Response generalization

The response changes while its function remains related

A person may be taught to ask, “Can you explain that?” and later say, “What does that mean?” without direct teaching. Both responses may obtain clarification. The second form may represent response generalization if it was genuinely untrained and occurs in the relevant context.

Minor pronunciation or movement variation may belong to one response class rather than a new generalized response. Define the unit before collecting data.

Stimulus generalization changes another dimension

Stimulus generalization occurs when a learned response happens in the presence of new stimuli. Asking the taught clarification question to a new instructor is stimulus generalization. Using a new clarification phrase with the original instructor is response generalization.

Both can occur together: a new phrase may appear with a new person. Report each dimension so “generalized” does not hide what changed.

Novel means absent from the teaching history

A phrase copied from a model, prompted during correction, or rehearsed at home is not untrained in the strict sense. Before probes, list trained forms, examples heard, and any acceptable novel class.

Exposure can be indirect. A communication partner’s repeated wording may shape later responses. Honest records can describe a response as “not directly trained by the team” when the wider history is unknown.

Generalization should remain useful

Variation is beneficial when new responses are understandable, efficient, safe, and aligned with the person’s goal. A broad response class can also include unclear or harmful forms. Plan discrimination and partner responses alongside generativity.

Communication partners should honor recognizable speech, sign, gesture, writing, and AAC. A novel spoken form should not receive more value merely because it looks more typical.

A fictional response-generalization record

Amara chooses a goal of asking for clarification during a craft workshop. Teaching uses one AAC message: “Show me that step again.” During nine independent novel opportunities, she uses three untrained forms: “Which part?”, a pointing-plus-question gesture, and “Again, please.”

At least one effective untrained form occurs in 6 of 9 opportunities. Partners respond within ten seconds in 5 of those 6. The response-generalization denominator is nine opportunities; partner responsiveness uses six messages. These counts do not show which teaching component produced the variation.

Program variation without scripting every form

Teach a functional core, accept multiple reliable forms, vary models, and avoid correcting harmless differences. Natural partner feedback can support forms that work in daily life.

Stokes and Baer describe a broad technology for programming generalization. Their categories support deliberate planning rather than one fixed procedure.

Marzullo-Kerth and colleagues reported unscripted sharing offers after a treatment package with multiple exemplars, video modeling, prompting, and reinforcement. Four participants and a package design limit claims about one procedure.

Language generativity is broader

Novel language can involve new words, sentences, relations, and context-sensitive meaning. Barnes-Holmes and colleagues argue that response generalization alone does not fully explain language generativity and discuss derived relational responding as another account.

Clinicians can report observable new forms without turning one term into a complete theory of language development.

Measure form, function, and opportunity

Predeclare eligible opportunities, trained forms, novel forms, prompts, partner response, and exclusions. Report distinct novel forms, total novel responses, and opportunities with any appropriate novel response.

Keep errors and unwanted generalization visible. Check maintenance and use across people, settings, and materials. Ask whether the person finds the new forms effective and comfortable.

The BACB BCBA Test Content Outline, 6th edition includes generalization, maintenance, and response classes in examination content. It is not a person-specific plan.

Variability and mastery answer different questions

A person can produce many forms inconsistently, or one form very reliably. Count response diversity and opportunity-level effectiveness separately. A diversity count might be three distinct novel messages; an effectiveness measure might be six successful clarification requests across nine eligible opportunities.

Avoid rewarding novelty for its own sake when a stable response works well. Some people prefer a consistent AAC selection or scripted phrase. Response generalization should expand useful options rather than make variation a compliance requirement.

When partners understand only one form, the barrier may be partner training. Record whether each recognizable response is acknowledged and honored within the defined window. Teaching the person more variants cannot compensate for a system that ignores their existing communication.

Check maintenance without demanding every form reappear. The relevant outcome may be continued access to at least one effective, chosen response.