Broek, P., Rohleder, L., & Narvaez, D. (1994).
Cognitive processes in the comprehension of
Literary texts. In H. Oostendorp & R. Zwaan (Eds.),
Naturalistic Text Comprehension, (pp. 229-246).
Norwood, NJ: Ablex Publishing Corp.
13
Cognitive Processes
in the Comprehension
of Literary Texts
Paul van den Brock
Lisa Robleder
Darcia Narvaez
University of Minnesota
When we read a novel, a newspaper article, or any other type of
naturalistic text, we do not perceive the individual sentences to
convey isolated pieces of information. Rather, we perceive them as
being interconnected, as forming a coherent whole. The perception
of coherence is the result of extensive inferential processes, some of
which are automatic while others are intentional and strategic. In
these processes, the reader draws on his or her general background
knowledge as well as on memory for the preceding text to establish
how the events. objects. and people described in the present state-
ment are related to those in statements that were described earlier
‘The vast majority of the connections that readers detect in a text
are between statements or events that follow each other closely. For
example, in Charles Dickens’ A Christmas Carol, Mr, Scrooge’s gift
of food to his employee is directly related to the fact that the family
rejoices in having a bountiful Christmas. Other connections, how-230 van den Brock etal
ever, may connect statements or events that are separated widely in
the text. For example, in Tolstoy's Anna Karenina, the fact that
Anna commits suicide by throwing herself infront ofa train attains
added significance when the reader realizes the connection to the
fact that she met her lover and nemesis. Prince Vronsky. in a train
‘scene hundreds of pages earlier. Thus, the perception of coherence
is not only necessary for minimal comprehension of what one reads,
but also is an essential component of the aesthetic and emotional
experience that the text evokes in the reader.
The purpose of the present chapter {s to discuss the coherence-
building activities that underlle comprehension of texts. The chap-
ter consists of three parts. The first part provides a description of
the inferential processes that take place during reading, based pri-
marily on research with nonnaturalistic, experimenter generated
texts, In the second and third sections, the applicability of this
general model to reading of naturalistic, in particular literary, texts
{s explored. in the second section, it is demonstrated that many of
the inferential processes that take place during reading of non-
naturalistic texts also occur curing the reading of naturalistic texts.
‘The third section, in contrast, focuses on the differences between
reading of naturalistic and nonnaturalistie texts. Here we suggest
‘hat naturalistic texts may involve more extensive. and in some
respects qualitatively different, inferential activities and that they
may provide a better window on the communicative process as a
whole, which includes the reader, the text, and the author.
INFERENTIAL PROCESSES DURING READING
A central component of successful reading is the construction of a
coherent mental representation of the text, Such a representation,
results from extensive inferential processes that take place during
reading, These inferences generally fall into two categories. First,
inferences can be associative. These inferences result when words.
concepts, or events are highly associated with the information con-
tained in the statement that presently is being read. Second, infer:
fences may be coherence-based. Coherence-based inferences are
generated in an atte*hpt to connect information in the present state
‘ment to the remainder of the text. These two types of inferences are,
of course. not independent; frequently, the associates that are act!
vated during reading of the statement may fulfll the function of
establishing coherence (cf, Kintsch, 1988). For example. consider
the following sentence pair (from van den Broek, Risden. & Husebye.
Hartmann, 1992):
Margle intently peered at the beehive
‘Suddenly. she screamed and slapped her arm ao
In this case, the reader is likely to establish coherence by inferring
that Margie was stung by a bee. This inference is supported both by
association ("beehive” is associated with “bee” and “sting") and by
coherence ("sting’ allows the reader to connect the two statements}
(See below for a detailed discussion of coherence.) Thus, in this case
an associated inference would, as a side effect, also establish a
coherence inference.
At other times, however, the establishment of coherence takes
more effort because associates of the words in the sentences do not
provide a basis for coherence. In this case, the reader may search his
or her memory for prior text or may access background knowledge in
order to resolve the ambiguity in the present statement (Gerns-
bacher. 1990; O'Brien, 1987: van den Broek, in press). Consider the
following sentences:
Margie peered intently at the nest
Suddenly. she screamed and slapped her arm, a
Here, the reader also is likely to infer that Margie was stung by a bee.
ut now this inference is only supported by the demand for coher:
nce not by a strong association (the association between “nest” and
ting” is weaker than between “beehive” and “sting’). In order to
understand that these sentences are related, the reader needs to
access his or her background knowledge.
‘Two situations in particular are likely to lead to extensive, co-
herence-based processing, namely anaphoric ambiguity and nsufli-
‘lent causal explanation. Anaphoric ambiguity occurs when an ob-
Ject or person mentioned in the present statement is not clearly
Identified. Consider, for example, the following sentence pair:
‘The editor had to make some decisions concerning the manuscript.
