Null Hypothesis
The default statistical claim a researcher tests, conventionally written H₀. In SLA and language-testing studies it usually states that there is no effect, no difference between groups, or no association between variables in the population — for example, that a treatment and control group come from populations with the same mean post-test score, or that two test versions yield the same difficulty.
Two Traditions
Fisher's Statistical Methods for Research Workers (1925) treated the null as a working hypothesis to be examined against the data, with a small p-value taken as evidence that the data and the null sit poorly together. Neyman and Pearson reframed testing as a decision rule between H₀ and an Alternative Hypothesis, with pre-specified error rates α and β. The hybrid taught in most applied-statistics textbooks — fix α, compute p, "reject" or "fail to reject" — fuses elements of both and has been criticised for obscuring their distinct logics.
Reject vs. Fail to Reject
A null is never "accepted" or "proven true". Failure to reject means the data do not provide enough evidence against it under the chosen α and design, not that an effect is absent. Underpowered studies — common in classroom quasi-experiments — routinely fail to reject false nulls (see Type I and Type II Error).
Specification in SLA
Nulls are often left implicit but should be stated. A Pre-test Post-test Design with treatment and control groups typically pairs a substantive H₁ ("explicit instruction raises gain scores") with the directly tested H₀ ("mean gain in the treatment group equals mean gain in the control group"). Reviews such as Plonsky and Oswald (2014) recommend pairing any null-test result with Effect Size and Confidence Interval reporting, since the null framework alone says nothing about magnitude.
References
- Fisher, R. A. (1925). Statistical Methods for Research Workers. Edinburgh: Oliver and Boyd.
- Larson-Hall, J. (2016). A Guide to Doing Statistics in Second Language Research Using SPSS and R (2nd ed.). New York: Routledge.
- Plonsky, L., & Oswald, F. L. (2014). How big is "big"? Interpreting effect sizes in L2 research. Language Learning, 64(4), 878–912.