A Functional Analysis of Stages of Spontaneous Impression Formation, Serving Nonconscious Affiliation Goals

We examined the effect of a nonconscious affiliation goal on implicit impression formation at two stages: implicit activation of traits and implicit binding of traits to actors.. We prop

A Functional Analysis of Stages of Spontaneous Impression Formation, Trait Inference in

Serving Nonconscious Affiliation Goals

SoYon RimNew York University

Kate E MinDuke University

James S UlemanNew York University

and Tanya L ChartrandDuke University

Keywords: impression formation, spontaneous trait inferences, affiliation motivation, attribution,social cognition, social perception

Address correspondence to

SoYon Rim, Department of Psychology, New York University, 6 Washington Place, Room 766, New York, NY 10003 E-mail: soyon.rim@nyu.edu, Phone: 212-998-7809

nonconscious desire to fulfill an affiliation goal led perceivers to suppress accessibility of

nontrait concepts, and to spontaneously and selectively form more positive trait inferences about others These studies uncover the underlying machinery of implicit impression formation and pinpoint when positivity biases occur as a result of an affiliation goal In addition, they are the first to provide evidence that STIs are sensitive to perceivers’ nonconscious goals and operate in functional ways in impression formation

A Functional Analysis of Spontaneous Trait Inference in Nonconscious Affiliation Goals

Imagine that you are meeting the parents of your fiancé for the first time at a family barbeque at their place You observe your future mother-in-law smiling and chatting with all of her guests, tending to everyone’s needs, and engaging all guests, including you But you also observe that she eats with her fingers, knocks over a cup of water, and trips over someone’s foot

Do you spontaneously draw the inference that she is sociable and friendly, and also that she is messy and clumsy? Clearly, you would like to become close to your fiancé’s mother How does this motivation automatically shape your attention and the inferences you draw from her

behaviors? The purpose of the present research is to investigate this intriguing question We examined the effect of a nonconscious affiliation goal on implicit impression formation at two

stages: implicit activation of traits and implicit binding of traits to actors We propose that an

affiliation goal initially leads to suppression of trait-unrelated concepts (thus facilitating attention

to trait-related concepts), and later to selective formation and binding of positive spontaneous trait inferences (STIs) to representations of actors In the above example, this means that in forming an implicit impression of your future mother-in-law, your attention will first be drawn away from your future mother-in-law’s incidental behaviors (e.g., clearing the plates, taking out the garbage) that do not inform your impression, thereby allowing you to focus on the trait-implications relevantof her behaviors, both positive and negative Subsequently, the positive traits (sociable and friendly), but not the negative traits (messy and clumsy), that you

spontaneously inferred would become more bound to the rose-colored representation of your future mother-in-law in memory

Affiliation Motive and Intentional Impressions

In line with the general notion that social thinking serves a practical purpose or that

“thinking is for doing” (Fiske, 1992; McArthur & Baron, 1983), past research has shown that goals affect the kinds of impressions that people form of others (Devine, Sedikides, & Fuhrman, 1989; Fein & Spencer, 1997; Clark & Wegener, 2008; Goodwin et al., 2002; Scholer & Higgins, 2008) Goals can lead a perceiver to be positively biased in their impressions as in the case of affiliation-related goals (Clark & Wegener, 2008; Goodwin et al., 2002) But they can also lead

to more negatively biased impressions when the perceiver is motivated to self-enhance, as demonstrated by the large body of literature on the effect of self-esteem threat and goals in general on stereotyping of others (Fein, Hoshino-Browne, Davies, & Spencer, 2003; Fein & Spencer, 1997; Gilbert & Hixon, 1991; Sinclair & Kunda, 2000; Spencer, Fein, Wolfe, Fong, & Dunn, 1998; for reviews, see Blair, 2002 and Kunda & Spencer, 2003) Most relevant, however, for the present research is work on the effect of affiliation-related goals on impression formation

Goodwin et al (2002) examined the effect of romantic goals on explicit evaluations of goal-relevant and –irrelevant characteristics Participants watched a video presentation made by aperson who they believed was a prospective date (vs a random person) and subsequently rated the target on various goal-relevant (e.g., social) vs goal-irrelevant (e.g., presentation

competence) qualities Half of the participants viewed a video of a competent actor (i.e., in terms

of presentation skills) and the other half saw an incompetent actor Subsequently, participants were asked to rate the actor on various dimensions related to competence and sociability Results showed that participants rated task-competent and –incompetent prospective dates as equally competent, thus showing a positivity bias Participants also rated the actor more favorably on various social qualities when the actor was (vs was not) a prospective date, irrespective of manipulated task-competence Murray and colleagues reported similar positivity biases in the

evaluation of romantic partners within the context of close relationships (Murray & Holmes, 1993; Murray, Holmes, & Griffin, 1996).

