Shaping: Differential Reinforcement of Successive Approximations — A Practitioner's Guide

Skinner's Foundational Account and What It Still Explains in 2026

Shaping as a formal procedure traces to Skinner's operant conditioning work, where the successive approximation method was used to establish novel behavior in pigeons and rats that would never have occurred spontaneously at sufficient rate to contact reinforcement. Skinner's framework explained that any response sharing topographic or functional dimensions with the terminal behavior could serve as a starting point, and that the practitioner's task was to use those shared dimensions as a bridge. In 2026, that account remains correct and is supported by contemporary applied data. Ghaemmaghami, Hanley, Jessel, and Landa demonstrated that a systematic shaping process applied within a changing-criterion design successfully developed complex functional communication responses (FCRs) from simple FCRs across four children with varying language and intellectual abilities, doing so without resurgence of problem behavior (Ghaemmaghami et al. (2018)). Maintaining low-effort, high-density reinforcement at each criterion step before advancing was identified as the key procedural safeguard — the mechanism that mirrors Skinner's original operant logic and remains the most common source of shaping failure when omitted (Ghaemmaghami et al. (2018)).

The Single Reinforcer Problem: Why Accidental Reinforcement Is Catastrophic During Shaping

Shaping is exquisitely sensitive to inadvertent reinforcement of off-target approximations. Hunter and Rosales-Ruiz modeled the "desperation-driven click" — the practitioner's reflex to deliver a reinforcer after a period of no reinforcement, even for a response that is not the target approximation — in a controlled analogue with college students (Hunter & Rosales‐Ruiz (2019)). When participants received a single reinforcer for interacting with a non-target object following a reinforcement pause, they subsequently interacted with that new object for a duration equal to or exceeding their interaction with an object that had a substantially longer reinforcement history (Hunter & Rosales‐Ruiz (2019)). The implication is stark: one accidentally delivered reinforcer during a shaping gap can produce a competing response that matches the strength of a deliberately shaped approximation. Practitioners who interpret inactivity during a shaping session as grounds for accepting any behavior they see will systematically generate these competing approximations, fragment the shaping trajectory, and then attribute the resulting stall to the learner rather than to the procedure.

Shaping Functional Communication: From Simple to Complex FCRs

The FCT-to-complex FCR progression studied by Ghaemmaghami and colleagues illustrates the core procedural constraint in shaping: response efficiency drives which topography the learner will choose (Ghaemmaghami et al. (2018)). Initial FCT succeeds precisely because the simple FCR — a single word, a card exchange, a device activation — is low-effort and immediately effective. Attempting to replace that simple FCR with a grammatically complete, socially appropriate sentence without shaping the transition typically produces resurgence of problem behavior, because the learner's behavioral history now favors the simpler, less effortful form. A changing-criterion design — in which the minimum acceptable FCR complexity is raised only after mastery at the current level — solves this problem procedurally: the learner is never asked to produce a response that is far from what they are currently producing and successfully contacting reinforcement for (Ghaemmaghami et al. (2018)). The practitioner's documentation job in this design is to operationally define each criterion step before the session begins, record whether the step was met, and document the threshold for advancement — not to make in-session judgment calls about when the behavior "looks close enough."

Percentile Schedules: A Data-Driven Approximation Tool

Milyko described K-schedules — shifting percentile reinforcement schedules — as a data-driven mechanism for shaping response frequency upward within precision-teaching frameworks (Milyko (2021)). The K-schedule sets the reinforcement criterion at a defined percentile of the learner's recent performance history (e.g., beat the 40th percentile of the last five timings), then recomputes after each session so the criterion rises only as performance actually improves (Milyko (2021)). This mechanism eliminates two common shaping errors simultaneously: it prevents criterion creep (advancing before the behavior is stable) by anchoring the shift to actual data, and it prevents reinforcement gaps (the condition that produces desperation-driven clicks) by keeping the schedule dense early and thinning it only proportionally to performance gains. The protocol is directly applicable to any fluency-building application where the target is rate of responding — academic tasks, vocal production, written responses — rather than topographic accuracy alone.

Shaping Eye Contact Without Prompts

Fonger and Malott demonstrated that shaping eye contact without the use of prompts produced socially significant increases and naturalistic generalization in two young boys with autism receiving early intensive behavioral intervention (Fonger & Malott (2019)). The key design feature was starting from the child's current level — brief, incidental gaze shifts — and reinforcing successively longer durations before granting access to a preferred toy. No verbal or physical prompts were used, which meant that the eye contact that emerged was unprompted, and that the generalization to natural contexts required no separate fading step (Fonger & Malott (2019)). Shaping without prompts produced a response free of prompt dependency — a critical advantage whenever the terminal behavior must occur spontaneously in a natural environment, because no separate prompt-fading step is required to produce the generalized form (Fonger & Malott (2019)).

Shaping in Feeding Intervention

Demchuk, Ibarra, and Hansen evaluated a stepwise pairing-and-shaping protocol for two children with autism and avoidant restrictive food intake disorder (ARFID) (Demchuk et al. (2026)). When pairing alone did not decrease inappropriate behavior or increase new-food acceptance, adding systematic shaping steps — moving from looking at the food, to touching it, to licking it, to taking a bite — produced acceptance of four new foods for one participant without requiring restrictive seating or spoon-nonremoval procedures (Demchuk et al. (2026)). The authors recommend a data-triggered decision rule: if pairing alone produces no movement within three to four sessions, introduce shaping by reinforcing the closest available approximation to acceptance, and track which step the learner stalls on to guide individualized sub-steps. For the second participant, shaping with pairing was ultimately insufficient and spoon-nonremoval procedures were required — a critical finding because it illustrates that shaping is not always the terminal procedure in feeding intervention; it may be a component that reduces the restrictiveness needed, rather than eliminating intensive procedures altogether.

