Effects of criterion-level probing on demonstrating newly acquired discriminative behavior.
Hard probe trials can trick you into thinking your client hasn't learned when they actually have.
01Research in Context
What this study did
The team ran a lab study with single-case design. They taught people to tell two similar pictures apart.
After each teaching round, they gave extra-hard probe trials. These probes were tougher than the training items.
They wanted to see if the probes showed what the learners really knew.
What they found
The hard probes caused lots of errors that would not go away. The mistakes stuck even when the learners had actually learned the task.
The probes made it look like no learning happened. But the learners could still do the easy items they were taught.
How this fits with other research
A-Bigby et al. (2009) saw the same problem in real interviews. When staff asked the same tricky question over and over, a large share of kids with ID changed their answers. Both studies show repeated hard questions can create false errors.
McMillan et al. (1999) also found that the tool you pick changes the answer you get. Their structured interviews gave fewer but truer reports than quick questionnaires. This matches our finding that probe type changes what you think the learner knows.
Hawley et al. (2004) and Cox et al. (2017) used single-case methods too, but in sports. They tracked how reinforcement rates shift in real games. Our lab work warns them: if you test players with harder shots than they practiced, you might miss real skill gains.
Why it matters
Stop using probe trials that are harder than your teaching set. If a learner fails a probe, do not assume they have not learned. Instead, test with items that match what you taught. This saves you from false negatives and keeps therapy moving forward.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Match your probe difficulty to your training items for the next mastery check.
02At a glance
03Original abstract
The purpose of the study was to determine whether probes of the final criterion-level discrimination administered during and after training provided an accurate measure of acquisition. Training and probe stimuli were designed to make training and probe trials initially very discriminable and then progressively less discriminable as training progressed. Initially, the discrimination required on probe trials was more difficult than the discrimination required on training trials. However, this difference in difficulty was gradually eliminated as training stimuli were topographically altered and made identical to probe stimuli by the end of training. Results showed that while correct responding was maintained throughout training, error patterns occurred on all probe trials administered during training. Error patterns developed regardless of whether probe trials occurred only at the beginning of training sessions (temporally discriminable probes) or were randomly interspersed in the training sessions (temporally indiscriminable probes). Probe error patterns seemed to be controlled by the stimulus properties of training and probe trials. Thus, probes did not measure acquisition as it occurred during training. Probe error patterns were maintained when probes were administered after completion of training. This final measure of acquisition did not agree with the demonstration of acquisition provided by the final training trial. The results suggest that probe trials can measure a different stimulus-response relationship from that trained when training starts with an easier or known discrimination and probes involve a final or criterion test of a more difficult or unknown discrimination. Stimulus control of correct responses versus error patterns is discussed.
Journal of the experimental analysis of behavior, 1983 · doi:10.1901/jeab.1983.39-479