133
837
THS

THES'S

IllmuIWilliIliliiliilliiwlll

3 1293 017

LIBRARY

Michigan State -
University

 

 

 

 

This is to certify that the

dissertation entitled

THE ROLE OF MENTAL SIMULATION
IN CAUSAL DECISION MAKING

presented by

Keith E. Niedermeier

has been accepted towards fulfillment
of the requirements for

Ph.D. degree in Psychology

 

7AM

Major professor

Date jl/J/éI/ff

MS U is an Affirmariw Action/Equal Opportunity Institution 042771

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1198 chlRC/DucOmpGG-p.“

 

THE ROLE OF MENTAL SIMULATION IN CAUSAL DECISION MAKING

By

Keith E. Niedermeier

A DISSERTATION

_ Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Psychology

1999

ABSTRACT
THE ROLE OF MENTAL SIMULATION IN CAUSAL DECISION MAKING

By

Keith E. Niedermeier

The theoretical importance of alternative hypotheses in
causal decision making has been recognized by several
scholars (Dougherty, Gettys, & Thomas, 1997; Einhorn &
Hogarth, 1986; Pennington & Hastie, 1992, 1993). However,
empirical work on the subject to date paints an incomplete
picture. The purpose of the present investigation was to
explore how the presence and properties of alternative
hypotheses can affect belief in a primary hypothesis. Four
experiments were conducted in which participants acted as
jurors in a case in which a bus company was being sued for
killing a dog. Across experiments, participants generally
ruled against the defendant less when they were able to
mentally simulate an alternative explanation than When
simulation was more difficult. In other words, the ability to
simulate an alternative hypothesis reduced belief in the

primary hypothesis.

This dissertation is dedicated to the many great teachers who
have influenced me over the years. Foremost among them are
Irwin Horowitz, Norbert Kerr, James Livecchi, Allen
McConnell, Lawrence Messé, Alexander von Eye, Kipling
Williams, and my parents. I further dedicate this work to

Michelle Niedermeier.

iii

ACKNOWLEDGMENTS

I would first like to thank my advisor and the chair of
my dissertation committee, Dr. Norbert Kerr. Dr. Kerr has
provided unparalleled direction and support throughout my
graduate career. Another person worthy of special thanks is
Dr. Lawrence Messé, who played a crucial role in this line of
research and often served as a surrogate advisor to me. I
would also like to thank Dr. Bruce Burns, Dr. Thomas Carr,
and Dr. Allen McConnell, all of whom served on my
dissertation committee. Additionally, Dr. Irwin Horowitz at
Oregon State University provided needed input and support.
Dr. Gwen Wittenbaum of the Department of Communications and
Dr. Robert Hymes of the University of Michigan—Dearborn
assisted with elements of this paper. Also, Kelly Sandor
deserves recognition for her assistance with data collection.

The following deserve special acknowledgment: Apple
Computer, Inc. for making products that work easily and
seemlessly, the Michigan State University Basketball and
Hockey teams for advancing to the final four of their
respective NCAA tournaments and providing a needed uplifting
diversion, Pennsylvania State University for hiring me early
and eliminating my job anxiety, my students for showing me
why I want to do this in the first place, Tala for being

there, and Michelle for everything.

iv

TABLE OF CONTENTS

LIST OF TABLES ............................................. Vi
INTRODUCTION ................................................ l
EXPERIMENT l ............................................... 19
Method ................................................ 22
Results ............................................... 26
Discussion ............................................ 28
EXPERIMENT 2 ............................................... 31
Method ................................................ 33
Results ............................................... 36
Discussion ............................................ 41
EXPERIMENT 3 ............................................... 44
Method ................................................ 46
Results ............................................... 49
Discussion ............................................ 51
EXPERIMENT 4 ............................................... 52
Method ................................................ 55‘
Results ............................................... 56
Discussion ............................................ 59
GENERAL DISCUSSION ......................................... 62
CONCLUSION ................................................. 67
APPENDIX A ................................................. 70
REFERENCES ................................................. 72

LIST OF TABLES

Table 1: Verdicts, Subjective Probabilities,
Verdict/Confidence, and Ease of Simulation by Condition

(Experiment 2) ...........................................

Table 2: Verdicts, Subjective Probabilities,
Verdict/Confidence, and Ease of Simulation by Condition

(Experiment 3) ...........................................

Table 3: Verdicts, Subjective Probabilities,
Verdict/Confidence, and Ease of Simulation by Condition

(Experiment 4) ...........................................

Table 4: Percentage of Participants Who Considered Gray...

vi

..37

..50

..57

.60

INTRODUCTION

Everyday we make judgments about the causes of events.
Often we are presented with a possible or likely cause and we
must decide if that explanation is, in fact, valid. For
instance, jurors are routinely given the primary hypothesis
that a defendant is responsible for a certain negative
outcome. The jurors must then decide whether to accept this
hypothesis as true or not. Much of the existing research on
this type of decision making has focused on the amount of
evidence or the likelihood that the primary hypothesis is
true. For example, a variety of juror decision making models
focus on the amount and the weight given to the evidence (see
Hastie, 1993, for a review). However, much less attention has
been paid to the effects of alternative hypotheses on the
perceived validity of a primary hypothesis. The purpose of
the present investigation was to explore how the presence and
properties of alternative hypotheses can affect belief in a
primary hypothesis.

Einhorn and Hogarth (1986), in their review of the
literature on judging causality, discussed the importance of
alternative explanations for the cause of a particular
outcome. Einhorn and Hogarth stated that the strength of a
primary explanation is discounted by the presence of specific
alternative explanations. Furthermore, they proposed that the
power of a specific alternative explanation lies in the

potential for it to replace the primary explanation. Although

Einhorn and Hogarth emphasized the importance of alternative
explanations in the process of causal reasoning and decision
making, they did not offer any empirical data to support
their notion.

Pennington and Hastie's (1992, 1993) story model
represents another attempt to integrate alternative
hypotheses into a model of causal reasoning. According to the
model, jurors initially encode evidence and implications in
long term memory. Pennington and Hastie state that the
evidence is stored in a meaningful structure that represents
a coherent, narrative summary of the testimony. This coherent
structure is referred to as the story. In the second stage,
jurors learn the verdict categories from the judge’s
instructions. Finally, the juror matches the story to the
category of best fit. The degree to which the story is
complete, matches a verdict category, and is coherent yields
a degree of confidence in the verdict.

Research by Pennington and Hastie (1986, 1988, 1992) has
generally supported the story model. For instance, the
researchers used extensive interviewing techniques to
establish that jurors were, in fact, creating narrative
structures of the evidence. In their studies, jurors who
reached similar verdicts possessed similar narrative accounts
of the evidence. Furthermore, jurors incorporated a
substantial amount of implied information into their stories.
The researchers also demonstrated that story summaries were

the key factor in reaching a verdict by varying the order of

evidence. Changing the order of evidence presumably altered
the construction of the jurors’ stories and ultimately the
verdicts.

Pennington and Hastie (1992, 1993) state that multiple
stories can be constructed, but one story will usually be
accepted as the best story. The emergence of the “best" story
is based on the principles of coverage, coherence, and
uniqueness. Coverage is the extent to which a story accounts
for all of the evidence. Coherence is the extent to which a
story is consistent, plausible, and complete. If only one
story is judged to be coherent, the story is said to be
unique. It is the concept of uniqueness that is of most
interest to the present investigation.

Pennington and Hastie (1992, 1993) state that the
presence of more than one coherent story will lessen the
belief in any one particular story. However, this idea is
given little attention in the various tests of this
perspective. It is assumed that one story will emerge as the
best and the other stories merely will be dismissed. It is
important to note that no explicit basis for alternative
stories are given in the trials in the cases used in
Pennington and Hastie’s (1992) research. In one case, for
example, a defendant was charged with embezzlement and there
were various pieces of incriminating and exonerating evidence
presented, but there was no explicit basis by which to create
an alternative story that could account for the missing

money. In other words, all of the evidence was either

incriminating or exonerating to the defendant. No evidence
introduced specific alternatives to the primary hypothesis
that the defendant embezzled the money.

In real court cases, explicit alternative hypotheses
often are introduced by the defense. For instance, the
defense attorney in a murder case may demonstrate that a
specific person other than the defendant had the motive and
ability to commit the murder in question. In a civil case, an
attorney may make the case that someone other than his
client, like the plaintiff or a specific third party, is
responsible for the outcome that led to the law suit. In both
examples, there is a specific basis for creating an
alternative story or hypothesis.

Empirical Inyestigatigns of Alternatiye Explanatigns

Most of the previous research on alternative hypotheses
has focused on the effect of multiple causal routes on the
perceived likelihood of different outcomes. Much of this
research has demonstrated that simulating many causal routes
to a number of different outcomes reduces likelihood
estimates of any particular outcome (Hirt & Markman, 1995;
Koehler, 1994). The focus of this paper, however, is on the
effect of multiple causal routes to a single outcome. In
other words, what is the effect of simulating alternatives to
the primary cause (hypothesis) of a given outcome?
Unfortunately, there have been few empirical tests of the
effects of alternative explanations on belief in a primary

hypothesis.

In a few studies, strength of belief in a given
hypothesis was found to be independent of belief in competing
hypotheses (Robinson & Hastie, 1985; Robinson Van Wallendael,
1989). In these investigations, subjects read murder
mysteries in which various incriminating and exonerating
clues were introduced about a primary suspect. At various
points, one or more additional suspects were introduced.
Across a number of experiments, the introduction of
alternative hypotheses in the form of additional viable
suspects did not reliably decrease the strength of belief in
a primary hypothesis.

In contrast, other research has found that the strength
of belief in a given explanation is directly related to the
number of available alternative explanations (Dougherty,
Gettys, & Thomas, 1997; Fischhoff, Slovic, & Lichtenstein,
1978; McDonald, 1998). For instance, McDonald presented
subjects with an unusual event and provided them with varying
numbers of equally coherent explanations for the cause. The
average probability assigned to a specified hypothesis
approximated the proportion of the number of hypotheses
presented. For example, if subjects were presented with three
possible explanations for an event, the average probability
assigned to a specified hypothesis was approximately 33%.
Similarly, if subjects were presented with five possible
explanations, the average probability assigned to the same

specified hypothesis was approximately 20%.

In a similar investigation, Fischhoff et a1. (1978)
presented subjects with organized lists of possible causes
for a car that failed to start. Omitting possible causes from
the lists increased estimates of probability given to the
remaining explanations, despite the presence of a category
labeled “All other problems." The researchers concluded that
increasing the number of specific alternative explanations
decreased the estimated probability of any given explanation
being the cause for the car's failure to start. Furthermore,
they stated that the subjects were quite insensitive to
possible explanations that were not specifically listed.

In a more recent investigation, Dougherty et al. (1997)
had subjects read short vignettes in which the cause of an
outcome was unclear. For instance, one vignette described a
firefighter who was killed on duty. Subjects were then asked
to judge the probability of a given cause (e.g., smoke
inhalation) and list all the thoughts they had while making
the probability judgment. The results showed that subjects
who considered multiple causes reported lower probabilities
for the given cause of an outcome.

