Single-subject design
Single-subject design, also known as single-case or single-subject experimental design, is a rigorous quantitative research methodology that evaluates the effects of interventions on individual participants by using each as their own control, through repeated measurements of behavior or outcomes across baseline and treatment phases to establish experimental control and demonstrate functional relationships.[1][2] This approach focuses on the unit of analysis at the individual level, distinguishing it from group-based designs by emphasizing intra-subject variability and replication within the same participant rather than statistical aggregation across groups.[3] Originating in the fields of applied behavior analysis and communication sciences and disorders in the mid-20th century, single-subject design gained prominence in the 1960s with early applications in fluency treatments and behavior therapy, and it flourished in special education and counseling from the late 1970s onward, particularly through pioneering work at institutions like the University of Kansas.[3][1] Key features include the establishment of baseline stability with at least three data points, systematic manipulation of the independent variable (e.g., introduction or withdrawal of an intervention), and visual analysis of graphed data to evaluate changes in level, trend, and variability of the dependent variable.[2][1] Common design types encompass the simple A-B design (baseline followed by intervention), reversal or withdrawal designs (e.g., A-B-A-B to reinstate baseline for comparison), multiple-baseline designs (staggered intervention across behaviors, settings, or subjects), and changing-criterion designs (gradually shifting performance criteria).[2][3] Single-subject designs are particularly valuable in evidence-based practice for their strong internal validity in demonstrating causal effects at the individual level, enabling quick data-driven adjustments in clinical or educational settings, and supporting generalization through replication across multiple participants or contexts, though they have limited external validity without such extensions.[1][3] Applications span psychology, special education, and rehabilitation sciences, such as assessing behavioral interventions for autism (e.g., positive reinforcement in picture exchange communication systems), increasing speech volume in selective mutism, or evaluating language training effects, often serving as a precursor to larger randomized controlled trials.[2][3] Despite their utility, challenges include the need for ethical considerations in withdrawal designs and a historical decline in training, underscoring the importance of standardized reporting guidelines like those from the What Works Clearinghouse.[1]Fundamentals
Definition and Purpose
Single-subject design is an experimental research method that focuses on a single participant, who serves as their own control, to evaluate the effects of an independent variable—such as an intervention—on a dependent variable, such as behavior. This approach employs repeated measurements over time to capture an individual's variability and assess changes attributable to the intervention.[1][3][4] The primary purpose of single-subject design is to demonstrate functional relationships between interventions and individual outcomes, emphasizing an idiographic perspective that prioritizes unique, person-specific responses over nomothetic generalizations derived from group data. By enabling detailed analysis of intra-individual changes, it supports evidence-based practices in fields like applied behavior analysis and special education, particularly when group studies are impractical due to small or heterogeneous samples.[5][6][3] At its core, the logic of single-subject design involves systematic manipulation of experimental conditions through within-subject comparisons, such as alternating baseline and intervention phases, to establish causality and rule out alternative explanations for observed effects. For example, in applied behavior analysis, researchers might use this method to test the impact of paced reading instruction on a third-grade student's acquisition of reading skills, tracking improvements in reading rate from baseline levels to post-intervention performance.[6]Key Principles
Single-subject designs rely on a set of foundational principles derived from baseline logic to establish experimental control and infer causality from intervention effects on individual behavior.[7] These principles—prediction, verification, and replication—enable researchers to systematically demonstrate that observed changes are due to the independent variable rather than extraneous factors.[8] The principle of prediction involves establishing a stable baseline phase through repeated measurement of the dependent variable prior to introducing the intervention, typically with at least three data points to ensure stability characterized by low variability and minimal trends. This baseline provides a reliable forecast of how the behavior would continue in the absence of any manipulation, allowing researchers to anticipate the expected pattern if no change occurs.[7][8] The principle of verification tests the accuracy of the baseline prediction by implementing the intervention and observing whether the dependent variable changes in a manner consistent with the expected impact. If the behavior shifts predictably upon intervention introduction—such as an increase in a target skill or decrease in an undesired response—this verifies that the intervention is responsible for the alteration, rather than coincidental events.