THE LEAST LIKELY OF BEHAVIORS The Least

The Least Likely of Behaviors 1
Running Head: THE LEAST LIKELY OF BEHAVIORS
The Least Likely of Behaviors:
When Memory for Atypical Past Behavior biases Behavioral Predictions
Carey K. Morewedge
Carnegie Mellon University
Alexander Todorov
Princeton University
WORKING PAPER: DO NOT CITE OR DISTRIBUTE
Corresponding Author:
Carey K. Morewedge
Department of Social and Decision Sciences
Carnegie Mellon University
208 Porter Hall
Pittsburgh, PA 15213
email: [email protected]
phone: 412.268.6079
Words:
References:
Figures:
Tables:
3,916
30
1
2
The Least Likely of Behaviors 2
ABSTRACT
Individuals’ atypical past behavior is both memorable and by definition unrepresentative
of how they typically behave. Despite explicitly considering an atypical past behavior to
be most unusual, observers were most likely to attend to, remember, and use that
behavior to predict an individual’s future action in three experiments. Observers did not
rely on an actor’s most unusual behavior, however, when that actor was a member of a
group. As group size increased, observers were less likely to use an actor’s atypical past
behavior and more likely to use the group’s normative behavior as a basis for their
predictions. The results elucidate the kinds of case-based information most likely to be
cognitively available and used to predict the behavior of individuals, and why behavioral
predictions made for individual persons are often less accurate than predictions made for
group members.
Word count: 138
Keywords: Attribution, Availability, Judgment and Decision Making, Social Prediction
The Least Likely of Behaviors 3
The Least Likely of Behaviors:
When Memory for Atypical Past Behaviors bias Behavioral Predictions
Many have wondered whether a colleague will remember to bring a bottle of wine to
their dinner party or have paused to think whether the baby sitting next to them in coach
will cry during the flight, despite knowing well what the majority of colleagues and
infants will do. Although people are generally accurate when predicting how the majority
of people will behave in any given situation, people are often inaccurate when predicting
the behavior of specific individuals, whether familiar or strangers (e.g., Davis, Hoch, &
Ragsdale, 1986; Epley & Dunning, 2000; Nisbett & Kunda, 1985). Even experts
inconsistently apply the information they consider most important to predictions made for
individuals, to the extent that experts are less accurate than linear models derived from a
small sample of their predictions (Dawes, 1979; Meehl, 1954).
How can people exhibit “impressive” accuracy when predicting the behavior of
groups of people (Nisbett & Kunda, 1985, p. 297), despite being inaccurate when
predicting the behavior of individuals? Different classes of information are used to
predict the behavior of individuals and groups of people (Kahneman & Tversky, 1973).
When predicting the behavior of an individual, people imagine what they would do and
correct for the peculiarities of that situation and individual (use case-based information),
whereas when predicting the behavior of a group, people rely more closely on base-rates
and the behavior people normally exhibit in similar situations (use distributional
information) (Epley & Dunning, 2000; Epley, Keysar, Van Boven, & Gilovich, 2004;
Nickerson, 1999; c.f., Hamill, Wilson, & Nisbett, 1981).
The Least Likely of Behaviors 4
Although knowing more about a situation or person should improve the accuracy of
predictions, inaccuracies may be caused by the kinds of information most readily
available when predictions are made (Hall, Ariss, & Todorov, 2007; Taylor, 1982;
Tversky & Kahneman, 1973). Situational constraints and norms thought to be most
influential in a situation are often misconstrued (e.g., Miller & Ratner, 1998; Monin &
Norton, 2003; Nisbett & Borgida, 1975; Van Boven, Dunning, & Loewenstein, 2000;
Wilson, Wheatley, Meyers, Gilbert, & Axsom, 2000; for a review, see Dunning, 2007).
Predictions may also be based on cognitively accessible traits and past behaviors that are
unrepresentative of an individual’s typical behavior.
