← All study notes
Study Notes

Stimulus vs. Response Generalization: Where Does the Novelty Live?

14 minute read

Across several BA Prep questions testing stimulus and response generalization, each item has been answered hundreds of times — and on multiple questions, fewer than half of those responses were correct. The pattern suggests that this distinction is harder in practice than the definitions make it look.

That tells us something important. This is not one obscure edge case tripping people up. The same conceptual boundary is producing difficulty even when the examples change.

And the problem usually is not that they have never seen the definitions. Most people studying for the BCBA exam can tell you something like this: stimulus generalization means behavior occurs with new stimuli, response generalization means new responses occur. That is a reasonable place to start. It is also not quite enough.

Because the harder questions change the person, setting, materials, response form, and function at the same time. Suddenly, asking only "did the stimulus change or did the response change?" can get you into trouble.

A better question is: what was directly trained — and what appeared without direct training?

That is where the distinction starts becoming useful.

The Short Version

Stimulus GeneralizationResponse Generalization
What extends beyond training?The stimulus conditionsThe response repertoire
What do you observe?Established responding under untrained conditionsRelevant untrained response forms
Common exampleA learned skill occurs with a new person or in a new settingA new way of accomplishing the trained function appears
Question to ask"Where else did the established behavior occur?""What else did the person do?"

A useful first-pass shorthand: stimulus generalization expands where established responding occurs. Response generalization expands how the relevant behavior can occur. That shorthand is useful — just do not turn it into an inflexible definition.

Stimulus Generalization: Established Behavior Under Untrained Conditions

Suppose a learner is taught to greet a therapist by saying, "Good morning." Later, without direct teaching, the learner says "Good morning" to a classroom teacher, a bus driver, and a receptionist.

The relevant response has already been established. What expanded? The stimulus conditions under which that responding occurs. The learner is now engaging in established responding with people and circumstances that were not part of the original teaching arrangement. That is stimulus generalization.

The same basic pattern can occur across people, settings, materials, examples, instructions, objects, or other relevant stimulus conditions. The important feature is not simply that something in the environment looks different. It is that established responding occurs under stimulus conditions that were not directly trained.

Physical Similarity Is a Clue, Not the Definition

Textbooks often introduce stimulus generalization with physically similar stimuli. A pigeon learns to respond to one wavelength and later responds to nearby wavelengths. A learner is taught to identify one red object and correctly identifies other red objects. A dog trained to respond to one whistle pitch responds to similar pitches.

Those are useful examples. But do not accidentally learn "looks similar = stimulus generalization." Physical similarity can influence generalization, and generalization gradients can show systematic changes in responding across a stimulus dimension. But physical resemblance is not the entire concept.

Did established responding occur under an untrained stimulus condition? If yes, stimulus generalization may be the relevant process.

Response Generalization: The Repertoire Expands

Now change the example. A learner is directly taught to request assistance by saying, "Help, please." After that response is established, the learner begins using other appropriate forms that were not directly taught: raising a hand for assistance, presenting a help card, or gesturing toward the difficult task while seeking help.

Now the important change is not primarily where the learned response occurs. The response repertoire itself has expanded. Relevant, untrained response forms have emerged following training. That is response generalization.

Stimulus generalization extends established responding across conditions. Response generalization extends performance across response forms. Notice the word untrained doing a lot of work there. If the analyst directly taught the vocal request, the hand raise, the help card, and the gesture one by one, the existence of four response forms would not by itself demonstrate response generalization. They were all trained.

The Most Important Word in This Article: Untrained

A surprising number of generalization questions can be solved by identifying what was not directly taught.

Imagine a learner receives direct instruction with five different community helpers: a teacher, police officer, firefighter, nurse, and librarian. The learner then responds correctly to each of those five people. That demonstrates successful learning. But you do not need generalization to explain responding to the exact examples that were directly trained.