‘She decided to take a closer look at it. @
In this pair of sentences, the word "she" Is disambiguated by con-
necting it to “the editor” in the previous sentence. This connection
hhas two results: on the one hand, it now is clear to whom the word
“she” refers: and on the other hand, the reader's knowledge of “the
editor” has become more detailed, in that the gender of the person ts
now specified. The resolution of this anaphor is relatively easy in
this situation, because the disambiguating information is provided
in the immediately preceding statement. As was noted above, how-252 van den Brock et al
ever, the disambiguating information may have been presented
‘much earlier in the text or may be absent from the text altogether. In
these cases. backward memory searches or activation of background
Knowledge will take place.
Causally insufficient explanation occurs when the event de-
scribed in the present statement requires explanation in the current
circumstances of the text. Consider, for example, the following sen-
tence pair:
‘John accidentally dropped the banana peel.
Danielle fell on her back. a
In this ease, the event of Danielle’ falling on her back ts infrequent
‘enough to warrant a search for explanation. Such an explanation is
readily found in the preceding sentence by making an inference that
it was the banana peel that made Danielle fall. Notice that in this
example, simple activation of the preceding statement is not sufli-
cient; in addition, the reader will have to access background know!
edge that banana peels are often associated with people's falling
(Singer. Halldorson, Lear, & Andrusiak, 1992). As with anaphoric
ambiguities, when the immediately preceding sentence does not
provide causally sufficient explanation, extensive memory or know!-
edge searches may be required for comprehension.
‘These inferential processes can be summarized as follows (sce
Table 13.1). As the reader proceeds through the text, each statement
that is being read evokes activation-based and coherence-based In-
erences, Activation-based inferences include the anticipation of
future events (for example, in Example 2, that Danielle may scold
ohn). likely preceding evens (Danielle's falling might evoke the
notion of “slipping”), or information that is contemporaneous with
the present statement (“editor” in Example 3 may activate “books”
or “Journals"). Coherence-based inferences take place when compre-
TaBLe 13.1
“Types of Inferences During Reading
1 Associate inferences
2 Aniipatjon of fae eves
1. Activation of preceding events
. Gontemporaneous information
2. Coherence Based inferences
|. Connecnginerences to activated information
5. Reinsaterets rom ing tem memory
Elaboration, based on baiground knowledge
Cognitive Processes 233
hension is inadequate. They can take the form of connecting Infer-
‘ences, which connect the presently read statement to events from
the prior text that are still activated (usually the preceding state-
ment as in Example 4}. They also may involve the reinstatement of
information that was encountered in the prior text from episodic
memory. Finally, they may constitute elaborative inferences by
‘means of the activation of background knowledge in semantic mem-
ory that explains or disambiguates what is being read at present.
For example, when one reads the following as part of a text: "The
secretary put the folder with papers under his arm and ran towards
his car. The wind blew open the folder, and the papers got all wet,
the reader is likely to make an elaborative inference (o explain the
papers getting wet, perhaps by inferring that it must have been
raining. Note that all three types of coherence-based inferences are
backward, in that they connect the current statement to prior
events, whereas association-based inferences can be backward, for
ward, and concurrent to the statement that is being read.
Empirical evidence for this general description of the reading
process comes from studies in which the activation of concepts and
events is measured. Among the most frequently used measures are
naming and lexical decision. In each case, the opeed with which
readers respond to target words (in the naming task by pronouncing,
the word. in the lexical decision task by indicating whether the
presented letter string is a word or not) is taken (o be an indication
Of the availability of the tested information. If the speed with which
subjects respond systematically is faster than in a contral situation,
it ts concluded that an inference was drawn. The results of such
studies indicate that assoclative Inferences are routinely generated
during reading (Sharkey & Sharkey, 1992; Till, Mross, & Kintsch,
1988). These activations generally are short-lived, however, unless
they establish coherence. When they do establish coherence. they
are likely to be maintained and possibly encoded in a relatively stable
‘memory representation of the text. Coherence-based inferences also
have been found to be generated. Connecting inferences (2a in Table
13.1) contribute to the memory for textual information (Myers &
"Weshould be pointed out thatthe rote of context diferences should be
‘considered and. if possible. minimized, before conclusion art drawm. ‘The reason
for this tg that responses in one condition also may be faster if the information
{fo which the eader has to respond Is supported more strongly by. or is more
‘compatible with. the textual context gen in that condition chai he other
condition (cf. Keenan, Potts. Golding. & Jennings. 1990, an Oastendorp. 1991)
‘The cited studies all conte for such fects,234 van den Broek et al
Duffy. 1990}, Information that is hypothesized to be reinstated (2b)
is more available immediately after such a reinstatement was to take
place than at other points in the text (Dopkins, Klin, & Myers, in
1993; Suh & Trabasso, 1993; Thurlow, 1991; Trabasso & Suh.
