Abstract
Recent studies have highlighted the important role of horizontal spatial information for many aspects of face processing, such as face detection (Balas, Schmidt & Saville, 2015), face identification (Goffaux & Dakin, 2010), and facial expression recognition (Huynh & Balas, 2014). One study has also reported an association between horizontal tuning for faces and face recognition ability (Pachai, Sekuler, & Bennett, 2013). However, these measures were obtained within the same task, which could have led to an overestimation of the true correlation. Therefore, in this study, horizontal tuning for faces and face processing ability of 37 subjects were measured with independent tasks. A face ability score was extracted based on performance in tree well known face processing measures (Cambridge Face Perception Test, Cambridge Face Memory Test +, and Glasgow Face Matching Task), using principal component analysis. Subjects also completed a task (600 trials) in which they were asked to identify face stimuli randomly filtered with orientation bubbles (Duncan et al., 2017). This method allowed us to extract individual orientation profiles for faces, and in turn, horizontal tuning scores. Orientation profiles were extracted on a subject basis by computing a weighted sum of orientation filters across trials, using standardized accuracies as weights. Horizontal tuning was then calculated as the weighted sum of orientation profile vectors dot-multiplied with a Von Mises distribution (FWHM = 42 deg) centered on the -90 deg horizontal axis. We then measured the association between horizontal tuning for faces and face processing ability scores, and observed a significant positive correlation, r= 0.4, CI 95%= [0.13; 0.64], p<0.05. Importantly, this relation could not be explained by factors such as horizontal tuning for cars, object-processing ability, or low-level sensitivity to horizontal gratings. Our results further reinforce the hypothesis according to which horizontal spatial structure is crucial for face processing.