Multiple-Baseline Designs Across Participants, Behaviors, or Settings: Which Should You Use?

Should You Stagger an Intervention Across People, Behaviors, or Environments?

Suppose you want to evaluate a digital prompting intervention that supports independent academic task completion. You have access to three students, each of whom could receive the intervention at a different time.

Alternatively, you could study one student across three target behaviors, introducing the intervention sequentially for task initiation, task completion, and assignment submission.

A third possibility is to introduce the intervention in different instructional settings, such as a classroom, computer laboratory, and independent study environment.

All three arrangements appear to follow multiple-baseline logic. Each involves repeated measurement and staggered intervention introduction.

However, the experimental implications differ. Students may share strategies, behaviors may influence one another, and learning acquired in one setting may transfer to another.

Which arrangement provides the most credible evidence that the intervention causes the observed change?

Choose Tiers That Permit Meaningful and Sufficiently Independent Experimental Contrasts

How Participant, Behavior, and Setting Tiers Differ

What does across participants mean?

A multiple-baseline design across participants uses different individuals as experimental tiers.

Each participant's outcome is measured repeatedly, and the intervention begins at a different planned time.

For example, Student A may begin digital prompting after five baseline sessions, Student B after eight, and Student C after eleven.

The researcher examines whether each student's performance changes following their own intervention introduction while students awaiting treatment remain near their existing baseline patterns.

This arrangement is particularly useful when the research question concerns whether an intervention produces comparable functional effects across different individuals.

However, participants should not receive meaningful intervention exposure before their scheduled introduction, and their outcomes must be measured consistently.

What does across behaviors mean?

A multiple-baseline design across behaviors uses different target behaviors as experimental tiers, often within the same participant.

For example, a researcher may measure task initiation, independent task completion, and assignment submission.

The intervention is introduced for one behavior while the others remain under baseline conditions. Subsequent introductions occur at different times.

This arrangement may be useful when the intervention can be applied selectively to distinct behaviors.

However, the behaviors must be sufficiently separable for the staggered comparisons to remain meaningful.

If improving task initiation automatically improves task completion and submission, untreated behaviors may change before their own intervention phases.

That generalization may be educationally desirable but can undermine the intended experimental contrast.

What does across settings mean?

A multiple-baseline design across settings uses different environments or contexts as experimental tiers.

For example, a researcher might introduce an instructional support procedure in a classroom, computer laboratory, and independent study environment at different times.

The same participant or group may be observed across these settings, depending on the research question.

This arrangement may be appropriate when the intervention can be implemented separately in each environment and outcomes can be measured comparably.

However, changes acquired in one setting may transfer to another.

For example, a student who learns a general task-management strategy in the classroom may begin using it independently in the laboratory before the laboratory intervention begins.

Researchers should anticipate such transfer when selecting settings as tiers.

How do the three arrangements compare?

The appropriate arrangement is the one whose tiers can provide meaningful demonstrations of intervention effects while answering the intended question.

Why does tier independence matter?

Multiple-baseline inference depends partly on observing outcomes in tiers that have not yet received the intervention.

If untreated tiers change because of treatment introduced elsewhere, the staggered contrast becomes less informative.

For example, students may share an instructional strategy, or a teacher may unintentionally apply a new procedure to every student.

Similarly, improving one behavior may influence related behaviors, and learning in one setting may transfer to another.

Researchers should therefore consider whether intervention effects can remain sufficiently localized during the experimental sequence.

Independence does not require participants or outcomes to be statistically unrelated in every respect. It requires that the intervention can be introduced and evaluated across tiers without uncontrolled exposure undermining the planned comparisons.

Does generalization always weaken a multiple-baseline design?

Generalization can complicate the experimental logic when it occurs before the intervention is formally introduced in an untreated tier.

For example, a student may learn a strategy for organizing assignments and begin applying it to several related tasks.

From an educational perspective, such transfer may represent a desirable outcome.

However, if the researcher planned to introduce the strategy separately for each task, improvement across all tasks after the first introduction reduces the ability to demonstrate staggered effects.

The researcher should distinguish an intervention that generalizes successfully from a design that provides adequate experimental control.

When generalization is a primary research objective, it may be better to measure transfer explicitly as an outcome rather than rely on untreated behaviors as independent baseline tiers.

When should researchers prefer participants over behaviors?

Across-participant designs may be preferable when the intervention is expected to influence several related behaviors simultaneously.

