6.1 Characteristics of Single-Case Experimental Designs & Experimental Control

Foundations of Single-Case Experimental Designs (SCEDs)

Single-case experimental designs (SCEDs)—also termed single-subject, within-subject, or intra-subject designs—represent the empirical backbone of applied behavior analysis (ABA). Originating from B.F. Skinner's experimental analysis of behavior (EAB) and formalized methodologically by Murray Sidman in his seminal 1960 text Tactics of Scientific Research, SCEDs provide a rigorous, scientific methodology for evaluating the effects of environmental variables on the behavior of individual organisms.

In stark contrast to traditional group experimental research, which relies on large participant cohorts ($N$), random assignment to treatment and control groups, and inferential statistical testing (e.g., $t$-tests, ANOVA), single-case methodology focuses intensely on the individual learner. Behavior is an individual physical phenomenon: groups of people do not behave; individuals behave. When data from twenty or fifty individuals are averaged together, the resulting group mean frequently creates an artificial statistical composite that does not accurately describe the behavioral trajectory, learning rate, or response pattern of any single participant in the study.

Single-case experimental designs are defined by three indispensable operational characteristics:

1. Participant Serves as Their Own Control (Intra-Subject Comparison): In an SCED, each participant's own behavioral responding under baseline conditions serves as the control against which their responding under intervention conditions is evaluated. Because the individual is compared directly against themselves across time, inter-subject variability (differences in genetics, developmental history, intelligence, temperament, and ambient home life) is completely held constant.
2. Continuous, Repeated Measurement of the Dependent Variable Across Time: Rather than capturing arbitrary pre-test and post-test snapshots, the investigator measures the dependent variable (DV) repeatedly across consecutive sessions, days, or weeks before, during, and after the introduction of the independent variable (IV). This high-density time-series measurement captures behavioral level, trend, variability, and immediacy of change, exposing the dynamic, real-time process of learning.
3. Systematic Manipulation of the Independent Variable: The researcher or clinician systematically introduces, alters, or withdraws specific environmental manipulations while actively holding all other extraneous ambient variables constant, directly testing whether changes in the target behavior are functionally related to the intervention.

---

The Steady-State Strategy and Baseline Logic

A hallmark of single-case research is the steady-state strategy. Formulated by Sidman (1960), the steady-state strategy is the experimental practice of exposing an individual to a given environmental condition (whether baseline or intervention) while actively holding extraneous factors constant, and repeatedly measuring the behavior until a stable, consistent pattern of responding—termed steady-state responding—is achieved.

Steady-state responding is defined as a pattern of responding that exhibits minimal variability, a stable level, and the absence of an active upward or downward trend over a succession of measurement intervals. Establishing steady-state responding provides the empirical foundation for baseline logic.

The Operational Role of Baseline Data

In behavior analysis, a baseline (Phase A) is not a period of passivity or merely the "absence of treatment." Baseline is an active, controlled experimental condition in which the independent variable is absent, allowing the investigator to establish an objective benchmark against which subsequent manipulations can be meaningfully evaluated. Baseline data fulfill four interrelated scientific functions:

• Description: Baseline data objectively quantify the current level, variability, and trend of the target behavior under naturalistic environmental contingencies, documenting the clinical necessity of intervention.
• Prediction: A stable baseline establishes an empirical trajectory that allows the analyst to predict the future course of the behavior if the independent variable is not introduced. Prediction operates on an inductive premise: if environmental conditions remain constant, steady-state responding will continue along its current trajectory.
• Verification: Verification occurs when subsequent data demonstrate that the baseline level would have remained unchanged had the intervention not been introduced. In reversal designs, verification is achieved when withdrawing the intervention causes the behavior to return to its original baseline level. In multiple baseline designs, verification is achieved when untreated tiers remain stable while the treated tier changes.
• Replication: Replication is the demonstration that the observed behavior change can be reproduced by reintroducing the independent variable (in reversal designs) or by introducing the independent variable across subsequent tiers (in multiple baseline designs). Replication provides convincing proof of reliability and experimental control.

Decision Rules for Introducing Interventions: Evaluating Baseline Trends

A critical competency for the Board Certified Assistant Behavior Analyst (BCaBA) is determining exactly when baseline stability has been achieved and whether it is methodologically sound and clinically safe to introduce an independent variable. Clinicians evaluate four distinct baseline patterns:

