The proportion of peripheral cues that predicted at which of two locations a discrimination target would occur was manipulated in three experiments. Near cues occurred adjacent to the target and far cues occurred adjacent to the other location. In Experiment 1, near and far cues were presented in separate blocks so each predicted target location. Both types of cue produced benefit relative to a neutral baseline. In Experiment 2 expectancy for near versus far cues was manipulated. In one group, 80% of the peripheral cues were near; in another, only 20% were near; in a third, near and far cues occurred randomly. Near cures always produced benefit. Far cues produced benefit when they were expected, had no effect in the random group, and sometimes produced cost when they were unexpected. Experiment 3 repeated the near-expected and far-expected conditions, but target duration was shortened to 50 msec. A similar pattern emerged, except that near cues produced less benefit at longer SOAs when they were unexpected. The data suggest that peripheral, abrupt-onset cues may produce benefit via an automatic process, but that such a process is more like location priming than the capture of attention. Expectancy determines whether and where attentional resources will be allocated in response to the cue.