? 2005 The Society for the Study of Evolution. All rights reserved.
Evolution, 59(8), 2005, pp. 1671–1684
THE EVOLUTION OF SPECIES’ DISTRIBUTIONS: RECIPROCAL TRANSPLANTS
ACROSS THE ELEVATION RANGES OF MIMULUS CARDINALIS AND M. LEWISII
A. L. ANGERT1AND D. W. SCHEMSKE
Department of Plant Biology, 166 Plant Biology Building, Michigan State University, East Lansing, Michigan 48824
do not evolve to allow range expansion. Hypotheses for the evolutionary stability of geographic ranges assume that
species are maladapted at the range boundary and unfit beyond the current range, but this assumption has rarely been
tested. To examine how fitness varies across species’ ranges, we reciprocally transplanted two species of monkey-
flowers, Mimulus cardinalis and M. lewisii, within and beyond their present elevation ranges. We used individuals of
known parentage from populations collected across the elevation ranges of both species to examine whether populations
are adapted to position within the range. For both species we found the greatest average fitness at elevations central
within the range, reduced fitness at the range margin, and zero or near-zero fitness when transplanted beyond their
present elevation range limits. However, the underlying causes of fitness variation differed between the species. At
high elevations beyond its range, M. cardinalis displayed reduced growth and fecundity, whereas at low elevations
M. lewisii experienced high mortality. Weak differences in performance were observed among populations within each
species and these were not related to elevation of origin. Low fitness of both species at their range margin and weak
differentiation among populations within each species suggest that adaptation to the environment at and beyond the
range margin is hindered, illustrating that range margins provide an interesting system in which to study limits to
Every species occupies a limited geographic area, but it remains unclear why traits that limit distribution
Distribution, elevation gradient, range limit, reciprocal transplant, survivorship analysis.
Received February 24, 2005.Accepted June 6, 2005.
Every species occupies a restricted geographic area. In
some cases, geographic ranges stop at an obvious barrier,
such as a land-water interface. However, more frequently,
ranges end at seemingly arbitrary points in space (Kirkpatrick
and Barton 1997). Historically, ecologists and biogeogra-
phers have correlated range boundaries with climate to iden-
tify environmental determinants of range boundaries (Griggs
1914; Good 1931; Dahl 1951). Subsequent analyses have
shown that range limits are associated with abiotic variables
such as temperature or precipitation (Root 1988; Cumming
2002), biotic factors such as competitors (Terborgh and Wes-
ke 1975; Bullock et al. 2000), or complex interactions be-
tween biotic and abiotic variables (Randall 1982; Taniguchi
and Nakano 2000).
Even a mechanistic understanding of the relationship be-
tween environmental variables and distribution limits pre-
sents an evolutionary conundrum. Natural selection should
continually improve adaptation at a range boundary and thus
overcome current geographic limits, causing species’ ranges
to ‘‘grow by a process of annual accretion like the rings of
a tree’’ (Mayr 1963, p. 524). Several hypotheses for the evo-
lutionary stability of range limits propose that populations at
range boundaries do not have sufficient genetic variation to
respond to natural selection (Bradshaw and McNeilly 1991;
Hoffman and Blows 1994; Gaston 2003). Other hypotheses
focus on other factors that may prevent populations from
adapting to the environment at the range margin, such as
genetic trade-offs among fitness-related traits in the marginal
environment (Antonovics 1976), genetic trade-offs between
fitness in central and border environments (Holt 2003), or
gene flow from populations adapted to the range center (Hal-
dane 1956; Garcia-Ramos and Kirkpatrick 1997; Kirkpatrick
and Barton 1997). These hypotheses are not necessarily mu-
tually exclusive, and may act synergistically to constrain
All of the above hypotheses are united by the assumption
that populations are maladapted at a range boundary and unfit
beyond the current range. A corollary of this generalization
is that concomitant environmental changes impose selection
for local adaptation to the range edge. Surprisingly, these
assumptions have rarely been directly tested.
Indirect evidence for a decline in fitness with distance from
the range center is provided by the observation that, in some
species, numerical abundance decreases with distance from
the range center, presumably in response to an increasingly
unfavorable environment (Brown 1984; Brown et al. 1996;
Sagarin and Gaines 2002). Other indirect evidence for chang-
es in fitness across species ranges comes from studies of
fluctuating asymmetry. Developmental instability may in-
crease when organisms are under genetic or environmental
stress, as is predicted for individuals at range boundaries, and
several studies of fluctuating asymmetry have found that pop-
ulations at range boundaries do have higher levels of fluc-
tuating asymmetry than central populations (Møller 1995;
Carbonell and Telleria 1998; Gonzalez-Guzman and Mehl-
A more critical test for reduced fitness in marginal pop-
ulations involves direct observation of fitness components
across species ranges. Such studies have often found lower
survival of certain life-history stages or reduced fecundity at
the range margin relative to the range center (Marshall 1968;
Pigott and Huntley 1981; McKee and Richards 1996; Garcia
et al. 2000; Hennenberg and Bruelheide 2003). Unfortunate-
ly, the demographic consequences for such reductions in fit-
ness are generally unclear. Perhaps the biggest stumbling
block to observations of fitness variation, however, is that
by definition, observations of extant populations cannot de-
termine fitness levels beyond present range boundaries
A. L. ANGERT AND D. W. SCHEMSKE
Reciprocal transplant experiments are a powerful way to
test for fitness variation both within and beyond present range
limits as well as for the presence of genetically based local
adaptation (e.g., Schemske 1984; Stanton and Galen 1997;
Verhoeven et al. 2004). Although many classic studies used
reciprocal transplants between areas within species ranges
(Turesson 1922; Clausen et al. 1940), few have transplanted
individuals beyond the range (Gaston 2003). We used recip-
rocal transplants to evaluate population and geographic var-
iation in fitness for sister species of monkeyflower, Mimulus
cardinalis and M. lewisii (Phrymaceae) across their elevation
ranges in California.
The study of closely related species with distinct distri-
butions offers a conceptual advantage for the investigation
of range limits. In a comparison of central versus border
populations of a single species, one could never reject the
possibility that border populations have not yet acquired the
right mutation(s) to extend the border. In a comparison of
parapatric sister species partitioning an environmental gra-
dient, evolution from the common ancestor toward each spe-
cies’ native environment has already occurred, and the ques-
tion of interest is what causes and constrains adaptation to
different ends of the gradient. Mimulus cardinalis and M.
lewisii have been the subject of ecological and genetic studies
for several decades and have many properties that make them
ideal research subjects, including high seed number, high
germination rates, and low transplant mortality (Vickery
1967, 1978; Hiesey et al. 1971; Bradshaw et al. 1998; Brad-
shaw and Schemske 2003; Ramsey et al. 2003). Pioneering
studies of M. cardinalis and M. lewisii by Hiesey et al. (1971)
revealed variation in performance across elevation, with M.
cardinalis displaying low survival and reproduction at high
elevation and M. lewisii displaying low survival and growth
in a coastal climate. Unfortunately, several features of this
study limit its usefulness for drawing definitive conclusions
about variation in fitness versus elevation. First, populations
were collected throughout the geographic ranges of both spe-
cies from Washington to Baja California, but transplanted at
only three sites (Stanford, elev. 30 m; Mather, elev. 1400 m;
and Timberline, elev. 3050 m) along a narrow elevation tran-
sect in northern California. The wide latitudinal and longi-
tudinal distances that separated most populations from the
transplant sites are not easily separated from the effects of
adaptation to elevation. Although the authors found signifi-
cant population differentiation within each species (e.g., be-
tween coastal Californian and montane Arizonan M. cardi-
nalis), regional and subspecies differences are not easily sep-
arated from differences related to elevation alone. Second,
the use of vegetatively propagated clones eliminated infor-
mation about the performance of early life-history stages that
may experience strong selection and be critical for population
establishment (Travis 1994; Caswell et al. 2003; Davis et al.
2003; Lee et al. 2003; Zacherl et al. 2003). Finally, the low
elevation transplant station at Stanford (30 m) potentially
conflated the effects of low elevation with a maritime climate.
