This study evaluated the effectiveness of a rival- model learning procedure on the acquisition of variations of color tacts and assessment of generalized tacting with children with autism. The participants IQs ranged from 49 to 63 and all were eleven years of age or younger. A multiple baseline across participants with embedded multiple probe was used. The child’s typical therapist functioned as an observational model but participants never received programmed consequences for responding related to the target behavior. The results of this study indicated that using a rival-model procedure produced accurate responding on trained stimuli tacting shades of color as well as generalized shades of color. This study adds to the limited amount of research published on the rival-model observational learning method with children with autism and related disabilities. Keywords: Autism, generalization, PEAK Relational Training System, rival-model,