Skip to content

Discrete trial training (DTT) in ABA

Discrete trial training teaches a skill through short, repeated learning opportunities, each with a clear beginning and end. That structure is what makes DTT data unusually clean - every trial is one scorable event - and it is also why DTT data are easy to record in ways that quietly change what the graph means. This guide covers the parts of a trial, how trials are scored, error correction, and how the data turn into a mastery decision.

  • Reviewed November 2026
  • 7 min read

What discrete trial training is

Discrete trial training (DTT), also called discrete trial teaching, is a structured teaching method within applied behavior analysis. A skill is broken into small units, and each unit is taught through brief trials: the clinician gives an instruction, the learner responds, and the clinician delivers a consequence before a short pause separates that trial from the next. Trials are typically run in blocks at a table or in another low-distraction setting, which gives the learner many practice opportunities in a short time.

DTT is a teaching format, not a curriculum. What gets taught - receptive labels, imitation, matching, early language - comes from the learner's assessment, such as the VB-MAPP, and treatment plan; DTT is one of the ways those targets are taught and measured. It is most closely associated with early intensive behavioral intervention for young autistic children, and it is usually one part of a program rather than all of it.

The five parts of a discrete trial

Every trial has the same anatomy. Some descriptions count four parts and treat the prompt as optional, because a learner who already responds correctly needs none. The intertrial interval is the part most often skipped in practice, and it is the part that makes a trial discrete: without it, trials blur into one another and neither the learner nor the person scoring can tell where one ends.

The parts of a discrete trial, with an example from teaching a receptive color label. The prompt is used only when the learner needs it, and is faded as the skill is acquired.
PartWhat happensExample: teaching "touch red"
Discriminative stimulus (SD)The instruction or cue that signals a response is expected. It is brief, clear, and delivered the same way each time.Three color cards are on the table and the clinician says, "Touch red."
PromptAdded help, given with or just after the SD, that makes a correct response more likely. It is planned in advance and faded over time.The clinician points to the red card.
ResponseWhat the learner does, scored against the target's operational definition.The learner touches the red card within a few seconds.
ConsequenceReinforcement after a correct response; a planned error-correction procedure or a neutral response after an incorrect one.Specific praise and a token toward a preferred activity.
Intertrial intervalA short pause, usually a few seconds, that separates one trial from the next.The clinician records the trial, rearranges the cards, and pauses before the next instruction.

How DTT data are recorded

Each trial is scored as it happens: correct and independent, correct with a prompt (and which prompt), incorrect, or no response. The session-level number is usually one of two percentages, and they answer different questions. Percent correct counts every correct response; percent independent counts only correct responses that needed no prompt. A session in which every response was correct but half were prompted is 100% correct and 50% independent.

Pick one and write the mastery criterion in the same terms. Percent independent is the stricter measure and usually the more informative one during acquisition, because the point of prompting is to stop needing it. Recording which prompt level was used, rather than a single prompted or unprompted mark, is what lets a BCBA see fading progress inside a target that has not yet moved on percent independent - see the ABA prompt hierarchy for how the levels are defined.

Programs also choose how much to record. Trial-by-trial data score every trial in the session and show learning within the session. First-trial or cold-probe data score only the first opportunity, before any teaching that day, and take far less effort. Because a probe samples a single opportunity, it can show a skill as mastered before it is reliable across a session, so many programs probe for monitoring and switch to trial-by-trial data when a mastery or phase-change decision is close. Whichever is used, record during the session rather than reconstructing it afterward.

Error correction and errorless teaching

Errorless teaching aims to prevent errors rather than correct them. For a new target, the clinician prompts immediately - with no delay between the instruction and the prompt - so the learner practices only the correct response, then fades the prompt as responding becomes reliable. This is the logic of most-to-least prompting, and it is the usual starting point for skills where errors tend to repeat.

