Differential effects of four token conditions on rate and choice of responding in a matching-to-sample task

In a two-phase experiment, four different token conditions were presented each session to retarded students in a TMR class. In Phase 1, students responded by matching-to-sample under the following conditions: (A) correct responses produced exchangeable tokens, (B) correct responses produced exchangeable tokens and incorrect responses lost tokens, (C) correct responses produced nonexchangeable tokens, and (D) no-tokens given. In Phase 2, the students chose the condition in which they worked. Results showed that responses cost decreased errors but did not influence correct responding, that each baseline condition produced less correct responding than the token conditions, and that Condition C produced more responding than D. Phase 2 results showed a preference for Condition A and no difference between Conditions B and C. Results were discussed in terms of baselines, response cost, and type of datum.