RESEARCH DOSSIER
Person Perception and Impression Formation
Abstract. Person perception and impression formation are the processes through which observers combine appearance, speech, behavior, contextual knowledge, and reputation into representations of other people. This fundamental research monograph examines the formation and revision of person-specific judgments through historical configural theories, algebraic information integration, social-cognitive category accessibility, multidimensional social evaluation, rapid facial impressions, and observational thin-slice research. It evaluates experimental evidence on perceived warmth, competence, morality, halo effects, image variability, and the diagnosticity of conduct, while distinguishing inter-rater consensus from judgment accuracy against independent criteria. The synthesis incorporates realistic accuracy theory, componential models of interpersonal judgment, longitudinal person knowledge, experiments on implicit evaluation reversal, and neural research on updating. It examines measurement validity, correlated cues, cultural transportability, ethical implications for institutional decisions, and novel testable questions about human–artificial social evaluation. The evidence supports an account of impressions as context-dependent and revisable inferences whose social influence may exceed their accuracy. This work is a critical narrative synthesis of published studies and established theory; it reports no original experiment, systematic-search protocol, or newly computed meta-analysis.
Keywords: person perception; impression formation; first impressions; social cognition; trait attribution; configural impression formation; information integration; facial impressions; warmth; competence; moral character; halo effect; thin slices; interpersonal accuracy; social accuracy model; realistic accuracy model; diagnosticity; impression updating; person knowledge; appearance bias; cultural variation; digital reputation; human–artificial interaction; Artificial Era
Scientific Object and Explanatory Scope
Person perception is the organization of observable and remembered information into an inference about another person. Impression formation is the construction of a relatively integrated representation of that person's qualities, motives, likely conduct, and social meaning. A person impression can combine facial appearance, voice, linguistic choices, behavioral episodes, third-party testimony, roles, and interaction history. Its scientific status depends on how inputs are measured and what external criteria can test the resulting inference. A perceiver's certainty, a group's agreement, and an impression's influence on behavior are distinct empirical properties. Classic information-integration research Asch (1946) Anderson (1965) and contemporary componential measurement Biesanz (2010) provide different instruments for understanding this construction.
The present dossier investigates cue integration, the formation of first and continuing impressions, the diagnosticity of evidence, predictive accuracy, and the revision of person knowledge. Its central questions ask what cues produce judgments, how multiple cues become coherent wholes, why observers converge or disagree, under which conditions judgments predict behavior, and how a previously established evaluation changes. Group stereotypes matter where they supply priors for interpreting an individual, while their independent history and political consequences belong to adjacent intergroup research Fiske & Neuberg (1990) Macrae & Bodenhausen (2000). Causal explanations of action and responsibility are examined only insofar as they enter a person representation.
Research on person perception traverses psychology of perception, social cognition, personality measurement, social neuroscience, and relational interaction. That breadth requires distinctions between fast categorical evaluations and extended interpersonal knowledge, between stable characteristics and situational behavior, and between accuracy for one property and accuracy across all properties. The monograph follows a critical narrative evidence-integration method: original laboratory experiments are interpreted alongside naturalistic studies, formal measurement models, reviews, and work on stimulus variation Funder (1995) Todorov et al. (2015). It presents a synthesis of available research without claiming a systematic-search protocol, a new experiment, or a newly calculated pooled effect.
Constructs, Criteria, and Levels of Representation
A cue is an observable feature of a person's appearance, action, speech, or circumstances that can inform judgment. A trait attribution describes a hypothesized relatively enduring disposition; an intention attribution refers to a current goal; an evaluation attaches favorable or unfavorable significance; a prediction concerns expected action under specified conditions. These constructs operate at different temporal scales. A perceived smile can support a momentary inference about friendliness, yet one expressive episode supplies limited grounds for assuming enduring trustworthiness. The distinction between appearance-based judgment and validated disposition is central in the face-evaluation literature Oosterhof & Todorov (2008) Todorov et al. (2015).
Operational definitions determine whether competing findings address the same question. Impression consensus is agreement between raters. Impression accuracy is agreement with a defensible criterion, ideally one that is independent of the initial judgment. Perceptive accuracy concerns a particular observer's relative skill; expressive accuracy concerns a particular target's legibility. Normative accuracy reflects correct recognition of widely shared trait patterns, whereas distinctive accuracy reflects correct identification of how one individual differs from others Biesanz (2010). These components cannot be recovered by treating the average correlation between judgments and criterion scores as the entire structure of accuracy.
An impression is also a representation about an interaction, not only a property assigned to a target. The perceived reliability of a colleague depends on task demands and the observer's experience with the colleague; the appearance of confidence depends on setting and communication norms. The judgment therefore has at least a perceiver, a target, a cue sample, a context, a dimension, and a criterion. Cross-classified modeling of facial judgments provides empirical evidence that these components contribute unequally across different impressions Hehman et al. (2017). A scientific account states which unit was sampled and which source of variance its inference concerns.
Historical Foundations: Configural Judgment
Solomon Asch's 1946 experiments established a durable problem: the meaning of one attributed trait can change when the surrounding traits change. His account of impression formation emphasized organization and the centrality of certain descriptors within a whole-person representation Asch (1946). The historical experiments should be read as demonstrations of how different verbal information structures influence judgment in a controlled setting, rather than as a proof that every trait has one fixed psychological weight. Their significance lies in showing that the same descriptive element may be interpreted through the structure of the person portrait in which it appears.
Harold Kelley's 1950 warm–cold study further tested the importance of a preceding interpersonal description for subsequent impressions Kelley (1950). A brief initial framing could change how participants understood an instructor they subsequently encountered. As a historical intervention, the design connects expectation to perceived ongoing conduct and moves beyond a wholly fictional list of characteristics. It does not establish that observers cannot update expectations when new behavior is available; rather, it identifies an early channel through which descriptions can organize the interpretation of an actual encounter.
The central-trait tradition motivated enduring questions about configural meaning, primacy, and coherence. Its experimental vocabulary should be separated from contemporary trait psychometrics: ratings of 'warmth' in a short vignette need not represent the same construct as a population-scale personality inventory or a judgment of moral character. What persists is the hypothesis that social information is interpreted relationally, through the way one quality changes the significance of another. Later work on morality, warmth, and competence revisits this question with different materials and measurement structures Fiske et al. (2002) Goodwin et al. (2014).
