| Abstract: | A 7-yr-old bilingual boy of normal intelligence, judged by his school to be deficient in carrying out complex requests, was trained to comply with five-component instructions, e.g., “Give me/the chip/behind/the block/on blue”. Three interchangeable words or phrases were used for each component. Training proceeded in stages. First, the child was trained to identify all individual objects and actions; then to carry out requests involving only the first component, then the first two, then three, etc. On every trial, the visual setting permitted every possible response in the set. A test for generalization to nonreinforced instructions was given at each stage by giving no feedback for all instructions that included one preselected phrase. The phrase selected at each stage was one of the three that was introduced at that stage. As a further test for generalization, nonreinforced instructions were also given that included one additional component: the next to be trained (Probes Ahead). As a test for generalization across settings and instructions, several five-component instructions were presented each session in an unused classroom. These instructions used phrases, most of which were different from those being trained, and which referred to familiar classroom objects. Results showed: acquisition occurred for each stage of training, including the full five-component instruction; almost complete generalization of responding occurred to the subset of nonreinforced instructions; little or no generalization occurred to the Probes Ahead, where an additional untrained component was included; and little or no generalization was seen to the five-component classroom instructions, until training began on the five-component instructions in the training sessions. Performance was also examined for each component. Results showed that when a new component was introduced, correct performance to previously trained components declined, and was little if any superior to performance on the new component. In summary, transfer was found to untrained sentences of the same form as those being trained, even in another setting, where most of the components were different; but poor transfer was found to more complex sentences, and performance declined for previously trained components during training of a more complex sentence. Some features of the training procedure that may have affected the degree and form of transfer were discussed: the necessity for prior training in an appropriate response to the component phrases, the importance of intermixing of reinforced and nonreinforced trials, and the effects of the abruptness with which more complex sentence forms were introduced. |
