Experimental analysis of human vocal behavior: applications of speech-recognition technology.
A computer can listen, score, and pay for vocal responses during equivalence training — and the new words still pop out.
01Research in Context
What this study did
Researchers used speech-recognition software to train adults to say nonsense words.
The computer listened for the right sounds and gave points right away.
Each person got two rounds of training to see if new words would pop out without more teaching.
What they found
Most adults started saying new words that belonged to the same group as the trained words.
The speech-recognition setup worked — it caught the words and paid off fast.
Only two short training rounds were enough to see this transfer.
How this fits with other research
Pérez‐González et al. (2026) took the same idea further. They used the same equivalence training but pushed adults to solve analogies like 'A is to B as C is to ?' after training.
Cox et al. (2025) also used computers to handle reinforcement, but instead of shaping speech they used AI to guess the next response. Both papers show that pairing tech with ABA can speed things up.
Becraft et al. (2020) gives you the tool to pool many small studies like this one. Their primer shows how to meta-analyze single-case data, so future replications of the speech-recognition method can be combined into one big picture.
Why it matters
You can run vocal equivalence drills without sitting next to the learner. The computer scores and pays for correct sounds in real time. This frees you to watch body language or run other kids at the same time. Try it next time you need clean data on emergent verbal relations.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Record three nonsense words into free speech-recognition software and set it to play a ding plus a point when the learner says each word — run a quick probe for emergent relations.
02At a glance
03Original abstract
Recent developments in speech recognition make it feasible to apply the technology to study vocal behavior. The present study illustrates the use of this technology to establish functional stimulus classes. Eight students were taught to say nonsense words in the presence of arbitrarily assigned sets of symbols consistent with three three-member experimenter-defined stimulus classes. Computer-controlled speech-recognition software was used to record, analyze, and differentially reinforce vocal responses. When the stimulus classes were established, students were taught to say a new nonsense word in the presence of one member of each stimulus class. Transfer of function was tested subsequently to determine if the novel stimulus names transferred to the remaining stimulus class members. Most subjects required two iterations of the training and testing procedures before transfer occurred. The data illustrate the usefulness of recording vocal behavior during stimulus control procedures and demonstrate the use of speech-recognition technology. The paper also describes the current state of speech-recognition technology and suggests several other areas of research that might benefit from using vocal behavior as its primary datum.
Journal of the experimental analysis of behavior, 2000 · doi:10.1901/jeab.2000.74-363