Shaping Complex Husbandry Behavior: Granularity Matters

Waite, Kodak, and Whang developed and validated a 24-step shaping protocol for teaching companion dogs to tolerate ear cleaning, using a caregiver-implemented multiple-baseline design across six dogs ((Waite et al., 2025)). The study found that the stringent 100%-correct-over-two-sessions mastery criterion and the large number of pre-programmed shaping steps — far more than typical behavior chains for motor tasks — were important for producing errorless acquisition of the ear-handling sequence ((Waite et al., 2025)). The take-home for clinical shaping is that the optimal number of steps is not determined by the practitioner's intuition about "how similar" adjacent approximations look; it is determined empirically, session by session, by whether the learner meets the mastery criterion cleanly before advancement. When they do not — especially when the task involves physical proximity or handling — more sub-steps and a stricter mastery criterion reduce both procedural errors and behavioral resistance.

Shaping Social Engagement: Level Systems as Scaffolding

Cihon and colleagues embedded a flexible shaping hierarchy within a five-level point system for two adolescents with autism learning synchronous face-to-face engagement (Cihon et al. (2019)). Shaping steps moved from brief eye contact (3 seconds) through coordinated back-and-forth interaction with increasing complexity, with advancement to the next level contingent on meeting the current step's performance criterion (Cihon et al. (2019)). Embedding shaping within a level system served two practical functions: it made the current criterion visible to the learner (reducing demand-avoidance by clarifying what the contingency was), and it preserved flexible clinical judgment — the "flexible shaping" language in the title refers to the practitioner's ability to individualize which specific skills constituted the next step within each level, rather than following a predetermined script that would not account for individual response patterns.

Shaping in Organizational Behavior Management

The application of shaping to staff and organizational performance is explicit in both the OBM literature and contemporary ABA agency-systems work. Townsend and colleagues described organizational practices within high-performing autism service organizations as "prompted and shaped" to increase staff access to reinforcement from stakeholders ((Townsend et al., 2024)). The LaMarca and LaMarca ADDIE framework for program development states explicitly that "the best way to improve most professional behaviors is through a shaping process," recommending that BCBAs building clinical programs select one manageable development step at a time and reinforce gradual expansion rather than attempting complete overhaul at once ((LaMarca et al., 2024)). Flynn and Wilder's evaluation of task clarification and feedback for teaching feedback reception skills in an OBM context showed that contingent performance feedback shapes accurate professional behavior even when the target involves meta-level skills (receiving feedback gracefully) rather than observable motor tasks (5b68158d-ac6a-4937-809b-e4b5fc9d7f1f).

Shaping Verbal Behavior: Mand Topography and Echoic Development

In verbal behavior applications, shaping operates on both the topographic form of vocalizations and the functional independence of communicative responses. Da Silva and Williams reviewed autoshaping of learner vocalizations through stimulus-stimulus pairing (SSP), distinguishing autoshaping — in which behavior is elicited by contingency-independent pairings — from operant shaping, in which differential reinforcement builds approximations toward a target vocalization (da Silva & Williams (2020)). The SSP literature documents that early vocal behavior is better established through respondent-based autoshaping procedures in prelingual learners, with operant shaping becoming the appropriate tool once the learner has established basic vocal repertoire and can contact differential reinforcement for phonemic approximations. Frampton, Axe, Davis, and colleagues demonstrated that shaping indicating responses (IRs) — the motor behaviors that accompany mands, such as pointing — is best accomplished through prompt fading from most-to-least assistance, with each decrease in physical guidance representing a successive approximation toward independent pointing ((Frampton et al., 2024)). The temporal-relational responding study by Barry, Neufeld, and Stewart extended shaping logic to abstract language learning, showing that beginning with highly nonarbitrary stimuli (photographs with clear before/after relationships) and fading systematically toward arbitrary symbolic stimuli produced mastery and generalization of temporal relational responding in five adolescents with autism ((Barry et al., 2024)).

Shaping Emotion Identification: Iterative Program Design as Applied Shaping

Linnehan's iterative teaching program for emotion labeling in autistic adolescents exemplifies shaping logic applied to curriculum design itself: the instructional sequence was systematically revised after each learner's errors identified which stimulus features were controlling responding, and those revisions represented micro-criterion shifts in the difficulty and perceptual salience of the teaching stimuli ((Linnehan, 2025)). Juxtaposing example and non-example video clips, starting with fully spelled-out language before abbreviations, and re-entering the design-test-revise cycle after each participant's data all parallel the within-session shaping logic: the current criterion is held constant until mastery, then shifted incrementally in the direction of the terminal behavior ((Linnehan, 2025)). O'Connor and colleagues confirmed the same approach in teaching multiple-control emotion-context relations in children with autism: when standard errorless teaching failed, reducing task complexity by starting with two stimulus classes and gradually adding one new class per block — a shaping-like fade in task demand — produced mastery for learners who had stalled (O’Connor et al. (2020)).

Changing-Criterion Design as the Natural Measurement Tool for Shaping

Diller and Li noted that the changing-criterion single-case design is the methodologically appropriate framework for demonstrating experimental control during shaping, because it requires behavior to track a succession of criterion shifts — if responding rises and falls with each shift, the data demonstrate that the contingency, not maturation or history, is responsible ((Diller & Li, 2025)). The same paper cautions that the design is unsuitable for behaviors that cannot tolerate gradual change or that are vulnerable to external disruptions that flatten the criterion-tracking pattern, which maps directly onto clinical situations where shaping is inappropriate: behaviors where any approximation of the terminal form is dangerous (e.g., some topographies of self-injury), where the environment will not maintain differential reinforcement consistently across criterion shifts, or where the terminal behavior needs to be established rapidly.

