The author investigates the extent to which data obtained from basic corpus-processing tools such as word-frequency lists, text statistics and concordances can be used to assess the quality of specialized translations. The article shows that the use of corpora in studies on translation provides innovative techniques for the analysis of features of translation products which can illuminate the nature of the translation process itself. To discover the correlation between basic corpus-processing tools and the assessment of translation quality, special translations in Italian by advanced translator trainees are first compared with their English source texts to identify significant syntactic and lexicogrammatical correlations between the two subcorpora. The results are then compared to the translation quality assessment grades given by the evaluators. Although these grades are based mainly on translation errors made by the translator trainees and reflect an inevitable degree of subjectivity inherent in quality assessment, they also shed light on broader stylistic issues relating to the translators’ use of lexicogrammatical cohesive devices at the level of the sentence. The frequency of some of these devices in “good” and “less good” translations is finally tested against an Italian reference corpus of non-translational specialist texts. These results are then used to determine whether the translation quality standards taught at the university level coincide with those required by the market.