14.2 Single-Case Experimental Designs in Clinical Practice

Single-Case Experimental Designs in Clinical Practice

Exam Tip: On the QASP-S exam, you must demonstrate a deep conceptual and applied understanding of Baseline Logic (Prediction, Verification, and Replication) across the primary single-case experimental designs. Be prepared to identify when a Reversal (ABAB) design is clinically or ethically contraindicated (e.g., severe self-injurious behavior or irreversible academic skills), how Multiple-Baseline Designs establish experimental control without withdrawing treatment, how to interpret Multielement / Alternating Treatments Designs (ATD) based on vertical path separation, and how Changing Criterion Designs use bidirectional criterion shifts to provide definitive proof of experimental control.

In Applied Behavior Analysis, establishing that a specific clinical intervention—and not an uncontrolled extraneous variable—caused a client's behavioral improvement requires rigorous experimental design. Unlike between-group research designs that evaluate aggregate differences between experimental and control groups, single-case experimental designs (SCEDs) employ intra-subject replication. In an SCED, the individual client serves as their own experimental control, with repeated measurements taken across time under systematically manipulated environmental conditions.

---

The Operant Engine: Baseline Logic

All single-case experimental designs derive their scientific validity from what Murray Sidman (1960) and Baer, Wolf, and Risley (1968) termed Baseline Logic. Baseline logic comprises three distinct, sequential experimental elements: Prediction, Verification, and Replication.

┌─────────────────────────────────────────────────────────────────────────────┐
│                         THE THREE ARCS OF BASELINE LOGIC                    │
├──────────────────┬──────────────────────────────────────────────────────────┤
│ 1. Prediction    │ Projecting future behavior if environmental conditions   │
│                  │ remain unchanged (extrapolating baseline trend line).    │
├──────────────────┼──────────────────────────────────────────────────────────┤
│ 2. Verification  │ Demonstrating that baseline levels of behavior would have│
│                  │ persisted unchanged had the intervention not occurred.   │
├──────────────────┼──────────────────────────────────────────────────────────┤
│ 3. Replication   │ Reproducing the behavioral change by reintroducing the   │
│                  │ independent variable, proving reliability and control.   │
└──────────────────┴──────────────────────────────────────────────────────────┘

1. Prediction

Prediction refers to the anticipated outcome of a presently unknown or future measurement. When a stable baseline (Phase A1) is established with minimal variability and zero trend, the clinician predicts that if no environmental changes occur, the behavior will continue along that same trajectory. When the independent variable is introduced (Phase B1) and the data path departs noticeably from the predicted baseline path, the clinician obtains initial evidence that the intervention altered the behavior.

2. Verification

Verification is the demonstration that prior baseline levels of behavior would have continued unchanged in the absence of the independent variable. In a reversal design, when the clinician withdraws the treatment (Phase A2) and the behavior returns to its original baseline levels, this return verifies the accuracy of the original baseline prediction. In a multiple-baseline design, verification is achieved when untreated tiers remain completely stable while treatment is actively applied to the first tier.

3. Replication

Replication involves reintroducing the independent variable (Phase B2) and observing whether the behavioral effect is reliably reproduced. When the target behavior changes in the same direction and magnitude as it did during the first intervention phase (Phase B1), the clinician demonstrates that the outcome was not an accidental coincidence, establishing a functional relation and experimental control.

---

Reversal / Withdrawal (ABAB) Designs

The reversal design (commonly referred to as an ABAB design or withdrawal design) is the gold standard of experimental control in single-case research. It involves the repeated introduction and removal of the independent variable across four distinct phases:

1. Baseline 1 (A1): Initial measurement of the target behavior under natural conditions; establishes prediction.
2. Intervention 1 (B1): Introduction of the independent variable; demonstrates intervention effect.
3. Withdrawal/Reversal (A2): Removal of the independent variable; provides verification of baseline prediction.
4. Intervention 2 (B2): Reintroduction of the independent variable; provides replication of treatment effect.

Variations of the Reversal Design:

• BAB Design: Used when a client presents with severe behavioral crises requiring immediate intervention without waiting for an initial baseline. The clinician implements treatment immediately (B1), withdraws it briefly (A), and reintroduces it (B2). Limitation: Lacks an initial baseline prediction arc.
• ABAC Multi-Treatment Design: Evaluates the relative efficacy of two distinct interventions (B and C) separated by baseline or reversal phases (e.g., Baseline -> Treatment B -> Baseline -> Treatment C -> Baseline -> Treatment B).

Critical Clinical & Ethical Limitations of Reversal Designs:

1. The Irreversibility Constraint: Reversal designs can never be used to evaluate behavioral interventions that teach learned skills, knowledge, or academic repertoires. Once a child learns to read sight words, ride a bicycle, or mand using a picture exchange system, withdrawing the teaching prompt will not cause the client to "unlearn" the skill. The behavior cannot revert to baseline levels, preventing verification.
2. Ethical Dilemmas with High-Risk Behaviors: It is strictly unethical and clinically hazardous to withdraw an effective behavior intervention plan if the target behavior is dangerous to the client or others. Withdrawing treatment for severe eye-gouging, head-banging, glass-shattering property destruction, or high-intensity physical aggression merely to "prove" experimental control violates QABA and BACB ethical codes.
3. Social and Educational Resistance: Parents, teachers, and residential staff frequently object to removing an intervention that is visibly helping a child succeed.

---

Multiple-Baseline Designs (MBD)

The Multiple-Baseline Design (MBD) is the most widely adopted experimental design in applied behavioral settings because it allows clinicians to establish experimental control without withdrawing the intervention. It is ideally suited for irreversible behaviors (learned skills) and severe, dangerous topographies.

