What Is Natural Environment Teaching?
Natural Environment Teaching (NET) is an applied behavior analytic teaching procedure in which the learner’s own motivation, interests, and ongoing activity set the occasion for instruction. Instead of pulling a child to a table and running 10 cold trials on “ball,” you watch for the moment they reach toward a ball, then you teach in that moment. The reinforcer is the ball itself. The environment is the classroom rug, the backyard, the bathtub, the snack table. The teaching is embedded in the flow of the child’s day.
NET sits inside the broader family of naturalistic developmental behavioral interventions (NDBIs), a category that includes Pivotal Response Training (PRT), the Early Start Denver Model (ESDM), Incidental Teaching, and Milieu Teaching. The lineage runs through Hart and Risley’s incidental teaching work in the 1960s and 70s, through Robert and Lynn Koegel’s Pivotal Response Training research at UC Santa Barbara, through Laura Schreibman’s contributions on motivation and child choice, and through Mark Sundberg and James Partington’s verbal behavior framework, which gave the field the VB-MAPP and a clean rationale for teaching language in the natural environment.
The key features of NET are consistent across these lineages: the learner initiates, the teaching opportunity is built around an existing motivating operation (MO), the response requirement is contextually appropriate, and the reinforcer is functionally related to the response. If the child asks for a bubble, the reinforcer is bubbles. If the child reaches for the swing, the reinforcer is the swing. The whole procedure is designed to look, to an outside observer, less like teaching and more like a really attentive, slightly choreographed play session.
That choreography matters. NET is not unstructured play. It is a clinically deliberate procedure, with operational targets, prompt fading plans, and data collection — it just happens to look natural.
NET vs DTT: Complementary, Not Opposed
One of the most common confusions in early BCBA training is the framing of NET and Discrete Trial Training (DTT) as opposing camps. They are not. They are two delivery formats for behavior analytic teaching, each with strengths the other lacks, and most clinically competent programs blend them.
DTT typically involves a contrived antecedent (often an instruction at a table), a learner response, and an arbitrary reinforcer — meaning a reinforcer not functionally related to the response (e.g., correct labeling earns a token, not the labeled item). Trials are massed, brief, and high-density. NET inverts most of those variables: the antecedent is the natural environment, the response is captured rather than evoked, and the reinforcer is the item or activity the learner was already pursuing.
The behavior chains differ in an important way. In DTT, the chain runs: instruction → response → arbitrary reinforcer → inter-trial interval. In NET, the chain runs: MO present → learner initiation → contrived opportunity → response → natural reinforcer. The MO does the heavy lifting in NET. The instruction does the heavy lifting in DTT.
| Dimension | DTT | NET |
|---|---|---|
| Antecedent | Adult-delivered instruction (SD) | Naturally occurring MO; learner initiation |
| Setting | Often table-based, controlled | Wherever the MO is — play area, kitchen, yard |
| Reinforcer | Arbitrary (often token, edible, praise) | Natural (functionally related to response) |
| Trial density | High (many trials per minute) | Lower (paced by learner motivation) |
| Best for | Acquisition of discrete responses, low-MO targets, foundational repertoires | Generalization, language, social-communicative behavior, mands |
| Prompting | Often most-to-least, errorless | Often least-to-most, time delay |
| Stimulus control | Tight, narrow | Broad, multiple exemplars |
| Generalization | Must be programmed explicitly | Often built into the procedure |
| Data collection | Trial-by-trial, easy | Probe-based or first-trial, harder |
A blended approach is what most modern BCBAs actually do. You might use DTT to acquire a discrete echoic repertoire when a learner’s vocal output is sparse, then immediately shift those responses into NET to bring them under natural MO control. You might run DTT for letter identification and NET for mands. You might do DTT in the morning when the learner is fresh and NET in the afternoon when they are engaged in preferred play. The question is rarely “DTT or NET?” It is “what does this target need, right now, for this learner?”