One’s most atypical past experiences come first to mind and are used to predict and
choose similar future experiences (Morewedge, Gilbert, & Wilson, 2005; Wirtz, Kruger,
Scollon, & Diener, 2003). We propose that similarly, individuals’ most atypical past
behaviors come first to mind and are used to predict their future behavior. Whereas
individuals are remembered in terms of their most unusual traits and features (Hastie &
Kumar, 1979; McGuire, McGuire, Child, & Fujioka, 1978), group members remembered
in terms of the group’s typical traits and behaviors (Hamilton & Rose, 1980; Rothbart,
Sriram, & Davis-Stitt, 1996; Susskind, Maurer, Thakkar, Hamilton, & Sherman, 1999).
Atypical past behaviors should thus be more likely to influence predictions made for
persons perceived as individuals than as group members.
The experiments reported tested two predictions drawn from these assumptions: First,
predictions of an individual’s behavior should correspond to their most atypical past
behavior because atypical past behaviors should be more available than more typical past
behaviors. Second, predictions made for a group member should be less likely to
The Least Likely of Behaviors 5
correspond to their most atypical behavior than predictions made for an individual.
Additionally, predictions were expected to exhibit diminishing sensitivity to a group
member’s atypical behavior as the size of their group increased, rather than only
exhibiting sensitivity to a person’s status as an individual or group member.
EXPERIMENT 1
Some participants observed an individual’s selection of consumer items, one
selection being particularly unusual. All participants then predicted which item she would
select from a novel choice set that included items corresponding to each of her previous
selections. Participants who observed her previous selections were expected to be more
likely to predict that she would select the item corresponding to her most unusual
previous choice than participants unaware of her previous selections.
Method
Participants
One hundred and thirty-seven people in Boston, MA (72 female; Mage = 23.3, SD
= 4.9) received $5 for participating.
Procedure
Participants in an informed condition saw a woman’s photograph alongside the
pens she selected from four different sets of expensive pens. Each set consisted of four
pens and was displayed for 6s. After 3s a frame appeared around the pen she selected. Of
her four selections, one was more atypical (i.e., pink with a tassel) than the other three
pens (i.e., black, blue, and silver). Participants in a control condition did not see her
selections.
The Least Likely of Behaviors 6
All participants were then shown a choice set containing four pens that
corresponded to the colors of the previously selected pens, and predicted the likelihood
that she would choose each of those pens on five-point scales with endpoints, Not at all
Likely (1) and Extremely Likely (5). Next, participants indicated which pen they thought
she would choose. Participants also assessed the typicality of every pen in the experiment
(including the pens from the choice set) on five-point scales marked with endpoints, Not
at all Unusual (1) and Extremely Unusual (5). Predictions and typicality judgments were
counterbalanced across participants in this experiment, but judgment order did not
influence any of the results. Set presentation, likelihood judgment, and typicality
judgment order were random in all experiments.
Results
One participant’s responses were more than three standard deviations from the
mean and were not included in any subsequent analyses.
Predictions
As non-parametric tests revealed asymmetries in the pens that participants in both
the informed and control conditions predicted that the woman would choose, χ2 (3, n =
65) = 39.19, p < .001 and χ2 (3, n = 72) = 14.78, p = .002, predictions were compared
between groups with ANOVA as suggested by Rosenthal and Rosnow (1991). Whereas a
majority of participants in the informed condition predicted that she would choose the
pen corresponding to her most atypical selection, a majority of controls predicted that she
would chose a more typical pen, F(1, 135) = 3.91, p = .05, ηp2 = .03 (see Table 1).
Surprisingly, participants in the informed condition were more than twice as likely
The Least Likely of Behaviors 7
(58.5%) to predict that she would choose the pen corresponding to her most atypical
previous choice as would be expected if predictions were random (25%).
Likelihood Judgments
The same pattern of results was obtained for likelihood ratings. A 2 (information:
informed vs. controls) x 2 (corresponding choice: atypical vs. typical) ANOVA with
repeated measures on the last factor revealed that although participants in both conditions
considered her most likely to choose the pen corresponding to her most atypical previous
selection, Fcorresponding choice(1, 135) = 76.33, p < .001, ηp2 = .36, participants in the
informed condition were more likely than controls to predict that she would choose that
pen and less likely to predict that she would choose a pen corresponding to her more
typical selections, Finteraction (1, 135) = 24.17, p < .001, ηp2 = .15.
