Abstract
Using the spatial frequency (SF) Bubbles technique, we recently revealed that the same SFs are used for the identification of upright and inverted faces (Willenbockel et al., in press; see also Gaspar, Sekuler, & Bennett, 2008). In these articles, the faces were presented through an elliptical aperture hiding contours. Given that contours do contain information useful for face identification, real-world differences between upright and inverted face SF processing might have been missed. Here, we examined the role of contour information in the SF tuning of upright and inverted face identification using SF Bubbles. We created a bank of 20 faces, and each face was randomly assigned either to set A or to set B. Six participants saw the faces from set A with contours and the faces from set B without contours (shown through an elliptical aperture), whereas six other participants saw the faces from set A without contours and faces from set B with contours. On each trial, a face was selected and its SFs were sampled randomly (for details, see Willenbockel et al., in press). Participants completed one thousand trials in each condition. Multiple linear regressions were performed on the random SF filters and response accuracy. Without contours, we closely replicated Willenbockel et al.: the same SFs correlated with accurate identification of upright and inverted faces (a single band beginning at ∼6 cycles per face (cpf) and ending at ∼15 cpf). The presence of contour information led to a similar increase in the diagnosticity of low spatial frequencies, irrespective of face orientation; and to a decrease in the diagnosticity of higher spatial frequencies for inverted faces (upright faces with contour: a single band beginning at ∼2.3 cpf and ending at ∼16.5 cpf; upright faces without contour: a single band beginning at ∼4 cpf and ending at ∼20 cpf).