Generalization in ABA: Making Skills Work in Real Life

Generalization means a skill works outside therapy, with new people, places, and things. See how BCBAs plan for it from day one.

Key takeaway

Generalization is when a skill learned in therapy also shows up with new people, in new places, and with new materials. A child who asks for help only with one therapist has learned a skill.

Rad N Bad: Beyond the Blue Square — Demystifying Discrete Trial Training, Stimulus Control, and Generalization for RBTs
Watch the full CEU recording

Rad N Bad: Beyond the Blue Square — Demystifying Discrete Trial Training, Stimulus Control, and Generalization for RBTs

Sean Yocum & Michael Carrero · 0.5 CEU · 44 min
Watch on openceu.com →

Free learning path on this topic

Verbal Behavior: Beyond WH-Questions

Dana Pettus's three-part series on building generative verbal behavior. Why WH-question programs stall, how to establish speaker-listener repertoires, and how to scaffold advanced verbal skills like recall and simple inference. Three hours, three CEUs.

Start the path →3 courses · 3 CEUs · Watch at your own pace

Experts who cover this

22 recordings on openceu cover generalization.

Generalization is when a skill learned in therapy also shows up with new people, in new places, and with new materials. A child who asks for help only with one therapist has learned a skill. A child who asks for help at home, at school, and at the park has generalized it.

Generalization matters because therapy ends, but life keeps going. A skill that only lives in the clinic room does not help much. The experts below agree on one big idea. You have to plan for generalization from the start. You cannot just teach and hope it spreads on its own.

What generalization really means#

Kaelynn Partlow pushes back on a narrow view of generalization. Many new staff, often RBTs (registered behavior technicians, the frontline staff who run therapy sessions), learn that it means using lots of different cards or cues. An SD is a cue or instruction that tells the learner to respond. Partlow says varied cues alone miss the point.

Contrary to an RBT's understanding, the point of generalization isn't to evoke the same behavior in response to a variety of contrived STs. Rather, generalization should be the process of returning the previously isolated skill back to the context in which it's needed.

From the talk — Kaelynn Partlow

She shares the story of Matthew, a seven-year-old boy. His team taught him over 25 sight words with flashcards and a whiteboard. He mastered every one. Then he started class. In books and on worksheets, the words made no sense to him. Partlow says his BCBA, the board-certified behavior analyst who designs his treatment plan, had the right idea when she picked those goals. But until the skills went back to the classroom, they were "effectively useless" for him.

Her fix is simple. It is fine to break a skill into small parts to teach it. But you need a plan to put the real setting back as the learner succeeds.

You may need to break down the components of a skill in order to teach, but we must have a plan for returning the context as the learner demonstrates success with the isolated pieces.

From the talk — Kaelynn Partlow

Train and hope is not a plan#

Sean Yocum and Michael Carrero return again and again to one classic paper. It is Stokes and Baer from 1977. That paper warned the field about a common habit. Clinicians teach a skill, then hope it shows up elsewhere later. Carrero calls the paper a wake-up call.

Stokes and Bayer, what was that? 77 with the generalization paper. I think that's a big gut check.

From the talk — Michael Carrero

Carrero says the field talks about programming for generalization all the time. Yet many programs still just "pray to the behavior analysis gods." He says this critique hits table teaching hard. That is true when the teaching is stripped of real context, uses contrived rewards, and has weak plans for transfer. Contrived rewards are ones the learner would not get in daily life, like tokens or candy for a label.

Yocum puts the point bluntly. He says the "train and hope" approach is the default way behavior change fails.

train and hope is like a default failure mode of behavior change. It's not a strategy.

From the talk — Sean Yocum

Yocum names the tools from that old research: programming common stimuli, using enough examples, and using mediators. Common stimuli means bringing things from the real setting into teaching. Mediators means teaching the learner a cue, like a self-reminder, that travels with them into new places. He adds that none of this is new science. It comes straight out of the literature.

Build the plan into every goal#

Yocum argues that most ABA (applied behavior analysis) programs have a structural flaw. They treat generalization like a phase that comes after a skill is learned. He wants clinicians to flip that thinking.

It's not, generalization's not a phase. It is a design constraint.

From the talk — Sean Yocum

A design constraint is a rule you build around from the very first step. For Yocum, no goal should exist without a generalization plan. Every goal must answer one question. How does this skill survive when therapy is gone? If you cannot answer that, the goal is not ready.

He also gives a short test for any program. It checks four things that a skill must cross.

If you cannot describe how the target will transfer across people, settings, materials, natural reinforcers, then you are not done designing the program.

From the talk — Sean Yocum

Yocum gets specific about the exit plan. If the skill only works because staff hand out rewards, it is still artificial. If staff never fade out, you are building dependence into the task. So he wants plans to spell out how prompts fade and how the reward schedule thins. They should also say how staff step back. His simple rule is that if you cannot explain how a skill survives without you, do not teach it yet. Anything less, he says, "is just rehearsal."

