Further evaluation of the accuracy of reinforcer surveys: a systematic replication.

The team asked staff to fill out a quick reinforcer survey. Then they watched the same clients pick items in a 5-minute concurrent-operants test.

They compared the two lists to see if the survey guessed right.

The survey was wrong 43 % of the time. It called an item "preferred" even when the client rarely touched it in the test.

Only 57 % of survey answers matched the real choices.

Ferrari et al. (1991) saw the same problem with the Motivation Assessment Scale. Staff rarely agreed on the function of behavior, so the survey scores were shaky.

Boyle et al. (2019) later showed that some free-operant checks also miss top items. Together these papers say the same thing: a quick paper survey is not enough.

Stolz (1977) warned that many studies skip daily reliability checks. Without those checks, both surveys and brief tests can look better than they really are.

Before you build a treatment, run a 5-minute concurrent-operants test. Let the client choose between items while you count responses. If the survey and the test clash, trust the test. This small step saves you from using weak reinforcers and speeds up learning.