Step 1: Define the Terminal Behavior in Observable, Measurable Terms

The terminal behavior must be defined in language a second clinician can replicate without verbal instruction from the original practitioner. For echoic shaping, this might be "production of the target phoneme /s/ in isolation without imitative prompt, across three consecutive sessions at ≥80% correct." For on-task duration in a classroom, it might be "maintaining work behavior for 10 consecutive minutes with no more than one 30-second off-task interval, without teacher prompting." Vague terminal behaviors ("speak more clearly," "stay on task longer") cannot anchor a shaping program because there is no operationally defined success criterion — which means there is no way to know when the final approximation has been reached or whether the program has succeeded.

Step 2: Define the Starting Approximation and All Intermediate Steps Before Beginning

Select the current response that the learner produces reliably under current conditions and that shares at least one dimension with the terminal behavior. The starting approximation must be reinforceable — if it never occurs, you cannot reinforce it. Map the full approximation sequence before the first session by asking: what is the smallest change in the response that the learner could plausibly make given one to three sessions of dense reinforcement at the current step? Write each step in operational terms. Having the full sequence documented ahead of time prevents the most common procedural error in clinical shaping: making ad hoc criterion decisions in-session under time pressure, which produces both criterion creep (advancing before stability) and inadvertent reinforcement of off-target forms.

Step 3: Set and Apply Criterion Shift Rules Prospectively

Define the mastery criterion — the performance level at which the current approximation is "stable enough" that the criterion can shift — before the session begins. Common options include: a percentage correct over a defined number of consecutive trials (e.g., 80% over two consecutive sessions), a rate criterion (responses per minute above a threshold), or a duration criterion (sustained behavior for N consecutive seconds). The K-schedule approach described by Milyko provides an empirically grounded, session-adaptive alternative that automatically adjusts the criterion to track the learner's percentile performance rather than a static threshold (Milyko (2021)). Avoid making criterion shifts based on subjective impressions of "readiness" without quantitative support; this is the primary pathway to premature criterion shift, which exposes the learner to extinction during a step they have not mastered.

Step 4: Deliver Differential Reinforcement Consistently Throughout

Differential reinforcement during shaping means: reinforce every instance of the current target approximation immediately and consistently; do not reinforce responses that fall short of the current criterion; and do not reinforce responses that exceed the current criterion if they are not being targeted for this step. The last element is counterintuitive but important: if a learner occasionally produces the terminal behavior spontaneously while the shaping program is still in early steps, those instances should be reinforced — but the practitioner should not advance to the terminal criterion simply because occasional instances occur. Stability at the current criterion matters because it builds behavioral momentum and prevents the fragile, inconsistent responding that results from skipping steps.

The three variables that determine approximation quality are dimensional distance from the terminal behavior, increment size between successive steps, and the degree of response variability the learner shows at the current criterion.

Dimensional distance refers to how different the current approximation is from the terminal behavior along the measured dimension (form, duration, rate, force, latency, or topographic accuracy). Beginning too far from the terminal behavior requires more steps, which increases total session time and opportunity for reinforcement gaps. Beginning too close requires the learner to produce behavior they cannot yet reliably emit at the first session, which produces immediate extinction and often problem behavior. A practical heuristic: the starting approximation should occur at least once in a five-minute baseline observation without any programmatic intervention.

Increment size is the single most consequential variable in whether a shaping program succeeds. Large increments — advancing by too much in a single criterion shift — expose the learner to extinction for responses that, moments earlier, produced reinforcement consistently. Smaller increments maintain a higher rate of reinforcement per unit time during the transition, which prevents the reinforcement gaps that produce desperation-driven clicks and competing approximations (Hunter & Rosales‐Ruiz (2019)). The Waite et al. ear-cleaning protocol used 24 steps for a behavior chain that some clinicians would have attempted in 6 to 8 steps, and the stringent mastery criteria and fine step-granularity were specifically identified as contributing to the protocol's errorless acquisition ((Waite et al., 2025)). When a learner stalls between steps, the correct first response is to add an intermediate step between the current and the next criterion, not to increase reinforcer magnitude or session duration.

Response variability at the current criterion is both a challenge and an opportunity. High response variability means the learner is already exploring behavioral space near the terminal behavior, and that the practitioner can "catch" instances of the next approximation without waiting for spontaneous drift. Low variability (highly stereotyped responding at the current criterion) means the learner's behavior is well-established but exploration of adjacent forms is unlikely without explicit variation-promoting procedures, which may include differential reinforcement of other (DRO) applied to response variability itself, or deliberate extinction of the most common current form to induce exploration.

At the introduction of any new criterion step, continuous reinforcement (CRF) — delivering a reinforcer after every instance of the approximation — is the appropriate schedule. CRF maximizes acquisition rate and keeps the reinforcement gap minimal, which is the condition under which the learner is least likely to contact extinction and least likely to emit desperation-driven competing responses (Hunter & Rosales‐Ruiz (2019)). Once the approximation is stable at CRF, the schedule can be thinned to a lean intermittent schedule before advancing the criterion, but the thinning should occur within the current criterion step — not simultaneously with a criterion shift. Changing both the reinforcement criterion and the schedule in the same step introduces two sources of potential extinction simultaneously, which typically produces significant disruption. After a criterion shift, return to CRF for the new approximation until it is stable, then thin again.

The K-schedule protocol provides a data-driven alternative to manually managing this CRF-to-thin-to-advance sequence: by tying the reinforcement criterion to a percentile of recent performance, K-schedules automatically start dense when performance is low and thin organically as performance improves (Milyko (2021)). This approach is particularly valuable in precision-teaching contexts where session-by-session data are already being graphed on Standard Celeration Charts and the operational infrastructure for percentile computation is available.

When to Shape, When to Prompt, and When to Do Both

Shaping and prompting solve different problems. Shaping is appropriate when the terminal behavior does not yet exist in the learner's repertoire and must be built from an existing response that shares dimensional features with the goal. Prompting is appropriate when the terminal behavior exists (or can be produced with assistance) and the goal is to bring it under the control of natural antecedent stimuli. These two situations frequently co-occur in practice, and many clinical programs blend both procedures — but the blend requires explicit procedural planning or the two techniques will interfere with each other.

