Exploring the Varieties of Single Subject Design in Research

Not all research questions can be answered by studying large groups of people. Sometimes, what matters most is what happens to one individual over time – how their behavior shifts when a treatment is introduced, removed, or gradually adjusted. This is where single subject design (SSD) becomes invaluable. Rather than averaging outcomes across many participants, single subject designs focus on tracking one person’s responses in detail, making the individual the unit of analysis. Within this broader framework, researchers have developed several distinct design types – each suited to different research questions and ethical constraints. Understanding these varieties helps clarify not just how interventions work, but when and for whom they work best.

The logic behind single subject research 🔗

Before exploring specific design types, it helps to understand the shared logic that underlies all single subject research. According to researchers at the University of Minnesota, single subject experimental designs (SSEDs) provide a flexible alternative to traditional group designs, particularly when studying the effects of interventions at an individual level. Because randomized control trials may not always be feasible – especially with rare conditions or highly individualized treatments – SSEDs offer a rigorous yet practical path to building evidence-based practice.

Three principles anchor all single subject designs: prediction (forming a hypothesis about what will happen when an intervention is applied), verification (demonstrating that baseline behavior would have remained stable without the intervention), and replication (repeating the intervention to show consistent results). Together, these principles allow researchers to establish that behavior changes are genuinely linked to the independent variable – not random fluctuation or outside events.

Withdrawal design 🔗

The withdrawal design – also called the reversal or ABA design – is arguably the most straightforward single subject approach. In this design, baseline data is first collected (Phase A), a treatment is then introduced and its effects are recorded (Phase B), and finally, the treatment is withdrawn and the study returns to baseline (back to Phase A). The researcher compares behavior across these phases to determine whether the intervention caused the observed change.

A more extended version, the ABAB design, adds another treatment phase after the second baseline. This repetition strengthens the case for experimental control – if behavior consistently improves during treatment phases and reverts during baseline phases, the evidence for a causal link is compelling. The APA Dictionary of Psychology defines the withdrawal design as one in which a treatment is removed during one or more periods, allowing researchers to observe whether the behavior reverts to its original pattern.

Strengths and limitations 🔗

The withdrawal design is powerful at demonstrating experimental control. However, it carries notable ethical concerns. It is not appropriate for high-risk or dangerous behaviors, and some behaviors simply cannot be reversed – if a child has learned to read, removing the reading instruction won’t erase that skill. This means the design is best used for behaviors that are both reversible and where temporarily withholding treatment poses no significant harm.

Alternating treatments design 🔗

When a researcher wants to compare the effectiveness of two or more interventions – not just test whether one works – the alternating treatments design (ATD) is the tool of choice. This design involves rapidly alternating between conditions within the same phase, with each condition having an equal opportunity to be present during measurement. The rapid switching allows multiple interventions to be evaluated simultaneously within a relatively short time frame.

For example, a researcher studying language acquisition in a child with autism might alternate between English-language instruction in one session and Spanish-language instruction in the next, tracking response accuracy under each condition. One study published in the Journal of Behavioral Education used exactly this approach, employing an ATD to compare the effects of language instruction on response accuracy and challenging behavior.

What makes it effective – and where it falls short 🔗

The ATD’s biggest advantage is efficiency. There is no need to withdraw a treatment to establish a comparison; instead, both treatments run concurrently in alternating fashion. It is particularly useful when treatments are fast-acting and when ethical considerations prevent withholding a potentially beneficial intervention. The main drawback is the risk of multiple treatment interference – the possibility that exposure to one treatment affects how the participant responds to another. To meet rigorous design standards, researchers using an ATD typically need to compare at least three conditions or include a direct baseline for adequate replication.

Multiple baseline design 🔗

The multiple baseline design is one of the most widely used single subject designs, largely because it sidesteps the ethical problems of treatment withdrawal entirely. Rather than removing an intervention, this design staggers its introduction across multiple behaviors, individuals, or settings. The intervention is applied to one baseline at a time while the others remain in the baseline condition. If behavior changes only after the intervention is introduced in each tier – and not before – a strong case for the treatment’s effectiveness is established.

