| |
Abstract:
In rare cases hemispatial neglect affects just one category
of stimuli, such as faces (Young, De Haan, Newcombe & Hay,
Neuropsychologia, 1990), words (e.g., Katz & Sevush, Brain and
Language, 1989), or numbers (Cohen & Dehaene, Cognitive
Neuropsychology, 1991). This has been interpreted by some (e.g.,
Umilta, J. Clin and Exp Neuropsychology, 1995; Young et al., 1990)
as evidence for category-specific spatial attention modules, e.g.,
a system of attention just for faces. Although such a system of
modules is possible, our preference for parsimony motivates the
search for alternative interpretations. We demonstrate that
content-specific neglect can be explained with a single
domain-general spatial attention system. The demonstration makes
use of 'belief nets', abstract models of cognitive information
processing based on Bayesian reasoning. The domain-general
attentional impairment is simulated by degrading the input to
higher-level perceptual systems. To account for the
content-specificity of the observed impairments, distortion is
introduced into the relevant higher-level perceptual output
pathways (e.g., face). Given the assumption that each stage of
processing converges slowly to a stable belief state over time,
then very slight degradation at each of the two loci in the model
(e.g., both attention and face pathway) produces a substantial
performance deficit when engaged together, despite little or no
performance deficit when engaged separately (e.g., attention to
objects other than faces and perception of faces in non-neglected
regions of space).
|
