Abstract
Categorization is very efficient. However it is unclear how multiple simultaneously active visual categories interact. Such interactions were tested in categorization tasks in which participants ignored a flashed prime, but responded to a 100ms target presented after an SOA of either 80 or 180ms. The prime’s category was either the same as the target category (congruent condition) or different (incongruent condition). In both experiments 1 (bird/non-bird categorization) and 2 (animal/non-animal categorization), responses to congruent trials were faster and more accurate than those to incongruent trials. More importantly, we tested two different types of incongruent primes: related and unrelated to the target. In Exp1, related incongruent primes belonged to the same superordinate category (e.g. dog), but unrelated ones did not (e.g. car). Related incongruent primes caused more interference than unrelated ones at both SOAs. In Exp2, each image belonged to one of three subclasses (air, water, ground). Whereas related primes belonged to the same subclass as the target, unrelated ones did not. As in Exp1, related incongruent primes caused more interference than unrelated ones, but only at long SOA. To explain these results, we propose a model based on the idea that categories are represented as hierarchical patterns of neuronal activity in infero-temporal cortex. When two objects share attributes (e.g. ‘animal-ness’), their patterns overlap. Since related incongruent primes share more attributes with the target than unrelated ones, their patterns overlap more, increasing the difficulty to determine the target’s category. Furthermore, target and incongruent related primes had far more overlap in Exp1 (e.g. bird primed by dog) than in Exp2 (e.g. bird primed by plane). Accordingly, interference was observed earlier in Exp1 than in Exp2. We conclude that residual activity in infero-temporal cortex affects subsequent categorization depending on the extent of overlap between activity patterns.
Meeting abstract presented at VSS 2014