Cheerleader Effect: Why Faces Look Better in a Group Than Alone

The explanation Walker and Vul offered joins three findings that were already established separately. The first is ensemble coding: shown a group of faces for a moment, the visual system quickly extracts an average of them, the average identity, expression or gender. The second is hierarchical encoding in visual memory: individual items are stored in relation to the group summary, so that each face is remembered as somewhat closer to the group average than it actually was. The third is that average faces are attractive, a finding Langlois and Roggman reported in 1990: composite faces made by blending many individuals were rated higher than most of the faces that went into them.

Put together, the three predict the cheerleader effect. A face seen in a group is pulled, in memory, toward the group's average face; the average face is more attractive than the typical member; so the remembered face gains a little attractiveness it did not have on its own. The gain should be small, because the pull toward the average is small, and it is: about 1.5 to 2 percent of the rating scale in Carragher and colleagues' 2018 and 2019 measurements.

Whether that is the right explanation is less settled than the effect itself. Carragher, Thomas, Gwinn and Nicholls tested it directly in 2019 by varying what the group was made of. Groups that could be summarized into an average face produced the largest effect, as the account predicts. But faces shown among non-face images also gained a little, which averaging into a face cannot explain, and the authors concluded that the evidence for hierarchical encoding is limited. The effect is real; the mechanism is a live question.

Two things follow. A point or two on a hundred-point scale is unlikely to change who in a photo looks attractive; the studies measured an average over many ratings. And in Carragher's 2018 experiments the effect appeared wherever the other faces were placed, so position in the photo does not matter to it. Whether the companions' looks change the size of the gain is a separate question those experiments left open.

Drew Walker and Edward Vul published "Hierarchical encoding makes individuals in a group seem more attractive" in Psychological Science in 2014, online in October 2013. The name of the effect came from a 2008 episode of the television comedy How I Met Your Mother, in which a character claims that a group of women looks attractive together and less so one by one; the paper tested whether the claim was true for individual faces.

Across five experiments, participants rated the attractiveness of faces presented either alone or in a group of the same gender. The design was within-subject: each target face was rated twice by the same rater, once in its group photograph and once as a cropped portrait, with the two presentations separated in the session. The faces were rated higher in the group. The paper reports the increase as consistent across experiments that varied the group's size and composition, and the authors interpreted it through the three-part hierarchical encoding account described above.

The effect has since been reproduced by independent laboratories, though not every attempt succeeded: Ojiro and colleagues ran two direct replications of Walker and Vul's fourth experiment with Japanese participants in 2015 and found a pattern resembling the original in one of them, small and not significant. Carragher, Lawrence, Thomas and Nicholls reported in 2018 that target faces were rated significantly more attractive in every group arrangement they tested, with the increase in the range of 1.5 to 2 percent of the scale. Carragher, Thomas, Gwinn and Nicholls followed in 2019 with three experiments that reproduced the effect at the same size while finding only limited support for the averaging mechanism.

A related but distinct finding, from van Osch, Blanken, Meijs and van Wolferen in 2015, is that a group as a whole is rated more attractive than the average of its members' individual ratings, with a medium to large effect across 33 comparisons. That is a judgment of the group, not of a face within it, and the authors attributed it to attention settling on the most attractive members.

The effect concerns ratings of faces in photographs, and the examples below stay close to that. They are illustrations, not findings from the studies cited here.

There is little to counter in a one- or two-point effect, and little to exploit. The habits below are about reading faces and research honestly.

• Compare faces in the same format. If you are judging between people, as an interviewer or a casting director, look at single portraits for everyone or group shots for everyone, not a mixture.
• Keep the size in mind. An effect of 1.5 to 2 percent of a scale is a small average shift across many ratings; claims that a group makes a person look much better are not what the research found.
• Do not take the mechanism as settled. The averaging account is elegant and partly supported; the 2019 tests found effects it cannot explain.
• Separate the two findings. A group rated as more attractive than its average member is a different result from a face rated higher inside a group, with a different proposed cause.
• Notice when a television joke is doing the work. The effect got its name from a comedy and its evidence from a laboratory, and the show's version, that a group looks good and its members do not, is not what the study showed.

Several effects concern how context changes the judgment of a face, and the cheerleader effect is often described in terms that belong to its neighbors.

• It is not the halo effect, in which one favorable quality of a person, such as warmth or looks, raises judgments of unrelated qualities. The cheerleader effect changes the rating of one quality, attractiveness, by changing what surrounds the face.
• It is not the contrast effect, in which a face looks better or worse next to a comparison. The cheerleader effect compares a face in a group with the same face alone, not with its neighbors. Contrast may still contribute: the 2019 study drew companions at random and could not test it, and its authors noted that the companions' attractiveness may change the size of the gain.
• It is not the mere exposure effect, in which repeated viewing makes a face more liked. In the 2014 design every face was seen alone and in a group, so exposure was equal.
• It is not the group attractiveness effect of van Osch and colleagues, a judgment of a whole group rather than of a member within it, though the two are often run together in popular accounts.