What Observer Agreement Adds to Big Five Traits

What does observer agreement add?

When different people describe a similar Big Five tendency in the same person, their agreement adds corroboration: the pattern is recognizable from more than one perspective. It is most useful when each observer can point to a concrete example from a different situation. Agreement does not make the description complete or objective. Several people may use the same word while relying on one shared episode or similar expectations about its meaning.

If one observer recalls someone speaking up in a planning meeting and another describes them initiating discussion in a different group, those examples offer broader support for visible social engagement than several retellings of the first meeting. The examples support a pattern across those settings; they do not establish private experience or behavior everywhere.

Connelly and Ones' meta-analysis treats observer agreement, self–observer comparisons, and behavior prediction as distinct research questions. This cautions against treating agreement as proof. use matching descriptions as provisional evidence: ask what behavior each person means, when they saw it, and whether their examples are separate. More relevant, varied evidence gives the shared description more weight.

Sources: An Other Perspective on Personality: Meta-Analytic Integration of Observers’ Accuracy and Predictive Validity

What are observers agreeing about?

When several people describe the same person in similar Big Five terms, it helps to ask what kind of agreement they have. In “An Other Perspective on Personality: Meta-Analytic Integration of Observers’ Accuracy and Predictive Validity,” Connelly and Ones separate three questions that can sound alike in everyday conversation. Consensus asks whether multiple observers give similar ratings of one target. Self–observer convergence asks whether the target’s self-rating resembles another person’s rating. Criterion accuracy asks whether a rating corresponds to a separately measured behavior or outcome. The first is agreement between observers; the second is agreement across viewpoints; the third checks a rating against something beyond the ratings themselves. These distinctions matter because a shared description is evidence about the relationship between reports, not by itself evidence that the description is true in every relevant sense. Suppose three people call someone conscientious. If they independently mean that this person usually prepares before meetings, they show consensus about a recognizable pattern. That still does not establish how the person sees their own habits, nor does it verify the label against a separate measure of behavior. If the observers all repeat one story they heard from the same colleague, their matching words may reflect one shared source rather than three independent observations. Even when observers directly saw similar conduct, the trait label summarizes behavior; it does not capture every situation or private experience. Connelly and Ones’ 2010 meta-analysis makes the separation concrete. It integrated ratings of 44,178 target individuals across 263 independent samples and reported three analyses: interrater consensus or reliability, correspondence between self- and observer ratings, and prediction of behavior. The authors treated these as separate criteria, rather than collapsing all three into one agreement score. That design is useful because a finding in one analysis answers only its own question. A high correspondence between observers does not tell us, on its own, whether self-ratings match them or whether either report predicts an independently assessed behavior. Likewise, prediction of a particular outcome does not show that observers agree about every trait or context. The study also warns that consensus can reflect more than accurate perception: shared stereotypes can contribute to interrater reliability. For everyday reflection, then, “people agree” is a starting description of the evidence, not a verdict. Ask whether the claim concerns observer-to-observer consensus, self–observer convergence, or a comparison with a distinct behavior criterion. Then ask what that criterion actually measured and whether it fits the question at hand. Keeping the categories separate lets agreement add perspective without turning a vote into proof.

Sources: An Other Perspective on Personality: Meta-Analytic Integration of Observers’ Accuracy and Predictive Validity

What does a second perspective add beyond self-description?

