Threats to Internal Validity

Factors that confound causal claims

Internal validity is the degree to which an observed outcome can be attributed to the independent variable rather than to extraneous factors. Campbell and Stanley (1963) systematically catalogued eight major threats: history, maturation, testing, instrumentation, statistical regression to the mean, selection, mortality (attrition), and their interactions. By identifying these threats in advance, researchers can design studies with control groups, random assignment, and repeated measurement to rule them out systematically.

What Is Internal Validity?

Internal validity addresses whether changes in the independent variable genuinely explain changes in the dependent variable. If an observed effect could plausibly stem from another source, the causal claim is undermined. Internal validity is a primary concern in intervention studies and experimental designs, where the researcher wishes to assert that X caused Y. The ability to rule out alternative explanations determines the scientific credibility of that assertion.

Major Threats

Campbell and Stanley identified eight major threat categories. History: external events unrelated to the treatment occur during the study and affect outcomes. Maturation: participants grow older, tire, or gain experience over time. Testing: exposure to a pre-test improves post-test performance. Instrumentation: observers or measuring instruments change across time. Statistical regression: extreme scores move toward the mean upon re-measurement. Selection: comparison groups differ at baseline. Mortality (attrition): participants drop out in a systematic, non-random pattern. Diffusion of treatment: the intervention spreads between groups.

A Concrete Example

Consider a researcher testing whether an intervention programme raises student achievement. If schools self-select into the treatment group, selection bias is at work — more motivated students may already populate those schools. If a nationwide exam reform occurs during the study, that is a history threat. If the study runs long enough for students to become more test-savvy, that is maturation. If initially very low scorers naturally improve toward the mean at post-test regardless of treatment, that is statistical regression. If control-group students gain access to intervention materials, diffusion of treatment has occurred.

Ruling Out Threats and Common Pitfalls

Random assignment most effectively eliminates selection bias and selection-maturation interaction. Control groups offset history and maturation threats. Standardised protocols guard against instrumentation drift; avoiding pre-tests or using a Solomon four-group design controls for testing effects. A common pitfall is assuming these threats apply only to experimental studies — observational and mixed-methods research faces them too, and researchers are expected to address them explicitly. Another frequent error is conflating internal validity with external validity (generalisability): design constraints that strengthen internal validity can sometimes limit the generalisability of findings.

Key terms

Internal Validity: Assurance that the observed effect stems from the independent variable alone.
Selection Bias: Systematic baseline differences between comparison groups before the study begins.
Maturation Threat: Natural changes in participants over time that affect the outcome independently.
Regression to the Mean: Tendency of extreme scores to move closer to the mean upon re-measurement.
Mortality / Attrition: Non-random, systematic dropout of participants during the study.