Academic journal article Cognitive, Affective and Behavioral Neuroscience

Visual Antipriming: Evidence for Ongoing Adjustments of Superimposed Visual Object Representations

Academic journal article Cognitive, Affective and Behavioral Neuroscience

Visual Antipriming: Evidence for Ongoing Adjustments of Superimposed Visual Object Representations

Article excerpt

A fundamental question of memory is whether the representations of different items are stored in localist/discrete or superimposed/overlapping manners. Neural evidence suggests that neocortical areas underlying visual object identification utilize superimposed representations that undergo continual adjustments, but there has been little corroborating behavioral evidence. We hypothesize that the representation of an object is strengthened, after it is identified, via small representational changes; this strengthening is responsible for repetition priming for that object, but it should also be responsible for antipriming of other objects that have representations superimposed with that of the primed object Functional evidence for antipriming is reported in young adults, amnesic patients, and matched control participants, and neurocomputational models. The findings from patients dismiss explicit memory explanations, and the models fit the behavioral performance exceptionally well. Putative purposes of priming and comparisons with other theories are discussed. Priming and antipriming may reflect ongoing adjustments of superimposed representations in neocortex.

A fundamental question of knowledge representation is whether the representations of different items are stored in separate or overlapping manners. Many cognitive models and theories posit localist/discrete representations (see, e.g., Grainger & Jacobs, 1996; Hintzman, 1986; Jacoby, 1983; Logan, 1990; Medin & Schaffer, 1978; Poggio & Edelman, 1990; Shiffrin & Steyvers, 1997), with different items stored in separate traces. Many other cognitive models and theories posit superimposed representations (see, e.g., Anderson & Hinton, 1981; Eich, 1982; Knapp & Anderson, 1984; Masson, 1995; McClelland & Rumelhart, 1985; Murdock, 1982; Seidenberg & McClelland, 1989), with different items stored in distributed and overlapping traces. In this article, we report a new effect, "visual antipriming," which provides functional evidence for superimposed representations of visual objects and a continualadjustment account of repetition priming effects.

Neural implementational evidence suggests that familiar visual shapes are represented in a distributed and partially overlapping manner in ventral temporal cortex. In neurophysiological studies, different whole shapes activate inferior temporal visual neurons to differing degrees (Gross, Rocha-Miranda, & Bender, 1972). Different neurons are maximally sensitive to different moderately complex features of whole objects (Tanaka, 1993); these features are simple enough that they are present in multiple whole objects but also complex enough that they are not present in all whole objects. Similarly, neuroimaging studies indicate that patterns of activation from different classes of shapes (e.g., houses, bottles, shoes, faces, chairs) overlap in human ventral temporal cortex (Haxby etal., 2001; Ishai, Ungerleider, Martin, Schouten, & Haxby, 1999). In addition, neurocomputational modeling indicates functional utilities for learning and storing partially superimposed representations (Hinton, McClelland, & Rumelhart, 1986). The overlap between representations not only increases the efficiency of storage, but can also be used to store systematic relationships between inputs and appropriate outputs that enable the system to efficiently generalize to novel inputs (e.g., identifying a previously unseen piano as a piano; McClelland & Rumelhart, 1985). Interestingly, though, this utility is attained at a price. In order for new categories to be learned in the system (e.g., a representation for a new gadget), trials of learning the new information must be interleaved with trials of relearning the old familiar representations (e.g., piano; McClelland, McNaughton, & O'Reilly, 1995). Otherwise, many trials of learning the new information all at once (without the interleaved relearning of old information) can cause old representations to be irretrievably lost (McClelland et al. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.