1. Stable Baseline (Zero Trend, Low Variability): Data points fluctuate within a narrow, predictable range with no upward or downward drift. This represents the gold standard for introducing an independent variable, as any subsequent shift in level or trend can be cleanly attributed to the intervention.
2. Ascending Baseline (Upward Trend): The target behavior is systematically increasing over time. The decision to intervene depends entirely on whether the target behavior is a behavioral deficit or a behavioral excess:
   • Behavioral Deficit (Targeted for Increase, e.g., Task Completion, Mands): An intervention must never be introduced into an ascending baseline. If academic engagement is already increasing spontaneously during baseline, introducing a token economy makes it impossible to determine whether subsequent increases were caused by the token economy or by the preexisting upward trajectory.
   • Behavioral Excess (Targeted for Reduction, e.g., Aggression, SIB): Introducing an intervention into an ascending baseline is methodologically acceptable and clinically urgent. If severe self-injury is escalating, waiting for stability is dangerous; furthermore, if the intervention reverses an ascending trajectory and forces behavior downward, experimental control is convincingly demonstrated.
3. Descending Baseline (Downward Trend): The target behavior is systematically decreasing over time:
   • Behavioral Excess (Targeted for Reduction, e.g., Tantrums, Property Destruction): An intervention must never be introduced into a descending baseline for research purposes. If tantrums are already diminishing spontaneously (perhaps due to natural habituation, peer extinction, or biological changes), introducing differential reinforcement of other behavior (DRO) will yield confounded data. The clinician cannot prove whether the ultimate reduction was caused by the DRO or by the continuation of the preexisting downward trend.
   • Behavioral Deficit (Targeted for Increase, e.g., Social Initiations): Intervening during a descending baseline is methodologically acceptable because an effective intervention must counteract and reverse the downward slide to demonstrate success.
4. Variable / Unstable Baseline (High Fluctuations): Data fluctuate wildly across sessions without a discernible pattern. Clinicians must withhold intervention and actively investigate the source of variability. The clinician must identify and control extraneous variables—such as erratic sleep schedules, intermittent teacher attention, medication administration fluctuations, or inconsistent task difficulty—until environmental stability yields behavioral stability.

---

Experimental Control and Establishing Functional Relations

In applied behavior analysis, scientific inquiry culminates in the demonstration of experimental control. Experimental control is achieved when an investigator reliably and convincingly modulates the occurrence of a dependent variable (turning behavior on and off, up or down) by systematically manipulating an independent variable while holding all potential confounding variables constant.

When experimental control is demonstrated, the investigator establishes a functional relation. A functional relation is an empirically confirmed, cause-and-effect relationship between an environmental event (the independent variable) and a specific class of behavior (the dependent variable). Formally, a functional relation exists when:

In plain terms: a change in the independent variable reliably produces a change in the dependent variable while extraneous variables are held constant, and the effect is replicated.

To establish a functional relation, single-case methodology demands more than a simple "before-and-after" correlation. A single pre-intervention baseline followed by a single intervention phase (an AB design) can demonstrate prediction (the behavior changed after intervention was added), but it can never demonstrate experimental control or a functional relation. Why? Because the AB design fails to rule out rival explanations: the observed change could have been caused by coincident historical events (e.g., a medication change, schedule modification), natural physical maturation, or observer reactivity. To confirm a functional relation, the effect must be systematically replicated within the experiment using established single-case architectures (such as ABAB reversals, multiple baselines, alternating treatments, or changing criterion designs).

---

Four Fundamental Types of Experimental Analyses

Behavior analysts design experiments to answer distinct clinical and scientific questions. On the BCaBA examination, candidates must differentiate among four foundational types of experimental analyses: nonparametric analysis, parametric analysis, component analysis, and comparative analysis.

                         [ EXPERIMENTAL ANALYSES ]
                                     |
     +------------------+------------+------------+------------------+
     |                  |                         |                  |
[ NONPARAMETRIC ]  [ PARAMETRIC ]           [ COMPONENT ]      [ COMPARATIVE ]
  Presence vs.      Varying values/          Dismantling or     Contrasting two
   Absence of        doses of the            adding parts of    distinct, rival
     an IV              SAME IV              a package           interventions

1. Nonparametric Analysis

A nonparametric analysis is an experimental evaluation of the presence versus the absence of an independent variable. It represents the foundational "treatment ON versus treatment OFF" experimental manipulation.

• Core Question: "Does this independent variable produce a meaningful behavioral change compared to when it is completely absent?"
• Mechanism: The experimenter alternates between a phase where the intervention is completely withheld (baseline or withdrawal) and a phase where the intervention is actively implemented.
• Clinical Exemplar: An assistant behavior analyst evaluates the impact of an antecedent visual activity schedule on classroom transitions. Phase A involves no visual schedule (treatment OFF); Phase B introduces the visual schedule (treatment ON); Phase A withdraws the visual schedule (treatment OFF); Phase B reinstates it (treatment ON). Comparing the presence of the schedule against its total absence constitutes a nonparametric analysis.

2. Parametric Analysis

A parametric analysis is an experimental evaluation that systematically investigates the differential effects of varying values, doses, schedules, or intensities of the SAME independent variable.

