Professional Documents
Culture Documents
Editors
Marcelo Dascal, Tel Aviv University
Raymond W. Gibbs, University of California at Santa Cruz
Jan Nuyts, University ofAntwerp
Editorial address
Jan Nuyts, University of Antwerp, Dept. of Linguistics (GER),
Universiteitsplein 1, B 2610 Wilrijk, Belgium.
E-mail: nuyts@uia.ua.ac.be
Volume8
Text Representation: Linguistic and psycholinguistic aspects
Edited by Ted Sanders, Joost schilperoord and Wilbert Spooren
Text Representation
Linguistic and
psycholinguistic aspects
Edited by
Ted Sanders
University of Utrecht
Joost Schilperoord
University ofTilburg
Wilbert Spooren
Free University, Amsterdam
Text Representation : Linguistic and psycholinguistic aspects I edited by Ted Sanders, Joost
Schilperoord, Wilbert Spooren.
p. em. (Human Cognitive Processing, ISSN 1387-6724; v. 8)
Based on papers presented at a conference held July 1997, Utrecht University.
Includes bibliographical references and indeL
l. Discourse analysis--Psychological aspects--Congresses. I. Sanders, Ted. II.
Schilperoord, Joost. III. Spooren, Wibert. IV. Series.
P302.8.T49 2001
401 ~41--dc21 2001035506
ISBN 90 272 2360 2 (Eur.) /1 58811 077 X {US) {Hb; alk. paper)
Preface vu
The chapters of this book volume are all based on papers presented at the
International workshop on text representation: Linguistic and psycholinguistic
aspects, held at Utrecht University in July 1997. We are indebted to the Centre
for Language and Communication (CLC) and the Faculty of Arts of Utrecht
University, the Center for Language Studies (CLS) and the Discourse Studies
Group of Tilburg University, the Netherlands Organization for Scientific Re-
search NWO and the Royal Academy of Sciences KNAW for their financial
support of this workshop and to the Utrecht Institute of Linguistics UIL OTS
(which now includes CLC) for logistic and financial support in the preparation
of the manuscript. We are very grateful to Gerben Mulder, Annelous van Rijn
and Martien Kroon for their invaluable help in editing this manuscript.
We would also like to thank the many colleagues who acted as reviewers of
the chapters:
Finally, we want to thank the editors of the Benjamins series Human Cognitive
Processing, especially Jan N uyts, for publishing this book volume in their series.
Ted Sanders
Wilbert Spooren
University of Utrecht/Free University, Amsterdam
~ 0
Q,o
0
•
Figure 1. Communication through written text: From cognitive representation to
text, to cognitive representation (as seen by Laura van Beek, 9 years old)
linguists pay more attention to the cognitive aspects of language use (compare
the emergence of cognitive linguistics), and psycholinguists give more serious
consideration to the linguistic complexity of the object under study. This
volume is intended to contribute to the information exchange among re-
searchers from these disciplines.
Before giving an overview of the content of the different contributions to
this volume, we first highlight themes we believe to be central to the cognitive
and linguistic study of text representation. We start with three general tenden-
cies in research on text representation. Then we introduce two linguistic
characteristics that constitute a text, which will be the object of study in this
volume.
So far, we have been using the term 'the text representation'. This term is in fact
an oversimplification. Both in linguistics and in psychology text representa-
tions are taken to be composite. In psycholinguistics we find this idea very
explicitly present in the work ofKintsch and associates (Kintsch 1998, but also
in Kintsch & van Dijk 1978 and in Van Dijk & Kintsch 1983; for an overview
see also Singer 1990). Their research focuses on the receptive side of communi-
cation and states that readers make multiple representations of the sentences of
a text: a surface code (a short-lived representation of the exact linguistic
material in the sentences), a text base (containing the propositions expressed
by the sentences and their interrelations), and a situation model (in which the
linguistic material is integrated with the background knowledge of the reader).
