TL;DR: Discrete Trial Training (DTT) only works when the trial itself is clean. That means a clear SD, a controlled prompt hierarchy, immediate consequences, and data you can actually use the next day. This post walks through the moving parts of a well-run DTT session and the small failures that quietly degrade learning.
DTT is one of the most procedurally specific tools in ABA, and that’s its strength. When a trial is clean, you can isolate what’s working and what isn’t. When it’s noisy — inconsistent SDs, drifting prompts, sloppy timing — your data tells you about your delivery, not the learner.
If you’re a BCBA writing programs, a supervisor running fidelity checks, or an RBT trying to tighten up your sessions, the difference between “running DTT” and “running DTT well” usually comes down to a small set of habits.
The Anatomy of a Clean Trial
Every discrete trial has the same skeleton. Get the skeleton right and the rest of the session falls into place.
- Antecedent (SD) — the discriminative stimulus. One clear instruction, delivered once. “Touch red.” Not “Can you touch red for me?”
- Prompt (if needed) — delivered at the level specified in the program, not improvised mid-trial.
- Response — the learner’s behavior, scored against the operational definition.
- Consequence — reinforcement for correct independent or correct prompted responses; error correction procedure for incorrect responses.
- Inter-trial interval (ITI) — a brief, consistent pause before the next trial.
A trial that skips a step or runs them in the wrong order isn’t a discrete trial. It’s a teaching attempt.
Five Habits That Quietly Degrade DTT Sessions
These are the small failures supervisors see most often during fidelity checks. None of them feel wrong in the moment, which is why they spread.
1. Repeating the SD
You deliver “Touch red.” The learner hesitates. You say “Touch red?” again, then “Come on, touch red.”
Now the trial is contaminated. You don’t know whether the learner responded to the first SD, the third SD, or your tone. If a prompt is needed, deliver it as specified. Don’t repeat the SD.
2. Drifting Prompt Levels
The program calls for a least-to-most prompt hierarchy with a 3-second delay. Halfway through the session, you’re prompting at 1 second because the learner is getting frustrated. The data still says “least-to-most, 3 seconds.”
Prompt drift is the most common reason a learner appears to acquire a skill in session and then “loses it” the following week. They never had it independently — your prompt was masking the gap.
3. Inconsistent Reinforcement Timing
Reinforcement delivered three seconds after a correct response reinforces whatever happened three seconds after the correct response. If you’re charting data on the clipboard between the response and the reinforcer, you’re teaching the learner to wait politely, not to respond accurately.
Deliver the reinforcer first. Score the trial during the ITI.
4. Mixed Mastered and Acquisition Targets Without a Plan
Interspersing mastered targets with acquisition targets keeps reinforcement density high and reduces frustration. But “interspersing” isn’t “randomly inserting easy ones when things feel hard.” The ratio should be specified in the program and held consistent across sessions, or your data won’t compare across days.
5. Skipping the ITI
A short, consistent ITI (typically 3–5 seconds) gives the learner a moment to disengage from the previous trial and orient to the next SD. Sessions that bleed one trial into the next tend to produce stimulus overselectivity and reduce response independence.
What Good Data Looks Like the Next Day
The point of DTT is the data. If you can’t open yesterday’s session and answer these three questions in under a minute, the data isn’t doing its job:
- Did the learner respond independently or with a prompt? At what level?
- Where in the session did errors cluster — early, middle, late?
- Did mastered targets stay mastered, or are we seeing drift?
If your sheet doesn’t capture prompt level per trial, you can’t answer question one. If it doesn’t include a trial-by-trial sequence, you can’t answer question two. Aggregate percentage correct hides almost everything that matters for programming decisions.
Where Practice Actually Builds These Habits
Most DTT errors aren’t knowledge errors. RBTs and new BCBAs can describe the prompt hierarchy in detail and still drift mid-session. The issue is reps under realistic conditions — a learner stalling, a parent watching, a session running long.
Reading a procedural manual doesn’t build pacing. Watching a video doesn’t build the muscle memory of holding a prompt delay through frustration. The only thing that builds those habits is running trials, getting feedback, and running more trials.
In most clinics, those reps happen with real learners. That’s the only option, but it’s a costly one — every trial run badly is a trial the learner experienced badly.
How Kipr Helps
Kipr lets BCBAs and behavior technicians practice DTT sessions with AI personas that simulate the moments where fidelity tends to break — the stall, the escape attempt, the unexpected response. You run the trial, the simulation responds, and you get immediate feedback on SD delivery, prompt level, reinforcement timing, and ITI consistency.
Kipr is not a replacement for supervision, fieldwork, or real client work. It’s a place to get reps before those reps cost a learner anything. The goal is the same as any deliberate practice tool: when you walk into a real session, the procedural mechanics are automatic, and your attention is free for the learner in front of you.
Join the Kipr waitlist for early access.