It is often claimed that the evaluation of translated poetry is a subjective affair, and that no method can account for what is kept and lost in translation. In this article, a relevance-based approach is adopted to analyse poems in translation from English into Spanish, using the notion of inference in order to determine whether the translations have respected the lines of thought and connotation of the original. The authors’ main objective is to furnish a tool for comparing different translations of an original poem. Notions originating from relevance theory are used to discuss controversial issues such as the components of poetic language, processing cost and imitation of style. Two translations of a poem by Emily Dickinson and a translation of a poem by Marianne Moore serve as practical examples of relevance-based criticism of poetry in translation.