Full text of DOI:10.3758/BF03196433

Memory & Cognition
2002, 30 (5), 779-788
Readers’ reality-driven and plot-driven analyses in
narrative comprehension
DAVID N. RAPP
Tufts University, Medford, Massachusetts
and
RICHARD J. GERRIG
State University of New York, Stony Brook, New York
We suggest that when readers experience narratives,their expectations about the likelihood of narra-
tive events are informed by two types of analyses. Reality-driven analyses incorporate real-world con-
straints involving, for example, time and space; plot-driven analyses incorporate concerns about out-
comes that emerge from the plot. We explored the interaction of these two types of analyses in the
application of temporal situation models. Participants read stories in which the final episode occurred
after a minute time shift (i.e., “A minute later . . .”) or hour time shift (i.e., “An hour later .. .”). Our exper-
iments assessedparticipants’judgments and reading times for statements describing the state of events
(e.g.,the possibilitythat characterscould carryout particularbehaviors)following eachtype of time shift.
Experiments 1A and 1B demonstrated that readers are appropriately sensitive to the real concomitants
of time shifts. Experiments 2A and 2B demonstrated, even so, that plot-driven preferences modify judg-
ments and reading times away from reality-drivenexpectations.Our resultshave implications for the role
of the reader in theories of narrative comprehension.
Thunderball
In the novel
(Fleming, 1961), James Bond of the scene supports a second, more positive outcomefor
finds himself at the mercy of an underwater assassin. Time Bond. Given readers’ fervent desire that Bond prevail over
is running out:
his (and the world’s) enemies, they may never give up hope
that Bond will, somehow, devise a way to escape the spear
gun.
Again the butt of the [spear]gun crasheddown on his head.
Now the water was full of black smoke, heavy, stringystuff
that clung to the glass of his mask. At last it cleared . . . the
gun was coming up again, slowly, agonizingly, as if it
weighed a ton, and the bright sting of the spear showed at
its mouth . . . . Bond couldn’t get his limbs to obey. They felt
like lead. He shook his head to clear it, but still his hands
and flippers moved only half consciously, all speed gone.
Now he could see the bared teeth around the other man’s
rubber mouthpiece. The gun was at his head, at his throat,
at his heart. Bond’s hands crept up his chest to protect him
while his flippers moved sluggishly, like broken wings,
below him. (p. 130)
Thunderball
This moment from
illustrates how the dif-
ferent analyses readers can undertake at the same narra-
tive juncture can lead them to anticipate opposite out-
analysis
comes. We use theterm
to suggestthatreaders are,
in some fashion, taking stock of the state of a narrative
world at a particular moment. Readers’ reality-driven
analyses are guided by appropriatelygeneral expectations
about properties of the real world, such as ordinary con-
Thun-
straints of space, time, and human behavior. In the
derball
example, readers should have preexisting notions
abouthow fast Bondcan move relativeto a spear fired from
a gun. In the real world, a human being could not hope to
outswim such a weapon. By contrast, readers’ plot-driven
analysesneed not be constrainedby physicalor behavioral
reality. Authors often create circumstances in which read-
ers become strongly committed to particularnarrative out-
comes. Readers’ plot-driven analyses will yield, as their
products, conceptualizations of what readers hope or pre-
fer will happenat a particularnarrativemoment.In thecase
Readers’ expectations for the outcome of this scene may
be informed by two sets of constraints. On the one hand, a
reality-driven analysis
would suggest that, given Bond’s
current physical condition,he will be unable to outmaneu-
plot-driven analysis
ver the gunman’s spear. However, a
This material is based on work supported by National Science Foun-
dation Grants IRI-9711974, ITR0082602, and IIS-9817484. We thank
Susan Brennan, Mark Aronoff, and Brenda Anderson for their insight-
ful advice during the development of this project. We also thank Holly
Taylor, Sid Horton, Emily Durbin, Rolf Zwaan, and two anonymous re-
viewers for helpful comments on the manuscript. Please address corre-
spondence to D. N. Rapp, Department of Psychology, 490 Boston Av-
enue, Tufts University, Medford, MA 02155 (e-mail: drapp@perseus.
tufts.edu).
Thunderball
, readers may hope that Bond will escape—
of
no matter how dire the “reality” of his situation.This plot-
driven analysis may partially be informed by expectations
concerning Bond’s propensity for escaping from danger-
ous situations.However, as we will describe in this article,
a plot-driven analysis need not be informed by past expe-
779
Copyright 2002 Psychonomic Society, Inc.

780
RAPP AND GERRIG
rience with specific characters. The goal of this article is
to demonstrate how reality- and plot-driven analyses in- reality-drivenanalyses facilitate the construction and ap-
teract during readers’ narrative experiences. plication of situation models for ordinary features of the
Results of this type provide an indication as to how
Many theories of text comprehension have taken as world, such as the normal passage of time. However, re-
their starting point readers’ reality-driven analyses of searchers have often neglected plot-driven analyses of
texts. These theories suggest that readers construct men- texts in their theories of situation models (Gerrig, 1993).
situation models
, that encode This omissionis somewhat problematic.Situationmodels
tal representations, termed
the normal passage of time, the normal organization of are supposed to serve as representations of “the state of
space, and normal expectationsfor how charactersand ob- affairs described in a text” (Zwaan & Radvansky, 1998,
jects interact. Situation models facilitate the comprehen- p. 162). Presumably, the products of what we are calling
sion of text, allowing readers to form inferences (Glen- plot-driven analyses would be an important component
berg & Mathew, 1992;Glenberg, Meyer, & Lindem, 1987; of that “state of affairs.” For example, the event-indexing
Perrig & Kintsch, 1985; van Dijk & Kintsch, 1983). The- model suggests that readers index characters’ intentions
orists have attempted to outline the processes involved in (Zwaan, Langston, & Graesser, 1995; Zwaan, Magliano,
constructing multidimensional situation models. One & Graesser, 1995). We suggest that readers encode not
event-
model that has gained empirical support is the
only characters’ intentions,but also their own preferences
indexingmodel
Thunderball
ex-
(Taylor& Tversky, 1997;Zwaan, Langston, with respect to those intentions. In the
& Graesser, 1995; Zwaan, Magliano, & Graesser, 1995). cerpt, the fellow with the spear gun very much intends to
According to this model, readers encode and retrieve in- kill Bond. However, we suspect that the modal reader
spatial temporal causal intentional
, and hopes that the assassin will fail.
