5.3 Baseline Logic and Data-Based Clinical Decision Making

The Epistemology of Baseline Logic in Single-Case Designs

Single-case experimental designs (SCEDs) are the hallmark of applied behavior analysis. Unlike between-group research designs that rely on randomized assignment and inferential comparisons between a treatment group and a control group, single-case methodology utilizes within-subject comparisons. Each individual serves as their own experimental control. Comparing an individual's behavior under the independent variable against their own baseline performance requires a systematic, rigorous inductive logic known as baseline logic.

Formulated by Murray Sidman (1960) in his classic text Tactics of Scientific Research and expanded by Baer, Wolf, and Risley (1968) and Cooper, Heron, and Heward (2020), baseline logic consists of three distinct, cumulative scientific operations: prediction, verification, and replication.

---

The Triad of Baseline Logic: Prediction, Verification, and Replication

To establish experimental control (demonstrating that the independent variable, and only the independent variable, is responsible for the observed change in the dependent variable), the assistant behavior analyst must master how prediction, verification, and replication operate across single-case designs.

1. Prediction

Prediction is the initial component of baseline logic. It is defined as the anticipated outcome of a presently unknown or future measurement.

• Mechanism: The clinician repeatedly measures the target behavior under constant baseline conditions until steady-state responding is achieved (stable level, stable trend, and low variability). Once stability is established, the clinician extrapolates the data path forward into the future, making the scientific prediction: If no environmental manipulations are introduced, the behavior will continue to occur along this projected trajectory.
• Scientific Necessity: Prediction serves as the baseline benchmark against which all future intervention data will be compared. Without an established baseline prediction, there is no standard for determining whether the intervention produced an effect.
• Prerequisites for Valid Prediction: A valid prediction requires a sufficient number of baseline sessions (typically a minimum of 3 to 5 data points) exhibiting steady-state responding. A baseline showing wild bounce or an improving trend cannot generate a believable prediction.

2. Verification

Verification is the second component of baseline logic. It demonstrates that the prior baseline level of responding would have continued along its predicted path had the independent variable not been introduced.

Verification rules out threats to internal validity—specifically extraneous historical events, biological maturation, testing effects, or coincidental environmental changes:

• Verification in an ABAB Reversal Design: Following the intervention phase ($B_1$), the clinician withdraws the independent variable and reinstates the baseline condition ($A_2$). If responding returns to the level and trend observed in the initial baseline phase ($A_1$), verification is achieved! The return to baseline levels confirms that the original baseline prediction was accurate, proving that the behavioral improvement observed in $B_1$ was not caused by maturation or extraneous variables.
• Verification in a Multiple Baseline Design: When the independent variable is applied to Tier 1, Tiers 2 and 3 remain in baseline. If responding in Tiers 2 and 3 remains stable at initial baseline levels while Tier 1 changes, verification is achieved! The untreated tiers prove that time, classroom maturation, or general clinic exposure did not alter the behavior in the absence of the independent variable.

3. Replication

Replication is the third and ultimate component of baseline logic. It is the repeating of the observed experimental manipulation to demonstrate the reliability of the behavior change.

Replication accomplishes two essential scientific objectives: it proves that the behavioral shift is replicable (not a one-time fluke or chance occurrence), and it definitively confirms experimental control:

• Replication in an ABAB Reversal Design: Re-introducing the independent variable in Phase $B_2$ causes behavior to shift once again to the therapeutic level observed in $B_1$. Reproducing the exact behavioral change across two distinct intervention phases provides convincing evidence of functional control.
• Replication in a Multiple Baseline Design: When the independent variable is introduced sequentially to Tier 2 and Tier 3, each tier displays the same immediate behavioral shift seen in Tier 1. Each staggered introduction represents an experimental replication across different behaviors, settings, or participants.

---

Clinical Decision Rules for Phase Transitions

In applied settings, assistant behavior analysts must make critical data-based decisions regarding when to conclude a baseline phase, when to advance an intervention, and when to modify a protocol.

Transitioning from Baseline to Intervention

The timing of the transition from baseline to treatment is governed by four clinical decision rules:

1. Rule of Stability (Steady State): The standard rule is to transition only after baseline demonstrates steady-state responding—characterized by stable level, zero trend, and minimal variability. A stable baseline provides the clearest predictive foundation.
2. Rule of Counter-Therapeutic Trend: If the target behavior is actively worsening during baseline (e.g., severe aggression is accelerating upward, or a functional skill is deteriorating), the clinician should introduce the intervention immediately. Waiting for stability while a client's behavior deteriorates is both clinically negligent and methodologically unnecessary. An intervention that immediately halts and reverses an escalating problem behavior convincingly demonstrates experimental control.
3. Rule of Improving Trend: If the target behavior is already improving during baseline (e.g., self-injury is dropping, or independent compliance is climbing), the clinician must NOT introduce the intervention. The clinician must maintain baseline until responding stabilizes or ceases to improve. Introducing an intervention during an improving baseline confounds the evaluation, making it impossible to determine whether continued progress was caused by the intervention or by pre-existing factors.
4. Ethical Overriding Mandate (Severe Crisis): When a target behavior poses acute physical danger to the client or others (e.g., retinal-detachment head banging, severe biting, running into traffic, or suicidal gestures), ethical responsibilities immediately supersede experimental design preferences. The clinician must implement immediate crisis intervention and safety management without waiting for baseline stability. Adhering to the BACB Ethics Code requires prioritizing client health and safety above research elegance.

