translations of the L1 words and phrases under scrutiny. Each of these L2 translations is then
translated back to L1. The resulting items are then translated to L2 and so on. This procedure may
go on indefinitely. The CN gives us a kind of semantic field which, in our case, may be regarded
as representing the function domain. In fact, typically some words and phrases will reoccur
whereas others will only appear sporadically. The following French-Danish CN quoted from
Nølke (1995: 315) illustrates this process:
Figure 2
52
Connectors in a cross-linguistic perspective
I shall briefly discuss some important questions raised by the use of CN analysis: What is the
status of the CN? How should we present it? How should we work it out? Regardless of our
answers to these questions, it is clear that the CN has to be applied with much caution.
Subsequently, I shall present the results of an application of French-Danish analysis to the two
connectors under scrutiny: donc and or. Twenty-five small text fragments were translated by
advanced students at the Aarhus Business School (which has a school of translation) and at the
University of Aarhus. Every text was translated to and fro five times by different groups. About
two-thirds of the texts were taken from Madame Bovary and one-third were non-literary texts. No
group saw the same text twice. On the basis of this material combined with the authentic
translations mentioned above, I dressed partial CNs as well as a global one. Afterwards the
results were evaluated against the individual contexts that gave rise to different paths in the
network, in order to see whether it is possible to find systematic contextual influence.
On the basis of these analyses I tried to ascribe function domains to each of the two connectors.
This procedure may be seen as the first step towards a proper system linguistic semantic
description insofar as it suggests the pragmatic effects (effets de sens) that the instructions coded
in the language system (langue) should be able to anticipate.
The last step will be to take up former analyses proposed for our two connectors. Whereas donc
has been submitted to a large number of such linguistic studies,1 this is not so for or.2 I shall try
to show that the adopted analytical method yields an excellent basis for scrutinising and
evaluating different descriptions suggested by my predecessors.
Finally, I shall present some conclusions in an attempt to answer the main questions posed in this
paper. I shall argue that it is possible to find acceptable translations, but hardly equivalent ones.
For instance, even apparently “elementary” connectors like the English but turn out to be
different from the French mais in several respects.3 This also means that dictionaries will never
work ideally. Are there any solutions? Two possible methods present themselves. We might have
recourse to descriptions by means of abstract forms or by means of function domains elucidated
by examples. Both methods are far from perfect. Abstract forms are generally difficult to
interpret and hence of limited use, and a list of functional domains will encounter the same
problem as traditional dictionary descriptions: they will never be exhaustive. I believe, however,
that a suitable mix of the two methods, possibly combined with the traditional one, might pave
the way for the creation of better lexical descriptions.
In answer to the last main question addressed in this paper, I shall argue that Contrastive Network
Analysis as well as authentic translations do function as heuristic devices, although one should be
extremely careful when interpreting the results. It would be very interesting to scrutinise the
methodological questions that these methods raise, but this must wait for future research.
1
E.g., Jayez & Rossari (1996), Rossari et al. (2004) and Zenone (1981).
See however XXX
3
Compare, e.g., Nyan (1998) and Blakemore (1987).
2
53
Henning Nølke
References
Blakemore, Diana. 1987. Semantic Constraints on Relevance. Oxford: Basil Blackwell.
Jayez, Jacques & Rossari, Corinne. 1996. Donc et les consécutifs. Des systèmes de contraintes
différentiels. Linguisticae Investigationes XX. 117-143.
Nyan, Thanh. 1998. Metalinguistic Operators. Bern: Peter Lang.
Nølke, Henning. 1995. Contrastive and argumentative linguistic analysis of the French
connectors donc and car. Leuven contributions in linguistics and philology 84/3: 313-328.
Nølke, Henning. 1999. Linguistique modulaire : principes méthodologiques et applications. In :
Nølke, Henning & Adam, Jean-Michel (eds.), Approches modulaires en linguistique : de la
phrase au discours. Lausanne/Paris: Delachaux & Niestlé. 17-73.
Nølke, Henning. 2001. Konnektorgrammatik. Ny Forskning i Grammatik 8. 191-205.
Rossari, Corinne, Beaulieu-Masson, Anne, Cojocariu, Corina & Razgouliaeva, Anna. 2004.
Autour des connecteurs. Réflexions sur l’énonciation et la portée. Bern : Peter Lang.
Zenone, Anna. 1981. Marqueurs de consécution : le cas de donc. Cahiers de linguistique
française 2. 113-139.
54
SPRIK Conference
Explicit and Implicit Information in Text – Information Structure across Languages
University of Oslo, June 8–10, 2006
Aspectual Influence on Temporal Relations
A Case Study of Experiential Guo in Mandarin
Jiun-Shiung Wu
National Chiayi University
[email protected]
This paper examines how the temporal relation between a clause the experiential guo and an
adjacent clause is determined. Mandarin is a language without tense (e.g. Lin 2006), and
therefore, tenses cannot help in determining the temporal relations in Mandarin. However,
Mandarin has a rich aspectual system. This paper argues that the experiential guo indirectly
influences temporal relations via rhetorical relations by either specifying a default rhetorical
relation, or by constraining which rhetorical relation can connect a clause with guo to an
adjacent clause. This paper also argues that the default rhetorical relation and the constraints
are determined by the aspectual properties of the experiential marker. Other information, such
as explicit discourse connectors, lexical information, etc, can override the default rhetorical
relation indicated by guo and specifies its own rhetorical relation. Therefore, this paper
concludes that in Mandarin aspect markers can indirectly influence temporal relations, a result
consistent with Wu’s (2005b) paper on the perfective marker le in Mandarin and Wu’s (2004)
work on the progressive marker zai and the durative marker zhe.
The semantics of the experiential guo has received much attention, e.g. Smith (1997), Yeh
(1996), Wu (2005a, 2006), etc. While they propose different semantics for the experiential
guo to explain its syntactic and semantic properties, such as recurrence, discontinuity, a class
meaning, etc, they share two points: guo presents a closed situation and this closed situation
was realized at an indefinite past. To capture the closedness property of a situation presented
by guo, Yeh (1996: 171) proposes that, inside a main DRS, the experiential guo triggers an
existential quantifier-led sub-DRS where the situation presented by guo is located. Arguing
against Yeh’s theory, Wu (2005a, 2006) proposes that guo presents an event type that was
realized at an unknown past time. Since the experiential guo presents a situation that has
completed or terminated and a situation presented by guo was realized at unknown past, the
experiential guo is hypothesized to have two effects on temporal relations:
(1) Hypothesis for the role of the experiential guo in temporal relations:
a. The event time of a situation presented by guo cannot be advanced unless its event time
is explicitly specified.
b. The internal process of a situation presented by guo cannot be accessed unless it is
forced open to be accessible.
c. A situation presented by guo cannot serve as a temporal background, which indicates
temporal overlapping.
Asher and Lascarides (2003) argue that, in addition to tenses, the temporal relations in
English are also determined by rhetorical relations. For the temporal relations in Mandarin,
there are three reasonable questions to ask. First, does the experiential guo directly determine
temporal relations? Second, if the question to the first question is negative, then do rhetorical
relations also play a role in deciding the temporal relations in Mandarin? Thirdly, to
determine the temporal relations, does the experiential guo interact with rhetorical relations in
any way?
55
Jiun-Shiung Wu
Lin (2003) and Smith and Erbaugh (2001) argue that aspectual information determines
temporal references. Therefore, it seems reasonable to assume that aspect markers also
directly determine temporal relations. However, this assumption is not accurate. See the
following examples.
(2)
a.
b.
c.
(3)
a.
b.
c.
(4)
a.
b.
!(lian nian qian) ta hui guo hunan laojia
two year before he return Exp Hunan hometown
‘Two years before, he returned to his hometown in Hunan.’
yijiuwuba nian di zaidu fan xiang shi
1958 year end again return home time
‘When he returned home again at the end of 1958,’
faxian renshiyifei
find everything change
‘(he) found that everything had changed.’
wo zhi ting guo qian xiaozhang gei huaxue xi tongxuemen
I only hear Exp Qian president for chemistry department classmates
zuo de youguan fenxi huaxue de yanshuo
make DE about analytic chemistry DE speech
‘I only heard once the speech on analytic chemistry that President Qian
delivered to the chemistry majors.’
ta yuzhongxinchangde mianli tongxue yi fan hua
he sincerely
encourage classmate one CL words
‘He sincerely encouraged the students he addressed to.’
zhi jin hai jiyiyouxin
to now still memory-fresh
‘The memory is still fresh so far.’
wo liang yiqi jingli guo wushu weinan
I both together experience Exp countless danger
‘We experienced countless danger together before.’
zhe ci huoxu ye neng taotuo
this time maybe also can escape
‘Maybe, this time, (we) can also escape.’
In the three examples above, each clause with the experiential guo has different temporal
relations with adjacent clauses. In (2), (2a) obviously occurs before (in the past of) (2b). In
(3), (3b) is temporally included in (3a). In (4), (4a) also occurs before (4b). If aspect markers
directly determined temporal relations, it would be difficult to explain why a clause with the
same aspect marker has different temporal relations. Hence, the answer to the first question is
negative.
If we examine the examples carefully, we can find that the rhetorical relations in the three
examples are different. (2b) is connected to (2a) by Narration because zaidu ‘again’ indicates
sequential repetition of an event, with one occurrence of the event before another. (3b) is
connected to (3a) by Elaboration because mianli ‘to encourage’ can be part of a speech. For
(4), though there is no explicit cue phrase, we can tell that (4a) provides information based on
which (4b) can be concluded. This kind of rhetorical relation is referred to as information
background, represented as BackgroundI, contra BackgroundT, meaning temporal background.
Since rhetorical relations can also determine the temporal relations in Mandarin, the answer to
the second question is positive.
In addition, the examples above illustrate a very important point, i.e. the experiential guo can
influence which rhetorical relation connects clauses. Hypothesis (1a) says that the event time
56
Aspectual Influence on Temporal Relations
of a situation presented by guo cannot be advanced unless its event time is explicitly
specified. So, even though in (2) zaidu ‘again’ indicates Narration, the event time of the
situation presented by guo needs to be explicitly specified so that this time can be advanced.
This is why in (2a) liang nian qian ‘two years before’ is obligatory.
Lexical information specifies the rhetorical relation that connects (3b) to (3a). Obviously, he
sincerely encouraged the students he addressed to is part of President Qian’s speech.
Therefore, (3b) is connected to (3a) by Elaboration, which indicates temporal inclusion.
Hypothesis (1b) suggests that the internal process of a situation presented by guo cannot be
accessed unless it is forced open to be accessible. Elaboration is exactly the rhetorical relation
that requires to access the internal process of an elaborated situation. Other rhetorical relations
that do not require so, such as Narration, cannot access the internal process of a situation
presented by guo.
In (4), there is no cue to indicate which rhetorical relation connects these two clauses together.
Intuitively, (4a) provides information based on which (4b) can be concluded. As pointed out
in Wu (2005b, 2004), the rhetorical relation that connects clauses with no cue phrases is the
default rhetorical relation specified by an aspect marker. Therefore, we propose that the
experiential guo specifies BackgroundI, information background, as its default rhetorical
relation. It is called ‘default’ because explicit information, such as the cue phrase in (2) and
lexical information in (3), can override it. The temporal relation indicated by BackgroundI is
temporal precedence, that is, the premise situation occurs before (in the past of) the
conclusion. This indication matches the temporal relation in (4) since (4a) occurs before (4b).
This temporal relation conforms to Hypothesis (1c), which states that a situation presented by
guo cannot serve as a temporal background, which indicates temporal overlapping.
To sum up, the aspectual properties of the experiential guo specifies a default rhetorical
relation and constrains which rhetorical relation can connect a clause with guo and an
adjacent clause, and what syntactic construction is required for a certain rhetorical relation to
perform its duty.
To capture the influence of the aspectual properties of guo on rhetorical relations, the
Segmented Discourse Representation Theory (SDRT, Asher and Lascarides 2003) is utilized.
An axiom, a constraint and a meaning postulate are proposed:
(5)
a.
b.
c.
Axiom for the Experiential guo
(?(α, β, λ) ∧ guo(….)(α)) > BackgroundI(β, α, λ)
Constraint on Narration
(Narration(α, β, λ) ∧ guo(…)(α)) → TMP_LOC(…)(α)
Meaning postulate for inferring temporal relations
ΦBackgroundI(β,α) ⇒ eα π eβ
(5a) says that if a clause marked as α is connected to a clause marked as β to form a discourse
λ, and α contains the experiential guo, then, by default, α serves as information background,
represented as BackgroundI, for β in the discourse λ. (5b) says that if α is connected to β by
Narration to form a discourse λ, and α contains the experiential guo, then there must be a
temporal location phase, such as three years before, in it. (5c) says that if α serves as
informational background for β, then α occurs before β.
Let’s take (4) as an example to briefly demonstrate how the axiom and the meaning postulate
work. When (4b) comes into the discourse, it has to be connected to (4a) by a rhetorical
relation. However, there is no information in the discourse that indicates any rhetorical
relation. Therefore, the axiom of guo kicks in and specifies that (4b) is connected to (4a) by
57
Jiun-Shiung Wu
BackgroundI. (5c) indicates that (4a) temporally precedes (4b). This inference matches native
speaker’s intuition about the temporal relation between (4a) and (4b).
This paper demonstrates a very interesting and intricate interaction between explicit
information, e.g. semantics of aspect markers, and implicit information in text, such as
rhetorical relations. Aspect markers in Mandarin do not directly determine temporal relations.
Instead, they do so indirectly via rhetorical relations. The semantics of guo defeasibly
specifies a rhetorical relation. Though explicit information, such as cue phrases or lexical
information, can override this default rhetorical relation, the semantics of guo still constrains
the form of a clause with guo connected to a following clause by Narration. However, only if
the function of the rhetorical relation requires so can a rhetorical relation ‘twist’ the semantics
of an aspect marker. This is why the event time of a clause with guo can be accessed only if
its following clause is connected to it by Elaboration, even though there is no phrase in the
clause that specifies its event time.
References
Asher, Nicholas, and Lascarides, Alex. 2003. Logics of Conversation. Cambridge: Cambrdige
University.
Lin, Jo-wang. 2003. “Temporal Reference in Mandarin Chinese.” Journal of East Asian
Linguistics 12: 259-311.
Lin, Jo-wang. 2006. “Time in a Language Without Tense: the Case of Chinese.” Journal of
Semantics 23: 1-53.
Smith, Carlota, and Erbaugh, Mary. 2001. “Temporal Information in Sentences of Mandarin.”
In: Xu, Liujong and Shao, Jingmin (eds). New Views in Chinese Syntactic Research –
International Symposium on Chinese Grammar for the New Millennium. Eds. Hangzhou:
Zhejiang Jiaoyu.
Smith, Carlota. 1997. The Parameter of Aspect. 2nd Edition. Dorecht: Kluwer Academic.
Wu, Jiun-Shiung. 2006. “Reversibility of Target State and the Semantics of the Experiential
Guo.” Paper submitted to Journal of East Asian Linguistics.
Wu, Jiun-Shiung. 2005a. “Discontinuity and the Semantics of the Experiential Guo.” The
Proceedings of 2005 Conference on Taiwan Culture: Linguistics, Literature, Culture and
Education. Taipei: Crane.
Wu, Jiun-Shiung. 2005b. “Aspectual Influence on Temporal Relations: Evidence from the
Perfective le in Mandarin.” LACUS Forum XXXI: Interconnections. Eds. Adam Makkai,
William J. Sullivan, and Arle R. Lommel. Houston: LACUS, pp181-188.
Wu, Jiun-Shiung. 2004. “Aspectual Influence on Temporal Relations: Evidence from the
Progressive zai and the Durative zhe.” Proceedings of the 16th North American
Conference on Chinese Linguistics. LA: USC.
Yeh, Meng. 1996. “An Analysis of the Experiential Guo in Mandarin. Journal of East Asian
Linguistics 5: 151-182.
58
SPRIK Conference
Explicit and Implicit Information in Text - Information Structure across Languages
University of Oslo, June 8-10 2006
Tense switch in Aspect/-Non-Aspect languages and its implications for
information structure
Barbara Schmiedtová, Mary Carroll, Christiane v. Stutterheim, Natasha Sahonenko
University of Heidelberg
[email protected]
In recounting a set of events, for example, language users have to select information from a
given knowledge base and decide what to say first, what next, (selection, segmentation,
linearization); the information on the set of events has also to be anchored in consistent terms
in space and time (specification of a spatio-temporal frame), and the selected components
have to be structured further with respect to informational status (topic, focus assignment),
mapped into form (main, subordinate clause) and linked in sequence via specific relations
(temporal, causal, etc.). These processes of information organization are viewed as proceeding
in both global (macro-planning) and local terms (micro-planning) in tasks which go beyond
the production of isolated utterances.
Cross-linguistic comparisons (Germanic, Romance, and Semitic languages, as well as learner
languages) show that languages with specific grammatical structures follow similar principles
in information organization in narrative and descriptive tasks (Carroll & Lambert 2003; v.