-------------------------------Insert Table 1 About Here
-------------------------------Typicality Judgments
As a manipulation check, the atypicality ascribed to each of the woman’s previous
selections was analyzed within a 2 (information: information vs. controls) x 2 (previous
choice: atypical vs. others) ANOVA with repeated measures on the last factor. A
significant main effect of choice suggested that participants considered the atypical pen
less typical than the other chosen pens, F(1, 135) = 471.71, p < .001 , ηp2 = .77. There
was no significant effect of information or interaction. To ensure that predictions were
based on the atypicality of her past rather than her potential choices, the atypicality
ascribed to each potential selection in the choice set was analyzed within a 2
The Least Likely of Behaviors 8
(information: information vs. controls) x 2 (corresponding potential choice: atypical vs.
others) mixed ANOVA, which yielded no significant effects or interaction, all Fs < 1.
Discussion
Participants predicted that an individual’s future behavior would most likely
correspond with their most atypical previous behavior, despite explicitly recognizing its
atypicality. The experiments that follow attempted to explain why and when this occurs.
EXPERIMENT 2
Experiment 2 varied the target’s group membership and the presentation format of
their past behavior to test two hypotheses. First, predictions of an individual’s behavior
were expected to be more likely to correspond to their most atypical behavior than
predictions of a group member’s behavior. Second, if atypical past behavior is more often
used to predict individuals’ future behavior because unusual stimuli are most likely to be
noticed and remembered (Morewedge et al., 2005; Wolfe & Horowitz, 2004), situations
in which perceivers can selectively allocate their attention to an individual’s past
behaviors (e.g., when behaviors are presented simultaneously) should exacerbate this
effect as less attention will be devoted to more typical behaviors than when perceivers
must allocate their attention equally (e.g., when behaviors are presented serially).
Method
Participants
One hundred and sixty-eight people in Boston, MA (81 female, Mage = 23.8, SD =
15.2) received $5 for participating.
Procedure
The Least Likely of Behaviors 9
In a between-subjects design, participants observing an individual saw the pens a
single woman selected and made predictions as in Experiment 1. Participants observing a
group member saw the pens that the woman selected from the four sets and saw the pens
seven other similarly dressed women selected from those four sets and a choice set. The
pens selected from the first four sets by the other women were randomly determined. The
pens selected from the choice set were randomly sampled from responses to a pretest in
which participants were shown the choice set and indicated pen they thought would be
most frequently selected (N = 71).
In a serial format condition each set was presented for 3s followed by the
presentation of each woman’s selection for 3s. In a simultaneous format condition each
set was presented for 3s in a random order and then all the selections made by each
woman were presented simultaneously for 12s. After seeing the pen selections, all
participants made predictions and typicality judgments on scales as described in
Experiment 1.
Results
Predicted Choice
As asymmetries were found with respect to the predicted choice, χ2 (3, n = 168) =
80.10, p < .001, a 2 (group status: individual vs. group member) x 2 (presentation format:
serial vs. simultaneous) between-subjects ANOVA was used to compare predictions
between conditions. It revealed a significant main effect of group status such that
participants were more likely to predict that the target would choose the pen
corresponding to her atypical previous choice when she was an individual than a group
member, F(1, 164) = 18.00, p < .001, ηp2 = .10. A significant main effect of presentation
The Least Likely of Behaviors 10
format was also found, such that participants in the simultaneous presentation condition
(M = .65, SD = .48) were more likely than participants in the serial presentation condition
(M = .43, SD = .50) to predict that the target would choose the pen corresponding to her
atypical previous choice, F(1, 164) = 10.33, p = .002, ηp2 = .06. There was no significant
group status x presentation format interaction, F < 1.
