Expectancies for reinforcement from alcohol appear to form during the grade-school years and play a causal role in problem drinking behavior. Using a sample of 268 second through fourth graders, we investigated whether children's expectancies could be modified successfully. Children were randomly assigned to one of four conditions: two experimental expectancy modification conditions (using 10-minute video interventions), one control condition involving a 10-minute video presenting facts concerning alcohol's deleterious effects, and one no-intervention control condition. Relative to controls, expectancies were increased by one experimental condition and decreased by the other; these changes were sustained at 4-week follow-up. This suggests expectancy modification should be tested as an adjunct to current prevention programs. Relations of gender and family history of problem drinking or alcoholism to expectancy endorsement are described.