Chaining: Forward, Backward, and Total Task Teaching for BCBAs and RBTs

The behavior chain as a unit of analysis

A behavior chain is a sequence of individual responses in which each response produces a stimulus change that functions simultaneously as a conditioned reinforcer for the previous response and as a discriminative stimulus (SD) for the next response. The terminal response produces the natural or programmed reinforcer that maintains the entire chain. This architecture has one critical practical implication: for chaining to work, the chain must be taught to mastery before it is used as the context for anything else — for example, the interrupted-chain procedure relies on a well-established chain so that a missing link functions as a motivating operation rather than simply as an obstacle (Frampton et al., 2024). If a learner routinely skips steps or substitutes alternatives, those stimulus changes do not reliably function as conditioned reinforcers or SDs, and the chain is not genuinely established.

Task analysis is the prerequisite for any chaining format. A defensible task analysis identifies every response in the chain, specifies the precise criterion for each step, and names the SD that signals the start of that step. Step size is a clinical judgment: steps that are too large produce errors early and slow acquisition, while steps that are appropriately small allow for errorless progress through each link. Research on forward chaining with caretaker-implemented protocols demonstrates that practitioners autonomously subdivide steps when pre-specified increments produce errors, suggesting that ongoing monitoring of step-level data is not just best practice but a natural implementation behavior when data systems are visible (Waite et al., 2025).

Forward chaining

In forward chaining, the practitioner teaches Step 1 first, reinforcing correct independent completion of that step, while completing all remaining steps for the learner (or providing physical guidance through them). Once Step 1 meets mastery criterion, Steps 1 and 2 are targeted together, and so on until the learner completes all steps independently. The learner always begins the chain from its natural starting point, which builds the early discriminative stimulus control that initiates the routine in natural environments.

Head-to-head research comparing forward and backward chaining provides nuanced guidance. Moore and Quintero (2019) used an alternating-treatments design to compare the two approaches when teaching Olympic weightlifting movements to novice lifters; participants showed more accurate performance on lifts taught with forward chaining than with backward chaining, a finding the authors attribute to the natural build-up of momentum and stimulus control from the chain's initiation point Moore & Quintero (2019). Similar logic applies to vocational and daily-living chains where the early SD (e.g., arriving at a workstation, entering the kitchen) is already a strong and reliable signal in the learner's environment.

Forward chaining is also well-supported when steps in the task analysis are tightly interdependent and errors at early links contaminate later ones. In a caretaker-implemented forward-chaining protocol for cooperative ear-cleaning in companion dogs, practitioners successfully built an 18-step chain using micro-step sizing; error-based retreating (moving back several steps rather than one) and choice-based consent at each step were built into the design, and all four learner-caretaker pairs reached mastery (Waite et al., 2025). While the population differs from human learners, the task-analytic and chaining mechanics are directly translatable and underscore the importance of fine-grained step calibration.

Backward chaining

In backward chaining, the practitioner completes all steps except the last, prompts or guides the learner through the terminal step, and delivers the natural reinforcer immediately after independent completion. Once the terminal step meets criterion, the second-to-last step is added, and teaching works backward through the chain until the learner initiates from Step 1. The core theoretical rationale is that the natural reinforcer is always immediate upon correct completion, which maximizes the reinforcement value at the exact step being learned.

Rubio and colleagues (2018) demonstrated backward chaining with a physical guidance delay to teach self-feeding of solids to a 4-year-old boy who had no prior self-feeding history Rubio et al. (2018). Using a multiple-probe across-steps design, mastery emerged first at the terminal step (placing food in the mouth) and moved progressively forward — precisely the pattern backward chaining predicts. A 3-second delay before physical guidance was sufficient to allow independent responding without creating prompt dependence Rubio et al. (2018).

For vocational tasks with adults with developmental disabilities, Kobylarz and colleagues (2020) conducted one of the most systematic procedural comparisons in the chaining literature: a multiple-probe across-behaviors design tested four variants of backward chaining — teacher-completion, participant-completion, no-completion, and a control Kobylarz et al. (2020). All three active conditions established vocational skills effectively. However, participant-completion — in which the instructor delivered least-to-most prompting for every untrained step so the learner actively rehearsed the full chain from the start — produced the fastest acquisition, highest maintenance, and was unanimously preferred by all three participants Kobylarz et al. (2020). The practical implication is direct: in backward chaining, having the learner actively participate in completing (with prompting) the steps not yet being formally taught appears to accelerate acquisition beyond passive teacher-completion variants.