When prompting is used inside a shaping program, the key risk is that the prompt becomes the maintaining antecedent for the approximation rather than the natural setting event. Fonger and Malott's eye-contact shaping study deliberately excluded prompts to avoid this outcome — the target behavior had to emerge from the learner's own behavioral variability in response to the differential reinforcement contingency, because unprompted eye contact was the clinical goal (Fonger & Malott (2019)). In contrast, Frampton and colleagues' mand training tutorial uses prompt fading as the shaping mechanism: the physical prompt for pointing is the starting approximation, and systematic reduction of the prompt — from full physical guidance through graduated physical guidance to gesture to no assistance — constitutes the successive approximations toward independent indicating ((Frampton et al., 2024)). This hybrid is entirely appropriate when the terminal behavior requires a specific topography that the learner cannot produce variably without some initial form guidance, and when the clinical goal includes eventually responding to natural EOs without artificial cues.

The decision rule: if the learner can produce a variable approximation in the absence of a prompt that shares a dimension with the terminal behavior, start with shaping alone. If no such variable approximation exists, introduce the terminal behavior via prompting first, then fade the prompt. Yarzebski and Dickson confirmed that caregivers can learn to implement graduated guidance with high procedural integrity after a brief video-modeling package, making prompt-fading-as-shaping a scalable procedure for caregiver-implemented home programs ((Yarzebski & Dickson, 2024)).

Fading Prompts Within Shaping Steps

When prompts and shaping co-occur, each reduction in prompt level should be treated as a criterion shift: the response must be stable at the current prompt level before reducing assistance. This prevents the learner from contacting extinction at the new (lower) prompt level before their behavior has stabilized in that condition. The mastery criterion for prompt fading should be operationally identical to the mastery criterion used for topographic shaping steps — a defined percentage correct over a defined number of consecutive sessions — not a clinician's subjective impression that the learner "seems ready."

When Shaping Is the Right Tool

Shaping is indicated when: (1) the terminal behavior does not exist in the learner's repertoire; (2) the terminal behavior can be approached along a continuous dimension (duration, form, frequency, intensity); (3) the learner shows at least some approximation that shares a measurable feature with the terminal behavior; and (4) the contingency environment permits consistent differential reinforcement (i.e., the practitioner can deliver reinforcers selectively without inadvertent reinforcement from other sources) (Ghaemmaghami et al. (2018)).

When Chaining Is the Right Tool

Behavioral chaining is indicated when the terminal behavior is a sequence of discrete responses that must occur in a fixed order, and each response in the chain is either already in the learner's repertoire individually or can be taught through DTT ((Frampton et al., 2024)). Chaining and shaping are frequently confused because both involve "building toward" a terminal behavior, but they differ in the unit of analysis: chaining builds a sequence of already-existing discrete responses; shaping builds a novel response form. The Waite et al. ear-cleaning protocol used shaping of each component handling step even though the overall procedure looks like a behavior chain — the distinction is that tolerating each level of ear-contact handling was not already in the dogs' repertoires and had to be built through differential reinforcement of successive approximations within each step ((Waite et al., 2025)). Frampton and colleagues similarly recommend that mand programs first teach behavioral chains to fluency before contriving establishing operations for mand training — because chain interruption requires each chain step to function as a conditioned reinforcer, which only happens if the step has been trained to mastery ((Frampton et al., 2024)).

When Discrete Trial Teaching Is the Right Tool

Discrete trial teaching (DTT) is indicated when the target response is already topographically established and the goal is to bring it under appropriate stimulus control ((LaMarca et al., 2024)). DTT's massed-trial format maximizes reinforced practice per unit time and provides clear antecedent structure. Shaping inside DTT is appropriate when a component response (e.g., a specific phoneme, a specific motor form) must be developed within the trial structure, but the trial format itself is not a shaping procedure ((LaMarca et al., 2024)). LaMarca and LaMarca describe progressive DTT variants — flexible prompting, varying instructions as soon as possible, instructive feedback — that incorporate aspects of shaping logic into the trial-by-trial clinical decision-making without constituting formal shaping programs ((LaMarca et al., 2024)).

The Decision Table

Goal	Appropriate Primary Procedure	Rationale
Build a novel response form (e.g., vocal approximation, gaze duration)	Shaping	Response does not exist; must be generated through differential reinforcement
Teach a sequence of existing discrete responses	Chaining	Each component is already established; sequencing is the target
Bring an existing response under SD control	DTT	Stimulus control, not response form, is the gap
Increase frequency/rate of existing response	Differential reinforcement of high rate (DRH) or shaping via schedule	Response exists; rate is the dimension to shape
Transition from prompted to unprompted responding	Prompt fading (± shaping)	Topography exists under prompt; independence is the target

Verbal Behavior: Echoic Shaping

Echoic shaping — differential reinforcement of successive vocal approximations toward a target phoneme or word — is the foundation of early speech-production programming in learners with limited vocal repertoires. The starting approximation is the learner's current best vocalization that shares any phonemic feature with the target: a vowel sound, a consonant initial position, a breath-only vocalization for learners at the earliest stages. Da Silva and Williams reviewed the conditions under which stimulus-stimulus pairing can establish the vocal precursors that make echoic shaping tractable, noting that autoshaping (respondent-based elicitation) and operant echoic shaping (differential reinforcement of approximations) operate through different mechanisms and may need to be sequenced rather than substituted for each other (da Silva & Williams (2020)). The Barry and colleagues temporal-relational responding study demonstrates that shaping also extends to abstract verbal repertoires, with nonarbitrary-to-arbitrary stimulus fading functioning as the approximation sequence for language concepts that cannot be approached through articulatory shaping ((Barry et al., 2024)).

