Identify the Defining Features of Single-Case Experimental Designs
If you’ve ever needed to know whether a specific intervention actually works for a specific client—not for a group, but for the one learner sitting across from you—you’ve already sensed why single-case experimental designs matter. SCEDs let practitioners test causality by measuring one person’s behavior repeatedly across phases, systematically changing what you do, and watching to see if behavior changes follow.
The core confusion is this: a graph that looks like it shows improvement doesn’t prove the intervention caused it. Before you scale a strategy up, change a treatment plan, or confidently tell a family “this works for your child,” you need to know the defining features of SCEDs—the structural elements that let you draw a causal conclusion instead of just guessing.
This article walks through what makes an SCED an SCED, why those features matter clinically, and how to recognize them in practice.
What Single-Case Experimental Designs Are
A single-case experimental design is a research method in which one person (or a small group) serves as their own control. Instead of comparing one group receiving an intervention against another group that doesn’t, you measure the same individual’s behavior repeatedly across different conditions—usually a baseline (no intervention) and one or more intervention phases.
The basic logic is simple: if behavior is stable in baseline, then changes reliably when you introduce an intervention, and changes again when you modify or withdraw that intervention, you’ve made a strong case that the intervention caused the change.
SCEDs rely on repeated, systematic measurement. You’re not taking a snapshot or relying on memory. You’re collecting data point after data point across weeks or months, creating a visual record of how behavior unfolds. That repeated measurement lets you see patterns—whether behavior is stable, trending up or down, or jumping around—and compare those patterns across phases.
The goal is to demonstrate experimental control: reliably producing a change in behavior by manipulating what you do, while keeping other things as constant as possible. When behavior changes predictably with your intervention and stays changed when the intervention stays in place, you’ve got evidence of control. That’s not correlation or coincidence—it’s evidence of cause and effect at the individual level.
Why Identifying These Features Matters for Your Practice
Making treatment decisions without experimental control can lead you to continue ineffective interventions, abandon strategies that actually work, or miss the real reason a client’s behavior is changing. If you rely only on intuition or a casual glance at data, you might keep investing time and resources in something that doesn’t work for this particular learner—or give credit for improvement to the wrong factor.
SCEDs protect your clients by ensuring they only receive interventions that demonstrably work for them. This is about more than research rigor—it’s about ethics and resource stewardship. Every session has a cost: time, energy, money, and opportunity. Using SCED logic to validate your approach ensures you’re spending that investment wisely.
For supervisors and clinic directors, understanding SCED features is also about quality assurance. It helps you and your team distinguish between “that graph looks like it’s improving” and “that graph provides evidence of experimental control.” Those are not the same thing. One can mislead you; the other can guide real decisions.
The Defining Features of Single-Case Experimental Designs
What actually makes an SCED an SCED? Here are the core elements:
Repeated, reliable measurement of the target behavior. You measure the same behavior, in the same way, across many time points. This creates a detailed record that lets you spot patterns rather than guessing from a handful of observations.
A clear baseline phase followed by intervention phase(s). Baseline is your reference point—it shows what the behavior looks like without your manipulation. Intervention is when you deliberately change what you’re doing. The contrast between these phases is where causal evidence lives.
Systematic manipulation of the independent variable. You don’t passively observe; you actively change something—a prompt strategy, a reinforcement schedule, a curriculum activity—and you plan when that change happens. This planned, intentional change separates experimental designs from simple observation.
Replication of effects. A single phase change might be coincidence. But if behavior changes again when you change the intervention (or across multiple clients, settings, or behaviors using the same logic), you’ve replicated the effect. Replication moves you from “maybe” to “likely.”
Phase-change logic. Changes in behavior should be tied in time to your intervention. When you introduce it, behavior changes relatively soon after. When you modify or withdraw it, behavior changes again. This temporal alignment is the fingerprint of experimental control.
Visual analysis as the primary interpretation method. Rather than relying solely on statistics, you look at the graph. Does the level of behavior shift between phases? Is there a trend within phases? How much do intervention data points overlap with baseline? These visual patterns tell you whether control was demonstrated.