1993). The evidence for elanorative inferences (2c) is less clear.
although some empirical support exists (Graesser & Clark, 1985:
Singer et al.. 1992: Whitney. Ritchie & Crane, 1992)
‘The end result of these inferential processes is the construction of
coherent mental representetion of the text. Such representations
include the individual pieces of information presented in the text,
the relations among these pieces of information, and information,
that was inferred in order to provide coherence. Thus, the mental
representation constitutes a situation model, including both explic-
Itly presented information and concepts that were inferred, rather
than a mere reflection of the text base. The representation can be
conceived of as a network of interconnected nodes. Many properties
of these networks have been found to have psychological validity.
Statements that have many causal connections to other statements
in the text tend to be recalleé more often than statements with few
connections (Fletcher & Bloom, 1988; Graesser & Clark, 1985:
‘OBrien & Myers, 1987: Trabasso & van den Brock, 1985). State.
mente that are on a path that leads through the network froin Ute
beginning to its end also are recalled more often, summarized more
frequently. and rated as more important than are statements that
are not on this path (Omanson, 1982). Finally, the more connec-
lions a statement has, the more quickly it will be remembered
(OBrien & Myers, 1987),
In sum, the results of research on experimental texts suggest that
the reader engages in extensive inferential processes. with some
inferences being based on assoclations and others on a need for
coherence, as dictated by tke reader's purposes for reading and
eriteria for comprehension. Together. these inferences result in a
coherent memory representation of the text that resembles a net
‘work of interconnected nodes. The representation provides a situa-
ton mode! that contains both explicily mentioned and inferred
information,
COHERENCE IN NATURALISTIC TEXTS
Recent evidence suggests that similar processes take place during
the reading of nonexperimental, naturalistic texts. In one study,
subjects read and recalled short literary stories (van den Broek,
Robleder, & Narviez, in press). The structure of these stortes, a
Grimm fairy tale and a shor: story from a novel by Maxine Hong
Cognitive Processes 235
Kingston. is considerably more complex than that of texts typically
used in experimental studies on reading. For example, they contain
‘multiple protagonists. have strongly interwoven causal chains. and
are characterized by many unexpected turns. For these stories, the
probability that a statement was recalled depended strongly on the
umber of its causal connections to other statements in the text
This suggests that the mental representation of these naturalistic
texts also resembles a network of causally connected nodes, Further
more, when a second group of subjects read the text but was fre-
quently interrupted by the presentation of a string of letters. their
speed at deciding whether the letter strings were English words was
considerably faster when the letter string represented a word that
‘was hypothesized to have been reinstated or elaborated than when
the letter string was either a nonword or a word that was not
required for coherence at that point in the text. Similar results have
been observed by Singer et al. (1992), who found that subjects
systematically inferred elaborative information when it was needed
{or comprehension.
‘Although these results suggest that there is similarity between
the processing of naturalistic and nonnaturalistic texts, many more
‘systematic explorations of this issue need to be conducted. Support
for the notion that the above inferential processes generalize beyond
‘experimenter-generated (ext comes from studies on the comprehen-
sion of naturalistic but nontextual events. however. For example,
‘children as well as adults have been found to represent events that
they encounter in the context of television programs as semantic
networks (van den Broek, Lorch, & Thurlow, 1993). Similarly. these
processes have been observed in the comprehension of plecure sto-
Fes as well (Trabasso & Nickels. 1992; Trabasso. Stein, Rodkin,
Munger. & Baugh, In press). As a final example, the structure of
descriptions of accounts of complicated events such as U.S. govern-
ment reports on the space shuttle Challenger disaster or the 1987
stock market crash reflect extensive causal inferences similar to
those described above (Shapiro, van den Broek, & Fletcher, 1993)
‘Thus, the inferential processes and the resulting mental representa
tions are a natural part of comprehension of events regardless of
whether these events are depicted in experimental texts, in natue
ralistic texts, or even in nontextual discourse,
‘THE IMPORTANCE OF NATURALISTIC TEXTS
IN THE STUDY OF READING COMPREHENSION
Although the commonalities between processing of nonnaturallstic
and naturalistic texts are substantial, there also are important dif-
ferences between the processes involved in each. These differences236 van den Brock et al
center around two Issues. The first issue concerns the extent to
which inferential activities take place as a regular part of reading.
‘The second issue concerns the unique communicative context that
is involved in the reading of naturalist texts.
