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

Why DTT Takes Both the Praise and the Blame

Discrete Trial Training (a tight teach and repeat format) splits the field in two. Some people swear by it. Others swear at it. The hosts name that split in the first few minutes. For one camp, DTT is the base of all good teaching. For the other, it is the face of all that is wrong with ABA. Both camps have a point.

The first fix is simple. ABA is the science. DTT is one method inside that science. Mike calls the mix up an overgeneralization (one label stretched too far). A child learns that a dog has four legs, ears, and a tail. Then every cat, bear, and horse is a dog too. Same shape, very different animal. We do the same thing with ABA and DTT.

That blur has a cost. Once a field treats one method as the whole science, the hard questions stop. Sean points to a 1980 paper by Hayes. It warned that ABA can drift into narrow technical ritual when it loses its own concepts. So the hosts hold a middle line all episode. DTT is not a hero. It is also not a villain.

DTT is a tool and like every tool in ABA, the real question is not, does it work? It's what does it work for?

From the talk — Sean Yocum & Michael Carrero

What DTT Is Still the Best Tool For

Start with the honest case for it. DTT gives you clean, countable teaching. You get tight stimulus control (what cue the behavior follows). You get many chances to practice. Prompts stay neat. Errors get fixed fast. Reward comes often. Now think of a learner with few early skills. Little attending. Little imitation. Weak discrimination. For that learner, the tight frame can make learning work at all.

The hosts point to reviews that log DTT across 58 single case designs. Those studies span talking, daily living, play, school readiness, social skills, and job skills. Sean puts the blunt version this way. If a learner is not picking up the skill in looser formats, DTT can be the line between hoping and teaching.

Mike's best example is street safety. A pretend crosswalk on the floor is missing a few key things. There is no two ton car. There is no horn. That is exactly why you teach the first steps there. You do not want a child's first try to come in live traffic.

He also notes that structure is not just a clinic thing. A general class runs on set rules, late rewards, and clear goals. Adults drill new work skills the same way. If you want to get better at running, you run more. It may not look pretty at first.

Pre-teaching using some DTT, you now have that skill.

From the talk — Sean Yocum & Michael Carrero

When the Table Itself Becomes the Cue

Here is the trap Mike names first. A table starts out as a thing you put stuff on. Run every lesson there and it turns into a cue. The child learns that work happens at that table. What follows can be rote answers, or problem behavior, or nothing at all once the chair is gone.

The same thing happens with people. Prompt dependency (waiting for your help first) creeps in when prompts never fade. Mike goes back to crossing the street. The child looks both ways only after an adult says to look both ways. That is not real independence. You have become the discriminative stimulus (the cue to respond now).

Sean's label for this is faulty stimulus control (the wrong cue runs the skill). The data can still look great. The score goes up. But the skill is tied to your voice, your table, and your token board (a chart that tracks earned points). Both hosts trace this to one habit. Teams camp out in acquisition (learning the skill at first). They camp out in maintenance (holding the skill over time). The fading never starts.

I also think when you reference prompt dependency, that makes it significantly more challenging in the long term.

From the talk — Sean Yocum & Michael Carrero

Train and Hope Is Not a Generalization Plan

A score can rise while nothing real changes. Sean's example is the word dog. A child can touch a dog card at a table for a token. If that child never looks at a real dog, talks about dogs, or plays with dog toys, the program bought a number. It did not buy a behavior.

Mike adds the context problem. You can drill farm animals with a child who lives in a big city. Nice to know. Not useful today. He also asks teams to finish the stimulus equivalence triangle (links that work both ways). Naming a cat card is one corner. Stop there and the skill is not done.

Generalization (skills that hold up everywhere else) was in the field's own charter from the start. Sean points to Baer, Wolf, and Risley, who put generality at the center of the work. He names Stokes and Baer in 1977 as well. The lesson has been in print for decades.

The comparison studies sting a bit. Sean cites a 2001 review of controlled language studies. More natural teaching beat discrete trial formats. Parent affect looked better too. He names a 2024 study that points the same way. Natural teaching paired with DTT beat DTT alone. That does not make DTT useless. It does mean the field got drunk on clean data.

