Perceptual errors: halo, primacy, Pygmalion, and what the studies really show

Perceptual errors are systematic distortions in how we perceive and judge other people: one overall impression colours every separate trait, the first information outweighs what comes later, and expectations shape what we see. They are the halo effect, the primacy effect, the Pygmalion effect and the rest of the list that turns up in organisational behaviour courses, communication textbooks and performance-appraisal training. What those lists rarely give you is the study behind each name, and how well it has held up.

• Perceptual errors is a working umbrella term, not a fixed catalogue. That is why every list you find has a different length.
• The halo effect is the best evidenced: Edward Thorndike described it in 1920 from army officers' ratings, and Nisbett and Wilson showed it experimentally in 1977.
• Solomon Asch demonstrated the primacy effect in 1946 by reading the same six traits in opposite orders.
• The Pygmalion effect is real but much smaller than its reputation: in the original study the average difference was about 4 IQ points.
• Leniency, severity, central tendency and contrast errors are rating tendencies described in measurement and appraisal practice, not findings of a single famous study.

What counts as a perceptual error

A perceptual error happens when a judgement of a person departs from what that person actually shows, and departs in a consistent direction rather than at random. The German Dorsch dictionary of psychology files the topic under observation error (in English, rater error or observer bias) and defines it as the deviation of an observation from the object's true characteristic. It separates unsystematic errors, which come from inattention or too little information and scatter in both directions, from systematic errors, which push the result up or down. The effects in this article are all of the systematic kind.

English uses several names for the same family, depending on the discipline. Organisational behaviour and interpersonal communication courses say perceptual errors, and communication textbooks also speak of perceptual distortions. Human resources and industrial-organisational psychology say rater errors or rating errors in performance appraisal. Research methods says observer bias. When you meet any of these terms, the effects underneath are largely the same.

Perceptual errors overlap with cognitive biases, but they are not the same thing. A cognitive bias is a systematic error in judgement under uncertainty, such as anchoring or confirmation bias, and it can distort any decision, including ones that have nothing to do with people. Perceptual errors are about judging a person: an applicant, an employee, a pupil, a patient. Some lists mix the two, and a few effects, such as the fundamental attribution error, appear in both. Our guide to ten cognitive biases covers the judgement side; this article covers the errors made when rating people.

Why every list is a different length

Whether a handout names seven, ten or fifteen perceptual errors depends on what it puts under the umbrella. English organisational behaviour lists often add stereotyping and selective perception; appraisal lists stick to the rating errors; some lists pull in confirmation bias and framing. A more useful order than a loose list is the one the Dorsch entry uses: it sorts observer errors by where in the process of observing they arise. Perception-based errors include the halo effect, projection and expectancy effects. Interpretation-based errors include central tendency, the tendency to extremes and contrast. Then come errors of memory and errors in how an observation is written up or reported. The sections below follow that idea.

Errors from the overall impression

Halo effect: one impression colours every trait

The halo effect is the tendency for an overall impression of a person to spill over onto the rating of their separate traits. Someone who comes across as likeable or competent also gets good marks for qualities that have nothing to do with it. The opposite case, where one poor impression drags every rating down, is usually called the horn effect in appraisal training.

Edward L. Thorndike named the effect in 1920 in a paper titled A constant error in psychological ratings. His data came from ratings of employees at two industrial corporations and from ratings made by army officers. One very conscientious officer rated 137 aviation cadets, with explicit instructions to judge each trait on its own. Even so, the intelligence ratings he gave correlated with his ratings of physique, leadership and character at 0.51, 0.58 and 0.64. Thorndike concluded that raters tend towards

Richard Nisbett and Timothy Wilson showed in 1977 that this happens even when raters do not notice it, in The halo effect: Evidence for unconscious alteration of judgments. Students watched one of two videotaped interviews with the same college instructor, who spoke English with a European accent. In one version he was warm and friendly, in the other cold and distant. Those who saw the warm version rated his appearance, mannerisms and accent as appealing; those who saw the cold version rated the same features as irritating. The students were unaware of the influence. According to the paper's abstract, those who saw the cold instructor even believed it ran the other way: they said their dislike of his attributes had lowered their opinion of him, not that their dislike of him had lowered their ratings of his attributes.

In an interview or an annual review, the halo effect looks like this: a candidate who answers the first question confidently gets credited with teamwork and attention to detail that no one has observed.

Errors from order

Primacy effect: first information weighs more

The primacy effect means that information received early shapes the overall impression of a person more than information received later. The classic demonstration is in Solomon Asch's 1946 paper Forming impressions of personality. One group heard a person described as intelligent, industrious, impulsive, critical, stubborn, envious. Another heard the same six traits in reverse order: envious, stubborn, critical, impulsive, industrious, intelligent. The information was identical, yet the two groups formed different impressions. Asch wrote: "Here we observe a factor of primacy guiding the development of an impression".

The same paper contains Asch's experiments with the traits warm and cold, which showed that a single central trait can change the picture of a whole person. For appraisal, the practical warning is that the first minutes of a job interview, or the first weeks of a new employee's work, can carry more weight than they deserve.

