August 2016
Volume 16, Issue 12
Open Access
Vision Sciences Society Annual Meeting Abstract  |   September 2016
The Influence of Facial-Feature Correlations on Face Perception
Author Affiliations
  • Carl Gaspar
    Center for Cognition and Brain Disorders, Hangzhou Normal University, China
Journal of Vision September 2016, Vol.16, 726. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Carl Gaspar; The Influence of Facial-Feature Correlations on Face Perception. Journal of Vision 2016;16(12):726. doi:

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements

One interpretation of holistic processing is that experience with faces leads to an interdependence of facial-feature appearance. But what would this perceptual interdependence be based on? Can we expect, on average, thick lips when we see round eyes? Here, we ask whether correlations between the top (brows and eyes) and bottom (nose and mouth) of the face are reliable and salient enough for observers to make explicit judgements about the match between the top- and bottom-halves of an unfamiliar, but own-race face. Two experiments used wavelet-based features that greatly diminished textures surrounding feature boundaries. All observers and all stimuli were native Chinese. The 'shape' experiment tested the accuracy of discrimination between faces composed of false and true combinations (bottom-halves were drawn from the same face). 12 observers judged 200 pairs of faces each. An 'adjust-and-match' paradigm ensured that abnormalities in the false combinations were absent, and that shape was critical. Prior MDS testing ensured high perceptual distinctiveness between each top-half. Binomial tests show only 3 subjects chose true-faces above chance. But strangely, half the subjects (6) incorrectly selected false-faces significantly beyond chance. The 'size' experiment measured the precision with which a face top could be manually scaled to match the size appropriate for a fixed-size bottom (8 observers), or vice versa (3 observers). Overall size, and the relative size and position of the adjustable half were randomized. Precision analysis compared true and adjusted log-distances: 10 different upper-face, or 5 different lower-face values. Scaling precision was measured by the mean distance along the 2nd principal component. Precision was impressive: 7 of the 11 observers had a precision less than the minimum predicted by any of our 50 top-to-bottom distance ratios.

Meeting abstract presented at VSS 2016


This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.