Stutterheim & Nüse 2003; Carroll & v. Stutterheim 2003). This also applies to the use of
tense and aspect forms in narratives (film renarrations) as well as descriptions of isolated
events.
The focus of the present study is on tense/aspect switching, mainly between aspectually
marked and unmarked forms. Our results show that speakers of aspect-dominant languages,
such as Czech and Russian or Polish, switch temporal perspective relatively often. These
changes take place at the main clause level and regularly co-occur with aspectual changes. In
German, by contrast, switches in temporal perspective as expressed in different tense forms
are rare and they are only used in side structures (e.g. relative clauses, personal comments).
We argue that these patterns of information construal are language-specific and driven by
grammaticalized features rooted in the linguistic system.
We will present and compare results from oral production data of Czech, German, Polish and
Russian native speakers and discuss the consequences of our findings with respect to general
principles of information structure in these languages. Further evidence for this general claim
will be gained from L2. We find that advanced learners show grammatically possible but
pragmatically unacceptable patterns of tense/aspect switching in the L2. These patterns are
driven by the respective L1.
The stimuli we used were 40 short video clips showing common everyday situations. The
speakers were requested to answer the question “What is happening?” in Czech “Co se děje?”,
in Russian “Cto proischodit” as soon as they were able to formulate their response. That is,
informants were asked to produce a short on-line narrative describing the scene. There were
30 speakers in every native speaker group (Czech, German, Russian) and 15 speakers in every
learner group (Czech learners of German, Russian learners of German).
As already mentioned, our study focuses on speakers who originate from an elaborate tempusaspect system (Czech, Russian) and who have to learn a less differentiated system in their
second language (German). In other words, while Czech and Russian show similarities, they
differ from the German systems a lot: Czech/Russian operate with a rich and productive
59
Barbara Schmiedtová, Mary Carroll, Christiane v. Stutterheim & Natasha Sahonenko
tempus-aspect system, the expression of aspect is obligatory, and the opposition between the
perfective and the imperfective is grammaticalized. The situation in German is completely
different: although German can convey the imperfective by lexical devices, aspect in German
is not grammaticalized whereas tense marking is.
Czech:
(1)
Pes v-běhl do zahrady.
Dog Pref-runPerfPast into the garden.
The dog ran into the garden. (perfective, past)
(2)
Pes v-bíhá/běží do zahrady.
Dog Pref-runImperf/runImperfPresent into the garden.
The dog is running/runs into the garden. (imperfective, present)
Russian:
(3)
Sobaka za-bežala v sad.
Dog Pref-runPerfPast into the garden.
The dog ran into the garden. (perfective, past)
(4)
Sobaka za-begaet/bežit v sad.
Dog Pref-runImerf/runImperfPresent into the garden.
The dog is running/runs into the garden. (perfective, past)
German:
(5)
Der Hund läuft/lief/ist in einen Garten gelaufen.
Dog runPresent/PastPreteritum/Participle into the garden.
The dog runs/ran into the garden. (past)
The data shows that there are different triggers for the tense switch to occur in German vs.
Czech/Russian. The default for German native speakers is to present the story in the present
tense (97%). Although the switch from present to past is rare (3%), it always occurs when the
narrator steps outside the referential framework of the story and hence interrupts the story
telling line.
(6)
Hier sieht man ein Gewächshaus im Garten
und in das Gewächshaus lief gerade ein Hund
Here one can see a greenhouse in a garden
and into the greenhouse ran just now a dog
As we can see, the narration in (6) is externalized and depicts the observer’s/narrator’s point
of view. The event that was shown in the stimulus is described in the past tense. Other
contexts where tense switch in German takes place is when subordination or personal
comments are produced. Note that in all cases the main event structure of the narrative is
interrupted.
In Czech and Russian by contrast, a switch in tense is more common: 60% for Czech and 43%
for Russian. Moreover, native speakers of these two languages change from present to past
tense only between main structures.
Czech:
(7)
Žena sedí (Imperf, Present) u stolu,
pije (Imperf, Present) sklenici s vodou.
Dopila (Perf, Past) ji,
teď to odložila (Perf, Past), tu skleničku
60
Tense switch in Aspect/-Non-Aspect languages and its implications for information structure
A woman is sitting at a table
(she) is drinking glass with water
(she) finished drinking it
now (she) (has) put it down, the glass
Russian:
(8)
devuška vzjala (Perf, Past) stakan so stola
i p’et vodu (Imperf, Present)
ona dopila (Perf, Past) ee do konca
i postavila (Perf, Past) stakan obratno na stol
A girl took a glass from the table
and is drinking water
she finished drinking it to the end
and (she) (has) put the glass down again on the table
Example (7) and (8) also demonstrate that in Slavic languages, switching in tense goes hand
in hand with changing aspect (from present-imperfective to past-perfective). From the point
of view of information structure the topic time (time of assertion) in Slavic languages is put in
relation to a particular phase of the event. The deictic origo is anchored implicitly via
aspectual means (e.g. imperfective) and shifts in narrator perspective are expressed on this
basis (aspectual). In German by contrast, the event structure is not segmented into phases and
shifts in perspective from the event line to the deictic origo are enabled explicitly, via the
speaker, by leaving the story telling line. These patterns are rooted in the temporal means
available in each language and the narrative strategies they allow when a longer retelling is
produced. To demonstrate this claim, we will present production data of Russian, German,
and Polish native speakers that are based on a seven minute animation movie (Quest). This
data set provides further evidence for our claim. To show how dominant the patterns of use
are, we will present and discuss data from Czech and Russian learners of German. Here it will
be shown that these learners change tense in German far more frequent than the native
speakers following the same pattern as in their respective L1. Furthermore, we will argue that
this happens because they attempt to express aspectual distinctions that German lacks.
Aspect
Pref – perfective
Imperf – imperfective
Pref – prefix
Tense
Past – past tense
Present – present tense
References
Carroll, Mary & von Stutterheim, Christiane (2003). Typology and information organization:
perspective taking and language-specific effects in the construal of events. In Anna
Ramat (ed.). Typology and Second Language Acquisition. (pp. 365-402) Berlin: de
Gruyter.
61
Barbara Schmiedtová, Mary Carroll, Christiane v. Stutterheim & Natasha Sahonenko
Carroll, Mary & Lambert, Monique & Natale, Silvie & Starren, Marianne & von Stutterheim,
Christiane (in press). Being specific: the role of aspect in event construal when grounding
events in context. A cross-linguistic comparison of advanced second language learners
(L1 French–L2 English, L1 German–L2 English, L1 Dutch–L2 French, L1 Italian–L2
German). In S. Haberzettl (ed.). Dynamics of learner varieties. Berlin: de Gruyter.
Carroll, Mary & Lambert, Monique (2003). Information structure in narratives and the role of
grammaticised knowledge: A study of adult French and German learners of English. In
Christine Dimroth & Marianne Starren (eds.). Information Structure and the Dynamics of
Language Acquisition (pp. 267-287). Amsterdam/Phil.: Benjamins.
Carroll, Mary & von Stutterheim, Christiane (2003). Typology and information organisation:
perspective taking and language-specific effects in the construal of events. In Anna
Ramat (ed.). Typology and Second Language Acquisition. (pp. 365-402) Berlin: de
Gruyter.
Klabunde, Ralf & von Stutterheim, Christiane (eds.): Representations and Processes in
Language Production. Wiesbaden 1999.
Schmiedtová, Barbara (in press). The development of the expression of simultaneity in L2
Czech: A special focus on (very) advanced learners. In Haberzettl, S. (Ed.) The end state
in SLA. Benjamin, Amsterdam.
Schmiedtová, Barbara & Sahonenko, Natasha. (2005). The role of grammatical aspect in
event encoding: A comparison of Czech and Russian advanced learners of German. A
Paper presented at the 15th Annual Conference of The European Second Language
Association (EUROSLA), University of Zagreb, Dubrovnik, Croatia,
Schmiedtová, Barbara (2004). At the same time... The expression of simultaneity in learner
varieties. Mouton de Gruyter, Berlin.
von Stutterheim, Christiane & Lambert, Monique (2005). Crosslinguistic analysis of temporal
perspective. In H. Hendricks (Hrsg.). The structure of learner varieties, 203-230. Berlin:
de Gruyter.
von Stutterheim, Christiane & Nüse, Ralf (2003). Processes of conceptualisation in language
production. Linguistics, vol. 41-5, (Special Issue: Perspectives in language production),
851-881.
62
SPRIK Conference
Explicit and Implicit Information in Text - Information Structure across Languages
University of Oslo, June 8-10 2006
Making implicit information explicit: Kortmann’s scale of informativeness and
the English V-ing free adjuncts in Catalan translations
Anna Espunya
Department of Translation and Philology, Pompeu Fabra University
Barcelona, Catalonia, Spain
[email protected]
Free adjuncts in V-ing –illustrated by the underlined clause in George obeyed, leaving the
room again (Agatha Christie, Third Girl)– belong to the class of converbs, one of the three
possible ways in which adverbial subordination is expressed (Nedjalkov, 1998: 421). A free
adjunct establishes semantic / discourse relationships with another clause, usually a finite one,
which is considered the matrix clause of the complex sentence, and which usually controls its
subject. Huddleston, Payne and Peterson (2002: 1353-1354) use the more suggestive term
semantic anchor. In a specific study of a corpus of texts belonging to various genres
(Kortmann, 1991), the following relationships were identified: Simultaneity, Anteriority,
Posteriority, Cause / Reason, Instrument, Result, Purpose, Manner, Concession, Contrast,
Condition, Exemplification / Specification, Accompanying circumstance and Addition. Some
of them are standard adverbial relations, whether temporal or causal (see Kortmann, 1995:
215-218), whereas others are better described in terms of discourse structure, such as
Exemplification / Specification, and Addition.
Free adjuncts are, in most instances, semantically indeterminate, which means that, unless
they are augmented by lexico-grammatical means such as conjunctions (e.g. while) and
connective adverbs (e.g. thus), the semantic / discourse relationship must be inferred. Besides
co-occurrence of linking items, several factors may condition the inferences, as described in
Stump (1985: 307-323): a) the aspectual type of the predicates of both free adjunct and
anchor; b) iconic ordering; and c) knowledge of the world.
Kortmann (1991: 115) adopts the view that the search domains for the individual process of
interpretation of free adjuncts are limited by a scale of specificness based on the principle of
informativeness (Atlas and Levinson, 1981): “the process of identifying some logical role(s)
for a given free adjunct/absolute is essentially determined by a scale on which the semantic
relations in principle available to these construction types can be arranged according to their
informativeness or, alternatively, specificness. […] Thus ‘more informative’ semantic
relations will be distinguished from ‘less informative’ ones on the basis that the former
require more knowledge or (co- or contextually substantiated) evidence in order to be
identified as the semantic relation holding between the proposition denoted by a given free
adjunct/absolute and some matrix proposition.” For example, Concession, Contrast and
Condition require more knowledge or (co-textual/contextual) evidence in order to be
identified and thus stand high on the informativeness scale, while Simultaneity, Addition and
Accompanying circumstance require less knowledge and thus stand low on the scale.
The scale proposed for ‘present-participial’ free adjuncts /absolutes is the following
(Kortmann 1991: 121, figure 9.1)
63
Anna Espunya
Most informative
Concession
(strongest)
Contrast
Condition
Instrument
purpose
Cause
result
Time before
time after
(anteriority)
(posteriority)
Manner
Exemplification /specification
Same time
(simultaneity/overlap)
Least
informative
Accompanying circumstance
(weakest)
Addition
In Kortmann’s study, the degree of informativeness was found to correlate with a tendency
toward linguistic augmentation or explicitness through lexico-grammatical means. For
instance, 4.4% of all adjuncts expressing ‘same time’ are augmented, whereas for Anteriority
and Posteriority, the ratio is 16.9%, and for Contrast / Concession, the percentage of
augmented adjuncts is as high as 30.2% (Kortmann, 1991: 121, 196). The author’s
interpretation of the data is that “these differences mirror the different degrees of
interpretative complexity and, consequently, processing effort for the hearer / reader that the
addresser attributes to these semantic relations when expressed by free adjuncts.” As a
consequence, “the more difficult the addresser believes the identification of the intended
interpretation to be, the more likely is it that he/she marks it by some lexical item
conventionally linked to the expression of the respective relation.”
The idea that informativeness is reflected in more intralinguistic explicitness may shed light
on the sources of interlinguistic explicitness, or explicitation in translation, a well-known
tendency that has been quantitatively identified in many language pairs by means of corpus
linguistics methodologies (see, e.g., Mauranen and Kujamäki, 2004). As an instance of
communication, translation requires text interpretation followed by text generation. It seems
sensible to ask whether there is more explicitation of more informative relations, i.e. whether
the effort required in the interpretation phase is reflected in the choices of the translator in the
production phase.
To answer this question, a study of a corpus of free adjuncts and their correspondences in
translation from English was devised which combined the quantitative with the qualitative
approach. The quantitative approach is used to confirm that the ratios of instances of
64
Making Implicit Information Explicit
explicitation can be ordered along a scale which parallels that in Kortmann. The qualitative
approach aims at describing the ways in which the specificness scale is reflected in
explicitation.
Right now the corpus of study consists of 269 instances of English free adjuncts and their
correspondences drawn from 4 published translations into Catalan (spoken mostly in NorthEastern Spain), each translation performed by a different translator. The instances that make
up the corpus have been manually extracted from popular fiction and from popularisation of
science. Variety of genres is justified by evidence of a correlation between genre and the
semantic relations holding between free adjuncts and their anchors provided in Thompson
(1983: 45-57) and Kortmann (1991: 138-139). The semantic relationships have been
identified introspectively. The translation solutions have been classified into one of the two
relevant categories: ‘explicitation’ vs. ‘non-explicitation’. The former includes any
correspondence in which a semantic / discourse relationship which was implicit in the source
text is linguistically encoded in the target text, regardless of the syntactic form of the
sentence. Non-explicitation includes any correspondence –not just the formally parallel “free
adjuncts in the gerund”– that requires readers of the target text to infer the relationship
between the two clauses unaided by explicit linguistic markers. The following is an instance
of explicitation [ST stands for Source Text, TT, for Target Text, and BT, for Back
Translation; the free adjunct is underlined, the linguistic items that encode the intended
inference are in bold]:
(1) ST ‘And now,’ said Mrs Oliver, as her guest finally replaced his cup on its saucer and sat
back with a sigh of satisfaction, wiping remnants of foaming cream from his
moustache, ‘what is all this about?’ [Third Girl: 20]
TT TT—I ara —féu Mrs. Oliver quan el convidat retornà la tassa al seu lloc, i s’aclofà a la
butaca amb un sospir de satisfacció, alhora que s’eixugava els regalims de nata que li
lliscaven bigoti avall—, digueu-me què passa. [La Tercera Noia: 15]
BT ‘And now,’ said Mrs Oliver, when the guest returned the cup in its place and sat back
with a sigh of satisfaction, while he was wiping the remnants of cream that were
trickling down from his moustache, ‘what is all this about?’
As far as the quantitative results are concerned, the most informative relations are
underrepresented in the corpus, which means that the sample is too small right now for
statistical significance tests. Nevertheless, the rate of explicitation seems to follow the scale in
three of the translations. For instance, in the translation of Third Girl, the rate of explicitation
for Cause, Result and Anteriority is higher than the rate for Manner, Specification and
Simultaneity. This suggests that the “scale of informativeness” hypothesis might be validated
by the translation data, as more instances are incorporated in the corpus.
Qualitatively speaking, support for the informativeness hypothesis is lent by the direction in
which semantic shifts occur. As the translator tries to render what was said in the source text,
the semantic / discourse relationship may be changed. In the corpus these shifts follow an
“upward” direction, i.e. the translator’s choices narrow down or even determine the possible
inferences for the reader of the translation to a relationship that stands higher on the scale than
the preferred source text inference. In its present state, the corpus provides two clear types of
shift. First, sequences which would allow a Cause and a Condition interpretation are rendered
by means of an explicitly Conditional relation, which stands higher on the scale, as illustrated
in (2).
(2) ST Mothers, being sometimes sensible people, don’t usually pay any attention. [Third
Girl: 153]
65
Anna Espunya
TT Les mares, si són assenyades, no en fan cas. [La Tercera Noia: 98]
BT Mothers, if they are sensible people, don’t pay any attention.
Most native English speakers that I have consulted interpret the relationship as Cause in the
first place although they accept the Condition reading. Following the approach in Stump
(1985: 58-59), the Cause relationship is the preferred inference, in modal contexts, for strong
adjuncts, i.e. those whose predicate is individual-level. Here the test for membership to this
class is provided by the fact that ‘be + sensible people’ is applied to mothers, which is a kindreferring NP. Nevertheless, a Condition interpretation is supported by various of the features
of (2). First, the sentence is negated. Secondly, it has the features of a Generic sentence
(Smith 2003: 79), i.e. the kind-referring NP, the frequency adverb usually, and the present
tense. Moreover, the free adjunct contains the adverb sometimes which may include mothers
in its scope (i.e. “mothers, some of which are not sensible people”).
