Join Free →
About Weekly CEU Resource Hubs ↳ CEU Library ↳ Research Library Blog Contact Join Free →
Topic Guide · Practitioner

Free-Operant Preference Assessment: Procedure, Indications, and Decision Logic for BCBAs

Query target: free operant preference assessment · BBC Editorial Team
★ Summary

A free-operant preference assessment is a procedure in which a behavior analyst makes an array of candidate stimuli available simultaneously and continuously to a learner in an unconstrained environment, then infers preference from the cumulative duration of engagement with each item rather than from discrete approach choices. Originally formalized by Roane, Vollmer, Ringdahl, and Marcus (1998), the format is procedurally distinct from forced-choice assessments — paired stimulus, MSWO, and single stimulus — in three specific ways: the learner is never required to discriminate between two presented options, never has a preferred item withheld, and never has to emit a choice response in any conventional sense. The practitioner simply lays out the array, starts a timer, and records what the learner contacts and for how long.

01What the Research Says

The conceptual difference between free-operant and forced-choice

Forced-choice formats — PS, MSWO, brief MSWO, single stimulus — work by imposing a structured choice opportunity and recording the learner's selection on each trial. They produce relative preference data: item A was selected over item B 80% of the time, ranked above item C, and so on Verriden & Roscoe (2016). Free-operant works on a different principle. The practitioner arranges all candidate items simultaneously in an unconstrained space, allows continuous noncontingent access for a fixed observation period, and records duration of engagement; preference is inferred from the absolute allocation of behavior across stimuli when no contingency is shaping choice Bigwood et al. (2026) Verriden & Roscoe (2016). The implication is conceptual, not just procedural: a forced-choice PA tells you what a learner picks under a discrimination demand; a free-operant PA tells you what a learner does when nothing is asking for a discrimination at all. For learners who cannot reliably discriminate between two presented options, the forced-choice rank is invalid by construction; the free-operant duration measure is not Morris & Vollmer (2019) Bigwood et al. (2026).

The format trade-off: stability versus evocation

The cleanest direct comparison of preference-assessment formats is Verriden and Roscoe's evaluation across six children and adults with developmental disabilities, which examined four formats — PS, MSWO, free operant, and response restriction — for stability and reinforcer efficacy Verriden & Roscoe (2016). Free-operant showed the lowest stability of rankings across repeated sessions, while PS and MSWO produced the most stable hierarchies. That single result has been repeatedly mis-summarized as "free-operant is invalid"; the same study showed free-operant consistently produced the lowest levels of problem behavior during assessment because it never withholds preferred items between trials, and reinforcer-test confirmation showed that free-operant-identified high-preference items did still function as reinforcers, even when their across-administration ranks shifted Verriden & Roscoe (2016). The trade-off is precision against evocation, not validity against invalidity.

Tung, Donaldson, and Kahng made the evocation argument explicit. In a direct alternating-treatments comparison of free-operant, PS, and MSWO for three children with tangible-maintained problem behavior, free-operant generated significantly lower rates of problem behavior than either trial-based format because it never required removing a preferred item from the learner Tung et al. (2017). For populations where item removal reliably evokes aggression, SIB, or property destruction, this is the difference between an assessment that completes safely and one that contraindicates itself.

Herbek and colleagues replicated and extended this finding using a response-restriction free-operant variant (RR-FO), in which the learner has continuous access to all items but a contacted item is removed from the array for the next session so the learner must allocate to a different stimulus Herbek et al. (2026). Across six children with severe challenging behavior, RR-FO produced preference hierarchies equivalent to a paired-stimulus format while simultaneously producing lower rates of problem behavior in four of six cases Herbek et al. (2026). RR-FO was longer to administer than PS, but the absence of evocation more than offset the time cost in this population. When stability matters and safety also matters, RR-FO closes most of the precision gap the original Roane format leaves open.

The "absolute preference under no constraint" interpretation

An additional finding that distinguishes free-operant from forced-choice appears in Bigwood, Staples, and Sharp's dementia-adaptation work Bigwood et al. (2026). Three older adults with dementia received both 10-minute free-operant assessments (seven leisure items arranged equidistantly on a table, all continuously available) and single-stimulus assessments. Free-operant quantified preference while avoiding the confusion produced by repeated discrete-choice arrays. Critically, free-operant duration data identified items that verbal report from both caregiver and participant had missed entirely — puzzles were never verbally nominated as preferred, yet were selected most by duration of engagement when continuously available Bigwood et al. (2026). Because item availability never changes during observation, refusal-style statements ("Why are you asking again?") that derail discrete-trial PA in this population disappeared. This is the "absolute preference under no constraint" claim made operational: free-operant catches what forced-choice and verbal report systematically miss.

Free-operant in conjunction with concurrent operants

Free-operant logic — measuring relative behavior allocation across simultaneously available stimuli — is the procedural backbone for concurrent-operant preference assessments, which extend the format from leisure items to social consequences and demands. Robinson, Desrochers, and Napolitano ran brief concurrent-operant social preference assessments with five adolescents with dual diagnosis and used the resulting socially preferred consequence to reduce task-initiation latency for daily-living routines; the identified social reinforcer, delivered for prompt task initiation, immediately shortened latencies for laundry, showering, and chores, with gains maintained over two weeks Robinson et al. (2019). Lloveras, Call, Bourret, and Slocum extended concurrent-operant logic to demand assessment in 17 children and adolescents with developmental disabilities; relative selection between paired demands ranked task aversiveness in a way that paralleled progressive-ratio escape break-points Lloveras et al. (2020). Morris and Vollmer's Social Interaction Preference Assessment (SIPA) is the formal codification of free-operant logic for social stimuli: across five children with autism, freely available adult-delivered social interactions (tickles, praise, singing) generated differential approach and time-allocation patterns that identified preferred social reinforcers without requiring the picture-tact prerequisites MSWO of social interactions presumes Morris & Vollmer (2019).

