Publications

Social impressions are often measured as stable endpoint judgments following a social encounter, despite naturally arising from temporally unfolding streams of social information (e.g., actions, speech, facial and bodily expressions). Here, we tested whether impressions emerge through progressively constructed, path-dependent trajectories of inferences made during social encounters. Study 1 (N = 404) established that impressions integrate incrementally during an encounter, rather than being assembled retrospectively at its conclusion. Impressions tracked moment-to-moment along a single dimension (e.g., warmth) closely corresponded to endpoint judgments of the same “get-to-know-you” videos truncated at matched timepoints. Studies 2–3 (N = 456) then examined how impressions are constructed from spontaneous inferences across multiple dimensions as an encounter unfolds. Participants generated free-response descriptions while viewing videos, allowing impressions to be modeled as trajectories through high-dimensional semantic space. Across an encounter, successive spontaneous inferences demonstrated incremental integration, whereby later inferences were increasingly semantically tethered to prior ones. Moreover, the unique temporal-semantic trajectories these inferences traced predicted downstream personality judgments and social decisions. Study 4 (N = 215) then offered the causal test of this trajectory framework by experimentally reordering the same video content presented to participants. Reordering reshaped impression trajectories and changed the inferences drawn from identical segments of content depending on their preceding content; order-induced differences in trajectory structure, in turn, predicted downstream differences in judgments and decisions. Our findings advance a view of impressions not as static accumulations of encoded social information, but as temporally-structured trajectories reflecting the evolving construction of social meaning.

Perceivers often draw inferences of personality from facial appearance, exerting pervasive influence on social cognition and behavior, despite their limited validity. These biases have proven remarkably resistant to mitigation efforts. This research examines two common interventions against facial stereotyping: counter-stereotype exposure and explicit instructions to control the bias. Across three experiments (N=1,632), participants made trust-based decisions about people whose facial appearance was manipulated to be more or less trustworthy, following counter-stereotype exposure, debiasing instructions, or a control procedure. Exposure reliably reduced facial stereotyping, remaining effective across a diverse set of faces, under high cognitive demand, and independent of explicit beliefs about facial stereotyping. Instructions, by contrast, consistently failed and became counterproductive under high cognitive demands. This dissociation suggests facial stereotyping is best modified through experience that may reshape the underlying associations between facial features and trait concepts, and helps explain why facial stereotyping often resists relatively deliberative debiasing strategies.

Opioid addiction poses a major public health challenge that is exacerbated by pervasive stigma toward people who use opioids. Four experiments (Ntotal = 1,551 participants) investigated whether and how impressions of people who use opioids differ based on how their use originated. People whose opioid use began for recreation evoked more negative reactions (e.g., less trustworthiness, more desired social distance) than people whose use began for managing physical pain. Opioid use for coping with emotional pain elicited intermediate impressions, potentially suggesting moral ambiguity. Both positivity impressions and blameworthiness attributions mediated the effect of opioid use origin on stigma toward physical pain-related versus recreation-related opioid use. Moreover, blameworthiness attributions causally influenced desired social distance, even when holding opioid use origin constant. These findings indicate that opioid use stigma varies depending on the initial reason for a person’s use.

Forming impressions of other people requires relying on multiple social cues. Prominent theories have proposed a competitive dynamic between the use of social categories and other cues (e.g., individuating cues, emotion expressions) to inform social judgments, with greater reliance on social categories inherently diminishing reliance on other cues and vice versa. Because of methodological limitations, however, this key component of person perception theorizing has remained largely untested. To overcome these limitations, we implement a novel multinomial processing tree framework to disentangle the unique contributions of multiple pieces of information (e.g., social categories, emotion expressions). If person perception is inherently cue competitive, we should find consistently negative covariation between social category use and the use of other cues. However, across 14 datasets (N = 12,239 participants) comprising a diversity of judgments (e.g., race, trustworthiness), methods (e.g., Implicit Association Test, speeded classification), and social cues (e.g., gender, emotion expression), we observed no consistent negative covariation between social category use and the use of other cues. Findings were characterized by substantial heterogeneity across methods, judgments, and social cues, suggesting that whether social cues compete, cooperate, or are integrated orthogonally is context- and perceiver-sensitive. Overall, this research challenges the position that the use of social categories and other cues in person perception is inherently competitive.

