David A. Kenny
April 5, 2021

Chapter 9 of Interpersonal Perception: The Foundation of Social Relationships.

The question of individual difference in accuracy of person perception has been of central interest in the behavioral and social sciences. In an important paper in the field of Interpersonal Perception, David Funder in 1995 proposed the “Good Judge” hypothesis suggesting that some perceivers are more accurate than others.  Additionally, he proposed the “Good Target” hypothesis that some targets are easier to read than others.  Though not specifically discussed by Funder, there might be the “Good Pair” hypothesis which is that some relationships lend themselves to more accuracy than others. Using the Social Relations Model, there are three individual difference questions:

        perceiver: Are some perceivers better at judging targets than other perceivers?

        target: Are some targets easier to judge than other targets?
        relationship: Are some relationship partners easier to judge than others.

The page, largely taken from Chapter 10 of Interpersonal Perception: The Foundation of Social Relationships, discusses these three different areas of individual differences in three different domains. The first is individual differences in the judgment of personality. The second is individual differences in the detection of emotion. The third is individual differences in lie detection. Also discussed is the Social Accuracy Model extents the analysis of individual differences to multiple traits and the extended Social Relations Model or eSRM, which allows for individual differences.

The hypothesis is that some perceivers are better at understanding targets’ personality than are other perceivers.  A natural way to test this hypothesis is to see if judges who good at judging targets in one domain (e.g., accurately judging extroversion) are also good at judging targets in another domain (e.g., accurately judging agreeableness).   Katja Schlegel, Thomas Boone, and Judith Hall in 2016 located 10 such studies and 103 measures of agreement across domain.  They found that if one is good judge in one domain, one is only likely to be a good judge in the other domain 53 percent of the time.

A value of 53 percent agreement across domains is very small and is only statistically detectable because there are so many studies.  To appreciate how low this value is, it is 94 percent for the relationship between two intelligence tests (the Wonderlic Personnel Test and the Wechsler Adult Intelligence Scale) and 88 percent for two measures of self-esteem (the Rosenberg and Single-Item Self-Esteem Scale).  Thus, finding only a 53 percent correspondence in the ability of to understand other people’s personality is pretty darn puny.  There are individual differences, but they are, just as Albright and I claimed 30 years ago, rather weak.

Again we look at Schlegel and her colleagues 2016 meta-analysis.  In their study, 24 studies involving 124 pairs of tests to measure accuracy in the perception of skill at reading the emotions of others.  Examples of tests to measure is the MSCEIT which is to measure a branch of Emotional Intelligence, the Profile of Nonverbal Sensitivity or PONS in which the target is Judith Hall who conducted the earlier mentioned meta-analysis, and Schlegel’s measure the Geneva Emotions Reactions Test or GERT.  If you want to take an online test of your skill in this area you can take the EYEs test (https://huxta.com/theory-of-mind-eye-test/).  The test was developed by Simon Baron-Cohen, the cousin of Sasha Baron-Cohen of Borat fame or infamy, depending on your point of view. 
Schlegel and colleagues find a 60 percent effect.  This is not large but it is much larger than the degree of individual differences in detecting personality and as we shall see in the next section in lie detection. Schlegel and colleagues did not look at target effects, which are generally called sender effects in this literature.   However, several studies do find larger sender effects.

Interesting, the correlation between perceiver accuracy and target accuracy in judging emotions is all over the place, varying from almost a perfect negative correlation in some studies (good receivers are poor senders) to a nearly perfect correlation in other studies (good receivers are also good senders).  +1 in others.  Hilliary Anger Elfenbein and Noah Eisenkraft found 40 studies and the average  relationship between sending and receiving was slightly positive: good receivers are slightly more likely to be good senders.  However, when they divided studies, they a positive correlation for nonverbal behaviors elicited as intentional communication displays but zero for spontaneous, naturalistic display.


Relatively few studies have looked for relationship effects in accuracy of emotional expression studies.   If there were large relationship effect variance, the PONS would be a poor measure as it has just one target.  The one study was conducting by Elfenbein and colleagues (2006) has and she found them.

