Download as pdf or txt
Download as pdf or txt
You are on page 1of 21

This article was downloaded by: [University of Oklahoma Libraries]

On: 23 January 2015, At: 03:14


Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number:
1072954 Registered office: Mortimer House, 37-41 Mortimer Street,
London W1T 3JH, UK

Discourse Processes
Publication details, including instructions for
authors and subscription information:
http://www.tandfonline.com/loi/hdsp20

Strategies for resolving


coherence breaks in
reading
a b
Christopher M. Hakala & Edward J. O'Brien
a
Department of Psychology , University of New
Hampshire , Durham, NH, 03824
b
University of New Hampshire ,
Published online: 11 Nov 2009.

To cite this article: Christopher M. Hakala & Edward J. O'Brien (1995) Strategies
for resolving coherence breaks in reading, Discourse Processes, 20:2, 167-185,
DOI: 10.1080/01638539509544936

To link to this article: http://dx.doi.org/10.1080/01638539509544936

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all
the information (the “Content”) contained in the publications on our
platform. However, Taylor & Francis, our agents, and our licensors
make no representations or warranties whatsoever as to the accuracy,
completeness, or suitability for any purpose of the Content. Any
opinions and views expressed in this publication are the opinions and
views of the authors, and are not the views of or endorsed by Taylor
& Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information.
Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities
whatsoever or howsoever caused arising directly or indirectly in
connection with, in relation to or arising out of the use of the Content.
This article may be used for research, teaching, and private study
purposes. Any substantial or systematic reproduction, redistribution,
reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access
and use can be found at http://www.tandfonline.com/page/terms-and-
conditions
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015
DISCOURSE PROCESSES 20, 167-185 (1995)

Strategies for Resolving Coherence


Breaks in Reading
CHRISTOPHER M. HAKALA
EDWARD J. O'BRIEN
University of New Hampshire
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015

In two experiments, participants read passages containing an elaborate description of a


main character that was either consistent, inconsistent, or neutral with respect to actions
carried out by that character later in the text. The descriptions and subsequent target
actions were separated by filler material that created a coherence break at either a local
level or a global level. In Experiment 1, reading times for critical sentences containing the
target actions were significantly longer when the early description and the critical sen-
tences were inconsistent. This was true for both the local and global coherence conditions.
In Experiment 2, resolution of global inconsistencies improved memory for the regions of
the text that involved the inconsistencies, whereas resolution of local inconsistencies did
not. The results are discussed in terms of the different strategies readers may use to resolve
local and global coherence breaks.

Most models of reading comprehension assume that a reader develops at least


two levels of representation of a text: a text base, which is a representation of the
text itself (e.g., Kintsch & van Dijk, 1978), and a situation model, which is a
representation of the situation described by the text (e.g., Johnson-Laird, 1983;
Just & Carpenter, 1987; van Dijk & Kintsch, 1983). Because of limited capacity
constraints, it is generally assumed that these levels of representation are devel-
oped over a series of cycles, with only a small portion of the text processed on
each cycle (e.g., Kintsch & van Dijk, 1978). On each cycle, the reader selects a
subset of information to be held over in active memory to provide connections
for information entering during subsequent cycles. The strategy for selecting a
subset of information is generally one that will maximize the likelihood that
incoming information can be easily integrated with, or mapped onto, the current
contents of active memory (e.g., Fletcher, 1981, 1986; Fletcher & Bloom,
1988).

Correspondence and requests for reprints should be sent to Christopher M. Hakala, Department of
Psychology, University of New Hampshire, Durham, NH 03824.
Portions of this research were conducted by Christopher M. Hakala in partial fulfillment of the
requirements for the master's degree at the University of New Hampshire under the direction of
Edward J. O'Brien.
We thank John Limber and Elizabeth Stine who also served as members of the master's commit-
tee. We also thank Michelle Rizzella for scoring the recall data, as well as Jason Albrecht, Art
Graesser, and three anonymous reviewers for helpful suggestions on improving this article.

167
168 HAKALA AND O'BRIEN

Mapping can be described as a set of processes designed to establish both


local coherence and global coherence (Albrecht & O'Brien, 1993; Glenberg &
Langston, 1992; McKoon & Ratcliff, 1992; O'Brien & Albrecht, 1992; van Dijk
& Kintsch, 1983). Local coherence involves mapping the currently processed
information onto the immediately preceding context (i.e., information in active
memory, generally the previous one to three sentences). Global coherence in-
volves mapping currently processed information onto contextually relevant infor-
mation that is no longer active in memory, as well as organizing local
Information into higher order structures (Graesser, Singer, & Trabasso, 1994).
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015

