Close
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
The effects of coherence -based reasoning on betting decisions
(USC Thesis Other)
The effects of coherence -based reasoning on betting decisions
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
THE EFFECTS OF COHERENCE-BASED REASONING
ON BETTING DECISIONS
Copyright 2003
by
Aaron Brownstein
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PSYCHOLOGY)
August 2003
Aaron Brownstein
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
U M I Number: 3116671
INFORMATION TO USERS
The quality of this reproduction is dependent upon the quality of the copy
submitted. Broken or indistinct print, colored or poor quality illustrations and
photographs, print bleed-through, substandard margins, and improper
alignment can adversely affect reproduction.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if unauthorized
copyright material had to be removed, a note will indicate the deletion.
®
UMI
UMI Microform 3116671
Copyright 2004 by ProQuest Information and Learning Company.
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
ProQuest Information and Learning Company
300 North Zeeb Road
P.O. Box 1346
Ann Arbor, Ml 48106-1346
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UNIVERSITY OF SOUTHERN CALIFORNIA
THE GRADUATE SCHOOL
UNIVERSITY PARK
LOS ANGELES, CALIFORNIA 90089-1695
This dissertation, written by
A o.ro a feroVMfxAfci r > ______________
under the direction o f h iS dissertation committee, and
approved by all its members, has been presented to and
accepted by the Director o f Graduate and Professional
Programs, in partial fulfillment o f the requirements fo r the
degree o f
DOCTOR OF PHILOSOPHY
Director
Date A u gu st 1 2 . 2003
Dissertation Committee
Chair
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table of Contents
List of Tables and Figures..............................................................................................iii
Abstract.......................................................................................................................... iv
Introduction........................................................................................................................1
Bolstering Probabilities at the Racetrack........................................................... 2
Hindsight B ias..................................................................................................... 8
Overview and Predictions................................................................................... 9
Study 1 .......................................................................................................................... 10
Introduction.........................................................................................................10
M ethods........................................................................................................... 12
Results................................................................................................................. 18
Discussion...........................................................................................................21
Study 2 ............................................................................................................................ 22
Introduction.........................................................................................................22
Methods..............................................................................................................23
Results..................... 25
Discussion ......................................................................................................28
Study 3 ............................................................................................................................ 30
Introduction.........................................................................................................30
M ethods..............................................................................................................33
Results.................................................................................................................36
Discussion...........................................................................................................41
General Discussion........................................................................................................ 43
Conclusions..................................................................................................................... 50
Alphabetized Bibliography.......................................................................................... 52
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
List of Tables and Figures
Figure 1. Network representation of decision between two alternatives, A and B ... 5
Figure 2. Ratings of chosen and non-chosen horses' chances of winning a race .... 19
Figure 3. Ratings of chosen and non-chosen horses' chances of winning a race .... 27
Figure 4. Ratings of chosen horse's chance of winning a race...................................28
Figure 5. Ratings of chosen horse's chance of winning a race................................... 38
Table 1. Proportion of participants who ranked their chosen horse first
by experimental condition and rating tim e...................................................................39
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
iv
ABSTRACT
This research demonstrates some ways in which coherence-based reasoning
affects decision making. Participants viewed information about horses in a simulated
race and rated each one’s chance of winning three separate times before placing a
bet, once after betting, and once again after learning the outcome of the race (in
hindsight). Ratings of the chosen horse were bolstered within the pre-decision period
and immediately after the decision. Pre-decision bolstering occurred even when
participants did not expect to bet, and increased with task importance and participant
expertise. In hindsight, ratings of the winner increased and ratings of the losers
(including the bet horse) decreased. I explain how coherence-based reasoning
accounts for these and other findings in decision making and other contexts.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1
Introduction
Early theories in social psychology, such as cognitive dissonance theory
(Festinger, 1957, 1964) and balance theory (Heider, 1958), depicted social thinkers
as imposing consistency on objects and events, in order to achieve a coherent
understanding of their world. Using principles of neural network modeling
(McClelland & Rumelhart, 1986), recent authors have developed theories of
constraint satisfaction that concretize and extend ideas pioneered by these earlier
consistency theories. Constraint satisfaction theory (CST) views cognitive processing
as involving the spreading of activation among interconnected elements, with
elements that are mutually consistent increasing each other’s activation and elements
that are mutually inconsistent decreasing each other’s activation, so that over time
people achieve an internally consistent or coherent understanding of their world
(Holyoak & Thagard, 1989; Read & Miller, 1994; Read, Vanman & Miller, 1997;
Simon & Holyoak, 2002; Shultz & Lepper, 1996; Thagard, 2000). CST has shown
how coherence-based reasoning accounts for findings on a wide array of topics,
including causal explanation, analogical reasoning, and attitude change (Read &
Marcus-Newhall, 1993; Read et al., 1997; Shultz & Lepper, 1996; Holyoak &
Simon, 1999; Spellman & Holyoak, 1992; Spellman, Ullman, & Holyoak, 1993;
Thagard, 2000). The present research looks at some of the ways coherence-based
reasoning can affect decision making, using the context of betting on a horse race.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2
Bolstering Probabilities at the Racetrack
One application of coherence-based reasoning concerns whether decision
makers bolster their chosen alternative. Cognitive dissonance theory suggests that
after people make a decision, they experience dissonance, which they relieve by
bolstering their chosen alternative. This prediction of post-decision bolstering of a
chosen alternative has been supported in several studies (e.g., Brehm, 1956;
Festinger, 1964; Gerard & White, 1983) including one dealing with the choice of
which horse to bet on in a horse race (Knox & Inkster, 1968). Knox and Inkster
asked bettors at a racetrack to rate the chance that the horse they bet on would win
the race, and found that bettors who had just placed a win-bet gave more optimistic
estimates of the chance that their horse would win than bettors who were about to
place a win-bet, suggesting that decision makers bolster their chosen alternative after
making a decision.
Although dissonance theory predicted that bettors bolster their chosen
alternative after betting, it maintains that they do not bolster their (eventually) chosen
alternative within the pre-decision phase, since it views the pre-decision period as
characterized by an “absence of any systematic, biasing re-evaluation of alternatives”
(Festinger, 1964, p. 153). Similarly, action phase theory, which views the pre
decision phase as characterized by “an orientation toward accurate and impartial
processing” of information (Gollwitzer, 1990, p. 65), and theories of rational choice,
which require estimates of probabilities associated with alternatives to be stable over
time and independent of the expected values of alternatives (Hogarth, 1987, von
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3
Neumann & Morgenstem, 1944), also suggest that bettors should not bolster their
chosen alternative before reaching a decision.
In contrast with dissonance, action phase, and rational decision theories, CST
suggests that coherence-based reasoning causes bolstering of a chosen alternative
even before a decision has been made. CST’s prediction of pre-decision bolstering is
based on an analysis of the relationships among salient concepts and the way
processing proceeds before a decision.
According to CST, a network of interconnected units can provide a symbolic
representation of the concepts involved in a decision and the relationships between
those concepts. In this type of network units represent salient concepts, so in the
ease of deciding which horse to bet on in a race, units might represent positive
information about a horse or estimates of its chance of winning a race. Units are
connected by links representing consistent or inconsistent relationships between
concepts. For example, positive information about a horse is consistent with an
estimate that it is likely to win a race, so those two concept units would be connected
by a positive link, but an estimate that one horse is likely to win the race is
inconsistent with positive information about another horse, so those two concept
units would be connected by a negative link.
In addition to using the structure of a network to represent salient concepts
and the relationships among them, CST simulates the operation of cognitive
processes by postulating that each unit affects the level of activation of other units
connected to it. The level of activation of a unit reflects the extent to which it is
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4
currently active in the network and the extent to which the concept it represents is
viewed as acceptable or believable. Each unit affects the level of activation of other
units linked to it, but the way in which one unit affects the activation of another unit
depends on the kind of link between them, so that units connected by positive links
increase each other’s level of activation and units connected by negative links
decrease each other’s level of activation.
When a network is constructed to represent the salient concepts in a decision
and the relationships between those concepts, and cognitive processing is simulated
by allowing activation to pass between units in the network, a subgroup of units
connected to each other by positive links can be expected to emerge as highly
activated while the other units, which are connected to them by negative links, can
be expected to end up with low levels of activation. Thus, CST predicts that as
cognitive processing proceeds a subgroup of concepts which are consistent with each
other emerge as the only ones which seem acceptable, and the decision maker
thereby achieves an internally consistent or coherent view of the concepts related to
the decision.
Figure 1 illustrates how a network could be used to simulate cognitive
processing during decision making, for the decision of which of two horses to bet on
in a race. If the decision maker is initially impressed by positive information about a
horse (e.g., its number of previous wins), the units representing the positive
information about that horse (+A in Figure 1) can be expected to be highly activated
and to increase the activation levels of units connected to them by positive links,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5
including other units representing positive information about the horse (other +A
units) and a unit representing the perceived probability that the horse will win the
race (pA). Meanwhile, as those positively linked units increase each other’s
activation, they can also be expected to reduce the activation levels of units
connected to them by negative links, including units representing positive
information about the second horse (+B) and a unit representing the perceived
probability that the second horse will win the race (pB). CST predicts that as long as
the decision maker continues to think about the information, processing will proceed
in this manner, with the perceived probability that the first horse will win the race
progressively increasing and the perceived probability that the second horse will win
the race remaining at low levels.
-A +A
-A +A
+B
+B
Figure I. Network representation of decision between two alternatives, A and B. Circles represent
units; those labeled +A represent positive information about alternative A, those labeled -A represent
negative information about alternative A, and the circle labeled pA represents the perceived
probability that alternative A will win a race. Solid lines represent positive links between consistent
units and dotted lines represent negative links between inconsistent units.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
6
The present research tested whether a favorite alternative emerges and is
bolstered during decision making as well as after a choice. Participants viewed
information about horses competing in a simulated race, and rated the chance each
horse had of winning, three separate times before betting on one to win, and then
again after betting. I expected to replicate Knox and Inkster’s finding that estimates
of the chosen horse’s chance increase after the bet is placed, and based on CST I also
expected to find that estimates of the chosen horse’s chance increase over time
within the pre-decision phase.