In the aforementioned studies, the primary dependent
variable was subjective probability, that is, subjects’
estimate of the likelihood that a given explanation was “the”
cause of the event. Although Robinson and Hastie (1985) and
Robinson Van Wallendael (1989) found no differences in
subjective probabilities based on the number of alternative

explanations, Fischhoff et al. (1978) and McDonald (1998)

did. A possible reason for this inconsistency in findings is
that the former studies specified a primary hypothesis, while
the latter studies did not. In the latter studies, various
numbers of explanations for particular events were all
introduced simultaneously and none was presented as the
primary hypothesis. In the former studies, the evidence
pertained to a primary suspect and alternatives were
introduced later.

The Dougherty et a1. (1997) study was different in that
it was unclear whether one explanation was primary. Subjects
were asked to judge the probability of a given explanation
only after reading the vignette. The researchers assumed that
the given explanation (such as smoke inhalation) was primary,
but it is possible that subjects had another hypothesis in
mind, making the judged explanation actually the alternative.
In any event, it is unclear if a primary hypothesis existed
for all subjects in this study.

Collectively, these studies suggest that the
presentation of one hypothesis as the primary hypothesis may
play an important role in the decision making process. If we
have no reason to consider one hypothesis as primary,
considering all available alternatives as more or less
equally likely is not an unreasonable strategy (Pennington &
Hastie, 1992, 1993). However, when we are presented with a
primary hypothesis, we are likely to consider that hypothesis

and retain it as the “default” explanation unless it can be

replaced or discounted by a plausible alternative (Einhorn &
Hogarth, 1986).
Primary Hypgtheses

Given this distinction between primary and alternative
hypotheses, what psychologically defines a hypothesis as
primary becomes an important question. In many situations,
one hypothesis may be overtly presented as primary. As
mentioned before, jurors are routinely given the primary
hypothesis that a defendant is the cause of a certain
negative outcome.

Alternatively, a situation may be structured such that
one explanation is presented prior to other possible
explanations, as in the Robinson and Hastie (1985) and
Robinson Van Wallendael (1989) studies. In these studies, one
suspect in a murder mystery was introduced and only later
were additional suspects revealed. In these situation, the
explanation presented first may be considered primary because
a story is constructed around this explanation and retained
in memory (Einhorn & Hogarth, 1986). Unless this explanation
is replaced or discounted later, it should ultimately be
accepted.

Furthermore, the mere mention of a single possibility
may be enough for it to be considered a primary hypothesis.
This line of reasoning is consistent with the conversational
convention literature, which suggests that things are not
generally stated in conversation unless they are relevant and

important. Therefore, the mention of one causal hypothesis in

a scenario may serve as cue that it should be primary
(Hilton, 1995; Schwarz, 1994).

Another possibility is that one explanation is
considered primary because of previous experience. Imagine,
for example, that a mother walks into the kitchen and finds
that the vase that normally sits on the table is broken.
There are numerous possible hypotheses for the broken vase,
but if the mother has a six-year old with a history of
playing ball in the house, this causal schema may be the most
accessible cognitively and thus emerge as a primary
hypothesis.

More generally, people may naturally construct one
primary hypothesis for a given outcome (Dougherty et a1.,
1997). One common example of this type of process is the
hindsight bias in which people overestimate the
predictability of an outcome (Fischhoff, 1975; Hawkins &
Hastie, 1990). This bias may occur because of the tendency to
focus on one causal route that occurred and the failure to
consider alternative routes. The confirmation bias is another
example of this tendency. This bias is the inclination to
seek out evidence that supports a given hypothesis (Wason,
1960). Similarly, the confirmation bias may occur because of
the tendency to focus on one primary hypothesis and the
failure to consider alternatives. A relevant condition in
which people do seek out disconfirming information is when

alternative causal hypotheses are made salient (Doherty,

Chadwick, Garavan, Barr, & Mynatt, 1996; Jou, Shanteau, &
Harris, 1996).
Continuous yarsua Categorical Maaaoraa

As previously mentioned, the primary dependent variable
in most studies of alternative hypotheses has been subjective
probability estimates. Subjective probabilities are a
continuous measure of people’s estimates of the likelihood
that an event occurred in a given manner. However, people
often are required to make categorical decisions about cause.
For instance, jurors must ultimately decide if the defendant
is guilty or not guilty, regardless of subjective
probability. In an investigation of jurors’ use of
statistical evidence, Wells (1992) collected both continuous
and categorical measures in a study that presented
participants with a primary hypothesis.

In Wells’s (1992) study, participants read about a civil
trial in which Mrs. Prob was suing the Blue Bus Company for
running over her dog. Participants were informed that Mrs.
Prob was walking her dog along a county road when a swerving
bus hit and killed her dog. Testimony revealed that only two
bus companies, the Blue Bus Company and the Gray Bus Company,
operated in the area. Both companies used the road to run
empty buses back to their respective terminals.

Unfortunately, Mrs. Prob was color blind so she could
not identify the color of the bus. In a condition termed the
tira—traoka version, a county transportation official

testified that tire tracks at the scene were compared to all

10

10 of the buses owned by the Blue Bus Company and all 10 of
the buses owned by the Gray Bus Company. The tracks matched
80% of the blue buses and 20% of the gray buses. Participants
reported continuous subjective probabilities and categorical
verdicts indicating whether or not they would find the Blue
Bus Company liable.

Like the murder mystery studies (Robinson & Hastie,
1985; Robinson Van Wallendael, 1989), a primary hypothesis
was implicit in that the Blue Bus Company was being sued by
Mrs. Prob. In the tire-tracks version described above,
participants reported than it was near 80% likely that a blue
bus was responsible for killing the dog. The participants
were instructed to follow the preponderance of evidence
decision rule and thus should have ruled against the Blue Bus
Company. However, Wells (1992) found that fewer than 20% of
the participants ruled against the Blue Bus company in this
condition.

In another condition termed the tire—track-baliaf
version, the transportation official testified that tire
tracks at the scene were compared to all 10 of the buses
owned by the Blue Bus Company and all 10 of the buses owned
by the Gray Bus Company using a method that is correct 80% of
the time. Based on the results of this matching technique,
the official testified that he believed that a blue bus
killed Mrs. Prob’s dog. Just as in the tire—tracks version,
participants reported that it was nearly 80% likely that a

Blue Bus Company bus killed Mrs. Prob’s dog. In this latter

ll

condition, however, participants ruled against the Blue Bus
Company nearly 70% of the time.

The Wells (1992) study illustrates the important point
that subjective probabilities do not necessarily translate
into functionally equivalent categorical choices. The tire-
tracks and tire—track-belief versions produced similar
subjective probabilities, a pattern that is consistent with
the Robinson and Hastie (1985) and Robinson van Wallendael
(1989) studies, which also presented participants with a
primary hypothesis. However, the two conditions produced very
different categorical verdicts indicating that continuous and
categorical measures do not necessarily produce parallel
results, at least in some conditions.

Furthermore, Wells (1992) reasoned that this pattern
disconfirmed the family of threshold models of decision
making. These models (e.g., Dane, 1985; Kaplan, 1977) View
continuous subjective probability as driving categorical
verdicts. Generally, these models maintain that an initial
prior probability is updated by entering each piece of new
information into a mental equation that continually revises
subjective probability of guilt. This process occurs over and
over until there is no new evidence to consider. The final
subjective probability is then compared to a decision
criterion or threshold. If subjective probability surpasses
the decision threshold, the juror will rule against the

defendant.

12

Wells (1992) stated that the mock jurors is his study
were all operating under the same decision rule
(preponderance of evidence), so they should have had the same
threshold. Given the same level of subjective probability but
the very different rates of verdicts against the Blue Bus
Company in the two versions of the case, Wells reasoned that
the data disconfirmed these types of threshold models. This
view of threshold models further suggests that continuous and
categorical measures do not necessarily produce parallel
results.

Additionally, Wells’s paradigm nicely lends itself to
the question at hand because there is a clear primary
hypothesis (a blue bus killed the dog), and there is a clear
alternative hypothesis (a gray bus killed the dog). Also,
both subjective probabilities and categorical decisions can
be examined while holding actual probabilities constant.
Conaideration of Algernatiyas: The Rola of Manta; Simolation

Wells (1992) investigated a number of explanations for
the very different verdicts against the Blue Bus Company in
the two different versions of the case. He ruled out a number
of explanations and proposed fact—to-evioence reasoning as
the operative mechanism. The fact-to-evidence explanation
states that, "in order for evidence to have a significant
impact on people's verdict preferences, one's hypothetical
belief about the ultimate fact must affect one’s belief about
the evidence" (p. 746). In other words, if the evidence

presented at the trial can be true regardless of what

13

actually happened, then that evidence is seen as non—
diagnostic. For instance, the evidence in the tire-tracks
version stated that the tire tracks matched 80% of the blue
buses and 20% of the gray buses. This evidence could be
accepted as true regardless of which bus actually killed the
dog. Thus, such evidence fails the fact—to-evidence criterion
and should not have a significant impact on the verdict.

The evidence in the tire-track—belief condition,
however, stated that the official believed a Blue Bus Company
bus killed the dog based an test that is reliable 80% of the
time. The official is correct in his belief if a blue bus
killed the dog, but would be incorrect if it were a gray bus
that did so. In this version of the case, which bus company's
bus actually killed the dog (the fact) bears on the validity
of the witness’s testimony (the evidence). Because the
witness’s testimony is true only if a blue bus killed the
dog, this evidence passes the fact-to—evidence test and
affects the verdict

In contrast to the fact—to-evidence reasoning
explanation, Niedermeier, Kerr, and Messé (1999) suggested
that mock jurors’ ability to mentally simulate a situation in
which the Blue Bus Company was go; guilty (i.e., an
alternative hypothesis) influenced their willingness to rule
against the defendant. For instance, in the tire—tracks
version, the matching technique produced a complete match
with 8 of 10 Blue Bus Company buses and 2 of 10 Gray Bus

Company buses. In this version, there indisputably are two

14

concrete gray buses that could have plausibly killed the dog.
Because it is easy to imagine that one of these two gray
buses may have killed the dog, jurors may be more reluctant
to hold the Blue Bus Company responsible. In the tire—tracks—
belief version, however, the transportation official
testified that he believed that a blue bus killed the dog
based on an 80%-reliable matching technique. This scenario
requires a much more complex cognitive mutation to plausibly
imagine that a gray bus killed the dog. Although this
testimony could create some doubt about a blue bus's guilt,
it offers no explicit basis to create a story consistent with
a gray bus's guilt. Thus, mock jurors were more willing to
rule against the Blue Bus Company.

Niedermeier et al. (1999) tested this ease—of—simulation
explanation by altering the critical testimony in the case
used by Wells (1992). In a condition termed the QQEDIBLQ:
EBLQE version, the transportation official testified that
tire tracks from the scene of the accident completely matched
8 of 10 blue buses and 2 of 10 gray buses, as in Wells tire-
tracks version. In another condition termed the oaroial-match
version, the official testified that the tire tracks
partially matched one blue bus and one gray bus, and test
results indicated an 80% probability that the blue bus killed
the dog and a 20% probability that the gray bus was the one.

