Data models provide a map of the components of an information system. Prior research has indicated that more expressive conceptual data models (despite their increased size) result in better performance for problem solving tasks. An initial experiment using logical data models indicated that more expressive logical data models also enhanced end-user performance for information retrieval tasks. However, the principles of parsimony and bounded rationality imply that, past some point, increases in size lead to a level of complexity that results in impaired performance. The results of this study support these principles. For a logical data model of increased but still modest size, users composing queries for the more expressive logical data model did not perform as well as users composing queries for the corresponding less expressive but more parsimonious logical data model. These results indicate that, when constructing logical data models, data modelers should consider tradeoffs between parsimony and expressiveness.