Distinguishing serial and parallel processing in visual search without depending on set size effect

Abstract

Two-stage models (e.g., Treisman & Gelad, 1980; Wolfe, 1994) for visual search commonly assume a serial attentive processing stage following a parallel object encoding stage. A novel experimental manipulation was devised in a series of experiments with increasing task difficulty, intended to selectively influence processes in the attentive processing stage of a two stage system. In Experiment 1, subjects searched for a red T among green Ts and red Os (4 stimuli). In some trials, one or two distracters were replaced independently by special distracters that require longer processing time than the ones replaced. In Experiment 2, subjects searched for a letter L among an upright T, a T rotated by 180° and two Os. In some trials, two Os were independently replaced by Ts rotated by 90° and 270°. In Experiment 3, subjects searched for upright or rotated Ts among Os, an upright T, and a T rotated by 180° whose vertical strokes were slightly shifted to the left or right. In some trials, Os were replaced by stroke-shifted Ts rotated by 90° and 270°. Patterns of interactions produced by the two special distracters (Experiment 1) and Ts rotated by 90° and 270° (Experiment 2) in terms of means and cumulative distribution functions of reaction times support parallel processing (sub-additivity) rather than serial processing (additivity). The serial processing assumption was only supported in experiment 3. Also the same conclusion was drawn from the same experiments with a large set size (8 & 12 stimuli) with an exception that can be attributed to salience. The results indicate that the assumption of serial attentive processing is rejected in for these situations and the set size effect should not be considered as a sole criterion for identifying the process organization in visual search (serial vs. parallel) even if its magnitude of effect is large.