Personality and affective forecasting: Trait introverts underpredict the hedonic benefits of acting extraverted

Personality and Affective Forecasting: Trait Introverts Underpredict theHedonic Benefits of Acting Extraverted

John M. Zelenski, Deanna C. Whelan, Logan J. Nealis, Christina M. Besner, Maya S. Santoro,and Jessica E. Wynn

Carleton University

People report enjoying momentary extraverted behavior, and this does not seem to depend on trait levelsof introversion–extraversion. Assuming that introverts desire enjoyment, this finding raises the question,why do introverts not act extraverted more often? This research explored a novel explanation, that traitintroverts make an affective forecasting error, underpredicting the hedonic benefits of extravertedbehavior. Study 1 (n � 97) found that trait introverts forecast less activated positive and pleasantaffect and more negative and self-conscious affect (compared to extraverts) when asked to imagineacting extraverted, but not introverted, across a variety of hypothetical situations. Studies 2–5(combined n � 495) found similar results using a between-subjects approach and laboratorysituations. We replicated findings that people enjoy acting extraverted and that this does not dependon disposition. Accordingly, the personality differences in affective forecasts represent errors. Inthese studies, introverts tended to be less accurate, particularly by overestimating the negative affectand self-consciousness associated with their extraverted behavior. This may explain why introvertsdo not act extraverted more often (i.e., they overestimate hedonic costs that do not actuallymaterialize) and have implications for understanding, and potentially trying to change, introverts’characteristically lower levels of happiness.

Keywords: extraversion, affective forecasting, negative affect, subjective well-being, judgment

The personality trait of extraversion is a strong predictor ofhappiness (Diener, Suh, Lucas, & Smith, 1999). Moreover, at thestate level of analysis, behaving in extraverted ways also seems tocause pleasant emotions, and this relationship does not depend ondispositional extraversion (Fleeson, Malanos, & Achille, 2002). Inother words, even people who tend to behave in an introverted waymore often (trait introverts) appear to enjoy their less frequentperiods of extraverted behavior. Such findings suggest the poten-tial of increasing happiness by increasing extraverted behavior,and this notion is bolstered by experimental studies showing thatextraverted behavior seems to cause positive affect even when

manipulated in the lab (e.g., Fleeson et al., 2002; McNiel &Fleeson, 2006; Zelenski, Santoro & Whelan, 2012).

Assuming that people want to be happy, such findings also raisethe question, why do introverts not act more extraverted? Affectiveforecasting errors, predictions about future emotions that lead tohedonically suboptimal decisions, may provide part of the expla-nation. To illustrate, imagine a time when you did not want toattend a social event, thinking it would not be very fun or evenstressful, and then had a fantastic time after being dragged there bya friend. We suspect that this happens to trait introverts more thantrait extraverts. That is, introverts may fail to fully anticipate howmuch fun social or extraverted behavior will be, even though theyare likely to enjoy it as much as extraverts. We examined theplausibility of this explanation across five studies.

The strong link between trait extraversion and positive affect isone of the most established findings in personality psychology andis found across many methods and sample characteristics, thoughthe magnitude of the relationship can depend on these factors(Lucas & Fujita, 2000; Steel, Schmidt, & Schulz, 2008; Zelenski,Sobocko, & Whelan, in press). Moreover, there are likely manythings that contribute to the relationship. For example, extravertsappear to be more sensitive to pleasant rewarding stimuli (Smillie,Cooper, Wilt, & Revelle, 2012; Zelenski & Larsen, 1999), main-tain pleasant moods longer (Hemenover, 2003), spend more timein pleasant social situations (Srivastava, Angelo, & Vallereux,2008), and may simply possess a higher set point for positive affect(Lucas & Baird, 2004).

Although such findings can be traced back to causal theories ofextraversion that emphasize stable individual differences (e.g.,Eysenck’s, 1967, arousal or Gray’s, 1981, reward sensitivity),

This article was published Online First April 15, 2013.John M. Zelenski, Deanna C. Whelan, Logan J. Nealis, Christina M.

Besner, Maya S. Santoro, and Jessica E. Wynn, Department of Psychology,Carleton University, Ottawa, Ontario, Canada.

Logan J. Nealis is now at the Department of Psychology, DalhousieUniversity, Halifax, Nova Scotia, Canada; Maya S. Santoro is now at theSan Diego State University/University of California, San Diego JointDoctoral Program in Clinical Psychology.

This research was supported by a Social Sciences and HumanitiesResearch Council of Canada Standard Research Grant awarded to John M.Zelenski. We thank Carleton University Happiness Laboratory membersfor assistance with this research, especially Nathalie Gillen and HelenaSillanpaa.

Correspondence concerning this article should be addressed to John M.Zelenski, Department of Psychology, Carleton University, 1125 ColonelBy Drive, Ottawa, ON K1S 5B6, Canada. E-mail: [email protected]

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

Journal of Personality and Social Psychology © 2013 American Psychological Association2013, Vol. 104, No. 6, 1092–1108 0022-3514/13/$12.00 DOI: 10.1037/a0032281

1092

Fleeson (2001) has highlighted the value of examining within-person variation in moment-to-moment thoughts, feelings, andbehaviors. People show wide variation in their behavior over time.Although trait extraverts behave in extraverted ways more oftenthan introverts do, trait introverts still spend much of their timeacting sociable, bold, and active (Fleeson & Gallagher, 2009). It iswhen we average these behaviors over time that trait differencesclearly emerge. Fleeson’s approach also assumes an isomorphismbetween traits and states such that relationships observed at onelevel should also exist at the other. In the domain of extraversion,this appears to be true, at least for positive affect. That is, in bothexperience-sampling and laboratory studies, acting extraverted inthe moment is strongly associated with positive affect, and thisbasic finding has now been replicated many times (Fleeson et al.,2002; Gallagher, Fleeson, & Hoyle, 2011; McNiel & Fleeson,2006; McNiel, Lowman, & Fleeson, 2010; Wilt, Noftle, Fleeson,& Spain, 2012; Zelenski et al., 2012). It seems most people enjoybehaving in extraverted ways more than behaving in introvertedways.

Assuming that introverted people seek pleasure, it is worthconsidering explanations for introverts’ seemingly suboptimal be-havioral choices. Research has already examined a few explana-tions, often considering potential costs of behaving counter toone’s disposition. For example, Little (2008) suggested that peoplewill behave against their first nature (i.e., physiological disposi-tions) to pursue personally meaningful goals (e.g., personal proj-ects). Although successful goal pursuit might enhance well-being,Little argued that it also detracts from well-being because it ispsychologically taxing and challenges the nervous system. Intro-verts may enjoy acting extraverted (particularly in service ofimportant goals), but without care, it may exhaust them physicallyand psychologically. Although an intuitively appealing idea, theresearch conducted to this point has not provided much empiricalsupport for costs associated with acting extraverted.

More specifically, the costs of concurrent negative affect, egodepletion (as cognitive fatigue), effort, and reduced subjectiveauthenticity have been explored in the domain of counterdisposi-tional behavior related to introversion–extraversion. Negative af-fect has received the most attention; it is worth noting that peoplecan experience simultaneous pleasant and unpleasant feelings(Larsen, McGraw, & Cacioppo, 2001), possibly something likestressful exuberance in the case of introverts acting extraverted.Across four published studies examining negative affect, nonefound evidence that acting counterdispositionally produces nega-tive affect (McNiel & Fleeson, 2006; McNiel et al., 2010; Zelenskiet al., 2012). Some of these studies lacked adequate power todetect the key interaction, but it seems unlikely that introvertsexperience elevated negative affect when acting extraverted, atleast over the course of 20 min in laboratory studies.

Counterdispositional behavior might exact nonaffective costs;for example, it may cause ego depletion, consuming self-regulatory resources (cf. Vohs, Baumeister, & Ciarocco, 2005). Inother words, it might take more effort or active self-control tobehave in nonhabitual ways. Zelenski et al. (2012) found someevidence for this idea using the Stroop task as a measure ofdepletion; however, effects were limited to trait extraverts assignedto behave like introverts. That is, following counterdispositionalbehavior in an interview or group task, trait extraverts performedworse on a Stroop task, but trait introverts did not. Gallagher et al.

(2011) observed a very similar asymmetric pattern when theyassessed subjective effort. Following group discussions, trait ex-traverts reported that behaving introverted was considerably moreeffortful than behaving extraverted, whereas trait introverts re-ported roughly equal effort when assigned to enact either intro-verted or extraverted behavior. Thus, some evidence supports theidea that behaving counterdispositionally requires effort and canproduce fatigue; however, the asymmetric patterns suggest thatthis may not be a problem for trait introverts.

Finally, even if trait introverts enjoy acting extraverted withoutaffective, cognitive, or self-control costs, they might still avoidsuch behavior because it feels inauthentic. Studies of authenticityhave yet to manipulate introverted or extraverted behavior in alaboratory setting, but the available evidence argues against sub-jective authenticity as a cost of introverts’ extraverted behavior.That is, Fleeson and Wilt (2010) found that although trait intro-verts perceived their introverted behavior as more authentic inretrospective reports, simultaneous ratings (using the experience-sampling method or ESM) suggest that most people, even traitintroverts, reported feeling more authentic when also reportingmore extraverted behavior. Somewhat counterintuitively, peopleseemed to feel more authentic (in the moment) when their behaviorwas extraverted, emotionally stable, agreeable, conscientious, andopen, and these associations did not depend on disposition. Thus,given the more intuitive retrospective results, introverts mightanticipate that behaving extraverted would feel inauthentic, butwhen doing so in the moment, they might actually feel differently.

Similar to our speculation about how people predict momentaryfeelings of authenticity, we suggest affective forecasts might helpexplain why introverts do not behave like extraverts more often,despite the fact that doing so might make them happier. People useanticipated emotions to make decisions (Mellers & McGraw,2001; Mellers, Schwartz, & Ritov, 1999), but they often mispredictthe intensity and duration of future emotions (i.e., make affectiveforecasting errors; Gilbert, Pinel, Wilson, Blumberg, & Wheatley,1998; Wilson & Gilbert, 2005). These errors then lead to decisionsthat fail to maximize happiness. For example, subjects predictedthat they would enjoy eating spaghetti more if they were hungrywhen making the prediction and if they were considering eating itin the evening, rather than the morning (Gilbert, Gill, & Wilson,2002). However, neither of these factors predicted the actualenjoyment of eating spaghetti the next day; people mispredictedtheir enjoyment (Gilbert & Wilson, 2000). Had they been requiredto make a decision in advance, it likely would have been subop-timal (Kermer, Driver-Linn, Wilson, & Gilbert, 2006; Loewen-stein, 1996; Loewenstein, O’Donoghue, & Rabin, 2003). Movingfrom breakfast to social behavior, Dunn, Biesanz, Human, andFinn (2007) found that subjects anticipated preferring interactingwith a romantic partner but that they actually preferred interactingwith strangers. They suggested that subjects underestimated thebenefits of positive self-presentation and that there was moreself-presentation when interacting with strangers. Such positiveself-presentation seems similar to acting extraverted (particularlyin a North American context), but Dunn et al. did not exploreindividual differences in their study.

In fact, very little research has examined the role of personalityin affective forecasts (Hoerger, Chapman, Epstein, & Duberstein,2012). Given the many links among personality, emotion, andjudgment (Rusting, 1998; Zelenski, 2008), personality differences

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

1093PERSONALITY AND AFFECTIVE FORECASTS

nonetheless seem likely. There is some evidence that people highin emotional intelligence make more accurate affective forecasts(Dunn, Brackett, Ashton-James, Schneiderman, & Salovey, 2007;Hoerger et al., 2012). On the other hand, Quoidbach and Dunn(2010) also suggested personality neglect, the notion that person-ality differences influence actual experience but that these differ-ences are often not anticipated by people, leading to forecastingerrors (see also Tomlinson, Carmichael, Reis, & Aron, 2010). Inother words, people neglect their dispositional levels of happinesswhen making predictions about future happiness, particularlywhen predictions are tied to a particular event, thus drawing onepisodic rather than semantic memory. In these studies, personalitywas unrelated to affective forecasts.

