Date of Award:


Document Type:


Degree Name:

Educational Specialist (EdS)



Committee Chair(s)

Andrew L. Samaha


Andrew L. Samaha


Donna Gilbertson


Sarah E. Bloom


Previous research suggests learning of children with autism often fails to successfully generalize across changes in settings and stimuli. Much of this research has assessed generalization by first teaching a behavior in one context and then measuring the transfer of the behavior to extra-treatment stimuli and settings. The present study measured generalization of learned behavior by systematically varying the tone of an auditory stimulus present during training to obtain generalization gradients. Generalization gradients are graphical representations of the strength of a response produced by stimuli that vary from the training stimulus along some stimulus dimension. By obtaining generalization gradients, this research may offer a more precise means of characterizing the extent of generalization and the basic processes underlying it. The study also went beyond previous research with children with autism by examining the effects of two different training procedures upon the resulting generalization gradients. Participants were first taught to discriminate between the presence and absence of a specific stimulus, and later, to discriminate between two stimuli varied along the same dimension. Gradients were measured following both trainings. In the first training procedure, three children with autism were taught to engage in a simple communicative request in the presence of a specific tone and to withhold the request when there was no tone. The researchers then measured the extent to which these children continued to engage in the request as the tone was changed in frequency. They graphed the resulting data in the form of a generalization gradient. Although the shape of resulting generalization gradients differed between participants, all three participants in the present study showed a pattern of responding consistent with generalization. Gradients for two of three participants were orderly and decremental. In the second training procedure participants were taught to discriminate between two tones of different frequencies. Generalization gradients were again obtained. Predictable changes in the shape of gradients, consistent with basic research on generalization gradients, were noted for two of three participants. Results are discussed with regard to stimulus control, the behavioral processes of reinforcement and extinction, and what has been called the "peak shift" effect.



Included in

Psychology Commons