Here’s how our visual system decides how cute a baby is

By Emma Young

What makes one baby seem cuter than another (parental bias aside)? Large and round eyes, a small nose and mouth, a high and protruding forehead, chubby cheeks and soft skin have all been associated with cuteness, not just in babies but puppies, kittens, dolls and Japanese anime and manga characters. But now a new paper in Emotion suggests that another factor has an impact: the “spatial frequency” of what we see.

Our visual system uses spatial frequencies to rapidly process variations in relative light and dark in an image. “Low” spatial frequencies convey coarse information — the location of a dark eye against pale skin, for example. “High” spatial frequencies convey fine detail such as edges. But other frequencies are important, too; research shows that medium spatial frequencies are most useful for recognising faces, for example. Spatial frequency (SF) processing is the basic mechanism for visual analysis, write Mengni Zhou at Okayama University, Japan, and colleagues. “Therefore understanding the effect of SFs is essential for a better understanding of the processing of cuteness,” they argue.

The team first asked 32 participants from Okayama University to rate the cuteness of 40 black and white images of infant faces (aged up to one). Using these ratings, the researchers assembled a set of four faces from the bottom cuteness quartile. They did the same for the second, third and top cuteness quartiles, generating a total of four sets. (Fourteen of the infants were identified as ‘Caucasian’, one ‘Asian’ and one ‘Indian’).  The team then manipulated the spatial frequency of each of the faces, generating four filtered versions, ranging from low to high SF.

Example of an unfiltered face (A), and filtered faces ranging from low SF (B) to high SF (E). From Zhou et al (2022)

In the next stage of the study, a fresh group of 30 students from the same university used a 7-point scale to rate each of these filtered images, as well as the originals, for cuteness.

As the team expected, the unfiltered original photos got higher cuteness ratings than their filtered versions. But within the filtered versions, some were seen as cuter than others.

The analysis showed that faces with the lowest SF always got the lowest cuteness ratings. When the team talked to participants after the study, some said that these faces looked “eyeless with dark empty orbits.” These versions were seen as “distasteful, even scary,” the team writes

In comparison with low SF, faces with medium and high SF got relatively high cuteness ratings. This is probably because the facial features associated with cuteness (small nose, chubby cheeks, etc) are more obvious at relatively high spatial frequencies, the team notes. 

But when the cuteness of the unfiltered faces was high, the medium SF versions got better ratings than the high SF versions. This might relate to earlier findings that medium SFs are most important for facial recognition, the team suggests. “One possible explanation is that the processing mechanism of positive cuteness” — i.e. finding a baby’s face definitely cute — “is strongly related to face identity,” they suggest.

“In the future, eye tracking and neuroimaging studies are also necessary to verify these results and explore the mechanism of cuteness processing in depth,” they add. For now, though, the team argues that this work provides “direct evidence that the perception of infant facial cuteness is affected by SF information”.

It’s hard to see how this finding might have any real world impacts — except perhaps in the production of cartoon, manga and anime characters. My own kids enjoy manga and anime. I — no doubt like the participants in this study, who were all at a Japanese university — am well aware of anime caricatures of cuteness. But perhaps SF could be manipulated to influence our perceptions of a character’s cuteness, too. 

Spatial frequencies affect cuteness perception of infant faces

Emma Young (@EmmaELYoung) is a staff writer at BPS Research Digest