Similarly, past research has shown that dependencey on another person for desired outcomes leads to a positivity bias in intentional impressions (Berscheid, Graziano, Monson, & Dermer, 1976; Clark & Wegener, 2008) Clark and Wegener (2008, Experiment 2) recently demonstrated that people evaluate others more positively when they expect to interact with them and when a smooth interaction ensures a larger monetary reward Also, Berscheid et al (1976) manipulated outcome dependency in a dating context and found that participants attended more

to targets they expected to date (as assessed by looking time), remembered more information about them, and rated them more positively on various trait measures (e.g., warm-cold, sensitive-insensitive) than targets they did not expect to date

Thus, past work on goals’ effects on explicit trait judgments and evaluations demonstratesthat perceivers’ momentary goals are important in shaping intentional impressions However, impression formation in daily life occurs, for the most part, spontaneously and passively, withoutexternal instructions to form impressions, and without perceivers ever having to explicitly verbalize those impressions In addition, in navigating a highly complex social environment, the perceiver must often make sense of social information quickly and efficiently, without draining

valuable cognitive resources Therefore, it is critical to examine the implicit processes that underlie the explicit judgments and evaluations that have been studied thus far When do

positivity biases in impression formation arise as a result of an affiliation goal? Do these biases surface at the level of automatic trait activation? Or do they manifest at the later stage of implicit trait binding to actor representations in memory? It is also possible that implicit trait impressions are impervious to such goal manipulations and that an affiliation goal only biases later,

intentional judgments, such as the ones that have been previously studied (Berscheid et al., 1976;Clark & Wegener, 2008; Goodwin et al., 2002).? Thus, the present research is a contribution in two respects: 1) it takes a process-focused approach to the study of an affiliation goal’s effects on

impressions and examines when positivity biases arise duringin the impression formation

process, and 2) it sheds light on whether implicit impressions, namely spontaneous trait

inferences formed from behaviors, are sensitive to temporarily activated social goals, such as the goal to affiliate

Spontaneous Trait Inferences

Spontaneous trait inferences (STIs; for a review, see Uleman, Saribay, & Gonzalez, 2008)are inferences that people form upon reading (e.g., Carlston & Skowronski, 1994; Todorov & Uleman, 2002) or observing (Fiedler & Schenk, 2001; Fiedler, Schenk, Watling, & Menges, 2005) other people’s trait-implying behaviors For instance, while reading the sentence, “He returned the wallet with all the money in it,” people have been shown to spontaneously generate the inference, “honest.” Such STIs form in the absence of any explicit task instructions to form impressions, as participants are usually given memory instructions, and without perceivers’ explicit awareness of having formed any inferences And STI effects are robust, occurring under various conditions thought to interfere with the trait-encoding process, such as concurrent cognitive load (Todorov & Uleman, 2003; Winter, Uleman, & Cunniff, 1984) STI formation has been indirectly assessed using a number of different paradigms, including the lexical decision tasks, recognition probes, savings-in-relearning, and the false recognition paradigms (for a review, see Uleman, Newman, & Moskowitz, 1996)

The apparent ubiquity and robustness of spontaneous trait inferences (Todorov &

Uleman, 2002; Winter, Uleman, & Cunniff, 1985) makes it unclear whether they are sensitive to

temporarily activated social motivations of the perceiver In fact, Uleman (1999) speculated that

“spontaneous impressions are guided more by chronically accessible constructs, while

intentional impressions are guided more by temporarily activated goal-relevant constructs and procedures” and that “spontaneous impressions are less focused, more wide-ranging, and more promiscuous, whereas intentional impressions are focused” (p 146) Indeed, STIs have been shown to be sensitive to chronically activated goal constructs (Zelli, Cervone, & Huesmann, 1996; Zelli, Huesmann, & Cervone, 1995; Moskowitz, 1993; Newman, 1991; Zárate, Uleman, &Voils, 2001) For example, Moskowitz (1993) found that people who scored high on personal need for structure (PNS; Neuberg & Newsom, 1993), a desire for certainty and aversion to ambiguity, formed more STIs from behaviors than those low on PNS And although past researchhas shown that STIs are affected by task instructions that alter the processing of STIs (Uleman &Moskowitz, 1994), these manipulations do not speak to the issue of functionality Thus, it is possible that spontaneous trait inferences are unaffected by temporarily activated social goals and that goals’ effects manifest only during at the stage of intentional impressions