Core Mechanics:

In an MBD, two or more concurrent baselines of varying temporal lengths are established simultaneously across multiple independent tiers. The independent variable is then introduced into Tier 1 while Baselines in Tier 2 and Tier 3 continue uninterrupted. Only after the behavior in Tier 1 demonstrates a clear, stable treatment effect is the independent variable introduced into Tier 2. This staggered, sequential implementation continues across all remaining tiers.

Baseline Logic in Multiple-Baseline Designs:

• Prediction: Continuous baseline data in each tier project future responding.
• Verification: When treatment is applied to Tier 1 and causes a behavioral shift, the fact that data in Tier 2 and Tier 3 remain unchanged at baseline levels verifies that the shift in Tier 1 was caused by the intervention and not by an extraneous historical or maturation event.
• Replication: When the independent variable is subsequently introduced into Tier 2 and Tier 3, and each tier shows the same behavioral change, experimental control is systematically replicated.

The Three Primary Variations:

1. Multiple Baseline Across Behaviors: Evaluates two or more distinct, independent behaviors emitted by the same individual in the same setting (e.g., mands, tacts, and intraverbals in a clinic). Threat: Behavioral covariation (if improving one behavior accidentally alters the others, experimental control is compromised).
2. Multiple Baseline Across Settings: Evaluates the same behavior of the same individual across two or more independent physical environments (e.g., classroom, playground, and home). Threat: Stimulus generalization across settings before treatment is officially introduced.
3. Multiple Baseline Across Participants (Subjects): Evaluates the same behavior in the same setting across two or more different individuals (e.g., three autistic students learning toothbrushing in a life skills classroom). This is the most common and robust MBD subtype because individuals cannot easily "covary."

Concurrent vs. Non-Concurrent Multiple Baseline:

• Concurrent MBD: All baselines across all tiers are measured simultaneously in real time. Provides the highest internal validity against historical confounds.
• Non-Concurrent MBD: Baselines across participants are measured at different chronological points in time (e.g., Participant 1 in January, Participant 2 in April) with predetermined baseline durations. Clinically convenient in community clinics where clients enroll at different times, but weaker because it cannot control for concurrent historical events.

Multiple Probe Design:

A highly practical variation of the MBD where baseline data are not collected continuously across every single session. Instead, intermittent probes are conducted. This prevents testing fatigue, extinction, or emotional behavioral bursts caused by repeatedly testing a student on skills they cannot yet perform.

---

Multielement / Alternating Treatments Design (ATD)

The Multielement Design (also called the Alternating Treatments Design [ATD]) is characterized by the rapid, semi-random alternation of two or more distinct independent variables or conditions across sessions, days, or times of day.

Structural Mechanics:

• Conditions are paired with distinct discriminative stimuli (e.g., different therapists, colored rooms, or visual cue cards) to facilitate discrimination.
• The alternation occurs rapidly (e.g., Condition A in the morning, Condition B in the afternoon; or Condition A on Monday, Condition B on Tuesday).
• No Initial Baseline Required: An ATD can compare interventions immediately from Day 1 without an extended baseline phase.

Demonstration of Experimental Control:

Experimental control is demonstrated by vertical separation (differentiation) between the data paths of the different conditions. If the data path for Treatment 1 consistently plots significantly higher (or lower) than Treatment 2 with minimal overlap, experimental control is confirmed.

                             MULTIELEMENT / ATD GRAPH
                 ▲
                 │   ●───────●───────●───────●───────●  (Condition A: High Rate)
  Target         │
  Behavior       │      (Clear Vertical Separation = Experimental Control)
                 │
                 │   ▲───────▲───────▲───────▲───────▲  (Condition B: Low Rate)
                 └────────────────────────────────────────►
                                    Sessions

Key Advantages:

• Does not require withdrawing treatment.
• Highly efficient: allows rapid comparison of multiple treatments within a short timeframe.
• Minimizes sequence effects (order effects) through semi-random alternation.
• Accommodates unstable baseline data.

Primary Limitation: Multiple-Treatment Interference

The primary threat in an ATD is multiple-treatment interference (also known as carryover effects or contrast effects), where the effects of one condition influence the client's responding in the adjacent condition. Clinicians mitigate this by using distinct visual $S^D$s and allowing sufficient temporal spacing between sessions.

---

Changing Criterion Design

The Changing Criterion Design is used to evaluate the effects of an intervention on behaviors that can be shaped or modified in a gradual, stepwise, incremental fashion. It is ideally suited for behaviors that change in rate, frequency, duration, or latency.

Mechanics:

Following an initial baseline phase, the clinician establishes a specific performance criterion for reinforcement (e.g., completing 5 math problems). Once the client demonstrates stable responding at that criterion level, the criterion is adjusted upward (e.g., 8 problems), and then upward again (e.g., 12 problems).

Demonstration of Experimental Control:

Experimental control is proven when the client's behavioral data closely match each successive criterion level, stabilizing at each shift before moving to the next.

• The Bidirectional Change Technique: To provide indisputable proof of experimental control, the clinician intentionally introduces at least one bidirectional change—temporarily shifting the criterion backward to a previous level. If the client's behavior immediately reverses to match this lower criterion and then climbs again when the criterion is increased, all alternative hypotheses (such as maturation or general practice effects) are conclusively ruled out.

Guidelines for Design Validity:

1. Length of Phases: Each criterion phase must be long enough to establish stable responding (typically at least 3-5 sessions).
2. Magnitude of Criterion Shifts: Shifts must be large enough to be visually detectable on a graph, but small enough to avoid ratio strain or client failure.
3. Varying Phase Lengths: Phase durations should vary (e.g., Phase 1 is 4 sessions, Phase 2 is 6 sessions) to prove that behavioral changes are tied to criterion shifts rather than arbitrary calendar time.

---

Comparative Analysis of Single-Case Experimental Designs