The Core Components of NET
NET is not a single procedure but a cluster of teaching tactics held together by a common philosophy: follow the motivation. Five components show up in nearly every well-run NET program.
Capturing the Motivating Operation
The MO is the engine of NET. Without an active MO, the natural reinforcer has no value, and the teaching trial collapses. Capturing the MO means noticing when the learner wants something — a snack, a toy, attention, an activity to continue — and using that brief window to evoke a response. You are not creating the MO. You are observing it and acting quickly.
Child-Led Initiation
The learner starts the interaction. They reach, they vocalize, they orient, they bring you an item. Your job is to be available and responsive, not to direct. This is the “child-led” piece that distinguishes NET from teacher-led formats — and it is the piece newer therapists most often skip.
Naturalistic Reinforcement
The consequence is functionally related to the response. If the target was a mand for “open,” the reinforcer is the container being opened. If the target was joint attention, the reinforcer is shared engagement with the object of interest. Arbitrary reinforcers (tokens, unrelated edibles) erode the natural contingency NET is built on.
Embedded Teaching
Teaching trials are inserted into ongoing activities rather than scheduled into a separate “teaching block.” A bath becomes a context for teaching body parts, action words, and requests. A snack becomes a context for mands, tacts, and turn-taking. Embedding requires you to know your targets cold so you can opportunistically run trials without breaking the flow.
Multiple Exemplars
NET programs in stimulus variation from the start. You teach “ball” with a basketball, a tennis ball, a foam ball, and a beach ball. You teach “open” with jars, doors, books, and packages. This is how NET builds generalization into the procedure rather than tacking it on later.
How NET Is Done: A Step-By-Step
1. Preference Assessment
Start with a current, learner-specific preference assessment. Free-operant observations work well for NET because they reveal what the learner actually approaches in an unconstrained environment. A multiple-stimulus-without-replacement (MSWO) assessment at the start of the session can refine the list. Preferences shift across days, sessions, and even within sessions — re-assess often.
2. Environmental Arrangement
Set up the space so preferred items are visible but not freely accessible. Place the bubbles on a shelf the learner can see but not reach. Put preferred snacks in clear containers with lids. Leave a few new items in the activity area to evoke curiosity. This is sometimes called contriving the MO, and it is a core NET skill — you are not creating motivation that does not exist, but you are arranging the environment so existing motivation has a chance to produce a teachable moment.
3. The Teaching Trial
When the learner initiates, you respond. If the response is in the learner’s current repertoire, you may simply reinforce. If you are working on acquisition, you prompt the target response — often using a time delay (wait 3-5 seconds, then prompt) or a least-to-most hierarchy. Deliver the natural reinforcer contingent on the response. Move on. The trial ends when the reinforcer is delivered, not when a timer goes off.
4. Generalization Probes
Across sessions, vary materials, settings, people, and times of day. Probe whether the response transfers without being directly taught in each new context. A response that only occurs with one therapist in one room is not yet functional.
5. Data Collection
NET data collection is genuinely harder than DTT data collection — there is no obvious trial boundary, and the rate of opportunities depends on the learner’s engagement. Common approaches include first-trial data (record only the first independent or prompted response per target per session), probe data (a scheduled brief assessment outside of teaching), and frequency counts of mands or initiations. Whatever you choose, decide before the session and be consistent across sessions.
NET for Verbal Behavior
NET is the natural home for verbal behavior interventions because most verbal operants are inherently MO-driven. Sundberg and Partington built the VB approach around the assumption that language is most efficiently taught where it is most functional — which, for most operants, is in the natural environment.
Mand Training
The mand — a request controlled by an MO and reinforced by the requested item — is the canonical NET target. You watch for the MO (the learner reaches for the swing), you evoke the response (“swing”), and you deliver the natural reinforcer (the swing). Mand training in NET should account for the bulk of language teaching time for early learners, because it produces functional communication faster than any other operant.