A second type of shift involves the rendering of an explicit Purpose relationship where a less
specific one was acceptable in the source text. This strategy is displayed especially when two
free adjuncts are coordinated. This is illustrated in (3):
(3) ST Someone would have to keep constant watch over the chronometer, noting all changes
in ambient temperature and figuring them into the longitude reading. [Longitude: 58]
TT Caldria que algú vigilés constantment el cronòmetre i anotés tots els canvis de la
temperatura ambient per traslladar-los a les lectures de longitud. [La longitud: 59]
BT Someone would have to keep constant watch over the chronometer, noting all changes
in ambient temperature in order to figure them into the longitude reading.
Possible objections to the methodology of the study concern, one the one hand, the impact of
contrastive stylistic facts on the application of translation strategies and, on the other, the
definition of what constitutes explicitness in free adjuncts. As far as the contrastive factors are
concerned, it might be claimed that Catalan written style is more explicit, so that explicitation
is the strategy that fulfils the publishers’ and readers’ expectations. In response to this
objection it may be said that this contrastive fact would account for explicitation in general,
but not for differences according to the informativeness degree of each semantic relationship.
Neither does it explain the “upward” direction of semantic shifts between ST and translation.
Another contrastive factor might concern differences in the interpretations available for the
Catalan gerund, so that the translator would be forced to choose a different structure, possibly
a more explicit one involving subordinating conjunctions, or a connective adverbs. It is true
that free adjuncts in the gerund in Catalan do not allow for a Posteriority reading. However,
the existing alternatives include not only clausal adjuncts introduced by conjunctions and
phrasal adjuncts introduced by prepositions, but also the coordinating conjunction and (in
fact, this use of the gerund is called copulative in Catalan traditional grammar.)
Regarding the definition of explicitness in free adjuncts, it might be adduced that semantic /
discourse relationships are made explicit not only by connectives but also by the lexical
choices in the predication. A number of V-ing adjunct predicates are in the process of
lexicalisation, such as being and seeing (see Stump, 1985: 314-316). In the corpus, certain Ving forms seem to assume a “conventionalised” meaning, at least in scientific discourse. Thus,
using determines an Instrument inference, while causing and producing determine a Result
interpretation. This objection may be overcome by discarding the instances as other
augmented free adjuncts are discarded.
66
Making Implicit Information Explicit
By way of conclusion, this paper reports work under way that attempts to extend Kortmann’s
(1991) scale to the description of interlingual explicitness in the encoding of interclausal
semantic / discourse relationships. The informativeness scale, which was proposed as an
account of intralingual explicitness, is based on the notion of cognitive effort. Its validity for
translation data would contribute in a specific way to a pragmatic theory of translation, which
focuses on the translator’s role as mediator in decreasing the amount of effort the readers have
to invest in inference (see Rosales Sequeiros, 2005: 65). In the reverse direction, the validity
of the scale for translation data lends support to it as a hypothesis for explicitness of
interclausal relationships within a single language.
References
Atlas, Jay David & Levinson, Stephen C. 1981. It-clefts, informativeness, and logical form.
In: Cole, Peter (ed.), Radical Pragmatics. New York: Academic Press. 1-61.
Huddleston, Rodney; Payne, John & Peterson, Peter. 2002. Coordination and
supplementation. In: Huddleston, Rodney and Geoffrey K. Pullum (eds.), The Cambridge
Grammar of the English language. Cambridge: Cambridge University Press. 1273-1364.
Kortmann, Bernd. 1991. Free adjuncts and absolutes in English. Problems of Control and
Interpretation. London: Routledge.
--- 1995. Adverbial participial clauses in English. In: Haspelmath, Martin & König, Ekkehard
(eds.), Converbs in cross-linguistic perspective. Structure and meaning of adverbial verb
forms – adverbial participles, gerunds–. Berlin: Mouton de Gruyter. 189-237
Mauranen, Anna & Pekka Kujamäki (eds.). 2004. Translation Universals: Do they exist?
Amsterdam / Philadelphia: John Benjamins.
Nedjalkov, Igor V. 1998. Converbs in the languages of Europe. In: van der Auwera, Johan
(ed.), Adverbial Constructions in the Languages of Europe. Berlin: Mouton de Gruyter.
421-455.
Rosales Sequeiros, Xosé. 2005. Effects of Pragmatic Interpetation on Translation. Munich:
Lincom.
Séguinot, Candace. 1988. Pragmatics and the Explicitation Hypothesis, TTR Traduction,
Terminologie, Rédaction 1/2: 106-114.
Smith, Carlota S. 2003. Modes of Discourse. The Local Structure of Texts. Cambridge:
Cambridge University Press.
Stump, Gregory T. 1985. The Semantic Variability of Absolute Constructions. Dordrecht:
D.Reidel.
Thompson, Sandra A. 1983. Grammar and Discourse: The English Detached Participial
Clause. In: Klein-Andreu, Flora (ed.), Discourse Perspectives on Syntax. Academic Press.
43-65.
67
SPRIK Conference
Explicit and Implicit Information in Text - Information Structure across Languages
University of Oslo, June 8-10 2006
Adverbial doch and the notion of contrast∗
Elena Karagjosova
University of Oslo
[email protected]
The German adversative discourse marker doch is notoriously versatile in its uses and functions:
it can be used as an adverb, conjunction, modal particle and response particle. Adverbial doch
is always accented and can occupy the middle field or the initial field of the German sentence.
In the latter case, it is categorised as a conjunct adverb (CA), cf. (1), where small caps denote
accent:
(1)
Draußen vor dem Zelt stand Artax, sein Pferd. Es war gefleckt und klein wie ein Wildpferd,
[seine Beine waren stämmig und kurz]C1 , und DOCH [war es der schnellste und ausdauerndste Renner weit und breit]C2 . (ME1)
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. (ME1TE)1
In spite of recent revived interest in the semantics and pragmatics of connectors in general and
adversative connectors in particular, I am not aware of any detailed studies of adverbial doch and
especially its CA use.2 Like conjunctions, CAs have sentence connecting function but behave
syntactically like adverbs in that they occupy the initial field, whereas conjunctions are positioned
before the initial field (Duden 2005). The present paper evolves from a comparison of CA doch
with the better understood conjunction doch and aims at reconsiling the following observations.
First, CA doch marks a contrastive discourse relation that can only be interpreted in terms of
concession, unlike the conjunction doch that can also mark the other kinds of contrastive relations
generally assumed in the literature: semantic opposition and concessive opposition.3 This point
is illustrated by the following examples, where the a-sentences contain the conjunction doch and
the b-sentences the CA doch:
∗ This
work is supported by the SPRIK project (Språk i Kontrast [Languages in Contrast], NFR 158447/530). I
would like to thank the members of the SPRIK group for stimulating discussions and Torgrim Solstad for valuable
comments on the paper.
1 Example from OMC (Oslo Multilingual Corpus, http://www.hf.uio.no/ilos/OMC/). C and C are the first and
1
2
the second conjunct linked by the CA respectively.
2 It is for instance completely ignored in a comprehensive work such as Pasch et al. (2003).
3 I will use the term “concession” for what is also called “denial of expectation”, e.g., König (1991), Lagerwerf
(1998), and, following Kruijff-Korbayová & Webber (2005), reserve Spooren’s (1989) “concessive opposition” for
the argumentative interpretation of contrast observed in Anscombre & Ducrot (1977) and Dascal & Katriel (1977).
69
Elena Karagjosova
Semantic opposition:
(2)
Der Geist ist willig, doch das Fleisch ist schwach.
The spirit is willing but the flesh is weak.
b. ?? Der Geist ist willig, und DOCH ist das Fleisch schwach.
a.
Concessive opposition:
(3)
(Wollen wir das Zimmer nehmen?)
(Shall we take this room?)
a.
Man hat einen tollen Ausblick, doch der Preis ist zu hoch.
It has a beautiful view, but it is very expensive.4
b. ?? Man hat einen tollen Ausblick, und DOCH ist der Preis zu hoch.
The intuition that CA doch is specialised in marking concession is confirmed by a corpus study
involving an analysis of all occurrences of CA doch in the OMC (Oslo Multilingual Corpus) with
respect to the discourse relation between the clauses conjoined by doch.
Second, in functioning as a CA, doch establishes an anaphoric link to the first conjunct,5 whereas
no such thing can be said about the conjunction doch. As a CA, doch is synonymous with the
concessive CA trotzdem (’despite that’). However, whereas anaphoricity is immanent in the
lexical semantics of trotzdem, doch cannot be seen as anaphoric in all its uses.
And third, concessive CAs and conjunctions such as trotzdem and obwohl are taken to contain
the logical negation as part of their meaning (e.g., König (1991), Umbach & Stede (1999)),
whereas adversative conjunctions such as aber, doch and but are generally assumed to have only
the logical conjunction in terms of their truth-functional semantics. This means that treating CA
doch in the same terms as trotzdem would force us to assume a separate meaning for CA doch,
which is not a satisfactory solution considering the close etymological relationship between the
various uses of doch (cf. e.g. Hentschel (1986)).
Against this background, the question that emerges is: Given that doch is not always anaphoric,
nor can be assumed to contain a negation, what is it that makes CA doch capable of, and at
the same time restricts it to, rendering the same interpretation as anaphoric concessive CAs like
trotzdem?
I base my attempt to answer this question on the fact that adverbial doch is accented, which invites a focus-theoretical perspective. I draw on recent work by Lang (2004), Lang & Adamı́ková
(2005) and Umbach (2001, 2004), who have convincingly argued for a focus-sensitive analysis
of the adversative connectors aber and but in terms of a notion of contrast based on the focussemantic concept of evoking alternatives.6 I argue that the behaviour of CA doch described above
can be accounted for by assuming that it acquires anaphoric properties and forces a concessive
interpretation due to a combination of factors related to the fact that it is accented:
First, the accent on doch makes it a focussed constituent evoking a focus set of alternatives (Rooth
1992). I argue that the set of alternatives that doch evokes contains two propositions, namely the
one expressed by the doch-conjunct and its polar opposite, i.e. the proposition expressed by the
doch-conjunct with reversed polarity. E.g. in (1), repeated below for convenience, this alternative
is (4b):
4 English
example from Lagerwerf (1998).
adverbs refer to the entire preceding sentence or clause and have thus a Pro-function (Duden 2005).
6 A similar approach is also taken in Sæbø (2003).
5 Conjunct
70
Adverbial 'doch' and the notion of contrast
(4)
a.
b.
Draußen vor dem Zelt stand Artax, sein Pferd. Es war gefleckt und klein wie
ein Wildpferd, [seine Beine waren stämmig und kurz]C1 , und DOCH [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 wide.
Second, the kind of focus doch constitutes is 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. In the case of CA doch, the relevant member of
the focus set of alternatives that can be seen as contrasted, is the polar opposite of the proposition
expressed by the doch-conjunct, i.e. (4b). This alternative is however not explicitly mentioned,
but is rather a consequent of the state of affairs expressed by the first conjunct (cf. also Zeevat
(2004, 2005)). In any case, contrastive focus signals that the evoked alternative should be linked
to the preceding context, i.e. a matching antecedent should be available.
And third, the accent on doch plays a role in specifying the interpretation of the relation between the conjuncts as concession. Lang & Adamı́ková (2005) argue that the focus-background
structure of the sentence (FBS) specifies which entities are contrasted, which in turn determines
the interpretation of the adversative relation as one of either concession or semantic opposition.7
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, as in (5a), then the relation is
interpreted as semantic opposition, if it is asymmetric, as in (5b), then the relation is concessive.8
(5)
a.
b.
[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.9
[Mein Vater ist ernsthaft krank]F , doch [meine Mutter geht arbeiten]F .
Returning to CA doch, the fact that it is accented seems to have consequences for the FBS of the
entire construction.10 Thus, it does not seem to allow a parallel FBS, cf. (6a), contrary to the
unaccented conjunction doch in (2), repeated below as (6b), which allows for both kinds of FSB
(as illustrated in (5)):
(6)
a.
b.
# [Der G EIST]T [ist WILLIG]F , und DOCH [ist das F LEISCH]T [SCHWACH]F .
[Der G EIST ]T [ist WILLIG]F , doch [ das F LEISCH]T [ist SCHWACH]F .
This suggests that although the asymmetric FBS of the sentence already determines a concessive
reading, the reason for this asymmetry of the FSB is the accent on doch.
I suggest that the anaphoricity of CA doch can be accounted for on the basis of the concept of
concession as contrast between two propositions promoted in both Lang & Adamı́ková (2005)
and Umbach (2001). This contrast leads to specifying the conjuncts linked by adversative connectors like aber and but as focus alternatives with respect to each other.
7 They
use the term “contrast” for what is referred to as “semantic opposition” here.
similar correlation is observed by Umbach (2001): “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.
9 The example is taken from Lang (2004) where the synonymous connector aber is used instead of doch.
10 This is reminiscent of Lang (2004), who suggests that the syntactic properties of adverbial connectors such as
trotzdem predestinate the prosodical asymmetry of the conjuncts. Taking CA doch into consideration, we can add
that the prosodic properties of the connector also play a role.
8A
71
Elena Karagjosova
The crucial idea, borrowed from Lang (2004), is that concession triggers the derivation of an
inference q from the first conjunct. This inference is a variable that must be instantiated on the
basis of world knowledge. It is furthermore required that q is such that it (logically) implies the
opposite of what is expressed by the second conjunct, i.e. p1 ⇒ q, such that q → ¬p2, where
p1 and p2 are the propositions expressed by the two conjuncts and ’p ⇒ q’ is read as ’given our
everyday experience, p is expected to have the consequence q’. I argue that the underspecified
inference q from p1 that a concessive interpretation invites, is a good candidate for providing a
matching antecedent of the alternative evoked by doch that needs to be bound to the preceding
context. In other words, the alternative evoked by doch instantiates the variable inferred from
p1.
To see how this works in an example, consider again (4), where the underspecified inference
that should be derived from the fact that the horse’s legs are short and stocky is instantiated by
the focus alternative evoked by doch, namely that the horse is not fast (cf. (4b)). The resulting
(default) inference relation is supported by world knowledge: normally, if a horse has short legs,
we do not expect it to be fast. I argue furthermore that due to this anaphoric behaviour, CA
doch forces a concessive interpretation even in cases where no concessive relation between the
conjuncts can be reconstructed on the basis of world knowledge. Consider (2b) which strongly
suggests that the fact that ’the flesh is not weak’ follows from the fact that ’the spirit is willing’
even though world knowledge does not provide support for such an inference, a fact which
accounts for the awkwardness of (2b). This means that doch allows instantiation of the inference
variable from structural information, prior to resorting to world knowledge, which is a case not
provided for in Lang’s model.
This analysis accounts for the intuition that CA doch requires that the alternative it evokes is
suggested by the first conjunct (cf. also Zeevat (2005)). The corpus study mentioned above
shows furthermore that the focus alternative has to be either locally bound to the first conjunct
in a conjunction of two clauses, or non-locally bound to some clause in the preceding discourse,
where doch is used sentence initially.
Finally, I suggest that the negation normally ascribed to concessive markers can be seen in the
case of CA doch to be a side effect of contrasting with doch which involves evoking and excluding a focus alternative. I follow Umbach (2001) who argues that the semantics of but, beyond
being a mere conjunction, requires the exclusion of a focus alternative. Umbach points out that
this does not mean that but introduces a negation, but rather requires a negation as a kind of
selectional restriction.
The analysis extends to the likewise accented middle-field use of adverbial doch and accounts
for the appropriateness of middle-field doch in contexts where it co-occurs with other adversative
markers.
References
Anscombre, J.-C. & Ducrot, O. (1977), ‘Deux mais en français?’, Lingua 43, 23–40.
Dascal, M. & 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, Dudenverlag, Mannheim.
Hentschel, E. (1986), Funktion und Geschichte deutscher Partikeln: ja, doch, halt und eben,
Reihe germanistische Linguistik; 63, Niemeyer, Tübingen.
72
Adverbial 'doch' and the notion of contrast
König, E. (1991), Konzessive Konjunktionen, in A. von Stechow & D. Wunderlich, eds, ‘Handbuch der Semantik’, Walter de Gruyter, Berlin, New York.
Kruijff-Korbayová, I. & Webber, B. L. (2005), Interpreting concession statements in light of
information structure, in H. Bunt & R. Muskens, eds, ‘Computing Meaning’, Vol. 3, Kluwer.
Lagerwerf, L. (1998), Causal Connectives have Presuppositions, PhD thesis, Tilburg.
Lang, E. (2004), Schnittstellen bei der Konnektoren-Beschreibung, in H. Blühdorn, E. Breindl
& U. H. Waßner, eds, ‘Brücken schlagen’, de Gruyter.
Lang, E. & Adamı́ková, 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’, Walter de Gruyter, Berlin-New York.
Pasch, R., Brauße, U., Breindl, E. & Waßner, U. H., eds (2003), Handbuch der deutschen Konnektoren, Walter de Gruyter, Berlin - New York.
Rooth, M. (1992), ‘A theory of focus interpretation’, Natural Language Semantics (1), 75–116.
