The hypothesis that abrupt visual onsets automatically capture attention was tested in four experiments. A central cue directed attention to one of several stimulus positions in preparation for identifying a target letter among distractors. In all experiments, one stimulus (either the target or a distractor) had an abrupt onset, while the rest did not. The effectiveness of the cue was manipulated to determine if abrupt onsets capture attention even when subjects are in a focused attentional state. Results showed that abrupt onsets do not necessarily capture attention in violation of an observer's intentions. A mechanism for partially automatic attentional capture by abrupt onset is proposed, and the diagnosticity of the intentionality criterion for automaticity is discussed. Introspective and empirical evidence suggest that abrupt visual onsets draw attention. A plausible account is that a mechanism tuned to abrupt onsets directs visual attention to their location, aiding efficient identification. This hypothesis has two components: a mechanism detecting abrupt onsets and signaling attentional resources, and automatic allocation of attention. The central thesis is that attention may be efficiently allocated to abrupt onsets under some circumstances, but not automatically. Evidence supports the existence of a mechanism for attentional capture by abrupt onsets. Electrophysiological and psychophysical results show visual mechanisms sensitive to abrupt onsets. Todd and Van Gelder (1979) found that onset stimuli were detected more rapidly than no-onset stimuli. Krumhansl (1982) confirmed this. Yantis and Jonides (1984) proposed that abrupt onsets capture attention automatically, supported by visual search experiments. Onset targets were processed first, and data fit a model of serial search. Jonides and Yantis (1988) tested if abrupt onsets capture attention automatically compared to color and intensity. Only abrupt onsets yielded results consistent with attentional capture. The intentionality criterion for automaticity requires that automatic processes are not subject to voluntary control. Kahneman and Treisman (1984) defined three levels of automaticity. Jonides (1981) found that peripheral cues capture attention automatically, satisfying the load-insensitivity and intentionality criteria. Posner and Cohen (1984) showed that attention is summoned to cues at short intervals but inhibited at longer intervals. Lambert, Spencer, and Mohindra (1987) found that subjects could not avoid attending to abrupt peripheral cues despite instructions. Müller and Rabbitt (1989) found that attention was involuntarily captured by irrelevant stimuli, even when directed to the cued location. In Experiment 1, abrupt onsets did not determine attention focus. Whether the target was cued or not modulated the effect of abrupt onsets. The results showed that when subjects attended to the target position, abrupt onsets had little effect, but when misled by the cue, targets with abrupt onsets were identified more rapidly. Experiment 2 showed that effectiveThe hypothesis that abrupt visual onsets automatically capture attention was tested in four experiments. A central cue directed attention to one of several stimulus positions in preparation for identifying a target letter among distractors. In all experiments, one stimulus (either the target or a distractor) had an abrupt onset, while the rest did not. The effectiveness of the cue was manipulated to determine if abrupt onsets capture attention even when subjects are in a focused attentional state. Results showed that abrupt onsets do not necessarily capture attention in violation of an observer's intentions. A mechanism for partially automatic attentional capture by abrupt onset is proposed, and the diagnosticity of the intentionality criterion for automaticity is discussed. Introspective and empirical evidence suggest that abrupt visual onsets draw attention. A plausible account is that a mechanism tuned to abrupt onsets directs visual attention to their location, aiding efficient identification. This hypothesis has two components: a mechanism detecting abrupt onsets and signaling attentional resources, and automatic allocation of attention. The central thesis is that attention may be efficiently allocated to abrupt onsets under some circumstances, but not automatically. Evidence supports the existence of a mechanism for attentional capture by abrupt onsets. Electrophysiological and psychophysical results show visual mechanisms sensitive to abrupt onsets. Todd and Van Gelder (1979) found that onset stimuli were detected more rapidly than no-onset stimuli. Krumhansl (1982) confirmed this. Yantis and Jonides (1984) proposed that abrupt onsets capture attention automatically, supported by visual search experiments. Onset targets were processed first, and data fit a model of serial search. Jonides and Yantis (1988) tested if abrupt onsets capture attention automatically compared to color and intensity. Only abrupt onsets yielded results consistent with attentional capture. The intentionality criterion for automaticity requires that automatic processes are not subject to voluntary control. Kahneman and Treisman (1984) defined three levels of automaticity. Jonides (1981) found that peripheral cues capture attention automatically, satisfying the load-insensitivity and intentionality criteria. Posner and Cohen (1984) showed that attention is summoned to cues at short intervals but inhibited at longer intervals. Lambert, Spencer, and Mohindra (1987) found that subjects could not avoid attending to abrupt peripheral cues despite instructions. Müller and Rabbitt (1989) found that attention was involuntarily captured by irrelevant stimuli, even when directed to the cued location. In Experiment 1, abrupt onsets did not determine attention focus. Whether the target was cued or not modulated the effect of abrupt onsets. The results showed that when subjects attended to the target position, abrupt onsets had little effect, but when misled by the cue, targets with abrupt onsets were identified more rapidly. Experiment 2 showed that effective