Beyond predicting that decision makers bolster their favorite alternative
before choosing, CST makes some interesting predictions about when decision
makers are more likely to bolster and which decision makers are most likely to
bolster.
Task importance. CST suggests that the spreading of activation among units
leads to bolstering of a favored alternative any time people think about a set of
alternatives, even in the absence of an explicit decision making goal (Simon, Pham,
Le, & Holyoak, 2001). Therefore, I predicted that participants who thought about a
set of horses about to compete in a race would bolster the horse they eventually bet
on even if they did not expect to place a bet.
CST also suggests that when people expect to make a decision, especially a
more important decision, they devote a greater amount of processing to the decision,
and that increased processing enables units to have a greater effect on each other’s
activation levels, so that the favored alternative will be bolstered to a greater degree.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Therefore, I predicted that participants who expected to make a decision (choose the
horse they think will win a race), even a strictly hypothetical, non-consequential
decision, would bolster their chosen alternative to a greater extent than participants
who thought about their alternatives without expecting to make a choice. I also
predicted that participants who expected to make a more important, consequential
decision (place a bet for a chance to win money) would bolster their chosen
alternative to a greater extent than participants who expected to make a non-
consequential decision or participants who did not expect to make a decision at all.
Previous research on the effects of decision importance on pre-decision re-evaluation
of alternatives has been inconclusive (Mills & Ford, 1995; O’Neal, 1971; Tyszka,
1998).
Individual expertise. Another intriguing issue concerns the effects of
individual differences in domain-specific expertise on bolstering of a chosen
alternative. Intuition suggests that experts should generally tend to be less biased, but
previous research shows that people with greater expertise have stronger correlations
among the attributes in their knowledge base resulting in more polarized attitudes
(Lusk & Judd, 1988; c.f., Judd & Lusk, 1984; Millar & Tesser, 1986). Similarly,
CST suggests that experts have stronger links between units, and that stronger links
enable units to affect each other’s activation levels more quickly and more
extensively, so that experts can be expected to bolster their favored alternative to a
greater extent than non-experts.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
8
Hindsight Bias
Another application of coherence-based reasoning concerns how outcome
knowledge affects evaluations of alternatives in hindsight. Research on hindsight
bias suggests that in hindsight, probability estimates conform to the actual outcome
of an event (Fischhoff, 1975; Hawkins & Hastie, 1990). The two major accounts of
hindsight bias view it either as a product of motivated self-presentation (people try to
appear as if they “knew it all along”) or as a cognitive learning process, in which
knowledge about an outcome affects “rejudgment” of the issue (Hawkins & Hastie,
1990). Within the cognitive viewpoint, Fischhoff (1975) suggested that outcome
knowledge affects rejudgment of the issue when it is “assimilated” with other
elements in an effort to perceive the entire system as a “coherent whole” (p. 297).
CST elaborates on Fischhoff s description, suggesting that it is the spreading of
activation in a network that enables a unit representing the outcome of an event to be
assimilated with other information in a coherent manner. When the outcome of a
horse race becomes known, a unit representing the winning horse’s victory is added
to the network. This highly activated unit increases the activation levels of positively
linked consistent units, including the perceived probability that the horse had been
likely to win all along, and decreases the activation levels of inconsistent negatively
linked units, including the perceived probabilities that the other horses had been
likely to win.
In the present research (Study 1), after the simulated race, participants
learned the outcome of each horse (first, second, etc.), and then gave hindsight
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
9
estimates of the chance each horse had of winning before the race was run. I
expected to find hindsight bias for horses that won or lost the race, so that when
participants found out that a particular horse won the race, their estimates of its
chance of winning would increase, and when participants found out that a horse lost
the race, their estimates of its chance of winning would decrease. In addition, I
expected to find hindsight bias for the chosen horse, so that participants who found
out that their chosen horse won the race would increase their estimates of its chance
of winning, and participants who found out that their chosen horse had not won the
race would decrease their estimates of its chance of winning.
Overview cmd Predictions
This paper reports three studies in which participants viewed information
about a set of horses about to compete in a simulated race, rated the chance each
horse had of winning the race three separate times, then bet on one of the horses to
win the race, and finally rated the chance each horse had of winning the race one
more time. In all three studies, I predicted that participants would report increasingly
optimistic estimates of the chance the horse they bet on would win the race,
bolstering it within the pre-decision phase and again after placing their bets.
Study 1 also investigated hindsight bias. After participants made their post
choice ratings of the four horses, they were told the outcome of the race, and then
asked to make hindsight ratings of the chance each horse had of winning before the
race was run. Based on CST, I expected hindsight estimates to conform to the actual
outcome of the race, so that ratings of the horse which won the race would increase
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
10
in hindsight, and ratings of the horses which did not win the race would decrease in
hindsight. I also predicted that among participants who found out that their chosen
horse won the race, estimates of its chance of winning would increase in hindsight,
and among participants who found out that their chosen horse lost the race, estimates
of its chance of winning would decrease in hindsight.
Studies 2 and 3 investigated the effects of individual expertise and task
importance on bolstering of a chosen alternative. They included participants with a
wide range of experience in betting on horse races, and based on CST, I predicted
that participants with greater expertise would bolster their chosen alternative to a
greater extent than participants with less expertise. Studies 2 and 3 also included
different experimental conditions, manipulating participants’ expectations about the
importance of the task they would be performing. Based on CST, I predicted that
participants in all task importance conditions would bolster their chosen alternative,
and that participants in conditions involving greater task importance would bolster
their chosen alternative to a greater extent than participants in conditions involving
lesser task importance.
Study 1
Introduction
In the first part of the study, participants viewed information on a set of
horses about to compete in a simulated race, and rated each one’s chance of winning
the race {baseline ratings). In the second part of the study, participants viewed
information on an identical set of horses about to compete in another simulated race,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
11
and were told that they would eventually be able to bet on one to win the race, and if
they bet on the horse which won the race, they would receive extra tickets in a cash
lottery. With this expectation of eventually making a consequential decision,
participants went on to rate the chance each horse had of winning the race, two
separate times (first and second pre-choice ratings), before finally betting on one of
the horses to win the race. After placing their bets, participants once again rated the
chance each horse had of winning the race (post-choice rating), and then after the
simulated race was run and they were told which horse had won, they rated the
chance each horse had seemed to have of winning the race before it was run
{hindsight rating).
I predicted that participants would bolster their eventually chosen alternative
throughout the decision making process, so that ratings of the chosen horse would
increase from baseline through the first and second pre-choice ratings, and again at
the post-choice rating. I also predicted that hindsight ratings would conform to the
actual outcome of the race, so that ratings of the horse that won the race (regardless
of whether it was chosen) would increase from the post-choice rating to the hindsight
rating, and ratings of the horses that did not win the race (regardless of whether they
were chosen) would decrease from the post-choice rating to the hindsight rating. I
also predicted that among participants whose chosen horse won the race, ratings of
the chosen alternative would increase from the post-choice rating to the hindsight
rating, and among participants whose chosen horse did not win the race, ratings of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
12
the chosen alternative would decrease from the post-choice rating to the hindsight
rating.
Methods
Participants. This study appeared on a website for psychology experiments.
Announcements were sent to the website’s mailing list, 180 people logged on, and 90
usable data sets were obtained.
M aterials and procedure. An introductory page invited people to participate
in a two-part study, explaining that participants who completed the first part of the
study were guaranteed one entry in a lottery for a $200 cash prize, and participants
who completed the second part of the study would have an opportunity to earn
additional entries in the same $200 lottery.
Baseline ratings. In the first part of the study, participants found instructions,
four sets of information charts corresponding to four racehorses, and dependent
measures. The instructions described the charts as the racing form for an imaginary
race. The charts were adapted from the Daily Racing Form, and contained a large
amount of data on each horse. There were four charts for each horse. One chart
displayed the horse’s profile, (color, sex, age, drugs, weight, odds, payoff) and a
second chart summarized its jockey’s overall and recent performance (wins, places,
shows, percent wins). A third chart summarized the horse’s performance overall, in
the last two years, and at specific racetracks (wins, places, shows, earnings, speed)
and a fourth chart displayed its performances in five recent races (date; track’s
location, condition, length; horse’s speed, position, margin, drugs, weight, jockey;
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
13
number in race). Participants could view explanations of the information in most
columns by clicking on the column header. The order of the charts for each horse,
and the order of presentation of horses, was the same for all participants.
The dependent measures were four questions asking for an assessment of the
chances each horse had of winning the race. The questions were customized for each
horse (for example, What chance does Radiance have o f winning the race?).
Responses were made using a 15-point radio button scale, arranged in a horizontal
line and anchored at slight (1), moderate (8), and excellent (15). Participants could
change their ratings while on this page.
Thus, up to this point participants were only asked to rate the four horses, and
were not told anything about the prospect of actually placing a bet. After making
their ratings they clicked a continue button which took them to the next page.
Throughout the experiment, once participants clicked continue to advance to the next
page, they were unable to return to the previous pages.
Invitation to the second part o f the study. On the next page, participants were
thanked for their participation in the first part of the study, assured that they had one
ticket in the $200 lottery, and invited to participate in the second part of the study.
Then they were invited to place a bet on a race, and were informed that they would
have something to gain from their decision. The instructions read:
In the second part of this study, we will run a simulated horse race. If you
choose to participate in the second part of the study, we will give you $2.00
in "virtual money" with which to bet on one of the horses to win the race. If
you bet on the horse which wins the race, we will convert its “payoff’ into
extra lottery tickets (the “payoff’ for each horse is listed in the information
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
14
charts). Winnings will be rounded up to the nearest whole dollar (so a $4.60
payoff becomes 5 extra tickets). If you bet on a horse which doesn't win the
race, you'll still have the single entry in the lottery which you received for
participating in the first part of the study.