‘The Extent of Inferential Processes
Although the research based on experimental texts indicates the
‘occurrence of various types of inferential processes during reading,
the frequency with which several of these inferences take place has
been found to be rather limited. For example, forward inferences, in
which future parts of the text are anticipated, have been found to
not occur at all {Keenan et al, 1990: Singer & Ferreira. 1983) or to
‘be made only when the text provides strong constraints toward that
particular inference (Murray. Klin, & Myers, 1993; van den Broek,
Risden. & Husebye-Hartmann, in press; Vonk & Noordman, 1990),
A cursory inspection of literary texts suggests that forward infer-
ences are generated routinely and with great frequency during the
reading of these materials, Part of the attraction of some of the great,
Russian novels, for instance, lies in the fact that the reader observes
something that the protagonists do not observe, namely, that the
protagonists are inevitably being drawn cowards a particular out:
come. Similarly, a crucial cemponent of tragedies, such as those
written by the Greek masters Sophocles, Aeschylus, and Euripides,
is the realization by the audience that actions performed at the
present moment have consequences, unbeknownst to the perpetra-
tor. that can span generations. In these tragedies, forward infer-
ences such as these often even extend beyond individual tragedies,
thereby creating coherence between several plays. As a final example
of forward inferences. the power of many novels rests to a large
extent on the element of surprise that occurs when the forward
inference made by the reader has to be revised in the face of new
information in the text (Brewer & Lichtenstein, 1982).
The findings on experimental texts suggest that backward, co-
herence-building inferences are generated much more consistently
than forward inferences. Even some of these inferences, however.
have been found te occur only under special circumstances.
Whereas connecting backward inferences are made with great re-
lability, the results on experimental texts indicate that reinstate.
ments and backward elaborative inferences frequently are made
‘only when local coherence fails (o be obtained (but see Albrecht &
OBrien. 1993; van den Broek & Lorch, 1993, for evidence that
reinstatements and elaborations occur even in the presence of local
Cognitive Processes 237
coherence}. That is. they are made when anaphoric ambiguity exists
or when the present event is not adequately explained by the imme.
diately preceding textual information. Again, it is likely that back-
ward inferences are gencrated with a much higher frequency during
the reading of naturalistic texts. For example, detective stories and
spy novels frequently conclude with a denouement in which the final
information results in a restructuring of one's understanding of
events that were described much earlier in the text, even though the
novel information is coherent with the immediately preceding text.
Likewise in Tolstoy's novel the dramatic impact of Anna Karenina’s
sulcide by throwing herself in front ofa train is caused in part by the
fact that her tragle circumstances began at the same railway sta-
tion. in this case, Anna’s suicide is coherent within the context of
the immediately preceding text, but the reader nevertheless (for
reasons to be discussed below) is drawn towards integrating it with,
distant text.
Finally. the inferences that are elicited by literary texts may in-
clude some that are of a different nature than those evoked by
experimenter-generated materials, Whereas experiinental texts of-
ten depict rather stereotypical, almost scriptlike events, literary
texts frequently describe events that are unexpected or emotionally
charged. or provide novel interpretations of familiar events, As a
result, literary texts bring about powerful. global and, frequently,
affective responses in the reader that rarely accompany the reading
of experimenter-generated texts.
‘Thus. even a cursory examination of literary texts suggests that
Inferential activities during reading are more plentiful and may
Include inferences of a different nature than the research based on.
‘experimental texts would lead one to believe, Recent investigations,
in which experimental methods are applied to naturalistic text, have
‘begun to provide empirical support for this conjecture. For example,
readers proceed through some naturalistic texts at a pace that is
much slower than that typically observed with experimental texts
(Carpenter & Just. 1992). In some types of naturalistic texts, the
reading speed may decrease by as much as 50 percent. These results
suggest that nacuralistic lexts evoke considerably more inferential
activity than do experimenter-generated texts—although some of
the increased time is spent primarily on surface features of the text
(ef. Zwaan, 1992)
‘The notion that the generation of inferences is even more impor-
tant in the reading of naturalistic texts than in that of experimental
texts is supported by the results of the eartier described study by van,
den Broek et al. (in press). As with experimental texts, statements
with many causal connections were recalled much more frequentlyr
238 van den Bock et al
than were statements with fewer causal connections. The relation
between number of connections and memorability was considerably
stronger in the titerary texts than had been found previously in
studies on experimental texts. however. This suggests that the
memory representations of naturalistic text are more strongly af-
fected by the causal structure of the text than are the representa.
tions of nonnaturalistic tex, either because the former are more
densely interconnected or because their causal connections are
more strongly encoded in memory. Furthermore, whereas previous
studies have shown little or no evidence that elaborative inferences
are made, the evidence that such inferences were generated during
reading of naturalist text is quite strong,
In sum, both Intuitive analyses of literary text and recent results
of experimental studies of reading of Ierary texts suggest that
inferential activities are coasiderably more extensive during. the
reading of naturalistic texts than during the reading of ‘ex:
perimenter-generated texts. Why might this be the case? There are
At least two possible reasons for this difference. First. reader tn-
tolvement is ikely to be higher for naturalistic texts than for exper
mental texts. Second, the sructure of naturalistic texts may elicit
more extensive and varied inferential processes
Reader involvement is high in naturalistic stories for several rea-
sons, One reason. for exampee. s that they tend to be more interest.
