Management of chronic psychiatric patients by a token reinforcement system.

Winkler (1970) set up a token economy on a locked adult ward. Patients earned plastic tokens for self-care and job tasks.

Later phases removed the pay or added fines. The team watched whether old problem behaviors returned.

While tokens were contingent, patients stayed on task and used fewer odd institutional behaviors.

When tokens stopped or were given free, the gains melted away within days.

Bennett et al. (1973) ran a near-copy on the same ward and saw the same lift in social talk, a conceptual replication.

Rogers-Warren et al. (1976) moved the idea to a prison and still got high work output, showing the rule extends outside hospitals.

Skrtic et al. (1982) looked like a contradiction: removing tokens did not hurt stutterers’ fluency. The key difference is the behavior. Fluency had already been shaped and was maintained by natural praise, while psychiatric ward tasks had no built-in payoff once tokens vanished.

The paper warns you to guard the contingency. If you fade tokens, first weave in other reinforcers like praise, activity choice, or social feedback. Check each client: when tokens disappear, does the skill stay or crash? If it crashes, keep the tokens or find a new payoff before you let go.