Internal Validity
Internal validity is the degree to which a study's results can be attributed to the treatment or independent variable rather than to confounding factors. A study with high internal validity allows the researcher to claim that the treatment caused the observed outcome. It is the fundamental requirement for causal inference.
Threats to Internal Validity
Campbell & Stanley (1963) and Shadish, Cook & Campbell (2002) identified several classic threats, all highly relevant to SLA classroom research:
| Threat | Description | SLA example |
|---|---|---|
| Selection | Groups differ before treatment begins | One intact class has higher proficiency than the other |
| Maturation | Participants change naturally over time | Students improve through general exposure, not the treatment |
| History | External events affect outcomes | One class gets extra tutoring outside the study |
| Testing | Taking the pre-test improves post-test performance | Familiarity with the test format inflates gains |
| Instrumentation | Measurement tool or rater changes | Raters become more lenient over time |
| Regression to the mean | Extreme scores move toward the average on re-testing | Low pre-test scorers appear to improve regardless of treatment |
| Attrition | Participants drop out non-randomly | Weaker students in the control group leave |
| Diffusion of treatment | Control group learns about and adopts the treatment | Students in different groups share materials |
Practice-Test Congruency as a Validity Threat
When the treatment activities closely resemble the post-test, apparent learning gains may reflect practice effects rather than genuine acquisition. This is a pervasive threat in SLA intervention research, particularly in studies of Form-Focused Instruction where the treatment tasks and the test items target the same isolated structures.
Internal Validity and Research Design
True experiments with random assignment provide the strongest internal validity because randomisation distributes confounds across groups. Quasi-experiments are more vulnerable because intact groups may differ systematically. Researchers mitigate threats through pre-testing, ANCOVA, delayed post-tests, and transparent reporting.
The Tension with External Validity
Maximising internal validity often means increasing control — using lab settings, standardised tasks, and homogeneous samples. But this reduces External Validity (generalisability to real classrooms) and Ecological Validity (resemblance to authentic conditions). Every study navigates this trade-off.
Key References
- Campbell & Stanley (1963) — original taxonomy of validity threats
- Shadish, Cook & Campbell (2002) — expanded and updated framework
- Norris & Ortega (2000) — internal validity concerns in SLA meta-analysis
- Plonsky (2013) — transparency in reporting threats