But this may not be true Recently, Rim, Uleman, and Trope (2009) provided indirect evidence that STI formation can be modulated by contextual information that makes STIs more

or less functional in forming impressions From the same behavioral information, participants formed STIs to a greater extent about actors who were from the distant (vs near) past or located

in a remote (vs proximal) place These results were consistent with construal level theory (CLT; Trope & Liberman, 2010), which predicts a functional and associative relationship between psychological distance and level of construal According to CLT, distal (vs proximal) others are represented more in terms of their high-level (i.e., abstract and central) features, which make STIs more likely for distal vs proximal entities Underlying this is a functional relationship

between distance and construal in which high-level, abstract construals are more useful for representing distal objects, events, and people Rim et al (2009) postulated that STIs are more functional for representing distant others because the specifics of the immediate situation (e.g., exact behaviors) may not always hold for those individuals Abstract traits are more stable and invariant with psychological distance and hence are more useful in representing distal people Although such a functional relationship seems reasonable, these studies did not unequivocally directly show that STIs can be flexibly formed according to their functionality in the moment

Stages of Spontaneous Trait Inference Formation

Spontaneous trait inference formation occurs in two stages: 1) activation of trait concepts from behavioral information implying traits (for a review of concept activation, see Förster & Liberman, 2007) and 2) “binding” or linking of inferred traits to actor representations in long term memory (Carlston & Skowronski, 1994; Skowronski, Carlston, & Hartnett, 2008; Zárate, Uleman, & Voils, 2001) Under certain conditions, traits might be inferred (activated) but not bound to actors, while in other cases they will be inferred and bound to the appropriate actors Atwhat stage of the impression formation process does might an affiliation goal have its influence? Our prediction is that an affiliation goal will affect both stages of spontaneous trait inference formation, but in different ways

Activation Traits are spontaneously activated upon exposure to trait-implying behaviors

of an actor, and this activation can be assessed using procedures such as lexical decision,

recognition probe, or word-stem completion (e.g., Ham & Vonk, 2003; Uleman, Hon, Roman, & Moskowitz, 1996; Zárate, Uleman, & Voils, 2001) When behaviors have both trait and

situational implications, research has shown that both meanings are simultaneously activated andselection based on contextual goals is made later in making intentional judgments (Todd,

Molden, Ham, & Vonk, 2010) Furthermore, in text comprehension, concepts (e.g., bug) with

multiple meanings initially activate all other concepts (e.g., ant and spy) related to them.and sSuppression of concepts irrelevant to a particular context occursonly after some delay (Swinney,1979; Gernsbacher & Faust, 1991) Based on this body of evidence on activation of concepts, wepredicted that an affiliation goal will lead to increased activation of traits, both positive and negative, which are relevant for forming impressions However, goals not only heighten

activation of goal-relevant stimuli, but suppress activation of goal-irrelevant stimuli (for a review, see Dijksterhuis & Aarts, 2010) Therefore, we also predicted that an affiliation goal will lead to increased suppression of trait-unrelated concepts

Binding Activated traits can then be bound (or not) to actor representations and stored in

long term memory Binding can be assessed using the savings-in-relearning paradigm (Carlston

& Skowronski, 1994) or the false recognition paradigm (Todorov & Uleman, 2002) In the false recognition paradigm, which we use in this research, participants initially study photographs of actors paired with trait-implying behavioral sentences about them Later, they are shown the same actors’ photos, some correctly paired with the pictured actor’s implied trait (from previous behaviors), and others incorrectly paired with another actor’s implied trait On a recognition test, participants are more likely to falsely recognize implied traits as having been presented in the earlier behavioral sentences when photos are correctly (vs incorrectly) paired with the right

actor This provides strong evidence that traits are bound to specific actor representations Unlike

simple activation, we predicted a positivity bias will to occur for in terms of implicit trait binding

as assessed using this procedure AThat is, an affiliation goal should will lead to greater positive spontaneous trait inference formation and binding, but have no effect on negative STIs

