Generalized instructional control and the production of broadly applicable relational responding.

Two experiments examined the performance of preschool children in tasks requiring the generalized matching of faces to faces and names to faces under the control of instructional stimuli (background color) that specified the basis by which faces were to be matched on a given trial. The children first learned to recite all the names, and to select all the faces, in a fixed order (the forward order). They then learned to select the faces in response to their spoken names. When the faces appeared on a white background, subjects selected the face named. On a gray background, they selected the face whose name was next forward. Subsequently, over a series of tests, when subjects were presented with novel, but similar stimuli with the same names, and with completely novel stimuli with novel names, control by the white and gray background colors generalized. In the second experiment, on trials with the gray background, when the face bearing the next-forward relation was not present, the children learned to select an appropriate substitute (two faces forward). This performance also generalized to novel stimuli. Together with earlier findings, these data suggest that widely generalized relational matching performances may arise because the labels for these relations are generically and metaphorically extended tacts.