The Modality-Specific Learning Style Hypothesis: A Mini-Review

1,215 records screened across six databases (EBSCO: ERIC, Academic Search Complete, PBSC; Ovid: Medline, EMBASE, PsycINFO), plus reference lists and Google Scholar citation searches. 10 studies met the review's methodological criteria, giving 13 experiments, 11 with means/SDs sufficient to compute Hedges g, and 22 post-test effect sizes. Visual-matching pool n = 484; auditory-matching pool n = 356. Only ONE study (Rogowsky et al. 2015) was fully aligned with Pashler's criteria.

Modality-specific (visual/auditory/kinaesthetic and verbalizer/visualizer) learning-style studies across ages; the included samples are predominantly adult and higher-education

States the Pashler criteria as the standard - screen participants for style, assign to matched or non-matched conditions, test everyone with the same instrument, and report the crossover interaction - and pools the studies that come closest to it. IMPORTANT CORRECTION to an earlier version of this record: the criteria were NOT applied as hard inclusion filters. The authors themselves write that "there still appear to be relatively few studies adhering strictly to the methodological criteria outlined by Pashler et al. (2008)" - in some included studies style was not established before group separation (Korenman and Peynircioglu 2007), and participants were assigned to one condition (Massa and Mayer 2006; Rogowsky et al. 2015) or to all conditions (Kassaian 2007). So the pool is of mixed methodological quality, and this is the source of the frequently quoted fact that across six databases and 1,215 records only ONE study was totally aligned with the required design. THE NULL IS UNDERPOWERED, which must be said plainly: the 95% CIs are [-0.74, 0.58] for visual matching and [-0.87, 0.32] for auditory matching, so the data are also consistent with a moderate positive matching effect. This is a null from a small, heterogeneous pool, not a tight zero. Two studies (Constantinidou and Baker 2002; Papanagnou et al. 2016) could not be converted to Hedges g and are described narratively; both reported no matching advantage. The authors also flag that some experimental tasks have a built-in visual or auditory component that would, if anything, INFLATE apparent matching effects - and the pooled effects are still negative.

Random-effects pooling of 22 post-test effect sizes gave a small, non-significant and NEGATIVE effect for visual matching (g = -0.09, 95% CI [-0.74, 0.58], p = 0.80, n = 484) and for auditory matching (g = -0.27, 95% CI [-0.87, 0.32], p = 0.37, n = 356) - i.e. matched instruction did no better, and on point estimate slightly worse, than mismatched. The only fully Pashler- compliant study, Rogowsky et al. (2015), reported no significant crossover interaction and effect sizes in line with the pool (visual-matching g = -0.11, auditory-matching g = -0.256). The authors conclude there is still no replicable statistical evidence for improved learning from aligning instruction to modality-specific style, and that the concept persists as a neuromyth. They contrast this with Newton (2015): 89% of learning-styles papers published 2013-2015 were positive toward the practice.

Low. The reviewed quantity is an interaction under assignment, not a correlation between family background and outcome.

The crossover/meshing null reproduces: Hattie & O'Leary (2025) report d = .04 across the four matching meta-analyses in Visible Learning MetaX, and Clinton-Lisell & Litzinger (2024) find the crossover interaction in only 26% of pooled outcome measures. NOTE THE ONE CONFLICT: Clinton-Lisell & Litzinger's pooled matched-vs-unmatched MAIN effect is g = +0.32, opposite in sign to this review's pooled per-modality estimates (g = -0.09 visual, -0.27 auditory). That is a different estimand - a main effect, not the crossover - but it is the same contrast pooled, and the sign disagreement is real and is not resolved in the literature.