dices for the
protagonist object
,
,
,
–
relations in a text. These indices con- In our view, readers may undertake plot-driven analy-
form to reality-drivenexpectationsaboutthe natural phys- ses any time a narrative has a plot that is sufficiently rich
ical and social limits of characters and events. to engage readers’ interest in particularoutcomes. Read-
Research on situation models has confirmed the view ers are at liberty, at any time, to take stock of the distance
that readers’ narrative experiences are structured through between the current state of a narrative and an outcome
appropriate expectations about reality. Consider studies they desire. Authors have availablea range of techniques
that have examinedhow readers encode informationabout to prompt reader involvement of this type. Authors may,
thepassage of timein a narrative(Graesser, Kassler, Kreuz, for example, create empathy for particular characters, so
& McLain-Allen, 1998; Magliano, Miller, & Zwaan, that readers wish for the characters to prevail. Authors
2001; Radvansky, Zwaan, Federico, & Franklin, 1998; may create suspense so that readers are prompted to con-
Rich & Taylor, 2000; Rinck, Hähnel, & Becker, 2001; template a range of possible outcomes. Authors may dis-
Zwaan, 1996; Zwaan, Madden, & Whitten, 2000). In one rupt the timeline of narrated events, through flashbacks
project, for example, participants read texts containing or flashforwards, so that readers maintain representations
a moment later an hour later
a day
) (Zwaan, 1996). Consider the following story:
time shifts (e.g.,
later
,
, or
from multipleperspectives.Each of these author practices
is likely to lead readers to engage different types of pro-
plot-driven
cesses or representations. We label them all
Today was the grand openingof Maurice’s new art gallery.
He had invited everybody in town, who was important in
the arts. Everyonewho had been invited,had said that they
would come. It seemed like the opening would be a big
success.At seven o’clock, the first guestsarrived.Maurice
was in an excellent mood. He was shaking hands and
analyses
, because we locate the origin of the processes
and representationsin readers’ strategic contemplationof
the possibilities of the plot.
Previous research has supported this suggestion that
readers’ plot-drivenresponses have an impact on text pro-
cessing and representation (Allbritton & Gerrig, 1991;
Prentice, Gerrig, & Bailis, 1997). For example, one set of
experiments demonstrated effects of readers’ outcome
preferences. Participantsread stories that began with out-
come statements such as “Sue was elated to hear her
name as the winning entry was announced”(Allbritton &
. A moment/an hour/a day later, he turned very
beaming
pale
. He had completely forgotten to invite the local art
critic. And sure enough, the opening was very negatively
reviewed in the weekend edition of the local newspaper.
Maurice decided to take some Advil and stay in bed the
whole day. (p. 1207)
After reading a version of the text, readers were asked Gerrig, 1991, p. 620). The texts continuedwith story ma-
beaming
to indicate whether a particularword (e.g.,
) had terial that created reader preferences with respect to such
appeared in the story. Their recognition latencies in- outcomes: e.g., “Sue didn’t really care about the award,
creased when the story included longer versus shorter but she did want the money. She plagiarizedan essay from
time shifts. Participantsalso took longer to read texts and a source no one would recognize and turned it in.” At the
answer questions from stories containing longer time end of each story, the participantswere asked to verify the
shifts. These effects of time shifts follow from reality- story’s actual outcome. The preferences that the stories
driven analyses. Specifically, time shifts of longer dura- imparted had a consistent effect on the time course with
tion usually move story events out of a particular tem- which readers were able to verify the outcomes: When
poral setting; readers shift focus to new scenarios. This readers’ preferences were nonnormative (i.e., they were
shift in reader focus appears to reduce the accessibility encouraged to construct a preference for a negative out-
of information from the earlier episode.
come), they took more time to verify actual story out-

REALITY- AND PLOT-DRIVEN ANALYSES
781
comes. These results demonstrate that what readers wish
to have happen has an impact on their representations of
what actually transpired in a text.
For the present experiments,we wished to make a more
direct contrast between the exigenciesof reality and plot.
We took as our domain readers’ assessments of the likely
concomitantsof the passage of time. We begin by demon-
strating the impact of reality-driven analyses in this do-
main. Consider the following story:
6. Billy looked at his perfect condition card.
7. He really wanted to get that autograph.
8. McGwire appeared and said hello to the fans.
9. A minute later he got into his limo and rode off.
In this version of the story, we intendedreaders to sympa-
thize with Billy’s plight.Accordingly, we intendedreaders
to encode the preference that Billy manage to get Mc-
Gwire’s autograph. In the contrasting case, we intended
readers to encode the preference that Billy not get the
autograph:
1. Billy was waiting outside the St. Louis Cardinals’ sta-
dium, with a baseball card for Mark McGwire to sign.
2. Billy stood in a large group of fans waiting to meet the
future hall of fame player.
2. Billy was a dishonest collectibles dealer who would
overchargechildrenwhen they boughtsports cards from
him.
3. McGwire’s fans were holding all types of baseball
memorabilia.
4. One child was holding a baseball.
5. Another adult was holdinga large framed photographof
McGwire.
6. Billy looked at his perfect condition card.
7. He really wanted to get that autograph.
8. McGwire appeared and said hello to the fans.
9. A minute later he got into his limo and rode off.
How do these preferences affect readers’ judgmentsabout
the likelihoodof outcomes?We proposethat readers’ pref-
erences can affect their mental simulations.In mental con-
templation, the activities that ought to take more than a
minute or less than an hour (according to a reality-driven
analysis) can be contracted or lengthened,or made less or
more precise, to fit the readers’ plot-driven needs.
Our research had two phases. In the first phase, we
After reading this brief text, what are readers likely to sought to demonstrate that readers encode and retrieve
think about the likelihood that Billy obtained an auto- temporal informationfrom situationmodels using reality-
graph? Presumably, readers are able to do some mental driven analysesof texts. Experiment1A demonstratedthat
simulation (cf. Kahneman & Tversky, 1982; see also readers make accurate judgments about the concomitants
Barsalou, 1999, for a recent discussion of mental simula- of the passage of time. Experiment 1B suggested that the
tion) to provide at least a tentative answer to this ques- match of story outcomes to readers’ expectations had a
tion. They are unlikelyto imagine that Billy could acquire consistent effect on participants’ reading times. We de-
McGwire’s autograph in a minute. Rather, they are likely signed the second phase of the research to demonstratethe
to expect that Billy would be unable to get the autograph, impact of readers’ plot-driven analyses. To do this, we al-
given the large crowd of fans and McGwire’s quick exit. tered our stories to create strong reader preferences for
Suppose, however, that Sentence 9 read as follows:
particularoutcomes. In Experiment 2A, those preferences
had a consistentimpacton readers’ judgmentsof what were
and were not likely outcomes. Experiment 2B demon-
strated a similar effect of preferences on the time course
of reading. Taken together, this set of experimentsdemon-
strates that readers are fully able to attend to the reality of
the passage of time. However, that reality is malleable
with respect to the imperatives of plot.