A second perspective can add information because a person's account and another person's account are related, but not interchangeable. In the meta-analysis The Convergent Validity between Self and Observer Ratings of Personality, researchers pooled findings on Big Five ratings across studies. Corrected mean self–observer correlations ranged from .46 for Agreeableness to .62 for Extraversion, with values of .56 for Conscientiousness, .51 for Emotional Stability, and .59 for Openness to Experience. The analysis also found substantial unique variance in both self and observer ratings. These are pooled self-versus-observer comparisons, not a direct test of adding a second independent observer. A correlation measures how ratings vary together; it cannot by itself establish which description is accurate, or whether either one captures the whole person. The findings support keeping both vantage points in view, rather than counting agreement as a verdict. In plain terms, the two sources tended to overlap, yet each retained information not captured by the other. The extra information may come from a difference in vantage point. A person has access to private reactions: what they felt before speaking, whether a decision required effort, or what they intended to do. Someone else may more readily notice visible conduct: whether the person spoke up in a group, followed through on a plan, or changed course when challenged. A quiet meeting, for example, might reflect a preference for listening, uncertainty about the topic, or a deliberate choice to wait. An observer can describe the silence; the person can report an inner reason. Both accounts may be useful, but they address different evidence. The study Who knows what about a person? The self-other knowledge asymmetry (SOKA) model illustrates why perspective can matter by trait aspect. In a study of 165 participants, each person rated themselves and was rated by four friends and up to four strangers; behavioral tests supplied separate criteria. The study reported that self-ratings best judged neuroticism-related traits, friends best judged intellect-related traits, and perspectives were similarly informative for extraversion-related traits. This is one study with its own participants and behavioral measures, not a rule for every person or every Big Five facet. When accounts conflict, neither self nor others deserve automatic priority. Compare what each could know, then keep the interpretation open where their evidence does not overlap.

Sources: The Convergent Validity between Self and Observer Ratings of Personality: A Meta-Analytic Review; Who knows what about a person? The self-other knowledge asymmetry (SOKA) model

Why do some trait descriptions produce more agreement?

Descriptions tend to align more when they concern visible behavior and less evaluative wording. Repeated conversational habits may be easier for several observers to notice than a private worry or an interpretation such as “selfish.” But visibility alone does not make a broad trait label precise, and agreement does not establish accuracy. In “Determinants of Interjudge Agreement on Personality Traits,” the authors examined how agreement varied with Big Five domain, observability of relevant behavior, evaluativeness of the trait, and whether the person being rated was among the judges. Findings were cross-validated in two samples. More observable and less evaluative traits elicited higher agreement. This is a pattern across the study’s ratings, not a rule for judging any one description. Observability means how readily the relevant behavior can be seen by the people making the judgment. Several people who regularly share conversations with someone may be able to say whether that person often starts discussions. They may have less access to how much the person privately replays a conversation afterward. Even the first behavior is only one possible part of a wider trait description. Seeing it does not show that the tendency appears in other settings or that observers interpret it alike. Evaluativeness concerns how much a trait word carries approval or disapproval. “Often offers to help” names an action people can compare. “Selfless” adds a judgment about motive and virtue; observers may agree about the action but differ over its meaning. The study found that self–peer agreement was lower than peer–peer agreement on average, but this difference was limited to evaluative traits. For neutral traits, self–peer agreement was as high as peer–peer agreement. This does not support a general claim that people cannot judge themselves as well as others can. The study also reported the highest agreement for Extraversion-related traits and the lowest for Agreeableness-related traits. These are domain-level patterns, not evidence that every social behavior is obvious or that agreeableness is unknowable. A broad domain contains different behaviors, and an adjective can bundle actions with a judgment about their meaning. For reflection, translate the adjective into an observable question before counting agreement. If several people call someone “dependable,” ask what each saw: arriving when agreed, completing a task, or responding when plans changed. Check whether examples come from distinct occasions and whether there are settings where behavior differs. This makes the description more specific and easier to examine. It still does not turn agreement into proof: observers can share a standard, and an action can support more than one interpretation. The useful addition is clearer corroboration about particular behavior, with the trait conclusion kept provisional.

Sources: Determinants of Interjudge Agreement on Personality Traits

Whose perspective can see the behavior?