Lambert and colleagues (2016) extended backward chaining to a complex 84-step athletic sequence — basketball-playing — with a 13-year-old boy with autism Lambert et al. (2016). The protocol subdivided the chain into eight sub-chains of 11–14 steps, taught each sub-chain with backward chaining plus conditional discrimination training (responding to coach/peer signals), then linked the sub-chains. The learner generalized the full chain to the school gymnasium and maintained performance at a 3-month follow-up Lambert et al. (2016). This case illustrates that backward chaining scales to long, complex behavioral sequences when a hierarchical architecture is used — teach sub-chains first, then link them into a master chain.

Backward chaining with leap-aheads

Backward chaining with leap-aheads is a modification designed for learners with severe disabilities or for situations where standard step-by-step backward progression is too slow. After the terminal step reaches criterion, rather than proceeding to the immediately preceding step, the practitioner skips ahead to a step several links earlier that the learner may already have in their repertoire or can acquire quickly. Steps the learner has not yet mastered are handled with least-to-most prompting during the intermediate steps. The result is faster movement through the chain without sacrificing the reinforcement-proximity advantage at the terminal link.

Spooner, Spooner, and Ulicny (1986) documented the original comparison of backward chaining with leap-aheads versus standard backward chaining and reverse chaining with leap-aheads. The leap-ahead variant showed efficiency advantages, particularly for learners who had isolated steps already in their repertoire but had not yet connected them into a chain Lambert et al. (2016). The mechanism is that learners with existing component skills do not need repeated trials on steps they already emit independently; leap-aheads allow those competencies to be credited and the instructional focus to fall on the genuinely missing links Kobylarz et al. (2020).

The practical decision rule is straightforward: if skills assessment reveals that a learner already performs several steps in isolation but cannot chain them, backward chaining with leap-aheads places instruction only on the steps where genuine deficits exist. This is especially relevant for learners with profound or multiple disabilities, where session time is limited and efficiency must be built into every procedural choice.

Total task chaining

Total task chaining presents the learner with the complete sequence on every trial Ozen et al. (2022). The instructor uses a prompt hierarchy (typically least-to-most or graduated guidance) to assist at each step where the learner is not yet independent, and delivers reinforcement at the terminal step or contingently throughout depending on design. There is no separate teaching phase for individual steps — the learner practices the full chain from the first session and receives assistance only where needed Ozen et al. (2022).

Total task is typically the format of choice when the learner already has some component steps in their repertoire, when the chain is relatively short (roughly 8–12 steps or fewer), or when the goal is rapid integration of component skills into fluent performance. The embedded teaching trials format — presenting chained prompting instruction within naturalistic activities rather than massed trials — has been validated as equivalent to massed formats for establishing dressing chains (snap-fastening and buttoning) with preschoolers with developmental disabilities Ozen et al. (2022). Ozen and colleagues (2022) showed that both simultaneous prompting and graduated guidance delivered within play-embedded total task trials produced comparable acquisition, generalization across materials, and maintenance — without requiring separate discrete-trial sessions Ozen et al. (2022).

When a chain is long and step acquisition is uneven, total task can mask which specific steps are creating bottlenecks unless step-by-step data are collected on every trial Ozen et al. (2022). Collecting step-level data on every session is the solution; without it, total task produces a global measure of chain completion that cannot guide targeted intervention when the chain stalls.

Progressive task analysis

A progressive task analysis is a variation of total task chaining used when the chain itself must be built incrementally because the full chain is too long or complex to introduce at once. Rather than beginning with all steps on Trial 1, the practitioner establishes a short initial chain (e.g., Steps 1–3), brings those steps to criterion, then extends the chain by adding the next block of steps Lambert et al. (2016). The chain grows forward in manageable blocks.

Progressive task analysis is most applicable when the natural environment does not afford repeated practice of the full chain, when the terminal reinforcer is not accessible until the full chain is complete, or when cognitive load or fatigue constrains the number of steps a learner can actively process in a single trial Kobylarz et al. (2020). Longitudinal data collection that tracks block-by-block mastery, rather than global chain-completion scores, is essential for determining when to extend to the next block.

Probe protocols: single-opportunity versus multiple-opportunity

A probe trial assesses which steps the learner can complete independently, without instructional prompting Lambert et al. (2016). Probes drive the training decision — specifically, which steps to target next, and whether mastery criteria have been met.

Single-opportunity probe: The session begins at Step 1. If the learner fails any step (or requires a prompt), the session is ended and no further steps are assessed. The advantage is speed and low-intrusion; the disadvantage is that steps occurring after the first error cannot be assessed, which underestimates independent performance for learners who have strong terminal steps but weak early steps. Single-opportunity probes are most efficient during active teaching phases.

