Computational Models of Visual Selective Attention: A Review
Dietmar Heinke, Glyn W. Humphreys
Abstract Draft version
We review some of the major computational models of visual selective attention in terms of how they apply to psychological data and to theoretical concepts derived from experiments concerned with stimulus filtering, visual search and spatial cueing. We also review attempts to model neuropsychological disorders of visual object recognition and attention, including visual agnosia and neglect. We highlight how computational models can unite several dichotomies in psychological models: space-based versus object-based selection; early versus late selection; representational versus attentional accounts for neurological deficits. A comparison of the models exposes weaknesses and advantages in different accounts, but it also hightlights that most models suggest that competitive interactions in visual selection are the basis for attentional effects in human data. There is also agreement across models on explaining neurological disorders as due to imbalances in competition following damage to certain areas of the brain. We propose that computational models clarify theoretical accounts of visual selective attention and integrate concepts across areas. The models provide a useful contribution to psychological research.
In G. Houghton (Ed.), Connectionist Models in Psychology , Psychology Press, in press.