Purchase this article with an account.
Lacey Perry, Leslie Blaha, James Townsend; Reassessing the architecture of same-different face judgments. Journal of Vision 2008;8(6):889. doi: 10.1167/8.6.889.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Given the holistic nature of faces, Wenger and Townsend (2001) proposed that face information processing would be supported by a facilitatory parallel system, meaning that all features are processed simultaneously. However, there is an apparent contradiction between this working definition and Bradshaw and Wallace's (1971) claim that face recognition is a serial process, wherein features from two faces are sequentially compared. Using a same-different face recognition paradigm, they found that as the number of diagnostic face features increased, processing time decreased. Although, these results are predicted by standard serial models, their primitive techniques could not distinguish true serial processing and the parallel models that mimic serial processing. With the methods of systems factorial technology (SFT; Townsend & Nozawa, 1995), we can characterize any human information processing system in terms of architecture, stopping rule, independence, and capacity. SFT provides a rigorous framework for assessing Bradshaw and Wallace's (1971) claim of seriality; given the intuitions of Wenger and Townsend's (2001) hypothesis we predict that same-different face recognition is a process of parallel comparison. In a set of experiments incorporating the double-factorial paradigm, we factorially manipulated the number of critical features, relative similarity of face pairs, and presense/absense of non-diagnostic context features. Different face pairs consisted of identical facial surrounds with internal features morphed to be more (66%) or less (33%) similar. Measuring the same-different judgment response times, we first replicated Bradshaw and Wallace's data. Information processing modeling revealed a positive survivor interaction contrast and an over-additive mean interaction contrast. This is the signature combination of parallel, self-terminating processing. Thus, our results resolve the apparent contradiction and add a new dimension to Wenger and Townsend's (2001) configural processing system; not only are the features within a configural object processed simultaneously, but also the comparisons between configural objects are performed in parallel.
This PDF is available to Subscribers Only