An ABA video modeling worksheet is most useful when it shows what the learner actually watched, what happened when the skill was attempted, and which conclusions the record cannot support. A note that says “used video modeling” leaves too much unstated. The clip may have shown the whole task or one step, the model may have been the learner or another person, the file may have changed, and a staff prompt may have occurred before anyone captured the initial response.

This copyable worksheet keeps the video plan, viewing events, performance probes, step data, maintenance checks, implementation fidelity, and client response in separate evidence layers. A qualified clinical team still has to decide whether any form of video modeling belongs in the authorized plan. The fictional example tests the form and its arithmetic; it is not a recommended dose, criterion, task, or privacy workflow.

Clinicians & ABA Professionals / Assessment and Treatment Planning.

Clinical boundary: Confirm the current plan, authorization, recording permissions, access supports, and local privacy and security rules before making or showing a recording. Pause when the learner withdraws assent, the video is inaccessible, the file or task version is uncertain, or distress or another unwanted effect appears. A completed form cannot establish medical safety, consent, payer coverage, skill acquisition, generalization, maintenance, or treatment effectiveness. The BACB ethics resources provide professional context, but case-specific review remains necessary.

The intervention and the evidence are different things

Video modeling generally shows a recorded demonstration before the learner has an opportunity to perform the target. Point-of-view video shows the task from the performer's perspective. Video self-modeling shows edited or selected footage of the learner. Video prompting usually presents a segment and then pauses for the corresponding action. Those formats create different exposure histories and implementation requirements; their labels should not be used interchangeably.

A meta-analysis of functional-living-skill studies found overall moderate effects for the populations and measures represented, while also identifying gaps and the need for more single-case experiments. In a small point-of-view video-modeling study, two young men met the study's skill criteria but did not retain the same criterion at a one-month follow-up, although performance remained above baseline. Those findings support direct measurement of the target skill and later maintenance probes rather than assuming that a clear clip or strong acquisition session settles the question.

The current BCBA Test Content Outline includes designing and evaluating modeling procedures, prompting and fading, generalization, maintenance, and data-based decisions about integrity and effectiveness. The BACB outline access page identifies the current certification outlines. These are useful terminology anchors, not an intervention recipe.

Purpose, format, and task version

The planning card starts with a meaningful use of the skill. “Increase independence” is too broad to reconstruct the reason for the work. Name the routine, setting, relevant people, accepted response forms, ordinary supports, and the observable performance that matters to the learner. A task analysis may be embedded in a whole-task video, divided across prompts, or used only for scoring; the form should identify which arrangement was actually selected.

Blank planning card

FieldCase-specific entryLearner-selected or agreed purposeTarget routine, setting, and meaningful outcomeObservable start and end of the performanceTask-analysis ID, version, and effective dateVideo format: whole-task / point-of-view / self-model / prompting / otherWhy this format fits this learner and contextModel, model role, and consent or authorization statusExact video file ID, version, checksum or controlled locationClip length, segments, captions, language, audio, speed, and display sizeOrdinary AAC, sensory, motor, visual, hearing, and processing supportsViewing opportunity and performance-probe sequenceInitial-response window and independent-performance definitionPrompt, correction, replay, and consequence rulesGeneralization and maintenance questionsAssent, dissent, pause, stop, and re-entry signalsClinical owner, privacy/security owner, approval date, and next review

Do not remove AAC or another ordinary accommodation to create an artificial independence test. ASHA's AAC Practice Portal describes AAC as multimodal and individualized. The worksheet should say which supports are part of ordinary performance and which added cues count as prompts under the case definition.

Recording identity, permission, and access

A video can include the learner, peers, caregivers, staff, the service setting, voices, names, screens, schedules, or other identifying details. HHS's Privacy Rule overview explains that the rule protects individually identifiable health information held by covered entities and requires safeguards and limits on uses and disclosures. It does not tell a particular practice that a specific recording workflow is permissible. The responsible privacy and legal reviewers must determine the applicable authorization, minimum-access, storage, sharing, retention, deletion, and breach-response requirements.

Video asset register

Video ID/versionFormat and modeled stepsPeople or protected details depictedPermission/authorization referenceAccessibility featuresApproved device/locationAccess listRetention/deletion ruleStatusdraft / approved / held / retireddraft / approved / held / retired

Store a durable file identifier rather than “tablet video.” If a caption, crop, voice-over, playback speed, segment boundary, model, or task step changes, create a new version. Retire the old version without erasing which sessions used it. A consent discussion, learner assent, organizational permission, and a legal authorization may answer different questions; the register should not collapse them into one checkbox.

Viewing and performance event ledger

Seeing the full clip is an implementation observation. Performing the skill afterward is a learner-performance observation. A replay can change the exposure condition. A prompt can produce successful completion without making the initial response independent. The event ledger preserves those distinctions.

Blank event record

Date/timeContextVideo ID/versionPlanned viewingActual viewing/segmentsAccess worked?Probe attempted?Initial responseIndependent under definition?Prompt/correction/replayFinal outcomeValid? reasonAssent/dissent or unwanted effectObserveryes / no / not scorableyes / no / not scorableyes / no / not scorable

A device failure, wrong file, interrupted task, unavailable material, inaccessible audio, unplanned staff cue, or changed response definition can make a probe invalid. That event still matters for implementation review, but it does not belong in the valid skill-performance denominator. A valid incorrect or incomplete performance remains client data and should not be rewritten as an implementation error.

Step record for a valid probe

Whole-task completion can conceal which steps are independent, prompted, omitted, completed out of sequence, or unavailable. A step record is useful when those differences matter to the clinical question. It should use the versioned task analysis named in the planning card.

