A connectionist model of visual search for simple form conjunctions
Evidence for qualitatively different processes involved in the coding of simple features and of feature conjunctions comes from studies of visual search. When targets have a unique distinguishing feature relative to distractors (i.e. with what we will term single-feature targets), the function for reaction times (RTs) relative to the number of distractors is non-linear or even flat. In contrast,search times for targets defined by a combination of features (i.e. combined-feature targets) characteristically increase linearly as the number of distractors increases. Furthermore, in the prototypical case, the slope for target-present decisions is about half that for absent decisions. The linear search functions and 1 : 2 present-absent slope ratios have been taken as indicative of serial, selfterminating search (Treisman and Gelade, 1980).