Video-based and digital free-operant variants

Free-operant logic translates cleanly to digital stimuli when the practitioner cannot or does not want to deliver contingent access. Brodhead, Kim, and Rispoli demonstrated that a video-only free-operant preference assessment — video clips presented without contingent reinforcer delivery, duration of viewing as the dependent measure — predicted relative reinforcer efficacy in a subsequent concurrent-operant reinforcer test for five children with autism; high-preference items functioned as reinforcers, low-preference items did not Brodhead et al. (2019). Curiel and colleagues' web-based MSWO of videos and Curiel et al.'s MSWO Preference Assessment Tool extend the same logic into rapid online procedures; these are technically MSWO rather than pure free-operant but inherit the duration-as-preference reasoning when videos are sampled simultaneously Curiel et al. (2018) Curiel et al. (2024). The MSWO PAT successfully predicted which videos would function as reinforcers for five of seven autistic children in concurrent-chains conjugate reinforcer tests Curiel et al. (2024). When the planned reinforcer is digital, a brief video-only free-operant or web-based brief MSWO is a defensible screen, provided a concurrent-operant reinforcer-validation step is layered on top.

The preference-versus-reinforcement gap

Free-operant produces a hierarchy of engagement; it does not by itself produce evidence of behavior change Frank‐Crawford et al. (2018). Frank-Crawford, Castillo, and DeLeon's behavioral-economic analysis used paired-stimulus PAs to rank 12-14 food items and tested whether the top four PA-selected items retained their reinforcing value when fixed-ratio response requirements increased; items that appeared similarly preferred under low-cost (FR 1) conditions did not always function equivalently at higher FR values Frank‐Crawford et al. (2018). The same logic applies one rung up: a free-operant duration ranking under no constraint does not guarantee the top-ranked item will function as a reinforcer when the learner must emit responses to access it. Gilroy, Waits, and Feck made this gap operational by combining free-operant SP assessments with a behavioral-economic operant-demand framework: after stimuli were ranked highly in a free-operant assessment, their elasticity of demand under progressive-ratio schedules quantified reinforcer efficacy beyond preference rank, with elastic items emerging as the strongest reinforcers Gilroy et al. (2021). Verriden and colleagues showed the same pattern for automatically reinforced challenging behavior: items chosen via a free-operant competing-stimulus assessment generally produced larger reductions in problem behavior under noncontingent reinforcement than items chosen via a trial-based PSPA Verriden et al. (2025). The methodological discipline is consistent: rank with a free-operant PA, then validate the top one or two items under a concurrent-operants or progressive-ratio reinforcer test that approximates the actual treatment context Frank‐Crawford et al. (2018) Gilroy et al. (2021).

Group, classroom, and practitioner-usage data

Free-operant logic scales to classrooms when class-wide reinforcer identification matters more than individual hierarchies. Radley, Dart, Battaglia, and Blake Ford's group preference scan with 19 elementary students used Plickers cards as a simultaneous-response tool; students simultaneously held up QR-coded cards indicating their choice from a small array, and the resulting group-level rankings matched the rankings produced by individual one-to-one MSWO sessions, with subsequent progressive-ratio testing confirming reinforcer potency of the group-scan top items Radley et al. (2019). Miranda and colleagues' qualitative interviews with eight preschool teachers found that only two had previously heard the term free-operant preference assessment, and neither could recall the procedure; teachers nonetheless rated the format as potentially useful for embedding student choice and monitoring preference over time, indicating the bottleneck is procedural training, not acceptability Miranda et al. (2025). Morris, Conine, Slanzi, Kronfli, and Etchison's survey of BCBAs providing intensive ABA showed that clinicians frequently swap reinforcers mid-session and most often do so by offering an informal choice among items — a "mini-preference assessment" that often takes the form of a single brief free-operant observation (Morris et al., 2024). McCammon, Wolfe, and Check's narrative review of mand-training studies makes the inverse argument: because preference fluctuates, single-shot pre-baseline PAs risk prompting children to mand for items they no longer want, and the recommended fix is brief daily preference checks or moment-to-moment behavioral indicators — a procedural prescription that maps directly onto brief free-operant probes (McCammon et al., 2024).

Decision-system positioning, assessor effects, and validation across species

Lill, Shriver, and Allen's Stimulus Preference Assessment Decision-Making System (SPADS) positions free-operant duration probes as the initial step for identifying highly preferred items: if a learner naturally approaches and engages with an item for at least 20 seconds, that item is moved to the top-ranked category and treated as high-preference without requiring a formal trial-based comparison Lill et al. (2021). The 20-second engagement gate is conceptual, not yet empirically validated, but it operationalizes a clinical intuition every experienced practitioner has — a learner who stays with a toy for 20 seconds in a free choice context is telling you something the rest of the array doesn't need to verify. Free-operant procedures preserve the assessor-effect problem that complicates social-stimulus PA generally; Huntington and Schwartz showed that one adult with autism selected different social interactions depending on whether his mother, a staff member, or an unfamiliar researcher delivered them, with the highly-ranked, assessor-specific interactions producing higher response rates in a follow-up reinforcer test Huntington & Schwartz (2022). When the planned reinforcer is social, the assessor during the free-operant scan must be a member of the same partner class as the treatment-delivery person, and the data sheet must document who that was Morris & Vollmer (2019) Huntington & Schwartz (2022). Cross-species replication adds further validity evidence: Payne, Fulgencio, and Aniga's comparison of PS and free-operant MSWO across 12 companion dogs found that free-operant MSWO produced a more reliable hierarchy of food items than PS and required fewer sessions to identify the same high-preference items Payne et al. (2023). Waite, Kodak, and Whang's caretaker-implemented ear-cleaning protocol explicitly recommends free-operant PA as the prerequisite step for less food-motivated companion dogs, reasoning that free-operant duration data identifies viable reinforcers without imposing the discrimination demand a paired-stimulus format would (Waite et al., 2025).