Probe IDStep ID/versionStep available?Initial actionIndependent?Prompt or correctionCompleted after support?Error or variationLatency/duration if relevantNotesyes / no / not scorableyes / no / not scorableyes / no / not scorable

Count step opportunities only within valid probes and only when the step was available. Do not substitute step accuracy for the whole-routine decision rule. A learner might independently complete most steps yet need support at a safety-critical step, or meet a routine-level rule while performing some steps through an accepted variation. Keep both views.

Viewing fidelity and implementation integrity

Fidelity measures what the implementer did relative to the current plan. It does not measure learner skill. A study of individualized video modeling for staff implementation found improved but variable integrity with video modeling alone; adding feedback increased performance for the three participating teachers. That small study does not establish a universal staff-training package, but it illustrates why a video should not be assumed to implement itself.

Blank fidelity checklist

EventCorrect video/versionPlanned segment and playbackAccessibility and ordinary supportsInitial response captured before helpPrompt/replay rule followedClient response and validity recordedCorrect componentsOpportunitiesNotes

Possible fidelity components must be operationalized for the specific plan. “Played video correctly” is not enough if volume, segment boundaries, captions, proximity, or response timing could change the exposure. A mismatch should lead to an implementation question, not an automatic judgment about the learner.

Review views and denominator rules

The summary page keeps planned coverage, valid performance, step accuracy, and fidelity apart. It also preserves prompted completion as useful but different information.

Review questionNumeratorDenominatorResultWhat it does not showValid probe coveragevalid performance probesplanned performance probesskill accuracyIndependent routine performancevalid probes meeting the independent rulevalid performance probesstep pattern or maintenancePlanned-probe yieldvalid probes meeting the independent ruleall planned probescause of invalid probesIndependent step performanceindependently completed available stepsavailable steps in valid probeswhole-routine qualityViewing fidelitycorrectly implemented componentsobservable component opportunitieslearner outcome

If the video was viewed eight times and the skill was probed five times, neither eight nor thirteen is a valid performance denominator. If a prompt follows the initial response, retain the supported final outcome without changing the initial score. The ABA video modeling worksheet should let a later reviewer recompute every percentage from visible counts.

Fictional community-lunch example

This example uses a fictional eight-step lunch-packing routine. The client agreed that preparing a preferred lunch for a community program was meaningful. The example's point-of-view file, response definition, viewing arrangement, and numbers are invented. They must not be copied into a real plan.

Eight performance probes were planned. One was unscorable because the device stopped during playback and the staff member substituted an unplanned live demonstration. Seven probes remained valid. Five met the fictional independent whole-routine rule. The record reports:

  • valid probe coverage: 7 / 8 × 100 = 87.5%;
  • independent performance among valid probes: 5 / 7 × 100 = 71.4%; and
  • independent yield across all planned probes: 5 / 8 × 100 = 62.5%.

Across the seven valid probes, the eight-step task analysis created 56 available step opportunities. Forty-five were completed independently, so independent step performance was 45 / 56 × 100 = 80.4%. The step result is not a replacement for the five-of-seven routine result. The invalid probe contributes to planned coverage but not valid performance or step accuracy.

The team also reviewed 16 viewing/probe events against six fidelity components. That created 96 component opportunities. Eighty-nine were correct and seven were mismatches, reconciling to 96. Component implementation was 89 / 96 × 100 = 92.7%. The 92.7% describes staff implementation, not the learner's skill.

Fictional decision record

Evidence layerFictional resultBounded interpretationPlanned probe coverage7 valid of 8 plannedOne planned observation is unresolvedWhole-routine independence5 of 7 validDescribes these valid probes onlyStep independence45 of 56 available stepsDoes not identify which step needs reviewViewing/probe fidelity89 of 96 componentsImplementation was not perfectGeneralizationnot testedNo claim across people, materials, or settingsMaintenancenot yet testedAcquisition-period performance cannot answer durabilityClient responsefictional acceptance varied by sessionCompletion is not assent

No causal conclusion follows from the summary alone. A decision note could call for review of the device failure, step-level pattern, unwanted effects, or a later probe. It should not silently change the video, task analysis, criterion, or prompting plan.

Maintenance, generalization, and change control

Generalization can concern people, materials, settings, cues, or response forms. Maintenance concerns performance after time or after teaching supports change. A new lunch container, different kitchen, delayed probe, or absent video may answer different questions. Label the dimension before collecting the observation.

Probe IDQuestionDeliberately changed dimensionDimensions held constantVideo available?Valid?Independent resultPrompts/supportClient responseInterpretation limitmaintenance / person / material / setting / cue / response

Create a new plan version when the modeled steps, response definition, accepted variations, recording, segment sequence, access supports, scoring rule, or intervention procedure changes. Link the new version to the old evidence. The CASP practice-guideline access page offers current professional context for qualified, individualized care; it does not validate this worksheet or determine whether a revised arrangement is appropriate.

Decision and exception record

FieldReview entryEvidence reviewed and date rangeVideo and task-analysis versions representedValidity or fidelity concernsLearner feedback, assent/dissent, burden, preference, or unwanted effectsStep pattern and possible alternative explanationsGeneralization or maintenance evidence still absentPrivacy, security, permission, or access issue requiring separate reviewHold / revise / continue / assess another option questionQualified decision owner and next review date

The Ethics Code for Behavior Analysts retains competence, client rights, confidentiality, documentation, and data responsibilities. This public artifact is not evidence that those duties were met. Preserve the decision owner, evidence considered, disagreement, unresolved questions, and version history rather than converting a percentage into an automatic action.

Related resources

Sources

The literature spans different participants, skills, video formats, settings, and measures. It supports explicit implementation and direct performance measurement, while leaving the appropriateness and effect of any particular plan to individualized evaluation.