We used reciprocal transplants within and beyond the el-
evation ranges of M. cardinalis and M. lewisii to examine
how survival, growth, and reproduction of each species
change with elevation. We used individuals of known par-
entage from populations collected across the elevation ranges
of both species along a narrow latitudinal transect to examine
whether populations are adapted to their position within the
elevation range. Specifically, we asked how fitness compo-
nents change from the center to the edge of ranges and be-
yond, and whether populations are locally adapted within
MATERIALS AND METHODS
Mimulus cardinalis and M. lewisii (Phrymaceae) are rhi-
zomatous perennial herbs that grow along seeps and stream
banks in western North America. The species are self-com-
patible and animal pollinated (Hiesey et al. 1971; Schemske
and Bradshaw 1999). Mimulus cardinalis grows from south-
ern Oregon to northern Baja California and from the coast
of California inland to Arizona and Nevada. Mimulus lewisii
is composed of two races, a northern form growing from
southern coastal Alaska to southern Oregon and eastward to
the Rocky Mountains, and a southern form, occurring pri-
marily in the Sierra Nevada Mountains of California (Hiesey
et al. 1971; Hickman 1993; Beardsley et al. 2003). The two
races are partially incompatible, and recent phylogenetic
analysis suggests that the two races are sister to one another
and together are sister to M. cardinalis (Beardsley et al.2003).
Here we study only the Sierran form of M. lewisii.
Mimulus cardinalis and M. lewisii segregate by elevation,
with M. cardinalis found from sea level to 2400 m and M.
lewisii found from 1200 m to 3100 m in California (Hickman
1993). In the Yosemite National Park region where this re-
search was conducted, the species coexist on larger water-
courses between 1200 and 1600 m elevation (A. Angert, un-
publ. data). Although the published Californian distributions
of M. cardinalis and M. lewisii extend to 2400 and 3100 m,
respectively, repeated attempts to locate extant M. cardinalis
populations above 1600 m and M. lewisii populations above
2900 m in the Yosemite region were unsuccessful. We con-
sider 1200–1600 m to be the shared mid-elevation distri-
bution limit for both species in the Yosemite region.
Genetic Material: Population Collection
and Crossing Design
Seeds from eight plants from each of six populations per
species were collected in September 1999 along an elevation
gradient from 590 m to 2750 m between 37.49 and 37.96?N
latitude (Appendix 1). One plant from each field-collected
family was grown to flowering in the University of Wash-
ington greenhouse under standard greenhouse conditions.
The eight plants from each population were crossed with one
another in a partial diallel mating design (one per population,
for a total of 12 partial diallels), where each plant served as
sire and dam twice with no self- or reciprocal pollinations.
Pollinations were performed by collecting all of the pollen
from one flower with a flat toothpick and fully saturating the
stigma of one flower. Seeds from four pollinations per full-
sib family were pooled. Our crossing design was intended to
provide a genetically variable, outcrossed seed pool for re-
ciprocal transplants rather than to accurately estimate genetic
variance components. Sire and dam effects were included in
MIMULUS ELEVATION RANGES
placement of reciprocal transplant gardens, after Clausen et al. (1948).
Transect of the central Sierra Nevada Mountains, California, showing Mimulus lewisii and M. cardinalis elevation ranges and
statistical models to account for the possible correlation of
error and nonindependence of individual measurements due
to their family structure.
Reciprocal Transplant Methods
varies across elevation ranges, we established experimental
gardens along an elevation transect on the western slope of
the Sierra Nevada Mountains. In June–July 2001, gardens
were planted near Jamestown, California (415 m), at Carnegie
Institution of Washington field stations at Mather (1400 m)
and Timberline (3010 m), and at the White Wolf Ranger
Station in Yosemite National Park (2395 m; Appendix 1).
These gardens were chosen to represent elevations for each
species that are central within the range, at the range bound-
ary, and beyond the range boundary in the Yosemite region
(Fig. 1). Specifically, for M. cardinalis, 415 m is at the range
center, 1400 m is at the upper elevation range boundary, and
2395 and 3010 m are beyond the upper range boundary. For
M. lewisii, 2395 m is at the range center, 1400 and 3010 m
are at the lower and upper range boundaries, respectively,
and 415 m is beyond the lower range boundary.
Due to the tiny seed size and partic-
ular microhabitat requirements for germination of M. car-
dinalis and M. lewisii, experimental gardens were established
with seedlings. Seeds from partial diallel crosses were sown
in flats in the University of Washington greenhouse five
weeks prior to transport to garden sites. The average age of
transplanted seedlings was approximately three weeks after
germination, corresponding closely to the size of plants ob-
served in natural populations at the time of planting. Two
seedlings from each full-sib family were planted at 10-cm
intervals in a randomized block design for a total of 384
seedlings per block (2 seedlings/family ? 16 full-sib fami-
lies/population ? 6 populations/species ? 2 species). During
June–July 2001, seedlings were planted in three blocks at
415 m (n ? 1152), four blocks at 1400 m (n ? 1536), four
blocks at 2395 m (n ? 1536), and three blocks at 3010 m (n
To examine how species’ performance
? 1152), for a total of 5376 seedlings across all four trans-
plant sites. Space and water limitations prevented planting a
larger number of blocks. Garden plots were covered in land-
scape fabric and irrigated daily to approximate conditions in
the species’ native riparian habitat and to standardize water
treatments across environments.
We collected soil samples from each garden
site and grew plants under uniformly favorable greenhouse
conditions to determine whether site differences in perfor-
mance were due to the effects of soils as opposed to other
environmental factors. We measured the performance of four
populations per species, using a subset of four independent
full-sib families per population from the partial diallel cross-
es. Plants were able to flower on all soil types in the green-
house environment and there was no evidence of local ad-
aptation to soil type (indicated by no significant species ?
soil type or population ? soil type interactions). We conclude
that differences in soil properties are not primarily respon-
sible for differences in fitness across elevation and we do not
consider soil type further.
To assess fitness within each garden, we
measured survival, growth, and reproduction. Plants grew at
vastly different rates among gardens. At 1400, 2395, and
3010 m, plants grew slowly and rarely attained a size at which
larger plants spread via rhizomes into neighboring plants’
space. However, at 415 m, M. cardinalis plants began to
spread via rhizomes into neighbors’ space after one growing
season, making it difficult to separate individuals and track
identity. For this reason, we truncated observations at 415 m
after one year, when all M. lewisii individuals were dead and
surviving M. cardinalis were very large. Individuals trans-
planted in a large preliminary study at 415 m displayed very
low mortality and continued rapid growth during the second
growing season, indicating that truncation after one year does
not bias our results (A. Angert, unpubl. data).
Survival was monitored from 2001 to 2002 at 415 m and
from 2001 to 2003 at 1400, 2395, and 3010 m. Survival was
recorded at approximately two-week intervals throughout each
A. L. ANGERT AND D. W. SCHEMSKE
growing season. Growth and reproduction were measured for
one growing season at 415 m and for two growing seasons at
1400, 2395, and 3010 m. To measure plant growth, we re-
corded the total stem number and length of all stems. Stem
number and total stem length were strongly correlated (M.
cardinalis: R2? 0.73, N ? 2065, P ? 0.0001; M. lewisii: R2
? 0.72, N ? 1790, P ? 0.0001). We present stem length data
because they better describe overall plant size at high eleva-
tions, where plants often have only one stem but differ in stem
length. Because permit restrictions prevented seed set at two
transplant sites, we use flower number rather than seed number
as a proxy for reproductive fitness. Flower number and fruit
number measured from 2000 to 2004 in demographic census
plots within natural central and border populations are highly
correlated (M. cardinalis: R2? 0.97, N ? 1132, P ? 0.0001;
M. lewisii: R2? 0.98, N ? 1064, P ? 0.0001), suggesting that
cumulative flower number is a good approximation of total
We estimated overall plant fitness, retaining zeros for
plants that failed to flower or failed to survive, as the cu-
mulative flower number over two growing seasons. We also
summed year 1 and year 2 total stem length to estimate cu-
mulative growth. For M. cardinalis grown at 415 m, only
first-year measurements of stem length and flower number
were available. To keep measures comparable across all sites,
we annualized measures of growth and fitness and compared
average annual stem length and average annual fitness. Com-
parisons of first-year growth and fitness at all sites as well
as cumulative growth and fitness with the 415 m site excluded
produced similar results; we present comparisons of annual
averages for brevity.
To examine fitness variation across species’ elevationrang-
es, we analyzed the relationships between transplant site and
the fitness components of survival and growth and between
transplant site and average annual fitness. Too few individ-
uals remained alive and flowering beyond their ranges to
allow analysis of flower number for surviving plants. To de-
termine whether populations are adapted to range position,
we analyzed the relationship between population origin and
performance within each transplant site. All analyses were
performed in SAS, version 8.2 (SAS Institute, Inc., Cary,
We used accelerated failure time models to
test for differences between species and populations within
species in patterns of survivorship across transplant sites.
