This paper will ask what modern poetic form does to scientific discourses, and what the presence of scientific terms in poems does to and for those poems. It will ask how – and how far – we can bring the history of science to bear on the interpretation of the poem, and how far its qualities as poetic discourse remove it from historical questions. The poems examine will date from approximately 1910 to 1975. ‘Form’ for the purposes of this paper is the gestalt of an individual poem, rather than fixed forms such the sonnet or the villanelle. While effects of rhyme and prosody may affect the significance of a scientific word within a poem – as for example with ‘etherised’ in the opening lines of T. S. Eliot’s ‘The Love Song of J. Alfred Prufrock’ (1917) – what looms larger in the post-1900 period is the relation of kinds of discourse and the ironies created through incompatible perspectives. The discovery of the poetic potential of scientific discourse was set in motion by Eliot’s valorisation of the seventeenth-century metaphysical poets, with the extended metaphor being foregrounded. The paper will discuss modernist and neo-metaphysical poetry by poets such as Mina Loy, John Rodker, Herbert Read, Michael Roberts, William Empson, and C. Day Lewis. In their works, scientific vocabulary is often represented by a small number of significant words, and the slightness of its presence raises the question of how far vocabulary can imply the presence of discourse; it raises the question of the readerly work needed to reconstruct the underlying concepts. Standing in contrast to these poets is the Scottish modernist Hugh MacDiarmid, whose strategy was one of appropriation (alternatively dubbed ‘plagiarism’) of much longer passages from prose, including specialist and popularising science writings and book reviews. MacDiarmid’s approach is interesting for the contrast it creates, and significant for its anticipation of later twentieth century ‘late modernist’ or ‘linguistically innovative’ poetries.