The current set of experiments was designed to address two issues concerning the
maintenance of local and global coherence: First, we wanted to further demon-
strate that readers maintain both local and global coherence online; second, we
wanted to examine the strategies that readers may utilize to reestablish both local
and global coherence and the effect that those strategies may have on subsequent
comprehension and memory.
Some process models have focused primarily on the maintenance of local
coherence (e.g., Fletcher, 1981; Kintsch & van Dijk, 1978). For example, ac-
cording to the minimalist hypothesis (McKoon & Ratcliff, 1992), as long as local
coherence is maintained, no additional processing is necessary; global coherence
is only attempted when local coherence fails or when global information is
readily available. In contrast, others have argued that readers maintain coherence
at both a local and global level (e.g., Albrecht & O'Brien, 1993; Glenberg &
Langston, 1992; Glenberg, Meyer, & Lindem, 1987; Graesser et al., 1994;
O'Brien, 1995; Singer, Graesser, & Trabasso, 1994; van Dijk & Kintsch, 1983).
According to these models, readers are simultaneously concerned with integrat-
ing or mapping current information onto information that immediately preceded
it (i.e., the current contents of working memory), as well as with relevant
information that appeared earlier in the text, even if that information is no longer
active in memory.
There are two general views of how a reader may maintain global coherence.
Albrecht and O'Brien (1993; see also O'Brien, 1995) suggested that one way that
global information becomes available to the reader is through a passive, non-
strategic process; information or trace elements that are active in memory reso-
nate with other active traces as well as inactive but contextually relevant traces
from earlier in the discourse and relevant traces from world knowledge (e.g.,
Gillund & Shiffrin, 1984; Hintzman, 1986; Ratcliff, 1978). All reactivated traces
are then mapped onto the active portion of the representation with little difficulty
unless they introduce some sort of coherence break.
Alternatively, Graesser et al. (1994; see also Singer et al., 1994) argued that
global information often becomes available to the reader through a more strategic
problem-solving process in which backgrounded global information is,reacti-
vated as a function of the search-after-meaning principle. That is, readers contin-
ually search both local and global information in an attempt to provide sufficient
RESOLVING COHERENCE BREAKS 169

reasons for currently read actions or events. The search for global explanations is
assumed to take place even when a text is locally coherent.
There is increasing evidence in support of the basic assumption of both views,
that readers routinely maintain coherence at both a local and global level (Al-
brecht & O'Brien, 1993; Myers, O'Brien, Albrecht, & Mason, 1994; O'Brien &
Albrecht, 1992). For example, O'Brien & Albrecht (1992) had participants read
texts describing a protagonist as being in a certain location (e.g., "Kim was
standing outside a health club"). Following several filler sentences designed to
background the location of the protagonist, a critical sentence described the
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015

protagonist as moving in a direction that was either consistent or inconsistent


with this location (e.g., "Kim decided to go outside and stretch her legs a little").
Despite the fact that the critical sentence was always locally coherent, reading
time increased when the location information in the critical sentence conflicted
with previously presented location information, suggesting that readers were
concerned with mapping the information at both a local and global level.
Albrecht and O'Brien (1993) had participants read passages that described a
specific characteristic of a protagonist (e.g., "Mary was a vegetarian" or "Mary
loves junk food"). After several backgrounding sentences, readers were pre-
sented with a critical sentence that was either consistent or inconsistent with
respect to the characteristic described earlier (e.g., "Mary ordered a cheeseburg-
er and fries"). Despite the fact that the critical sentence was always locally
coherent, reading times for the critical sentence were significantly longer when it
was inconsistent with the earlier described characteristic of the protagonist than
when it was consistent. Again, readers attempted to maintain global coherence
even when the critical sentence was locally coherent.
An important issue to examine is how readers resolve coherence breaks and
the impact they have on subsequent memory for a text. For example, O'Brien and
Myers (1985) found that when information presented in a text was "unexpected"
given the context presented earlier, memory improved. O'Brien and Myers sug-
gested that the unexpected information produced a coherence break that readers
attempted to resolve by reprocessing earlier portions of the text in an attempt to
integrate the unexpected information. When reprocessing was successful, memo-
ry improved. In contrast, there is evidence that if the coherence break is too
difficult to resolve, memory may suffer. For example, Myers, Shinjo, and Duffy
(1987) found that memory for text improved only if the coherence break was
moderately difficult to resolve. If integration required too much effort, memory
decreased (see also Duffy, Shinjo, & Myers, 1990; Keenan, Baillet, & Brown,
1984).
Assuming that readers are motivated to resolve coherence breaks, they have
two options: They can draw an inference that allows the conflicting information
to be integrated or they can edit or distort portions of the text to remove the
conflicting information. These two options lead to very different predictions
concerning comprehension and subsequent memory. For example, Albrecht and
170 HAKALA AND O'BRIEN

O'Brien (1993) found that resolution of a global coherence break led to an


improvement in memory. Because there were several sentences separating the
inconsistent portions of the text, the initial information concerning the target
characteristic of the protagonist was not immediately contradicted, thereby al-
lowing the reader to use this information to facilitate comprehension of subse-
quent actions described in the text. This would result in the target characteristic
being both well integrated and strongly encoded into the existing representation.
It would then be difficult to edit or distort that information in order to be able to
integrate a subsequent action that was inconsistent with it. Instead, readers drew
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015

inferences to facilitate the integration of the inconsistent information.


In contrast, Albrecht and O'Brien (1991) found that when local coherence
was violated, participants tended to edit or delete the portions of the passage that
contained the coherence break, leading to subsequent decrements in memory
performance. Because the conflicting pieces of information were both available
to the reader with little or no intervening material that needed to be checked, it
may have been easier to edit or delete portions of the text. Additionally, Black,
Turner, and Bower (1979) demonstrated that when a text consisting of two
sentences was incoherent, and the incoherence was unresolvable, readers dis-
torted one portion of the text to allow for integration. The distortions led to a
decrease in overall recall of text.
Thus, when contradictory information produces a global coherence break,
readers may attempt to reestablish coherence by elaborating or generating plausi-
ble inferences (e.g., Albrecht & O'Brien, 1993). In contrast, when a text con-
tains a contradiction that produces a local coherence break, readers can utilize
one of two strategies to resolve the inconsistency: First, they may simply inte-
grate the inconsistency into the representation. This would lead to an increase in
memory, as when global coherence is violated. Alternatively, participants may
either edit the representation by distorting contradictory information or they can
delete the conflicting portions of information so that what remains is coherent
(Albrecht & O'Brien, 1991; Black et al., 1979).
However, these assumptions may be unwarranted. Albrecht and O'Brien
(1991, 1993) used different types of materials to violate coherence. The different
recall results may have occurred because of this difference in materials rather
than due to violations of coherence at a local or global level. Albrecht and
O'Brien (1993) violated global coherence by having the protagonist engage in an
action that was unlikely given previous physical or personality characteristics.
This type of coherence break made it possible for readers to resolve the inconsis-
tency by generating plausible inferences. However, in Albrecht and O'Brien
(1991, see also Black et al., 1979), the local coherence breaks involved a direct
contradiction. Because there was no simple way to resolve the contradiction,
readers may have chosen to edit or distort the inconsistent information rather than
attempting to integrate it.
The current experiments used materials from Albrecht and O'Brien (1993) to
RESOLVING COHERENCE BREAKS 171