Like in a real horse race, the odds and payoffs were based on the relative
amounts bet on the horses in the race (Smith, 1998). The simulation assumed a
$100,000 betting pool in which $42,000 had been bet on one horse (so its odds were
1.4 to 1 and its payoff for a $2.00 bet was $4.80), $29,000 was bet on a second
(2.5:1, $7.00), $18,000 was bet on a third (4.6:1, $11.20) and $11,000 was bet on a
fourth (8.1:1, $18.20).
First pre-choice ratings. Participants who continued to the second part of the
study advanced to the next page, where they found instructions, information charts,
and dependent measures. The instructions explained that the charts contained
information about the horses competing in the race they would bet on. The
information in the charts was the same throughout both parts of the study, but in the
second part of the study the horses were renamed and presented in a different order
(so that they appeared to be a different set of horses), in order to prevent participants
from simply repeating their earlier ratings (at baseline) when they made subsequent
ratings (when they entered the second part of the study). The order of the charts for
each horse, and the order of presentation of horses, was the same for all participants.
The dependent measures were customized for each horse and appeared between the
charts, and participants responded using the 15-point scale.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
15
One concern in this type of research is that participants might reach a private
decision before placing a bet. To help prevent such premature decisions, the
instructions emphasized that the time to place the bet was still fairly far away. The
instructions explained that the charts contained only information that was “currently
available,” and that the information “may be changed or updated as new
developments come in before the race” because “a horse may be injured, prompting
the posting of a special notice or withdrawal of a horse from the race, a jockey may
be replaced, and the odds may change as bets come in.” The instructions then asked
participants to report their initial inclinations about the chance of each horse winning
the race.
As you study the available information you may find that you have formed
some inclinations about the chances of each horse winning the race. We
would like to know what these initial inclinations are. After having reviewed
the information, please provide your initial ratings of the chance of each
horse winning the race by clicking on one of the dots on each line. You can
change your ratings by clicking a different dot.
Participants were assured that they would have plenty of time to place their
bets once the information became final. After they made their ratings and clicked
continue, they advanced to the next page.
Second pre-choice ratings. On the next page, the instructions asked
participants to continue looking through the charts and again rate the chance each
horse had of winning the race. As on the previous page, the instructions warned
participants that it was too early to reach a decision about their bet and assured them
that they would have plenty of time to place their bet when the information became
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
16
final. The charts and dependent measures were the same as on the previous page.
After participants made their ratings and clicked continue, they advanced to the next
page.
Placing the bet. On the next page, the instructions told participants that it was
time to place their bets. Participants were informed that no new information was
coming, and were encouraged to review the charts before placing their bets. The
charts that appeared on the page were the same as on the previous page. Participants
placed their bets by clicking on a radio button next to the name of the horse they
wanted to bet on, and then clicked continue to advance to the next page.
Post-choice ratings. On the next page, the instructions thanked participants
for their bets, and announced that the betting windows were closed and the race
would start momentarily. The instructions then asked participants to review the
charts again and rate the chance each horse had of winning. The charts and
dependent measures were the same as on the previous page. After participants made
their ratings and clicked continue, they advanced to the next page. A computer
program selected the order of the horses for each race, with each horse’s chance of
winning proportional to its odds in the charts.
Outcome o f race and hindsight ratings. On the next page, participants were
told the outcome o f the race (which o f the four horses came in first, second, third,
and fourth), given a customized summary of their own racing outcome (“You bet on
, so your ‘virtual payoff would have been $ and you get extra
tickets in the lottery”), and then asked to rate the horses again:
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
17
Take another look at the information that was available just before you
placed your bet, and think back to the time before the race was run. Based
only on the information you had at that time, please rate what chance did
each horse have for winning the race.
The charts and dependent measures were the same as on previous pages,
except that the dependent measures were phrased in the past tense (for example,
Quicksilver’ s chance o f winning was:). After participants made their ratings and
clicked continue, they advanced to the next page.
Experience. I had originally intended to include individual differences in
experience with horse racing as a factor in Study 1; therefore, on the next page
participants were asked to report the extent of their experience with horse racing.
One item asked how many times they had bet on a horse race prior to this experiment
(less than five, 5-20 times, more than 20 times). A second item asked how they
would have described themselves with regard to betting on horses (novice, know a
little, know quite a lot, expert). However, I found that participants reported so little
experience with horse racing that it was not possible to test this variable in Study 1 .
As will be seen below, Studies 2 and 3 did obtain reasonable samples of more
experienced horse players, making it possible to test the effects of individual
expertise in those studies.
Debriefing. On the last page, participants viewed a debriefing statement,
which explained the experiment and the lottery, and were able to e-mail the
experimenters.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
18
Results
I predicted that participants would bolster the horse they chose, reporting
increasingly optimistic estimates of the chance it would win the race, and that this
bolstering would occur both before and after placing the bet. I did not expect
participants to bolster the horses they did not choose, so that estimates of their
chances of winning the race would remain stable throughout the procedure.
To test for bolstering of the chosen horse within the pre-decision period, I
compared estimates of the chosen horse with mean estimates of the three non-chosen
horses at three points in time: (1) baseline ratings (base), (2) first pre-choice ratings
( pre-1), and (3) second pre-choice ratings (pre-2), using a 2(chosen horse, non-
chosen horse) by 3(base, pre-1, pre-2) repeated measures ANOVA. I found a
significant main effect of alternative, F (l, 82) = 34.67,/? < .001, such that the chosen
horse was rated higher than the non-chosen horses, a significant main effect of time,
F(2, 164) = 10.68,/? < .001, such that ratings increased over time, and a significant
alternative by time interaction, F(2, 164) = 9.63, p < .001, such that ratings of the
chosen horse increased over time but ratings of the non-chosen horses remained
stable (Figure 1).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
19
11.5
9 1 1
| 10.5
£ 1 0
® 9.5
1 9
® 8.5
♦ Chosen
- • -o- - - Non-chosen
o '
7.5
Post Base P re-1 Pre-2
Rating Time
Figure 2. Ratings of chosen and non-chosen horses' chances of winning a race. Ratings were made at
baseline (base), the first pre-choice rating time (pre-1), the second pre-choice rating time (pre-2), and
the post choice rating time (post).
The horse that was eventually chosen was rated significantly more likely to
win the race than the other horses at base, 1(85) = 2.10, p < .04, a difference that
probably reflects an initial preference for a particular horse rather than pre-decision
bolstering. But then at pre-1, when participants evaluated an apparently new but
equivalent set of horses with the expectation of betting on one to win the race,
ratings of the chosen horse increased significantly from base, 1(81) = 3.21, p < .01,
suggesting that the expectation of making a consequential decision can strengthen a
preference. As participants continued through the decision making process, the
perceived advantage of their favorite horse continued to increase, as evidenced by a
marginally significant increase in ratings of the chosen horse from pre-1 to pre-2,
t(81) - 1.68, p < . 1 . After participants placed their bets, the chosen horse was
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
20
bolstered again, so that its ratings increased from pre-2 to post, t{89) = 2.51, p < .02.
A linear trend analysis confirmed that ratings of the chosen horse increased within
the pre-decision phase, F{1, 87) = 20.65, p < .001, but ratings of the non-chosen
horses did not change, p = .62.
After participants were informed of the outcome of the race, they made
hindsight ratings of how likely each horse had been to win the race before it was run.
As predicted, hindsight ratings conformed to the actual outcome of the race, so that
ratings of the horse that won the race (regardless of whether it was bet or not)
increased from post (just before the outcome was known) to hind (when the outcome
was known), /(83) = 2.18,/; < .04, and ratings of the horse that did not win the race
decreased from post to hind, t(80) = -3.0, p < .01.
I also examined how ratings of the chosen horse changed from post to hind,
separately for participants whose chosen horse won the race and for participants
whose chosen horse did not win the race. Among participants whose chosen horse
won the race, ratings did not increase from post to hind, p < .3. This finding probably
reflects a kind of ceiling effect, since other research using this paradigm (Studies 2
and 3 below) has found that participants with little experience betting on horse races
seldom rate their chosen horse higher than 12. Among participants whose chosen
horse did not win the race, ratings decreased significantly from post to hind, t{64) = -
2.90,/? < .01, once again showing that hindsight estimates are sensitive to the actual
outcome of the race. However, even after participants found out their chosen horse
did not win the race, they still rated it higher than they had at base, t(64) = 2.78, p <
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
21
.01, and marginally higher than they had at pre-1, f(63) = 1.96, p <.06, suggesting
that the tendency for hindsight ratings to conform to the actual outcome was not
strong enough to completely reverse the bolstering effect.
Discussion
CST suggests that coherence-based reasoning causes bolstering of
probabilities associated with a favored alternative before a decision as well as after a
decision, and indeed, ratings of the chosen alternative increased throughout the pre
decision period and again after the choice had been made, but ratings of the non
chosen alternatives did not change across time. The finding that preference for the
chosen alternative increased over time suggests that the cognitive processes active
during decision making can be seen as having a biasing effect; in this case, causing
an initially tentative preference to become a strong conviction that a bet is worth
making.
CST suggests that coherence-based reasoning also causes hindsight ratings to
conform to the outcome of an event. As predicted, ratings of the horse that won the
race increased from post to hind, and ratings of horses that did not win decreased
from post to hind, thereby demonstrating hindsight bias in a decision making task, a
context rarely used in research on hindsight bias (Creyer & Ross, 1993; Louie,
1999). Also as predicted, in cases where the chosen horse did not win the race,
hindsight ratings of the chosen horse conformed to the outcome of the race,
decreasing from post to hind. Yet, although participants moved toward “correcting”
their inflated ratings of their chosen horse, hindsight ratings of the chosen horse did
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
22
not decrease to the level they had been at base or pre-1, suggesting that the bolstering
effect can be sufficiently robust as to persevere in the face of disconfirming
evidence.
