Computer-based video provides a valuable tool for HIV prevention in hospital emergency departments. However, the type of video content and protocol that will be most effective remain underexplored and the subject of debate. This study employs a new and highly replicable methodology that enables comparisons of multiple video segments, each based on conflicting theories of multimedia learning. Patients in the main treatment areas of a large urban hospital's emergency department used handheld computers running custom-designed software to view video segments and respond to pre-intervention and postintervention data collection items. The videos examine whether participants learn more depending on the race of the person who appears onscreen and whether positive or negative emotional content better facilitates learning. The results indicate important differences by participant race. African American participants responded better to video segments depicting White people. White participants responded better to positive emotional content.