What does research say makes a face attractive?
Three properties recur consistently across studies, though the picture is more complicated than the popular summaries suggest.
Averageness. Composite faces — produced by digitally averaging many individual faces — are reliably rated more attractive than most of the individual faces used to make them. This is counterintuitive but well replicated since Francis Galton's nineteenth-century photographic composites. One explanation is processing fluency: average faces are easier for the visual system to process, and ease of processing is experienced as pleasantness. Another is that averageness signals genetic diversity or developmental stability.
Symmetry. Faces closer to bilateral symmetry tend to be rated higher. The proposed reason is that developmental disturbance — illness, malnutrition, genetic load — produces asymmetry, so symmetry signals developmental stability. The effect is real but smaller than commonly claimed, and some studies find it weak once averageness is controlled for. Nobody is actually symmetrical, and manipulated perfectly symmetrical faces often look unsettling.
Sexual dimorphism — features typical of one sex. Findings here are the least consistent. Preferences for masculine male features vary by context, culture, and the rater's own circumstances, and several widely reported effects have replicated poorly.
What complicates all of it:
Familiarity matters substantially. The mere exposure effect means repeated exposure increases liking, so people you see often become more attractive to you.
Culture and era shift standards considerably, particularly for features other than the three above.
Non-facial factors dominate in practice. Expression, eye contact, perceived warmth, and — in real interactions — personality and behaviour affect attractiveness ratings substantially. Studies find ratings of the same face change markedly with expression.
Individual variation is large. Averaged across many raters, patterns appear; any individual's preferences deviate considerably.
Beware overclaiming. Face-perception research has had replication problems, and "golden ratio" claims have no scientific basis.