Comparison of modelling and reinforcement procedures in increasing question-asking of mildly retarded children.
Tokens alone are enough; modeling just speeds the first few days.
01Research in Context
What this study did
The team worked with children who had mild intellectual disability. They wanted more student questions during lessons.
Three groups were run. One group earned plastic tokens for each question. Another group watched an adult model asking questions. A third group got both tokens and the model.
The design flipped back and forth. When tokens stopped, questions dropped. When tokens returned, questions rose again.
What they found
Tokens alone pushed question counts up fast. Tokens plus modeling started a little quicker, but ended at the same high level.
Modeling without tokens barely moved the numbers. The praise-only condition looked flat on the graph.
Bottom line: the tokens did the heavy lifting. The model was just a booster seat.
How this fits with other research
Azrin et al. (1969) saw the same pattern seven years earlier. Tokens beat praise for group instruction-following. The new study shows the same rule works for question-asking.
Scull et al. (1973) kept tokens running for a full school year. Kids with Down syndrome held their academic gains long after the study ended. Together these papers show tokens stick for many school skills.
Kaiser et al. (2022) meta-analysis found huge effect sizes in today’s K-5 rooms. The 1976 result is not a relic; it is the first brick in a wall that keeps growing.
Why it matters
You do not need fancy modeling scripts to grow curiosity. Start with a clear token board and backup reinforcers. Add a quick model only if you want faster take-off. This keeps therapy time cheap and teacher effort low, yet still gives big gains in student talk.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Put a 5-token strip on the desk; hand one token each time the student asks a new question.
02At a glance
03Original abstract
Questioning is an important skill, particularly for retarded children with deficits in accumulated information. Recent strategies for modifying a variety of verbal behaviors have included both token reinforcement systems and modelling procedures. However, the relative effectiveness ( i.e. , the rate of behavior change) of certain reinforcement procedures, especially with low base‐rate behavior (used alone) have recently been questioned. The present study, using an ABAB design, compared the effect of a Token reinforcement system, a Trainer‐Model procedure, and a combination of the two procedures on increasing the rate of question‐asking. Each condition was in effect for six sessions. Groups of four mildly retarded children (mean age: 10.1 yr) were prompted to ask questions about large photographs, and each of their questions was immediately answered. For the Model group, the Trainer modelled four questions for each subject for each photograph. For the Token group, each question earned the subject a point, which was exchangeable after the session for various items ( e.g. , candy, ice cream) from a “store”. For the Modelling plus Token group, the procedures were combined. The Model and the Control groups were yoked to receive the same amount of candy as the Token and the Model plus Token groups for behaviors other than question‐asking ( i.e. , being polite). The results indicated that during Baseline conditions, performance across groups was comparable and that the performance of the Control group did not change across time. The Model procedure had only minimal effects on the rate of question‐asking, while both the Model plus Token and the Token groups showed significant performance increments. The only significant difference between the Model plus Token and the Token groups was a faster rate of increase early in the first Training condition for the Model plus Token group. This difference, and the low level of performance change for the Model group, was interpreted as suggesting a facilitory effect of modelling procedures on reinforcement contingencies. The discrepancy between previous findings showing relatively minimal changes following reinforcement of low base‐rate responses and the present results were discussed in terms of various procedural differences. Educational implications of the present results were also discussed.
Journal of applied behavior analysis, 1976 · doi:10.1901/jaba.1976.9-108