An interesting in-press paper from FSU anthropologist Dean Falk links language and music evolution to baby-talk ("motherese"). According to the abstract:

The evolutionary underpinnings that preceded the emergence of language are investigated by comparing mother-infant vocal and gestural interactions in chimpanzees and humans, and modeling those of early hominins. These data suggest that melodious vocalizations that heralded protolanguage evolved as the trend for enlarging brains in late australopithecines/early Homo progressively increased the difficulty of parturition, thus selecting for females that gave birth to relatively undeveloped neonates. It is hypothesized that hominin mothers responded by adopting new foraging strategies that entailed putting down babies that were developmentally unable to cling to their bodies, and silencing, reassuring, and controlling them with ‘motherese.’

Steve Pinker, quoted in Science (306, pg. 1122), isn't convinced:

"The idea that music evolved to soothe babies might explain why mothers sing to their babies," he says, "but it doesn't explain why older children and adults listen to music."