Where table teaching fits#

Discrete trial training, or DTT, is short and repeated teaching at a table. Yocum and Carrero do not throw it out. They use a crosswalk example to show where it fits. You do not want a child learning to cross in the middle of a busy street. So you pre-teach the steps in a safe place first.

Once the child has those parts, the next step is key. You move the skill into the real world on purpose. Carrero says that is a good use of DTT, because the learner now has the skills taught in isolation. Those skills help them succeed in the natural setting.

The hosts also point to a 2001 review by Del Prado. They say it found that more naturalistic language programs did better than discrete trial ones. Studies of parent affect, or how parents felt, also favored the natural approach.

Yocum sums it up this way. DTT is a precision tool used inside a bigger, more humane "generalization first" model. You pick DTT because the learner needs it, not out of loyalty to the method. You fade from contrived to natural rewards as fast as the learner can handle. And if the learner hates DTT, Yocum says, you avoid it unless you need to teach a life-saving skill.

Natural teaching and built-in transfer#

Natural environment teaching, or NET, means teaching during real play and routines. Sean Yocum sees a big built-in win here. In NET, the skill is learned under the same sights and sounds where it must work later.

He paints a common picture. A child learns to label a picture of a shoe at a quiet table. Then the team spends six months running checks, hoping the child spots their own sneaker by the door. Stimulus control means a behavior happens reliably when a certain cue is present. Yocum says that control should not be stuck in one quiet room.

Stimulus control isn't glued to like a silent therapy room

From the talk — Sean Yocum

With real NET, he says, you skip that wait. Say you teach a child to ask for a jacket at the door when it is cold. The real setting does the heavy lifting for you. This can save families months of billable hours.

And you bypass that classic trap that Stokes and Bear warned us about decades ago. You don't train in hope.

From the talk — Sean Yocum

In another talk, Yocum connects this to natural rewards. When a request leads to getting the real item, you do not need tokens or prizes. He says that if we learn within the environment, chances are we are already programming for generalization. For a full walk through of this framework, see NET in the Wild Natural Environment Teaching & In-Vivo Data Collection.

Generality, the seventh dimension#

ABA has seven core dimensions. One of them is generality. Sean Yocum links this dimension straight to real-world outcomes. Showing a change once in one corner of a clinic does not prove much, he says. You need the effect repeated across at least three demonstrations. These can be across behaviors, settings, or the people who carry out the plan.

that's where the seventh dimension comes in with generality because a behavior change only has generality if it proves durable over time, carries across environments, holds up on everyday people in the learner's world and run the contingency.

From the talk — Sean Yocum

In the same talk, Mike Carrero warns against a shallow version of this. He calls it treating generalization like "a circus trick." A team may see a skill in the room on the left, the room on the right, and the room in back. Then they call it generalized. Carrero says that does not translate to home or the community.

For Carrero, the biggest test is whether the skill works in places without support. That, he says, is what true lifelong independence looks like. He also says honest data should guide decisions to reduce services. A clinician should be able to say plainly that a skill did or did not spread to new places.

Caregivers carry the skill home#

Many speakers say parents are the bridge between therapy and daily life. Kelly Brzak, speaking about two to five year olds, could not stress it enough.

We should be planning to generalize always. Please include parents and caregivers in programming and running trials.

From the talk — Kelly Brzak

Brzak warns against just handing parents a goal sheet to sign. She suggests covering one area with the parent each session. You could even let the parent pick from three or four topics. She also wants teachers and preschool staff involved.

Michael Carrero ties caregiver training to what "true" generalization means. In his words, programming for true generalization means programming for the least restrictive environment, meaning the most typical, everyday setting, one that does not depend on extra levels of support or reinforcement beyond what is normal for that environment.

His co-host Sean Yocum lists the gains of parent-led ABA. He says skills generalize more quickly, kids grow more independent, and rewards happen in real time.

Nicole Stewart gives an example from her ADHD work. She builds a child up to playing alone for five minutes. Then she checks if it carries over to home. If it never does, she asks, "what good is it?"

the generalization piece is really, really critical.

From the talk — Nicole Stewart

Yadi, from an interdisciplinary case talk, makes the same point. Caregivers are the ones who will carry the program across every setting. Brian Middleton goes further. He says behavior analysts should teach clients and families the science, not guard it.

Generalization requires that we empower. Generalization requires that we give power away.

From the talk — Brian Middleton

When values do not line up#

Pauli Catanese raises a quieter risk. Sometimes a parent agrees with the clinician in the meeting. But their actions at home show the goal was not a priority. It may clash with their values or culture.