hold constant the nature of the inconsistency causing the coherence breaks.1 To
create coherence breaks at both a local and global level, the passages were
modified to vary the distance between the elaboration region and the critical
sentences, In the local coherence conditions, the described characteristics of the
protagonist and the critical sentences were separated by only one to three short
sentences. In the global coherence conditions, six intervening sentences sepa-
rated the described characteristics of the protagonist and the critical sentence.
Two example passages are presented in the Appendix.
Experiment 1 ensured that readers were sensitive to an inconsistency that
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015

occurred at either a local or global level. Experiment 2 then examined recall to


determine whether readers would attempt to integrate the inconsistent informa-
tion when the coherence break occurred at a global level and edited or distorted
the inconsistent information when the coherence break occurred at a local level.

EXPERIMENT 1
The goal of Experiment 1 was to ensure that readers would be sensitive to
coherence breaks at both a local and a global level using materials similar to
those developed by Albrecht and O'Brien (1993). Given our previous work with
these materials, we expected that readers would detect coherence breaks at the
global level, resulting in increased reading times in the inconsistent conditions
(Albrecht & O'Brien, 1993). Presumably, readers should also be sensitive to the
coherence breaks when they occur at a local level, and there should be a corre-
sponding increase in reading time in the inconsistent conditions of those passages
as well (cf. Albrecht & O'Brien, 1991).

Method
Participants. The participants were 60 University of New Hampshire under-
graduates from the Department of Psychology participant pool who received
course credit for their involvement.

Materials. The materials were 30 narrative passages. Eighteen passages were


from Albrecht and O'Brien (1993) and 12 additional passages were written
according to the same criteria used by Albrecht and O'Brien, Each passage was
divided into five sections: (a) an introduction, (b) an elaboration section, (c) filler
information, (d) two critical sentences, and (e) a conclusion. In the global coher-
ence condition, each passage began with two or three introductory sentences that
served to introduce the protagonist and the situation. Immediately following the

'The use of contradictions is only one way that incoherence could be introduced into text. For
example, anaphors without antecedents would also produce incoherence. However, because much of
the previous literature has used contradictions and because of the ease of producing incoherence with
this methodology, we have also adopted the procedure.
172 HAKALA AND O'BRIEN

introduction, one of three elaboration regions was presented: (a) consistent, (b)
inconsistent, or (c) neutral. The elaboration regions were composed of two to
five sentences with average word lengths of 45.3, 45.5, and 45.4 words for the
consistent, inconsistent, and neutral conditions, respectively. The consistent
elaboration region contained information that would support actions or charac-
teristics described later in the two critical sentences. The inconsistent elaboration
region contained information that would conflict with actions or characteristics
described later in the two critical sentences. Finally, the neutral elaboration did
not contain information that was related to the action or characteristics described
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015

later in the two critical sentences.


Following the elaboration, six sentences were presented that shifted the focus
away from the elaboration, but continued the story line developed in the intro-
ductory region. The filler material was included to ensure that local coherence
was maintained, while at the same time ensuring that information from the
elaboration region would no longer be active in memory. Following the filler
material, the two critical sentences were presented. The mean length of the
critical sentences were 38.5 and 38.5 characters, respectively. The second criti-
cal sentence always immediately followed the first critical sentence and was
included to detect possible carry-over effects from the first critical sentence.
Finally, the last section of each passage was the conclusion that was composed of
two to three sentences. Passages ranged from 20 to 25 lines. Each line was no
longer than 53 characters on the computer screen. In the local coherence condi-
tions, the six filler sentences were replaced by one to three sentences that contin-
ued the story line. Each passage was followed by a comprehension question. The
questions were designed to probe specific details of the passages but did not
address information concerning characteristics of the protagonist. Examples are
presented in the Appendix.
Six material sets were constructed: Each set contained five passages in each of
the six conditions. Across the six sets, each passage occurred once in each of the
six conditions.

Procedure. Participants were tested individually in an experimental session


that lasted approximately 45 min. All materials were presented on a video moni-
tor controlled by a Zenith Z-100 microcomputer. The experiment began with the
prompt Ready displayed on the monitor. Participants were instructed to press a
line-advance key that erased the word ready and presented the first line of a text.
Participants read this line and pressed the line-advance key to read the next line.
Comprehension time was measured as the time between key presses. At the end
of the passage, the word question appeared on the screen for 2,000 ms. This was
replaced by a row of question marks for 500 ms, after which the question
appeared. The question was phrased so the participant could respond with a yes
or a no answer. Participants were informed that this was the most important
aspect of the experiment and that they were to answer as quickly and accurately
RESOLVING COHERENCE BREAKS 173

as they could. On those trials in which participants made errors, the response was
followed by the word error for 750 ms. Three practice passages were presented
first to ensure that the participants fully understood the procedure.

