Single-Case Design: How ABA Proves a Treatment Works
Single-case design tests whether a treatment changed one person's behavior. Learn why behavior analysts use it in research and everyday practice.
Key takeaway
Single-case design is a way to test whether a treatment caused a change in one person's behavior. You measure the behavior many times, change one thing, and watch the graph.

No Man's Land: Is Your Data Real? Clock Time Without IOA and Procedural Integrity Is Unfalsifiable
Free learning path on this topic
Supervision, Done Right
Four supervision CEUs that build on each other. What reflective supervision is, how to build a supervision system, how to assess supervisee competence instead of counting hours, and how supervision affects provider outcomes. Each course is free and issues its own certificate.
Experts who cover this
2 recordings on openceu cover single-case design.
- Featured talk1 CEUs (0.5 Learning, 0.5 Supervision)
No Man's Land: Is Your Data Real? Clock Time Without IOA and Procedural Integrity Is Unfalsifiable
Sean YocumHickory Learning Group
The hosts anchor the whole episode in single-case design standards (Baer, Wolf & Risley 1968; Horner et al. 2005), arguing a valid DV and IV are the…
- 1 Learning CEU
genArete: Milestone based comparison criteria in Skill Assessment
Mark MaladyThe Behaviorist Bookclub
Cited as the historical foundation of behavior analysis, arguing the field's research base was built on studying the individual organism in context rather than…
On this page · 10 sections▾
Single-case design is a way to test whether a treatment caused a change in one person's behavior. You measure the behavior many times, change one thing, and watch the graph.
Behavior analysts also call it single-subject design or SCD. It is the main research method in the field. But the experts here say it is not just for researchers. The same logic helps clinicians make better choices for each client, every week.
What single-case design means#
In a single-case design, each person serves as their own comparison. You do not compare a treated group to an untreated group. Instead, you look at one learner over time. First you collect baseline data. Baseline means the behavior before you start the treatment. Then you add the treatment and keep measuring.
If the behavior changes right when the treatment starts, that is a clue. But a clue is not proof. Something else might have changed at the same time. Maybe a new teacher started. Maybe a medicine changed. So single-case designs add structure. They repeat the change across people, settings, or behaviors. Or they remove the treatment and add it back. When the behavior follows the treatment each time, you can be more sure. Researchers call this a functional relation. That means the treatment, and not some other thing, is driving the change.
The two key parts have formal names. The dependent variable is the behavior you measure. The independent variable is the treatment you change.
Research and practice share the same logic#
Amber Valentino has spent her career working on both sides of this line. In a talk on doing research as a practitioner, she pushes back on the idea that research is a separate world. She says she might even redraw an old flow chart of hers. She now sees more overlap between applied research and applied practice than it showed.
The reasons we use single case design and research are not in any meaningful way different than we would use them in practice.
From the talk — Amber Valentino
Valentino gives a simple example. Say you find an intervention that lowers challenging behavior. Or it builds a functional communication skill. Now you plan to roll it out. You will train every teacher and team member. You will spend real resources on it. Before you do that, Valentino says, you want to be convinced. You want to know nothing else caused the change. That is the exact question a single-case design answers. So a clinician checking their work is already thinking like a researcher.
Why the graphs persuade people#
A big part of clinical work is convincing other people. Parents, teachers, and funders all need to trust the plan. Valentino calls this engaging and convincing stakeholders. She sees single-case graphs as one of the field's best tools for it.
This is the beautiful situation, the beautiful thing about a single case design and behavior analysis is that our graphs are incredibly convincing.
From the talk — Amber Valentino
Her point is that you do not need training to read these graphs. Someone may not know behavior analysis. They may rarely look at data. Valentino says a quick tour of the X and Y axes is usually enough. Then the picture speaks for itself. The behavior was here before treatment, and it moved after.
She admits people sometimes say "the data are wrong." But she says that when she shows a clean single-case graph about client care, nobody has ever pushed back on the spot. A clear design makes the result hard to argue with. That matters when you need a team to buy in fast.
Cleaner decisions than a pile of data#
Valentino is honest about the mess of real practice. She calls herself a lifelong practitioner. She knows things change all the time. Staff change. Schedules change. You cannot always control what goes into the electronic data system.
Many clinicians try to make sense of that mess after the fact. They point to one spot on the graph and guess what happened. Valentino says this almost never works. Without an experimental design, a stack of data points is just a stack. There is no real way to decide what to do next from it.
When you introduce a single case design, the decision-making is just so much cleaner. It's quicker, it's more efficient.
From the talk — Amber Valentino
With a design in place, she says, someone can show her a graph and she can answer fast. Keep going with this. Or stop, you are done, and move on to generalization. A co-presenter summed up her point well. Your job is to take the mess, form a guess about what is going on, and then test it. Single-case design is how you run that test.
Built on the individual learner#
Mark Malady, BCBA, looks at single-case design from a different angle. He traces it to the roots of the science. Behavior, he says, is shaped by a very complex set of variables. Those variables differ a lot from one person to the next. So the true subject of the science is the individual person in the context of their own life.
our research foundations were really built on N of one research methodologies and looking at the individual learner in context, rather than using group aggregate data, statistical approaches or methodologies
From the talk — Mark Malady
"N of one" means a study of one person. Group aggregate data means averages taken across many people. Malady says this focus on the individual is "core and critical to the basic concept of the operant." An operant is a behavior shaped by its consequences.
