EXPERIMENTAL CONTROL AND THREATS TO INTERNAL VALIDITY OF CONCURRENT AND NONCONCURRENT MULTIPLE BASELINE DESIGNS

Introduction to Single-Case Research Designs

  • Application in school psychology.

  • Focus on multiple baseline (MB) designs: concurrent and nonconcurrent.

  • Aim of the article: to evaluate their capacity to assess internal validity and establish experimental control.

Review of Scientific Merit of MB Designs

  • Distinction between AB replications and nonconcurrent MB studies based on Watson and Workman's (1981) conception.

  • Outcomes suggest that internal validity threats can be assessed with both MB designs.

  • Nonconcurrent designs can assess intervening history effects but may be more prone to mortality threats.

  • Recommendations for ensuring internal validity are provided.

Experimental Control and Internal Validity

Definition of Internal Validity

  • Campbell and Stanley (1963): "internal validity is the basic minimum without which any experiment is uninterprettable".

  • Internal validity threats offer alternative explanations for observed effects.

Threats to Internal Validity

  • Eight threats identified by Campbell and Stanley (1963):

    1. History: External events that influence the dependent variable (DV) (e.g., changes in a student’s bedtime affecting behavior).

    2. Maturation: Changes in participant responses over time resulting from aging or development (e.g., cumulative educational experiences).

    3. Testing: Influences of previous measurements on subsequent performance (e.g., practice effects from repeated tests).

    4. Instrumentation: Inconsistencies in measurement tools (e.g., observer drift, varying difficulties in test forms).

    5. Statistical Regression: Extreme scores moving toward mean over time, especially in populations with extreme behaviors.

    6. Mortality: Participants leaving the study prematurely, affecting the results.

    7. Interactions: Effects of multiple threats on internal validity.

Multiple Baseline Design (MB Design)

Conceptual Framework

  • MB design described as a concurrent series of AB designs with staggered phases (Baer et al., 1968; Barlow & Hersen, 1984).

  • Selection of multiple behaviors/settings/subjects that independently respond to manipulations is key.

  • Data collected continuously until stable baselines are established, followed by the introduction of manipulated variables.

Functioning of MB Design

  • Experimental manipulation is engaged for one data series while the others remain in baseline.

  • The goal is to analyze changes in DV relative to interventions across multiple series, demonstrating experimental control through stability in baseline versus changes following intervention.

Concurrent vs. Nonconcurrent MB Designs

Concurrent Design

  • Traditional approach involving simultaneous data collection across multiple subjects or settings.

  • Provides greater experimental control due to simultaneous measurement.

Nonconcurrent Design

  • Nonconcurrent MB design is characterized by temporal disparity between data series (Watson & Workman, 1981).

  • Baseline durations are established a priori and participants are randomly assigned to these baselines.

  • Challenges regarding participant exclusion criteria and the potential for bias when statistically controlling for participants with erratic responses.

Assessing Internal Validity and Experimental Control

Importance of Formative Assessment

  • Ongoing assessment is crucial for establishing robust data sets within MB designs.

  • Should emulate conditions across measures to reduce instrumentation threats and enhance reliability.

Concerns with Mortality and Statistical Regression

  • Mortality as a concern in nonconcurrent designs related to baseline data variability.

  • Statistical regression less a concern with thorough formative assessment since extreme behaviors in clinical populations may not reflect systematic regression.

History Effects and Data Series

  • Concurrent designs can more readily assess history effects as changes across data series can indicate external influences.

  • Nonconcurrent designs face criticism regarding historical threats as data is collected at separate times.

Summary and Recommendations

  • Both concurrent and nonconcurrent MB designs contribute meaningfully to experimental research.

  • Importance of a priori establishment of experimental manipulations, hypotheses, data collection durations, and schedules is emphasized.

  • The article supports that both designs have merits for contributing robust findings to scientific literature despite the ongoing debates.

References

  • Comprehensive list of literature supporting the content:

    • Baer, Wolf, & Risley (1968)

    • Barlow & Hersen (1984)

    • Campbell & Stanley (1963)

    • Carr (2005)

    • Cuvo (1979)

    • Harris & Jenson (1985)

    • Hayes (1976, 1981)

    • Johnston & Pennypacker (1993)

    • Kazdin (1981, 1982)

    • Kazdin & Kopel (1975)

    • Murphy & Bryan (1980)

    • Tawney & Gast (1984)

    • Watson & Workman (1981)