Discrete Trial Training: A Practitioner's Guide to DTT, Modern Variants, and Blended NET and DTT Programming

What DTT actually is — the five-component cycle

DTT is operationally defined by a five-part trial sequence: a discriminative stimulus (SD) delivered cleanly without distractors, a prompt at the appropriate level of the prompting hierarchy, a defined response window, a programmed consequence (reinforcer for correct responses, error correction or extinction for incorrect responses), and a brief inter-trial interval (ITI) before the next trial begins (LaMarca et al., 2024). LaMarca and colleagues' ADDIE-based programming guide is the cleanest contemporary statement of this architecture — analyze prerequisites, design each component, develop materials, implement with fidelity probes, and evaluate against integrity checklists tied to each step (LaMarca et al., 2024). The cycle's components are individually editable: revise prompt level when the learner errs, tighten ITI when momentum drops, swap reinforcers when satiation appears, and rotate stimulus topography to prevent restricted stimulus control (LaMarca et al., 2024). Frank-Crawford and colleagues' scoping review of 134 DTT studies (2,036 learners, ~77% with autism, mostly ages 0–6, mostly clinic or school) confirms this cycle is the dominant form across the published corpus Frank‐Crawford et al. (2024).

Lineage from Lovaas to modern variants

DTT in current practice descends from Lovaas's intensive early-intervention model and the experimental analysis of behavior tradition that bracketed the operant unit into discrete, repeatable trials (LaMarca et al., 2024). The contemporary version is broader than the original 40-hour table-bound EIBI image. Modern DTT now appears in three operationally distinguishable forms: traditional table-top DTT; embedded DTT, where the same cycle runs inside a play activity or routine; and progressive DTT, which layers instructive feedback, multiple-exemplar trials, and strategic stimulus arrays into the cycle Haq & Aranki (2019) Ferguson et al. (2022). Haq and Aranki's alternating-treatments comparison of traditional and embedded DTT in a single preschooler with autism showed comparable accuracy across formats — the choice is one of behavior management rather than acquisition efficiency Haq & Aranki (2019). Ferguson and colleagues' progressive DTT comparison against equivalence-based instruction across eight preschoolers with ASD found progressive DTT reached mastery in 16–39 trials and required fewer sessions than EBI for seven of eight participants Ferguson et al. (2022).

Massed vs. distributed trials — what the evidence actually says

The "massed-trial" critique is partly a definitional artifact (LaMarca et al., 2024). ITI duration is a programmable variable, and the field has converged on the view that brief, deliberate ITIs are part of the procedure rather than a flaw of it Dhadwal et al. (2021). Dhadwal and colleagues' replication across three children with autism (ages 4–10) used 30–60 second ITIs deliberately to reduce prompt dependency, allow stimulus rearrangement, and present novel stimuli on every trial — producing reliable acquisition of false-belief responding Dhadwal et al. (2021). Torelli and colleagues' kindergarten life-skills tiered model used the same range, beginning with 30–60 second ITIs in 1:1 DTT and thinning progressively as the child showed acquisition (Torelli et al., 2026). The blocked-versus-mixed question is more clinically active. McKeown and colleagues' alternating-treatments comparison across four children with autism (ages 3–6) showed blocked-trial DTT produced faster acquisition of foundational listener-response targets than mixed-trial DTT, with the difference emerging within the first sessions and the format maintaining discriminated responding without block-size fading McKeown et al. (2025). Ingvarsson and colleagues' earlier blocked-trials demonstration with four boys with autism (ages 5–10) reached the same conclusion for intraverbal discriminations Ingvarsson et al. (2016), with later question pairs requiring fewer trials as a learning-set effect emerged Ingvarsson et al. (2016). The procedural rule: start with larger blocks (6–10) and fade to mixed presentation only after mastery McKeown et al. (2025).

Errorless learning and prompt fading within DTT

Errorless learning — using prompts at sufficient intensity to prevent learner errors during acquisition, with systematic fading as responding stabilizes — is the prompting framework most DTT programs operate inside Foran‐Conn et al. (2021). Foran-Conn and colleagues' parallel-treatments comparison across three children with autism (ages 3–6) compared most-to-least, no-no, and responsive prompt-delay procedures. Responsive prompt delay produced acquisition at least as rapid as the established methods and avoided the trial-and-error exposure inherent in no-no prompting — aligning more directly with errorless guidelines than the comparator procedures Foran‐Conn et al. (2021). Bergmann and colleagues' matrix-training arrangements across three boys with autism (ages 5–7) operationalized errorless instruction inside a 0-second model prompt schedule with VR2 reinforcement thinning Bergmann et al. (2022), producing mastery of component and combined tacts in 10–20-trial sessions Bergmann et al. (2022). Halbur and colleagues' tact comparison reinforced the prompt-rotation principle: alternating auditory and visual trials from the outset was sufficient for emergent intraverbal-tacts, and compounding stimuli offered no incremental benefit (Halbur et al., 2025). Choose the prompt level that prevents errors during acquisition, fade systematically, and rotate trial features to keep stimulus control broad Foran‐Conn et al. (2021).

When add-ons don't add value

Practitioners routinely layer procedural elements they assume will accelerate learning, and the corpus is uncomfortably honest that some add-ons don't pay off. Dell'Aringa and colleagues' adapted alternating-treatments design across three boys with autism (ages 4–6) compared standard prompt-delay DTT for picture tacts to the same procedure plus a supplemental unprompted "transfer trial" after each correct response Dell’Aringa et al. (2021). Acquisition rates were virtually identical (mean sessions to mastery 6.8 vs. 7.0), and maintenance was comparable Dell’Aringa et al. (2021). The headline is not that transfer trials are bad — it is that practitioners should collect data on add-ons rather than assume they help Dell’Aringa et al. (2021). The same skeptical lens applies to other refinements: compound stimuli are no faster than single-feature presentations for tact acquisition, and responsive prompt delay matches traditional most-to-least without the added training overhead (Halbur et al., 2025) Foran‐Conn et al. (2021).

