This study examined the representation selection preference patterns of participants in a database query task. In the database task, participants were provided with a choice of information-equivalent data representations and chose one of them to use in answering database queries. A range of database tasks were posed to participants - some required the identification of unique entities, some required the detection of clusters of similar entities, and some involved the qualitative comparison of values, etc. Participants were divided, post hoc, into two groups on the basis of a pre-experimental task (card sort) designed to assess 'knowledge of external representations' (KER). Results showed that low and high KER groups differed most in terms of representation selection on cluster type database query tasks. Participants in the low group tended to change from more 'graphical' representations such as scatterplots to less complex representations (like bar charts or tables) from early to late trials. In contrast, high KER participants were able to successfully use a wider range of ER types. They also selected more 'appropriate' ERs (ie. ones that the diagrammatic reasoning literature predicts to be well-matched to the task).