In both the complete-match and the partial-match
conditions, mock jurors reported that it was about 80% likely

that a blue bus killed the dog. However, jurors were much

15

less willing to hold the Blue Bus Company responsible for
killing the dog in the complete-match condition (25%) than in
the partial-match condition (47%). In the complete-match
condition, there are two gray buses that completely matched
the tire tracks and could have killed the dog. In this
condition it is not difficult to imagine that one of these
gray buses killed the dog. In the language of the story model
(Pennington & Hastie, 1992, 1993), the possibility of a blue
bus being responsible lacks uniqueness because the
possibility of a gray bus being responsible has coherence. In
the partial—match condition, however, there is only a vague
20% probability match to a single gray bus. In this latter
condition, it is more difficult to simulate a situation in
which a gray bus could have killed the dog. The possibility
of a blue bus being responsible has uniqueness because the
possibility of a gray bus being responsible lacks coherence.
In fact, mock jurors reported that it was significantly
easier to imagine a scenario in which a gray bus killed the
dog in the complete-match condition than in the partial-match
condition. Thus, jurors’ ability to mentally simulate an
alternative hypothesis, that is, a gray bus being responsible
for killing the dog, significantly affected jurors’
willingness to rule on the primary hypothesis, that is, the
liability of the Blue Bus Company. Using a very similar
methodology to Niedermeier et al. (1999), Sykes and Johnson
(in press) found comparable patterns of verdicts and reached

a similar conclusion.

l6

It has been suggested that the ease of mental simulation
of alternatives may be used as a heuristic to determine the
validity of a given hypothesis (McDonald, 1998; Miller,
Turnbull, & McFarland, 1989). In a series of studies, Miller,
et al. demonstrated that the more easily people could imagine
something occurring by one route, the less suspicious they
were that it occurred by some other route. Miller et al.
presented subjects with a scenario in which a child takes
only chocolate chip cookies from a cookie jar while leaving
oatmeal cookies to go stale. To avoid stale cookies, the
child’s mother instructs him to close his eyes before
reaching into the cookie jar and take whichever cookie he
grabs. The child agrees to the method, goes to the kitchen
and returns with a coveted chocolate chip cookie. Subjects
were asked to rate their own suspicion that the child peeked.
In one condition, subjects were informed that there was 1
chocolate chip cookie and 19 oatmeal cookies in the jar. In
another condition, subjects were informed that there were 10
chocolate chip cookies and 190 oatmeal cookies.

Subjects expressed greater suspicion when there was only
1 chocolate chip cookie than when there were 10, despite the
fact that drawing a chocolate chip cookie at random was
equally likely in both conditions. When there was only one
chocolate chip cookie, there was only one way in which the
outcome could have occurred by chance. In contrast, it was
easier to imagine that the outcome had occurred by chance

when there were 10 chocolate chip cookies. The researchers

l7

concluded that the easier it was simulate that an outcome
occurred in one manner (chance), the less suspicious subjects
were that it occurred in another manner (peeking).

The Miller et al. (1989) study demonstrates that the
relative frequency of alternatives may play a role in the
decision making process. When presented with the primary
hypothesis that the child is peeking, participants were less
suspicious of the infraction when there were relatively more
alternative possibilities than when there were fewer
alternatives. Although the researchers suggested that it was
easier to imagine drawing a chocolate chip cookie when there
were 10 such cookies than when there was one, they provided
no direct evidence of ease of simulation.

General ijactives

To summarize, the theoretical importance of alternative
hypotheses in decision making has been recognized by several
scholars (Dougherty et al., 1997; Einhorn & Hogarth, 1986;
Pennington & Hastie, 1992, 1993). However, empirical work on
the subject to date paints an incomplete picture. The current
investigation sought to address four issues related to the
role of mental simulation of alternative hypotheses in causal
decision making.

First, the notion that mental simulation plays a role in
casual decision making was further tested by attempting to
make alternative hypotheses more or less easily simulated.
Second, the importance of categorically eliminating potential

alternatives was investigated by altering the order in which

18

participants considered various alternative explanations.
Third, the relative frequency of alternative possibilities
were altered to determine if one coherent alternative is
sufficient to reduce belief in the primary hypothesis as
opposed to several alternatives. Lastly, the heuristic nature
of mental simulation was investigated by manipulating
participants’ motivation to consider alternatives in order to
determine if mental simulation is heuristic or a systematic
but intrinsically flawed method of reasoning. In all
experiments, both categorical and continuous measures were

evaluated.

Experiment 1

Using Wells's (1992) paradigm, it was demonstrated that
the ability to mentally simulate alternatives to a primary
hypothesis reduced jurors’ willingness to believe the primary
hypothesis (Niedermeier et al., 1999; Sykes & Johnson, in
press). As previously described, jurors in the Niedermeier et
al. study were much less willing to hold the Blue Bus Company
responsible for killing a dog when it was easy to simulate a
gray bus doing so than when it was more to difficult to
simulate, as in the partial—match version. Again, there was
only a vague 20% match to a single gray bus in the partial-
match condition compared to full matches to two gray buses in
the complete-match condition.

In Experiment 3 of the Niedermeier et al. (1999) study,

one group of mock jurors read the partial-match version of

19

the blue bus case while another group read the partial-match
case with additional testimony in which the defense attorney
noted that the accident scene was not closed off to traffic.
The attorney asked if a second bus could have come by after
the accident and obscured the tracks of the responsible bus,
leaving its own tracks in place of the responsible bus’s
tracks, thus making the innocent bus look responsible. The
transportation official, who was testifying about the tire
tracks, acknowledged that this was possible, but stated that
there was no indication of a second set of tracks on the dog.

The scenario proposed by the defense attorney provided a
means for simulating a situation in which a blue bus appears
responsible when another bus actually killed the dog, without
adding any substantive evidence that changed actual (or
subjective) probabilities. However, when jurors heard the
scenario posited by the defense attorney, they were much less
willing to rule against the Blue Bus Company and reported
that it was easier to think of a way that a bus from the Gray
Bus Company could have been responsible for killing the dog
than when they heard the partial—match case alone.

The data from Niedermeier et al. (1999, Experiment 3)
suggest that, given a framework to simulate a coherent
alternative hypothesis, people will be less willing to rule
against the defendant. It follows that anything that will
increase jurors’ ability to simulate a coherent alternative
to the primary hypothesis should reduce rulings against the

defendant. For instance, consider some of the work on

20

pretrial publicity by Greene and others (Greene, 1990; Greene
& Loftus, 1984; Greene & Wade, 1988). In these studies, mock
jurors read newspaper articles about eyewitness testimony
that led to a conviction in a rape trial. The testimony was
later recanted and the convicted defendant was shown to be
innocent. Jurors then were asked to render verdicts in a
similar trial based on eyewitness testimony. Subjects who
were exposed to the pretrial publicity had significantly
lower conviction rates than subjects who were not exposed.

Although Greene and her colleagues (1984, 1988, 1990)
interpreted their findings differently, it is plausible that
the pretrial publicity provided jurors with a means to
simulate a scenario in which the defendant was not guilty
when they rendered verdicts in a similar trial. Research has
demonstrated that hearing stories, scenarios, and analogies
can create causal schematic structures that guide the
interpretation of information provided in a similar framework
(Johnson & Seifert, 1994; McKoon, Ratcliff, & Seifert, 1989;
Schank, 1982).

This possibility was tested using the partial-match
version of the blue bus/gray bus case used in Niedermeier et
al. (1999). It was predicted that given the same evidence,
mock jurors would rule against a defendant less when provided
a framework to mentally simulate an alternative explanation
than when simulation was more difficult. In other words, the
ability to simulate an alternative hypothesis should reduce

belief in a primary hypothesis. As in Niedermeier et al.

21

(1999) and Wells (1992), it was predicted that this
difference would be manifest in categorical verdicts, but not
in continuous measures of subjective probability.
Method

Participants

Participants were 72 undergraduate psychology students
who took part in the study as a classroom demonstration. The
mean age of participants was 21.9 years and 63% were female.
Desigo and Procedura

Participants were informed that they were taking part in
a study on juror decision making and that they would act as
jurors in a civil case and make several judgments. They were
further told that the case they were going to read included
some technical evidence about tire track matching. They were
then given a technical information sheet that described how
tire tracks can be lifted from an accident scene and later
matched to actual tires on a vehicle. Although the text was
fictitious, no participant questioned the veracity of the

information. The text read as follows:

Unknown vehicles can be identified by tire—track
identification. This technique involves matching the
tracks left at an accident scene to the tires on a
vehicle. Tire-track identification is effective because
each vehicle’s tires display unique patterns of wear in
addition to the fact that there are hundreds of

different types of tires and tread patterns.

22

Tire-tracks are usually lifted from the pavement using
photosensitive paper and an ultraviolet light emission
gun. The paper is laid over the track and exposed to the
ultraviolet light. The light is absorbed by the rubber
left by a recent tire—track but reflected by the
pavement. Thus, the exposed paper produces a negative
image of the track. This image is scanned into a
computer, which prints out a positive image of the track
that can then be compared to any suspect vehicles'
tires. Matches are graded from 0% (no match) to 100%

(perfect match).

For approximately half of the participants, text was
added describing situations in which tire tracks can be
obscured by vehicles passing by after the accident has
occurred, sometimes making uninvolved vehicles appear
responsible. This text was added to facilitate simulation of
an alternative explanation. The additional text read as

follows:

It is important that the track being lifted is the most
recent track on the pavement. If a second vehicle passes
over the track in question, the technique can produce a
mistaken match. For instance, if a car involved in a hit
and run is trying to be identified, it is important that

no other cars drive through the scene. If an innocent

23

car takes the same path as the offending car before the
tracks are lifted, the technique could produce a better
match with the innocent vehicle than with the offending

vehicle.

All participants then read the partial-match version of

the blue bus/gray bus trial. The text was as follows:

Mrs. Prob is suing the Blue Bus Company for having
caused the death of her dog. At the trial, the following

evidence was given:

Mrs. Prob testified that she was walking her dog along
county road #37 when she heard a large vehicle behind
her. She turned around and saw a bus serving recklessly
down the road. She jumped out of the way but the bus
swerved and hit her dog, killing him instantly. The
incident occurred at 11:40 A.M. The bus continued at a
high speed down the road. Unfortunately, Mrs. Prob is

color blind and thus does not know the color of the bus.

A county transportation official took the stand, was
sworn in as a witness, and testified that there are only
two bus companies that travel in the county: the Blue
Bus Company and the Gray Bus Company. Each company uses
the road to run empty buses back to their stations after

dropping off passengers. Therefore, one of the two bus

24

companies had to be responsible for the death of Mrs.

Prob’s dog.

A second county transportation official, who was the
transportation department’s chief investigator, took the
stand, was sworn in as a witness, and reported that he
examined the dead dog and took prints of the tire
tracks. These prints were then transferred onto paper
and compared to all 10 of the 10 buses owned by the Blue
Bus Company, and all 10 of the 10 buses owned by the
Gray Bus Company. He testified that the technique used
for matching produced a partial match with one of the
Blue Bus Company buses and a partial match with one of
the Gray Bus Company buses. The results of the test
indicated that there was an 80% chance that the blue bus
was the one that ran over the dog, and a 20% chance that
the gray bus was the one. Based on this matching
evidence, the official concluded that it was a blue bus

that killed the dog.

