Between the 1890s and the 1930s field ecologists in the United States (and elsewhere) looked to classification to make their fledgling science an exacting and respectable one. Taking plant taxonomy as their model, ecologists expected that more comprehensive empirical knowledge of vegetation types would produce robust systems of classifications, as it did with species taxonomy. In the event, however, scaled-up data-gathering in the field led ecologists to conclude that vegetation types were not natural units, as species are. Most ecologists then abandoned classification for agendas borrowed from causal sciences such as chemistry or physiology. This cycle of expectation and despair is examined in the practical fieldwork of four ecologists: Henry Cowles, Frederic Clements, Henry Gleason, and Arthur Vestal. Their experiences reveal how perceptions of categories depend on the density and geographical scope of data. Cycles of optimism and disillusionment probably characterize all the classifying sciences in the modern period: because in the "Age of Progress" all sciences sought to advance by expanding and perfecting their empirical base. Comparative study of collecting and classifying practices across the sciences is in order.