Likelihood Judgments
A 2 (group status: individual vs. group member) x 2 (presentation format: serial
vs. simultaneous) x 2 (corresponding choice: atypical vs. typical) mixed ANOVA with
repeated measures on the last factor revealed a main effect of corresponding choice, such
that participants considered the pen corresponding to her atypical choice to be most likely
to be chosen, F(1, 164) = 89.54, p < .001, ηp2 = .35. A significant group status x
corresponding choice interaction revealed that individuals were perceived to be more
likely than group members to choose the pen corresponding to their previous atypical
choice and less likely to choose one of the other pens, F(1, 164) = 18.03, p < .001, ηp2 =
.10. The atypical pen appeared more likely to be chosen in the simultaneous than serial
format conditions, t(166) = 1.64, p = .10 (see Table 2), although the interaction of pen
and format did not reach significance, F(1, 164) = 2.40, p = .12. No other significant
effects were found.
-------------------------------Insert Table 2 About Here
-------------------------------Typicality Ratings
The Least Likely of Behaviors 11
The typicality ascribed to the target’s previous selections was analyzed within a 2
(group status: individual vs. group member) x 2 (presentation format: serial vs.
simultaneous) x 2 (previous choice: atypical x typical) ANOVA with repeated measures
on the last factor, which revealed a significant main effect of choice such that the atypical
pen chosen (M = 4.29, SD = 1.02) was considered to be more atypical than the other pens
chosen (M = 1.67, SD = .60), F(1, 164) = 867.29, p < .001, ηp2 = .84. Interestingly, a
format x pen interaction, F(1, 164) = 3.91, p = .05, ηp2 = .02, suggested that the atypical
pen was considered more atypical when presented in the simultaneous (M = 4.45, SD =
.84) than serial format (M = 4.12, SD = 1.16), t(166) = 2.14, p = .03. No other significant
effects or interactions were found.
Discussion
Predictions were influenced by both the target’s group status and the presentation
format of the target’s past behavior as expected. Participants were more likely to predict
that a target’s behavior would correspond to her most atypical past behavior when she
was depicted as an individual rather than a group member, and when her past selections
were displayed simultaneously rather than serially. Participants were thus more likely to
rely on an atypical past behavior to guide their predictions for an individual as the
relative availability of that behavior increased, but were considerably less likely to rely
on an atypical past behavior to guide their predictions for a group member.1 Indeed,
participants were again more than twice as likely (67.7%) as chance (25%) to predict that
an individual would choose the pen corresponding to their most atypical previous choice,
whereas participants were only slightly more likely (37.7%) than chance to predict that a
The Least Likely of Behaviors 12
group member would choose the pen corresponding to their most atypical previous
choice.
EXPERIMENT 3
The previous experiments suggest that predictions made for individuals are likely
to be based on that person’s most atypical behavior because it is most readily available,
whereas predictions made for group members are more likely to be based on the group’s
normative behavior. Experiment 3 examined when predictors cease to base their
predictions on a person’s most atypical behavior. Predictors may either use a group’s
normative behavior to guide their predictions when the size of a group reaches a specific
threshold (e.g., if a group consists of eight or more individuals), or perceivers may be
more likely to use the normative behavior of a group to guide their predictions as group
size increases in a linear fashion.
Method
Participants
One hundred and forty-three people in Alston, MA (74 female, Mage = 23.1, SD =
4.6) received $5 for participating.
Procedure
In four conditions differing according a target’s group status, participants
observed women select pens and predicted a target’s future selection in a serial
presentation format. In an individual condition, participants observed a target’s previous
selections and predicted which pen she would select from a choice set as in Experiments
1-3. In other conditions, participants saw the target selections in addition to the selections
other women made from same four sets as the target and from the choice set, before
The Least Likely of Behaviors 13
predicting which pen the target would select from the choice set. In a dyad condition,
participants saw the choices of the target and one other woman. In a small group
condition, participants saw the choices of the target and three other women, and in a
large group condition a participants saw the choices of the target and seven other woman.
The other women made n-1 choices different from the target in the first four pen sets (i.e.,
1, 3, or 7, respectively). None chose the atypical pen from the choice set. All chose the
same different pen, in order to observe how a strong group norm influenced predictions.
