Current approaches to assessing agrammatism use data from restricted contexts, such as picture description and story telling tasks. There is evidence in the conversation analysis literature to suggest that conversational grammar may differ markedly from the grammar of such elicited language samples. The disparity between conversational and test grammar suggests that it is possible for the form of an agrammatic utterance to be motivated by the context in which it occurs. Thus, behaviours previously considered to be symptoms of agrammatism may be adaptations to talking in different environments. This study analyses two distinctive test response forms made by two agrammatic speakers, and discusses whether they might be strategic adaptations to testing rather than symptoms of impairment. (Contains 1 table.)