The Present Research

The purpose of the present research is to 1) examine different stages of the implicit impression formation process to pinpoint if and when positivity biases arise as a result of an affiliation goal and 2) to provide direct evidence that spontaneous trait inferences form flexibly depending on their functionality for activated perceiver goals In Experiment 1, we primed people with the goal to affiliate or no goal and measured implicit activation of trait vs nontrait (control) words and nonwords in a lexical decision task We predicted that an affiliation goal (vs

no goal) would lead to greater activation of trait concepts and greater suppression of

trait-unrelated concepts

In Experiment 2, we examined implicit binding of traits to actors In addition to the affiliation goal and no goal primes, we also included a positive semantic prime as a way to distinguish between the positivity of the affiliation goal and the goal itself Since our prediction was that an affiliation goal would lead to a positivity bias in implicit impressions, we wanted to rule out the possibility that this would occur simply because the affiliation goal primes general positivity rather than the goal to affiliate Participants were in one of three prime conditions: affiliation goal, no goal, and positive semantic Subsequently, we assessed binding of positive and negative traits to specific actors We predicted that at this later stage of implicit impression formation, perceivers with the goal to affiliate would spontaneously bind to actors more of the positive traits they inferred relative to those with no goal or those primed with positivity This effect of prime was not predicted for STIs from negative behaviors

Experiment 3 was conducted to provide unequivocal evidence that an affiliation goal leads to a positivity bias in implicit binding of spontaneous trait inferences to actors In this study, we examined the effect of a goal prime and goal fulfillment on STI formation from

positive and negative behaviors We predicted that the positivity bias found in Experiment 2

would be eliminated if the perceiver’s goal to affiliate was fulfilled prior to the impression formation (STI) task; when left unfulfilled, a positivity bias in selective trait binding would occur To summarize, we had two main hypotheses:

Hypothesis 1: Nonconscious priming of an affiliation goal (vs no goal) leads to greater

activation of trait concepts and greater suppression of trait-unrelated concepts during encoding ofbehaviors (Experiment 1)

Hypothesis 2: Nonconscious priming of an affiliation goal (vs no goal) leads to

selectively greater binding of positive (but not negative) spontaneous trait inferences to specific actors in memory (Experiments 2 and 3)

Experiment 1: Effect of Affiliation Goal on Implicit Trait Activation

Experiment 1 tested the effect of an affiliation goal on implicit trait activation

Participants were primed with the goal to affiliate or not Then they completed a lexical decision task where implicit activation of trait vs nontrait (control) words and nonwords were measured

We hypothesized that a nonconscious affiliation goal (vs no goal) would lead to greater

activation of traits and greater suppression of nontraits

Method

Participants Twenty-eight Duke University undergraduate students participated in the

study for monetary compensation Participants were randomly assigned to one of two

between-Ss conditions

Stimuli To prime the goal, a word-search puzzle containing a 9 x 9 matrix of letters was

presented Words could appear with letters in a horizontal or vertical line, either from left to right

or from right to left, reading down or reading up, and diagonally reading either down or up Both

of the prime conditions contained the same set of three neutral words to be found (eagles,

calendar, plant) with the remaining six words relevant (or not) to the concept of affiliation In the nonconscious affiliation goal prime condition, these words were companion, supporter, partner, together, affiliate, and kindness In the no goal prime condition, these words were stripe,

building, keyboard, table, paper, and window.

In the lexical decision task, there were a total of 12 trials, which were comprised of 12 trait-implying sentences followed by four words: one trait-implying word, one control word, and two nonwords Twelve trait-implying sentences were selected from those used by Uleman (1988) Six of these sentences implied positive traits and another six sentences implied negative traits Each of these positive and negative trait-implying sentences was matched with 1) a trait-implying word, 2) a control word that was equal in length and lexical frequency to the trait-implying word (Kučera & Francis, 1967), and 3) two nonwords that were equal in length to the trait-implying word

All 12 trait words had an overall consensus rate of 50% or greater on an earlier pretest in which participants were explicitly asked to generate a trait that was implied by each of the sentences (Uleman, 1988)

Procedure An experimenter, blind to prime condition, greeted participants upon their

arrival in the lab The experimenter told them that they were completing two separate

experiments, the first of which was a word search task (actually a supraliminal priming task) Wefollowed the priming procedure used by Bargh, Gollwitzer, Lee-Chai, Barndollar, and Trötschel (2001; see also Bargh & Gollwitzer, 1994) Participants were seated at a desk and randomly assigned to one of two prime conditions (affiliation goal or no goal) In the affiliation goal condition, the goal was primed by a word-search puzzle in which synonyms of affiliation were

presented (or not, in the no goal condition) When participants had found all of the nine words

embedded in the letter matrix, the experimenter instructed them to set the word-search task aside,and continue on to the next, supposedly unrelated, experiment, which allegedly involved

linguistic processing on the computer (actually a lexical decision task used to assess spontaneoustrait inference formation; e.g., Todd et al., 2010; Zárate, Uleman, & Voils, 2001)