Study 2
Introduction
Study 1 showed that participants bolster their chosen alternative within the
pre-decision phase as well as after making a decision. Study 2 drew on CST to
predict how two variables might moderate pre-decision bolstering of a chosen
alternative.
CST suggests that individuals with greater domain-specific expertise have
stronger links between units, which enable units to affect each other’s activation
levels more quickly and more extensively, so that the chosen alternative can be
bolstered to a greater extent. In Study 2 ,1 recruited participants with different
degrees of experience betting on horse races, and predicted that participants with
greater expertise would bolster their chosen alternative to a greater extent than
participants with less expertise.
Further, CST suggests that increasing task importance leads to more
extensive processing, so that units are able to have a greater effect on each other’s
activation levels, and the chosen alternative can be bolstered to a greater extent.
Study 2 included three experimental conditions differing in task importance. The
decision/consequences condition (a replication of Study 1) most closely resembled a
real horse race, because participants expected to bet on a simulated race which had
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
23
tangible consequences (if they won the race, they would receive extra tickets in a
lottery). In the decision/no-consequences condition, participants expected to make a
decision (bet on a horse in the race), but they did not expect any consequences from
their decision (they did not even expect to be told the outcome of the race). In the no
decision condition, participants rated the horses three times without expecting to
place a bet at all. I predicted that participants would bolster their chosen alternative
in all three conditions, but that they would bolster it to a greater extent in the
decision/no-consequences condition than in the no-decision condition, and that they
would bolster it to a greater extent in the decision/consequences condition than in the
other two conditions.
Methods
Participants. This study was conducted on a website for psychology
experiments. I obtained 152 usable data sets from the website’s mailing list. To
obtain a sample of experienced horse players, I advertised the study on a website
featuring news and discussions of real horse races, and obtained another 139 usable
data sets.
Decision/consequences condition. The decision/consequences condition was
identical to Study 1, except that the page which took hindsight ratings was omitted.
Decision/no-consequences condition. The decision/no-consequences
condition was identical to the decision/ consequences condition, with the following
exceptions.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
24
The second part of the study was described in a way that led participants to
believe that their decision would have no consequences:
The second part of this study involves some more short tasks related to horse
racing. If you choose to participate in the second part of this study, you will
review another set of charts about four horses running in a simulated race,
respond to questions about those horses, and choose the horse you would like
to bet on - though you won’t actually find out the results of the race.
To motivate participants to proceed to the second part of the study, I offered
them a chance to obtain more tickets in the lottery, even though I did not want them
to associate the chance with betting on the race. Therefore, the instructions
continued, “After that, you will have an opportunity to participate in a game which is
also related to horse racing, in which you may win additional tickets in our $200
lottery.”
On the next two pages, where participants rated the horses (pre-1 and pre-2),
the number of extra lottery tickets associated with each horse was omitted from the
information charts.
On the third page, participants were asked to bet on the race. They were told
for the first time that they would receive feedback on the results of the race, and that
if the horse they bet on won the race, they would obtain extra tickets in the lottery.
The instructions explained that “this is the opportunity we promised you earlier to
earn extra tickets in the lottery” and then explained how a winning horse’s payoff
would be converted into extra lottery tickets, using the explanation that was
presented to participants in the decision/feedback condition.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
25
No-decision condition. The no-decision condition was identical to the
decision/no-consequences condition, with the following exceptions.
The second part of the study was described in a way that led participants to
believe that they would not be expected to make a decision at all:
The second part of this study involves some more short tasks related to horse
racing. If you choose to participate in the second part of this study, you will
review another set of charts about four horses running in a simulated race and
respond to questions about those horses.
On the next two pages, references to betting on the race were deleted. Thus,
on the first page of the second part of the study, in the parts of the instructions that
warned that the charts may be updated with new information, phrases advising
participants to refrain from making a “decision about the bet” and promising that
“you will later have as much time as you like to make your bet” were deleted.
Similarly on the next page, the parts of the instructions warning participants that it
was “too early to make any decision about the bet” and assuring them that “you will
later have as much time as you like to make your bet” were omitted.
Results
Study 2 tested three major predictions. First, I predicted that participants
would bolster their chosen alternative within the pre-decision phase and then again
after placing their bets. Second, I predicted that participants with more experience
betting on horse races would bolster their chosen alternative to a greater extent than
participants with less experience betting on horse races. To test that prediction, I
classified participants as experts (n = 123, 7% from mailing list) if they had bet on 20
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
26
or more horse races and described themselves as a seasoned expert or as knowing
quite a lot about handicapping, and as non-experts (n = 135, 95% from mailing list)
if they had bet on less than 20 races and described themselves as a novice or as
knowing a little about handicapping (and excluded 31 participants who could not be
classified). Third, I predicted that bolstering would increase with task importance, so
that participants would bolster their chosen horse to a greater extent in the
decision/consequences condition than in the decision/no-consequences condition,
and would bolster their chosen horse to a greater extent in the decision/no
consequences condition than in the no-decision condition.
A 2(chosen, non-chosen) by 3 (base, pre-1, pre-2) by 3 (no-decision,
decision/no-consequences, decision/consequences) by 2(expert, non-expert)
ANOVA revealed a main effect of alternative, F( 1, 258) = 313.57, p < .001, a main
effect of time, F{2, 516) = 20.06, p < .001, and an alternative by time interaction,
F{2, 516) = 16.29,/? < .001. As predicted, linear trend analyses showed that ratings
of the chosen alternative increased over time within the pre-decision period, F( 1,
278) = 33.72,/? < .001, but ratings of the non-chosen alternatives did not change over
time, p < .7. Also as predicted, after the bet had been placed, ratings of the chosen
alternative increased again, /(280) = 7.10,/? < .001, but ratings of the non-chosen
alternatives did not change,/? < .7.
As predicted, linear trend analyses showed that ratings of the chosen
alternative increased over time among experts, F{1, 120) = 19.95,/? < .001, and
among non-experts, F (1, 157) = 16.04,/? < .001, and an alternative by expertise
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
27
interaction, F (1, 258) = 48.58,/? < .001, revealed that experts bolstered their chosen
alternative to a greater extent than non-experts bolstered their chosen alternative,
<280) = 6.75, p < .001 (Figure 2).
£ 10.5 -
E?peit Chosen
Non-ejpert Chosen
E xp ert Non-chosen
Non-ejpert Non-chosen
Base Pre-1 Pre-2
Rating Time
Post
Figure 3. Ratings of chosen and non-chosen horses' chances of winning a race. Ratings were made by
experts and non-experts at baseline (base), the first pre-choice rating time (pre-1), the second pre
choice rating time (pre-2), and the post choice rating time (post).
As predicted, linear trend analyses showed that ratings of the chosen
alternative increased over time in the no-decision, F (l, 102) = 10.23,/? < .01,
decision/no-consequences, F (l, 87) = 5.21,/? < .03, and decision/consequences
conditions, F (l, 87) = 21.98,/? < .001, and a time by condition interaction, F(4, 516)
= 2.36,/? < .06, revealed that participants in the decision/consequences condition
bolstered their chosen alternative to a greater extent than participants in the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
28
decision/no-consequences and no-decision conditions (Figure 3). At pre-1 the chosen
alternative was rated higher in the decision/consequences condition than in the
decision/no-consequences condition, /(169) = 22%, p < .03, or the no-decision
condition, /(191) = 2.40, p < .02, and at pre-2 the chosen alternative was rated higher
in the decision/consequences condition than in the decision/no-consequences
condition, /(169) = 2 2 3 ,p < .03.
125
e
♦ Decision/Consequences
— 0— Dec is iori/No-consequences
No-Decision
9.5
Base Pre-1 Pre-2 Post
Rating Time
Figure 4. Ratings of chosen horse's chance of winning a race. Ratings were made by participants in
no-decision, decision/no-consequences, and decision/consequences conditions at baseline (base), the
first pre-choice rating time (pre-1), the second pre-choice rating time (pre-2), and the post choice
rating time (post).
Discussion
As predicted by CST, participants bolstered their chosen alternative before
they had reached a decision as well as after deciding, suggesting that decision
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
29
making involves coherence-based reasoning. Also as predicted by CST, experts
bolstered their chosen alternative to a greater extent than non-experts, suggesting that
experts have stronger links enabling activation to spread more quickly and more
extensively between units. Finally, as predicted by CST, participants who expected
to make a consequential decision bolstered their chosen alternative to a greater extent
than participants who did not expect to receive consequential feedback on their
decision or who did not expect to make a decision, suggesting that increasing task
importance leads to a greater degree of processing, enabling activation to transfer
between units to a greater degree.
I had also predicted that participants who expected to make a non-
consequential decision would bolster their chosen alternative to a greater extent than
participants who did not expect to make a decision, but they did not. The failure of
this prediction may suggest that the expectation of making a non-consequential
decision does not increase bolstering beyond merely thinking about a set of
alternatives without any expectation of making a decision. However, I considered the
alternative possibility that my prediction may have been theoretically valid but the
experimental manipulation may have failed. Looking back at the instructions in the
no-decision condition, it seemed plausible that participants might have guessed that
they would eventually have an opportunity to indicate which horse they thought
would win the race (i.e., make a decision). Therefore, in Study 3 I introduced some
changes to the instructions in the no-decision condition, so that participants would be
less likely to anticipate being asked to make a decision.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
30
Study 3
Introduction
Study 2 found that participants who expected to make a consequential
decision bolstered their chosen alternative to a greater extent than participants who
expected to make a non-consequential decision or participants who did not expect to
make a decision, suggesting that increasing the importance of a decision by
increasing the consequences associated with it increases pre-decision bolstering.
