Stimulus fading

The goal is natural stimulus control

An extra cue can make the relevant feature easier to detect at first. Fading systematically reduces that cue while preserving success. The learner should eventually respond to the cue found in everyday use.

If performance only occurs with the added highlight, enlarged icon, or exaggerated sound, the transfer is incomplete. Calling a prompt faint does not show that natural control has developed.

Stimulus and response prompts differ

A stimulus prompt changes or adds something to the instructional stimulus. Examples include a brighter border, larger picture, closer placement, or partially completed word.

A response prompt acts on the learner's response, such as a model, gesture, verbal hint, or physical guidance. Both can be faded, though their dimensions and decision rules differ. The current BACB BCBA Test Content Outline includes selecting stimulus and response prompts and fading them through procedures such as prompt delay and stimulus fading.

Define the final cue first

Write the natural discriminative stimulus before designing the added cue. Ask what the person will encounter outside instruction, which features matter, and which irrelevant features will vary.

For a transit symbol, the final cue may include ordinary size, colors, viewing distance, lighting, and surrounding signs. Teaching against one oversized card can create narrow control by size or position.

Change one measurable dimension

A fading plan works best when each step can be reproduced. “Make the cue less obvious” leaves too much judgment to the implementer.

Specify the dimension and values, such as border widths of 12, 8, 4, 2, and 0 millimeters. State the presentation duration, location, materials, comparison stimuli, response window, reinforcement, and error procedure.

Use performance rules for movement

Predefine when to move forward, hold, or return to an earlier step. A criterion might require at least 8 correct independent responses across 10 eligible trials in two sessions, with procedural integrity above a defined threshold.

Avoid advancing from one unusually strong session. Also avoid staying at an easy step after stable success, which can strengthen dependence on the added feature.

A fictional teaching example

Tariq wants to locate the “messages” icon on a communication device. With Tariq's agreement, the clinician adds a thick orange border around the ordinary icon and keeps every other icon unchanged. The border shrinks across five planned steps.

At step three, Tariq selects correctly in 9 of 10 eligible trials. During three natural-cue probes with no border, he selects correctly in 1 of 3. The team holds the step and changes the practice mix instead of claiming transfer. After further teaching, natural-cue performance reaches 8 of 9 probes across two layouts.

These counts describe this fictional sequence. They do not prove that fading caused the change because practice, reinforcement, layout exposure, and time changed too.

Probe natural control separately

Teaching trials show performance under the current fading step. Independent probes show whether the natural cue works without the added feature. Label the two data streams separately.

Probe across relevant people, materials, settings, distances, and examples. Keep the skill accessible and avoid testing by removing a communication tool, mobility support, or needed accommodation.

Plan for errors and preference changes

When errors rise, check whether the step changed too much, the natural cue is unclear, the response requirement shifted, or motivation and access changed. Use the defined error procedure and return rule.

Honor pauses, withdrawal, discomfort, and changing preferences. Stimulus fading is a teaching arrangement, not permission to continue an unwanted goal.

An implementation record should show

• the socially meaningful target and the person's input
• natural cue and added stimulus prompt
• fading dimension and exact step values
• response definition, opportunity, and prompt code
• advance, hold, return, and stop rules
• teaching data and independent probe data
• procedural integrity and observer agreement when important
• generalization, maintenance, access, and adverse-effect checks

The record should let another trained person reproduce the current step and identify why it changed.

Train implementers against the same examples

Before client use, have each implementer identify the natural cue, current prompt value, next step, response code, and return rule from sample trials. Compare scoring on the same examples and correct disagreements. During teaching, sample whether the planned value was actually presented and whether feedback followed the defined response. Fidelity data can distinguish a learner-performance change from an unplanned jump in cue size or timing.

Plan maintenance after the prompt reaches zero. Schedule checks after ordinary delays and across materials that matter in daily use. If accuracy weakens, inspect stimulus control and access before automatically restoring the most intrusive prompt.