Mrs. Prob’s attorney argued that the jury must find the
Blue Bus Company liable for damages because in all
likelihood, it was the Blue Bus Company that killed Mrs.

Prob’s dog.

After reading the trial, participants indicated whether
they would rule against the Blue Bus Company and make them

pay damages to Mrs. Prob. Next, they reported subjective

25

probabilities by estimating the probability that the Blue Bus
Company actually killed the dog. Participants also rated
confidence in their individual verdicts on an 11—point scale
ranging from not at all confident (1) to completely confident
(11). In order to assess the ease-of-simulation, participants
were asked to rate how easy or difficult it was to think of a
situation in which a bus from the Gray Bus Company could have
been responsible for killing the dog on an eleven point scale
ranging from very easy (1) to very difficult (11).
33.51.1125.

The verdicts were in the predicted direction although
the trend was not statistically significant. Participants who
read the simulation framework ruled against the Blue Bus
Company less (24%) than participants who read the partial—
match case alone (39%), x2(l, N = 72) = 1.71, p < .20, ¢ =
.15. Effect sizes for categorical measures were calculated
using phi (Rosenthal & Rosnow, 1991). A one-tailed test was
also not significant, 2 = 1.31, o < .10.

Participants who read the simulation framework reported
that it was easier to think of a way that a bus from the Gray
Bus Company could be responsible for killing the dog (M =
3.80) than participants who read the partial-match case alone
(a = 5.13), 2(1,70) = 4.39, p_ < .05, co?- = .05. All effect
sizes for continuous measures were calculated using omega
squared (Tabachnick & Fidell, 1996). Additionally, a
significant relationship between the ease—of-simulation

measure and verdicts was revealed such that the easier

26

participants reported it was to simulate that a Gray Bus
Company bus may have killed the dog, the less willing they
were to hold the Blue Bus Company responsible, r(69) = .39, p
< .01.

Verdicts were combined with the confidence measure to
yield a scale that ranged from 1 (completely confident in a
verdict against liability of the Blue Bus Company) to 22
(completely confident in a verdict for liability of the Blue
Bus Company). Again, the means were in the predicted
direction in that participants who read the partial—match
case alone had more pro-conviction verdicts against the Blue
Bus Company (M = 10.16) than participants who read the
simulation framework (M = 7.85). However, the trend was not
significant, 51,70) = 1.86, as. m2 = .01.

Participants who read the simulation framework also
reported that it was less likely that a bus from the Blue Bus
Company killed the dog (M = 61.83) than participants who read
the partial-match case alone (M = 70.13), F(1,70) = 4.82, p <
.05, a9 = .05. A significant relationship between the
subjective probabilities and verdicts was revealed such that
participants who reported lower subjective probabilities were
less likely to rule against the Blue Bus Company, r(69) =
.42, p < .001. Additionally, a significant relationship
between ease-of-simulation and subjective probability was
revealed such that participants who reported that it was
easier to think of a way that a bus from the Gray Bus Company

could be responsible for killing the dog reported that it was

27

less likely that the Blue Bus Company killed the dog, r(69) =
.33, p < .01.

As the significant difference in subjective
probabilities was unexpected, a regression of the ease-of-
simulation measure on the verdicts was performed while
controlling for the effects of subjective probability. This
procedure was carried out to demonstrate that it was not
simply subjective probabilities driving the categorical
verdicts. Rather, ease-of-simulation was operating
independently. The ease-of—simulation measure still accounted
for a significant amount of the variance of verdicts after
controlling for subjective probability, t(69) = 2.52, p <
.02, 32 = .08.

Discussion

The data from Experiment 1 demonstrated that the easier
it was to simulate that a Gray Bus Company bus killed the
dog, the less willing participants were to hold the Blue Bus
Company responsible. Although this verdict pattern was not
significant between conditions, a significant correlation
showed that the relationship was evident across conditions.
Thus, participants’ ability to simulate an alternative
hypothesis (that a gray bus killed the dog) was related to
belief in the primary hypothesis (that a blue bus killed the
dog).

It is possible that the manipulation intended to make
simulation easier was not strong enough. Given that the

potential effect size is “small,” power analysis revealed

28

that the ability to detect the effect with the current sample
size would be less than .15 (Cohen, 1988). It would take a
sample of 400 to detect the effect at o = .05 with power of
.50. Thus, strengthening the ease-of—simulation manipulation
and increasing the sample size would aid in detecting the
effect.

Furthermore, it may be that the effect was limited by a
functional floor effect. It seems reasonable to assume that
some people will rule against the Blue Bus Company simply
because of the extreme likelihood that they are responsible,
regardless of other factors. Additionally, the rate of
verdicts against the Blue Bus Company for the partial—match
case alone was 39% in this experiment. This rate was the
lowest observed rate using the partial—match case in this
study and in previous research (Niedermeier et al., 1999).
This unusually low rate of verdicts against the Blue Bus
Company in the partial-match condition and the likely
functional floor effect may have suppressed the magnitude of
the effect in this experiment, thus making it difficult to
detect with the current sample size.

However, subjective ease—of—simulation may be more
important than the particular manipulation in that
individuals who were able to easily simulate a Gray Bus
Company bus being responsible for killing the dog ruled
against the Blue Bus Company less than those who found

simulation more difficult, regardless of condition.

29

Although it was unexpected, participants who read the
simulation framework reported that it was less likely that
the Blue Bus Company killed the dog than participants who
read the partial—match case alone. Moreover, a significant
relationship between the ease-of—simulation measure and
subjective probability was revealed such that participants
who reported that it was easier to think of a way that a bus
from the Gray Bus Company could be responsible for killing
the dog reported that it was less likely that the Blue Bus
Company killed the dog. These findings are inconsistent with
Niedermeier et al. (1999) and Wells (1992), who both found
that subjective probabilities did not differ across
conditions and were unaffected by ease-of—simulation.

In the current framework, both continuous and
categorical measures of the Blue Bus Company’s responsibility
are viewed as outcome variables, and in fact, both variables
were related to ease-of—simulation. As mentioned before, some
threshold models of decision making (e.g., Dane, 1985;
Kaplan, 1977) could view the continuous subjective
probabilities as driving the categorical verdicts. Although
the pattern of data from this experiment could support such a
model, ease—of-simulation still independently accounted for a
significant amount of the variability when the effects of
subjective probability were partitioned out. This is
consistent with the argument that mental simulation affects
categorical judgments above and beyond any effect of

subjective probability.

3O

Experiment 2

Given that the presence of alternative hypotheses can
decrease the belief in a primary hypothesis, it follows that
the elimination of such alternatives should increase belief
in the primary hypothesis. As previously noted, the current
paradigm is useful for investigating alternative hypotheses
because actual probabilities can be held relatively constant
at 80% (or any other value) for the Blue Bus Company and 20%
for the Gray Bus Company. In Experiment 2, mock jurors made
judgments about both the Blue Bus Company and the Gray Bus
Company after reading either the complete-match or partial—
match versions of the case. However, it was not made explicit
that the Blue Bus Company was being sued. Jurors were merely
told that they were going to read about a civil case in which
a woman's dog was killed. Participants were asked to make
continuous and categorical judgments while considering either
the Blue Bus Company first or the Gray Bus Company first.

Considering the Gray Bus Company first may suggest to

jurors that a gray bus being responsible is the primary
hypothesis. As it is clearly unlikely that a gray bus is
responsible (20%), jurors should easily reject this
hypothesis. Thus, when asked to rule on the Blue Bus Company,
the Gray bus hypothesis will already have been dismissed,
increasing rulings against the Blue Bus Company. This should
produce a main effect for order such that ruling against Blue
will increase when Gray is considered first. This effect

should be particularly evident in the complete—match

31

conditions because rulings against the Blue Bus Company
should be very low when Blue is considered first, as in
Niedermeier, et al. (1999). Furthermore, it was expected that
previous findings (Niedermeier et al.) would be replicated
such that jurors would find it easier to simulate a gray bus
being responsible in the complete-match conditions and rule
against the Blue Bus Company less than jurors in the partial-
match conditions. As in Experiment 1, a relationship also was
expected such that the easier participants report it is to
simulate that a Gray Bus Company bus killed the dog, the less
willing they would be to hold the Blue Bus Company
responsible.

The predicted main effect for order, however, could
occur because of a simple contrast effect. It could be that
considering the Blue Bus Company’s likely (80%)
responsibility after the unlikely (20%) Gray Bus Company
could increase rulings against the Blue Bus Company merely
because of the contrast of probabilities. To test this
notion, additional participants read a separate case prior to
reading either the complete-match or partial—match case. This
separate case involved a man who was suing a vitamin company
because he suffered seizures after using the company’s
product. Testimony in the case revealed that there was only a
15% chance that the product could have caused seizures. If
considering the Gray Bus Company first raises rulings against

the Blue Bus Company because of a contrast effect, the

32

consideration of a separate, unlikely case should also
increase ruling against the Blue Bus Company.
Mathod

Participants

Participants were 87 undergraduate psychology and
communications students who received class credit for taking
part in the study. The mean age of participants was 21.2
years and 72% were female.
Design and Procedure

Participants were told that they would be taking part in
a study on juror decision making and that they would act as
jurors in a case and make several judgments. Half of the
participants read the partial-match case, as in Experiment 1.
Again, the critical testimony in the partial-match case read

as follows:

A second county transportation official, who was the
transportation department’s chief investigator, took the
stand, was sworn in as a witness, and reported that he
examined the dead dog and took prints of the tire
tracks. These prints were then transferred onto paper
and compared to all 10 of the 10 buses owned by the Blue
Bus Company, and all 10 of the 10 buses owned by the
Gray Bus Company. He testified that the technique used
for matching produced a partial match with one of the
Blue Bus Company buses and a partial match with one of

the Gray Bus Company buses. The results of the test

33

indicated that there was an 80% chance that the blue bus
was the one that ran over the dog, and a 20% chance that
the gray bus was the one. Based on this matching
evidence, the official concluded that it was a blue bus

that killed the dog

The other half of the participants read the complete-
match case. The complete-match case was identical to the
partial-match case except for the fourth paragraph which

reads as follows:

A second county transportation official, who was the
department's chief investigator took the stand, was
sworn in as a witness, and reported that he examined the
dead dog and took prints of the tire tracks. These
prints were then transferred onto paper and compared to
all 10 of the 10 buses owned by the Blue Bus Company,
and all 10 of the 10 buses owned by the Gray Bus
Company. The tracks matched 8 of the 10 Blue Bus
Company's buses and 2 of the 10 Gray Bus Company’s

buses.

Participants were not told that Mrs. Prob was suing the
Blue Bus Company. They were merely told that they were going
to read about a civil case in which a woman’s dog was killed.
After reading the trial, participants indicated whether or

not they would rule against the Blue Bus Company and make

34

them pay damages to Mrs. Prob and estimated the probability
that the Blue Bus Company actually killed the dog. As in
Experiment 1, confidence and ease—of-simulation measures were
also collected. Participants made the same judgments about
the Gray Bus Company. The measures for the different bus
companies were on separate sheets and participants considered
either the Blue Bus Company first or the Gray Bus Company
first.