9. An hour later he got into his limo and rode off.
In this case, readers shouldbe better able to simulate men-
tally a course of events that would allow Billy to get the
autograph. This mental simulation is what we mean, in
this case, by a reality-driven analysis. Readers, presum-
ably, are able to use their knowledge of the real-life dura-
tions of activities to simulate what could happen in a
minute or an hour. Thus, our initial predictionis that, con-
sistent with earlier research on temporal encoding in situ-
ation models, readers will make accurate judgmentsabout
what can transpire in a minute versus an hour.
However, we also predict that those judgments will be
affected by the preferences generated through plot-driven
analyses. Consider our earlier example with a change in
Sentence 2:
EXPERIMENT 1
Readers’ Reality-Driven Analyses
We designed Experiment1 to demonstrate that readers’
judgments about story outcomes are reality-driven with
respect to the passage of time. We wrote stories that ended
a minute later
with a time shift of either short duration (
)
an hour later
or long duration ( ). We also wrote outcome
sentences for each story. These outcome sentences were
consistentwith eithera short time shift or a longtime shift
(for examples, see Table 1). The participantsread each of
these stories, followed by an outcome sentence. In Exper-
iment 1A, we asked the readers to judge whether the out-
come sentence was sensible following the story. Our pre-
diction for this experiment relied on the suggestion that
readers develop and apply reality-driven temporal con-
straints as they understand these stories. As such, partici-
1. Billy was waiting outside the St. Louis Cardinals’ sta-
dium, with a baseball card for Mark McGwire to sign.
2. Billy had been diagnosed with terminal cancer, and his
lifelong dream was to meet his baseball idol.
3. McGwire’s fans were holding all types of baseball
memorabilia.
4. One child was holding a baseball.
5. Another adult was holdinga large framed photographof
McGwire.

782
RAPP AND GERRIG
Table 1
strated that participants exhibited slower reading times
when a characters’ actions late in a text (e.g., “Mary or-
dered a cheeseburger and fries.”) contradicted informa-
tion given early in the text (e.g., “Mary, a health nut, has
been a strict vegetarian for 10 years.”). In Experiment 1B,
the participants read stories that ended with explicit out-
come sentences. We predicted that readers would gener-
ate reality-driven expectations to a sufficient criterion
that their reading times would be increased for outcome
sentences that did not match reality-driven expectations
for time shifts.
Sample Stories and Outcome Sentences From Experiment 1
Story and Time Shifts
Margaret had to hurry when the hotel forgot to give her a wake-up call.
Margaret had been staying in the hotel as part of a short overnight lay-
over before the next morning’s early flight to her final destination.
Margaret threw all of her clothes into her luggage bag.
Then she took the elevator down to the lobby of the hotel.
Margaret finished gulping down her coffee and ran out to the taxi stand.
A cab driver put her luggage in the trunk and she took a seat inside the
cab.
“To the airport, and step on it,” she said.
AlthoughExperiment1A addresses whether readers are
able to make accurate judgments about what types of
events are likely to transpire in different amounts of time
(i.e., a minute vs. an hour), Experiment 1B need not pro-
vide a parallel finding. Specifically, it could easily be the
case that participants would read and accept an outcome
without expendingthe effort to determinewhetherthe out-
come was compatiblewith an overarchingtemporal model.
Readerscould, for example, accept the narrator’s assertion,
“Billygot Mark McGwire’s autograph”withoutattempting
to integrate that outcome with the previous statement, “A
minute later he got into his limo and rode off.” As we have
indicated, we predict that readers will attempt such an in-
tegration. If our prediction is confirmed, Experiment 1B
will provide evidence that readers are, in fact, engaging in
some type of analysisto determineif the episodesare tem-
porally coherent.
The cab drove sixty miles per hour the entire route and approached the
airport terminal.
(M) A minute later the airplane took off. (minute time shift)
(H) An hour later the airplane took off. (hour time shift)
Story Outcomes
Margaret missed her flight. (minute-consistent)
Margaret made her flight. (hour-consistent)
Comprehension Question
Margaret drank some coffee while riding in the taxi. (False)
Story and Time Shifts
In the 1950s, a group of four men decided to form a singing troupe
called the Jingles.
The Jingles were a group of four young men that sang pop songs dur-
ing the early days of television.
Early in the Jingles’ careers, they had a chance to perform on the Amer-
ican Bandstand television program.
The show, hosted by Dick Clark, was literally watched by millions
around the country.
Method
Being seen on the program would guarantee the success of an up and
coming band.
Participants. Twenty-four Stony Brook undergraduates partici-
pated in Experiment 1A, and twenty-four in Experiment 1B, for class
credit. All participants were native speakers of English.
Television was beginning to become available in households across
America.
The same day, the President of the United States was going to interrupt
television programming to discuss current developments in the Ko-
rean War.
Apparatus. The experiments were run on two IBM-compatible
486 personal computers that recorded responses and judgment la-
tencies. The participants were seated in front of a MicroScan color
monitor with their hands resting on the keyboard. They used but-
tons on the keyboard to make appropriate responses. The sentences
were displayed in the center of the screen in standard upper- and
lowercase type.
Materials. To begin this series of experiments, we wrote 24 sto-
ries. These stories were each 10 sentences long. The first 8 sentences
introduced a character (or group of characters) and a scenario (e.g.,
Brad and a malfunctioning computer, Charles trying to get airline
tickets, the Jingles trying to perform on national television). The 9th
sentence indicated a time shift by using either a minute later or an
hour later. A final sentence followed the time shift, describing the
outcome of the story. There were two versions of the outcome sen-
tence for each story. A minute-consistent outcome was more likely
to have occurred if only a minute of story time had passed; an hour-
consistent outcome was more likely if an hour of story time had
passed. We equated the number of words for the two versions of the
outcome sentence.
We conducted an off-line norming study to ensure that readers
found the outcomes to be reasonable given the time constraints (e.g.,
that the minute outcome could, in reality, transpire in a minute). We
asked 54 native English-speaking Stony Brook undergraduates to
read one version of each story and indicate whether they thought the
outcome sentence accurately described what they felt could happen
next. There were six versions of each item (minute vs. 10-minute vs.
hour time shift, minute-consistent vs. hour-consistent outcome sen-
tence). (We included a 10-minute time shift to serve as baseline data
for a future experiment.) Using a Latin-square design, we placed one
Dick Clark introduced the Jingles, and they came out for a brief inter-
view early in the program.
A minute later the President interrupted the program. (minute time
shift)
An hour later the President interrupted the program. (hour time shift)
Story Outcomes
The Jingles never received the opportunityto sing. (minute-consistent)
The Jingles’ performance was shown across the nation. (hour-consistent)
Comprehension Question
The Jingles gave a short interview before performing. (True)
pants should more readily accept outcome sentences that
are consistent with a matching time shift.
