Grammatical pairs are surface markers which encode different processing strategies but seem to work in free variation. For translation trainees and foreign language learners, these pairs often become a recurring nightmare not only because of their close connection but also because most have no direct equivalents in their native language. The long list of English grammatical pairs includes such formal markers as nearly/almost, as if/as though, will/shall, may/might, must/have to, whether/if, yet/already, enough + noun / noun + enough, because/for, barely/hardly… and verbal patterns like (v1+v2), (v1 to v2) and (v1-v2-ing). In Arabic, the list includes dichotomies such as 'inna/laqad, lam/ma:, la:/lan, sa-/sawfa, faqat/faḥasb, na:hi:ka/fadhlan, la:/kalla:, naعam/'ajal, etc. This study, based on corpus analysis, claims that if grammatical surface similarities often induce Arab translation trainees to under- and mistranslation, this has less to do with the absence of direct equivalents in L1 than with the approach adopted in pedagogical grammar intended to account for the working of such markers in both languages. In fact, present-day foreign language pedagogy has been hampered not only by a descriptive sentencegrammar, which has perpetuated static binaries between Arabic and other languages, but also by a monolingual bias which prevents any insight into the working of natural languages. Findings suggest that an updated contrastive Arabic-English grammar, tailored for translation training and derived from real languages at work, is a prerequisite for effective training and successful interlingual transfer.