Motor Skills

Motor shaping follows the same four-step procedure but operates on movement dimensions: force, range of motion, bilateral coordination, accuracy, or speed. The husbandry-handling literature is particularly instructive because it requires tolerating an aversive stimulus (handling) rather than producing a positive motor response, and because the errorless mastery criteria in those protocols directly prevent the escape-motivated behavior that physical-prompt-based motor teaching frequently generates ((Waite et al., 2025)). For learners with motor impairments, shaping is often the only option — prompting cannot produce a movement the musculature cannot yet execute — and the increment size must be calibrated against the learner's motor capacity, not against a population-level developmental milestone.

Social Behavior

The eye-contact shaping protocol by Fonger and Malott and the level-system shaping program by Cihon and colleagues both illustrate the social behavior application: a practitioner reinforces progressively longer, more complex, or more contextually appropriate social responses (Fonger & Malott (2019)) (Cihon et al. (2019)). St. Clair and colleagues used sequential shaping phases to develop a generalized repertoire of playful deception in autistic children aged 8 to 11, beginning with foundational trick components under rules-and-modeling and progressing to multiple-exemplar training for novel flexible trick creation ((St. Clair et al., 2024)). The critical feature in social shaping is that the natural environment must maintain differential reinforcement after the programmed contingency is withdrawn — social behavior that is shaped in clinic but never contacts natural peer reinforcement will extinguish in the community regardless of how well the shaping program was executed.

Eating and Feeding

As documented by Demchuk and colleagues, shaping in feeding intervention works through a food acceptance hierarchy: the practitioner identifies the closest the learner currently comes to accepting the target food (e.g., tolerates the food being on the table) and reinforces that approximation consistently, then systematically advances the criterion through proximity, visual contact, touching, smelling, licking, and biting (Demchuk et al. (2026)). The practical advantage of shaping in feeding contexts is that it avoids physical prompt procedures (spoon-nonremoval) that have significant ethical and assent concerns, and it tends to produce higher caregiver acceptability ratings because the procedure is graduated and the learner's distress is minimized at each step. The clinical limit, as Demchuk and colleagues show, is that shaping may not be sufficient for the most severe feeding presentations, particularly when a child's history of aversive feeding experiences has established strong conditioned emotional responses that require more intensive procedures to override.

Vocational Fluency and Organizational Performance

In OBM contexts, shaping is applied to work behavior rate and accuracy, staff feedback reception, and organizational systems development. Flynn and Wilder's evaluation demonstrated that task clarification specifies the current criterion while contingent feedback functions as the differential reinforcement mechanism that shapes accurate professional behavior over time (5b68158d-ac6a-4937-809b-e4b5fc9d7f1f). The ADDIE framework applied to ABA program development explicitly models practitioner skill development as a shaping trajectory: start anywhere, reinforce the next small step, and build systematically over years rather than attempting complete adoption at once ((LaMarca et al., 2024)).

Clinic-Based Skill Acquisition

The clinic setting provides the practitioner the greatest control over antecedent conditions, reinforcer delivery timing, and the differential reinforcement contingency. In clinic, shaping programs should be manualized with operationally defined approximation steps, mastery criteria, advancement rules, and data sheets that a second clinician can implement with high procedural fidelity. The Ghaemmaghami et al. FCT shaping protocol and the Waite et al. ear-cleaning protocol are both described as suitable for "typical teaching contexts" — meaning they are not laboratory-only procedures but do require systematic documentation (Ghaemmaghami et al. (2018)) ((Waite et al., 2025)).

Classroom Applications

In school settings, shaping is most commonly applied to on-task duration, handwriting accuracy, assignment completion, and classroom social behavior. The Torelli and colleagues kindergarten life skills study used progressive intertrial interval lengthening — a schedule-shaping tactic that gradually thins reinforcement density while maintaining performance — as a component of Tier-3 instruction ((Torelli et al., 2026)). Neef and colleagues showed that BST targeting specific praise statements produced teacher verbal behavior change that generalized to improved student behavior management — with teacher instructional behavior itself shaped through the BST sequence (Neef et al. (2024)). The classroom constraint that most limits shaping is reinforcement continuity: if the teacher or paraprofessional who will implement differential reinforcement is not present consistently, or if other adults in the room inadvertently reinforce off-criterion approximations, the shaping contingency is disrupted at both ends simultaneously.

Home-Based Programs

In home settings, caregiver implementation of shaping requires the same fidelity as clinic-based programming but with a different training model. Yarzebski and Dickson's video-modeling study showed that a brief (4-minute) video plus procedural rehearsal was sufficient to train accurate graduated guidance (a prompt-fading shaping procedure) in parents of preschool children with autism ((Yarzebski & Dickson, 2024)). For toilet training, mealtime behavior, and home hygiene skills — all common home-shaping targets — the practitioner's role is to write the shaping protocol clearly enough that the caregiver can implement it without same-session supervision. The protocol must specify what counts as the current criterion approximation, what constitutes reinforcement, what to do when the learner does not produce the approximation within the session, and exactly how many sessions to run at each criterion step before contacting the clinician about advancement.

Athletic Coaching

The athletic coaching context is the clearest example of shaping outside ABA's traditional population: coaches systematically reinforce successive approximations toward target form in throwing, batting, gymnastics, and swimming through differential performance feedback. The dimensional shaping here often operates on force, trajectory, timing, and bilateral symmetry simultaneously, with the coach attending to whichever dimension is currently farthest from criterion and holding others constant. OBM-trained sports psychologists have applied formal shaping protocols to athletic skill acquisition, with the shared conclusion that coaches who provide contingent, behavior-specific feedback produce faster acquisition and more durable form than those who provide general encouragement or punishment of error.

Organizational Behavior Management

The OBM application of shaping is documented in staff training, performance improvement, and organizational systems change. The Townsend et al. description of agency practices being "prompted and shaped" toward stakeholder reinforcement ((Townsend et al., 2024)) and the Flynn and Wilder feedback-reception shaping study (5b68158d-ac6a-4937-809b-e4b5fc9d7f1f) both illustrate that shaping is not restricted to skill acquisition with clients — it is an equally appropriate procedure for any behavior where the terminal performance does not currently exist and must be built through differential reinforcement of successive approximations.