Accelerated failure time models assume that factors affect
failure time (e.g., time to mortality) multiplicatively, shifting
the time periods when failures occur (for a general discussion
of failure time analyses, see Fox 2001). For our study, ac-
celerated failure time models were biologically appropriate
because environmental differences among transplant treat-
ments were expected to shift the distribution of time to failure
(Jones and Sharitz 1998; Keith 2002; Denham and Auld
2004). We performed two kinds of analyses. In the first anal-
ysis, we tested for interspecific differences across sites using
a model with fixed effects of site, species, and the interaction
of site and species. In the second analysis, we fit separate
models for each species and tested for intraspecific differ-
ences across sites using models with fixed effects of popu-
lation, site, and the interaction of population and site. These
two kinds of analyses were necessary because the coding of
categorical variables sets one species’ regression coefficients
to zero in analyses with two species, making it impossible
to assess between-site differences for one species. Because
populations were chosen deliberately to span the elevation
range of each species and were not drawn at random from
within each species’ distribution, we consider population as
a fixed effect. Standard statistical packages do not incorporate
random effects in survival time analyses, so for these analyses
we were not able to include sire, dam, or block effects.
To apply the accelerated failure time model, we used PROC
LIFEREG with an underlying Weibull distribution of failure
time (measured in days after transplantation). Survivorship
was described using the function:
S(t) ? e, (1)
where the scale parameter ? scales the model to a baseline
rate of mortality, t is the time since transplantation, and p is
a dimensionless shape parameter that describes change in
failure hazard over time, such that when p ? 1 hazard mono-
tonically decreases with time and when p ? 1 hazard mono-
tonically increases with time (Dudycha and Tessier 1999;
Fox 2001; Keith 2002). We also ran models using an alter-
native plausible distribution, the exponential, which is a spe-
cial case of the Weibull with the shape parameter p ? 1,
indicating a constant risk of mortality (Fox 2001). The ex-
ponential distribution gave a significantly poorer fit to the
data than the Weibull according to likelihood ratio tests (spe-
cies combined: ?2? 288.9, P ? 0.0001; M. cardinalis: ?2
? 288.9, P ? 0.0001; M. lewisii, ?2? 31.8, P ? 0.0001)
but yielded qualitatively similar results, indicating that our
results are robust to the underlying distribution. Observations
were right censored if the individual remained alive at the
end of the observation period. For each categorical variable,
one level was arbitrarily chosen as the reference level and
its regression coefficient was set to zero. Regression coef-
ficients and significance of all other levels were determined
relative to the reference, but this did not reveal whether dif-
ferences among nonreference levels existed. For analyses of
intraspecific differences across transplant sites, multiple com-
parisons were necessary to examine differences among sites
other than the reference. We constructed Z-tests for multiple
comparisons from estimated regression coefficients and the
asymptotic covariance matrix according to the methods of
Fox (2001). Because effects act multiplicatively on failure
time, regression coefficients less than zero can be interpreted
as shrinking the time to failure relative to the reference level,
whereas positive regression coefficients expand the expected
time to failure relative to the reference (Dudycha and Tessier
To examine the relationship between growth and
transplant site, we performed mixed model analysis of var-
iance on log-transformed data with PROC MIXED, which
uses the restricted maximum-likelihood method (REML) to
estimate variance components. Only two M. lewisii individ-
uals remained alive for stem length measurements at 415 m,
causing the full model containing all sites to contain many
MIMULUS ELEVATION RANGES
nonestimable parameters. To remedy this, we ran separate
analyses for each species and excluded the 415 m site from
the M. lewisii stem length analysis. We tested for variation
in average annual stem length of each species with respect
to transplant site, population of origin, sire within population
of origin, dam within each population of origin, block within
site, and the interaction of site and population. For this and
all subsequent models, we consider transplant site and pop-
ulation of origin as fixed effects and sire, dam, and block as
random effects. To evaluate the significance of fixed effects,
we used Type III estimable functions with denominator de-
grees of freedom obtained by Satterthwaite’s approximation.
When fixed effects were statistically significant, differences
among levels were evaluated with Tukey-Kramer adjusted
comparisons of least square means. We used the PDMIX800
macro to convert pairwise differences between least square
means to letter groupings, where means sharing the same
letter code are not significantly different (Saxton 1998). We
used likelihood-ratio tests (comparing each reduced model
to the full model including all effects) to evaluate the sig-
nificance of all random effects.
We used mixed linear models to test for variation
in average annual fitness with respect to transplant site, spe-
cies, population within species, sire within population, dam
within population, and all interactions among fixed effects.
A model containing the random effect of block within site
failed to converge, so we also analyzed residuals after ac-
counting for the block effect. The block effect was highly
significant (P ? 0.0001). However, analyses of raw data and
of residuals after removing the effect of block gave quali-
tatively similar results. We present analysis of raw data, rath-
er than residuals, so that differences in least square means
are easily interpretable. The distribution of fitness was highly
non-normal due to an excess of zeros and a long right tail.
Examination of residuals in preliminary analyses revealed
significant departures from parametric assumptions, and
transformations only slightly improved the distribution of
residuals. Therefore, we used two approaches to model annual
fitness. First, we performed mixed model analysis of variance
on log-transformed data (after adding 1/6 to each observation;
Kuehl 2000) with PROC MIXED. Second, we used the
GLMM800 macro of PROC MIXED to fit generalized linear
models, which are appropriate for a wider range of error
structures than traditional linear models (Kuehl 2000). Gen-
eralized linear models extend traditional linear models in two
key ways. First, they allow the distribution of the response
variable to be any member of the exponential family of dis-
tributions (e.g., gamma, Poisson, binomial). Second, they re-
late the response variable to a set of linear predictor variables
through a nonlinear link function (SAS Institute 1999). The
GLMM800 macro uses restricted/residual pseudolikelihood
(REPL) estimation to fit a generalized linear model with ran-
dom effects. We modeled variation in average annual fitness
using a gamma distribution with a log link function, which
is appropriate for positive, continuous data (SAS Institute
1999; Juenger and Bergelson 2000). Observations were first
transformed by adding one to each observation. Significance
of fixed effects was assessed as described above for stem
length. To evaluate the significance of random effects, we
used the covtest option to obtain Z-tests, which tested whether
the Z-value of each effect (its variance parameter divided by
its approximate standard error) was different from zero (Juen-
ger and Bergelson 2000). Because results obtained from
PROC MIXED and GLMM800 did not differ qualitatively
and because the data violated the assumptions of traditional
linear analysis, we present only results from GLMM800.
To evaluate whether populations
are adapted to their elevation of origin, we used two ap-
proaches. First, we examined population-by-site interactions
in the analyses described above. A significant population-by-
site effect indicates that populations differ in their response
to elevation. If a significant population-by-site effect was
found for failure time, we compared the confidence intervals
of regression coefficient estimates to determine which pop-
ulation and site combinations were significantly different
from one another. If a significant population-by-site effect
was found for growth or fitness, we used Tukey-Kramer ad-
justed comparisons of least square means to determine which
population and site combinations were significantly different
from one another. Second, if populations are locally adapted
to their elevation of origin, then fitness should decrease as
the difference between elevation of origin and transplant site
elevation increases. For each transplant site, we examined
the rank correlations of population average annual fitness
with the absolute value of the difference between origin and
transplant elevations using PROC CORR.
vivorship across the elevation gradient were detected for M.
cardinalis and M. lewisii (species, ?2? 187.09, P ? 0.0001;
site, ?2? 3399.33, P ? 0.001; species ? site, ?2? 2022.04,
P ? 0.0001). Each species showed greater survivorship than
the other at its range center (Jamestown, 415 m: M. cardinalis
? M. lewisii, ?2? 1364.96, P ? 0.0001; White Wolf, 2395
m: M. lewisii ? M. cardinalis, ?232.79, P ? 0.0001). At the
shared middle elevation range boundary, M. cardinalis dis-
played greater survivorship than M. lewisii (?2? 469.02, P
For both species, transplant site had a highly significant
effect on survival time in analyses of intraspecific differences
in survivorship (Table 1). All sites (Jamestown, 415 m; Math-
er, 1400 m; and Timberline, 3010 m) were significantly dif-
ferent from the reference site, White Wolf (2395 m), as in-
dicated by regression coefficients different from zero (Table
1). To examine differences among nonreference sites, we
constructed Z-tests for comparisons of regression coefficients
and found that all pairwise differences among nonreference
sites were also significant, although for M. cardinalis the
difference between Mather (1400 m) and Timberline (3010
m) was only marginally significant after correcting for mul-
tiple comparisons (Table 2). Population did not affect sur-
vival time for either species (Table 1). For M. cardinalis, the
population-by-site effect was significant, indicating that pop-
ulations differ in their response to elevation (Table 1). There
was no population-by-site interaction for M. lewisii survi-
vorship (Table 1). For both species, the Weibull shape pa-
rameter was significantly greater than one (M. cardinalis, 1.57
Significant interspecific differences in sur-
A. L. ANGERT AND D. W. SCHEMSKE
for Mimulus cardinalis, 1339 uncensored values and 1073 right-censored values for M. lewisii, and a Weibull distribution.