For example, teaching a comprehensive self-regulation strategy may improve task initiation, completion, and submission together.

Using those behaviors as separate tiers could be problematic because the intervention mechanism predicts cross-behavior generalization.

Staggering the intervention across different students may provide more distinct contrasts, provided students do not receive the strategy indirectly.

However, differences in student characteristics and response patterns must still be considered.

The objective should determine whether replication across individuals or selective change across behaviors is more informative.

When should researchers prefer behaviors over participants?

Across-behavior designs may be appropriate when only one participant is available and several sufficiently distinct behaviors can be targeted independently.

For example, an intervention may be designed to address three unrelated classroom behaviors using separately delivered procedures.

Researchers can introduce the intervention sequentially while measuring all behaviors throughout the study.

However, the behaviors should be selected because they support the experimental logic, not merely because three measurable outcomes are available.

Highly related behaviors may change together even without separate intervention introductions.

The researcher should examine the likely mechanisms linking the behaviors before selecting them as tiers.

When should researchers prefer settings?

Across-setting designs may be useful when the same behavior occurs in distinct environments and intervention delivery can be restricted to one setting at a time.

For example, a student may perform independent learning tasks in a classroom, laboratory, and supervised study area.

The researcher could introduce prompting separately in each environment.

However, the settings should offer comparable opportunities for the target behavior and sufficiently distinct intervention exposure.

If the student learns a transferable strategy in the first setting, subsequent baseline measurements elsewhere may change before treatment introduction.

Researchers should also consider whether differences in instructors, tasks, or environmental demands complicate comparisons.

How many tiers should be included?

Three tiers are common in multiple-baseline experiments because they can provide several opportunities to demonstrate an effect at different points in time.

However, three tiers do not automatically satisfy rigorous design standards.

Each tier must contribute sufficiently informative observations, and intervention introductions must be separated in a way that supports meaningful contrasts.

The What Works Clearinghouse single-case design standards emphasize demonstrations of effect at different points in time, along with relevant phase and measurement requirements.

Researchers should consult the applicable standards when determining the number of tiers and observations.

Adding tiers may strengthen replication but also increases measurement demands and may delay intervention for some participants.

Must the outcome be identical across tiers?

Not necessarily, particularly in an across-behavior design where the behaviors are inherently different.

However, each outcome should be operationally defined and measured consistently within its tier.

For example, task initiation may be measured through response latency, while assignment completion may be measured as a proportion of required steps.

These measures should not be treated as directly interchangeable merely because both are displayed on graphs.

Researchers should evaluate whether the expected intervention effect is interpretable across tiers and whether measurement differences could explain apparent patterns.

For across-participant designs, comparable outcome definitions are often particularly useful for evaluating replication.

How should baseline lengths differ?

Baseline lengths should be staggered to create distinct intervention introduction points.

For example, one participant may receive treatment after five sessions, another after eight, and another after eleven.

However, these values are illustrative rather than universal minimum requirements.

The spacing should reflect outcome variability, anticipated treatment response, measurement frequency, and ethical considerations.

Researchers should avoid choosing intervention start points retrospectively because particular outcomes happen to improve.

Prespecified or appropriately randomized start points may strengthen design credibility.

Further discussion of the broader design logic appears in when multiple-baseline designs are preferable.

What if untreated tiers begin improving before intervention?

Improvement before intervention introduction may indicate an existing baseline trend, generalization, contamination, or another concurrent influence.

For example, if task completion improves across all three behaviors immediately after an intervention targeting the first behavior, the untreated tiers no longer provide unchanged baselines.

Researchers should investigate whether the pattern reflects the intervention mechanism or unrelated events.

Such improvement should not automatically be described as a methodological failure, particularly when generalization is educationally desirable.

However, it may weaken the intended demonstration of experimental control.

The final interpretation should distinguish evidence of broad improvement from evidence that the staggered design successfully isolated the intervention effect.

Can participants be studied during different calendar periods?

Yes. Nonconcurrent multiple-baseline arrangements may be used when participants become available at different times.

For example, Student A may be studied during one semester, while Students B and C enter the program in later semesters.

Such designs can improve feasibility when simultaneous recruitment is impossible.

However, different calendar periods may involve different instructional practices, institutional policies, or other historical conditions.

These differences can weaken protection against history-related explanations.

Researchers should document relevant contextual changes and consider whether randomizing intervention start points within permissible windows is feasible.

Nonconcurrent arrangements should not be described as equivalent to concurrent designs without examining their inferential limitations.