In Experiment 1B, we assessed the extent to which
readers’ temporal analyses have an influence on reading
times. Previous research has suggested that reading times
are sensitive to the consistency between events early in a
text and subsequent information (Albrecht & O’Brien,
1993; Cook, Halleran, & O’Brien, 1998; O’Brien, Riz-
zella, Albrecht, & Halleran, 1998; Rapp, Gerrig, & Pren-
tice, 2001). For example, Albrecht and O’Brien demon-

REALITY- AND PLOT-DRIVEN ANALYSES
783
of the six versions of each item on one of six questionnaires. Each sentences. After participants pressed the next key following the
participant read 24 stories, 6 of each type. The instructions read, “At ninth sentence, a beep sounded from the computer, and an outcome
the end of each story, you’ll be asked to decide the likelihood of the sentence appeared. The participants were instructed to decide whether
final outcome. . . . Using a scale rating from 1 to 9, we would like you the outcome sentence made sense, given the story: “It will be your job
to tell us how accurate you think the sentence is in describing what to decide whether you think this sentence accurately describes what
happens next.” The endpoints of the scale were labeled 1 (not at all you feel would happen next in the story.” The participants indicated
what happens next) and 9 (exactly what happens next).
YES (i.e., “I agree [that this would happen next]”) or NO (i.e., “I dis-
Mean agreement rates for the time shift sentences are presented in agree [that this would happen next]”) by pressing “J” or “K” on the
Table 2. The data suggest that we were successful at creating time keyboard. If participants took longer than 3 sec to read and respond to
shifts that differentiated the outcome sentences. To assess the relia- the outcome sentence, a message reading “* * * too slow * * *” was
bility of this finding, we carried out analyses with both participants
(F1) and items (F2) as random variables. Those analyses revealed a
reliable interaction between time shifts and outcome sentences
[F1(2,96) 5 67.71, MSe 5 2.603, p , .001; F2(2,46) 5 114.30,
displayed on the screen for 4 sec.
The procedure was identical for Experiment 1B, with the following
modifications, eliminating the overt judgment task. The participants
pressed the next key after reading each of the 10 sentences in the en-
MSe 5 0.658, p , .001]. Matching conditions (e.g., minute time tire story. After participants pressed the next key following the 10th
shift with minute-consistent outcome and hour time shift with hour- sentence (the outcome sentence), a beep sounded from the computer,
consistent outcome) yielded higher agreement ratings than did non- and a prompt read “*** Think of a title for the story ***” (cf. Gerrig,
matching conditions (e.g., minute time shift with hour-consistent
1989; Rapp et al., 2001). We asked participants to carry out this task
outcome and hour time shift with minute-consistent outcome). The so that they would pay appropriate attention while reading each story.
10-minute time shifts yielded agreement ratings that fell between After participants pressed the next key (to indicate that they had
these matching and nonmatching conditions. There were no signif-
icant main effects for either time shifts or outcome sentences (all tion appeared on the screen. The participants were required to press ei-
Fs , 1.65). These data suggest that our stories successfully instan- ther a YES (i.e., “Yes, that is true”; “J”) or a NO (i.e., “No, that is
tiated appropriate reality-driven expectations about the passage of false”; “Z”) key in response to the comprehension question. There was
thought of a title), another beep sounded, and a comprehension ques-
time.
We also wrote 24 filler stories without time shifts. The filler sto-
no time limit for responding to the comprehension question.
ries included outcomes that were not dependent on the passage of
time. There was only one version of each filler item. In addition, for
Experiment 1B, we wrote comprehension questions for each of our
experimental and filler stories, to test the readers’ basic knowledge
for the events in the stories (see Table 1 for a sample).
Results and Discussion
We will presentthedataandanalysesfor Experiment1A,
followed by those for Experiment 1B. Table 3 presents
mean agreement rates for Experiment 1A. We eliminated
data from trials on which participants’ response latencies
were more than three standard deviationsabove the mean,
which included “too slow” trials.1 This resulted in a loss
of 0.86% of the data.
Design. For Experiments 1A and 1B, the stories and outcome sen-
tences were the same as those used in the norming study. Overall,
there were four versions of each story based on the 24 stories, as a
function of time shift and outcome sentence (see Table 1 for exam-
ples). Using a Latin-square design, we constructed four lists of sto-
ries so that each story appeared in a different version on each list. The
filler stories were added to each of the lists. Each list was split in half
and presented in random order to the participants in two blocks, one
immediately following the other. The lists were split in order to allow
the participants a short break after reading half of the stories. For both
Experiments 1A and 1B, each participant read one version of each
story and all filler items in a different random order (within each
half), for a total of 48 stories. In Experiment 1B, each story con-
ForExperiment1A, we predictedthat participantswould
be more likely to respond
sistent with a preceding time shift. To test this, we con-
ducted an analysisof agreement ratings carried out on the
yes
to outcome sentences con-
yes
proportion of
responses provided by the participants.
As is shown in Table 3, participants were considerably
more likely to accept outcomesthat were temporally con-
F
5
MS 5
cluded with a comprehension question. Comprehension questions for sistent with time shifts [ 1(1,20) 199.11,
e
0.039,
p ,
0.026, .001].
each list were counterbalanced so that half of each list contained 12
true and 12 false statements.
p ,
F
5
MS 5
.001; 2(1,20) 312.27,
Simple effects tests confirmed this pattern both for stories
e
Procedure. For Experiment 1A, participants began with 12 prac-
tice stories and outcome sentences in order to become acquainted
with the stimulus format and keyboard controls. Each story began
with the words “press next for the next story” appearing on the mon-
itor. The participants pressed the “A” key, labeled next, to proceed
to the next story. The first sentence of a story appeared on the screen.
After reading it, the participants pressed the next key to advance to
the next sentence. This sequence was repeated for each of the nine
F
5
with minute time shifts [a 64.3% difference; 1(1,20)
MS 5 p , MS 5
F 5
.001; 2(1,20) 190.82,
127.22,
p ,
0.039,
.001] and for those with hour time shifts [a
MS 5 p ,
e
e
0.026,
49.9% difference; 1(1,20) 76.62,
MS 5 p ,
0.026,
F
5
0.039,
.001]. Partic-
ipants also tended to accept more outcomes following
M 5
e
F
5
.001; 2(1,20) 114.92,
e
hour time shifts (
M 5
58.7%) than following minute time
MS 5 p 5
F
5
shifts (
54.3%) [ 1(1,20) 3.99, 0.031,
MS 5 p ,
0.051, .05]. This sug-
e
F
5
.069; 2(1,20) 8.38,
gests that participantswereesomewhat more likelyto judge
that an outcome still would not occur after an hour (e.g.,
after an hour shift, “Billy didn’t get the autograph”) than
that an outcome could occur with insufficient time (e.g.,
after a minute shift, “Billy got Mark McGwire’s auto-
graph”).