[7] Verification strengthens internal validity by confirming the intervention's role in disrupting the predicted baseline trajectory.[8] The principle of replication further bolsters evidence by repeating the conditions of prediction and verification multiple times, either within the same phase sequence or across different elements of the study. This repetition demonstrates the consistency of intervention effects, reducing the likelihood of spurious results and building a robust case for causality.[8] Seminal work by Sidman emphasized replication as the cornerstone of single-subject research, arguing that repeated demonstrations within and across studies are necessary to evaluate behavioral relations reliably. Replication occurs in two primary forms: intra-subject replication, which involves repeating the intervention effects within the same individual across successive phases (e.g., alternating baseline and intervention conditions), and inter-subject replication, which extends the effects across multiple individuals, behaviors, or settings to generalize findings.[7] Intra-subject replication provides strong evidence of control for a single case by showing consistent behavioral changes tied to phase transitions, while inter-subject replication enhances external validity by confirming the intervention's efficacy beyond one participant.[8] Together, these forms ensure that effects are not idiosyncratic but reflect functional relations applicable in applied contexts.[7] Control in single-subject designs is achieved primarily through rigorous baseline measurement, which requires ongoing, frequent data collection to establish a clear reference point for comparison. Stable baselines, obtained via repeated probes under consistent conditions, allow researchers to isolate intervention influences by contrasting pre- and post-manipulation data patterns.[8] Without such controlled baselines, confounding variables like maturation or history could mimic intervention effects, undermining the design's ability to demonstrate causality.[7]Types of Designs
Reversal Design
The reversal design, also known as the withdrawal or ABAB design, is a single-subject experimental approach that systematically alternates between a baseline phase (A) and an intervention phase (B) to evaluate the effects of an independent variable on behavior.[9] The structure typically progresses from an initial baseline phase, where the target behavior is measured without intervention, to an intervention phase, followed by withdrawal back to baseline conditions, and finally reinstatement of the intervention to replicate observed effects.[10] This sequence, often requiring at least four phases with multiple data points per phase for rigor, enables within-subject comparisons to isolate the intervention's impact.[9] Procedures for implementing the reversal design emphasize establishing baseline stability prior to intervention introduction, typically through collection of at least three data points demonstrating consistent levels, trends, or variability in the dependent variable.[9] Once stability is confirmed, the intervention is applied, and ongoing measurement tracks changes in behavior; if a reliable effect emerges, the intervention is withdrawn to assess reversion toward baseline levels.[9] The intervention is then reintroduced to verify replication, with data visually inspected for immediate, clear shifts across phases to confirm experimental control.[10] Stability in each phase is crucial, often requiring extended data collection to rule out extraneous influences. The rationale underlying the reversal design centers on demonstrating functional relations, wherein predictable behavioral changes occur contingent on the introduction, withdrawal, and reintroduction of the intervention, thereby establishing the intervention as the controlling variable.[10] It is particularly suited to reversible behaviors, such as those modifiable through contingent reinforcement, as it provides strong internal validity through intra-subject replication without needing multiple participants.[9] This design aligns with the analytic dimension of applied behavior analysis, prioritizing experimental demonstration of causality over correlational evidence.[10] A representative example involves evaluating a token reinforcement program to decrease disruptive classroom behaviors, such as out-of-seat and talking-out incidents, among elementary students in an adjustment class. During the baseline phase, observations revealed high disruption rates (e.g., averaging over 80% of intervals for talking-out); introduction of tokens exchangeable for privileges reduced these to near-zero levels, with behaviors returning to baseline upon withdrawal and decreasing again during reinstatement, thus confirming the program's efficacy. Variations of the reversal design address practical constraints, such as brief reversals that limit withdrawal duration to ethical minimums or non-reversal adaptations (e.g., using differential reinforcement of alternative behaviors) when full withdrawal risks harm or irreversibility.[9] These modifications maintain the design's core logic while accommodating real-world applications in educational or clinical settings.[9]Alternating Treatments Design