Results and Discussion


Reading times for the critical sentences of each passage were recorded. Reading
times greater than three standard deviations from the mean were discarded. This
eliminated less than 1% of the data. In what follows, F t refers to tests against an
error term based on participant variability and F 2 refers to tests against an error
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015

term based on item variability. All analyses reported are significant at the .05
alpha level unless otherwise specified. All planned comparisons use the Bonfer-
roni procedure with a family-wise error rate equal to .05 and an error term based
on participant variability.
The reading times for the critical sentences are presented in Table 1. Separate
analyses of variance (ANOVAs) were conducted for first and second critical
sentences. Because the patterns for the first and second critical sentences did not
differ, they are presented together. Critical sentences were read more slowly
when they followed an inconsistent elaboration. This was true for both the first
critical sentence, F,(2, 108) = 36.2, MSE = 53347, F 2 (2, 48) = 16.2, MSE =
56454 and the second critical sentence, F,(2, 108) = 19.2, MSE = 66016, F 2 (2,
48) = 13.7, MSE = 45808.
Planned comparisons confirmed that for both the first and second critical
sentences in the global coherence conditions, reading times were longer follow-
ing an inconsistent elaboration than following a consistent elaboration, /(59) =
4.48 for the first critical sentence and r(59) = 3.37 for the second critical
sentence, or a neutral elaboration, f(59) = 3.43 for the first critical sentence and
/(59) = 4.06 for the second critical sentence. Reading times for critical sentences

TABLE 1
Experiment 1: Mean Reading Times for Critical Sentences 1 and 2
in both the Local and Global Conditions

Passage Condition

Critical Sentence Consistent Inconsistent Neutral

Global Coherence Condition


First 1,901 2,119 1,931
Second 1,807 1,973 1,769
M 1,854 2,046 1,853

Local Coherence Condition


First 1,818 2,085 1,898
Second 1,735 1,921 1,763
M 1,776 2,003 1,830
174 HAKALA AND O'BRIEN

following a consistent elaboration and a neutral elaboration did not differ (p >
.05).
In the local coherence conditions, reading times were also longer following
the inconsistent elaboration than following the consistent elaboration, f(59) =
5.79 for the first critical sentence and f(59) = 4.28 for the second critical
sentence, or the neutral elaboration, /(59) = 3.58 for the first critical sentence
and /(59) = 3.09 for the second critical sentence. Reading times for critical
sentences following a consistent elaboration and a neutral elaboration did not
differ (p > .05).
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015

To ensure that the slowdown in reading on the critical sentences in the incon-
sistent conditions was isolated to only those two sentences, we examined reading
times for the sentence that immediately preceded the critical sentences, as well as
for the sentence that immediately followed the critical sentences.2 Reading times
for the sentence preceding the critical sentences did not differ reliably across the
three elaboration conditions (p > .5). Additionally, there were no reliable differ-
ences in the reading times for the sentence that followed the critical sentences, (p
>.5).
Thus, Experiment 1 confirmed that with the current set of materials, readers
detected coherence breaks at both a local and a global level. Reading times for
both the first and second critical sentences were longest for the inconsistent
passages in both the local and global conditions. The fact that readers detected
the coherence break in the local coherence condition is not surprising; that
finding has been well established. The finding that readers detected the global
coherence break replicated the findings of Albrecht and O'Brien (1993) and adds
to a growing body of literature indicating that readers monitor global coherence
even when local coherence remains intact (for an excellent review of this litera-
ture, see Graesser et al., 1994).

EXPERIMENT 2
The goal of Experiment 2 was to examine the strategies that readers may use to
resolve coherence breaks that occur at either a local or a global level. Consistent
with earlier findings (Albrecht & O'Brien, 1993), global coherence breaks
should be resolved by readers integrating the inconsistent information into the
existing representation. This should result in increased recall for the regions of
the passage involved in the coherence breaks, namely, the elaboration region and
the critical sentences. For example, for the first passage in the Appendix, the
reader may recall that "even though Mary almost never ate junk food, for some
reason, she decided to order a cheeseburger and fries."
In contrast, when there is a local coherence break, we expect readers to distort

2
Because these sentences varied in length, we transformed the reading times into the average
number of milliseconds per characters and submitted those times to an ANOVA.
RESOLVING COHERENCE BREAKS 175

or delete portions of the text to force integration. For example, the reader may
recall that "Mary was not a vegetarian" or the reader may eliminate recall of the
elaboration region entirely and only recall that "Mary ordered a cheeseburger and
fries." If correct, recall in the inconsistent condition should not increase relative
to the consistent condition. Furthermore, readers should show a higher distortion
rate of information from the local inconsistent condition than in the global
inconsistent condition (e.g., Albrecht & O'Brien, 1991; Black et al., 1979).

Method
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015

Participants. The participants were 72 University of New Hampshire under-


graduates from the Department of Psychology participant pool who received
course credit for their involvement.

Materials. The materials were the same used in Experiment 1.

Procedure. The procedure was the same as Experiment 1, with the following
exception. Participants were not told about the recall phase, but were told that the
experiment consisted of two parts. After completing the reading phase of the
experiment, participants were given a recall booklet. Each page of the booklet
provided a recall cue for a particular passage. The cue for each passage was a
phrase or paraphrase from the introduction of the passage. For example, the
recall cue for the first passage in the Appendix was, "Mary meeting her friend for
lunch." The passages were recalled in the same order that they were read.
Participants were instructed to recall all they could about each passage; once they
finished recall for a given passage and turned the page, they were asked not to
return to the previous pages.

Results
Reading times for the critical sentences of each passage were recorded. The same
cutoffs were used as in Experiment 1, which eliminated less than 3% of the data.