One of psychology’s most ubiquitous findings is that of robust ingroup preference, an explicitly stated human preference for the groups to which we belong. The advent of implicit measures of social cognition introduced a paradoxical result: Compared to members of higher-status groups, members of lower-status groups often show no implicit ingroup preference, sometimes even showing outgroup preference, a result rarely observed on explicit measures. Prevailing theories have provided macro-level accounts, often in terms of system-justifying tendencies, but the various cognitive processes that could produce such an absence of ingroup preference among lower-status groups have not been directly decomposed. In this paper, we address this gap via formal process modeling applied to data from 5.7 million respondents collected across 16 years and 6 social categories. We find that, among lower-status groups, ingroup-affirming evaluations are accompanied by more pronounced self-regulatory control than outgroup-affirming evaluations, a regulatory asymmetry absent among higher-status groups. For example, evaluations affirming Black and overweight ingroups were accompanied by heightened self-regulatory control, whereas a comparable self-regulatory pattern was not observed for evaluations affirming White and non-overweight ingroups. The evidence points to a previously unrecognized role for self-regulatory control in the absence of ingroup preference, suggesting that the cognitive imprint of social hierarchies may extend beyond evaluative associations to the regulatory processes that govern their expression.

Mental representations of other people are informed by their intersecting social identities. In two experiments (N total = 1,592 participants), we tested whether ethnicity and binarized gender interact to shape how Middle Eastern (vs. ethnically unspecified) women and men are mentally represented. Results revealed that spontaneously generated character and facial appearance representations of Middle Eastern women differed from those of Middle Eastern men (i.e., more trustworthy and less dominant in character and facial appearance) and ethnically unspecified women (i.e., less trustworthy in character and facial appearance, less dominant in character but more dominant in facial appearance). Notably, these results differed somewhat from self-reported impressions, suggesting some, but not complete, correspondence between spontaneous and deliberate reactions to these groups. This work complements growing evidence that mental representations of ethnicity and gender constrain each other.

Racially biased weapon identification, wherein guns are identified more easily after seeing Black (vs. White) face primes, is a robust and replicable phenomenon. Mounting evidence suggests that introducing additional facial information (e.g., varying age cues) does not meaningfully alter this racial bias. Only when augmenting its relative salience does the additional, nonrace information appear to mitigate racially biased weapon identification. Even when reducing racial bias by enhancing nonrace facial cues, social information is typically communicated via the face, a context in which race may be particularly salient. Two experiments (N total = 590 participants) using a sequential priming task tested whether broadening the contextual information in primes to include both faces and bodies moderates racially biased weapon identification (gun vs. tool) decisions. Replicating past findings, racial bias was evident when primes cued age and race via facial information only. However, this behavioral effect disappeared when primes included both faces and bodies, providing richer social context. Diffusion decision modeling revealed that race cues shifted the starting point of the decision-making process toward stereotype-consistent responses (e.g., "gun" after Black primes) with face-only primes, but this processing bias disappeared with face-and-body primes. Multinomial processing tree modeling further revealed attenuated attention to race in face-and-body (vs. face-only) primes, whereas attention to age remained intact across conditions. These findings advance theory on the operation of racially biased decision making in richer social contexts.

Forming impressions of others is a fundamental aspect of social life. These impressions necessitate the integration of many and varied sources of information about other people, including social group memberships, apparent personality traits, inferences from observed behaviors, and so forth. However, methodological limitations have hampered progress in understanding this integration process. In particular, extant approaches have been unable to measure the independent contributions of multiple features to a given impression. In this article, after describing these limitations and their constraints on theory testing and development, we present a multinomial processing tree model as a computational solution to the problem. Specifically, the model distinguishes the contributions of multiple cues to social judgment. We describe an empirical demonstration of how applying the model can resolve long-standing debates among person perception researchers. Finally, we survey a variety of questions to which this approach can be profitably applied.