The Boundary: Not Every Single-Subject Graph Is an SCED
Not every time you track one person’s behavior over time are you conducting an SCED. A descriptive single-case design might involve one person and lots of data points, but without systematic manipulation and phase logic, it’s observation—not experimentation. An SCED requires that you deliberately change what you do and track behavior across those planned changes.
Similarly, SCEDs are not the same as group experimental designs. Group designs average across many people and use statistics to infer whether an effect is “significant” across the sample. SCEDs focus on the individual and use replication and visual analysis.
Both are valid, but they answer different questions: group designs ask “does this usually work?” while SCEDs ask “does this work for this person?”
Common SCED Formats
Here’s a brief overview so the terminology makes sense as you read further.
ABAB reversal designs involve baseline, intervention, withdrawal of the intervention, and reintroduction. The repeated reversal strengthens the causal claim because you’re showing the effect multiple times.
Multiple-baseline designs introduce the intervention at different times across settings, participants, or behaviors. Rather than withdrawing the intervention, you stagger when you introduce it, so you can see that behavior changes only when and where the intervention starts.
Alternating treatments (multi-element) designs rapidly switch between two or more interventions across sessions. This lets you compare which works better without waiting months between phases.
Changing-criterion designs involve stepwise, graduated changes in the criterion for success. You’re not comparing “no intervention” to “intervention”; you’re showing control by meeting increasingly challenging benchmarks.
Each format has strengths and fits different clinical situations. The right choice depends on what you need to know, what’s ethical, and what’s practical in your setting.
How Clinicians Use Visual Analysis to Judge Effects
When you look at an SCED graph, you’re inspecting six key patterns:
Level is the average value within a phase. Does behavior happen more or less often in the intervention phase compared to baseline?
Trend is the direction over time. Is behavior moving up, down, or staying stable within each phase?
Variability is how much behavior bounces around. Are the data points tightly clustered or scattered?
Immediacy is how quickly behavior changes after a phase change. Does it shift right away, or does it take a while?
Overlap is how many intervention-phase data points fall within the baseline range. No overlap strengthens your claim; high overlap suggests the intervention didn’t change much.
Consistency is whether the pattern repeats across multiple replications. If you see the same change each time you use the same logic, that consistency supports your conclusion.
Together, these visual features tell a story. A sharp level change with no overlap, immediate timing, and consistent replication is strong evidence of experimental control. Modest changes with high variability and significant overlap require more caution.
When and How to Use This in Practice
Think about the moments in your work when you need an SCED mindset.
When you introduce a new behavior-reduction strategy and want to know if it’s actually responsible for the change before rolling it out clinic-wide, you’re in SCED territory. When you’re comparing two prompt-fading approaches for one learner and need a quick, evidence-based answer, an alternating-treatments design makes sense. When you’re working with a skill that can’t be reversed (like reading fluency) but you still want experimental evidence, a multiple-baseline design across settings or skills gives you that proof without ever withdrawing a working intervention.
The key is matching the design to the clinical reality. If the behavior is reversible and withdrawal is safe, an ABAB reversal is powerful. If reversing the intervention risks harm or the skill is permanently acquired, reach for multiple-baseline or changing-criterion logic instead.
Baseline Stability: The Foundation
Before you move from baseline to intervention, your baseline data should be stable enough that you can reasonably predict what would happen if you continued doing nothing. If baseline is trending sharply upward or wildly variable, you can’t clearly see what the intervention does. It’s like trying to measure the effect of medication when the patient’s symptoms are already changing on their own.
The rule of thumb: collect enough baseline data points (often 3 to 7 or more, depending on context) that you can see a clear pattern. If you get 3 points and they’re all different, you don’t have stability yet. Keep measuring.
Yes, this takes time and patience, but it saves you from making wrong conclusions later.
Replication: Why One Phase Change Is Not Enough
Imagine a learner’s on-task behavior increases right after you introduce a token economy. That’s encouraging—but it could be coincidence. Maybe that week was just a good week. Maybe something else changed. You need replication to rule out luck.
Within-subject replication means you show the effect happens repeatedly for the same person (like in ABAB). Between-subject replication means you show the same logic works for multiple learners or across multiple settings. Each replication strengthens the case that the intervention, not chance, caused the change.
