Tundra regions have experienced regional warming and climate models predict continued warming as a result of the greenhouse effect. This study examines the potential effects of future warming by observing the response of tundra vegetation to variation in temperature due to natural temperature gradients, interannual variability, and experimental warming at four sites in northern Alaska. The four sites spanned a temperature gradient from warmer Atqasuk to cooler Barrow and moisture gradients from dry heaths to wet meadows. At each site 24 warmed and 24 control plots were monitored for 5-7 years. The warming treatment (small open-top chambers) increased air temperatures throughout the growing season between 0.6-2.2 oC depending on the year and site. There was generally a consistent relationship between growing season thawing degree-day totals and plant response irrespective of treatment; thus, it was concluded that experimental differences in plant traits were primarily due to temperature and that the open-top chambers were a reasonable analog of regional climate warming. The plant traits measured were phenological events (leaf emergence, visible buds, and flowering), growth traits (leaf length and change in overall size), and reproductive traits (inflorescence length and number of inflorescences per plot). Plants responded to temperature in 130 of 267 observations (49%). The most common response to warming was earlier phenological development and increased growth and reproductive effort; however, when the response of multiple traits was examined each species response was individualistic and varied among sites. The trajectories of species composition and cover change due to warming were different for each site; nevertheless, the general response to warming was a trend toward lower diversity, an increase in canopy height, an increase in standing dead plant matter, and a decrease in lichens. These findings demonstrate that the response of tundra plant species to warming is complex and varies greatly by species and habitat type. Therefore, it is concluded that forecasts of tundra vegetation change at the regional and species level derived from in situ experimental manipulation will be more accurate than forecasts based on mechanistic or correlational modeling.