Outbreeding depression, a decline in offspring fitness resulting from mating between distantly related individuals, can jeopardize the success of habitat restoration projects. Concern about outbreeding depression has led to the recommendation that restoration projects use seed collected from geographically proximate populations. However, this recommendation assumes that (1) geographically proximate populations are more ecologically similar than distant populations, and (2) offspring of crosses between individuals from ecologically similar populations have higher fitness than offspring of crosses between ecologically dissimilar populations. We tested these assumptions in Lobelia siphilitica, a species commonly used in ecological restoration. We grew F₁ offspring of within- and between-population crosses in the greenhouse and measured four components of performance: seeds per fruit, seed size, germination success, and final aboveground biomass. We found mixed support for the recommendation that restoration projects use seed collected from geographically proximate populations because of concerns about outbreeding depression. Geographically proximate L. siphilitica populations had more similar annual mean temperatures than distant populations, as expected if geographically proximate populations are more ecologically similar. However, there was no relationship between geographic distance and the difference in annual precipitation. We also found that crosses with climatically dissimilar populations decreased the performance of L. siphilitica, as expected if there is outbreeding depression. However, we detected outbreeding depression in only a subset of populations. Consequently, our results support the recommendation that factors other than geographic distance, particularly climatic distance, need to be considered when selecting seed sources for ecological restoration projects.