Multiple-opportunity probe: When the learner fails a step, the instructor completes the step for them (or resets the environment) and then asks the learner to attempt the next step. All steps receive an independent assessment opportunity. The advantage is a complete picture of the learner's step-by-step repertoire across the full chain; the disadvantage is longer session time and the possibility of inadvertent prompting through the instructor's step-completion model. Multiple-opportunity probes are indicated for initial baseline assessment, when making decisions about which format to use, and for periodic reassessment Kobylarz et al. (2020).

A critical practical note: the multiple-probe design (discrete from the multiple-opportunity probe) refers to the single-subject experimental design used to evaluate chaining across steps or behaviors — not a session-level measurement format. When practitioners read the chaining literature, they should not confuse the measurement protocol (single vs. multiple-opportunity) with the experimental design (multiple-probe vs. multiple-baseline) Lambert et al. (2016).

Interrupted-chain procedure and behavior chain interruption strategy

Mastered behavior chains have a secondary application beyond the chain itself: they serve as contexts for teaching new verbal behavior. The interrupted-chain procedure (ICP) deliberately removes or breaks an item needed at a specific chain link, creating a temporary establishing operation that increases the value of the missing item and reliably evokes a mand for it (Frampton et al., 2024). The chain must already be firmly established — each step reliably functioning as a conditioned reinforcer for the preceding step and as an SD for the following one — before the interruption will function as an effective motivating operation. If the chain is not mastered, the interruption will not reliably evoke communication about the missing item.

Frampton and colleagues (2024) provided a comprehensive clinical tutorial on capturing and contriving establishing operations using the ICP (Frampton et al., 2024). Their guidance includes varying which item is missing across trials (e.g., bowl versus water in a soup-making chain) so the learner does not simply mand "bowl" on every trial regardless of what is missing (Frampton et al., 2024). Chains used in the ICP should naturally produce potent reinforcers — food, preferred toys, movies — so that the interrupted consequence remains valuable enough to evoke communication.

Jessel and Ingvarsson (2022) demonstrated the ICP with two boys aged 4–5 years with autism, embedding mand training for both known and unknown items within incomplete toy-assembly chains Jessel & Ingvarsson (2022). Participants successfully manded for known missing items and spontaneously generalized autoclitic mand frames ("What is it?") to obtain names of unknown items, without additional teaching for each new item. The chain context provided the motivating operation that made each new instance of manding both necessary and functional Jessel & Ingvarsson (2022).

Thompson and Hanson (2024) extended the behavior chain interruption strategy (BCIS) to a 65-year-old adult with severe intellectual disability and deaf-blindness Thompson & Hanson (2024). Using mastered self-care routines (toileting, dressing, eating), they contrived EO trials (items missing or inoperable) alternated with AO trials (intact chain) in a multielement design Thompson & Hanson (2024). The participant independently used a paging device to request help on EO trials and never used the device on AO trials — exactly the pattern indicating the response was under mand control rather than SD control Thompson & Hanson (2024).

Chained schedules of reinforcement and their clinical applications

Distinct from chaining as a skill-acquisition procedure, chained schedules of reinforcement arrange contingencies in a sequence: completing the requirements of a terminal link earns access to a specific reinforcer, and a preceding link is required to open access to the terminal link. Research on chained schedules has produced clinically relevant findings for managing stereotypy and problem behavior.

Slaton and Hanley (2016) compared chained and multiple schedules for two children with autism who exhibited automatically reinforced stereotypy Slaton & Hanley (2016). In the chained schedule, access to stereotypy was contingent on prior functional item engagement; in the multiple schedule, stereotypy access was time-based. The chained schedule produced less stereotypy and more consistent item engagement, established stimulus control over the stereotypy, and was preferred by both participants in a concurrent-chains assessment Slaton & Hanley (2016).

Sloman and colleagues (2022) systematically replicated and extended this finding with two children with autism exhibiting vocal stereotypy Sloman et al. (2022). Embedding response interruption and redirection (RIRD) within chained schedules produced greater vocal stereotypy reduction for one participant than multiple-schedule RIRD; for the second participant, both were effective. A component analysis showed no single element alone accounted for the reduction — it was the contingent chained structure that mattered Sloman et al. (2022).

