His technique of digitally manipulating faces first attracted attention four years ago, when he showed that women prefer more masculine faces during the fertile period of their menstrual cycle. Faces were made more masculine by strengthening the jawline and brows and more feminine by widening the face and raising the eyebrows.
Now Professor Perrett is using the same techniques to investigate the connections between facial features and personality. Perrett is using the most widely accepted model of human personality: the five-factor model. This consists of:
Openness to experience - creative, original, independent
Conscientiousness - careful, hard-working, conscientious
Extraversion - affectionate, talkative, sociable
Agreeableness - forgiving, sympathetic, warm
Neuroticism - nervous, worrying, highly strung
For the time being, Perrett has decided to focus his attention on the best understood of the 'big five' personality factors: extraversion and its opposite state, introversion. Extraverts are talkative, fun-loving and sociable, while introverts tend to be reserved, quiet and retiring.
In previous experiments, Perrett and Little have found that digitally altering the masculinity and femininity of a face affects how people perceive aspects of their personality.
"As we manipulate female faces to make them more feminine, people see them as more extravert," says Perrett.
But masculinity and femininity is only part of the story. Pinning down the essence of an introvert or extravert face is more complicated.
Perrett and Little found that there was little data on what constituted an extravert or an introvert face.
However, Perrett and his team came up with an ingenious solution.
After showing a group of volunteers 15 carefully chosen faces, the team asked them to complete a 20-item questionnaire. The questionnaire asked the volunteers to say which faces best represented certain character traits.
The team then carried out a statistical technique known as factor analysis on the results. This allowed them to draw out the features in a face that people regard as extravert and introvert.
Volunteers rate face composites in a university computer laboratory.
With this information, they created average extravert and average introvert faces from the same 15 images by using computer software to amplify some features and suppress others. These composite images were then used to transform other faces, making them either more introvert or more extravert.
"A lot of the things that we're seeing in extravert and introvert faces are transient things like how likely you are to smile," says Dr Tony Little, of the Perception Lab at St Andrews. Indeed, while the withdrawn look of introvert faces is instantly recognisable, extravert faces seem to be fixed in the earliest stages of a grin.