Study 3 tested whether further increasing the consequences associated with a choice
would further increase pre-decision bolstering, by introducing a new condition in
which participants were told that, if they bet on the horse that won the race, they
would receive extra lottery tickets and would also receive the cash payoff associated
with their winning horse. I predicted that participants in this new decision/payoff
condition would bolster their chosen alternative to a greater extent than participants
in the decision/consequences condition, who were not offered the opportunity to
obtain the cash payoff.
Research on decision making has shown that anticipated regret affects
choices people make (Zeelenberg, 1999) and some authors have debated how
anticipated regret may affect pre-decision re-evaluation of alternatives (Janis &
Mann, 1977; Svenson, 1992; Svenson, Rayo, Andersen, Sandberg, & Svahlin, 1994;
Tyszka, 1998). According to CST, an imagined regrettable choice is inconsistent
with optimistic probability estimates, and people can be expected to reduce the
inconsistency by further bolstering their expectations for success. Study 3 tested
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
31
whether introducing anticipated regret would increase bolstering of the chosen
alternative, using a new condition in which participants expected to make a
consequential decision and were instructed to think about how they might feel if they
made a poor choice. I predicted that participants in this new decision/regret
condition would bolster their chosen alternative to a greater extent than participants
in the decision/consequences condition, who were not given anticipated regret
instructions.
Study 2 found that participants in the no-decision condition rated their chosen
alternative as highly as participants in the decision/no-consequences condition. That
finding may suggest that the expectation of making a non-consequential decision
does not increase the extent to which the chosen alternative is bolstered, but I
considered the alternative possibility that participants in the no-decision condition
behaved like participants in the decision/no-consequences condition because they
guessed that they would eventually make a decision. Therefore, in Study 3 I
tightened the instructions in the no-decision condition, primarily by emphasizing the
similarity of the second part of the study to the first part of the study, so that
participants would be more likely to think that their task merely involved rating sets
ofhorses.
I also strengthened the manipulation in the decision/no-feedback condition to
make it more difficult for participants to guess that their decision would involve
consequences. I amended the wording of the instructions to make it clearer to
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
32
participants that they were making a hypothetical decision without any consequences
associated with their choice.
Overview and predictions. I predicted that participants would bolster their
chosen alternative during the pre-decision phase and again in the post-decision
phase. Like Study 2, Study 3 included participants with varying degrees of
experience betting on horse races, and I predicted that experts would bolster their
chosen alternative to a greater extent than non-experts.
Study 3 included the same three experimental conditions as Study 2, though I
attempted to strengthen the manipulations in the no-decision and decision/no
consequences conditions. Thus, in the no-decision condition participants rated the
horses without expecting to make a decision, in the decision/no-consequences
condition they expected to make a non-consequential decision, and in the
decision/consequences condition they expected to make a consequential decision. I
predicted that participants in all conditions would bolster their chosen alternative, but
participants in the decision/no-consequences condition would bolster their chosen
alternative to a greater extent than participants in the no-decision condition, and
participants in the decision/consequences condition would bolster their chosen
alternative to a greater extent than participants in the decision/no-consequences
condition.
Study 3 also included two new experimental conditions. In the
decision/payoff condition participants expected to make a more highly consequential
decision (which could result in winning additional money), and in the decision/regret
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
33
condition they expected to make a consequential decision and were encouraged to
think about anticipated regret for a poor choice. I predicted that participants in these
two conditions would bolster their chosen alternative to a greater extent than
participants in the decision/consequences condition.
Methods
Participants. This study appeared on a website for psychology experiments. I
obtained 219 usable data sets from the website’s mailing list and 192 from an
advertisement on a horse racing website.
Materials. Two changes were made to all experimental conditions. First, to
satisfy a new Human Subjects Committee regulation, the introductory information
page told potential participants that they could obtain a single entry in the lottery
without participating in any part of our experiment. Second, in order to make the
payoffs seem valuable to participants in the decision/payoff condition, I doubled the
cash payoff associated with each horse (to $9.60, $14.00, $22.40, and $32.40), and
correspondingly doubled the number of lottery tickets associated with each horse
(10, 14, 23, 33).
No-decision condition. The no-decision condition was the same as in Study 2,
except that the instructions were modified to better ensure that the task in the second
part of the study appeared to be the same as the task in the first part of the study.
As in Study 2, the page inviting participants to enter the second part of the
study told them, “if you choose to participate in the second part of this study, you
will review another set of charts about four horses running in a simulated race and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
34
respond to questions about those horses”, and now added in parentheses, “like in the
first part of the study.” On the next page, the instructions which introduced the
second part of the study in all of the other conditions were replaced with excerpts
from the instructions used in the first part of the study. I used the same text as in the
first part of the study to tell participants that the horses were competing in a race, ask
them to browse through the charts, and tell them how to make their ratings. I thought
that repeating these simple instructions gave participants no reason to think that the
task in the second part of the study would be any different from the task they had just
completed in the first part of the study.
Decision/no-consequences condition. The decision/no-consequences
condition was the same as in Study 2, except that a few changes were made to the
instructions to help ensure that participants understood that their decision would not
involve any consequences.
As in Study 2, the page inviting participants to enter the second part of the
study told them that they would be able to choose the horse they would like to bet
on, but now explicitly stated that they “won’t actually get to bet in this race”, before
telling them that they would not find out the results of the race. On the next two
pages, the parts of the instructions reminding participants not to make a decision
about “the bet” or “your bet” because new information might becom e available were
re-phrased to remind them not to “make a decision about which horse you would bet
53
on.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
35
Decision/consequences condition. The decision/consequences condition was
the same as in Study 2.
Decision/payoff condition. The decision/payoff condition was the same as the
decision/consequences condition, except that the page inviting participants to enter
the second part of the study explained the opportunity to win a payoff in addition to
extra lottery tickets:
If you bet on the horse which wins the race,
(1) You will receive the monetary “payoff’ listed for your winning horse in
the information charts (i.e., we will send a check to your mailing address for
the amount you won in the virtual race), and
(2) We will convert your winning horse’s “payoff’ into extra lottery tickets,
by rounding up to the nearest whole dollar (so a $9.60 payoff becomes 10
extra tickets in the lottery).
Decision/regret condition. The decision/regret condition was the same as the
decision/consequences condition, except that in the pre-decision phase of the second
part of the study, participants were asked to think about the regret they would
experience if they made a poor choice.
On the first page of the second part of the study, the following paragraph was
added to the end of the instructions:
While you are looking over the charts and considering your bet, think about
the regret you’ll feel if you bet on a horse that doesn’t win the race.
Remember, if you bet on the horse that wins the race, w e’ll round its virtual
payoff up to a whole number and give you that many extra tickets in the
lottery. But if you bet on a horse that doesn’t win the race, you’ll miss out on
the extra tickets and the satisfaction of betting on the winner. So try your best
to pick the winner - and avoid regretting a poor choice after the race!
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
36
On the next page, the following reminder was added to the end of the
instructions: “Keep in mind the regret you’ll feel if you bet on a horse that doesn’t
win the race. Remember, you have to bet on the horse that wins the race in order to
get extra tickets in the lottery!”
Results
I predicted that participants would bolster their chosen alternative during the
pre-decision phase and again after placing their bets. A 2(chosen, non-chosen) by
3(base, pre-1, pre-2) by 2(expert, non-expert) by 5 (no-decision, decision/no
consequences, decision/consequences, decision/payoff, decision/regret) ANOVA
revealed a main effect of alternative, F{1, 380) = 339.72,/? < .001, a main effect of
time, F(1, 760) = 41.48,/? < .001, and an alternative by time interaction, F(2, 760) =
27.81,/? < .001. As predicted, linear trend analyses showed that ratings of the chosen
alternative increased over time within the pre-decision phase, F (l, 397) = 80.23,/? <
.001, but ratings of the non-chosen alternatives did not change over time, p < .2. Also
as predicted, after the bet had been placed, ratings of the chosen alternative increased
again, t{397) = 8.82,/? < .001, and ratings of the non-chosen alternatives tended to
decrease,/? < .1.
I predicted that experts (n = 158; 7% mailing list) would bolster their chosen
alternative to a greater extent than non-experts (n = 209; 95% mailing list; 40 could
not be classified). As predicted, linear trend analyses showed that ratings of the
chosen alternative increased overtime among experts, F (1, 156) = 32.31,/? < .001,
and among non-experts, F( 1, 240) = 47.78,/? < .001, and an alternative by expertise
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
37
interaction, F (l, 380) = 43.07,/? < .001, confirmed that experts bolstered their
chosen alternative to a greater extent than non-experts bolstered their chosen
alternative, 7(379) = 6.16, p < .001.
As predicted, linear trend analyses showed that ratings of the chosen
alternative increased over time in the no-decision, F {\, 68) = 12.06,/? < .01,
decision/no-consequences, 7^(1, 91) = 22.51,/? < .001, decision/consequences, F (l,
77) = 11.40,/? < .01, decision/payoff, F{1, 77) = 13.60,/? < .001, and decision/regret
conditions, F (1, 80) = 21.98,/? < .001, and an alternative by condition interaction,
F(4, 380) = 5.04,/? < .01, revealed that participants bolstered their chosen alternative
to a greater extent when the task was more important. As seen in Figure 4,
participants in the three consequential decision conditions (decision/consequences,
decision/payoff, decision/imagine) gave their chosen alternative very similar ratings
at each rating time. Pooling responses from the three consequential decision
conditions (and using t-tests which do not assume equal variances), I found that the
Chosen alternative was rated significantly higher in the consequential decision
conditions than in the no-decision condition at base, 7(108) = 3.31,/? < .01, pre-1,
7(108) = 2.66, p < .01, pre-2,7(96) = 2.84,/? < .01, and collapsing across the three
fating times, 7(109) = 3.71,/? < .001. The chosen alternative was rated marginally
higher in the consequential decision conditions than in the decision/no-consequences
condition at base, 7(153) = 1.82,/? < .08, and collapsing across the three rating times,
7 ( 149) = 1.66, /? < . 1, but not at pre-1, /? < .4, or pre-2, p < .2. The chosen alternative
was rated marginally higher in the decision/no-consequences condition than in the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
38
no-decision condition, collapsing across the three rating times, *(154) = 1.70,/? < .1,
but not at the individual rating times, jps < .3.
