The dynamics of pattern identification and categorization

Abstract

Whether identification and categorization of visual stimuli are supported by separate or identical neural circuits remains a matter of debate. Similarly, the dynamics of these processes are poorly understood: Does identification precede categorization, or vice-versa? To answer these questions, we measured how fast visual patterns are identified and categorized. In our identification task, subjects were initially trained to recognize a given pattern. In subsequent sessions, they had to indicate whether each of a series of patterns was identical to the learned pattern or not. In the categorization task, subjects trained with similar patterns defining an arbitrary category. They then had to indicate whether subsequently presented patterns belonged to the category or not. We measured accuracy and reaction times in both tasks. Performance improved during the first 3 to 4 sessions, then stabilized. While accuracy in both tasks showed similar dynamics with practice, the reaction times behaved differently: In the first session, the average reaction time for categorization was 716±193 ms, and 550±78 ms for identification. With training, categorization reaction times decreased more rapidly than in the identification task. Both types of reaction times then converged around 467 ms (500±81 ms categorization; 434±24 ms recognition). These results suggest that identification and categorization are two independent processes.