Rad N Bad: DTT Is Not a Permanent Residence — Integrating Discrete-Trial and Naturalistic Teaching for Generalization

ABA Is the Science. DTT Is One Method Inside It.

Plenty of people still use ABA and DTT as if they mean the same thing. Sean Yocum says that is simply wrong. ABA is the science. Discrete-trial teaching (teaching in small, planned trials) is one method within it. Mike Carrero compares the mix-up to a child who learns that a dog has four legs, ears, and a tail. The child walks outside, sees a cat, and says dog. Same parts. Very different animal.

Here is the short version of the format. You break a skill into small teaching units. You deliver a clear cue, prompt the response, and follow it with reinforcement (something that makes a behavior happen more). Then you run that trial again and again. Error correction (fixing a wrong answer right away) keeps the loop tight. That control is the whole point of the design.

The hosts say trouble starts when a field confuses one procedure with the entire science. They point to a 1980 article by Hayes. It warned that ABA can drift into narrow technical ritual when it loses contact with its conceptual system. Decades later, the table, the chair, the card, and the token board still read as ABA to many families. Sean's fix is to stay close to radical behaviorism (the philosophy behind this science). From there, he keeps asking what a procedure is actually selecting.

the real question is not does it work it's what does it work for under what conditions at what cost us toward what goals.

From the talk — Sean Yocum, MA, BCBA, LBA & Michael Carrero, BCBA, LBA

The Honest Case for Discrete-Trial Teaching

DTT earns its place. It is one of the clearest, most analyzable teaching tools the field has. You get tight stimulus control (when a cue reliably sets off a response). You get many learning chances per minute. Prompting is clean. Error correction is fast. The learner contacts reinforcement often.

For some learners, that structure is what makes learning possible at all. Sean names the profile: limited attending, weak imitation, shaky discrimination. When looser teaching has not produced progress, DTT can be the difference between hoping a skill sticks and actually building one. He also points at the evidence base. Reviews report that DTT has met evidence-based practice criteria across 58 single-case designs. Target areas include communication, play, adaptive skills, school readiness, social skills, and job skills.

Mike adds two more points. High-volume practice speeds up learning. If you want to run better, you run more. It will not look pretty at first, and that is fine. Structured teaching also lives far outside the clinic. A general education classroom runs on cues, criteria, and delayed reinforcement. Adults use the same loop when they learn a new task at work. Mike attaches one asterisk to all of it: this only holds if the format is implemented correctly.

I think that's actually great instructional engineering from a pro standpoint.

From the talk — Sean Yocum, MA, BCBA, LBA & Michael Carrero, BCBA, LBA

Prompt Dependency and the Day You Became the Cue

The real risk is not that DTT fails to teach. The risk is that it teaches the wrong thing very well. The format is adult led, prompt heavy, and often run with contrived reinforcers (rewards the real world will not give). Run it long enough without a plan and you build prompt dependency (waiting for a cue before acting). The learner finishes the programs. The learner gets the thing at the end. Independence never shows up.

Mike walks through how the setting itself takes over. A table starts as a plain object with four legs. Then it becomes the one place where learning happens. Now the table signals the whole routine, including rote answers and problem behavior. The arrangement is doing the work, not the skill.

His sharpest example is the crosswalk. A learner looks both ways only after an adult says a crosswalk is coming. That adult has become the SD (the cue that sets off a response). That is faulty stimulus control, not true independence. The hosts blame timing more than format. Teams sit in acquisition (early learning of a skill) and maintenance far too long. Then they fade prompts late, or never.

Sean offers an audit you can run on any DTT program. What contingencies is this arranging? What repertoire is it selecting? What stimulus control is being built? Which motivating operations (things that change what a reward is worth) are present or absent? Does the response survive when all the theater goes away? If the learner responds only with you, in your trials, the procedure may be technically behavioral and still weak.

You are serving as the SD to have that behavior change, right?

From the talk — Sean Yocum, MA, BCBA, LBA & Michael Carrero, BCBA, LBA

The Generalization Test: Can the Skill Leave the Room?

Stokes and Baer called for programming generalization (using a skill in new places) in 1977. Baer, Wolf, and Risley put generality at the center of the field long before that. The hosts treat both as a gut check. You do not train a skill and then hope it travels. You design the travel from day one.

Mike uses the stimulus equivalence triangle (linking word, picture, and object). Symmetry, transitivity, and reflexivity all have to hold. A learner can name a picture of a cat and still fail the rest of the triangle. Sean's version is the dog. A child touches dog on a card for a token. That same child never orients to a real dog, never comments on one, never plays with a dog toy. The score went up. Nothing socially meaningful changed.

Relevance matters too. Teaching farm animal sounds to a child in a dense city is nice to know. It carries low value for that life. The toy crosswalk makes the same point. The painted street is missing the two-ton car, the horn, and the real risk. Mike's answer is not to drop DTT here. Pre-teach the parts somewhere safe. Then move the skill into the real setting on purpose, and keep widening from there.