Racial stereotypes are commonly activated by informational cues that are detectable in people's faces. Here, we used a sequential priming task to examine whether and how the salience of emotion (angry/scowling vs. happy/smiling expressions) or apparent race (Black vs. White) information in male face primes shapes racially biased weapon identification (gun vs. tool) decisions. In two experiments (N total = 546) using two different manipulations of facial information salience, racial bias in weapon identification was weaker when the salience of emotion expression versus race was heightened. Using diffusion decision modeling, we tested competing accounts of the cognitive mechanism by which the salience of facial information moderates this behavioral effect. Consistent support emerged for an initial bias account, whereby the decision process began closer to the "gun" response upon seeing faces of Black versus White men, and this racially biased shift in the starting position was weaker when emotion versus race information was salient. We discuss these results vis-à-vis prior empirical and theoretical work on how facial information salience moderates racial bias in decision-making.

People sometimes prefer groups to which they do not belong (outgroups) over their own groups (ingroups). Many long-standing theoretical perspectives assume that this outgroup favorability bias primarily reflects negative ingroup evaluations rather than positive outgroup evaluations. To examine the contributions of negative ingroup versus positive outgroup evaluations to outgroup bias, we examined participants' data (total n > 879,000) from Implicit Association Tests [A. G. Greenwald, D. E. McGhee, J. L. K. Schwartz, J. Pers. Soc. Psychol. 74, 1464–1480 (1998)] measuring intergroup attitudes across four social domains in exploratory and preregistered confirmatory analyses. Process modeling [F. R. Conrey, J. W. Sherman, B. Gawronski, K. Hugenberg, C. J. Groom, J. Pers. Soc. Psychol. 89, 469–487 (2005)] was applied to the responses of participants who demonstrated implicit outgroup bias to separately estimate the contributions of negative ingroup and positive outgroup evaluations. The outgroup biases of lower-status group members (i.e., Asian, Black, gay and lesbian, and older people) consistently reflected greater contributions of positive outgroup evaluations than negative ingroup evaluations. In contrast, the outgroup biases of higher-status group members (i.e., White, straight, and younger people) reflected a more varied pattern of evaluations. We replicated this pattern of results using explicitly measured intergroup evaluations. Taking these data together, the present research demonstrates a positive–negative asymmetry effect of outgroup bias, primarily among members of lower-status groups.

Extending evidence for the rapid revision of mental representations of what other people are like, we explored whether people also rapidly revise their representations of what others look like. After learning to ascribe positive or negative behavioral information to a target person and generating a visualization of their face in a reverse-correlation task, participants learned new information that was (a) counter-attitudinal and diagnostic about the person's character or (b) neutral and non-diagnostic, and then they generated a second visualization. Ratings of these visualizations in separate samples of participants consistently revealed revision effects: Time 2 visualizations assimilated to the counter-attitudinal information. Weaker revision effects also emerged after learning neutral information, suggesting that the evaluative extremity of visualizations may dilute when encountering any additional information. These findings indicate that representations of others' appearance may change upon learning more about them, particularly when this new information is counter-attitudinal and diagnostic.

In this article, we describe four theoretical and methodological problems that have impeded implicit attitude research and the popular understanding of its findings. The problems all revolve around assumptions made about the relationships among measures (indirect vs. versus direct), constructs (implicit vs. explicit attitudes), cognitive processes (e.g., associative vs. propositional), and features of processing (automatic vs. controlled). These assumptions have confused our understandings of exactly what we are measuring, the processes that produce implicit evaluations, the meaning of differences in implicit evaluations across people and contexts, the meaning of changes in implicit evaluations in response to intervention, and how implicit evaluations predict behavior. We describe formal modeling as one means to address these problems, and provide illustrative examples. Clarifying these issues has important implications for our understanding of who has particular implicit evaluations and why, when those evaluations are likely to be particularly problematic, how we might best try to change them, and what interventions are best suited to minimize the effects of implicit evaluations on behavior.