Sæbø, K. J. (2003), Presupposition and contrast: German aber as a topic particle, in Weisgerber,
ed., ‘Proceedings of Sinn und Bedeutung 7’.
Spooren, W. P. M. S. (1989), Some Aspects of the Form and Interpretation of Global Contrastive
Coherence Relations, PhD thesis, University of Nijmegen.
Umbach, C. (2001), Contrast and contrastive topic, in ‘Proceedings of the ESSLLI 2001 Workshop on Information Structure, Discourse Structure and Discourse Semantics’.
Umbach, C. (2004), ‘On the notion of contrast in information structure and discourse structure’,
Journal of Semantics 21(2), 155–175.
Umbach, C. & Stede, M. (1999), Kohärenzrelationen: Ein Vergleich von Kontrast und Konzession, KIT-Report 148, Technische Universität Berlin.
Zeevat, H. (2004), ‘Contrastors’, Journal of Semantics 21(2), 95–112.
Zeevat, H. (2005), A dynamic approach to discourse particles, in K. Fischer, ed., ‘Approaches to
Discourse Particles’, Studies in Pragmatics 1, Elsevier, Amsterdam.
73
SPRIK Conference
Explicit and Implicit Information in Text - Information Structure across Languages
University of Oslo, June 8-10 2006
Evidence for a Scalar Analysis of Result in SDRT from a Study of the French
Temporal Connective alors
Myriam Bras°, Anne Le Draoulec° & Nicholas Asher*
° ERSS, Université Toulouse - Le Mirail / * IRIT, Université Paul Sabatier, Toulouse
[email protected], [email protected], [email protected]
This paper presents a preliminary analysis of the French temporal connective alors (generally
translated in English by then, at that time, so). It is part of a broader project to provide a
systematic analysis of French temporal connectives within Asher’s formal framework of
Segmented Discourse Representation Theory (Asher 1993, Asher and Lascarides 2003). Among
the linguistic markers that establish a temporal relation between the eventualities introduced by
two clauses (henceforth discourse constituents) to be discourse linked, temporal connectives are
distinguished by the fact that they introduce at the same time some sort of discourse relation.
In previous work some of us showed that puis, which induces a temporal connection between
constituents, is just such a connective; its role in SDRT is to impose a relation of Narration 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, 1988a&b; Franckel 1987; Gerecht
1987; Hybertie 1996; Reyle 1998; Gosselin in press). Most authors distinguish three major uses
of alors in assertions: temporal uses (with or without a consequential value), merely
consequential uses (close to donc ‘therefore’), and other uses where alors is a kind of
« structuration » marker.
Le Draoulec and Bras (2004) studied the temporal uses of alors when it relates two assertions
describing events. They showed that the temporal value is necessarily associated with a
consequential value only when alors is in clause initial position. 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). Moreover, the
temporal value itself depends on the sentential position: clause initial alors gives rise to a relation
of temporal succession between the events; clause internal or final alors denotes a temporal
relation of concomitance or coincidence.
In this paper we focus on uses of alors in initial position, still restricting the study to clauses
describing events (states are left for future research). We investigate which discourse relation(s)
is (are) able to express the consequential value necessarily involved by alors.
In (1), SDRT predicts a discourse relation of Result between the two constituents.
(1)
Olivier a fait tomber la carafe. Alors elle s'est cassée.
Oliver dropped the carafe. Alors1 it broke.
1
In this example alors could be translated by so or then. We prefer not to choose a translation, so as not to blur the
problem.
75
Myriam Bras, Anne Le Draoulec & Nicholas Asher
SDRT allows us to deduce the discourse relation Result when one can infer from lexical or
domain information the predicate causeD (we simplify the axiom of Asher and Lascarides (2003)
as we will not consider different aspectual classes).
Axiom_Result
(? (α, β, λ ) ∧ causeD(α, β)) > Result (α, β, λ)
In (1), the information needed to infer causeD is readily available – tomber(x) is a permissible
cause of se casser(x). Initial alors goes very well with this inference. However its role is not
completely evident in this example, as we would have the same inference for (1) without alors.
The role of alors is clearer in examples like (2) or (3):
(2)
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.
(3)
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.
From a strict SDRT point of view, the requisite information needed to infer causeD for (2) or (3)
is lacking. So we cannot infer Result. It also seems improbable that occasion, the relevant
predicate on eventuality types needed to infer Narration, holds between the two constituents2. So
the appropriate axiom (Axiom_Narration) can not be used to infer Narration. As SDRT does not
yet account for the role of alors, it would predict Narration by default.
Axiom_Narration
(? (α, β, λ ) ∧ Occasion(α, β)) > Narration (α, β, λ)
Following Hybertie (1996), Le Draoulec and Bras (2004) show that in examples such as (2) and
(3), alors triggers a discourse relation requiring that ‘the event expressed by the first constituent
is a necessary condition for the event described by the second constituent’. This relation differs
from the extant, similar SDRT relations of Result and Narration.
Asher and Lascarides (2003) don't give a complete definition of Result but they take Result(α,β)
to imply that the main eventuality in α is the cause of the main eventuality in β.
It thus seems necessary to introduce a new relation, which we will call Weak-Result. Alors is a
trigger for this relation, which expresses a causal link close to that proposed by Lewis (1973):
An event eα associated with a description Kα (in a discourse constituent α) is a necessary
cause for an event eβ (associated with a description Kβ in a discourse constituent β) iff
(┐Kα □→ ┐Kβ) ∧ (Kα ∧ Kβ)
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.
In less formal terms, eα is a necessary cause for eβ means that:
(i)
if eα had not occurred, eβ wouldn’t have occurred either, in all the worlds closest
to α’s world, and
(ii)
Kα and Kβ are true.
2
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.
76
Evidence for a Scalar Analysis of Result in SDRT from a Study of the French Temporal Connective alors
This relation is weaker than the one according to which the first event is the cause of the second:
eα is the cause of eβ if and only if eα is a necessary cause for eβ, and Kα > Kβ
thanks to the expression Kα > Kβ (normally if Kα obtains then Kβ).
We thus define Weak-Result as a discourse relation on constituents as follows:
Definition_Weak-Result
Weak-Result (α, β, λ) ↔ ((┐ Kα □→ ┐ Kβ ) ∧ (Kα ∧ Kβ ∧ eα < eβ))
Axiom_Weak-Result
(? (α, β, λ) ∧ alors(β)) → Weak-Result (α, β, λ)
This formulation of Weak-Result yields an immediate consequence: there is a temporal
succession between the events, which corresponds to what Le Draoulec and Bras (2004) describe
for clause initial alors.
As we said before, we restrict ourselves here to events. There is still a lot of work to do on states:
in many uses of alors, in particular consequential uses, the eventualities involved are states (with
different temporal implications for temporal uses). Dealing with states will also imply accounting
for logical consequence uses, as:
(4)
Puisque A est vrai, alors A ou B est vrai
As A is true, alors A or B is true
But we can also have events in the uses of alors that express logical consequence, as in:
(5)
Toutes les filles sont arrivées à l’heure, alors Myriam est arrivée à l’heure
All the girls arrived on time, alors Myriam arrived on time
In order to be able to account for these cases, we should add a disjunction in our formula:
Definition_Weak-Result_bis
Weak-Result (α, β, λ) ↔ ((┐ Kα □→ ┐ Kβ ) ∧ (Kα ∧ Kβ ∧ eα < eβ)) ∨ □(Kα → Kβ )
Weak-Result now is defined to hold between clauses that involve logical consequential uses of
alors. We could have separated out these consequential uses, but including them within WeakResult will enable us to define another result relation that entails Weak-Result, as we shall see in
a minute.
Let us see how we can now account for our examples with the definition and axiom on WeakResult given above. The relation of Weak-Result as defined is the appropriate one linking the
clauses in (2) and (3). For example, in (3): 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 (1): the causal
link at stake is stronger than what the predicate ‘is a necessary cause for’ expresses: it
corresponds to the predicate ‘is the cause of’ 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 StrongResult, which can be defeasibly inferred from causeD, and which we define using the predicate ‘is
the cause of’.
Definition_Strong-Result
Strong-Result (α, β, λ) ↔ (Weak-Result (α, β, l) ∧ Kα > Kβ )
77
Myriam Bras, Anne Le Draoulec & Nicholas Asher
Axiom_Strong-Result
(? (α, β, λ ) ∧ causeD(α, β)) > Strong-Result (α, β, λ)
For example (1), both Axiom_Weak-Result and Axiom_Strong-Result will apply: both WeakResult and Strong-Result will be inferred. For the same example without alors, Strong-Result
would be inferred too.
We finally examine cases when alors combines with occasion, as in (6):
(6)
Il est tombé. Alors je l'ai aidé à se relever.
He fell. Alors I helped him up.
In such a case, the extant axioms lead to infer Narration and Weak-Result. However, we feel that
Weak-Result is not sufficient, and that the relation of Strong-Result would be more appropriate to
capture the interpretation of (6). This leads to add the following axiom which has the
consequence in following theorem.
Axiom_Strong-Result2
(? (α, β, λ) ∧ alors(β) ∧ occasion(α, β)) → Strong-Result (α, β, λ)
Theorem
(? (α, β, λ ) ∧ alors(β) ∧ occasion(α, β)) → (Kα > Kβ )
From these definitions and axioms, we can deduce that both Strong-Result and Weak-Result are
veridical relations in the sense of Asher and Lascarides (2003).
We finally come back to the uses of alors expressing a logical consequence. In our definition of
Weak-Result they are described by the second disjunct (□(Kα → Kβ)). We now see that these
uses of alors are now entailed to be cases of Strong-Result as well, which is intuitively what is
desired: this is the case because we can infer the theorem Kα > Kβ, from □(Kα → Kβ). We added
□(Kα → Kβ) as a disjunct in the right part of Definition_Weak-Result_bis because we wanted to
guarantee the scalarity of Result : as it is defined now, Strong-Result is stronger than WeakResult thanks to the conjunct Kα > Kβ. We thus have a scalar relation with a strong form and a
weak one.
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. We hope that this work
promises to be helpful in unravelling the intricacies of other discourse connectives also related to
causality such as donc, and the Explanation markers like parce que, puisque and car.
78
Evidence for a Scalar Analysis of Result in SDRT from a Study of the French Temporal Connective alors
References
Asher, N. 1993. Reference to Abstract Objects in Discourse. Dordrecht: Kluwer.
Asher, N. & Lascarides, A. (2003). Logics of Conversation. Cambridge: Cambridge University
Press.
Borillo, A.; Bras, M.; De Swart, H.; Le Draoulec, A.; Molendijk, A.; Verkuyl, H.; Vet, C.;
Vetters, C. & Vieu, L. 2004. Tense and Aspect. In Corblin, F. & de Swart, H. (eds.),
Handbook of French Semantics. Stanford: CSLI Publications. 233-348.
Bras, M.; Le Draoulec, A. & Vieu, L. 2001. French Adverbial Puis between Temporal Structure
and Discourse Structure. In Bras, M. & Vieu, L. (eds.), Semantic and Pragmatic Issues in
Discourse and Dialogue: Experimenting with Current Theories, CRiSPI series, vol. 9.
Amsterdam: Elsevier. 109-146.
Delort, L. & Danlos, L. 2005. Coordination of Causal Relations in Discourse'. In Aurnague, M.;
Bras, M.; le Draoulec, A. & Vieu, L. (eds.), Proceedings of SEM’05. 75-84. http://www.univtlse2.fr/erss/sem05/
Franckel, J.J. 1987. Alors, alors que. Bulag 13.
Gerecht, M.-J. (1987). Alors: opérateur temporel, connecteur argumentatif et marqueur de
discours. Cahiers de linguistique française 8: 69-79.
Gosselin, L. (in press). Contraintes pragmatico-cognitives sur l'ordre des constituants. Le cas de
séquences de connecteurs exprimant la consécution temporelle. In: Lane, P.
(ed.), Linguistique du texte et du discours. Rouen: Presses Universitaires de Rouen.
Hybertie, C. 1996. La conséquence en français. Paris-Gap: Ophrys.
Jayez, J. 1981. Etude des rapports entre l’argumentation et certains adverbes français. Thèse de
3ème cycle. Aix-Marseille 2.
Jayez, J. 1988a. Alors, descriptions et paramètres, Cahiers de Linguistique Française 9.
Jayez, J. 1988b. L’inférence en langue naturelle. Paris : Hermès.
Le Draoulec, A. & Bras, M. 2004. Rôles de alors temporel dans la structuration du discours.
Chronos 6, Genève, 22-24 sept. 2004.
Lewis, D. 1973. Causation., Journal of Philosophy, 70: 556-567, Reprinted in Lewis, 1986
Philosophical Papers II. Oxford: Oxford University Press. 159-213.
Reyle, U. 1998. A note on enumerations and the semantics of puis and alors. Cahiers de
Grammaire 23: 67-79.
79
SPRIK Conference
Explicit and Implicit Information in Text - Information Structure across Languages
University of Oslo, June 8-10 2006
Contrastive Lexical Pragmatics: A relevance-theoretic approach to lexical
narrowing and broadening in English and Norwegian original texts and translations
Ingrid Lossius Falkum
University of Oslo
[email protected]
This paper examines and compares the uses of the English words ‘happy’, ‘dry’, ‘bald’ and ‘flat’
and their corresponding Norwegian expressions in the English-Norwegian Parallel Corpus
(ENPC). The purpose of the present study was to examine and compare the workings of lexical
narrowing and broadening in closely related languages such as English and Norwegian, with the
aim of providing cross-linguistic evidence for or against the relevance-theoretic account of
lexical pragmatics (e.g. Carston 1997, 2002, Wilson 2003), as well as giving insights into the
relations between these languages with regard to the ways in which lexical-pragmatic processes
apply. How is the semantics/pragmatics distinction manifested in the relation between original
text/translation? To what extent does the English/Norwegian translator choose to explicitly
narrow/broaden a linguistically underspecified concept encoded by the original text, and what
may this tell us about the relation between English and Norwegian with regard to pragmatic
narrowing/broadening?
Relevance theory (Sperber & Wilson 1986/1995, Wilson & Sperber 2004), presents a unified
account of the pragmatic processes involved in on-line concept construction. The fact that the
concept expressed may be narrower (e.g. fish used to express ‘shark’) or broader than the
encoded one, as in the case of approximation (e.g. circular used to express ‘approximately
circular’), hyperbole (e.g. genius used to express ‘very intelligent’) and metaphor (e.g. dry used
to express ‘boring’) is explained in terms of “a single pragmatic process which fine-tunes the
interpretation of virtually every word.” (Wilson 2003: 1). On the relevance-theoretic account,
comprehension of a word involves constructing an ad hoc concept, a process that is assumed to
apply in the interaction between decoding and inference at the word level.
The corpus investigations of the word sets ‘happy’, ‘dry’, ‘bald’ and ‘flat’ and their Norwegian
equivalents raised some interesting issues. The investigation of ‘happy’ in the English originals
revealed that the Norwegian translator was forced, by lack of a comparable general expression in
Norwegian, to use a more specific expression in the translation, which would consequently
explicitly reflect the translator’s choice of direction and degree of narrowing, as illustrated by (1):
(1) a. When he licks you that’s his way of sayin’ I likes you and you makes me happy. (MM1)
Når han slikker deg, sier han egentlig “jeg liker deg, og du gjør meg glad”. (MM1T)
b. She would pass through the rooms where she’d been so happy all these years. (AT1)
Hun vil gå igjennom de rommene der hun hadde vært så lykkelig alle disse årene. (AT1T)
c. She had never wanted children, and would have been quite happy on her own, with her
numerous activities to fill up her spare time. (AB1)
Hun hadde aldri ønsket seg barn, og ville ha vært fornøyd med å leve alene, med sine
utallige aktiviteter å fylle fritiden med. (AB1T)
81
Ingrid Lossius Falkum
In the opposite case, the investigation of ‘happy’ in English translations showed that the relation
between the translation and the original was in most cases one of broadening, although English
readers do, of course, narrow in the course of interpretation. However, the result seemed to be
that there in some cases was a genuine loss of specificity in the translation from Norwegian into
English, as illustrated by the following example:
(2)
Når han kommer tilbake har bestefaren allerede spist fire fiskekaker og virker fornøyd.
(LSC1)
When he comes back, Grandfather has already eaten four fish cakes and looks happy.
(LSC1T)
In this example, the concept encoded by the Norwegian word, ‘fornøyd’ will be narrowed down
to the ad hoc concept FORNØYD* (the asterisk indicates an ad hoc concept) communicating that
the Grandfather has had enough food, that he is ‘full’. This sense does not seem to be equally
accessible in the English translation although it would be one of the possible interpretations
primed by use of the word ‘happy’. In this case one may therefore speak of a genuine loss of
specificity in the English translation.
The corpus investigations showed that there are many cases in which both lexical broadening and
narrowing are required in forming an ad hoc concept. The analysis of ‘dry’ showed that
approximation or hyperbole (i.e. broadening) may combine with narrowing in forming an ad hoc
concept; in the interpretation of collocations such as ‘dry hair’, there is both an element of
approximation or hyperbole (in that (parts of) living kinds may never be said to be strictly free of
moisture), and an element of narrowing, in that the hearer has to decide whether the object in
question is free from internal or surface moisture. This is illustrated by (3):
(3) a. I pulled my hood down and my hair was dry; Oliver’s was still soaking wet. (JB1)
Jeg tok av meg hetten og var tørr i håret, mens Oliver fremdeles var søkkvåt.