Criterion Creep

Criterion creep is the pattern of advancing the shaping criterion before the current approximation is stable, driven by practitioner eagerness, time pressure, or the absence of a prospectively defined mastery threshold. It is the most common shaping error and produces characteristic data: a learner who appears to "hit a wall" partway through the shaping sequence, with variable responding that does not improve across sessions. The fix is not more sessions at the current (too-advanced) criterion; it is returning to the last step where responding was stable, rebuilding mastery there, and then advancing more slowly with smaller increments.

Premature Criterion Shift

Closely related to criterion creep, premature criterion shift refers specifically to advancing the criterion after a single session that appears successful rather than requiring stability across multiple sessions. This error is encouraged by clinical pressure to demonstrate progress and by data systems that score only percentage correct without tracking stability across sessions. The Ghaemmaghami et al. changing-criterion FCT protocol addressed this explicitly by requiring mastery across multiple sessions before advancement (Ghaemmaghami et al. (2018)). Define "stable" before the program begins: how many consecutive sessions, at what percentage correct, constitutes mastery?

Inadequate Reinforcer Pairing and Satiation Management

A shaping program requires dense reinforcement delivery, particularly at the introduction of each new criterion step. Dense delivery creates satiation risk, especially when edible reinforcers are the primary medium. Inadequate satiation management produces apparent extinction — the learner stops responding not because the shaping step is too large but because the reinforcer has lost its value mid-session. Morris and colleagues' survey of clinician practices around reinforcer changing during teaching sessions confirms that practitioners do change reinforcers during sessions when responding drops, but not always systematically, and that the decision is often based on impression rather than pre-defined satiation criteria ((Morris et al., 2024)). Identify backup reinforcers before the shaping session, rotate primaries if sessions are long, and monitor session-level reinforcer effectiveness by tracking the rate of responding on the initial trials of each session ((Morris et al., 2024)).

Missing Approximation Operationalization

A shaping program that defines the terminal behavior but not each intermediate approximation in operational terms will produce inconsistent criterion application across sessions and clinicians. The result is that a given learner's approximation may be reinforced by one RBT and placed on extinction by another, depending on each person's subjective threshold. Document each approximation step in language that does not require the practitioner to make a topographic judgment call: "the learner's gaze meets the clinician's eyes for ≥2 seconds" is operationalized; "the learner makes eye contact" is not.

Conflating Shaping with Prompting and Fading

The most conceptually important pitfall is treating all gradual skill development as shaping when some of it is actually prompt fading. If the learner is producing the terminal behavior correctly under a full physical prompt, the relevant procedure is graduated guidance (or another prompt-fading strategy), not shaping — because the terminal response form already exists under the prompt. Shaping and prompt fading often co-occur within the same program, as in the IR mand-training tutorial by Frampton and colleagues ((Frampton et al., 2024)), but calling prompt fading "shaping" leads to incorrect procedural decisions: specifically, it leads practitioners to search for "intermediate approximations" when the issue is actually the antecedent prompt level, not the response form.

The Desperation Click: Accidental Reinforcement During Pauses

Hunter and Rosales-Ruiz documented the specific mechanism by which shaping programs go wrong during inactivity: after a period of no reinforcement, the practitioner delivers one reinforcer for a non-target response (Hunter & Rosales‐Ruiz (2019)). The correct response to a reinforcement gap during shaping is to simplify the current criterion — accepting a less precise approximation temporarily to restore the reinforcement flow — not to deliver a reinforcer for whatever behavior is occurring at that moment. Distinguish between "backing up the criterion" (procedurally appropriate) and "accepting any behavior to end the gap" (procedurally incorrect).

The shaping evidence base in this corpus rests primarily at the single-subject experimental design (SCED) layer, with supporting methodology and tutorial papers that provide procedural guidance without independent empirical validation (Ghaemmaghami et al. (2018)) (Fonger & Malott (2019)). The Ghaemmaghami et al. FCT shaping study and the Fonger and Malott eye-contact study used multiple-baseline designs — the strongest SCED evidence for a functional relation (Ghaemmaghami et al. (2018)) (Fonger & Malott (2019)). The Hunter and Rosales-Ruiz single-reinforcer study used a laboratory analogue with college students — its ecological validity is limited to informing procedural principles rather than demonstrating treatment outcomes in applied populations (Hunter & Rosales‐Ruiz (2019)). The Waite et al. ear-cleaning protocol and the Demchuk et al. feeding study are applied SCEDs with small samples (6 and 2 participants respectively) that provide existence-proof evidence for the procedures but cannot speak to the range of learner characteristics across which the protocols will generalize ((Waite et al., 2025)) (Demchuk et al. (2026)). The K-schedule protocol by Milyko is a methodology paper without an independent empirical test — it provides a logically coherent framework grounded in existing reinforcement theory but should be treated as a promising clinical tool pending systematic replication (Milyko (2021)). What is notably absent from this corpus is a large-scale comparative study examining shaping against alternative procedures (e.g., modeling, direct instruction, errorless teaching) across matched learner populations — the kind of evidence that would allow evidence-based selection of shaping versus alternatives for a defined learner profile (O’Connor et al. (2020)).

1. Define the terminal behavior in observable, measurable terms before writing any approximation step. A vague goal cannot anchor a criterion shift, and a criterion you cannot operationally define is a criterion you will apply inconsistently across clinicians and sessions.

2. Write all intermediate approximation steps before the first session, not during it. In-session criterion decisions under time pressure are the primary pathway to both criterion creep and desperation-driven clicks (Hunter & Rosales‐Ruiz (2019)).

3. Begin every new approximation step on CRF, then thin the schedule before advancing. Do not change both the response criterion and the reinforcement schedule simultaneously — doing so introduces two sources of extinction at once.