When an error does occur, a planned correction procedure follows. A common sequence re-presents the instruction with a prompt, then runs a transfer trial with the prompt reduced or removed so the learner has a chance to respond more independently, sometimes followed by a distractor trial before returning to the target. The original trial stays scored as incorrect, and correction trials are not counted as independent correct responses.

Consistency matters more than the particular procedure. If one clinician scores a corrected trial as correct and another scores it as incorrect, the graph is mixing two definitions, and the same disagreement will show up as low interobserver agreement. Writing the correction procedure and its scoring rule into the program is what keeps a target's data comparable from one session and one clinician to the next.

Writing a mastery criterion for DTT

A usable mastery criterion names four things: the metric (percent independent or percent correct), the threshold, the number of sessions it must hold for, and whether those sessions must be consecutive. "80% independent across three consecutive sessions" is a common form; some programs add a condition such as "across two instructors" to build generalization into the criterion itself.

Criteria set too low tend to produce skills that do not maintain, which is why a threshold is usually required across more than one session rather than reached once. Mastery in acquisition is also not the end of the target: maintenance probes check that the skill holds after teaching stops, and generalization probes check that it appears with new people, materials, and settings.

DTT and natural environment teaching

DTT and natural environment teaching (NET) are complements, not rivals. DTT gives many controlled, uniform opportunities and the cleanest acquisition data; NET embeds teaching in play and routines, follows the learner's motivation, and gives far better evidence that a skill is used outside the teaching setting. A common pattern is to teach a new skill in DTT and probe or extend it in NET.

They are also measured differently. DTT data are scored per trial against a set number of opportunities, while NET captures opportunities as they arise, so they are fewer and less uniform. The eight ABA measurement systems guide covers how each format's data are summarized and when each is the right choice.

How Cadence records DTT

In Cadence, discrete trial training is one of the 8 measurement systems built into the data model. Clinicians record trials by voice or by tapping an eyes-up grid that keeps working with no connection, prompted trials are scored against the clinic's own prompt hierarchy, and a mis-scored trial is corrected with one tap. A DTT target's graph and its mastery criterion share units - percent independent or percent correct - so the threshold and the plotted value always mean the same thing.

On the Clinical and Practice plans, mastery is checked against each target's rule nightly and whenever a client is reviewed. When the data meet the criterion, Cadence surfaces a phase-change proposal with the sessions it rests on, and the BCBA decides whether to accept it.

FAQ

Common questions

What are the steps of discrete trial training?
A discrete trial has five parts: the discriminative stimulus (the instruction or cue), a prompt if the learner needs one, the learner's response, the consequence (reinforcement for a correct response or a planned error correction for an incorrect one), and a brief intertrial interval that separates one trial from the next.
Is discrete trial training the same as ABA?
No. Applied behavior analysis is the broader science and clinical practice. Discrete trial training is one teaching method used within ABA programs, alongside others such as natural environment teaching and task analysis, and it is usually one part of a learner's program rather than all of it.
How do you take data during discrete trial training?
Score each trial as it happens: correct and independent, correct with a prompt (noting the prompt level), incorrect, or no response. The session value is usually percent independent - unprompted correct trials divided by total trials, times 100 - or percent correct. Some programs score only the first trial of a session as a probe, which takes less effort but samples a single opportunity.
What is a typical mastery criterion for DTT?
A common form is 80% or 90% independent across two or three consecutive sessions, sometimes across more than one instructor. The criterion should name the metric, the threshold, the number of sessions, and whether they must be consecutive, and it must use the same units as the data being graphed.
What is the difference between DTT and NET?
Discrete trial training delivers many structured, closely spaced trials, usually at a table, and produces uniform acquisition data. Natural environment teaching embeds teaching in play and daily routines and follows the learner's motivation, which produces fewer but more natural opportunities and better evidence of generalization. Many programs use both.
Keep reading

Data collection that holds up

Cadence captures every ABA measurement system natively, measures interobserver agreement in-product, and drafts the billable note from the data you just committed - for a clinician to review and sign.