(JB1T)
b. … the oven that will bring profit, money, fresh bread, meat and fish to the listless
youngsters around her, bring back the shine to the dry reddish hair… (TB1T)
Bakerovnen som skal gi profitt, penger, ferske brød, kjøtt og fisk til de apatiske
ungene rundt henne, som skal gi dem glansen tilbake i det tørre, rødlige
håret… (TB1)
In interpretation of the collocation ‘dry hair’ in (3) a. and b., the direction of narrowing is crucial
in order to determine which proposition that is communicated by the speaker, both in the English
and the Norwegian version of the sentence. The fact that narrowing and broadening may combine
in forming an ad hoc concept may provide support for the relevance-theoretic view that lexical
narrowing and broadening both are the outcomes of one single pragmatic process that fine-tunes
the interpretation of every word. If these processes in fact constitute a continuum, we should
expect to find cases in which both processes apply in the course of interpretation, as noted by
Carston (2002).
The comparison of original texts and translations on the basis of the chosen word sets also raises
the question of whether the English and Norwegian readers may be said to end up with similar
concepts as a result of relevance-driven pragmatic inference, in cases where the concepts encoded
in the original and the translation differed in terms of specificity. In the case of ‘dry’ and ‘tørr’,
the corpus analyses revealed a remarkably high correspondence between the English and
82
Contrastive Lexical Pragmatics
Norwegian originals and translations, as well as a diverse range of corresponding metaphorical
uses, some of which are presented in (4) a.-c.:
(4) a. … still he was a tempting, enigmatic man, not dry or beaten like the other men she knew.
(JC1)
… fortsatt var han en tiltrekkende, spennende mann, ikke noen tørr taper, slik som de
andre mennene hun kjente. (JC1T)
b. … that caused Dorothy to make dry remarks, which Harriet felt like accusations. (DL1)
… som fikk Dorothy til å komme med noen tørre bemerkninger, som Harriet følte som en
anklage. (DL1T)
c. Kjente hvor hul og tørr jeg var. (EHA1)
Felt how empty and dry I was. (EHA1T)
This seems to indicate that the concepts encoded by ‘dry’ and ‘tørr’ may give access to largely
similar encyclopaedic information in the two languages, and that there may, in many cases, be a
considerable overlap in the concepts constructed by English and Norwegian readers. The
investigation of ‘bald’, however, revealed some problems in the English translation of the
Norwegian word ‘måne’, which the translator on one occasion had chosen to translate into ‘bald’,
and not ‘bald spot’, which would be the English equivalent to ‘måne’:
(7) a. Du har måne! (LSC1)
b. You are bald! (LSCT1)
The utterances in (7) seem to yield the construction of substantially different ad hoc concepts in
the original and in the translation, and hence give rise to different cognitive effects for English
and Norwegian readers, and one might ask to what extent the translator has succeeded in
producing a translation that is adequately relevant to the target audience in this case.
In the case of ‘flat’, however, some differences were observed between English and Norwegian
with regard to the metaphorical uses of the word, as illustrated by (8):
(8) a. She felt very flat, going to bed after all the gaiety… (RDA1)
Hun følte seg veldig tom ved tanken på å gå til sengs etter lystigheten… (RDA1T) (’tom’:
empty)
b.
Jess’s voice was confident and flat, without resonance, as if everything he might say
would be the simple truth. (JSM1)
Stemmen til Jess var trygg og rolig, som om alt han sa ganske enkelt var sant. (JSM1T)
(‘rolig’: calm)
The use of ‘flat’ as a translation of the English word ‘flat’ in (8) a. and b. would not yield the
intended cognitive effects for the Norwegian readers, In (8) a., the use of ‘flat’ in the Norwegian
translation would have communicated that the person in question had no particular feelings at all
about ‘going to bed after all the gaiety’, while in (8) b. the use of ‘flat’ as a collocate of ‘trygg’
(‘confident, secure’) would give an odd result in Norwegian. Hence the translator has chosen a
word that s/he judges to be more apt to communicate the meaning expressed in the original text.
According to the relevance-theoretic account of metaphor, the hearer constructs, via metaphorical
extension, an ad hoc concept on the basis of encyclopaedic information made accessible by the
concept, guided by expectations of relevance raised by the utterance itself. If the Norwegian word
‘flat’ cannot be used in the contexts of (8) a. and b. to convey the metaphorical senses
83
Ingrid Lossius Falkum
communicated by the English original, this may be because there is no information made
accessible by the concept encoded by this word that is compatible with or relevant enough to the
overall interpretation of the utterance, or because this information is not accessible enough to be
appropriately used. This indicates that the concepts encoded by this word give access to different
encyclopaedic information, or to similar information organised in different ways, in English and
Norwegian. It seems therefore likely that English and Norwegian readers will end up constructing
different concepts in some of the cases where the word ‘flat’ occurs in both the original and the
translation.
Conclusion
In summary, the investigations of ‘happy’, ‘bald’, ‘dry’ and ‘flat’ and their Norwegian
equivalents in the ENPC showed a remarkable flexibility of lexical narrowing and broadening in
both languages. The analyses revealed that each of these words may give rise to a number of
different ad hoc concepts, and that narrowing and broadening may combine in forming a
communicated concept. The inferential account of interpretation proposed by relevance-theorists
predicts such flexibility in the construction of ad hoc concepts. The results of this study may
therefore provide further support for the relevance-theoretic view according to which narrowing
and broadening are the outcomes of a single process that fine-tunes the interpretation of virtually
every word, and for their claim that both processes contribute to the explicit content of a given
utterance.
This study also revealed that parallel corpora present useful data for the study of lexical
pragmatics, not only because they facilitate and make accessible the comparison of original texts
and translations, but also because they reveal important facts about the pragmatic inferences
made by the translators in the course of translation. These valuable sources of data should be
further explored in future studies of lexical narrowing and broadening.
References
Carston, Robyn. 2002. Thoughts and Utterances: The Pragmatics of Explicit Communication.
Oxford: Blackwell.
Carston, Robyn. 1997. Enrichment and loosening: complementary processes in deriving the
proposition expressed? Linguistische Berichte 8: 103-127.
Sperber, D. & Wilson, D. 1986/1995. Relevance: Communication and Cognition. 2nd edition.
Oxford: Blackwell.
Wilson, D. 2003. Relevance theory and lexical pragmatics. Italian Journal of Linguistics/ Rivista
di
Linguistica
15.2:
273-91.
(Cited
version
downloaded
from
http://www.phon.ucl.ac.uk/home/deirdre/)
Wilson, D. & Sperber, D. 2004. Relevance Theory. Horn, L.R. & Ward, G. (eds.) The
Handbook of Pragmatics. Oxford: Blackwell: 607-632.
84
SPRIK Conference
Explicit and Implicit Information in Text – Information Structure across Languages
University of Oslo, June 8-10 2006
Explaining connections in Akan discourse: the role of discourse markers
Nana Aba Appiah Amfo
Norwegian University of Science and Technology
[email protected]
Relevance theory (Sperber and Wilson 1986, 1995) explains utterance interpretation as
not just a matter of identifying the proposition expressed, but crucially it involves the
recovery of intended cognitive or contextual effects. These contextual effects could either
come in the form of a communicated assumption combining with an already existing one
to yield a contextual implication, strengthening of an existing assumption and finally
contradicting an existing assumption which leads to further elimination. The contextual
effects need to be obtained within a context. Consequently access to an appropriate
context is crucial in retrieving the intended contextual effects. Context as defined within
the relevance-theoretic framework is not confined to the co-text. It is a psychological
construct involving a subset of the hearer’s assumptions about the world. A particular
utterance may contain certain linguistic items which help in making connections between
different parts of the discourse or of a single utterance, by triggering certain contextual
assumptions which subsequently form part of the context within which an utterance is
interpreted.
Blakemore (1992) suggests that connections between parts of a discourse can be made
either by intonation or by the use of discourse connectives or markers. The use of these
markers practically ensures the correct context selection, which subsequently leads to
retrieval of the intended contextual effects.
Blakemore (1992, 2002) examines the role of discourse markers as constraints on the
interpretation process by exemplifying almost exclusively from the English language. She
suggests that discourse connectives can be classified into three categories, parallel to the
kind of contextual effects that can be achieved by processing an ostensive verbal
stimulus. Connectives such as after all, besides, moreover and furthermore introduce
additional evidence in a bid to strengthen an earlier held assumption by the addressee.
Connectives which direct the addressee to drop an earlier held assumption in favour of
the one which is just about to be introduced include however, nevertheless and but.
Finally, connectives such as too and also interact with the phenomenon of focus to direct
the addressee to derivation of the contextual effects which are parallel to some already
derived contextual effects, possibly derived from the immediately preceding discourse.
This paper demonstrates that in spite of language specific grammatical and cultural
features which may consequently lead to certain idiosyncratic pragmatic implications, a
general cognitively based theory of communication such as Relevance theory is enough
to account for the use of discourse markers cross-linguistically. At the same time, it
draws attention to certain language internal details which ought not to be lost in the
pursuit of a general theory of utterance interpretation.
85
Nana Aba Appiah Amfo
This paper focuses on the use of three discourse markers in Akan, a Niger Congo (Kwa
branch) language. They are nso, na and nanso. Nso is one of a number of focus markers
in Akan. Na is the mundane clausal coordinating connective in Akan and nanso is the
contrastive connective which basically communicates some form of contrast between the
two propositions it conjoins. These markers make distinct contributions to the
interpretation process by highlighting certain connections which need to be made
between parts of an utterance or the discourse as a whole.
Nso may be classified as an additive focus marker (cf. König 1991) which may take
scope either over a noun phrase as in (1), or a clause as in (2). Its presence in an utterance
signals to the interlocutor to look for parallelisms of context within the immediately
preceding discourse. The immediately preceding utterance shares an entailment with the
nso-utterance. In (1) the use of nso signals that some people in addition to the speaker
were at the location referred to by the locational adverb, hç. The proposition expressed in
the second conjunct of (2) is to be interpreted against the background that the referent of
the third person pronoun, Abena, has already performed some domestic chores, namely
cooking. The relevance of why a speaker encourages this kind of parallelism, by the use
of nso, is contextually determined. It may be for instance to encourage the addressee to
derive an implicature such as “Abena is tired and cannot perform any more domestic
chores”.
(1)
Me nso me kç-ç
hç bi.
I nso I go-COMPL there some.
‘I was also there.’
(2)
Abena noa-a
aduane na
ç-si-i nneEma nso.
Abena cook-COMPL food CONJ she-wash things nso
‘Abena cooked and did the laundry as well’
However, the use of nso in (3) does not give access to any parallel context. What it does
is to communicate a speaker attitude of disdain, disapproval, possibly even disgust. The
speaker of (3) in addition to communicating the proposition that Kofi is pompous,
certainly succeeds in communicating the fact that she is disgusted at such an attitude
displayed by Kofi.
(3)
Kofi nso ç-kyerE neho papa.
Kofi nso he-show REFL well.
‘Kofi is indeed pompous.’
These two uses of nso appear unrelated and should probably be considered simply as
homonyms. However given that so-called additive focus markers such as English too and
Norwegian også may sometimes communicate a speaker attitude of disdain, possibly
even protest (as in (4) and (5) respectively), albeit in colloquial or even child speech, it
may be worthwhile investigating the kind of relationship existing between these two uses
of nso.
(4)
I did too!
(5)
Det gjorde jeg også!
86
Explaining connections in Akan discourse: the role of discourse markers
Na is the ordinary clausal coordinating connective in Akan, equivalent to English and. As
noted in Amfo (2006), its use in an utterance indicates to the interlocutor to look out for
certain kinds of inferential relations between the conjuncts. These relations could be
temporal, causal, parallel, contrastive or even explanatory as illustrated in examples (6) to
(10) below.
(6) Me wie-e
m’adwuma no, na
me pue-E.
I finish-COMPL my’work DCM, CONJ I go.out-COMPL
‘When I finished my work, and then I went out’
(I went out after I had finished my work)
fam wç sukuu
na ç-nya-a kuro.
(7) Adjoa bç-ç
Adjoa hit-COMPL ground be school, CONJ she-get-COMPL sore.
‘Adjoa fell at school and (as a result) got a sore’
(8) ç-n-ni
abotare
na ç-yE dwEE.
He-NEG-has patience CONJ he-be arogant
‘He is impatient and he is arrogant.’
(9) Na me-m-pE n’asEm,
na
sesei deE me-a-hu sE ç-yE nipa papa.
Then I-NEG-like his’matter, CONJ now FM I-PERF-see COMP he-be person good
‘I didn’t use to like him but now I have realized he is a good person.’
(10)
To
wo
po na aha yE toro.
Throw your chest CONJ here be slippery
‘Take your time because it is slippery over here.’
Combined with contextual information, the presence of this connective, gives the
addressee access to contextual information which aids in arriving at the intended
interpretation. The fact that a given token of a na-conjunction can give rise to an
explanatory inferential relation between the individual conjuncts and specifically that the
second conjunct is considered as an explanation of the proposition expressed in the first
conjunct, is in contradistinction to Carston (2002), Blakemore (1992) and Blakemore and
Carston’s (2005) claim that “one conjunct cannot function as an explanation for the state
of affairs described in the other” (Carston 2002: 245), and especially not in such a way
that the explanation appears in the final conjunct.
Na not only indicates what kind of connections need to be made between two or more
conjuncts of a particular utterance, but also it indicates the kind of inferential relations
which exists between parts of a discourse, the latter role highlighted by its utteranceinitial use as in (11).
(11)
Na
w’a-tua
ne
ka no nyinaa a-wie?
CONJ she’PERF-pay POSS debt DEF all PERF-finish
‘And has she finished settling all her debt?’
Finally, nanso can be argued to consist of two morphemes, na and nso, which have
possibly lexicalized over time and is now considered as a single word rather than a
sequence of two words, an assumption corroborated by the way the word is written. The
use of nanso indicates to the interlocutor that what follows is contrary to what may be the
87
Nana Aba Appiah Amfo
usual expectation as illustrated in (12), where the speaker expected that due to her early
arrival she will meet the referents of the third person plural object pronoun, wçn.
(12)
Meduru-u sukuu hç ntEm, nanso m’a-n-kç to wçn.
I-arrive- COMPL school there early, nanso I’COMPL-NEG-go reach them.
‘I arrived at the school in good time but I didn’t meet them’
This study seeks to show that coherence relations within single utterances and between
parts of a discourse can generally be achieved by the use of discourse markers since their
use provides procedural information (cf. the relevance-theoretic conceptual/procedural
distinction) leading to the identification of the kind of contextual information the
addressee needs to access, in his bid to arrive at the intended interpretation. However in
pursuit of a general cognitive theory of utterance interpretation we need to proceed
cautiously recognizing certain language-internal facts, as illustrated for instance by the
use of na in (10), where the inferred logical relation between the conjunct propositions is
one that is alien to and-conjunctions in European languages.
References
Amfo, Nana Aba Appiah 2006. Clausal Conjunction in Akan. ms. NTNU
Blakemore, Diane 1992. Understanding Utterances: an introduction to pragmatics,
Oxford: Blackwell Publishing.
Blakemore, Diane 2002. Relevance and Linguistic meaning: the semantics and
pragmatics of discourse markers. Cambridge: Cambridge University Press.
Cambridge Studies in Linguistics 99.
Blakemore, Diane and Robyn Carston 2005. The Pragmatics of Sentential Coordination
with and. Lingua, Vol 115: 4, 569-590.
Carston, Robyn 2002. Thoughts and Utterances: the pragmatics of explicit
communication. Oxford: Blackwell Publishing.
König, Ekkehard 1991. The Meaning of Focus Particles: a comparative perspective.
London: Routledge.
Sperber, Dan and Deirdre Wilson 1995 Relevance: communication and cognition,
Oxford: Blackwell Publishing, 1st edition 1986.
88
SPRIK Conference
Explicit and Implicit Information in Text - Information Structure across Languages
University of Oslo, June 8-10 2006
The metarepresentational use of main clause phenomena in embedded clauses
Thorstein Fretheim
Norwegian University of Science and Technology
[email protected]
Consider the direct report in (1) and the indirect report in (2).
(1) "Mick is just as scared as me," she thought.
(2) She realized that Mick was as scared as her.
An utterance of (1) metarepresents (cf. Sperber 2000) a thought attributed to the woman referred
to. What looks like a quotation, a case of direct speech, is just an imitation of what could have
been uttered ― a metarepresentation of something the woman could have said at the time when
the thought reported in (1) entered her mind, but which was never vocalized. What I've already
described as an 'indirect report' presented in (2) has a rather more open speech act potential. It
could be used to perform an indirect report of the woman's thought, as in (1), but it could also
report something the woman had said (a true case of 'indirect speech'), or it could be used to
affirm the speaker's belief that the complement clause expresses a true proposition and that the
woman had acknowledged that fact. In the latter situation the speaker of (2) confirms the main
clause proposition, or confirms just the complement proposition and volunteers the main clause
information about the woman's attitude to the embedded proposition.
