Functional Communication Training: A Practitioner's Guide to FCT for BCBAs, RBTs, and School Teams

FCT itself is a single core procedure: teach a functional communication response (FCR) that produces the reinforcer the problem behavior used to produce, then extinguish or thin the old contingency while protecting the new one. The variation in this literature lives downstream of that basic package, in how the schedule thins, which response you teach, and what you layer on top once reinforcement gets leaner.

Two decisions this page assumes rather than works through: that a functional assessment has already identified the reinforcer maintaining the problem behavior, since FCT is the treatment that follows that finding rather than the assessment itself, and that you have settled on a communication modality for the FCR, whether vocal, sign, picture exchange, or a speech-generating device. The study set behind this grade concentrates on what comes after those two decisions, so that is where this page goes deep: schedule thinning, response selection within a chosen modality, and the layers that protect the new contingency as reinforcement leans out. For the assessment that precedes FCT, see the functional analysis material; modality selection is a separate clinical decision this evidence set does not adjudicate.

Fixed-lean (jump-to-terminal) schedule thinning

The traditional thinning path moves through intermediate schedules on the way to a leaner terminal one, FR1 to FR2 to FR3 and so on. An alternative skips the intermediate steps and moves directly to the terminal schedule. Chesbrough et al. (2024) found the jump-to-terminal approach reached the target schedule faster, with no loss in clinical effect, the FCR held and problem behavior stayed low. This is a single comparison, not a settled default, and a severe or dangerous topography still warrants the more conservative gradual path until it replicates.

Demand fading

For escape-maintained behavior, thinning isn't only about reinforcement density, it's often the work requirement itself. Gerow et al. (2020) compared demand fading against a dense reinforcement schedule alone and found demand fading kept problem behavior low while also raising task completion, an outcome a reinforcement-only plan doesn't reach on its own.

Multiple schedules (mult FCT)

A multiple schedule pairs the FCR contingency with a discriminative stimulus, a colored card, for example, that signals when the mand will and won't be reinforced. Fisher et al. (2015) found this arrangement produced rapid transfer of treatment effects to new therapists and settings once the signal was consistently displayed. It's the strongest tool here for generalization, but only if every implementer is trained to show the stimulus the same way.

Lag schedules for response variability

A Lag schedule reinforces a mand only if it differs from the immediately preceding response (Lag 1) or from some number of preceding responses. Adami et al. (2017) found that moving from Lag 0, reinforcing every mand form, to Lag 1 immediately increased varied manding without raising problem behavior. This is the tool for once acquisition is solid but communication has narrowed to a single rigid form.

Response selection and shaping complexity

Two decisions sit under this heading: which response to teach, and how to build it. Matter et al. (2017) found that a brand-new, unpaired topography performed comparably to one already carrying the problem behavior's reinforcement history, and learners preferred the novel one, so clinicians aren't obligated to recycle an existing response. Ghaemmaghami et al. (2018) showed complex, multi-component FCRs can be built in learners with intellectual disability through gradual changing-criterion shaping without provoking resurgence, so complexity is a shaping target, not a reason to simplify the plan up front.

Resurgence mitigation during acquisition

Two tactics target resurgence before it happens. Lambert et al. (2017) extended serial DRA logic to FCT, teaching multiple functionally equivalent mands in sequence, and found this shrank the rebound of problem behavior once reinforcement stopped. Fisher et al. (2018) found brief exposure to the establishing operation before starting FCT trials reduced extinction bursts during acquisition.

Augmenting the reinforcement contingency

Several additive layers each solve a distinct problem during thinning rather than replace the core package. A brief competing-activities assessment before a thinning step identifies the highest-engagement item to embed during no-reinforcer periods and reduces resurgence risk (Fuhrman et al., 2018). Stocking the wait interval itself with moderately preferred tangibles, attention, or easy demands sped up thinning and raised learner preference for the arrangement (Simmons et al., 2022). Where stereotypy is the maintaining reinforcer, pairing FCT with arbitrary noncontingent reinforcement outperformed either component alone (Boyle et al., 2018). And when mand training runs alongside classroom instruction, a Social Story matched to the behavior's function reduced problem behavior and taught the mand more effectively than a non-matched story (Pane et al., 2015).

The decision a BCBA faces with FCT is rarely whether to use it once function is known, the field treats FCT as the default DRA application for problem behavior with a clear social function. The real decisions come after that: which thinning approach protects the gains, which response to teach, and what to layer on when the base package alone isn't holding.

Match the thinning approach to the maintaining function. For escape-maintained behavior, don't thin reinforcement density alone, pair it with demand fading so the work requirement moves with the schedule, which produced both low problem behavior and rising task completion against a reinforcement-only plan (Gerow et al., 2020). For cases where speed matters and the behavior isn't safety-critical, jump-to-terminal thinning is worth piloting over the traditional gradual ladder, it reached the lean schedule faster with no loss of effect in the one comparison available (Chesbrough et al., 2024), but keep the conservative gradual path for severe or dangerous topographies until it replicates.

