The participants were an opportunity sample of 56 female and 44 male alcohol servers. The research tested the participants in their workplaces, which included supermarkets, off-licenses, corner shops, and public houses. The participants viewed 80 head-and-shoulders photographs of individuals ages 13, 16, 20, and 22 years and rated their ages on an integer scale from 10 to 25. The participants ages 13 and 16 had previously used a photo identification card containing the same photo to enter licensed premises and attempt to purchase alcohol. They had approximately 5 seconds for each rating. Results revealed consistent overestimates of the ages of the 13-year-olds and 16-year-olds. The errors were substantially greater for the females photographed and increased with the age of the server. However, the same participants underestimated the ages of the 20-year-olds and 22-year-olds. A response of age 18 or greater occurred in 3 percent of the estimates for the 13-year-old males, 18 percent for 13-year-old females, 38 percent for 16-year-old males, and 56 percent for 16-year-old females. A comparison of earlier data on actual alcohol sale to the same 13-year-olds and 16-year-olds whose photographs the study used suggested that misperceptions of age accounted for only about half the underage alcohol sales. Findings suggest that training servers in age perception may reduce sales of alcohol to underage individuals. Figures, tables, and 16 references