The most useful perspective depends on what is being judged and who has had a chance to see it. A person may know private reactions that friends cannot observe; friends may notice recurring conduct the person does not readily recall or treat as distinctive. Neither vantage point wins across every trait. Observer agreement is more informative when we ask what evidence each perspective could access, rather than assuming that either the person or the people around them must be right. The study “Who knows what about a person? The self-other knowledge asymmetry (SOKA) model” tested this conditional view with 165 participants. Each participant rated themselves and was rated by four friends and as many as four strangers in a round-robin design. Participants also completed behavioral tests, which supplied separate criteria for comparing the ratings. The results varied by trait aspect: self-ratings were the best judgments for neuroticism-related traits, friends’ ratings were best for intellect-related traits, and self, friend, and stranger perspectives were similarly informative for extraversion-related traits. These are findings from one study and its particular behavioral-test battery, not a rule for deciding whose account is right in every life or relationship. The contrast helps explain why access matters. A person’s private worry, for example, may not be visible in a group conversation; a friend who sees only that conversation has little direct evidence about it. Conversely, repeated actions across shared activities may be easier for an observer to recall than for the person to summarize about themselves. These are illustrations of differences in access, not claims that a particular trait always belongs to one source. The study’s pattern suggests that a rating should be interpreted alongside the kind of evidence the trait calls for: private experience, conduct others can observe, or judgments shaped by social evaluation. When two descriptions conflict, first ask what each person could actually see and over what period. When they match, ask the same question: did both perspectives have relevant opportunities, or are they echoing one narrow slice of behavior? The SOKA findings support conditional perspective-taking, not a universal ranking of self, friends, and strangers. Agreement can strengthen a tentative interpretation when it draws on pertinent evidence; it cannot settle a trait description by itself. Keep the conclusion proportionate to the access behind it, and leave room for examples from settings neither account covers.

Sources: Who knows what about a person? The self-other knowledge asymmetry (SOKA) model

When does familiarity help, and when can it blur independence?

Familiarity can improve a rater's access to recurring behavior, but it does not guarantee independent evidence. Two close observers may know someone well and still draw on the same episodes, conversations, or expectations. Matching descriptions then show a shared view, not necessarily two separate checks. Ask what each could observe and whether their examples are distinct. The 2007 meta-analysis, “The Convergent Validity between Self and Observer Ratings of Personality,” examined self–observer correlations across Big Five studies. Duration of acquaintance moderated convergence, while observer type, such as work peers versus relatives, did not. Knowing someone longer can matter when comparing an observer’s account with the person’s own. But this analysis does not measure how much a second outside observer adds to the first, or whether two familiar observers are independent. Convergence with self-description and agreement among observers remain different questions. The 2010 meta-analysis, “Observer Reports of Personality,” distinguishes interaction frequency from interpersonal intimacy. Frequency means how often people encounter one another; intimacy concerns relationship closeness and available information, including self-disclosure. Across 263 independent samples and 44,178 targets, the authors analyzed observer consensus, self–other correlations, and behavior prediction separately. They reported that more frequent interaction improved rating accuracy, while intimacy was needed for substantial increases in other-rating accuracy. Intimacy mattered especially for less visible traits, with smaller gains for highly evaluative traits. These pooled findings do not decide whether particular friends are right. A colleague who sees someone daily may know their meeting habits but have little access to private thoughts or behavior in unfamiliar settings. A close friend may know more about private experience, yet two friends who discuss the same disagreement may recall one shared episode. This is illustrative, not a study result. More exposure can broaden evidence; shared exposure can make accounts less independent. That is an inference from the research distinction, not a score or measured adjustment. The 2025 review, “Self- and Observer Reports of Personality,” says close informants’ ratings often converge substantially with self-ratings, though agreement is somewhat lower for cooperation-related traits, including Big Five Agreeableness. It also notes some similarity or assumed similarity between closely acquainted people for Openness and, to some extent, Agreeableness. This complicates interpretation; it does not show that close observers are generally biased or matching ratings false. Familiarity may improve access while shared assumptions shape interpretation. To judge what agreement adds, ask whether each observer saw the behavior in a different setting and described separate occasions, or repeated a story they heard together. Were they rating an observable action or inferring a broad quality? What period does each have in mind? Distinct examples offer more corroboration. When accounts depend on one context or shared interpretation, keep the description narrower and provisional.

