Self-report and informant ratings are complementary methods for studying personality. Self-report asks individuals to describe their own recurring thoughts, feelings, motives, and behavior. Informant report asks someone else—such as a partner, friend, parent, sibling, teacher, or coworker—to judge the target person from what that observer has seen and inferred. Each source is a measurement perspective with distinctive access to information, interpretive habits, and error. The scientific question is therefore how ratings converge, when they diverge, what each perspective predicts, and how researchers can distinguish information about personality from characteristics of the rater and relationship. Ashton and Lee, 2025; Connelly and Ones, 2010.
The central empirical conclusion is that knowledgeable informants often agree substantially with self-descriptions, while both perspectives can contain distinctive valid information. Agreement varies across traits, relationships, populations, measures, and criteria. Average differences in score levels, agreement about who scores higher than whom, and accuracy against independently observed behavior are separate outcomes. None can substitute for the others. A rigorous personality assessment identifies which of these outcomes matters for its purpose and selects evidence accordingly. Kim, Di Domenico, and Connelly, 2019; Vazire, 2010.
What Self-Report and Informant Ratings Measure
Personality traits are inferred from patterns, not observed as objects
A personality trait summarizes a relatively enduring tendency in how a person thinks, feels, desires, or behaves across occasions. The trait itself is a theoretical construct; a questionnaire score is an estimate derived from reports about a sample of experiences or actions. A state describes functioning at a particular moment or in a particular situation. A respondent's current irritation may influence a questionnaire answer without demonstrating a long-term change in irritability. Repeated measures, multiple settings, and thoughtfully selected informants can help separate momentary impressions from recurring tendencies. Funder, 1995; Ashton and Lee, 2025.
A self-report item such as “I usually hesitate before speaking in unfamiliar groups” can draw on memories, private experiences, goals, and behaviors invisible to companions. An informant version asks whether the person tends to hesitate. The informant may have witnessed many group interactions but may not know whether the person felt afraid, was choosing to listen, or was unfamiliar with the topic. The same observable action can have different motives. Consequently, agreement is evidence that perspectives converge on a description; disagreement invites examination of observations and meanings rather than an automatic verdict on either respondent.
Private experience, public behavior, identity, and reputation
Self-report offers direct access to a person's consciously available experience and self-concept. People know many of their unspoken worries, intentions, fantasies, and preferences, while memories and self-interpretations can be selective. Informants observe speech, interpersonal conduct, routines, and consequences in settings they share with the target, but they infer private motivation indirectly. An individual can also present different aspects of personality at work and at home. Distinct ratings may be accurate descriptions of distinct relationships and occasions.
The trait–reputation–identity framework treats shared trait information, socially held reputation, and self-perceived identity as distinguishable sources of variation. A rater's perspective can contain substantive information as well as evaluative coloring. This is a useful theoretical framework for multi-rater data, rather than proof that every instrument can isolate three independent components perfectly. McAbee and Connelly, 2016.
What Does Self–Other Agreement Mean?
Rank-order agreement
Rank-order agreement asks whether people whom self-reports place relatively high on a trait are also placed relatively high by informants. It is commonly investigated through correlations across a sample of targets. A positive self–informant correlation indicates shared ordering and is compatible with different mean levels. A correlation describes covariation across sampled people and cannot determine whether two particular respondents agree about one target. Its magnitude depends in part on the trait variation present in the sample, the reliability and content of both instruments, and the range of targets recruited. Connelly and Ones, 2010; Ashton and Lee, 2025.
For instance, if a group of close friends rates the same set of people as generally more sociable than those people rate themselves, the two groups of scores could still show strong rank-order agreement. Conversely, identical average scores could conceal poor agreement about the relative positions of individuals. Claims such as “self and observer reports agree” should therefore specify the statistical criterion, the population, and the trait.
