This paper comparatively analyses human perception of language and music. Its purpose is to elevate the historically worn-out music/language analogy to a scientific level within contemporary cognitive science. The research: (1) compares the postulates of the mental grammar of music with the findings of modern transformational-generative grammars of natural languages. All relevant theories stemming from  Lerdahl and Jackendoff's Generative Theory of Tonal Music (1983) to the present day are discussed. (2) tackles the issue of musical semantics, by comparing a number of traditional and formal concepts of linguistic semantics with contemporary psychological research of the meaning of music (denotation, connotation, metaphor, formalization). (3) offers an overview of the findings of contemporary neuropsychology, where the reactions of zones of the human brain's cortex active during music / language perception are discussed. Common grounds for neurolinguistics and neuropsychology of music are hypothesized. The research points to a much more abstract specialization for higher cognitive processes than it was until recently believed. Shared 'temporal processing' and 'tonal-grammatical' modules are postulated, accountable for some common reactions in music and language processing. Music perception findings strongly contradict the famous Chomskian thesis proposing the uniqueness of human linguistic competence (strong genetic specialization for language only).