Table 2
Mean Agreement Ratings in Norming Study for Experiment 1
Minute-Consistent Hour-Consistent
Outcome
Outcome
Mean
Minute time shift
10-minute time shift
Hour time shift
Mean
6.27
4.86
3.66
4.93
3.68
4.56
6.17
4.80
4.97
4.71
4.91
Table 4 presents mean reading times for outcome sen-
tences in Experiment 1B. We eliminated outlying data

784
RAPP AND GERRIG
Table 3
2. Billy had been diagnosed with terminal cancer, and his
lifelong dream was to meet his baseball idol.
Agreement Rates (Percent Yes Responses)
in Experiment 1A
This version of the story with this material should gen-
erate a reader preference that Billy obtain the autograph
(i.e., that Billy will have acquired McGwire’s signature
before McGwire leaves the stadium). We know from our
earlier experiments that readers will generally accept the
assertion “Billy didn’t get the autograph.” However, the
overall prediction for Experiment 2 is that the readers’
plot-driven preferences will prompt them, on some oc-
casions, to reject outcomes that are inconsistent (or ac-
cept outcomes that are consistent)with those preferences.
In essence, we are predicting that readers’ plot-driven
analyses—what they need or wish to take place—will ren-
der time somewhat more flexible than reality allows.
As in the earlier experiments, participants read stories
that ended with either minute or hour time shifts. We ex-
pected to find the same general consistencybetween time
shifts and outcomes that we obtained in Experiment 1.
However, we also expected a relative change in the data
in the direction of the Experiment 2 preferences. For ex-
ample, in Experiment 2A, we expected the majority of
readers to accept the statement “Billy didn’t get the auto-
graph” after a minute time shift. However, we expected
that majority to be smaller when the narrative had im-
parted a reader preference that Billy get the autograph(an
hour-consistentpreference), by comparison with circum-
stances in which the narrative imparted a reader prefer-
ence that Billy fail to receive the autograph (a minute-
consistent preference). For Experiment 2B, we expected
to find the same pattern of preferences overlaid on Ex-
periment 1B’s reality-driven reading time results. Specif-
ically, we expected longer reading times when story out-
comes were inconsistent with readers’ preferences.
Minute-Consistent
Outcome
Hour-Consistent
Outcome
Mean
Minute time shift
Hour time shift
Mean
86.4
33.7
60.1
22.1
83.6
52.9
54.3
58.7
falling more than three standard deviations above the
mean. This resulted in a loss of 1.21% of the data.
For Experiment 1B, we predicted that participants’
reading times would be slowed when they encountered an
inconsistency between reality-driven temporal expecta-
tions and the actual story outcome. Specifically, we ex-
pected that participants’ reading times would increase
when they read an outcome sentence that was inconsis-
tent with a previous time shift. Analysis of reading times
supported this prediction: Participants were, on average,
140 msec slower to read the outcome sentence if it was
inconsistent with an earlier time shift than if it was con-
F
5
MS 5
p ,
sistent [ 1(1,20)
11.03,
58,468,
ple effects tests revealed that the 212-msec difference for
MS 5
43,996,
.005;
p ,
.01]. However, sim-
e
F
5
MS 5
2(1,20) 9.10,
e
F
5
minute time shifts was reliable [ 1(1,20) 12.26,
p , MS 5 p ,
58,468,
e
F
5
43,996,
.01], whereas the 67-msec difference for hour time shifts
MSe 5 p .
.005; 2(1,20) 9.22,
e
F
5
was not [ 1(1,20)
1.22, .20;
p .
58,468, .20]. These simple
43,996,
F
5
MS 5
2(1,20) 0.92,
effects tests suggesteed that, in moment-by-moment read-
ing, people marshaled stronger expectations for the con-
comitants of minute time shifts than for those of hour
time shifts: Events following an hour’s passage of time
are less constrained than those following a minute’s pas-
sage. Recall that, in Experiment 1A, we found symmetri-
cal effects on people’s judgments. This contrast suggests
that somewhat different constraints were at work during
the exigenciesof reading. There were no significant main
Method
Participants. Forty-eight Stony Brook undergraduates partici-
pated in Experiment 2A, and 48 in Experiment 2B, for class credit.
All the participants were native speakers of English.
Apparatus. The apparatus was identical to that used in Experi-
ment 1.
F ,
s 1.5).
effects of time shift or outcome sentence (all
Overall, these resultsprovidestrong evidencethat read-
ers engage in reality-driven analyses of the temporal co-
herence of narrative episodes. They complement Zwaan’s
(1996) results by showing that readers construct models
that are influencedby the temporal coherenceof narrative
events. We intended the second experiment to confirm
that readers’ involvementwith narratives’ plots can bring
about modifications of reality-driven judgments.
Materials. To evaluate the impact of plot-driven analyses, we
added preference information to the narration of each experimental
story. We wrote two new versions of the second sentence of each story.
These sentences were designed to instantiate either a preference for
the minute-consistent outcome (a minute-consistent preference) or a
preference for the hour-consistent outcome (an hour-consistent pref-
erence) (see Table 5). We equated the lengths of the two versions of
the second sentence.
We pretested the revised stories to determine whether the preference
statements had the intended impact on readers’ preferences. We asked
24 English-speaking Stony Brook undergraduates to read one version
EXPERIMENT 2
Readers’ Plot-Driven Analyses
Table 4
We designed the second phase of our project to demon-
strate that plot-driven analyses have a consistent impact
on readers’ use of temporal expectations. We wrote ver-
sions of each of our stories that imparted strong reader
preferences for particular story outcomes. Recall the
sample story from the introduction, which had this sec-
ond sentence in one of its two versions:
Mean Reading Times (in Milliseconds)
for Outcome Sentences in Experiment 1B
Minute-Consistent
Outcome
Hour-Consistent
Outcome
Mean
Minute time shift
Hour time shift
Mean
1,881
2,081
1,981
2,093
2,014
2,054
1,987
2,048

REALITY- AND PLOT-DRIVEN ANALYSES
785
Table 5
between positive and negative outcomes (M 5 6.14). These means
differed reliably from one another [F1(2,42) 5 19.53, MSe 5 4.759,
p , .001; F2(2,44) 5 65.78, MSe 5 1.815, p , .001]. These pretest
data suggest that the preference statements successfully instantiated
suitable preferences for outcomes.
Sample Stories and Outcome Sentences From Experiment 2
Minute-Consistent Preference
Story and Time Shifts
Andre and his wife had been arguing again, but this time things had
gone completely crazy.
Andre continually abused his helpless wife Karen, and now he was
going to really give her a beating.
Andre managed to grab the gun out of Karen’s hands.
He wiped sweat from his brow and began stepping backwards until his
back hit the bedroom wall.
Her eyes grew wide as she realized what was about to happen.
Andre’s finger gently pressed the trigger of the gun.
Karen fell to the floor in a pool of sticky blood.
