Abstract
It is now well established that face identification is supported specifically by information contained in the horizontal orientation band (Dakin and Watt, J Vis 2009), and we have shown that observers who selectively utilize this information demonstrate greater overall identification performance (Pachai et al, Front Psychol, 2013). Still unknown, however, is the extent to which the horizontal tuning supporting identification is deployed throughout stimulus exposure. To this end, we examined the rapid processing of identity information by dynamically masking the orientation content of a face in a 10AFC identification paradigm. In five 53ms intervals of a 265ms stimulus presentation, we generated independent white noise masks for which we randomly modulated the power contained in the horizontal and vertical bands. Logistic regression models assessed the effect of noise orientation power and time interval on identification accuracy for each observer. These models fit the data well, so we conducted an analysis of individual regression coefficients as a function of time and orientation. This analysis revealed significant effects of time, demonstrating the greater influence of information collected earlier in presentation, and orientation, demonstrating the greater influence of horizontal structure relative to vertical. On average, the effect of orientation was significant in all time windows except the first (i.e., 0-53ms), which suggests that horizontal tuning evolves through stimulus exposure. However, some individual observers exhibited horizontal tuning even in the earliest time window, highlighting the importance of individual differences. We currently are examining the relationship between the temporal evolution of orientation tuning effects and individual differences in speeded identification accuracy. Together, these results further demonstrate the importance of horizontal structure for accurate face identification and suggest that diagnostic orientation structure for face perception is extracted within the first 100ms of stimulus presentation.
Meeting abstract presented at VSS 2014