4. When a learner stalls between two steps, add an intermediate step rather than extending time at the current criterion. The correct diagnosis of a stall is "the increment was too large," not "the learner isn't ready." ((Waite et al., 2025))

5. Never deliver a reinforcer during a shaping pause for a non-target response. A single reinforcer for an off-target approximation can establish a competing behavior with the same strength as a deliberately shaped approximation (Hunter & Rosales‐Ruiz (2019)). Back up the criterion instead.

6. Operationally define every approximation step in language that does not require topographic judgment. "Gaze meets eyes for ≥2 seconds" is operationalized; "makes eye contact" is not. Inconsistent criterion application across implementers will fragment the shaping trajectory.

7. Distinguish shaping from prompt fading. If the terminal behavior exists under a full physical prompt, the procedure is graduated guidance, not shaping. Misidentifying prompt fading as shaping leads to searching for "approximations" when the issue is antecedent support.

8. For FCT programs, maintain response efficiency at each criterion step before advancing topographic complexity. Ghaemmaghami et al. show that attempting to advance FCR complexity without a systematic shaping transition reliably produces resurgence of problem behavior (Ghaemmaghami et al. (2018)).

9. Use K-schedules or percentile-based criteria when fluency (rate) is the target dimension. Fixed aim-line criteria produce lean reinforcement schedules for learners whose current rate is far below the aim, which generates the exact reinforcement-gap conditions that derail shaping (Milyko (2021)).

10. Monitor for satiation mid-session, especially during dense CRF delivery at new approximation steps. Identify backup reinforcers and rotation protocols before the session begins, not after the learner's responding drops.

11. In home and classroom settings, confirm that the implementer understands what the current criterion is and what not to reinforce. Inadvertent reinforcement of off-criterion responses by a second caregiver or paraprofessional can disrupt shaping as effectively as a deliberately delivered desperation click (doi:10.1007/s40617-024-00969-3).

12. Use the changing-criterion design when you need to demonstrate experimental control over a shaping program. The design is the natural measurement tool for shaping because it requires behavior to track each successive criterion shift — if it does, the data demonstrate that the contingency, not maturation, is responsible ((Diller & Li, 2025)).

Q: What is the difference between shaping and chaining?

Shaping builds a novel response form through differential reinforcement of successive approximations — the target response does not yet exist in the learner's repertoire. Chaining assembles a sequence of responses that are individually established, using the completion of each step as both the reinforcer for that step and the discriminative stimulus for the next. A practitioner may use shaping to build each individual chain step and then chain the steps into a sequence — the two procedures are not mutually exclusive, but they are conceptually distinct operations on different behavioral units.

Q: When should I use shaping instead of prompting?

Use shaping when the terminal behavior does not exist in the learner's repertoire and cannot be physically prompted into existence (e.g., a specific vocal sound, a spontaneous social initiation, a tolerating response to aversive stimuli). Use prompting when the terminal behavior can be physically assisted into occurring and the goal is to bring it under natural antecedent control through fading. Use both together when the topography must be initially guided (prompt) but the independence of the response also requires building from current behavioral variability (shaping). The Fonger and Malott eye-contact study demonstrates that excluding prompts and relying solely on shaping can produce more naturalistic generalization for behaviors that need to occur spontaneously (Fonger & Malott (2019)).

Q: How do I know if my criterion shift was premature?

The clearest signal of a premature criterion shift is variable responding that does not stabilize across sessions at the new criterion: the learner contacts the advanced step inconsistently, and your session-level data show no upward trend after three to five sessions. A secondary signal is an increase in problem behavior or task avoidance, which is characteristic of extinction — the learner has been moved into a condition where their previous (now below-criterion) responses are not producing reinforcement and the new criterion response is not yet stable enough to produce it either. Return to the last stable criterion step and advance more slowly.

Q: Can shaping be used with learners who have very low response variability?

Yes, but it may require deliberate variability-induction procedures before shaping proper can begin. Options include: extinction of the most common current response form (which typically increases variability through extinction-induced variation), lag schedules of reinforcement (reinforcing only responses that differ from the previous N responses), and high-probability instructional sequences that precede shaping trials and increase behavioral momentum. Response variability is both a prerequisite for shaping and a dimension that itself can be shaped.

Q: What is the changing-criterion design and why is it used to evaluate shaping?

The changing-criterion design is a single-case experimental design in which the criterion for reinforcement is shifted across phases and the learner's behavior is expected to track each shift. It is the natural measurement tool for shaping because it operationalizes the shaping logic directly into the research design: if behavior rises and falls in correspondence with each criterion change, and if this happens predictably across multiple shifts, the data demonstrate that the reinforcement contingency — not maturation, practice effects, or history — is responsible for the behavior change ((Diller & Li, 2025)). The design does require that the target behavior tolerate gradual change; it is not appropriate for behaviors that need rapid transition (e.g., dangerous behaviors where any occurrence is unacceptable).

Q: How is shaping used in feeding therapy, and when is it not enough?

In feeding therapy, shaping uses a food acceptance hierarchy as the approximation sequence: reinforce the closest the learner currently comes to accepting the target food, then incrementally advance the criterion toward full acceptance (looking → proximity → touching → smelling → licking → biting → chewing → swallowing). The Demchuk et al. study shows that for children with autism and ARFID, adding shaping steps to a pairing intervention produced four-food acceptance for one participant without aversive procedures (Demchuk et al. (2026)). For the other participant, shaping was insufficient, and nonremoval of the spoon was required — indicating that shaping alone may not be the terminal procedure for the most severe feeding presentations, particularly when conditioned emotional responses to food are strong.

Q: How do reinforcement schedules interact with shaping progression?

At each new criterion step, begin on CRF to maximize acquisition rate and minimize reinforcement gaps. Once the approximation is stable at CRF, thin the schedule before advancing the criterion — but do not thin and advance simultaneously. K-schedules provide a data-driven alternative: they automatically track the learner's percentile performance and adjust reinforcement density proportionally, preventing both lean schedules at low performance levels and artificially dense schedules as performance approaches the aim (Milyko (2021)). The core principle remains: the schedule during shaping should keep the reinforcement density high enough at each step that reinforcement gaps are rare, because gaps are the primary condition under which accidental reinforcement of competing approximations occurs (Hunter & Rosales‐Ruiz (2019)).