Reading Time. Reading times for both critical sentences in the local and
global conditions are presented in Table 2. As in Experiment 1, critical sentences
were read more slowly when they followed the inconsistent elaborations. This
was true for both the first critical sentence, F,(2, 132) = 36.8, MSE = 89210,
F 2 (2, 48) = 24.3, MSE = 56779 and the second critical sentence, Fx{2, 132) =
29.7, MSE = 65735, F2(2, 48) = 21.5, MSE = 32486.
Planned comparisons confirmed the same pattern of results that were found in
Experiment 1. For both the first and second critical sentences in the global
coherence conditions, reading times were longer following the inconsistent elab-
oration than following the consistent elaboration, f(71) = 4.97 for the first
critical sentence and r(71) = 5.16 for the second critical sentence, or a neutral
176 HAKALA AND O'BRIEN

TABLE 2
Experiment 2: Mean Reading Times for Critical Sentences 1 and 2
in both the Local and Global Conditions

Passage Condition

Critical Sentence Consistent Inconsistent Neutral

Global Coherence Condition


First 1,879 2,141 1,900
Second 1,770 1,968 1,784
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015

M 1,824 2,054 1,842

Local Coherence Condition


First 1,874 2,143 1,865
Second 1,788 2,008 1,793
M 1,831 2,075 1,829

elaboration t(7l) = 4.52 for the first cntical sentence and f(71) = 3.92 for the
second critical sentence. Reading times for critical sentences following a consis-
tent elaboration and a neutral elaboration did not differ (p > .05).
In the local coherence condition, reading times were also longer following the
inconsistent elaboration than following the consistent elaboration ?(71) = 5.23
for the first critical sentence and ?(71) = 5.11 for the second critical sentence, or
the neutral elaboration, t(7l) = 5.29 for the first critical sentence and /(71) =
3.58 for the second critical sentence. Reading times for critical sentences follow-
ing a consistent elaboration and a neutral elaboration did not differ (p > .05).
As in Experiment 1, we examined reading times for the sentence that pre-
ceded and followed the critical sentences to ensure that reading times were not
generally slower in the inconsistent conditions. Reading times for these sen-
tences did not differ across conditions. This was true for both the sentence that
preceded the critical sentences (p > .5) and for the sentence that followed the
critical sentences (p > .5).

Recall. Each passage was broken down into idea units (e.g., Trabasso &
Sperry, 1985). A lenient scoring criterion was employed; participants were given
credit for an idea unit if their recall captured the meaning of the respective idea
unit. Two independent judges scored all the recall, and interrater reliability was
.95. Discrepancies were resolved by a third independent judge. The idea units for
each passage were grouped into the following regions: introduction, elaboration,
filler,3 first critical sentence, second critical sentence, and close. Because each

3
To create the global incoherence conditions, we utilized six filler sentences to separate the
elaboration region and the inconsistent critical sentences. In the local coherence conditions, these
sentences were removed. Thus, there was no measure of recall for filler material in the local
coherence condition.
RESOLVING COHERENCE BREAKS 177

passage was composed of a different number of idea units, all analyses were
conducted on proportion of idea units recalled. Table 3 presents the proportion of
idea units recalled for all passages as a function of condition and region in the
passage. Separate ANOVAs were performed on each region of the passage.
For both the local and global coherence conditions, there were no reliable
differences in the proportion of idea units recalled from the introductory portion
of the passage, the filler sentences (global condition only), the second critical
sentence, or the closing sentences (p > . 13 in all cases). The only regions of the
passage showing differences in recall were the elaboration region and the first
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015

critical sentence. First, consider these two regions for the passages in the global
coherence conditions. There was significant variability in the proportion of recall
of the elaboration regions, F,(2, 132) = 26.3, MSE = .008; recall was greater in
the inconsistent condition than in either the consistent, f(71) = 4.39, or the
neutral condition, f(71) = 7.01. Also, recall in the consistent condition was
higher than in the neutral, t(l\) = 2.56. There was also significant variability in
recall of the first critical sentence in the global coherence condition, F,(2, 132)
= 13.3, MSE = .036; recall of the first critical sentence was greater following
the inconsistent elaboration than following the consistent, t(7\) = 3.66, or
neutral condition, ?(71) = 4.83. There was no reliable difference in recall be-
tween the consistent and neutral conditions, f(71) = .185.
Next, consider the pattern of recall for the elaboration region and the first
critical sentence in the local coherence condition. Again, there was significant
variability in recall of elaboration regions, F,(2, 132) = 17.6, MSE = .009;
however, planned comparisons revealed that there was no reliable difference in

TABLE 3
Experiment 2: Mean Proportion of Idea Units Recalled in Global and Local Passages
as a Function of Condition and Location in the Passage

Passage Condition

Location in the Passage Consistent Inconsistent Neutral

Global Coherence Condition


Introduction .22 .23 .25
Elaboration .25 .32 .21
Filler .23 .23 .23
Critical Sentence 1 .47 .61 .46
Critical Sentence 2 .23 .23 .25
Close .21 .19 .21

Local Coherence Condition


Introduction .26 .23 .26
Elaboration .29 .29 .21
Critical Sentence 1 .51 .58 .44
Critical Sentence 2 .19 .21 .25
Close .22 .22 .20
178 HAKALA AND O'BRIEN

the proportion of idea units recalled between the consistent and inconsistent
conditions, /(71) = .336. Proportion of idea units recalled in the neutral condi-
tion showed a significant decline relative to both the consistent, f(71) = 4.70,
and inconsistent conditions, f(71) = 4.70. Recall for the first critical sentence in
the local coherence conditions also varied, F,(2, 132) = 10.9, MSE = .037; the
proportion of first critical sentences recalled was greater following the inconsis-
tent elaboration than following the consistent, f(71) = 2.01, or neutral condition,
?(71) = 4.67. The first critical sentence was also better recalled in the consistent
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015

condition than in the neutral condition, f(71) = 2.03.


