C.8 Generalization

What the outline says

Task C.8 in content area C, Behavior Acquisition (19 of the 75 scored questions): "Implement generalization procedures (e.g., conduct intervention procedures across settings, people, and stimuli)."

What generalization is

A skill that occurs only at one table, with one RBT, using one set of cards, has limited value to the client. The goal of teaching is for the client to use the skill wherever it is needed. Behavior analysts describe this in two main ways.

Stimulus generalization is when a learned response occurs in the presence of stimuli that were not part of teaching. A child taught to say "cup" with a red plastic cup at the clinic who then says "cup" for a blue mug at home is showing stimulus generalization. The TCO task lists the three you will work across most: settings (new places), people (new adults, peers or family members) and stimuli or materials (new examples of the same thing). Stimulus generalization is strongest for stimuli that share many features with the teaching stimulus.

Response generalization is when the client performs untaught responses that serve the same function as the taught one. A client taught to greet by saying "Hi" who begins saying "Hello" and waving, with the same people, is showing response generalization.

Response maintenance, the skill continuing over time after teaching is reduced or ended, is closely related and is covered in task C.9. Keep the terms straight: new stimuli is stimulus generalization, new forms of the response is response generalization, and the same skill lasting over time is maintenance.

Why generalization has to be planned

Generalization often does not happen on its own. During teaching, features that do not matter, such as the RBT's exact words, the table, the time of day or the specific picture, can gain control over the response. Then the client answers "What's your name?" at the table but not in the hallway, or only when one RBT asks. Teaching a skill and simply expecting it to appear elsewhere is called "train and hope."

For this reason supervisors write generalization steps into the plan from the start, and those steps are part of what you implement with fidelity.

Strategies you will carry out

Your supervisor chooses the strategies; you need to recognize them and carry them out as written:

• Multiple exemplar training: teach with many different examples of the stimulus, such as photos of many breeds, toy dogs and real dogs for "dog."
• General case analysis: teaching examples are chosen to cover the range of variation the client will meet, such as vending machines with coin slots, card readers, buttons and keypads.
• Training loosely: vary the parts of teaching that should not matter, such as instruction wording, your position, the room and the time of day.
• Programming common stimuli: include in teaching some stimuli from the setting where the skill is needed, such as using the same type of picture schedule at the clinic as at school.
• Natural contingencies: target skills that everyday life will reinforce, such as asking for help, and shift from contrived reinforcers like edibles toward natural ones like getting the requested item.
• Indiscriminable contingencies: once a skill is learned, reinforce on an unpredictable intermittent schedule so the client cannot tell which response will be reinforced.
• Multiple response examples: teach several forms that serve one function, such as saying "Break, please," handing over a break card or signing.
• Reinforce generalization when it happens: when the client uses a skill in a new situation, deliver the reinforcement the plan specifies.
• Sequential modification: if a skill improves in one setting but not another, the plan may introduce the procedure in each new setting in turn.

Doing it step by step

• Read the program before the session and note any list of materials, instructions, people or settings to rotate.
• Rotate exactly what the plan lists. Do not narrow it to the easiest cards or add new wordings or settings of your own.
• Run generalization probes as written. A probe tests whether the skill occurs under untrained conditions, so do not add prompts or teaching unless the plan says to.
• Record the conditions of each probe (setting, person, materials) and the client's response, and keep probe data separate from teaching data.
• Report patterns to your supervisor, such as a skill that occurs only with you or only in one room.
• Pass on requests from caregivers or teachers, such as practicing at a grandparent's house, rather than adding a setting yourself.

When generalization goes too far

Not all generalization is wanted. A child who learns "dog" and then calls cats, horses and cows "dog" is overgeneralizing. A client who learns to hug family as a greeting and then hugs strangers at the store is showing generalization that is socially unsafe. Generalization and discrimination work together: the client needs to respond the same way to stimuli that belong together and differently to stimuli that do not. Report overgeneralization to your supervisor, who may add discrimination training (task C.6).

Generalization applies to behavior reduction too. A replacement behavior such as using a break card only helps if it is honored in every setting where the client needs it, so plans often ask teachers and caregivers to respond to it as well.

How this connects to other tasks

Naturalistic teaching (C.4) supports generalization because skills are taught with the real materials, people and settings of daily life. Discrimination training (C.6) keeps generalization from spreading too far. Maintenance (C.9) asks whether skills last over time. Accurate, well-labeled data (A.4, A.8) are what let your supervisor judge generalization.

Common mistakes

• Treating a skill as fully learned because the client performs it perfectly with you. Mastery in teaching conditions does not show generalization; probes with other people, places and materials do.
• Narrowing the plan for convenience by using one set of materials or one wording. This can let irrelevant features control the response.
• Adding new settings, people or wordings on your own. Generalization steps are part of the plan, so pass suggestions to your supervisor.
• Mixing up the terms. New stimuli is stimulus generalization; new response forms is response generalization; lasting over time is maintenance.
• Reinforcing every instance of generalization, even unsafe or inappropriate ones such as hugging strangers. Report overgeneralization instead.

How it's tested

Expect short scenarios that ask you to name what is happening (stimulus generalization, response generalization or maintenance), to identify a strategy (multiple exemplars, training loosely, programming common stimuli, natural or indiscriminable contingencies) or to choose what the RBT should do. Watch whether the change is in the stimuli (new person, place or material), in the response form, or over time. In 'what should the RBT do' items, the right answer usually follows the plan as written, records data accurately and reports to the supervisor.