Changing-criterion design

The criterion changes while the intervention continues

A changing-criterion design, or CCD, is useful when a measurable dimension of an existing response can change gradually. The target might be frequency, duration, latency, accuracy, or magnitude. After baseline, the intervention remains available while the required or planned response level changes across phases.

For example, a sequence may move from baseline to criteria of 10, 15, 20, and 25 minutes. If responding settles near each value only after that criterion begins, the repeated correspondence supports a functional relation. The WWC Single-Case Design Technical Documentation places changing-criterion designs within the family of designs that use structured repetition to support inference.

The criterion and the measured outcome are separate. A line drawn at 15 minutes records the planned value. Direct observation records what the person actually did. Meeting the criterion cannot be scored as the outcome because that hides magnitude, variability, and near misses.

Choose the response and criterion range carefully

The response should already occur within the person's repertoire and be able to move in controlled steps. Skill acquisition with abrupt, durable learning may fit another design. Outcomes with delayed feedback, long carryover, medical instability, a hard floor or ceiling, or unacceptable risk can also weaken the design.

Use baseline evidence, direct client input, health and contextual information, and qualified judgment to set the first criterion. Each shift should be:

• attainable enough to remain acceptable and feasible
• large enough to distinguish from ordinary variability
• within safe and clinically justified limits
• defined in the same unit and observation window as the outcome

A single-point criterion may be too rigid when ordinary variation is expected. A range-bound CCD defines an acceptable band, such as 12 to 15 minutes. State how values inside, above, and below the band will be interpreted.

Predeclare when and how criteria will shift

Define the rule before data collection. A phase might change after a stated number of observations within the criterion range, after a visual-analysis decision with named features, or according to a randomized window. Record any departure and its reason.

Avoid moving the criterion after seeing each outcome merely to keep the line near the data. That makes the criterion follow the response and reverses the intended logic. Phase duration can vary, but the reason should be visible.

A review of changing-criterion best practices emphasizes criterion magnitude, phase length, number of shifts, and variation across shifts. It found wide variation in published implementation and presents its recommendations as practices to strengthen inference, not fixed clinical cutoffs.

Bidirectional changes strengthen the test when appropriate

Progressive steps in one direction can coincide with practice, maturation, or another gradual event. A safe temporary shift toward an earlier criterion, often called a mini-reversal, tests whether responding changes in the opposite direction too. Bidirectional correspondence is harder to explain through a simple upward or downward trend.

A mini-reversal is an optional design feature. It should never require a harmful increase or decrease, remove an essential support, or override the person's preference. Variation in criterion size and phase length can also help show that the response follows the planned values rather than a steady calendar trend.

A review of randomized single-case designs describes a randomized range for eligible phase-change timing as one option. Randomization must be planned in advance and remain clinically acceptable. It does not cure weak measurement, poorly chosen criteria, or implementation drift.

Current research standards set a floor, not a treatment rule

The current WWC handbook page identifies Version 5.0 as the current standards. It applies reversal or withdrawal standards to CCDs with a modification: baseline, intervention, and criterion changes count as phase changes, and at least three criterion changes provide three attempts to demonstrate an effect. Its highest research rating ordinarily requires six initial-baseline observations and five or more observations in later qualifying phases, with a zero-variability exception.

These counts govern WWC ratings. A clinical team still needs a phase long enough to interpret level, trend, variability, immediacy, overlap, and correspondence with the criterion. More observations add little when the underlying measure or condition is invalid.

Measure implementation, benefit, and unwanted effects

Define the response, opportunity, observation window, prompt, exclusion, missing-data rule, and criterion range. Graph every observation, criterion line, and phase boundary. Measure whether the intervention and criterion were presented as planned, and report procedural integrity separately from the outcome.

Train and calibrate observers when data drive phase changes. Sample interobserver agreement across conditions. Also track effort, discomfort, injury, fatigue, access, preference, and practical benefit. Criterion matching alone cannot establish social importance or acceptable burden.

The BACB BCBA Test Content Outline, 6th edition covers measurement, graphing, procedural integrity, single-case design logic, changing-criterion designs, and data-based intervention decisions. It is examination content rather than authority to implement a plan.

The current BACB Ethics Code applies to BCBA and BCaBA certificants and people who completed an application. It addresses competence, client involvement, informed consent and assent when applicable, risk, preferences, data, and evaluation. A meaningful withdrawal signal, adverse effect, or change in risk can override the planned next phase.

A fictional changing-criterion example

Jules is a fictional adult who chooses to increase time in a preferred community-gardening activity. A health professional has cleared the activity and limits, and Jules can stop or change the goal. Minutes gardening within a scheduled visit are measured directly. A chosen planning card and end-of-visit feedback form the intervention.

Across six baseline visits, Jules gardens 7, 8, 6, 7, 8, and 7 minutes. Under a 10-minute criterion, five observations are 10, 11, 10, 12, and 10. A 15-minute criterion produces 14, 15, 16, 15, and 15. A safe mini-reversal to 12 minutes produces 12, 13, 12, 12, and 13. A final 20-minute criterion produces 19, 20, 21, 20, and 20.

The response changes with three criterion shifts and moves in both directions. That pattern supports a functional relation for Jules if measurement, integrity, health, and contextual evidence are credible. It establishes neither a standard activity target nor benefit for another person. Jules's enjoyment, fatigue, preference, and real-world feasibility remain separate outcomes.