‘ng (MeDaniel, Hines, Wadaill, & Einstein, 1992), Elements of sur.
Prise. twists in the flow of the narrative, and the choice of topic al
Combine to create a more conducive environment for inferential
activity. Second, naturaliste texts usually are longer than non-
naturalistic stories. One consequence is that the reader ts drawn
into the hypothetical world deseibed in the text and hence becomes
involved in the events that take place. Third, naturalistic texts re
quently provoke a unique literary experience such as the perception
of closure, aesthetic appeal, and moral implication, which divectly
depends on and therefore encourages inferential processes (f. Tan,
this volume). As a final example, the reader's purposes in reading
natutaliste texts differ from those typically found with respect to
artificial texts (van Oostendorp, 199%. this volume). As a result of
these and other properties, reader involvement in naturalist texts
generally Is greater than in nonnaturalistic texts
Likewise, the structure of naturalistic texts prompts more infer
ential processes than that ef nonnaturalistic texts. Again, several
contributing factors can be discussed. First, the length of naturals
tic texts noted above not ony evokes involvement, it also makes It
{Imperative for the reader to make inferences in order to Keep track of
the flow of the narrative, Second, naturalistic texts tend to be com
Cognitive Processes 239
plicated in structure. Unlike experimental texts, naturalistic texts
usually Involve multiple protagonists with multiple plans that all
intertwine and are mutually dependent. The resulting complexity of
naturalistic texts forces the reader to engage in considerable infer-
‘ential activity. Third, as mentioned above, literary texts often
describe unfamiliar or unexpected patterns of events whereas
‘experimenter-generated texts usually are quite conventional. Liter:
ary texts therefore may evoke more inferential and integrative pro-
cesses (cf. van Oostendorp. this volume) than typically are found
‘with experimenter-generated texts. Fourth, the author of a literary
text may control the flow of information in order to elicit specific
responses on the part of the reader, for example by foregrounding
information or by selecting particular linguistic constructions. As a
result, the Informational structure is less straightforward and may
even violate expectations by the reader, thereby requiring problem:
solving activities by the reader.
‘The examples illustrate how reader involvement and structural
‘complexity together may prompt inferential processes usually not,
found in the reading of nonnaturalistic texts. As a result, research
based on nonnaturalistic text typically will underestimate the extent
of inferential activity that takes place during regular reading. Only
the use of naturalistic texts will allow a proper appreciation of these
‘inferences. In addition, the use of naturalistic text allows the con.
‘sideration of various factors that Influence the generation of differ
‘ent types of inferences. These factors might never be studied sf the
‘occurrence of the inferences is itself dismissed because they do not
occur in reading of experimenter-generated texts.
Coherence and Communicative Discourse
Although much psychological research on reading focuses on the
comprehension processes that take place in the reader (much like
reader-response theory does in literary analysis), reading typically ts
embedded in the larger context of communicative discourse. We
propose that the act of communication through discourse is the
result of an interaction between author, text, and reader. Although
cach of these elements can be and has been investigated in isolation,
{tis only when they are considered jointly that the full complexity of
communicative discourse is revealed. Examples of aspects of each of
the elements that contribute to the communication are given in
Table 13.2,
‘These aspects interact extensively during the communicative pro-
cess. For example, the extent to which the author's message will be
conveyed adequately to the reader is limited by the extent to which240 van den Broek et al,
ABLE 13.2
Aspects of Author, Text, and Reader that Contribute
to Communicative Discourse
Autor
TT Mesage 1 be conveyed
4 Wea ei
4. Background krowledge
1 Suc complesity
2. Galt vip formation
5: Linguiic cue to inlerenal poceses eg. comectvs)
4 Cee
Reader
1 Mesvaioneading gos!
2 Background knowledge
3. Reading selleeateges
5 MlrteanonyEnaviral imokement wi petagenst
the author and reader bring similar background knowledge to bear
upon the situation. For example. in Euripides Btectra, Electra rolls
out a red carpet for her returning husband, Agamemnon. The Greek
audience would have realized the fact that the color red is associated
with death and therefore would have grasped Immediately the dra-
‘matic relevance of Agamemnon’s stepping onto the carpet. For a
‘modern-day western audience, the implication of Agamemnon’ act,
may very well remain elusive. Similarly. the author ofa literary work
implicitly or explicitly has the properties of the reader in mind when
‘writing, Information that can be inferred readily, m the judgment of
the author, may be omitted from the text: elements of surprise.