Approximately one third of the subjects participated in
the contrast control conditions. These participants first
read a case in which a man was suing the Herbrite Vitamin

Company. The text of this control case read as follows:

Mr. Smith had a very severe chest cold. His lungs were
congested and he was coughing. In order to alleviate his
symptoms, he took a herbal medicine that he bought at a
local nutrition store. The medication, manufactured by
the Herbrite Company, claimed it could help alleviate
chest cold symptoms. Approximately six hours after he
took the medicine, he began having respiratory seizures

and was rushed to the hospital.

The doctor who treated Mr. Smith testified that the
patient was placed on a respirator and given medication
to control the seizures. He further testified that Mr.

Smith would likely have permanent damage to his lungs.

35

An independent biochemist conducted tests on the herbal
medicine and Mr. Smith and concluded that there was only
a 15% chance that the herbal medicine could have caused
the seizures. In other words, he stated that there was
an 85% chance that the seizures were caused by something

other that than the herbal medicine.

Participants who read this control case uniformly
reported that the probability that the herbal medicine caused
Mr. Smith’s seizures was quite low (M = 29%). Furthermore,
most participants in the control conditions reported that
they would not rule against the vitamin company (83%). After
reading this case, participants read either the complete—
match or partial-match case and filled out the dependent
measures for only the Blue Bus Company.

Resolts

Table 1 reveals the verdicts against the Blue Bus
Company, subjective probabilities, verdict/confidence means,
and ease-of-simulation means for all conditions.

Initially, the verdicts against the Blue Bus Company in
the complete—match version were compared to those in the
partial—match version only for the participants who
considered the Blue Bus Company first. This test was
conducted as a replication of the basic contrasts in previous
research (Niedermeier et al., 1999: Wells, 1992). As
expected, participants who read the complete-match condition

ruled against the Blue Bus Company significantly less than

36

Table 1
Verdicts, Subjective Probabilities. VerdithConfidence. and
Ease of Simulation by Condition (Ezoeriment 2)

 

 

 

 

 

Blue First Gray First Control
DV Complete Partial Complete Partial Complete Partial
n 19 11 15 13 13 16
V 21% 64% 40% 54% 31% 44%
SP 74. 82.6 75.3 78.3 77.3 75.0
'V/C 7.7 13.4 9.7 12.0 10.6 11.2
EOS 6.1 7.2 6.1 7.6 5.8 6.8

 

Note. The numbers refer to percentage of verdicts against the

Blue Bus Company (V), mean subjective probability (SP), mean

verdict/confidence (V/C),

and mean ease-of-simulation (EOS).

Higher numbers on the verdict/confidence measure represent

more confident verdicts against the Blue Bus Company. Higher

numbers on the ease-of-simulation measure represent more

reported difficulty.

37

those who read the partial-match version, x2(1, N = 30) =
5.44, p < 05, ¢ = .42 (z = 2.46, p < .01, one—tailed). This
pattern did not hold when the Gray bus company was considered
first, as participants ruled against the Blue Bus company at
the same rate in the complete—match condition as the partial—
match condition, x2(1, n = 28) = .54, _s, 0 = .14.

Next, the pattern of verdicts when the participants
considered the Blue Bus Company first was compared to the
pattern of verdicts when the participants read the control
case prior to ruling on the Blue Bus Company. If contrast
with a very low probability case is enough to raise verdicts
against the Blue Bus Company, a main effect should emerge
such that participants who first read the control case should
rule against the Blue Company more often than those who did
not. However, participant ruled against the Blue Bus Company
at the same rate when they read the control case (38%) as
when they did not (37%), x2(1, u = 59) = .01, g, 0 = .00.
The data from the control conditions were not considered in
any further analyses.

The verdicts when participants considered the Blue Bus
Company first were then compared to the conditions in which
the Gray Bus Company was considered first. A main effect was
expected such that participants who considered the Gray Bus
Company first should rule against the Blue Bus company at a
higher rate. Although the pattern was in the predicted
direction, participants ruled against the Blue Bus Company at

statistically the same rate when they considered the Gray Bus

38

Company first (46%) as they did when they considered the Blue
Bus Company first (37%), x2(1, u = 58) = .56, as. o = .10, (z

= .70, _a). Next, the verdicts for those who considered the
Blue Bus Company first were compared to those who considered
the Gray Bus Company first only for participants who read the

complete-match case. Although the trend was in the predicted
direction, the effect was not significant, x2(1, M = 58) =
1.45, go, o = .21, (z = 1.20, go).

There was a main effect for case such that participants
who read the complete—match case ruled against the Blue Bus
Company significantly less (29%) than those who read the
partial—match case (58%), x2(1, a = 58) = 4.85, n < 05, ¢ =
.29 (z = 2.20, o < .02, one-tailed). Case version and order
did not interact, 12(1, u = 58) = 1.38, ns, 0 = .15.

Considering the possibility of a ceiling effect for
verdicts against the Blue Bus Company in the partial-match
conditions, it is plausible that the only cell that would
have a low conviction rate is the complete-match case when
Blue is considered first. Convictions rates should be high in
both partial—match conditions, so there may not have been
room for the partial-match conviction rate to get any higher
when Gray was considered first. Also, it was predicted that
conviction rates should be high in the complete-match when
Gray is considered first. Therefore, it is reasonable to
compare the complete—match case when Blue was considered

first to the other three cells. Participants who considered

Blue first and read the complete-match case in fact ruled

39

against the Blue bus Company significantly less than

participants in the other three conditions, x2(1, M = 58) =

4.81, p_ < .05, q) = .29.

The combined verdict/confidence measures revealed a
similar pattern to verdicts in that a main effect was
revealed such that participants who read the partial-match
case reported more pro-conviction verdicts against the Blue
Bus Company (M = 12.63) than those who read the complete-
match case (M = 8.89), 31,54) = 4.27, e < .05, m2 = .06. No
other significant effects were revealed on this measure.

As Table 1 shows, subjective probabilities did not
differ and were appropriately high across conditions.
However, a significant relationship between the subjective
probability and verdicts was revealed such that participants
who reported lower subjective probabilities were less likely
to rule against the Blue Bus Company, ;(56) = .28, o < .05.

Next the ease—of—simulation measure was evaluated. Those
who read the partial-match case reported that is was more
difficult to think of a situation in which the Gray Bus
Company could be responsible for killing the dog (M = 7.33)
than those who read the complete—match case (M = 6.10),
F(1,54) = 4.64, o < .05, d9 = .06. Ease-of—simulation was
unaffected by order. Furthermore, a significant relationship
between the ease-of—simulation measure and verdicts was
revealed such that the easier participants reported it was to
simulate that a Gray Bus Company bus may have killed the dog,

the less willing they were to hold the Blue Bus Company

40

responsible, £(56) = .33, p < .02. Additionally, ease-of-
simulation still accounted for a significant amount of
variance in the verdicts after controlling for the effects of
subjective probability, C(56) = 2.83, o < .01, 32 = .09.
There was no relationship between ease-of—simulation and
subjective probabilities, ;(56) = .15, na.

Lastly, the intervening role of ease—of-simulation was
tested by entering case version, ease-of-simulation, and
verdicts into a general linear model. When the effects of

ease—of-simulation were controlled, the significant effect of

case version (o .03) was reduced below the level of

significance (p .10). Furthermore, the effect size of case

version ((1)2 = .06) was reduced by 50% ((1)2 = .03).
Diocusaion

First, the basic pattern of findings in previous
research (Niedermeier et al., 1999: Wells, 1992) was
replicated. Participants who read the complete-match
condition ruled against the Blue Bus Company significantly
less than those who read the partial-match version. MOreover,
participants’ ability to simulate an alternative hypothesis
(that a gray bus killed the dog) was related to belief in the
primary hypothesis (that a blue bus killed the dog) across
conditions.

Furthermore, it was demonstrated that ease-of—simulation
plays an intervening role between case version and verdicts.
When the effects of ease—of-simulation were partialed out of

the model, the effect of case on verdicts was no longer

41

significant. This finding is suggestive that ease—of-
simulation mediated verdicts.

Unlike Experiment 1, subjective probabilities did not
differ across conditions and were appropriately high across
conditions. This pattern is consistent with previous research
(Niedermeier et al., 1999: Wells, 1992). There was an
unexpected relationship between subjective probability and
verdicts, but ease—of-simulation was still a significant
predictor of verdicts after controlling for the effects of
subjective probability. Furthermore, there was no
relationship between subjective probability and ease-of-
simulation, indication that both mechanisms independently
affected verdicts. Overall, these findings further suggest
that continuous and categorical measures do not necessarily
produce parallel results and are not necessarily measuring
the exact same thing.

The main hypothesis in this experiment was not supported
in that verdicts against the Blue Bus Company were not higher
when participants considered the Gray Bus Company first.
However, the pattern of verdicts were in the predicted
direction and may have become evident with additional power.
Given that the potential effect size is small, power analysis
revealed that the ability to detect the effect with the
current sample size would be less than .10 (Cohen, 1988). It

would take a sample of 400 to detect the effect at o = .05

with power of .50.

42

However, two findings did indicate that considering the
alternative hypothesis first may have had an effect. First,
more participants ruled against the Blue Bus Company in the
complete-match condition than the partialematch condition
when Blue was considered first, but there was no difference
when Gray was considered first. Thus, considering Gray first
did reduce the effect of case, despite the fact that a
significant interaction failed to emerge. Second, the
contrast of the complete-match condition when Blue was
considered first to the other three conditions suggests that
the consideration of Gray did affect verdicts by raising the
conviction rate in the complete-match condition when Gray was
considered first.

Additionally, a pilot study was conducted which was
exactly the same as Experiment 2, but did not include ease-
of-simulation measures. This pilot revealed the predicted
main effect for order such that participants who considered

Gray first ruled against the Blue Bus Company significantly

less (26%) than those who considered Blue first (45%), x2(1,
M = 126) = 4.63, o < 05, ¢ = .19 (z = 2.25, o < .02, one—
tailed). The unusual fact that the pilot study had more
participants than the actual experiment was due to an
unforeseen shortage of subjects at the time the experiment
was conducted. The combined order effect for the pilot and
Experiment 2 was 2 = 2.09, p < .02.

Experiment 2 clearly replicated the basic contrast

between the easily simulated alternative in the complete—

43

match case and the less easily simulated alternative in the
partial—match case. Second, the relationship between ease-of—
simulation and verdict was again demonstrated and ease-of-
simulation was shown to play a mediating role. Lastly,
Experiment 2, when combined with the data from the pilot
study, provided some suggestive evidence that eliminating an
hypothesis by considering it first and categorically
rejecting it may increase acceptance of the primary

hypothesis.

Experiment 3

The research by Miller et al. (1989) suggests that
increasing the number of alternatives will undermine belief
in a given hypothesis. In their study, the proportion of the
alternatives is constant (10%), only the absolute number is
changed. For instance, the existence of 10 chocolate chip
cookies reduced subjects’ suspicion that the child peeked in
the cookie jar compared to when there was one chocolate chip
cookie. Similarly, other research (Fischhoff et al., 1978;
McDonald, 1998) also suggests that the likelihood that people
will accept a given explanation for an event is directly
related to the number of available alternatives. Again, none
of these studies collected categorical measures.