Analysis of accelerated failure-time models for survival time, using 1339 uncensored values and 1273 right-censored values
Species VariabledfEstimate SE
(Jamestown, 415 m)
(Mather, 1400 m)
(White Wolf, 2395 m)
(Timberline, 3010 m)
(Levels not shown)
Site ? population
(Levels not shown)
(Jamestown, 415 m)
(Mather, 1400 m)
(White Wolf, 2395 m)
(Timberline, 3010 m)
(Levels not shown)
Site ? population
(Levels not shown)
TABLE 2. Pairwise differences of transplant site regression coefficients from accelerated failure-time analyses.
Jamestown (415 m) vs. Mather (1400 m)
Jamestown (415 m) vs. Timberline (3010 m)
Mather (1400 m) vs. Timberline (3010 m)
* After correcting for multiple comparisons, only Z-scores ?2.1 remain significant at the 0.05 level.
? 0.04, M. lewisii, 1.13 ? 0.02), indicating that the risk of
mortality increased monotonically with time.
Mimulus cardinalis survival during the first year was high-
est at the 1400 m range border, intermediate at high elevations
beyond the upper range limit (2395 and 3010 m), and lowest
at the 415 m range center, although first-year survival was
relatively high across all sites (Fig. 2A). There was an early
decrease in survival during the first growing season at 415
m, whereas survival at 2395 and 3010 m was high during
the first growing season and declined over the first winter.
During subsequent years, survivorship remained highest at
the 1400 m upper range boundary and was reduced at high
elevations beyond the range (2395 and 3010 m). Examination
of regression coefficient confidence limits for each site and
population indicated that the M. cardinalis population-by-site
interaction arose because of differences in elevation response
between the low elevation Mariposa Creek population (590
m) and the midelevation Tenaya Creek population (1210 m;
data not shown). At 1400 m, the Mariposa Creek population
survived longer than the Tenaya Creek population, and the
converse was true at 3010 m.
Mimulus lewisii survivorship was highest at high elevations
(2395 and 3010 m) and intermediate at the lower elevation
range boundary (1400 m; Fig. 2B). At 415 m, beyond the
lower range limit, M. lewisii suffered high mortality during
the first growing season. The few individuals surviving after
one growing season at 415 m died over the winter, resulting
in 100% mortality within one year. At 1400 m, M. lewisii
experienced pulses of mortality at the end of the second and
third growing seasons. At the high elevation range center
(2395 m) and upper range boundary (3010 m), mortality rates
were roughly constant and low.
Transplant site had a highly significant effect on
growth of both species (Table 3). There were no significant
population or population-by-site effects for M. cardinalis
growth, but both population and population-by-site effects
were significant for M. lewisii growth (Tables 3, 4). Mimulus
cardinalis growth was greatest at the range center (415 m),
intermediate at the upper elevation range boundary (1400 m),
and greatly reduced at high elevations beyond the range (2395
and 3010 m; Fig. 3A). The difference in growth between the
415 m range center and the 1400 m range margin was not
statistically significant in Tukey-Kramer adjusted post-hoc
contrasts. Growth of M. lewisii peaked at the 2395 m range
center and was reduced at both the lower (1400 m) and upper
(3010 m) range margins (Fig. 3B). High mortality resulted
in small sample size for M. lewisii at 415 m (n ? 2). The M.
lewisii population main effect is due to the differencebetween
the Warren Fork population (2750 m) and all other popula-
tions except for the Tuolumne River population (1320 m;
Table 4). The M. lewisii population-by-site interaction in-
dicates that populations differ in their growth response to
elevation. This difference is driven in part by the greater
increase in growth at the range center (2395 m) versus the
upper range margin (3010 m) for two middle elevation pop-
ulations (Tuolumne River, 1320 m, and Tamarack Creek,
1900 m) relative to two high elevation populations (Snow
Creek, 2690 m, and Warren Fork, 2750 m; Table 4).
Significant inter- and intraspecific differences in
fitness across the elevation gradient were detected (Table 5).
MIMULUS ELEVATION RANGES
Jamestown; MA, Mather; WW, White Wolf; TI, Timberline.
Survivorship at each transplant site. (A) Mimulus cardinalis. (B) M. lewisii. Transplant site abbreviations as follows: JA,
constructed by SAS MIXED procedure, with denominator degrees of freedom (dfD) obtained from the Satterthwaite approximation.
Significance of random effects (indicated by [R]) determined by likelihood ratio tests.
Linear mixed model analysis of variance summary for log-transformed average annual stem length. F-tests for fixed effects
Site ? population
Block (site) [R]
Sire (population) [R]
Dam (population) [R]
ns, P ? 0.05; ** P ? 0.01; *** P ? 0.001; **** P ? 0.0001.
Each species was more fit than the other at its range center
(Fig. 4). At the shared midelevation range boundary (1400
m), M. cardinalis outperformed M. lewisii. The species did
not differ in fitness at 3010 m, the upper range limit for M.
lewisii and beyond the upper range limit for M. cardinalis.
Mimulus cardinalis fitness was highest at the 415 m range
center, reduced at the 1400 m upper range border, and zero
or near zero at high elevations beyond the upper range limit
(Fig. 4). The significant main effect of population arose from
the difference between a population from midelevation (Ten-
aya Creek, 1210 m) and all other populations except for the
Tuolumne River population (1320 m; Tables 5, 6). The pop-
ulation-by-site interaction was driven in part by populations
differing in the degree of decrease in fitness from 1400 m to
2395 m. The Moore Creek (830 m), Bear Creek (860 m), and
Snow Creek (950 m) populations displayed a greater decrease
in fitness from 1400 to 2395 m than did the Tenaya Creek
(1210 m) and Tuolumne River (1320 m) populations (Table
6). Similarly, the Mariposa Creek population (590 m) dis-
played a greater decrease in fitness from 415 m to 1400 m
than the Tenaya Creek population (1210 m; Table 6).
Mimulus lewisii fitness was highest at the 2395 m range
center and zero or near zero at the lower (1400 m) and upper
(3010 m) range borders and beyond its lower range limit (415
m; Fig. 4). Population did not significantly affect M. lewisii
fitness (Table 6), but the significant population-by-site effect
indicated that M. lewisii populations differ in reaction norms
of fitness versus elevation (Table 6). This interaction was the
result of populations differing in the degree of increase in
fitness at 2395 m, relative to the uniformly low fitness at
other sites. The Tuolumne River (1320 m), Tamarack Creek
(1920 m), and Tioga Road (2580 m) populations showed a
large increase in fitness at 2395 m, whereas the Porcupine
Creek (2400 m), Snow Creek (2690 m), and Warren Fork
(2750 m) populations did not show a statistically significant
increase in fitness at 2395 m (Table 6).
To determine whether populations are adapted to their po-
sition within the elevation range, we also examined the rank
correlation between average fitness and the difference in el-
evation between transplant site and population origin. If pop-
ulations are adapted to position within the elevation range,
then the correlation between fitness and the difference be-
tween origin and transplant elevations should be negative,
indicating that fitness declines as the transplant environment
becomes more different from the native environment. No
correlations were statistically significant, suggesting that fit-
ness variation among populations is not caused by differences
in elevation of origin (Table 7).
Geographic Variation in Fitness
The results of this reciprocal transplant experiment support
the hypothesis that species are most fit at their range center
A. L. ANGERT AND D. W. SCHEMSKE
significantly different. Number superscripts indicate significant differences among levels of the population main effect, where populations
sharing the same number are not significantly different. Differences determined by Tukey-Kramer adjusted comparisons of least square
means in analysis of log-transformed data. High mortality resulted in small sample size (N ? 2) for M. lewisii at Jamestown.