It's not something that you train and hope that it's going to happen later.

From the talk — Sean Yocum & Michael Carrero

DTT as a Culture Instead of a Procedure

Sean's sharpest point is about why DTT spread so wide. It is easy to train. Give the cue. Prompt. Reward or correct. Repeat. It is easy to score, easy to audit, and easy to sell. None of that proves it is the right call for the learner in front of you.

He names the pattern that grew out of that ease. Whole programs built on sitting, on compliance, on eye contact, on blocked escape, and on long adult led chains. The workflow asked for it more than the science ever did. That is how a teaching method turns into a treatment identity.

Both hosts take lived experience seriously here. Autistic adults have called early trial based therapy compliance training. Sean names a 2024 journal article that asked autistic adults about social validity (does this matter to the people served). Early discrete trial work came up as the prime example. He does not use that to throw out the tool. He uses it to indict bad use of the tool.

I think DTT is dangerous when structure becomes like our field's excuse for like intellectual laziness.

From the talk — Sean Yocum & Michael Carrero

That Is Not NET, That Is DTT on the Rug

Supervisors should read this part twice. Natural Environment Teaching (teaching inside real play routines) is not a place. Sean watches teams walk around a clinic or a house, place demand after demand, and call it natural. The chair is gone. The adult control is not.

Mike shows the difference with a toy barn. You are playing with the horse. You cluck like a chicken. The child says no, the horse says neigh. You agree and keep playing. The lesson rides inside the game. Nobody holds up a token board.

The fake version sounds like this. You are both playing with the barn. Then the adult says one, two, three, what is this, and holds up a car. The car has nothing to do with the play. The play was the motivating operation (what makes a reward matter). Cut it off and you throw away the one thing you had going for you.

Mike ties this to branding. A trainer who claims one method only is selling a product, not a plan. We only do DTT and we only do NET are both slogans. Real care pushes and pulls between the two for one person. That is what the word individual means.

It's just DTT walking around.

From the talk — Sean Yocum & Michael Carrero

Run This Audit Before Your Next Trial

Sean offers a short list of questions to ask first. What is this trial arranging? What repertoire is it building? What stimulus control are you building? Which MOs are present, and which are missing? Which rewards are doing the real work?

Then comes the last one. Does the response survive when all the theater goes away? If the learner only responds with you there, leading tight trials, you have a problem. The plan can be technically behavioral and still be thin where it counts. That is the gap between using reward and thinking like a behavior analyst.

Mike adds the long game check. Real life has noise, no tokens, and no trained adult standing by. Sometimes the person next to your learner is learning the same skill too. If your plan quietly assumes a clean room and a skilled helper, the skill will not hold up outside that room.

What are we arranging for with this? What repertoire is it actually selecting?

From the talk — Sean Yocum & Michael Carrero

The Generalization Ladder, Step by Step

Mike's door safety example shows the shape of a good plan. Step one is small. Hear the knock and do not open the door. Then there is step two, three, four, and on. The top step is a learner who peeks, asks who is it, and has a few ways to handle it. The first trial was never meant to be the last one.

He warns against calendar thinking. You do not spend one quarter on DTT, the next on NET, and the third on transfer. The mix shifts inside the session. Use graduated guidance (hands on help you slowly lighten) to thin prompts right before a real event. Prime the skill, then step into the real setting. Then go find a new one.

Sean's design rule is to plan transfer on day one. If you cannot say how the skill will move across people, settings, materials, and natural rewards, the program is not finished. Fade contrived reinforcers (rewards we add on purpose) to natural ones as fast as the learner can take it. Treat assent (the learner's willing yes) as a real variable, not decor. If the learner hates the drill, skip it, unless the target keeps them safe.

He also names when DTT gets hard to defend. Free talk, social play, and flexible language rarely come out of rigid adult led trials. For those he points to natural teaching models like the Early Start Denver Model (a play based early model). Those were built around shared control and child starts. And measure more than acquisition. Watch free use of the skill, how fast the learner starts, variety in responses, what caregivers can teach, and quality of life.

DTT should serve as a ladder. It's powerful and important but meant to be climbed off of onto something else.

From the talk — Sean Yocum & Michael Carrero