Abstract
We have shown that prior knowledge impacts our ability to discriminate intact images from noise. For example, images that are more probable (Greene et al, 2015) or more familiar (Yang et al, 2020) were better discriminated than less probable or unfamiliar images, respectively. The current study aims to extend the familiarity findings of Yang et al (2020) to faces and inverted stimuli. In Experiment 1, we used two manipulations of familiarity: famous vs. computer-generated faces and upright vs. inverted faces. The task required participants to discriminate briefly presented and masked intact images from scrambled (diffeomorphed) versions of the images. A 2-way repeated-measure ANOVA on d prime scores revealed significant main effects of fame (F(1,55)=8.95, p=.004) and orientation (F(1,55)=5.84, p=.019), indicating that famous faces were more easily discriminated from noise than CG faces, as were upright faces compared to inverted faces. Interestingly, there was a significant interaction of fame x orientation (F(1,55)=5.13, p=.027), such that the fame effect was larger for upright faces compared to inverted faces, suggesting that famous faces lose their advantage when inverted. To test whether the inversion effect was specific to faces, in Experiment 2 we ran the same intact/scrambled task with upright and inverted logos. The results replicated the face experiment results with main effects of fame (i.e. highly familiar logos) (F(1,22)=23.24, p<.001) and orientation (F(1,22)=12.66, p=.002). However, for logos, there was no interaction of fame x orientation (F(1,22)=2.09, p=.163), suggesting that the interaction in Experiment 1 was specific to faces and may stem from the well-established detrimental effect of inversion on face recognition. Together, these data indicate that, although all the intact images are easily recognizable as intact under normal viewing times, familiarity is a powerful modulator of discrimination under brief exposure.