Population differences in average annual stem length. For Mimulus lewisii, population means sharing the same letter are not
Jamestown, 415 mMather, 1400 m
2395 m Timberline, 3010 m
Mariposa Ck., 590 m
Moore Ck., 830 m
Bear Ck., 860 m
Snow Ck., 950 m
Tenaya Ck., 1210 m
S. Fork, 1320 m
S. Fork, 1320 m1,2
Tamarack Ck., 1920 m1
Porcupine Ck., 2400 m1
Tioga seep, 2580 m1
Snow Ck., 2690 m1
Warren Fork, 2750 m2
(B) M. lewisii. For each species, site means sharing the same letter are not significantly different based on Tukey-Kramer adjusted
comparisons of least square means in analysis of log-transformed data. Note that species are graphed on different scales. Transplant site
abbreviations as in Figure 2.
Species’ average annual stem length ? SE at each transplant site (mean values given within each bar). (A) Mimulus cardinalis.
and become increasingly maladapted as the distance from the
range center increases. Both species exhibited the greatest
average fitness at elevations central within their range (415
m for M. cardinalis, 2395 m for M. lewisii) and reduced fitness
at elevations at the range margin (1400 m for both species,
3010 m for M. lewisii). Furthermore, both species exhibited
zero or near-zero fitness when transplanted beyond their pres-
ent elevation range limits (2395 and 3010 m for M. cardinalis,
415 m for M. lewisii).
However, the underlying causes of this fitness variation
differed between the species. Mimulus cardinalis survival in
the first year was relatively high across all elevations. High
fitness at the low elevation range center arose due to high
growth and fecundity. At higher elevations beyond the upper
range boundary, although many M. cardinalis were able to
survive, growth was greatly reduced and few individuals were
able to reach reproductive maturity. The few M. cardinalis
able to flower at 2395 m did so in September, after most M.
lewisii stopped flowering, and did not bear mature seeds be-
fore senescence. By contrast, M. lewisii encountered a strong
survival barrier beyond its lower elevation range limit. Mor-
tality during the first growing season at 415 m was rapid;
most individuals died within one month of planting and all
were dead within one year. Because experimental planting
was timed to match the phenology of natural populations,
transplanted seedlings were exposed to the climate they
would have encountered if naturally dispersed to low ele-
vation. A large preliminary study conducted at 415 m in June
2000 produced nearly identical results (M. cardinalis surviv-
al: 85.8% after four months, 76.3% after 10 months, N ?
962; M. lewisii survival: 6.2% after four months, 0% after
10 months, N ? 953), supporting the inference that observed
patterns of mortality are not exaggerated by unusually harsh
conditions in 2001.
Our findings are largely congruent with the patterns of
variation in performance across elevation described by Hie-
MIMULUS ELEVATION RANGES
nual fitness, using a gamma distribution and a log link function. F-
tests for fixed effects constructed by SAS MIXED procedure, with
denominator degrees of freedom (dfD) obtained from the Satterth-
waite approximation. Significance of random effects (indicated by
[R]) determined by Z-tests.
Generalized linear mixed model analysis of average an-
Site ? species
Site ? population (species)
Sire (population, species) [R]
Dam (population, species) [R]
ns, P ? 0.05; ** P ? 0.01; *** P ? 0.001; **** P ? 0.0001.
year) ? SE at each transplant site (mean values given within each
bar). Means sharing the same letter are not significantly different
based on Tukey-Kramer adjusted comparisons of least square means
in generalized linear mixed model analysis. Transplant site abbre-
viations as in Figure 2.
Species’ average annual fitness (in units of flowers per
the same letter are not significantly different. Number superscripts indicate significant differences among levels of the population main
effect, where populations sharing the same number are not significantly different. Differences determined by Tukey-Kramer adjusted
comparisons of least square means in generalized linear mixed model analysis.
Population differences in average annual fitness (in units of flowers per year). For each species, population means sharing
415 m Mather, 1400 m
Mariposa Ck., 590 m1
Moore Ck., 830 m1
Bear Ck., 860 m1
Snow Ck., 950 m1
Tenaya Ck., 1210 m2
S. Fork, 1320 m1,2
S. Fork, 1320 m
Tamarack Ck., 1920 m
Porcupine Ck., 2400 m
Tioga seep, 2580 m
Snow Ck., 2690 m
Warren Fork, 2750 m
sey et al. (1971) in their landmark reciprocal transplant study
of M. cardinalis and M. lewisii. They demonstrated low sur-
vival and reproductive capacity of M. cardinalis at high el-
evation and low survival and growth of M. lewisii in a coastal
climate. However, in their study, M. cardinalis displayed the
highest survivorship in the low elevation Stanford transplant
garden (30 m), whereas we observed greatest survivorship in
the middle elevation Mather garden, which highlights the
important difference between the low elevation maritime en-
vironment and the low elevation foothills environment. A
second difference between the present findings and the pre-
vious study is the relatively poor performance of M. lewisii
that we observed at Timberline, where Hiesey et al. (1971)
found that M. lewisii achieved its highest performance. This
difference is likely due to several factors, including our ad-
dition of White Wolf as an intermediate transplant site be-
tween Mather and Timberline, exclusion of populations from
the northern race of M. lewisii, and use of seedlings rather
than vegetatively propagated clones. Our use of seedlings
provided missing information about the performance of early
life-history stages, which may be critical for population es-
tablishment (Lee et al. 2003; Zacherl et al. 2003). It is also
important to note that none of the transplant sites used by
Hiesey et al. (1971) were central within the elevation range
of M. lewisii.
Although many reciprocal transplants have been conducted
between areas within species’ ranges, far fewer have trans-
planted individuals beyond the range (Gaston 2003). How-
ever, transplants beyond the range often display reductions
in fitness components such that long-term persistence of pop-
ulations is unlikely (see references in Gaston 2003 and Geber
and Eckhart 2005). Several other experiments have demon-
strated reduced growth, delayed phenology, and, as a result,
reduced fecundity of plant species transplanted beyond their
northern or high elevation range margins, as we found for
M. cardinalis (Prince 1976; Davison 1977; Woodward 1990;
Asselin et al. 2003). Analogous patterns of delayed devel-
opment have been reported for aphids (Gilbert 1980) and
butterflies (Crozier 2004) transplanted beyond their latitu-
dinal range limits. In these examples, fitness reductions gen-
erally are not due to a single environmental event such as a
A. L. ANGERT AND D. W. SCHEMSKE
annual fitness and ?transplant elevation ? population origin ele-
Spearman rank correlation between population average
r Prob ? ?r?
r Prob ? ?r?
frost or to a single vulnerable life-history stage, but rather
result from the gradual accumulation and cascading effects
of fitness reductions at many stages.
In contrast to expectations for northern or upland range
limits, it is generally assumed that climate becomes more
permissive for most organisms and that biotic interactions
become relatively more important in setting southern or low-
land distribution limits (MacArthur 1972; Woodward 1975;
Sievert and Keith 1985; Hersteinsson and Macdonald 1992;
Richter et al. 1997; Scheidel et al. 2003; Cleavitt 2004).
Similarly, a recent study of fitness variation across the par-
apatric ranges of Clarkia xantiana subspecies along a west-
to-east gradient implicated biotic interactions such as com-
petition, herbivory, and pollination as causes of fitness de-
clines when subspecies were transplanted beyond the range
boundary (Geber and Eckhart 2005). Few studies of southern
or lowland distributions limits find severe abiotic limitation
as we have documented for M. lewisii at low elevations. Many
plants showed signs of heat stress such as leaf scorching and
reduced leaf size, and subsequent growth chamber studies
have demonstrated strikingly similar patterns of mortality
when M. lewisii are grown under the high temperatures char-
acteristic of low elevation (A. Angert, unpubl. data).
Population Variation in Fitness
Although population and population-by-site effects were
frequently statistically significant, they were of much smaller
magnitude than site effects. We detected differences among
M. cardinalis populations but not M. lewisii populations in
survivorship at different elevations. For M. cardinalis, the
low elevation Mariposa Creek population (590 m) displayed
greater survivorship at middle elevation than the mideleva-
tion Tenaya Creek population (1210 m), but at high elevation
the Tenaya Creek population had greater survivorship than
the Mariposa Creek population. The direction of reversal in
survivorship is consistent with adaptation of the range margin
Tenaya Creek population to higher elevations, but it is not
entirely consistent with adaptation to position within the
range because of the poor relative performance of the mid-
elevation population at middle elevation. No other differ-
ences among populations were significant, indicating that dif-
ferentiation among populations for survivorship is low.