Finally, an analysis of the number of distortions was conducted. A response
was scored as a distortion if it contradicted the information provided in the
elaboration region or the critical sentence. Distortions took one of two general
forms: Explicitly stated information was changed either through negation or
substitution. Examples of distortions from the passages presented in the Appen-
dix that involved negation included changing "Mary was a vegetarian" to "Mary
was not a vegetarian" or changing "Peggy's daughter . . . was barely able
to . . . crawl from place to place" to "Peggy's daughter . . . was able to
. . . crawl from place to place." Examples of substitution included changing
"Mary ordered a cheeseburger and fries" to "Mary decided to get cauliflower" or
changing "Janie stood up and walked to her mom" to "Peggy brought the phone
over to her daughter." Participants generated significantly more distortions in the
local inconsistent condition than in the global inconsistent condition (14% vs.
3%), f(71) = 4.04. Participants showed no distortions in the consistent or neutral
conditions.

Discussion
The reading time results of Experiment 2 replicated Experiment 1; reading times
were longer for passages in the inconsistent condition than for passages in either
the consistent or neutral conditions. This was true in both the local and global
coherence conditions. However, the result of primary interest involved recall for
the passages following either a local or global coherence break. For passages in
the global coherence condition, recall for both the first critical sentence and the
elaboration region were significantly higher in the inconsistent condition than in
either the consistent or neutral conditions. This finding suggests that when read-
ing the inconsistent passages, participants attempted to reprocess the elaboration
portion of the passage in order to integrate the critical sentences and reestablish
coherence.
In contrast, in the local conditions there was no overall difference in the level
of recall for the elaboration regions in the consistent and inconsistent passages.
This-lack of a difference occurred despite the fact that recall of the first critical
sentence was greater in the inconsistent passages. A more detailed analysis of the
recall patterns revealed that participants were still attempting to reprocess the
elaboration portion of the inconsistent passages in order to integrate the critical
RESOLVING COHERENCE BREAKS 179

sentences. However, instead of generating an elaborative inference, participants


tended to distort information in the elaboration region in order to make the
critical sentence consistent with the elaboration. This was confirmed by the
significantly greater distortion rate in the inconsistent condition.
Participants also recalled more from the elaboration regions in the consistent
passages than in the neutral passages in both the local and global conditions. This
result fits well within a resonance framework. Because the elaboration in the
consistent passage was contextually relevant to the critical sentences, those sen-
tences should reactivate and provide additional contact with the elaboration
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015

region, resulting in additional processing of that information. In contrast, the


critical sentence in the neutral condition had little to do with the elaboration
region in those conditions and as a result, would not prompt any additional
processing of that information.
The second critical sentence also took longer to read following an inconsistent
elaboration in both the local and global conditions. However, recall of the second
critical sentence was unaffected by this increase in reading times. This increase
in reading time is probably the result of spillover effects from the first critical
sentence; when readers encountered the second critical sentence, the inconsisten-
cy from the first critical sentence had not yet been resolved. Because the extra
processing time reflected continual attempts to integrate the first critical sen-
tence, there was no benefit for recall of the second critical sentence.

GENERAL DISCUSSION
The current set of results add to a growing number of studies demonstrating that
readers are simultaneously concerned with the maintenance of both local and
global coherence (e.g., Albrecht & O'Brien, 1993; Garrod & Sanford, 1988;
Glenberg & Langston, 1992; O'Brien & Albrecht, 1992). In both Experiments 1
and 2, participants experienced comprehension difficulty when they read critical
sentences that were inconsistent with earlier described characteristics of the
protagonist. This difficulty occurred when the critical sentences produced a local
coherence break; it also occurred when the critical sentences produced a global
coherence break despite the fact that the passages were locally coherent.
The recall results from Experiment 2 indicated that the resolution of these
coherence breaks and their impact on memory depended on whether the coher-
ence break occurred at a local or global level. When global coherence was
violated, memory for the elaboration region and the first critical sentence im-
proved, replicating previous work by Albrecht and O'Brien (1993). Presumably,
readers reactivated and reprocessed the contextually relevant information in the
elaboration region and generated the inferences necessary to integrate the critical
sentences with the elaboration region. In contrast, when local coherence was
violated, memory for the elaboration region did not improve. An examination of
the recall protocols indicated that rather than reprocessing and integrating the
180 HAKALA AND O'BRIEN

inconsistent information, readers tended to distort information from the elabora-


tion region in order to reestablish coherence. This was confirmed by the signifi-
cantly higher distortion rates for the inconsistent passages. This result is
consistent with Black et al. (1979) who found that readers distorted location
information in sentences to ensure that it was consistent with previously pre-
sented location information (see also Albrecht & O'Brien, 1991).
Why would readers utilize one strategy to resolve global coherence breaks and
another to resolve local coherence breaks? In the global coherence conditions,
there are two reasons why a reader may have chosen to generate an inference to
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015