Mean-level agreement and systematic discrepancies
Mean-level agreement concerns whether a self-report group and an informant-report group give similar average scores on comparable scales. The 2019 meta-analysis of 152 samples comprising 33,033 targets reported little overall average difference in Big Five self- and informant-report means (mean standardized discrepancy approximately −0.038). Differences were more evident when the comparison involved strangers rather than knowledgeable informants. These findings do not establish equal scores for every participant or rule out discrepancies in specific facets, settings, or cultural groups. Kim, Di Domenico, and Connelly, 2019.
A 2024 HEXACO investigation of 5,405 Dutch university students, drawing on reports from partners, parents, siblings, and friends, similarly found strong correlations across the studied dimensions, with mean correlations at least .59 in the reported comparisons and the highest agreement for partners. Openness showed higher average self-ratings than informant ratings (reported d = 0.37). Parents and siblings displayed distinct characteristic patterns when judging the same general trait domains. These values describe that study's design and population; they are not universal calibration constants. Lee, de Vries, and Ashton, 2024.
Absolute agreement, reliability, and accuracy
Absolute agreement requires closeness of scores for the same person on a compatible metric. It differs from correlation, which can remain high despite systematic shifts between raters. Reliability concerns the reproducibility or consistency of a measurement under a specified design. Accuracy concerns correspondence with an independent and defensible criterion. A reliable, highly consensual set of reports may reproduce a shared stereotype; a measure that agrees weakly with another source may nevertheless predict a relevant outcome. These distinctions are central to multitrait–multimethod reasoning, but detailed general reliability and scale-construction problems belong to psychometric methodology. Campbell and Fiske, 1959; Funder, 1995.
Why People Agree About Personality
Observable and recurring behavioral cues
Certain tendencies leave frequent public traces: arriving on time, initiating conversation, speaking forcefully, cooperating with others, or finishing tasks. Informants who see many relevant episodes can infer stable patterns from repeated behavior. Others require access to internal experiences—private anxiety, rumination, self-doubt, or unexpressed wishes—that may be less accessible through observation. Funder's Realistic Accuracy Model describes accurate judgment as dependent on relevant behavioral cues being available, detected, and used appropriately. The model provides a conceptual explanation for variation in agreement; direct empirical studies are required to test its predictions. Funder, 1995.
Judgment is often more robust when observations span occasions. A partner who has seen someone negotiate disagreements, manage unexpected demands, and maintain everyday commitments has a different information base from a stranger who watches a brief interaction. The content of the trait matters, however: even long acquaintance cannot guarantee access to private experiences or establish that a favorable impression is accurate. Connelly and Ones, 2010.
Acquaintance, closeness, and opportunities to observe
Connelly and Ones synthesized three meta-analyses drawn from 263 independent samples involving 44,178 targets. The research separated consensus among informants, self–other correlations, and prediction of outcomes. Interacting with a target more frequently generally improved observers' accuracy, while interpersonal intimacy could provide additional value, especially for relatively less visible traits. The evidence supports informed observer assessment and cautions against treating sheer contact hours as equivalent to accurate knowledge. Connelly and Ones, 2010.
Closeness can also complicate judgment. Partners and friends share routines and private conversations, but may protect a relationship, project their own preferences, or focus on familiar roles. A work colleague can know occupational diligence while remaining unaware of behavior outside work. Parents and adolescents may interpret the same action through different expectations. These mechanisms generate specific testable hypotheses, not a ranking in which one relationship type is always the best rater.
The influence of trait visibility and evaluativeness
Visibility refers to the extent to which a trait is manifested in accessible behavior. Evaluativeness refers to how socially desirable, undesirable, or morally charged the trait content seems. These dimensions can affect judgments differently. A visibly talkative person is comparatively easy to assess, whereas opinions about another person's fairness may be influenced by reputation, relationship history, and disagreement about what counts as fair. The large meta-analytic observer literature and the 2025 integrative review both identify differences across traits and sources rather than a uniform agreement coefficient. Connelly and Ones, 2010; Ashton and Lee, 2025.
When Does the Self Know More, and When Do Others Know More?