02Evidence Tier Breakdown

The free-operant preference assessment literature is dominated by single-subject experimental designs anchored by a handful of format-comparison studies, with one decision-making model, one survey, one qualitative interview study, and a small set of narrative reviews providing supporting context Verriden & Roscoe (2016) Tung et al. (2017).

Single-subject experimental designs. The format-comparison evidence sits here. Verriden and Roscoe (n=6) anchor the four-format stability comparison and the case for free-operant's lower problem-behavior rates Verriden & Roscoe (2016). Tung, Donaldson, and Kahng (n=3) provide the strongest direct demonstration that free-operant evokes less problem behavior than PS or MSWO for tangible-maintained populations Tung et al. (2017). Herbek et al. (n=6) extend the safety-equivalence argument to RR-FO with rankings comparable to PS Herbek et al. (2026). Bigwood et al. (n=2) demonstrate the dementia adaptation Bigwood et al. (2026). Robinson et al. (n=5), Lloveras et al. (n=17), and Morris and Vollmer (n=5) ground the concurrent-operant social and demand extensions Robinson et al. (2019) Lloveras et al. (2020) Morris & Vollmer (2019). Brodhead et al. (n=5) and Curiel et al. (n=7) cover video and web-MSWO predictive validity Brodhead et al. (2019) Curiel et al. (2024). Frank-Crawford et al., Gilroy et al. (n=3), and the Verriden et al. 2025 CSA paper (n=5) cover the preference-versus-reinforcement gap and progressive-ratio demand Frank‐Crawford et al. (2018) Gilroy et al. (2021) Verriden et al. (2025). Huntington and Schwartz (n=1) document assessor effects Huntington & Schwartz (2022). Kronfli et al. (n=3) extend MSWO embedding to fruit-and-vegetable reinforcer comparisons in applied teaching (Kronfli et al., 2024). Waits and Gilroy (n=2) cover bilingual extensions that border on free-operant logic (Waits & Gilroy, 2025). Payne et al. (n=12 dogs) and Waite et al. (n=3 dogs) provide cross-species replication Payne et al. (2023) (Waite et al., 2025). O'Handley et al. (n=4) document trainability via brief BST O’Handley et al. (2021).

Quasi-experimental. Radley et al.'s classroom group-scan comparison (n=19) is the field's main classroom-feasibility study Radley et al. (2019).

Methodology and decision-model papers. Lill, Shriver, and Allen's SPADS is the only published decision-making model that explicitly positions free-operant within a structured selection logic; the model awaits empirical validation Lill et al. (2021). Curiel et al.'s web-based MSWO procedure is a methodology demonstration with embedded efficacy data Curiel et al. (2018).

Survey and qualitative. Morris, Conine, et al.'s BCBA survey describes how free-operant logic actually appears in EIBI day-to-day (Morris et al., 2024). Miranda et al.'s qualitative interviews with eight preschool teachers describe procedural-knowledge gaps that block classroom adoption Miranda et al. (2025).

Narrative review. McCammon, Wolfe, and Check's review of 35 mand-training studies frames free-operant logic as the corrective for stale single-shot pre-baseline preference data (McCammon et al., 2024).

Bottom line. Convergent evidence is strong for: (1) free-operant produces less stable rankings than PS or MSWO across repeated administrations Verriden & Roscoe (2016); (2) free-operant produces dramatically lower problem-behavior rates for tangible- and item-removal-evoked populations Tung et al. (2017) Herbek et al. (2026); (3) RR-FO closes most of the precision gap with PS while preserving the safety advantage Herbek et al. (2026); (4) free-operant logic extends cleanly to concurrent-operant social and demand assessments Robinson et al. (2019) Lloveras et al. (2020) Morris & Vollmer (2019); (5) video-only and web-based brief variants reproduce free-operant predictive validity when prerequisites are intact Brodhead et al. (2019) Curiel et al. (2024). Evidence is weaker for empirical validation of SPADS's 20-second engagement gate Lill et al. (2021) and for any prospective trial comparing treatment outcomes when clinicians do versus do not validate top free-operant items via concurrent-operants tests Frank‐Crawford et al. (2018).

03Decision Logic

The free-operant-versus-forced-choice question is not "which is better?" — both formats are valid for what they measure. The right question is: what is this client's repertoire, what is the safety profile, and what is the assessment for?

  1. Severe problem behavior evoked by item removal (particularly tangible-maintained). Default to free operant or RR-FO. FO produces dramatically lower problem-behavior rates than PS or MSWO Tung et al. (2017); RR-FO closes most of the precision gap with PS while preserving the safety advantage Herbek et al. (2026).

  2. Very limited choice-making repertoire, non-discriminated reaching, or cannot reliably select between two items. Use free operant. Forced-choice requires a discriminated selection response; if that response is unreliable, the resulting hierarchy is invalid by construction Verriden & Roscoe (2016) Morris & Vollmer (2019).

