Ethanol has been shown to have a relatively greater effect on error rates in speeded tasks than temazepam, and this may be due to a differential effect on the speed-accuracy trade-off (SATO). This study used different instruction sets to influence the SATO. Forty-nine healthy volunteers (24 males, aged 18-41 years) were allocated at random to one of three instruction conditions--emphasising accuracy, neutral, and emphasising speed. After familiarisation, they took part in two sessions spaced at least 4 days apart in which they received either ethanol (0.8 g/kg, max 60 g males, 50 g females) or placebo in randomised order. Tests were administered starting at 30 and 75 min postdrug. Instructions significantly affected performance. In two maze tasks, one on paper, the other on a pen computer, the pattern of instruction effects was as expected. A significant increase in errors with ethanol was seen for both maze tasks, and there was a tendency to speed up with ethanol (significant only for the pen computer task). Responses to fixed stimulus sequences on the Four-Choice Reaction Test also showed a tendency to speed up and an increase in errors with ethanol, while all other tests showed both slowing and increases in errors with ethanol compared to placebo. Error scores are consistently increased by ethanol in all test situations, while the effects of ethanol on speed are variable across tests.