One possible response of plant populations to heterogeneous environmen
ts is genetic adaptation resulting in the formation of distinct ecotyp
es. Genetic adaptation to stressful environments may affect both the l
imits to species boundaries and the potential for response to a changi
ng environment. Reciprocal transplant experiments have frequently been
used to describe ecotypic differentiation and to infer the role of na
tural selection when there is evidence for home site advantage. The de
monstration of a home site advantage, however, does not reveal which p
lant characters are responsible for conferring increased fitness on po
pulations planted in their native site. Here, we combine the classic r
eciprocal transplant experiment with multivariate regression analysis
of selection to ask a series of questions relevant to understanding ad
aptive genetic differentiation in natural plant populations. Impatiens
pallida plants from a mesic floodplain and a dry hillside site were r
eciprocally transplanted. We initially presumed the hillside to be a s
tressful site for Impatiens given its sparser population of consistent
ly smaller individuals. This study describes the two environments from
the perspective of the plant to ask whether it is stressful. In addit
ion, we investigate genetic differentiation between populations and as
k whether the two populations are distinctly adapted to their home sit
es. To identify traits that may be important for conferring home site
advantage, we quantify present-day natural selection in these sites an
d ask whether the observed selective forces can explain genetic differ
ences. Finally, because phenotypic correlations may play an important
role in a population's response to its environment, we investigate rel
ationships among traits to determine the extent to which they are gene
tically and/or environmentally controlled. The large reduction in tota
l seed production when plants from both populations were grown on the
hillside supported our initial bias that this site was stressful to Im
patiens. In addition, the higher relative fitness of each population p
lanted in its native site demonstrated that these populations represen
t distinct ecotypes. Genetic differences between populations were obse
rved for several life history and morphological characters. In particu
lar, plants from the hillside population were smaller and produced cle
istogamous flowers earlier than floodplain plants. Selection analysis
revealed that, while there is strong selection favoring early flowerin
g on the hillside, there is no advantage to early flowering fnr plants
grown on the floodplain. An increased developmental rate, which allow
s plants to produce seeds before they succumb to drought stress, appea
rs to be the most important mechanism responsible for the greater rela
tive fitness of the hillside population in its native site. While grea
ter total plant leaf area is favored by selection on the floodplain, t
here is no evidence for selection on this trait on the hillside. Pheno
typic covariances among traits differed between sites and populations,
resulting in differences in the action of indirect selection. There i
s evidence that indirect selection on correlated traits is responsible
for some of the observed genetic differences.