In linguistics, the concept of multiple representations has also been devel-
oped. In formal semantics in the work ofKamp (see Kamp 1981; Heim 1982;
Kamp & Reyle 1993) in cognitive linguistics most explicitly in the work by
Fauconnier ( 1985, 1994) on mental spaces. In general, the idea is that linguistic
expressions are instructions to update the current mental representation, that
is based on previous discourse, background knowledge and inferencing. Thus,
expressions are considered to have a procedural meaning. Or, as Fauconnier
( 1994, p. xviii) has put it:
Language does not itself do the cognitive building - it "just" gives us minimal,
but sufficient, clues for finding the domains and principles appropriate for build-
4 Ted Sanders and Wilbert Spooren
ing in a given situation. Once these clues are combined with already existing
configurations, available cognitive principles, and background framing, the ap-
propriate construction can take place, and the result far exceeds any overt explicit
information.
The quote from Fauconnier brings us to the second important theme of recent
research on text representations, the underspecification of mental representa-
tions. Contrary to what is maintained in the standard coding theory of
meaning, it is fairly generally accepted in many branches of linguistics and
psychology that what an utterance means cannot in any easy, transparent and
compositional way be connected to the meaning of the individual elements in
the utterance and their interrelation. An utterance explicitly codes only part of
the meaning of the utterance into explicit linguistic material, the rest having to
be provided by inferencing.
In linguistics this point has been formulated forcefully by Sperber and
Wilson ( 1992, Chapter 1) and Fauconnier ( 1994, Introduction), but it is in fact
prominent in most descriptive accounts of coherence (see Bublitz, Lenk &
Ventola 1999, for a recent overview). It is even one of the major tenets, in a
much more radical 1 form, in, e.g., conversation analysis (cf. Pomerantz & Fehr
1997). The same linguistic items will be interpreted differently in different
situations and contexts, and hence any adequate theory of meaning will have to
allow for rich inferential mechanisms. The Gricean program (e.g., Grice 1975)
is one such attempt to provide the necessary inferential power: Under the
assumption of cooperativeness, participants in a conversation will generate
implicatures in order to extend the literal meaning of an utterance and be able
to make a coherent representation of the discourse. This inferential mecha-
nism has been explored systematically by Levinson (1991 ), for instance to
account for the binding properties of anaphors. In the field of text linguistics,
Spooren ( 1997) uses a similar mechanism to account for so-called underspeci-
fied uses of connectives (cases where coherence relations are not (completely)
matched by the meaning of the connectives that occur in the text).
In psychology there are numerous findings demonstrating the under-
specification of linguistic material. For instance, Graesser, Singer, and
Trabasso ( 1994) have suggested that in narrative texts causal connections
function as a default, thus allowing for coherence relations to remain un-
marked. Noordman and Vonk (1998) and Sanders and Noordman (2000),
Text representation as an interface between language and its users S
suggest that such a view is promising for expository texts as well. In the latter
study it is found that causal relations lead to faster processing of the connected
information, but still lead to a more integrated representation. Taking the role
of non-linguistic factors, such as reader's characteristics, into account,
Noordman and Vonk (Noordman, Vonk & Kempff 1992; Noordman & Vonk
1992) have shown that depth of processing of causal relations is dependent on
the reader's goals and knowledge.
In this volume the role of both types of coherence in text processing is dis-
cussed. Now that we have identified the types of representation and the consti-
tuting principle of text, we can state a central theme of this volume in different
terms: A major issue is the relationship between the linguistic surface code
(what Giv6n 1995, calls 'grammar as a processing instructor') and the meaning
representations. Both coherence phenomena under investigation - referen-
tial and relational coherence - have dear linguistic indicators that can be
taken as processing instructions, which will typically affect the surface repre-
sentation. For referential coherence these are anaphoric devices such as pro-
nouns, and for relational coherence these are connectives and (other) lexical
markers of relations.
