Environmental measurements show that secondary forest succession is a complex temporal gradient of light and soil moisture. Populations of the grass Danthonia spicata are found across a seventy year span of succession following fire in the aspen-pine forests of northern lower Michigan. The populations include four collections from a single site 44 days before, 10 days after, 10 months after, and 13 months after an experimental burn on August 19, 1980. Four other populations are from sites which were part of an experimental study and burned in 1954, 1948, and 1936. The populations were collected 26, 32, and 44 years after fire. Three populations are from an area that burned naturally in 1911 and were collected 69, 70, and 70 years after fire. During this period the environment changed from being unshaded and dry immediately following clear-cutting and burning to a shaded, moist pine and hardwood forest. Three processes, microhabitat selection, physiological tolerance, and genetic adaptation, are postulated to be responsible for persistence across this gradient. Measurement of the microenvironment of D. spicata in the field shows that populations within sites with the lowest mean daily total photosynthetic photon flux density (PPFD) are found in locations that have a higher mean daily PPFD than would be expected at random. Populations in the four experimental burn sites are found during drought in locations with a lower mean soil water potential than would be expected at random. Greenhouse experiments using clonal replicates and a factorial design of three light treatments and two watering treatments show that individual plants of D. spicata can grow and reproduce in light levels lower than those found in the field. The populations are genetically distinct from each other but the differences are small relative to the range of physiological plasticity. Individuals from some populations, including those collected the year following the fire, ahve a higher growth rate under high light conditions. However, individuals from all populations have similar growth rates when grown under low light conditions. There is no evidence for the existence of any differentially shade-adapted populations or individuals. These results were confimed in an experiment using transplant gardens placed at the ends of the environmental gradient. Multivariate analysis of the structure of the phenotypic, genetic, and residual covariance matrices shows that there is a negative correlation between growth and reproduction. It is suggested that this may be due to selection for a range of life-history patterns. At one extreme is a pattern that includes the following phenotypic properties: (1) germination under the favorable conditions of high light following a fire, (2) a high growth rate under those conditions, and (3) a low relative, but high absolute, reproductive rate. The other extreme is a pattern that includes: (1) germination under a range of conditions but predominating under the poor and deteriorating conditions of late succession, (2) a low growth rate under those conditions, and (3) a high relative reproductive rate. Comparisons of three populations from sites of the same age suggest that genetic drift may have been important in differentiating the populations. A multivariate analysis using twelve characters also produced evidence that genetic drift has been more important than selection in differentiating the populations.