Many severely subnormal patients talk little to each other. In this experiment, three pairs of subjects were reinforced for talking to each other, and learned to do so quite quickly. Whether social speech would continue to occur without the benefit of external reinforcement was examined by observing the subjects through a one-way mirror in a bare interview room adjacent to the teaching room immediately after each training session. On some occasions untrained subjects were observed in the bare room with the trained subjects.

The reinforcement of social speech was demonstrated to be effective by the use of a reversal design (baseline, reinforcement, no reinforcement, reinforcement), where the rate of speech increased considerably when reinforcement was available but decreased when it was discontinued. Generalization of the increased social speech, however, was very poor and only significantly above baseline levels with the pair who seemed responsive to social as well as material reinforcement. The implications of this for training programmes are discussed.