Data collection in DTT

DTT's defining data signature is per-trial response data — plus/minus on each trial, with prompt level, response latency, and stimulus rotation tracked across the session Frank‐Crawford et al. (2024). Frank-Crawford and colleagues' scoping review documents an 80% correct mastery criterion across three consecutive sessions as the de-facto standard, although underlying mastery definitions remain inconsistent across studies Frank‐Crawford et al. (2024). Two practical refinements have entered the corpus recently. Ferguson and colleagues' alternating-treatment design with three children (ages 5–7) showed estimation data collection — a brief post-session rating scale rather than continuous recording — yielded comparable accuracy and learner acquisition rates while reducing data-collection burden Ferguson et al. (2020). Halbur and colleagues' enhanced-data-sheet study used sheets that embedded procedural prompts (counterbalancing cues, target-rotation reminders, dedicated data fields) and significantly improved treatment-fidelity implementation over standard sheets — demonstrating the data system itself is a teaching variable (Halbur et al., 2024). Brand and colleagues' sequential analysis adds the high-resolution version: one-step Markov transition matrices on video-recorded sessions flagged within-trial procedural errors (extraneous prompts, premature reinforcement, prompt-removal omissions) that conventional percent-correct integrity checks missed entirely Brand et al. (2017).

The mastery-criterion question is moving past the 80%-across-three rule Ramos (2025). Ramos's MIEBL tool translates baseline accuracy and target mastery into a probability-based minimum-correct count per session, replacing arbitrary 80%-or-three-in-a-row thresholds with a data-driven decision rule, though the tool has not yet been empirically validated against fixed-percent criteria Ramos (2025). Jones and colleagues' translational study is the procedural counterweight: even after mastery is reached under perfect integrity, introducing consequence errors (omitted reinforcers, mistimed delivery) produced immediate declines in conditional-discrimination accuracy (Jones et al., 2026). Fidelity monitoring needs to continue past mastery, not stop at criterion.

Generalization concerns and modern blended approaches

The classical critique — that table-bound massed-trial instruction produces rote responding that fails to generalize — is part true and part procedural artifact Frank‐Crawford et al. (2024). The corpus supports the concern when DTT is run without an explicit generalization plan, and provides several evidence-based remedies Levesque-Wolfe et al. (2021) Dhadwal et al. (2021). Levesque-Wolfe and colleagues' multiple-baseline across three preschool boys with autism (ages 4–5) embedded discrete-trial picture-discrimination sets inside in-vivo behavioral skills training for abduction-prevention responses, with mastery defined as 100% in one session or ≥91% across two consecutive sessions, followed by community probes within 1 week — producing generalization to new settings and 3-month maintenance Levesque-Wolfe et al. (2021). Dhadwal and colleagues' multiple-exemplar DTT for false-belief tasks went further, using novel stimuli on every trial during acquisition to prevent the learner from latching onto a single exemplar Dhadwal et al. (2021). St. Clair and colleagues' "Making Deception Fun" study with four autistic children (ages 7–16) embedded DTT components inside brief playful "trick-playing" sessions, producing acquisition that generalized across novel tricks and partners without coercion or escape extinction (St. Clair et al., 2024). The remedy is to build multiple-exemplar instruction, in-vivo probes, and naturalistic embedding into the plan from the start.

Discrimination, problem behavior, and reinforcer rotation

DTT is not only a skill-acquisition format. Lambert and colleagues' multiple-probe study with a 10-year-old boy with autism, complex communication needs, and severe problem behavior used multiple-operant discrimination training inside short, clearly cued discrete trials to teach precise S-Δ responding — sharpening contingency detection while keeping challenging behavior near zero Lambert et al. (2021). Cividini‐Motta and colleagues' systematic review of differential reinforcement in skill acquisition reaches a parallel conclusion: DTT remains the primary instructional format for teaching discriminations (Cividini-Motta et al., 2024). McCammon and colleagues' systematic review of mand-training environmental variables documents DTT as the prevailing format for mand instruction, with structured trials, clear discriminative stimuli, immediate reinforcement, and rotating instructors emerging as the dominant pattern (McCammon et al., 2024).

Morris and colleagues' survey documented a practice the corpus had previously left implicit: clinicians frequently swap reinforcers mid-session, driven by moment-to-moment cues like reduced approach or slower responding, but decision rules for when to switch are inconsistent across staff (Morris et al., 2024). Pre-identify a backup pool of potent reinforcers, program a switch decision rule (e.g., "after three consecutive low-effort responses"), and document switches so within-trial integrity audits can interpret the consequence sequence (Morris et al., 2024) Brand et al. (2017).

Staff training to deliver DTT

The DTT staff-training corpus is unusually deep, and brief, structured packages reliably bring entry-level staff to ≥90% procedural integrity Clayton & Headley (2019) Zheng et al. (2025). Clayton and Headley's multiple-probe study with three paraprofessionals showed a 10-minute behavioral skills training (BST) package lifted DTT accuracy from ≤70% to ≥95%, with gains maintained at 30 days Clayton & Headley (2019). Zheng and colleagues' replication with four staff used BST plus a 2-minute video model and reached >90% integrity after a single session, maintained 7 days later Zheng et al. (2025). Olaff and colleagues' multiple-baseline across three teaching assistants showed BST was specifically required for staff to apply DTT skills to novel instructional programs — instruction-only baseline produced variable, low performance Olaff et al. (2025). Higbee and colleagues' interactive computer training reached mastery across eight Brazilian undergraduates and special educators in 2–3 hours of self-paced instruction without in-person coaching Higbee et al. (2016). Lionello‐DeNolf and colleagues' computer-based Train-to-Code system raised post-training procedural fidelity in role-played delivery Lionello‐DeNolf et al. (2025). Bartle and colleagues' multi-element design showed video modeling with exemplars produced procedural integrity only after structured DTT training (Bartle et al., 2025). Chen and Lerman's remote train-the-trainer model with six Taiwanese educators combined synchronous and asynchronous instruction and let initial teachers train colleagues Chen & Lerman (2024). Brief BST plus a video model gets new hires to mastery in a single session; computer-based options scale where in-person coaching is impractical.

