Understanding how learners talk to themselves during complex tasks can shape how clinicians teach and test derived relations. This study examined whether spoken self-talk corresponds with performance on auditory matching tasks—a question with real implications for how we interpret results and support learners with different verbal repertoires.
What is the research question being asked and why does it matter?
The study asked a straightforward question: when people complete a go/no-go successive matching-to-sample (S-MTS) task with sounds, does their spoken self-talk correspond with their accuracy?
In this task, the learner hears one sound, then another, and must either click (go) if the sounds “go together” or wait (no-go) if they do not. The researchers wanted to know whether correct responding might be supported by verbal behavior—like naming the sounds or stating rules such as “these two go together.”
This matters because many learners, especially those with strong verbal skills, may talk to themselves while solving hard discrimination tasks. In clinical work, that might look like a client whispering labels, repeating a rule, or quietly listing group members. If self-talk helps, you might allow it, teach helpful self-talk, or track it as part of why the learner is succeeding or struggling. If self-talk interferes—or if the situation requires silence—you may need to adjust how you test for truly “learned” relations.
It also matters because S-MTS is one way to teach auditory-to-auditory relations, which are difficult in standard matching-to-sample since you cannot present multiple sounds simultaneously. If verbal behavior is doing part of the work, then S-MTS outcomes may not mean the same thing for learners with limited speech or naming skills. Clinicians need to know when results are tied to language abilities.
What did the researchers do to answer that question?
Eight college students completed the study remotely on a laptop. The stimuli were short clips of common sounds (a toaster, a sliding door, a razor). The researchers arranged the sounds into three “classes,” each with three members (A, B, and C sounds within a class). Training taught AB and AC relations using the go/no-go S-MTS format, with points and prompts for correct “go” responses.
Before training, participants completed pretests for symmetry (BA and CA) and transitivity/equivalence (BC and CB). They had to score low enough to show they did not already know the relations. After training, they completed posttests without feedback. The researchers used set accuracy criteria to determine whether “emergence” occurred on the symmetry and equivalence tests.
The key addition was a talk-aloud “protocol analysis” during posttests. Participants were instructed to think out loud throughout. The researchers transcribed and coded statements as class-consistent (matched the intended relations), class-inconsistent (stated the wrong relation), or irrelevant. They also noted whether participants used standard names for the sounds or their own unique labels.
The researchers then compared how often participants made class-consistent or class-inconsistent statements with their S-MTS accuracy. This was correlational—they examined how the two moved together, not whether one caused the other.
How you can use this in your day-to-day clinical practice
When you see strong performance on derived relations tests, consider that some learners may be “solving” the task with self-talk rather than relying solely on trained stimulus relations. In this study, participants who made more class-consistent statements had more correct responses, and those who made more class-inconsistent statements made more errors. Pay attention to what the learner is doing while responding, not just percent correct. If a learner is labeling stimuli, repeating class members, or stating rules, that behavior may be supporting performance.
You can add simple, dignity-centered observation during hard discrimination or derived relations testing. If the learner speaks, quietly note whether they are naming stimuli, repeating a sequence, or stating “go together/not go together.” You do not need to stop it automatically. Instead, clarify your goal. If you want functional categorization in real life, self-talk may be a helpful strategy worth allowing. If you want to measure whether relations are strong without extra support, you may want a separate condition where you reduce overt self-talk and see if performance holds.
Be careful not to over-read the findings. These were college students, not children or clinical clients. They were likely verbally skilled and used language easily. Do not assume the same pattern will appear with learners who have limited vocal speech, limited naming, or different communication modes. Also, the talk-aloud instruction may have changed how participants performed. In your setting, requiring a client to talk during testing could add extra demands, increase stress, or change the skill you are actually measuring.
If you suspect helpful self-talk is supporting correct responding, you can shape it respectfully—but only when it matches the learner’s goals and preferences. Some learners benefit from learning stable labels for stimuli (consistent tacts) before complex auditory discrimination work. In this study, everyone used tacts in their statements, either common labels or their own. Clinically, that suggests learning may improve when the learner can reliably label stimuli in a way that makes sense to them. You can accept the learner’s own label if it is consistent and useful, rather than forcing your label, as long as it supports communication and does not create confusion across people and settings.
Use the study as a reminder to separate teaching conditions from testing conditions. During teaching, allowing or prompting a learner to label and state rules might reduce errors and help them contact patterns faster. During testing, keep conditions steady and minimal so you know what is maintaining performance. If performance drops when self-talk is unavailable, that does not mean the learner failed. It may mean the skill is still dependent on a strategy, and your next step is to decide whether that dependency is acceptable for the learner’s life.
If you run go/no-go tasks that include both “click” and “do not click,” watch for incorrect verbal rules developing. This study showed that wrong statements tracked with wrong responses. A learner might say “this one never goes with that one” and then get stuck avoiding a correct response. When you see persistent errors, consider an error analysis that includes what the learner is saying or signing, not just what they are clicking or touching. Your intervention might be clarifying labels, reducing confusing similarity between stimuli, or reteaching relations with clearer contrasts—rather than simply adding more trials.
Finally, treat the findings as a cue to individualize, not a rule to follow. For some learners, adding spoken strategies may be supportive and empowering. For others, it may be distracting or uncomfortable, especially if prompted to perform speech outside their goals. Use caregiver and learner input, observe assent and stress signs, and choose the smallest change that answers your clinical question.
Works Cited
Sordello, J. C., Hanson, R. J., Miguel, C. F., Angulo, A., & Dingus, C. S. (2024). Correspondence between vocal-verbal behavior and go/no-go responses during the successive matching-to-sample procedure. The Analysis of Verbal Behavior, 40, 76–87. https://doi.org/10.1007/s40616-024-00205-7
