Purchase this article with an account.
Elinor McKone, Galit Yovel; A single holistic representation of spacing and feature shape in faces. Journal of Vision 2008;8(6):163. doi: 10.1167/8.6.163.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Background. A common view is that spacing information in faces (i.e., distance between centres of the major feature) is coded differently from local feature information (e.g., shape or colour of eyes), with only the former being considered configural. Classically, this view was supported by findings that, while sensitivity to spacing changes is substantially impaired by inversion of the face, inversion had little or no effect on perception of feature changes. Method. We reviewed the 17 published studies relevant to this claim. Results. The size of the feature inversion effect varies substantially across studies. Several studies show very large features inversion effects - both relative to the response scale (i.e., independent of spacing inversion effects) and relative to spacing inversion effects in the same study. We evaluated various proposals that have been made to explain discrepancies across studies. We found patterns of feature inversion effects could not be explained by task, by whether feature trials were blocked or mixed with spacing trials, or by whether performance level for upright was matched to that for spacing. We found a strong relationship between the size of the feature inversion effect and the extent to which the feature change includes colour/brightness: dramatic colour-only changes produce no inversion effect while shape-only changes produce large inversion effects. We also found evidence of stimulus set size effects: for moderate colour changes, inversion effects can occur with large stimulus sets but are absent with small sets (e.g., the “Jane” faces). Overall, feature tasks produce no inversion effect only when the task can be easily solved by attention to nonface information. Conclusions. A holistic representation of face identity codes both spacing-between-blobs and exact feature shape. We review fMRI evidence suggesting the location of this convergent representation might be the Fusiform Face Area.
This PDF is available to Subscribers Only