---

Data-Based Rules for Modifying Behavioral Interventions

Once an intervention is launched, behavior analysts do not rely on subjective impressions or emotional reactions to judge efficacy. Instead, they establish predetermined decision rules to govern plan revisions.

Setting Predetermined Decision Rules

Predetermined decision rules establish objective, quantitative triggers for action. Common clinical rules include:

• Aim Line Rules: An aim line (target trajectory line) is drawn from baseline to the terminal behavioral goal. A standard decision rule dictates: If 3 to 5 consecutive data points fall below the aim line (for acceleration targets) or above the aim line (for deceleration targets), conduct an immediate procedural review.
• Stall Rules: If no measurable progress toward criterion is observed across two consecutive weeks (or 10 consecutive sessions), initiate a clinical review.

The Integrity-First Mandate: Checking Treatment Fidelity

When graphed data reveal that an intervention is stalling, failing, or exhibiting an adverse trend, the assistant behavior analyst must follow a mandatory sequence of clinical steps:

• Step 1: Assess Treatment Fidelity First: Directly observe and score procedural integrity using an objective fidelity checklist. Evaluate whether staff or caregivers are implementing antecedent modifications, prompting sequences, and consequence delivery accurately.
• If Treatment Fidelity is Low (< 80% to 85%): DO NOT modify the behavioral intervention. The intervention has not failed; it simply has not been tested. The assistant behavior analyst must provide immediate Behavioral Skills Training (BST), model correct implementation, deliver performance feedback, and reassess fidelity until performance meets competency standards (> 90%).
• If Treatment Fidelity is High (> 90% to 95%) and Progress Stalls: Systematically modify intervention parameters. Because the protocol is being executed correctly, the behavioral failure is attributable to the intervention itself. Clinicians systematically adjust reinforcement magnitude, increase reinforcement density, shorten delay to reinforcement, reduce task difficulty or response effort, or reassess motivating operations (MOs).

CRITICAL EXAM PRINCIPLE: A behavior analyst must NEVER modify or discard a behavioral intervention until treatment fidelity (procedural integrity) has been directly measured and confirmed. Altering the intervention package or escalating to more restrictive procedures (such as punishment) when poor implementation fidelity is the root cause violates ethical and behavioral standards.

---

Mastery Criteria, Generalization, and Schedule Thinning

Behavioral programs do not conclude the moment an intervention produces an initial therapeutic shift. Clinicians must establish clear, objective mastery criteria and systematically program for maintenance and schedule fading.

Defining Objective Mastery Criteria

A mastery criterion must be objective, observable, and measurable. Ambiguous phrases such as "client will understand classroom rules" or "client will usually cooperate" are unacceptable. A rigorous mastery criterion specifies:

• Quantitative Accuracy / Rate: E.g., "80% to 90% correct independent responding, or a rate of 40 correct digits per minute."
• Consistency Across Time: E.g., "Across 3 consecutive instructional sessions."
• Generality Across People and Settings: E.g., "Across at least 2 distinct therapists and 2 novel educational environments."

Evaluating Mastery Criteria (Task C.10)

A mastery criterion is a prediction: "if the learner meets this standard, the skill will maintain and generalize." Treat it as something to evaluate, not a fixed rule.

• Accuracy level matters. Studies comparing criteria (e.g., Fuller & Fienup, 2018; Richling, Williams, & Carr, 2019) found that skills mastered to lower accuracy criteria such as 80% were more likely to deteriorate than skills mastered to 90% or higher.
• Number of sessions and observations. One good session can be chance; requiring the criterion across consecutive sessions, or on the first trial of the day ("cold probe"), is more conservative.
• Rate and fluency. For academic and vocational skills, accuracy alone can hide slow, effortful responding. Adding a rate criterion (e.g., correct responses per minute) supports retention, endurance, and application.
• Generalization and maintenance checks. Include different instructors, materials, and settings, and schedule follow-up probes.

If maintenance probes show that "mastered" skills are dropping, the criterion was too lax: raise it for future targets and re-teach the skills that slipped.

Systematically Fading and Thinning Reinforcement

Once mastery criteria are met, the clinician transitions the client from dense, artificial instructional contingencies to naturally occurring reinforcement schedules:

1. Thinning Reinforcement Schedules: Systematically transition from continuous reinforcement (CRF / FR1) to intermittent schedules (FR3, VR5, VI-10 min) to build resistance to extinction and mirror the unpredictability of natural community environments.
2. Fading Prompts: Utilize systematic time delay or graduated guidance to transfer stimulus control from artificial prompts to natural discriminative stimuli ($S^D$).
3. Programming for Maintenance: Periodically conduct maintenance probes over weeks and months to ensure the client maintains behavioral repertoires in the absence of intensive clinical intervention.

---

Baseline Logic Component Matrix

The following table outlines the three core components of baseline logic, their formal definitions, implementation methods in single-case experimental designs, clinical examples, and the threats to internal validity they eliminate: