Indiscriminable contingency

Maintenance means performance after teaching changes

A skill is maintained when it continues after teaching has ended, been reduced, or changed. Frequent immediate reinforcement can establish a response, while everyday environments often provide outcomes less consistently.

An indiscriminable arrangement can help bridge that gap. The person cannot reliably identify one response after which support disappears, so responding may persist across leaner conditions.

Several arrangements can reduce predictability

Intermittent reinforcement delivers a consequence after some responses or intervals. Delayed reinforcement separates the response and outcome in time. Varied outcomes reduce dependence on one specific item. Natural consequences place the skill under everyday contingencies.

These procedures differ. Record which feature changed rather than treating “indiscriminable” as one standard schedule.

Transparency and predictability can coexist

A person can know the general arrangement without knowing the exact next reinforced response. For example, “Points come at different times during the activity, and you can check the total afterward” is transparent about the rule while leaving timing variable.

Avoid false promises, hidden losses, fabricated scarcity, or changing criteria after a response. Trust is a clinical outcome worth monitoring.

Thin from a stable baseline

Define the current schedule, response, consequence, mastery evidence, generalization, preference, and side effects before thinning. Change one meaningful feature at a time when interpretation matters.

Predeclare advance, hold, return, and stop criteria. The BACB BCBA Test Content Outline includes maintenance, schedule thinning, naturally occurring reinforcers, and mitigation of unwanted effects.

A fictional maintenance example

Samira chooses a goal of checking a visual work plan and requesting clarification when needed. During acquisition, feedback follows every completed check. After stable use, the schedule changes across three planned steps.

At the second step, Samira uses the plan independently in 11 of 12 eligible transitions. Clarification requests receive a useful partner response in 4 of 4 opportunities. During six maintenance probes one week later, independent use occurs in 5 of 6.

Report all three denominators. The data describe performance under the sequence and cannot isolate which schedule feature caused maintenance.

Keep natural reinforcers real

Many useful responses produce ordinary outcomes: a request gets information, a schedule check clarifies what comes next, and a repaired misunderstanding improves communication. Arrange teaching so those consequences can take over when possible.

Do not withhold a needed answer, communication access, safety support, or basic need merely to create intermittency. Some consequences should remain reliable.

Watch for ratio strain and relapse

Rapid thinning can produce longer pauses, lower response rates, errors, avoidance, distress, or abandonment of the task. Track these alongside the target response.

Maintenance can weaken after long delays, context changes, illness, staff turnover, or loss of the natural outcome. Plan booster teaching and a clear route back to denser support.

Measure both sides of the contingency

For the person's response, track eligible opportunities, independence, latency, accuracy, and maintenance. For the environment, track scheduled and delivered reinforcement, delay, magnitude, partner response, and procedural integrity.

A high client score with unreliable partner follow-through can conceal a fragile system. Report missed outcomes as implementation events rather than removing those trials silently.

A planning checklist

• Which meaningful response should maintain?
• What consequence currently supports it?
• Which aspect will become less predictable?
• What general rule will the person understand?
• Which consequences must remain reliable?
• What data permit advancement or require return?
• How will preference, distress, burden, and trust be checked?
• When will natural-setting and delayed maintenance probes occur?

The smallest effective change is often easiest to interpret and repair.

Build a maintenance ledger

Create one row for each schedule phase. Record the programmed schedule, obtained consequence rate, delay, outcome type, implementer, setting, eligible opportunities, independent responses, prompts, errors, pauses, and withdrawal signals. Add maintenance probes with their exact delay from teaching.

Review the ledger before every thinning step. If client performance stays strong while implementer delivery falls, repair implementation before making the schedule leaner. If performance falls while delivery matches plan, return to the last effective step and inspect task effort, outcome value, natural reinforcement, access, and recent context. Version changes instead of rewriting the old schedule.

Pair quantitative data with a brief check-in from the person. Ask whether the arrangement is understandable, fair, and worth continuing. A maintained response accompanied by rising frustration may still signal poor fit.

Keep acquisition, thinning, maintenance, booster, and natural-context phases visually distinct in the graph and permanent record.