125
11.5
fl
1
S t
o
a
10.5
1 0
- - -o- - - Decision/Regret
------- Decision/Pay off
Decision/Consequences
— 0— Dec is ion/No-consequences
— No-decision
9.5
as
Base Pre-1 Pre-2 Post
Rating Time
Figure 5. Ratings of chosen horse's chance of winning a race. Ratings were made by participants in
no-decision, decision/no-consequences, decision/consequences, decision/payoff, and decision/regret
conditions at baseline (base), the first pre-choice rating time (pre-1), the second pre-choice rating time
(pre-2), and the post choice rating time (post).
I had expected participants in conditions involving greater task importance to
rate their chosen alternative higher than participants in conditions involving lesser
task importance, but I was surprised that they did so even at base, because the
instructions were the same in all conditions at base. To better understand the data, I
determined the proportion of participants who rated their ultimately chosen
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
39
alternative higher than their other alternatives (i.e., ranked it first) in each condition
at each rating time (Table 1).
Table 1
Proportion o f Participants Who Ranked Their Chosen Horse First by Experimental
Condition and Rating Time
Rating time
Condition Base Pre-1 Pre-2 Post
No-Decision 0.25 0.29 0.43 0.76
Decision/No-consequences 0.33 0.54 0.62 0.77
Decision/Consequences 0.43 0.49 0.62 0.80
Decision/Payoff 0.50 0.55 0.63 0.85
Decision/Regret 0.36 0.55 0.70 0.83
Table 1 reveals that the differences in ratings of the chosen alternative reflect
differences in the likelihood that the chosen alternative was ranked first. That is,
participants in the consequential decision conditions rated their chosen alternative
significantly higher than participants in the no-decision condition at each rating time
because they were significantly more likely to rank their chosen alternative first at
base, X 2 (1, N = 307) = 6.77, p < .01, pre-1, X 2 (1,N = 308) = 11.73 p < .01, and pre-
2, X2 (1, N = 301) = 9.92, p < .01. Similarly, participants in the consequential
decision conditions rated their chosen alternative marginally higher than participants
in the decision/no-consequences condition at base but not at pre-1 or pre-2 because
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
40
they were marginally more likely to rank their chosen alternative first at base, X 2 (1,
N = 321) = 3.38,/? < .07, but not at pre-1 or pre-2,p s > .3.
Although it was difficult to understand how the task importance instructions
introduced at pre-1 could have affected the earlier ratings of the chosen alternative at
base, it is easier to understand how they could have affected the likelihood of the
chosen alternative being ranked first at base. It seems likely that increasing task
importance increased rankings of the chosen alternative at base by increasing the
likelihood of participants choosing the alternative they had initially favored. Thus,
baseline rankings of the chosen alternative may have been higher in the
consequential decision conditions than in the no-decision condition because
participants who expected to make a consequential decision were more likely to
identify a favorite at base, bolster it during the pre-decision phase, and ultimately
choose it, while participants who did not expect to make a decision were more likely
to change their preferences over time, and ultimately choose an alternative they had
not preferred at base. Similarly, baseline rankings of the chosen alternative may have
been somewhat higher in the consequential decision conditions than in the
decision/no-consequences condition because participants who expected to make a
consequential decision were somewhat more likely to choose the alternative they had
ranked first at base.
The results of Study 3, then, suggest that increasing task importance increases
the likelihood of choosing an initial preference, as well as increasing the extent to
which the chosen alternative is bolstered during the pre-decision phase. Interestingly,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
41
Table 1 also shows that participants in the decision/regret condition were less likely
than participants in the other consequential decision conditions to rank their chosen
alternative first at base, suggesting that anticipated regret may prevent increasing
task importance from increasing the likelihood of maintaining an initial preference
and ultimately choosing it.
Discussion
As predicted by CST, participants bolstered the probabilities associated with
their chosen alternative during the pre-decision phase, and experts bolstered their
chosen alternative to a greater extent than non-experts. Participants in all conditions
bolstered their chosen alternative, and participants in conditions involving greater
task importance bolstered their chosen alternative to a greater extent than participants
in conditions involving lesser task importance. As in Study 2, participants who
expected to make a consequential decision bolstered their chosen alternative to a
greater extent than participants who did not expect to make a decision, and tended to
bolster to a greater extent than participants who expected to make a non-
consequential decision. Although in Study 2 participants who expected to make a
non-consequential decision did not bolster any more than participants who do not
expect to make a decision, when the experimental instructions were improved in
Study 3, participants in the decision/no-consequences condition did tend to bolster
their chosen alternative to a greater extent than participants in the no-decision
condition.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
42
I had also predicted that further increasing the potential consequences
associated with the choice, or introducing anticipated regret, would further increase
bolstering of the chosen alternative, but participants in the decision/payoff and
decision/regret conditions did not bolster their chosen alternative any more than
participants in the decision/consequences condition. The failure to increase
bolstering beyond the level found in the decision/consequences condition may be
partially attributable to a practical ceiling effect in the horse race paradigm. As noted
earlier in regard to the hindsight findings of Study 1 and seen across Studies 2 and 3,
participants seldom give their chosen horse pre-decision ratings higher than about
11,5 (novices) or 12 (experts) on the 15-point scale, suggesting that they may
understand that there are no sure bets at the race track and so may be unwilling to
claim maximal confidence in any horse. This finding, in turn, suggests that reality
constraints can impose limitations on the extent to which decision makers feel
comfortable bolstering their chosen alternative.
Table 1 suggests that the expectation of making a decision focused
processing on a single promising alternative. When participants did not expect to
make a decision (all conditions at base, and no-decision condition at pre-1 and pre-
2), less than half preferred the alternative they eventually chose (with the exception
o f the decision/payoff condition at base, which w ill be discussed below ). This was
particularly dramatic in the no-decision condition, where the eventually chosen
alternative was preferred by only 25% of participants at base, 29% at pre-1, and 43%
at pre-2. But when participants were told that they would eventually make a decision
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
43
(pre-1 and pre-2 in all conditions except no-decision), half or more preferred their
chosen alternative.
Table 1 also suggests that introducing the expectation of making a more
important decision focuses processing on an alternative which had been preferred
earlier (before the decision maker expected to make a decision). Notice that
participants who were informed that they would make a highly consequential
decision (decision/payoff) were about as likely to favor the alternative they
eventually chose at base (50%) as at pre-1 (55%). But participants who were
informed that they would make a less consequential decision
(decision/consequences) were somewhat less likely to favor the alternative they
eventually chose at base (43%) than at pre-1 (49%). And participants who were
informed that they would make a non-consequential decision (decision/no
consequences) were much less likely to favor the alternative they eventually chose at
base (33%) than at pre-1 (54%). The exception to the pattern occurs among
participants who expected to make consequential decisions and also anticipated
regretting a poor choice; those participants were much less likely to favor the
alternative they eventually chose at base (36%) than at pre-1 (55%), suggesting that
anticipated regret may lead decision makers to be less likely to maintain their initial
preference.
General Discussion
The present research shows how CST provides a single theoretical
framework that can account for many aspects of processing during decision making.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
44
By constructing a network representing concepts salient in a decision and the
relationships between them, and considering how activation would spread among
them, it is possible to predict several major findings in decision making behavior.
A network analysis of a horse racing bet decision (Figure 1) suggests that as
pre-decision processing proceeds and activation spreads among interconnected units
a coherent pattern can be expected to emerge, so that units which are consistent with
an emerging preference for one alternative become more highly activated. Consistent
with this view of decision making, in the present research estimates of the chance a
chosen horse would win a race were progressively bolstered throughout a pre
decision phase. Although this finding of pre-decision bolstering was predicted by
CST, it runs counter to the predictions of cognitive dissonance theory (Festinger,
1957, 1964) and action phase theoiy (Gollwitzer, 1990), which maintain that
alternatives are not re-evaluated before a decision, and rational decision theory
(Hogarth, 1987, von Neumann & Morgenstem, 1944), which maintains that the
probability estimates associated with an alternative should be independent of its
expected value and should not change during decision making.
Beyond making the basic point that a chosen alternative may be bolstered
before a decision, CST also informs our understanding of individual difference
factors which can affect pre-decision bolstering of a chosen alternative. CST
suggests that people with greater domain-specific expertise have stronger links
among concepts, enabling units to affect each other’s activation levels to a greater
extent, and thereby leading to greater bolstering of a chosen alternative. As predicted
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
45
by CST, in the present research participants with greater experience betting on horse
races bolstered their chosen alternative to a greater extent than participants with less
experience betting on horse races.
The finding that increasing individual expertise increased pre-decision
bolstering of a chosen alternative is consistent with previous research suggesting that
people who can be considered experts in a domain use more correlated attributes to
think about that domain and, as a result, express more polarized attitudes (Lusk &
Judd, 1988; c.f., Judd & Lusk, 1984; Millar & Tesser, 1986). The present findings
are also consistent with research on calibration in horse racing, which has found that
the racing public, and particularly racing experts, overestimate the chance
“longshots” have of winning races (Koehler, Brenner, & Griffin, 2002; Snyder,
1978; Thaler, & Ziemba, 1988). However, research on calibration is concerned with
how the racing public’s probability estimates compare with the predictions of an
efficient market or the actual probability of winning, while the present research is
concerned with how an individual’s probability estimates change over time.
