Normality Ignored: The Misuse of Parametric Statistics in Postgraduate Psychology Research

Parametric procedures such as the t-test, ANOVA, Pearson’s r, and ordinary least squares regression anchor most quantitative work submitted at the postgraduate level in psychology. Students find them familiar, the software makes them painless, and reviewers expect them. Each procedure carries distributional commitments, though, and normality is the one most often invoked and least often examined with care. This commentary sets out how that neglect appears in dissertation practice, why it persists, and what might be done about it. Five recurring problems are described: assumption checks are skipped outright; skewness and kurtosis thresholds are read in isolation; Shapiro–Wilk output is treated as a verdict rather than as evidence; single-item Likert responses are analysed as interval scores without justification; and sample sizes above an arbitrary cut-off are taken to settle the question of distributional fit. These habits are then placed against the institutional pressures that sustain them, including reliance on departmental templates, the way software fluency is mistaken for statistical reasoning, and the continuing premium attached to statistical significance. Recommendations follow at programme level and at supervisor level, with attention to transparent reporting and to the difference between teaching thresholds and teaching judgement. The argument throughout is that assumption checking belongs inside the analysis rather than alongside it, and that postgraduate training in psychology should be redesigned with that point in mind.