Now present a paramedic who was never part of instruction, and the learner responds appropriately. That untrained performance provides evidence of generalization.

The same logic applies to response generalization. If three communication forms were individually taught, observing all three later does not prove response generalization. If one was trained and another relevant form emerges without direct teaching, now response generalization becomes a much stronger interpretation.

Exam Trap

Novel-looking behavior is not automatically generalized behavior. Ask whether the relevant stimulus condition or response form was actually outside direct training.

Why "Stimulus Changed vs. Response Changed" Works — Until It Doesn't

You may have learned a shortcut like this: stimulus changes, response stays the same → stimulus generalization; response changes, stimulus stays the same → response generalization. Keep it — it is a good heuristic for clean examples. Just do not mistake it for a law.

Real behavior does not politely change only one variable at a time. Suppose a learner is taught to request a break vocally during table work with one therapist. Later, in an art class with a teacher, the learner hands over a break card.

Now what changed? The person changed. The activity changed. The setting changed. The response form changed. Trying to force that entire event into stimulus or response generalization simply because an exam mnemonic told you only one thing is allowed to change would be a mistake. The data may support both.

Yes, Both Can Occur

Stimulus and response generalization are not mutually exclusive boxes. Consider this progression: a learner is taught to say "Can I have a break?" during academic work with one therapist. Later, the learner says the same phrase to a classroom teacher during an untrained activity, and the learner also independently uses an untrained break card to request the same outcome.

The first pattern provides evidence of stimulus generalization: established responding occurs under untrained conditions. The second provides evidence of response generalization: an untrained relevant response form appears. Nothing in behavior analysis requires us to pretend only one occurred. Natural behavior is frequently messier than textbook examples.

Think About It This Way
New conditions, established response look for stimulus generalization
New relevant response form look for response generalization
Both dimensions expand evidence may support both

The exam may still ask you for the best description of one particular part of the pattern. Read the stem carefully enough to identify which relation it is actually asking you to classify.

Response Class Is Not the Same Thing as Response Generalization

This is one of the strongest distractors. Suppose a child says "toy," points toward the toy, and guides a caregiver toward the toy. Each response reliably produces access to the same toy. Those different topographies may belong to the same response class because they share a relevant functional relation to the environment.

But that fact alone does not tell you that response generalization occurred. Why? Because response class and response generalization answer different questions. A response class asks: which responses can be grouped together based on a relevant shared function or effect? Response generalization asks: did relevant response forms occur that were not directly trained?

If the child already used all three forms before teaching began, you might have a response class with no evidence that the intervention produced response generalization. If only one form was taught and the other relevant forms then emerged without direct teaching, you may have both a response class containing multiple forms, and response generalization demonstrated by the emergence of untrained forms.

A class describes how behavior is organized. Generalization describes extension beyond direct training.

Stimulus Discrimination and Stimulus Generalization Can Coexist

Another common trap is treating discrimination and generalization as though one must disappear for the other to exist. That would make effective behavior rather difficult.

Suppose a child learns the tact "dog." We would generally want the child to say "dog" in the presence of a Labrador, a poodle, a German shepherd, and unfamiliar dogs never used during teaching. That is desirable generalization. But we probably do not want the child saying "dog" to cats, horses, squirrels, or every four-legged animal.

Effective behavior therefore often requires generalization within relevant boundaries and discrimination outside those boundaries. The learner generalizes across appropriate examples while discriminating nonexamples. These processes can coexist.

Exam Trap

Do not reason "the learner discriminates, so generalization cannot be occurring," or "the learner generalized, so discrimination must have failed." The useful question is whether responding has extended appropriately without becoming indiscriminate.

Generalization Is Not Automatically Good

The word generalization sounds positive because treatment plans often aim to promote it. But generalization is a behavioral relation, not a gold star.

A child taught to identify one dog may initially call wolves, cats, and stuffed animals "dog." Responding has extended beyond the original training stimulus — but it has extended too broadly for the instructional goal.