  3. Dementia or neurocognitive disorder where repeated discrete-choice arrays evoke confusion or refusal. Use a 10-minute free-operant procedure with continuous availability of all items Bigwood et al. (2026). Single-stimulus assessment is a strong alternative.

  4. Severely impaired learner who does not respond to forced-choice formats. Free operant. The learner allocates whatever behavior they have available; the assessor records duration; no discrimination response is required Morris & Vollmer (2019).

  5. Observation-based screening before forced-choice. SPADS positions free-operant duration probes as the gate that elevates items into the high-preference category before structured trial-based comparisons are run on the remaining items Lill et al. (2021).

  6. Setting where prompted choice is impractical — residential, severe SIB, in-home assessment. Free operant. The procedural footprint is one timer, one clipboard, one observer; the format tolerates the variability of natural environments Bigwood et al. (2026) Robinson et al. (2019).

  7. Time-pressed in-session reinforcer adjustment. A 1-3 minute brief free-operant probe with a small array; the 20-second engagement threshold from SPADS lets the practitioner select a next-block reinforcer without a formal data sheet Lill et al. (2021) (Morris et al., 2024).

  8. Whole-classroom group reinforcer identification. Use a group simultaneous-response free-operant variant such as Plickers cards Radley et al. (2019).

  9. Reinforcer is a social interaction. Run a SIPA-style free-operant assessment with the assessor who will deliver the consequence in treatment; document the assessor Morris & Vollmer (2019) Huntington & Schwartz (2022).

  10. Reinforcer is digital. Use a video-only free-operant or web-based brief MSWO; pair with a concurrent-operant reinforcer-validation step Brodhead et al. (2019) Curiel et al. (2024).

  11. Need a precise full-rank hierarchy across mid-range items. Free-operant rankings are less stable than PS or MSWO across administrations Verriden & Roscoe (2016). Use free-operant to gate the candidate set, then forced-choice for the precision step.

  12. Top items identified — validate before treatment commitment. Run a concurrent-operants reinforcer test under conditions resembling the actual treatment schedule before committing to a BIP or skill-acquisition plan; preference rank does not always predict reinforcer substitution at higher response costs Frank‐Crawford et al. (2018) Gilroy et al. (2021). For automatically maintained problem behavior, run a free-operant CSA instead and use those items for NCR Verriden et al. (2025).

The decision is not free-operant versus forced-choice; it is which format to run first. For most cases involving severe disability, item-removal-evoked problem behavior, severely restricted response repertoires, or unconstrained-environment assessment, free-operant is the first format. Forced-choice (PS or MSWO) follows when precision among mid-range items is needed.

04Across Settings

Clinic intake

In an outpatient or specialty clinic, free-operant is the format of choice during the first session with a learner whose history of severe problem behavior — particularly aggression or SIB evoked by item removal — would contraindicate paired-stimulus or MSWO. The full procedure runs in 5-10 minutes per session across 2-3 sessions; the practitioner uses the resulting top-tier items as session reinforcers for subsequent functional-analysis or skill-acquisition work Tung et al. (2017) Herbek et al. (2026). For learners whose history suggests escape-maintained problem behavior, a concurrent-operant demand assessment (Lloveras-style) added to the intake protocol identifies the least-aversive demand class for early teaching trials Lloveras et al. (2020). Clinic environments are also where RR-FO earns its keep: the controlled space allows sequential sessions with item removal without disrupting natural routines Herbek et al. (2026). For automatically reinforced challenging behavior, switch to a free-operant competing-stimulus assessment rather than a standard PA; CSA-derived items often produce larger reductions under NCR than PSPA-derived items Verriden et al. (2025).

Residential care and severe-disability environments

Residential settings concentrate the populations for whom free-operant is uniquely well-suited: severe and profound intellectual disability, dual diagnosis, severe behavior histories, dispersed staff who cannot run trial-based formats reliably. Robinson et al.'s adolescent dual-diagnosis work demonstrates the model: a brief concurrent-operant social preference assessment identifies the social consequence each resident will work for, and that consequence — delivered for prompt task initiation — immediately shortens latency to begin daily-living routines without extra prompting or token systems Robinson et al. (2019). Because residential staff turnover and dispersion make centralized PA training hard, the preferred installation pattern is brief BST or VMVO-style training of free-operant data collection; O'Handley et al. demonstrated mastery via a 45-minute behavioral skills training session with maintenance at 30 and 60 days O’Handley et al. (2021).

Severe-disability classrooms and special education

Special-education classrooms serving learners with severe disabilities face the same population characteristics as residential settings, with the added constraint that teacher time per learner is short. Three deployment patterns work. First, embed a 5-minute free-operant probe into a free-play period at the start of the school day, recording duration across a small array and using the top item as the day's primary reinforcer Miranda et al. (2025). Second, for class-wide reinforcer identification, use the Radley group simultaneous-response free-operant scan (Plickers cards): a 30-second scan produces class-level rankings that match individual MSWO data while preserving instructional time Radley et al. (2019). Third, for individual programming with a learner whose response repertoire defeats PS or MSWO, use a 5-minute free-operant probe at the start of each session with a small array drawn from caregiver/teacher report — the format teachers in Miranda et al.'s qualitative interviews described as feasible once they had procedural training Miranda et al. (2025).

