Written examinations represent one of the most common assessment tools in education. Though typically perceived as measurement instruments, written examinations are primarily texts that perform a communicative function. To complement existing research, this study viewed written examinations as a distinct form of communication (i.e. 'register'). Its starting point was sociolinguistic theory that suggests that the linguistic features of a text are not arbitrary, but are dictated by the context of communication (e.g. who is writing, for whom, for what purpose). Drawing on this theory, this study sought to identify the most distinctive linguistic features of written examinations, as well as their contextual drivers. The investigation, which involved an analysis of 3036 examination questions, revealed the interrelationship between communication and measurement in assessment. More importantly, it highlighted the need to render explicit the currently implicit 'social contract' that governs the linguistic design of examination questions.