‘emotion, and so on, can be inserted in the text on the basis of the
author's intuitions of the reader's reading processes. Thus, proper
tues of the reader affect the author and, vice versa, the author's
Incuitlons affect the experience in the reader (Brewer & Lichten:
stein, 1982; Britton & Gulgoz, 1991; van den Broek et al., in press)
All of the above mentioned effects are mediated by the text. The
text is the end produict of the author's intentions as well as of his or
her intuitions about the reader's processes. From the reader's point
fof view. however, the text is the starting point. Regardless of the
author's intentions or intuitions, the content and structure of the
text will elicit inferential processes on the part of the reader. AS has
been described extensively above, aspects of the reader and those of
the text will interact to yield a unique memory representation of the
text in the reader.
Cognitive Processes 241
Although both psychological research and theories of literary crit-
{ism (for example, constructivism and reader-response theory] tend
to focus primarily or exclusively on one element or another, commu-
nicative discourse can be understood only by investigation of all
elements (consider the diagram of the interactions between the
author. text. and reader elements in Figure 13.1). Across instances
of communicative discourse, the amount of overlap between the
three elements varies. For example, although a text will capture
parts of the author's intentions, thoughts, and so on, as indicated
by the intersections labeled B and D in Figure 13.1. substantial
aspects of the author's thoughts are not represented in the text and
the text may contain information not intended or shared by the
author. as indicated by the nonoverlapping parts of the author and
text circles in Figure 13.1. Similarly, the reader may extract only
some parts of the text, indicated by Segments D and C. Some parts
of the text are not grasped by the reader and, conversely, the text
‘may not touch upon many aspects of the reader's world of thought,
as indicated by the nonoverlapping segments of the reader and text
circles. Finally. the author and reader may share many properties,
intentions, and thoughts. as indicted by Segments A and D, but
they also have many unique properties, as indicated by the non-
overlapping author and reader circles
Only the segment indicated by D represents an instance of com:
FIGURE 13.1 Conmunictve course at an Inieacton of Author, Test, an Reader
Author, Reader
Text242 van den Broek et al
muntcative discourse that involves all three elements, The larger the
segment, the more extensively the author, reader. and text are
united. A text perfectly communicates the author's intended mean-
ing to the reader if neither assigns meaning to the text that is not
shared by the other—that 's, if the Segments B and C are empty.
This is unlikely to occur for various reasons. If the author and
reader have very little in common, as may happen when they live in
different cultures or have very different belief systems, then the
Ukelihood of true communication is low. Similarly. the common-
allties between the reader and the text. and the author and the text,
respectively, may vary substantially. Finally, note that each element
may share with another element without the involvement of the
third element. Thus, the reader may extract the information from
the text even though that was not part of the author's intentions or
beliefS, as indicated by Segment C. Likewise, the author and reader
‘may share beliefs or properties that are not conveyed by the text (as
Indicated by Segment A), and the text may convey information in-
tended by the author but not grasped by the reader (as indicated by
Segment B).
‘These considerations represent some of the implications of the
hypothetical model of communicative discourse. Others could be
pursued in a similar fashion, but the main point is that in order to
understand the full complexity of communicative discourse i 1s
essential to investigate the role of all three elements. In order to do
50, it is imperative to use naturalistic texts. The author's intentions,
in experimenter-generated texts usually are limited to simply creat-
ing a text that possesses certain structural properties. This has
‘several important implications for the validity of any conclusions
about reading and communicative discourse. On the one hand, an
important element of communicative discourse. the author's inten
tons, is omitted from the picture. On the other hand. and perhaps
more importantly from the point of view of psychological theories of
reading, the obvious absence of intentions. purposes. or hidden
meanings on the part of the author are likely to affect the compre-
hension processes in which the reader will engage (ef. Gibbs, Kush-
nner. & Mills, 1992; Zwaan, 1992), Asa consequence. a deseription of
the cognitive processes during reading that is based on experiments
‘with nonnaturalistte texts only will be an impoverished one because
{tis likely to omit central inferences that the reader makes about the
author's purposes and consequently about the inferences that are
generated in response to the textual information itself.
‘The Communicative Discourse view of reading (and writing) sug
‘gests unique directions for future research. For example. the simul
taneous consideration of text, author. and reader properties allows,
Cognitive Processes 243
fone to determine whether readers’ intuitions about the author's
Intentions affect the comprehension processes in which they engage
and, vice versa, whether the author's writing process Is affected by
intuitions about the reader. Likewise, this depiction of the commu-
nicative process predicts that the author's/reader’s reliance on text
properties in writing and understanding is diminished as their
shared background knowledge increases, A final example of the
Implications of this theoretical framework is that memory for iter-
ary texts, compared to memory for experimental texts, will be deter-
mined even more strongly by the structural properties of the reader's
‘mental representation of the texts. Also, literary texts this represen-
tation would include affective components in addition to the an-
aphoric and causal ones typically found in the research on experi
mental texts.
