We tested theories of links between musical expertise and language ability in a sample of 6- to 9-year-old children. Language ability was measured with tests of speech perception and grammar. Musical expertise was measured with a test of musical ability that had 3 subtests (melody discrimination, rhythm discrimination, and long-term memory for music) and as duration of music training. Covariates included measures of demographics, general cognitive ability (IQ, working memory), and personality (openness-to-experience). Music training was associated positively with performance on the grammar test, musical ability, IQ, openness, and age. Musical ability predicted performance on the tests of speech perception "and" grammar, as well as IQ, working memory, openness, and age. Regression analyses--with other variables held constant--revealed that language abilities had significant partial associations with musical ability and IQ but not with music training. Rhythm discrimination was a better predictor of language skills compared with melody discrimination, but memory for music was equally good. Bayesian analyses confirmed the results from the standard analyses. The implications of the findings are threefold: (a) musical ability predicts language ability, and the association is independent of IQ and other confounding variables; (b) links between music and language appear to arise primarily from preexisting factors and not from formal training in music; and (c) evidence for a special link between rhythm and language may emerge only when rhythm discrimination is compared with melody discrimination.