The Masai Mara Game Reserve of southwestern Kenya forms the northernmost extension of the Serengeti ecosystem and provides the critical dry season range for approximately 1.5 million migratory wildebeest. Over the past 100 years, major ecological changes have occurred. The area has experienced a transformation from open grassland to dense woodlands and back. This study addressed the transition in the Mara woodlands from both an historical and a contemporary perspective. The study focused on two central questions: 1) what factor (or factors) were responsible for the decline of the woodlands in the 1960s? And 2) what factor (or factors) are currently responsible for inhibiting woodland regeneration? In the 1880s an introduced disease, rinderpest, decimated wild and domestic ungulates in the Serengeti-Mara region. Local pastoralists, dependent on their livestock for survival, succumbed to disease and starvation. Elephant numbers had also been greatly reduced by indigenous hunters. Explorers, slave traders, and hunters described the area as an open grassland by 1900. In the following decades, conditions were conducive to the establishment of woodlands; burning rates were low and elephant browsing was negligible. By the 1930s, the area was covered by dense woodland. These woodlands began a steady decline several decades later. Unusually high rainfall, high grass productivity, and severe fires characterized the period of greatest decline (1961 1967). Although woodland losses were initially viewed as "elephant problems", findings from this study suggest that fire was the primary factor in the disappearance of woodlands, while elephants merely accelerated the rate of decline. Elephants preferred open grasslands, swamps, and relict thickets in the wet season. However, in the dry season, elephants selected wooded habitats. Average group size was significantly higher in the wet season than the dry. Mara elephants fit the same feeding patterns reported for many African elephant populations. Elephants concentrated on grasses and herbs in the rainy season and browse in the dry season. In general, males browsed more than females, while females ate more diverse diets containing more herbaceous matter. Elephants utilized seedlings under 1m more than any other height class of trees throughout the year. This pattern of selective feeding significantly reduced seedling survivorship. Large-scale field experiments subjected plants to three treatments: browsed only, browsed and burned, and neither browsed nor burned. Although fire, at current fuel loadings and intensities, produced an almost total topkill, the majority of burned individuals resprouted within six months. Elephants removed a significant proportion of seedlings and severely damaged others. Wildebeest inhibited seedling growth through trampling, thrashing, and accidental browsing. Only those seedlings protected from both burning and browsing increased in height. Woodland dynamics in the Mara are currently more affected by elephants, wildebeest, and other browsers than by fire. Elephants can be considered a "keystone" species in this system. I concluded that elephants were not capable of initiating the woodland declines which started over two decades ago. However, once tree densities had been reduced by previous perturbations (such as increased burning rates following a reduction in wildebeest numbers and an increased frequency of man-made fires), elephants accelerated the rate of decline. My findings did not support Caughley's "stable limit cycle" hypothesis. Today, elephants are holding the Mara in a grassland phase, despite low burning rates. This pattern suggests that the Mara may have two locally stable states, woodland or grassland, and that an external factor such as fire is necessary to move the system between the two. Elephants, alone, apparently cannot move the system from one state to another, but once it is in the grassland phase, they can hold it there.