Information Integration: Weighting, Averaging, and Coherence
Norman Anderson's information-integration program represented impressions as quantitative combinations of evaluative inputs. The 1965 comparison of averaging and adding rules asked how a person judgment changes as the number and value of descriptors change Anderson (1965). An additive model predicts accumulation of evidence; an averaging model predicts that a newly introduced moderate descriptor can reduce an extreme existing evaluation even while adding information. Weighted formulations permit cues to differ in importance. These are measurement models for specified judgment tasks and should not be confused with claims that subjective interpersonal perception literally performs one exact arithmetic operation in every setting.
Configural and algebraic findings can be complementary at different explanatory levels. The meaning assigned to a descriptor may depend on other descriptors, while a response scale may still exhibit orderly integration once those meanings are fixed. This allows investigators to ask separately whether an impression changes because cue interpretation changed, cue weight changed, or a response transformation changed. It also explains why one global rating cannot identify the unique cognitive computation that produced it. Comparing competing models requires multiple stimulus combinations and out-of-sample prediction, not simply an appealing fit to a small set of judgments.
The integration problem is further complicated by dependence among cues. Clothing, speaking style, expressed confidence, and institutional role may reflect a single underlying setting, so treating them as independent observations can exaggerate the amount of evidence. Impression strength should rise with informational diversity and reliability rather than the raw number of correlated descriptors. This is a methodological implication of cue-sampling and realistic-accuracy approaches Funder (1995) and of the evidence that seemingly diverse facial photographs may be highly variable observations of one person Jenkins et al. (2011).
From Trait Lists to Social Cognition
The social-cognition tradition redirected attention toward the mental categories accessible when an observer receives person information. Higgins, Rholes, and Jones showed that prior exposure to applicable trait words influenced later impressions of an ambiguously described individual; inapplicable words did not produce equivalent effects Higgins et al. (1977). This interaction is important because it links an experimentally manipulated cognitive state with the interpretation of a subsequent target. It supports conditional category accessibility, not a general rule that any prior positive or negative word automatically determines the next judgment.
Fiske and Neuberg articulated a continuum from relatively category-based impressions toward individuating processing as relevant information and motivational conditions change Fiske & Neuberg (1990). The model treats processing strategies as conditional on the perceiver's goals and the evidence available. An observer may use a first accessible category to organize expectations, notice attributes that challenge the category, and allocate more sustained attention when personal consequences warrant it. Such processes are scientifically distinguishable from both stable discriminatory attitudes and the objectively correct assessment of an individual.
Macrae and Bodenhausen summarized evidence for activation, use, control, and inhibition of social categories Macrae & Bodenhausen (2000). Their review places person perception in an architecture of attention and memory: categories can accelerate information processing and can direct which behavior is encoded as meaningful. This conceptual link does not make every particular individual impression reducible to intergroup stereotypes. It identifies a mechanism by which broad representations and target-specific observations interact. The distinctive outcome of this dossier remains the resulting person-specific judgment and its evidential warrant.
The Research Tradition and its Methodological Turning Points
The field's experimental history can be read as a series of changes in the evidence available to the perceiver. Trait-list experiments manipulated verbal descriptors; classroom and encounter experiments introduced visible actors; social-cognitive paradigms manipulated accessible categories and processing resources; thin-slice studies sampled seconds of recorded behavior; computational work varied facial appearance systematically; modern multilevel designs separated target, perceiver, and image variance Asch (1946) Ambady & Rosenthal (1993) Oosterhof & Todorov (2008) Hehman et al. (2017). Each advance introduced new observational opportunities alongside new confounds.
Apparent contradictions often reflect incompatible tasks. Agreement after viewing the same standardized portrait measures a different quantity from agreement after viewing different photographs of one individual. Accurately predicting teacher evaluations is different from accurately estimating enduring kindness. Updating a newly learned fictional target is different from revising impressions of a long-known family member. These distinctions determine which theoretical claims a result can support. The relevant methodology progresses by explicitly aligning target populations, cue exposure, response dimensions, and criteria rather than announcing one universal ability to 'read people' Ambady & Rosenthal (1992) Todorov & Porter (2014) Mann & Ferguson (2015).
The development of person perception research also reveals the need to revisit foundational experiments under contemporary sampling and open-science expectations. Early influential tasks often used small, convenient samples, restricted written descriptions, and locally familiar language. Later tasks sampled additional modalities but frequently retained artificial rating scales and narrowly drawn facial datasets. Historical importance and current transportability must therefore be evaluated separately. An enduring experimental result can reveal a cognitive mechanism under its manipulation while leaving its population distribution, effect size, and long-term consequences uncertain.
Rapid Face-Based Impressions and their Temporal Limits
First impressions from facial appearance can be generated remarkably quickly. Willis and Todorov manipulated viewing time for unfamiliar faces and reported that judgments after 100 milliseconds were strongly aligned with judgments made without a time limit Willis & Todorov (2006). The finding establishes rapid availability of evaluative responses and their relative stability across exposure durations in that paradigm. It does not establish that a hundred milliseconds can reveal a stranger's actual honesty, ability, or moral character. A research account must distinguish rapid generation, inter-rater consistency, subjective confidence, and criterion-based predictive validity.
The same experiments found changes in speed, confidence, and differentiation as exposure increased Willis & Todorov (2006). This matters because additional time can strengthen commitment to an initial interpretation without proportionately improving accuracy. Repeated sampling of the same visually salient cues may produce a more detailed report while leaving the original inference's reliability unchanged. In a real encounter, additional seconds may provide new behavioral information; in a static-face task they often provide only more inspection of the original photograph. Exposure duration therefore interacts with the type of information acquired.
Rapid face judgments are especially consequential in selection and trust decisions because the visual stimulus can precede all relevant behavioral evidence. Reviews have documented associations between face-based inferences and consequential social choices Olivola et al. (2014) Todorov et al. (2015). That such choices can be influenced by an impression demonstrates behavioral power; it does not verify the impression's target-level truth. The appropriate scientific question is how a fast and potentially uncertain evaluative response is subsequently weighted against independently validated information.
Dimensions of Facial Evaluation
Oosterhof and Todorov analyzed judgments of emotionally neutral faces and identified a useful two-dimensional structure of facial evaluation involving valence-related trustworthiness and dominance Oosterhof & Todorov (2008). Data-driven models connected these evaluation axes with visual configurations that resemble affective expressions or apparent physical strength. The study explains aspects of how faces come to look approachable or threatening to observers. It does not convert an appearance-derived coordinate into a clinical measurement of personality or a stable latent moral disposition.
A dimensional model summarizes patterned covariance across judgments collected with particular stimuli and observers. It can clarify why ratings of competence, aggression, dominance, and trustworthiness cluster in particular ways. However, the axes belong to representations generated by the experimental procedure. An individual's face also changes with expression, lighting, pose, health, camera angle, and image selection. The inferred dimension can thus be a reliable description of an image's social appearance even where it is an unreliable description of the depicted person Jenkins et al. (2011) Todorov & Porter (2014).
