| This research models the cognitive process by which end user professionals formulate database queries in response to realistic scenarios and empirically shows that an interface designed to support the natural process improves query performance. Specifically, strategies utilized to navigate a chain of related database tables are studied via a two-experiment methodology.; The first experiment uses verbal protocol analysis to infer underlying processes used by ten Food and Drug Administration (FDA) review chemists as they engage in structured information search tasks. A model is developed based on this exploratory study, and context-independent interventions are designed to promote effective strategies and prevent ineffective strategies. These interventions, which enforce an explicit sequence of query elements, efficient searching, and a generate-and-test strategy, are built in to a prototype interface.; The experimental procedural query interface is compared to a commercial query-by-example interface in a controlled within-subjects experiment involving 18 FDA review chemists. The results of this second study indicate that subject performance is significantly better on the experimental interface for complex queries in terms of accuracy. Since both experimental designs preserve external validity, this dissertation contributes to the practical goal of designing more usable query interfaces for end users. |
