This intervention-based study focuses on the relation between music and its graphic representation from a meta-representational point of view. It aims to determine whether middle school students show an increase in meta-representational competence (MRC) after an educational intervention. Three classes of 11 to 14-year-old students participated in the teaching experiment: one experimental class (E) and two control classes (C). An intervention on MRC was carried out on the E class during the hours that were allocated for the regular music lessons, while students from the C classes followed the regular music curriculum. E and C classes were given the same pretest and posttest, which measured students' MRC by means of six representational criteria. One month after the posttest, all classes completed a retention test. The results reveal an overall effect in favour of the E group, despite the negative results for two representational criteria. Moreover, the overall gain, to a great extent, was due to a decrease in the score of the C classes, while the gains in the E class were rather small. Theoretical, methodological and educational implications are discussed.