Discourse Comprehension

and Memory

Chapter 7 by carroll

Instructor: Dr Dana

1
 Main points to be covered
• Introduction
• Comprehension of Discourse
• Memory for Discourse
• Schemata and Discourse Processing
• Educational Implications

2
 introduction
• This chapter deals with the ways we comprehend
and remember connected discourse—that is units
of language larger than the sentence.

• In our everyday lives, we process a number of

different types of discourse—for example, stories,
lectures, and sermons.

3
 Reasons behind discourse studies
• First, because we usually do not speak in isolated
sentences, discourse is a more natural unit of
language to investigate.
• Second, sentences are ambiguous, when we
consider them out of their discourse context.
• Finally, discourse provides a rich source of material
for those interested in the cognitive processes
involved in language use.
Discourse imposes a considerable burden on working
memory, while at the same time drawing heavily from
our permanent memory.
4
 Comprehension of discourse

• Local and Global Discourse Structure

• Cohesion

• Strategies Used to Establish Coherence

• Role of Working Memory

5
 Local and Global Discourse Structure

Comprehension of any connected discourse depends

on the arrangement of its sentences, not just the

meaning of its individual sentences.

Note that we could have a group of meaningful

sentences tied together, while having no sense at all.

6
 Local and Global Discourse Structure

• Local structure (sometimes called microstructure) is the

relationship between individual sentences in the discourse.

• global structure (or macrostructure) is our knowledge of

the structure corresponding to the related topic that
enables us to comprehend and remember the shorter
passage about that particular topic.

7
• Both levels of structure (local and global) contribute to the
coherence of a text, i.e., the degree to which the different
parts (words, sentences, paragraphs) of a text are
connected to one another.

Cohesion
• a discourse is coherent, if there are semantic relationships
between successive sentences.
• the range of possibilities that exist for linking something
with what has gone before is called cohesion (Halliday and
Hasan, 1976).
8
 Categories of Cohesion
• Reference: reference is a semantic relation whereby
information needed for the interpretation of one item is
found elsewhere in the text. It includes pronominal,
comparative and demonstrative types. (see table 7.1, p.
160)

• In substitution, we replace one lexical item with another

as an alternative to repeating the first.

• Ellipsis is a form of cohesion that is really a special case of

substitution in which we ‘‘substitute’’ one phrase with
nothing.
9
 Categories of Cohesion
• In conjunctive cohesion, we express a relationship
between phrases or sentences by using conjunctions such
as and, or, but, yet, and so.

• In lexical cohesion, a tie is made between one sentence or

phrase and another by virtue of the lexical relationships
between certain words in the sentence.

10
11
 Anaphoric and Cataphoric Reference

• When we use an expression to refer back to

something previously mentioned in discourse, the
referring expression is called an anaphor, and the
previous referent is called an antecedent.

• Alternatively, we sometimes use referring

expressions to point forward, which is called
cataphoric reference.

12
 Anaphoric and Cataphoric Reference
Anaphoric reference has been of the greatest
interest among psychologists. One reason is
that anaphoric reference enables us to explore
the role of working memory in discourse
comprehension.

13
 Anaphoric and Cataphoric Reference
To understand a simple pair of sentences, we must
hold the antecedent in working memory long enough
to link it with the anaphor.

Sometimes the distance between antecedent

and anaphor is much longer; long distances generally
impose a burden on working memory and ultimately
disrupt comprehension.

14
 Strategies Used to Establish Coherence
A great deal of the research has been done by the Clark
and Haviland (1977) on the distinction between given
(old) and new information.
• Given information refers to information that an author
or speaker assumes the reader or listener already
knows.
• new information is information that the comprehender
is assumed to not know. Most sentences contain both
given and new information.
15
Given/New Strategy
According to this strategy, the process of understanding a
sentence in discourse context consists of three sub-processes
or stages:

(1) identifying the given and new information in the current

sentence,

(2) finding an antecedent in memory for the given

information, and

(3) attaching the new information to this spot in memory.

16
Given/New Strategy
• The primary usefulness of this model has been in
examining the various possibilities that can occur
during stage 2 (i.e., finding an antecedent in
memory for the given information)

• Sentences that mark information as given, but have

no obvious antecedent from previous sentences
should pose comprehension difficulties.

17
Strategies used to establish coherence: Direct matching

Here, the given information in the target sentence directly

matches an antecedent in the context sentence.

(9) We got some beer out of the trunk.

(10) The beer was warm.