Catanese says this kind of mismatch blocks generalization. If the family is not carrying the skill forward, it cannot become a generalized skill for the client. She links this to low buy-in that comes from poor cultural fit. That can lead to low implementation fidelity, which means the plan is not carried out the way it was written.

...that can carry on and become a generalized skill for the client.

From the talk — Pauli Catanese

This point fits well with what Brzak and Carrero say about parent training. Training alone is not enough. The family has to value the goal. If the goal does not match what matters to them, the skill may stall at the clinic door. So asking families what they care about is not a side task. It is part of planning for transfer.

The whole team has to row together#

Eric Soltysik describes how generalization breaks down when providers work alone. In his case study, a boy named Jacob communicates well in speech sessions. But those skills do not show up in class. Skills from his social group do not reach home or school either. Soltysik also gives a simple daily example: a client may put their shoes on for one paraprofessional (a trained aide who supports the client day to day) but refuse to do it for a different paraprofessional. That gap itself is a sign the skill never generalized.

Soltysik asks a useful question. If he can do it in one setting, what is stopping him in others? He says too often "the left hand doesn't know what the right hand's doing."

Sean Yocum names the payoff of good teamwork. He calls it the generalization gains.

I call it the generalization gains. Like when you got everyone rowing in the same direction, that's when the client wins.

From the talk — Sean Yocum

Yocum adds a twist. If we expect clients to generalize skills across settings, BCBAs should generalize their own professionalism across teams. Matt Harrington makes a related point about schools. He calls school the best testing ground because it is "completely uncontrollable." It shows which plans hold up and which only work when everything goes perfectly.

Nicole Stewart adds that a skill gap in class may not mean the child lacks the skill. Kids with ADHD often have performance deficits, which means they can do it but do not in that setting. So she looks at data from the real world, not just from session. She asks how to help the teacher bring the skill into class.

Safety skills and the generalization assumption#

Carolyn Broner looks at body safety and abuse prevention lessons. She notes that most programs use stories, videos, games, and role play. Underneath, they work much like DTT and behavioral skills training. Behavioral skills training, or BST, means teaching with instructions, modeling, practice, and feedback. Broner says that teaching logic is sound. The weak spot is somewhere else.

The issue isn't the instructional logic itself. The issue is the generalization assumption. The expectation that a child who learns and rehearses a rule in a structured setting will apply it in a real life situation.

From the talk — Carolyn Broner

Broner explains when that transfer is hardest. It may fail when a situation feels overwhelming or does not look unsafe. It may also fail when it involves someone the child loves and trusts. The evidence shows these programs can teach a rule. The harder question is what a child needs to use that rule when it matters most.

She sees the BCBA's role as cutting across home, school, and direct teaching. That includes checking for prerequisite skills and training caregivers. It also includes building help-seeking and refusal skills.

we program for generalization because a skill demonstrated in therapy is not the same as a skill used in real-world relationships.

From the talk — Carolyn Broner

Language skills that spread to new questions#

Dana Pettus shows how a teaching method can create generalization in language. Her focus is WH-questions, like "what color" versus "what is it." Telling these apart is a conditional discrimination. That means the right answer depends on which question is asked.

Pettus describes a method using autoclitic frames. These are short sentence starters that help the learner shape a full answer. She cites a 2020 study that tested this method. In that study, two children with autism learned "what is it" versus "what color." They then answered correctly for colors and animals that were never taught. They could even answer "what number" versus "what color" without extra teaching.

Pettus says a 2021 replication with three more children also worked. A later 2021 study compared teaching with and without the frames. Two learners only got the skill with the frames. For all four, the frames were faster.

even though learners that got it eventually, they got it a lot quicker with this.

From the talk — Dana Pettus

Her takeaway is practical. Therapy hours are limited. Methods that build generalized answers help learners reach harder skills sooner.

Generalization for staff, not just clients#

The same rules apply to adults. Ally, who teaches instructional design, says trainers should plan for transfer the same way they would for clients. In her field, "transfer" is just another word for generalization.

When you see transfer in instructional design, all that means is generalization. So we always want to be thinking about transfer and how this is going to generalize outside of the instruction that we're actually providing.

From the talk — Ally

Ally gives an example about session notes. Training starts in a team meeting and gets reinforced in supervision. But the skill is used every day in the electronic record after each session. Some staff have a history of writing notes wrong with no feedback. A good training plan bridges the gap between where staff learn and where they work.

Sean Yocum applies this to BCBA supervision. Anyone can show a trainee a behavior plan or the right terms, he says. That is not the point.

the goal isn't and should not be. Did they see it? The goal is, can they freaking do it?

From the talk — Sean Yocum

For Yocum, a supervisee should be able to write a goal and train a team. They should also defend decisions and revise when things go wrong. He calls the generalization of those skills "the metric."

