In line with the proposal of de Roder (1999), we will draw an analogy between the structure of ritual, poetic language and natural language, exploiting Frits Staal’s conception of ritual as a set of recursively applicable formal procedures, and the biological ramifications of the Chomskyan postulate of Universal Grammar. The central hypothesis is that, in terms of evolution, poetry takes a position between age-old rituals and natural languages, as a sort of missing link. We will argue that the building blocks and mechanisms of natural language developed out of ritual acts and the associated rhythmic sound sequences. The formal principles underlying rituals turned out to be useful for communication and thus natural language could evolve, an example of exaptation in the sense of Gould and Vrba (1982). Under this view, the rhythmic layer of poetry — like syntactic structure — is a semantically empty, autonomous pattern, going back to the structural principles underlying ritual. Rhythmic patterns in poetry are thus instances of pure acts in the ritual sense, and as a consequence poetry is a form of language use in which the ritual basis of language is experienced. This puts T.S. Eliot’s famous dictum “Genuine poetry can communicate before it is understood” in a new perspective. The paper thus focuses on the ritualistic substrate of language, not on the synchronic role of ritual in language. We will also discuss neurological evidence which independently supports this idea: Broca’s area — an area in the left hemisphere of the brain that has traditionally been associated with language comprehension and production — is activated when syntactically complex sentences are being processed, but it is also activated in tasks involving the perception of rhythmic patterns in music, thus supporting the idea that both (and by implication ritual too) have the same origins evolutionarily.