Andre knew the police would be here soon, but it was hard to clear his
head.
Design. There were eight versions of each of the 24 stories, as a
function of outcome preference, time shift, and outcome sentence
(see Table 5 for an example). Using a Latin-square design, we con-
structed eight lists of stories so that each story appeared in a differ-
ent version on each list. The filler items were added to each of the
lists. Each list was split in half and presented in random order to par-
ticipants in two blocks, one immediately following the other. The lists
were split in order to allow the participants a short break after read-
ing half of the stories. For both Experiments 2A and 2B, each partic-
ipant read one version of each story and all filler items in a different
random order (within each half), for a total of 48 stories. In Experi-
ment 2B, each story concluded with a comprehension question. Com-
prehension questions for each list were counterbalanced so that half
of each list contained 12 true and 12 false statements.
(M) A minute later the cops barged into the room. (minute time shift)
(H) An hour later the cops barged into the room. (hour time shift)
Procedure. The procedure for Experiment 2A was identical to
that used in Experiment 1A, and the procedure for Experiment 2B
was identical to that used in Experiment 1B.
Story Outcomes
Andre was caught holding a smoking gun. (minute-consistent)
Andre had cleaned the room of evidence. (hour-consistent)
Results and Discussion
Hour-Consistent Preference
Table 6 presents mean agreement rates for Experi-
ment 2A. We eliminated trials on which participants’ re-
sponse latencies were more than three standard deviations
above the mean. This resulted in a loss of 0.87% of the
data.
Story and Time Shifts
Andre and his wife had been arguing again, but this time things had
gone completely crazy.
Andre knew that Karen would hurtthe children after murdering him, so
he acted quickly to save them.
Andre managed to grab the gun out of Karen’s hands.
He wiped sweat from his brow and began stepping backwards until his
back hit the bedroom wall.
Her eyes grew wide as she realized what was about to happen.
Andre’s finger gently pressed the trigger of the gun.
Karen fell to the floor in a pool of sticky blood.
Andre knew the police would be here soon, but it was hard to clear his
head.
(M) A minute later the cops barged into the room. (minute time shift)
(H) An hour later the cops barged into the room. (hour time shift)
The agreement ratings analyses were carried out on the
yes
proportion of participants’
responses. As in Experi-
ment 1A, we predicted that participants would prefer out-
comes that matched time shifts. The data confirmed this
prediction:On average, participantswere 35% more likely
to agree with story outcomesthat matched preceding time
F
5
MSe 5
p ,
0.080, .001;
shifts [ 1(1,40)
F
140.78,
MS 5
5
p ,
2(1,16) 120.24,
0.045, .001]. Simple ef-
fects tests confirmed tehis pattern both for stories with
F
5
minute time shifts [a 46.5% difference; 1(1,40)
Story Outcomes
Andre was caught holding a smoking gun. (minute-consistent)
Andre had cleaned the room of evidence. (hour-consistent)
MS 5
0.045,
p ,
F
5
143.32,
0.080,
.001; 2(1,16) 254.79,
MS 5
e p ,
.001] and for those with hourtime shifts
MS 5 p ,
e
F
5
[a 22.2% difference; 1(1,40) 59.56, 0.080,
MS 5 p ,
0.045, .001].
e
Comprehension Question
Andre grabbed the knife out of Karen’s hands. (False)
F
5
.001; 2(1,16) 105.89,
e
However, we also predicted that participants would
provide higher agreement ratings when story outcomes
matched preferences. Again, the data confirmed this pre-
diction:On average, participantswere 17% more likely to
agree with story outcomes if they were consistent with
of each story containing a sentence imparting a minute-consistent, an
hour-consiste nt, or a neutral preference. (The neutral versions used the
original sentences from Experiment 1.) The pretest stories ended be-
fore the time shifts. In parallel to procedures used by Allbritton and
Gerrig (1991) to norm their preference stimuli, each story was followed
by both outcome sentences. Participants were asked to decide which of
the two possible outcomes, labeled “A” and “B,” they would rather have
happen. Outcomes were randomly ordered between items. One version
of each story was placed into one of three questionnaires . Each partic-
ipant read 24 stories, 8 of each type. The instructions read, “Using a
scale rating from 1 to 9, we would like you to tell us which outcome you
would rather have happen.” The scale was labeled 1 ( prefer A), 5 (neu-
tral), and 9 ( prefer B).
The data were transformed so that 1 represented a preference for
a minute-consistent outcome and 9 represented a preference for an
hour-consistent outcome. Participants preferred hour-consistent out-
comes following hour-consistent preferences (M 5 6.98) and
minute-consistent outcomes following minute-consistent preferences
(M 5 4.26). Participants’ preferences following neutral statements fell
Table 6
Agreement Rates (Percent Yes Responses)
in Experiment 2A
Minute-Consistent
Outcome
Hour-Consistent
Outcome
Mean
Minute-Consistent Preference
Minute time shift
Hour time shift
Mean
89.5
56.2
72.9
29.9
57.6
43.8
59.7
56.9
Hour-Consistent Preference
Minute time shift
Hour time shift
Mean
69.0
43.6
56.3
35.6
86.6
61.1
52.3
65.1

786
RAPP AND GERRIG
F
5
p ,
preferences than if they were inconsistent [ 1(1,40)
.001]. This result replicates Experiment 1B. How-
ever, unlike Experiment 1B, simple effects tests con-
.001]. Simple effects tests suggested that the firmed this pattern both for stories with minute time shifts
MS 5
p ,
F
5
MS 5
.001; 2(1,16) 24.27,
e
38.94,
0.052,
0.072,
e
p ,
F
5
MS 5
MS 5
effect of preference was more robust when the stories im- [a 284-msec difference; 1(1,40) 57.33,
e
206,929,
p ,
102,171, .001]
p ,
F
5
parted preferences for the minute-consistent outcomes[a
.001; 2(1,16) 116.10,
F
5
MS 5 p ,
29.2% difference; 1(1,40) 112.91,
0.072,
.001] than for
the hour-consistent outcomes [a 4.8% difference; 86.49,
and for those with hourtime shiftse[a 157-msec difference;
e
F
5
MS 5
p ,
F
5
MS 5
p ,
F
5
.001; 2(1,16) 156.33,
0.052,
1(1,40) 42.71,
MS 5
206,929,
p ,
.001; 2(1,16)
e
e
102,171,
.001].
e
F
5
0.052,
MS 5
.10].
p ,
F
5
1(1,40) 3.07,
MS 5 p ,
0.072,
.10; 2(1,16) 4.25,
We also predicted that participants’ reading times
would increase with mismatches between reader prefer-
e
e
We also obtained a reliable interaction between pref- ences and story outcomes. The data supported this pre-
F
5
MS 5
erences and time shifts [ 1(1,40) 6.94, 0.079, diction: On average, the participants were 302 msec
p , MS 5 p ,
0.040, .01]. The slower to read outcome sentences that were inconsistent
e
F
5
.05; 2(1,16) 8.88,
e
three-way interaction of preference, time shift, and story with the reader preference than if they were consistent
F , MS 5 p ,
F
5
F
5
outcome failed to reach significance (
pants were somewhat more likely to agree with minute- 26.35,
M 5
2.1). Partici-
[
1(1,40) 30.83,
272,737,
.001; 2(1,16)
.001]. Simple effects tests
64.7%) versus hour-consistent suggest that the effect of preferences was reliable both for
MS 5
e
MS 5
p ,
168,911,
e
consistentoutcomes(
M 5
F
5
outcomes (
52.5%) [ 1(1,40) 23.43,
0.060, stories that imparted preferences for minute outcomes [a
e
p ,
F
5
MS 5
p 5
F
5
MS 5
.05; 2(1,16) 4.01,
0.154,
.062].
