The Evidence on Teaching

Mozart effect-Shmozart effect: A meta-analysis

Pietschnig J, Voracek M, Formann AK · 2010

grade Bmeta-analysisindependentfailed
Sample
Nearly 40 studies, over 3,000 subjects, including unpublished research
Population
Mixed ages; spatial-task performance after musical exposure
Design
The definitive meta-analysis of the claim, 16 years after Rauscher et al. (1993). Deliberately includes unpublished work, tests for publication bias, and - the decisive move - tests LAB AFFILIATION as a moderator.
Key findings
Mozart KV448 versus non-musical or no stimulus: d = 0.37 (95% CI 0.23 to 0.52). But ANY other musical stimulus versus non-musical or none gave a similar d = 0.38 (95% CI 0.13 to 0.63), and Mozart versus other music was negligible, d = 0.15 (0.02 to 0.28) - i.e. there is an arousal effect of music, not a Mozart effect. Formal tests found confounding publication bias requiring downward correction, and the central finding is a systematically higher effect in studies by Rauscher and colleagues than by other researchers. Little evidence remains for a specific Mozart effect.
Genetic confound
Low; experimental within- and between-subject exposure contrasts.

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Spatial task performance, Mozart vs non-musical/no stimulusCohen dd = 0.37 (95% CI 0.23-0.52), before downward correction for publication biasstandardizedimmediately post-exposurebusiness-as-usualend-of-treatmentg
Spatial task performance, any other music vs non-musical/no stimulusCohen dd = 0.38 (95% CI 0.13-0.63) - indistinguishable from Mozartstandardizedimmediately post-exposurebusiness-as-usualend-of-treatmentg
Mozart vs other music (the specific effect)Cohen dd = 0.15 (95% CI 0.02-0.28)standardizedimmediately post-exposureactive-alternativeend-of-treatmentg
Lab affiliation as moderatormoderator testeffects systematically higher in the originating labstandardizedn/anonenot-applicableg

Cited by