It's all in your head: A meta-analysis on the effects of music training on cognitive measures in schoolchildren

21 studies, k = 100 effect sizes. The reported N is contested - Cooper's abstract states N = 5,612 while Sala & Gobet (2020) describe the same meta-analysis as covering 1,767 participants, most likely a unique-participants versus summed-across-effect-sizes discrepancy. Neither figure is treated as verified here.

Schoolchildren; age range not obtainable without the paywalled full text. Academic- achievement measures were excluded by design as "too different from cognitive ability."

A third, independent synthesis of the same literature, useful because it was run by a single author outside both camps of the Sala-Gobet versus Román-Caballero dispute and reaches the same structural conclusion as Sala & Gobet without adopting their framing. Paywalled with no repository copy, so everything here comes from the abstract plus two independent descriptions of it (Bigand & Tillmann 2022; Sala & Gobet 2020). Note the year ambiguity: online-first October 2019, print issue August 2020, and both years appear in the citing literature.

Naive pooled effect g = 0.28 [0.21, 0.34], p < .001. With active control groups it falls to g = 0.21. Once methodological-quality moderators are entered it becomes g = 0.08 [-0.04, 0.20], p = .19 - non-significant. No area of cognitive function showed a clear advantage over any other, and results did not differ by geography or by type of music intervention. The author's conclusion is deliberately double-edged: "music training may be a positive cognitive intervention for schoolchildren; however, advantages as to the utility of music training compared to other cognitive interventions were less empirically supported." Bigand & Tillmann document substantial effect-size coding disagreements between Cooper and Sala & Gobet on the same primary studies (six positive effect sizes taken from Bilhartz et al. 1999 where Sala & Gobet took one; ten of Rickard et al. 2012 where Sala & Gobet took all 26), which is a useful reminder of how much meta-analytic discretion sits behind these estimates.

Not addressed directly; handled only through the methodological-quality moderator, which is where the effect disappears.

Independently reproduces the central structural finding of Sala & Gobet (2017, 2020) - a small naive positive that does not survive design-quality adjustment - using a different author, a different corpus definition and different coding decisions. That convergence is why replication is recorded as `replicated` even though the effect being replicated is the disappearance of an effect.