We followed the lexical decision procedure used by Todd et al (2010) Participants determined as quickly and accurately as possible whether or not a series of letter strings

represented actual English words Each block of trials started with a trait-implying sentence briefly presented for 3000 ms on a computer screen, followed by a trait word, a control word, and two nonwords The presentation of these blocks, and words within each block, were

randomized When a letter string appeared on the screen, participants classified it as a word or nonword by pressing one of two keys labeled “Yes” and “No.” Once participants had responded, the next block of trials began automatically Following the final block of trials, participants were asked to complete a funnel debriefing questionnaire that probed for awareness or suspicion regarding the priming manipulation (Bargh & Chartrand, 2000) They were asked (1) what they thought the purpose of the experiment had been, (2) whether they thought any of the tasks had been related, (3) whether anything they had done on one task had affected what they had done onany of the other tasks, (4) whether they had ever seen or completed a word search task for another experiment, and (5) whether they attempted to form an impression of the actor No participant reported any awareness or suspicion of a connection between the different tasks of theexperiment or indicated that completing one task might have affected responses on another Further, none of the participants reported intention or awareness of having made inferences from the behaviors Thus, their inferences were spontaneous (unintended and nonconscious)

Results and Discussion

Response latencies Participants’ lexical decision latencies served as our primary index

of implicit trait activation Shorter latencies for traits as compared to matched control words would indicate that traits had been activated while reading the trait implying behavior

descriptions (Hassin, Aarts, & Ferguson, 2005)

Incorrect responses in identifying words vs nonwords were excluded from the analyses (5.3% of all responses) Given that reaction times tend to be positively skewed, we also

eliminated all response latencies faster than 200 ms and slower than 2000 ms and subjected the remaining latencies to an inverse (1/x) transformation (see Ratcliff, 1993) For ease of

interpretation, means are reported in milliseconds

Activation of trait inferences A repeated-measures analyses of variance (ANOVA) on

word type (trait, control, and nonword) on participants’ transformed lexical decision latencies

revealed a main effect of word type, F(2, 26) = 41.90, p =<.001,  = 76, (see Figure 1), and all p2pairwise comparisons were significant (ps < 001) Importantly, participants were faster to respond to trait words (M = 607.63, SD = 68.43) than both control words (M = 659.92, SD = 83.26; t(27) = -5.17, p = <.001, d = 98) and nonwords (M = 742.71, SD = 94.52 ; t(27) = -9.07,

p = < 001, d = 1.72) This provides evidence that traits were activated spontaneously (Todd et

al., 2010)

Effect of an affiliation goal on activation of trait inferences More critical to the

central hypothesis was an analysis of the transformed lexical decision latencies in a 2 (Prime:

affiliation goal vs no goal) x 2 (Word Type: trait and control) mixed-model ANOVA with Prime between-Ss and Word Type within-Ss This revealed a significant main effect of word type with

faster responses to trait words (M = 607.64, SD = 68.43) compared to control words (M =

659.92, SD = 83.26), F(1, 26) = 43.69, p <.001,  = 63 There was also the predicted significant2p

2-way interaction between prime and word type, F(1, 26) = 12.48, p = 002,  = 32 (see Figure 2p2) Participants were faster to respond to trait words than control words in both affiliation goal,

F(1, 26) = 45.01, p < 001,  = 63, and no goal, F(1, 26) = 5.521, p < 05, 2p  = 18, prime 2p

conditions, thereby showing trait activation across conditions However, as indicated by the significant interaction, this difference in activation between trait and control words was

significantly greater when participants were primed with the goal to affiliate (vs no goal)

Pairwise comparisons revealed that having an affiliation goal led to slower responses to

control words (M = 713.44, SD = 58.92) relative to having no goal (M = 619.77, SD = 77.71),

F(1, 26) = 14.93, p = 001,  = 37 Having an affiliation goal, however, did not lead to faster 2p responses to trait words (M = 619.96, SD = 58.92) compared to not having a goal (M = 598.40,

