A graphical exploratory investigation of CLAMP data reveals further serious and unaddressed statistical issues with the standard procedures used to analyze such data. Exploration and estimation are different goals. If the only utility envisioned for fossil leaf floras is the production of ever-more-precise but possibly inaccurate climate estimates, then the current methods of publication and analysis of CLAMP data are satisfactory. In order, however, to understand the ways in which plant leaves respond to environmental stimuli in the context of real communities, we need application of looser, more flexible tools for data analysis, an appraisal of uncertainty that accounts for systematic bias and unquantifiable noise as well as trivial stochastic errors, and the publication of raw data in a form that can be compared between studies. Graphical techniques like pairs plots are effective methods of exploratory analysis of multivariate data, but theories of biological interest like mechanistic models of leaf response to environmental variables cannot be tested against such data unless the standard forms in which the data are currently published are extended to include the raw (by species) scores.

Leaf morphology remains a valuable and under-exploited source of multivariate data. The CLAMP method is not ideal, but it it is the best source of data currently available. It can give satisfactory results if the data it produces are published and analyzed appropriately. From a biological as opposed to a strictly paleoclimatological perspective, appropriate analysis consists of taking the "climate" out of CLAMP and allowing multivariate data on leaf architecture to illuminate broader ecological questions. The pairs plot as a tool for graphical exploratory analysis provides information on complex covariation among leaf-physiognomic variables, and allows evaluation of systematic errors in CLAMP data, neither of which can be done with eigenvector methods of data reduction or with hierarchical clustering. This has the potential to make multivariate leaf-physiognomic data interesting not only to paleoclimatologists, but also to plant morphologists and functional ecologists. Moreover, the exploratory graphical approach advocated here may prove valuable in other paleontological data sets where current analyses obscure interesting detail in complex, multivariate data.