Assess Generalized Imitation With Novel Models

To assess generalized imitation with novel models, separate Paloma's reinforced training examples from probe models that lack direct training at the relevant response level. Define novelty, response-class boundaries, access, and consequence arrangements before testing. Balance difficulty and preserve probe integrity. Report matching within each response type and across types. Success with familiar actions or one novel topography cannot establish a universal imitation repertoire.

Define novelty at the right level

A new video of a trained act is a novel stimulus, while the response may remain trained. A new act sharing all essential components may be a within-class probe. State which dimension is untrained.

Separate training and probe records

Mark each model as trained, previously observed, familiar outside the program, or probe. Record direct reinforcement, correction, prompts, and later reuse so probe status can change over time.

Sample response classes deliberately

Include safe examples across relevant object, gesture, vocal, movement, sequence, live, and video classes only when they matter to Paloma's goals. Avoid treating results from one class as universal.

Protect access and comparability

Match visibility, duration, materials, complexity, motor demand, opportunity, and response window as closely as the decision allows. Record any meaningful mismatch.

Use this sequence to assess generalization

Define the useful repertoire, audit prior exposure, create trained and probe sets, verify access, lock consequences, record each trial, analyze by response class, and state the narrowest supported conclusion.

Build Paloma's novel-model generalization probe

Create one versioned novel-model generalization probe for the community garden tool lesson. Include Paloma's priority, model and response definitions, relevant matching dimensions, equivalent forms, trained or novel status, opportunity and clock rules, access, prompts, rules, task cues, direct outcomes, observed model outcomes, context, history, client experience, choice, invalidity, missingness, withdrawal, alternative control, evidence limits, qualified owner, correction, and reassessment trigger. Store only decision-relevant information with role-limited access.

Validate Paloma's evidence

Reproduce 18 trained trials, 12 novel probes, 16/18 trained matches, 8/12 probe matches, 6/6 hand-tool probes, and 2/6 whole-body probes. Verify novelty evidence, response classes, probe consequences, access, prior exposure, order, carryover, and individual trial history.

Connect Paloma's evidence to a decision

Paloma's result can guide whether more varied models, direct teaching, access repair, or a different response class deserves attention. The clinician should avoid programming novelty merely to test a label when Paloma already has an effective way to learn the useful task.

Work through Paloma's example

Paloma receives 18 trained-model trials and 12 novel probes. She matches 16/18 trained acts and 8/12 probes. Probe matching is 6/6 for new hand-tool actions and 2/6 for new whole-body actions. The data support generalization within the sampled hand-tool class and weaker evidence across the whole-body class. Preserve every planned and valid opportunity, model version, accessible presentation, response, equivalent, prompt, rule, task cue, consequence, timing, access and health state, withdrawal, invalid event, correction, and unresolved item. This fictional example demonstrates one assessment control. It supplies no diagnosis, universal imitation hierarchy, treatment effect, legal conclusion, coverage decision, payment promise, or outcome guarantee for Paloma.

Address Paloma's main interpretation risk

Calling every unreinforced trial novel ignores prior demonstrations, everyday experience, or shared components with trained actions. Pooling 8/12 hides the response-type boundary. Delivering correction or reinforcement after a probe can change its status for later presentations. Review model access, matching criteria, motor and sensory demands, communication, prompts, rules, task products, direct outcomes, prior learning, context, health, fatigue, preference, direct experience, and design strength separately. A convincing sequence cannot make an unsafe, inaccessible, unwanted, unauthorized, or irrelevant imitation goal clinically appropriate.

Keep scope and authority clear for Paloma

For Paloma's novel-model generalization probe, the CASP public summary supplies high-level ABA behavioral-health-treatment scope for autistic people, with licensed detail outside this page. The current BACB Ethics Code addresses competence, client involvement, consent and assent when applicable, assessment, medical needs, risk, data, documentation, and evaluation for covered people. The BACB outline is examination content. These sources supply no individualized imitation assessment, motor diagnosis, legal authority, or requirement that Paloma copy another person.

Use generalized-imitation studies narrowly for Paloma

Young and colleagues studied four children with autism using reinforced training and nonreinforced probes across vocal, toy-play, and pantomime response types. Imitation generalized within trained response types and did not generalize across types in that experiment. A three-participant facial-model study reported generalization to unreinforced facial probes for two participants and inconsistent probe responding for one. These small studies show why Paloma's trained examples, probe history, response classes, and consequences must remain visible. They supply no universal probe ratio, mastery criterion, or prognosis.

Separate monitoring and later learning for Paloma

Taylor, DeQuinzio, and Stine evaluated three children with autism in a sight-word arrangement. Test performance was more accurate when the study required monitoring responses related to a peer's response than when participants only encountered the peer's response. The authors described the analysis as preliminary. The result motivates measuring accessible monitoring and later performance separately for Paloma; it does not establish a universal component sequence or prove learning from another person's consequence in every context.

Expect model and format effects to vary for Paloma

Marcus and Wilder compared peer and self video models while teaching textual responses, and later conversation performance differed by child. An older peer-model study involved four children and five discrimination tasks; correct responding increased after peer modeling in that specific design. These studies support documenting model, task, media, and person-level results. They do not establish that a peer, self, live, or video model is generally superior for Paloma.

Preserve assent and communication access for Paloma

Breaux and Smith propose individualized assent and withdrawal practices while describing an evolving evidence base. Their paper is practice guidance rather than a separate BACB mandate. ASHA's AAC portal says AAC users should always have access to their tools or devices. Give Paloma an accessible way to ask, correct, accept, pause, decline, or withdraw. Do not require eye contact, speech, copying, or removal of communication access as the price of participation.

Choose Paloma's next bounded action

Paloma asks to keep the hand-tool demonstrations and receive direct instruction for balance-heavy whole-body actions. Future probes use genuinely new, safe actions and a documented exposure history. Record the qualified owner, source evidence, effective date, current model and task version, ordinary supports, access and health state, implementation check, accessible explanation, disagreement route, and reassessment trigger. Preserve the earlier record when models, responses, settings, systems, equipment, consequences, health, or priorities change. A repaired arrangement creates a dated phase rather than an error in Paloma's prior performance.

Close Paloma's assessment

Review the novel-model generalization probe with Paloma, the qualified behavior analyst, relevant partners, access owners, and specialists named in the manifest. Confirm that model exposure, accessible monitoring, immediate matching, later performance, observed-consequence learning, generalization, prompt or rule control, direct outcomes, and theoretical hypotheses remain separate; every denominator is reproducible; AAC and basic access remain protected; urgent needs received action; and conclusions stay bounded to sampled conditions. Keep this page draft and noindex until every required review is complete.

How to Assess Generalized Imitation With Novel Models