• Core Question: "What is the optimal, most efficient, or most cost-effective dose or parameter of this specific independent variable?"
• Mechanism: Rather than turning the IV on and off, the experimenter keeps the IV present but systematically adjusts a quantitative parameter across sequential phases or alternating conditions.
• Clinical Exemplars:
  • Evaluating Reinforcement Schedules: Systematically evaluating the rate of vocational task completion when tokens are delivered on an FR 2, FR 5, FR 10, and FR 20 schedule.
  • Evaluating Delay to Reinforcement: Assessing problem behavior when the delay to functional communication reinforcement is set to 0 seconds, 10 seconds, 30 seconds, and 60 seconds.
  • Evaluating Reinforcer Magnitude: Comparing the evocative effect of 1 minute, 3 minutes, and 5 minutes of iPad access following compliance.
  • Evaluating Response Cost Fines: Comparing the suppressive effect of losing 1 token versus losing 3 tokens versus losing 5 tokens for aggression.

3. Component Analysis

A component analysis is an experimental analysis designed to systematically identify the active, necessary, sufficient, or redundant elements of a multi-component treatment package.

Behavioral interventions are frequently implemented as comprehensive "packages." For example, a classroom behavior support plan might include: (1) a visual schedule, (2) differential reinforcement of alternative behavior (DRA), (3) a token economy, and (4) functional communication training (FCT). While the full package may successfully reduce disruptive behavior, implementing four simultaneous components requires substantial staff effort and financial resources. A component analysis determines which specific parts are doing the heavy lifting.

Component analyses are conducted using two primary methodological strategies:

• Dismantling (Drop-Out) Strategy: The investigator begins by implementing the entire multi-component package until steady-state responding is established. Successive phases then systematically withdraw (drop out) one individual component at a time. If the behavior remains fully suppressed when the visual schedule is eliminated, the visual schedule is proven redundant. If disruptive behavior immediately surges when the token economy is withdrawn, the token economy is identified as a necessary, active component.
• Additive (Add-In) Strategy: The investigator begins with baseline or a single core component, and then systematically adds individual components one by one (e.g., Phase A: Baseline; Phase B: DRA alone; Phase C: DRA + Token Economy; Phase D: DRA + Token Economy + Response Cost). This strategy reveals the incremental therapeutic value added by each successive component.

4. Comparative Analysis

A comparative analysis is an experimental evaluation that directly contrasts two or more distinct, conceptually independent interventions against one another to determine their relative efficacy, efficiency, or social acceptability.

• Core Question: "Which of these two completely different interventions is superior for treating this target behavior in this individual?"
• Mechanism: Two or more independent behavioral technologies are evaluated either in an alternating treatments design or in sequential multi-element phases.
• Clinical Exemplars:
  • Comparing Functional Communication Training (FCT) versus Noncontingent Reinforcement (NCR) to determine which produces faster suppression of escape-maintained aggression.
  • Comparing Discrete Trial Teaching (DTT) versus Naturalistic Environment Teaching (NET) for the acquisition of expressive labeling repertoires.
  • Comparing Most-to-Least Prompting versus Least-to-Most Prompting to identify which leads to fewer learner errors during shoe-tying instruction.

---

Comparative Analysis: Single-Case Designs vs. Group Experimental Designs

To excel on the BCaBA examination, practitioners must thoroughly understand the methodological divergence between single-case experimental designs and between-subject group experimental designs:

---

Common BCaBA Exam Traps: Single-Case Characteristics and Experimental Control

• Trap 1: Confusing Parametric Analysis with Component Analysis: If a question describes altering the dose, schedule, delay, or magnitude of a single intervention (such as evaluating FR 2 vs. FR 5 vs. FR 10 token delivery), it is a parametric analysis. If the scenario describes adding, removing, or pulling apart distinct intervention strategies (such as testing whether praise alone works as well as praise plus a token board), it is a component analysis.
• Trap 2: Intervening During a Therapeutic Baseline Trend: Never select an option that introduces an intervention into an improving baseline trend. If problem behavior is already decreasing during baseline, introducing a reduction intervention destroys experimental control. If a skill deficit is already increasing, introducing an acquisition intervention destroys experimental control.
• Trap 3: Believing Group Statistical Significance Equals Clinical Control: Exam questions often present administrators or outside professionals arguing that a randomized group trial with a $p$-value $< .05$ is superior to an SCED. The correct behavioral answer emphasizes that group averages mask individual trajectories, fail to identify non-responders, and do not demonstrate functional control over an individual's operant repertoire.
• Trap 4: Confusing Prediction with Experimental Control: Demonstrating that behavior changes following the introduction of an intervention in an AB design demonstrates prediction, but it does not demonstrate experimental control. Experimental control requires convincing verification and replication.