Four experiments tested a new hypothesis that involuntary attention shifts are contingent on the relationship between the properties of the eliciting event and the properties required for task performance. In a variant of the spatial cuing paradigm, the relation between cue property and the property useful in locating the target was systematically manipulated. In Experiment 1, invalid abrupt-onset precues produced costs for targets characterized by an abrupt onset but not for targets characterized by a discontinuity in color. In Experiment 2, invalid color precues produced greater costs for color targets than for abrupt-onset targets. Experiment 3 provided converging evidence for this pattern. Experiment 4 investigated the boundary conditions and time course for attention shifts elicited by color discontinuities. The results of these experiments suggest that attention capture is contingent on attentional control settings induced by task demands.

When the stimuli from two tasks arrive in rapid succession (the overlapping tasks paradigm), response delays are typically observed. Two general types of models have been proposed to account for these delays. Postponement models suppose that processing stages in the second task are delayed due to a single-channel bottleneck. Capacity-sharing models suppose that processing on both tasks occurs at reduced rates because of sharing of common resources. Postponement models make strong and distinctive predictions for the behaviour of variables slowing particular second-task stages, when assessed in single- and dual-task conditions. In Experiment 1, subjects were required to make manual classification responses to a tone (S1) and a letter (S2), presented at stimulus onset asynchronies of 50, 100, and 400 msec, making R1 responses to S1 as promptly as possible. The second response, R2, but not R1, was delayed in the dual task condition, and the effects of two S2 variables (degradation and repetition) on R2 response times in dual- and single-task conditions closely matched the predictions of a postponement model with a processing bottleneck at the decision/response-selection stage. In Experiment 2, subjects were encouraged to emit both responses close together in time. Use of this response grouping procedure had little effect on the magnitude of R2 response times, or on the pattern of stimulus factor effects on R2, supporting the hypothesis that the same underlying postponement process was operating. R1 response times were, however, dramatically delayed, and were now affected by S2 difficulty variables. The results provide strong support for postponement models of dual-task interference in the overlapping tasks paradigm, even when response times are delayed on both tasks.

Two experiments used the locus-of-cognitive-slack method to determine whether dual-task interference occurs before or after the response selection stage. The experiments used the overlapping tasks paradigm, in which two signals, each requiring a different speeded choice response, are presented in rapid succession. In Experiment 1, stimulus-response (S-R) compatibility was manipulated by varying whether Task 2 stimuli were mapped onto their responses by a rule or arbitrarily. Compatibility effects were additive with the effects of degree of task overlap, manipulated by varying the stimulus onset asynchrony between the signals. Experiment 2 examined 2 additional forms of S-R compatibility: symbolic compatibility (arrows vs. letters) and spatial compatibility (the "Simon" effect). Effects of symbolic compatibility were additive with effects of degree of task overlap, whereas the effects of spatial compatibility and degree of task overlap were underadditive. It is argued that only a central-bottleneck model provides a consistent account of these results. The nature of the central bottleneck is considered. People are severely limited in their ability to perform two or more tasks at the same time (Pashler & Johnston, 1989; Vince, 1948; Welford, 1952). The study of this kind of performance limit is important both theoretically and practically. At a theoretical level, understanding multitask interference provides clues to cognitive architecture and the control of mental processes (Keele, 1973). At a practical level, interference between tasks severely limits the functioning of operators in multitask environments such as air traffic control towers. A better understanding of multitask limitations could help improve the performance of such operators. Two broad classes of models have been advanced to explain interference between tasks. One class (Broadbent, 1971; Pashler & Johnston, 1989; Welford, 1952) holds that interference arises because certain cognitive operations of each task demand simultaneous access to a processor (or processors) that can only service one task at a time. During the time that one task is occupying the bottleneck process(es), there is postponement of processing on the other task. The other major class consists of capacity models (Kahne

Subjects made lexical decisions on a target letter string presented above or below fixation. In Experiments 1 and 2, target location was cued 100 ms in advance of target onset. Responses were faster on validly than on invalidly cued trials. In Experiment 3, the target was sometimes accompanied by irrelevant stimuli on the other side of fixation; in such cases, responses were slowed (a spatial filtering effect). Both cuing and filtering effects on response time were additive with effects of word frequency and lexical status (words vs. nonwords). These findings are difficult to reconcile with claims that spatial attention is less involved in processing familiar words than in unfamiliar words and nonwords. The results can be reconciled with a late-selection locus of spatial attention only with difficulty, but are easily explained by early-selection models.

Early and late selection models of attention disagree about whether visual objects are identified before or after selection, with recent evidence of interference from to-be-ignored stimuli favoring late selection over early selection accounts. However, these tests may not have permitted optimal attentional focusing. In 4 experiments subjects identified an attentionally cued target letter embedded among distractors. Only minimal effects of information appearing in to-be-ignored locations were observed. This striking efficiency of selection provides support for early selection theories and calls into question some late selection theories holding that stimuli throughout the display are immediately and fully identified prior to attentional selection. In order to explain the larger pattern of results across a variety of focused-and divided-attention paradigms, a hybrid model is advanced with a flexible locus for visual selection.