We detected differences among M. lewisii populations but
not M. cardinalis populations in average annual stem length.
Local adaptation of growth traits may take two possible
forms. First, populations could exhibit genetically based clin-
al differences in growth in which populations originating
from higher elevations display reduced growth rates or short
stature across all environments (Clausen et al. 1940). Alter-
natively, populations could show decreasing growth with in-
creasing distance from population origin. We find some slight
evidence that the former scenario is true for M. lewisii. The
population from the highest elevation of origin was signifi-
cantly smaller than other populations based on differences
among levels of the population main effect. Also, populations
from the midelevation range margin showed a greater in-
crease in growth at the range center than two high elevation
populations. Hiesey et al. (1971) also found some evidence
for genetically based clinal growth differences among M.
lewisii populations. However, because in their study popu-
lations were collected from throughout the geographic ranges
of both species, the wide latitudinal and longitudinal dis-
tances that separated most populations from the transplant
sites are not easily separated from the effects of adaptation
For both species, we detected variation among populations
in reaction norms for average annual fitness versus transplant
site, but these differences were not consistent with the hy-
pothesis that populations are adapted to their elevation of
origin. For example, at Mather (1400 m), the nearby Tuol-
umne River populations of both species (1320 m) were not
more fit than the distant M. cardinalis Mariposa Creek (590
m) or M. lewisii Warren Fork (2750 m) populations. The M.
cardinalis population-by-site interaction resulted from pop-
ulations differing in the degree of decrease in fitness with
increasing elevation. The significant M. lewisii population-
by-site interaction arose because three populations (Tuol-
umne River, 1320 m; Tamarack Creek, 1920 m; and Tioga
Road seep, 2580 m) displayed significantly increased fitness
at 2395 m versus other elevations and three did not (Por-
cupine Creek, 2400 m; Snow Creek, 2690 m; and Warren
Fork, 2750 m). Reaction norms for fitness did not cross, but
instead differed in the slope of decrease from the range center
to range margins, suggesting that populations do not exhibit
symmetrical ‘‘home’’ elevation advantages. This conclusion
is supported by the lack of significant correlations between
population mean fitness and the difference in elevation be-
tween population origin and transplant site.
Gene Flow and Selection
Range limits arise where populations are no longer able
to adapt sufficiently to local environmental conditions. Low
fitness of both species at their range margin suggests that
adaptation to the marginal environment is hindered. Like-
wise, weak differentiation among populations within each
species indicates that populations from the range margin have
been unable to adapt to environmental conditions at the range
A similar lack of strong regional adaptation was observed
in a recent study of Clarkia xantiana transplanted within and
beyond the range boundary, in which populations showed
much greater fitness declines when moved beyond the range
than when transplanted to a different location within the range
(Geber and Eckhart 2005). The weak adaptation to position
of origin within the range observed in these studies is striking
given the number of documented examples of adaptive dif-
ferentiation both among populations at geographic scales
MIMULUS ELEVATION RANGES
(e.g., Clausen et al. 1940; Grant 1963) and within populations
at extremely local spatial scales (e.g., Bradshaw 1960;
Schemske 1984). Many species display ecotypic variation
along elevation gradients (Clausen et al. 1940; Oleksyn et
al. 1998; Jonas and Geber 1999). The populations used in
the present experiment were sampled along an elevation gra-
dient that imposes variation in several important abiotic en-
vironmental variables, including length of growing season
and temperature. Species may not be able to adapt to envi-
ronmental conditions at the range margin if they lack appro-
priate genetic variation upon which selection can act or if
differential natural selection is weak relative to the homog-
enizing effects of gene flow (Mayr 1963; Kirkpatrick and
The interplay of gene flow and selection along environ-
mental gradients or between discrete environments is im-
portant to several models of range or niche evolution (Holt
and Gaines 1992; Kawecki 1995; Kirkpatrick and Barton
1997; Gomulkiewicz et al. 1999; Holt 2003). For example,
Kirkpatrick and Barton (1997) modeled the evolution of a
quantitative character determining fitness across a one-di-
mensional environmental gradient. The character evolved un-
der stabilizing selection toward an optimum phenotype that
varied with the environmental gradient. Population density
in their model depended on dispersal, density-dependent pop-
ulation regulation, and the degree of mismatch between the
optimum and population mean phenotypes. Stable range lim-
its arose when gene flow imposed a strong constraint on local
adaptation, as when dispersal was high or the environmental
gradient was steep.
Although the focus of the Kirkpatrick and Barton model
was on the swamping effects of gene flow, it also modeled
adaptive trade-offs between environments because no single
phenotype was optimal across the entire environmental gra-
dient. Models of niche evolution explicitly consider the role
of trade-offs between habitats in limiting species distribu-
tions, finding that selection to improve adaptation to envi-
ronments outside of the niche may be weak due to the de-
mographic asymmetry between habitats within versus outside
of the niche (Kawecki 1995; Holt 1996; Gomulkiewicz et al.
1999). In a recent model of range evolution, Holt (2003)
explicitly modeled the feedback between the evolution of
dispersal and the evolution of habitat specialization (i.e.,
trade-offs) in a two-habitat model where neither habitat was
initially outside of the niche. In this model the evolutionary
dynamics of the geographic range depended on the shape of
adaptive trade-offs between habitats and the initial habitat
distribution of the population. For instance, a species initially
specialized to one habitat may evolve habitat generalization
if mutations that increase adaptation to a new habitat have
little cost to fitness within the present habitat. Conversely,
if a linear and symmetrical trade-off in fitness between two
habitats exists, evolution will favor increased specialization
to whichever habitat the species initially resides in. These
models highlight the need to understand the relative roles of
dispersal, adaptive trade-offs, and demographic asymmetries
between habitats in range evolution. Further work is neces-
sary to understand how these components interact to deter-
mine the elevation range limits of Mimulus cardinalis and M.
perimental analog to latitudinal distributions at larger spatial
scales, because both arise along continuous environmental
gradients and encompass multiple populations. The environ-
mental gradient from the center to the edge of elevation and
latitude ranges is also similar, with temperature and length
of growing season decreasing to the north and at higher el-
evations, although the rate of change in environmental pa-
rameters across space is greater for elevation than for latitude
gradients. Indeed, a change of 100–200 m in elevation is
roughly equivalent to a change of one degree in latitude (Crid-
dle et al. 1994; Flebbe 1994). Due to the steepness of the
environmental gradient across elevation, for a given dispersal
distance, individuals encounter a more different environment
than if dispersing across latitude, making it more likely that
marginal populations may be swamped by centrally adapted
phenotypes at elevation range boundaries than at latitudinal
range boundaries (Kirkpatrick and Barton 1997).
Little is known about mechanisms of dispersal of M. car-
dinalis and M. lewisii seeds. Because both species are found
in riparian habitats, it is possible that seed dispersal via down-
stream currents provides a mechanism for primarily unidi-
rectional long-distance dispersal among populations, setting
up an interesting dichotomy between M. cardinalis and M.
lewisii at their shared midelevation range boundary. A net
flux of migrants downstream would imply that the M. lewisii
midelevation range limit may be subject to swamping gene
flow from high elevation central populations, but that the M.
cardinalis midelevation range limit is not. However, gene
flow via pollen may show the opposite pattern due to the
greater flight distance of hummingbirds, the primary polli-
nator of M. cardinalis, compared to bumblebees, the primary
pollinator of M. lewisii. Estimations of FSTamong popula-
tions of each species are in progress to begin to identify
patterns of gene flow among central and marginal populations
of each species.
Because central and marginal pop-
ulations of each species display few adaptive differences ver-
sus elevation, interspecific comparisons are necessary to un-
derstand adaptive trade-offs across the elevation gradient.
Since their recent common ancestor, M. cardinalis and M.
lewisii have evolved differences that restrict their distribu-
tions to different areas of the complex environmental gradient
associated with elevation. Specialization to different eleva-
tion ranges suggests that different phenotypes are necessary
for fitness at low versus high elevations. Estimation of the
strength and direction of selection on phenotypic traits across
the elevation gradient, in combination with genetic mapping
of quantitative trait loci, will identify traits under selection
at high versus low elevation and the underlying genetic ar-
chitecture of those traits (A. L. Angert, H. D. Bradshaw, and
D. W. Schemske, unpubl. data). Experimental evolution of
segregating hybrid populations at low and high elevation will
also illuminate whether there are fitness costs of speciali-
zation to low versus high elevation (A. L. Angert, H. D.