SD = 75.30) F (1, 26) = 1.51, p = 23 Thus, an affiliation goal (vs no goal) led to greater

suppression of trait-unrelated words but not greater activation of trait words

Effect of an affiliation goal on positive and negative trait inferences To demonstrate

that the goal to affiliate leads to greater activation of both positive and negative trait inferences from others’ behaviors, we conducted the same analysis on latencies to positive and negative traits separately We found the identical pattern: there was a significant Prime x Word Type 2-

way interaction for both positive traits, F(1, 26) = 4.50, p < 05,  = 15, and negative traits, 2p

F(1, 26) = 17.21, p < 001,  = 40 In both cases, the effect was driven by responses to control 2p

words, F(1, 26) = 14.93, p = 001,  = 37 Based on these analyses, we can conclude that there 2p

is no dissociation in the effect of an affiliation goal prime on positive and negative traits at the level of activation

Experiment 1 provides evidence that the goal to affiliate (vs no goal) leads to greater suppression of trait-unrelated concepts While traits were activated across all conditions, as evidenced by shorter latencies to trait vs nontrait words in both the affiliation goal and no goal conditions, this difference was greater under the influence of an affiliation goal relative to a no goal prime due to longer latencies to control words Partially supporting our hypotheses, the goal

to affiliate (vs no goal) led to greater suppression of nontrait concepts that could potentially

interfere with the goal but not necessarily to greater activation of trait concepts This suggests floor effects in the activation of traits, as mean latencies to trait words were extremely short (<

608 ms) Further discussion of these results is provided in the General Discussion in light of past research using this paradigm

At the level of activation, we found that an affiliation (vs no goal) goal leads to greater

suppression of goal-irrelevant, nontrait concepts However, at the level of trait binding,

impressions should become more fine-tuned and an affiliation goal (vs no goal) prime should lead to greater binding of spontaneously inferred positive traits but not negative traits, thereby resulting in more positive implicit impressions when motivated to affiliate We tested this

possibility in Experiment 2 Furthermore, in Experiment 1, we primed words related to the affiliation goal construct, but it is also possible that other constructs, such as the “positivity” of the affiliation goal, were activated and thus affected STI formation Another purpose of

Experiment 2 was to separate the “positivity” of the affiliation goal prime from the goal itself, and thus we included a positive semantic prime in addition to the affiliation goal and no goal primes

Experiment 2: Effect of Affiliation Goal on Positive Spontaneous Trait Inferences

Experiment 2 tested the effect of an affiliation goal on spontaneous trait inferences at the level of trait binding to specific actors (as opposed to mere trait activation as in Experiment 1) Participants were either in an affiliation goal, no goal, or positive semantic prime condition Theycompleted the false recognition task, which assesses not only trait inferences formation but trait binding to actors We hypothesized that an affiliation goal prime (vs no goal prime and positive semantic prime) would lead to greater formation and binding of positive spontaneous trait inferences This effect of priming an affiliation goal was not expected for negative spontaneous trait inferences

Method

Participants Forty-eight Duke University undergraduate students participated in the

experiment for monetary compensation Participants were randomly assigned to one of six between-Ss conditions

Stimuli To prime the goal, a word-search puzzle packet containing one 9 x 9 matrix and

five separate 6 x 6 matrices of letters were presented All three of the prime conditions contained

the same set of three neutral words to be found in each of the matrices (e.g., eagles, calendar, plant) and the remaining six words were either relevant to affiliation (e.g., companion, partner, together), or were positive and unrelated to affiliation (e.g., butterfly, ice-cream, sunshine) or neutral words (e.g., table, chair, pencil)

In the false recognition paradigm, there were 36 study phase trials comprised of implying sentences selected from Uleman (1988) Eighteen sentences implied positive traits and

trait-18 sentences implied negative traits about the actor All 36 trait words had an overall consensus

of 50% or greater in pretesting for agreement on the trait best exemplified by the sentences (Uleman, 1988) All of these traits were used as probes in the second part of the STI task

Procedure An experimenter, blind to prime condition, greeted participants upon their

arrival in the lab The experimenter told them that they would complete two separate

experiments, the first of which was a word search task (actually a supraliminal prime task as in Experiment 1) Participants were seated at a desk and randomly assigned to one of three prime conditions: affiliation goal, positive semantic, or no goal In all conditions, participants

completed a word search task in which they had to find nine words embedded in a 9 x 9 letter matrix Six of the nine words were affiliation-related, neutral or positive depending on the condition When participants had found all of the nine words embedded in the 9 x 9 letter matrix,the experimenter instructed them to set the task aside, and continue onto the next “experiment,” which allegedly involved a linguistic processing task on the computer (actually the false

recognition paradigm used to assess spontaneous trait inference formation; Todorov & Uleman, 2002)