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

From the talk — Sean Yocum, MA, BCBA, LBA & Michael Carrero, BCBA, LBA

What the Comparison Research Actually Shows

Sean points to a 2001 review by Del Prado. Across controlled language studies, more normalized or naturalistic behavioral interventions outperformed discrete-trial approaches. Studies that also looked at parent affect favored the normalized conditions. That critique lands hardest on DTT run as massed trials with contrived reinforcers and a thin transfer plan.

Newer work points the same direction. Sean cites a 2024 study he credits to Janczyk on discrete-trial and natural environment teaching. As he reports it, natural environment teaching, and natural environment teaching paired with DTT, produced better outcomes than DTT alone. None of that makes DTT useless. It does mean DTT by itself is a weak plan.

Then there is social validity (whether people value the results). Sean points to a 2024 paper in the Journal of Autism and Developmental Disorders. He credits it to Baden, and says it sought input from autistic adults on behavior interventions. Early discrete-trial training came up directly. The hosts do not wave that away. Their read is that the tool is not the harm. Poor reasons and poor implementation are.

I do think we have to acknowledge lived experience of individuals

From the talk — Sean Yocum, MA, BCBA, LBA & Michael Carrero, BCBA, LBA

Natural Environment Teaching on the Floor Is Still DTT

Sean's daily complaint is fake NET. Natural environment teaching (teaching inside real daily activities) is not DTT with walking. Plenty of teams roam a clinic or a home, place demands, and call it naturalistic. It is still adult led. It is still the same trial loop with a different backdrop.

Mike gives the clean contrast with a toy barn. You and the learner are playing. You pick up the horse and start clucking like a chicken. The learner corrects you and tells you what the horse says. You agree and keep playing. The teaching rides inside the activity. No sterile question. No token board held up in the middle of it.

Now the bad version. The learner is deep in the barn play. You hold up a card and say one, two, three, what's this. It is a car. The card has nothing to do with what is happening. That is DTT on the floor wearing a NET label. Mike compares this to a trainer who advertises as positive only. That is branding, not a plan. The job is matching supports to the person in front of you, not selling a format.

we're embedding this naturally into the activity

From the talk — Sean Yocum, MA, BCBA, LBA & Michael Carrero, BCBA, LBA

A Decision Map for Choosing the Teaching Format

Sean lays out when DTT is justified. Use it when a learner needs dense, repeated chances to acquire a skill. Use it when the target is a clear discrimination or a foundation skill. Use it when naturalistic teaching has not produced efficient progress. Use it when prompting and error correction need real control. Use it to pre-teach a safety skill before the learner faces live risk.

He is just as clear about when to stop. DTT is hard to defend when the target is spontaneous language, social engagement, or flexible, context-sensitive play. It is also wrong when the learner already has the prerequisite skills. If the team keeps grinding massed trials (many trials back to back) out of habit, that is not science either. Sean points to naturalistic ABA models such as NDBI (naturalistic developmental behavioral intervention) and the Early Start Denver Model. Both were built for shared control, child initiation, and natural reinforcement.

One condition sits under all of it: a written plan to thin prompts, transfer stimulus control, and move fast. The target is always natural contingencies. Mike warns against scheduling that by quarter. The blending happens inside the session. Use priming (a quick preview before the real thing). Use graduated guidance (hands-on help that fades fast). Then step into the natural setting you know is coming. Learner preference counts as a real variable, not a courtesy. If a learner hates DTT, Sean says avoid it unless the skill is life-saving.

You pick DTT because the learner actually needs DTT.

From the talk — Sean Yocum, MA, BCBA, LBA & Michael Carrero, BCBA, LBA

Measure More Than Acquisition

Acquisition data is seductive. A rising graph feels like proof of work. The hosts argue that acquisition was never the mission. Meaningful behavior change in the natural environment is the mission, and a clean graph can hide its absence.

So widen the dashboard. Track spontaneous use of the skill with no cue from you. Track latency to initiate (how long before they start). Track variability, meaning how many forms the response can take. Track caregiver implementation: can the family actually run this. Track quality of life. Before the first trial, write down how the target will transfer across people, settings, and materials. If you cannot name the natural reinforcers it has to reach, the program is not finished being designed.

Sean's closing line is that structure is not the problem. Structure becomes a problem when it stands in for thinking. Robotic responding, weak generalization, low joy, and high adult control are the warning signs. Watch for them especially when everyone is still celebrating the graph. Mike's image is a ladder. It is powerful and worth climbing. It is also meant to be climbed off of.

I think if it builds prerequisite repertoires and then gets out of the way, then it's good behavior analysis.

From the talk — Sean Yocum, MA, BCBA, LBA & Michael Carrero, BCBA, LBA