Why you look different in the mirror, in photos, and to other people

There was a consistent pattern: women consistently preferred their own mirror-reversed image, while their friends and partners preferred the genuine, unreversed picture. There is a well-documented reason for this split, covered in the next section. While the original 1977 samples were biased toward female subjects and small in size, a 2024 follow-up (Suchow et al., 2024) tested men and women together and found the same pattern. Irrespective of gender, everyone preferred their own mirror-reversed reflection. That newer study did not repeat the friend-comparison part of the experiment, so we cannot say whether friends’ preference for the true image holds up in the same way.

The actual test was more strict than either preference in isolation. Both halves of the pair had to choose as predicted simultaneously, with the woman picking her mirror print and her partner picking her true one, which occurred in 45% of pairs in the first study and 50% across the five-trial replication, against the 25% expected by chance (Mita et al., 1977).

This mismatch is not purely psychological. In fact, it has been wired into your phone by design. Every iPhone’s front camera viewfinder shows a mirrored preview, the same reversed orientation as a bathroom mirror to make framing intuitive. However, up to 2020, the photo saved to your camera roll was the true, unmirrored image, meaning your picture never matched the preview you saw. Finally, iOS 14 added a “Mirror Front Camera”, but this only applies to the Camera app’s photos and videos, thus excluding video calls. For example, Zoom mirrors your self-view by default but has always transmitted the true, unmirrored orientation to everyone else on the call; FaceTime does the same.

Later researchers went even further by using stimuli flashed so briefly that participants could not consciously recognize them, finding that the liking boost still occurred. Kunst-Wilson and Zajonc (1980) showed subjects a series of irregular shapes for one millisecond each, revealing that subjects guessed correctly only 48% of the time (no better than chance) when asked to pick out which shapes they had seen before. However, when asked which shapes they preferred, 60% of the time they preferred the previously shown ones, suggesting their brains registered and favored the familiar shape without ever knowing they had seen it.

The usual explanation is evolutionary; a never-before-seen stimulus could easily be a threat, so our brains treat novelty cautiously. However, as the novelty wears off through exposure, our guard slowly drops. Researchers coined this concept ‘perceptual fluency’: your brain finds it easier to process something it has seen before, and this ease of processing is misinterpreted as likeability. We can measure this physiological guardedness directly too: when researchers showed people the same pictures while recording how aroused their nervous system was (via skin conductance response), the response dropped most sharply within the first several viewings, and the difference between pleasant and unpleasant pictures disappeared completely after that same short window (Bradley et al., 1993).

This is not a one-off study either. In 1989, psychologist Robert Bornstein set out to test whether Zajonc’s 1968 findings genuinely held up. His paper “Exposure and Affect: Overview and Meta-Analysis of Research, 1968-1987,” published in Psychological Bulletin, compiled 134 studies, testing 208 separate effects, conducted over the 20 years since Zajonc’s 1968 paper. This spanned diverse labs, participant groups, and types of stimuli, such as words, sounds, images, faces, and others. Across the body of research, people preferred items to which they were more frequently exposed, positioning the mere-exposure effect as one of the most consistently replicated findings in all of psychology. This applied to faces too: identical faces were rated more likeable by groups who saw them more often.

A 2017 re-examination in the same journal (Montoya et al., 2017) went further, statistically modeling 268 separate effects drawn from 81 articles published in the decades after Bornstein’s review. Liking still rose with exposure overall, but it plateaued at around 35 exposures on average across the reviewed studies.