ERIC Number: ED335213
Record Type: RIE
Publication Date: 1991-Jan-25
Reference Count: N/A
Neuropsychological Components of Imagery Processing, Final Technical Report.
Kosslyn, Stephen M.
High-level visual processes make use of stored information, and are invoked during object identification, navigation, tracking, and visual mental imagery. The work presented in this document has resulted in a theory of the component "processing subsystems" used in high-level vision. This theory was developed by considering neuroanatomical, neurophysiological, and computational constraints. The theory has led to two kinds of empirical work. First, specific hypotheses about individual processing subsystems have been tested. For example, the analysis of the representation of spatial relations led to the prediction that the subsystems are used to encode this information, and a set of experiments was conducted that provided support for this distinction. This work has involved a combination of divided-visual-field experiments with normal subjects and detailed examinations of patients with focal brain damage. Second, the subsystems have been implemented in a large computer simulation model, which has been used to generate predictions about specific neurological syndromes. The model can be damaged in a variety of ways, and its performance on a set of tasks then observed. Predictions from the model have been tested, the results generally support its underlying assumptions and specific claims. In addition, individual subsystems have been implemented as "neural network" simulation models, which have been related directly to properties of the neural substrate assumed to underlie processing. The experiments conducted to date are summarized in the context of the theory in this report, and the utility of the theory for understanding the effects of brain damage is illustrated by reviewing a single case in detail. (Author)
Publication Type: Reports - Research
Education Level: N/A
Sponsor: Air Force Office of Scientific Research, Washington, DC.
Authoring Institution: Harvard Univ., Cambridge, MA.
Identifiers: Computational Models