Purchase this article with an account.
Jonathan Folstein, Kelly Fuller, Thomas DePatie, Dorothy Howard; The effect of category learning on the temporal dynamics of object similarity. Journal of Vision 2015;15(12):1166. doi: 10.1167/15.12.1166.
Download citation file:
© 2017 Association for Research in Vision and Ophthalmology.
Category learning causes objects to become less similar. This is hypothesized to have a diverse set of consequences: 1) categorized objects become more perceptually discriminable, (psychological distance increases) 2) decisional similarity, thought to be a non-linear function of psychological distance, decreases, and 3) early rule-based categorization strategies are supplanted by increasingly rapid similarity-based strategies. To investigate the effect of category learning on similarity at multiple decision stages, a category learning paradigm was combined with a rare target detection ERP task. Participants learned a logical rule for distinguishing two categories of cartoon animals, each with typical members and ambiguous members. ERPs were recorded while subjects categorized the objects. To manipulate the P300 component, one category was designated the rare target category while the other was the frequent non-target, holding feature-frequency constant. Typical members of the target category had to most target features, followed by ambiguous targets and non-targets, followed by typical non-target category members. The initial ERP session was followed by two days of categorization practice, followed by a second ERP session. The results suggested two stages of similarity representation. ERPs from 300 to 330 ms showed a linear pattern of target similarity on both days: positivity decreased linearly from typical targets to typical non-targets, the pattern becoming steeper and more linear during session 2. The peak of the P300, from about 330 to 750 ms, showed a very different pattern. During the first session, typical targets and non-targets were about equal, and more positive than ambiguous targets and non-targets. After category training, a similarity gradient emerged that was similar to the early stage, but curvilinear. These results suggest that 1) similarity is represented in multiple processing stages, 2) similarity decreases with category learning, and 3) similarity based strategies may replace rule based strategies.
Meeting abstract presented at VSS 2015
This PDF is available to Subscribers Only