Evaluating translations of poetry will always be difficult. The paper focuses on the problems posed by French surrealist poetry, where the reader was held to be as important as the writer in creating interpretations, and argues that evaluations involving these poems inevitably require reader-response data. The paper explores empirically, in the context of André Breton’s “L’Union libre”, whether a modification of Think-Aloud procedure, called Note Down, applied both to the original text and to three English translations, can contribute useful information to a traditional close reading approach. The results suggest that comparative Note-Down protocols permit simple cost-benefit analyses and allow one to track phenomena, like the persistence of an effect through the text, which might be hard to obtain by other methods.