Abstract. We used patch-specific matrix models to test the influence of neighboring plants on the demography of Ambrosia dumosa, a dominant perennial shrub of the Colorado Desert in southern California. In the desert literature, the presence or absence of neighboring plants is reported to influence plant growth and survival and has long been associated with plant-plant interactions that range from extreme competition to facilitation. Here we consider the less addressed question of how neighbor-defined population subsets contribute differently to overall population dynamics.Demographic data collected from Ͼ6000 individual plants from a permanently mapped hectare over 10 yr were used to divide this A. dumosa population into subsets that were defined by the presence or absence of neighbors. By partitioning the population in this way, we documented differences in population growth, elasticity structure, and stable size structure among the distinct patch types, and evaluated the contributions of each patch type to overall population dynamics. Population growth was consistently higher for population subsets of plants that were isolated throughout their ontogeny, compared to those with close neighbors throughout their ontogeny. Further, overall population growth was proportionally more sensitive to perturbations involving isolated adults, despite the projected persistence of plants with close neighbors. In short, changes in survival of isolated adults had a greater influence on population growth than changes in survival of adults with neighbors.We used life table response experiments (LTREs) to test for spatiotemporal effects of neighbors. The LTRE was consistent with the elasticity analyses in showing that dynamics among adult members made the greatest contribution to the observed differences in population growth among the neighbor-defined population subsets during each census period. Neighbor effects among adults were greater and more important than temporal variation in drought levels in decreasing population growth relative to a pooled 10-yr reference matrix that ignored neighbor effects. Although neighbor effects contributed greatly to differences in population growth among the models projected, the LTRE revealed that, relative to the reference matrix, transitions among juvenile size classes decreased population growth for the predominantly drought free 1984-1989 census interval and increased population growth for the 1989-1994 interval, which included four winters of extreme drought. We hypothesize that higher mortality during periods of high rainfall was due to increased competition among plants, especially in favor of adults at the expense of juveniles. The drought period was also characterized by increased recruitment of new adults, probably reflecting diminished competition from adults for well-established juveniles capable of growing into reproductive condition after elimination of smaller juveniles during 1984-1989. Our habitat-specific partitioning of this population revealed dramatic differences in the ...