More hours do not equal generalization#

Michael Carrero, speaking on treatment hours, says many studies miss a key outcome. They rarely track long-term generalization.

Long-term outcomes of generalization, right? We've seen this firsthand across a variety of different places.

From the talk — Michael Carrero

Carrero lists the barriers. How well can you set up and support the new environment? Did you prepare the skill for the new setting? Too often, he says, the team just says "we did our job" and moves on. Filling 30 hours on a calendar is not a way to change behavior. What you teach, how you teach it, and whether caregivers can use it all matter more.

Sean Yocum is even harsher about clinic-only work full of token boards and edibles.

if your ABA therapy isn't resulting in meaningful generalization of skills outside of therapy, you are not practicing.

From the talk — Sean Yocum

Carrero also flags a limit of telehealth. Most telehealth happens at a clinic or at home. He says some research suggests those skills may not transfer to community settings. Following a person around town on camera is not practical.

Where experts disagree#

The speakers agree that generalization matters. They do not fully agree on how to judge it or how to get it.

The first split is about standards. Right after Sean Yocum said that therapy without meaningful generalization is "not practicing" ABA, his own show had pushback. A co-host played devil's advocate. The point was that you do not have to achieve generalization by a certain time. What you must do is program for it and see a path forward. So one view judges the outcome. The other judges whether the plan is in place.

The second split is between fields. Matt Harrington names a real tension between BCBAs and speech-language pathologists, or SLPs.

One of the complaints that SLPs have about BCBAs is that we're too strict and are not doing a good job on generalization with language. And vice versa, BCBAs say SLPs are too naturalistic and aren't making enough progress on language fast enough.

From the talk — Matt Harrington

The third split is about what counts. Kaelynn Partlow says many RBTs are trained to see generalization as varied cues and materials. She argues it should mean returning the skill to its real context. Carrero makes a similar point about the "three rooms" habit. Both push back on a common way the field checks the box.

Finally, there is a tension about NET. Yocum says natural teaching mostly builds in transfer. Carrero, in a separate talk, lists many barriers to generalization even with good setups. Both can be true. But clinicians should not assume any method does the work for them.

What the research says#

Recent studies show that how you teach shapes how well skills spread.

A 2025 study taught three children with autism to act out play with real objects, such as a real cup or brush. Those actions then spread to pretend play with toy versions of the same objects (Agana et al., 2025). The authors note that scripted, drilled play teaching can backfire by producing stiff, repetitive play instead of real pretend play.

A 2024 study taught three children with autism safety and social communication skills using a virtual reality headset (Gayle et al., 2024). In every case, the skills held up over time and spread to the real, natural setting.

A 2025 study used matrix training, a method that teaches a grid of word combinations so new, untaught combinations emerge on their own, with four children on the autism spectrum in China (Lee et al., 2025). The children learned to label combinations of objects and locations, then showed untaught labels and listener skills without direct teaching. Two children kept these skills over time. For the other two, labeling accuracy dropped during later checks.

A basic-research study trained pigeons on a choice task to see how a reward for one response spreads, or generalizes, to similar but different stimuli (Davison & Cowie, 2025). It is not a clinical study, but it points to the same underlying principle: generalization follows predictable patterns, even in a simple lab setting.

FAQ#

What is generalization in ABA? Generalization is when a learned skill shows up beyond where it was taught. That includes new people, places, materials, and natural rewards. Kaelynn Partlow describes it as returning a skill to the real context where it is needed.

Why do skills fail to generalize? Speakers name several reasons. Teaching may use only one setting, one person, or contrived rewards. Staff may never fade out. Caregivers may not be trained, or the goal may not fit family values. Sean Yocum says the biggest cause is "train and hope," which means teaching first and planning for transfer later.

How do you program for generalization? Start on day one. Sean Yocum says every goal should explain how the skill survives without the therapist. Plan how it will transfer across people, settings, materials, and natural rewards. Involve parents, teachers, and other providers early. Then check the skill in real settings, like home and school.

Turn this topic into a CEU

You just studied this. Now get credit for it.

Watch Rad N Bad: Beyond the Blue Square — Demystifying Discrete Trial Training, Stimulus Control, and Generalization for RBTs with Sean Yocum & Michael Carrero and earn 0.5 free BCBA CEUs. Detailed certificate, delivered the moment you finish.

Watch and earn the CEU →Free account · No card · detailed cert

Keep going with a free learning path

Verbal Behavior: Beyond WH-Questions

Dana Pettus's three-part series on building generative verbal behavior. Why WH-question programs stall, how to establish speaker-listener repertoires, and how to scaffold advanced verbal skills like recall and simple inference. Three hours, three CEUs.

Start the path →3 courses · 3 CEUs · Watch at your own pace