The notion of personality neglect notwithstanding, there are stillgood reasons to suspect that differences in dispositionalintroversion–extraversion will influence at least some affectiveforecasts. First, Quoidbach and Dunn (2010) did not report onextraversion, but rather general happiness, neuroticism, and opti-mism. Although the term personality neglect suggests a broadergeneralizability, it is possible that effects do not extend to extra-version. Trait extraversion predicts other judgments about futureaffective events and experiences. For example, introverts predictlower probabilities for positive events occurring in the future(Zelenski & Larsen, 2002), evaluate hypothetical positive eventsless positively (Uziel, 2006), and anticipate experiencing lesshappiness in pleasant situations (Lucas & Diener, 2001) comparedto extraverts. These studies did not measure actual experience andthus are somewhat ambiguous with regard to the extent of error,but they do suggest personality differences in forecasts. We areaware of only one study that examined trait extraversion with bothforecasts and experiences; participants were asked to earlier fore-cast and then report on experience 2 days after Valentine’s Day(Hoerger & Quirk, 2010). Extraversion predicted both forecastsand experiences of pleasant emotions, and it remained a significantpredictor of forecasts even after controlling for experience, furthersuggesting that personality influences forecasts beyond any actualdifferences in experience. Thus, despite some conflicting results, itseems plausible that personality could influence affective fore-casts.

Going beyond previous work, we suggest that context mightalso be important. That is, whether introverts over- or underesti-mate positive affect may depend on the particular situation orbehavior they are forecasting. More specifically, we suggest thatintroverts may act less extraverted or select less-social situationsbecause they do not anticipate these as being quite as positive astrait extraverts do. Similarly, trait introverts might view momen-tary introverted behavior as more positive than trait extraverts do.To the extent that such affective forecasts determine choices aboutbehavior, they may help explain why introverts do not act extra-verted more often despite seeming to enjoy it.

To explore this idea, we conducted a series of studies that assessdispositional introversion–extraversion and affective forecasts formomentary introverted and extraverted behavior. Study 1 took awithin-person approach with participants forecasting experiencefor both introverted and extraverted behavior across a series ofsituations. In Studies 2–5, participants were randomly assigned(between-subjects) to act introverted or extraverted, and then theymade affective forecasts for the studies’ activities. (Some studiesalso included a control condition with no acting instructions.)

These studies also assessed participants’ actual experience, allow-ing us to confirm the hedonic benefits of acting extraverted andassess the objective accuracy of forecasts. We hypothesized thatpersonality would predict affective forecasts such that extravertswould more strongly prefer extraverted behavior and that intro-verts would predict more positive affect for introverted behavior,at least compared to extraverts—the forecasting preference forintroverted behavior may only be relative. Affective forecastingdifferences are often found in magnitude rather than direction—reality also has some influence—and data strongly indicate thatacting extraverted is more enjoyable. We assessed affect in termsof the widely used constructs of activated positive affect andnegative affect (cf. the Positive and Negative Affect Schedule[PANAS]; Watson, Clark, & Tellegen, 1988). Although widelyused, these scales focus on high arousal affects, states that mightbe more or less appealing depending on dispositional extraversion(Kuppens, 2008; Rusting & Larsen, 1995). Thus, we includedanother pleasant affect scale (i.e., happy, pleased) to assess a morearousal-neutral form of pleasure. We also created an ad hoc mea-sure of self-conscious affect (i.e., self-conscious, embarrassed), asthis might be what introverts want to avoid. Self-consciousnesscould also be understood as a proxy for feelings/predictions ofinauthenticity (cf. Fleeson & Wilt, 2010). We predicted that per-sonality differences would be conceptually similar across positiveaffect, negative affect, pleasantness, and self-consciousness indi-cators.

Study 1

Method

Participants. A total of 137 participants from our universitysubject pool were recruited for this online study titled Personalityand Predicting Emotions. All participants completed an onlinemass testing questionnaire (including trait extraversion) earlier asa prerequisite for this study. A test was implemented to ensure thatparticipants were reading the instructions. Specifically, for onesection, participants were instructed to indicate a “7” for allresponses on this page. Participants with completion times of lessthan 8 min (the average was 16), failure on the test, or missing/invalid mass testing (trait) data were removed from analyses. Thefinal sample consisted of 97 participants (72 female, 24 male, oneunknown). All received course credit for their participation.

Materials.Acting condition. Two different acting instructions served as

a within-subject manipulation. Similar to Fleeson et al. (2002),acting extraverted instructions asked participants to imagine actingbold, talkative, energetic, active, assertive, and adventurous,whereas the acting introverted instructions asked participants toimagine acting reserved, quiet, lethargic, passive, compliant, andunadventurous.

Settings. Five different settings were presented to participantswith both introverted and extraverted instructions. Three of thesettings were described as a psychology lab setting (cf. Studies2–5) where the differing activities include being interviewed byanother participant, completing a jigsaw puzzle with three otherparticipants, and a group task involving planning a day together.The remaining two settings were chosen to increase generalizabil-

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

1094 ZELENSKI ET AL.

ity to particularly social and nonsocial environments: a silent areaof the library and a party that is very crowded.

Affective forecasting questionnaire. Each scenario was de-scribed (a setting plus acting instruction) and participants wereasked to forecast how they would feel during that scenario. Par-ticipants were instructed to take a moment to think about thescenario and to anticipate their thoughts and feelings within thescenario. Affective forecasts were made with participants rating 12mood adjectives on a 7-point Likert-type scale (from 1 � Veryslightly or not at all to 7 � Extremely or a lot). We created forecastscales for activated positive affect as the average of excited andinterested; activated negative affect as the average of worried,fearful, upset, nervous, and distressed (cf. the PANAS constructs);pleasantness as the average of happy and pleased; and self-consciousness as the average of embarrassed and self-conscious.Average Cronbach’s alphas across settings were .82 for positiveaffect, .88 for pleasantness, .92 for negative affect, and .79 forself-consciousness, with ranges between settings of .58 to .85, .82to .92, .88 to .93, and .76 to .89, respectively.

Trait extraversion. Trait extraversion was assessed in masstesting, along with many other unrelated measures. Mass testing isan online survey that occurs at the beginning of the academic yearand takes approximately 50 min to complete. Trait extraversionwas measured using the eight-item extraversion subscale of Gold-berg’s Mini-Markers (Saucier, 1994). Participants rated how welleach adjective described them on a 7-point Likert-type scale (from1 � Not at all to 7 � Very Well).

Procedure. The study was completed online via the Psychol-ogy Department’s SONA software portal. Following informedconsent and collection of student numbers (to link to mass testing),participants completed the affective forecasting questionnaire for11 scenarios. Each setting was listed twice, once for each actingcondition (5 settings � 2 acting conditions). The 11th affectiveforecasting questionnaire was the test to determine if participantswere carefully reading instructions. The order of these 11 pageswas randomized across participants.

Results

Multilevel modeling (MLM) was used to examine the within-person comparison of acting conditions. Intraclass correlation co-efficients (ICCs) revealed a significant percentage of variability atLevel 2: 18%, 10%, 20%, and 16% for positive affect, pleasant-ness, negative affect, and self-consciousness, respectively (all ps �.001). The significant ICCs validate the use of MLM. HLM 6(Raudenbush, Bryk, & Congdon, 2004) was used to examine 4two-level models with participants as the Level 2 unit. The depen-dent variables were forecasted emotions (positive affect, pleasant-ness, negative affect, and self-consciousness), one per model.

The Level 1 variable was acting condition, and it was dummycoded and uncentered. To determine the best model, randomintercepts only and random intercepts with random slopes modelswere compared with likelihood ratio tests. Across the four depen-dent variables, the random intercepts with random slopes modelssignificantly improved model fit over random intercepts onlymodels, �2(2) � 83.32, �2(2) � 98.94, �2(2) � 115.78, and�2(2) � 105.32, for positive affect, pleasantness, negative affect,and self-consciousness, respectively (all ps � .001). The Level 2

variable was trait extraversion, and it was grand-mean centered.The final model for interpretation is as follows:

Level 1 : Yij � �0j � �1j�Acting Condition� � e.

Level 2 : �0j � �00 � �01�Trait Extraversion� � r0.

�1j � �10 � �11�Trait Extraversion� � r1.

We hypothesized an interaction (�11) between acting instruc-tions and trait extraversion on each of the dependent variables (i.e.,the relationship between forecasted affect and personality willdiffer by acting condition). Furthermore, for positive affect andpleasantness, we posited that the simple slope for the actingextraverted condition would be positive (i.e., extraverts forecastmore positive affects compared to introverted participants). Fornegative affect and self-contentiousness, we posited a negativeslope for the acting extraverted simple slope (i.e., introverts fore-cast more negative affects relative to extraverts).

For positive affect, there was a significant main effect of traitextraversion on forecasts (� � .18, SE � .08, p � .03), indicatingthat extraverted participants forecast more positive affect relativeto introverted participants. There was also a main effect of condi-tion such that participants forecast more positive affect in theacting extraverted condition (� � 1.29, SE � .14, p � .001)relative to the acting introverted condition. More important to thecentral hypothesis, there was a significant interaction betweencondition and trait extraversion (� � .47, SE � .16, p � .004). Inthe acting introverted condition, disposition was unrelated to fore-casts of positive affect; the simple slope was flat (� � .01, SE �.11, p � .50). In contrast, the acting extraverted simple slope waspositive (� � .49, SE � .11, p � .001; see Figure 1); participantsscoring high on dispositional extraversion forecast more positiveaffect for acting extraverted than the more introverted participants.The results for positive affect fully support our hypotheses.

For pleasantness, there was a significant main effect of traitextraversion on forecasts (� � .19, SE � .07, p � .01), indicatingthat extraverted participants forecast more pleasantness relative tointroverted participants. There was also a main effect of conditionsuch that participants forecast more pleasantness in the acting

1

2

3

4

5

6

Introverts Extraverts

Fore

cast

ed P

ositi

ve A

ffect

Acting Introverted Acting Extraverted

Figure 1. Trait extraversion and acting condition interaction on fore-casted positive affect (Study 1).

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

1095PERSONALITY AND AFFECTIVE FORECASTS

extraverted condition (� � .80, SE � .16, p � .001) relative to theacting introverted condition. More important to the central hypoth-esis, there was a significant interaction between condition and traitextraversion (� � .70, SE � .17, p � .001). In the acting intro-verted condition, disposition was unrelated to forecasts of pleas-antness; the simple slope was flat (� � �.11, SE � .11, p � .30).In contrast, the acting extraverted simple slope was positive (� �.54, SE � .11, p � .001; see Figure 2); participants scoring highon dispositional extraversion forecast more pleasantness for actingextraverted than the more introverted participants. The results forpleasantness fully support our hypotheses.

For negative affect, there was a significant main effect of traitextraversion on forecasts (� � �.18, SE � .07, p � .01), indi-cating that extraverted participants forecast less negative affectrelative to introverted participants. There was a marginally signif-icant main effect of condition (� � .26, SE � .16, p � .10) suchthat participants forecast more negative affect in the acting extra-verted condition. Most importantly there was also a significantinteraction between acting condition and trait extraversion (� ��.51, SE � .16, p � .002). In the acting introverted condition,disposition was unrelated to forecasted negative affect; the simpleslope was flat (� � .04, SE � .09, p � .50). In contrast, the actingextraverted simple slope was negative (� � �.47, SE � .11, p �.001; see Figure 3); participants scoring high on dispositionalextraversion forecast less negative affect for acting extravertedthan more introverted participants. Said another way, introverts,relative to extraverts, forecast more negative affect for actingextraverted. The results for negative affect fully support our hy-potheses.