The self–other knowledge asymmetry model proposes that the self often has an advantage for less observable experiences, whereas knowledgeable others may be better judges of highly evaluative attributes that people have difficulty appraising impartially. In Vazire's empirical study, 165 participants received ratings from friends and strangers and completed behavioral criterion tasks. Evidence supported a self advantage for neuroticism-related traits, a friend advantage for intellect-related traits, and comparable accuracy across perspectives for extraversion-related traits in that design. These findings are informative tests of the model, not a rule determining which reporter is correct for every person or instrument. Vazire, 2010.
The distinction between a judgment's accuracy and a rater's confidence matters. People may feel especially certain about their own intentions or moral standing, while an observer feels certain about habitual behavior. Accuracy is established by how the judgment relates to well-chosen independent information, which itself has limitations. Some behaviors are easier to detect than stable internal motives, and some criteria are susceptible to the same social desirability pressures as the ratings being tested.
A separate study comparing self-ratings and other-ratings with everyday behavioral information found that both sources can have unique predictive utility. This supports the principle of complementary information without establishing that a recorded behavioral sample is a perfect ground truth: recording windows, privacy settings, activity selection, and observer coding all constrain what the criterion represents. Vazire and Mehl, 2008.
Bias in Self-Report Personality Assessment
Self-presentation and social desirability
Self-report respondents may answer in ways they regard as socially acceptable, especially when stakes are high, the setting feels evaluative, or confidentiality is uncertain. Other respondents genuinely endorse favorable traits as part of an organized self-concept. These mechanisms should be distinguished because intentional impression management and sincerely held self-perceptions are different psychological phenomena. A low score on a favorable trait can also reflect candor, strict standards, modesty, or differing item interpretation. Neither high nor low self-ratings alone establish deliberate deception.
The empirical picture resists a simple allegation that individuals systematically exaggerate how positively they should be judged. The 2019 meta-analysis found virtually no overall self-enhancement in Big Five mean comparisons. The 2024 HEXACO study also failed to demonstrate a general self-favoring mean advantage, although its domain-specific pattern included higher self-rated Openness. The conclusions concern observed average scores, not a proof that social desirability never influences responses. Kim, Di Domenico, and Connelly, 2019; Lee, de Vries, and Ashton, 2024.
Acquiescence, extreme responding, and item wording
Acquiescence is a tendency to endorse statements regardless of content. Other response tendencies involve avoiding extreme categories, using endpoints freely, responding carelessly, or treating positively and negatively worded items differently. These tendencies can affect self-reports and observer reports, although the form and context of each measure change the risk. Reverse-worded items are sometimes used to counter response styles, but can create comprehension burdens. Balanced scales and statistical corrections require instrument-specific validation rather than automatic assumptions about improvement. Ashton and Lee, 2010; Ashton and Lee, 2025.
Retrospective self-description also depends on memory and summary rules. Respondents may retrieve especially emotional or recent episodes, or use a personal standard instead of estimating behavior across an appropriate time period. A person who recalls one recent conflict may temporarily describe themselves as argumentative, while a friend relying on years of interactions assigns a lower rating. The discrepancy may arise from time frame and sampling rather than an enduring difference in perceived personality.
Identity, insight, and interpretive standards
People use personal narratives to organize experiences: “I am usually dependable,” “I am shy at first,” or “I handle criticism well.” These narratives shape which observations seem relevant. They can capture long-run experience that an observer lacks and can also make counterexamples easier to overlook. Observers have analogous narratives about the target. Explicitly defining the reporting interval, trait content, and behavior examples helps distinguish disagreement about the construct from disagreement about what happened.
Bias in Informant and Observer Ratings
Halo, liking, and global evaluation
A halo effect occurs when a broad favorable or unfavorable evaluation colors distinct characteristic judgments. Someone who likes a colleague may rate that colleague as warm, competent, conscientious, and emotionally steady without independent evidence for each dimension. A dissatisfied partner may interpret otherwise ordinary behavior less positively. This is a threat when items are highly evaluative, but it does not mean observers invariably exaggerate or that all covariance across ratings is artifactual. Studies that distinguish common trait variation from source-related factors offer a more discriminating analysis. Ashton and Lee, 2010; McAbee and Connelly, 2016.