Livingston and colleagues (2023) used a break-to-choice chained schedule to treat multiply maintained problem behavior in three boys with autism Livingston et al. (2023). Rather than developing separate treatment conditions for each maintaining function, the chained schedule allowed one functional communication response (requesting a break) to serve as the initial link, with access to a reinforcer menu as the terminal link. Problem behavior decreased for all three participants without training separate communication topographies for each function identified in the functional analysis Livingston et al. (2023).

Self-instructional chaining and verbal rehearsal

Research on joint control and covert verbal behavior provides a mechanistic account of how learners navigate multi-step chains under reduced prompting Clough et al. (2016). Clough, Meyer, and Miguel (2016) demonstrated that accurate sequencing of novel stimulus sets depended critically on verbal rehearsal — when participants were blocked from rehearsing (by singing or tapping), sequencing accuracy dropped substantially Clough et al. (2016). Verbal rehearsal of step names or relational statements (e.g., "shoes after socks") functioned as precurrent behavior that produced the discriminative stimuli controlling each subsequent response in the chain Clough et al. (2016).

Barry, Neufeld, and Stewart (2024) extended this line to adolescents with autism, showing that relational statements specifying temporal order of actions (e.g., "red after blue, blue after yellow") could govern multi-step action sequences without explicit response-by-response prompting (Barry et al., 2024). Participants who learned the relational statements could execute the chains controlled solely by the verbal rules, suggesting that teaching temporal relational responding may scaffold self-instructional chaining for learners who cannot yet generalize step sequences from physical prompts alone (Barry et al., 2024).

The practical translation: when a learner completes a chain correctly during prompted trials but consistently fails during probe trials, teaching explicit self-instruction (verbal rules about step order) or joint control (echoic-then-tact sequences for each step) may bridge the gap between prompted and independent chain performance Clough et al. (2016) (Barry et al., 2024).

Generalization planning for chained behavior

Chains learned in one context show only modest untrained generalization to novel settings, materials, or trainers. Practitioners should build generalization into chaining programs from the outset rather than addressing it as an afterthought.

Lambert and colleagues (2016) reported that an 84-step basketball chain trained in one setting generalized to the school gym after a single structured generalization probe Lambert et al. (2016). Ozen and colleagues (2022) found that embedded total task chaining instruction produced maintenance and generalization across novel materials and settings without additional training trials Ozen et al. (2022). Both outcomes reflect deliberate generalization programming: multiple trainers, multiple materials, and naturalistic activity contexts were built into the training protocol rather than added at the end.

Specific strategies supported by the literature include: (a) conducting training trials across settings from early in the teaching phase, (b) using natural stimuli as SDs rather than clinician-generated cues when possible, (c) ensuring the terminal reinforcer is one that occurs naturally in the generalization environment, and (d) fading trainer presence systematically rather than abruptly removing the instructor at criterion Lambert et al. (2016) Ozen et al. (2022).

McCammon, Wolfe, and Check (2024) reviewed 119 mand training studies and found that only 21% explicitly manipulated or analyzed motivating operations during chain-interruption procedures (McCammon et al., 2024). The same gap applies to generalization programming in chaining studies: establishing operations and discriminative stimuli that govern the chain must be present in generalization environments or the chain will not transfer (Frampton et al., 2024). Practitioners who build skill-acquisition programs on chaining should explicitly document the SDs and reinforcers expected in target settings and verify their presence before fading support Ozen et al. (2022).

Behavioral skills training of mediators

Parents, teachers, paraprofessionals, and other direct-care staff frequently implement chaining procedures on behalf of clinicians. The quality of that implementation directly determines whether the chain is established, maintained, and generalized. Behavioral skills training (BST) — instruction, modeling, rehearsal, and performance feedback — is the evidence-based format for training mediators in behavioral procedures (Waite et al., 2025). The standard BST architecture applies directly: instruct on the task analysis, model the procedure, rehearse with feedback, and verify integrity before the mediator conducts independent sessions.

For chaining specifically, the mediator must be able to: complete a task analysis correctly, distinguish independent responses from prompted ones, deliver consequences contingently at the step level rather than the trial level, collect step-by-step data, and make data-based decisions about when a step has met mastery criterion (Waite et al., 2025). Each of these competencies can be addressed in BST using the chaining task itself as the teaching context — the trainer models the procedure with the actual task analysis, observes the mediator performing it, provides corrective feedback at the step level, and repeats until procedural integrity meets standard (typically ≥80% across three consecutive opportunities) Kobylarz et al. (2020).

The chaining literature sits primarily at the single-subject experimental design (SCED) level, with several methodologically strong designs: multiple-probe across behaviors or steps, multiple-baseline across participants, and alternating or adapted alternating treatments. No large randomized trials have compared chaining formats, and the procedural-variant literature has focused mostly on backward chaining variants Kobylarz et al. (2020).