Home-based assessment of in-home reinforcers

The home environment is the natural habitat of free-operant assessment: a small array of items that already exist in the home, an observer (the practitioner or trained caregiver), and a timer. Bigwood et al.'s dementia-adaptation procedure — seven items equidistantly arranged on a table, 10 minutes of continuous observation, four sessions across days — translates directly to in-home assessment of leisure preferences for any population that benefits from continuous availability Bigwood et al. (2026). For bilingual families, pair the free-operant procedure with a culturally-informed RAISD interview that identifies idiosyncratic candidate items and language-paired stimuli; Waits and Gilroy's bilingual paired-stimulus protocol is the closest published example, and the same logic applies to free-operant duration measurement of bilingual reinforcers (Waits & Gilroy, 2025). For caregivers running the assessment without a practitioner present, brief telehealth coaching with real-time feedback brings direct-care staff and parents to mastery in 3-4 sessions O’Handley et al. (2021).

Adult day programs and supported employment

For adults with developmental disabilities in day programs and supported employment, free-operant has two specific applications. Gilroy, Waits, and Feck combined free-operant SP assessments with a behavioral-economic operant-demand framework for three adults attending a day program; elasticity of demand under progressive-ratio schedules quantified reinforcer efficacy beyond preference rank, allowing the team to break ties among top-ranked items for vocational reinforcement Gilroy et al. (2021). The Robinson et al. concurrent-operant social preference work generalizes here: for adults with limited verbal repertoires, a brief free-operant social preference probe identifies the staff-delivered consequence that will reduce task-initiation latency for vocational and self-care routines Robinson et al. (2019).

05Case Examples

Original procedure (Roane et al. 1998 and standard implementation)

The canonical free-operant preference assessment was formalized by Roane, Vollmer, Ringdahl, and Marcus (1998) and has remained procedurally stable across the format-comparison literature. The standard implementation has the following components, adapted across studies in the corpus to specific populations and stimulus classes Verriden & Roscoe (2016) Bigwood et al. (2026) Tung et al. (2017).

Pre-session arrangement. Identify a candidate item array, typically 5-10 stimuli drawn from caregiver/RAISD interview, indirect inventories, prior preference data, or items observed in the environment. Items should be drawn from the same general stimulus class (edibles, leisure items, social interactions) where possible Lill et al. (2021).

Item placement. Arrange items equidistantly in a layout the learner can scan and reach without prompting — a table, a mat, a shelf at eye level. Bigwood et al.'s dementia-adaptation work specifies seven items arranged equidistantly on a table for 10 minutes Bigwood et al. (2026). Standardize placement across administrations to reduce proximity bias.

Observation duration. Two to ten minutes of continuous observation per session is the published range. Tung et al. and Verriden & Roscoe used 5 minutes; Bigwood et al. used 10 minutes; Brodhead et al.'s video procedure used 5-10 minutes Tung et al. (2017) Verriden & Roscoe (2016) Bigwood et al. (2026) Brodhead et al. (2019). Two minutes is the floor for any usable measurement; 5 minutes is the modal default; 10 minutes is recommended for populations with slow-stabilizing engagement bouts (older adults with NCD, learners with very limited approach repertoires) Bigwood et al. (2026).

Data collection. The primary dependent variable is duration of engagement with each item across the observation period, captured via continuous timing on a paper or digital data sheet, partial-interval recording, or video review Bigwood et al. (2026). Secondary measures include latency to first contact, repeated approaches (separate engagement bouts per item), and total proportion of session time allocated. Engagement must be operationally defined per stimulus class — for edibles, consumption; for leisure items, manipulating, holding, or playing; for social interactions, oriented attending plus reciprocal engagement with the partner Morris & Vollmer (2019). If the only relevant response is olfactory or covert (the learner sniffs an item without subsequent visible contact), the protocol must be modified to record that response or the data are incomplete Bigwood et al. (2026).

Hierarchy construction. Rank items by total duration (or proportion of session time). SPADS adds an engagement-duration threshold — items contacted for ≥20 seconds within an unconstrained free-access period are designated high-preference and elevated to the top category for the next decision step Lill et al. (2021).

Repeated administration. Run multiple sessions to stabilize rankings. Bigwood et al.'s 4-session multiple-probe design captures across-day variability that single-session data cannot Bigwood et al. (2026). For populations with stable engagement bouts, 2-3 sessions across 1-2 days is typical Verriden & Roscoe (2016).

The response-restriction free-operant (RR-FO) variant

RR-FO adds a single procedural change: after a session, the item the learner contacted most (or any item that crossed an engagement threshold) is removed from the array for the next session, forcing differential allocation across remaining stimuli over time Herbek et al. (2026). The hierarchy is built by combining engagement durations across sessions. Herbek et al.'s procedure produced rankings comparable to a paired-stimulus PA across six children with severe challenging behavior while preserving the safety advantage of never withholding the array as a whole during any single session Herbek et al. (2026). RR-FO takes longer than a single free-operant series, but the time cost is offset for populations who would otherwise produce unusable PS data due to escalation.