Common Mistakes and Pitfalls
One of the most frequent errors is treating a short time series with one phase change as definitive proof. It’s not. You need planning, stability, and ideally replication.
Another mistake is ignoring baseline instability and forging ahead anyway. If your baseline is all over the place, you genuinely cannot see what the intervention does. It’s not laziness to extend baseline—it’s scientific integrity.
A third pitfall is choosing the wrong design for the situation. Selecting an ABAB reversal when withdrawal would cause harm or when a skill is irreversible puts you in an ethical bind. Multiple-baseline and changing-criterion designs exist precisely to give you experimental control without those risks.
Finally, watch out for misreading variability. A noisy graph doesn’t mean there’s no effect—it means there’s variability. Replication and clear phase logic can show effects even with messy data. The key is whether behavior consistently changes in line with the phase changes.
Ethical Considerations and Client Welfare
Using SCED logic responsibly means keeping client safety and dignity at the center.
If you’re using a reversal design, you’re withdrawing an intervention that may be working. If the behavior is dangerous or related to safety, withdrawal creates risk. In those cases, a different design is not just preferable—it’s ethically required. You might use a B-A-B approach in crisis situations, or design a multiple-baseline that never requires withdrawal.
Informed consent is also critical. Clients and families deserve to understand what phase changes will happen, why you’re measuring so closely, and what benefit they can expect. This isn’t just a compliance box; it’s part of respecting agency.
Similarly, social validity matters. The intervention should address goals that matter to the client and family, not just what’s easy to measure. And your data-collection methods should be respectful of privacy and not unduly burdensome.
Practical Decision Points
Choosing a design comes down to a few questions:
Is withdrawal safe and ethical? If yes, reversal designs are strong. If no, choose multiple-baseline, alternating-treatments, or changing-criterion.
Can the behavior be reversed? Skills like reading or math fluency, once learned, don’t typically go backward. For acquired skills, multiple-baseline is usually better than reversal.
How many tiers (subjects, settings, behaviors) can you access? Multiple-baseline requires at least two or three tiers to be credible. If you only have one learner and one setting, reversal or changing-criterion may be your option.
How quickly do you need an answer? Alternating treatments are faster than extended baseline plus phases. Reversals take longer but build stronger evidence.
FAQ: Questions You Might Have
How many baseline data points do I need? Often 3–5 is a starting point, but 6–7 is preferred for confidence. If your baseline is variable, collect more. Document your reasoning in the clinical file.
What if baseline is trending before intervention? That’s a real problem because you can’t tell if the intervention caused the change or if the pre-existing trend continued. Extend the baseline, try a design like multiple-baseline that mitigates trend effects, or document the limitation and interpret results cautiously.
Is statistics required? Visual analysis is primary. Statistics can supplement—effect sizes, for example—but they don’t replace careful design and replication. A crystal-clear phase change with solid replication doesn’t need a p-value to be convincing.
Can I use SCED logic with group-level data? Not in the traditional sense. SCEDs are fundamentally about within-subject replication. You can conduct multiple SCEDs across individuals and then aggregate results, but that’s different from a group design.
Bringing It Together: What Defines an SCED
At its core, a single-case experimental design lets you know whether an intervention caused a change for a specific person. It does this through repeated measurement, planned phase changes, systematic manipulation of what you do, and visual analysis of the resulting patterns. When those elements are in place and replication is solid, you have evidence. When they’re absent or weak, you have observation—useful, but not experimental proof.
For clinicians making real-time decisions, this distinction is everything. It’s the difference between continuing an intervention because “it seems to be working” and continuing it because the data prove it’s working for that learner. It’s the foundation of evidence-based practice at the individual level.
Key Takeaways
SCEDs reveal causal relations by systematically changing what you do and watching whether behavior follows. The defining features—repeated measurement, baseline stability, planned phase changes, replication, and visual analysis—are what make a graph tell an evidence-based story.
Choosing the right design means matching your approach to ethics, reversibility, and what you need to know. Baseline stability is non-negotiable; it’s the foundation on which valid conclusions stand.
Experimental control is about client welfare. Using SCED logic carefully and ethically ensures that every intervention you persist with truly benefits the individual you’re serving.