365-msec difference; 1(1,40) 47.02,
272,737,
p ,
168,911, .001]
As in earlier experiments,ethese results confirm the
.001; 2(1,16) 75.92,
e
p , MS 5
F
5
e
view that readers’ judgments aboutthe likelihoodof story and for those that imparted preferences for hour out-
F
5
MS 5
outcomes are constrained by temporal reality. However, comes [a 238-msec difference; 1(1,40) 19.94,
e
p ,
F
5
MS 5
.001; 2(1,16) 32.19, 168,911,
e
the data also supportour suggestionthat plot-drivenpref- 272,737,
p ,
erences will, on some occasions, modify those temporal
judgments. For example, in Experiment 1A, the readers
.001].
There was also a significantinteractionbetween reader
MS 5
F
5
preferred minute-consistent story outcomes (e.g., “Billy preferences and time shifts [ 1(1,40) 45.32,
p , MS 5
e
F
5
p ,
didn’t get the autograph”), on average, 86% of the time if 131,986,
.001; 2(1,16) 35.93,
94,152,
e
they followed a matching time shift (e.g., “A minute later .001]. Table 7 suggests the origin of this result: Partici-
he got into his limo and rode off.”). However, in Experi- pants were slowest reading the outcome sentences when
ment 2A, these agreement ratings fell to 69% after the the time shift was consistentwith the preference (e.g., the
story instantiateda preference for an hour-consistentout- readers preferred and the story delivered a minute time
come.
shift) but the outcome was consistent with neither (e.g.,
Table 7 presents mean reading times for the outcome the stated outcomewas the hour outcome). There were no
sentences in Experiment 2B. We eliminated data falling other reliable effects.
more than three standard deviationsabove the mean. This
resulted in a loss of 0.26% of the data.
The overall results of Experiment 2 support our sug-
gestion that readers’ plot-driven preferences have an im-
For Experiment 2B, we predicted that participants’ pact on the expectationsthey generate for story outcomes.
reading times would be slowed when they were presented In Experiment 2A, readers’ overt judgments about the re-
with an outcome sentence that was inconsistentwith their lationship between time shifts and likely outcomes were
temporal models. This predictionwas confirmed: On av- affected by plot-drivenpreferences. In Experiment2B, al-
erage, participantswere 220 msec slower to read outcome thoughparticipants’reading times displayedsensitivityto
sentences that were inconsistent with the preceding time the consistency between time shifts and outcomes, those
F
5
MS 5
shift than if they were consistent[ 1(1,40) 18.05,
e
reading times also provided evidence of an overlaid im-
p ,
F
5
MS 5
.001; 2(1,16) 22.22, 102,171, pact of consistency with preferences.
e
206,929,
GENERAL DISCUSSION
Table 7
The goal of this research was to demonstrate an inter-
action between readers’ reality-driven and plot-driven
analyses of likely narrative outcomes. In our first pair of
experiments, readers used reality-driven analyses to
structure their expectations: Both their overt judgments
and reading times reflected assessments of the concomi-
tants of the passing of time that were consistentwith real-
world constraints. When, for example, only a minute had
passed but the narrative asserted that an hour’s worth of
events had transpired, participants showed consistent in-
creases in their time to understand the stories. Our second
pair of experiments demonstrated that, even so, readers’
Mean Reading Times (in Milliseconds) for Outcome Sentences
in Experiment 2B
Minute-Consistent
Outcome
Hour-Consistent
Outcome
Mean
Minute-Consistent Preference
Minute time shift
Hour time shift
Mean
1,995
1,964
1,980
2,565
2,125
2,345
2,280
2,045
Hour-Consistent Preference
Minute time shift
Hour time shift
Mean
1,990
2,524
2,257
1,988
2,050
2,019
1,989
2,287

REALITY- AND PLOT-DRIVEN ANALYSES
787
plot-driven preferences for particular outcomes wield an readers did not remove themselves from the narrative
influenceon narrativeexperiences:Again, bothovert judg- worlds of our experimental stories to encode and apply
ments and reading times reflected a shift such that readers generic expectations about “good guys” versus “bad
showed a propensityto accept that their preferred outcome guys.” In parallel to earlier research, we suggest that read-
might have happened despite reality-driven constraints. ers’ hopes and preferences heighten feelings of suspense
We take the overallpattern of data to supportour view that and the urgency of the outcomes (Gerrig, 1993).
readers’ narrative experiences are affected by the impera-
tives of both reality and plot.
When we defined plot-driven analyses in the introduc-
tion, we suggested that readers are likely to undertake
Our experimentstook as their starting point theories of these types of analyses quite regularly, but that different
situation models, such as the event-indexing model, that instances of plot-driven analyses will require different
make concrete predictions about the dimensions readers mental processes and representations. We conclude by
regularly encode during narrative experiences (Zwaan, briefly returning to that observation.For our experiments,
Langston, & Graesser, 1995; Zwaan, Magliano, & Graes- we have suggested that readers’ preferences prompted
ser, 1995). In that context, we believe that our first pair of them to act toward identicaltexts (i.e., the parts of the texts
experiments further contributes to the literature that sug- subsequent to the preference material) in quite different
gests that readers are attentive to temporal relationships ways. However, we categorized our preferences as only
positive negative
or . Suppose we varied the intensity of
when they create representations of texts. The question,
then, becomes how exactly the preferences our stories in- preferences within categories. We would be likely to find
stantiated (in Experiment 2) brought about the shifts in that readers would expend more effort (or different types
readers’ judgmentsand reading times. In the introduction, of effort) when the preferences were particularly strong.
we suggested that preferences might have an impact on We would imagine, for example, that readers would be
the way in which readers construct their mental simula- particularly intent on their plot-driven analyses in life-or-
Thunderball
tions: Readers could imagine fitting less or more activity death circumstances, such as James Bond’s
into an interval of time, depending on their plot-driven dilemma. We use this example to reinforce our suggestion
predilections. A related alternative would be that, in the that plot-drivenanalyseswill give rise to diverse processes
face of preferences, readers work less diligently to simu- and representations. An understanding of this diversity
late counterevidence. Consider, for example, circum- must be a precursor to a completecognitivepsychological
stances in which McGwire leaves the stadium after a account of readers’ experiences of narratives.
minute. Suppose the narration asserts “Billy got Mark
McGwire’s autograph.” In the absence of a preference, we
would expect readers to be skeptical (as they were in Ex-
periment 1) because they cannot easily imagine how, in a
minute’s time, that could be the case. In the presence of a
preference, we are suggesting either that readers might
simulatea scene in which Billy stands out and/or that they
might not work as hard to simulate the scene at all. When
they get the outcome they desire, readers might disengage
the ordinary impulse to construct or evaluate the coher-
ence of a text along a particular dimension (in this case,
the temporal dimension).