Recency effect: the last thing sticks

The recency effect is the reverse: the most recent information weighs most. In rater-error lists it usually stands for the case where the last few weeks before an annual review decide the rating for the whole year. Which of the two effects dominates in a real appraisal cannot be read off Asch's experiments. They show primacy for lists of traits, not a general law that the first impression always wins.

Halo or primacy?

The two are easy to confuse because both describe a first impression taking over. The difference is what does the distorting. In the halo effect, a global impression distorts ratings of specific traits, whatever order you learned them in. In the primacy effect, the order of information decides which pieces shape the impression. A rater can fall for both at once: an early impression (primacy) that then spreads to every trait on the form (halo).

Errors from expectations

Pygmalion effect: smaller than its reputation

The Pygmalion effect is a self-fulfilling prophecy in the classroom: what teachers expect of a child influences how the child develops. Many course materials use Rosenthal effect as a synonym, and so does this article, although the Rosenthal effect is sometimes reserved for the narrower case of experimenters' expectations influencing their results.

The basis is Robert Rosenthal and Lenore Jacobson's 1968 book Pygmalion in the Classroom. At one elementary school, children took a nonverbal IQ test disguised as a "Test of Inflected Acquisition". Teachers were told that about 20% of the children were "late bloomers" about to spurt ahead. In fact, those children had been chosen at random.

Lee Jussim and Kent Harber traced what the results actually showed in a 2005 review of 35 years of research on teacher expectations:

• Both groups gained: the supposed bloomers by about 12 IQ points, the other children by about 8.
• The average difference between the groups was about 4 points, an effect size of about 0.30.
• Almost all of the effect came from the first and second grades, with differences of about 15 and 10 points. It was significant in 2 of 6 grades in the first year and 1 of 5 in the second.

Their conclusion: self-fulfilling prophecies in the classroom do occur, but they are typically small and do not accumulate greatly across teachers or over time. They may be stronger for students from stigmatised groups. And teacher expectations may predict student outcomes more because the expectations are accurate than because they are self-fulfilling. The effect is worth knowing; the popular version, in which a teacher's belief transforms a child's IQ, is not what the data show.

Expectancy effects and prior information

More general than the Pygmalion effect is the expectancy effect: if you expect something, you notice what fits it. A common trigger is prior information, such as a CV, a reference, a previous appraisal or a colleague's remark about a new team member. The Dorsch entry lists the influence of prior information and expectancy effects as separate perception-based sources of error. The difference from the Pygmalion effect is where the change happens. With an expectancy effect, the expectation changes the observer's judgement. With the Pygmalion effect, it also changes the person being judged, through the way the teacher treats them.

Errors from the rater's own yardstick

The next errors do not go back to one famous study. They are tendencies described in measurement and appraisal practice, and taught in rater training. The Dorsch entry confirms central tendency, the tendency to extremes and contrast as categories of their own.

• Leniency error: everyone is rated too high.
• Severity error: the reverse, a yardstick that is too harsh for everyone.
• Central tendency error: the rater sticks to the middle of the scale and avoids clear judgements either way.
• Tendency to extremes: the rater uses the ends of the scale instead.
• Contrast effect: a person is judged against someone else rather than against the standard, typically the candidate interviewed just before.
• Similar-to-me error: people who resemble the rater get better ratings.

Projection: one word, two concepts

In rater-error lists, projection means that the rater reads their own traits, feelings or motives into the person being rated. That is not the projection of psychoanalysis, where it is a defence mechanism: unacceptable impulses of one's own are unconsciously attributed to others to keep them out of awareness. Our guide to twelve defence mechanisms covers that sense. The rating error needs no defence at all, only a rater who infers someone else's experience from their own.

The main perceptual errors in one table

The table gathers the effects above with the name each one goes by and where it comes from. The German column is for readers who meet these terms in German course material, where the umbrella is Wahrnehmungsfehler or Beurteilungsfehler. Terms like these that do not line up neatly between the two languages are collected in our list of psychology terms that do not translate cleanly into German.

What actually reduces perceptual errors

Thorndike's data hold the least comfortable lesson of the whole topic. His officer had been told to rate each trait separately, and the halo appeared anyway. Knowing an error by name does not switch it off.

Where observations are collected for research, the answer is procedure rather than good intentions. The Dorsch entry names flaws in the observation system itself, such as missing or unsuitable categories, as a source of error. It recommends that every observational study check how well its observers agree with each other before the main data collection, and improve that agreement with observer training where needed. It also lists memory as a source of error in its own right. Carried over to interviews and appraisals, that means rating against criteria fixed in advance, writing observations down when they happen instead of reconstructing a year from memory, and comparing your ratings with a second rater where you can.

Keeping the names apart

Perceptual errors are an exam topic where almost everything hangs on the exact name: halo versus primacy, Pygmalion versus expectancy effect, projection as a rating error versus projection as a defence mechanism. Near-synonyms like these stick best when you learn them against each other rather than one at a time, as our guide on how to memorise psychology terminology explains. In Psychology World, you can save the terms you look up to one list and drill them as flashcards, as multiple-choice questions once the list has four or more words, or by typing the answer out, in English and German. What each study mode does.