How should intervention fidelity be monitored across tiers?

Researchers should verify that the intervention is implemented as intended in treated tiers and that untreated tiers do not receive meaningful elements prematurely.

For example, instructors may use a standardized procedure to document when digital prompts are activated for each student.

In an across-setting design, fidelity monitoring may establish whether the intervention remains restricted to the designated environment.

In an across-behavior design, researchers may need to document whether the intervention procedures are genuinely behavior-specific.

Fidelity evidence helps distinguish an ineffective intervention from inadequate implementation and helps identify contamination.

However, high fidelity alone does not establish a causal effect. The observed outcome patterns must still support the experimental contrasts.

What ethical considerations differ across arrangements?

Across-participant designs may require some individuals to wait longer than others before receiving a potentially beneficial intervention.

Across-behavior designs may delay intervention for certain target behaviors within the same participant.

Across-setting designs may leave a participant without the intervention in some environments while receiving it elsewhere.

These arrangements can have different consequences for participant welfare.

For example, withholding support for a behavior associated with serious risk may be inappropriate even if other behaviors are being treated.

Researchers should assess the consequences of each planned delay and preserve necessary ordinary care or educational accommodations.

The absence of intervention withdrawal does not automatically make staggered introduction ethically acceptable.

Can multiple-baseline designs be combined with generalization assessments?

Yes. Researchers may use one set of tiers to establish experimental control and collect additional measurements to examine whether intervention effects transfer to other behaviors or settings.

For example, an across-participant design could evaluate independent task completion while separately monitoring whether acquired strategies generalize to unfamiliar tasks.

The generalization measures would address an additional outcome rather than necessarily functioning as untreated experimental tiers.

This distinction may be useful when the intervention is explicitly designed to promote transfer.

Researchers should identify which measurements support causal demonstration and which evaluate generalization or maintenance.

Combining these purposes without distinction can create confusion about what the experimental design establishes.

How should the data be analyzed?

Visual analysis commonly examines level, trend, variability, immediacy, overlap, and consistency of intervention-related changes across tiers.

Researchers should graph observations in chronological order, identify intervention start points, and preserve the relevant time scale.

For across-participant designs, the analysis examines whether individual outcomes change following their respective introductions.

For across-behavior or across-setting designs, it examines whether targeted behaviors or environments change in the planned sequence while untreated tiers remain interpretable.

Statistical methods may supplement visual analysis, but they must respect the temporal structure and experimental arrangement.

Researchers should not treat repeated observations as independent participants or assume that a difference between phase means establishes experimental control.

How should researchers decide which arrangement is strongest?

The strongest arrangement is the one that creates credible experimental contrasts for the intended intervention and outcome.

Across-participant designs may be preferable when distinct individuals can receive treatment separately and cross-behavior generalization is expected.

Across-behavior designs may be useful when target behaviors can be manipulated independently within one participant.

Across-setting designs may be appropriate when intervention delivery can be restricted to distinct environments.

However, no arrangement is universally superior.

Researchers should consider tier independence, outcome measurement, intervention fidelity, ethical constraints, and the intended scope of inference.

Selection should occur before data collection, not after examining which arrangement would make the observed effects appear most convincing.

Choosing the Most Defensible Tiers for a Digital Learning Intervention

The across-participant design is preferable in this hypothetical situation because the intervention is expected to influence several behaviors and settings simultaneously.

However, the choice remains conditional on students being sufficiently distinct experimental tiers. If they share intervention materials or receive prompting indirectly, contamination may still undermine the comparison.

The example illustrates why tier selection should follow the intervention mechanism rather than the convenience of having three participants, three behaviors, or three settings available.

Common Misconceptions About Multiple-Baseline Tier Selection

Select Tiers According to the Intervention's Expected Effects

Begin by identifying whether the intervention is expected to influence individuals, particular behaviors, or performance in specific environments.

Then evaluate whether treatment can be introduced separately across the proposed tiers without uncontrolled exposure or unacceptable delays.

Tier selection should also be coordinated with baseline stability planning and replication requirements in single-case research.

The strongest arrangement is not necessarily the easiest to organize. It is the one that provides the clearest defensible relationship between intervention timing and outcome change.

Before Choosing Participants, Behaviors, or Settings as Tiers

Frequently Asked Questions About Multiple-Baseline Tier Arrangements

Choose the Tier Structure That Best Preserves Experimental Control

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.