Head-to-head comparisons. Moore and Quintero (2019) compared forward and backward chaining in an alternating-treatments design with novice weightlifters, showing a forward chaining advantage for movement accuracy Moore & Quintero (2019). Kobylarz and colleagues (2020) compared four procedural variants of backward chaining in adults with developmental disabilities, finding participant-completion most efficient across acquisition, maintenance, and preference Kobylarz et al. (2020). Ozen and colleagues (2022) compared simultaneous prompting and graduated guidance within total task trials for dressing chains, finding equivalent outcomes Ozen et al. (2022).

Single-case demonstrations. Lambert and colleagues (2016) documented an 84-step backward chaining program with generalization and 3-month maintenance Lambert et al. (2016). Rubio and colleagues (2018) showed backward chaining with a physical guidance delay establishing a 10-step self-feeding chain in a child with no prior self-feeding Rubio et al. (2018). Thompson and Hanson (2024) showed BCIS using mastered chains to establish help-seeking mands in an adult with dual sensory loss and severe intellectual disability Thompson & Hanson (2024). Jessel and Ingvarsson (2022) showed ICP establishing autoclitic mand frames without step-by-step teaching for each new item Jessel & Ingvarsson (2022).

Chained schedule research. Slaton and Hanley (2016), Sloman and colleagues (2022), and Livingston and colleagues (2023) provide SCED evidence that chained reinforcement schedules (contingent access to a terminal-link reinforcer) produce better stimulus control and behavior reduction than time-based multiple schedules for certain populations Slaton & Hanley (2016) Sloman et al. (2022) Livingston et al. (2023).

Verbal behavior and self-instruction. Clough and colleagues (2016) and Barry and colleagues (2024) provide mechanistic evidence that covert verbal rehearsal and temporal relational responding govern multi-step chain performance Clough et al. (2016) (Barry et al., 2024).

Systematic reviews and tutorials. Frampton and colleagues (2024) provide a comprehensive tutorial on ICP design and EO manipulation (Frampton et al., 2024). McCammon and colleagues (2024) reviewed 119 mand training studies and identified that EO manipulation in chain-interruption contexts was systematically under-described, limiting analytic precision (McCammon et al., 2024).

Overall weight. The evidence strongly supports chaining as an effective and efficient skill-acquisition method across populations, settings, and skill domains Kobylarz et al. (2020) Moore & Quintero (2019). The comparative evidence is sufficient to inform format selection decisions, though no head-to-head trial has compared all three formats in the same study with the same learner profile Ozen et al. (2022). Evidence for self-instructional and relational variants is emerging and promising but requires broader replication Clough et al. (2016) (Barry et al., 2024).

Which format to use

The choice among forward, backward, and total task chaining is driven by three factors: the learner's existing step-level repertoire, the structure of the chain's natural reinforcer, and the cognitive and physical demands of the full chain during early trials Kobylarz et al. (2020) Moore & Quintero (2019).

1. Learner has no component steps. Start with backward chaining — the learner always contacts the natural reinforcer immediately after their response, maximizing reinforcement value during acquisition of each new step.

2. Learner has some component steps but not others. Use backward chaining with leap-aheads to credit existing steps and focus instruction only on missing links. This reduces unnecessary repetition and accelerates progress.

3. Learner has most steps but cannot integrate them. Use total task chaining with graduated prompting across all steps. Distributed practice of the full chain with decreasing assistance is faster than re-teaching individual steps in isolation.

4. Chain involves complex interdependent early steps. Use forward chaining — early steps establish the discriminative stimulus control that initiates the routine in natural settings, and accurate early-step performance reduces error-propagation through later steps Moore & Quintero (2019).

5. Chain is long (more than 15 steps) or requires complex conditional discriminations. Break the chain into sub-chains of manageable length, teach each sub-chain with the appropriate format (typically backward chaining), then link sub-chains together Lambert et al. (2016).

6. Goal includes mand training via chain interruption. Establish the full chain to mastery first using any format; only then introduce ICP interruptions. Interrupting an unmastered chain does not reliably evoke mands (Frampton et al., 2024).

7. Multiple functions of problem behavior are complicating skill instruction. Assess whether a break-to-choice chained schedule — where one communication response accesses a reinforcer menu — can address multiple maintaining functions before introducing complex chaining instruction Livingston et al. (2023).