Later critical reviews emphasized that cue-driven visual evaluations have substantial downstream consequences while their diagnostic validity for complex personal traits has often been overstated Todorov et al. (2015). A sound interpretation of a face-trustworthiness scale is therefore 'how trustworthy this face appears to this class of observers under these conditions.' Reinterpreting the same scale as 'how trustworthy this person is' silently changes the construct. Preserving that distinction is essential when translating face evaluation into hiring, safety, justice, or machine-assisted judgments.
Identity, Photographs, and Within-Person Variation
One photograph is a sample from a person's changing appearance, not an exhaustive representation of that person. Jenkins, White, Van Montfort, and Burton used naturally varying images to show how within-person visual variation can complicate judgments and recognition Jenkins et al. (2011). The methodological implication extends beyond face identification: if images are unrepresentative of stable appearance, social trait ratings obtained from those images may also exaggerate between-person differences. Researchers should sample multiple images per identity before attributing variation to stable target-level properties.
Todorov and Porter demonstrated directly that different images of the same person can produce markedly different social impressions; the within-person image variance was comparable to or greater than between-person variance for multiple judgment dimensions Todorov & Porter (2014). Context-sensitive image selection altered which pictures were preferred for different social purposes. These results undermine an inference sometimes embedded in single-portrait studies: that one photograph is a transparent readout of a person's physiognomic social meaning.
The scientific design follows from this observation. Crossed samples of perceivers, persons, photographs, expressions, and contexts allow image-specific, person-specific, and observer-specific variance to be estimated separately. Without such crossing, an apparently robust face-to-trait association can be an artifact of one image, one photographer, one platform, or one selection rule. Image diversity is not merely a cosmetic issue of external validity; it is part of identifying the proper target of explanation Hehman et al. (2017).
Expressive Behavior, Thin Slices, and Ecological Criteria
Thin-slice research asks what can be predicted from brief samples of expressive behavior. Ambady and Rosenthal's 1992 meta-analysis assembled results relating observations lasting less than five minutes to a range of interpersonal outcomes Ambady & Rosenthal (1992). The synthesis supported nontrivial prediction under specified conditions and showed that longer observations did not always improve correlations in the collected studies. This finding concerns predictive relationships across multiple observational tasks; its pooled result cannot be treated as a universal probability that a given stranger's personality will be correctly inferred.
In a classroom application, observers' ratings derived from very brief silent video samples of teachers predicted later evaluations made by students or school administrators Ambady & Rosenthal (1993). Such a criterion is ecologically meaningful because it arises outside the original thin-slice session. Yet evaluations are themselves social judgments, potentially shaped by warmth, presentation style, attractiveness, and expectations. Predicting one impression from another should be distinguished from detecting a teacher's actual instructional effectiveness or a stable disposition. The study motivates a stronger question about which target properties are causally represented by the sampled conduct.
An extensive research program has therefore moved from asking whether brief observation predicts any outcome to asking which outcomes, which modalities, and which populations support valid inference. Nonverbal behavior can convey interactional competence, affective state, or social engagement, while leaving hidden motives and conduct in other situations less accessible. Prediction must be evaluated separately for each criterion and task. Thin slices are evidence of available social information, with an information limit established by the cue's diagnostic relation to the desired property Funder (1995).
Appearance, Embodiment, and First-Encounter Behavior
Person perception includes more than a face. Whole-body appearance, posture, dress, expression, gait, and situational behavior can communicate different cues. Naumann and colleagues compared judgments from constrained and spontaneous full-body photographs against personality criteria incorporating reports from people who knew the targets Naumann et al. (2009). Accuracy depended on trait and photographic condition; naturalistic posture and expression carried additional information. Their results cannot be generalized to a claim that all appearance-based trait judgments are accurate, because cue availability and diagnostic value vary with the dimension under examination.
Back, Schmukle, and Egloff examined why people with elevated narcissistic traits could be positively received at zero acquaintance Back et al. (2010). Their work shows how first-meeting popularity can emerge from observable presentation cues while carrying different implications for longer-run relationships. Attractiveness in a first encounter, immediate confidence, and durable interpersonal reliability can diverge. This study is especially instructive because the criterion is an initial peer response, whereas the personality measure concerns another construct; a positive impression can itself be an outcome requiring explanation.
Context affects the informativeness of appearance. Spontaneous expressions in a relaxed encounter differ from performances at an interview, scripted introduction, portrait session, or online profile. Observers frequently infer stable qualities from behavior staged under powerful situational constraints. Valid research must specify whether the target's cue was freely produced, strategically managed, assigned by an experimenter, or imposed by a platform. The causal chain from disposition to visible cue is part of the scientific question, not a background assumption Gilbert et al. (1988) Naumann et al. (2009).
Warmth, Competence, and Morality as Judgment Dimensions
The content of an interpersonal impression is structured by the questions it is expected to answer. Perceived warmth can concern anticipated intentions toward others, competence can concern capacity to act, and morality can concern character expressed in honesty, fairness, or responsibility. Fiske, Cuddy, Glick, and Xu developed a warmth–competence framework in research on group stereotypes Fiske et al. (2002). Its relevance to individual impressions lies in the generality of these evaluative dimensions, while a group's stereotype distribution must not be silently treated as a measurement of a particular person's character.
Goodwin, Piazza, and Rozin argued, through empirical person-evaluation research, that moral character can have special importance for the overall assessment of another person Goodwin et al. (2014). This finding refines accounts that subsume morality under broad sociability or warmth: perceived benevolence, honesty, or integrity may bear differently on trust than pleasantness or social fluency. The relative influence of any dimension still depends on the task. Selecting a collaborator for a technical assignment, evaluating a friend after a betrayal, and making an immediate approach decision can place different demands on competence, communion, and morality.
Dimensions should be treated as operationally specific variables, not timeless labels with guaranteed common meanings. A face judged 'dominant' in one task, a colleague judged 'competent' in another, and a friend judged 'morally good' after extensive interaction are assessed through different cue channels and criteria. Clear conceptual separation permits researchers to ask when an impression dimension predicts later cooperation or avoidance and when it merely reproduces the wording of the prompt. It also prevents the empirical success of one warmth–competence model from being interpreted as a complete taxonomy of personal knowledge.
