Thirty-five children diagnosed with autism were randomly assigned to either a joint attention or a symbolic play intervention. During the 5-8 week treatment, three novel probes were administered to determine mastery of joint attention skills. The probes consisted of auditory and visual stimuli, such as a loud spider crawling or a musical ball bouncing. The current study examined affect, gaze, joint attention behaviors, and verbalizations at three different time points of intervention. Results revealed that children randomized to the joint attention group were more likely to acknowledge the probe and engage in shared interactions between intervener and probe upon termination of intervention. Additionally, the joint attention group improved in the proportion of time spent sharing coordinated joint looks between intervener and probe. These results suggest that generalization of joint attention skills to a novel probe did occur for the group targeting joint attention and provides further evidence of the effectiveness of the joint attention intervention.