The different uses of a sentence such as the one shown in (2) will typically occur with different
intonation patterns when spoken. A speech act that confirms the truth of the complement
proposition or the main clause proposition will normally be produced with a pitch accent on the
verb form realized, because the main clause verb conveys the new information in the utterance
and the following complement contains topical information. However, if the pragmatic import of
an utterance of (2) is meant to equal that of the report in (1), there will be a pitch accent on at
least one item in the complement clause, for instance one on Mick and maybe also one on her, or
one on the first token of as and one on her. These are prosodic features that may be said to
mimic certain accentual aspects of an imagined utterance performed by the female referent.
When two strong accents are located inside the complement, that part of the complex sentence
has its own topic-comment structure with the subject Mick as topic. It is a kind of utterance
prosody indicative of free indirect speech, or what Banfield (1993) refers to as 'represented
speech or represented thought', though the echoic elements are not the standard morpho-syntactic
features typical of free indirect speech, as illustrated by (3), from Banfield (1993: 342).
(3) Oh, she was so pleased to see him ― delighted!
The day was so charming, didn't he agree?
(Katherine Mansfield, "Miss Brill", in Stories, 1956)
89
Thorstein Fretheim
The co-occurrence of a specific syntactic form and certain intonational features can trigger a
processing of the stimulus as a metarepresentation of a thought attributed to some contextually
retrievable individual. Pitch-based vocal gestures can be used to mimic structural features that
could have appeared in a linguistic expression, imagined or real, of the thought metarepresented
by the reporting speaker.
The present paper first examines the interaction of intonational form, word order and certain
main clause phenomena which can appear in an embedded Norwegian clause that metarepresents
a thought or a speech act, and then I am going to look at the interaction of intonational form and
certain modal particles in embedded clauses.
Consider (4a) and (4b).
(4) a.
Selv om hun faktisk syntes det var greit, hun, så visste hun om folk som var opprørte.
even if
she actually thought it was ok
she, then knew she about people who were upset
"Although she actually felt it was ok, herself, she knew people who were upset."
b.
*Selv om hun syntes faktisk det var greit, hun, så visste hun om folk som var opprørte.
even if
she thought actually it was ok
she, then knew she about people who were upset
The right-dislocated resumptive copy of the subject pronoun hun ('she') reveals that the
concessive clause metarepresents the female referent's thought, as interpreted by the speaker. The
right dislocation construction within the concessive clause in (4a) is a 'main clause phenomenon',
so it may prompt the hearer to process the concessive clause as an instance of free indirect
speech. (4a) could echo the woman's performance of an utterance such as (5), whose structural
elements resemble those found in (4a), but it could just as well be used to metarepresent an
utterance of (6) performed by the same woman, even though there are hardly any formal features
in (4a) that would remind one of (6). The way that the concessive clause in (4a) is structured, it
would serve as a faithful interpretation of either one of (5) or (6).
(5) Jeg syns faktisk det er helt greit, jeg.
I
think actually it
is quite ok
I
"I actually think it's quite ok, myself."
(6) Jeg har ikke noe imot det.
"I have nothing against it."
There is a difference in word order between (5) and the concessive clause in (4a). The sentence
adverb faktisk precedes the finite verb syntes in (4a), and this is the Norwegian subordinate
clause word order. A sentence like (5), on the other hand, will necessarily have the main clause
word order: the verb precedes the sentence adverb. Observe that (4b) is ill-formed precisely
because the verb syntes in (4b) precedes the adverb faktisk, as if the embedded concessive clause
were a main clause.
Though the main clause word order is disallowed when the connective is concessive selv om, the
dislocated pronoun hun mimicking a 'main clause feature' typical of colloquial speech is
permitted in (4a). The apparent mismatch between embedded clause word order and main clause
right dislocation is fully acceptable. Notice, however, that the dislocated pronoun in (4a)
90
The metarepresentational use of main clause phenomena in embedded clauses
constrains the speaker's choice of intonation pattern for an utterance of (4a). There must be a
phrase accent, manifested as a fundamental frequency (F0) peak in East Norwegian, on the topic
hun in the subject position, and then another phrase accent on the final item belonging to the
(information) focus. In such right dislocation constructions, the focus is structurally bounded by
the two tokens of hun (Fretheim 2001).
(4a')
[Selv om HUN faktisk syntes det var GREIT, hun]Intonation Unit (IU)
[så VISSTE hun om folk som var OPPRØRTE]IU
"Although SHE actually felt it was OK, herself, she KNEW people who were UPSET."
The presence of two phrase accents in a single Intonation Unit (henceforth IU), as opposed to
just one, is a main clause phenomenon. Whenever an IU with two phrase accents coincides with
an embedded clause, the clause will be interpreted as the linguistic vehicle of a separate
illocutionary act, most typically an act with assertoric force, or else as a report on someone's
performance of an assertion. Thus the double occurrence of a phrase accent in the concessive
clause of (4a') informs the hearer that the proposition expressed in that clause is not presupposed
but is asserted to be true ― in spite of the linear order of the two clauses. Of the two phrase
accents (or focal accents) in each of the IUs in (4a'), one represents the topic and the other one
the focus of the construction. The linear order of the phrase accents does not indicate the location
of the topic and the focus; rather, whatever is more likely to be the topic, depending on
contextual premises as well as the category membership of the phrase-accented grammatical
constituent, will be processed as topic. There is no phonological difference between a
prosodically highlighted topic constituent and a prosodically highlighted focus constituent
(Gundel and Fretheim 2004). The phrase meaning 'people who were upset' which is highlighted
by a phrase accent in the main clause of (4a') will be processed as a topic that contrasts with the
topical female referent whose sentiment is metarepresented in the preceding clause (and IU).
That woman's recognition of the fact that some people were upset is the main clause focus.
A modal particle is another main clause phenomenon, quite obviously, because modal particles
encode a range of attitudes to the proposition expressed and/or to the speech act performed. In
Figure 1 on the next page (the notation underneath the intonation contours will be explained in
my presentation) the location of the utterance-final modal particle vel outside the final Accent
Unit (AU) cues a processing of vel which relates it to the speaker's speech act and helps the
hearer identify it as a request for confirmation. In Figure 2, whose string is identical to the one in
Figure 1, vel is inside the final AU and may therefore be parsed either as in Figure 1 or as a
particle belonging to the embedded clause proposition. An occurrence of vel inside the embedded
clause opens for a processing of that clause as the vehicle of a metarepresented thought, but
Figure 2 could also be interpreted in the same way as Figure 1, depending on the context
accessed by the hearer. Finally, the particle vel is accented in Figure 3 (it even has a phrase
accent), which shows that vel belongs to the embedded clause and the speaker metarepresents the
woman's wondering whether her headache would disappear, a case of 'free indirect thought'.
91
Thorstein Fretheim
Figure 1
Hun tenkte at hodepinen forsvant vel
230
F0 (Hz)
160
80
0
2.4
H* L H
2
H* L
H
2
1
H— H%
L
[[hun [ tenkte-at]AU [ hodepinen-for]AU [ SVANT]AU vel ]FP ]IU
"She thought that the headache would disappear, didn't she?"
Figure 2
Hun tenkte at hodepinen forsvant vel
230
F0 (Hz)
160
80
0
2.4
H* L
[[hun
2
H
H* L
H
2
[ tenkte-at]AU [ hodepinen-for]AU
1
L
H—/H%
[ SVANT-vel]AU ]FP ]IU
"She thought that the headache would disappear, wouldn't it?", or
"She thought that the headache would disappear, didn't she?"
Figure 3
Hun tenkte at hodepinen forsvant vel
230
F0 (Hz)
160
80
0
2.4
H* L
2
H
2
H* L
H
1
[[hun [ tenkte-at]AU [ hodepinen]AU for [
L
H— L H—/H%
SVANT]AU ]FP
[[1VEL]AU ]FP ]IU
"She thought that the headache would disappear, wouldn't it?"
92
The metarepresentational use of main clause phenomena in embedded clauses
A grammatical construction in which a right-detached pronoun copies a subject NP at the
main clause level does not tolerate an intervening embedded clause that contains main
clause elements of its own. Thus (7) is a well-formed structure but (8) is not.
(7)
Hun tenkte at [HODEPINENi forSVANT nok, deni]
she thought that the headachei disappeared ModPart iti
"She thought that the headache would probably disappear."
(8)
*Huni tenkte at [HODEPINEN forSVANT nok], huni
shei thought that the headache disappeared ModPart shei
The Modal Particle nok, here glossed as 'probably', is to the right of the finite verb in the
complement clause, and this main clause feature is compatible with the lower clause right
dislocation in (7) where the pronoun den ('it') corefers with the NP hodepinen ('the
headache'), but it is not compatible with the upper clause dislocation in (8) where the
coreferential items are the two tokens of hun ('she'). (8) is bad because whatever occupies
the slot between the two coreferential items in a Norwegian right dislocation construction
must be the focus of information (Fretheim 2001), but the string bounded by such
coreferential items cannot contain a full topic-focus structure like the one in (8) where
both the complement subject NP and the following verb get a phrase accent.
While many languages permit a range of main clause phenomena in complements of
verba sentiendi or dicendi as well as in causal clauses that follow the main clause, a
Norwegian main clause intonation pattern can even be imposed on a restrictive relative
clause in an act which metarepresents someone's thought. The relative clause in (9) is
intoned in a way that echoes Jens' original assertion.
(9)
[De som Jens FØLTE han hadde SVIKTET den gangen]IU, [FØLTE ikke det SELV]IU
"Those whom Jens felt he had let down at that time, didn't feel that themselves."
I must confess I do not know to what extent this liberty is something that even speakers
of non-Scandinavian languages can avail themselves of.
References
Banfield, Ann. 1993. The development of represented speech and thought. In: Lucy, John
A. (ed.), Reflexive Language: Reported Speech and Metapragmatics. Cambridge:
Cambridge University Press.
Fretheim, Thorstein. 2001. The interaction of right-dislocated pronominals and
intonational phrasing in Norwegian. In: van Dommelen, Wim A. & Fretheim,
Thorstein (eds.), Nordic prosody: Proceedings of the VIIth Conference. Frankfurt am
Main: Peter Lang, 61-75.
Gundel, Jeanette K. & Fretheim, Thorstein. 2004. Topic and Focus. In: Horn, Laurence
R. & Ward, Gregory (eds.), The Handbook of Pragmatics. Oxford: Blackwell
Publishing, 175-196.
Sperber, Dan (ed.). 2000. Metarepresentations: An Interdisciplinary Perspective. Oxford:
Oxford University Press.
93
SPRIK Conference
Explicit and Implicit Information in Text - Information Structure across Languages
University of Oslo, June 8-10 2006
Discourse Structure in Optimality Theoretic Pragmatics
Henk Zeevat
University of Amsterdam
[email protected]
Typical of the influential approaches to parsing discourse structure (the discourse grammar of
Scha and Polanyi and the SDRT approach of Asher and Lascarides) is that they contain a
number of apparently unmotivated defaults or preferences. The talk tries to unify these
defaults or preferences with the well-known preferences or defaults in presupposition
resolution and accommodation by proposing a single system of pragmatic constraints, in the
tradition of optimality theory.
The crux is the assumption that all assertions address a topic and that the interpreter must
identify that topic much in the same way as (other) given material that is introduced in the
sentence.
In the interaction with the system of constraints specified below, it gives the desired effect.
FAITH > CONSISTENCE > *NEW > RELEVANCE
FAITH is the principle that the current sentence must be optimal (using a separate syntactic
system) for the candidate interpretation. CONSISTENCE minimises the conflict of the
interpretation with the context, RELEVANCE is the principle that whereever the sentence can
address an activated topic the speaker is taken to intend to settle it and wherever the utterance
can be interpreted as relevant for an activated goal the speaker is taken to intend to contribute
to reaching that goal.
The constraint that gives the discourse structure preferences is *NEW. It militates in all things
against constructing new things and - where that is unavoidable - still tries to make the new
things maximally related to old things. Applied to topics, it tries to preserve the old topic and
thereby gives rise to a preference for topic preserving discourse relations and in combination
with relevance to the principle of maximising discourse coherence.
This way of looking at discourse structure also gets rid of two artificial features of existing
approaches: the assumption that every element of a discourse is linked together with
rhetorical relations (the identification of a given topic can be impossible) and the postulation
of topics where none are clearly given or could be recovered by formal means.
Topics have been identified as the question that the current utterance answers, as what the
current sentence is about and as the part of the current utterance that is given. On all of these
notions, recognising the topic is essential for getting the point of an utterance and a fortiori for
the interpretation of the utterance. It is thereby even more central than resolving pronouns or
temporal anaphora or finding the correct interpretation of the lexical words: if one has a
picture of what the speaker is trying to do one can tolerate some errors in identifying the
content, a correct understanding of the content is pointless if one cannot figure out what the
speaker wants to do with it.
95
Henk Zeevat
*NEW on NPs is the preference for the most activated referents with FAITH possibly
overruling it (indefinites, other-marking). This leads to subsectional and bridging anaphora
when activated referents are not available.
In much the same way, *NEW on topics prefers the most activated topics, preferably the topic
of the last sentence, with as second-best a super-topic of that topic. And in the absence of that,
it tries to create a subtopic of the most activated topic or to anchor the topic of the current
utterance in whatever is given.
This leads to the well-known discourse relations: the reformulation, the continuation of a
story or a list, the creation of an elaboration or a list (subtopic) and finally to explanations,
backgrounds and results, where the topic is anchored in the event of the last utterance or other
activated participants.
The preferences ("defaults") that have been standardly assumed to obtain in approaches to
inferring rhetorical structure can be reduced to *NEW assuming a component that decides
between different constructions on the basis of the content of the utterance, an admittedly
knowledge-intensive procedure, but well within the abilities of the semantic memory usually
attributed to human language users.
Marking in this area is optional and typically uses FAITH to override the workings of *NEW.
The relation of contrast -marked by contrastive and adversative particles and connectors and
by intonation- indicates a change in topic -often in the opposite direction. Causal and
temporal markers have to be used when the defult is other than intended by the speaker. When
that is not the case they are clumsy.
What is often called relevance maximisation is largely reduced to the workings of *NEW on
topics. For example, Grice's petrol station should be open because the speaker is taken to
address the question how the addressee can get some petrol. This is established as a topic by
the other speaker putting an explicit question. The answer is interpreted as adressing that topic
and does so on the basis of *NEW. The relevance principle merely makes it the case that
where other open questions and goals seem to be addressed, they are indeed taken to be
addressed. In that way, certain kinds of pragmatic strengthening such as exhaustive
interpretations, interpreting "when"-clauses as causal, going for the logically strongest
accommodations or interpreting next events as results when possible are taken care of. In
some these cases, *NEW forces a minimally less activated topic and Relevance decides which
one that is.
The main claim of the paper is that a simple system of OT pragmatics is able to capture three
sets of preferences in exactly the same way: in presupposition, in anaphora and in discourse
structure. Discourse structure arises from the sole assumption that topic is an important
discourse parameter and managed in a way that is quite similar to the discourse objects that
play a role in nominal and temporal anaphora.
The full paper derives the constraints and their ordering from standard views on
presupposition and addresses one possible counterargument: the existence of pragmatic
marking that is not directed against *NEW but rather seems to promote it, like e.g. particles
marking certain material as old.
References
Reinhard Blutner and Henk Zeevat. 2004. Pragmatics in Optimality Theory. Palgrave
Macmillan.
96
Discourse Structure in Optimality Theoretic Pragmatics
Alex Lascarides and Nicholas Asher. 2003. Logics of Conversation. Cambridge University
Press.
Remko Scha and Livia Polanyi: "An Augmented Context Free Grammar for Discourse." In
Proceedings of the 12th International Conference on Computational Linguistics. Budapest,
August 1988. Pp. 573-577
Hub Prüst, Remko Scha and Martin van den Berg. 1994. "Discourse Grammar and Verb
Phrase Anaphora." Linguistics and Philosophy 17, pp. 261-327.
97
SPRIK Conference
Explicit and Implicit Information in Text - Information Structure across Languages
University of Oslo, June 8-10 2006
Entailment, assertion, and textual coherence: a case study
Patrícia Amaral
Ohio State University, USA/Universidade de Coimbra, Portugal
[email protected]
This paper focuses on the relation between entailed content, asserted content, and textual
coherence, by analyzing the semantic-pragmatic properties of the approximative adverbs
(henceforth AA) almost and barely, a pair of adverbs which have been the subject of much debate
in the literature (cf. Sevi 1998, Horn 2002). The current study contributes to this research by
providing experimental evidence for the role that AAs’ entailments pertaining to the truth or
falsity of the modified predicate play in textual coherence.
