Abstract
Prior expectation may influence our perception of people at the first meeting. For example, it has been shown that we are more likely to categorise a person as 'male' or to estimate their age as closer to our own. This could be due to perceptual effects and/or cognitive response bias. Here we use a forced choice task to reduce cognitive influence and a Bayesian modelling approach to estimate perceptual biases inherent in our decisions about the sex and age of faces. Two identities were presented to participants simultaneously, both of the same age or sex. One face was presented for 1000msec while the other was presented for either 250ms or 500ms (more/less uncertainty). Participants were asked to indicate which face appeared more male/female (n=40) or younger/older (n=20). The proportion of trials on which the briefer stimulus was chosen as more male/older indicated that under conditions of increasing uncertainty participants were more likely to respond 'male' and 'older'. For both judgements this is consistent with the operation of a Gaussian prior with a peak towards male and older faces and with the prior becoming more influential under conditions of increased uncertainty. This finding was supported by the results of a rating task. Here uncertainty was induced by jittering the phase spectrum of the face images. The equivalent shift in the rating of the noiseless version of each face that would be required to match the ratings of the phase jittered faces was also consistent with a Gaussian prior with a peak towards male and older faces. This demonstrates a biased expectation, operating at a perceptual level, that faces will be male and older. As this effect is not due to cognitive response bias, it represents a true inaccuracy in the experienced percept which we anticipate will be impenetrable to cognitive control.
Meeting abstract presented at VSS 2015