River-floodplain ecosystems offer some of the most diverse and dynamic environments in the world. Accordingly, floodplain habitats harbor diverse fish assemblages. Fish biodiversity in floodplain lakes may be influenced by multiple variables operating on disparate scales, and these variables may exhibit a hierarchical organization depending on whether one variable governs another. In this study, we examined the interaction between primary variables descriptive of floodplain lake large-scale features, suites of secondary variables descriptive of water quality and primary productivity, and a set of tertiary variables descriptive of fish biodiversity across a range of floodplain lakes in the Mississippi Alluvial Valley of Mississippi and Arkansas (USA). Lakes varied considerably in their representation of primary, secondary, and tertiary variables. Multivariate direct gradient analyses indicated that lake maximum depth and the percentage of agricultural land surrounding a lake were the most important factors controlling variation in suites of secondary and tertiary variables, followed to a lesser extent by lake surface area. Fish biodiversity was generally greatest in large, deep lakes with lower proportions of watershed agricultural land. Our results may help foster a holistic approach to floodplain lake management and suggest the framework for a feedback model wherein primary variables can be manipulated for conservation and restoration purposes and secondary and tertiary variables can be used to monitor the success of such efforts. ?? 2011 Springer Science+Business Media B.V.