reestablish coherence rather than attempt to edit or distort portions of the text.
First, there was a significant amount of material separating the inconsistent
portions of the text; as a result, the target characteristic of the protagonist was not
immediately contradicted. This information was therefore free to be used by the
reader to explain future actions. This would result in the target characteristic
being both well integrated and strongly encoded into the existing representation,
making it increasingly difficult to edit or distort that characteristic in order to
integrate a subsequent action which was inconsistent with it. Second, distorting
or editing the representation to facilitate integration would require considerable
effort; all the information in the elaboration region as well as the entire set of
intervening material would have to be checked to ensure that it was consistent
with any altered structure.
A more reasonable strategy would be to generate an inference or elaboration
to facilitate integration. For example, a typical elaboration from the recall proto-
cols for the first passage in the Appendix was, "Mary ordered a cheeseburger and
fries even though she was a vegetarian and didn't eat anything cooked in grease."
For the second passage in the Appendix, a typical elaboration was, "Janie had
only been crawling, but she must have learned to walk because while Peggy was
sitting down in a chair in the kitchen, her daughter walked over and pulled at her
pant leg." The extra inferential processing accorded to the elaboration region in
this condition presumably resulted in additional retrieval routes rendering it more
easily accessible at recall (Albrecht & O'Brien, 1993; Myers et al., 1987;
O'Brien & Myers, 1985). Because the inconsistency was only between the
critical sentence and the elaboration region, any additional inferencing or elab-
any benefit in recall should have been restricted to those two regions as well: the
result obtained.
In contrast, when local coherence is violated, all information that needs to be
checked is currently active in memory. Under these conditions, distorting one
fact is an efficient and effective strategy for reestablishing coherence. For exam-
ple, a typical distortion from the recall protocols for the first passage in the
Appendix was, "Mary really loved junk food" and for the second passage in the
Appendix a typical distortion was, "Janie crawled over to her mom." This is
further supported by the finding that the increase in reading times following a
local coherence break was similar to the increase following a global coherence
RESOLVING COHERENCE BREAKS 181

break. However, unlike the global inconsistent condition, recall of the elabora-
tion region in the local inconsistent condition did not improve; instead, partici-
pants generated significantly more distortions.
The inconsistencies used in this set of experiments did not involve a direct
contradiction. Although it was unlikely that the protagonist would engage in
actions described in the critical sentences, it was not impossible or unreasonable.
When an inconsistency is encountered that is resolvable, readers have the option
to select a strategy to reestablish coherence. However, if an inconsistency in-
volves a direct contradiction, reestablishing coherence through elaboration may
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015

not be possible. Readers may only be able to distort or delete the inconsistent
information, or they may not be able to engage in any successful strategy.
Presumably, this is what occurs in the studies in which comprehension difficulty
is high and subsequent recall is poor (e.g., Duffy et al., 1990; Keenan et al.,
1984; Myers et al., 1987).
Albrecht and O'Brien (1993) demonstrated that global coherence breaks re-
sulted in increased reading time as well as increased recall for conflicting por-
tions of text. The current research has extended this finding and demonstrated
that readers select different strategies for reestablishing coherence, depending on
the level of coherence that was violated. By further exploring the conditions
under which participants select different strategies and the factors that influence
those strategies, we may be able to predict more precisely the conditions under
which coherence breaks may actually improve memory (e.g., Albrecht &
O'Brien, 1993; O'Brien & Albrecht, 1992; O'Brien & Myers, 1985) and the
conditions under which coherence breaks may disrupt memory (e.g., Keenan et
al., 1984; Myers etal., 1987).

REFERENCES
Albrecht, J.E., & O'Brien, E.J. (1991, November). Comprehension strategies in the development of
a mental model. Poster session presented at the 32nd annual meeting of the Psychonomic
Society, San Francisco, CA.
Albrecht, J.E., & O'Brien, E.J. (1993). Updating a mental model: Maintaining both local and global
coherence. Journal of Experimental Psychology: Learning, Memory and Cognition, 19,
1061-1070.
Black, J.B., Turner, E., & Bower, G.H. (1979). Point of view in narrative comprehension memory
and production Journal of Verbal Learning and Verbal Behavior, 18, 187-198.
Duffy, S.A., Shinjo, M., & Myers, J.L. (1990). The effect of encoding task on memory for sentence
pairs varying in causal relatedness. Journal of Memory and Language, 27, 235-244.
Fletcher, C.R. (1981). Short-term memory processes in text comprehension. Journal of Verbal
Learning and Verbal Behavior, 20, 564-574.
Fletcher, C.R. (1986). Strategies for the allocation of short-term memory during comprehension.
Journal of Memory and Language, 25, 43-58.
Fletcher, C.R., & Bloom, C. (1988). Causal reasoning in the memory in the comprehension of
simple narrative texts. Journal of Memory and Language, 27, 235-244.
Garrod, S.C., & Sanford, A.J. (1988). Thematic subjecthood and cognitive constraints on discourse
structure. Journal of Pragmatics, 12, 519-534.
182 HAKALA AND O'BRIEN

Gillund, G., & Shiffrin, R.M. (1984). A retrieval model for both recognition and recall. Psychologi-
cal Review, 97, 1-67.
Glenberg, A.M., & Langston, W.E. (1992). Comprehension of illustrated text: Pictures help to build
mental models. Journal of Memory and Language, 31, 129-151.
Glenberg, A.M., Meyer, M., & Lindem, K. (1987). Mental models contribute to foregrounding
during text comprehension. Journal of Memory and Language, 26, 69-83.
Graesser, A.C., Singer, M., & Trabasso, T. (1994). Constructing inferences during narrative text
comprehension. Psychological Review, 101, 371-395.
Hintzman, D.L. (1986). "Schema abstraction" in a multiple-trace model. Psychological Review, 93,
411-428.
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015