We have suggested that our plot-driven results arise
from readers’ hopes and preferences. An alternative pos-
sibility is that the preferences were creating “good guys”
and “bad guys” and that the readers were engaginggeneric
expectations for who should win or lose. Recall, for ex-
REFERENCES
Albrecht, J. E., & O’Brien, E. J. (1993). Updating a mental model:
Maintaining both local and global coherence. Journal of Experimen-
tal Psychology: Learning, Memory, & Cognition, 19, 1061-1070.
Allbritton, D. W., & Gerrig, R. J. (1991). Participatory responses in
text understanding. Journal of Memory & Language, 30, 603-626.
Barsalou, L. W. (1999). Perceptual symbol systems. Behavioral &
Brain Sciences, 22, 577-660.
Cook, A. E., Halleran, J. G., & O’Brien, E. J. (1998). What is read-
ily available during reading? A memory-based view of text process-
ing. Discourse Processes, 26, 109-129.
Fleming, I. (1961). Thunderball. New York: Viking.
Gerrig, R. J. (1989). Suspense in the absence of uncertainty. Journal
of Memory & Language, 28, 633-648.
Gerrig, R. J. (1993). Experiencing narrative worlds. New Haven: Yale
University Press.
Glenberg, A. M., & Mathew, S. (1992, December). When minimal-
ism is not enough: Mental models in reading comprehension. Psy-
coloquy [On-line serial], 3(64).
Glenberg, A. M., Meyer, M., & Lindem, K. (1987). Mental models
contribute to foregrounding during text comprehension. Journal of
Memory & Language, 26, 69-83.
Thunderball
. We sug-
ample, the opening excerpt from
gested that readers’ plot-driven analyses lead them to
hope, against the reality of his situation, that Bond will
survive. Perhaps what is at work here is not a hope but the Graesser, A. C., Kassler, M. A., Kreuz, R. J., & McLain-Allen, B.
(1998). Verification of statements aboutstory worlds that deviate from
normal conceptions of time: What is true about Einstein’s dreams?
Cognitive Psychology, 35, 246-301.
strong expectation,based on past experience,that Bond is
indestructible. However, if this were the case, it would be
hard to explain why readers experience suspense—which
they do (Gerrig, 1989, 1993). Readers’ ongoingfeelings of
suspense suggestthat the expectationthat Bond cannotdie
does not function in their moment-by-moment experi-
ence of the narrative. To access the “fact” that Bond can-
not perish, readers must purposefully remove themselves
from the narrative world. We argue, similarly, that the
Kahneman, D., & Tversky, A. (1982). The simulation heuristic. In D.
Kahneman, P. Slovic, & A. Tversky (Eds.), Judgment under uncer-
tainty: Heuristics and biases (pp. 201-208). New York: Cambridge
University Press.
Magliano, J. P., Miller, J., & Zwaan, R. A. (2001). Indexing space
and time in film understanding. Applied Cognitive Psychology, 15,
533-545.
O’Brien, E. J., Rizzella, M. L., Albrecht, J. E., & Halleran, J. G.

788
RAPP AND GERRIG
(1998). Updating a situation model: A memory-based text process-
ing view. Journalof Experimental Psychology: Learning, Memory, &
Cognition, 24, 1200-1210.
Perrig, W., & Kintsch, W. (1985). Propositional and situational rep-
resentations of text. Journal of Memory & Language, 24, 503-518.
Prentice, D. A., Gerrig, R. J., & Bailis, D. S. (1997). What readers
bring to the processing of fictional texts. Psychonomic Bulletin & Re-
view, 4, 416-420.
Zwaan, R. A. (1996). Processing narrative time shifts. Journal of Exper-
imental Psychology: Learning, Memory, & Cognition, 22, 1196-1207.
Zwaan, R. A., Langston, M. C., & Graesser, A. C. (1995). The con-
struction of situation models in narrative comprehension: An event-
indexing model. Psychological Science, 6, 292-297.
Zwaan, R. A., Madden, C. J., & Whitten, S. N. (2000). The presence
of an event in the narrated situation affects its availability to the com-
prehender. Memory & Cognition, 28, 1022-1028.
Radvansky, G. A., Zwaan, R. A., Federico, T., & Franklin, N.
(1998). Retrieval from temporally organized situation models. Jour-
nalof Experimental Psychology:Learning,Memory, & Cognition, 24,
1224-1237.
Rapp, D. N., Gerrig, R. J., & Prentice, D. A. (2001). Readers’ trait-
based models of characters in narrative comprehension. Journal of
Memory & Language, 45, 737-750.
Zwaan, R. A., Magliano,J. P., & Graesser,A. C. (1995). Dimensions
of situation model construction in narrative comprehension. Journal
of Experimental Psychology: Learning, Memory, & Cognition, 21,
386-397.
Zwaan, R. A.,& Radvansky,G. A. (1998). Situationmodels in language
comprehension and memory. Psychological Bulletin, 123, 162-185.
Rich, S. S., & Taylor, H. A. (2000). Not all narrative shifts function
NOTE
equally. Memory & Cognition, 28, 1257-1266.
Rinck, M., Hähnel, A., & Becker, G. (2001). Using temporal infor-
mation to construct, update, and retrieve situation models of narra-
tives. Journal of Experimental Psychology: Learning, Memory, &
Cognition, 27, 67-80.
1. We recorded participants’ response times in Experiment 1A and
Experiment 2A to enable us to eliminate trials that were unusually long.
We do not report those data because we had no predictions with respect
to them, nor did they produce stable patterns.
Taylor, H. A., & Tversky, B. (1997). Indexingevents in memory: Ev-
idence for index dominance. Memory, 5, 509-542.
van Dijk, T. A., & Kintsch, W. (1983). Strategies of discourse com-
prehension. New York: Academic Press.
(Manuscript received July 11, 2001;
revision accepted for publication April 22, 2002.)