Historical Context & Motivation
The formalization of research design as a distinct methodological concern arose from the need to separate credible scientific evidence from anecdote, opinion, and coincidence. For centuries, clinicians relied on case observations and philosophical reasoning to guide treatment, but the absence of systematic controls meant that conclusions about cause and effect remained deeply unreliable. The behavioral and health sciences were particularly vulnerable to this limitation because human behavior is influenced by an enormous number of interacting variables—biological, psychological, social, and contextual—making causal attribution uniquely challenging.
As experimental methods from the natural sciences were adapted for the study of human behavior, researchers confronted ethical and practical constraints that demanded new design strategies. You cannot randomly assign participants to experience trauma, withhold effective treatments indefinitely, or manipulate variables like socioeconomic status at will. These constraints fueled the development of multiple design types—each representing a different trade-off between internal validity (the ability to infer causation) and external validity (the ability to generalize findings to broader populations and settings).
The central question that this lesson addresses is deceptively simple: How do we decide what kind of study to conduct—or how to evaluate a study that has already been conducted—when the goal is to understand human behavior and inform clinical practice? Answering this question requires a thorough understanding of four major design types—experimental, quasi-experimental, correlational, and case study—and the conditions under which each is appropriate, powerful, or limited.
Core Principles & Definitions
Before differentiating among the four major designs, it is essential to understand the foundational principles that define and distinguish them. Every research design can be evaluated along several dimensions: the degree of control the researcher exercises over variables, the method of participant assignment, the type of conclusions that can be drawn, and the ethical and practical constraints that shape its use. These dimensions form a coherent framework for understanding why certain designs are preferred for specific research questions and why no single design is universally superior.
Manipulation of Variables
Random Assignment
Control Over Confounds
Causal Inference
Ecological Validity & Practicality
Visual Explanation: The Design Continuum
The continuum depicted above is not a strict hierarchy of quality; rather, it illustrates the trade-offs inherent in selecting a research design. A true experiment maximizes internal validity by controlling for confounds through randomization and manipulation, but it may occur in an artificial laboratory setting that limits generalizability. Conversely, a case study conducted in a clinical setting captures behavior as it naturally occurs, offering strong ecological validity and clinical richness, but it cannot rule out alternative explanations for the observed outcomes. The quasi-experimental and correlational designs occupy intermediate positions, each with distinctive strengths and vulnerabilities that are explored in the sections that follow.
How Each Design Works: Structural Deep Dive
True Experimental Design
The true experimental design (also called a randomized controlled trial or RCT in clinical contexts) is defined by three features: (1) the researcher manipulates at least one independent variable, (2) participants are randomly assigned to conditions, and (3) a control group is included for comparison. Random assignment is the critical distinguishing element because it is the only procedure that, on average, equates groups on all possible confounds—both measured and unmeasured. Classic examples in behavioral health include randomizing clients to CBT versus a wait-list control and measuring symptom reduction at post-treatment.
Quasi-Experimental Design
A quasi-experimental design resembles a true experiment in that the researcher manipulates an independent variable, but it lacks random assignment. Participants arrive in pre-existing, intact groups—such as classrooms, hospital wards, or demographic categories—and the researcher applies the intervention to one group while using another as a comparison. The most common quasi-experimental designs include the nonequivalent control group design and the interrupted time-series design. Because groups may differ systematically before the intervention, the primary threat to internal validity is selection bias, along with selection-by-maturation and selection-by-history interactions.
Correlational Design
In a correlational design, the researcher measures two or more variables as they naturally occur, without manipulating any of them, and examines the statistical relationship between them. Correlational research can reveal the direction (positive or negative) and magnitude of an association. However, because there is no manipulation and no random assignment, the design cannot establish causation. The classic adage—correlation does not imply causation—captures the central limitation: a third variable may be responsible for the observed association, and the direction of influence is ambiguous.
Case Study Design
The case study is an in-depth investigation of a single individual, family, group, or organization. It typically employs multiple data sources—interviews, behavioral observations, psychometric testing, archival records—to construct a comprehensive portrait. Case studies are invaluable for studying rare phenomena, generating hypotheses, and illustrating theoretical principles in clinical practice. Freud's case of 'Little Hans,' Broca's study of patient 'Tan,' and Sacks' neurological narratives are canonical examples. The case study's principal limitation is its inability to support generalization to other individuals or contexts, a weakness rooted in its lack of sampling and absence of controlled comparison.