Halo, Valence, and the Dependence of Ratings
A halo effect occurs when an overall evaluation influences ratings of attributes that could in principle be assessed separately. Nisbett and Wilson manipulated a videotaped instructor's warm or cold interpersonal behavior and found corresponding differences in ratings of qualities such as appearance, mannerisms, and accent Nisbett & Wilson (1977). The result illustrates how a global interpretation may reorganize the perceived value of component cues. The study's evidence rests on its manipulated video contrast and ratings; it does not establish that all favorable appearances generate the same magnitude of generalized evaluation.
Evaluation measures are often mutually dependent by construction. A perceiver who decides that a speaker is likable may subsequently rate competence, sincerity, and intelligence higher, even when the viewed behavior provided little direct evidence about those properties. Agreement among such ratings may therefore reveal shared affective appraisal or response style rather than several independent discoveries about the target. Researchers can test this by counterbalancing rating order, separating judgments temporally, and providing objective behavioral outcomes that are not derived from the same observer.
Valence also shapes the interpretation of ambiguous acts. An initially admired colleague's delay may be read as carefulness; the same delay may be taken as negligence when the colleague is disliked. These are candidate mechanisms consistent with configural and halo findings Asch (1946) Nisbett & Wilson (1977). The precise claim must remain tied to manipulations or longitudinal designs that can discriminate selective attention, reinterpretation, memory, and response bias. One rating profile alone cannot tell which mechanism occurred.
Spontaneous Trait Inference and Semantic Accessibility
Behavior descriptions can evoke trait concepts without a deliberate instruction to judge character. Work on accessibility demonstrated that previously activated, applicable categories could color interpretations of ambiguous target descriptions Higgins et al. (1977). This mechanism helps explain why the same behavior may enter person memory with different meanings depending on linguistic context and prior exposure. It also shows why the mere presence of evaluative vocabulary is insufficient: a category's application to a concrete target depends on its perceived relevance to the observed conduct.
Spontaneous person-knowledge formation requires distinguishing an inference about an actor from a learned association between an actor and a word. In behavioral memory experiments, savings in relearning trait associations may arise even when the stimulus actor was incidentally linked to a behavioral description. This finding warns against equating every indirect-memory index with a conscious and correctly localized inference about a person's enduring quality. A rigorous design distinguishes remembering the behavior, associating a concept with a face, and attributing the behavior to the actor who actually performed it.
Language brings further complications through conversational pragmatics. Evaluative terms are not merely measurements of observed conduct; they can be recommendations, accusations, jokes, or descriptions shaped by social goals. The same sentence about a person may supply evidence about the target, the speaker, or the relationship between them. For source credibility, researchers should vary the informant's knowledge, incentive, and accuracy history, then test whether the observer weights testimony appropriately. The need for such controls follows from the study of accessible interpretations and context-dependent evaluation Higgins et al. (1977) Fiske & Neuberg (1990).
Behavioral Attribution and the Problem of Situation
Observers often convert a visible action into an apparent character attribute. Gilbert, Pelham, and Krull proposed a processing distinction among identifying an action, characterizing the actor, and correcting an initial characterization for situational constraints Gilbert et al. (1988). In their experiments, concurrent cognitive demands reduced appropriate use of information about situational circumstances. These results identify a vulnerability of correction under the studied conditions and illuminate why a pressured observer might interpret constrained public conduct as a voluntary expression of personality.
Yet perceiving behavior is more than tracing a trait to an action. People act under institutional roles, incentives, social surveillance, fatigue, audience expectations, and strategic communication demands. A public interview may show how someone performs under interview conditions without revealing how that person behaves when unobserved. The scientifically informative contrast uses repeated situations, manipulates constraints, and compares person-specific behavior across contexts. A behavioral cue becomes more diagnostic when alternative situational explanations are constrained or explicitly measured.
Person perception studies and the distinct research object of causal attribution intersect at precisely this point. Impression formation asks how evidence changes a person representation; attribution asks which cause produced a behavior and how responsibility should be assigned. The same observed event can update a trait inference only if the observer takes the actor to be responsible for it. Research on the revision of implicit evaluation demonstrates how moral diagnosticity and personal responsibility shape which information receives decisive weight Cone & Ferguson (2015).
Interpersonal Consensus, Perceiver Contributions, and Individual Variation
Social agreement is a phenomenon to be explained, not a substitute for accuracy. Albright, Kenny, and Malloy studied personality judgment at zero acquaintance and demonstrated consensus under specified observational conditions Albright et al. (1988). Agreement can emerge from shared perceptual sensitivities, shared cultural expectations, or common response conventions. Each possibility makes different predictions about whether independent behavioral criteria will confirm the judgments. Agreement between observers is most informative when its source and target trait are identified.
Hehman, Sutherland, Flake, and Slepian fitted cross-classified models to a large collection of facial ratings and found that perceivers, targets, and their interactions contributed differently across judgment dimensions Hehman et al. (2017). A perceiver may assign similar trustworthiness ratings to many faces, another may be selectively sensitive to particular images, and targets may evoke relatively consistent responses for some visual dimensions. These components should be modeled rather than averaged away. In particular, the aggregate face rating may conceal substantial disagreement about the same apparently neutral stimulus.
Individual differences in person perception can include familiarity, motives, task goals, cultural learning, and criteria for applying evaluative words. Interactions between observer and target are not merely statistical noise. They may reflect genuine differences in what counts as trustworthy, threatening, helpful, or competent for particular relationships. The next methodological step is to relate these differences to independently observable consequences. Without that step, componential variance estimates describe the structure of judgments but cannot by themselves establish which observer is more accurate Biesanz (2010) Hehman et al. (2017).
Cultural Semantics and the Limits of Transportability
People interpret expressions and interpersonal signals within cultural practices of communication. Elfenbein and Ambady's meta-analysis of emotional-expression recognition found both above-chance cross-cultural recognition and an in-group advantage in recognition accuracy Elfenbein & Ambady (2002). The result concerns recognizing expressions under the tasks studied. It provides a methodological warning for broader person perception: emotional expressions cannot simply be translated into universal personality traits without examining shared contexts, normative display practices, and the meanings assigned to observed conduct.
A person-description scale developed in one language or population can change meaning after translation or when applied to other occupational and social contexts. The words 'warm,' 'confident,' 'modest,' and 'assertive' may refer to different behavioral indicators, carry different desirability, and vary in conversational use. Cross-cultural claims consequently require measurement-invariance checks, locally appropriate criterion measures, and evidence beyond a restricted set of face images. Importantly, the definition of accuracy must not rely solely on the evaluative norms of the sample doing the judging.
