Revisiting the association between music lessons and intelligence: Training effects or music aptitude?

133 undergraduates (65 women), of whom 62 had five or more years of music training (mean 13.91 years)

Canadian university students; measures were Raven's Advanced Progressive Matrices, the Musical Ear Test (melody and rhythm aptitude) and maternal education as an SES proxy

Small and cross-sectional, so grade C, but it is included because it isolates the variable that actually explains the music-IQ correlation, and it is the only design that does. The key move is including MUSIC APTITUDE alongside music training. Nearly every observational study in this literature controls socioeconomic status and stops; this one shows SES is not the operative confound at all.

Music training correlates with nonverbal intelligence at r = 0.24, and the association SURVIVES controlling maternal education (partial r = 0.26, p = .002). It does not survive controlling music APTITUDE: in the hierarchical regression, music training beta = 0.09 (partial r = 0.09, p > .3, non-significant) while melody aptitude beta = 0.21 (p = .027) and rhythm aptitude beta = 0.30 (p = .001). Variance decomposition: rhythm alone 7.14%, melody alone 3.02%, melody and rhythm shared 5.44%, aptitude and training shared 6.20%, and MUSIC TRAINING ALONE 0.63%. No mediation by training (bootstrap CIs all spanning zero) and no moderation. The companion 2020 study in 91 children aged 6-9 reproduces the pattern for language: musical ABILITY predicts speech perception (r = .43) and grammar (r = .56) independently of IQ, while music TRAINING contributes nothing once ability and IQ are in the model, and a Bayesian analysis finds adding training makes the data 1.36 times LESS likely.

This is the confound identified precisely rather than assumed. Music aptitude is substantially heritable (Mosing 2014; Niarchou 2022 puts SNP heritability of beat synchronisation at 13-16% on the liability scale with 69 genome-wide-significant loci), so "aptitude explains the correlation" is a hereditarian claim with a molecular footing. Note the companion finding that socioeconomic status does NOT attenuate the music-IQ link - the archive should not repeat the common claim that family income explains it. Aptitude does.

The authors reproduce the pattern in a second, independent sample of children in Swaminathan and Schellenberg (2020, Journal of Experimental Psychology: Learning, Memory, and Cognition 46(12), 2340-2348, doi:10.1037/xlm0000798), using language rather than nonverbal intelligence as the outcome, with the same result. Consistent in direction with Wesseldijk, Gordon, Mosing and Ullén (2023), whose co-twin control found no causal effect of early music training on verbal ability at 18.