Memory model-based expectancy studies have shed light on the process by which expectancies may influence drinking but have not related changes in expectancy activation to drinking changes. In the present study, 38 undergraduates completed a drinking measure and factor-based and memory model-based expectancy measures, before and after an expectancy challenge intervention designed to alter expectancies. Expectancies were mapped into memory network format with individual differences scaling and likely paths of activation were modeled with preference mapping. Results indicated that exposure to the expectancy challenge led to a change in likely activation of expectancies for men, but not for women. In the 30 days after the intervention, alcohol use among men decreased significantly but did not change among women. Therefore, changes in likely activation corresponded to changes in drinking. These findings support a memory model conceptualization of expectancy influence on drinking.