Finally, all participants made predictions and typicality judgments on scales as described
in Experiment 1.
Results
Predicted Choice
As asymmetries were found with respect to the predicted choice, χ2 (3, n = 143) =
23.52, p < .001, predictions were compared with a between-subjects ANOVA, which
revealed a significant effect of group status on predicted choice, F(1, 139) = 4.19, p =
.007, ηp2 = .08. As illustrated by Figure 1, as the size of the group increased, participants
were less likely to predict the target would choose the pen corresponding to her most
atypical previous choice, Flinear contrast(1, 139) = 12.44, p = .001.
-------------------------------Insert Figure 1 About Here
-------------------------------Likelihood Ratings
The same pattern of results was obtained for likelihood ratings. A 4 (group status)
x 2 (corresponding pen: atypical vs. typical) ANOVA with repeated measures on the last
The Least Likely of Behaviors 14
factor yielded significant main effects of group status and corresponding pen, F(1, 139) =
3.27, p = .02, ηp2 = .07 and F(1, 139) = 8.92, p = .003, ηp2 = .06, respectively. As
illustrated by Figure 1, a significant group status x choice interaction indicated that the
predicted likelihood of the target choosing the atypical pen decreased as group size
increased, whereas the predicted likelihood of the target choosing another pen increased
as group size increased, F(1, 139) = 7.46, p < .001, ηp2 = .14; Flinear contrast (1, 139) = 7.46,
p < .001, ηp2 = .14 (see Figure).
Typicality Judgments
A 4 (group status) x 2 (previous choice) mixed ANOVA revealed significant main
effects of group status and previous choice on atypicality judgments, but no interaction,
F(3, 139) = 2.64, p = .05, ηp2 = .08 and F(1, 139) = 4.19, p = .007, ηp2 = .08, and F < 1,
respectively. Although participants considered the atypical pen more unusual than the
pens previously chosen irrespective of condition as in previous experiments, target’s
previous choices also appeared more unusual in the individual (M = 3.16, SE = .08) and
large group (M = 3.15, SE = .09) conditions than in the dyad condition (M = 2.87, SE =
.09), ps ≤ .025 (Fisher’s LSD), which was not predicted.
Discussion
Again, participants were twice more likely (51.2%) than chance to predict that an
individual would select the pen corresponding to her most atypical previous choice,
whereas participants were no more likely than chance to predict that a member of a large
group would make that selection (16.7%). More important, predictions of a person’s
future behavior corresponded less with their most atypical behavior and corresponded
more with their group’s behavior as group size increased linearly rather than at a specific
The Least Likely of Behaviors 15
threshold. The correspondence between predictions and a target’s most atypical past
behavior was inversely related to the size of the group she belonged to, whereas the
correspondence between predictions and the normative behavior of the group increased
with the size of the group she belonged to.
GENERAL DISCUSSION
It may be easiest to remember and describe a person using the traits and behaviors
they do not share with other people, but the unusual traits and behaviors of an individual
may not be the best information to use to predict their future behavior. A nephew who
once spilled wine on a couch or was caught shoplifting in high school is unlikely to spill
or steal anything from one in the near future. Yet the ease with which these instances are
retrieved from memory may tempt one to cover the furniture and hide the silverware
before the next family gathering. Although atypical behaviors may serve a good index of
what behavior a person is capable of (Skowronski & Carlston, 1992), they are by
definition unrepresentative of how a person will behave in most situations. Despite
explicitly identifying a person’s past behavior as their most atypical, however,
participants in three experiments were most likely to remember and use atypical past
behavior to predict an individual’s future behavior.
We did not employ a design measuring the accuracy of predictions, but a reliance
on the base-rates of behavior in a situation is likely to engender greater accuracy
(Kahneman & Tversky, 1973).2 Predictors are willing to use base-rates to predict the
behavior of a population and generally accurate when they do, but are quick to ignore
base-rates when provided with individuating information (Nisbett & Borgida, 1975). The
distinction between predictions made for populations and individuals, however, may not
The Least Likely of Behaviors 16
be as dichotomous as has been previously suggested (Epley & Dunning, 2000).