Culture can organize interactional expectations without determining every individual's interpretation. The evidence permits investigations of variation across environments and experience, not a fixed portrait of how an entire population perceives people. Longitudinal contact and repeated shared tasks can change which cues observers learn to trust. Researchers should record recruitment, local stimulus distributions, familiarity between observers and targets, and opportunities for correction. These design requirements directly follow from the observed contextuality of interpersonal judgment and expression recognition Elfenbein & Ambady (2002) Todorov et al. (2015).
Accuracy as an Empirical Problem
The scientific problem of impression accuracy begins with the criterion. David Funder's realistic accuracy model organizes successful personality judgment around the availability of relevant cues, their accessibility to the observer, their detection, and their appropriate utilization Funder (1995). A person may have a characteristic that rarely becomes visible in the sampled setting; another may reveal a useful cue that an observer fails to notice; a third may be judged through an irrelevant cue. These alternatives imply different explanations and different interventions. Accuracy therefore depends on an evidence-generating encounter, not solely on the intuitive talent of the person doing the judging.
Accuracy also changes with the level of analysis. A judgment can correctly classify a broad tendency while mispredicting a particular act, or accurately predict a single observed outcome while misunderstanding the person's motives. Self-report, peer report, direct behavioral observation, and administrative records offer overlapping but imperfect criteria. Ideally, several independent indicators are integrated while their limitations are reported. A comparison against a target's self-description alone can be contaminated by self-presentation, memory, or differences in scale interpretation Biesanz (2010) Naumann et al. (2009).
Research must also distinguish interpersonal prediction from trait attribution. A portrait that reliably predicts a popularity judgment tells us that viewers share a visual standard; it cannot alone establish that the portrait predicts long-term interpersonal competence. An apparently accurate forecast of a teacher's student ratings may predict the qualities that those students reward rather than instructional learning Ambady & Rosenthal (1993). This analytic discipline preserves the value of thin-slice and first-impression evidence without inflating its inferential reach.
Components of Interpersonal Accuracy
Biesanz's social accuracy model decomposes interpersonal judgment into normative and distinctive accuracy, with perceiver and target components Biesanz (2010). Normative accuracy refers to judgments that track a generally shared and criterion-supported profile of traits. Distinctive accuracy concerns whether an observer detects a target's deviations from that profile. The distinction is critical because a perceiver can seem highly accurate by assigning most people plausible average characteristics while failing to perceive what makes a particular individual unusual.
Design determines what these components can mean. Round-robin interactions, in which several participants judge one another, can help disentangle who tends to judge accurately from who tends to be legible. Multiple judges evaluating multiple targets with independent trait criteria allow cross-classified effects to be estimated. If each observer rates only one person, familiar statistical correlations can conflate a perceptive observer's skill with an especially readable target. Multi-trait evaluation adds another layer: some dimensions are more visible and agreed upon than others Albright et al. (1988) Biesanz (2010).
A componential model makes it possible to quantify uncertainty rather than claiming a global ability. Yet its interpretation still depends on how the target traits were measured and which kinds of interaction were sampled. A formal accuracy score is conditional on available cues, a criterion instrument, a social relationship, and the population of possible targets. It does not represent a context-free measure of human nature. Comparison across studies should therefore specify whether results concern consensus, normative accuracy, distinctive accuracy, interpersonal sensitivity, or later behavior.
Why Consensus May Outrun Validity
People can agree on a face's apparent trustworthiness because they share conventions about the visual meaning of facial configurations. Such consensus is compatible with weak or absent evidence that the face predicts honest behavior. The face-attribution review by Todorov and colleagues explicitly cautions that the predictive validity of appearance-based social inferences has been overstated Todorov et al. (2015). A clear report separates agreement among raters, model prediction of ratings, prediction of actual conduct, and any causal role of the appearance in shaping how a target is subsequently treated.
Image selection supplies a particularly strong counterexample to assuming stable target characteristics. Different photographs of the same person can provoke different impressions Todorov & Porter (2014). Furthermore, an image-based rating system may reproduce evaluative preferences embedded in the collection procedure rather than discover a stable human disposition. When researchers optimize models to fit average ratings, successful prediction can show that the model has learned rating conventions. This is a worthwhile cognitive finding, but the target variable remains perceived character, not independently established character.
The social consequence can nevertheless be substantial. Appearance-based impressions influence choices, as reviewed by Olivola, Funk, and Todorov Olivola et al. (2014). A false appearance-derived belief may change trust, attention, or opportunities, thereby altering the very behavior later observed. Accuracy studies should distinguish prediction of preexisting qualities from outcomes partially produced by perceivers' prior judgments. This creates a need for temporal designs and interventions that separate cue effects from social feedback effects.
Diagnosticity and Asymmetric Updating
A first impression is a revisable representation. Evidence does not receive equal weight merely because it is recent or salient. Mende-Siedlecki, Baron, and Todorov studied asymmetric updating in morality and ability domains and showed the importance of perceived diagnostic value for impression revision Mende-Siedlecki et al. (2013a). A single episode may be more informative in one domain than another because different behaviors have different implications for stable qualities. These patterns can be interpreted as context-dependent judgments about how likely behavior would be under competing hypotheses about a person's character.
Diagnosticity requires an explicit comparison. A cooperative act may be common under strong social incentives, while a costly and freely chosen act may contain more information about commitment. A displayed ability under highly favorable conditions differs from ability demonstrated under challenge. These examples are analytic implications rather than reported experimental effect sizes. The empirical question is whether the relevant observers correctly incorporate the contingencies that make an act more or less diagnostic, or instead treat any emotionally salient event as decisive Cone & Ferguson (2015).
Morality and competence also support different updating asymmetries. Evidence of a serious wrongdoing may challenge a favorable moral impression even when many ordinary benign behaviors preceded it; evidence of high competence may be less decisive if a task admits strong situational assistance. Such inferences depend on base rates, the perceived agency of the actor, and reliability of the report. Science requires measuring these contingencies rather than universally labeling negative evidence stronger. The key finding is the structured weighting of new information, not the immutability of positive or negative reputations Mende-Siedlecki et al. (2013a).
Implicit Evaluation and Reinterpretation
Impression change concerns both reportable beliefs and indirect evaluations measured through tasks that do not require an explicit judgment of the target's character. Cone and Ferguson found in six experiments that newly learned implicit evaluations could be reversed by highly diagnostic contradictory information Cone & Ferguson (2015). Their results oppose a simple assumption that indirect evaluations are always impervious to a single subsequent revelation. The experiments concern defined learning procedures and measures of evaluative response; they do not demonstrate that every lifelong prejudice or every emotionally charged relationship can be reset in the same way.