A person can also become known through reputation. Information received from others may influence later ratings even without direct observation. Reputation can be socially consequential in its own right: someone regarded as reliable may be assigned responsibilities, regardless of whether every element of that reputation is accurate. This creates a distinction between whether reports track dispositions and whether beliefs held by others predict social outcomes.
Projection and assumed similarity
Informants sometimes assume other people share their own preferences or characteristic reactions. Similarity may be real, particularly in self-selected friendships and partnerships, and it may also be overestimated. Researchers therefore distinguish actual similarity between two people's traits from assumed similarity in a rater's depiction of the target. The 2025 review documents both resemblance and assumed similarity in some reports of Openness and Honesty-Humility among closely acquainted people. The magnitude and meaning of these processes depend on measurement designs that assess both people's traits independently. Ashton and Lee, 2025.
Role-specific observations and relationship effects
Parents, siblings, romantic partners, classmates, and supervisors encounter the same person under different norms and incentives. Relationships shape what behavior is observable, which actions are interpreted as relevant, and how warmly a target is viewed. In the Dutch HEXACO study, parents tended to view children as more honest and less anxious than some other relationship categories, while siblings tended to judge each other as less agreeable. These are group patterns conditioned on that study's sampling and measures. They should not be extrapolated to an individual household without supporting evidence. Lee, de Vries, and Ashton, 2024.
Personality Models Beyond the Big Five
HEXACO and evaluative source factors
Self–informant methods apply to many personality structures. HEXACO organizes variation into Honesty-Humility, Emotionality, Extraversion, Agreeableness, Conscientiousness, and Openness to Experience. Research using HEXACO measures is especially informative because Honesty-Humility and cooperation-related dimensions raise questions about morally charged descriptions. Ashton and Lee's 2010 study of 563 well-acquainted pairs found evidence for six shared trait factors alongside two source factors, interpreted in terms of differing evaluative emphases of self and observer perspectives. The conclusion arises from the specified factor-analytic models and sample rather than from direct observation of moral character. Ashton and Lee, 2010.
The later 2025 integrative review reports relatively strong self–observer correspondence for full-length Big Five and HEXACO scales with close acquaintances, with lower agreement in cooperation-related domains than in several other factors. Researchers must verify a given scale's observer form and norms before making comparisons; the existence of a model does not make every questionnaire a validated multi-informant measure. Ashton and Lee, 2025.
Temperament and narrower dispositions
Temperament research asks related but developmentally distinctive questions about reactivity, attentional control, and regulation. Parent and teacher observations often provide access to behaviors that young children cannot reliably describe in self-reports. The meaningful reporter changes with developmental level, language comprehension, classroom exposure, and family context. For a small child, disagreement between two adult observers may capture genuine setting-specific behavior. With adolescents, an additional self-report may reveal emerging reflective capacities and private emotion, while parents may know long-run patterns. The logic of multiple sources is relevant, but adult trait-report findings cannot simply be transferred to early childhood as if instruments and abilities were unchanged.
Measures of motivational sensitivity, interpersonal style, sensation seeking, and other normal-range dimensions likewise differ in observability. Certain private feelings or incentives may be better represented by self-report; outward behavior may be characterized efficiently by knowledgeable observers. Specific conclusions require studies of the chosen instrument, not analogy alone. The broader construct map is developed in Cluster Personality Traits, Temperament and Individual Differences.
Clinical personality information is a separate application
Clinical personality assessment sometimes uses collateral histories or informant measures when an individual has limited insight, but clinical appropriateness and informed consent require an independent framework. A normal-range trait questionnaire does not establish a personality disorder, psychiatric diagnosis, level of impairment, or need for treatment. Discrepancies between a parent and adolescent, a patient and relative, or two colleagues are data requiring context; they are not evidence that one respondent is ill or dishonest. The clinical canonical questions belong to dedicated diagnostic and assessment frameworks, not to interpretations of ordinary self–other correlations.