CONCLUSIONS
‘The past decade has seen tremendous advances in our understand-
Ing of the cognitive processes that take place during reading. It has
become clear that a crucial component of successful reading is the
construction of a coherent mental representation of the text. This
representation resembles an elaborate network of interconnected
pieces of information. The connections in this network are the
result of inferential processes that take place as the reader proceeds
through the texts. Some of these inferences are based on assocta-
tions, while others are based on the need for coherence, particularly
that of anaphoric clarity and causally sufficient explanation.
‘The inferential and representational processes that take place
during the reading of nonnaturalistic texts have been found to take
place during the reading of naturalistic texts as well. There are
important differences, however. First, inferential activities are likely
to be more frequent and in some respects different during reading of
naturalistic texts than during that of nonnaturabistic texts. This
‘may reflect both greater reader involvement and greater structural
complexity. Second. the use of experimenter-generated texts dimin-
ishes the importance of the triad of central elements of communica-
tive discourse during natural reading, namely author, text, and
reader. Above, we proposed that it is exactly the interactions be-
tween all three components that are central o reading. AS a result,
the picture of reading processes and comprehension that results
from research on nonnaturalistic texts will yield a simplistic render:
Ing of the richness of reading and underestimate the powerful, and
often long-lasting, effects that the text may have on the reader. The244 van den Brock eta.
complexity of the reading process and its impact on the reader can
only be gauged by extending existing experimental paradigms to the
study of comprehension of naturalistic texts.
REFERENCES
Albrecht, J. E., & OBrien &. J.{1993}. Updating a mental model: Maintain-
ng both local and global esherence, Journal of Experimental Psychol.
ogy: Learning, Memory. and Cognition, 19(5). 1061-1070.
Brewer, W. F., & Lichtenstein. E, H. (1982). Stories are to entertain: A
structural-alfect theory ofstortes. Journal of Pragmatics, 6, 473-486.
Briton. B. K. & Gulgoz, S. (1991). Using Kintseh’s computational model to
improve instructional tex! Eflets of repairing inference calls on recall
and cognitive structures. Journal of Educational Psychology. 3.
329-345.
Carpenter, P.A.. & Just, M. (May. 1992). How mental capacity constrains
werdry comprehension. Paper presented atthe third annual meeting
‘of the International Association for the Empirical Study of Literature,
Memphis, TN.
Dopkins,S.. Klin. Ce Myers, J. L. (1993). The accessibility of information
about goals during the processing of narrative texts. Joumal of Exper
imental Psychology: Leaming, Memory and Cognition. 1941), 70-80,
Fletcher. C. R.. & Bloom, C. P. 1988), Causal reasoning in the comprehen
sion of stimple narrative texts, Journal of Memory and Language, 27,
235-284.
Gemsbacher, M. A. (1990). Language comprehension as structure bulld-
ing. Hillsdale. NJ: Erlbaum.
Gibbs, RW., Kushner, J. M., & Mills, W.R.,l(1992). Authorial intentions
and metaphor comprehension. Journal of Psycho-linguistte Re
search, 30, 11-20.
Graesser. A. C., & Clark, L. F. (1985), The structures and procedures of
“implicit knowledge. Norwood, NJ: Ablex.
‘Keenan, J. M., Potts, G. .. Golding, J. M., & Jennings. T. M, (1990), Which
elaborative inferences are drawn during reading? A question of meth
edologies. tn D. A, Balote. G. B. Flores d'Arcals, & K. Rayner (Eds.),
Comprenension Processes tn Reading (pp. 377-898}, Hillsdale, NJ
Erlbaum.
‘Kintsch, W. (1988)’ The role o” knowledge in discourse comprehension: A
construction-ntegration model. Psychological Review. 95. 163-183.
MeDantel, M.A., Hines, R. J, Wadi P. J. & Einstein, G. 0. (May, 1992),
What produces unique memory patterns in folktales: Content faml-
(arty, causal structure, scripts. or superstructures? Paper presented
at third annual meeting ofthe International Association for the Es
pirteal Study of Literature, Memphis. TN,
Murray, J... Klin, C. M., & Myers, J. L- (1993), Forward inferences about
Cognitive Processes 245
specific events during reading, Journal of Memory and Language,
52, 464-473,
Myers. J. L., &Duffy, S. A, (1990). Causal inferences and text memory. In
‘A.C. Graesser & G. H, Bower (Eds.), Psychology of learning and
motivation: Inferences and text comprehension, 25. 35-31, Sun
Diego. CA: Academie
OBrien, E. J.. & Myers, J. L. (1987). The role of causal connections in the
retrieval of text. Memory and Cognition, 15, 413-427,
OBrien, E. J. (19871, Anapharic inference during reading, Journal of Ex:
perimental Psychology: Learning, Memory. and Cognition. 12,
346-352.
