Translation scholars have long noted that assessment, a key component in translator training, is performed in a generally arbitrary manner. The use of corpora to document differences within language pairs, however, can provide an empirical basis for the formulation of assessment tools. The present study, based on data collected from Russian and English comparable corpora organized by text type, offers a case study in the development of an assessment tool designed to evaluate three isolatable, but nevertheless frequently ignored, features of textual cohesion: punctuation, sentencing, and paragraphing. Because novice translators tend to focus on the level of the word or phrase, ignoring textual elements occurring above the sentence level, focusing assessment on such textual elements can encourage novices to consider the target text globally, as a professional product composed of various features above and beyond lexis. A granular tool for assessing punctuation in Russian>English texts is provided, which can be replicated for other language pairs as well.