John Hattie

New Zealand educational researcher (John Allan Clinton Hattie, born 1950), Laureate Professor and Director of the Melbourne Education Research Institute at the University of Melbourne. Trained at the University of Otago, ran an early career in educational measurement at the University of North Carolina at Greensboro, then at the University of Auckland from the early 1990s until the move to Melbourne in 2011. The work for which he is most widely known — and the work that matters for this topic — is the Visible Learning programme: a meta-meta-analysis project synthesising effect sizes across more than a thousand individual meta-analyses to produce a ranking of educational interventions by their average effect on student achievement.

Stake§

Hattie's stake is the public-figure one of an academic whose popular reach has substantially exceeded the academic-research community's level of consensus on his methods. Visible Learning (Routledge, 2008) ranked roughly 138 educational interventions by effect size; the book and its sequels became one of the most widely sold works in education through the 2010s, with school districts in several countries citing the Hattie rankings in their improvement strategies. The methodological objections to the meta-meta-analytic approach — raised by Pierre-Jérôme Bergeron and others through the 2010s — have not had the policy reach that the rankings themselves have. The 2007 Power of Feedback paper that matters for the topic predates the Visible Learning programme and is more modest in its claims.

The work that anchors Hattie's presence in this topic is the 2007 Review of Educational Research paper The Power of Feedback, written with Helen Timperley. The paper is a synthesis of the feedback-effects literature organised around two questions: which kinds of feedback help, and what conditions determine whether feedback that should help actually does. The answer the paper gives is a four-level model. Feedback can address the task (was the answer correct, what was missing); the process of doing the task (what strategy was used, how could it be improved); the self-regulation behind the work (how the learner monitored their own progress, when to seek help); or the self as a whole (praising or criticising the person, not the work). The first three help; the fourth, on the empirical evidence the paper reviews, hurts.

The paper's policy reach has been substantial. The four-level model is in the vocabulary of most contemporary teacher-education programmes and most school-improvement initiatives that put feedback at the centre. The construct is in regular conversation with Sadler's gap-closing account of formative assessment and with Ruth Butler's 1988 study showing that scores attached to feedback shift students toward the self level and undermine the learning effect. The Hattie–Timperley synthesis is the consolidating reference; the empirical results are not original to it but the organising frame has held.

The later Visible Learning programme has been more contested. Visible Learning (Routledge, 2008) reported effect sizes for 138 educational interventions averaged across roughly 800 meta-analyses, with a recommended hinge point (an effect size of 0.4) above which interventions were claimed to be worth doing. The methodology has been criticised on multiple grounds: that averaging effect sizes across studies that measured different outcomes is conceptually problematic; that the original meta-analyses varied substantially in quality and were not adjusted for; that the hinge point is a heuristic rather than a defensible statistical threshold; that the rankings ignore the substantial variation in effects across student populations and contexts. Hattie has answered these in subsequent volumes (Visible Learning for Teachers in 2011, Visible Learning MetaX in 2017); the methodological debate remains active in the journals.

The two bodies of work — the 2007 feedback paper and the Visible Learning programme — should be read separately; the methodological literature on each is different, and the critiques of Visible Learning do not transfer to the 2007 synthesis.