In the false recognition task, all participants were told that this was a study of how individuals memorize information about other people Participants worked individually in sound-proof cubicles, and instructions were presented on the computer Participants were told that the experiment consisted of two parts In the first part (study phase), they would be shown pictures

of people with information about them, and in the second part (test phase), their memory for this information would be tested The experiment began with a practice trial and, if everything was clear to the participant, continued with the study phase of the false recognition test Participants were presented with 36 photo/behavior pairs (trials) The order of the 36 trials was randomized for each participant In 12 of the trials, the sentences contained the trait implied by the behavior, and these served as filler sentences In the remaining 24 trials, the sentences only implied traits The duration of each trial was 8 s, and the inter-trial delay was 2 s After every 6 trials, during the

inter-trial delay, participants were asked to stop and complete one of the five 6 x 6 letter

matrices, similar to the 9 x 9 matrix completed at the very start of the experiment This was a way to keep the prime (affiliation goal, no goal, or positive semantic) activated in participants’ minds

Upon completion of the study phase, participants were reminded that in the second part, they would be tested for their memory of the photo/behavior pairs that they were presented with

in the first part of the experiment They were presented with faces from the first part of the experiment, each paired with a single trait word Their task was to decide whether they had seen the word in the sentence about this person during the first part of the experiment Participants

pressed the key (/) labeled old if they believed that they had seen the word in the study phase, or the key (Z) labeled new if they believed that they had not seen the word Participants completed

8 practice trials in which feedback about the correct response was given, and continued with the test phase if they were clear about the instructions

The test phase comprised 36 trials where participants saw photos from the study phase paired with trait words The trait word was below the photo, and each trial followed participants’ responses automatically In 12 of the trials, photos presented with filler sentences during the study phase were correctly paired with traits that had been in the sentences earlier (filler trials) Another 12 trials consisted of photos systematically and correctly paired with traits that were implied about those people (experimental trials), and the remaining 12 consisted of photos randomly and incorrectly paired with traits that were implied about another person (control trials) There were two counterbalanced sets such that a photo correctly paired with a trait in one set was incorrectly paired with a trait in the other set This ensured that any effects would not be due to a particular photo or trait word pairing Thus the overall design was a 3 (Prime: affiliation

goal vs positivity vs no goal) x 2 (Counterbalancing Set) x 3 (Trial Type: filler, experimental, and control) mixed-model ANOVA with the first two factors between-Ss and the last factor within-Ss

Following the final block of trials, participants completed a funnel debriefing

questionnaire that probed for awareness or suspicion regarding the priming manipulation (as in Experiment 1) No participant reported any awareness or suspicion of a connection between the different tasks of the experiment or indicated that completing one task might have affected responses on another Further, none of the participants reported intentions or awareness of havingmade inferences from the behaviors Thus, their inferences were spontaneous

Results and discussion

Evidence of spontaneous trait inferences The proportion of false recognition of

implied traits was calculated for the experimental and control trials for each participant For fillertrials, the proportion of correct recognition was calculated

Correct recognition of traits that had actually been presented with the photo during the

study phase was 68 (SD = 16) This significantly exceeded the chance level of 50, t(47) = 8.40,

p < 001, d = 1.17 More importantly, false recognition of traits was greater on experimental trials where traits and photos were systematically paired (M = 32, SD = 25) than on control trials where they were randomly paired (M = 13, SD = 13), t(47) = 7.28, p < 001, d = 1.05 (see

Figure 3) This provides evidence that spontaneous trait inferences had been formed (Todorov & Uleman, 2002)

Effect of an affiliation goal on positive trait inferences AWe did a 3 (Prime:

affiliation goal vs positive semantic vs no goal) x 2 (Counterbalancing Set) x 2 (Trial Type: experimental and control) mixed-model ANOVA, with the first two factors between-Ss and the

last factor within Ss, on the positive trait trials This revealed a significant main effect of trial type, with greater false recognition of positive traits on experimental (M = 37, SD = 26) vs

control (M = 13, SD = 13) trials, F(1, 42) = 64.65, p < 01,  = 61 More importantly, as p2

predicted, there was a significant Prime x Trial Type interaction on false recognition of positive

traits, F(2,42) = 3.70, p = 03,  = 15 (see Figure 4) This effect was driven by differences in 2p false recognition rates on experimental, F(2, 42) = 2.75, p = 075, and not control trials, F(2, 42)

< 1, p = ns No other effects were significant (Fs < 1.75, ps > 18)

Pairwise comparisons by prime confirmed that false recognition of positive traits was

significantly greater on experimental trials in the affiliation goal prime condition (M = 48, SD

= 29) as compared to the no goal prime condition (M = 31, SD = 24), F(1, 42) = 3.83, p < 05 False recognition of positive traits in the positive semantic prime (M = 31, SD = 21) did not differ from the no goal condition, F(1, 42) < 1, p = ns None of the pairwise comparisons for false recognition on control trials was significant (Fs < 1).

