Alcohol servers were asked to rate the age of individuals shown in head and shoulders photographs. The subjects of the photographs were males and females aged 13, 16, 20 and 22. Respondents consistently overestimated the ages of the 13- and 16-yearolds; these errors were substantially greater for the girls, and increased with the age of the server. However, the same respondents underestimated the ages of the 20- and 22-yearolds. A response of 18 or greater was made in 3% of estimates of the age of 13-year-old boys. This figures increased to 18% for 13-year-old girls, 38% for 16-year-old boys, and 56% for 16-year-old girls. These data suggest that alcohol servers may sometimes sell alcohol to underage customers in good faith, and that training servers in age perception may decrease underage alcohol sales. However, a comparison of the present data with earlier data on actual alcohol sales, to the same 13- and 16-year-olds whose photographs were used in the present study, suggests that only around half of underage alcohol sales can be accounted for by misperceptions of age.