• Diodia teres;
  • ecotypic differentiation;
  • environmental variation;
  • genetic drift;
  • local adaptation;
  • population divergence;
  • reciprocal transplant


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • • 
    Local adaptation is common, but tests for adaptive differentiation frequently compare populations from strongly divergent environments, making it unlikely that any influence of stochastic processes such as drift or mutation on local adaptation will be detected. Here, the hypothesis that local adaptation is more likely to develop when the native environments of populations are more distinct than when they are similar was tested.
  • • 
    A reciprocal transplant experiment including two populations from each of three habitats was conducted to determine the pattern of local adaptation. In addition to testing for local adaptation at the population level, the hypothesis was tested that local adaptation is more common between populations from different habitats than between populations from the same habitat.
  • • 
    Local adaptation was not common, but more evidence was found of local adaptation between populations from different habitats than between populations from the same habitat. Two instances of foreign genotype fitness advantage confirm that stochastic processes such as drift can limit local adaptation.
  • • 
    These results are consistent with the hypothesis that stochastic processes can inhibit local adaptation but are more likely to be overwhelmed by natural selection when populations occur in divergent environments.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The study of adaptive population divergence has been a major focus in evolutionary biology and reciprocal transplant experiments have been a critical tool in this endeavor. Common reports of greater fitness of native relative to transplanted foreign genotypes in reciprocal transplants suggest that natural populations are frequently adapted to their local environments (reviewed by Linhart & Grant, 1996; Schluter, 2000; Geber & Griffen, 2003). A caveat to this generalization is that reciprocal transplant studies may be biased towards finding local adaptation because they are often designed as studies of adaptation to particular environmental features. Comparison of populations from strongly divergent environments is appropriate for such a purpose, as for example, in the comparison of plant populations on serpentine soils and nonserpentine soils (Brady et al., 2005). However, because reciprocal transplant studies are often designed to test for adaptive divergence, they may not consider evidence that not all population differentiation is adaptive and that processes other than natural selection can also contribute to population divergence in fitness.

A number of reciprocal transplant studies have not found consistent evidence for local adaptation (Antonovics & Primack, 1982; Schmidt & Levin, 1985; Rice & Mack, 1991; Galloway & Fenster, 2000; Callahan & Pigliucci, 2002; Angert & Schemske, 2005; Geber & Eckhart, 2005) and some report instances of greater fitness of foreign than of native genotypes (Schmidt & Levin, 1985; Rice & Mack, 1991; Galloway & Fenster, 2000). These patterns that are not consistent with ubiquitous local adaptation, especially significant foreign genotype fitness advantage, raise the question of why the most fit genotypes are not always the ones that occupy a site.

Theory treats adaptation as the result of the interplay between selection and nonselective processes including gene flow (reviewed in Endler, 1977; Slatkin, 1985), mutation (Phillips, 1996) and genetic drift (Crow & Kimura, 1970). Laboratory studies of mutation accumulation have shown that mutation and drift can give rise to lineages that differ in fitness (reviewed by Lynch et al., 1999), but empirical evidence supporting the contribution of nonselective processes to divergence in fitness of natural populations is rare. Data from natural populations have shown that selection can overcome the effects of even fairly strong gene flow (reviewed by Lenormand, 2002; McKay & Latta, 2002, cf Sambatti & Rice, 2006). Similarly, comparisons of the magnitude of population divergence in phenotypic traits and neutral markers suggest that selection nearly always accounts for more variation than drift (Merila & Crnokrak, 2001). However, these generalizations may also be affected by a bias toward study systems in which selection regimes are likely to be very different (Merila & Crnokrak, 2001). Even a significant pattern of local adaptation does not rule out contributions of nonadaptive process to population differentiation for fitness. For example, Hendry & Taylor (2004) showed that the magnitude of differences in adaptive traits between pairs of three-spine stickleback populations from lakes and adjacent streams was negatively correlated with estimates of gene flow between them, suggesting that local adaptation to lake and stream environments would be more pronounced in the absence of gene flow.

When both adaptive and nonadaptive processes contribute to population differentiation in fitness, a logical expectation is that when the selective environments of populations contrast strongly, selection will overwhelm stochastic effects and produce a pattern of local adaptation (Lande, 1976; Turelli et al., 1988). Conversely, when differences in the selection regimes experienced by populations are more modest, nonadaptive processes may be able to weaken or prevent local adaptation. Reciprocal transplant studies that include one population from each of two strongly contrasting environments do not allow a test of the hypothesis that greater environmental divergence is more likely to produce a pattern of local adaptation, and are unlikely to reveal evidence for a role of stochastic factors. Some previous studies have included multiple populations from strongly contrasting environments to examine the repeatability of adaptive divergence (Via, 1991; Schluter, 1995, cf Kawecki & Ebert, 2004). Recently, Becker et al. (2006) performed reciprocal transplants among multiple populations from each of three geographic regions and found evidence of local adaptation of populations to regions but not to sites in close proximity to one another within a region. These results are consistent with a greater influence of nonadaptive processes on variation among populations within regions than between different regions. Here we extend this approach to examine adaptive divergence among multiple populations from similar vs contrasting environments to test the hypothesis that populations from more different environments have diverged more in fitness than those from more similar environments.

We conducted a reciprocal transplant among six populations of the annual plant Diodia teres (Rubiaceae) comprising two populations from each of three contrasting habitats to test the hypothesis that local adaptation is more likely to develop between populations from contrasting habitats than between populations from more similar habitat types. We also quantified several environmental variables in the home sites of the study populations to describe the pattern of environmental differences within and among habitats.

Materials and Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Diodia teres Walt. (Rubiaceae) is a self-compatible annual plant ranging from tropical to temperate climates (Kearney & Peebles, 1964). Its small, inconspicuous flowers, along with high rates of autonomous seed set in the glasshouse, and greater than expected homozygosity within populations (J. Hereford, unpublished) suggest a high rate of self-fertilization. In the south-eastern USA, it is found in a range of habitat types from coastal sand dunes to inland forests, where it occurs along roadsides and in canopy gaps. Previous studies have demonstrated local adaptation of populations of D. teres to inland agricultural and coastal habitats in North Carolina (Jordan, 1992). The present study included two populations from each of three habitat types in the Florida panhandle, designated Dunes, Sandhills and Inland (Fig. 1). Distances between the study populations range from 31 to 185 km, making contemporary gene flow between populations very unlikely. This conclusion is supported by analyses of allozyme allele frequencies, which indicate pronounced structure overall (FST = 0.46; J. Hereford, unpublished) and between populations from the same habitat type (FST = 0.11; J. Hereford, unpublished). The habitats of the study populations span an environmental gradient from the coastal plain (Dunes habitat) to the Piedmont (Inland habitat) along which we expected soil type to become less sandy and the vegetation to become denser and more dominated by overhead canopy.


Figure 1. The locations of the six study populations in northern Florida and southern Georgia, USA. Two were located in Dunes (DU) habitat, two in Inland (IN) habitat and two in Sandhills (SA) habitat.

Download figure to PowerPoint

To quantify environmental differences among the native sites of our six study populations, soil particle size and three aspects of plant competition were measured at each site. These are the environmental variables that appear to most consistently differentiate these habitat types and that were expected to contribute most to differences in the patterns of selection among habitats, though the existence of other variables that may also contribute cannot be ruled out. Soil particle size has a substantial influence on the water-holding capacity of soil (Barbour et al., 1987), and soil water availability has been shown to be an important agent of natural selection on annual plants (Dudley, 1996; Huber et al., 2004). Competition from conspecific and heterospecific plants has also been demonstrated to exert selection on plants (Dudley & