Imitation training

Imitation is defined by a model-response relation

In an imitation trial, another person or a recorded model demonstrates an action. The learner performs an action with relevant similarity after seeing it, and a history of reinforcement for matching models helps explain that response.

The definition requires more than two people moving alike. They might respond to the same music cue, written instruction, or environmental event. A clinician needs a comparison showing that the modeled action changes what the person does.

Start with a reason to teach it

Imitation can support learning actions that the person wants to use in recreation, work, daily routines, communication, or community participation. Examples include copying a new chord shape, following a dance warm-up, reproducing a repair step, or joining a chosen group activity.

Assessment should ask whether imitation is the best route. Written directions, pictures, video, physical setup, or direct instruction may be clearer and less effortful. A different learning route can be the better accommodation.

Models need clear boundaries

Define the model, response, timing window, similarity criterion, and eligible opportunity. “Do what I do” is too broad for reliable measurement.

For a drum pattern, the criterion might specify the same two surfaces in the same order within five seconds. It may allow either hand if laterality is irrelevant. If exact hand position matters for safe tool use, that feature belongs in the definition.

Teaching often moves from familiar to varied actions

A clinician may begin with actions already in the person’s repertoire, provide a clear model, wait, and use an assessed prompt if needed. Reinforcement should fit the activity and the person’s preference. Prompts then fade.

Varied models test whether the person responds to the modeled action rather than memorizing a small set. Novel probes should be reserved for actions that are safe, accessible, and not previously reinforced as exact targets.

Generalized imitation is a broader outcome

Generalized imitation refers to imitating new modeled actions without direct teaching of every action. Success on several trained examples does not establish that broader repertoire.

Probe untrained actions across people, materials, and settings. Keep probe results separate from teaching trials. A novel response can also reflect a familiar instruction, routine cue, or physical affordance, so the model-control interpretation still needs care.

Imitation differs from observational learning

Imitation involves correspondence with a model’s action. Observational learning is broader: a person’s behavior changes after observing what happens to someone else, and the later response need not copy the observed response exactly.

The current BACB Test Content Outline explicitly distinguishes imitation from observational learning and includes modeling procedures. It is examination content rather than an individualized protocol.

A fictional recreation example

Talia chooses a goal of joining a community hand-drumming warm-up. The team defines eight eligible modeled patterns, each containing two strikes. During baseline, she reproduces the defined pattern after 2 of 8 models.

Teaching uses her preferred drum, a side-by-side model, a five-second wait, and a visual sequence when requested. Across eight later opportunities, Talia reproduces 6 of 8 independently, requests the visual on one, and declines one. The declined opportunity remains reported and is not scored as an error.

The phase difference supports continued assessment. It does not isolate imitation training because familiarity, visual access, the instructor, and practice changed together.

Access matters more than topographic neatness

Motor range, pain, fatigue, vision, hearing, sensory load, and assistive technology can change what counts as accessible correspondence. Consult the person and relevant professionals before selecting the response form.

Avoid goals aimed at suppressing harmless movement, forcing eye contact, manufacturing facial expressions, or making someone appear less disabled. A model can be visible without requiring gaze at a face. Preserve AAC, breaks, mobility supports, and a reliable way to stop.

Measure the learner and the teaching system

Report independent imitations divided by eligible models, prompted responses by prompt level, latency, declines, inaccessible models, and untrained probes. Also measure model clarity, wait time, correct prompt delivery, reinforcement, and whether required supports were present.

The ABAI basic-principles page provides a broad training context for behavioral principles. Qualified clinical judgment, client involvement, and actual performance remain necessary.

Questions families can ask

• What useful outcome did the person choose?
• Which features of the action truly matter?
• How were motor and sensory access assessed?
• Are prompted and independent responses reported separately?
• Which untrained models test generalization?
• How does the person pause, decline, or choose another format?