Telehealth, assent, and adult populations

Lindgren and colleagues' adapted alternating-treatments comparison across three preschool-aged children with autism is the corpus's strongest evidence that DTT can be delivered effectively via telehealth — telehealth produced equal or faster skill acquisition and required fewer trials to mastery than traditional in-person DTT (Lindgren et al., 2024). Track trials-to-criterion to detect efficiency gains across modalities, and use identical prompting and error-correction protocols across formats (Lindgren et al., 2024). Weber and colleagues' multiple-probe with twelve school instructional assistants demonstrates assent and assent-withdrawal opportunities can be embedded into DTT without disrupting acquisition — typically a 2-second post-prompt latency, materials held outside the visual field until visual assent, and brief differential reinforcement for calm orientation (Weber et al., 2025). The cycle's natural pause points support assent monitoring without procedural overhead (Weber et al., 2025). For adults, Radogna and colleagues' multiple-baseline across three Italian adults (ages 19–32) with autism or intellectual disability embedded DTT format steps inside BST cycles for vocational social skills — token plus descriptive feedback for correct responses, re-modeling for incorrect — producing rapid acquisition and generalization across novel supervisors and settings, with brief boosters at shift start maintaining performance (Radogna et al., 2024). Huba and Belfiore's flashcard-format DTT in a postsecondary culinary program showed stimulus-feature discrimination statements during error correction accelerated acquisition relative to feedback alone (Huba & Belfiore, 2024). The procedure's component structure scales across age — the SD remains the SD, the prompt is fitted to the learner, and the consequence is contextually appropriate.

The DTT literature is unusually wide and deep, with one large scoping review at the corpus level, multiple systematic reviews of specific applications, and a dense band of single-subject experimental designs across acquisition, staff training, and modality variants Frank‐Crawford et al. (2024) (Cividini-Motta et al., 2024).

Scoping and systematic reviews. Frank-Crawford and colleagues' scoping review of 134 DTT studies (2,036 learners, ~77% with autism, mostly ages 0–6, mostly clinic or school) is the most comprehensive synthesis of DTT parameters and mastery criteria currently available Frank‐Crawford et al. (2024). Cividini‐Motta and colleagues' systematic review of differential reinforcement in skill acquisition documents DTT's continued primacy as the format for teaching discriminations (Cividini-Motta et al., 2024). McCammon and colleagues' systematic review of mand-training environmental variables documents DTT as the dominant operational pattern for mand instruction (McCammon et al., 2024).

Single-subject experimental designs (acquisition and procedural comparisons). Most DTT evidence is SCED: McKeown and colleagues (n=4) on blocked vs. mixed trials McKeown et al. (2025); Ingvarsson and colleagues (n=4) on blocked-trials intraverbal acquisition Ingvarsson et al. (2016); Foran-Conn and colleagues (n=3) comparing most-to-least, no-no, and responsive prompt-delay Foran‐Conn et al. (2021); Dell'Aringa and colleagues (n=3) on the transfer-trial null finding Dell’Aringa et al. (2021); Bergmann and colleagues (n=3) on matrix-training arrangements Bergmann et al. (2022); Halbur and colleagues (n=4) on simple vs. compound stimuli for tact instruction (Halbur et al., 2025); Dhadwal and colleagues (n=3) on multiple-exemplar DTT for false-belief tasks Dhadwal et al. (2021); Haq and Aranki (n=1) on traditional vs. embedded DTT Haq & Aranki (2019); Ferguson and colleagues (n=8) on progressive DTT vs. EBI Ferguson et al. (2022); Levesque-Wolfe and colleagues (n=3) on embedded DTT within in-vivo BST for abduction prevention Levesque-Wolfe et al. (2021); St. Clair and colleagues (n=4) on DTT format components inside playful "trick" sessions (St. Clair et al., 2024); Lambert and colleagues (n=1) on multiple-operant discrimination training inside DTT Lambert et al. (2021); Huba and Belfiore (n=1) on flashcard-format DTT in a vocational program (Huba & Belfiore, 2024); Radogna and colleagues (n=3) on DTT with embedded BST for vocational social skills (Radogna et al., 2024); Yassa and colleagues (n=3) on DTT format scripting for FCT-with-multiple-schedules Yassa et al. (2024); Lindgren and colleagues (n=3) on telehealth vs. in-person DTT (Lindgren et al., 2024); Ferguson and colleagues (n=3) on estimation vs. trial-by-trial data Ferguson et al. (2020); Brand and colleagues' sequential analysis of within-trial integrity Brand et al. (2017); and Jones and colleagues' translational study of post-mastery consequence-fidelity errors (Jones et al., 2026).

Staff-training and implementation studies. Clayton and Headley (n=3) on 10-minute BST Clayton & Headley (2019); Zheng and colleagues (n=4) on BST plus 2-minute video model Zheng et al. (2025); Olaff and colleagues (n=3) on BST for cross-program generalization Olaff et al. (2025); Higbee and colleagues (n=8) on interactive computer training Higbee et al. (2016); Lionello‐DeNolf and colleagues on computer-based Train-to-Code Lionello‐DeNolf et al. (2025); Halbur and colleagues on enhanced data sheets (Halbur et al., 2024); Bartle and colleagues on video modeling with exemplars (Bartle et al., 2025); Chen and Lerman (n=6) on remote train-the-trainer Chen & Lerman (2024); and Weber and colleagues (n=12) on assent-withdrawal embedded in DTT (Weber et al., 2025).

Methodology and conceptual papers. LaMarca and colleagues on the ADDIE cycle (LaMarca et al., 2024); Ramos's MIEBL mastery tool Ramos (2025); Morris and colleagues' survey of reinforcer-switching (Morris et al., 2024); and Torelli and colleagues' kindergarten life-skills evaluation (Torelli et al., 2026).

Bottom line. The convergent picture supports the operational claims this page makes — that the five-component cycle is the procedural unit, blocked-then-mixed sequencing accelerates acquisition for early learners, errorless prompting via responsive prompt delay or matrix schedules works as well as classical most-to-least, brief BST plus a video model trains staff to mastery in a single session, telehealth DTT is operationally viable, and adding procedural elements without data is a common practitioner error McKeown et al. (2025) Clayton & Headley (2019) (Lindgren et al., 2024) Dell’Aringa et al. (2021). The evidence is thinner for adult populations, telehealth at scale, and the strong "DTT generalizes worse than NET" claim — which sits on individual SCED demonstrations rather than head-to-head comparative-effectiveness trials Frank‐Crawford et al. (2024).

