This paper describes the design of a 192-text Spanish-English specialized corpus of biomedical research articles (RAs) divided into three 64-text subcorpora (English texts, their corresponding Spanish translations, and Spanish comparable texts) for use in quantitative contrastive analysis. The paper also presents an exploratory study analysing theme–rheme structure in these subcorpora. Two definitions of theme were used: Halliday’s ideational theme and preverbal theme (i.e., all clause constituents before the finite verb of the main clause). The study adopted a target-oriented approach and assessed the acceptability of the translated texts with regard to the statistical norm of the comparable native-speaker Spanish subcorpus. Statistically significant differences were found for marked theme and its different syntactic manifestations (prepositional phrase adjunct and subordinate clauses) and there was evidence of a different thematic distribution within the semantic category person (researcher, patient, first person). The most striking results were found for different measures of theme length, suggesting a consistent information overload in the thematic zone in the whole RA and in the individual rhetorical sections except for the Introduction. The translated texts occupy a kind of no-man’s land half-way between the source articles and the independently created Spanish RAs. A refined three-stage model of the study design is proposed for future target-oriented quantitative and qualitative research into translation.