Abstract
Recent work by Grill-Spector and Kanwisher (2005) discovered a tight temporal coupling between object detection and categorization. Specifically, in an experiment where participants performed a task requiring both a detection and categorization judgment on the same trial, they found entirely dependent performance between detection and categorization. Categorization did not occur without successful detection and vice versa. Grill-Spector and Kanwisher concluded that object detection and basic-level categorization are both carried out by an initial stage of visual processing that precedes categorization at more subordinate and abstract levels.
In the present study, we decoupled performance on detection and categorization by manipulating the level of specificity of the categorization judgment. The categorization in Grill-Spector and Kanwisher's study (person vs. car) was at the superordinate level (not the basic-level, as asserted in their paper), a level of categorization that can occur very rapidly (Thorpe, Fize, & Marlot, 1996) and is available early in processing (Rogers & Patterson, 2007). With superordinate-level categorization, we found a dependent link between detection and categorization, replicating Grill-Spector and Kanwisher. In contrast, when the categorization decision was at the basic level (e.g., car vs. boat), the coupling between detection and categorization was severed such that successful detection occurred without successful categorization. Evidence of this decoupling of detection and basic-level categorization was observed across a range of presentation durations.
Similar dissociations of the time courses of detection and categorization have previously been shown with stimulus inversion and degradation (Mack et al., in press). The current study significantly extends earlier results by providing a dissociation with fully intact and canonically-oriented stimuli.
The closely linked performance between detection and superordinate-level categorization remains a provocative finding. However, our selective manipulation of categorization performance casts doubt on any claim for an initial processing stage that both detects objects and categorizes objects at the basic level.