Alternatively, the Niedermeier et al. (1999) study
suggests that a single coherent alternative is enough to
undermine the primary hypothesis. In their study, the lone

possibility of a second bus company was enough to reduce

44

rulings against the Blue Bus Company in the complete-match
condition. Additional alternatives would not necessarily
enhance ease of simulation, especially if the total
probability of all alternatives were held constant.

In Experiment 3, participants read an adaptation of the
complete—match trial in which the Blue Cab Company was being
sued for killing Mrs. Prob’s dog. There were either one or
five other cab companies that produced matches to the tire
tracks at the scene. Cabs were used instead of buses in this
experiment because it is more plausible that multiple cab
companies would operate in one area than multiple bus
companies. Additionally, some subjects read a version in
which they were aware of five alternative matches, but were
unaware to which company the matches belonged. In this study,
participants were told that there were 100 cabs that matched
the tire tracks at the scene. Ninety—five Blue Cab Company
cabs matched the tracks, as did five cabs from other
companies. The overall probability of a blue cab being
responsible was 95% in this experiment (compared to 80% in
the previous studies) in order to avoid a possible floor
effect and better differentiate between the conditions. This
was deemed necessary because pretests in which 80% of the
implicated cabs were blue produced very few findings against
the Blue Cab Company in any condition.

If increasing the absolute number of alternatives
reduces belief in a primary hypothesis, participants should

have ruled against the Blue Cab Company less when there were

45

five alternative cab companies that produced matches than
when there was only one. Research would suggest that
increasing alternative routes to a particular outcome should
reduce belief in given hypothesis (Fischhoff et al., 1978;
McDonald, 1998; Miller et al., 1989). However, if a single
easily simulated alternative is all that is necessary,
rulings against the Blue Cab Company should be equally low in
all three conditions. As in Experiments 1 & 2, a relationship
was expected such that the easier participants report it is
to simulate that an alternative cab company may have killed
the dog, the less willing they would be to hold the Blue Cab
Company responsible. Furthermore, no differences were
expected on the continuous measure of subjective probability.
Mothod

Paroicioants

Participants were 141 undergraduate psychology students
who took part in the study as a classroom demonstration. The
mean age of participants was 20.0 years and 74% were female.
Design and Procedore

Participants were told that they would be taking part in
a study on juror decision making and that they would act as a
juror in a civil case and make several judgments.
Approximately one third of the subjects read a case entitled
the one—alternative version. The case was the same as the
case in Experiment 1, except that it referred to cabs instead

of buses. The critical fourth paragraph read as follows:

46

A second county transportation official, Mr. Jones, took
the stand, was sworn in as a witness, and reported that
he examined the dead dog and took prints of the tire
tracks. The tire tracks were checked against all cabs
that operate out of the central dispatching station on
County Road #37. Mr. Jones produced a table that showed
the results of his investigation. As the table shows,
the prints matched 95 of the cabs owned by the Blue Cab
Company and 5 of the cabs owned by the Gray Cab Company.
None of the cabs from the other companies produced a

match [see Appendix A].

Approximately one third of the participants read the
fiyo-aeoarate-alternatiyes version. The only difference in
this case was the critical fourth paragraph which read as

follows:

A second county transportation official, Mr. Jones, took
the stand, was sworn in as a witness, and reported that
he examined the dead dog and took prints of the tire
tracks. The tire tracks were checked against all of the
cabs that operate out of the central dispatching station
on County Road #37. Mr. Jones produced a table that
showed the results of his investigation. As the table
shows, the prints matched 95 of the cabs owned by the
Blue Cab Company and one cab from five other cab

companies (that is, one match each from the Gray,

47

Yellow, Red, Orange, and Purple Cab Companies) [see

Appendix A].

The remaining third of the participants read the fixa—
grogood-alternatiyoa version. The only difference in this
case was the critical fourth paragraph which will read as

follows:

A second county transportation official, Mr. Jones, took
the stand, was sworn in as a witness, and reported that
he examined the dead dog and took prints of the tire
tracks. The tire tracks were checked against all of the
cabs that operate out of the central dispatching station
on County Road #37. Mr. Jones produced a table that
showed the results of his investigation. As the table
shows, the prints matched 95 of the cabs owned by the
Blue Cab Company and 5 of the cabs owned by five other

cab companies [see Appendix A].

After reading the trial, participants indicated whether
or not they would rule against the Blue Cab Company and make
them pay damages to Mrs. Prob and estimated the probability
that the Blue Cab Company actually killed the dog. As, in
Experiment 1 and 2, confidence measures were also gathered.
To measure ease-of—simulation, participants were asked to
rate how difficult it was to think of a way that a cab other

than one from the Blue Cab Company could have been

48

responsible for killing the dog on an 11-point scale from
very easy (1) to very difficult (11).
Raoults
Table 2 reveals the results for Experiment 3. Rulings

against the Blue Cab Company did not differ among conditions,

x2(2, N = 141) = 1.63, go, o = .10. Likewise, there were no

differences among conditions on the verdict/confidence

measure, F(2,138) = .73,

s, «9 = .00. Also, subjective
probabilities did not differ and were appropriately high
across conditions, £(2,138) = 1.13, MS, «Q = .00.
Furthermore, there was no relationship between subjective
probability and verdicts, ;(139) = .15, ES-

Participants who read the one-alternative version
reported that it was more difficult to think of a way that a
cab from a company other than the Blue Cab Company could have
been responsible for killing the dog than those who read the
five—separate-alternatives version or the five-grouped-
alternatives version, F(2,138) = 12.41, p,< .001, «fl = .14.

Finally, a significant relationship between the ease-of-
simulation measure and verdicts was revealed such that the
easier participants reported it was to simulate that a cab
from a company other than the Blue Cab Company may have
killed the dog, the less willing they were to hold the Blue
Cab Company responsible, r(139) = .34, p < .001. There was no
relationship between ease-of—simulation and subjective

probability, £(139) = .11, ns.

49

Table 2

 

 

-roi . b" i ‘ ’r0019'_: i‘- ‘r01 _ - fiq‘n.‘ : d

Eas 0 im la i n r'm n
Alternatives

Meaaure One Five Five grouped
n 48 47 46
V 29% 36% 24%
SP 90.9 90.0 85.9
V/C 9.00 9.94 8.09
EOS 6.46a 3.77b 4.15b

 

Moooo The numbers refer to percentage of verdicts against
the Blue Bus Company (V), mean subjective probability (SP),
mean verdict/confidence (V/C), and mean ease-of-simulation
(EOS). Higher numbers on the verdict/confidence measure
represent more confident verdicts against the Blue Bus
Company. Higher numbers on the ease-of-simulation measure
represent more reported difficulty. Means with different

subscripts are different at o < .01.

50

Di i n

Rulings against the Blue Cab Company were equally low in
all conditions, indicating that any coherent alternative is
enough to suppress verdicts of liability. Across all three
conditions, participants ruled against the Blue Cab Company
at a rate of 30%. This seems quite low when compared to the
partial—match cases in Experiment 2 and Niedermeier et a1.
(1999) in which participants ruled against the defendant at a
rate of approximately 55%. Also, keep in mind that the actual
probability of Blue’s responsibility was 80% in the previous
studies compared to 95% in the current experiment.

An informative condition would have been a partial—match
version of the cab case in which there were incomplete
matches with the Blue Cab Company and alternative cab
companies that resulted in a 95% chance the Blue was
responsible. This would have allowed a clear contrast with
the three existing conditions. Provided that the partial—
match condition was high, as in previous studies, this
contrast would bolster the conclusion that any number of
easily-simulated coherent alternatives would reduce
willingness to rule against the Blue Cab Company.

The differences between conditions on the ease-of—
simulation measure was unexpected. However, the measure in
this experiment concerned simulation of any company other
than blue, whereas the previous experiments specified
simulation of gray. Therefore, the object of simulation in

this experiment was not constant among the conditions. For

51

- annulus aw i. run. 13min!

 

instance, participants must pick out the Gray Cab Company to
answer the question in the one-alternative condition, but
they may consider any of five companies in the five—separate-
alternatives condition. Therefore participants in the one-
alternative condition must take the extra mental step of
separating the Gray Cab Company from the others.
Alternatively, participants in the one—alternative condition
may have mentally averaged across all of the available
alternatives, which included five companies with no matches.
In either case, this could account for the elevated ease-of-
simulation scores in the one-alternative condition. However,
the basic relationship between ease-of-simulation and
verdicts held in this experiment, despite the potential

ambiguity in the ease—of-simulation measure.

Experiment 4

Much of the research on lay hypothesis testing assumes
that reliance on the availability of alternatives to make
judgments is a heuristic (Kahneman & Tversky, 1982; McDonald,
1998; Niedermeier et al., 1999). Kahneman and Tversky dubbed
the phenomenon currently under study “the simulation
heuristic.” Additionally, McDonald claims that beliefs in
many extraordinary phenomena such as telekinesis and alien
crop circles are bolstered by the inability to easily
simulate alternative explanations. McDonald termed this

effect the “can—you—think-of-anything-else heuristic.”

52

This raises the question of whether the reduction of
belief in a primary hypothesis when alternative hypotheses
are easily simulated is indeed a heuristic rather than
inherently flawed, mindful processing. Popular dual—
processing models, such as Petty and Cacioppo's (1984, 1986)

elaboration likelihood model or Chaiken's (1987)

 

F
systematic/heuristic model of attitude change may lend some
insight. One might conceive of evidence as a persuasive
communication and presume that the recipients of this
communication will only process it systematically (i.e., I

carefully and thoroughly analyzing all available relevant
information) if they have both the capacity and the
motivation to do so. Otherwise, they might be expected to
process the communication heuristically, that is, to utilize
cues to simplify judgment and decision making.

In the Blue Bus Company case, it is plausible that mock
jurors process the available evidence heuristically.
Specifically, mock jurors may be using an easily simulated
alternative as a heuristic cue to find the Blue Bus Company
not liable. If this is true, the stronger their motivation to
consider the available evidence systematically, the less they
would be expected to rely upon such a heuristic cue. So given
motivation to process the case systematically, jurors would
be less willing to rely on the simulation heuristic and
become more likely to rule against the Blue Bus Company. If

participants process in a mindful fashion, verdicts should be

53

more in line with actual and reported subjective
probabilities.

In this experiment, participants read either the
standard complete—match case, in which there is an easily
simulated alternative hypothesis, or the partial-match case,
in which there is not an easily simulated alternative. Half
of the participants were told at the outset of the study
that, after they made their individual choices, they would be
asked to report and defend their private verdict choices in
front of the entire group in order to promote systematic
processing. Previous research (e.g., Petty, Cacioppo, &
Goldman, 1981) has demonstrated the efficacy of such a
"public accountability" treatment for increasing systematic
processing. The other half was given no such instructions and
should presumably rely more on mental simulation as in the
previous experiments. If mental simulation is indeed used as
a heuristic cue, an interaction is expected such that rulings
against the Blue Bus Company should be more common when
jurors are cognitively engaged than when they are not or when
mental simulation is difficult. When engagement was low, the
original pattern of few rulings against blue in the complete-
match should have been observed. Alternatively, if mental
simulation is part of an intrinsically flawed reasoning
process, then verdicts against the Blue Bus Company should be
lower in the complete-match cases than the partial-match

cases, regardless on engagement.

