Purchase this article with an account.
Mirta Stantic, Michael A Cohen, George A Alvarez; Differences in representational geometries of prosopagnosics and neurotypical controls. Journal of Vision 2019;19(10):23. doi: https://doi.org/10.1167/19.10.23.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
In the human visual system, broad categories of visual stimuli form a representational space remarkably consistent across people. Previous research studied the relationship between these neutrally and behaviorally derived representational spaces – “representational similarity matrices” (RSMs) – and suggested methods for analyzing these similarity structures. The current study explores similarity of RSMs between two developmental prosopagnosics and a group of neurotypical controls. Participants (Ncontrols = 78, Mage = 22.41, SDage = 3.65, range = 18–35; Nprosopagnosic = 2, ages: 52, 51) completed two types of visual search task. In the between-category trials, they searched for an object from a particular category presented amongst objects from a different category (e.g. search for a body surrounded by chairs as distractors). In within-category trials, they searched for a particular exemplar of one category amongst other exemplars of the same category (e.g. search for face A surrounded by distractors of face B). Representational similarity matrices were then constructed based on pairwise similarities inferred from reaction times on each task (such that shorter RTs implied that categories are more similar as they are less confusable). Prosopagnosics’ accuracy did not differ from the control group on any task. Crucial differences emerged in comparing the organization of RSMs. For one prosopagnosic, the within-category RSM was significantly different from that of an average control participant, but only for faces. The other prosopagnosic participant showed an RSM similar to the average control participant on all tasks. This finding suggests that the face recognition impairment in some prosopagnosics might stem from a differently organized feature space for faces, but not for other categories of visual stimuli. We discuss potential clusters of prosopagnosic population characterized by different strategies (e.g., compensatory mechanisms more closely aligned to holistic face processing) or neural activation patterns and explore further ideas for representational similarity analysis research.
This PDF is available to Subscribers Only