The DTT decisions a senior practitioner makes are not "DTT or not" so much as "which DTT variant, with which prompt and trial structure, for which target, in which setting, with which data system" (LaMarca et al., 2024). A defensible logic, drawn directly from the corpus:

1. Novel arbitrary discrimination, learner with prerequisite attending. Default to traditional table-top DTT with a 2- to 5-second prompt delay, blocked trials of 6–10 same-target presentations, error correction via responsive prompt delay, and 80%-across-three-sessions mastery McKeown et al. (2025) Foran‐Conn et al. (2021) Frank‐Crawford et al. (2024).
2. Foundational listener-response or tact target with early learner. Start with blocked-trial DTT and fade to mixed only after mastery McKeown et al. (2025). Skip block-size fading; large blocks maintain discriminated responding without the procedural step McKeown et al. (2025).
3. Intraverbal or complex stimulus-control target. Use a five-step blocked-trials procedure with prompt-delay and error correction; later question pairs typically require fewer trials as a learning-set effect emerges Ingvarsson et al. (2016).
4. Tact program where transfer-trial add-ons are being considered. Run the standard prompt-delay cycle first and collect data — supplemental unprompted transfer trials may not accelerate acquisition Dell’Aringa et al. (2021).
5. Recombinative language target (color + object tacts). Use a 0-second model prompt with VR2 reinforcement thinning inside 10–20-trial matrix-training sessions; if recombinative generalization does not emerge, systematically overlap diagonals within the same DTT format Bergmann et al. (2022).
6. Equivalence-based learning target. Progressive DTT with embedded instructive feedback, multiple-exemplar presentations, and strategic stimulus-array pairing can equal or exceed specialized equivalence-based instruction in sessions-to-criterion Ferguson et al. (2022).
7. Generalization to community settings is a primary goal. Embed DTT picture-discrimination sets inside in-vivo BST sessions; require ≥91% across two sessions before community probes; schedule generalization probes within 1 week of mastery Levesque-Wolfe et al. (2021).
8. Behavior is sensitive to instructional context and table-top DTT triggers escape. Switch to embedded DTT in a preferred activity — accuracy is comparable across formats Haq & Aranki (2019).
9. Discrimination training during problem-behavior treatment. Use multiple-operant discrimination training inside short, clearly cued discrete trials, and rely on differential reinforcement of correct S-Δ responses to keep challenging behavior near zero Lambert et al. (2021).
10. Adult or vocational learner. Embed DTT format steps inside BST cycles — token plus descriptive feedback for correct responses, re-modeling for incorrect — and use a brief booster at the start of each shift rather than full BST every session (Radogna et al., 2024).
11. Telehealth delivery. Track trials-to-criterion rather than only percent-correct, prepare prompting and error-correction protocols identical across modalities, and use telehealth DTT when in-person staffing is constrained (Lindgren et al., 2024).
12. Data system. Use enhanced data sheets that embed counterbalancing and target-rotation prompts; switch to estimation data collection for routine work; reserve trial-by-trial recording for novel acquisition and fidelity probes (Halbur et al., 2024) Ferguson et al. (2020).
13. Mastery criterion. The 80%-across-three default works for most targets Frank‐Crawford et al. (2024); for cases where individual baseline accuracy and risk tolerance argue for tighter or looser thresholds, use Ramos's MIEBL tool to derive an evidence-based minimum-correct count Ramos (2025).
14. Within-trial fidelity monitoring. Code within-trial sequences (Markov-style transitions) on video-recorded sessions when a program is failing or staff drift is suspected — session-level percent-correct integrity checks miss procedural lapses Brand et al. (2017).
15. Post-mastery monitoring. Continue fidelity probes after mastery — consequence errors introduced post-mastery produce immediate accuracy declines (Jones et al., 2026).
16. Staff training. Start with a 10-minute BST package plus a 2-minute video model; expect ≥90% integrity after a single session, with maintenance for 7–30 days Clayton & Headley (2019) Zheng et al. (2025). For cross-program generalization, full BST is required Olaff et al. (2025).
17. Cross-geography or under-resourced training context. Use remote synchronous-plus-asynchronous train-the-trainer or self-paced computer-based instruction; both produce mastery without in-person coaching Chen & Lerman (2024) Higbee et al. (2016).
18. Assent monitoring. Embed a 2-second post-prompt latency for assent withdrawal; hold materials outside the visual field until visual assent; reinforce calm orientation (Weber et al., 2025).
19. Reinforcer rotation. Pre-identify a backup pool; program a switch decision rule (e.g., "after three consecutive low-effort responses"); document switches in session notes (Morris et al., 2024).

Clinic-based DTT

Clinic-based DTT is where most published procedures were developed and validated. Frank-Crawford and colleagues' scoping review documents that the bulk of published DTT studies sit in clinic or school settings with early learners, running prompt and consequence variations within the standard five-component cycle Frank‐Crawford et al. (2024). Bergmann and colleagues' matrix-training arrangements at a university research clinic illustrate the cycle's productivity Bergmann et al. (2022): 10–20-trial sessions with 0-second model prompts and VR2 reinforcement thinning produced mastery of component and combined tacts in three boys with autism Bergmann et al. (2022). McKeown and colleagues' clinic-based blocked-vs-mixed comparison gives the operational rule for foundational targets: start blocked, fade to mixed only after mastery McKeown et al. (2025). The clinic substrate also supports the within-trial fidelity work that's harder to run in homes or schools — Brand and colleagues' Markov-transition analyses of video-recorded sessions are most feasible when sessions are routinely recorded as part of the clinic's quality system Brand et al. (2017).

School DTT

School DTT works inside classroom routines and tiered service-delivery models rather than at isolated 1:1 tables. Torelli and colleagues' kindergarten life-skills evaluation operationalized a three-tier hierarchical model in which 1:1 brief, massed DTT with prompted trials and 30–60-second ITIs was reserved for children who failed to master skills in larger group conditions (Torelli et al., 2026). Olaff and colleagues' study with three teaching assistants and one 7-year-old boy with ASD in a special-education classroom is the cleanest evidence that classroom staff can be trained to deliver DTT across novel programs at ≥90% integrity using full BST Olaff et al. (2025). Weber and colleagues' multiple-probe with twelve K-12 instructional assistants demonstrates the same school-staff population can be trained to embed assent-withdrawal pause points into DTT without disrupting acquisition (Weber et al., 2025). Chen and Lerman's remote training of six Taiwanese educators with subsequent train-the-trainer dissemination shows the school DTT model scales internationally Chen & Lerman (2024).