The Multitrait–Multimethod Logic
A sound comparison separates constructs from sources of information. If each of several traits is rated by both self and informants, the analyst can examine convergence for the same trait across sources and discriminability between different traits. Campbell and Fiske's multitrait–multimethod framework remains influential because agreement is more meaningful when a measure also distinguishes its intended characteristic from neighboring constructs. A correlation between self- and friend-rated sociability, for example, carries less interpretive weight if both ratings correlate equally with broad favorable impressions of the person. Campbell and Fiske, 1959.
Statistical models may incorporate trait factors, reporter-specific method factors, and residual error, but model identification, item selection, number of raters, and sample size affect what those components mean. The trait–reputation–identity framework proposes a theoretically richer decomposition in which self-perceived identity and socially shared reputation can explain meaningful source-specific variance. It must be applied with clear assumptions and independent validation, not treated as a universal formula for calculating the “true person.” McAbee and Connelly, 2016.
Designing Informant Studies
Selecting people who can actually judge the target
The most useful informant is the one who has repeated, relevant, and recent opportunities to observe the behaviors in question. Researchers should specify the selection rule: spouse or partner, friend with a minimum acquaintanceship period, classmate, coworker, teacher, or other defined relationship. Recruiting only agreeable friends risks selective coverage; relying on strangers may limit cues for private traits. The observer's closeness, frequency of interaction, relationship quality, and direct exposure should be recorded when feasible. The meta-analytic observer literature shows why acquaintance is part of the measurement design rather than an incidental demographic detail. Connelly and Ones, 2010.
More than one informant can reveal whether an impression is shared or tied to a particular relationship. Additional reporters, however, may not provide independent information if they belong to the same social network, discuss the subject with one another, or share the same cultural stereotype. Aggregation can reduce some source-specific noise while obscuring authentic context-specific differences. A study should state what aggregation is intended to estimate.
Equivalent questions, instructions, and time frames
Self and informant instruments should target the same trait domain and a comparable time frame. Switching grammatical person from “I” to “she” is necessary for many observer forms but does not, by itself, establish equivalence. A question about “feeling tense” may require an observer to infer internal experience, while a question about “appearing tense” may represent a different behavioral indicator. Differences in item wording, scoring, reference groups, and familiarity can change apparent agreement. Interpret absolute scores only when norms and scale comparability support doing so.
Instructions should identify whether ratings concern typical behavior, recent months, a particular setting, or a specific developmental period. Researchers should record who completed the scale, whether target and observer conferred, and whether the observer had access to the target's self-report. Independent completion is usually preferable when the aim is to examine agreement between perspectives.
Multiple outcomes and transparent analysis
A well-designed paper reports the distinction between mean differences, correlations, and predictive performance, and uses confidence intervals rather than only declaring statistical significance. The design should also address clustering when multiple informants rate a target, repeated observations across time, and the possibility that some data are systematically missing. Reporting only the strongest agreement coefficient can misrepresent a construct whose facets or settings vary sharply. Preregistration and independent replication strengthen conclusions when the hypothesis and analytical choices permit them.
What External Behavioral Validation Can Establish
Observed behavior, digital traces, task performance, or consequential life outcomes can serve as independent criteria, but their evidential value depends on relevance and data quality. A single social interaction is a narrow sample of behavior, a work evaluation can reflect organizational bias, and an online trace may reflect platform-specific habits. A questionnaire could validly describe tendencies beyond any one criterion. Investigators should therefore use representative observations and avoid declaring the highest-correlating source the universally accurate one. Funder, 1995; Vazire and Mehl, 2008.
Longitudinal research provides an additional test: do the self- and informant-based estimates relate to later outcomes after considering their overlap? In a German study following 186 participants from adolescence toward age 29, informant reports collected at ages 12 and 17 showed unique predictive associations beyond self-reports for several adult outcomes and later personality ratings, with particularly consistent results in some visible or evaluative trait domains. The sample, timing, and outcome measurements limit extrapolation; the design supports prognostic associations rather than proving that the observer's description caused later outcomes. Luan and colleagues, 2019.