Participants used the normative behavior of a group to predict the behavior of an
individuated person (Experiments 2 and 3), and were more likely to do so as the group’s
size increased in a linear rather than dichotomous fashion. Furthermore, the findings
suggest that predictors do not update behavioral predictions made for individuals when
provided with base-rate information because their typical behaviors are unlikely to be the
most readily available behaviors.
Considering multiple rather than single events decreases the impact of recalled
atypical events on forecasts of one’s future affect (Morewedge et al., 2005). Observing
multiple persons’ behavior appears to similarly decrease predictors’ reliance on atypical
past behavior when predicting a person’s future behavior. Perhaps the notorious
inaccuracy of clinicians, spouses, and admissions committees (Davis et al., 1986; Dawes,
1979; Meehl, 1954) could be improved by having them consider what any group of
clients, partners, or students would do under those circumstances than having them
predict the fate of an individual. Ironically, it appears that the lens through which people
are perceived widens as one thinks of them less as individuals and more as one in a
multitude.
The Least Likely of Behaviors 17
FOOTNOTE
1. To ensure that the difference in predictions made by participants exposed to
simultaneous and serial present formats was due to between condition differences in the
availability of the target’s previous choices, an ancillary study was conducted (N = 52).
Participants were only presented with the target’s previous choices in a serial or
simultaneous format exactly as in Experiment 2, and were later asked to identify which
pens she did and did not choose after a delay in which they complete unrelated tasks (M =
25.72 min, SD = 5.93). A 2 (presentation format: serial vs. simultaneous) x 2 (previous
choice: atypical vs. typical) ANOVA with repeated measures on the last factor revealed a
main effect of previous choice, such that participants were more likely to correctly
identify the atypical pen as having been previously chosen than the typical pens F(1, 50)
= 32.17, p < .001, ηp2 = .39. A significant presentation format x previous choice
interaction suggested that whereas participants were equally likely to correctly identify
the atypical pen in both formats (Mserial = .89, SD = .31; Msimultaneous = .88, SD = .34),
participants were more likely to correctly identify the typical pens when presented
serially than simultaneously (Mserial = .73, SD = .26; Msimultaneous = .41, SD = .36), F(1, 50)
= 7.49, p = .01, ηp2 = .12; t(50) = .20, p = .85, r = .03 and t(50) = 3.62, p = .001, r = .46,
respectively. No other significant effects were found.
2. A reliance on stereotypes should not improve the accuracy of predictions, however, as
they are unlikely to be corrected when inaccurate (Hamilton & Rose, 1980).
The Least Likely of Behaviors 18
AUTHORS’ NOTE
We gratefully acknowledge the support of NIA grant P30 AG024361 to the Princeton
Center for the Demography of Aging. We thank Dana Graef, Amy Krilla, Amit Kumar,
Michael Norton, and Jill Swencionis for their assistance in the execution of these
experiments, and Daniel Kahneman and Robyn Dawes for helpful suggestions.
The Least Likely of Behaviors 19
References
Dawes, R. M. (1979). The robust beauty of improper linear models in decision making.
American Psychologist, 34, 571-582.
Dunning, D. (2007). Prediction: The inside view. In A. W. Kruglanski & E. T. Higgins
(Eds.), Social Psychology, Handbook of Basic Principles, (pp. 69-91). New York:
Guilford.
Davis, Hoch, & Ragsdale, (1986). An anchoring and adjustment model of spousal
predictions. Journal of Consumer Research, 13, 25-37.
Epley, N., & Dunning, D. (2000). Feeling “Holier than thou”: Are self-serving
assessments produced by errors in self- or social prediction? Journal of
Personality and Social Psychology, 79, 861-875.
Epley, N., Keysar, B., Van Boven, L., & Gilovich, T. (2004). Perspective taking as
egocentric anchoring and adjustment. Journal of Personality and Social
Psychology, 87, 327-339.
