Face detection deficits in Prosopagnosia assessed by eye-movements and detection thresholds Xiaokun Xu 1* , Irving Biederman 1,2 1 Department of Psychology, 2 Neuroscience Program, University of Southern California ( * xiaokunx@usc.edu) Background Conclusions Prosopagnosics have been reported not to have deficits in face detection (de Gelder & Rouw, 2000; Rossion et al., 2003). Would more rigorous testing, however, reveal face detection deficits? And if so, how could these deficits be characterized? Reference: Crouzet, S., Kirchner, H., & Thorpe, S. (2010). Fast saccades toward faces: face detection in just 100 ms. Journal of vision. 10, 16.1-17. Dakin, C., Hess, F., Ledgeway, T., & Achtman, L. (2002). What causes non-monotonic tuning of fMRI response to noisy images? Current biology. 12, 476-477. Horner, A. J., & Andrews, T. J. (2009). Linearity of the fMRI response in category-selective regions of human visual cortex. Human Brain Mapping. 30, 2628-2640. Oppenheim, A., & Lim, S. (1981). The importance of phase in signals Proceedings of the IEEE. 69, 529-541 Rossion, B., et al. (2003). A network of occipito-temporal face-sensitive areas besides the right middle fusiform gyrus is necessary for normal face processing. Brain. 126, 2381-95. Sadr, J., & Sinha, P. (2004). Object recognition and Random Image Structure Evolution. Cognitive Science. 28, 259-287. Torralba, A., & Oliva, A. (2003). Statistics of natural image categories. Network: Computation in Neural Systems. 14, 391–412. Watson, A. B. & Pelli, D. G. (1983) QUEST: a Bayesian adaptive psychometric method. Percept Psychophysics. 33, 113-20. Supported by NSF BCS 04-20794, 05-31177, 06-17699 to I.B. Exp1. Saccade Choice RTs to Faces vs. Vehicles Reveal Deficits in Face Detection • Single detection task: When only a single stimulus (face or vehicle) randomly appeared in one of the four possible locations, saccades were fastest and almost error free for both MJH and Controls, indicating that he does not have a general deficit in saccading to a stimulus. • Choice task: With a simultaneous display of a face and a vehicle, controls replicated the Cruzet et al’s (2010) result of higher saccade accuracy and shorter saccadic RTs for faces than vehicles, regardless of the instruction (saccade to a face or vehicle in separate blocks). • In contrast, MJH did not showed an advantage in saccade to faces. His saccadic RTs for faces and vehicles were identical. His accuracy was close to chance for both types of targets, with slightly higher accuracy for faces, suggesting a small residual preference of faces. Exp 2. Detection Threshold Defined by Phase-Integrity Exp 3. Detection with Equal Power Spectrum • Because shape identifiability is largely a function of spatial phase (Openheim & Lim. 1981), by introducing external noise into the phase spectrum, we could vary the detectability of faces and cars (Sadj & Sinha, 2003; Dakin et al., 2002). • The QUEST method is used to measure the threshold of phase SNR to achieve 75% detection accuracy in a 2AFC task with two exposure durations. Both controls and MJH showed a lower detection threshold for faces than cars in terms of phase SNR. When scaled according to the S.D.s of the controls (to yield a z-score ), it was apparent the MJH’s detection threshold was markedly higher than that of controls, especially at the 100 msec exposure, suggesting a deficit in face detection. • MJH’s threshold was higher than controls for both type of stimuli and durations. The Z score showed that MJH’s threshold was markedly higher than controls at both durations for faces, but not so much for cars. • In comparison with Exp. 2, his (absolute) detection threshold was increased for faces, but reduced for cars, especially for the 100ms exposure duration. • It is possible that MJH was relying more on the power spectra than control subjects, such that equalizing the spectrum of faces and cars brought his detection of cars close to that of controls, whereas his detection of faces remained significantly impaired. • Normal subjects’ perception of faces and their FFA’s activation are correlated with the visibility of faces, which could be manipulated by phase- scrambling (Horner & Andrews., 2009). In contrast, manipulating power spectrum did not significantly affect face detection in normal subjects, but did so in MJH. • MJH indeed manifests a deficit in face detection (in addition to his severe face identification deficit), reflected in his higher detection threshold, and the lack of preferential saccades to faces. These impairments are likely a consequence of his bilateral occipito-temporal lesions. • MJH, a 46-year-old prosopagnosic with bilateral lesions to (what would be) FFA and OFA but with an intact STS suffered from a fall at age 5. • Has normal contrast sensitivity and expresses no subjective difficulty in face detection, suggesting that his lesions to posterior face areas might not mediate face detection (as opposed to face individuation). Torralba & Oliva (2003) reported that the statistic of the power spectrum could predict detection of animals in natural images. Could the lower SNR threshold for faces vs. cars be a function of the power spectrum difference between them? We tested this hypothesis by using the grand mean power spectrum of all stimuli in the phase-blending process, and then repeated the procedure from Exp 2. Again, controls showed lower thresholds for faces than cars, suggesting only a minimal contribution of power spectra to face detection in normal subjects. . Scan this for PDF download