Cross-Cultural Interpretation and Reference Groups
Questionnaire labels such as “assertive,” “modest,” “organized,” or “emotionally expressive” are understood against culturally situated behavioral standards. Respondents may implicitly compare themselves with familiar peers rather than a uniform international reference group. A national sample's lower mean self-report score can therefore reflect a stricter comparison standard, translation differences, response styles, or real trait differences. A peer-rating system also operates within cultural expectations and does not automatically escape these problems. Heine, Buchtel, and Norenzayan, 2008.
The research literature contains differing results that must be examined together. A 2008 analysis found serious concerns about cross-national interpretations of Conscientiousness averages when comparing self-reports, peer reports, and external indicators. A 2005 study of knowledgeable informant ratings across 51 cultures found broadly replicable factor structures and informative aggregate patterns. These papers asked different questions with different methods; one does not nullify the other. Claims about cultures require transparent translation, scale equivalence, sampling, comparison groups, and independent validation. Heine, Buchtel, and Norenzayan, 2008; McCrae and Terracciano, 2005.
Cross-cultural analyses have also identified replicable differences in self-versus-observer profile judgments in several settings, indicating that systematic perspective effects deserve measurement rather than an assumption that one culture is unusually biased. These are statistical findings about respondents and instruments, not character assessments of national populations. Allik and colleagues, 2010.
Children, Adolescents, and Changes Across the Lifespan
Development changes both the target's behavior and the quality of information available to reporters. Young children may struggle with abstract self-evaluations and extended retrospective time frames. Parents may observe long stretches of home behavior, while teachers observe peer interactions, frustration during assignments, and rule-following in classrooms. Such reports can disagree because different dispositions are expressed under different demands. Reports can also differ because one observer sees an unusually constrained sample of situations, not because the trait necessarily changed.
With adolescents, growing introspective abilities and privacy change the balance of information. The longitudinal study by Luan and colleagues illustrates the value of combining adolescent self and parent or friend perspectives when predicting later outcomes, but its small regional sample places limits on universal inference. For adults, partners, friends, and colleagues provide different windows into ordinary routines. Late-life contexts may alter reporting opportunities again through retirement, changing social networks, or health-related constraints. Developmental comparability should be demonstrated for the instrument, respondent and outcome rather than inferred from the continuity of a trait label. Luan and colleagues, 2019; Connelly and Ones, 2010.
Between-Person and Within-Person Evidence
Most self–informant agreement correlations concern between-person variation: people ranked higher by one source tend to rank higher by another. They provide limited information about whether the same individual's behaviors and feelings rise and fall together from day to day. Within-person research asks how a particular person changes across situations and times. To examine this, researchers may collect repeated self-reports of states and repeated informant observations of behavior in the same periods. These are analytically distinct levels and should be modeled accordingly.
A partner may rate someone's general social confidence consistently while the person reports daily confidence fluctuations that track work stress. The scores need not contradict one another: one is an estimate of average tendency, while the other samples states under changing demands. Repeated measurement may also expose informant lag, changes in acquaintance, or genuine personality development. A cross-sectional disagreement is insufficient to establish any of these mechanisms on its own.
Research Evidence: What Is Established, Limited, and Contested?
Established across multiple designs
The broad empirical findings are that self- and well-acquainted observer ratings frequently converge, that agreement is trait- and relationship-dependent, that multiple sources can have incremental predictive value, and that mean-level comparison is different from correlational agreement. These conclusions rest on large meta-analyses, a 2025 scientific synthesis, controlled judgment studies, and multi-rater longitudinal designs. Their strength concerns general patterns in studied samples. Connelly and Ones, 2010; Ashton and Lee, 2025; Kim, Di Domenico, and Connelly, 2019.
Model-dependent and still developing explanations
Explanatory frameworks about reputation, identity, method factors, visibility, evaluativeness, and assumed similarity are scientifically productive. The relative weight of each mechanism may change with instrument, trait, informant composition, culture, and research criterion. A theoretical model offering a plausible reason for disagreement is not a direct identification of that reason in any particular dataset. Even an apparently clear discrepancy between scores can result from multiple interacting processes. McAbee and Connelly, 2016; Vazire, 2010.