For self-consciousness, there was no significant main effect oftrait extraversion on forecasts (� � �.11, SE � .08, p � .17),indicating that forecasts of self-consciousness were similar be-tween introverted and extraverted participants. There was a maineffect of condition (� � .46, SE � .19, p � .02) such thatparticipants forecast more self-consciousness in the acting extra-verted condition. There was a significant interaction between traitextraversion and acting condition (� � �.57, SE � .16, p � .001).In the acting introverted condition, disposition was unrelated to

forecasted self-consciousness; the simple slope was flat (� � .13,SE � .11, p � .26). In contrast, the acting extraverted simple slopewas negative (� � �.44, SE � .12, p � .001; see Figure 4);participants scoring high on dispositional extraversion forecast lessself-consciousness for acting extraverted than the more introvertedparticipants. The results for self-consciousness fully support ourhypotheses.

Discussion

The primary goal of Study 1 was to explore the idea that traitextraversion would predict differences in people’s affective fore-casts of enacting introverted and extraverted behaviors. Resultsstrongly supported hypotheses, with trait introverts forecasting lesspositive affect, less pleasantness, more negative affect, and moreself-consciousness for extraverted behavior, compared to trait ex-traverts. In addition, although acting extraverted was generallyforecast to be higher in positive affect and pleasantness (comparedto acting introverted), trait introverts also forecast it to be substan-

1

2

3

4

5

6

Introverts Extraverts

Fore

cast

ed P

leas

antn

ess

Acting Introverted Acting Extraverted

Figure 2. Trait extraversion and acting condition interaction on fore-casted pleasantness (Study 1).

1

2

3

4

5

6

Introverts Extraverts

Fore

cast

ed N

egat

ive

Affe

ct

Acting Introverted Acting Extraverted

Figure 3. Trait extraversion and acting condition interaction on fore-casted negative affect (Study 1).

1

2

3

4

5

6

Introverts Extraverts

Fore

cast

ed S

elf-c

onsc

ious

ness

Acting Introverted Acting Extraverted

Figure 4. Trait extraversion and acting condition interaction on fore-casted self-consciousness (Study 1).

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

1096 ZELENSKI ET AL.

tially higher on negative affect and self-consciousness, comparedto introverted behavior. In other words, introverts seem to haveanticipated that the extraverted behavior could be unpleasant forthem. These findings might help explain why introverts do notpursue extraverted behavior more often. Although this study didnot assess actual experience, previous work suggests that theseaffective forecasts were likely errors; in both lab and experience-sampling studies, virtually everyone enjoys extraverted behavior,with very few personality differences in this pattern (and nosubstantial negative affect). To the extent that introverts are mak-ing affective forecasting errors about extraverted behavior, theymay avoid it and thus experience suboptimal levels of happiness.

Although suggestive, Study 1 has some limitations. For exam-ple, as an online study, participants were speculating about situa-tions and behaviors that were obviously hypothetical. Moreover,because they rated many situations and varying behavior withinthose situations, it is possible that they did not give careful con-sideration to each. Additionally, the many repeated assessmentsmay have magnified subtle differences by implicitly suggestingthat each variation deserved a different answer (see Schwarz,1999). Finally, although affective forecasts likely influencechoices regardless of their accuracy, this study cannot tell uswhether or not the personality differences were errors because itdid not assess experience.

Studies 2–5

Studies 2–51 addressed many of Study 1’s limitations. Acrossthese studies, participants came to the lab and were randomlyassigned to behave in an introverted or extraverted way. Afterreceiving these behavior instructions and a description of forth-coming tasks, participants made affective forecasts about how theyexpected to feel. They then completed the tasks and reported ontheir actual experience. In contrast to Study 1, comparisons be-tween introverted and extraverted behavior were entirely betweensubjects (likely reducing bias), and the within-person differencesbetween forecast and experience were compared to assess accu-racy. Despite these method changes, our predictions remained thesame. We hypothesized that trait introverts would forecast fewerpositive affects and more negative affects for acting extraverted,compared to trait extraverts. Because Studies 2–5 share methodsand hypotheses, we present them together. Only the specific con-texts (tasks) for enacting the assigned behavior differ, and we haveno systematic predictions about these contextual differences (whilerecognizing that context may have some effect).

Method

Participants. Across studies, undergraduate students were re-cruited from the university pool, and participants were givencourse credit or a cash incentive for participation. Trait extraver-sion was assessed prior to the testing sessions to reduce demandeffects. Studies 2–4 collected disposition through mass testing atthe beginning of the academic year. Study 5 had participantscomplete an online survey a couple of days prior to in-personparticipation. Studies 3 and 4 selected participants who scored inthe top or bottom 30% of the distribution of trait extraversion; thisremoved people �0.5 from the mean on a 7-point Likert-typescale.

We recruited 514 participants across all studies. Of these, 20participants were excluded from analysis due to abnormalitiesduring the testing session (e.g., not following instructions). Anal-yses included 495 participants (29.9% male, 70.1% female). Abreakdown of participants included in analyses for each study isshown in Table 1.

Materials.Trait extraversion. As in Study 1, trait extraversion was as-

sessed with the eight-item scale from Goldberg’s Big Five Mini-Markers (Saucier, 1994).

Acting instructions. Across all studies, participants were ran-domly assigned to act in introverted or extraverted ways. Studies2, 3, and 5 also included a control condition with no instructions tomodify behavior. Instructions to modify behavior were very sim-ilar to Fleeson et al.’s (2002) and Study 1. Instructions included anexplicit request to avoid discussing them with other participantsduring the study. Although not reported in detail here, manipula-tion checks based on both self- and observer reports (includingblind confederates and naïve control-condition participants)clearly confirm compliance with acting instructions.

Affective forecasting. Prior to any social interaction, partici-pants completed the forecasting questionnaire. Participants were in-structed to rate how they anticipated feeling during the interactionusing mood adjectives and a 7-point Likert-type scale (7 � Extremelyor a lot, 1 � Very slightly or not at all). Across studies, the numberof adjectives listed varied slightly between 12 and 16. From theseadjectives, we created four scales similar to Study 1: activated positiveaffect (excited, interested, strong), negative affect (worried, fearful,upset, nervous, distressed), pleasantness (happy, pleased), and self-consciousness (embarrassed, self-conscious). Cronbach’s alphasranged from .56 to .80 for forecast positive affect (.77 across studies),from .72 to .83 for forecast pleasantness (.76 across studies), from .80to .86 for forecast negative affect (.83 across studies), and from .64 to.80 for forecast self-consciousness (.73 across studies). Finally, theforecasting questionnaire also included statements about anticipatedenjoyment. We omit these data for brevity, but results are generallysimilar to the adjectives.

Affective experience. The questionnaire to assess actual experi-ence included additional mood items (27–30 adjectives) but wasotherwise similar to the forecasting questionnaire; participants wereasked to rate how they were feeling right now using the 7-pointLikert-type scale. Experienced positive affect, pleasantness, negativeaffect, and self-consciousness were calculated in parallel to theirforecasted counterparts. Cronbach’s alphas ranged from .76 to .84 forexperienced positive affect (.79 across studies), from .70 to .86for experienced pleasantness (.80 across studies), from .71 to .83for experienced negative affect (.80 across studies), and from .68to .78 for experienced self-consciousness (.74 across studies). Wealso created an arousal scale (stimulated and full of energy) tofurther explore the experience of acting extraverted in some sup-plementary analyses (cf. McNiel et al. 2010), and alphas rangedfrom .68 to .80 (.76 across studies). Participants also completed aposttask questionnaire that included two statements about enjoy-ment rated on a 7-point scale of agreement (The interaction was

1 Other data from Studies 2 and 3 were presented in Zelenski et al.(2012) as Studies 1 and 2, respectively, but no affective forecasting datahave been previously reported.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

1097PERSONALITY AND AFFECTIVE FORECASTS

enjoyable, I had moments of fun during the interaction). Thesewere combined into an additional scale with Cronbach’s alphasranging from .81 to .82 (.82 across studies). The enjoyment state-ments overlapped less with the adjective-based acting instructionsand could further test the extent to which participants judged theinteraction to be fun.

Affective accuracy. To determine if affective forecasts wereaccurate, we calculated difference scores (experience minus fore-cast). Thus, positive numbers indicate underprediction whereasnegative numbers indicate overpredictions.2

Procedure. Studies 2–5 shared the same general procedure.Participants arrived at the lab individually and were ushered to aseparate small room. They completed informed consent and weregiven a description of the study’s activity (e.g., an interview, agroup task; see below). They also were instructed to act intro-verted, were instructed to act extraverted, or received no actinginstructions for the activity, depending on their randomly assignedcondition. Next, they completed affective forecasts for the activity(with the knowledge of their acting instruction). Participants werethen brought together to enact behavior consistent with their ran-domly assigned condition. Following about 20 min of the activity,participants rated their current affect, their behavior during theactivity, and the affect and behavior of other participants (some-times a confederate).

In Study 2, groups of three participants (each assigned to adifferent acting condition) engaged in a group discussion on twotopics (rank ordering 10 items in importance to winter survivalafter an airplane crash and planning a day together). Each topictook approximately 10 min. In Study 3, a confederate, who wasblind to the hypotheses of the study, was introduced as anotherparticipant and was instructed to conduct a “get to know you”interview. The confederate’s role was to ask a series of questionsand maintain conversation with the participant for a 20-min period,and the interviews were videotaped with a camera in full view. InStudy 4, each participant and a (blind) confederate completed twotasks together: planning a day (as in Study 2) and using children’sblocks to build items from a list (e.g., something that is usuallyblue). The researcher kept listing items until the total time of bothactivities took 20 min. Study 5 was conducted with groups of threeor four participants, one assigned to each acting condition, or twoto the acting introverted condition with groups of four. Participantswere brought to a large room with a single jigsaw puzzle andinformed that they had 20 min to complete as much as possible.

Results

Data on affective forecasts, experience, and forecasting accu-racy are presented separately for each study in Tables 2, 3, and 4,respectively. We also present two sets of confidence intervals(CIs): 80% as a reasonable criterion for difference as suggested byCohen (1990) and 95% to aid inferences about the statisticalsignificance of differences within a study. The tables providemeans or correlations depending on whether extreme/categoricalpersonality groups were selected (Studies 3 and 4) or we recruitedregardless of disposition (Studies 2 and 5). In addition to correla-tions, Table 4 provides mean difference scores (by condition) forStudies 2 and 5. For differences scores, the mean provides impor-tant information about accuracy (i.e., values near zero suggestaccuracy, etc.) otherwise not available from the correlations.

Our key hypotheses involve personality differences in forecastssuch that trait introverts in the acting extraverted condition willforecast more negative affects and fewer positive affects than traitextraverts. Significance tests of such hypotheses can be assessedindividually for each study by comparing the means and whetherthey fall inside or outside the comparison means’ 95% CIs orwhether the correlation 95% CI includes zero or not.

To provide a comprehensive test of hypotheses, we also com-bined data across studies and conducted multiple regression anal-yses. The three conditions were represented by two dummy-codedvariables with control condition as the comparison (i.e., coded aszero for the two dummy-coded variables). To examine the generaltrends across the four studies while controlling for differencesbetween them, we included three effect-coded study variables(Aguinis, 2004).

Despite two of the four studies selecting participants with moreextreme scores on extraversion, we examined disposition as acontinuous (rather than categorical) variable in the regressions.Raw disposition scores provided additional gradation information(we had actual scores), increasing power relative to a categoricalapproach. Skew and kurtosis values demonstrated that dispositionwas approximately normal though there was, unsurprisingly, asomewhat bimodal pattern with fewer participants in the center of

2 Inadvertently, Study 2 omitted strong from the forecasting question-naire, and Study 5 omitted fearful from the experience questionnaire.Difference scores were calculated by truncating the corresponding experi-ence/forecast scales so that means on the exact same items were subtracted.