Hamill, R. C., Wilson, T. D., & Nisbett, R. E. (1980). Ignoring sample bias: Inferences
about populations from atypical cases. Journal of Personality and Social
Psychology, 39, 578–589.
Hamilton, D. L., & Rose, T. L. (1980). Illusory correlation and the maintenance of
stereotypic beliefs. Journal of Personality and Social Psychology, 39, 832-845.
Hall, C. C., Ariss, L., & Todorov, A. (2007). The illusion of knowledge: When more
information reduces accuracy and increases confidence. Organizational Behavior
and Human Decision Processes, 103, 277-290.
The Least Likely of Behaviors 20
Hastie, R., & Kumar, P. A. (1979). Person memory: The processing of consistent and
inconsistent person information. Journal of Personality and Social Psychology,
37, 25–38.
Kahneman, D., & Tversky, A. (1973). On the psychology of prediction. Psychological
Review, 80, 237-251.
McGuire, W. J., McGuire, C. V., Child, P., & Fujioka, T. (1979). Salience of ethnicity in
the spontaneous self-concept as a function of one’s ethnic distinctiveness in the
social environment. Journal of Personality and Social Psychology, 36, 511-520.
Meehl, P. E. (1954). Clinical versus statistical prediction: A theoretical analysis and a
review of the evidence. Minneapolis: University of Minnesota Press.
Miller, D. T., & Ratner, R. K. (1998). The disparity between the actual and assumed
power of self-interest. Journal of Personality and Social Psychology, 74, 53-62.
Monin, B., & Norton, M.I. (2003). Perceptions of a fluid consensus: Uniqueness bias,
false consensus, false polarization and pluralistic ignorance in a water
conservation crisis. Personality and Social Psychology Bulletin, 29, 559-567.
Morewedge, C. K., Gilbert, D. T., & Wilson, T. D. (2005). The least likely of times: How
remembering the past biases forecasts of the future. Psychological Science, 16 (8),
626-630.
Nickerson, R. (1999). How we know—and sometimes misjudge—what others know:
Imputing one’s own knowledge to others. Psychological Bulletin, 125, 737-759.
Nisbett, R. E., & Borgida, E. (1975). Attribution and the psychology of prediction.
Journal of Personality and Social Psychology, 32, 932-943.
The Least Likely of Behaviors 21
Nisbett, R. E., & Kunda, Z. (1985). Perception of social distributions. Journal of
Personality and Social Psychology, 48, 297-311.
Rosenthal, R., & Rosnow, R. L. (1991). Essentials of Behavioral Research: Methods and
Data Analysis (2nd Edition). Boston, MA: McGraw-Hill.
Rothbart, M., Sriram, N., Davis-Stitt, C. (1996). The retrieval of typical and atypical
category members. Journal of Experimental Social Psychology, 32, 309-336.
Sherman, S. J., Castelli, L., & Hamilton, D. L. (2002). The spontaneous use of a group
typology as an organizing principle in memory. Journal of Personality and Social
Psychology, 82, 328-342.
Skowronski, J. J., & Carlston, D. E. (1989). Negativity and extremity biases in
impression formation: A review of explanations. Psychological Bulletin, 105,
131–142.
Susskind, J. L., Maurer, K., Thakkar, V., Hamilton, D. L., & Sherman, S. J. (1999).
Perceiving individuals and groups: Expectancies, dispositional inferences and
causal attributions. Journal of Personality and Social Psychology,76, 181-191.
Taylor, S. E. (1982). The availability bias in social perception and interaction. In D.
Kahneman and A. Tversky (Eds.) Judgment under uncertainty: Heuristics and
biases, (pp. 190-200). New York: Cambridge University Press.
Tversky, A., & Kahneman, D. (1973). Availability: A heuristic for judging frequency and
probability, Cognitive Psychology, 5, 207-232.
Van Boven, L., Dunning, D., & Loewenstein, G. (2000). Egocentric empathy gaps
between owners and buyers: Misperceptions of the endowment effect. Journal of
Personality and Social Psychology, 79, 66–76.
The Least Likely of Behaviors 22
Wilson, T. D., Wheatley, T. P., Meyers, J. M., Gilbert, D. T., & Axsom, D. (2000).
