Music therapy for autistic children: helps social communication, no measurable effect on daily function or quality of life

Music therapy for autistic children has been studied for decades, usually by asking whether scores on autism rating scales go down. Lu and colleagues argue that is the wrong single question, because those scales measure how close a child looks to a non-autistic norm rather than whether the child communicates, participates or is doing well. So they searched ten databases to August 2026, kept the clinically interpretable measures, and sorted the results into five families: social communication and interaction, language, adaptive function and participation, quality of life and well-being, and autism-related scale scores. Thirty-three controlled studies qualified, covering 1,880 participants under 18.

The answer differs by family, which is why we grade this mixed rather than moderate. Social communication and interaction showed a moderate pooled benefit (10 studies, Hedges’ g = 0.64, 95% CI 0.25 to 1.03), and it held up when the authors restricted the pool to the studies with the strongest methods (7 studies, g = 0.57, CI 0.02 to 1.12). Language showed a small benefit overall (11 studies, g = 0.32, CI 0.10 to 0.54) that did not survive the same restriction (g = 0.19, CI -0.17 to 0.55). Autism-related scale scores produced the biggest number of all (16 studies, g = 1.23) but with substantial disagreement between studies (I² = 77 percent), and it stayed significant in the six studies that described a reproducible randomisation method (g = 0.95).

Adaptive function, participation, quality of life and well-being did not reach significance. Those are the outcomes closest to what families usually want, and the honest reading is that the current evidence does not show music therapy moving them. None of the moderator tests came out significant either: not technique, not duration, not parent involvement. Longer programmes were not better (-0.034 per month, p = 0.55).

Limitations. The included studies differ widely in what the therapist actually did and in what was measured, and the authors say so. Blinding is close to impossible for a music intervention, and most outcomes are rated by parents or clinicians who know which group a child is in, which tends to inflate effects. Geography is narrow enough that cultural context is not well represented. The authors’ own main caveat is the one worth keeping: many of the measures were built to score autism severity, not to capture what music therapy is trying to do, so both the positive and the null results are partly a statement about the instruments.