• autoassociative recall
• backpropagation algorithm
• biased competition
• bottom-up vs. top-down attention
• Broadbent’s filter theory
• convolutional neural networks
• Deco and Rolls’ multimodular neurodynamical model
• early selection vs. late selection
• feature integration theory
• feedback model of biased competition
• Fukushima’s selective attention model
• global saliency map hypothesis
• gradient descent
• guided search 2.0
• hebbian learning
• hmax model
• Neisser’s two-stage model
• neocognitron
• object recognition
• overt vs. covert attention
• selective visual attention
• self-organization algorithm
• shunting equation
• spatial vs. feature-based attention
• trace learning rule
• visual search
• wta neural networks
• Yarbus’ eye-tracking experiments

This thesis provides a comprehensive analysis of selective visual attention by integrating findings from the fields of neuroscience, cognitive psychology, and artificial intelligence. It explores different types of selective visual attention, including top-down versus bottom-up, spatial versus feature-based, and overt versus covert attention. Leading neuroscientific theories of visual attention are examined, such as feature integration theory (FIT), guided search theory, and biased competition theory. Special emphasis is placed on the mathematical analysis of several neurally-inspired computational models developed from these theories. The role of visual attention in visual search tasks is investigated with reference to FIT-based models such as Itti, Koch, and Niebur's global saliency map model and Wolfe's Guided Search 2.0. For object recognition tasks, the effects of visual attention on convolutional and convolutional-based neural networks are evaluated, including Riesenhuber and Poggio’s HMAX Model and Fukushima's Selective Attention Model. Fukushima's model is also used to investigate the role of visual attention in autoassociative recall processes. Furthermore, models based on biased competition are explored at different levels to explain both single-cell recordings (via shunting equation models and feedback-based models) and fMRI data (using Deco and Rolls' Multimodular Neurodynamical Model). Despite the advancements, there remains a notable gap in the scientific literature: the lack of a neurophysiologically plausible, unified computational model that integrates all known types of selective visual attention. Future research must address this gap to deepen our understanding of the complex mechanisms underlying visual attention.