The nature and specificity of the language coding deficit in autistic children.

The subjects, 12 autistic, 12 retarded and 12 normal children, range-matched on digit span and mental age, were asked to learn and recall sequences of words. Four types of sequence were used: sequences were constructed to be either grammatically well formed or not and were constructed so that the elements had a high or low degree of semantic relatedness. Sequences were presented in an auditory-verbal serial recall task and performance was assessed with three measures: serial recall, cluster recall and free recall. In general, recall was better when the degree of semantic relatedness was high rather than low, and the improvement was similar for all groups. Similarly, recall was better for sequences which were grammatically well formed than for sequences which were not, but the improvement of normal children was significantly greater than that of autistic children. While contemporary theories view the linguistic deficit evidenced by autistics as a deficit in the processing of semantic information, these results show that autistics have a deficit in processing syntactic information. However, this study does not resolve the issue of whether the deficit is specific to autism or a more general, developmental deficit, for while autistics differed from matched normal children, they did not differ from matched retarded children. Methodological problems in the construction of natural language sequences which vary independently in semantic and syntactic features are discussed, and it is suggested that sequences with artificial syntactic and semantic (associational) structures should be used to resolve the question of whether autistics have a deficit in processing semantic information.