Purchase this article with an account.
Vladislav Ayzenberg, Stella Lourenco; Three-dimensional objects are preferentially categorized using their medial axes. Journal of Vision 2017;17(10):800. doi: 10.1167/17.10.800.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Much work has stressed the importance of shape for object categorization (Wagemens et al., 2008), yet it is unclear what properties of shape allow for successful categorization. Here we tested whether the medial axis (MA), a summary representation of object shape, supports object categorization. There were two goals: first, to create a stimulus set of 3D objects with which to investigate MA processing, and second, to test whether adults preferentially categorize objects using the MA. One hundred and fifty objects were created such that some objects were identical in MA but differed in surface form (SF) and others were identical in SF but differed in MA (Figure 1). Using a forced-choice discrimination task (N = 83), we identified 20 objects whose MA and SF were matched for discriminability. Then, using a match-to-sample task (N = 25), we examined categorization on this novel stimulus set. On each trial, participants were shown a sample object and two choice objects, judging which of the two choice objects was most similar to the sample. Choice objects could match the sample in terms of MA, SF, or both. Critically, some trials consisted of a conflict between MA and SF, forcing participants to choose either the MA or SF for categorization. In the absence of conflict, participants successfully matched objects using both MA and SF information (ps < 0.001, ds > 2.95). In the conflict condition, participants preferentially chose the object that matched the sample in MA (p = 0.003, d = 0.66). These findings suggest that despite sensitivity to SF, 3D objects are categorized using the MA. This work lends support to the hypothesis that the MA is central to both shape perception (e.g., Firestone & Scholl, 2014) and object categorization.
Meeting abstract presented at VSS 2017
This PDF is available to Subscribers Only