Token Economy With 42% Treatment Integrity
Jordan is a 7-year-old diagnosed with ADHD who attends a clinic three days per week for two-hour sessions targeting on-task behavior and aggression. Your token economy specifies a token every two minutes of on-task behavior and 15 tokens exchanged for five minutes of iPad access. An independent observer completed treatment-integrity checklists across 12 sessions: tokens were delivered contingently in only 42% of intervals, prompts were delivered on 78% of opportunities, and the exchange menu was missing in three sessions. Four new behavior technicians were hired in six weeks. The clinical director wants to abandon the token economy and try 'relationship-based therapy' because the parent reports no improvement. You have four weeks before a funding review, and the parent is asking whether to continue.
1. What does the treatment-integrity data suggest, and what is the immediate decision?
Read the worked answer
Integrity is the first hypothesis. The token economy was designed to deliver a token every two minutes of on-task behavior and to make 15 tokens exchangeable for iPad. At 42% contingent token delivery, the programmed contingency was absent more often than present; at 78% prompt delivery, discriminative stimuli were inconsistent; missing exchange menus in 3 of 12 sessions means the backup reinforcer was unavailable. A client cannot respond to a contingency that is not delivered.
Immediate decision: Do not abandon the token economy and do not attribute nonresponse to the client. Treat this as an implementation failure until integrity is corrected. Four new technicians in six weeks make skill deficits and lack of feedback more likely than motivational failure.
But verify outcome data: Check whether on-task behavior and aggression data are reliable and whether the 42% figure is stable across technicians. If one technician accounts for most lapses, target that technician; if lapses are agency-wide, revise training and supervision. The funding review is a reason to document the integrity problem, not to change the intervention prematurely.
2. Design a performance-feedback and training package to improve integrity, specifying measures and criteria.
Read the worked answer
Package: behavioral skills training plus graphed feedback.
- Instructions: One-page checklist with operational definitions of on-task, token interval, prompt, and exchange.
- Modeling: Supervisor demonstrates a 10-minute session with a confederate, including a transition and exchange.
- Rehearsal: Each technician role-plays two sessions; supervisor scores with the same integrity checklist.
- Feedback: Immediate specific feedback plus a graph of the technician's own integrity across sessions.
- Self-monitoring: Technician tallies token intervals on a clipboard for the first week; supervisor checks agreement.
Criteria: 90% integrity across three consecutive sessions before independent sessions; if below 80%, repeat rehearsal. Integrity checks twice weekly for four weeks, then once weekly. Include exchange-menu availability as a checklist item. Use a performance graph posted privately, not publicly. Track both integrity and client outcome (on-task and aggression) on the same graph. A decision rule: if integrity averages at least 85% for two weeks and on-task remains flat, then modify the intervention; if integrity is below 85%, the intervention is not yet testable.
3. What data would justify modifying versus maintaining the plan at the four-week review?
Read the worked answer
Maintain or continue if: Integrity is at least 85–90% for two weeks, the token schedule is actually in place, and on-task shows an improving trend or aggression decreases. Continue with periodic integrity checks; do not change contingencies just because the parent is impatient.
Modify if: Integrity is high and stable but on-task remains low or aggression is unchanged after a reasonable exposure (for example, 10–15 sessions). Then examine function: Are tokens conditioned reinforcers? Is two minutes too thin? Is iPad access actually reinforcing? Consider increasing token density, adding choice, or pairing tokens with praise.
Do not modify solely because the parent reports no improvement: Check whether parent data match clinic data and whether generalization is programmed. The four-week review should present integrity data, outcome data, and a decision rule. If integrity remains below 80%, the correct decision is retraining, not a new therapy brand. Document all decisions and share the integrity graph with the director.