Johnson-Laird, P.N. (1983). Mental models. Cambridge, MA: Harvard University Press.
Just, M.A., & Carpenter, P.A. (1987). The psychology of reading and language comprehension.
Newton, MA: Allyn & Bacon.
Keenan, J.M., Baillet, S.D., & Brown, P. (1984). The effects of causal cohesion on comprehension
and memory. Journal of Verbal Learning and Verbal Behavior, 23, 115-126.
Kintsch, W., & van Dijk, T.A. (1987). Toward a model of text comprehension and production.
Psychological Review, 85, 363-394.
McKoon, G., & Ratcliff, R. (1992). Inference during reading. Psychological Review, 99, 440-
466.
Myers, J.L., O'Brien, E.J., Albrecht, J.E., & Mason, R.A. (1994). Maintaining global coherence
during reading. Journal of Experimental Psychology: Learning, Memory and Cognition, 20,
876-886.
Myers, J.L., Shinjo, M., & Duffy, S.A. (1987). Degree of causal relatedness and memory. Journal of
Memory and Language, 26, 453-465.
O'Brien, E.J. (1995). Automatic components of discourse comprehension. In R.F. Lorch & E.J.
O'Brien (Eds.), Sources of coherence in reading (pp. 159-176). Hillsdale, NJ: Erl-
baum.
O'Brien, E.J., & Albrecht, J.E. (1992). Comprehension strategies in the development of a mental
model. Journal of Experimental Psychology: Learning, Memory and Cognition, 18, 777—
784.
O'Brien, E.J., & Myers, J.L. (1985). When comprehension difficulty improves memory for text.
Journal of Experimental Psychology: Learning, Memory and Cognition, 11, 12-21.
Ratcliff, R. (1978). A theory of memory retrieval. Psychological Review, 85, 59-108.
Singer, M., Graesser, A.C., & Trabasso, T. (1994). Minimal or global inference during reading.
Journal of Memory and Language, 33, 421-441.
Trabasso, T., & Sperry, L. (1985). Causal relatedness and importance of story events. Journal of
Memory and Language, 24, 595-611.
van Dijk, T., & Kintsch, W. (1983). Strategies of discourse comprehension. New York: Academic.

APPENDIX
EXAMPLES OF THE PASSAGES USED
IN EXPERIMENTS 1 AND 2
Experiment 1

Introduction
Today, Mary was meeting a friend for lunch. She arrived early at the restaurant
and decided to get a table. After she sat down, she started looking at the menu.
RESOLVING COHERENCE BREAKS 183

Consistent Elaboration
This was Mary's favorite restaurant because it had fantastic junk food. Mary
enjoyed eating anything that was quick and easy to fix. In fact, she ate at
McDonald's at least three times a week. Mary never worried about her diet and
saw no reason to eat nutritious foods.

Inconsistent Elaboration
This was Mary's favorite restaurant because it had fantastic health food. Mary, a
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015

health nut, has been a strict vegetarian for ten years. Her favorite food was
cauliflower. Mary was so serious about her diet which she refused to eat anything
that was fried or cooked in grease.

Neutral Elaboration
This was Mary's favorite restaurant because it had a nice quiet atmosphere.
Mary frequently ate at the restaurant and had recommended it to all of her
friends. She especially liked the cute tables and the country style table cloths on
them. It made her feel right at home.

Global Filler
After about ten minutes, Mary's friend arrived. It had been a few months since
they had seen each other. Because of this they had a lot to talk about and chatted
for over a half hour. Finally, Mary signaled the waiter to come take their orders.
Mary checked the menu one more time. She had a hard time deciding what to
have for lunch.

Local Filler
After about ten minutes, Mary's friend arrived. Mary signaled the waiter to come
take their orders. Mary had a hard time deciding what to have for lunch.

Critical Sentence 1
Mary ordered a cheeseburger and fries.

Critical Sentence 2
She handed the menu back to the waiter.

Close
Her friend didn't have as much trouble deciding what she wanted. She ordered
and they began to chat again. They didn't realize there was so much for them to
catch up on.

Comprehension Question
Was Mary meeting her husband for lunch?
184 HAKALA AND O'BRIEN

Experiment 2

Introduction
Peggy had gained 40 pounds during her pregnancy. Recently, she started dieting
and exercising daily in order to regain her youthful figure. She wanted to fit into
her old clothes.

Consistent Elaboration
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015

Peggy's daughter, Janie, who was now two years old and could walk, loved to
run around while Peggy did her sit-ups. In fact, Janie use to follow her mommy
everywhere. This made life easier for Peggy because she always knew Janie was
right behind her.

Inconsistent Elaboration
Peggy's daughter, Janie, who was five months old, was barely able to pull herself
up and still crawled from place to place. Janie always stayed on a mat while
Peggy was exercising. This made life easier for Peggy because she always knew
where Janie was.

Neutral Elaboration
Janie, Peggy's daughter, always sat on the floor and watched her mother do her
sit-ups. Janie thought her mom looked funny as she puffed and groaned. Usually,
Janie chuckled and Peggy would begin to laugh as well. Then Peggy would lean
over and tickle Janie.

Global Filler
One day while exercising, the telephone began to ring. Quickly, Peggy got up to
answer it. First, she walked over and turned off the exercise video. After Peggy
turned it off, she walked into the kitchen to pick up the phone. Janie's eyes
followed her mommy over to the telephone in the kitchen. Peggy picked up the
phone and sat down on a chair.

Local Filler
One day while exercising, the telephone began to ring. Peggy picked up the
phone and sat down on a chair.

Critical Sentence 1
Janie stood up and walked to her mom.

Critical Sentence 2
She tugged on her mommy's pants leg.
RESOLVING COHERENCE BREAKS 185

Close
Peggy bent over and placed the phone next to Janie's ear. Her dad was on the line
and gave her a cheery hello. She smiled when she heard his voice.

Comprehension Question
Was Janie's uncle on the phone?
Downloaded by [University of Oklahoma Libraries] at 03:14 23 January 2015

You might also like