Focalism: A source of durability bias in affective forecasting. Journal of
Personality and Social Psychology, 78, 821-836.
Wirtz, D., Kruger, J., Scollon, C. N., & Diener, E. (2003). What to do on spring break?
The role of predicted, on-line, and remembered experience in future choice.
Psychological Science, 14, 520–524.
Wolfe, J. M., & Horowitz, T. S. (2004). What attributes guide the deployment of visual
attention and how do they do it? Nature Reviews Neuroscience, 5, 495-501.
The Least Likely of Behaviors 23
Table 1
Predicted and Previous Choices by Information about Previous Choices in Experiment 1 (N = 137)
Information
No Information
_____________________
_____________________
Measure
Atypical Pen Typical Pens
Atypical Pen Typical Pens
Predicted Choice
38 (58.5%)
27 (41.5%)
30 (41.7%)
42 (58.3%)
Predicted Likelihood of Choosing
4.05 (.98)a
2.51 (.60)b
3.39 (1.19)c
2.96 (.69)d
Atypicality of Previous Choices
4.00 (1.08)a
1.65 (.71)b
3.74 (1.13)a
1.65 (.71)b
Atypicality of Corresponding Potential Choices
2.22 (1.14)a
2.13 (.81)a
2.12 (1.12)a
2.11 (.77)a
Note: If predictions were made randomly, 25% of participants should choose the atypical pen and 75% of participants should choose
one of the three typical pens. Means within rows that do not share a common subscript differ significantly (p ≤ .05) according to
simple effects t-tests. Standard deviations and percentages appear in parentheses.
The Least Likely of Behaviors 24
Table 2
Predicted and Previous Choices by Presentation Format and Group Status in Experiment 2 (N = 168)
Simultaneous Format
Group Status
Serial Format
Mean
Atypical Pen
Typical Pens
Atypical Pen
Typical Pens
Atypical Pen
Typical Pens
Individual
36 (78.3%)
10 (21.7%)
27 (57.4%)
20 (41.6%)
63 (67.7%)
30 (32.3%)
Group Member
19 (50.0%)
19 (50.0%)
9 (24.3%)
28 (75.7%)
28 (37.3%)
47 (62.7%)
Combined
55 (65.5%)
29 (34.5%)
36 (42.9%)
48 (57.1%)
Individual
4.20 (.98)a
2.55 (.69)b
3.94 (1.09)a
2.57 (.63) b
4.06 (1.04)
2.57 (.66)
Group Member
3.66 (1.28)b
2.88 (.71)c
3.30 (1.24)ab
2.93 (.62)ac
3.48 (1.27)
2.91 (.67)
Mean
3.95 (1.15)
2.71 (.71)
3.65 (1.20)
2.73 (.65)
Predicted Choice
Predicted Likelihood of Choosing
Atypicality of Previously Chosen
The Least Likely of Behaviors 25
Individual
4.43 (.94)a
1.59 (.61)b
4.17 (1.05)a
1.75 (.57) b
4.30 (1.00)
1.72 (.60)
Group Member
4.47 (.73)a
1.64 (.60)b
4.05 (1.29)a
1.59 (.61) b
4.27 (1.06)
1.62 (.60)
Mean
4.45 (.84)
1.67 (.61)
4.12 (1.16)
1.68 (.59)
Note: If predictions were made randomly, 25% of participants should choose the atypical pen and 75% of participants should choose one of the
three typical pens. Means within rows that do not share a common subscript differ significantly according to t-tests (p ≤ .05). Standard deviations
and percentages are in parentheses.
The Least Likely of Behaviors 26
Figure 1. Participants in Experiment 3 were more likely to predict and more often predicted that an individual (n = 41) would choose a
pen corresponding to her most atypical previously choice than a member of a dyad (n = 38), small group consisting of 4 people (n =
32), or large group consisting of 8 people (n = 38). Note that if predictions were made randomly, 25% of participants should choose
the atypical pen and 75% of participants should choose one of the three typical pens.