Sources: The Convergent Validity between Self and Observer Ratings of Personality: A Meta-Analytic Review; An Other Perspective on Personality: Meta-Analytic Integration of Observers’ Accuracy and Predictive Validity; Self- and Observer Reports of Personality

What can agreement predict—and what can it not prove?

Observer ratings can predict some independently measured behavior in studied settings, but prediction is separate from agreement. It does not certify an individual trait description or define a person's identity. Ask not only “Did observers agree?” but “What separately measured outcome did the descriptions help predict, and does that outcome matter to this question?” In “An Other Perspective on Personality: Meta-Analytic Integration of Observers’ Accuracy and Predictive Validity,” researchers combined evidence on 44,178 target individuals from 263 independent samples. They conducted three meta-analyses, treating agreement between observers, correspondence between self and observer ratings, and prediction of behavior as distinct criteria. In the behavior-prediction analysis, observer ratings predicted behavior; for academic achievement and job performance, they had predictive validities substantially greater than and incremental to self-ratings. “Incremental” means observer information contributed predictive value beyond self-ratings in those analyses. The result supports a bounded point: other people's descriptions can carry information about some outcomes beyond what a person reports about themselves. This does not mean that a group that agrees has verified a trait in the same way a separate outcome can test a prediction. Observers may use similar words because they noticed the same behavior, share a setting, or hold overlapping expectations. Agreement can show that a description is recognizable across perspectives; prediction asks whether ratings relate to a separately measured result. The meta-analysis treats these as distinct questions. The criterion must also be relevant. Academic achievement and job performance were outcomes in the reported analysis; they do not establish what a trait means in friendship, at home, or during a difficult conversation. Nor do group-level predictive findings establish what one person's behavior will be. A criterion is an operational measure: a defined way to represent an outcome. It can be useful without being exhaustive. Performance measures, for example, answer a narrower question than whether someone is considerate, dependable, or conscientious in every part of life. Treat predictive evidence as support for observer reports under the conditions studied, not as a vote that settles identity. Ask what behavior or outcome was measured, whether it was distinct from the ratings, and whether it matches the situation you want to understand. If it is distant from that question, the finding may support the method generally while adding little to your interpretation. Agreement can prompt a hypothesis; a relevant, separately measured outcome can add another kind of support, but neither turns a dimensional tendency into a complete verdict about a person.

Sources: An Other Perspective on Personality: Meta-Analytic Integration of Observers’ Accuracy and Predictive Validity

How should disagreement change the interpretation?

Disagreement between self-description and another person's view does not show that either account is dishonest or mistaken. It shows that descriptions differ; ask what evidence each rests on. A person may describe private worry, while an observer describes conduct in a setting the person rarely considers. Those accounts may not concern the same behavior. “Who knows what about a person? The self-other knowledge asymmetry (SOKA) model” gives a reason to inspect access before choosing a side. In a sample of 165 participants, each person rated themselves and was rated by four friends and up to four strangers; behavioral tests supplied criteria. Self-ratings best judged neuroticism-related traits, friends best judged intellect-related traits, and perspectives were similarly informative for extraversion-related traits. This is one study with specific measures. Ask whether the disputed trait concerns private experience, visible conduct, or both. Differences may reflect setting or time. One observer may mean behavior in a familiar group, another in a new situation. “Usually quiet” might refer to one meeting or interactions across settings. Ask each person for an action and when it occurred. Different contexts may give both accounts narrower meaning. Trait words can create disagreement. “Considerate” carries evaluation, and observers may apply it to different actions. Ask what the person did that prompted the label. They may have noticed different behavior or interpreted the same act differently. Assuming people rate themselves more favorably is unwarranted. “Self-Other Agreement in Personality Reports” compared self- and informant-report means for the same targets in 33,033 people across 152 samples. The average difference was small, inconsistent with a general self-enhancement effect; moderate differences appeared in comparisons with strangers. This meta-analysis concerns group averages. It cannot explain one person's motive or settle a disagreement. Use it to resist an accusation, not to conclude every perspective is equally informative. State what would change the interpretation. If the claim is that someone rarely speaks up, look for examples in another setting where participation is possible. A counterexample may narrow the description to unfamiliar groups or rapid discussion. Similar examples from distinct observers broaden support, but still describe a tendency, not a complete account. Do not average the accounts or count votes yet. Ask each person for an observable example, its setting, and when it occurred. Compare access and the meaning of the trait word, then consider a counterexample. Decide whether to retain the description provisionally, limit it to a context, or leave the question open.