Finding an antecedent for given information in a target sentence

resembles searching for a concept, which is more than searching
for a word.

when we speak of direct matches, we are talking of matches of

underlying concepts previously introduced into the discourse .
18
 Strategies Used to Establish Coherence: Bridging

In some cases, we do not have a direct antecedent for the

given information, but can still tie the sentences together .

(13) Last Christmas Eugene went to a lot of parties.

(14) This Christmas he got very drunk again.

Here, we must make a bridging inference, such as that Eugene

got very drunk at last year’s parties, to make sense of the
word “again”.
19
 Strategies Used to Establish Coherence: Reinstating old
information
When a sentence refers to something or someone already
introduced, but no longer in the foreground, the
comprehender must reinstate the information that is to be
matched with the target information. Several studies have
shown that reinstatements increase comprehension time.
(see examples in page 164)

Example:
Yesterday a black dog bit a little girl. It got away, and we are
still trying to find it. He is short and has a dog tag on his neck
that says Fred. She was scared, but she wasn’t really hurt.

20
 Strategies Used to Establish Coherence: Identifying New
Topics of Discourse
All of the strategies used to establish coherence share the implicit
assumption that part of a target sentence should relate to earlier
information, but sometimes the information is all new and the target
sentence is meant to establish a new topic of discourse (topic shift). This
is easy to detect, when explicit markers such as the following are used:

“Now”,

“I want to move on to . . .” or

“this concludes our discussion of . . .”

21
 Strategies Used to Establish Coherence: Identifying New
Topics of Discourse

Until now we have focused on stage 2 of the strategies used

to establish coherence (i.e., finding an antecedent in memory
for the given information).

Let us now consider stage 3,i.e., the process of attaching new

information to the memory location defined by antecedents.

22
 Strategies Used to Establish Coherence: Identifying
New Topics of Discourse

Note that the process of adding new information to given

information subordinates the former to the latter. That is, the
new information is generally taken as an elaboration,
sometimes a small detail, of the given information.

Once introduced, this new information may itself serve

as an antecedent for later sentences, which are subordinated
to it. The natural result of this integration process is a
hierarchical structure in episodic memory. 23
 The Role of Working Memory

• individual differences in working memory might

influence how we comprehend discourse .

• The limited resources of working memory are

allocated to processing certain tasks as well as to
temporarily storing the results of these tasks. As a
result, we sometimes find ourselves in a trade-off
position.
24
 The Role of Working Memory

Daneman and Carpenter developed a complex reading

span task to examine this trade-off between processing
certain tasks and temporarily storing the results of these
tasks.
Their interpretation of the results was that
individuals with smaller reading span had smaller
working memory capacity, which made it difficult for
them to comprehend references more than a few
sentences back. 25
• They also found that their reading span measure
correlated significantly with their participants’
verbal SAT scores.

• both the reading comprehension test that the

researchers devised, and the verbal SAT, tap
working memory processes.
• A question: what does SAT stands for?

26
The SAT is defined as the Scholastic Assessment Test,
now called the SAT Reasoning Test, which is a test
that measures the reading, writing and math levels of
high school juniors and seniors. An example of the
SAT is the test many students take to get into college.

27
 Memory for discourse

If we approach a discourse in order to recall it

later on, what we understood from that
discourse is strengthened and reinforced
through some processes .

28
 Memory for discourse
Memory for discourse exists on three distinct levels :

1. surface representation of the discourse,

2. we construct a propositional representation of

the discourse,

3. we construct a situational model of the discourse

29
 Memory for discourse
memory for discourse exists on three distinct levels :

1. surface representation, in which we remember the exact

words that we encountered.

2. a propositional representation of the discourse, which

specifies the meaning apart from the exact words used.

These two levels are obviously similar to the corresponding levels

in our memory for sentences.

3. we construct a situational model of the discourse, which is a

model of the state of affairs in the world as described in the
30
passage.
• the surface or verbatim form of a sentence is stored
in working memory only until its meaning is
understood, then it would be eliminated to make
room for the next sentence.

• Apparently, we can remember the exact wording of

some material, when it is distinctive and easily
separable from the rest of the discourse.

• If we indeed eliminate the exact wording of a

sentence from working memory, what is left is the
propositional structure of the sentence.
31
• the number of propositions influences the time
required to read a passage when preparing to recall
it.

• discourse is stored as a network of propositions.

• we have a propositional structure in episodic

memory, not a verbal representation, after we have
understood a passage.