Likewise, the appearance of new response forms is not automatically response generalization. Suppose a learner is taught an appropriate way to request escape and later begins screaming, throwing materials, and running from the room. Those forms may represent response variability, and they may belong to a common functional class if the relevant behavior–environment relations are demonstrated. But novelty by itself does not establish response generalization. The important question is whether relevant, untrained response forms emerged in relation to the trained performance.

Generalization vs. Maintenance: Where Else vs. How Long

Maintenance is another tempting neighbor. Suppose a learner is taught to independently check a work schedule. Three months later, in the same work environment under essentially the same relevant conditions, the learner still checks the schedule independently. That is evidence of maintenance.

Now suppose the learner also begins checking schedules in a different classroom and later at a volunteer site where the skill was never directly trained. That provides evidence of stimulus generalization. The learner could show maintenance without demonstrated generalization, generalization without long-term maintenance, or both.

Generalization asks where else or what else. Maintenance asks whether the change persists across time after acquisition conditions change.

Do not use "they still do it later" as automatic evidence of stimulus generalization.

A Generalization Procedure Is Not Proof That Generalization Happened

This distinction matters especially because the 6th Edition TCO separates identifying stimulus and response generalization in B.14 from designing and evaluating procedures to promote them in G.15.

Behavior analysts can deliberately program for generalization. Stokes and Baer's classic review argued that generalization should not simply be left to "train and hope." They organized strategies such as training sufficient exemplars, training loosely, programming common stimuli, and arranging natural maintaining contingencies.

But here is the exam trap: using a procedure intended to promote generalization does not prove that generalization occurred. If a teacher trains multiple exemplars, that may increase the likelihood that performance will extend. The outcome still has to be demonstrated. You would need evidence that the relevant behavior actually occurs with untrained stimuli, settings, people, or response forms, depending on what type of generalization you are evaluating. Procedure and outcome are not the same thing.

Put the Common Neighbors Side by Side

PatternBest ConceptWhy
A trained response occurs with an untrained personStimulus generalizationEstablished responding transferred to a new stimulus condition
An untrained relevant response form emergesResponse generalizationThe response repertoire expanded beyond direct teaching
Several forms already produce the same outcomeResponse classThey share a relevant functional relation; emergence is not established
Responding differs appropriately across conditionsStimulus discriminationThe learner responds differentially rather than simply extending responding
The skill is still present months laterMaintenanceThe key dimension is persistence over time
Multiple exemplars were deliberately taughtGeneralization programmingA procedure was used; untrained performance still must be demonstrated

The words around generalization are similar enough that memorizing them independently is inefficient. Study the relation each term describes.

Five Examples, Worked

Same Skill, New Person

A technician teaches a learner to scan an employee badge before entering a work area. During an untrained probe, the learner performs the same sequence when a different supervisor is present. Stimulus generalization — the established response occurs under a stimulus condition not used during training.

New Form of the Same Useful Behavior

A learner is taught to request clarification by saying, "Can you explain that again?" Later, without direct teaching, the learner raises a hand and asks, "What does that mean?" Response generalization — a relevant untrained response form has emerged.

Both Dimensions Change

A learner is taught to request assistance vocally from a therapist during desk work. Later, at a community activity with an unfamiliar staff member, the learner independently shows the staff member a help symbol. Both may be demonstrated — the performance extends to new stimulus conditions and includes an untrained relevant response form.

Several Responses, But All Were Taught

A learner is directly taught to request a break by speaking, signing, and exchanging a card. Later the learner uses all three. Not enough information to call this response generalization — the observed response forms were directly taught.

Same Performance Much Later

A learner masters a handwashing routine. At a follow-up three months later in the original training environment, the routine remains intact. Maintenance — the passage of time is the relevant dimension. You would need additional evidence to conclude that the routine generalized to new conditions.