There are three main variants. A multiple baseline across behaviors design tests the same intervention on different behaviors in the same person. A multiple baseline across participants design applies the intervention to different individuals at different time points. A multiple baseline across settings design examines whether the treatment produces consistent effects in different environments, such as home, school, and clinic.

Why researchers favor it 🔗

Multiple baseline designs are often appealing to researchers and clinicians precisely because they do not require the behavior to be reversible and do not necessitate withdrawing an effective treatment. This makes them ethically suitable for serious therapeutic interventions – such as reducing self-injurious behavior or improving communication skills – where removal of treatment would be harmful. Internal validity is strengthened by the staggered replication across tiers, and findings are more likely to generalize across conditions. A related variant, the multiple probe design, reduces data collection burden by measuring baseline only sporadically instead of continuously, which is helpful when time and resources are limited.

Changing criterion design 🔗

The changing criterion design (CCD) takes a different approach altogether. Rather than comparing the presence versus absence of a treatment, it evaluates whether behavior tracks a series of progressively shifting performance standards. The CCD uses step-wise benchmarks to manipulate a dimension of a behavior – such as its frequency, duration, or accuracy – that is already present in the individual’s repertoire. The treatment criterion is raised (or lowered) incrementally, and the researcher observes whether behavior consistently matches each new standard.

This design is especially suited to goals that require gradual shaping rather than an immediate, large-scale change. Classic applications include reducing cigarette smoking step by step, increasing daily exercise, or helping a student build up the number of math problems completed within a set time. Each criterion shift acts as a mini-replication: if behavior consistently adjusts to each new criterion, the evidence mounts that the intervention is driving the change.

Key considerations 🔗

One important constraint of the CCD is that it can only be used when the target behavior is already in the participant’s repertoire. It does not require withdrawal of treatment, and it does not allow for direct comparison between two different interventions. It is also worth noting that experimental control can optionally be strengthened by temporarily reversing a criterion – dropping back to a previous performance standard – to confirm that behavior tracks the criterion rather than improving independently. First formally described in the Journal of Applied Behavior Analysis by Hartmann and Hall in 1976, the CCD has since been applied across education, clinical psychology, sport science, and behavioral health.

Comparing the four designs: when to use which 🔗

Each design fits a different research scenario. The withdrawal design is best when strong experimental control is the priority and the target behavior is ethically reversible. The alternating treatments design shines when two or more interventions need rapid comparison within a single study. The multiple baseline design is the go-to when treatment withdrawal is impractical or unethical, and when the researcher wants to observe effects across multiple tiers. The changing criterion design is the right fit when the goal is gradual behavior shaping across incremental steps.

What all four share is the core feature of single subject research: the participant serves as their own control. Rather than comparing an individual to a group average, the researcher compares that person’s behavior under different conditions. This sensitivity to individual differences is precisely what makes these designs so valuable in clinical, educational, and behavioral research – and why they remain a cornerstone of evidence-based practice in psychology and applied behavior analysis.

What do you think? If you were designing a study to evaluate a new classroom intervention for students with attention difficulties, which of these single subject designs would you choose – and what ethical considerations would guide that choice? And do you think the gradual, stepwise approach of the changing criterion design could be more effective than an all-or-nothing intervention for building long-term behavioral habits?

References

1. https://en.wikipedia.org/wiki/Single-subject_design
2. https://pmc.ncbi.nlm.nih.gov/articles/PMC3992321/
3. https://behavioranalyststudy.com/single-subject-experimental-design/
4. https://uta.pressbooks.pub/advancedresearchmethodsinsw/chapter/15-2-2/
5. https://learningbehavioranalysis.com/d-5-use-single-subject/
6. https://pmc.ncbi.nlm.nih.gov/articles/PMC5352619/
7. https://onlinelibrary.wiley.com/doi/10.1901/jaba.1976.9-527