Abstract
PURPOSE. Prior research has shown that belief about contour connectedness alters global shape integration but not illusory contour formation. Here, we aim to replicate those results while also independently verifying that subjects adopted the instructed belief. METHOD. Subjects discriminated briefly-presented, partly-visible fat and thin shapes, the edges of which either induced or did not induce illusory contours (relatable and nonrelatable condition, respectively). Half of trials of each condition incorporated task-irrelevant distractor lines, which disrupt contour filling-in. Half of the observers were asked to treat the visible parts of the target as belonging to a single thing (group strategy); the other half were asked to treat the parts as disconnected (ungroup strategy). A strategy was encouraged by giving subjects pictures of fat and thin response templates in the experiment instructions, and after every 20 trials. These pictures depicted either unitary shapes or fragmented shapes, depending on strategy. After each half of the experiment, subjects were asked how they cognitively regarded the stimulus. Task-naïve judges evaluated the responses. RESULTS. The results confirmed previous findings. Distractor lines impaired performance more in the relatable condition than in the nonrelatable condition (p=.001). Strategy did not alter the effects of distractor lines in the relatable trials or the nonrelatable trials (ps>.25). The attempt to group relatable fragments improved performance (p=0.004), while the attempt to group nonrelatable fragments did not (p>.5). Finally, 90% of the subjects were evaluated as having adopted the instructed strategy on both halves of the experiment. CONCLUSIONS. These new results show once again that a) filling-in effects during illusory contour formation cannot be easily removed via top-down strategy; b) filling-in effects cannot be easily manufactured from stimuli that fail to elicit interpolation; c) actively grouping fragments can readily improve discrimination performance, but only when those fragments cause interpolation.
Meeting abstract presented at VSS 2012