APA PROOFS How Object-Specific Are Object Files? Evidence for Integration by Location Wessel O. van Dam and Bernhard Hommel Leiden University Institute for Psychological Research and Leiden Institute for Brain and Cognition Given the distributed representation of visual features in the human brain, binding mechanisms are necessary to integrate visual information about the same perceptual event. It has been assumed that feature codes are bound into object files—pointers to the neural codes of the features of a given event. The present study investigated the perceptual criteria underlying integration into an object file. Previous studies confounded the sharing of spatial location with belongingness to the same perceptual object, 2 factors we tried to disentangle. Our findings suggest that orientation and color features appearing in a task-irrelevant preview display were integrated irrespective of whether they appeared as part of the same object or of different objects (e.g., 1 stationary and the other moving continuously, or a banana in a particular orientation overlaying an apple of a particular color). In contrast, integration was markedly reduced when the 2 objects were separated in space. Taken together, these findings suggest that spatial overlap of visual features is a sufficient criterion for integrating them into the same object file. Keywords: Visual information processing occurs in a distributed fashion— different features of a visual object are processed in spatially and functionally distinct cortical areas (Zeki, 1976). Given that event representations depend on distributed neural codes, a mechanism needs to be in place that integrates the codes representing the perceptual (Treisman, 1996) and response features (Stoet & Hom- mel, 1999; Wickens, Hyland, & Anson, 1994) of an event. Kah- neman, Treisman, and Gibbs (1992) suggested that encountering a particular combination of visual features leads to the creation of what they call an object file—a cognitive structure that cross- references all the features belonging to a particular object (cf. Allport, Tipper, & Chmiel, 1985). They reasoned that binding features into an object file may produce specific aftereffects that benefit performance if this particular combination is repeated. They presented participants with visual arrays of randomly ar- ranged letters in constantly visible boxes (the preview display or prime) followed by the presentation of a to-be-named target letter in one of the boxes (the probe). Two types of effects were observed: Performance was better if the probe letter had already appeared in the preview display than if it had not, but it was particularly good if the letter had appeared in the same box. Hence, there was both a nonspecific identity-priming effect (a benefit due to shape or letter repetition) and what Kahneman et al. called an object-specific preview benefit (a benefit due to the repetition of both shape and location). Kahneman et al. (1992) interpreted object-specific preview ef- fects as benefits resulting from the fact that repeating a feature combination (a shape and a location in their case) allows the receiver to make use of an already existing object file that already contains this particular combination. That is, if the probe matches the prime, it can in a sense take over and reuse the object file that was created to represent the prime. However, Hommel (1998) suggested an interpretation in terms of partial-overlap costs: Re- peating one or more features may lead to the automatic retrieval of the just created episodic binding, which creates code confusion if the retrieved feature codes do not entirely match with the features of the present stimulus–response episode. In the Kahneman et al. study, only three cells of what one may consider a two-by-two design were investigated: the repetition of letter identity and loca- tion, the repetition of letter identity only, and the alternation of both identity and location. 1 In contrast, Hommel’s design, which differed mainly in presenting only one prime stimulus at a time, allowed looking into all four combinations of identity and location repetition and alternation and to dissociate the effects of several nonspatial features. It turned out that repeating the combination of two features, such as a particular shape and a location or a particular shape and a color, yields no better performance than if both shape and location or shape and color were changed. Instead, performance is hampered if some but not all features repeat. Along the lines of Kahneman et al. (1992), this suggests that codes of co-occurring features are spontaneously bound, and that repeating at least one feature leads to the retrieval of the whole 1 Following common practice in research on repetition effects, we use the term alternation as an antonym of repetition to refer to change (of features, stimuli, or responses) as such but not necessarily regular changes. Wessel O. van Dam and Bernhard Hommel, Leiden University Institute for Psychological Research and Leiden Institute for Brain and Cognition, Leiden, the Netherlands. Wessel O. van Dam is now affiliated with the Radboud University, Donders Centre for Brain, Cognition and Behaviour, Nijmegen, the Neth- erlands. Correspondence concerning this article should be addressed Bernhard Hommel, Leiden University, Department of Psychology, Cognitive Psy- chology Unit, Wassenaarseweg 52, 2333 AK Leiden, The Netherlands. E-mail: hommel@fsw.leidenuniv.nl Journal of Experimental Psychology: © 2010 American Psychological Association Human Perception and Performance 2010, Vol. ●●, No. ●, 000–000 0096-1523/10/$12.00 DOI: 10.1037/a0019955 1 AQ: 1 AQ: 3 Fn1 tapraid5/zfn-xhp/zfn-xhp/zfn00410/zfn2502d10z xppws S=1 6/1/10 15:53 Art: 2007-0973