The relevant linguistic indicators for referential coherence are pronouns and
other devices for anaphoric reference. Ever since the seminal work of linguists
like Chafe (1976) and Prince (1981), both functional and cognitive linguists
have argued that the grammar of referential coherence can be shown to play an
important role in the mental operations of connecting incoming information
to the existing mental representations. For instance, referent NPs are identified
as either those that will be important and topical, or as those that will be
unimportant and non-topical. Hence, topical referents are persistent in the
mental representation of subsequent discourse, whereas the non-topical ones
are non-persistent. Recently, more and more empirical data from corpus
studies become available which underpin this cognitive interpretation of refer-
ential phenomena, following a route guided by functional linguists such as
DuBois (1980).
In a distributional study, Giv6n ( 1995), for instance, shows that in English,
the indefinite article a( n) is typically used to introduce non-topical referents,
whereas topical referents are introduced by this. In addition, there is a clear
interaction between grammatical subjecthood and the indefinite article this:
most this-marked NPs also appear as grammatical subjects in a sentence, while
a majority of a(n)-marked NPs occurred as non-subjects. Across languages
there appears to be a topic persistence of referents; in active-transitive clauses the
topic persistence of subject NPs is systematically larger than that of object NPs.
In several publications Ariel (1988, 1990) has argued that regularities in
grammatical coding should indeed be understood to guide processing. She has
studied the distribution of anaphoric devices and she has suggested that zero
Text representation as an interface between language and its users 9
(Britton, Glynn, Meyer & Penland 1982; Meyer 1975). At the same time, on-
line data suggest that the presence of linguistic markers facilitates processing
(Britton et al. 1982; Deaton & Gemsbacher, in press).
Thus far, we have discussed the role of connectives and signaling phrases in
discourse processing. A preliminary conclusion might be that they can be
treated as linguistic markers which instruct readers in how to connect the new
discourse segment with the previous one (Britton 1994). In the absence of such
instructions readers have to determine for themselves what coherence relation
connects the incoming segment to the previous discourse. Such an inference
process requires additional cognitive energy and results in longer processing
times. If this idea has any validity, it implies that the coherence relations
themselves would have a major influence on discourse processing as well. One
might expect that the type of relation that connects two discourse segments, be
it causal, additive, contrastive etc., affects the discourse representation.
Here we move into another area where the combination of text linguistic
and discourse psychological insights has lead to significant progress: the cat-
egorization of coherence relations. In the last decade, a significant part of
research on coherence relations has focused on the question how the many
different sets of relations should be organized (Hovy 1990; Knott & Dale 1994;
Pander Maat 1998; Redeker 1990; Sanders 1997a). Sanders et al. ( 1992, 1993)
have started from the properties common to all relations, in order to define the
'relations among the relations', relying on the intuition that some coherence
relations are more alike than others. For instance, the relations in ( 4), (5) and
(6), all express (a certain type of) causality, whereas the ones in (7) and (8) do
not. Furthermore, a negative relation is expressed in (7), as opposed to all other
examples, and (8) expresses an enumeration or addition.
(4) The buzzard was looking for prey. The bird was soaring in the air for
hours.
(5) The bird has been soaring in the air for hours now. It must be a buzzard.
(6) The buzzard has been soaring in the air for hours now. Let's finally go
home!
(7) The buzzard was soaring in the air for hours. Yesterday we did not see it
all day.
(8) The buzzard was soaring in the air for hours. There was a peregrine
falcon in the area, too.
presentational, internal and pragmatic relations, on the other hand. In the first
type of relations, segments are related because of their propositional content,
i.e. the locutionary meaning of the segments. They describe events that cohere
in the world. The relation in (9) can be interpreted as semantic because it
connects two events in the world; our knowledge allows us to relate the
segments as coherent in the world. A relation like (9) could be paraphrased as
"the cause in the first segment (Sl) leads to the fact reported in the second
segment (52)" (Sanders 1997a).
(9) The neighbours suddenly left for Paris last friday. So they are not at
home.
(10) The lights in their living room are out. So the neighbours are not at
home.