CST also sheds light on the conditions under which people are more likely to
bolster their chosen alternative. CST suggests that any time people think about a set
of alternatives a preference emerges for one alternative and activation spreads to
consistent units in a coherent manner. As predicted, the present research found that
even participants who did not expect to make a decision bolstered their chosen
alternative within the pre-decision period. This finding converges with previous
research on legal decision making, which found coherence shifts even among
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
46
participants who thought about a legal case without expecting to render a verdict
(Simon et al, 2001).
CST also suggests that greater task importance will lead to more extensive
processing, so that units will have a greater affect on each other’s activation levels
and the chosen alternative will be bolstered to a greater extent. As predicted, in the
present research participants bolstered their chosen alternative to a greater extent
under conditions of greater task importance. Participants who expected to make a
consequential decision bolstered their chosen alternative to a greater extent than
participants who did not expect to make a decision, and in Study 3, participants who
expected to make a non-consequential decision took an intermediate position,
tending to bolster their chosen alternative to a greater extent than participants who
did not expect to make a decision but less than participants who expected to make a
consequential decision. This finding joins research based on choice certainty theory
(Mills, 1968), which found that increasing decision importance increased pre
decision re-evaluation of alternatives (Mills & Ford, 1995; O’Neal, 1971), against a
single study by Tyszka (1998), which found that increasing decision importance
decreased pre-decision re-evaluation of alternatives.
The present research on the effects of task importance on pre-decision
bolstering of a chosen alternative also led to some unanticipated findings concerning
the focus of pre-decision processing. An analysis of ranking of alternatives (Table 1)
suggested that when people expect to make a decision, they tend to focus on a
preferred alternative and eventually choose it, and when they expect to make a more
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
47
important decision, they are more likely to focus on and choose the alternative they
had initially preferred (unless they experience anticipated regret). The finding that
when people enter a decision making process they focus on a favorite alternative and
bolster it suggests that goals activated during decision making can increase biased
processing, in contrast with research on accountability (Lemer & Tetlock, 1999;
Tetlock, 1992) and accuracy goals (Lundgren & Prislin, 1998; Salvemini, Reilly, &
Smither, 1993), which suggest that goals decrease biased processing.
CST suggests that the effects of coherence-based reasoning extend beyond
pre-decision re-evaluation of alternatives to re-evaluation of alternatives in hindsight.
CST suggests that a unit representing outcome knowledge increases activation of
positively linked consistent units and decreases activation of negatively linked
inconsistent units, thereby increasing coherence within the system. In the case of a
horse race, a unit representing the outcome of the race increases activation of
consistent units, such as the perceived probability that the winning horse was going
to win, and decreases activation of inconsistent units, such as the perceived
probability that the other horses were going to win.
Consistent with CST’s view, when participants found out that a horse had
won a race they increased their estimates of the chance it had of winning the race
before it was run, and when participants found out that a horse had not won a race
they decreased their estimates of its chance of winning the race. Also consistent with
CST, participants who found out that their chosen horse did not win the race adjusted
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
48
down their estimates of the chance it had of winning the race, though they did not
bring their estimates all the way down to baseline levels.
The results of the present research can be related to findings from other lines
of work, which can also be explained by CST as instances of coherence-based
reasoning. The increasingly optimistic assessments of the favored alternative in the
present research are consistent with the negative relation between perceived risk and
benefit found in other research (Slovic, 2000; Slovic, Finucane, Peters, &
MacGregor, 2002). That is, the finding in the present research that gamblers
perceived progressively higher probabilities for a desirable outcome is consistent
with Slovic et al’s finding that people who perceive greater benefit in an object or
event tend to also perceive less risk in the same object or event. Slovic et al.
suggested that a reason why perceived benefits and risks are negatively correlated
may be that they are both produced by a single affect-based judgment, so that objects
which are judged favorably are perceived as having high benefits and low risks, and
objects which are judged unfavorably are perceived as having low benefits and high
risks. CST would suggest that the process by which a single affect-based judgment
affects both perceived risks and benefits is one of coherence-based reasoning,
whereby activation spreads from the single affect-based judgment to bolster risk and
benefit assessments which are consistent with it and to inhibit risk and benefit
assessments which are inconsistent with it.
The finding that participants reported increasingly optimistic assessments of
their chosen alternative may initially seem discrepant with research on optimistic
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
49
bias (Weinstein, 1980) which found that optimism decreased closer to the time for
performance or feedback (Gilovich, Kerr, & Medvec, 1993; Savitsky, Medvec,
Charlton, & Gilovich, 1998; Shepperd, Ouellette, & Fernandez, 1996; Taylor &
Shepperd, 1998). However, those experiments involved measures of confidence or
expected performance in potentially difficult or self-threatening tasks (e.g., coping
with a potential threat of medical illness, expected grades on an exam, prospects for
future jobs), where participants may have imagined a poor outcome which decreased
activation of inconsistent units representing confidence and strong performance
expectations. In contrast, in the present research participants may have focused on a
favorable event which increased activation of consistent units representing optimistic
performance expectations
Other research has shown other ways in which coherence-based reasoning
affects processing during decision making. CST suggests that activation spreads
from an emerging decision to units representing different interpretations of
ambiguous issues related to the decision, increasing activation of interpretations
which are consistent with the emerging decision and suppressing interpretations
which are inconsistent with the emerging decision. In their research on legal decision
making, Simon and his colleagues asked participants to indicate their interpretations
o f ambiguous factual issues first in isolated vignettes and then in the context o f
deciding a verdict in a legal case. It was found that the interpretations shifted towards
greater coherence with the verdict (Holyoak & Simon, 1999; Simon, et al., 2001;
Simon, Snow, & Read, 2003). Coherence shifts were similarly observed in a decision
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
50
task that emulated a multi-attribute decision task (Simon, Krawczyk & Holyoak, in
press; c.f., Keeney & Raiffa, 1976).
In the present research, I had expected that further increasing the
consequences associated with the task or introducing anticipated regret would further
increase the extent to which the chosen alternative was bolstered, but participants in
the decision/payoff and decision/regret conditions did not bolster their chosen
alternative more than participants in the decision/consequences condition. Future
research might further explore the effects of high consequences and anticipated
regret on pre-decision bolstering, using a paradigm less likely to involve an effective
ceiling on probability estimates. The possibility that increasing task importance may
increase bolstering of a favored alternative seems worthy of further investigation
because of its potential importance in real life applications. Distortion of preferences
can be unpleasant if it leads to misjudging one’s attitudes in everyday situations
(Tyszka, 1998), and distortion of preferences at the racetrack, where large sums of
money may be at stake, can produce costly consequences if it leads bettors to
overestimate a horse’s chance of winning a race.
Conclusions
CST provides an elegant explanation for many aspects of processing during
decision making. In the present research, CST correctly predicted that participants
would bolster the probability associated with the alternative they eventually chose,
that pre-decision bolstering would occur even when participants thought about their
alternatives without expecting to make a choice, and that increasing individual
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
51
expertise or task importance would increase bolstering. CST also suggests that
coherence-based reasoning helps assimilate outcome knowledge with other
information, so that hindsight estimates conform to the actual outcome of an event.
In the present research, CST correctly predicted that estimates of winners would
increase in hindsight and estimates of non-winners decreased in hindsight, even if
they had been bet to win.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
52
Alphabetized Bibliography
Brehm, J. W. (1956). Postdecision changes in the desirability of alternatives. Journal
o f Abnormal and Social Psychology, 52, 384-389.
Creyer, E., & Ross, W. T. (1993). Hindsight bias and inferences in choice: The
mediating effect of cognitive effort. Organizational Behavior and Human
Decision Processes, 55, 61-77.
Festinger, L. (1957). A theory o f cognitive dissonance. Stanford, CA: Stanford
University Press.
Festinger, L. (1964). Conflict, decision, and dissonance. Stanford, CA: Stanford
University Press.
Fischhoff, B. (1975). Hindsight is not equal to foresight: The effects of outcome
knowledge on judgment under uncertainty. Journal o f Experimental
Psychology: Human Perception and Performance, 7(3), 288-299.
Gerard, H. B., & White, G. L. (1983). Post-decisional reevaluation of choice
alternatives. Personality & Social Psychology Bulletin, 9(3), 365-369.
Gilovich, T., Kerr, M., & Medvec, V. H. (1993). Effect of temporal perspective on
subjective confidence. Journal o f Personality and Social Psychology, 64(4),
552-560.
Gollwitzer, P. M. (1990). Action phases and mind-sets. In E. T. Higgins & R. M.
Sorrentino (Eds.), Handbook o f motivation and cognition: Foundations o f
social behavior (Vol. 2, pp. 53-92). New York: Guilford Press.
Hawkins, S. A., & Hastie, R. (1990). Hindsight: Biased judgments of past events
after the outcomes are known. Psychological Bulletin, 107(3), 311-327.
Heider, F, (1958). The psychology o f interpersonal relations. Oxford: John Wiley.
Hogarth, R. M. (1987). Judgement and choice (2n d ed.). New York: John Wiley &
Sons.
Holyoak, K. J., & Simon, D. (1999). Bidirectional reasoning in decision making by
constraint satisfaction. Journal o f Experimental Psychology: General, 128, 3-
31.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
53
Holyoak, K. J. & Thagard, P. (1989). Analogical mapping by constraint satisfaction.
Cognitive Science, 13(3), 295-355.
Janis, I. L., & Mann, L. (1977). Decision making: A psychological analysis o f
conflict, choice, and commitment. New York: Free Press.
Judd, C. M. & Lusk, C. M. (1984). Knowledge structures and evaluative judgments:
Effects of structural variables on judgmental extremity. Journal o f
Personality and Social Psychology, 46(6), 1193-1207.