Brief free-operant and concurrent-operant variants

Clinicians in EIBI report using brief free-operant probes mid-session as informal "mini-preference assessments" — a small array offered for an unconstrained observation, with the item the learner approaches and engages selected as the next reinforcer (Morris et al., 2024). This is procedurally a 1-3 minute single-session free-operant scan and is the operational answer to McCammon et al.'s concern about stale pre-baseline preference data (McCammon et al., 2024). SPADS's 20-second engagement gate makes brief probes defensible without formal data collection: if the learner spontaneously contacts an item for ≥20 seconds, treat it as high-preference for the next teaching block Lill et al. (2021). The free-operant procedure scales to social interactions and demands. For SIPA-style social preference, the assessor delivers different social interactions (tickles, praise, singing) on the learner's approach, and the practitioner records duration of contact with each Morris & Vollmer (2019). For concurrent-operant demand assessment, pairs of demand stimuli (work cards, work-task pictures) are presented and relative selection across pairs ranks task aversiveness in a way that parallels progressive-ratio escape break-points Lloveras et al. (2020). Both are free-operant in structure but extend the dependent-variable definition to fit the stimulus class. For digital stimuli, Brodhead et al.'s video-only procedure presents clips simultaneously without contingent reinforcer access and records duration of viewing per clip Brodhead et al. (2019); Curiel et al.'s web-based brief MSWO compresses this further into a sub-60-second online sequence Curiel et al. (2018) Curiel et al. (2024). Pair every alternative-modality free-operant or brief MSWO with a concurrent-operant reinforcer-validation step using the actual digital stimulus contingent on responding before committing to it in a treatment plan.

06Common Pitfalls

  • Insufficient observation duration. A 1- or 2-minute observation window does not allow engagement bouts to stabilize, especially for populations with slow approach repertoires (older adults with NCD, learners with severe disabilities, learners with limited play repertoires). The published modal duration is 5 minutes; for slower-engaging populations, 10 minutes is recommended Bigwood et al. (2026) Verriden & Roscoe (2016). Brief in-session probes are valid for reinforcer adjustment but should not be the basis for formal hierarchy construction.

  • Poor placement and proximity bias. Items placed nearer to the learner, in the line of dominant sight, or at hand-height are more likely to be approached regardless of preference. Standardize placement: equidistant from the learner, at a consistent height, with the same orientation across sessions Bigwood et al. (2026). Rotate item position across sessions to detect placement effects.

  • Satiation timing. A high-preference item identified at 9 a.m. with a hungry learner is not the same item at 11 a.m. after free-access snack time. Free-operant data is sensitive to motivating-operation state — schedule the assessment when the relevant MO is most likely to be active, document the time and any recent consumption on the data sheet, and re-assess any time MOs meaningfully shift (McCammon et al., 2024).

  • Missed engagement (covert or non-observable response). If the learner's primary mode of contact with an item is olfactory, auditory, or visual without subsequent observable manipulation, a duration-of-manipulation measure will silently miss the engagement and bias the hierarchy toward items that produce visible motor contact Bigwood et al. (2026). Operationalize engagement per stimulus class: for example, include sustained visual orienting plus consummatory contact for olfactory or visual stimuli, and document the operational definition in the protocol.

  • Treating the free-operant hierarchy as the reinforcer hierarchy. Free-operant duration ranks engagement under no constraint; it does not by itself produce evidence that contingent access to a top-ranked item will increase responding. Validate the top one or two items with a concurrent-operants reinforcer test under conditions resembling the actual treatment schedule Frank‐Crawford et al. (2018) Gilroy et al. (2021).

  • Treating one free-operant session as a stable hierarchy. Across-administration stability of free-operant rankings is lower than PS or MSWO; a single session should be treated as a screen, not a hierarchy Verriden & Roscoe (2016). Run 2-3 sessions before committing to a rank, and consider RR-FO if PS-comparable precision is needed Herbek et al. (2026).

  • Running free-operant when the question is competing-stimulus or demand-aversion. Free-operant produces a duration-based preference rank; it does not measure whether free access to an item will suppress a target behavior. For automatically maintained problem behavior, the right procedure is a free-operant competing-stimulus assessment — same procedural structure, different dependent variable (rate of target behavior under noncontingent stimulus access) Verriden et al. (2025). For task aversion, a concurrent-operant demand assessment ranks demands rather than items Lloveras et al. (2020).

  • Ignoring the assessor effect for social free-operant. Social interactions selected in a free-operant scan depend on who delivers them. A praise type ranked high when the parent delivers it may be lower-preferred and a weaker reinforcer when a staff member delivers the same form. Run social free-operant with the partner who will deliver the consequence in treatment, and document the assessor on every data sheet Huntington & Schwartz (2022) Morris & Vollmer (2019).

  • Using a video-only or digital free-operant without verifying prerequisites. Alternative-modality free-operant assessments only approach tangible accuracy when the learner can discriminate the video, icon, or symbol from competitors. Missing prerequisites produce silent invalid hierarchies Brodhead et al. (2019) Curiel et al. (2024).

  • Failing to define engagement operationally per stimulus class. "Duration of engagement" means different things for an edible (consumption), a leisure item (manipulation, holding, playing), and a social interaction (oriented attending, reciprocal engagement). Specify the operational definition in the protocol before running the assessment, or different observers will produce different rankings Morris & Vollmer (2019).

  • Single-shot pre-baseline preference data. McCammon and colleagues' review of mand-training studies identified the systemic error of running one PA before baseline and then prompting children to mand for items they no longer want; preference fluctuates (McCammon et al., 2024). Plan brief in-session free-operant probes into every 30-40 minute teaching block, particularly for mand training and other procedures where MOs shift rapidly (Morris et al., 2024).

07When to Refer Out

  • Suspected medical or biological substrate driving apparent preference shifts. Sudden item rejection, food refusal, or new aversion patterns in a previously stable hierarchy can mask GI, dental, or sensory-related issues; document a medical referral before re-running or re-validating the free-operant assessment.

  • Severe automatically maintained problem behavior unresponsive to free-operant CSA-derived items. When competing-stimulus assessments fail to identify suppressing items across replications, refer for specialist consultation with a team experienced in matched-stimulation and sensory-enrichment programming Verriden et al. (2025).