8. Stereotypy or problem behavior is competing with chain performance. Evaluate whether a chained schedule (task engagement → access to stereotypy) can be arranged before teaching the chain, to establish task engagement as the initial link that unlocks a preferred terminal reinforcer Slaton & Hanley (2016).

Probe protocol selection

Use multiple-opportunity probes at initial baseline to get a full step-level picture. Transition to single-opportunity probes during active teaching to reduce session duration. Return to multiple-opportunity probes at any point when a learner's step-level data become difficult to interpret due to variability across steps.

Mediator training

Use BST — not lecture alone — to train any mediator who will implement chaining procedures. Verify procedural integrity at ≥80% before the mediator conducts sessions independently. Schedule fidelity re-checks during ongoing supervision, not only at initial training.

Clinic and center-based skill acquisition

Center-based programs most commonly use total task or forward chaining for daily-living and language skill programs, and backward chaining for motor, self-care, and grooming chains where the terminal step (e.g., completing a fastener, placing food in mouth) is a strong natural reinforcer. The embedded teaching trials format — practicing chains within naturalistic play or activity contexts rather than in massed discrete-trial sessions — has been validated for dressing skills with preschoolers with developmental disabilities and produced maintenance without additional sessions Ozen et al. (2022). For chains used to teach manding (via ICP), clinic settings afford the controlled arrangement of EO and AO trials and careful alternation of which item is missing to build discrimination (Frampton et al., 2024).

School and inclusive classroom

School teams apply chaining most commonly to self-care, transition, and vocational readiness skills for students with intellectual disabilities or autism. Total task with least-to-most prompting embedded in naturally occurring classroom transitions is the most feasible format given typical classroom logistics. The literature supports that step-by-step data collection is feasible in classroom settings when data sheets are well-designed and staff are trained to collect data without interrupting instruction Ozen et al. (2022). For complex academic sequences or athletic skill chains, backward chaining with sub-chain architecture is supported by at least one peer-reviewed generalization demonstration in a school gymnasium Lambert et al. (2016).

Vocational and community settings

The strongest evidence for vocational chaining comes from Kobylarz and colleagues (2020), whose comparison of backward chaining variants in adults with developmental disabilities was conducted in a community-based day-training and residential facility with actual job tasks Kobylarz et al. (2020). Participant-completion backward chaining — where the learner actively rehearses all steps with least-to-most prompting from Day 1 — produced superior acquisition and maintenance. For vocational settings where staff time is limited, participant-completion also reduces the trainer's completion burden compared to teacher-completion, because the learner begins practicing every step from the first session.

Home and parent training

Parents and family members can implement chaining procedures effectively when trained via BST on the specific task analysis their child is learning. The caretaker-implemented forward chaining protocol demonstrated by Waite and colleagues (2025) illustrates that non-specialist mediators can implement chaining with fidelity when given a scripted protocol with micro-steps and a clear decision rule for when to retreat (back several steps, not one, after an error) (Waite et al., 2025). For self-feeding and self-care chains with young children, backward chaining is particularly well-suited to parent-implemented instruction because the natural reinforcer at the terminal step (completing the meal, finishing dressing) is immediately meaningful to both child and parent Rubio et al. (2018).

Residential and adult disability services

For adults with intellectual or developmental disabilities in residential settings, vocational task chaining with participant-completion procedures and active prompting is supported by the Kobylarz and colleagues (2020) vocational study Kobylarz et al. (2020). Adults with severe or complex disabilities — including those with dual sensory impairments — can be reached through the BCIS applied to mastered self-care chains, where the chain context provides the motivating operation for communication training without requiring direct mand instruction in a stripped context Thompson & Hanson (2024).

• Task analysis with steps that are too large. Steps that span multiple behaviors require the learner to solve sub-problems the task analysis has not accounted for. When a learner stalls at a specific step repeatedly, the correct intervention is to break that step into smaller approximations, not simply to increase prompting level (Waite et al., 2025).

• Ambiguous step descriptions. A task analysis entry like "wash hands" is not a step — it's another chain. Every step should specify the observable response criterion so two trainers independently reading the task analysis would identify the same response as correct or incorrect Kobylarz et al. (2020).

• No probe protocol specified. Without a standardized probe protocol, "probe data" become impressionistic. Specify single-opportunity or multiple-opportunity before the program begins, and collect data on every probe session with the same format Lambert et al. (2016).

• Prompt-level confusion across steps. Using different prompt types (gestural for one step, physical for another) without a documented rationale creates inconsistency across trainers Ozen et al. (2022). A single prompt hierarchy applied uniformly across all chain steps is the default; documented exceptions should be justified by step-specific data.

