A central issue in language testing is the choice of norms, and the need to reconcile notions of "standard" English with local language norms and features. Data from studies of international intelligibility indicate that some features of "standard" language descriptions, based on native-speaker language use, are not essential for successful communication. A specific question in test design is thus the extent to which descriptors and rater decisions reflect native-speaker, as opposed to local, language norms. This article takes a case study approach by focusing on a speaking test used in Hong Kong. It analyses published examiner comments from 12 examination sessions between 2003 and 2011 and compares them with criteria for international intelligibility. It also analyses the distribution of the comments among various segmental and suprasegmental phonological categories and relates this to intelligibility research and to theoretical considerations. The study critically evaluates the concept of "intelligibility" and considers its relevance for language testing, and revisits certain aspects of the "local norms" debate. The findings of the study suggest that what examiners notice is generally also what is important for international intelligibility, although the importance of certain suprasegmental features is questionable from an intelligibility standpoint.