Ethanol may increase the willingness to take risks, but this issue remains controversial. We used a risk-taking paradigm in which volunteers answered a series of general knowledge questions with numerical answers and were asked to judge the length of a line that would just fit into a given gap. A maximum score was given for an exactly correct answer. For answers that were less than the correct value, the score was reduced gradually to zero, while answers even slightly over the correct value were penalized considerably. Total points were rewarded by cash payments, so volunteers were taking real risks when making their responses. Performance was assessed in a two-period, double-blind crossover study, comparing ethanol (0.7 g/kg) with placebo in 20 female volunteers aged 19-20 years. Tests were carried out before and at 45 min after dosing. Mean (SD) ethanol blood alcohol concentrations were 65 (10.5) mg/100 ml. Ethanol impaired the skill/ability measure of the length estimation test (SD of difference between length of line and gap), which increased from 5.9 to 6.6 (p < 0.05), indicating a reduced accuracy of estimation. The risk measures in both tasks were not significantly affected. The skill/ability measure in the general knowledge task was not significantly affected. Other performance tests showed that ethanol produced the expected impairment of both speed and accuracy. These results suggest that risk-taking is not increased by ethanol at doses approaching the UK legal limit for driving.