Rats were trained to respond to the lever above which a light stimulus was briefly (0.5 s) presented at unpredictable times. Once the task had been learned to criterion, subjects were injected, intra-peritoneally, with arginine8-vasopressin, desglycinamide arginine8-vasopressin (AVP or DGAVP: 0, 5, 10 or 20 micrograms/kg) or D-amphetamine (AMP: 0, 0.75, 1.5 or 3 mg/kg) prior to test. Attention performance was assessed using several different indices, including percent corrects, sensitivity and responsivity measures derived from signal detection theory, the recently described probability of (response) repetition and switching, and latency to respond. AVP had a disruptive effect on percent corrects at the highest dose and increased response latencies, but DGAVP, which lacks pressor activity, had no behavioral effects. AMP markedly impaired most aspects of performance, and was the only substance to alter response strategies by inducing bias and repetitive responding. It is concluded that (1) contrary to some recent reports, visual attention is disrupted, not improved, by peripherally injected AVP, (2) these effects reflect pressor potency, (3) the disruption induced by AMP reflects response alterations, while the peptide probably affects more cognitive mechanisms, and (4) certain recently described indices are more sensitive than others in detecting response bias.