ABSTRACT
A major tenet of infl uential theories of human vision is that feature binding represents a critical limiting step, operating at only a single attended location at a time (Treisman, 1998; Treisman & Gelade, 1980). Typical evidence supporting this view is that search for targets comprising a conjunction of different features is slow (often taking around 25 ms/item or longer) and linearly related to the number of nontargets present. Ineffi cient search for conjunction targets occurs even when search for each feature making up the conjunction is effi cient and minimally affected by nontargets (search rates ~10 ms/item or less; Treisman & Gelade, 1980). It has been argued that search rates of under 10 ms/item are inconsistent with a biologically plausible serial search process, with search slopes above 20 ms/item refl ecting the operation of serial selection processes (Treisman & Gormican, 1988). Such results suggest that, to code the conjunction of features that distinguishes targets from nontargets, attention must be serially applied to each stimulus in turn (Treisman, 1998; Treisman & Gelade, 1980).