Malady raises this to point out a tension. Many tools in the field still compare a child to group norms. Those tools push clinicians to act as if the goal is to make a person "normal." That sits awkwardly next to a science built on each person's own pattern. Malady does not reject assessment. He is asking the field to square its practice with its roots. His full talk on this is genArete: Milestone based comparison criteria in Skill Assessment.
Two studs in the wall: measurement you can trust#
Sean Yocum takes the hardest line on quality. He uses a house metaphor. How much of our science holds the weight, he asks, and how much is fresh paint over rot? He says the answer does not depend on your service model. It applies to two hours of parent coaching or 40 hours of clinic therapy.
if you claim to practice single case design, you need two non-negotiable studs in the wall.
From the talk — Sean Yocum
The first stud is a reliable dependent variable. That means good data on the learner's behavior. Yocum points back to Baer, Wolf, and Risley (1968). Their first dimension of ABA was "behavioral." We measure what people do, not what someone remembers. He also cites Hawkins and colleagues (1977). They said a behavior definition must be objective, clear, and complete.
Yocum then cites Horner and colleagues (2005), who set quality markers for single-case research. For the dependent variable, the standard was repeated measurement. It also needed interobserver agreement of at least 80% across at least 20% of sessions in every condition. Interobserver agreement, or IOA, means two people score the same behavior and match.
The second stud is a measured independent variable. That means checking the treatment was actually delivered as planned. This is called procedural integrity or procedural fidelity. Both IOA and fidelity have their own pages, so they get only a brief note here.
Yocum warns about "car seat data." That is when a tired staff member fills in the whole session from memory at the end of the day. Without the two studs, he says, a line graph is just a story you cannot test.
Common designs you will see#
Single-case design is a family of methods, not one method. The studies covered here show several types.
A multiple baseline design starts treatment at different times for different people, behaviors, or settings. If each one changes only when its own treatment starts, that is strong evidence. Bosley and colleagues (2024) used a multiple baseline across six young children with language delays. They tested a reading method called RECALL, which uses a prompting order with picture cards.
A multiple probe design is a close cousin. It checks the behavior now and then instead of in every session. Balikci (2026) used one with four teacher and child pairs. Teachers used an AI-supported self-coaching tool, and their teaching accuracy went up.
A reversal design removes the treatment and then brings it back. If the behavior goes back and forth with the treatment, that shows control. Grekov and Pustejovsky (2026) re-analyzed one reversal study and one multiple baseline study.
Designs can also grow large. One planned trial by LB and colleagues (2023) uses a multiple single-case design with 40 students. It has four randomized baselines and runs for 18 months.
Limits and open questions#
None of the experts quoted here argue against single-case design. But the research literature does raise fair questions about what it can and cannot do.
D'Agostino and Pinkelman (2025) note that SCD is the method behavior analysts are mostly trained in. It is also the most common method in the field's journals. But they argue it may not answer every question society cares about. They point back to Baer and colleagues (1968). Good research, in that view, is judged by society's interest in the problem, not by the method. They urge more input from clients and families in research.
Burnham Riosa (2025) suggests a way to blend methods. A study could run a single-case design and add short interviews during each phase. The interviews would capture how people felt during baseline, treatment, and after. That could add a richer picture of the lived experience behind the graph.
These ideas extend single-case design rather than replace it. They also echo Malady's point about keeping the individual person at the center.
What the research says#
Walker and Carr (2021) tackle a common myth about single-case generality. Many people think you need a big sample to apply findings to a new client. The authors explain that single-case research builds generality through a line of replication studies instead. They also offer a sample "elevator speech" for explaining this idea to others.
Grekov and Pustejovsky (2026) note that researchers mainly use visual analysis to judge single-case data. Statistical models are now being added to support it. The authors describe a way to check whether a model fits by simulating fake data and comparing it to real data. They suggest focusing those checks on the same features people look for in visual analysis.
Bosley and colleagues (2024) studied the RECALL reading method with autistic children and children with language delay. During the intervention, children responded more and gave more correct answers.
D'Agostino and Pinkelman (2025) argue for adding qualitative and community-based methods alongside single-case design.
FAQ#
Is single-case design only for published research? No. Amber Valentino argues the reasons to use it in research and practice are basically the same. Any time you want to be sure a treatment caused a change, the design helps. It is especially useful before you train a whole team on a new plan.
Can you apply single-case findings to other clients if the study had so few people? Yes, with care. Walker and Carr (2021) explain that generality comes from many replication studies over time. It does not come from one large study. Each new study that repeats the result adds confidence.
What makes single-case data trustworthy? Sean Yocum names two things. You need reliable data on the behavior, checked with interobserver agreement. You also need proof the treatment was delivered as planned. Without both, he says, the graph cannot be tested.
Turn this topic into a CEU
You just studied this. Now get credit for it.
Watch No Man's Land: Is Your Data Real? Clock Time Without IOA and Procedural Integrity Is Unfalsifiable with Sean Yocum and earn 1 free BCBA CEU. Detailed certificate, delivered the moment you finish.
Keep going with a free learning path
Supervision, Done Right
Four supervision CEUs that build on each other. What reflective supervision is, how to build a supervision system, how to assess supervisee competence instead of counting hours, and how supervision affects provider outcomes. Each course is free and issues its own certificate.