Reach for a multiple schedule when the priority is fast, reliable transfer across therapists, caregivers, and settings, that's the specific outcome the discriminative-stimulus arrangement produced (Fisher et al., 2015), but only commit to it if every implementer, related service providers included, can be trained to display the signal consistently. An inconsistently applied signal undercuts the mechanism that makes the schedule work.

On response selection, don't assume you have to reuse the mand topography closest to the problem behavior's own history. A novel, unpaired response performed as well and was preferred by learners (Matter et al., 2017), so pick whatever's easiest to teach and maintain. Once the FCR is stable and narrowed to a single rigid form, add a Lag schedule to build variability rather than accepting a one-note request as finished (Adami et al., 2017). For learners with intellectual disability who need a more complex, multi-word FCR, build it through gradual criterion shifts instead of expecting the full form on day one (Ghaemmaghami et al., 2018).

Build resurgence protection into the plan before you thin, not after a relapse. Teaching at least two functionally equivalent mands in sequence shrinks the rebound once reinforcement is withheld (Lambert et al., 2017), and a brief exposure to the establishing operation before early trials reduces extinction bursts during acquisition (Fisher et al., 2018). Layer a competing-activities assessment before any thinning step regardless of function, it's a cheap addition that lowers resurgence risk broadly (Fuhrman et al., 2018).

Reach for the remaining additive layers based on what's actually maintaining the case, not as blanket upgrades. Add arbitrary noncontingent reinforcement on top of FCT specifically when stereotypy is the function (Boyle et al., 2018). Add a function-matched Social Story when mand training runs alongside classroom instruction (Pane et al., 2015). When thinning itself is the sticking point, moderately preferred items in the wait interval sped the process and the learner preferred it (Simmons et al., 2022).

Stay honest about the grade. FCT carries a Moderate rating built on a corpus that's overwhelmingly about refining an already-effective procedure, not about whether the core package works. Treat single-study findings like jump-to-terminal thinning as strong leads worth piloting, not settled defaults, and keep the more established tactics as first-line choices on higher-stakes cases.

What separates an FCT plan that holds up on a real caseload from one that collapses the first time reinforcement thins is almost entirely operational: what you do in the first few trials, how you structure the thinning steps, and who else is running the contingency besides you.

Before the first session, run a brief exposure to the establishing operation, thirty seconds to a minute of the relevant motivation, before delivering the first opportunity to mand. That small step reduced extinction bursts during acquisition compared to starting cold (Fisher et al., 2018), and it costs nothing beyond a slightly slower start. If the case is escape-maintained, build the demand-fading ladder into the same data sheet as the reinforcement schedule from day one, thinning both together held problem behavior down while raising task completion in a way reinforcement thinning alone didn't (Gerow et al., 2020).

Before you take the first thinning step, run a five-minute competing-activities screen and identify the highest-engagement item available, that item goes into the no-reinforcer periods once thinning starts and measurably lowers resurgence risk (Fuhrman et al., 2018). Stock the wait interval itself with a few moderately preferred tangibles, brief attention, or an easy demand rather than leaving it empty, thinning moved faster and learners preferred it when the wait wasn't just silence (Simmons et al., 2022).

Pick your thinning shape deliberately and brief every implementer on which one you're running. If you're moving to a jump-to-terminal schedule, tell staff the change is planned, not a data-sheet error, and watch the next several sessions closely for stability, the approach reached the lean schedule faster with no drop in effect in the study available, but it's one comparison, not a default for every learner (Chesbrough et al., 2024). If you're running a multiple schedule instead, the discriminative stimulus has to look and behave identically across every person who implements the plan, parents, other therapists, and any related service provider who shares the caseload. That consistency is the mechanism behind the fast transfer of gains to new settings and new people (Fisher et al., 2015).

Once the FCR is stable, don't stop at a single mand form. Move to a Lag 1 requirement in one session rather than waiting for an arbitrary mastery window, varied manding increased immediately without a rise in problem behavior (Adami et al., 2017). If the target FCR needs to be more complex than a single word or gesture, break it into three or four criterion steps and hold each until stable across two consecutive sessions, that sequence built complex responses without triggering resurgence (Ghaemmaghami et al., 2018). And don't assume the mand closest to the problem behavior's own history is the one to teach, a clean, unpaired topography works just as well and learners often choose it over the older one (Matter et al., 2017).

Build in resurgence insurance before you need it. Teach a second functionally equivalent mand alongside the first, that sequencing shrank the rebound once reinforcement was withheld (Lambert et al., 2017). If the case involves automatically maintained stereotypy, layer a short arbitrary noncontingent reinforcement break after FCT mands rather than relying on the mand contingency alone to suppress it (Boyle et al., 2018). In a classroom, pair the FCT contingency with a Social Story written to the specific function you identified, not a generic template, matched narratives outperformed non-matched ones on both behavior reduction and mand teaching (Pane et al., 2015).