Changing Criterion Design
- Introduced by
- Single-case methodology tradition; design standards by Kratochwill et al.
- Year
- 2013
- Type
- Within-subject experimental pipeline for shaping behavior toward a goal in graded steps
- Family
- Single-case / single-subject experimental methodology
What does it do?
The changing criterion design (CCD) is a single-case experimental method in which a behavior is gradually shaped toward a terminal goal through a series of stepwise performance criteria.
Instead of expecting a behavior to leap from baseline to its final target, the analyst sets an initial subgoal, reinforces performance that meets it, and then ratchets the criterion up (or down) once behavior stabilizes at each step. Experimental control is demonstrated when the behavior repeatedly and closely tracks each successive criterion change — changing only when, and to the degree that, the criterion changes — so that each new step functions as another demonstration of effect. The design is especially well suited to behaviors that should change incrementally, such as increasing exercise tolerance, reducing cigarettes smoked, or building a new skill in graded approximations. It belongs to the single-case design family codified by Kratochwill and colleagues in 2013, sharing their requirements for systematic manipulation and replicated demonstrations of effect.
The intuition behind the method
Think about helping someone gradually build up the number of minutes they can walk during rehabilitation. Going straight from two minutes to a thirty-minute goal would be unrealistic and might fail, and even if it worked you could not be sure the program caused it. The changing criterion design works like a staircase. You set a first step that is a little above current ability — say five minutes — and reinforce reaching it until performance settles there. Then you raise the bar to ten minutes, then fifteen, each time waiting for stability before stepping up. The persuasive part is the pattern: if walking time rises to roughly five minutes and plateaus, then rises to ten and plateaus, then fifteen, repeatedly matching each new target, it becomes very unlikely that anything other than the program is driving the change. The behavior is, in effect, following the criterion staircase the experimenter built.
History
The changing criterion design was articulated within applied behavior analysis as a way to demonstrate experimental control over behaviors that are shaped progressively, drawing on the shaping and successive-approximation principles of operant conditioning. Early applications targeted health and skill behaviors — such as graduated reductions in smoking and increases in productivity — where reversal designs were impractical or undesirable, and methodologists added the bidirectional criterion change to rule out maturation as a rival explanation. The 2013 What Works Clearinghouse single-case intervention research design standards by Kratochwill and colleagues incorporated the changing criterion design into the broader single-case framework, clarifying how its successive criterion changes map onto the requirement for at least three demonstrations of effect and how it should be appraised alongside reversal, multiple-baseline, and alternating treatments designs.
When to use it
Use the changing criterion design when the target behavior is best changed gradually and already exists in the individual's repertoire, so that it can be shaped upward or downward in steps rather than introduced wholesale. It fits goals like increasing exercise duration, academic productivity, or social initiations, and reducing behaviors such as smoking or caloric intake, where a stepwise approach is natural and where a return-to-baseline reversal would be undesirable or unethical. The design requires a behavior that can be measured repeatedly and that responds to a contingency tied to a performance criterion, and it assumes the behavior can hold at each step long enough to establish stability before advancing. It is a poor choice for behaviors that change all-at-once, that cannot plausibly be partitioned into graded levels, or that drift so much that tracking of a specific criterion cannot be discerned.
Limitations and considerations
- Works only for behaviors already in the repertoire that can be incrementally raised or lowered, not for newly acquired skills.
- Difficult to interpret when the behavior changes gradually and continuously rather than stepping cleanly with each criterion.
- Choosing step sizes and subphase lengths is judgment-laden; poor choices weaken or obscure the demonstration of control.
- If behavior overshoots or fails to plateau at each step, the close criterion tracking that the design relies on is lost.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Demonstrates experimental control while shaping a behavior gradually toward a goal, matching how many skills and habits actually change.
- Avoids withdrawing or reversing an effective intervention to a harmful baseline, since control is shown through criterion tracking.
- Each successive criterion the behavior follows provides another demonstration of effect, building the three-demonstrations evidence base.
- An optional bidirectional criterion change strengthens causal inference by ruling out a simple maturation or trend explanation.
Applications
- Gradually increasing exercise duration or physical activity in rehabilitation programs.
- Reducing cigarette smoking or caloric intake through stepwise contingency criteria.
- Building academic or vocational productivity in graded increments for learners with disabilities.
- Shaping social or communicative initiations upward in small, reinforced approximations.
- Demonstrating that an incremental skill-acquisition program causes change without a return-to-baseline reversal.
How it works
How it works
The pipeline begins by measuring a stable baseline level of the behavior and specifying the terminal goal that defines success. Establishing the baseline documents where the individual starts and provides the reference against which all later progress is judged, while the terminal goal anchors the whole staircase. Because the design assumes incremental change, the analyst confirms that the gap from baseline to goal can sensibly be traversed in steps rather than in a single jump. A clear, repeatedly measured baseline and a well-defined endpoint are prerequisites, since the credibility of later criterion-tracking rests on knowing both where the behavior began and where it is heading.
Formula
y_0 = \text{baseline level},\quad y^{*} = \text{terminal goal},\quad y_0 \to y^{*} \text{ to be approached in graded steps}Next the analyst sets an initial criterion — a first subgoal that lies between the baseline and the terminal goal and that is challenging yet attainable. The size of this step matters: too small and the change may be indistinguishable from baseline variability, too large and the behavior may not reliably reach it, weakening the demonstration. The criterion is the explicit performance standard the individual must meet to earn reinforcement during the first subphase. Choosing it thoughtfully sets up the first opportunity to show that behavior moves to the level specified by the criterion, which is the foundation of the design's logic of control.
Formula
c_1 \text{ with } y_0 < c_1 < y^{*},\quad c_1 \text{ a stretching yet attainable first criterion}During each subphase, reinforcement or the relevant contingency is delivered only when performance meets the current criterion, so the behavior is pulled toward exactly the level the criterion specifies. This contingency is the systematic manipulation that gives the design its experimental status: the experimenter deliberately ties consequences to a stated standard rather than letting behavior drift. Holding the criterion constant within a subphase, and applying the contingency consistently, allows the behavior to stabilize at that step. The clarity and fidelity of this contingency are what make it credible that subsequent changes in behavior are responses to the changing criterion rather than to extraneous factors.
Formula
\text{reinforce iff } y_t \ge c_k\ (\text{or } \le c_k),\quad \text{contingency tied to the current criterion } c_kOnce the behavior has stabilized at the current criterion, the analyst steps the criterion to its next value and repeats the contingency, producing a staircase of increasingly demanding subphases that culminate at the terminal goal. The decision to advance is made only after performance is stable at the present step, because a stable subphase establishes the prediction that behavior would stay there absent a new criterion, against which the next step's change is evaluated. Each successive criterion change that the behavior follows constitutes another demonstration of effect, and the standards' three-demonstrations logic is met when behavior tracks several successive steps, making coincidental explanations implausible.
Formula
c_{k+1} = c_k + \Delta_k \text{ once } y_t \approx c_k \text{ is stable over a subphase};\ c_1 < c_2 < \dots < c_K = y^{*}Finally, the analyst verifies that the behavior is genuinely under the control of the criterion by examining how tightly it tracks each step and, optionally, by introducing a bidirectional criterion change that reverses direction. If the criterion is briefly lowered and the behavior follows it back down, then resumes climbing when the criterion rises again, the design rules out the rival explanation that the behavior was simply improving on its own due to maturation or a monotonic trend. This bidirectional check sharpens the demonstration of experimental control, because only criterion-controlled behavior would move down as well as up in step with the experimenter's standard.
Formula
\text{control} \iff y_t \text{ tracks each } c_k;\ \text{optional reversal } c_{k+1} < c_k \text{ tests bidirectional control}Common pitfalls
- Setting steps so small that criterion changes are swamped by normal variability, or so large that behavior cannot reliably reach them.
- Advancing to the next criterion before the behavior has stabilized, breaking the prediction logic that makes each step a demonstration.
- Relying on unidirectional increases alone, leaving a maturation or monotonic-trend explanation unaddressed when a bidirectional check was feasible.
- Treating loose, drifting tracking of the criteria as a demonstration of control rather than acknowledging the ambiguous evidence.
Frequently asked questions
How does a changing criterion design demonstrate experimental control?
It shows control through criterion tracking. The behavior is reinforced for meeting a stated performance criterion, and once it stabilizes at that level the criterion is stepped to a new value. If the behavior repeatedly changes to match each new criterion — and only when the criterion changes — it becomes implausible that anything other than the contingency is driving it. Each successive criterion the behavior follows acts as another demonstration of effect, satisfying the single-case standards' requirement for at least three demonstrations within the study.
What is a bidirectional criterion change and why use it?
A bidirectional criterion change temporarily reverses the direction of the criterion — for example, lowering a target the behavior had been climbing toward — to test whether the behavior follows it down as well as up. If the behavior tracks the reversal and then resumes when the criterion is raised again, this rules out the alternative explanation that the behavior was simply improving on its own through maturation or a monotonic trend. The bidirectional check is optional but greatly strengthens the causal claim of the changing criterion design.
When is a changing criterion design preferable to a reversal design?
It is preferable when the behavior should change gradually and when returning it to baseline would be undesirable or unethical — for instance, you would not want to reverse gains in exercise tolerance or reductions in smoking. The changing criterion design demonstrates control without a harmful reversal, by showing the behavior tracks a staircase of criteria. A reversal design is better when the behavior can ethically return toward baseline and you want to isolate a single intervention's on-off effect rather than shape behavior incrementally.
Related methods
Sources
Kratochwill, T. R., Hitchcock, J. H., Horner, R. H., Levin, J. R., Odom, S. L., Rindskopf, D. M., & Shadish, W. R. (2013). Single-case intervention research design standards. Remedial and Special Education, 34(1), 26-38.
Open DOI referenceCopied.