Purchase this article with an account.
Maria Nordfang, Jeremy M. Wolfe; Nonlinear effects of target-distractor feature sharing in triple conjunction visual search. Journal of Vision 2014;14(10):918. https://doi.org/10.1167/14.10.918.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Suppose you are searching for one of the 27 stimuli created by the conjunctions of three colors, three shapes, and three orientations (say your target is a red, vertical, rectangle, leaving the remaining 26 stimuli as possible distractors). Models like Feature Integration and Guided Search predict that the efficiency of that triple conjunction search will depend on the features shared by the distractors and the target, not on the specific conjunctions of these features. However, conjunction effects have been found (e.g., Wolfe, 2010). We propose that a primary cause of this effect is a non-linearity in the effects of target-distractor feature sharing. A triple conjunction target can share zero, one, or two features with any distractor. In Experiment 1, we varied the number of features the distractors shared with the target. We found that compared to a baseline where all distractors share one feature with the target, the cost of sharing two features is greater than the benefit of sharing zero. In Experiment 2, we extended this nonlinearity to six-fold conjunctions. In Experiment 3, distractors always shared one feature on average with the target but, in some displays, distractors shared zero, one, or two features while in other displays all distractors shared one feature. This allowed us to keep the distribution of individual features constant. Because the cost of sharing two is greater than the benefit of sharing zero, the Share012 condition is slower than the AllShare1 condition. In Experiment 4, with brief exposures (110 and 200 msec), we found a subsidiary effect of grouping. Share012 conditions were slower if they contained 12 distinct distractor types compared to three types. These Sharing and Grouping effects are not predicted by previous accounts of GS; however, they can be accommodated in a GS framework.
Meeting abstract presented at VSS 2014
This PDF is available to Subscribers Only