Table 1Details for Studies 2, 3, 4, and 5, Including Participant Gender, Acting Condition, Disposition, and Study Tasks

Study

N Acting condition Disposition

TaskTotal Male FemaleAct

extravertedAct

introverted Control M SD

2 113 39 74 38 38 37 4.57 0.93 Groups of 3; discussion (plan aday, winter survival)

3 127 39 88 46 42 39 4.35 1.31 Confederate; “get to know you”interview

4 96 21 75 48 48 4.49 1.36 Confederate; interaction (buildingblocks, plan a day)

5 159 49 110 49 61 49 4.63 1.02 Groups of 3–4; complete singlejigsaw puzzle

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

1098 ZELENSKI ET AL.

the distribution. Normality of predictor variables is not a require-ment for multiple regression; thus, we examined residuals(Tabachnick & Fidell, 2007). Residuals for affective forecasts anddiscrepancy scores were approximately normal. Residuals for ex-perienced negative affect and experienced self-conscious affectwere positively skewed. Square-root transformations addressed theskewed residuals while not changing significance. We report orig-inal nontransformed results for ease of interpretation.

For each dependent variable, a four-step hierarchical regressionanalysis was conducted such that Step 1 of the regression con-tained the three effect-coded study variables, Step 2 added dispo-sition, Step 3 added the two acting condition variables, and finallyStep 4 added the interaction terms (two variables created bymultiplying the centered disposition by dummy-coded conditionvariable).3

Forecasted affect. For forecasted positive affect in the com-bined data, there was a positive relationship with disposition(b � .17, R2 � .026, p � .001), indicating that extravertspredicted more positive affect than introverts. There were alsocondition differences, R2 � .121, F(2, 488) � 34.91, p �.001; participants in the acting introverted condition forecastsignificantly less positive affect compared to those in the actingextraverted and control conditions (which did not differ). Most

importantly, the Condition � Disposition interaction was sig-nificant, R2 � .016, F(2, 486) � 4.76, p � .009. Consistentwith hypotheses, introverts forecast less positive affect thanextraverts in the acting extraverted condition (i.e., the simpleslope was positive, b � .34, p � .001). In both control andacting introverted conditions, the relationship between disposi-tion and forecasted positive affect was not significant (i.e., thesimple slopes were flat, b � .02, p � .50; b � .11, p � .11,

3 We performed another series of regressions that dummy-coded (ratherthan effect-coded) individual studies to also assess the two- and three-wayinteractions of study by disposition and acting condition. These interactionsassessed the consistency of effects across the studies. Of the 27 R2 forinteraction steps examined, three were significant. In no case was thethree-way (Study � Acting Condition � Disposition) interaction statisti-cally significant. In other words, the omnibus tests provided no evidence ofa personality-related forecasting error that varies by acting instructiondepending on the study context. Three of the two-way interactions weresignificant, suggesting that acting instructions and disposition can occa-sionally operate differently in different experimental contexts. These wereCondition � Study for positive affect forecasts, Trait � Study for pleas-antness forecasts, and Condition � Study for actual enjoyment statements.The nature of these interactions is described by comparing individual studyeffects in Table 2 (forecasts) and main text (for enjoyment, where Study 5was atypical).

Table 2Means and Confidence Intervals of Forecasted Affect

StudyActing

condition Disposition

Positive affect Pleasantness Negative affect Self-consciousness

80% CI 95% CI 80% CI 95% CI 80% CI 95% CI 80% CI 95% CI

2 E .29 .08 �.47 �.40.11, .48 �.01, .53 �.06, .22 �.14, .30 �.63, �.29 �.70, �.17 �.58, �.22 �.66, �.09

I �.23 �.18 �.08 �.26�.41, �.05 �.48, .05 �.40, .06 �.48, .17 �.29, .13 �.41, .26 �.45, �.06 �.55, .06

N �.26 �.09 �.22 �.32�.43, �.08 �.52, .01 �.29, .15 �.39, .24 �.38, �.07 �.45, .03 �.51, �.13 �.58, �.01

3 E E 5.06 4.63 2.08 3.194.73, 5.40 4.53, 5.60 4.28, 4.97 4.07, 5.18 1.64, 2.51 1.38, 2.77 2.70, 3.68 2.41, 3.97

I 4.20 3.95 3.26 5.053.93, 4.47 3.78, 4.62 3.68, 4.22 3.52, 4.38 2.97, 3.55 2.80, 3.72 4.74, 5.36 4.57, 5.53

I E 3.37 3.55 2.58 3.743.04, 3.69 2.85, 3.88 3.16, 3.94 2.94, 4.16 2.29, 2.87 2.12, 3.04 3.32, 4.16 3.08, 4.40

I 3.25 3.24 3.14 4.173.03, 3.48 2.90, 3.61 2.95, 3.52 2.79, 3.69 2.82, 3.46 2.64, 3.64 3.79, 4.54 3.58, 4.75

N E 3.97 4.46 2.41 3.763.67, 4.27 3.50, 4.44 4.15, 4.76 3.98, 4.93 2.13, 2.69 1.97, 2.85 3.42, 4.10 3.23, 4.29

I 3.94 3.94 2.28 3.383.65, 4.23 3.48, 4.40 3.68, 4.20 3.53, 4.35 1.84, 2.71 1.59, 2.96 2.87, 3.88 2.57, 4.18

4 E E 4.63 4.42 2.08 3.064.37, 4.88 4.22, 5.03 4.17, 4.67 4.04, 4.80 1.81, 2.35 1.66, 2.50 2.72, 3.40 2.53, 3.59

I 3.80 3.50 3.24 4.963.48, 4.11 3.30, 4.29 3.27, 3.73 3.13, 3.87 2.93, 3.56 2.75, 3.74 4.60, 5.31 4.40, 5.51

I E 3.89 4.26 2.26 3.223.63, 4.16 3.48, 4.30 3.97, 4.56 3.80, 4.72 1.99, 2.54 1.84, 2.69 2.82, 3.62 2.59, 3.85

I 3.64 3.52 2.87 4.203.33, 3.94 3.16, 4.12 3.15, 3.89 2.94, 4.10 2.52, 3.22 2.32, 3.42 3.73, 4.66 3.47, 4.92

5 E .37 .12 �.44 �.41.17, .55 .06, .62 �.06, .30 �.15, .39 �.59, �.31 �.64, �.24 �.57, �.27 �.62, �.17

I .20 .22 �.16 �.13.04, .35 �.07, .43 .04, .39 �.05, .46 �.33, �.00 �.42, .08 �.31, .06 �.42, .15

N .14 .03 �.38 �.56�.04, .31 �.17, .39 �.15, .23 �.22, .31 �.55, �.17 �.63, �.05 �.68, �.43 �.74, �.35

Note. Studies 2 and 5 are Pearson correlations with bootstrapping for CIs, using 1,000 samples; Studies 3 and 4 are means with CIs. CI � confidenceinterval; E � extraversion; I � introversion; N � no instruction (control).

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

1099PERSONALITY AND AFFECTIVE FORECASTS

respectively), indicating that both introverts and extraverts fore-cast similar levels of positive affect. Thus, in support of ourhypotheses, extraverts, compared to introverts, expected to ex-perience more positive affect but only when instructed to actextraverted.

For forecasted pleasantness, there was a general positive trendof disposition (b � .20, R2 � .041, p � .001), indicating thatextraverts predicted more pleasantness than introverts. There werealso condition differences, R2 � .071, F(2, 488) � 19.89, p �.001; participants in the acting introverted condition forecast sig-nificantly less pleasantness compared to those in the acting extra-verted and control conditions (which did not differ). Counter to ourhypotheses, the Condition � Disposition interaction was not sig-nificant for pleasantness, R2 � .003, F(2, 486) � 0.89, p � .41.Thus, the acting condition and disposition effects did not appear todepend on one another with regard to pleasantness.

For forecasted negative affect, there was a negative relationshipwith disposition, indicating that introverts expected more negativeaffect than extraverts (b � �.35, R2 � .126, p � .001). The maineffect of condition was also significant, R2 � .02, F(2, 488) �6.23, p � .002, with the control condition having the least forecastednegative affect and acting extraverted the most, although the actingintroverted and acting extraverted conditions did not differ signifi-

cantly. Finally, the interaction of disposition and condition was alsosignificant, R2 � .018, F(2, 486) � 5.22, p � .006. All threeconditions maintained the trend of introverts expecting more negativeaffect than extraverts, but the relationship was strongest in the actingextraverted condition (i.e., all simple slopes were significant, controlb � �.18, p � .04; acting introverted b � �.28, p � .001; and actingextraverted b � �.50, p � .001). That is, introverts forecast morenegative affect in general, but they forecast considerably more nega-tive affect when instructed to act extraverted, again consistent withhypotheses.

For forecasted self-consciousness, there was a negative relationshipwith disposition, indicating that introverts expected to feel moreself-conscious than extraverts (b � �.51, R2 � .137, p � .001).Condition was significant, R2 � .039, F(2, 488) � 12.17, p �.001, with the acting extraverted condition having the most expectedself-conscious affect and control having the least. Acting introvertedand control conditions did not differ significantly. The interaction ofdisposition and condition was also significant, R2 � .013, F(2,486) � 4.07, p � .02. All three conditions maintained the negativerelationship of disposition and forecasted self-consciousness, but therelationship was again strongest in the acting extraverted condition(i.e., all simple slopes were negative; for control, b � �.37, p � .002;for acting introverted, b � �.37, p � .001; and for acting extraverted,

Table 3Means and Confidence Intervals of Experienced Affect

StudyActing

condition Disposition

Positive affect Pleasantness Negative affect Self-consciousness

80% CI 95% CI 80% CI 95% CI 80% CI 95% CI 80% CI 95% CI

2 E .30 .04 �.05 �.11.07, .52 �.05, .61 �.17, .26 �.25, .35 �.23, .12 �.33, .20 �.29, .08 �.39, .18

I �.17 �.23 �.01 �.21�.39, .03 �.49, .15 �.46, �.04 �.56, .11 �.21, .22 �.30, .31 �.40, .03 �.52, .12

N .08 �.01 �.30 �.35�.19, .31 �.33, .42 �.29, .25 �.39, .34 �.47, �.13 �.53, �.03 �.48, �.23 �.55, �.16

3 E E 5.02 5.41 1.90 2.884.70, 5.34 4.51, 5.53 5.05, 5.76 4.84, 5.98 1.56, 2.24 1.37, 2.43 2.32, 3.43 2.00, 3.75

I 4.56 4.48 2.55 4.054.27, 4.85 4.10, 5.01 4.23, 4.74 4.09, 4.88 2.28, 2.83 2.12, 2.99 3.62, 4.48 3.38, 4.72

I E 3.19 3.88 2.11 3.402.74, 3.64 2.48, 3.90 3.43, 4.33 3.17, 4.59 1.77, 2.46 1.58, 2.65 2.90, 3.91 2.60, 4.20

I 3.11 3.24 2.79 3.882.85, 3.37 2.70, 3.52 2.94, 3.53 2.77, 3.70 2.41, 3.16 2.19, 3.38 3.37, 4.39 3.08, 4.68

N E 4.41 4.80 2.05 3.614.10, 4.72 3.92, 4.89 4.49, 5.12 4.31, 5.30 1.84, 2.27 1.72, 2.39 3.18, 4.04 2.93, 4.29

I 4.00 4.31 1.76 2.693.49, 4.51 3.19, 4.81 3.84, 4.79 3.55, 5.07 1.47, 2.06 1.30, 2.23 2.21, 3.16 1.93, 3.45

4 E E 5.07 5.12 1.53 2.064.77, 5.37 4.60, 5.54 4.85, 5.39 4.70, 5.54 1.37, 1.68 1.28, 1.77 1.69, 2.43 1.48, 2.64

I 4.41 4.46 2.23 3.044.17, 4.66 4.03, 4.80 4.13, 4.78 3.95, 4.97 1.94, 2.51 1.77, 2.68 2.63, 3.46 2.39, 3.70

I E 3.28 3.69 1.92 3.212.93, 3.63 2.73, 3.82 3.35, 4.03 3.16, 4.22 1.61, 2.23 1.43, 2.40 2.65, 3.77 2.33, 4.09

I 3.49 3.83 2.32 4.303.12, 3.86 2.91, 4.08 3.42, 4.23 3.20, 4.46 2.00, 2.64 1.82, 2.82 3.80, 4.81 3.51, 5.10

5 E .21 .24 �.16 �.27.03, .37 �.10, .45 .10, .39 .00, .47 �.36, .05 �.44, .17 �.45, �.06 �.54, .06

I .04 .14 �.13 �.19�.13, .21 �.21, .28 �.02, .31 �.11, 40 �.30, .05 �.39, .15 �.36, �.02 �.43, .10

N .15 .01 �.29 �.53�.01, .30 �.11, .39 �.21, .21 �.31, .35 �.48, �.03 �.60, .11 �.67, �.37 �.75, �.26

Note. Studies 2 and 5 are Pearson correlations with bootstrapping for CIs, using 1,000 samples; Studies 3 and 4 are means and associated CIs. CI �confidence interval; E � extraversion; I � introversion; N � no instruction (control).

