| Abstract: | Previous work has amply demonstrated divided attention “costs” both in single-task multistimulus visual processing and in performance of multiple simple tasks; however, the relationship between the two has not been clarified. This article postulates two distinct causes: (1) visual processes that commence without delays and proceed simultaneously, but show lingering mutual interference dependent upon complexity, and (2) discrete queueing of the response selection stage. The first has resource-like properties, while the second has bottleneck-like properties. Either or both can generate performance costs observed in any particular situation, accounting for a variety of previous results. To test this theory, the effects of stimulus onset asynchrony (SOA) on accuracy and speed in performing dual choice tasks were examined. The first two experiments involved a choice response to a tone as a first task, and a second task requiring complex perceptual decisions (digit identification or conjunction search) with masked displays and unspeeded second responses. Reducing the SOA had negligible effects upon second-task accuracy, and performance in the two tasks was virtually independent. However, when speeded manual (Exp. 3) or vocal (Exp. 4) responses were required on the same second task, dramatic interference was observed, with strong positive dependencies between reaction times (RTs) on the two tasks. When both tasks involved complex visual displays, SOA reductions produced dramatic interference, but no dependencies between performance, whether the first task involved a speeded (Exp. 5) or unspeeded (Exp. 6) response. The results reject pure late-selection accounts and general capacity sharing models, and support the two-component theory. They also suggest that standard use of the term “attention,” suggesting a single resource or mechanism, is highly misleading. |
