Summary
This article follows up the debate on inferential statistics in Volumes 19 and 20 of Educational Research. It is particularly addressed to the researcher working with a non‐random sample, perhaps in a single institution. Attention is drawn to the suitability of randomization tests in such a situation. In these circumstances, researchers sometimes use statistical tests which can only be regarded as descriptive: accordingly a rationale is offered for the non‐inferential use of statistical tests.