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

1100 ZELENSKI ET AL.

b � �.69, p � .001). That is, introverts expected to feel moreself-conscious in all conditions but they expected to feel most self-conscious when instructed to act extraverted.

Overall, analyses on the combined data produced results that sup-ported hypotheses for positive affect, negative affect, and self-consciousness, but not pleasantness. Examining individual studiesreveals similar results, though with some minor variations acrossstudies (see also footnote 3). Although there are many possible com-parisons in Table 2, our key prediction is that trait introverts wouldpredict fewer positive and more negative affects in the acting extra-verted conditions, compared to trait extraverts. In Studies 3 and 4,with categorical personality groups, we found fully supportive results.Introverts forecast significantly more negative affect and self-consciousness and significantly less positive affect and pleasantnessin the acting extraverted condition. The magnitude of these effectswas medium to large, ranging from d � .62 (r � .29) for pleasantnessin Study 3 to d � �1.47 (r � �.59) for self-consciousness in Study4. The results for Studies 2 and 5 were consistently significant fornegative affects; trait extraversion was significantly negatively asso-ciated with negative affect and self-consciousness in the acting extra-verted condition (rs from �.40 to �.47), as predicted. The positiveaffects showed a similar trend (rs from .08 to .37) but were not alwaysstatistically significant and were especially weak for pleasantnesscompared to positive affect.

In sum, introverts expected less positive affect and more negativeand self-conscious affect than extraverts when instructed to act extra-verted. This may help explain why introverts do not choose to actextraverted more often. Results for pleasantness were less consistentand not statistically significant in the combined data set, however.

Experienced affect. Previous studies have found that partici-pants report experiencing more positive and pleasant affect whenacting extraverted, compared to acting introverted, and this has notdepended on (interacted with) disposition. Differences in negativeaffect have typically not been observed, but no previous studies haveexamined self-consciousness specifically. Our combined data set pro-vides a unique opportunity to reveal any possible interactions withsignificantly more power than previous attempts. Parallel to the fore-casting analyses, hierarchical regression models were run to exploreexperienced positive affect, pleasantness, negative affect, and self-

4 Some studies were run with small groups of participants. Althoughgroup members’ affective forecasts were independent (because no groupinteraction had occurred yet), their experienced affect was not. Groupedparticipants’ affect may have been linked due to the shared experience, thussuggesting an MLM approach to analyses. We investigated this possibilitybut retained the regression approach because the amount of variability atthe group level was not statistically significant.

Table 4Mean Difference Scores in Affect (Experience � Forecast)

StudyActing

condition Disposition

Positive affect Pleasantness Negative affect Self-consciousness

80% CI 95% CI 80% CI 95% CI 80% CI 95% CI 80% CI 95% CI

2 E .04 (.29) �.04 (.67) .41 (�.77) .25 (�1.36)�.15, .24 �.24, .31 �.24, .15 �.36, .22 .26, .56 .17, .63 .03, .42 �.06, .52

I �.00 (�.24) �.13 (�.08) .08 (�.49) .07 (�.54)�.26, .23 �.36, .36 �.35, .08 �.44, .21 �.15, .30 �.27, .42 �.20, .32 �.35, .41

N .30 (.03) .07 (.27) �.03 (�.64) �.04 (�.68).10, .47 �.01, .54 �.21, .30 �.32,.41 �.21, .14 �.32, .23 �.22, .13 �.34, .23

3 E E �.04 .78 �.18 �.31�.27, .19 �.40, .32 .49, 1.07 .32, 1.25 �.56, .21 �.79, .44 �.66, .03 �.86, .24

I .36 .53 �.71 �1.00.14, .58 .01, .70 .28, .79 .13, .93 �.95, �.46 �1.09, �.32 �1.27, �.73 �1.42, �.58

I E �.17 .33 �.47 �.33�.57, .22 �.80, .45 �.01, .67 �.20, .87 �.71, �.23 �.85, �.09 �.74, .07 �.97, .30

I �.14 .00 �.36 �.29�.44, .16 �.62, .33 �.22, .22 �.35, .35 �.60, �.11 �.74, .02 �.61, .04 �.79, .22

N E .43 .35 �.36 �.15.25, .62 .14, .73 .10, .59 �.03, .73 �.56, �.16 �.67, �.04 �.46, .15 �.63, .32

I .06 .38 �.51 �.69�.35, .48 �.60, .72 .01, .74 �.20, .95 �.89, �.13 �1.12, .10 �1.29, �.08 �1.65, .28

4 E E .44 .70 �.55 �1.00.15, .73 �.01, .89 .49, .91 .37, 1.03 �.74, �.37 �.84, �.26 �1.31, �.69 �1.48, �.52

I .62 .96 �1.02 �1.91.31, .92 .14, 1.09 .61, 1.30 .42, 1.49 �1.31, �.73 �1.47, �.56 �2.38, �1.44 �2.65, �1.18

I E �.67 �.63 �.35 .00�1.13, �.20 �1.39, .06 �1.00, �.25 �1.22, �.03 �.65, �.05 �.83, .13 �.49, .49 �.76, .76

I �.14 .30 �.55 .11�.43, .14 �.59, .30 �.04, .65 �.23, .84 �.85, �.25 �1.02, �.08 �.34, .56 �.60, .82

5 E �.10 (�.33) .15 (�.13) .30 (�.42) .17 (�1.12)�.27, .07 �.35, .17 �.04, .32 �.10, .39 .14, .46 .04, .54 �.01, .35 �.10, .45

I �.14 (�.14) �.05 (�.20) .01 (�.01) �.05 (�.28)�.29, .01 �.37, .10 �.20, .13 �.29, .22 �.16, .18 �.27, .27 �.21, .11 �.29, .19

N .04 (�.27) �.01 (�.36) .05 (�.21) .19 (�.87)�.12, .21 �.20, .29 �.24, .21 �.35, .32 �.16, .26 �.27, .35 �.00, .41 �.11, .50

Note. Studies 2 and 5 are Pearson correlations with bootstrapping for CIs, using 1,000 samples, and mean difference score in parentheses; Studies 3 and4 are means and associated CIs. CI � confidence interval; E � extraversion; I � introversion; N � no instruction (control).

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

1101PERSONALITY AND AFFECTIVE FORECASTS

consciousness, and we added the enjoyment scale as an additionalindicator in experience.4

Somewhat different from previous research, there was a maineffect of trait extraversion for all four primary dependent variables(b � .13, p � .01; b � .18, p � .001; b � �.22, p � .001; b ��.39, p � .001, for experienced positive affect, pleasantness,negative affect, and self-consciousness, respectively), indicatingthat, in general, introverts reported experiencing fewer positiveaffects and more negative affects. Similar to previous acting re-search, there was a significant and substantial main effect ofcondition—R2 � .175, F(2, 488) � 53.26, p � .001, forpositive affect; R2 � .13, F(2, 488) � 37.91, p � .001,for pleasantness; R2 � .027, F(2, 486) � 7.38, p � .001, fornegative affect; and R2 � .025, F(2, 488) � 7.31, p � .001, forself-consciousness—such that participants reported more positiveaffects when acting extraverted and more negative affects whenacting introverted. These results were mirrored in the enjoymentitems with a small but significant effect of disposition whereextraverts reported enjoying themselves more (b � .09, p � .05)and a clear effect of acting condition with acting extraverted beingmost enjoyable, R2 � .053, F(2, 488) � 13.86, p � .001.

Our key focus was determining whether disposition and condi-tion interacted. For all four primary dependent variables, positiveaffect, pleasantness, negative affect, and self-consciousness, theinteraction of disposition by condition was nonsignificant, R2 �.006, F(2, 486) � 1.92, p � .15; R2 � .001, F(2, 486) � 0.23,p � .80; R2 � .003, F(2, 486) � 0.75, p � .47; and R2 � .004F(2, 486) � 1.07, p � .35, respectively. This was also true forthe enjoyment statements (R2 � .002, p � .64). Therefore,despite our large sample size and increased power, the effects ofdisposition and acting condition did not appear to depend on oneanother.5

Examining studies individually reveals similar results (see Table3 for means and Table 5 for inferential tests and effect sizes).Acting extraverted clearly and substantially increased positiveaffect and pleasantness. There were occasional and inconsistentdifferences across negative affect and self-consciousness. Moreimportantly, of the 16 interactions tested, only one was statisticallysignificant (Study 3 self-consciousness); the effects of acting ex-traverted did not appear to depend much on disposition. We foundsimilar results with the enjoyment scale. Across all studies, actingextraverted produced the most enjoyment, and no significant in-teractions with disposition emerged. The condition effect wassubstantial for Study 2 (R2 � .11, p � .001), Study 3 (2 � .15,p � .001), and Study 4 (2 � .12, p � .001), but virtuallynonexistent in Study 5 (R2 � .01, p � .98). This isolated nullresult is surprising but is corroborated somewhat by a smallerpleasantness effect in Study 5, compared to other studies (seeTable 5).

In sum, we found that participants experienced substantiallymore positive affects and fewer negative affects when actingextraverted and did not find evidence for the idea that dispositionmoderated these effects. Considered in the context of significantinteractions in forecasted affect, these results suggest that someforecasts are indeed errors. We examined this possibility moredirectly with difference scores.

Accuracy of forecasts. We calculated difference scores (ex-perience minus forecast) for each affect scale. Positive differencescores indicate experiencing more affect than was forecast (an

underestimate), whereas negative scores indicate experiencing lessthan forecast (an overestimate). We hypothesized that introvertswould overestimate negative affects and underestimate positiveaffects when asked to act extraverted. Similar to forecast andexperience variables, regressions using the combined data pre-dicted difference scores with disposition, condition, and theirinteractions.

For positive affect, the constant was approximately zero, indi-cating that overall, participants were accurate in predicting positiveaffect. Disposition was nonsignificant; in general, introverts andextraverts did not differ significantly in predictive accuracy (b ��.04, R2 � .001, p � .48). There was a significant effect ofcondition, R2 � .018, F(2, 488) � 4.62, p � .01; actingintroverted differed from acting extraverted such that the positiveaffect of acting extraverted was underestimated. Interestingly, theinteraction of disposition by condition was also significant, R2 �.012, F(2, 486) � 3.02, p � .05. For the control condition, therewas a marginal positive slope (b � .19, p � .06), with introvertsoverestimating and extraverts underestimating. For acting intro-verted and acting extraverted conditions, disposition was not sig-nificantly associated with accuracy (b � �.08, p � .11; b � �.10,p � .18, for acting introverted and acting extraverted conditions,respectively). Nonetheless, the overall pattern of the interactionacross studies suggests that trait introverts are indeed underesti-mating how much positive affect acting extraverted will produce(see Figure 5).

