1. Environmental change can affect species directly by altering their physical environment and indirectly by altering the abundance of interacting species. A key challenge at the interface of community ecology and conservation biology is to predict how direct and indirect effects combine to influence response in a changing environment. In particular, little is known about how direct and indirect effects on biodiversity develop over time or their potential to influence ecosystem function. 2. We studied how nitrogen (N), winter precipitation (snow), and warming influenced diversity and ecosystem function over six years in alpine tundra. We used path analyses to partition direct effects of environmental manipulations from indirect effects due to changes in the abundance of two dominant plants. We hypothesize that 1) indirect effects will develop more slowly but will become stronger than direct effects over time, and 2) after six years, indirect effects will more strongly influence diversity while direct effects will influence ecosystem function. 3. Indirect effects of N on diversity were consistently stronger than direct effects and actually developed quickly, prior to direct effects. Direct effects of snow on diversity were detected in year two but then subsequently were reversed, while indirect effects were detected in year four and grew stronger over time. Overall in year six, indirect effects were much stronger than direct effects. 4. Direct effects predominated for three of four ecosystem functions we measured (productivity, N mineralization, winter N availability). The only indirect effects we found were that N and snow indirectly affected microbial biomass N by influencing Geum abundance. Across all four ecosystem measures, indirect effects were infrequent and weaker than direct effects. 5. Synthesis. Increasing indirect effects on diversity over time indicate that short-term experiments or monitoring of natural systems may underestimate the full magnitude of global change effects on plant communities. Explicitly accounting for changes in dominant plant abundance may be necessary for forecasting plant community response to environmental change. Conversely, weak indirect effects for ecosystem processes suggest that predicting ecosystem function without knowledge of plant responses to global change may be possible.