A meta-analysis of lie detection was conducted by Charles Bond Jr. and Bella DePaulo in 2008 and provided a wealth of information that will be reviewed.  I consider myself very fortunate to have collaborated on other projects with both of these scholars and they are master scholars.  A meta‑analysis is quantitative summary of multiple studies.  The Bond and DePaulo meta-analysis involves 247 studies involving 19,801 perceivers and 2,945 targets.

In measuring lie detection, it is crucial to control for each perceiver’s base rate which is the percentage of time that a perceiver thinks the target is lying.  In everyday life for most of us, we have a very low base rate for lying.  In a 1966 study, Bella DePaulo, the “The Deception Maven,” and Deborah A. Kashy had people of all ages keep a diary of all the lies they told over the course of a week. Most people, she found, lie once or twice a day.  Perhaps this is more lying than we would have thought, but it does document the fact in reality most of the time people are telling the truth.  Because of the low base rate for lying, most of the time we are likely to be wrong when we guess that someone is lying because we think generally that people are not lying.  In a parallel fashion, we shall be right most of the time when someone is telling the truth.  Adjustment for base rates is critical for measuring lie detection.

The first question to ask from this meta-analysis is what is the level of lie detection?  The answer is 54 percent.  Thus, if are asked if someone is lying, you will have a 54.05 percent chance of being right.  This is above chance, but just barely so.

Bond and DePaulo propose four components of lie detection.  The first, and what most people are interested in, is perceiver ability.  Are some perceivers better than other perceivers at knowing which targets are lying and which are telling the truth?  The second, akin to readability, concerns how good a liar someone is, or what will be called target credulity.  Are some targets better at lying, i.e., less likely to be caught, than others?  The third component refers the perceiver’s base rate.  Do some perceivers think everyone is lying and do other perceivers think that everyone is telling the truth?  The fourth component is called demeanor bias.  Some of us appear to be lying, even when we are not, and others of us appear to be telling truth even when we are not.  Just how important are these four components in terms of how large their standard deviations?  

Jeremy Biesanz in 2010 proposed a very general model of measuring individual differences in accuracy which he calls the Social Accuracy Model or SAM.  With SAM, it is possible to measure the level of differences in accuracy due to perceiver, target, and relationship.  SAM is a very complicated statistical model that builds on the Social Relations Model that was introduced in Chapter 1 and has been used in this book.  However, SAM is much gorier statistical model than the SRM, comparable to the amount of gore in a Wes Craven movie.  A key feature of the SAM is that is multivariate in that it looks across multiple traits, e.g., all five of the Big Five, whereas the SRM looks at just one trait at a time.  In SAM, accurate knowledge of a target is measured not only for target’s standing on a personality trait but also the target’s relative standing across traits.  Thus, does the perceiver know that the target is more extroverted than he or she is agreeable?

For SAM, perceivers make judgments of the same targets on multiple traits.  There are two predictors, one being the measure of the truth and the other the mean score on the truth on the given trait.  The effect of the mean judgment is referred to as normative accuracy and the effect of target standing on the trait is referred to as distinctive accuracy.  In Chapter 5 the concept of normative accuracy was introduced.  Normative accuracy refers to ability of a perceiver to know what the average person is like whereas distinctive accuracy refers to accuracy in understanding how targets differ from the average.

With SAM, normative and distinctive accuracy can be measured for perceivers, targets, and relationships, resulting in six different types of accuracy.  Considering normative accuracy for perceivers first, it asks whether a judge knows what the average person is like.   Thus, it examines the correspondence between the perceiver’s profile of average judgments across the traits with the profile of average values of the truth for the targets or the extent to which a perceiver’s judgments match the prototypical pattern of the truth.  For instance, if the targets being judged were on average truly highly extroverted and agreeable, but low on conscientiousness and openness, then perceivers who generally viewed targets in that way would get positive normative accuracy scores.  However, if a perceiver had the opposite pattern (saw targets as not extroverted and agreeable, but high on conscientiousness and openness), then that perceiver would have low (actually negative) accuracy scores.  Distinctive perceiver accuracy measures the extent to which a perceiver knows who has a higher or a lower standing on a given trait relative to the group mean on that trait.  For instance, does the perceiver know who has relatively high extroversion scores and who has relatively low ones?

