Quasi-Experimental Design
A quasi-experimental design shares the logic of a true experiment — treatment group, comparison group, pre/post measurement — but lacks random assignment. Participants are allocated to conditions based on pre-existing groupings, typically intact classes. This is the most common design in classroom SLA research because random assignment of individual students is rarely feasible in school settings.
Why Quasi-Experiments Dominate Classroom Research
Schools assign students to classes for administrative, logistical, or pedagogical reasons — not research purposes. Researchers must work with these intact groups. The result: groups may differ systematically before the treatment begins (e.g., one class may have higher motivation, a different teacher, or more hours of prior instruction). These pre-existing differences are the central threat to Internal Validity.
Common Designs
| Design | Structure | Notes |
|---|---|---|
| Non-equivalent control group | Intact treatment + intact comparison, pre/post-test | The workhorse of classroom SLA research |
| Time-series | Repeated measures before and after treatment | Useful when no comparison group is available |
| Counterbalanced | Groups receive treatments in different orders | Controls for sequence effects |
| Regression discontinuity | Assignment based on a cut-off score | Rarely used in SLA but strong for causal inference |
Threats to Internal Validity
Because groups are not randomly formed, several confounds can explain post-treatment differences:
- Selection bias — groups differ before the study begins
- Maturation — natural development over time, independent of treatment
- History — events outside the study affect one group differently
- Testing effect — taking a pre-test improves post-test performance (see Pre-test Post-test Design)
- Instrumentation — changes in the measurement tool or rater standards
- Attrition — differential dropout between groups
- Practice-Test Congruency — treatment activities resemble the post-test, inflating apparent gains
Mitigating Threats
Researchers strengthen quasi-experimental studies through:
- Pre-testing — documenting baseline equivalence (or using ANCOVA to adjust for differences)
- Delayed post-tests — testing durability of effects beyond immediate post-treatment
- Multiple comparison groups — adding a second control or alternative treatment
- Triangulation — combining quantitative outcomes with qualitative data (see Triangulation)
- Transparent reporting — disclosing group differences and potential confounds
Reporting Standards
Norris & Ortega (2000) and Plonsky (2013) called for SLA researchers to report effect sizes, confidence intervals, and detailed participant information to make quasi-experimental findings interpretable and meta-analysable.
Key References
- Campbell & Stanley (1963) — taxonomy distinguishing true from quasi-experimental designs
- Shadish, Cook & Campbell (2002) — comprehensive treatment of validity threats
- Mackey & Gass (2005) — Second Language Research: Methodology and Design