Mann and Ferguson investigated another route to revision: new information that changes the meaning of earlier evidence Mann & Ferguson (2015). Rather than merely accumulating a competing episode, reinterpretation can reorganize previously learned events into a different causal story. Their experiments documented reversals in newly learned implicit person evaluations when participants actively reconsidered earlier material. The mechanism implies that a person impression can change substantially even without forgetting the evidence on which the earlier impression was based.
These studies clarify why counting positive and negative anecdotes is an insufficient model of revision. New evidence may be credible, surprising, contradictory, identity-relevant, causally explanatory, or revealing of the actor's degree of responsibility. A credible reinterpretation can transform the status of many prior observations simultaneously. Such revision is neither automatic in every setting nor reducible to a fixed number of exposures. The durable research question concerns what information changes perceived diagnosticity and the explanatory structure linking actions to the person Cone & Ferguson (2015) Mann & Ferguson (2015).
Neural Studies of Impression Updating
Social neuroscience investigates neural responses during person evaluation and updating. Mende-Siedlecki, Cai, and Todorov examined the neural dynamics of revising person impressions Mende-Siedlecki et al. (2013b). Mende-Siedlecki, Baron, and Todorov later studied how diagnosticity contributes to asymmetric updating in morality and ability judgments Mende-Siedlecki et al. (2013a). These experiments connect psychological models of evaluation with measurements of brain activity while maintaining the need to identify which behavioral judgments the neural signals track.
A neural response associated with surprising information can be consistent with prediction error, attention, salience, motivational relevance, or reinterpretation, depending on design and analysis. Correlations between activity in a named region and a behavioral update cannot independently establish that the region computes a uniquely defined social trait or that its activity causes the new judgment. Robust interpretation depends on task contrasts, computational alternatives, measurement reliability, and independent behavioral validation.
Neural evidence thus adds a level of process measurement rather than replacing person-perception theory. The strongest research programs connect an explicit hypothesis about cue integration or diagnostic value, a manipulated input, a behavioral updating result, and a neural signal that supplies incremental explanatory information. Longitudinal interactions and repeated measures can then ask whether the identified processes generalize to real personal knowledge. The gap between constrained scanner tasks and natural social life remains an important scientific limit.
Longitudinal Relationships and the Accumulation of Person Knowledge
First impressions become part of a relationship when observers repeatedly interact with a target, encounter the consequences of earlier predictions, and share contextual information. Repeated encounters create opportunities to sample conduct across circumstances, but also opportunities for selection and confirmation. An observer may preferentially seek evidence consistent with a favorable interpretation or react differently to behavior already framed as threatening. The scientific question is whether a growing impression approaches a more accurate representation of context-sensitive behavior and stable tendencies, or becomes a more elaborate justification for its earliest assumptions Funder (1995) Fiske & Neuberg (1990).
Longitudinal research can separate memory from prediction by eliciting an initial judgment, independently recording behavior across situations, and measuring revision after new episodes. The target's opportunities, roles, and incentives should be recorded because behavior is co-produced by persons and settings. A global reputation score may remain stable while different components change, such as perceptions of competence, benevolence, reliability, and conversational fit. Measures of change need to distinguish revisions of what a person believes from revisions of how important a belief is for a particular relationship Goodwin et al. (2014) Mende-Siedlecki et al. (2013a).
Relationships also introduce reciprocal causation. Someone treated as untrustworthy may become guarded; someone supported as capable may attempt more difficult work. If the initial impression changes the conditions of later encounters, observed behavioral differences cannot automatically be attributed to preexisting personality. Naturalistic research should measure interaction histories and opportunities for feedback, recognizing the epistemic difficulty of holding social treatment constant while allowing relationships to develop. Such designs bring person perception into contact with developmental, organizational, and relational psychology.
Consequential Judgments in Institutions
First impressions matter in employment, education, professional assessment, and interpersonal gatekeeping because early evaluations can structure later access to resources. Evidence that facial judgments influence social choices is documented in reviews of decision-making Olivola et al. (2014) Todorov et al. (2015). A decision's dependence on appearance is a causal possibility requiring direct examination, especially where decision-makers have access to more predictive evidence such as demonstrated work, standardized tasks, or a documented history of conduct. The relevant standard is not whether the initial impression feels coherent; it is whether the selected cue improves valid prediction for the decision being made.
Teacher-rating research demonstrates the distinction between predicting another person's evaluations and predicting an independently defined performance outcome Ambady & Rosenthal (1993). Employment ratings pose similar problems when colleagues' assessments share visual expectations or institutional norms. Studies of selection should include objective criteria where feasible, compare alternative evidence channels, assess differential error across groups, and report whether additional diagnostic information changes the effect. When a score is produced by the same people influenced by the impression, its apparent predictive validity may contain circularity.
An institutional safeguard follows from this evidence: high-stakes decisions benefit from role-relevant, independently assessed information and opportunities to correct unsupported assumptions. Such a safeguard is an applied inference from accuracy and face-bias research, not a claim that one single standardized procedure eliminates all social bias. Researchers must test its effect against concrete outcomes, including opportunity, fairness, error, and unintended costs. Person perception is therefore a substantive area of organizational and educational evaluation as well as laboratory social cognition.
Digital Identity, Profiles, and Reputation
Digital environments expose observers to selected photographs, headlines, short video clips, usernames, and algorithmically ordered feedback. These are not neutral samples of a target. Profile owners may curate favorable evidence, platforms may favor vivid or emotionally expressive material, and a third party's comment can reshape the perceived meaning of an otherwise ordinary act. Experiments on image choice demonstrate that changing the photograph of the same individual can shift the resulting impression Todorov & Porter (2014). The same logic makes profile-based evaluation sensitive to which portion of a person's conduct a platform displays.
Computer-generated rankings of apparent trustworthiness or competence can reproduce consensual ratings from a dataset without verifying the person's underlying quality. The relevant target variable in such a model is often perceived appearance under particular recording and labeling conventions. A valid assessment of real-world trust would require independent behavioral criteria, representative sampling, tests of robustness across images and observers, and monitoring of downstream effects. High inter-rater reliability of a digital appearance score does not solve this validation problem Hehman et al. (2017) Todorov et al. (2015).
Online reputation further involves source reliability. A publicly repeated allegation, a context-free quotation, or a coordinated rating campaign may influence impressions without supplying independent evidence. Investigations should vary the credibility of claims, the availability of corrections, the temporal ordering of disclosures, and the target's opportunity to respond. Researchers should examine how corrections propagate through the same communication environment that produced the initial interpretation. This extends established experimental problems of accessibility, information weighting, and reinterpretation into a contemporary media setting Higgins et al. (1977) Mann & Ferguson (2015).
Person Perception in Human–Artificial Interaction
In the Artificial Era, an artificial interlocutor may present linguistic fluency, apparent social responsiveness, stable naming, personalized memory, and a consistent interface identity. Human observers can respond to these signals by forming impressions of a system's apparent warmth, competence, reliability, intentions, or continuity. These impressions are psychological facts about the perceiver and encounter. Their relation to the architecture and capabilities of an artificial system requires separate operational testing. The research traditions on trait integration and cue diagnosticity provide conceptual tools for that distinction Asch (1946) Funder (1995).
A fluent response is an observable output and may be informative about linguistic capability in a given task; it is not an independent demonstration of every property an observer may attribute to the system. Likewise, a friendly style may influence approach and trust without establishing durable reliability across adversarial, unfamiliar, or high-stakes conditions. This is an explicitly interdisciplinary extrapolation from human person-perception research, not an experimental conclusion automatically applicable to all artificial architectures. It generates testable hypotheses about how voice, interface embodiment, reputation, error history, and continuity influence calibration of human judgments.
A responsible experimental design would independently manipulate an artificial partner's response accuracy, visible confidence, conversational style, and disclosure of limitations. It would then measure users' impressions, their predictions of future performance, actual performance across held-out tasks, and willingness to revise beliefs after discrepant outputs. Repeated interaction allows study of impression persistence and correction; cross-platform comparison can reveal interface effects. The philosophical program of Artificial Era supplies questions about emerging nonbiological subjects, while empirical person perception specifies what observable measurements are needed to investigate judgments made by humans about those systems.
Experimental, Psychometric, and Computational Methods
Traditional trait-list studies offer precise control over descriptor sequence and co-occurrence; laboratory encounters add behavior at the cost of greater variability; thin-slice studies provide ecologically recognizable stimulus streams; longitudinal field designs measure experience over time. Each method targets a different segment of the perception process. To estimate causal influence, researchers randomize cue presentation while holding competing cues constant. To estimate out-of-sample accuracy, they compare impressions with independent criteria. To study naturally occurring reputations, they record repeated observations and information provenance Asch (1946) Ambady & Rosenthal (1992) Funder (1995).
Multilevel designs are especially important because each rating is a relation among a perceiver, target, stimulus sample, and judgment dimension. Hehman and colleagues demonstrate how crossed random-effects models quantify the distinct contributions of these sources Hehman et al. (2017). A model that ignores multiple photographs per target or repeated judgments by one observer can exaggerate precision through pseudoreplication. Measures should test internal consistency only where the construct warrants it and should examine whether a score's interpretation survives changes in task instructions, populations, and modalities.
Computational models can formalize alternative hypotheses: weighted cue integration, category-based inference, dimensional mapping of faces, and diagnosticity-sensitive learning. Their evaluation requires separating model fit to observed ratings from predictions about independently observed traits or actions. Training and testing on different images of the same individuals can leak identity-specific visual information; testing on held-out individuals, images, and contexts gives a stronger account of generalization. The statistical sophistication of a model never substitutes for the validity of its criterion Oosterhof & Todorov (2008) Todorov & Porter (2014) Biesanz (2010).
Evidence Quality, Reproducibility, and Open Questions
The evidence base is heterogeneous in strength. The Asch and Anderson paradigms offer experimental control over verbal information; face-rating studies often provide strong evidence about consistency and representational geometry; thin-slice research links expressed behavior to observational outcomes; correction studies show that learning can revise evaluation. The inferential boundary of each method differs. A standardized facial photograph is strong for investigating appearances and weak for inferring unseen ethical conduct. A field association is useful for prediction but may leave causal pathways open Anderson (1965) Ambady & Rosenthal (1992) Todorov et al. (2015).
Three recurring risks deserve explicit attention: restricted populations, correlated observations, and outcome contamination. Convenience samples may not represent unfamiliar cultures or occupational settings. Multiple images and ratings from the same identities require hierarchical treatment. Socially produced criterion scores may reproduce the very bias under investigation. Other risks include publication bias, flexible analytic choices, selective image presentation, measurement noninvariance, and unequal exposure to relevant cues. Stronger evidence accumulates through preregistered hypotheses, openly documented materials, independent replications, and realistic holdout tests, with protection of participant privacy.
Open questions include how impressions generalize between online and in-person settings, which classes of behavior are genuinely diagnostic of long-term characteristics, and how an observer's history shapes cue weighting. Further research should measure calibration, uncertainty disclosure, the speed and durability of correction, interactional feedback, cultural transfer, and the consequences of algorithmic amplification. The field especially needs designs that can distinguish a shared aesthetic reaction, a valid behavioral prediction, and a socially produced outcome. These are research priorities derived from the documented methodological contrasts Jenkins et al. (2011) Hehman et al. (2017) Mann & Ferguson (2015).
Conclusion
Person perception and impression formation constitute a systematic process of transforming social cues into person-specific representations, evaluations, and expectations. Its foundational empirical traditions reveal both the organization of trait information and the dependence of interpretation on accessible categories, context, and goals Asch (1946) Anderson (1965) Fiske & Neuberg (1990). Research on rapid face evaluations demonstrates how quickly social meaning can be attached to appearance, whereas work on photograph variability shows why such meaning cannot automatically be equated with stable target properties Willis & Todorov (2006) Todorov & Porter (2014).
The most durable synthesis is an evidential one. A good account specifies the source of a cue, its diagnostic relevance for a defined construct, the way the perceiver integrates it, the independent criterion against which the inference is evaluated, and the conditions under which new evidence changes the impression. Consensus, confidence, predictive validity, and causal influence remain separate dimensions of analysis. Componential accuracy models, critical reviews of facial attributions, and experimental work on diagnosticity show how those distinctions can be researched rather than presumed Funder (1995) Biesanz (2010) Cone & Ferguson (2015).
The resulting research agenda reaches beyond the first meeting. Interpersonal relationships, institutional judgments, digital profiles, and human–artificial encounters all produce impressions whose social consequences may exceed their evidential reliability. Scientific progress depends on combining experimental manipulation, ecologically valid observation, longitudinal correction, representative cue sampling, and explicit assessment of uncertainty. Person perception is best understood as a revisable and context-dependent inference process whose accuracy is earned through relevant evidence and tested against independently specified realities.
