Stimulus overselectivity, or the selective responding to narrow or irrelevant cues, may partially account for difficulties that students with autism face with generative language production. A body of literature exists that provides suggestions for its remediation. However, much of the research in this area has been basic in nature, with arbitrary stimuli presented in a discrete trial format. Research questions remain as to what role overselectivity plays in the natural environment. If overselectivity does play a significant role in language production in that setting, procedures need to be developed to evaluate, remediate, and prevent stimulus overselectivity in the natural environment. This article reviews the literature on overselectivity and makes suggestions for future research. ********** There are several goals that educators and clinicians set for children with autism when teaching them a new skill. The first goal is that the response should be reliably emitted in the presence of the discriminative stimuli. The second is that the skill generalizes to new environments and similar stimuli. At times, the skill is acquired, but does not generalize. For example, a student learning to sort items such as spoons may be able to do so with 100% accuracy in the classroom, but when the task is presented in the cafeteria, responding quickly decreases. Professionals then seek the root of the breakdown. It is possible that the stimuli that control responding in one environment are not present in the new environment. In the example presented earlier, the items to be sorted are functionally similar (e.g., they are still spoons); however, perhaps the spoons sorted in the classroom are metal while the spoons in the cafeteria are plastic. In this case, the metal, not the shape of the spoon, was the discriminative stimuli for responding. The problem may also lie in the prompting strategy selected. Individuals acquiring new skills are often prompted in order to insure accuracy. For some students, this prompting becomes the discriminative stimulus for responding. Attention is given to the prompting stimulus as opposed to the relevant stimulus. These are examples of stimulus overselectivity; a narrow or irrelevant component of a complex stimulus gains stimulus control over responding to the exclusion of other stimulus features (Rosenblatt, Bloom, & Koegel, 1995). OVERSELECTIVE RESPONDING DEFINED Stimulus overselectivity was first identified in the literature by Lovaas, Schreibman, Koegel, and Rehm (1971). Groups of children with autism, mental retardation (MR) and typical development were trained in a discrimination task involving complex stimuli. The participants with autism required more trials to learn the discrimination than those with MR or typical development. When each of the stimulus components was presented individually, it was revealed that individuals with autism often responded to only one component of the complex stimulus to the exclusion of other stimulus components (Lovaas, et al, 1971). A number of studies have replicated the Lovaas et al. (1971) finding that children with autism often selectively respond to a limited number of stimuli in the environment--however, this phenomenon has also been seen in young, typically developing children, individuals with severe and profound MR, and individuals with learning disabilities (Bailey, 1981; Lovaas, Koegel, & Schreibman, 1979). Wilhelm and Lovaas (1976) found that groups of children with lower IQs were more likely to respond to narrow cues than those with higher IQs. Rincover and Ducharme (1987) also found that developmental level was associated with overselective responding. These studies indicated that the lower the individual's developmental level, regardless of diagnosis, the more likely overselective responding will be observed. Gersten (1983) found that as chronological age increased, however, stimulus overselectivity decreased for participants with autism, MR, and typical development (this was independent of developmental level). …