  • Persistent inability to identify any preferred item across formats and re-assessments. If free-operant, brief MSWO, and single-stimulus all fail to surface any reliably approached stimulus across multiple sessions, escalate to a specialist team rather than committing to a treatment plan without an empirical reinforcer.

  • Client repertoire requirements beyond local staff training capacity. When the client population requires SIPA, concurrent-operant demand assessment, or operant-demand-framework analyses and local staff cannot reach mastery via brief BST or telehealth packages, refer to a regional consultation team rather than running an underpowered assessment in-house Morris & Vollmer (2019) Gilroy et al. (2021) O’Handley et al. (2021).

08Future Research Directions

The corpus has strong operational claims about free-operant's safety advantages and procedural feasibility, but several gaps remain. SPADS's 20-second engagement gate is the field's clearest written articulation of how to operationalize a free-operant duration probe, but the threshold itself is conceptual and lacks empirical validation against subsequent reinforcer-test data Lill et al. (2021). A prospective comparison of treatment outcomes when clinicians do versus do not validate top free-operant items via concurrent-operants tests would settle the practical importance of the preference-versus-reinforcement gap for this format Frank‐Crawford et al. (2018) Gilroy et al. (2021). The intra-administration stability of free-operant rankings across multiple sessions has been documented as lower than PS or MSWO but has not been quantified with the same rigor that Melanson and colleagues applied to MSWO; an analogous rank-correlation study within free-operant is overdue Verriden & Roscoe (2016). Concurrent-operant demand assessment outside inpatient settings is under-replicated Lloveras et al. (2020). The institutional-adoption gap identified by Miranda and colleagues — preschool teachers who have never heard the term free-operant preference assessment despite rating the format as acceptable — needs implementation-science work to solve, particularly because the procedural-fidelity literature shows the format can be taught in under an hour Miranda et al. (2025) O’Handley et al. (2021).

09Practitioner Takeaways

  1. Free-operant measures absolute preference under no constraint; forced-choice measures relative preference under a discrimination demand. They are not interchangeable. Use free-operant when the question is what the learner allocates behavior toward in an unconstrained environment; use PS or MSWO when the question is which of two simultaneously presented items the learner picks Verriden & Roscoe (2016) Bigwood et al. (2026).

  2. Default to free-operant or RR-FO when problem behavior is evoked by item removal. Tangible-maintained problem behavior is the canonical case; FO produces dramatically lower problem-behavior rates than PS or MSWO during assessment, and RR-FO closes most of the precision gap with PS Tung et al. (2017) Herbek et al. (2026).

  3. Use free-operant for severely impaired learners who do not respond to forced-choice formats. Forced-choice requires a discriminated selection response; free-operant does not Morris & Vollmer (2019) Bigwood et al. (2026).

  4. Use 5-10 minute observation windows; longer for slow-engaging populations. Five minutes is the modal published duration; 10 minutes is recommended for older adults with NCD and learners with limited approach repertoires Verriden & Roscoe (2016) Bigwood et al. (2026).

  5. Run multiple sessions before treating a free-operant ranking as stable. Free-operant rankings are less stable across administrations than PS or MSWO; 2-3 sessions is the minimum for a defensible hierarchy Verriden & Roscoe (2016). Use RR-FO when PS-comparable precision is required Herbek et al. (2026).

  6. Treat the SPADS 20-second engagement threshold as a high-preference gate. Items contacted for ≥20 seconds in a free-operant probe are designated high-preference and elevated to the next decision step without requiring a formal trial-based comparison Lill et al. (2021).

  7. Use brief 1-3 minute free-operant probes for in-session reinforcer adjustment. This is what experienced BCBAs in EIBI actually do mid-session, and it solves the McCammon stale-PA problem for mand training and other procedures where MOs shift (Morris et al., 2024) (McCammon et al., 2024).

  8. Validate top free-operant items with a concurrent-operants reinforcer test before BIP commitment. Free-operant produces a duration ranking; reinforcer status requires evidence that contingent access actually increases responding under treatment-like schedules Frank‐Crawford et al. (2018) Gilroy et al. (2021).

  9. For social free-operant, document the assessor. Social preference is assessor-specific; run the assessment with the partner who will deliver the consequence in treatment Huntington & Schwartz (2022) Morris & Vollmer (2019).

  10. For automatically maintained problem behavior, switch from free-operant PA to free-operant CSA. A standard PA produces a preference rank; a CSA produces evidence of behavior suppression under noncontingent stimulus access. CSA-derived items often outperform PSPA-derived items for NCR Verriden et al. (2025).

  11. Operationalize engagement per stimulus class before running the assessment. Edibles, leisure items, social interactions, and digital stimuli each require a stimulus-class-specific operational definition of engagement, or different observers will produce different rankings Morris & Vollmer (2019) Bigwood et al. (2026).

  12. Train staff with brief BST or VMVO; the trainability evidence is encouraging. A 45-minute BST package brings preservice graduate students to MSWO and free-operant mastery with maintenance at 30 and 60 days; the institutional-adoption barrier is workload pressure, not teachability O’Handley et al. (2021) Miranda et al. (2025).

10Frequently Asked Questions

What exactly is a free-operant preference assessment?

A free-operant preference assessment is a procedure in which the practitioner makes an array of candidate stimuli simultaneously and continuously available to the learner in an unconstrained environment, and infers preference from the cumulative duration of engagement with each item across an observation period (typically 2-10 minutes) Bigwood et al. (2026) Verriden & Roscoe (2016). The learner is never required to discriminate between two presented options, never has a preferred item withheld, and never has to emit a discrete choice response; preference is inferred from how the learner allocates behavior under no constraint Tung et al. (2017) Herbek et al. (2026). The format was formalized by Roane, Vollmer, Ringdahl, and Marcus (1998) and has remained procedurally stable across the format-comparison literature.