Tact Training
Tacts — labels controlled by non-verbal stimuli and reinforced socially — can be taught in NET by capitalizing on attentional MOs. When the learner orients to a bird outside, you tact “bird,” and the reinforcer is shared attention and your enthusiastic response. Tact training in NET tends to generalize better than tact training in massed-trial formats, but it requires the learner to have a baseline social reinforcement repertoire.
Intraverbal Expansion
Intraverbals — verbal responses controlled by verbal antecedents — are often taught in DTT, then expanded in NET. Once a learner can answer “what do you eat?” with “pizza” at the table, you bring that response into the kitchen during snack prep, into the car ride home, into conversation with a sibling. NET is where intraverbals become conversation.
Echoic in Context
Echoic teaching in NET typically uses model-prompts within ongoing activities. The learner reaches for juice, you model “juice,” they echo, you deliver. The echoic is transferred to a mand within the same trial — a transfer procedure Sundberg and Partington describe extensively.
NET for Social and Play Skills
NET is well-suited to social and play targets because social behavior is, almost by definition, naturalistic. Pulling social skills to a table and running massed trials rarely produces a learner who initiates with peers at recess.
Peer-Mediated NET
Peer-mediated arrangements use trained peers as the teaching agents. The peer provides the SD (“want to build?”), the learner responds, and the peer delivers the natural reinforcer (joint play). Research on peer-mediated interventions for autistic learners — much of it from the Strain and Odom labs — shows robust effects on social initiations and durable generalization.
Joint Attention
Joint attention is taught in NET by responding to and prompting bids for shared engagement — pointing, showing, alternating gaze between an object and a partner. The reinforcer is the partner’s responsiveness. This is delicate teaching and depends heavily on the partner’s pacing.
Symbolic Play
Symbolic play (pretending the block is a phone, feeding a doll) is often taught in NET through modeling and expansion. The therapist models a play scheme, waits for the learner to imitate or vary, and reinforces variation. PRT in particular has a long research base for symbolic play targets.
Social Initiations
Initiations — greetings, invitations, comments — are NET targets by necessity. They cannot be evoked in a contrived antecedent and still count as initiations. Programming for initiations means arranging environments where the MO for social contact is present and the partner is responsive.
A Worked Case Example
To make this concrete, here is a fictional but realistic session.
Maya is a 4-year-old in a home-based program. Her BCBA has identified mands for action verbs (“open,” “push,” “go”) as a priority target. Her RBT, Jordan, arrives for a 2-hour session at 9 a.m.
Preference assessment. Jordan does a brief free-operant scan: Maya goes straight to a sealed container of magnetic tiles, then to a wind-up dog, then to a bubble wand with the cap on. All three involve a closed container or activated mechanism — perfect MOs for “open” and “go.”
Environmental arrangement. Jordan places the magnetic tiles in a clear bin with a tight lid on the play rug. The bubble wand sits on a low shelf with the cap screwed on. The wind-up dog is in Jordan’s lap, not wound. Maya can see all three.
Teaching trial 1. Maya walks to the magnetic tile bin and tugs the lid. The MO is captured. Jordan kneels beside her, waits 3 seconds (time delay), and when Maya looks up, models “open.” Maya echoes “open.” Jordan opens the bin and hands her two tiles. Maya plays for 90 seconds.
Teaching trial 2. Maya brings the bubble wand to Jordan. Jordan waits. Maya vocalizes “buh.” Jordan prompts: “open.” Maya says “open.” Jordan opens the cap and blows one set of bubbles. Maya laughs and reaches.
Embedded teaching. Across the next 20 minutes, Jordan runs “open” trials with the wind-up dog (a separate latch), a snack pouch, a book with a flap, and a sticker sheet. Five exemplars, naturally distributed.
Generalization probe. At the end of the session, Jordan hands Maya a new closed item she has not seen before — a small treasure chest with a hinged lid. Maya says “open” without a prompt. Jordan records the probe as independent.