Keeney, R. L., & Raiffa, H. (1976). Decisions with multiple objectives: Preferences
and value tradeoffs. New York: Cambridge University Press.
Knox, R. E., & Inkster, J. A. (1968). Postdecision dissonance at post time. Journal o f
Personality and Social Psychology, 5(4), 319-323.
Koehler, D. J., Brenner, L., & Griffin, D. (2002). The calibration of expert judgment:
Heuristics and biases beyond the laboratory. In T. Gilovich, D. Griffin & D.
Kahneman (Eds.), Heuristics and biases: The psychology o f intuitive
judgment (pp. 686-715). New York: Cambridge University Press.
Lemer, J. S., & Tetlock, P. E. (1999). Accounting for the effects of accountability.
Psychological Bulletin, 125(2), 255-275.
Louie, T. A. (1999). Decision makers’ hindsight bias after receiving favorable and
unfavorable feedback. Journal o f Applied Psychology, 84(1), 29-41.
Lundgren, S. R. & Prislin, R. (1998). Motivated cognitive processing and attitude
change. Personality & Social Psychology Bulletin, 24(7), 715-726.
Lusk, C. M., & Judd, C. M. (1988). Political expertise and the structural mediators of
candidate evaluations. Journal o f Experimental Social Psychology, 24, 105-
126.
McClelland, J. L., & Rumelhart, D. E. (1986). Parallel distributed processing:
Explorations in the microstructure o f cognition. Vol. 2. Psychological and
biological models. Cambridge, MA: MIT Press.
Millar, M. G., & Tesser, A. (1986). Thought-induced attitude change: The effects of
schema structure and commitment. Journal o f Personality and Social
Psychology, 51(2), 259-269.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
54
Mills, J. (1968). Interest in supporting and discrepant information. In R. P. Abelson,
E. Aronson, W. J. McGuire, T. M. Newcomb, M. J. Rosenberg, & P. H.
Tannenbaum (Eds.), Theories o f cognitive consistency: A source hook.
Skokie, IL: Rand McNally.
Mills, J., & Ford, T. E. (1995). Effects of importance of a prospective choice on
private and public evaluations of the alternatives. Personality and Social
Psychology Bulletin, 21(3), 256-266.
O’Neal, E. (1971). Influence of future choice importance and arousal upon the halo
effect. Journal o f Personality and Social Psychology, 19(3) 334-340.
Read, S. J. & Miller, L. C. (1994). Dissonance and balance in belief systems: The
promise of parallel constraint satisfaction processes and connectionist
modeling approaches. In: R. C. Schank & E. Langer (Eds.), Beliefs,
reasoning, and decision making: Psycho-logic in honor o f Bob Abelson (pp.
209-235). Hillsdale, NJ: Lawrence Erlbaum Associates.
Read, S. J. & Marcus-Newhall, A. (1993). Explanatory coherence in social
explanations: A parallel distributed processing account. Journal o f
Personality & Social Psychology, 65(3), 429-447.
Read, S. J., Vanman, E. J., & Miller, L. C. (1997). Connectionism, parallel constraint
satisfaction processes, and Gestalt principles: (Re)introducing cognitive
dynamics to social psychology. Personality & Social Psychology Review,
1(1), 26-53.
Salvemini, N. J., Reilly, R. R , & Smither, J. W. (1993). The influence of rater
motivation on assimilation effects and accuracy in performance ratings.
Organizational Behavior & Human Decision Processes, 55(1), 41-60.
Savitsky, K., Medvec, V. H., Charlton, A. E., & Gilovich, T. (1998). “What, me
worry?”: Arousal, misattribution, and the effect of temporal distance on
confidence. Personality and Social Psychology Bulletin, 24(5), 529-536.
Shepperd, J. A., Ouellette, J. A., & Fernandez, J. K. (1996). Abandoning unrealistic
optimism. Performance estim ates and the temporal proximity o f self-relevant
feedback. Journal o f Personality and Social Psychology, 70(4), 844-855.
Simon, D., & Holyoak, K. J. (2002). Structural dynamics of cognition: From
consistency theories to constraint satisfaction. Personality & Social
Psychology Review, 6(4), 283-294.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
55
Simon, D., Krawczyk, & Holyoak, K. J. (in press). Construction of preferences by
constraint satisfaction. Psychological Science.
Simon, D., Pham, L. B., Le, Q. A., & Holyoak, K. J. (2001). The emergence of
coherence over the course of decision making. Journal o f Experimental
Psychology: Learning, Memory, and Cognition (Special Issue), 27(5), 1250-
1260.
Simon, D., Snow, C. J., & Read, S. J. (2003) The Redux o f Cognitive Consistency
Theories: Evidence Judgments by Constraint Satisfaction. Unpublished
manuscript, University of Southern California, Los Angeles, CA.
Shultz, T. R. & Lepper, M. R. (1996). Cognitive dissonance reduction as constraint
satisfaction. Psychological Review, 103(2), 219-240.
Slovic, P. (2000). The perception o f risk. London: Earthscan Publications Ltd.
Slovic, P., Finucane, M., Peters, E., & MacGregor, D. G. (2002). The affect
heuristic. In T. Gilovich, D. Griffin & D. Kahneman (Eds ), Heuristics and
biases: The psychology o f intuitive judgm ent (pp. 686-715). New York:
Cambridge University Press.
Smith, S. B. (1998). The complete idiot’ s guide to betting on horses. New York:
Alpha Books.
Snyder, W. W. (1978). Horse racing: Testing the efficient markets model. The
Journal o f Finance, 53(4), 1109-1118.
Spellman, B. A., & Holyoak, K. J. (1992). If Saddam is Hitler then who is George
Bush? Analogical mapping between systems of social roles. Journal o f
Personality & Social Psychology, 62, 913-933.
Spellman, B. A., Ullman, J. B., & Holyoak, K. J. (1993). A coherence model of
cognitive consistency: Dynamics of attitude change during the Persian Gulf
War. Journal o f Social Issues, 49(4), 147-165.
Svenson, O. (1992). Differentiation and consolidation theory of human decision
making: A frame of reference for the study of pre- and postdecision
processes. Acta Psychologica, 80, 143-168.
Svenson, O, Rayo, A. O., Andersen, M, Sandberg, A, & Svahlin, I. (1994). Post
decision consolidation, as a function of the instructions to the decision maker
and of the decision problem. Acta Psychologica, 87, 181-197.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
56
Taylor, K. M., & Shepperd, J. A. (1998). Bracing for the worst: Severity, testing, and
feedback timing as moderators of the optimistic bias. Personality and Social
Psychology Bulletin, 24(9), 915-926.
Tetlock, P. E. (1992). The impact of accountability on judgment and choice: Toward
a social contingency model. Advances in Experimental Social Psychology,
25, 331-376.
Thaler, R. H., & Ziemba, W. T. (1988). Parimutual betting markets: Racetracks and
lotteries. Journal o f Economic Perspectives, 2(2), 161-174.
Thagard, P. (2000). Coherence in thought and action. Cambridge, MA: The MIT
Press.
Tyszka, T. (1998). Two pairs of conflicting motives in decision making.
Organizational Behavior and Human Decision Processes, 74(3), 189-211.
von Neumann, J, & Morgenstem, O. (1944). Theory o f Games and Economic
Behavior. Princeton: Princeton University Press.
Weinstein, N. D. (1980). Unrealistic optimism about future life events. Journal o f
Personality & Social Psychology, 59(5) 806-820.
Zeelenberg, M. (1999). Anticipated regret, expected feedback and behavioral
decision making. Journal o f Behavioral Decision Making, 12, 93-106.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Abstract (if available)
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Using a causal network to model memory for an unscripted social sequence
PDF
The effects of increasing cognitive load on self -report and dual -task measures of mental effort during problem solving
PDF
The effects of anxiety and age on attentional processes: Emotional Stroop and dot -probe in younger and older adults
PDF
Solving real problems in adulthood: Effects of age and knowledge on complex problem -solving performance
PDF
The out-group must not be so bad after all: Generalization of the effects of disclosure, typicality and salience on intergroup attributions
PDF
Countering the effects of selective exposure
PDF
Parents' psychological distress and children's temperament as moderators of the effects of daily work stress on family interactions
PDF
The impact of state anxiety on categorization: Categorical inclusiveness, stability, and inferential perceptions
PDF
The effects of race and gender on respondent attitudinal perceptions of acquaintance rape
PDF
Collective efficacy, anxiety, creativity/innovation and work performance at the team level
PDF
Effects of test interpretation style and need for cognition on elaboration, favorability, and recall
PDF
The correspondence bias: The applicability of contextual information
PDF
The effect of standardized test feedback on high school students
PDF
Interhemispheric interaction in bilateral redundancy gain: Effects of physical similarity
PDF
The impact of parental attachment, expressiveness, parental substance use, and childhood maltreatment on young adults' interpersonal relations
PDF
The mismatched couple: Effects on depressive symptomatology and romantic relationship functioning
PDF
Prostate cancer patients and their partners: Effectiveness of a brief communication enhancement intervention prior to undergoing radical prostatectomy
PDF
Task factors in attentional startle reflex modification
PDF
The impact of working memory and rule complexity on classification strategy adoptions
PDF
Plausibility and grammatical agreement
Asset Metadata
Creator
Brownstein, Aaron
(author)
Core Title
The effects of coherence -based reasoning on betting decisions
Degree
Doctor of Philosophy
Degree Program
Psychology
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
OAI-PMH Harvest,psychology, cognitive,psychology, social
Language
English
Contributor
Digitized by ProQuest
(provenance)
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c16-637162
Unique identifier
UC11340320
Identifier
3116671.pdf (filename),usctheses-c16-637162 (legacy record id)
Legacy Identifier
3116671.pdf
Dmrecord
637162
Document Type
Dissertation
Rights
Brownstein, Aaron
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA
Tags
psychology, cognitive
psychology, social