How is free-operant different from MSWO or paired-stimulus preference assessment?

Forced-choice formats (PS, MSWO) require the learner to emit a discriminated selection response between simultaneously presented items on each trial; preference is inferred from the percentage of opportunities each item was selected over its competitors Verriden & Roscoe (2016). Free-operant requires no such selection response — the learner allocates behavior across continuously available items, and preference is inferred from duration of engagement Bigwood et al. (2026). Forced-choice produces relative preference data (item A over item B); free-operant produces absolute preference data (item A engaged with for X seconds out of Y) Verriden & Roscoe (2016). Forced-choice produces more stable rankings across repeated administrations; free-operant produces dramatically lower problem-behavior rates during assessment for tangible-maintained populations Verriden & Roscoe (2016) Tung et al. (2017) Herbek et al. (2026).

When should I use free-operant instead of MSWO?

Use free-operant when the client has a history of severe problem behavior evoked by item removal (particularly tangible-maintained), when the client has a very limited choice-making repertoire or non-discriminated reaching, when the setting makes prompted choice impractical (residential, severe SIB, in-home assessment of in-home items), when the learner has dementia or NCD where repeated discrete-choice arrays evoke confusion, or when you want a fast initial screen before deciding whether to run a more structured forced-choice format Tung et al. (2017) Herbek et al. (2026) Bigwood et al. (2026) Lill et al. (2021). Use MSWO when the client has scanning and selection prerequisites, the item class is well-defined, you need a precise hierarchy across mid-range items efficiently, and there is no item-removal-evoked problem behavior in the history Verriden & Roscoe (2016).

How long should the observation period be?

Two minutes is the floor for any usable measurement; 5 minutes is the modal published duration across format-comparison studies; 10 minutes is recommended for populations with slow approach repertoires (older adults with NCD, learners with severe disabilities, learners with limited play repertoires) Verriden & Roscoe (2016) Bigwood et al. (2026). For brief in-session reinforcer adjustment, a 1-3 minute probe is acceptable and matches what experienced BCBAs in EIBI actually do, with the SPADS 20-second engagement threshold serving as the high-preference gate (Morris et al., 2024) Lill et al. (2021).

Is free-operant valid for severely impaired learners or learners who don't respond to forced-choice?

Yes — that is the population for which free-operant is uniquely well-suited. Forced-choice formats require a discriminated selection response between simultaneously presented options; if the learner cannot reliably emit that response, the resulting hierarchy is invalid by construction Verriden & Roscoe (2016). Free-operant requires no such response; the learner allocates whatever behavior they have available, and the assessor records duration of allocation Morris & Vollmer (2019) Bigwood et al. (2026). Bigwood et al.'s dementia-adaptation work specifically demonstrated that free-operant identified preferred items (puzzles) that both caregiver and participant verbal report had missed entirely Bigwood et al. (2026).

Does a high free-operant ranking guarantee a stimulus will function as a reinforcer?

No. Free-operant produces a hierarchy of engagement under no constraint; it does not by itself produce evidence that contingent access to the top-ranked item will increase responding when the learner must work for access Frank‐Crawford et al. (2018) Gilroy et al. (2021). The standard methodological discipline is to validate the top one or two items with a concurrent-operants reinforcer test under conditions resembling the actual treatment schedule, or to combine free-operant ranking with a behavioral-economic operant-demand analysis that quantifies elasticity of demand under progressive-ratio schedules Gilroy et al. (2021). Items that look similarly preferred at low response cost (FR 1) do not always retain their advantage at higher response costs Frank‐Crawford et al. (2018).

Can I run a free-operant preference assessment via telehealth or with digital stimuli?

Yes, with verified prerequisites. Brodhead and colleagues' video-only free-operant procedure presents video clips simultaneously without contingent reinforcer access and records duration of viewing per clip; items selected as high-preference functioned as reinforcers in subsequent concurrent-operant tests Brodhead et al. (2019). Curiel et al.'s web-based MSWO and the Curiel et al. 2024 MSWO Preference Assessment Tool extend this into rapid online procedures that predict reinforcer efficacy in concurrent-chains tests Curiel et al. (2018) Curiel et al. (2024). The procedural rule across digital variants: the learner must be able to discriminate the digital stimulus from competitors, and a concurrent-operant reinforcer-validation step using the actual digital stimulus contingent on responding should follow before committing to it in a treatment program Brodhead et al. (2019). For telehealth caregiver implementation, brief BST and live-feedback packages bring direct-care staff to mastery in 3-4 sessions O’Handley et al. (2021).

What's the difference between free-operant PA and a free-operant Competing Stimulus Assessment?

The procedural structure is similar — items continuously available, simultaneous engagement opportunity, duration-based dependent variable — but the dependent variable and the question differ. A free-operant PA measures duration of engagement with each item to rank preference. A free-operant Competing Stimulus Assessment measures the rate of a target problem behavior under noncontingent access to each item, ranking items by their suppressive effect on the target behavior Verriden et al. (2025). For automatically maintained problem behavior specifically, CSA-derived items often produce larger reductions under NCR than items derived from a paired-stimulus preference assessment Verriden et al. (2025). The clinical question determines which procedure to run: PA for "what does this learner prefer?", CSA for "which items, delivered noncontingently, suppress this target behavior?"

12References

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