Overview of Multiple Baseline Designs
Multiple baseline designs are among the most widely used single-subject research designs in applied behavior analysis. They are especially valuable because they do not require withdrawal of treatment, making them practical and ethical choices for clinical settings. The core logic involves collecting baseline data on two or more baselines simultaneously and introducing the intervention to each baseline at staggered intervals.
If behavior changes in each baseline only when the intervention is introduced to that specific baseline (and not before), this provides a convincing demonstration that the intervention, rather than extraneous variables, is responsible for the change.
The Logic of Staggered Introduction
The power of the multiple baseline design depends on the staggered introduction of the independent variable. Baseline data are collected concurrently on all baselines. The intervention is first applied to one baseline while the others remain in baseline conditions. When the treated baseline shows improvement and the untreated baselines remain stable, the researcher has initial evidence of a functional relationship.
The intervention is then applied to the second baseline. If change occurs only upon introduction of treatment, and the remaining baseline(s) stay at baseline levels, the evidence grows stronger. This process continues until all baselines have received the intervention. Three or more baselines are generally recommended to establish a convincing demonstration of experimental control.
The key requirement is that baselines must remain independent of one another. If treating one baseline causes changes in an untreated baseline (covariation), the demonstration of experimental control is compromised.
Multiple Baseline Across Behaviors
In this variation, the researcher measures two or more different behaviors of the same individual. The intervention is applied to each behavior in a staggered fashion. For example, a practitioner might target hand raising, on-task behavior, and homework completion for a single student. The intervention is first applied to hand raising while monitoring the other two behaviors. Once hand raising improves, the intervention is extended to on-task behavior, and so on.
Key requirement: The behaviors must be functionally independent. This means that treating one behavior should not cause changes in the others. If behaviors share the same function or are part of a response class, changes may generalize across behaviors, weakening the design.
Practical considerations: Selecting truly independent behaviors can be challenging. Consider the maintaining variables for each behavior. If two behaviors are both maintained by teacher attention, treating one may affect the other.
Multiple Baseline Across Settings
Here, the same behavior is measured for the same individual across two or more settings. The intervention is introduced in each setting at different times. For instance, a behavior analyst might measure disruptive behavior during math class, reading time, and recess. The intervention is first implemented during math, then reading, then recess.
Key requirement: The behavior should not generalize across settings without direct intervention. If treating the behavior in one setting causes improvement in untreated settings, the staggered introduction loses its value as a control. This can occur if the settings share common stimuli or if the individual generalizes the contingency.
Practical applications: This design is useful when you want to demonstrate that an intervention works in multiple environments. It is commonly used in school settings where the same student exhibits problem behavior in different classrooms or activities.
Multiple Baseline Across Participants
This variation applies the same intervention for the same behavior across two or more individuals. The intervention is introduced to each participant at different times while all participants' behaviors are measured continuously.
Key requirement: The participants should be in relatively independent situations. If participants are in the same classroom and can observe each other receiving the intervention, vicarious learning could cause premature changes in untreated participants' behavior.
Practical applications: This is one of the most common designs in clinical practice because it allows practitioners to demonstrate effectiveness across multiple clients simultaneously. It is also useful for training studies in which staff members are taught a new procedure at staggered intervals.
When to Choose a Multiple Baseline Over a Reversal
Multiple baseline designs are preferred over reversal designs in several situations:
- Irreversible behaviors: When the target behavior is unlikely to return to baseline levels after treatment (such as learned skills), a reversal design will not demonstrate control effectively.
- Ethical concerns: When withdrawing an effective treatment would pose a risk to the client, such as with aggression or self-injurious behavior.
- Practical constraints: When stakeholders (parents, teachers, administrators) are unwilling to withdraw a treatment that is working.
- Multiple targets: When you are already planning to treat multiple behaviors, work across settings, or serve multiple clients, a multiple baseline design takes advantage of this naturally occurring situation.
Strengths and Limitations
Strengths: No withdrawal of treatment is needed. The design is flexible and practical for applied settings. It can demonstrate generality across behaviors, settings, or participants. It aligns well with typical clinical practice where multiple targets are addressed sequentially.
Limitations: It requires that baselines remain independent, which is not always achievable. Extended baselines for the later tiers can be problematic: individuals may lose motivation, or stakeholders may become frustrated waiting for treatment. If baselines covary, the demonstration of control is weakened. Additionally, it typically requires at least three baselines for a convincing demonstration.
Interpreting Multiple Baseline Data
When analyzing a multiple baseline graph, look for these indicators of experimental control: each baseline should remain stable (or show a predictable pattern) until the intervention is introduced, each baseline should show a clear change coinciding with the introduction of the intervention, and the untreated baselines should not show similar changes at the same time.
If you see a baseline improving before the intervention is introduced to it, consider whether generalization, maturation, or some other confounding variable might be responsible. This pattern weakens the case for a functional relationship.