Sources: Who knows what about a person? The self-other knowledge asymmetry (SOKA) model; Self-Other Agreement in Personality Reports

How can matching descriptions be checked in daily life?

Matching descriptions are most useful when they point to examples, not just the same adjective. Write down what happened, where and when, who could see it, and whether each observer is drawing on a separate occasion. Then look for a boundary: a setting in which the behavior was absent, or a different behavior that complicates the label. It is a reflection aid, not a validated measure. Take “often speaks up in groups,” an illustrative description, not a report about a particular person. Ask each observer what they mean. One may recall asking questions in a familiar discussion; another may remember presenting an idea in a new group. If both refer to the same meeting, their agreement may repeat one episode. Separate occasions provide broader corroboration, but do not show the behavior appears everywhere. A brief evidence map keeps the comparison concrete: identify the description; note the action behind it; record setting and time period; ask whether the observer could see that behavior; check whether the example is independent or a shared story; then note a counterexample or boundary. “Asked two questions before the group moved on” is more observable than “is outgoing.” A broad label can mean different things to each observer; translation separates observation from interpretation. In two cross-validated samples, “Determinants of Interjudge Agreement on Personality Traits” found higher agreement for more observable and less evaluative traits. Agreement was higher for Extraversion-related traits and lower for Agreeableness-related traits. These group-level results do not mean every visible action maps neatly to Extraversion or that agreement proves a description true. Access matters as much as wording. “Who Knows What About a Person? The Self-Other Knowledge Asymmetry (SOKA) Model” studied 165 participants rated by themselves, friends, and strangers, then compared ratings with behavioral-test criteria. Results varied by trait aspect: the self was the best judge for neuroticism-related traits, friends for intellect-related traits, and perspectives were similarly informative for extraversion-related traits. The question is who could observe the behavior at issue: a private worry and a visible meeting contribution call for different vantage points. Keep the time window consistent. “The Convergent Validity between Self and Observer Ratings of Personality” found that pooled self–observer correlations varied across Big Five dimensions and that duration of acquaintance moderated convergence. These findings do not quantify what one additional observer adds, but support asking whether observers had relevant contact with a recurring tendency. If examples cluster in one setting, narrow the claim to that setting. If separate examples recur across settings, retain it provisionally and keep the boundary visible.

Sources: The Convergent Validity between Self and Observer Ratings of Personality: A Meta-Analytic Review; Determinants of Interjudge Agreement on Personality Traits; Who knows what about a person? The self-other knowledge asymmetry (SOKA) model

What should you do with agreement now?

Keep a matching description as a working hypothesis when each observer can give a specific example and had a distinct opportunity to see the behavior. Narrow it when examples come from one shared setting or the pattern does not appear elsewhere. The fit of the examples matters more than the number of people repeating an adjective.

Use four checks. Ask what the person did, rather than which label an observer would use. Note when and where it happened, and whether that observer saw that part of the person’s life. Ask whether examples are separate or trace back to one event. Then look for a counterexample: a setting in which the behavior was absent or took another form. The 2025 review, “Self- and Observer Reports of Personality,” describes convergence across close informants and discusses assumed similarity among acquaintances. This supports checking access and independence, but gives no formula for scoring an individual description.

Decide whether the pattern helps you understand a tendency, needs a setting attached, or remains unresolved pending varied examples. To consider how this tendency sits alongside others, the private Context Profile at /assessment is an optional exercise. It is non-validated; results stay local unless you explicitly request anonymous AI synthesis. For more explanations of personality frameworks and everyday behavior, explore the learning library at /topics.

Sources: Self- and Observer Reports of Personality