For pleasantness, the constant was positive (b � .20), indicatingthat overall, participants expected the studies to be less pleasantthan they were. Disposition was not statistically significant; ingeneral, introverts and extraverts did not differ significantly inpredictive accuracy (b � �.02, R2 � .000, p � .64). There wasa significant effect of condition, R2 � .037, F(2, 488) � 9.77,p � .001; acting extraverted differed from both acting introvertedand control such that the pleasantness of acting extraverted wasunderestimated. The interaction of disposition by condition was,however, not statistically significant, R2 � .002, F(2, 486) �0.50, p � .61; both introverts and extraverts were similar in thedegree of accuracy across conditions, contrary to hypotheses (seeFigure 6) but consistent with the null pleasantness results found inforecasts.

For negative affect, the constant was negative (b � �.50),indicating that participants expected the studies to be more nega-tive than they were. Disposition was related to accuracy (b � .11,R2 � .019, p � .002). The negative constant and positive slopeindicate that extraverts were more accurate (less overestimation ofnegative affect) than introverts. Condition was marginally signif-icant, R2 � .011, F(2, 488) � 2.82, p � .06, with only actingintroverted differing from acting extraverted. Participants in theacting extraverted condition overestimated negative affect morethan those in the acting introverted condition. The interaction ofdisposition and condition was also significant, R2 � .015, F(2,486) � 4.02, p � .02. For the acting introverted and controlconditions, disposition was not significantly associated with accu-

5 Although we had more participants in these analyses, our brief affectmeasures (chosen to mirror the forecast scales) may be less reliable thanthe longer measures used in previous research. Examining longer positiveaffect and negative affect scales in these data revealed similarly nonsig-nificant interactions.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

1102 ZELENSKI ET AL.

racy (b � .04, p � .50; b � .03, p � .50, respectively). For actingextraverted, introverts were less accurate than extraverts by overpre-dicting negative affect when instructed to act extraverted (b � .26,p � .001), a pattern consistent with our hypotheses (see Figure 7).

For self-consciousness, the constant was negative (b � �.69),indicating that participants expected to feel more self-consciousthan they actually did. Disposition was positively related to accu-racy, with extraverts being more accurate than introverts (b � .12,R2 � .008, p � .05). Condition was significant, R2 � .068,F(2, 488) � 18.11, p � .001, with all three conditions differingfrom each other. Acting extraverted was least accurate, and actingintroverted was most accurate. The interaction of disposition andcondition was significant, R2 � .018, F(2, 486) � 5.00, p �.007. Acting introverted and control conditions had similar accu-racies across disposition (b � �.11, p � .28; b � .12, p � .30,respectively). For acting extraverted, trait extraverts were moreaccurate than introverts (b � .29, p � .001); consistent withhypotheses, introverts were particularly likely to overestimate self-consciousness when instructed to act extraverted (see Figure 8).

Results for individual studies are reported in Table 4, where the95% CIs can be interpreted as tests of statistically significantdifferences. Similar to forecasts, the key accuracy comparisoninvolves trait introverts’ accuracy relative to trait extraverts’ in theacting extraverted condition. Although the pattern of findingsgenerally fits with predictions, the significance of these differencesis not ubiquitous across individual studies. Hypothesized forecast-ing and accuracy effects did not materialize for pleasantness, andthe difference scores are thus unsurprisingly also not significantfor each study individually. Somewhat more surprisingly, theresults for positive affect were also not particularly supportive,with the positive affect difference score in Study 3 the only one toreach statistical significance. In contrast, the hypothesized discrep-ancies in negative affect were significant in all studies, suggestingthat trait introverts tended to experience substantially less negativeaffect than they predicted. A similar pattern was found for self-consciousness but was only statistically significant in Studies 3and 4.

In sum, although the accuracy results in individual studies weresomewhat inconsistent in meeting the p � .05 criterion, the pat-

Table 5Inferential Statistics and Effect Sizes for Experienced Affect for Studies 2–5 Separately

Study Effect

Positive affect Pleasantness Negative affect Self-consciousness

F 2/R2 F 2/R2 F 2/R2 F 2/R2

2 Acting condition 10.42��� .16 12.47��� .19 4.56� .08 1.28 .02Disposition 0.43 .01 0.71 .01 0.79 .01 5.08� .04Interaction 1.94 .03 0.98 .02 0.36 .01 0.49 .01

3 Acting condition 18.89��� .24 14.50��� .19 2.54 .04 0.87 .01Disposition 2.00 .02 9.69�� .07 3.11 .03 0.63 .01Interaction 0.29 .01 0.33 .01 2.54 .04 3.95� .06

4 Acting condition 32.80��� .26 17.08��� .16 1.38 .02 11.70��� .11Disposition 0.51 .01 1.07 .01 7.36�� .07 8.85�� .09Interaction 3.85 .04 2.61 .03 0.50 .01 0.03 .00

5 Acting condition 8.85��� .10 4.95�� .06 1.16 .02 3.42� .04Disposition 2.33 .01 2.43 .02 5.90� .04 15.95��� .09Interaction 0.34 .01 0.63 .01 0.20 .01 0.48 .01

Note. F is from the analysis of variance for Studies 3 and 4 and, for R2, from regression in Studies 2 and 5.� p � .05. �� p � .01. ��� p � .001.

-2

-1.5

-1

-0.5

0

0.5

1

Introverted Extraverted

Pos

itive

Affe

ct A

ccur

acy

Control Acting Extraverted Acting Introverted

Figure 5. Trait extraversion and acting condition interaction on positiveaffect accuracy (Studies 2–5).

-2

-1.5

-1

-0.5

0

0.5

1

Introverted Extraverted

Ple

asan

tnes

s A

ccur

accy

Control Acting Extraverted Acting Introverted

Figure 6. Trait extraversion and acting condition interaction on pleasant-ness accuracy (Studies 2–5; interaction not statistically significant).

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

1103PERSONALITY AND AFFECTIVE FORECASTS

terns were usually consistent with hypotheses. In the combineddata, we found significant differences in accuracy across actingconditions and disposition for positive affect, negative affect, andself-consciousness (but not pleasantness). Overall, the patternssuggest that introverts are likely to erroneously underestimate thehedonic benefits of acting extraverted, particularly by overestimat-ing the potential for negative and self-conscious affect.

General Discussion

Research suggests that people enjoy acting in extraverted waysand that this phenomenon does not depend on trait extraversionlevels (e.g., Fleeson et al, 2002). This raises a question: why dointroverts not act extraverted more often? In this research, weexplored one possible explanation; perhaps trait introverts’ affec-tive forecasts underlie unduly negative expectations about extra-verted behavior. Across five studies, we found support for thisidea. That is, trait introverts forecast significantly more negativeaffect and self-consciousness for extraverted behavior (and some-times less positive affect and pleasantness), compared to traitextraverts. This general finding held in both within- and between-subjects comparisons and across multiple contexts. To the extentthat affective forecasts influence behavior, these results help ex-plain why introverts might avoid extraverted situations and behav-ior and ultimately why they tend to report lower levels of happi-ness.

We were also able to explore the emotional consequences ofacting introverted and extraverted more comprehensively thanprevious research. That is, by combining many samples, we hadvastly more power to detect potentially subtle personality differ-ences in the experience of enacting introverted or extravertedbehaviors. Despite this increased power and breadth in consideringpositive affect, pleasantness, enjoyment, negative affect, and self-consciousness, we found no consistent personality interactions.This provides even more confidence that people enjoy actingextraverted regardless of their disposition. It is worth noting thatall the behavior and experience occurred in the lab setting andwere randomly assigned. On one hand, assigning a person tobehave counterdispositionally might increase the chances of an

unpleasant experience; however, having the excuse of this instruc-tion might also provide an emotional buffer. Future research mightsearch for personality differences outside the lab (though considerFleeson et al.’s, 2002, ESM results) or examine counterdisposi-tional behavior over longer periods of time (cf. Little, 2008). It isalso possible that systematic study of different contexts couldreveal additional boundary conditions. We observed a hint of thisin Study 5 (working on a puzzle), where acting extraverted wasassociated with a more modest increase in pleasantness and nosignificant difference on the enjoyment items. Despite these cave-ats, our data strongly support the conclusions of previous work;acting extraverted (usually) seems like fun for everyone.

Given that we found personality differences (interactions) in affec-tive forecasts but not in actual experience, it is unsurprising thatanalyses of difference scores revealed some significant affective fore-casting errors. Overall, our studies were more enjoyable than partic-ipants anticipated (i.e., more pleasantness and substantially less neg-ative affect and self-consciousness were forecast than experienced),but personality moderated the magnitude of these errors. As hypoth-esized, introverts underpredicted positive affect and overpredictednegative affect and self-consciousness when forecasting responses toextraverted behavior. (Forecasting errors in pleasantness were notsignificant, however.) Trait extraverts were somewhat more likely tomake accurate predictions (difference scores closer to zero). Whereastrait introverts made poor predictions for counterdispositional behav-ior, trait extraverts seemed comparatively more accurate in predictingthe affective consequences of introverted behavior. It is not clearwhether this might reflect an overall greater degree of emotionalintelligence (cf. Hoerger et al., 2012) or if the relatively mundane (i.e.,potentially less variable) emotional consequences of introverted be-havior are simply easier to predict.

This research is among the first to closely examine personalitydifferences in affective forecasting. Our results are generally con-sistent with previous research suggesting that extraverts judge thefuture more positively than introverts (e.g., Lucas & Diener, 2001;Zelenski & Larsen, 2002) and that extraverts both forecast andexperienced more positive affect in the one study that examinedboth (Hoerger & Quirk, 2010). Going beyond these suggestions,however, we also find that personality differences in forecasts (and

-2

-1.5

-1

-0.5

0

0.5

1

Introverted Extraverted

Neg

ativ

e A

ffect

Acc

urac

y

Control Acting Extraverted Acting Introverted

Figure 7. Trait extraversion and acting condition interaction on negativeaffect accuracy (Studies 2–5).

-2

-1.5

-1

-0.5

0

0.5

1

Introverted Extraverted

Sel

f-Con

scio

us A

ccur

acy

Control Acting Extraverted Acting Introverted

Figure 8. Trait extraversion and acting condition interaction on self-consciousness accuracy (Studies 2–5).

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

1104 ZELENSKI ET AL.

their accuracy) depend substantially on the context of (manipu-lated) behavior. In addition, our findings diverge from the notionof personality neglect, the idea that personality has an influence onexperienced emotions that is not fully appreciated when makingforecasts (Quoidbach & Dunn, 2010). In fact, our participants’actual experience did not differ dramatically by personality, but wefound clear personality differences in how they forecast theirexperience. The methodological context of our studies likely con-tributed to the differing results. That is, we explicitly told partic-ipants that they would act introverted or extraverted immediatelybefore they made forecasts, likely making both disposition andbehavior salient. (On the other hand, we used adjectives that werenot unambiguously personality labels, e.g., talkative, and person-ality was not assessed during experimental sessions.) Additionally,the situational manipulation (behaving as instructed) has a strongmain effect on experience that often overwhelms personality dif-ferences in experience (and behavior). Although our methods mayhave contributed to finding personality differences in forecasts, itis also possible that personality neglect (failing to see personalitydifferences in forecasts) is not ubiquitous. Many trait affect-congruent judgment biases have been found (Rusting, 1998; Zel-enski, 2008). Moreover, the personality neglect effects may de-pend on instructions that focus attention on a discrete event, ratherthan disposition, at the time of forecasts. Additional research isneeded to determine the prevalence and moderators of these di-vergent patterns and generalizability to daily life.

Affective forecasting errors become more than a quirk of humanjudgment when they predict important decisions and behavior inthe real world. However, the links among forecasts, decisions, andbehavior are often assumed rather than explicitly tested in thisliterature. The assumed links seem reasonable and when tested areempirically supported (e.g., Mellers et al., 1999). Nonetheless,personality differences provide a somewhat novel context foraffective forecasting research, and it is thus worth reconsideringwhether or how individual differences in forecasts predict subse-quent differences in behavior (cf. Mathieu & Gosling, 2012). Forexample, we would like to interpret our results as an (at leastpartial) explanation for differences in extraverted behavior; that is,because introverts overestimate its hedonic costs, they may avoidthe extraverted behavior that would make them happier. On theother hand, explanations about the causes of personality differ-ences in extraverted behavior have rarely relied on affective fore-casts (or much conscious deliberation at all); causal theories typ-ically appeal to biological differences in arousal (Eysenck, 1967)or reward sensitivity (Gray, 1981). This may be a theoreticaloversight; differences in the cognitive networks that develop overa lifetime of experience with these temperamental sensitivitiesmay become important for behavior (Rusting, 1998, 1999). None-theless, the causal role of forecasts in this context requires empir-ical confirmation.

