Spatial frequency tuning of upright and inverted face identification

Previous research suggests that observers use information near the eyes and eyebrows to identify both upright and inverted faces [Sekuler, A. B., Gaspar, C. M., Gold, J. M., & Bennett, P. J. (2004). Inversion leads to quantitative, not qualitative, changes in face processing. Current Biology, 14(5), 391-396]. Here we ask whether more significant differences between upright and inverted face processing exist in the spatial frequency domain. Thresholds were measured in a 1-of-10 identification task with upright and inverted faces presented in no noise, white Gaussian noise, and in low-pass and high-pass filtered noises with various cutoff frequencies. In Experiment 1, all faces were presented in fronto-parallel view; in Experiment 2, viewpoint varied across trials. Thresholds were higher for inverted faces, but the magnitude of the inversion effect did not vary across conditions or experiments. Moreover, the shapes of the noise-masking functions obtained with low-pass and high-pass noise were the same for upright and inverted faces, did not vary between experiments, and revealed that identification was based on information carried by a 1.5 octave wide band of spatial frequencies centered on approximately 7 cycles per face width. Finally, individual differences in the magnitude of the inversion effect were not related to individual differences in the frequency selectivity of face identification. The results indicate that the face inversion effect for identification judgments is not due to subjects using different bands of spatial frequencies to identify upright and inverted faces.