Home and parent-implemented DTT

Home DTT typically runs as parent- or caregiver-implemented under BCBA coaching rather than direct technician delivery. Lindgren and colleagues' telehealth study leans heavily on caregiver-implemented DTT — caregivers present materials and reinforcers using items already in the home while the BCBA coaches via the camera (Lindgren et al., 2024). The procedural rule for home DTT is that the cycle is unchanged but the delivery agent and the materials shift: the caregiver becomes the trial-runner, household items become the discriminative stimuli, and the BCBA's role is to verify prompting and consequence fidelity through video review or live coaching (Lindgren et al., 2024). The staff-training literature transfers directly: Clayton and Headley's 10-minute BST package and Zheng's BST plus video model approach work with parent-implementers as cleanly as with paraprofessionals Clayton & Headley (2019) Zheng et al. (2025).

Telehealth DTT

Telehealth DTT is feasible and, in the most direct comparison currently available, produced equal or faster acquisition than in-person DTT for three preschool-aged children with autism (Lindgren et al., 2024). The procedural takeaways are concrete: track trials-to-criterion to detect efficiency differences across modalities, prepare brief prompting and error-correction protocols that are identical across formats, and use telehealth as a supplement when face-to-face staffing is limited rather than as a fallback (Lindgren et al., 2024). The infrastructure assumption is non-trivial — the published comparison occurred at a university clinic with strong technical setup — so practitioners replicating in lower-resource contexts should plan for connectivity, hardware, and caregiver-mediation explicitly (Lindgren et al., 2024).

Adult, vocational, and residential DTT

Adult DTT data are sparser than child data but operationally clean. Radogna and colleagues' multiple-baseline with three Italian adults (ages 19–32) in a vocational program demonstrated that DTT format steps embedded inside BST cycles produce rapid acquisition and generalization of workplace social skills across novel supervisors and settings, with brief boosters at shift start maintaining performance (Radogna et al., 2024). Huba and Belfiore's flashcard-format DTT in a postsecondary culinary program with one 19-year-old student with moderate intellectual disability is a single-case demonstration that the cycle scales to vocational learners, with descriptive-feature discrimination statements during error correction accelerating acquisition relative to feedback alone (Huba & Belfiore, 2024). The procedural implication for adult-services BCBAs is that the SD-prompt-response-consequence-ITI structure transfers cleanly — the SD becomes a kitchen tool or social scenario, the prompt is fitted to the learner's repertoire, the consequence is a token plus descriptive feedback, and the ITI is brief but defined.

The most common practitioner question about DTT is when to use it instead of (or alongside) Natural Environment Teaching, and the corpus supports a structural rather than ideological answer (LaMarca et al., 2024).

Use DTT when: the target requires high trial density to establish (novel arbitrary discriminations, precise prompt fading on stimuli that don't exist in the natural environment), the learner doesn't yet have the prerequisite imitation, attending, or approach repertoire to engage productively with naturalistic teaching, or the target is a precision discrimination where errors during acquisition would be costly to correct later McKeown et al. (2025) Ingvarsson et al. (2016). DTT also gives the cleanest substrate for per-item mastery tracking, within-trial fidelity coding, and probability-based mastery criteria — the data infrastructure that makes acquisition decisions defensible at the supervisor level Frank‐Crawford et al. (2024) Brand et al. (2017).

Use NET when: the target is a mand, tact, intraverbal, social, perspective-taking, or life-skills repertoire where the natural environment already supplies the establishing operation and the natural reinforcer; generalization to untrained people, settings, or stimuli is part of the goal rather than a separate phase; or the learner's engagement and motivation rise materially when instruction happens inside their preferred activity. (See the Natural Environment Teaching practitioner guide (link pending) for the procedural details.)

Use blended programming when: DTT establishes the discrimination and NET generalizes it — the dominant pattern in modern early-intensive programming (LaMarca et al., 2024). Lindgren and colleagues' telehealth study illustrates the structure operationally — separate targets taught in DTT and NET inside the same overall package (Lindgren et al., 2024). Haq and Aranki's traditional-versus-embedded comparison sits at the boundary: when behavior is sensitive to instructional context, embedded DTT inside a play activity preserves the cycle's acquisition properties while reducing escape-driven problem behavior Haq & Aranki (2019). Ferguson and colleagues' progressive DTT comparison reinforces the larger point: small procedural upgrades within DTT (instructive feedback, multiple-exemplar sampling, strategic stimulus arrays) can reach the efficiency of specialized teaching paradigms without requiring a wholesale curriculum change Ferguson et al. (2022).

The decision is target-driven, not philosophical: a clinician running only DTT loses the generalization-promoting properties NET supplies, and a clinician running only NET loses the trial density, discrimination control, and per-item data DTT supplies (LaMarca et al., 2024) McKeown et al. (2025). The blended approach — DTT to establish, NET to generalize — is the operational default in 2026 EIBI programming.

• Rote responding without a generalization plan. The classical DTT critique is real when the procedure is run without explicit generalization programming. Build multiple-exemplar instruction, in-vivo probes, and naturalistic embedding into the DTT plan from the start Dhadwal et al. (2021) Levesque-Wolfe et al. (2021) (St. Clair et al., 2024).
• Prompt dependency from poorly faded prompts. Responsive prompt delay produces acquisition rates at least as fast as most-to-least without the trial-and-error exposure of no-no prompting Foran‐Conn et al. (2021). Track prompt level explicitly across sessions; if prompt level is not fading after 3–5 sessions of mastery, redesign the prompt schedule.
• Mixing targets from session one. Begin with blocks of 6–10 same-target presentations and fade to mixed after mastery; rate differences emerge in the first sessions and modified blocked-trial formats maintain discriminated responding without block-size fading McKeown et al. (2025) Ingvarsson et al. (2016).
• Adding procedural elements without data. Supplemental unprompted transfer trials did not accelerate tact acquisition (mean sessions to mastery 6.8 vs. 7.0), and compound stimuli were no faster than single-feature presentations for tact instruction Dell’Aringa et al. (2021) (Halbur et al., 2025). Run the basic cycle first; add complexity only when data justify it Dell’Aringa et al. (2021).
• Within-trial fidelity drift. Session-level percent-correct integrity checks miss procedural lapses inside the trial. Code within-trial sequences on video-recorded sessions when fidelity is suspect Brand et al. (2017). Use enhanced data sheets that embed counterbalancing and target-rotation prompts to raise fidelity at the data-collection layer (Halbur et al., 2024).
• Stopping fidelity probes at mastery. Mastery is not a license to stop monitoring. Consequence-based fidelity errors introduced post-mastery produced immediate declines in conditional-discrimination accuracy (Jones et al., 2026).
• Reinforcer satiation without programmed rotation. Pre-identify a backup pool of potent reinforcers, program a switch decision rule, and document switches so audits can interpret the consequence sequence (Morris et al., 2024).
• One-shot staff training. Brief BST plus a video model gets new hires to ≥90% in a single session, but cross-program generalization requires full BST — instruction-only baselines do not produce reliable implementation Zheng et al. (2025) Olaff et al. (2025).
• Neglecting assent and assent withdrawal. A 2-second post-prompt latency, materials held outside the visual field until visual assent, and brief differential reinforcement for calm orientation integrate cleanly into the cycle (Weber et al., 2025).
• Choosing DTT globally for ideological reasons. Format selection should be target-driven. Running only DTT loses the generalization-promoting properties of NET; running only NET loses the trial density and discrimination control of DTT McKeown et al. (2025) (LaMarca et al., 2024).