32
 Inferences and Propositional Representations

• Inferences are intrinsic to discourse structure. Authors leave out

information that they think readers will be able to figure out.

• The ability to restore discourse coherence requires more than the

ability to make connections between explicit propositions. We
should also be able to detect when an inference should be drawn.

• From a communication standpoint, an inference is a proposition in

the underlying discourse structure that is intended, but not
explicitly expressed by the author and thus must be drawn by the
reader.
33
We automatically draw inferences during reading only
when two conditions are present:
1. The inference must be necessary to make a text locally
coherent.
2. the information on which the inference is based, must
be easily activated (either from explicit statements in
the text or from general knowledge).
when we draw inferences from a text, we store the
implicit propositions right alongside the explicit
propositions we have derived from the text itself. 34
 Situational Models
• The assumption is that as we comprehend the
propositions of a text, we construct a mental or
situational model of the world as described by the
text.
• One possibility is a spatial layout. The phrasing of a
text may lead in either the development of a
propositional text base or a situational model.
• The construction of a situational model influences the
accessibility of previous information in discourse. 35
Simultaneous Investigations of all Three Levels (i.e., surface,
propositional and situational levels)
• We form surface, propositional, and situational representations
during the course of comprehending a discourse and we can
reliably distinguish between different levels of representation.
• Fletcher and Chrysler (1990) found that recognition memory was
worst when the test sentence and its distractor differed only at the
surface level, intermediate when they differed at the surface and
propositional levels, and best when they differed at all three
levels. Thus, students can reliably distinguish between different
levels of representation (see page 174).

36
 Simultaneous Investigations of all Three Levels
• surface memory is strong only in the immediate
context and falls to chance level shortly after that. It is
short-lived except when the wording is pragmatically
significant.
• Propositional recognition are much better retained
and include the meaning of presented information
along with any inferences we have drawn.
• Situational models are retained the best and are based
on spatial or causal relations between parts of a text.37
 Simultaneous Investigations of all Three Levels

We form surface, propositional, and situational

representations during the course of comprehending
discourse.
Most of the studies have tried to isolate one of these
levels or to distinguish between different levels.
It is helpful to set up a study that attempts to
investigate how each of the levels operates in the
same experiment. 38
 Simultaneous Investigations of all Three Levels
Fletcher and Chrysler (1990) have reported such a study.
They used passages (such as the one shown in Figure 7.6)
and the students were then given a recognition memory
test.
The items on the test were carefully constructed to probe
the surface, propositional, and situational levels.
For example, the distinction between rug and carpet taps
the surface level, because the meaning (propositional
level) is the same. In contrast, the distinction between
carpet and painting is at the propositional level.
39
 Simultaneous Investigations of all Three Levels
Fletcher and Chrysler (1990) found that recognition
memory was worst when the test sentence and its
distractor differed only at the surface level,
intermediate when they differed at the surface and
propositional levels, and best when they differed at all
three levels.
Thus, students can reliably distinguish between
different levels of representation.

40
41
 Schemata and discourse processing
• A schema (plural: schemata) is a structure in semantic
memory that specifies the general or expected
arrangement of a body of information.

• In some studies, Bartlett tried to show that

remembering is not a rote or reproductive process but
rather a process in which we retain the overall gist of
an event and then reconstruct the details from this
overall impression. 42
 Schemata and discourse processing
Bartlett suggested that when we encounter an event
that that we could not understand it easily, in fact, we
have difficulty fitting it into our existing schemata and
subsequently either we do not remember it ,or
maybe we try to ‘‘normalize’’ it,i.e., we try to alter its
details, until it becomes in accordance with the
existing schemata.
43
 Activation of Appropriate Schemata

• When we do not have a schema that corresponds to

the current story or event, comprehension and
memory would be poor; because it is impossible to
see the significance of the events being described.
• Sometimes we may have an appropriate schema in
memory, but fail to activate it for some reason.

44
 Reconstruction of Schema-Specific Details

• the schema that is active during comprehension has

a powerful organizing effect on recall.
• information central to the schema is well
remembered, but other details maybe misplaced,
although they can be reconstructed with a shift in
perspective.

45
 Genres
• Genre is a type of discourse that has a characteristic
structure.
• We have genres for lectures, sermons, opinion articles,
presidential opening speeches, and comedy monologues.
• Genres are important because they provide us with general
expectations regarding the way information in a discourse will
be arranged.
• One genre that has been studied a great deal in discourse
research has been narrative discourse.
46
• Typically, stories begin with the introduction of
characters and setting. The main character sets out
with some sort of goal, runs into some obstacles,
and ultimately resolves the dilemma /dailema/.