Detailed Classification & Decision Flowchart
Accurately classifying a study design requires systematic evaluation of its structural features. The decision flowchart below provides a step-by-step algorithm that mirrors the reasoning process tested on the EPPP. By answering a series of yes/no questions about manipulation, assignment, and sample characteristics, you can reliably determine which of the four major design types applies to any given study scenario.
| Feature | True Experiment | Quasi-Experiment | Correlational | Case Study |
|---|---|---|---|---|
| IV Manipulation | Yes | Yes | No | No |
| Random Assignment | Yes | No | No | No |
| Causal Inference | Strong | Moderate | Weak | Very Weak |
| Typical Sample Size | Moderate to large | Moderate to large | Large | 1 to few |
| Primary Threat | Low ecological validity | Selection bias | Third-variable problem | Poor generalizability |
| Example in BH | RCT comparing CBT vs. medication for anxiety | Comparing schools that adopted an anti-bullying program vs. those that did not | Survey examining the relationship between stress and burnout in clinicians | Detailed analysis of treatment for a patient with rare dissociative disorder |
Worked Example: Classifying a Study
Consider the following research scenario, similar to what you might encounter on the EPPP:
Strengths, Limitations, and When to Use Each Design
No single research design is inherently superior to another; rather, each is suited to particular research questions, ethical constraints, and practical circumstances. The table below summarizes the principal strengths and limitations of each design type, followed by guidance on when each is most appropriately used in behavioral health research.
| Design | Strengths | Limitations |
|---|---|---|
| True Experiment | Strongest basis for causal inference; controls for known and unknown confounds through randomization; considered the gold standard for evaluating treatment efficacy | May be ethically or practically impossible (e.g., cannot randomly assign people to abuse); often conducted in artificial settings reducing ecological validity; can be expensive and time-consuming |
| Quasi-Experiment | Feasible when randomization is impossible; can be conducted in real-world settings; permits evaluation of programs and policies as naturally implemented; higher ecological validity than lab experiments | Vulnerable to selection bias and related threats; cannot fully establish causation; requires careful statistical control and awareness of plausible alternative explanations |
| Correlational | Can study variables that cannot be ethically manipulated; efficient for large samples; useful for identifying patterns, making predictions, and generating hypotheses; often higher ecological validity | Cannot establish causation (third-variable problem, directionality problem); correlations can be misleading if the relationship is nonlinear; often relies on self-report measures |
| Case Study | Provides rich, detailed data; ideal for rare conditions or phenomena; can challenge or support existing theories; clinically practical and directly relevant to practice | Cannot generalize to populations; highly susceptible to researcher bias; no control group; cannot establish causation; findings are idiographic rather than nomothetic |
Connection to Validity Frameworks and Advanced Designs
The four design types discussed in this lesson are the foundation upon which more advanced and nuanced designs are built. Understanding these basic categories positions you to appreciate the sophisticated methodological variations you will encounter in the research literature and on the EPPP. Cook and Campbell's (1979) validity framework provides a lens through which any design can be evaluated, and more recent developments—particularly in the evidence-based practice movement—have expanded the conversation about what constitutes rigorous evidence in behavioral health.
| Basic Design | Advanced Extensions | EPPP Relevance |
|---|---|---|
| True Experiment | Factorial designs (2×2, 2×3), Solomon four-group design, crossover/within-subjects designs, dismantling studies, dose-response experiments | Recognizing factorial notation, understanding main effects vs. interactions, identifying placebo-controlled RCTs |
| Quasi-Experiment | Regression discontinuity, propensity score matching, interrupted time-series with comparison, switching-replications design | Distinguishing quasi-experimental subtypes, understanding that regression discontinuity can approach causal strength of RCTs |
| Correlational | Multiple regression, path analysis, structural equation modeling (SEM), longitudinal panel designs, cross-lagged panel models | Understanding that advanced correlational techniques can model directionality but still cannot prove causation without manipulation |
| Case Study | Single-case experimental designs (ABAB, multiple baseline, alternating treatments), qualitative case analysis, mixed-methods case designs | Recognizing that single-case experimental designs include manipulation and can demonstrate functional relationships, unlike traditional descriptive case studies |
Looking forward, the field of behavioral health increasingly values mixed-methods approaches that integrate quantitative designs (experiments, correlational studies) with qualitative methods (case analyses, phenomenological interviews) to capture both the causal mechanisms and the lived experience of participants. Understanding the foundational design types covered in this lesson is essential for evaluating and conducting such integrative research. Additionally, the hierarchy of evidence used in evidence-based practice places systematic reviews and meta-analyses of RCTs at the top, followed by individual RCTs, quasi-experiments, correlational studies, and finally case studies—reinforcing the importance of knowing where each design sits in terms of evidential strength.
Practice Problems
Summary
Research designs in behavioral health exist along a continuum defined by the degree of researcher control over variables. True experimental designs (RCTs) provide the strongest basis for causal inference because they incorporate both manipulation of an independent variable and random assignment of participants to conditions, thereby controlling for both known and unknown confounds. Quasi-experimental designs retain manipulation but lack random assignment, making them susceptible to selection bias yet practical for evaluating real-world programs where randomization is unfeasible. Correlational designs measure naturally occurring associations without manipulation, making them efficient for large-scale surveys and hypothesis generation but unable to establish causation due to the third-variable and directionality problems.
Case studies provide the richest clinical detail through in-depth analysis of one or very few participants, ideal for rare conditions and hypothesis generation, but they lack the controls needed for causal or generalizable conclusions. When classifying a study on the EPPP, ask two key questions: (1) Was an IV manipulated? and (2) Were participants randomly assigned? These two questions, combined with attention to sample size and study goals, will reliably identify the design type. Remember that single-case experimental designs (e.g., ABAB) are classified as experiments, not case studies, because they incorporate systematic manipulation and replication logic to demonstrate experimental control.