In previous research, the meaning of AAs has been accounted for as the conjunction of two
propositional components, named by Horn (2002) the polar and the proximal components,
defined below (following Sevi 1998, Rapp and von Stechow 1999, Horn 2002):
(1) Let w be any world:
(I) ALMOST (w)(p) = 1 iff (a) and (b) hold:
(a) ∃ w’ s.t. CLOSE (w’,w) & p(w’) = 1 [Proximal component]
(b) p(w) = 0 [Polar component]
(II) BARELY (w)(p) = 1 iff (a) and (b) hold:
(a) ∃ w’ s.t. CLOSE (w’,w) & p(w’) = 0 [Proximal component]
(b) p(w) = 1 [Polar component]
There is robust evidence for the asymmetric status of the two components. On the one hand, only
the proximal component can be the target of a yes/no question, and support a causal or a
concessive link (Schwenter 2002, Ziegeler 2000); on the other hand, the distribution of negative
polarity items (NPIs) goes against what would be predicted, given the polar component of almost
and barely (Horn 2002). Cross-linguistically, several analyses have pointed to the scalar
orientation of AAs and to their bearing on textual coherence. In particular, it has been noted that
AAs induce inference patterns oriented towards an expected or ideal standard provided by the
proximal component (Ducrot 1973 and Champaud and Bassano 1987 for French, Lundquist and
Jarvella 1994 for Danish and Finnish, Schwenter 2002 for Spanish, Li 1976 for Mandarin
Chinese, Klein 1997 for Dutch, Sevi 1998 for Hebrew and English).
These observations pose a challenge for the analysis of AAs, in particular with respect to the
theoretical status of the polar component. Proponents of the conversational implicature approach
(Sadock 1981, Atlas 1984) face a major difficulty, namely that the purported status of
conversational implicature is hard to reconcile with the resistance to explicit cancellation
exhibited by the polar component (Hitzeman 1992, Horn 2002). The implicature view is unable
99
Patrícia Amaral
to account for the fact that (2) is contradictory, which would be predicted if the polar component
were considered an entailment:
(2) #Mary almost finished dinner and she finished dinner.
On the other hand, standard tests for presupposition do not seem conclusive with respect to AAs.
As for the entailment approach, the “backgroundedness” of the polar component is problematic.
In discourse, a sentence containing an AA licenses continuations that would coherently follow
the proximal, not the polar component (i.e. are argumentatively oriented toward the proximal
component, in Anscombre and Ducrot’s terms). This asymmetry can be conceived in terms of
context update; only the proximal component is added to the common ground and presented as
potentially controversial, i.e. only the proximal component is asserted (Stalnaker 1978), although
both the proximal and the polar components are entailed (Horn 2002). This is the approach
adopted in this paper.
The experimental design of the present study builds on an analogy with research on quantifiers
like few, not many, a few, most (Moxey and Sanford 1993, Moxey and Sanford 1998, Paterson et
al. 1998), which is reminiscent of some of the issues discussed above. These studies show that
the meaning of quantifiers can be differentiated by the inference patterns that they license. In
production and comprehension experiments, speakers tend to agree about the possible
continuations that are felicitous after a sentence with a quantified expression, as in (3)-(5)
(Moxey and Sanford 1993: 60):
(3) Not many of the lecturers are fast talkers.
(4) They should slow down for the introductory classes. [?]
(5) They should speed up to be less boring. [OK]
While (5) is a felicitous continuation of (3), (4) is less acceptable. The use of “negative” (e.g. not
many) or “positive” quantifiers seems to involve a “focal bias” in that the set of entities being the
focus of attention in interpretation is either the set of which the predicate in (3) is false (the
complement set, or compset), or the set of which it is true (the reference set, or refset).
In the present study, a similar methodology is used in reading and judgement tasks to test the
interpretation of sequences of sentences, the former containing an AA, the latter containing a
connective that retrieves one of the meaning components of the AA. This contextual
manipulation makes it possible to target each of the propositions entailed by the sentence that
contains an AA, in order to test the inference pattern licensed by the adverb, as exemplified in (6)
and (7):
(6) John can almost swim. Therefore he’ll enjoy a day at the pool with his friends./#Therefore
he’s afraid of drowning.
(7) John can barely swim. #Therefore he’ll enjoy a day at the pool with his friends./Therefore
he’s afraid of drowning.
The main hypothesis is that almost behaves like a “positive” quantifier in that it favours
continuations coherent with the truth of the predicate over which it has scope, whereas barely
licenses continuations consistent with the falsity of the predicate, therefore showing the
“compset” pattern of negative quantifiers.
100
Entailment, assertion, and textual coherence: a case study
Despite the fact that almost modification entails the negation of the modified predicate (Hitzeman
1992, Sevi 1998, Rapp and Von Stechow 1999), and barely modification entails the truth of the
predicate, the felicitous continuation in (6) and (7) is not the sentence that would coherently
follow the respective entailment (i.e. the “polar” component), as shown in (8) and (9):
(8) John cannot swim. #Therefore he’ll enjoy a day at the pool with his friends./Therefore he’s
afraid of drowning.
(9) John can swim. Therefore he’ll enjoy a day at the pool with his friends./#Therefore he’s afraid
of drowning.
The hypothesis was tested using online tasks in two different experiments. In the first experiment,
subjects were asked to read pairs of sentences as in (6) and (7) above. Then a “makes sense”
judgement was elicited, and subjects rated the pair using a quantitative scale. In the second
experiment, subjects were provided with a sentence containing an AA followed by a sentence
consisting either of the polar or of the proximal component, and were asked to make a True/False
judgement.
The first experiment was intended to test the role of each of the components in textual coherence.
The prediction was that continuations consistent with the proximal component would be ranked
higher than continuations consistent with the polar component. In other words, the coherent
continuation was predicted to be the one that builds on the asserted content. As for the second
experiment, subjects were predicted to answer in accordance to the polar component, the
entailment which pertains to the truth of the modified predicate in the world of evaluation (recall
(1)). According to the “entailment” approach adopted in this paper (Horn 2002), this component
is part of the conventional, truth-conditional content of the adverb. The two tasks were intended
to provide insight into the different use of entailed information depending on its assertoric force;
in the first experiment, the sentence containing the AA occurs in contextual integration and in the
second experiment it is interpreted in isolation.
The results confirmed the first hypothesis; for both adverbs, continuations consistent with the
proximal component were ranked higher than continuations consistent with the polar component,
with a significant effect of the interaction adverb × continuation (F1[1,44] = 63.367, p < .01,
F2[1,16] = 37.223, p < .01). Moreover, the asymmetry of the two components with respect to
textual coherence is more notorious for barely, since the difference between the ratings for the
possible continuations was significantly greater for the sentences with this adverb than for
sentences with almost (F1[1,44] = 59.381, p < .01).
The results also confirmed the second hypothesis, showing that subjects respond in accordance to
the polar component when the sentence is interpreted in isolation, as shown in Figure 1 below.
The interaction adverb × component was significant (F1[1,36] = 36.126, p < .01), and a
significant effect of adverb (F1[1,36] = 45.294, p < .01) and component (F1[1,36] = 5.041, p <
.05 (.031)) were also obtained. For almost, when the targeted component was the polar
component, subjects had a greater difficulty in determining the truth conditions (55,5% of “True”
responses vs 44,5% of “False” responses), a result which is further supported by the delayed RTs
obtained in the respective condition (“A cannot” in Figure 1).
101
Patrícia Amaral
Proportion of T
responses per subject
Figure 1: Proportion of T responses per subject per
condition (means)
1
0,8
0,6
0,4
0,2
0
Acan
Acannot
Bcan
Bcannot
Conditions
The results reported in this study provide psycholinguistic evidence for the asymmetric status of
the entailments that have been identified in the interpretation of AAs. It was shown that the
proximal component plays a crucial role in textual coherence and that the pattern of continuations
licensed by AAs clearly contrasts with the pattern obtained for the polar component. These
results confirm the hypothesis that AAs display a behavior similar to the behavior of positive and
negative quantifiers like most and few, as defined above. On the other hand, subjects respond in
accordance to the polar component (the “backgrounded” entailment) when they are explicitly
asked about the truth conditions of the sentence that contains an AA. This difference was
accounted for in terms of the distinction between asserted and entailed (but not asserted) content,
following the proposal in Horn (2002).
References
Ancombre, Jean-Claude and Ducrot, Oswald. 1983. L’Argumentation dans la langue. Bruxelles:
Mardaga.
Champaud, Christian and Bassano, Dominique. 1987. Argumentative and Informative Functions
of French Intensity Modifiers “presque” (almost), “à peine” (just, barely) and “à peu près”
(about): an Experimental Study of Children and Adults. Cahiers de Psychologie
Cognitive/European Bulletin of Cognitive Psychology 7/6: 605-631.
Ducrot, Oswald. 1973. La preuve et le dire. Paris: Mame.
Hitzeman, Janet. 1992. The Selectional Properties and Entailments of “Almost”. Papers from the
28th Annual Meeting of the Chicago Linguistics Society. 225-238.
Horn, Laurence R. 2002. Assertoric inertia and NPI licensing. Proceedings of CLS 38-2: The
Panels. 55-82.
Jarvella, Robert J. and Lita Lundquist. 1994. Scales in the Interpretation of Words, Sentences,
and Texts. Journal of Semantics 10: 171-198.
Li, Charles. 1976. A functional explanation for an unexpected case of ambiguity (S or ~S). In:
Juilland, Alphonse (ed.), Linguistic Studies Offered to Joseph Greenberg. Saratoga, CA:
Anma Libri. 527-535.
102
Entailment, assertion, and textual coherence: a case study
Lundquist, Lita and Robert J. Jarvella. 1994. Ups and Downs in Scalar Inferences. Journal of
Semantics 11: 33-53.
Moxey, Linda M. and Anthony J. Sanford. 1993. Communicating Quantities. A Psychological
Perspective. Hillsdale: Lawrence Erlbaum.
Moxey, Linda M. and Anthony J. Sanford. 1998. Complement Set Reference and Quantifiers.
Proceedings of the 20th Annual Conference of the Cognitive Science Society: August 1-4,
1998, University of Wisconsin-Madison. London: Lawrence Erlbaum. 734-739.
Paterson, Kevin B.; Sanford, Anthony J.; Moxey, Linda M. & Dawydiak, Eugene. 1998.
Quantifier Polarity and Referential Focus during Reading. Journal of Memory and Language
39: 290-306.
Rapp, Irene and Arnim von Stechow. 1999. Fast ‘Almost’ and the Visibility Parameter for
Functional Adverbs. Journal of Semantics 16: 149-204.
Sadock, Jerrold M. 1981. Almost. In: Cole, Peter (ed.), Radical Pragmatics. New York:
Academic Press. 257-271.
Schwenter, Scott A. 2002. Discourse context and polysemy: Spanish casi. In: Wilshire, Caroline
R. and Camps, Joaquim (eds.), Romance Phonology and Variation. Selected Papers from the
30th Linguistic Symposium on Romance Languages, Gainesville, Florida, February 2000.
Amsterdam: John Benjamins. 161-175.
Sevi, Aldo. 1998. A semantics for almost and barely. Unpublished M.A. thesis, Tel-Aviv
University.
Stalnaker, Robert C. 1978. Assertion. Syntax and Semantics 9: Pragmatics. 315-332.
Ziegeler, Debra. 2000. What almost can reveal about counterfactual inferences. Journal of
Pragmatics 32: 1743-1776.
103
SPRIK Conference
Explicit and Implicit Information in Text - Information Structure across Languages
University of Oslo, June 8-10 2006
The role of discourse topic and proximity for demonstratives in German and
Russian
Maria Averintseva / Manfred Consten
Univ. Tübingen / Univ. Jena
[email protected] / [email protected]
1. Demonstratives in German and Russian
Demonstratives, like personal pronouns, are textual means of coherence within the manifold
field of ‘domain-bound reference’, i.e. referential relations that have to be established dependent on features of the linguistic or extra-linguistic context. In our talk, we discuss the textual
functions of Russian and German demonstratives in terms of discourse topicality and
proximity, thus covering a broad range of referential phenomena with a unified approach.
1.1 Lexical forms
Demonstratives are a subclass of lexical means of marking definiteness. In both languages,
demonstratives can be used either as pronouns or as determiners, while Russian lacks articles
in the proper sense like German der/die/das. We discuss the pronominal use and the use as
determiner analogously, when possible.
(1)
GERMAN
pers. pron. dem. pron.
(weak)
determiners
er/sie/es
der/die/das
he, she, it
this / the
demonstr. pron.
(strong)
dem. determiners
dieser/-e/-(es) jener/-e/-es
this
that
RUSSIAN
pers. pron
demonstr. pron.
(strong)
dem. determiners
on/ona/ono ėtot/ ėta /ėto tot/ta/to
he, she, it the / this
the / that
1.2 Present notions of demonstrativity: topicality and proximity
For German, Zifonun / Hoffmann / Strecker (1997: 555ff) assign following functions to
demonstratives: 1) demonstratives evoke a ‘new orientation of the addressee’ (in contrast to
personal pronouns which serve as means of thematic continuity) and 2) ‘point from right to
left’ in the textual space while personal pronouns operate in the text as a whole.
For Russian, demonstratives are traditionally analyzed as having the function of ‘pointing’ to
a referent (Švedova 1982) which corresponds primarily to the deictic use of the
demonstratives. Besides, additional functions like distinguishing between different ‘points of
view’ (e.g. the speaker vs. somebody the speaker is quoting) or signalling ‘empathy’ are
proposed (cf. the overview in Weiss (1988)).
Thus, in both languages, demonstratives are assumed to refer to a 1) newly introduced
referent 2) spatially nearest referent or 3) emphatically marked referent. Any dependency
between these functions is not accounted for.
However, some data (both for Russian and German) cannot be explained by these approaches,
e.g.:
- Demonstrative continuation of the reference to an already established thematical referent,
that is obviously motivated by emotional emphasis, which Zifonun / Hoffmann / Strecker
(1997: 558-560) inconsistently describe as ‘new orientation to the same referent’.
105
Maria Averintseva & Manfred Consten
(2) Wer zweifelt noch an der Schuld der Angeklagten? Diese Frau hat einen Mord begangen, damit
Who doubts
yet
at the guilt
of-the accused ?
This woman has a murder commited so-that
sie an das Vermögen ihres Gatten kommen konnte. Somit ist diese Frau eine habgierige Mörderin!
she the assets o f-her husband achieve was-able-to.
Thus is this woman a
greedy
murderer!
- very restricted use of demonstratives with indirect deixis and anaphora (see 2.1.2 and 2.2.2)
In order to explain these data, we propose an account that shows the interrelation of different
functions of demonstratives. Some preliminaries: we define discourse topicality (DT-ity) as a
property of a certain discourse referent a given discourse segment is ‘about’ (in the sense of
Reinhart (1981)). Discourse segment is understood intuitively as a thematically contiguous
part of a discourse. We assume that in a given segment there is only one topical referent in the
sense of DT-ity, which we call ‘discourse topic’ (DT).
As for ‘proximity’, it is first understood in the literal sense of the term as short spatial distance
between the speaker and the intended referent, sometimes contrasting with another, more
distant referent (especially in the case of deictic reference). Proximity can be applied to
anaphora if text / discourse is considered analogous to physical space as suggested by
Bühler’s (1934), ‘Zweifelder-theory’. Then, spatial proximity means a low distance between
an anaphor and its antecedent. In section 3. we show how the concept of proximity is
extended to cover the cases of emotional emphasis.
2. Phenomena of demonstrative reference
Like other deictically or anaphorically used expressions, demonstratives are a means of
domain bound reference, their different discourse functions depend on the domain of
reference, deixis operating in a non-textual domain, anaphora in the textual one.
2.1 Deixis
2.1.1 Direct deixis
With direct deixis, demonstratives are used to refer to distant versus near referents (here,
proximity is defined in terms of physical space), as the opposition dieses –jenes1 in German
and ėtot – tot in Russian shows, cf. (3) and (4):
(3) Dieses Café (wo wir sind) gefällt mir besser als jenes dort drüben auf der anderen Straßenseite.
This
café
(where we are) like
I
better than that there over the road
Here, the use of demonstratives is deictically motivated with respect to the place of utterance
(denoted by dies-+ N) in contrast with another place (denoted by jen-).
(4) [...] Dorodnyx vyšel
D
On vpolgolosa
he
iz-za
kustov na bereg, k samoj reke.
came-out from-behind bushes to shore
to itself
river
otdaval prikazanija [...] gotovit'sja k brosku na tot bereg.
in-a-low-voice gave
orders
prepare
to rush
to that shore
[Tübinger Russische Korpora]
Here, the shore the protagonist is standing upon is referred to with a bare noun, while the
other one, i.e. the more distant shore, is referred to with tot+N.
2.1.2 Indirect deixis
By indirect deixis we understand deictic reference where the referent cannot be seen directly
but can be derived from visual features of the current discourse space.
1
However, jenes has become quite peripheral in German usage, so that the meaning of physical distance is more
often lexicalized in some other way, e.g. with der andere (‘the other one’).
106
The role of discourse topic and proximity for demonstratives in German & Russian
While dies- in German as well as ėtot/tot in Russian are the most typical means of direct
deixis, they are very constrained with indirect deixis.