Additionally, we conducted pairwise comparisons by trial type to examine whether STIs had been formed in all priming conditions False recognition of positive traits on experimental trials was greater than on control trials in all three prime conditions: affiliation goal, F(1, 42) =

50.05, p <.001,  = 54), no goal, F(1, 42) = 12.31, p = 001, 2p  = 23), and positive semantic, p2

F(1, 42) = 12.64, p = 001,  = 23), prime conditions Therefore, STIs had been formed in all 2pconditions, though tjeu were it was greatest in the affiliation goal prime condition, as indicated

by the significant interaction

Effect of an affiliation goal on negative trait inferences We did a A 3 (Prime:

affiliation goal vs positive semantic vs no goal) x 2 (Counterbalancing Set) x 2 (Trial Type:

experimental and control) mixed-model ANOVA, with the first two factors between-Ss and the last factor within-Ss, on negative trait trials This revealed only a significant main effect of trial

type, F(1, 42) = 17.64, p < 001, = 30, with greater false recognition of negative traits on

experimental vs control trials Importantly, the Prime x Trial Type 2-way interaction was not

significant, F(1, 42) < 1, p = ns, and neither were any other effects (Fs < 1)

Thus an affiliation goal (vs no goal) prime led to selectively greater positive,(relative to but not negative,) spontaneous trait inference formation and binding In addition, the positive semantic prime had no effect relative to the no goal prime condition, which rules out the

possibility that the affiliation goal simply primed general positivity rather than the goal to affiliate As predicted, goal priming had no effect on formation and binding of negative STIs to actors

Experiment 3

Unlike simple activation of traits, spontaneous trait inference formation as assessed by

the false recognition paradigm signifies binding of traits to specific actors in memory

Experiment 2 demonstrates that at this later stage of implicit impression formation, an affiliation goal (vs no goal) prime leads to greater binding of spontaneously inferred positive, but not negative, traits to actors These results are consistent with our functional hypothesis that the goal

to affiliate leads to a more positive implicit impression of others

In Experiment 3, we wanted to demonstrate unequivocally that the effect of the affiliationgoal prime on positive STI formation was indeed a goal effect We primed participants using the same procedure as in Experiment 1 Then participants were either fulfilled or unfulfilled in their goal (if one existed) via a purported online interview with another student Subsequently, they completed the same STI task as in Experiment 2 Insofar as the affiliation goal prime actually

primes a goal, its effects should persist until the goal is fulfilled (Dijksterhuis & Aarts, 2010; Förster & Liberman, 2007; van den Bos & Stapel, 2009), and its effects should decrease once thegoal has been fulfilled (Atkinson & Birch, 1970; Chartrand, Huber, Shiv, & Tanner, 2008) Specifically, our prediction was that when the goal to affiliate is activated and then fulfilled, participants will not show a positivity bias in terms of increased positive STI formation and binding On the other hand, when the goal remains unfulfilled, the goal will persist and result in the positivity bias demonstrated in Experiment 2.

Method

Participants Sixty-two Duke University undergraduate students participated in the

experiment for monetary compensation Participants were randomly assigned to one of eight between-Ss conditions

Stimuli In an online interview, participants (as the interviewers) asked the confederate

(supposedly the interviewee) eleven questions about being a student at Duke University (e.g.,

“Do you use any of the recreation facilities on campus?”) The answers to the interview

questions were scripted for the goal-fulfilled and goal-unfulfilled conditions, and thus the confederate answered accordingly The answers were friendly in the goal-fulfilled condition (e.g., “I occasionally use them – I should probably use them much more than I do!”) and

unfriendly in the goal-unfulfilled condition (e.g., “Occasionally)

Procedure An experimenter blind to prime condition greeted participants upon their

arrival in the lab The experimenter told them that they would complete three separate

experiments In the first “experiment,” participants were initially primed with an affiliation goal

or not using the same word search task as in Experiment 1 Subsequently, in the second

“experiment,” in an adaptation of an interview procedure used by Lakin and Chartrand (2003),