Maintenance

Maintenance asks whether change lasts

A teaching criterion describes acquisition; a maintenance check measures later performance after a named delay and teaching change. Ordinary cues, access, practice, and consequences may support the response.

Maintenance never requires removing useful access. Record AAC, visual, sensory, and mobility supports separately from plan-defined prompts. Follow-up results apply only to sampled conditions; a lower score invites assessment rather than a judgment about effort, compliance, or worth.

Maintenance, mastery, generalization, and independence differ

A later probe in a new setting samples both time and setting variation. Report maintenance and generalization separately rather than merging them as “mastered.”

How maintenance is planned

A useful plan names the response and outcome, criterion date, teaching components that continue or change, first and later checks, measure, prompts, supports, exclusions, and action threshold.

Intervals should fit the decision: two days cannot answer what happens after two months. Rare natural opportunities may need longer observation; safety-related responses may need closer qualified oversight. Do not manufacture opportunities that distort the skill.

The BACB BCBA Test Content Outline (6th ed.) includes response maintenance and procedures intended to maintain desired change, such as schedule thinning and transfer to naturally occurring reinforcers. This is entry-level examination content, not a universal interval or fading plan.

A fictional measurement example

Diego is a fictional sixteen-year-old who chooses a goal of asking a supervisor for clarification during a volunteer shift. Before teaching, the team defines an eligible opportunity as a naturally occurring instruction missing a task, place, or timing detail that the written volunteer protocol identifies as necessary; repeated wording of the same instruction counts once. The predeclared acquisition criterion is an independent request by speech or typed message in 8 of 10 total eligible opportunities across four shifts. Diego meets it, and rehearsal of the exact scenarios ends.

Four weeks after the criterion was met and rehearsal ended, five eligible opportunities occur. Diego independently asks for clarification in four; two of those four requests use his phone-based communication tool. In the fifth opportunity, he asks after a verbal prompt. The learner result is an independent clarification request in 4 of 5 eligible maintenance opportunities (4/5, 80%) at four weeks, plus one prompted request.

The supervisor provides the defined clarification within the plan's response window after all five requests. The partner result is 5 of 5 emitted requests received the defined partner response (5/5, 100%). The learner metric uses eligible opportunities as its denominator; the partner metric uses emitted requests. The denominators happen to equal five here because Diego requested clarification in every opportunity.

Maintenance data need context and denominators

“The skill maintained at 80%” hides the most useful information. A quality metric should include:

• Numerator: eligible follow-up opportunities with the response at the defined support level
• Denominator: all eligible follow-up opportunities in the stated window
• Interval: elapsed time from one named anchor, such as the date the criterion was met or the last active teaching session; use the same anchor when comparing checks
• Conditions: people, setting, materials, instructions, and ordinary supports sampled
• Exclusions: predeclared invalid opportunities, reported by reason; describe missing expected supports separately rather than silently removing those opportunities
• Decision rule: what action follows stable, improving, or declining performance

Report independent and prompted responding separately and preserve raw counts beside percentages. If no eligible opportunity occurs in the planned window, report 0 eligible opportunities; rate not calculated. Do not record 0% or claim that maintenance was or was not demonstrated. Extend or redesign observation only when clinically appropriate and safe.

Segment by condition when aggregation hides a problem. For example, 9 of 10 overall (90%) can combine 9 of 9 at the clinic (100%) with 0 of 1 in the community (0%). Those unequal denominators and the missing community communication support belong in the report.

McDougale, Richling, Longino, and O'Rourke descriptively coded 157 included skill-acquisition articles published from 2015 through 2017 in three selected behavior-analytic journals. Twenty-six percent did not report a specific mastery-criterion method, 58% did not report maintenance probes, and 1% were unclear about maintenance probes. This was a review of reporting practices, not an experimental test of one best mastery criterion or follow-up interval. It supports transparent criteria and direct follow-up measurement.

Podlesnik, Kelley, Jimenez-Gomez, and Bouton reviewed basic and translational renewal research, much of it involving nonhuman subjects and extinction arrangements. Renewal is the recurrence of previously reduced behavior after a context change; it is not the same outcome as retention of a newly taught useful response. The review supports planning for context-related relapse when relevant. It does not predict an individual's outcome or show that every context change will affect skill maintenance.

Maintenance should protect useful access

The current BACB Ethics Code for Behavior Analysts applies to BCBA and BCaBA certificants and people who have completed an application for either certification. The BACB states that it does not have separate jurisdiction over organizations or corporations. For behavior analysts within its jurisdiction, the Code addresses client and stakeholder involvement, informed consent and assent when applicable, individualized intervention, minimizing risk, and continual evaluation.

The public CASP Version 3.0 summary covers ABA assessment and treatment standards for autism; full guideline access requires a license. CASP's resources page adds organizational and practice materials. Neither supplies a universal maintenance schedule.

A plan can be technically durable and still be poorly chosen. Maintenance matters only while the response serves communication, autonomy, safety, participation, relationships, or another person-valued outcome. Follow applicable consent and assent requirements, preserve AAC and sensory, mobility, and visual access, and treat refusal, dissent, or changed preferences as reasons to review the plan. Never create deprivation, unsafe opportunities, or unwanted practice to produce a favorable maintenance rate. Revise or retire goals when preferences, needs, or environments change.

Questions to ask about maintenance

• When was the check anchored, and how had teaching changed?
• What are the numerator, denominator, window, exclusions, and zero-opportunity rule?
• Were plan-defined prompts separated from communication, sensory, mobility, and visual supports?
• Were partner responses and relevant settings reported rather than hidden in one aggregate?
• What will trigger reassessment, teaching, an environmental change, or retirement of the goal as preferences change?