Does Attention Serve to Integrate Features?

Theoretical implications of findings reported by Treisman and her colleagues (e.g., Treisman, 1982, 1985; Treisman & Gelade, 1980) are discussed, and 3 possible hypotheses about the encoding of conjunctions are presented. Two plausible versions of the hypothesis, which posits attentional feature integration, are shown to predict that the time to search for a feature combination grows as a positively accelerated function of display size. The typically obtained linear functions are consistent with either of 2 alternative views: (a) that features are properly conjoined from the outset or (b) that features are extracted de novo for every stimulus being focally attended to. Perhaps visual attention does not act to integrate but rather enables testing of whether a stimulus matches a conjunctive target.