In (10) however, the two discourse segments are related because we under-
stand the second part as a conclusion from evidence in the first, and not
because there is a causal relation between two states of affairs in the world: It is
not because the lights are out that the neighbours are not at home. The causal
relation (10) could be paraphrased as "the description in 51 gives rise to the
conclusion or claim formulated in the 52." Hence, in the second type of
relation the discourse segments are related because of the illocutionary mean-
ing of one or both of the segments. The coherence relation concerns the speech
act status of the segments.
If this distinction is applied to the set of examples above, the causal relation
(4) is semantic, whereas (5) and (6) are pragmatic. This systematic difference
between types of relations is noted by many students of discourse coherence.
Still, there is quite a lot of discussion about the exact definition of a distinction
like this (see e.g., Bateman & Rondhuis 1997; Degand 1996; Hovy 1990; Knott
& Dale 1994; Knott 1996; Knott & Sanders 1998; Martin 1992; Moore &
Pollack 1992; Oversteegen 1997; Pander Maat 1998; Sanders 1997a; Sanders &
Spooren 1999 and several contributions to Couper-Kuhlen and Kortmann
(2000)). At the same time, several researchers have come up with highly similar
distinctions, and there seems to be basic agreement on the characteristics of the
prototypical relations (Sanders 1997a). Moreover, very similar distinctions
have been shown useful in describing the differential meaning of conjunctions
(Sweetser 1990). Also, the saliency of categorizations like these has been shown
in experiments in which, among other tasks, language users were asked to
judge the similarity of relations. Still, the discussion on this issue is clearly
continued in this volume, especially in Section 2 (Chapters 6 to 9).
Text representation as an interface between language and its users 15
1981; Trabasso & Van den Broek 1985; Trabasso & Sperry 1985). Keenan,
Baillet and Brown ( 1984) and Myers, Shinjo and Duffy ( 1987) demonstrated
that the effect of causal connectedness on memory for sentences is greatest for
moderate levels of causality. Also, causally related sentences are read faster
(Haberlandt & Bingham 1978), and the reading time decreases when the
causality increases (Keenan et al. 1984, Myers et al. 1987).
Fewer studies exist for expository text. One example is a study by Meyer
and Freedle (1984). They claimed that differences exist in the amount of
organization of different types of text structure. The better organizing types are
CoMPARISON, CAUSATION and PROBLEM-SOLUTION, whereas COLLECTION is a
weaker organizing type. These structure types are rather similar to the often
distinguished types of coherence relations. In a free recall experiment, Meyer
and Freedle expected readers to reproduce more information from better
organized types than from less organized ones. The results show that recall of
the CAUSATION and COMPARISON passages was indeed superior to the recall of
the coLLECTION passage. 3 However, there are some problems with the Meyer
and Freedle study, see Horowitz (1987) and Sanders & Noordman (2000) for
further details.
Sanders and Noordman (2000) embedded a similar text segment in two
different contexts. In one case it was a Solution to a Problem, in the second case
the same segment was part of an addition. It was found that PROBLEM-SOLU-
TION relations lead to faster processing, better verification and superior recall.
The authors conclude that the processing of a text segment depends on the
relation it has with preceding segments. Perhaps the most interesting finding in
this experiment is the contrast in the effect of the two independent variables:
linguistic markers (implicit or explicit) and type relations (PROBLEM-SOLU-
TION or LIST). Explicit marking of the relations resulted in faster processing,
but did not affect recall. However, verification data concerning the representa-
tion immediately after reading, show an effect of the linguistic marker. This
finding is quite similar to the effect Millis and Just (1994) found for the
influence of because.
Hence, it can be concluded that the relational marker has an effect during
on-line processing, but that its influence decreases over time. This contrasts
with the effect of the coherence relation, which is also manifest in the recall.