Contested or inadequately supported shortcuts
Claims that self-reports are universally inflated, friends always know a person better, informant consensus proves objective truth, or one discrepancy exposes a hidden disorder go beyond available evidence. The aggregate absence of overall self-enhancement does not eliminate individual response distortions. Conversely, evidence of a particular response distortion does not justify discounting an entire measurement source. Cross-national mean comparisons, casual internet ratings, and individual profile interpretations require particularly careful study-specific evidence. Kim, Di Domenico, and Connelly, 2019; Heine, Buchtel, and Norenzayan, 2008.
Practical Interpretation of Disagreeing Ratings
When two respondents provide different personality ratings, first inspect whether they used comparable measures and time frames. Next examine the trait's observable indicators and ask which situations each person had in mind. A score about conscientiousness may emphasize punctuality for one respondent and persistence with complex tasks for another. A score about social engagement may refer to unfamiliar events for one and close friendships for another. Such questions help interpret what each rating represents rather than treating a broad domain score as a complete personality account.
If two scores must be compared numerically, use the validated instrument's instructions, scoring, norms, and uncertainty estimates. Comparing raw totals from unrelated scales is generally misleading. For consequential professional uses, interpretation requires evidence of validity for the exact population, purpose, and decision. A low level of agreement is itself a result to investigate, not a diagnosis or an accusation of bias. Ashton and Lee, 2025; Campbell and Fiske, 1959.
Ethical Considerations and Privacy
Informant ratings contain information about relationships as well as a target person. Research involving identifiable personality profiles should address consent, confidentiality, storage, the conditions under which respondents can decline, and the risk of interpersonal consequences if answers become visible. A partner's confidential assessment can be useful scientific data while being inappropriate for unrestricted disclosure. Recruitment from hierarchical relationships, such as supervisor and employee or parent and dependent adolescent, requires attention to pressure and unequal power.
Personality ratings should not be used to stigmatize people, assert permanent moral worth, infer clinical conditions, or make unsupported deterministic predictions about relationships and work. Trait associations describe distributions and tendencies; they do not establish inevitable behavior or causal mechanisms. Ethical reporting specifies the limits of what the particular sources can support.
Frequently Asked Questions
Are personality self-reports scientifically valid?
They can provide useful, replicable information when administered and interpreted through instruments with appropriate evidence. Their validity depends on the construct, items, sample, response conditions, and intended use. The same applies to informant reports. The strongest approach is to examine converging sources and external evidence relevant to the question. Ashton and Lee, 2025.
Do friends know someone's personality better than that person does?
Sometimes knowledgeable friends provide stronger prediction for particular observable or evaluative traits, while self-reports can be especially informative for private experience. No source has a universal advantage for every characteristic, criterion, or individual. Vazire, 2010; Vazire and Mehl, 2008.
What is a good self–informant agreement correlation?
There is no context-free cutoff. Interpretation depends on the trait, instrument reliability, source relationship, sample variability, study design, and purpose. Published correlations are study-specific benchmarks, not universal pass–fail standards. Connelly and Ones, 2010.
Can two ratings have the same average but disagree about individuals?
Yes. Similar group means concern mean-level agreement, whereas correlational agreement concerns whether sources rank individual targets similarly. Both are distinct from absolute agreement on each target's score. Kim, Di Domenico, and Connelly, 2019.
Is an informant report free from self-presentation or social bias?
No. An observer can be influenced by liking, stereotypes, assumed similarity, selective exposure, incentives, and concern for the relationship. Self and observer ratings offer complementary perspectives rather than one pristine and one contaminated channel. Ashton and Lee, 2010.
Should several informant ratings be averaged together?
Aggregation can reduce some idiosyncratic variation and improve coverage, but it may also conceal meaningful differences between work, home, school, and other settings. Whether aggregation is appropriate depends on the construct and research question, the degree of informant dependence, and the available validation evidence.
Can disagreement reveal a personality disorder?