• Severe problem behavior that interrupts DTT sessions repeatedly. Refer for a function-based assessment and behavior plan before continuing skill-acquisition work; mixing acquisition and behavior reduction without a clear plan inflates risk.
• Suspected medical or sensory substrate driving low engagement. Behavior that may involve pain, sleep, sensory, or seizure issues should be referred for medical evaluation before assuming the DTT cycle needs procedural revision.
• Learner who consistently fails to engage with the SD across multiple staff and settings. Refer for preference assessment, prerequisite-skill review (joint attention, imitation, basic instructional control), and possibly an embedded DTT or NET trial before returning to traditional table-top DTT Haq & Aranki (2019).
• Staff cannot reach mastery on DTT fidelity after two BST cycles. When BST plus video modeling does not produce ≥90% procedural integrity, refer the case to a more experienced clinician or restructure training Clayton & Headley (2019) Zheng et al. (2025).
• Risk-sensitive or socially complex target where DTT alone is the wrong tool. Pedestrian safety, abduction prevention, and similar consequential skills should be established with embedded DTT inside in-vivo BST sequences with explicit community probes — and when the risk profile exceeds outpatient capacity, refer to a specialized clinic Levesque-Wolfe et al. (2021).

The honest read of the corpus is that DTT's operational claims sit on a remarkably broad SCED base plus one large scoping review, while the comparative-effectiveness layer against alternative procedures (NET, EBI, naturalistic developmental behavioral interventions) is thinner than the field needs Frank‐Crawford et al. (2024) Ferguson et al. (2022). The largest single research opportunity is a head-to-head DTT versus NET comparative-effectiveness trial of the kind that would settle the generalization question prospectively rather than by aggregating individual SCED demonstrations within each format Frank‐Crawford et al. (2024).

Several smaller-scale gaps are more tractable. Ramos's MIEBL tool offers an individualized, evidence-based mastery-criterion alternative to the 80%-across-three rule, but the tool has not yet been empirically validated against fixed-percent criteria — a prospective comparison would convert the recommendation into evidence-grade procedure Ramos (2025). Brand and colleagues' Markov-transition method for within-trial integrity is conceptually strong but logistically heavy; a streamlined version producing actionable feedback in near-real-time would close the gap between session-level and within-trial fidelity monitoring Brand et al. (2017). Lindgren and colleagues' telehealth-versus-in-person comparison is the only direct modality study currently available; a multi-site replication varying clinic infrastructure and caregiver mediation would tell the field whether the equal-or-faster-than-in-person finding generalizes (Lindgren et al., 2024). Adult and vocational DTT remains under-studied relative to the volume of adult ABA practice; replication of Radogna and colleagues' BST with embedded DTT model across more vocational programs would strengthen the adult-services evidence base (Radogna et al., 2024). Finally, Jones and colleagues' translational finding on post-mastery consequence-fidelity errors deserves field replication in applied DTT contexts (Jones et al., 2026).

1. Define DTT by the five-component cycle, not by the table. SD → prompt → response → consequence → ITI, run as a discrete unit with per-trial data (LaMarca et al., 2024). The cycle is portable: traditional table-top, embedded in a play activity, and progressive variants all run the same five components Haq & Aranki (2019).
2. Start blocked, fade to mixed. For early learners and foundational discrimination targets, blocked trials of 6–10 same-target presentations produce faster acquisition than mixed trialing from session one. Skip block-size fading McKeown et al. (2025) Ingvarsson et al. (2016).
3. Default to responsive prompt delay for error correction. Acquisition is at least as fast as classical most-to-least, without the trial-and-error exposure of no-no prompting Foran‐Conn et al. (2021).
4. Mastery = 80% across three consecutive sessions, unless data argue otherwise. The 80%-across-three rule is the de-facto standard Frank‐Crawford et al. (2024); for cases warranting tighter or looser thresholds, use Ramos's MIEBL tool Ramos (2025).
5. Use enhanced data sheets and estimation data collection. Embed counterbalancing and target-rotation prompts directly on the data sheet to raise treatment fidelity; use estimation data for routine sessions to recover instructional time (Halbur et al., 2024) Ferguson et al. (2020).
6. Run within-trial fidelity coding when programs underperform. Code video-recorded sessions for within-trial transitions when staff drift is suspected Brand et al. (2017). Continue fidelity monitoring past mastery (Jones et al., 2026).
7. Plan generalization from session one. Multiple-exemplar instruction (entirely novel stimuli on every trial), in-vivo probes within 1 week of mastery, and embedded DTT inside preferred activities prevent the rote-responding failure mode Dhadwal et al. (2021) Levesque-Wolfe et al. (2021).
8. Don't add procedural elements without data. Transfer trials did not accelerate tact acquisition, and compound stimuli were no faster than single-feature presentations Dell’Aringa et al. (2021) (Halbur et al., 2025).
9. Pre-program reinforcer rotation. Pre-identify a backup pool, set a switch decision rule, document switches in session notes (Morris et al., 2024).
10. Train staff with brief BST plus a 2-minute video model. A 10-minute BST package gets new paraprofessionals to ≥95% DTT accuracy with 30-day maintenance Clayton & Headley (2019) Zheng et al. (2025). For cross-program generalization, full BST is required Olaff et al. (2025).
11. Use computer-based training when in-person coaching is constrained. Interactive computer training and Train-to-Code coding tools both produce DTT mastery without in-person coaching Higbee et al. (2016) Lionello‐DeNolf et al. (2025) Chen & Lerman (2024).
12. Embed assent and assent-withdrawal opportunities into the cycle. A 2-second post-prompt latency, visual-field discipline, and reinforcement for calm orientation integrate without disrupting acquisition (Weber et al., 2025).
13. Use embedded DTT when behavior is sensitive to instructional context. Traditional and embedded DTT produce comparable accuracy Haq & Aranki (2019).
14. Telehealth DTT is operationally viable. Telehealth produced equal or faster acquisition than in-person; track trials-to-criterion across modalities (Lindgren et al., 2024).
15. Choose DTT and NET by target, not ideology. DTT for precision discriminations, dense trial-count tact and intraverbal acquisition; NET for mand, social, perspective-taking, and life-skills work; blended programming — DTT to establish, NET to generalize — for most contemporary EIBI McKeown et al. (2025) (LaMarca et al., 2024).