This contrast is similar to another frequently observed finding in language
comprehension: initially a reader or listener constructs the surface representa-
tion of a sentence, but after a short time interval only the meaning or gist of the
message is retained. Sachs ( 1967) found this effect for the form and meaning of
Text representation as an interface between language and its users 17
these restrictions are and how they interplay with the meaning expressed by the
connected segments {cf. Knott & Dale 1994; Pander Maat & Sanders 2000 and
several contributions to Risselada & Spooren 1998). It is this type of insights
that underlines the importance of further cooperation of text linguists and
psycholinguists working on discourse (Sanders 1997b).
What can readers expect of this volume? The collected chapters typically
present a cross-disciplinary account of text representation, by both linguists
and psycholinguists. This implies that linguistic analyses of textual characteris-
tics ultimatdy aim at accounting for the cognitive interpretation they can
receive. At the same time, psycholinguistic studies focus on the relevance of text
characteristics for theories of text processing, where text processing concerns
both production and interpretation. An important benefit of this combination
of text linguistics and psycholinguistics, and of production and understanding
is that we will encounter various methodologies, which are complementary:
linguistic analysis, text analysis, corpus linguistics, computational linguistics,
argumentation analysis, and the experimental psycholinguistic study of text
processing. A final focus of this book is the comparison and further testing of
linguistic and processing theories of text representation.
The following 12 chapters are divided in four sections. Section 1 deals with
referential coherence in text and text representation, and especially with acces-
sibility: how can the notion of varying accessibility explain for different refer-
ential forms, and what is the evidence for such a dynamic account of the
cognitive representations language users have?
In Section 2 focus shifts from referential to relational coherence in text and
text representation, when the classification of coherence rdations and connec-
tives is discussed in a closely connected cluster of chapters, combining various
theoretical approaches (from Relevance Theory to Argumentation Theory and
cognitive accounts of coherence relations) and different empirical methods
(from text-analysis to reading experiments).
Section 3 focuses on the cognitive representations of discourse and its
relation to knowledge representations: how are they related? How large is the
role of linguistic factors?
Finally, Section 4 discusses an issue typically neglected in the previous
sections: when coherence is said to exist, it exists between something, for
Text representation as an interface between language and its users 19
instance, discourse units or text segments. But how are these segments defined?
And when we distinguish between different linguistic levels of representation
(word, clause, sentence, paragraph), do we know that these levels have any
psychological validity?
Together, the chapters in these four sections present an overview of a
growing field of interest, at the intersection of linguistics and psychology, the
study of a phenomenon that is crucial to our behavior because it is the mostly
used vehicle of communication: that of text and its cognitive representation.
Notes
1.More radical in that radical conversation analysists deny the possibility of attributing any
meaning to an utterance without regard of its context.
1..Stricdy speaking there is no logical connection between dynamic systems and incremen-
tal systems, in that dynamic systems can be constructed that are not incremental and
incremental systems that are not dynamic. Yet in every serious language interpretation
system that we know of the two go together.
3· However, there are some problems with the Meyer and Freedle study, see Horowitz
( 1987) and Sanders and Noordman (2000) for further details.
References
Andriessen, J., De Smedt, K., 8c Zock, M. ( 1996). Discourse planning: Empirical re-
search and computer models. InT. Dijkstra, 8c K. De Smedt (Eds.), Computational
linguistics. AI and connectionist models ofhuman language processing(pp. 247-278).
London etc.: Taylor 8c Francis.
Anscombre, J., & Ducrot, 0. ( 19n). Deux MAIS en Fran~ais? Lingua, 43, 23-40.
Ariel, M. (1988). Referring and accessibility. Journal ofLinguistics, 24,65-87.
Ariel, M. {1990). Accessing noun-phrase antecedents. London: Roudedge.
Bateman, }.A., 8c Rondhuis, K. J. (1997). Coherence relations: towards a general
specification. Discourse Processes. 24, 3-49.
Birkmire, D. P. {1985). Text processing: The influence of text structure, background
knowledge and purpose. Reading Research Quarterly, 20, 314-326.
Black. J. B., & Bern, H. {1981). Causal coherence and memory for events in narratives.