Disagreement on a normal-range personality inventory is not a diagnostic criterion. Clinical evaluation requires a separate evidence base, appropriate instruments, and assessment of clinically relevant functioning and context.
Do self and informant ratings predict future behavior?
Both sources can show predictive associations, sometimes adding information beyond one another. Predictive validity varies with traits, outcomes, measurement conditions, and time intervals. Correlation with a later outcome does not prove the rating caused that outcome. Luan and colleagues, 2019.
Related Articles
Cluster Personality Traits, Temperament and Individual Differences — The scientific foundations of traits, temperament, development, and research on individual differences.
Individual Differences in Psychology: Definition, Examples and Research — How psychological variation is described, compared, and investigated.
Self-Report vs Observer Ratings in Big Five Personality Assessment — The instrument-specific and domain-specific Big Five measurement question.
References
Allik, J., Realo, A., Mõttus, R., Borkenau, P., Kuppens, P., and Hřebíčková, M. (2010). How people see others is different from how people see themselves: A replicable pattern across cultures. Journal of Personality and Social Psychology, 99, 870–882. https://doi.org/10.1037/a0020963
Ashton, M. C., and Lee, K. (2010). Trait and source factors in HEXACO-PI-R self- and observer reports. European Journal of Personality, 24, 278–289. https://doi.org/10.1002/per.759
Ashton, M. C., and Lee, K. (2025). Self- and observer reports of personality. Annual Review of Psychology, 76, 771–795. https://doi.org/10.1146/annurev-psych-020124-115044
Campbell, D. T., and Fiske, D. W. (1959). Convergent and discriminant validation by the multitrait–multimethod matrix. Psychological Bulletin, 56, 81–105. https://doi.org/10.1037/h0046016
Connelly, B. S., and Ones, D. S. (2010). An other perspective on personality: Meta-analytic integration of observers' accuracy and predictive validity. Psychological Bulletin, 136, 1092–1122. https://doi.org/10.1037/a0021212
Funder, D. C. (1995). On the accuracy of personality judgment: A realistic approach. Psychological Review, 102, 652–670. https://doi.org/10.1037/0033-295X.102.4.652
Heine, S. J., Buchtel, E. E., and Norenzayan, A. (2008). What do cross-national comparisons of personality traits tell us? The case of conscientiousness. Psychological Science, 19, 309–313. https://doi.org/10.1111/j.1467-9280.2008.02085.x
Kim, H., Di Domenico, S. I., and Connelly, B. S. (2019). Self–other agreement in personality reports: A meta-analytic comparison of self- and informant-report means. Psychological Science, 30, 129–138. https://doi.org/10.1177/0956797618810000
Lee, K., de Vries, R. E., and Ashton, M. C. (2024). Self/observer agreement in personality assessment by observers' relationship types. Journal of Research in Personality, 113, 104529. https://doi.org/10.1016/j.jrp.2024.104529
Luan, Z., Poorthuis, A. M. G., Hutteman, R., Denissen, J. J. A., Asendorpf, J. B., and van Aken, M. A. G. (2019). Unique predictive power of other-rated personality: An 18-year longitudinal study. Journal of Personality. https://doi.org/10.1111/jopy.12413
McAbee, S. T., and Connelly, B. S. (2016). A multi-rater framework for studying personality: The trait-reputation-identity model. Psychological Review, 123, 569–591. https://doi.org/10.1037/rev0000035
McCrae, R. R., and Terracciano, A. (2005). Personality profiles of cultures: Aggregate personality traits. Journal of Personality and Social Psychology, 89, 407–425. https://doi.org/10.1037/0022-3514.89.3.407
Vazire, S. (2010). Who knows what about a person? The self–other knowledge asymmetry model. Journal of Personality and Social Psychology, 98, 281–300. https://doi.org/10.1037/a0017908
Vazire, S., and Mehl, M. R. (2008). Knowing me, knowing you: The accuracy and unique predictive validity of self-ratings and other-ratings of daily behavior. Journal of Personality and Social Psychology, 95, 1202–1216. https://doi.org/10.1037/a0013314