Confirmation bias in visual attention

This dissertation evaluates the proposal that visual attention is biased towards stimuli that could potentially verify a perceptual hypothesis, compared to stimuli that might falsify it. To do so, I use a visual search task with two, mutually exclusive, coloured targets (e.g., a red p or a blue p). Before each set of searches, one target colour is designated positive, or confirmatory, information and one is designated negative information by requiring affirmative responses or negative responses to one of the two targets (e.g., press Z if the p is red, press M if it is not). The first set of experiments, reported in Chapter 2, showed that confirmatory search patterns were observed for both novel and practiced hypotheses, were observed despite instructions to the contrary, and were mitigated by a search colour-preview. Chapter 3 measured the ocular correlates of the search bias, showing that more fixations indeed go towards confirmatory stimuli, but showed that when stimulus inspections took longer, search became les confirmatory and more efficient. Finally, Chapter 4 demonstrated that confirmatory search did not occur when hypotheses referred to visually heterogeneous stimuli and also did not occur when hypotheses included a negation, suggesting that a visual matching bias underlies confirmatory search. Confirmation bias thus appears to be a side-effect of goal-directed cognition in general, and the present results are discussed in relation to theories likening reasoning to foraging and information search.Ph.D