• Starting ICP too early. Interrupting a chain that is not yet mastered does not evoke a mand — it evokes a pause, a confused look, or problem behavior. The chain must be established to mastery first (Frampton et al., 2024).

• Using teacher-completion backward chaining when participant-completion is feasible. Research shows participant-completion — active prompting through unpracticed steps — produces faster acquisition and higher preference in adults with developmental disabilities Kobylarz et al. (2020). Teacher-completion should be a deliberate choice for learners who cannot tolerate physical guidance, not a default.

• Failing to plan generalization from the start. Chains trained exclusively in clinic or one-to-one settings with a single trainer do not automatically transfer to naturalistic environments Ozen et al. (2022). Build multiple trainers, multiple materials, and natural SD exposure into the teaching protocol from the first training session.

• Collecting only chain-completion scores. Global completion scores (e.g., "8/10 steps correct") obscure which steps are causing failures Ozen et al. (2022). Step-level data on every trial are required for good clinical decisions, especially in total task chaining where errors can occur at any step.

• Skipping step-level data under chained schedule programs. When using chained reinforcement schedules for behavior management, ensure that the initial-link response (e.g., completing a task chain) has a well-defined task analysis so criterion for terminal-link access is clear and applied consistently by all staff Slaton & Hanley (2016).

1. Build the task analysis before choosing a chaining format. Step size determines everything; steps that are too large produce errors that slow the chain and require remediation.

2. Use backward chaining as the default for learners with no component steps. Immediate natural reinforcement at the terminal link maximizes reinforcement value during early acquisition.

3. Switch to backward chaining with leap-aheads when skills assessment shows the learner already has isolated steps in their repertoire. Credit existing competencies; teach only the missing links Kobylarz et al. (2020).

4. Use total task chaining when the learner has most component steps and the goal is rapid integration. Collect step-level data every session to identify bottleneck steps Ozen et al. (2022).

5. Use forward chaining when accurate early-step performance is critical to later steps. Evidence from Olympic weightlifting shows forward chaining produced more accurate performance than backward chaining for multi-step movements where early link accuracy governs downstream form Moore & Quintero (2019).

6. For long chains, divide into sub-chains and teach each sub-chain independently before linking. An 84-step basketball chain was mastered and generalized using this architecture Lambert et al. (2016).

7. For vocational chains with adults with developmental disabilities, use participant-completion backward chaining — it is faster, produces higher maintenance, and is preferred by learners over teacher-completion variants Kobylarz et al. (2020).

8. Use multiple-opportunity probes at baseline and whenever step-level data are ambiguous. Single-opportunity probes are efficient during active teaching once the full baseline is established.

9. Master the chain before using it for ICP mand training. An incompletely mastered chain will not function as an effective motivating operation when interrupted (Frampton et al., 2024).

10. Vary which item is missing in ICP sessions across trials to build mand discrimination and prevent rote responding under SD control rather than EO control (Frampton et al., 2024).

11. Consider chained schedules of reinforcement for learners with stereotypy or multiply maintained problem behavior. Contingent access to preferred reinforcers after task engagement produces stronger stimulus control than time-based access alone Slaton & Hanley (2016) Livingston et al. (2023).

12. Teach verbal step-by-step self-instruction alongside chain instruction for learners who struggle with independence. Verbal rehearsal of step order has been shown to govern accurate sequencing in the absence of physical prompts Clough et al. (2016).

13. Train mediators with BST on the specific task analysis being taught (Waite et al., 2025). Procedural integrity at the step level — not just completion of the session — is the standard.

14. Build generalization trials into the teaching protocol from Session 1 Ozen et al. (2022). Novel trainers, novel settings, and natural SDs should appear early, not only after criterion in the training context.

15. Conduct periodic fidelity checks of mediator prompt delivery, data collection, and mastery-criterion application throughout the program — not only at initial training (Waite et al., 2025).

Primary research cited in this article.

