TWELVE CHRONIC HOSPITALIZED FEMALE PATIENTS RECEIVED TOKEN REINFORCEMENT CONTINGENT ON TWO SEPARATE CLASSES OF VERBALIZATIONS: (a) positive statements about optional activities available in the hospital setting, and (b) positive statements about people. Cross-class generalization of reinforced verbal responses about activities to overt behavior was tested by actual participation in activities; within-class generalization of verbal responses about people to verbalizations in another stimulus setting was assessed in a structured interview situation. A multiple baseline design with contingency reversals was employed to demonstrate experimental control of both classes of verbalizations in the group sessions. Positive statements about activities generalized to actual participation in activities, while generalization of positive statements about people to verbalization in the extragroup setting did not occur.