Normative target accuracy measures the extent to which a target is typical, i.e., conforms to the average profile of the traits on the measure of truth.  Distinctive target accuracy refers to extent that perceivers do especially well at judging the target on a given trait, e.g., perceivers know the target’s relative standing on the trait.  For instance, if the target tends to have high scores on extroversion and agreeableness and low scores on conscientiousness and openness, perceivers on average tend to know this.  

For relationship accuracy, normative relationship accuracy measures the extent to which a given perceiver is able (or unable) to recognize that target is typical, more (or less) so than other perceivers and other targets.  This is not an easy concept to get a handle on.  Distinctive relationship accuracy refers to extent that a given perceiver does especially well (or poor) at judging a particular target on a given trait.  By exceptionally well or poor is meant that perceiver-target pair does better or worse than would be expected by the perceiver’s distinctive accuracy score and the target’s distinctive accuracy score.

Megan Goldring, Taeyun Jung, and I (2024) have proposed extending the SRM to allow for individual differences. We introduce four new parameters:

   Sensitivity or c: Some perceivers have more target variance than do others.
   Prototypicality or a: Some targets have more perceiver variance than do others.
   Differentiation or d: Some perceivers have more relationship or error variance than do others.
   Volatility or v: Some targets have more relationship or error variance than do others.

The model is called the extended Social Relations Model or eSRM. Standardized Sensitivity is the correlation of a perceiver's judgments with the estimated target effect and can be interpreted as a measure of Accuracy. Estimation of the eSRM is not yet completely worked out but we develop a relatively simple approximation, using slopes and standard deviations of errors. I am currently exploring a Bayesian analysis of the eSRM..

In the paper we re-analyze data gathered by Jung (1999) in which 160 participants provided judgments of 40 different targets on 20 different traits. Perceivers were University of Connecticut undergraduates and targets were celebrities who were familiar.  All 160 perceivers also judged how acquainted they were with each target and how much they liked each target. Among the key findings are that there is evidence of extreme responding; that is some perceivers use a wider range of the scale than do others. Women show higher levels of Accuracy than to men and most of this is due to lower levels of relationship and error variance, i.e., lower levels of Differentiation. Also Prototypicality correlates positively with Familiarity and Liking, whereas Volatility correlates negatively.

The model is similar to SAM in that it looks at individual differences, but it is different in that the eSRM looks at a single trait, and SAM looks across traits. Also SAM uses a measure of truth whereas the eSRM does not.

Biesanz, J. C. (2010). The social accuracy model of interpersonal perception: Assessing individual differences in perceptive and expressive accuracy. Multivariate Behavioral Research, 45, 853–885.

Bond, C. F., Jr., & DePaulo, B. M. (2008). Individual differences in judging deception: Accuracy and bias. Psychological Bulletin, 134, 477–492.

DePaulo, B. M., Kashy, D. A., Kirkendol, S. E., Wyer, M. W., & Epstein, J. A. (1996). Lying in everyday life. Journal of Personality and Social Psychology, 70, 979–995.

Elfenbein, H. A., Der Foo, M., Boldry, J. G., & Tan, H. H. (2006). Dyadic effects in nonverbal communication: A variance partitioning analysis. Cognition and Emotion, 20, 149–159.

Funder, D. C. (1995). On the accuracy of personality judgment: A realistic approach. Psychological Review, 102, 652–670.

Kenny, D. A., Goldring, M. R., & Jung, T. (2024).  The Extended Social Relations Model:  Understanding dissimilation and dissensus in the judgment of others.  European Journal of Personality, 27, 114-139.

Schlegel, K., Boone, R. T., & Hall, J. A. (2017). Individual differences in interpersonal accuracy: A multi-level meta-analysis to assess whether judging other people is one skill or many. Journal of Nonverbal Behavior, 41, 103–137.

Book logo
Go to the Interpersonal Perception menu.

Go back to my homepage.