Halo Effect: How One Good Trait Colors Everything Else You See

People form an overall impression of a person quickly, from whatever information arrives first or stands out most, and then read the rest of the evidence through it. A trait that is central, such as warmth, does not sit alongside the other traits in a list; it changes what the other traits mean. An intelligent and warm person is wise. An intelligent and cold person is calculating. The word intelligent is the same in both cases, but the impression it produces is not.

Edward Thorndike named the effect in 1920 after noticing that army officers who rated the men under them gave correlated scores on physique, intelligence, leadership and character, as if each man were simply good or bad overall. He called it a halo because the overall impression seemed to surround every specific judgment. Solomon Asch showed in 1946 that changing a single central trait changed the whole impression, though he read the result as an organized impression rather than a simple halo error, and Nisbett and Wilson showed in 1977 that people can be unaware of it: their participants said the lecturer's warmth had not affected their ratings of his appearance, mannerisms and accent, even though those ratings had moved with it.

The effect has a dark twin, sometimes called the horn effect, in which one negative impression spreads the same way. It also runs on appearance. Dion, Berscheid and Walster found in 1972 that people judged attractive from a photograph were also expected to have more desirable personalities, better jobs and happier marriages, a pattern they summarized as "what is beautiful is good."

What makes the halo hard to notice is that the judgments it produces feel like observations. The rater does not experience choosing to see generosity; the generosity seems to be there. One way to see it is to compare the ratings of people who differ in one quality and nothing else, which is what Asch did with two groups and what the demo above adapts for one reader. A gap in your two ratings is consistent with the word having moved them, though one reader's pair of ratings cannot separate the word from ordinary variation in how a scale gets used; the evidence that the word is the cause comes from Asch's comparison between groups. Whether such a gap is an error is a separate question, since warmth may genuinely say something about generosity.

Solomon Asch's "Forming impressions of personality" appeared in the Journal of Abnormal and Social Psychology in 1946. In the first and best-known experiment, students heard a list of seven traits describing a person: intelligent, skillful, industrious, warm, determined, practical, cautious. A second group heard the same list with cold in place of warm. Each group then wrote a short sketch of the person and chose, from 18 pairs of opposite traits, which of each pair fitted the person better.

The single word moved most of the judgments. Of those who heard warm, 91 percent chose generous rather than ungenerous; of those who heard cold, 8 percent did. Good-natured was chosen by 94 percent in the warm group and 17 percent in the cold group; humorous by 77 percent against 13; wise by 65 percent against 25. Some traits did not move: reliable, important, good-looking and persistent were chosen at nearly the same rates in both groups. Asch concluded that warm and cold were central traits that organized the whole impression, while others were peripheral.

A follow-up replaced warm and cold with polite and blunt. The sketches barely changed, which showed that centrality was a property of the particular trait in the particular context, not of any pleasant or unpleasant word. Asch also showed an order effect: a list that ran from positive to negative traits produced a more favorable impression than the same list reversed.

Harold Kelley took the manipulation out of the laboratory in 1950. Students in three sections of a psychology course were handed a short description of a guest instructor before he arrived; within each section, half of the descriptions called him "very warm" and half "rather cold," and two instructors took the sections between them. The instructor then led a 20-minute discussion. The students who had read warm rated him more sociable, considerate and humorous, and more of them took part in the discussion, 56 percent against 32 percent.

The halo operates wherever a person is judged on several qualities by someone who has formed an overall impression. The rows below are illustrations, not findings from the studies cited on this page.

Because the halo is hard to see from the inside, the counter-measures are structural rather than mental.

• Rate traits separately and in sequence, with the overall rating last, so that a summary impression is not formed before the parts are scored.
• Use specific, behavioral anchors. "Delivered three of four milestones on time" can be checked against a record; "reliable" cannot. Which records get chosen and how they are read can still be colored by an impression, so the anchor is a way to check the basis of a rating rather than a guarantee.
• Compare across people on one dimension at a time. Grading every student's answer to question one before reading question two makes it easier to see whether a mark rests on that answer or on the impression the earlier ones left.
• Mask what you can. Blind auditions, anonymized applications and blind tastings remove the cue that would otherwise start a halo; other cues can remain, so masking narrows the effect rather than ending it.
• Ask what the evidence for each judgment is. If the answer to "why do I think this person is honest" is "they seem warm," the honesty rating is a halo.
• Expect it in yourself. Nisbett and Wilson's participants showed the effect and, when asked, said it had not happened.

The halo effect is one of several ways an impression spreads beyond its evidence, and it is often confused with its neighbors.

• It is not stereotyping, which fills in traits from a person's group membership. The halo fills them in from a quality of the individual, such as warmth or looks, and it applies within a group as readily as between groups.
• It is not the fundamental attribution error, the tendency to explain behavior by character rather than circumstance. The halo is about how traits color each other once attributed, not about where the attribution came from.
• It is not the cheerleader effect, in which people look more attractive in a group than alone. That one is about averaging faces, not about one trait spreading to others.