A skill that only happens with one RBT, in one room, with one set of cards is not finished. The point of teaching is for the learner to use the skill wherever it matters: at home, at school, in the store, and with the people in their life. This lesson covers how that spread happens and what you do to help it along.
Why it matters
Generalization often does not happen on its own. If a learner is taught to say "cup" with the same blue cup at the same table every day, the word may become tied to that cup and that table. When a different cup shows up at grandma's house, the learner may say nothing. Behavior analysts sometimes call the hands-off approach "train and hope": teach in one narrow way and hope the skill spreads. It often doesn't.
For RBTs, generalization is part of everyday work. Your supervisor writes generalization into the plan, and you carry it out: varying materials, rotating settings, running probes, and reporting what you see.
Key ideas
- Plan for it from day one. Generalization is programmed, not assumed. Good plans list the settings, people, and materials the skill should eventually occur with.
- Vary what does not matter, keep what does. The relevant feature (the "cup-ness" of a cup) stays constant, while irrelevant features (color, size, material) change across teaching.
- Teach multiple exemplars. Use enough different examples of the target that the learner responds to the shared feature, not to one particular item, person, or place.
- Program common stimuli. Bring features of the place where the skill is needed into teaching, such as the same type of visual schedule the classroom uses.
- Train loosely. Vary noncritical parts of the session (where you sit, your tone, the time of day, background noise) so no single detail controls the response.
- Lean on natural reinforcers. Skills that contact reinforcement in daily life, like getting juice after asking for it, are more likely to keep going outside sessions.
- Make reinforcement less predictable. Once a skill is solid, an unpredictable intermittent schedule (an indiscriminable contingency) makes it hard for the learner to tell when reinforcement is available, so responding is likelier to continue in new situations.
- Probe, don't assume. A generalization probe checks whether the skill occurs in an untrained condition, usually before any teaching happens there.
Definitions
- Stimulus generalization: a learned response occurs in the presence of stimuli that were not part of teaching but share features with the trained ones, such as a new person, setting, or set of materials.
- Response generalization: the learner emits untrained responses that serve the same function as the trained response. A learner taught to say "hi" starts saying "hello" too.
- Multiple exemplar training: teaching with several varied examples of a stimulus or response so the skill extends to new, untaught examples.
- Programming common stimuli: including stimuli from the generalization setting in the teaching setting so the two share features.
- Training loosely: deliberately varying noncritical aspects of instruction so that no single irrelevant feature gains control over responding.
- Generalization probe: a test of the skill under conditions that were not used in teaching, typically without prompts, to see whether it occurs there.
- Overgeneralization: the response occurs in the presence of stimuli where it does not belong, such as calling every four-legged animal "dog."
Generalization is about new conditions. Maintenance is about time: does the skill keep happening after teaching ends? The two are related, and the next topic covers maintenance in detail.
Examples from a session
Settings, people, and materials
- Materials (clinic): Leo, 5, is learning to label "shoe." His RBT, Hana, teaches with sneakers, sandals, boots, and photos of shoes, rotating them every session as the plan says.
- People (home): Ruth, 72, is relearning to use a picture board to ask for help after a stroke. As the plan directs, her RBT shows her daughter and the evening caregiver how to prompt and respond the same way, so the skill does not depend on one person.
- Settings (school): Andre, 12, learned to check his visual schedule at the clinic. His BCBA has the RBT use the same schedule format his classroom uses, then run a probe in the classroom to see whether Andre checks it there without help.
- Non-example (maintenance): Two months after teaching ended, Leo still labels the same sneakers at the same clinic table. No condition is new, so this shows maintenance, not generalization.
In each case, the RBT is not inventing a new program. They are carrying out the generalization steps written into the plan and recording data on them.
Tip
If you notice a skill happening in session but not at home or school, that is valuable information. Write it down and tell your supervisor. Don't change the program on your own.
Common mistakes / exam traps
- Confusing generalization with maintenance. New setting, person, or material points to generalization. Weeks later, same conditions, points to maintenance.
- Teaching during a probe. If you prompt or teach before recording, you are measuring teaching, not generalization.
- Waiting until mastery to think about generalization. Most plans build in varied exemplars and settings during acquisition, not only afterward.
- Varying the wrong thing. Changing the critical feature (using a bowl when teaching "cup") teaches the wrong discrimination. Vary only what your plan identifies as irrelevant.
- Treating overgeneralization as success. A response showing up everywhere is not the goal. If "dog" is used for cats and cows, the learner needs discrimination training, which your supervisor will plan.
Exam trap
"Train loosely" changes the setting details around the target. "Multiple exemplars" changes the examples of the target itself. Read which one the scenario varies.
Exam trap
Stimulus generalization: the same trained response shows up under new conditions (a new person, place, or material). Response generalization: a new, untaught form of the response appears that serves the same function. Ask what is new, the situation or the behavior.
Quick check
An RBT teaches a learner to request 'help' with a hard-to-open jar. A week later, at school, the learner asks a teacher for 'help' with a stuck zipper, which was never taught. What happened?