54

Method

Partioioanga

Sixty introductory psychology students received extra
class credit for taking part in the study. Participants were
run in groups of three to thirteen. Group size showed no
effects on any individual measures. The mean age of
participants was 20.1 years and 83% were female.
Design and Procedoro

The cases used in this experiment were the standard
complete-match and partial-match cases used in Experiment 2.
Crossed with this treatment was a cognitive engagement
(motivation to process systematically) treatment. All
participants were told that they would be taking part in a
study on juror decision making and that they would act as
jurors in a case and make several judgments. Half of the
participants were told at the outset of the study that after
they made their individual choices, they would have to report
and defend their private verdict choices in front of the
entire group. The other half was given no such instruction.

Thus, the presence of an easily simulated alternative
(complete-match) or less easily simulated alternative
(partial—match) was crossed with the cognitive engagement
manipulation (high engagement vs. low engagement) to yield a
2 X 2 factorial design.

After reading the trial, participants indicated whether
or not they would rule against the Blue Bus Company and make

them pay damages to Mrs. Prob and reported subjective

55

probability by estimating the likelihood that the Blue Bus
Company actually killed the dog. As in the previous
experiments, confidence and ease-of-simulation measures were
also gathered. In addition, participants responded to an
open—ended question that asked them to explain how they
reached their verdict.

Rim

Table 3 reveals the verdicts against the Blue Bus
Company, subjective probabilities, verdict/confidence means,
and ease—of—simulation means for all conditions.

Initially, the verdicts against the Blue Bus Company in
the complete—match version were compared to those in the
partial—match version only for the participants who did not
receive the engagement instructions. This test was conducted
as a replication of the basic contrasts in previous research
(Experiment 2, Niedermeier et al., 1999: Wells, 1992). An
effect was revealed such that participants who read the

complete-match condition ruled against the Blue Bus Company

less than those who read the partial—match version, x2(1, M =
31) = 3.48, o < 07, ¢ = .33 (z = 1.96, o < .05). As Table 3
reveals, the pattern was the same in the high engagement
condition, x2(1, n = 31) = 6.43, o < 02, ¢ = .46 (z = 2.70, p
< .01). When participants in both engagement conditions were
considered, those who read the complete—match condition ruled

against the Blue Bus Company significantly less (14%) than

those who read the partial-match version (52%), x2(1, N = 60)

= 9.64, p < .001, ¢ = .40 (z = 3.33, p < .001).

56

Table 3

 

 

 

 

eroi _- . o'-ct' é P o-s;7_' '- ero' - 0nf:!‘! - .nd

Ease of Simulation My Conditioa (Ezparimant 4)
Complete-match Partial-match

Measure High Low High Low
n 14 15 15 16
V 14% 13% 60% 45%
SP 77.9 73.3 74.3 69.1
V/C 6.1 7.1 13.1 11.0
EOS 3.7 5.0 5.5 5.1

 

Mogao The numbers refer to percentage of verdicts against the
Blue Bus Company (V), mean subjective probability (SP), mean
verdict/confidence (V/C), and.mean ease-of-simulation (EOS).
Higher numbers on the verdict/confidence measure represent
more confident verdicts against the Blue Bus Company. Higher
numbers on the ease-of-simulation measure represent more

reported difficulty.

57

As Table 3 also reveals, participants ruled against the
Blue Bus Company at statistically the same rate when they
received engagement instructions (38%) as when they did not
(29%), x2(1, u = 60) = .53, as, q) = .09. Additionally, there
was no interaction between case type and engagement
instructions, x2(1, M = 60) = .32, Ma, ¢ = .07.

There were no significant effects for subjective
probability, as estimates were appropriately high across
conditions. There was no relationship between subjective
probability and verdicts, £(58) = .02, na.

The verdict/confidence measure revealed a main effect
such that participants in the partial—match conditions had
more pro-conviction verdicts against the Blue Bus Company (M
= 12.03) than participants in the complete-match conditions
(14. = 6.62), E(1,56) = 8.62, 19 < .01, (02 = .12. There were no
other significant effects on this measure.

There were no main effects or interaction involving the
ease—of-simulation measure. However, there was a significant
relationship between the ease—of—simulation measure and
verdicts such that the easier participants reported it was to
simulate that a bus from the Gray Bus Company may have killed
the dog, the less willing they were to hold the Blue Bus
Company responsible, £(58) = .29, o < .05. There was no
relationship between ease-of—simulation and subjective
probability, £(58) = .05, oa.

The open—ended question was coded for consideration of

the Gray Bus Company. Participants were scored as either

58

considering The Gray Bus Company in their decision or not.
Any mention of the Gray Bus Company (or a reference to the
“other” company) was scored as considering Gray. Table 4
reveals the percentage of participants who considered Gray in
each condition. There was a non-significant trend that
suggested participants were more likely to consider Gray when
they received engagement instructions (69%) than when they
did not (52%), x2(1, a = 60) = 1.88, Ma, 0 = .18.
Participants who read the complete—match case considered the
Gray equally under both engagement conditions, x2(1, M = 60)

= .01, go, o = .00. However, participants who read the

partial—match case were more likely to consider Gray when
they received engagement instruction than when they received
no such instructions, x2(1, M = 60) = 3.89, o < 05, ¢ = .25
(z = 2.02, o < .05). Although these findings suggest an
interaction, the trend did not reach significance, x2(1, M =
60) = 2.42, p = .20, ¢ = .33 (z = 1.54, p < .13).
Discussion

First, the basic pattern of findings in previous
research (Niedermeier et al., 1999: Wells, 1992) and in
Experiment 2 was replicated in that participants were more
willing to rule against the Blue Bus Company when they read
the complete-match case than when they read the partial-match
case. Furthermore, increased ability to simulate an
alternative hypothesis (that a gray bus killed the dog) was
again reliably related to belief in the primary hypothesis

(that a blue bus killed the dog).

59

Table 4

 

 

 

 

Percontaga of Pagtioioaats who Considered Cray
Complete—match Partial-match

Engagement

High ‘ 79 60

Low 80 25

 

60

Additionally, subjective probability was consistently
high in all conditions and showed no relationship with
verdicts, indicating that continuous and categorical measures
do not necessarily produce parallel results.

The analysis of the open ended question did suggest that
participants were more likely to consider an alternative
hypothesis when they received engagement instructions than
when they did not. This pattern was evident for participants
who read the partial-match case. Consideration of Gray was
high in both engagement conditions for those who read the
complete—match case, which stands to reason as simulating
Gray in this case has been shown to be easy (Experiment 2,
Niedermeier et al., 1999). These results suggest that the
engagement manipulation did motivate participants to process
more systematically and consider more of the evidence.

A potentially informative condition might be to limit
engagement by making participants cognitively busy. This
manipulation may limit participants ability to simulate the
alternative hypothesis. Without the ability to simulate
alternatives, participants may rely more heavily on the
probabilistic information and rule against the Blue Bus
Company in the complete-match condition.

Overall, participants in Experiment 4 ruled against the
Blue Bus Company no differently when they were highly engaged
than when they were less motivated to consider the evidence.

This pattern suggests that the basic difference in verdicts

61

is not solely due to reliance on a heuristic. Rather, it
suggests that the verdict differences are the result of
inherently flawed reasoning, as the pattern held when
engagement was high and systematic processing was likely.
However, it should be noted that the sample size was small

limiting the power the detect any potential effect.

General Discussion

This study began with four objectives. First, the notion
that mental simulation plays a role in casual decision making
was tested by making alternative hypotheses more or less
easily simulated. The trend in Experiment 1 and significant
main effects for case version in Experiments 2 and 4 support
the notion that an easily simulated alternative reduces
willingness to accept a primary hypothesis. Furthermore, all
four experiments revealed significant relationships between
ease-of—simulation and verdicts. Additionally, the effects of
ease-of-simulation on verdicts was consistently shown to be
independent of subjective probability estimates.

Second, the importance of categorically eliminating
potential alternatives was investigated by altering the order
in which participants considered various alternative
explanations. The results of Experiment 2 (along with the
pilot data) suggest that considering one alternative first
and categorically rejecting it increased willingness to

accept an hypothesis presented subsequently.

62

Third, the relative frequency of alternative
possibilities were altered to determine if one coherent
alternative is sufficient to reduce belief in the primary
hypothesis as opposed to several alternatives. The results of
Experiment 3 suggest that a single coherent alternative
hypothesis is sufficient to reduce belief in a primary
hypothesis. Increasing the number of alternatives beyond a
single one did not have a significant effect on the
willingness to accept the primary hypothesis.

Lastly, the heuristic nature of mental simulation was
investigated by manipulating participants’ motivation to
consider alternatives in order to determine if mental
simulation is simply a cue or a systematic but intrinsically
flawed method of reasoning. Experiment 4 demonstrated that
verdicts were not altered by increasing motivation to
carefully process evidence, indicating that the use of mental
simulation is not so much a reliance on a heuristic as it is
intrinsically flawed reasoning.

Although manipulating moderators of simulation proved
more difficult than expected across experiments, it was clear
in all four experiments that ease-of-simulation was related
to the acceptance of a primary hypothesis. More specifically,
the easier it was to simulate that a Gray Bus Company bus
killed the dog (the alternative hypothesis), the less willing
participants were to hold the Blue Bus Company responsible
(the primary hypothesis). The combined effect of ease-of—

simulation for all four experiments was L = .35, indicating a

63

significant and reliable relationship. Whatever made
simulation easy or difficult for individuals across
experiments clearly related to verdicts and was independent
of any effect of subjective probability.

Furthermore, the results of Experiments 2 and 4 directly
replicated previous work in this area (Niedermeier et al.,
1999, Wells, 1992). Participants who read the complete—match
case, in which simulation of the alternative was easy, were
more likely to rule against the Blue Bus Company than those
who read the partial-match case, in which simulation was more
difficult.

Although ease-of—simulation is theoretically viewed as
driving verdicts in this context, it could be argued that
participants were simply justifying their verdicts through
the ease-of—simulation measures. However, this argument seems
implausible for a number of reasons. First, when ease-of-
simulation was controlled for in Experiment 2 and Niedermeier
et al. (1999), the effect of case version on verdicts was
reduced, indicating that ease-of-simulation indeed plays an
mediating role. Second, it seems unlikely that participants
in experiments such as these would feel so compelled to
justify their verdicts that they would do so on a measure
that is not obviously related to the verdict. If
justification were a motivating factor, one would think this
would become evident on the subjective probability measure,

which is more clearly linked to the verdict.

64

Another contribution of this study is the demonstration
that continuous and categorical measures do not necessarily
produce parallel results. In much of the previous work
(Dougherty et al., 1997; Fischhoff et al., 1978; McDonald,
1998; Robinson & Hastie, 1985; Robinson Van Wallendael,

1989), only continuous measures of belief in the primary

hypothesis were collected. Some of these studies showed
effects of alternative hypotheses and other did not. The
current experiments demonstrate that continuous subjective

probabilities do not necessarily translate into like

 

categorical responses, as subjective probabilities had no
relationship with categorical verdicts in Experiments 3 and
4. There was such a relationship in Experiments 1 and 2, but
it was shown to be independent of the effects of mental
simulation on categorical verdicts. These findings, along
with those of Niedermeier et al. (1999) and Wells (1992),
converge to show that a seemingly precise continuous measure
may fail to capture a psychological effect evident in a
dichotomous categorical measure, at least in cases in which
there is a clear primary hypothesis.