Data. Jordan records first-trial data for “open” (prompted), and a frequency count of independent “open” mands across the session (4 independent, 6 prompted, across 5 exemplars). She notes the new exemplar generalization on the probe sheet.
This is a single target in a single session. A real NET program for Maya would have 4-8 active targets across operants, with similar choreography for each.
Common Mistakes and How to Avoid Them
Over-Prompting
The most common NET error is jumping in too fast. The therapist sees the MO, gets excited, and prompts before the learner has a chance to initiate. Over time, this produces prompt dependence and erodes the learner’s spontaneous responding. Fix: build in a deliberate time delay — count to three before you prompt — and track independent versus prompted responses separately.
Missing the MO
The opposite error is missing the window entirely. The learner reached, the therapist was looking at the data sheet, the MO passed, the moment is gone. Fix: minimize within-session data sheet attention, use clipboards or apps that allow quick tally, and assume you will miss some — that is the cost of NET.
Contriving Too Obviously
Contrivance is fine; obvious contrivance is not. If you put the bubbles on the shelf, screw the cap on tighter than necessary, then stare at the child waiting for them to mand, the procedure stops feeling natural and starts feeling extractive. Fix: contrive lightly, and if the learner does not engage with the contrived item, move on.
Poor Data Collection
NET data is harder to take well than DTT data. Many therapists drift into ratings (“good session”) rather than counts. Fix: pick one or two operationally defined targets per session for frequency data, use probes for the rest, and review data weekly with your BCBA.
Ignoring Generalization
NET is supposed to generalize — but it does not generalize automatically. If you teach “open” only on the bubble container, “open” stays a bubble-container response. Fix: program multiple exemplars from day one, vary settings and people, and run periodic novel-stimulus probes.
Where NET Is Not the Right Tool
Being honest about where NET underperforms is part of being a competent BCBA. NET is genuinely not the right tool in several situations.
Acquisition of new motor responses. Imitation of fine motor sequences, handwriting, or specific functional motor skills often requires more repetitions per unit time than NET can deliver. DTT or structured task analysis produces faster acquisition for these targets.
Learners with very low motivation in natural contexts. Some learners — particularly early in programming — have such a narrow range of preferred items that captured MOs are rare. For these learners, structured pairing and preference-expansion work in a controlled setting often comes first; NET comes online once the reinforcer pool is broader.
Skill deficits requiring high-density teaching. Learners with extremely limited verbal repertoires sometimes benefit from a period of high-density echoic and mand training using transfer procedures in a controlled setting, before generalization to NET. The argument is not that NET cannot do this — it can — but that some learners acquire faster with the trial density DTT provides.
Targets with safety implications. If the response you are teaching has safety consequences (e.g., crossing a street, responding to “stop”), you want tight stimulus control before you take it into the natural environment.
Assessment-style data demands. Some funders and programs require trial-by-trial data that NET cannot produce cleanly. This is a programmatic constraint, not a clinical one, but it is real.
Ethical Considerations and Assent in NET
NET has a built-in advantage on the assent front: because the learner initiates, much of the teaching happens with the learner’s active cooperation. But that advantage can mask real ethical complexity, and “the kid kept playing” is not the same as informed assent.
Modern BCBA practice — informed by the BACB Ethics Code 2.11 (Documenting Professional Activity) and 2.15 (Minimizing Risk of Behavior-Change Interventions), and the broader assent literature — treats assent as ongoing, not a one-time checkbox. In NET, that means watching for assent withdrawal signals: the learner turns away, pushes the materials, stops responding, leaves the area. When you see those signals, you back off. You do not chase, you do not block escape, you do not contrive a stronger MO to override the withdrawal.
There is a low-coercion ceiling on good NET. If the learner is not engaging despite well-arranged opportunities, the answer is not to make the opportunities more aversive (locking items away harder, withholding longer). The answer is to widen the reinforcer pool, lower the response requirement, and reassess. NET works because the learner wants to play. The moment it stops working that way, it has drifted into something else.