Primary research synthesized in this guide. DOIs link to the original source.

• LaMarca, V. J., & LaMarca, J. M. (2024). Using the ADDIE model of instructional design to create programming for comprehensive ABA treatment. Behavior Analysis in Practice, 17, 371–388. https://doi.org/10.1007/s40617-024-00908-2 https://doi.org/10.1007/s40617-024-00908-2
• Lindgren, N. A., Higbee, T. S., Osos, J. A., Nichols, B., & Campbell, V. E. (2024). Comparing the effectiveness of discrete trial training delivered via telehealth and in‑person on skill acquisition. Behavior Analysis in Practice, 17, 783–795. https://doi.org/10.1007/s40617-023-00855-4 https://doi.org/10.1007/s40617-023-00855-4
• Frank‐Crawford, M. A., Borrero, J. C., Fisher, A., Talhelm, P., & Fernandez, N. (2024). Discrete‐trial teaching: A scoping review. Behavioral Interventions, 39(3). https://doi.org/10.1002/bin.2012 https://doi.org/10.1002/bin.2012
• Cividini-Motta, C., Livingston, C., & Efaw, H. (2024). Systematic review of differential reinforcement in skill acquisition. Behavior Analysis in Practice, 17, 401–416. https://doi.org/10.1007/s40617-023-00903-z https://doi.org/10.1007/s40617-023-00903-z
• Dell’Aringa, A. R., Juanico, J. F., & Harrison, K. L. (2021). Using Transfer Trials to Teach Tacting to Children With Autism Spectrum Disorder. Behavior Analysis in Practice, 14(1), 120-130. https://doi.org/10.1007/s40617-020-00507-x https://doi.org/10.1007/s40617-020-00507-x
• Olaff, H. S., Koch, I. B., & Brand, D. (2025). Using Behavioral Skill Training to Extend Discrete-Trial Teaching Skills Across Instructional Programs. Behavior Analysis in Practice. https://doi.org/10.1007/s40617-025-01094-5 https://doi.org/10.1007/s40617-025-01094-5
• Bergmann, S., Van Den Elzen, G., Kodak, T., Niland, H., & Dawson, D. (2022). Comparing Matrix-Training Procedures with Children with Autism Spectrum Disorder. The Analysis of Verbal Behavior, 38(1), 24-53. https://doi.org/10.1007/s40616-022-00167-8 https://doi.org/10.1007/s40616-022-00167-8
• Ferguson, J. L., Cihon, J. H., Majeski, M. J., Milne, C. M., Leaf, J. B., McEachin, J., & Leaf, R. (2022). Toward Efficiency and Effectiveness: Comparing Equivalence-based Instruction to Progressive Discrete Trial Teaching. Behavior Analysis in Practice, 15(4), 1296-1313. https://doi.org/10.1007/s40617-022-00687-8 https://doi.org/10.1007/s40617-022-00687-8
• McCammon, M. N., Wolfe, K., & Check, A. R. (2024). A review of the environmental variables included in mand training interventions. The Analysis of Verbal Behavior, 40, 345–378. https://doi.org/10.1007/s40616-024-00211-9 https://doi.org/10.1007/s40616-024-00211-9
• Haq, S. S. & Aranki, J. (2019). Comparison of Traditional and Embedded DTT on Problem Behavior and Responding to Instructional Targets. Behavior Analysis in Practice, 12(2), 396-400. https://doi.org/10.1007/s40617-018-00324-3 https://doi.org/10.1007/s40617-018-00324-3
• Bartle, G. E., Ruby, S. A., & DiGennaro Reed, F. D. (2025). The effects of video modeling containing different exemplar types on procedural integrity. Journal of Organizational Behavior Management. https://doi.org/10.1080/01608061.2025.2476425 https://doi.org/10.1080/01608061.2025.2476425
• Torelli, J. N., Snyder, S. K., Griffin, M. L., Hellemn, M. P., Cagliani, R. R., & Morgan, G. A. (2026). Life skills evaluation in a kindergarten special education classroom. Journal of Applied Behavior Analysis, 59(1), e70041. https://doi.org/10.1002/jaba.70041 https://doi.org/10.1002/jaba.70041
• Halbur, M., Reidy, J., Kodak, T., Cowan, L., & Harman, M. (2024). Comparison of enhanced and standard data sheets on treatment fidelity and data collection for tact training. Behavior Analysis in Practice, 17, 533–543. https://doi.org/10.1007/s40617-023-00869-y https://doi.org/10.1007/s40617-023-00869-y
• Brand, D., Mudford, O. C., Arnold‐Saritepe, A., & Elliffe, D. (2017). Assessing the Within‐Trial Treatment Integrity of Discrete‐Trial Teaching Programs Using Sequential Analysis. Behavioral Interventions, 32(1), 54-69. https://doi.org/10.1002/bin.1455 https://doi.org/10.1002/bin.1455
• Levesque-Wolfe, M. A., Rodriguez, N. M., & Niemeier-Beck, J. J. (2021). Consideration of Both Discriminated and Generalized Responding When Teaching Children with Autism Abduction Prevention Skills. Behavior Analysis in Practice, 14(2), 396-409. https://doi.org/10.1007/s40617-020-00541-9 https://doi.org/10.1007/s40617-020-00541-9
• Lionello‐DeNolf, K. M., Eckerman, D., Hise, R., Pinzino, E., & Ray, R. (2025). Improvement of procedural fidelity in discrete‐trial programs using computer‐based instruction to teach skilled observation. Journal of Applied Behavior Analysis, 58(4), 821-840. https://doi.org/10.1002/jaba.70027 https://doi.org/10.1002/jaba.70027
• Higbee, T. S., Aporta, A. P., Resende, A., Nogueira, M., Goyos, C., & Pollard, J. S. (2016). Interactive computer training to teach discrete-trial instruction to undergraduates and special educators in Brazil: A replication and extension. Journal of Applied Behavior Analysis, 49(4), 780-793. https://doi.org/10.1002/jaba.329 https://doi.org/10.1002/jaba.329