Methodologically, it is important to recognize the non—
parallel nature of the measures. Clearly, studies that look
at continuous or categorical judgments alone may be missing
an important piece of the decision making puzzle.
Furthermore, a dichotomous measure, in some instances, may
actually be more precise and meaningful than a continuous

subjective probability. For instance, decisions to vote

65

guilty or not guilty, to marry or not marry, or to buy or not
buy, are the types of categorical choices that have meaning
and impact in our lives, regardless of our subjective
probability estimates.

As alluded to earlier, many of the models of juror
decision making fail to adequately consider the power of
alternative explanations and the non-parallel nature of
continuous and categorical measures. For instance, the story
model (Pennington & Hastie, 1992, 1993) assumes that one
story will emerge as the best and the other stories merely
will be dismissed. However, the current research suggests
that the power of a given explanation relies in large part on
the presence and ease—of—simulation of alternative
explanations. It stands to reason that the story model could
benefit by considering ease-of—simulation of alternative
stories.

Additionally, threshold models of decision making (e.g.,
Dane, 1985; Kaplan, 1977) generally view continuous
subjective probability as driving categorical verdicts. These
models maintain that if subjective probability surpasses some
psychological threshold of guilt, jurors will rule against
the defendant. This approach fails to capture the pattern of
results found in this study and previous research in which
categorical verdicts changed while continuous subjective
probabilities remained static (Niedermeier et al., 1999,
Wells, 1992). A way in which these findings could be

rectified with the family of threshold models is if mental

66

simulation affects the level of the threshold criterion. It
may be that the ability to easily simulate an alternative
hypothesis raises the threshold that subjective probability
must surpass in order to rule against a defendant. Thus,
subjective probability may more accurately predict verdicts
if the effect of mental simulation of alternatives on
thresholds could be assessed. This is a proposition that
merits further investigation.

Although the paradigm used in this study has many merits
and has been an effective tool for studying causal decision
making, it may inject some unwanted elements into the
process. For instance, the fact that there are legal
standards of liability that are consider by many participants
may create a default verdict of not liable. Although there
are likely subjective thresholds and defaults involved in any
decision, the legal standards inherent in this paradigm may
limit generalizability. Another problem with the current
study is power. Some of the trends that are suggested by the
data may have been clearer and emerged as significant given
larger sample sizes. Future research should attempt to both
replicate these findings with greater power and generalize

the ease—of—simulation phenomenon using non-legal scenarios.

Conclusion
The purpose of the present investigation was to explore
how the presence and properties of alternative hypotheses

affect belief in a primary hypothesis. Across experiments,

67

participants generally ruled against the defendant less when
they were able to mentally simulate an alternative
explanation than when simulation was more difficult. In other
words, the ability to simulate an alternative hypothesis was
associated with reduced belief in the primary hypothesis.
However, the factors that influence individuals’ ability to
simulate alternative hypotheses remain unclear. Further work
is needed to identify specific moderators of mental

simulation and refine current models of decision making.

68

APPENDIX

69

APPENDIX A

STIMULI USED IN EXPERIMENT 3

One—Alternative yersion

Cab C Total # of matches

Blue
Gra
Yellow

Red
Or e
le

 

Five—Separate-Alternatives Version

Cab C Total # of matches

Blue
Gra
Yellow
Red
Oran e

le
Green

 

Five—Crouoed-Altornatives Version

Cab C Total # of matches

Blue 95
Gra
Yellow

Red
e
le

 

70

 

 

 

ace-en's EMF .. H ..- ,

(61': b... V

 

 

___1 ._ ,_.__. _.._..._.—..

REFERENCES

71

REFERENCES

Chaiken, S. (1987). The heuristic model of persuasion.

In M.P. Zanna, J.M. Olsen, & C.P. Herman (Eds.), Sootal

influenoa: The Qntario sympooium (Vol. 5). Hillsdale, NJ:
Erlbaum.
Cohen, J. (1988). Stattatioal ooflar analyais for tho

behavioral aoionoea (2nd ed.). Hillsdale, NJ: Erlbaum.

Dane, F. (1985). In search of reasonable doubt: A
systematic examination of selected quantification approaches.
Law and Human Behayior, 9, 141-158.

Doherty, M.E., Chadwick, R., Garavan, H., Barr, D., &
Mynatt, C.R. (1996). On people’s understanding of the
diagnostic implications of probabilistic data. Mamo;y_aod
Cognition, 24, 644—654.

Dougherty, M.R.P., Gettys, C.F., & Thomas, R.P. (1997)

The role of mental simulation in judgments of likelihood.

Crganizational Sehayio; and Hagan Daoision Etooaasaa. 70,
135—148.

Einhorn, H.J., & Hogarth, R.M. (1986). Judging probable
cause. Paychologioal Bollatin. 99, 3-19.

Fischhoff, B. (1975). Hindsight foresight: The effect
of outcome knowledge on judgment under uncertainty. Qogrnal

of Eagerimental Psyohology: Human Percaotion and Parformance.
to 288-299.

72

Fischhoff, B., Slovic, P. & Lichtenstein, S. (1978).
Fault trees: Sensitivity of estimated failure probabilities
to problem representation. Qoprnal of Exporimontal
Psychology: Human Perooption and Performance, 4, 330—344.

Greene, E., & Loftus, E.G. (1984). What's new in the
news? The influence of well—publicized news events on
psychological research and courtroom trials. Basio and
Applied Social Psychology, 5, 211-221.

Greene, E. (1990). Media effects on jurors. Lag apd
Human Bohavior, 14, 439-450.

Greene, E., & Wade, R. (1988). Of private talk and
public print: General pre-trial publicity and juror decision—
making. Appliod Cogpitiye Psyohology. 2. 123—135.

Hastie, R. (1993). Inoido tho jorot; Tho psyohoiogy of
juror deoioion making. New York: Cambridge University Press.

Hawkins, S.A., & Hastie, R. (1990). Hindsight: Biased
judgments of past events after the outcomes are known.
Psychological Bollotin, 1Q7, 311-327.

Hilton, D.J. (1995). The social context of reasoning:
Conversational inference and rational judgment. Psyohologioal
Bulletin, 118, 248-271.

Hirt, E.R., & Markman, K.D. (1995). Multiple
explanations: A consider-an—alternative strategy for

debiasing judgments. n l P al' cial

Psychology. 69, 1069-1086.

73

 

 

Johnson, H.M., & Seifert, C.M. (1994) Sources of the

continued influence effect: When misinformation in memory

affects later inferences. l E ri n P h l :
Learning, Momoty, and Cogpitioa. 20, 1420-1436.

Jou, J., Shanteau, J., & Harris, R.J. (1996). An
information processing View of framing effects: The role of
causal schemas in decision making. Memoty & Cognition. 24,
1-15.

Kahneman, D., & Tversky, A. (1982). The simulation
heuristic. In D. Kahneman, P. Slovic, & A. Tversky (Eds.),
W New York:
Cambridge University Press.

Kaplan, M.F. (1977) Judgment by juries. In M.F. Kaplan &
s. Schwartz (Eds.), WW
appltoo_oott;ago. San Diego, CA: Academic Press.

Koehler, D.J. (1994). Hypothesis generation and
confidence in judgment, Journal of Ezporimontal PoyohoLogy:
Learning, Momoty, apd Cognition, 20, 461—469.

McDonald, J. (1998). 200% probability and beyond: The
compelling nature of extraordinary claims in the absence of
alternative explanations. Skoptioal tngpiror, 22. 45-49, 64.

McKoon, G., Ratcliff, R., & Seifert, C.M. (1989). Making
the connection: Generalized knowledge structures in story
understanding. Joatpai_otaMomoty_apd_tapggagoy_zgy 711—734.

Miller, D.T., Turnbull, W., & McFarland, C. (1989). When

a coincidence is suspicious: The role of mental simulation.

Journal of Personality and Social Psychology, 57, 581—589.

74

Niedermeier, K.E., Kerr, N.L., & Messé, L.A. (1999)

Juror’s use of naked statistical evidence: Exploring the

bases and implications of the Wells effect. Journal of
Personality and Social Poychology, 76, 533-542.

Pennington, N., & Hastie R. (1986). Evidence evaluation
in complex decision making. ourna f Personali ocial
Poyohology, 51. 242—258.

Pennington, N., & Hastie R. (1988). Explanation-based
decision effects: Effects of memory structure on judgment,
Journal of Exporimontal Eoyohology; Learning, Momoty. and
CognitionI 14, 521-533.

Pennington, N., & Hastie R. (1992). Explaining the
evidence: Tests of the story model for juror decision making.
Journal of Personality apd Social Poyohologyll 62, 189-206.

Pennington, N., & Hastie R. (1993). The story model for
juror decision making. In R. Hastie (Ed.), Inside_the_iureri
The psychology of jorot dooision making, (pp.192-221), New
York: Cambridge University Press.

Petty, R.E., & Cacioppo, J.T. (1984). The effect of
involvement on responses to argument quantity and quality:
Central and peripheral routs to persuasion. Jogtpal_of
Personality and Sooial Poyohology. 46. 69-81.

Petty, R.E., Cacioppo, J.T., & Goldman, R. (1981).
Personal involvement as a determinant of argument-based
persuasion. W412
847-855.

75

Robinson, L.B., & Hastie, R. (1985). Revision of beliefs

when a hypothesis is eliminated from consideration. Joprnal

of Ezperimontal Psychology;_fluman Peroeption and Performanco,
11, 443-456.

Robinson Van Wallendael, L.R. (1989). The quest for
limits on noncomplimentarity in opinion revision.
Organizational Behavior and Human Decision Processes. 43,
385—405.

Rosenthal, R. and Rosnow, R.L. (1991). Essontialo of
behavioral research: Methods and data analysis (2nd ed.). New
York: McGraw—Hill.

Schank, R.C. (1982). Dypamio memoty: A theoty of
reminding and learning in computers and people. New York:
Cambridge University Press.

Schwarz, N. (1994). Judgment in social context: Biases,
shortcomings, and the logic of conversation. Advancoo in
Eyperimental Social Psyohology, 26, 123—162.

Sykes, D.L., & Johnson, J.T. (in press). Probabilistic

evidence vs. the representation of an event: The curious case

of Mrs. Prob’s Dog. Basic and Applied Social Psychology.
Tabachnick, B.G., & Fidell, L. (1996). sin

multivariate statistios (3rd ed.). New York: Harper & Row.

Wason, P.C. (1960). On the failure to eliminate
hypotheses in a conceptual task. Tho Coartorly Jopraal of
Eaperimental Psyohology, 12, 129—140.

76

 

 

Wells, G.L. (1992). Naked statistical evidence of
liability: Is subjective probability enough? Jogrnal of
Personality and Sooial Poyohology, 62. 739-752.

77

"I7111111)]llllllllll