• There are many different genres for stories; in fact,

there are different ones for detective stories, fairy
tales, and romances.
47
• Narrative discourse can be contrasted with expository
discourse, in which the goal of the writer is not to tell a
story, but rather to convey information about the
subject matter. This is the form of discourse that we
encounter when reading a textbook or listening to a
lecture.

• In the expository discourse, the emphasis is on

presenting the information in an organized and logical
manner. 48
 Narrative Discourse Processing
• A story grammar is a schema in semantic memory that
identifies the typical or expected arrangement of
events in a story.

• story grammars view narratives as consisting of a

setting, one or more episodes, and then an ending.

• story grammars correspond to several aspects of how

comprehenders process simple stories. 49
 Narrative Discourse Processing

• episodes are an important unit in our memory for

stories. They tend to be recalled in an all-or-none
fashion, as if they are stored in separate chunks in
working memory.

• An implication of the view that episodes are processed

as chunks is that the boundaries between episodes
should be areas of high processing load. 50
• Haberlandt and colleagues studied the recall of stories
and found that some story constituents are recalled
better than others, i.e., beginnings, attempts, and
outcomes are recalled better than reactions, goals, and
endings.

• These results suggest that a story grammar could be a

universal schema rather than one that is specific to our
culture.
51
 Inaccessibility of Knowledge
• We have been discussing how we activate appropriate
knowledge bases during the course of comprehending
narratives. Now, we deal with some cases in which we could
not activate the appropriate knowledge.
• Note that when knowledge was not activated,
comprehension was severely impaired.
• It is also possible to comprehend a passage and still not
activate the relevant body of knowledge. Gerrig (1993) calls
this situation anomalous suspense (‫ تعلیق غیرعادی‬see page 183)

52
 Inaccessibility of Knowledge
For anomalous suspense consider, for example, the
following passage from Garrison Keillor (1987):
In Uncle Lew’s story, a house burned down on a cold
winter night and the little children inside ran barefoot
into the snow of 1906—some were pitched out the
bedroom window by their father—and all were safe. But
although I heard the story dozens of times, whenever he
told it again I was never sure they’d all get out. And since
these children grew up to be my ancestors, I had an
interest in their survival. (pp. 220–221) 53
 Educational Implications
• the comprehension activities we engage in when reading or
listening to a prose play a crucial role in the way we
understand (and misunderstand) what is being said.

• To learn a text’s content, we must store its underlying

structure.

• Much of what we study is material for which we have no

available schema.
54
 Educational Implications
• without the appropriate perspective, appreciating
the significance of even those concepts that are
learned and remembered is often difficult.

• When we deal with familiar material, we are

scarcely aware of the multiple ambiguities, missing
elements, and irrelevant distracting details, because
we are able to resolve most of them rather easily.
55
Five strategies to improve comprehension and memory

Actively Processing Discourse:Active processing refers to a

collection of activities that includes relating new information to
information we have in permanent memory .
• when we process information at the deeper semantic level, we
remember more of what we read
• self-reference effect is the tendency to remember information
better when we relate it to ourselves .

Connecting Propositions in Discourse: a strategy of explicitly

looking for relationship between concepts in discourse. 56
This strategy (explicitly looking for relationship between
concepts in discourse) leads to several beneficial results.
• First, it produces a network of interrelated
propositions in which each concept may serve as a
retrieval cue for many others.
• Second, even if we do not have the information
needed to draw an inference, explicitly searching for
the relationship between propositions deepens the
level of processing and hence promotes the retention
of individual propositions. 57
Finally, as propositions are connected to one another,
they are also subordinated or super-ordinated to one
another, thus leading to a hierarchical memory
structure that may be used to organize our recall of
the text or to summarize it.

58
Identifying the Main Points: the difficulty in determining
main points may be traced to the presence of distracting and
often confusing details.

• when the key points of a passage are signaled explicitly,

performance improves.

Building Global Structures: we need to identify important

points even when they are not so explicitly marked. As we
become more familiar with the content and structure of an
author’s prose, we can gradually deduce the author’s schema.

59
Tailoring Comprehension Activities to Tests

We should try to match our comprehension activities

to the types of tests we have to take.

• Memory researchers have established that

retention is best when we study material in a
manner similar to the way we must encode it at the
time of a test.
60
End of Chapter 7

61