• McKeown, C. A., Walker, S. F., & Peters, K. P. (2025). Comparison of Blocked Versus Mixed Trialing When Teaching Foundational Skills to Early Learners. Behavioral Interventions, 40(1). https://doi.org/10.1002/bin.2071 https://doi.org/10.1002/bin.2071
• Morris, S. L., Conine, D. E., Slanzi, C. M., Kronfli, F. R., & Etchison, H. M. (2024). A survey of why and how clinicians change reinforcers during teaching sessions. Behavior Analysis in Practice, 17, 815–830. https://doi.org/10.1007/s40617-023-00847-4 https://doi.org/10.1007/s40617-023-00847-4
• Chen, N. & Lerman, D. C. (2024). Remote training of educators in Taiwan to disseminate discrete trial training for students with developmental disabilities. Behavioral Interventions, 39(1). https://doi.org/10.1002/bin.1980 https://doi.org/10.1002/bin.1980
• Ramos, M. L. F. (2025). MIEBL: Measurement of Individualized, Evidence-Based Learning Criteria Designed for Discrete Trial Training. Behavior Analysis in Practice, 18(4), 1283-1289. https://doi.org/10.1007/s40617-025-01058-9 https://doi.org/10.1007/s40617-025-01058-9
• Clayton, M. & Headley, A. (2019). The use of behavioral skills training to improve staff performance of discrete trial training. Behavioral Interventions, 34(1), 136-143. https://doi.org/10.1002/bin.1656 https://doi.org/10.1002/bin.1656
• Ferguson, J. L., Milne, C. M., Cihon, J. H., Dotson, A., Leaf, J. B., McEachin, J., & Leaf, R. (2020). An evaluation of estimation data collection to trial‐by trial data collection during discrete trial teaching. Behavioral Interventions, 35(1), 178-191. https://doi.org/10.1002/bin.1705 https://doi.org/10.1002/bin.1705
• Foran‐Conn, D., Hoerger, M., Kelly, E., R. Cross, R., Jones, S., Walley, H., & Firth, L. (2021). A comparison of most to least prompting, no‐no prompting and responsive prompt delay procedures. Behavioral Interventions, 36(4), 1024-1041. https://doi.org/10.1002/bin.1808 https://doi.org/10.1002/bin.1808
• Dhadwal, A. K., Najdowski, A. C., & Tarbox, J. (2021). A Systematic Replication of Teaching Children With Autism and Other Developmental Disabilities Correct Responding to False-Belief Tasks. Behavior Analysis in Practice, 14(2), 378-386. https://doi.org/10.1007/s40617-020-00531-x https://doi.org/10.1007/s40617-020-00531-x
• Zheng, T., Albright, L., & Mahoney, A. (2025). The Use of Behavioral Skills Training to Teach Staff Discrete Trial Teaching. Behavioral Interventions, 40(1). https://doi.org/10.1002/bin.2070 https://doi.org/10.1002/bin.2070
• St. Clair, M., Massoudi, K., Tarbox, J., Najdowski, A., Simchoni, L., Jackson, M., & Persicke, A. (2024). Making deception fun: Teaching autistic individuals how to play friendly tricks. Behavior Analysis in Practice, 17, 1134–1146. https://doi.org/10.1007/s40617-024-00935-z https://doi.org/10.1007/s40617-024-00935-z
• Huba, K., & Belfiore, P. J. (2024). Using flashcards and descriptive feedback to teach industrial kitchen equipment to a student with intellectual disabilities in a postsecondary education program. Behavior Analysis in Practice, 17, 544–552. https://doi.org/10.1007/s40617-023-00853-6 https://doi.org/10.1007/s40617-023-00853-6
• Ingvarsson, E. T., Kramer, R. L., Carp, C. L., Pétursdóttir, A. I., & Macias, H. (2016). Evaluation of a Blocked-Trials Procedure to Establish Complex Stimulus Control over Intraverbal Responses in Children with Autism. The Analysis of Verbal Behavior, 32(2), 205-224. https://doi.org/10.1007/s40616-016-0071-5 https://doi.org/10.1007/s40616-016-0071-5
• Weber, J., Weiss, M. J., & Ferris, K. (2025). Training instructors to support assent and assent withdrawal during instruction for students with disabilities. Behavior Analysis in Practice. https://doi.org/10.1007/s40617-025-01122-4 https://doi.org/10.1007/s40617-025-01122-4
• Halbur, M., Kodak, T., & Reidy, J. (2025). A comparison of training procedures on the emergence of intraverbal-tacts. The Analysis of Verbal Behavior, 40, 379–402. https://doi.org/10.1007/s40616-024-00214-6 https://doi.org/10.1007/s40616-024-00214-6
• Jones, L., Brand, D., Bensemann, J., Heinicke, M. R., Penrod, B., & Burlison, S. (2026). A translational approach to investigating the effects of consequence-based procedural fidelity errors postmastery. Journal of Applied Behavior Analysis, 59(1), e70042. https://doi.org/10.1002/jaba.70042 https://doi.org/10.1002/jaba.70042
• Iversen, I. H. (2025). Levels of Replication. Perspectives on Behavior Science, 48(1), 41-58. https://doi.org/10.1007/s40614-024-00431-3 https://doi.org/10.1007/s40614-024-00431-3
• Lambert, J. M., Lamers, M. E., Copeland, B. A., & Banerjee, I. (2021). Multiple operant discrimination training to increase the precision of stimulus deltas following treatment of challenging behavior. Behavioral Interventions, 36(3), 606-619. https://doi.org/10.1002/bin.1788 https://doi.org/10.1002/bin.1788
• Radogna, C., D’Angelo, G., & Lerman, D. C. (2024). Assessing and teaching job‑related social skills to adults with neurodevelopmental disorders in Italy. Behavior Analysis in Practice, 17, 803–814. https://doi.org/10.1007/s40617-023-00873-2 https://doi.org/10.1007/s40617-023-00873-2