Startle eyeblink modification was measured in 20 relatively asymptomatic medicated schizophrenia outpatients and 18 matched controls in order to test for deficits in early and later stages of attentional processing during a memory-load version of the Continuous Performance Test. Participants viewed a series of digits and pressed a button after the digit 7 of each 3–7 sequence. On some trials, a startling noise burst was presented either 120 or 1200 ms following cues that a response might be needed soon (the digit 3) and also following noncues. Controls showed greater startle inhibition at 120 ms following cue than noncue prepulses, whereas patients showed equal inhibition to both, suggesting a deficiency in allocation of controlled attentional resources in early stages of processing. The patients, however, did show large startle inhibition at 120 ms when a distracting stimulus accompanied the task-relevant cue, unlike the controls, who ignored the distracting stimulus. In contrast, both groups showed equal startle inhibition 1200 ms following cue and noncue prepulses, indicating that later modality-specific attention processes are not impaired in patients during this paradigm. Both groups also showed equal inhibition at 120 ms during passively attended prepulses, suggesting that automatic attentional processes were not impaired in these patients.