Another Trap: "New" Does Not Tell You What Is New

Consider the statement: "the learner performed the skill in a new situation." That sounds like stimulus generalization. But new situation is vague. Did the same established behavior occur with a new person? Did a new response form appear? Was the behavior actually trained in that setting last month? Did both the stimulus conditions and response form change?

The word new is not a diagnosis. Break the example into its components.

Complete the Sentence
The directly trained performance was __________
The untrained part of the later performance was __________

Then classify that untrained dimension. That is more reliable than scanning for words like new, different, novel, or generalized.

The 10-Second Decision Rule

The useful shorthand is not simply stimulus changed versus response changed. It is: what did training establish, and what extended beyond it?

Try It Yourself

Before reading the explanation, classify this one the way you would on exam day.

Quick check — try it before you keep reading

A learner is directly taught to request a break by saying "break" during table work with one therapist. During an untrained classroom activity with a different teacher, the learner says "break." Later in that same classroom, the learner independently hands the teacher a break card, a request form that was not directly taught in the program. Which interpretation is BEST supported?

The vocal request occurring with a different teacher during an untrained classroom activity provides evidence that established responding extended to new stimulus conditions. The later use of an untrained break-card response provides evidence that the relevant response repertoire expanded beyond the directly taught vocal form. Maintenance does not capture the critical relation, since the stem is not simply showing that the original skill persisted across time. Most importantly, the presence of one form of generalization does not cancel the other — the scenario supplies evidence for both dimensions.

The Bigger Lesson

The hard part of generalization questions is rarely memorizing what the two words mean. The hard part is identifying which aspect of performance extended beyond direct training. That is why easy examples help initially but can become dangerous if you turn them into rigid formulas.

"Same response, new stimulus" is useful. "New response, same stimulus" is useful. But behavior does not owe your mnemonic anything. People encounter new settings, new materials, new listeners, new motivating conditions, and new opportunities to respond — often at the same time. Your job is to identify the relation the evidence actually demonstrates.

Final Takeaway

Stimulus generalization: established responding extends to untrained stimulus conditions. Response generalization: relevant untrained response forms emerge beyond direct training.

Then ask one more question: what was directly trained — and what generalized beyond it? If the answer is new conditions, look toward stimulus generalization. If the answer is new response forms, look toward response generalization. If both changed in ways the evidence actually supports, do not be afraid to say both. And if every example was directly trained, slow down before calling anything generalization at all.

One More Round

ScenarioBest Interpretation
A trained greeting occurs with unfamiliar staff membersStimulus generalization
A new appropriate greeting form emerges without direct teachingResponse generalization
The learner uses several greeting forms that were all explicitly taughtDirectly trained repertoire; not yet response generalization
Appropriate greetings occur with adults but not peers after differential trainingStimulus discrimination
The trained greeting remains intact six months laterMaintenance
The skill occurs with new people and a new untrained response form appearsBoth forms of generalization may be present

Don't classify generalization from what looks different. Identify what training established, then find what occurred beyond direct training.

Knowing the definitions is only the starting point. BA Prep questions are designed to make you separate stimulus and response generalization when people, settings, response forms, and neighboring concepts change together — the kind of discrimination required by the BCBA 6th Edition Test Content Outline.

Start Free Trial

References

Behavior Analyst Certification Board. (2022). BCBA test content outline (6th ed.). Behavior Analyst Certification Board. Updated September 2024. The current outline places identifying and distinguishing between stimulus and response generalization in B.14 and designing and evaluating procedures to promote generalization in G.15. https://www.bacb.com/wp-content/bcba-outline-6thEd/

Cooper, J. O., Heron, T. E., & Heward, W. L. (2020). Applied behavior analysis (3rd ed.). Pearson.

Stokes, T. F., & Baer, D. M. (1977). An implicit technology of generalization. Journal of Applied Behavior Analysis, 10(2), 349–367. https://doi.org/10.1901/jaba.1977.10-349