In this paper we present the results of a study that investigates the relationships between search tasks, information architecture, and interaction style. Three kinds of search tasks (simple lookup, complex lookup and exploratory) were performed using three different user interfaces (standard web site, hierarchical text-based faceted interface, and dynamic query faceted interface) for a large-scale public corpus containing semi-structured statistical data and reports. Twenty-eight people conducted the three kinds of searches in a between-subjects study and twelve others conducted the three kinds of searches on all three systems in a within-subjects study. Quantitative results demonstrate that the alternative general-purpose user interfaces that accept automated structuring of data offer comparable effectiveness, efficiency, and aesthetics to manually constructed architectures. Qualitative results demonstrate the manual architectures are favored.