Q: How does shaping apply to organizational and staff performance contexts?

In OBM contexts, shaping applies the same four-step procedure to staff behavior: define the terminal performance, identify what the staff member currently does that resembles it, set incrementally higher criteria, and deliver differential reinforcement (feedback, praise, recognition, tangible rewards) contingent on meeting each criterion. The LaMarca and LaMarca ADDIE framework models this explicitly, noting that practitioner skill development in comprehensive ABA program design should proceed as a shaping trajectory rather than an all-at-once overhaul ((LaMarca et al., 2024)). Flynn and Wilder's feedback-reception shaping study demonstrates that even meta-professional behaviors — how staff receive corrective feedback from supervisors — are amenable to shaping through task clarification and contingent performance feedback (5b68158d-ac6a-4937-809b-e4b5fc9d7f1f).

• Ghaemmaghami, M., Hanley, G. P., Jessel, J., & Landa, R. (2018). Shaping complex functional communication responses. Journal of Applied Behavior Analysis, 51(3), 502–520. https://doi.org/10.1002/jaba.468

• Hunter, M., & Rosales-Ruiz, J. (2019). The power of one reinforcer: The effect of a single reinforcer in the context of shaping. Journal of the Experimental Analysis of Behavior, 111(3), 449–464. https://doi.org/10.1002/jeab.517

• Milyko, K. (2021). K-Schedules Meet Precision Measurement: A Protocol for Intervention. Behavior Analysis in Practice, 14(3), 582–594. https://doi.org/10.1007/s40617-020-00459-2

• Demchuk, N. C., Ibarra, A., & Hansen, B. A. (2026). A Stepwise Evaluation of Pairing and Shaping for Children With Autism and Feeding Difficulties. Behavioral Interventions, 41(1). https://doi.org/10.1002/bin.70064

• Fonger, A. M., & Malott, R. W. (2019). Using Shaping to Teach Eye Contact to Children with Autism Spectrum Disorder. Behavior Analysis in Practice, 12(1), 216–221. https://doi.org/10.1007/s40617-018-0245-9

• Waite, M. R., Kodak, T. M., & Whang, A. J. (2025). Development and validation of a caretaker-implemented ear cleaning teaching protocol for companion dogs. Behavior Analysis in Practice. https://doi.org/10.1007/s40617-025-01135-z

• Cihon, J. H., Ferguson, J. L., Leaf, J. B., Leaf, R., McEachin, J., & Taubman, M. (2019). Use of a Level System with Flexible Shaping to Improve Synchronous Engagement. Behavior Analysis in Practice, 12(1), 44–51. https://doi.org/10.1007/s40617-018-0254-8

• da Silva, S. P., & Williams, A. M. (2020). Translations in Stimulus-Stimulus Pairing: Autoshaping of Learner Vocalizations. Perspectives on Behavior Science, 43(1), 57–103. https://doi.org/10.1007/s40614-019-00228-9

• Barry, D., Neufeld, J., & Stewart, I. (2024). Teaching nonarbitrary temporal relational responding in adolescents with autism. The Analysis of Verbal Behavior, 40, 135–152. https://doi.org/10.1007/s40616-024-00210-w

• Frampton, S. E., Axe, J. B., Davis, C. R., Meleshkevich, O., & Li, M.-H. (2024). A tutorial on indicating responses and their importance in mand training. Behavior Analysis in Practice, 17, 1238–1249. https://doi.org/10.1007/s40617-024-00965-7

• Frampton, S. E., Davis, C. R., Meleshkevich, O., & Axe, J. B. (2024). A clinical tutorial on methods to capture and contrive establishing operations to teach mands. Behavior Analysis in Practice, 17, 1270–1282. https://doi.org/10.1007/s40617-024-00985-3

• Linnehan, A. M. (2025). Teaching autistic adolescents to identify fear and anger: a preliminary study. Behavior Analysis in Practice. https://doi.org/10.1007/s40617-025-01129-x

• O'Connor, M. T., Belisle, J., Stanley, C. R., & Dixon, M. R. (2020). Establishing Multiple-Control Responding of Children with Autism to People and Emotions in Context by Utilizing Derived Stimulus Relations. Behavior Analysis in Practice, 13(1), 192–196. https://doi.org/10.1007/s40617-019-00339-4

• 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

• Yarzebski, V., & Dickson, C. (2024). Teaching caregivers to use graduated guidance using video modeling. Behavior Analysis in Practice, 17, 1198–1203. https://doi.org/10.1007/s40617-024-00969-3

• Diller, J. W., & Li, A. (2025). Incorporating qualitative data when training behavior analysts. Behavior Analysis in Practice. https://doi.org/10.1007/s40617-025-01124-2

• 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

• 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

• Flynn, K. E., & Wilder, D. A. (2025). An evaluation of task clarification and feedback to teach feedback reception skills. Journal of Organizational Behavior Management. https://doi.org/10.1080/01608061.2025.2468190

• Townsend, D. B., Brothers, K. J., MacDuff, G. S., et al. (2024). Alliance for Scientific Autism Intervention: System components and outcome data from high-quality service delivery organizations. Behavior Analysis in Practice, 17, 565–580. https://doi.org/10.1007/s40617-023-00898-7

• Torelli, J. N., Snyder, S. K., Griffin, M. L., et al. (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

• Neef, N. A., Kranak, M., Shapiro, M., Xu, Z., & Catania, A. C. (2024). Changing What Teachers Say Changes What They Do: Increasing Their Specific Praise of Student Behavior. Journal of Behavioral Education, 33(2), 344–357. https://doi.org/10.1007/s10864-022-09488-z