Additionally, even if we remain cautious about the influenceof affective forecasts in day-to-day trait-relevant behavior, af-fective forecasting errors could still be important in understand-ing behavior change or lack thereof. Noting the dramatic effectson positive affect, Fleeson and colleagues (Fleeson et al., 2002;McNiel et al., 2010) suggested that an acting extraverted inter-vention could make people happier and possibly even be used asa treatment for depression and other psychological maladies.Such an approach might be similar to behavioral activation

interventions that appear beneficial in both clinical and non-clinical populations (see Mazzucchelli, Kane, & Rees, 2010).The suggestion remains appealing given that subsequent re-search that investigated potential costs to trait introverts, suchas stress, effort, and self-control deficits, found none. However,implementing an intervention to act extraverted might be trick-ier than simply giving the instruction. Here, we have shown thatthe people who might benefit the most from an intervention arelikely to anticipate significant unpleasantness (i.e., negativeaffect and embarrassment) for this counterdispositional behav-ior. In other words, their affective forecasting errors may createa barrier to behavior change by decreasing willingness to en-gage in counterdispositional behavior outside the lab (thoughcompliance in the lab is excellent). Moreover, affective fore-casting errors tend to resist change, especially for positiveevents (Wilson, Meyers, & Gilbert, 2001). We conducted anunpublished study where the same positive mood induction thathas made people feel more social (see Whelan & Zelenski,2012) failed to have any effect on affective forecasts for actingextraverted (Besner, 2009). Cajoling people into extravertedbehavior to observe the hedonic benefits firsthand is also un-likely to be sufficient; introverts already behave in extravertedways frequently and experience high levels of positive affectwhen doing so (Fleeson et al., 2002). If experience was enough,we should not see the forecasting differences we do. Thus,although acting extraverted interventions may be able to in-crease happiness (cf. Fordyce, 1983; Mazzucchelli et al., 2010),such efforts might benefit by accounting for, and possiblyattempting to explicitly refute, trait introverts’ forecasting er-rors. That is, similar to cognitive-behavioral approaches likemetacognitive therapy, unhelpful and inaccurate forecastsmight be adjusted, perhaps by examining how individuals arriveat their forecasts (see Wells, 2007). In addition, developingmethods that can experimentally alter affective forecasts in thisdomain would also be a useful research tool in establishing theircausal role in extraverted behavior. That is, by manipulatingforecasts, we could become more confident that they play acausal role in extraverted behavior.

Despite our speculation about acting extraverted interven-tions, we feel compelled to explicitly state that we do not viewintroverts as inferior or in need of psychological change. (Manyof us have pronounced introverted tendencies of our own!)Some introverts might desire the potential happiness boost thatan acting extraverted exercise might provide. We believe thatstudying the potential of such a tool and providing accurateinformation about its effects may ultimately produce hedonicbenefits for some, but we have no interest in forcing it onanyone. It is also possible that most introverts would have littleinterest in such a tool. Although the finding that introverts areless happy than extraverts is very clear (see Steel et al., 2008;Zelenski et al., in press), there are also hints that introvertsdesire happiness less than extraverts.

First, as the title of Diener and Diener’s (1996) classic papersuccinctly stated, “most people are happy.” Introverts may not beparticularly motivated to pursue pleasure beyond an already pleas-ant state (cf. theories of reward sensitivity). This is particularlytrue when operationalizations of happiness rely on high arousalpleasant states such as those assessed by the widely used PANAS.For example, Kuppens (2008) found that reward responsiveness (a

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

1105PERSONALITY AND AFFECTIVE FORECASTS

correlate and potential cause of extraversion) moderated the rela-tionship between momentary experiences of pleasantness andarousal; those high on reward responsiveness felt pleasure withhigh arousal whereas those low felt pleasure with low arousal.6

Similarly, Rusting and Larsen (1995) found that extraverts re-ported desiring high arousal states more than introverts. It is worthnoting that the experience of acting extraverted includes increasesin both pleasure and arousal (McNiel et al., 2010). Thus, introvertsmight avoid acting extraverted because they prefer lower arousalstates.7 That said, Rusting and Larsen’s data did not show intro-verts desiring low arousal states more than extraverts, and anothersimilar study found that trait extraversion was significantly asso-ciated with a higher desire for all pleasant states, even low arousalones (Kämpfe & Mitte, 2009). In terms of affect, it appears thatintroverts do not want to be happy as much as extraverts do(though this may be less true in some Asian cultures; see Tsai,Knutson, & Fung, 2006). Said another way, it seems that introvertsdesire the comparatively lower (though still generally positive)levels of happiness they experience. On the other hand, meetingthese (lower) affective expectations does not seem ideal; introvertsalso report lower levels of life satisfaction and broader psycholog-ical well-being (e.g., Ryff’s scales) than extraverts (see Steel et al.,2008, for a meta-analysis). From an objective perspective, higherlevels of subjective well-being are associated with many benefits(e.g., career success, health), and a causal role for happinessappears probable (Diener & Chan, 2011; Lyubomirsky, King, &Diener, 2005). We reiterate our discomfort with being prescriptiveand recognize important gaps in the literature, but evidence to dateseems to suggest that introverts might plausibly benefit by actingextraverted more often, even if they do not anticipate these bene-fits.

A few methodological limitations deserve mention. As we havenoted, much of this research took place in the lab, a context thatincludes somewhat limited situational variability and that might beunfamiliar to participants. That is, although they have experienceacting introverted or extraverted in general, the novel contextmight have contributed to forecasting errors. Although plausible,we suspect that the psychological situation was familiar to partic-ipants (i.e., small group interaction with peers at a university), andunfamiliarity cannot easily explain the personality interactions(familiarity would have to differ by personality and/or behavior).In addition, Study 1 asked participants to forecast in other contexts(e.g., a party), and we found similar results. Experience was notcollected in Study 1, but Fleeson et al.’s (2002) experience-sampling studies suggest that introverts enjoy extraverted behavioracross contexts in the real world, and thus, the substantial fore-casting differences likely indicate errors.

Ideally, affective forecasting studies use a mixed design tocompare forecasts both within and between subjects. In our stud-ies, all participants made forecasts, and it is thus possible that theact of making the forecasts somehow altered experience. Althoughwe cannot rule out this possibility, it seems unlikely because manyother studies have found the same effects of acting extravertedmanipulations (i.e., large main effects and null personality inter-actions) when affective forecasts were not collected. Finally, wemust acknowledge that the effect sizes in this research are notenormous, particularly for positive affect. Small differences inforecasts might argue against them being a likely cause of differ-ences in extraverted behavior. On the other hand, we suspect that

the laboratory situation may have also contributed to small effects.Study 1’s larger differences with nonlab forecasts are again sug-gestive. Outside the lab, behavior variation might appear morerisky due to stronger situations or reputational concerns (e.g.,consider an introvert vs. extravert deciding whether or not to meetfriends at a karaoke bar).

Conclusion

This research was motivated by a counterintuitive pattern offindings regarding personality and happiness: Trait introverts seemto enjoy acting extraverted more than acting introverted. Researchhas explored costs of counterdispositional behavior, such as effort,ego depletion, negative affect, and inauthenticity, but has foundvery little to argue against trait introverts pursuing extravertedbehaviors more often. It is possible (and intuitively appealing) thatcosts might materialize after extended periods of counterdisposi-tional behavior, but this remains to be tested. In contrast, ourstudies suggest that some of the intuitive costs may be errors injudgment. Introverts’ consistently underestimated the hedonic ben-efits of acting extraverted, and this was particularly apparent intheir overestimates of negative emotions. This unique affectiveforecasting error is among the first to account for individualdifferences, particularly across behavioral contexts (i.e., Person �Situation interactions). Moreover, these forecasting differences,regardless of veridicality, may ultimately help explain personalitydifferences in behavior. That is, part of what causes or maintainsdispositions may be judgments about the potential for pleasure andpain with various behavioral choices and subsequent regulatorydecisions. Of course, much further research is needed to explainthe complex causes of personality differences in behavior and therole affective forecasting plays in day-to-day life. In the meantime,we feel confident in suggesting to our introverted readers that afew more moments of extraverted behavior might be good for theirhappiness (even if they do not think so).

6 We examined the interaction between dispositional extraversion and astate arousal in predicting pleasantness but found null results in thesestudies. This null finding should be treated with caution, however, becauseour data come from a relatively narrow set of situations.

7 In our combined data, acting extraverted was associated with signifi-cant increases in arousal compared to both control and acting introvertedconditions. Acting conditions’ effect size on arousal (R2 � .185) wassimilar to that of positive affect (R2 � .172) and somewhat larger than forpleasantness (R2 � .126). Nonetheless, the pleasantness difference issubstantial, and trait introverts experienced it as much as extraverts with noconcomitant differences in negative affects, that is, recall the null Dispo-sition � Condition interactions for all affect indices.

References

Aguinis, H. (2004). Regression analysis for categorical moderators. NewYork, NY: Guilford Press.

Besner, C. M. (2009). The role of personality and current mood inpredicting behaviour and self-reported affective forecasts for imaginedsocial events (Unpublished master’s thesis). Carleton University, Ot-tawa, Ontario, Canada.

Cohen, J. (1990). Things I have learned (so far). American Psychologist,45, 1304–1312. doi:10.1037/0003-066X.45.12.1304

Diener, E., & Chan, M. Y. (2011). Happy people live longer: Subjectivewell-being contributes to health and longevity. Applied Psychology:Health and Well-Being, 3, 1–43. doi:10.1111/j.1758-0854.2010.01045.x

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

1106 ZELENSKI ET AL.

Diener, E., & Diener, C. (1996). Most people are happy. PsychologicalScience, 7, 181–185.

Diener, E., Suh, E. M., Lucas, R. E., & Smith, H. L. (1999). Subjectivewell-being: Three decades of progress. Psychological Bulletin, 125,276–302. doi:10.1037/0033-2909.125.2.276

Dunn, E. W., Biesanz, J. C., Human, L. J., & Finn, S. (2007). Misunder-standing the affective consequences of everyday social interactions: Thehidden benefits of putting one’s best face forward. Journal of Person-ality and Social Psychology, 92, 990–1005. doi:10.1037/0022-3514.92.6.990

Dunn, E. W., Brackett, M. A., Ashton-James, C., Schneiderman, E., &Salovey, P. (2007). On emotionally intelligent time travel: Individualdifferences in affective forecasting ability. Personality and Social Psy-chology Bulletin, 33, 85–93. doi:10.1177/0146167206294201

Eysenck, H. J. (1967). The biological basis of personality. Springfield, IL:Charles C Thomas.

Fleeson, W. (2001). Toward a structure- and process-integrated view ofpersonality: Traits as density distributions of states. Journal of Person-ality and Social Psychology, 80, 1011–1027. doi:10.1037/0022-3514.80.6.1011

Fleeson, W., & Gallagher, M. P. (2009). The implications of Big Fivestanding for the distribution of trait manifestation in behavior: Fifteenexperience-sampling studies and a meta-analysis. Journal of Personalityand Social Psychology, 97, 1097–1114. doi:10.1037/a0016786

Fleeson, W., Malanos, A. B., & Achille, N. M. (2002). An intraindividualprocess approach to the relationship between extraversion and positiveaffect: Is acting extraverted as “good” as being extraverted? Journal ofPersonality and Social Psychology, 83, 1409–1422. doi:10.1037/0022-3514.83.6.1409

Fleeson, W., & Wilt, J. (2010). The relevance of Big Five trait content inbehavior to subjective authenticity: Do high levels of within-personbehavioral variability undermine or enable authenticity achievement?Journal of Personality, 78, 1353–1382.

