The purpose of the present study was to investigate the differential effects of video modeling (VM) and picture-based interventions (PI) for teaching children with ASD to recognize emotions in context. The study used an adapted alternating treatments design. One girl and three boys in China, 4.5-5 years old with autism spectrum disorder, participated. Both methods were effective, but three of the four children demonstrated a higher percentage of accurate responses with VM than with PI. All children required less time overall and less time per trial to reach criterion with PI than VM. They required fewer trials to reach criterion with VM, but VM trials were longer. All children displayed longer on-task behavior with VM than with PI. Overall, the two methods can be considered comparable in efficiency, with the choice of one rather than the other depending on learners' and instructors' needs, preferences, and contexts.