• 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 https://doi.org/10.1007/s40617-024-00985-3
• Kobylarz, A. M., DeBar, R. M., Reeve, K. F., & Meyer, L. S. (2020). Evaluating backward chaining methods on vocational tasks by adults with developmental disabilities. Behavioral Interventions, 35(2), 263-280. https://doi.org/10.1002/bin.1713 https://doi.org/10.1002/bin.1713
• Lambert, J. M., Copeland, B. A., Karp, E. L., Finley, C. I., Houchins-Juarez, N. J., & Ledford, J. R. (2016). Chaining Functional Basketball Sequences (with Embedded Conditional Discriminations) in an Adolescent with Autism. Behavior Analysis in Practice, 9(3), 199-210. https://doi.org/10.1007/s40617-016-0125-0 https://doi.org/10.1007/s40617-016-0125-0
• Slaton, J. D. & Hanley, G. P. (2016). Effects of multiple versus chained schedules on stereotypy and item engagement. Journal of Applied Behavior Analysis, 49(4), 927-946. https://doi.org/10.1002/jaba.345 https://doi.org/10.1002/jaba.345
• Thompson, H. E. & Hanson, R. J. (2024). The Use of a Behavior Chain Interruption Strategy to Teach Mands for Help with an Adult with Intellectual Disability and Deaf-Blindness. The Analysis of Verbal Behavior, 40(1), 88-98. https://doi.org/10.1007/s40616-024-00204-8 https://doi.org/10.1007/s40616-024-00204-8
• 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
• Rubio, E. K., Pichardo, D., & Borrero, C. S. (2018). Using backward chaining and a physical guidance delay to teach self‐feeding. Behavioral Interventions, 33(1), 87-92. https://doi.org/10.1002/bin.1504 https://doi.org/10.1002/bin.1504
• Basile, C. D., Tiger, J. H., & Lillie, M. A. (2021). Comparing paired‐stimulus and multiple‐stimulus concurrent‐chains preference assessments: Consistency, correspondence, and efficiency. Journal of Applied Behavior Analysis, 54(4), 1488-1502. https://doi.org/10.1002/jaba.856 https://doi.org/10.1002/jaba.856
• Livingston, C. P., Melanson, I. J., Martinez, S., & Anderson, H. (2023). Evaluation of a break‐to‐choice chained‐schedule intervention for multiply maintained problem behavior. Behavioral Interventions, 38(3), 836-848. https://doi.org/10.1002/bin.1941 https://doi.org/10.1002/bin.1941
• Sloman, K. N., McGarry, K. M., Kishel, C., & Hawkins, A. (2022). A comparison of RIRD within chained and multiple schedules in the treatment of vocal stereotypy. Journal of Applied Behavior Analysis, 55(2), 584-602. https://doi.org/10.1002/jaba.906 https://doi.org/10.1002/jaba.906
• 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 https://doi.org/10.1007/s40617-025-01135-z
• Moore, J. W. & Quintero, L. M. (2019). Comparing forward and backward chaining in teaching Olympic weightlifting. Journal of Applied Behavior Analysis, 52(1), 50-59. https://doi.org/10.1002/jaba.517 https://doi.org/10.1002/jaba.517
• 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 https://doi.org/10.1007/s40616-024-00210-w
• Jessel, J. & Ingvarsson, E. T. (2022). Teaching two children with autism to mand for known and unknown items using contrived motivating operations. Behavioral Interventions, 37(1), 139-152. https://doi.org/10.1002/bin.1777 https://doi.org/10.1002/bin.1777
• Grace, R. C. & McLean, A. P. (2019). Effects of schedule type and terminal‐link duration on choice in concurrent chains. Journal of the Experimental Analysis of Behavior, 112(2), 192-209. https://doi.org/10.1002/jeab.544 https://doi.org/10.1002/jeab.544
• Davis, C. R. & Axe, J. B. (2021). Analyzing Consequence Variables Within the High-Probability Instructional Sequence for a Child Diagnosed With CHARGE Syndrome. Behavior Analysis in Practice, 14(2), 352-359. https://doi.org/10.1007/s40617-020-00524-w https://doi.org/10.1007/s40617-020-00524-w
• Ozen, A., Genc‐Tosun, D., & Tekin‐Iftar, E. (2022). Response prompting procedures delivered within embedded teaching trials for teaching chained skills. Behavioral Interventions, 37(4), 1096-1117. https://doi.org/10.1002/bin.1887 https://doi.org/10.1002/bin.1887
• 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 https://doi.org/10.1007/s40617-024-00965-7
• Clough, C. W., Meyer, C. S., & Miguel, C. F. (2016). The Effects of Blocking and Joint Control Training on Sequencing Visual Stimuli. The Analysis of Verbal Behavior, 32(2), 242-264. https://doi.org/10.1007/s40616-016-0067-1 https://doi.org/10.1007/s40616-016-0067-1
• Torres‐Viso, M., Strohmeier, C. W., & Zarcone, J. R. (2018). Functional analysis and treatment of problem behavior related to mands for rearrangement. Journal of Applied Behavior Analysis, 51(1), 158-165. https://doi.org/10.1002/jaba.437 https://doi.org/10.1002/jaba.437