Bradshaw, and D. W. Schemske, unpubl. data). Together,
these studies will help elucidate mechanisms of adaptive
trade-offs between low and high elevation environments. In
conjunction with estimates of gene flow between central and
Elevation distributions offer a tractable ex-
A. L. ANGERT AND D. W. SCHEMSKE
marginal populations, we hope to understand what causes
and constrains adaptation to different elevation ranges.
We thank J. Anderson, K. Brady, B. Clifton, D. Ewing, D.
Grossenbacher, B. Igic, and A. Wilkinson for help in the field
and in the greenhouse; S. Beatty, L. Ford, C. Millar, P.
Moore, and P. Sterbentz for assistance in obtaining permits;
and the Stone family for generous donation of their property
and limitless hospitality. Financial support was provided by
grants from the California Native Plant Society, the North-
west Orchid Society, Sigma Xi, and the National Science
Foundation (Graduate Research Fellowship to AA and DEB-
0075660). J. Conner, K. Gross, J. Ramsey, J. Sobel, and two
anonymous reviewers provided many helpful comments on
Antonovics, J. 1976. The nature of limits to natural selection. Ann.
MO. Bot. Gard. 63:224–247.
Asselin, H., S. Payette, M. J. Fortin, and S. Vallee. 2003. The
northern limit of Pinus banksiana Lamb. in Canada: explaining
the difference between the eastern and western distributions. J.
Beardsley, P. M., A. Yen, and R. G. Olmstead. 2003. AFLP phy-
logeny of Mimulus section Erythranthe and the evolution of hum-
mingbird pollination. Evolution 57:1397–1410.
Bradshaw, A. D. 1960. Population differentiation in Agrostis tenuis
Sibth. III. Populations in varied environments. New Phytol. 59:
Bradshaw, A. D., and T. McNeilly. 1991. Evolutionary response to
global climatic change. Ann. Bot. 67:5–14.
Bradshaw, H. D., and D. W. Schemske. 2003. Allele substitution
at a flower colour locus produces a pollinator shift in monkey-
flowers. Nature 426:176–178.
Bradshaw, H. D., K. G. Otto, B. E. Frewen, J. K. McKay, and D.
W. Schemske. 1998. Quantitative trait loci affecting differences
in floral morphology between two species of monkeyflower (Mi-
mulus). Genetics 149:367–382.
Brown, J. H. 1984. On the relationship between abundance and
distribution of species. Am. Nat. 124:255–279.
Brown, J. H., G. C. Stevens, and D. M. Kaufman. 1996. The geo-
graphic range: size, shape, boundaries, and internal structure.
Annu. Rev. Ecol. Syst. 27:597–623.
Bullock, J. M., R. J. Edwards, P. D. Carey, and R. J. Rose. 2000.
Geographical separation of two Ulex species at three spatial
scales: Does competition limit species’ ranges? Ecography 23:
Carbonell, R., and J. L. Telleria. 1998. Increased asymmetry of
tarsus-length in three populations of blackcaps Sylvia atricapilla
as related to proximity to range boundary. Ibis 140:331–333.
Caswell, H., R. Lensink, and M. G. Neubert. 2003. Demography
and dispersal: life table response experiments for invasion speed.
Clausen, J. C., D. D. Keck, and W. M. Hiesey. 1940. Experimental
studies on the nature of species. I. Effect of varied environments
on western North American plants. Carnegie Institute of Wash-
ington Publ. no. 520, Washington, DC.
———. 1948. Experimental studies on the nature of species. III.
Environment responses of climatic races of Achillea. Carnegie
Institution of Washington Publ. no. 581, Washington, DC.
Cleavitt, N. 2004. Comparative ecology of a lowland and a sub-
alpine species of Mnium in the northern Rocky Mountains. Plant
Criddle, R. S., M. S. Hopkin, E. D. McArthur, and L. D. Hansen.
1994. Plant distribution and the temperature coefficient of me-
tabolism. Plant Cell Environ. 17:233–243.
Crozier, L. G. 2004. Field transplants reveal summer constraints on
a butterfly range expansion. Oecologia 141:148–157.
Cumming, G. S. 2002. Comparing climate and vegetation as limiting
factors for species ranges of African ticks. Ecology 83:255–268.
Dahl, E. 1951. On the relation between summer temperature and
the distribution of alpine vascular plants in the lowlands of Fen-
noscandia. Oikos 3:22–52.
Davis, A. S., P. M. Dixon, and M. Liebman. 2003. Cropping system
effects on giant foxtail (Setaria faberi) demography: II. Retro-
spective perturbation analysis. Weed Sci. 51:930–939.
Davison, A. W. 1977. The ecology of Hordeum murinum L.: III.
Some effects of adverse climate. J. Ecol. 65:523–530.
Denham, A. J., and T. D. Auld. 2004. Survival and recruitment of
seedlings and suckers of trees and shrubs of the Australian arid
zone following habitat management and the outbreak of rabbit
calicivirus disease (RCD). Austral. Ecol. 29:585–599.
Dudycha, J. L., and A. J. Tessier. 1999. Natural genetic variation
of life span, reproduction, and juvenile growth in Daphnia. Evo-
Flebbe, P. A. 1994. A regional view of the margin: salmonid abun-
dance and distribution in the Southern Appalachian Mountains
of North Carolina and Virginia. Trans. Am. Fish. Soc. 123:
Fox, G. A. 2001. Failure time analysis: studying times to events
and rates at which events occur. Pp. 235–266 in S. M. Scheiner
and J. Gurevitch, eds. Design and analysis of ecological exper-
iments. Oxford Univ. Press, Oxford, U.K.
Garcia, D., R. Zamora, J. M. Gomez, P. Jordano, and J. A. Hodar.
2000. Geographical variation in seed production, predation and
abortion in Juniperus communis throughout its range in Europe.
J. Ecol. 88:436–446.
Garcia-Ramos, G., and M. Kirkpatrick. 1997. Genetic models of
adaptation and gene flow in peripheral populations. Evolution
Gaston, K. J. 2003. The structure and dynamics of geographic rang-
es. Oxford Univ. Press, Oxford, U.K.
Geber, M. A., and V. M. Eckhart. 2005. Experimental studies of
adaptation in Clarkia xantiana. II. Fitness variation across a
subspecies border. Evolution 59:521–531.
Gilbert, N. 1980. Comparative dynamics of a single-host aphid. I.
The evidence. J. Anim. Ecol. 49:351–369.
Gomulkiewicz, R., R. D. Holt, and M. Barfield. 1999. The effects
of density dependence and immigration on local adaptation and
niche evolution in a black-hole sink environment. Theor. Popul.
Gonzalez-Guzman, L. I., and D. W. Mehlman. 2001. Developmental
stability across the breeding distribution of the scissor-tailed
flycatcher (Tyrannus forficatus). Ecol. Lett. 4:444–452.
Good, R. D. O. 1931. A theory of plant geography. New Phytol.
Grant, V. 1963. The origin of adaptations. Columbia Univ. Press,
Griggs, R. F. 1914. Observations on the behavior of some species
at the edges of their ranges. Bull. Torrey Bot. Club 41:25–49.
Haldane, J. B. S. 1956. The relation between density regulation and
natural selection. Proc. R. Soc. Lond. B 145:306–308.
Hennenberg, K. J., and H. Bruelheide. 2003. Ecological investi-
gations on the northern distribution range of Hippocrepis comosa
L. in Germany. Plant Ecol. 166:167–188.
Hersteinsson, P., and D. W. Macdonald. 1992. Interspecific com-
petition and the geographical distribution of red and Arctic foxes
Vulpes vulpes and Alopex lagopus. Oikos 64:505–515.
Hickman, J. C. 1993. The Jepson manual: higher plants of Cali-
fornia. Univ. of California Press, Berkeley, CA.
Hiesey, W. M., M. A. Nobs, and O. Bjo ¨rkman. 1971. Experimental
studies on the nature of species. V. Biosystematics, genetics,
and physiological ecology of the Erythranthe section ofMimulus.
Carnegie Institute of Washington Publ. no. 628, Washington,
Hoffman, A. A., and M. W. Blows. 1994. Species borders: ecolog-
ical and evolutionary perspectives. Trends Ecol. Evol. 9:
Holt, R. D. 1996. Adaptive evolution in source-sink environments:
MIMULUS ELEVATION RANGES
direct and indirect effects of density-dependence on niche evo-
lution. Oikos 75:182–192.