References
- Albright, L., Kenny, D. A., & Malloy, T. E. (1988). Consensus in personality judgments at zero acquaintance. Journal of Personality and Social Psychology, 55(3), 387–395. https://doi.org/10.1037/0022-3514.55.3.387
- Ambady, N., & Rosenthal, R. (1992). Thin slices of expressive behavior as predictors of interpersonal consequences: A meta-analysis. Psychological Bulletin, 111(2), 256–274. https://doi.org/10.1037/0033-2909.111.2.256
- Ambady, N., & Rosenthal, R. (1993). Half a minute: Predicting teacher evaluations from thin slices of nonverbal behavior and physical attractiveness. Journal of Personality and Social Psychology, 64(3), 431–441. https://doi.org/10.1037/0022-3514.64.3.431
- Anderson, N. H. (1965). Averaging versus adding as a stimulus-combination rule in impression formation. Journal of Experimental Psychology, 70(4), 394–400. https://doi.org/10.1037/h0022280
- Asch, S. E. (1946). Forming impressions of personality. Journal of Abnormal and Social Psychology, 41(3), 258–290. https://doi.org/10.1037/h0055756
- Back, M. D., Schmukle, S. C., & Egloff, B. (2010). Why are narcissists so charming at first sight? Decoding the narcissism–popularity link at zero acquaintance. Journal of Personality and Social Psychology, 98(1), 132–145. https://doi.org/10.1037/a0016338
- Biesanz, J. C. (2010). The social accuracy model of interpersonal perception: Assessing individual differences in perceptive and expressive accuracy. Multivariate Behavioral Research, 45(5), 853–885. https://doi.org/10.1080/00273171.2010.519262
- Cone, J., & Ferguson, M. J. (2015). He did what? The role of diagnosticity in revising implicit evaluations. Journal of Personality and Social Psychology, 108(1), 37–57. https://doi.org/10.1037/pspa0000014
- Elfenbein, H. A., & Ambady, N. (2002). On the universality and cultural specificity of emotion recognition: A meta-analysis. Psychological Bulletin, 128(2), 203–235. https://doi.org/10.1037/0033-2909.128.2.203
- Fiske, S. T., Cuddy, A. J. C., Glick, P., & Xu, J. (2002). A model of (often mixed) stereotype content: Competence and warmth respectively follow from perceived status and competition. Journal of Personality and Social Psychology, 82(6), 878–902. https://doi.org/10.1037/0022-3514.82.6.878
- Fiske, S. T., & Neuberg, S. L. (1990). A continuum of impression formation, from category-based to individuating processes: Influences of information and motivation on attention and interpretation. Advances in Experimental Social Psychology, 23, 1–74. https://doi.org/10.1016/S0065-2601(08)60317-2
- Funder, D. C. (1995). On the accuracy of personality judgment: A realistic approach. Psychological Review, 102(4), 652–670. https://doi.org/10.1037/0033-295X.102.4.652
- Gilbert, D. T., Pelham, B. W., & Krull, D. S. (1988). On cognitive busyness: When person perceivers meet persons perceived. Journal of Personality and Social Psychology, 54(5), 733–740. https://doi.org/10.1037/0022-3514.54.5.733
- Goodwin, G. P., Piazza, J., & Rozin, P. (2014). Moral character predominates in person perception and evaluation. Journal of Personality and Social Psychology, 106(1), 148–168. https://doi.org/10.1037/a0034726
- Hehman, E., Sutherland, C. A. M., Flake, J. K., & Slepian, M. L. (2017). The unique contributions of perceiver and target characteristics in person perception. Journal of Personality and Social Psychology, 113(4), 513–529. https://doi.org/10.1037/pspa0000090
- Higgins, E. T., Rholes, W. S., & Jones, C. R. (1977). Category accessibility and impression formation. Journal of Experimental Social Psychology, 13(2), 141–154. https://doi.org/10.1016/S0022-1031(77)80007-3
- Jenkins, R., White, D., Van Montfort, X., & Burton, A. M. (2011). Variability in photos of the same face. Cognition, 121(3), 313–323. https://doi.org/10.1016/j.cognition.2011.08.001
- Kelley, H. H. (1950). The warm–cold variable in first impressions of persons. Journal of Personality, 18(4), 431–439. https://doi.org/10.1111/j.1467-6494.1950.tb01260.x
- Macrae, C. N., & Bodenhausen, G. V. (2000). Social cognition: Thinking categorically about others. Annual Review of Psychology, 51, 93–120. https://doi.org/10.1146/annurev.psych.51.1.93
- Mann, T. C., & Ferguson, M. J. (2015). Can we undo our first impressions? The role of reinterpretation in reversing implicit evaluations. Journal of Personality and Social Psychology, 108(6), 823–849. https://doi.org/10.1037/pspa0000021
- Mende-Siedlecki, P., Baron, S. G., & Todorov, A. (2013). Diagnostic value underlies asymmetric updating of impressions in the morality and ability domains. Journal of Neuroscience, 33(50), 19406–19415. https://doi.org/10.1523/JNEUROSCI.2334-13.2013
- Mende-Siedlecki, P., Cai, Y., & Todorov, A. (2013). The neural dynamics of updating person impressions. Social Cognitive and Affective Neuroscience, 8(6), 623–631. https://doi.org/10.1093/scan/nss040
- Naumann, L. P., Vazire, S., Rentfrow, P. J., & Gosling, S. D. (2009). Personality judgments based on physical appearance. Personality and Social Psychology Bulletin, 35(12), 1661–1671. https://doi.org/10.1177/0146167209346309
- Nisbett, R. E., & Wilson, T. D. (1977). The halo effect: Evidence for unconscious alteration of judgments. Journal of Personality and Social Psychology, 35(4), 250–256. https://doi.org/10.1037/0022-3514.35.4.250
- Olivola, C. Y., Funk, F., & Todorov, A. (2014). Social attributions from faces bias human choices. Trends in Cognitive Sciences, 18(11), 566–570. https://doi.org/10.1016/j.tics.2014.09.007
- Oosterhof, N. N., & Todorov, A. (2008). The functional basis of face evaluation. Proceedings of the National Academy of Sciences, 105(32), 11087–11092. https://doi.org/10.1073/pnas.0805664105
- Todorov, A., Olivola, C. Y., Dotsch, R., & Mende-Siedlecki, P. (2015). Social attributions from faces: Determinants, consequences, accuracy, and functional significance. Annual Review of Psychology, 66, 519–545. https://doi.org/10.1146/annurev-psych-113011-143831
- Todorov, A., & Porter, J. M. (2014). Misleading first impressions: Different for different facial images of the same person. Psychological Science, 25(7), 1404–1417. https://doi.org/10.1177/0956797614532474
- Willis, J., & Todorov, A. (2006). First impressions: Making up your mind after a 100-ms exposure to a face. Psychological Science, 17(7), 592–598. https://doi.org/10.1111/j.1467-9280.2006.01750.x