‘Omanson. R. C, (1982). The relation between centrality and story category
vartation. Journal of Verbal Learning and Verbal Behavior. 21
326-397.
‘Sharkey, A. J. C., & Sharkey, N. B. (1992), Weak contextual constraints in
text and word priming, Journal of Memory and Language. 31
543-572.
Shapiro. BP. van den Broek. P.. & Fletcher, C. R. (1999). The understand:
ng of complex real-ife events. Unpublished manuscript
Singer. M., Halldorson. M., Lear, J.C. & Andrusiak, P. (1992). Validation of
causal bridging inferences in discourse understanding. Joumal of
Memory and Language. $1, 507-524
Singer. M., & Ferreira, F, (1983). Inferring consequences in story compre-
hhension. Journal of Verbal Learning and Verbal Behavior. 22.
437-448,
Sub, S. ¥., & Trabasso, T. (1999). Inferences during reading: Converging
evidence form discourse analysis. tall-aloud protocols, and recogat=
tion priming. Journal of Memony and Language, J2. 275-300,
‘Thurlow. R. E. (1991). The inference of causal antecedents dunng the
reading of narratives. Unpublished doctoral dissertation, University
‘of Minnesota. Minneapolis, MN.
Tl, R. E., Mross. E. F.. & Kintsch, W. (1988). Time course of priming for
associate and inference words in a discourse context. Memory &
Cogntiion. 16(4}, 283-298,
‘Trabasso. T., & Nickels. M. (1992). The development of goal plans of action
tn the narration ofa picture story, Discourse Processes. 15, 249-275.
‘Trabasso, T., Stein. N. L.,.Rodkin. P.C., Munger, M, P.. de Baughn, CR. (in
Dress). Knowledge of goals and plans in the orine narration of
events. Cognitive Development
‘Trabasso, T...& Suh, S. ¥. (1993). Understanding text: Achleving explana
tory coherence through on-line inferenees and mental operations in
working memory. Discourse Processes, 16, 3-34
‘Trabasso. T.. & van den Broek, P. W. (1985), Causal thinking and the
Fepresentation of narrative events, Joumal of Memory and
Language, 24, 612-630,
van den Broek, P. W. (in press). Comprehension and memory an narrative
{exts: inferences and coherence. in M. G. Gersbacher [Ed]. Hand:
book of Psychelinguisties. New York: Academie.246 van den Brock et al
van den Broek, P.. & Lorch, R. F (1993). Network representations of causal
Felations in memory for narrative texts: Evidence from primed recog
nition. Discourse Processes, 16(1~2). 75-98,
van den Broek. P. W.. Lorch, E. P., & Thurlow. R. (1993). Children’s and
Adults’ memory for television stories: The role of causal factors, story:
‘grammar categories and hierarchical level. Unpublished manuscript.
van den Brock. P.. Risden. K., Husebye-Hartmann, E. (Nov, 1992), Causal
and assoetative constraints in the generation of bridging inferences,
Paper presented at the 33rd annual meeting of the Psychonomic Soci.
ety. St. Louis, MO.
van den Broek, P. W., Risden, K.. Husebye- Hartmann, E. (in press). Com:
prehension of narrative events: Maintaining sufficient explanation. In
R. F. Lorch, Jr. & OBrien, E. (Eds.). Sources of coherence in text
comprehension. Hillsdale, NJ: Erlbaum,
van den Broek. P. W., Rohleder, L., & Narvaez, D. (in press). Causal infer
‘ences in the comprehensten of literary texts. In R. J. Kreuz & M. S,
McNealy (Eds.). The empirical study of literature. Norwood, Ni
Ablex
van Oostendorp, H. (1991). Inferences and integrations made by readers of
script-based texts. Journal of Research in Reading, 14(1), 3-20,
Vonk, W.. & Noordman. L. (1990). On the control of inferences in text
understanding. In D. A. Balota, G. B. Flores d’Arcais, & K. Rayner
(Eds.). Comprehension processes in reading (pp. 447-465), Hills-
dale, NJ: Eribaum,
Whitney. P.. Ritchie, B. G., & Crane, R. S, (1992). The effect of foreground-
Ing on readers’ use of predictive inferences. Memory and Cognition.
20, 424-432,
‘2waan, R. A, (1992). Aspects of lterary comprehension: A cognitive ap-
proach. Doctoral dissertation. The Netherlands, Riksuniversiteit te
Utrecht.