———. 2003. On the evolutionary ecology of species’ ranges. Evol.
Ecol. Res. 5:159–178.
Holt, R. D., and M. S. Gaines. 1992. Analysis of adaptation in
heterogeneous landscapes: implications for the evolution of fun-
damental niches. Evol. Ecol. 6:433–447.
Jonas, C. S., and M. A. Geber. 1999. Variation among populations
of Clarkia unguiculata (Onagraceae) along altitudinal and lati-
tudinal gradients. Am. J. Bot. 86:333–343.
Jones, R. H., and R. R. Sharitz. 1998. Survival and growth of woody
plant seedlings in the understorey of floodplain forests in South
Carolina. J. Ecol. 86:574–587.
Juenger, T., and J. Bergelson. 2000. The evolution of compensation
to herbivory in scarlet gilia, Ipomopsis aggregata: herbivore-
imposed natural selection and the quantitative genetics of tol-
erance. Evolution 54:764–777.
Kawecki, T. J. 1995. Demography of source-sink populations and
the evolution of ecological niches. Evol. Ecol. 9:38–44.
Keith, D. A. 2002. Population dynamics of an endangered heathland
shrub, Epacris stuartii (Epacridaceae): recruitment, establish-
ment and survival. Austral. Ecol. 27:67–76.
Kirkpatrick, M., and N. H. Barton. 1997. Evolution of a species
range. Am. Nat. 150:1–23.
Kuehl, R. O. 2000. Design of experiments: statistical principles of
research design and analysis. Duxbury Press, Pacific Grove, CA.
Lee, C. E., J. L. Remfert, and G. W. Gelembiuk. 2003. Evolution
of physiological tolerance and performance during freshwater
invasions. Integr. Comp. Biol. 43:439–449.
MacArthur, R. H. 1972. Geographical ecology: patterns in the dis-
tribution of species. Harper and Row, New York.
Marshall, J. K. 1968. Factors limiting the survival of Corynephorus
canescens (L.) Beauv. in Great Britain at the northern edge of
its distribution. Oikos 19:206–216.
Mayr, E. 1963. Animal species and evolution. Harvard Univ. Press,
McKee, J., and A. J. Richards. 1996. Variation in seed production
and germinability in common reed (Phragmites australis) in Brit-
ain and France with respect to climate. New Phytol. 133:
Møller, A. P. 1995. Patterns of fluctuating asymmetry in sexual
ornaments of birds from marginal and central populations. Am.
Oleksyn, J., J. Modrzynski, M. G. Tjoelker, R. Zytkowiak, P. B.
Reich, and P. Karolewski. 1998. Growth and physiology ofPicea
abies populations from elevational transects: common garden
evidence for altitudinal ecotypes and cold adaptation. Funct.
Pigott, C. D., and J. P. Huntley. 1981. Factors controlling the dis-
tribution of Tilia cordata at the northern limits of its geographical
range. III. Nature and causes of seed sterility. New Phytol. 87:
Prince, S. D. 1976. The effect of climate on grain development in
barley at an upland site. New Phytol. 76:377–389.
Ramsey, J., H. D. Bradshaw, and D. W. Schemske. 2003. Com-
ponents of reproductive isolation between the monkeyflowers
Mimulus lewisii and M. cardinalis (Phrymaceae). Evolution 57:
Randall, M. G. M. 1982. The dynamics of an insect population
throughout its altitudinal distribution: Coleophora alticolella
(Lepidoptera) in northern England. J. Anim. Ecol. 51:993–1016.
Richter, T. A., P. I. Webb, and J. D. Skinner. 1997. Limits to the
distribution of the Southern African ice rat (Otomys sloggetti):
thermal physiology or competitive exclusion? Funct. Ecol. 11:
Root, T. 1988. Energy constraints on avian distributions and abun-
dances. Ecology 69:330–339.
Sagarin, R. D., and S. D. Gaines. 2002. The ‘‘abundant centre’’
distribution: to what extent is it a biogeographical rule? Ecol.
SAS Institute. 1999. SAS OnlineDoc, Ver. 8. SAS Institute Inc.,
Saxton, A. M. 1998. A macro for converting mean separation output
to letter groupings in Proc Mixed. Proceedings of the 23rd SAS
Users Group International, Nashville, TN.
Scheidel, U., S. Rohl, and H. Bruelheide. 2003. Altitudinal gradients
of generalist and specialist herbivory on three montane Aster-
aceae. Acta Oecol. 24:275–283.
Schemske, D. W. 1984. Population structure and local selection in
Impatiens pallida (Balsaminaceae), a selfing annual. Evolution
Schemske, D. W., and H. D. Bradshaw. 1999. Pollinator preference
and the evolution of floral traits in monkeyflowers (Mimulus).
Proc. Natl. Acad. Sci. USA 96:11910–11915.
Sievert, P. R., and L. B. Keith. 1985. Survival of snowshoe hares
at a geographic range boundary. J. Wildl. Manag. 49:854–866.
Stanton, M. L., and C. Galen. 1997. Life on the edge: adaptation
versus environmentally mediated gene flow in the snow butter-
cup, Ranunculus adoneus. Am. Nat. 150:143–178.
Taniguchi, Y., and S. Nakano. 2000. Condition-specific competi-
tion: implications for the altitudinal distribution of stream fishes.
Terborgh, J., and J. S. Weske. 1975. The role of competition in the
distribution of Andean birds. Ecology 56:562–576.
Travis, J. 1994. Ecological genetics of life-history traits: variation
and its evolutionary significance. Pp. 187–204 in L. A. Real, ed.
Ecological genetics. Princeton Univ. Press, Princeton, NJ.
Turesson, G. 1922. The genotypical response of the plant species
to the habitat. Hereditas 3:211–350.
Verhoeven, K. J. F., T. K. Vanhala, A. Biere, E. Nevo, and J. M.
M. Van Damme. 2004. The genetic basis of adaptive population
differentiation: a quantitative trait locus analysis of fitness traits
in two wild barley populations from contrasting habitats. Evo-
Vickery, R. K., Jr. 1967. Experimental hybridizations in the genus
Mimulus. VI. Section Erythranthe. Proc. Utah Acad. Sci. 44:
———. 1978. Case studies in the evolution of species complexes
in Mimulus. Evol. Biol. 11:404–506.
Woodward, F. I. 1975. The climatic control of the altitudinal dis-
tribution of Sedum rosea (L.) Scop. and S. telephium L. II. The
analysis of plant growth in controlled environments. New Phytol.
———. 1990. The impact of low temperatures in controlling the
geographical distribution of plants. Philos. Trans. R. Soc. Lond.
Zacherl, D., S. D. Gaines, and S. I. Lonhart. 2003. The limits to
biogeographical distributions: insights from the northward range
extension of the marine snail, Kelletia kelletii (Forbes, 1852). J.
Corresponding Editor: M. Geber
1684 Download full-text
A. L. ANGERT AND D. W. SCHEMSKE
Source populations and reciprocal transplant sites. Abbreviations: M, Mariposa county; T, Tuolumne county; YNP, Yosemite National
Park; SNF, Stanislaus National Forest; INF, Inyo National Forest; na, not applicable.
(county, nearest landmark)Waterway Coordinates Elevation (m)
M, SNF, Buck Meadows
M, Midpines County Park
M, Triangle Rd. bridge
M, YNP, Yosemite Valley
T, YNP, Carlon Day Use Area
T, YNP, Carlon Day Use Area
M, YNP, Tamarack Flat
M, YNP, Porcupine Flat
T, YNP, Tioga Rd.
M, YNP, May Lake
Mono, INF, Tioga Pass
T, Quartz Mountain
T, Camp Mather
T, White Wolf Ranger Station
Mono, Saddlebag Lake
S. Fork Tuolumne R.
S. Fork Tuolumne R.
Warren Fork Lee Vining R.
N 37.4867?, W 119.9690?
N 37.7770?, W 120.0635?
N 37.5258?, W 119.9185?
N 37.5171?, W 119.8374?
N 37.7427?, W 119.5616?
N 37.8152?, W 119.8657?
N 37.8152?, W 119.8657?
N 37.7572?, W 119.7399?
N 37.8072?, W 119.5478?
N 37.8129?, W 119.5035?
N 37.8365?, W 119.4944?
N 37.9520?, W 119.2261?
N 37.9173?, W 120.4212?
N 37.8855?, W 119.8553?
N 37.8718?, W 119.6507?
N 37.9615?, W 119.2808?