Journal of Verbal Learning and Verbal Behavior, 20, 267-275.
Britton, B. K., Glynn, S.M., Meyer, B.}. F., & Penland, M. J. (1982). Effects of text
structure on use of cognitive capacity during reading. Journal of Educational Psy-
chology, 74, 51-61.
Britton, B. K. ( 1994). Understanding expository text: Building mental structures to
induce insights. In M.A. Gemsbacher {Ed.), Handbook of psycholinguistics (pp.
~o Ted Sanders and Wilbert Spooren
In W. Bublitz, U. Lenk, & E. Ventola (Eds.), Coherence in text and discourse (pp.
235-250). Amsterdam etc.: John Benjamins.
Sanders, T. J. M., Spooren, W. P.M., & Noordman, L. G. M. (1992). Toward a tax-
onomy of coherence relations. Discourse Processes, 15, 1-35.
Sanders, T. }. M., Spoorcn, W. P.M., & Noordman, L. G. M. (1993). Coherence rela-
tions in a cognitive theory of discourse representation. Cognitive Linguistics, 4, 93-
133.
Sanders, T ., & van Wijk, C. ( 1996). PISA- A procedure for analyzing the structure of
explanatory texts. Text, 16,91-132.
Sandra, D., & Rice, S. (1995). Network analyses of prepositional meaning: Mirroring
whose mind- the linguist's or the language user's? Cognitive Linguistics, 6, 89-
130.
Sanford, A.}., & Garrod, S.C. (1994). Selective processes in text understanding. In
M.A. Gemsbacher (Ed.), Handbook ofpsycholinguistics (pp. 699-720}. San Diego
etc.: Academic Press.
Schilperoord, }. (1996). It's about time. Temporal aspects of cognitive processes in text
production. Amsterdam: Rodopi.
Schilperoord, }., & Sanders, T. ( 1997). Pauses, cognitive rhythms and discourse struc-
ture. An empirical study of discourse production. In A. Liebert, G. Redeker, & L.
Waugh (Eds.}, Disourse and Perspective in Cognitive Linguistics (pp. 247-249).
Amsterdam: Bcnjamins.
Schilperoord, }., & Sanders, T. (1999). How hierarchical structure affects retrieval
processes; implications of pause and text analysis. In D. Galbraith, & M. Torrance
(Eds.), Conceptual processes in writing (pp. 13-30). Amsterdam: Amsterdam Uni-
versity Press.
Singer, M. (1990). Psychology of language: An introduction to sentence and discourse
processes. Hillsdale, NJ: Erlbaum.
Sperber, D., & Wilson, D. (1992}. Relevance: Communication and cognition (2nd ed.).
Oxford: Blackwell.
Spooren, W. P.M. S. ( 1989). Some aspects of the form and interpretation ofglobal con-
trastive coherence relations. Ph.D. Dissertation Nijmegen University.
Spooren, W. (1997). The processing of underspecified coherence relations. Discourse
Processes, 24, 149-168.
Spooren, W., Mulder, M., & Hoeken, H. (1998). The role of interest and text structure
in professional reading. ]oumal ofResearch in Reading, 21, 109-120.
Spooren, W., & Sanders, T. (in prep.). What does children's discourse tell us about the
nature ofcoherence relations? Paper VU Amsterdam I UiL OTS UtrechL
Spooren, W., Sanders, T., & Visser, J. (1994). Taxonomic van cohcrcnce-rclaties:
Evidentie uit taalverwervingsonderzoek? (Taxonomy of coherence relations: Evi-
dence from language acquisition research?) GrammafiTI, 3, 33-54.
Spyridakis, f. H., & Standal, T. C. ( 1987). Signals in expository prose. Reading Research
Quarterly, 22, 3, 285-298.
Stein, D., & Wright, S. (Eds.). (1995). Subjectivity and subjectivisation. Cambridge,
Cambridge University Press.
Text representation as an interface between language and its users 2.5