The alternating treatments design, also referred to as the multielement design, enables the comparison of two or more interventions within a single subject through rapid alternation of conditions, often beginning with an initial baseline probe rather than a prolonged baseline phase. This structure facilitates direct side-by-side evaluation of treatment effects without the need for sequential phases typical in other designs. For example, conditions might alternate in a pattern such as A-B-A-B or a counterbalanced sequence like A-B-B-A to ensure each intervention is applied multiple times. Unlike the reversal design, which involves introducing and withdrawing a single intervention sequentially, the alternating treatments design applies multiple treatments concurrently across sessions, thereby avoiding complete withdrawal of effective interventions.[9] Procedures for implementing this design emphasize minimizing biases through careful sequencing and data collection. Treatments are alternated rapidly—often daily or even within the same day (e.g., morning versus afternoon sessions)—using random or systematic orders with counterbalancing to reduce sequence effects, such as order of presentation influencing outcomes. Distinct discriminative stimuli, like different settings or materials, signal each condition to the subject. Data on the target behavior are gathered per session and plotted separately for each treatment condition, allowing for visual inspection of differences in levels, trends, or variability. This approach supports replication across conditions by providing multiple exposures to each intervention within the same subject.[9] The rationale for the alternating treatments design lies in its efficiency for identifying the most effective intervention among options, particularly when reversal is impractical due to ethical concerns or irreversible behavior changes. By comparing effects in close proximity, it accelerates decision-making in applied settings like education or therapy, reducing the time and resources needed compared to designs requiring stable baselines or staggered introductions. In contrast to the multiple baseline design, which delays interventions across behaviors, subjects, or settings to demonstrate control, this design focuses comparisons within a single behavior or subject for quicker differentiation.[9] A representative example involves evaluating reinforcers for a child's low task completion rates in a classroom setting. One condition applies verbal praise contingent on completed tasks, while the other uses corrective feedback; these are alternated across sessions, revealing praise as more effective if completion rates consistently higher under that condition, guiding selection of the optimal strategy. Specific limitations include potential confounding from sequence effects, where the order of treatments impacts results, or carryover effects, where one intervention's influence persists into the next—though these can be mitigated via counterbalancing or spacing.[9]Multiple Baseline Design
The multiple baseline design involves the staggered introduction of an intervention across multiple concurrent baselines, typically three or more, to establish experimental control in single-subject research.[3] This approach originated as an alternative to reversal designs in early applied behavior analysis, as described by Baer, Wolf, and Risley in their seminal 1968 paper outlining dimensions of the field. By applying the intervention sequentially to different baselines while keeping others unchanged, the design demonstrates that behavior changes occur only in response to the intervention, not extraneous factors.[11] Procedures for implementing a multiple baseline design begin with simultaneous data collection to establish stable baselines across all selected conditions, ensuring sufficient data points (often 3–5) to demonstrate variability and trend stability before intervention.[12] Once stability is achieved in the first baseline, the intervention is introduced there, while data collection continues on the remaining baselines without change.[3] The process repeats sequentially for each subsequent baseline, with the timing of intervention staggered to allow replication of effects, typically after 3–7 sessions of stability in prior conditions.[13] This staggered application verifies control through the temporal alignment of behavior changes with intervention onset across baselines. The rationale for the multiple baseline design lies in its ability to control for threats to internal validity, such as maturation, history, or external events, by showing that effects are specific to the intervened baseline and do not generalize prematurely to untreated ones.[11] It is particularly suitable for behaviors where reversal or withdrawal of the intervention would be unethical or impractical, such as skill acquisition that is irreversible once learned.[3] Unlike reversal designs, it avoids potential carryover effects from treatment removal, making it ideal for demonstrating functional relations in applied settings like education or clinical practice.[13] Multiple baseline designs are categorized by the dimension along which baselines are applied:- Across behaviors: The intervention targets different but related behaviors within the same subject, such as improving social initiations and compliance sequentially in a child with autism.[3]
- Across subjects: The same intervention is applied to identical behaviors in multiple participants, staggered by individual, as in evaluating a bullying prevention program across three schools where treatment reduced aggressive incidents only after implementation in each.[14]
- Across settings: The intervention addresses the same behavior in one subject across different environments, such as increasing on-task performance in a student from classroom to hallway to cafeteria contexts.[11]