A use of self-instruction to extend the generalization of a self-instructed in-common discrimination

Three typically developing preschool children were presented with an in-common sorting problem involving colored and marked shapes; the task was to see what two pictures had in common and to sort a deck of pictures according to that feature. The children failed to sort accurately. They were then taught to name the common features of any pair and to answer the question, "What are you looking for?" before sorting, to produce the form of a self-instruction (e.g., "I'm looking for blue triangles"). They still failed to sort accurately until they were taught to link their sorting to that potentially self-instructive answer. They then showed perfect accuracy in sorting and occasional spontaneous overt self-instructions, when told only, "Put here what these pictures have in common," across ever-changing pairs (much like the children of prior reports). The present report asks whether that finally correct performance would generalize to new stimuli. Accordingly, the children were probed with steadily changing sample pairs of three new stimulus sets-recombinations of the colored, marked shapes used in training; letters; and pictures. One child showed near-perfect generalization to all three of these new sets (like many children in prior reports). But the other two children showed near-perfect generalization to only two of the sets and not to the third Set-letters. Merely reintroducing the content-free question, "What are you looking for?" and acknowledging correct answers to it (i.e., self-instructions about letters) yielded largely accurate sorting of letter problems; this way of remediating failures of generalization had not been studied before. In general, sorting problems with two elements in common proved more difficult than those with one element in common; this difference diminished in the child told to self-instruct about letters after that self-instruction.