Practically, this means documenting assent withdrawal in your session notes, having a plan for how the team responds when a learner opts out, and treating opt-outs as clinical data rather than as behavior problems. A learner who opts out of bubbles today is telling you something useful about bubbles, your delivery, the day, or all three.
FAQs
What’s the difference between NET and incidental teaching?
Incidental teaching, as developed by Hart and Risley, is a specific procedure: the learner initiates, the adult elaborates, the adult prompts a slightly more advanced response, and the natural reinforcer is delivered. NET is a broader umbrella term that includes incidental teaching, mand training in the natural environment, embedded discrete trials, and other naturalistic tactics. Incidental teaching is one tool in the NET toolbox.
Is NET evidence-based?
Yes. NET and the broader NDBI family have a substantial evidence base, including multiple randomized controlled trials of PRT and ESDM, single-case research on mand training in natural environments, and consensus reviews (Schreibman et al., 2015 on NDBIs; the National Standards Project and the National Clearinghouse on Autism Evidence and Practice on naturalistic behavioral interventions).
Can RBTs do NET?
Yes — NET is within the RBT scope when implemented under BCBA supervision and according to a written program. The skill ceiling on NET is higher than on DTT, though, because the therapist has to make more in-the-moment judgments. Expect a longer onboarding curve.
Do I need data sheets for NET?
Yes, you need data — but you do not need a trial-by-trial sheet. First-trial probes, frequency counts of independent mands, and weekly generalization probes are common formats. The data system should match the procedure, not fight it.
How does NET compare to DTT for autism intervention?
Both have evidence bases for autistic learners. DTT tends to produce faster acquisition of discrete targets; NET tends to produce better generalization and stronger social-communicative outcomes. Most contemporary programs use both, matched to target and learner. Framing it as an either/or is usually a sign of program inflexibility rather than clinical judgment.
Can NET work for a non-vocal learner?
Yes. NET works with any communication modality — AAC devices, PECS, sign, gestures. The principles (capture MO, learner-initiated, natural reinforcer) are modality-agnostic. Mand training in NET is one of the most common entry points for AAC users.
Is NET the same as play-based ABA?
Overlapping but not identical. “Play-based ABA” is a marketing term more than a clinical one and can refer to anything from rigorous NET to loosely structured play without operationally defined targets. NET is a specific procedural framework with operational targets, prompting plans, and data. If a program describes itself as play-based, ask what the targets and data system look like.
How many targets should I run in a NET session?
It depends on the learner, but 4-8 active targets across operants is a reasonable working range for many early learners. Too few and you waste captured MOs; too many and your prompting and data collection degrade.
What if the learner doesn’t initiate?
Reassess the reinforcer pool first. Then look at environmental arrangement — are preferred items visible? Are they actually preferred today? Then consider whether the learner needs a period of pairing and preference expansion before NET becomes productive. A learner who is not initiating is giving you information; use it.
How is NET supervised?
Live observation is essential because NET is hard to evaluate from data sheets alone. BCBAs supervising NET should be watching pacing, prompt latency, MO capture, and assent — not just whether the data look clean.
Where to Learn More
If you want to keep building your NET skills, the most useful next step is watching clinically relevant CEUs on verbal behavior, NDBIs, and naturalistic teaching from clinicians who actually run programs. BBC’s free ABA CEUs include sessions on assent, motivation, and naturalistic teaching that are actually worth watching — recorded by practicing BCBAs, not corporate training departments. For broader topic coverage, our full CEU course library includes deeper dives on verbal behavior assessment, PRT, and ethics in naturalistic intervention.
NET rewards the BCBAs and RBTs who keep refining the craft. Watch good clinicians work. Take data on yourself. Re-read Sundberg, Partington, Koegel, and Schreibman every couple of years — they get better the more sessions you have under your belt. And keep the learner’s assent at the center of the procedure. That is what separates NET that works from teaching that just happens to look like play.