What is generalization in ABA?
Generalization is a skill holding up outside the conditions it was taught in - with different people, in different places, with different materials, and over time - which is the real goal of ABA teaching rather than performance in session.
A skill that only appears with one technician, at one table, with one set of materials is not yet a usable skill. Generalization is the difference between a client who can label pictures on a flashcard and one who asks for what they want at home.
Teams plan for it rather than hope for it: teaching with varied materials and instructions, involving multiple people, practicing in the settings where the skill is actually needed, and training caregivers so it is supported outside of session.
Maintenance is the same idea across time - whether the skill is still there weeks or months later. Both are tracked with data, which is why programs keep probing mastered targets instead of dropping them the moment criterion is met.
It helps to separate the kinds. Stimulus generalization is the skill appearing with new people, places, or materials. Response generalization is the learner producing variations that serve the same purpose - asking for help in several different ways rather than one memorized phrase. A program can succeed at one and fail at the other, and they are addressed differently.
Planning for generalization means building variation into teaching rather than adding it afterwards. Using several examples of a concept instead of one, varying the wording of instructions, rotating who runs the target, and choosing goals that will meet natural reinforcement outside of session all make it more likely the skill transfers. Teaching a single exemplar to mastery and then hoping tends not to work.
The failure mode is a skill bound to its teaching conditions - correct with one technician, in one room, with one set of materials, and absent everywhere else. Because in-session data can look excellent while this is happening, teams check by probing deliberately in untrained conditions rather than inferring transfer from session performance.
Maintenance needs the same deliberate attention. Skills that are never revisited after mastery can quietly degrade, so programs schedule periodic probes of mastered targets and treat a drop as a reason to re-teach. Caregiver training is central to both: a skill practiced daily by the people around a client is far more likely to persist than one practiced only in session.