Growth rates of mature trees of two broad-leaved evergreen species (Quercus albocincta and Q tuberculata) and one deciduous species (Lysiloma divaricata) were monitored across a gradient in nutrient availability (soil pH in this study) and vegetation abundance in the Sierra Madre Occidental of northwestern Mexico. Seedlings of two of the species were also grown in the greenhouse in soil from extremes of the pH gradient. The abundance of Q. albocincta is negatively correlated with soil pH, while that of L. divaricata is positively correlated. Quercus tuberculata is most abundant at intermediate pH levels. Growth rates of seedlings and mature trees were not significantly different between evergreen and deciduous species, contrary to the predictions of a general model relating the distribution of growth forms to nutrient availability. Quercus tuberculata showed a significant positive correlation between growth and soil pH, while adult trees of the other two species showed positive correlations between growth and a measure of vegetation density, but not between growth and soil pH. However, seedlings of L. divaricata grown in the greenhouse did show a strong positive response to soil pH. The results are discussed in terms of the usefulness of correlations of growth with environmental parameters in understanding the causes of a distribution pattern. INTRODUCTION The association of broad-leaf evergreen plants with nutrient-poor soils in regions where deciduous plants dominate on relatively nutrient-rich soils is a well-documented pattern (Beadle, 1966; Monk, 1966; Specht, 1963, 1969; Webb, 1968; Whitmore, 1975; Goldberg, 1981). The advantage of the evergreen habit on nutrient-poor soils has been hypothesized to be due to the greater amount of carbon assimilated per unit of nutrient in long-lived leaves (Small, 1972; Schlesinger and Chabot, 1977). The disadvantage of evergreens on relatively nutrient-rich soils has been hypothesized to stem from low maximum growth rates due to both high carbon costs per unit surface area and low photosynthetic rates per unit surface area (see review of evidence in Goldberg, 1981). Low growth rates, in turn, are predicted to result in an inability to successfully compete in dense, rapidly growing deciduous vegetation on fertile soils (Grime, 1977; Goldberg, 1981). This general model predicts that deciduous species are excluded from nutrient-poor soils due to intolerance of low nutrient availability, while evergreen species are excluded from nutrient-rich soils through competition, and, in the absence of competition, should respond positively or not at all to increased nutrient supply. In the present paper I use data on growth rates of naturally occurring mature individuals of evergreen and deciduous tree species found over a gradient of nutrient availability and vegetation density to test two components of the model outlined above: (1) Maximum potential growth rate of evergreen trees is lower than that of deciduous trees. (2) Growth rate of deciduous trees is positively correlated with nutrient availability, while growth rate of evergreen trees is either not correlated or positively correlated with nutrient availability. These two predictions were also tested with seedlings of one evergreen and one deciduous species in a greenhouse experiment using soils from the extremes of a nutrient-availability gradient. Because plant species abundances are often determined at the seedling stage (Grubb, 1977; Werner, 1979), it is critical to know how well adult response to environmental factors matches seedling response. 'Present address: W. K. Kellogg Biological Station, Michigan State University, Hickory Corners 49060.