(5) a. Ist dér
is
/ *dieser nicht da? (showing towards an empty office)
héDEM / *this
b. Netu
not
there
ego / ?ėtogo / *togo?
is-not-there he / ?this
/ *that?
2.2 Anaphora
2.2.1 Direct nominal and complex anaphors
Like with deixis, the concept of proximity can be applied in order to explain anaphoric
demonstrativity. However, DT-ity is crucial for anaphors as well. As for nominal anaphors,
non-DT, but ‘near’ antecedent in textual space prefers demonstratives, while DT with any
antecedent position prefers personal pronouns (as stated in 1.2).
(6) Odnaždy papa privel [...]
once
kakogo-to čeloveka, [...] ėtot čelovek vse vremja sprašival
father1 brought-home some
man2
[...] this man2
all time
asked
(7) Hast du schon das Neueste von SPD-Chef Müntefering gehört?
have you already the latest
Er wollte
about SPD-leader Müntefering1 heard?
seinen Vertrauten zu seinem Stellvertreter machen.
He1 intended his
confident2
to his
Dieser Mann / Dieser bekam aber
This
man2
/ This2
deputy
make.
keine Mehrheit.
obtained however no
majority.
In (6), the whole segment is about the speaker’s father1; the demonstrative NP refers to the
non-topical referent2. Similarly, in (7), the demonstrative full NP as well as the demonstrative
pronoun is assigned to the non-topical referent2. In both (6) and (7) the personal pronoun er /
on (“he”) would be read as coreferent with the discourse topic NP1.
Complex anaphors are NPs referring to propositionally structured referents by condensing a
larger text segment which serves as their antecedent (cf. Consten / Knees 2006). They are a
special and clear case of non-DT-ity, since the referent is created not until the act of anaphoric
reference. Thus, our claim that demonstratives function as means of non-topical reference
explains why demonstratives are preferred for complex anaphora while personal pronouns are
ruled out:
(8) Meine Freundin wird bald vierzig.
Dies / Das / *Es
deprimiert sie sehr.
Mojej podruge skoro ispolnitsja sorok let. Ėto
/ *Ono eje očen' ugnetajet.
[my
friendfem
will-be soon forty]event
Thisevent
/ *Itevent her very depresses.
2.2.2 Direct versus indirect anaphora
A distinction of direct and indirect reference is made with respect to anaphora like with deixis
(see 2.1.1 versus 2.1.2). Thus, indirect anaphors are anaphors without an explicit antecedent
(cf. Consten 2004). They are preferably realised by lexical NPs (s.(9)) or (under certain
conditions) by personal pronouns.2
(9) Am
U
Straßenrand stand ein Auto. Der
obočiny
stojala
mašina.
At-the roadside
stood (a) car.
Motor / *Dieser Motor war noch warm,
Motor / *Ėtot
motor byl ešče tëplym
(The) engine / *this
engine was
still
warm,
aber vom Fahrer / *von diesem Fahrer fehlte jede Spur.
2
Pronominal indirect anaphors are possible in case of a close relationship to their anchor, i.e., when the anaphor
denotes an argument of an anchoring verb, e.g.: Do not park at the teacher’s park lot – next time I will have it
(*this) towed away (“car”, Consten 2004). Only for these, resumption through pronominal forms is possible at
all.
107
Maria Averintseva & Manfred Consten
no
šofëra
but (the) driver
/ *ėtogo
šofëra
bylo ne vidat’.
/ *this
driver
was not to-see
Like with deixis, dies- and ėtot / tot are very restricted for indirect anaphora. This fact cannot
be explained in terms of DT-ity, since indirect reference introduces new (rhematic) referents
and, therefore, demonstratives should fit. We offer an explanation in 3.3.3.
With indirect complex anaphors (which are quite rare), the use of demonstrative lexical NPs
is possible only when the situations talked about are at the same temporal or epistemic level,
unlike (10), where the events referred to in the anchoring text are present and real within the
text world, whereas the event denoted by the anaphor is hypothetical:
(10) (Anchoring text, the speaker is a little boy:). I was seen dragging a big, old umbrella and I was caught just
in time when I tried to hide in an airplane. [Stanisław Lem, Der Planet des Todes, 83, German translation]
Aus dem / *diesem improvisierten
Fallschirmabsprung wurde nichts.
Iz ∅
/ *ėtogo improvizirovannogo pryžka s parašytom ničego ne vyšlo.
From the
/*this
improvised
parachute-jump
became nothing
We will show that this change of level is the reason for the inacceptability of demonstratives.
2.3 To sum up
In (11), different means of domain bound reference are assigned to their most typical textual
functions. Except for the personal pronouns er / on, this overview is intended to be valid for
determiners of lexical NPs as well as for pronouns.
(11)
GERMAN
RUSSIAN
direct deixis: distant ref.
jener
tot
direct NA, non-DT-referent
dir. / indir. CA
dieser
ėtot
direct deixis: near ref.
indirect deixis
der
indirect NA with central referent
on
direct NA, DT-referent
er
3. Discussion – Towards a unified model of demonstrativity
DT-ity and discourse segmenting and the choice of the referential means are interdependent:
Not only the choice of referential means is fixed through the discourse structure, but also
discourse structure is defined through the way a referent is referred to. So, e.g. a possible
beginning of a new discourse segment might cause ambiguities in the interpretation of a
demonstrative reference (cf. ():
(12) Ja rasskažu tebe pro Vanju. On sovsem uže
I tell
xodit v novyj klub. Petja včera
goes to new
club
a. On
ego
tože tam byl.
Petja2 yesterday also there was
srazu uvidel i
podošël pozdorovat'sja.
perspronNOM perspronAKK gleich sah
und kam
demonstrNOM perspronAKK gleich sah
und kam
b. Tot
spjatil. On každyj večer
you about Vanja1 he1 absolutely already got-mad he1 every
ego
srazu uvidel i
begrüßen
podošël pozdorovat'sja.
108
begrüßen
evening
The role of discourse topic and proximity for demonstratives in German & Russian
In (12) a., the personal pronoun on has two readings: 1) referring to Vanja who is regarded as
the DT within the ongoing discourse segment. 2) referring to Petja if the recipient thinks that
the sentence introducing Petja opens a new (sub-)segment with Petja as a DT. However, the
function of continuous reference to the most salient referent, which is typical for personal
pronouns, remains the same. The different readings are motivated by different kinds of
segmenting the discourse.
Analogously, in (12) b. tot will be related to Petja as a non-DT referent if the whole text is
seen as one discourse segment while it will be related to Vanja if Petja is considered a new
DT. Again, the function of indicating a non-DT holds with both readings.
As noticed in 1.2., as yet no relation between DT-ity and proximity as factors determining the
form of the reference resumption has been stated. However, the proximity-factor interacts
with DT-ity. Basically, demonstrativity indicates proximity. But, firstly, DT-ity overrides
proximity in a physical sense of the term as we are dealing with in deictic and textual
‘pointing’. For an interaction between deictic pointing and DT-ity think of a case where a
referent is physically present and, at the same time, becomes DT (e.g. talking about a cat
which is in the same room). Here, it is most plausible to introduce this referent deictically by
a demonstrative (combined with a gesture of pointing) and to continue with a chain of
personal pronouns.
(13) Look at this cat! She has been sleeping for twelve hours and now she is crying for food although
she is too fat already.
However, an anaphoric chain with repeated demonstrative NPs is possible as well if the
speaker wishes to give an emotional emphasis to his statement about the discourse referent. In
most of the cases a negative evaluation of the referent is given. Here, the use of demonstrative
NPs does not result from spatial proximity (which is overriden by DT-ity) but from the
speaker's emotional involvement with his topic, which is a common kind of ‘cognitive
proximity’. Thus, the term of proximity has to be used in a more broader sense as ‘cognitive
proximity’ with respect to emotional attitudes and epistemic levels (cf. indirect complex
anaphora). In contrast to the dominance of the DT-factor over physical proximity, cognitive
proximity can allow for demonstrativity regardless of DT-status.
As a result, we gain a hierarchy of features matching demonstrativity:
(14) physical proximity (deixis: in space, anaphora: in text) < Non-DT-ity < cognitive proximity
References
Bühler, Karl. 1934 / 19823. Sprachtheorie. Die Darstellungsfunktion der Sprache. Stuttgart: Gustav Fischer.
Consten, Manfred. 2004. Anaphorisch oder deiktisch? Zu einem integrativen Modell domänengebundener
Referenz. Tübingen: Niemeyer.
Consten, Manfred/Knees, Mareile. to appear. Complex Anaphors in Discourse. In: Benz, A./Kühnlein, P. (eds.)
Constraints in Discourse. Amsterdam: Benjamins.
Reinhart, Tanya. 1981. Pragmatics and Linguistics: An Analysis of Sentence Topics. Philosophica 27: 53-94.
Švedova, Natalija et. al. 1982. (eds). Russkaja Grammatika. Tom I. Moskva: Nauka.
Weiss, Daniel. 1988. Zum substantivisch-anaphorischen Gebrauch von russ. ėtot. Zeitschrift für slavische
Philologie 1988: 249-269.
Zifonun, Gisela/Ludger Hoffmann/Bruno Strecker .1997. Grammatik der deutschen Sprache. Bd. 1. Berlin/New
York: WDEG (Schriften des Instituts für deutsche Sprache, Bd. 7.1)
109
SPRIK Conference
Explicit and Implicit Information in Text - Information Structure across Languages
University of Oslo, June 8-10 2006
Some remarks on minimal sufficient conditions across languages
Richard Zuber
CNRS, Paris
[email protected]
Making hypothesis and thus expressing sufficient conditions is essential in coherent text
interpretation. Languages have a variety of means of expressing sufficient conditions for a goal to
be achieved or a state to obtain. The purpose of this talk is a preliminary and partially crosslinguistic study of specific constructions expressing the so-called minimal sufficient conditions
(MSCs). Minimality of MSCs is usually implicitly given. Among constructions expressing MSC
one finds conditional sentences with a specific conditional connector. Such conditionals, called
minimal sufficient condition conditionals, MSCC in short, will be analyzed in some details in
Polish. Furthermore, given great cross-linguistic variations of MSCC and their semantic
equivalents I will provide data from French, Spanish and Japanese and make some comparisons
with other constructions. I will outline an (almost) algebraic semantics for them, based on the
algebraic semantics for even and especially.
In a MSCC the conditional connective is the complex expression of the form IF ONLY and thus
MSCCs which are studied here have the form P IF ONLY Q, where P and Q are sentences (not
quite freely chosen). Essentially I use the indicative mood in MSCCs even if this is not an
empirical constraint. It is important to notice that MSCCs of the above form cannot contain the
the “adverbial” corresponding to THEN usually occurring in conditionals.
An English example of MSCC would be He would be happy if only he had a bottle of wine. In
fact such constructions seem rather restricted in English since apparently IF ONLY clauses are
preferably used in English in counterfactuals and in "incomplete conditionals" to express wishes
(as in If only she were intelligent). In addition some MSCCs expressed in other languages are not
simply translatable into English. For this reason I study basically MSCCs in Polish where the
(corresponding) IF ONLY clauses appear much more productive. In Polish there exist also
temporal MSCCs containing WHEN ONLY clauses.
Take the example in (1):
(1)
Kazio jest zadowolony jesli tylko ma butelke wodki.
K.
is satisfied IF ONLY has bottle vodka (GEN).
Sentence (1) carries double information:
1. Simple Sufficiency Condition: For K. to be satisfied it is enough to have a bottle of vodka
(assertion), and
2. (Relative) Minimality: To have a bottle of wine (for K.) is a minimal sufficient condition
to be satisfied.
The information in 2 corresponds probably to a presupposition.
Sentence (1) is an indicative conditional and thus in (1) we have an indicative MSCC. Quite
similar decomposition is possible, mutatis mutandis, for counterfactual MSCCs. Furthermore, in
111
Richard Zuber
Polish there are also imperative and interrogative MSCCs. Thus, generalizing a bit we can say
that a MSCC of the form P IF ONLY Q, whether indicative or counterfactual, corresponds to the
conjunction P IF Q and Q is minimal in some sense (for instance easy to obtain). The minimality
condition is clear in the case of temporal MSCCs, that is when, roughly, IF ONLY is replaced by
WHEN ONLY. In that case the corresponding connector in Polish (which is kiedy tylko) means
roughly right when and thus the minimality is determined by temporal ordering. Imperative
MSCCs express orders which have to be satisfied when same conditions, considered as minimal,
are satisfied. In particular imperative temporal MSSCs express orders which have to be satisfied
at a specific time.
An important point to keep in mind is that Q is clearly not the necessary condition for P to hold.
This means that ONLY IF and IF ONLY are different. Furthermore Q is clearly not the necessary
condition for P to hold.
In Polish and in other languages there are semantically (quasi) equivalent constructions
expressing MSCs, which are not conditional in form. I will mention the so-called sufficiency
modal constructions studied by Sæbæ (2001) von Fintel and Iatridou (2005) and others. The
consider conditional meaning of sentences like To find a good wine in France you only have to
go to a nearby wine shop). Von Fintel and Iatridou show, cross-linguistically, that such
constructions always contain a specific modal and an exceptive clause.(for instance “ne que
VERB” in French). There does not seem to be an obvious solution to various problems to which
such specific conditionals give rise (cf. Werner 2006). A related scalar use of only in conditional
sentences is studied in Krasikova and Zhechev (2005). The exact relation between the
constructions they study and the MSCCs studied here is not clear except that MSCCs I study here
do not to contain explicitly any expression of modality.
One observes that, cross-linguistically, among non-conditional constructions expressing MSCs
one finds Spanish constructions with con que/con tal que} and their French equivalent pourvu
que (as in Il est content pourvu qu'il ait une bouteille de vin).
An interesting observation is that in Japanese there are conditional constructions expressing
MSCs and that in this case the MSCCs are expressed, roughly, with the help of IF EVEN clauses
(-BA SAE). I take this observation as a starting point for my preliminary proposal.
Indeed, EVEN is a typical item involving an extreme point on a scale provoking a surprise effect,
similarly to IF ONLY. The data suggest that the MSCCs should include a surprise effect in their
interpretations (implicitly given): P IF ONLY Q can be paraphrased by the conjunction (P IF Q)
AND (IT WOULD BE SURPRISING THAT Q AND NOT- P). Thus we have a surprise effect
which, however, in this case is "inverse" to the one induced by even and is the same as the one
induced by especially/in particular, as in (2): the Polish correspondent of (2) entails (1),
including the minimality condition:
(2)
Kazio is often happy, in particular when he has a bottle of vodka.
Example in (2) shows, that other particles can induce minimally sufficient conditions in
conditional constructions.
Since only is categorically polyvalent and can apply to expressions of various categories it is
important to see what of the argument of IF ONLY in MSCCs. Now, since the following
sentence has a non-contradictory reading IF ONLY applies to sentences (if IF ONLY (or ONLY)
applied to "Kazio" the sentence would have only a contradictory reading):
112
Some remarks on minimal sufficient conditions across languages
(3)
Kasia bedzie tanczyc jesli tylko Kazio bedzie tanczyl
(Kasia will dance IF ONLY Kazio will dance)
It is not obvious what pragmatic "scale" should be used to order propositions except the ordinary
semantic entailment (in the pragmatically restricted set of possible worlds). So I propose to
consider as "minimal" atomic propositions, that is atoms in the corresponding algebra of
propositions. Such a move has proved to be useful for the analysis of various focus particles: only
denotes an atomic absolute modifier, and even - a restrictive (non-absolute) atomic modifier (in
the denotational algebra whose type is determined by the type of the argument to which these
items apply). This explains why such items are categorially polyvalent (an atom is a type
independent notion). Under this analysis Even Leo danced means that Leo is the only dancer who
has a property pragmatically incompatible with dancing, hence the surprise effect (cf. Zuber
2004). To apply this idea to MSCCs we observe that IF ONLY clauses (and other clauses
inducing MSCs) behave like atomic expressions: they cannot be conjoined (notice the
impossibility of the following conjunction: *even Leo and even Lea). Thus the conjunction of
two IF ONLY clauses is impossible in Polish (and two conjoined pourvu que clauses are
impossible in French). Facts like these suggest that IF ONLY denotes an atomic function which
applies to propositions and gives a propositional atom as result. Such a propositional atom should
be considered as a minimal sufficient condition.
References
Von Fintel and Iatridou, S. 2005. Anatomy of a Modal, MIT, unpublished
Krasikova, S. and Zhechev, V. 2005. Scalar uses of only in conditionals, in Proc. Of the Fifteenth
Amsterdam Coloquium: 137-143.
Werner, T. 2006 An analysis of existential anakastics, a paper presented at SALT 16.
Sæbø, K. J. 2001. Necessary Conditions in a Natural Language, in Fery, C. and Sternfeld , W.
(eds.) AudiaturVox Sapientiae: A Festschrift for Armin von Stechow, Akademie Verlag: 427449.
Zuber, R. 2004. Boolean Semantics and Categorial Polyvalency, in B. Schmeiser et al. (eds.)
WCCFL 23 Proceedings, Somerville, MA: Cascadilla Press: 829-42.
113