Fordyce, M. W. (1983). A program to increase happiness: Further studies.Journal of Counseling Psychology, 30, 483–498. doi:10.1037/0022-0167.30.4.483

Gallagher, P., Fleeson, W., & Hoyle, R. H. (2011). A self-regulatorymechanism for personality trait stability: Contra-trait effort. Social Psy-chological & Personality Science, 2, 335–342. doi:10.1177/1948550610390701

Gilbert, D. T., Gill, M. J., & Wilson, T. D. (2002). The future is now:Temporal correction in affective forecasting. Organizational Behaviorand Human Decision Processes, 88, 430–444. doi:10.1006/obhd.2001.2982

Gilbert, D. T., Pinel, E. C., Wilson, T. D., Blumberg, S. J., & Wheatley,T. P. (1998). Immune neglect: A source of durability bias in affectiveforecasting. Journal of Personality and Social Psychology, 75, 617–638.doi:10.1037/0022-3514.75.3.617

Gilbert, D. T., & Wilson, T. D. (2000). Miswanting: Some problems in theforecasting of future affective states. In J. Forgas (Ed.), Feeling andthinking: The role of affect in social cognition (pp. 178–197). Cam-bridge, England: Cambridge University Press.

Gray, J. A., (1981). A critique of Eysenck’s theory of personality. In H. J.Eysenck (Ed.), A model for personality (pp. 246–276). New York, NY:Springer-Verlag. doi:10.1007/978-3-642-67783-0_8

Hemenover, S. H. (2003). Individual differences in the rate of affectchange: Studies in affective chronometry. Journal of Personality andSocial Psychology, 85, 121–131. doi:10.1037/0022-3514.85.1.121

Hoerger, M., Chapman, B. P., Epstein, R. M., & Duberstein, P. R. (2012).Emotional intelligence: A theoretical framework for individual differ-ences in affective forecasting, Emotion, 12, 716–725. doi:10.1037/a0026724

Hoerger, M., & Quirk, S. W. (2010). Affective forecasting and the BigFive. Personality and Individual Differences, 49, 972–976. doi:10.1016/j.paid.2010.08.007

Kämpfe, N., & Mitte, K. (2009). What you wish is what you get? Themeaning of individual variability in desired affect and affective discrep-ancy. Journal of Research in Personality, 43, 409–418. doi:10.1016/j.jrp.2009.01.007

Kermer, D. A., Driver-Linn, E., Wilson, T. D., & Gilbert, D. T. (2006).Loss aversion is an affective forecasting error. Psychological Science,17, 649–653. doi:10.1111/j.1467-9280.2006.01760.x

Kuppens, P. (2008). Individual differences in the relationship betweenpleasure and arousal. Journal of Research in Personality, 42, 1053–1059. doi:10.1016/j.jrp.2007.10.007

Larsen, J. T., McGraw, A. P., & Cacioppo, J. T. (2001). Can people feelhappy and sad at the same time? Journal of Personality and SocialPsychology, 81, 684–696. doi:10.1037/0022-3514.81.4.684

Little, B. R. (2008). Personal projects and free traits: Personality andmotivation reconsidered. Social and Personality Psychology Compass,2, 1235–1254. doi:10.1111/j.1751-9004.2008.00106.x

Loewenstein, G. (1996). Out of control: Visceral influences on behavior.Organizational Behavior and Human Decision Processes, 65, 272–292.doi:10.1006/obhd.1996.0028

Loewenstein, G., O’Donoghue, T., & Rabin, M. (2003). Projection bias inpredicting future utility. Quarterly Journal of Economics, 118, 1209–1248. doi:10.1162/003355303322552784

Lucas, R. E., & Baird, B. M. (2004). Extraversion and emotional reactivity.Journal of Personality and Social Psychology, 86, 473– 485. doi:10.1037/0022-3514.86.3.473

Lucas, R. E., & Diener, E. (2001). Understanding extraverts’ enjoyment ofsocial situations: The importance of pleasantness. Journal of Personalityand Social Psychology, 81, 343–356. doi:10.1037/0022-3514.81.2.343

Lucas, R. E., & Fujita, F. (2000). Factors influencing the relation betweenextraversion and pleasant affect. Journal of Personality and SocialPsychology, 79, 1039–1056. doi:10.1037/0022-3514.79.6.1039

Lyubomirsky, S., King, L., & Diener, E. (2005). The benefits of frequentpositive affect: Does happiness lead to success? Psychological Bulletin,131, 803–855. doi:10.1037/0033-2909.131.6.803

Mathieu, M. T., & Gosling, S. D. (2012). The accuracy or inaccuracy ofaffective forecasts depends on how accuracy is indexed: A meta-analysisof past studies. Psychological Science, 23, 161–162. doi:10.1177/0956797611427044

Mazzucchelli, T. G., Kane, R. T., & Rees, C. S. (2010). Behavioralactivation interventions for well-being: A meta-analysis. Journal ofPositive Psychology, 5, 105–121. doi:10.1080/17439760903569154

McNiel, J. M., & Fleeson, W. (2006). The causal effects of extraversion onpositive affect and neuroticism on negative affect: Manipulating stateextraversion and state neuroticism in an experimental approach. Journalof Research in Personality, 40, 529–550. doi:10.1016/j.jrp.2005.05.003

McNiel, J. M., Lowman, J. C., & Fleeson, W. (2010). The effect state ofextraversion on four types of affect. European Journal of Personality,24, 18–35. doi:10.1002/per.738

Mellers, B. A., & McGraw, A. P. (2001). Anticipated emotions as guidesto choice. Current Directions in Psychological Science, 10, 210–214.doi:10.1111/1467-8721.00151

Mellers, B. A., Schwartz, A., & Ritov, I. (1999). Emotion-based choice.Journal of Experimental Psychology: General, 128, 332–345. doi:10.1037/0096-3445.128.3.332

Quoidbach, J., & Dunn, E. W. (2010). Personality neglect: The unforeseenimpact of personal dispositions on emotional life. Psychological Sci-ence, 21, 1783–1786. doi:10.1177/0956797610388816

Raudenbush, S. W., Bryk, A. S., & Congdon, R. (2004). HLM 6 forWindows [Computer software]. Lincolnwood, IL: Scientific SoftwareInternational.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

1107PERSONALITY AND AFFECTIVE FORECASTS

Rusting, C. L. (1998). Personality, mood, and cognitive processing ofemotional information: Three conceptual frameworks. PsychologicalBulletin, 124, 165–196. doi:10.1037/0033-2909.124.2.165

Rusting, C. L. (1999). Interactive effects of personality and mood onemotion-congruent memory and judgment. Journal of Personality andSocial Psychology, 77, 1073–1086. doi:10.1037/0022-3514.77.5.1073

Rusting, C. L., & Larsen, R. J. (1995). Moods as sources of stimulation:Relationships between personality and desired mood states. Personalityand Individual Differences, 18, 321–329. doi:10.1016/0191-8869(94)00157-N

Saucier, G. (1994). Mini-Markers: A brief version of Goldberg’s unipolarBig-Five markers. Journal of Personality Assessment, 63, 506–516.doi:10.1207/s15327752jpa6303_8

Schwarz, N. (1999). Self-reports: How the questions shape the answers.American Psychologist, 54, 93–105. doi:10.1037/0003-066X.54.2.93

Smillie, L. D., Cooper, A. J., Wilt, J., & Revelle, W. (2012). Do extravertsget more bang for the buck? Refining the affective-reactivity hypothesisof extraversion. Journal of Personality and Social Psychology, 103,306–326. doi:10.1037/a0028372

Srivastava, S., Angelo, K. M., & Vallereux, S. R. (2008). Extraversion andpositive affect: A day reconstruction study of person–environment trans-actions. Journal of Research in Personality, 42, 1613–1618. doi:10.1016/j.jrp.2008.05.002

Steel, P., Schmidt, J., & Schulz, J. (2008). Refining the relationshipbetween personality and subjective well-being. Psychological Bulletin,134, 138–161. doi:10.1037/0033-2909.134.1.138

Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics (5thed.). Boston, MA: Pearson.

Tomlinson, J. M., Carmichael, C. L., Reis, H. T., & Aron, A. (2010).Affective forecasting and individual differences: Accuracy for relationalevents and anxious attachment. Emotion, 10, 447–453. doi:10.1037/a0018701

Tsai, J. L., Knutson, B., & Fung, H. H. (2006). Cultural variation in affectvaluation. Journal of Personality and Social Psychology, 90, 288–307.doi:10.1037/0022-3514.90.2.288

Uziel, L. (2006). The extraverted and the neurotic glasses are of differentcolors. Personality and Individual Differences, 41, 745–754. doi:10.1016/j.paid.2006.03.011

Vohs, K. D., Baumeister, R. F., & Ciarocco, N. J. (2005). Self-regulationand self-presentation: Regulatory resource depletion impairs impressionmanagement and effortful self-presentation depletes regulatory re-sources. Journal of Personality and Social Psychology, 88, 632–657.doi:10.1037/0022-3514.88.4.632

Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and vali-dation of brief measures of positive and negative affect: The PANASscales. Journal of Personality and Social Psychology, 54, 1063–1070.

Wells, A. (2007). Cognition about cognition: Metacognitive therapy andchange in generalized anxiety disorder and social phobia. Cognitive andBehavioral Practice, 14, 18–25. doi:10.1016/j.cbpra.2006.01.005

Whelan, D. C., & Zelenski, J. M. (2012). Experimental evidence thatpositive moods cause sociability. Social Psychological & PersonalityScience, 3, 430–437. doi:10.1177/1948550611425194

Wilson, T. D., & Gilbert, D. T. (2005). Affective forecasting: Knowingwhat to want. Current Directions in Psychological Science, 14, 131–134. doi:10.1111/j.0963-7214.2005.00355.x

Wilson, T. D., Meyers, J., & Gilbert, D. T. (2001). Lessons from the past:Do people learn from experience that emotional reactions are short-lived? Personality and Social Psychology Bulletin, 27, 1648–1661.doi:10.1177/01461672012712008

Wilt, J., Noftle, E. E., Fleeson, W., & Spain, J. S. (2012). The dynamic roleof personality states in mediating the relationship between extraversionand positive affect. Journal of Personality, 80, 1205–1236. doi:10.1111/j.1467-6494.2011.00756.x

Zelenski, J. M. (2008). The role of personality in emotion, judgment, anddecision making. In K. Vohs, R. Baumeister, & G. Loewenstein (Eds.),Do emotions help or hurt decision making? A Hedgefoxian perspective(pp. 117–132). New York, NY: Russell Sage Foundation.

Zelenski, J. M., & Larsen, R. J. (1999). Susceptibility to affect: A com-parison of three personality taxonomies. Journal of Personality, 67,761–791. doi:10.1111/1467-6494.00072

Zelenski, J. M., & Larsen, R. J. (2002). Predicting the future: Howaffect-related personality traits influence likelihood judgments of futureevents. Personality and Social Psychology Bulletin, 28, 1000–1010.

Zelenski, J. M., Santoro, M. S., & Whelan, D. C. (2012). Would introvertsbe better off if they acted more like extraverts? Exploring emotional andcognitive consequences of counterdispositional behavior. Emotion, 12,290–303. doi:10.1037/a0025169

Zelenski, J. M., Sobocko, K., & Whelan, D. C. (in press). Introversion,solitude, and subjective well-being. In R. J. Coplan & J. C. Bowker(Eds.), A handbook of solitude: Psychological perspectives on socialisolation, social withdrawal, and being alone. New York, NY: Wiley-Blackwell.

Received September 11, 2012Revision received December 13, 2012

Accepted February 19, 2013 �

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

1108 ZELENSKI ET AL.