Abstracting phonological generalizations
Evidence from children with disorders
The purpose was to document the trials to induce first generalization in children with phonological disorders enrolled in treatment. Archival data from 65 preschoolers were examined, with three kinds of generalization documented: treated, within- and across-class gains in production accuracy. Overall, an average of 185 trials was sufficient to induce first generalization, but this varied based on the kind of generalization that occurred. Across-class generalization required the fewest trials and generalization to the treated sound, the most trials. Results bear on applied issues associated with lexical diffusion in clinical treatment and theoretical issues associated with error-driven learning algorithms and abstraction of phonological generalizations from the input.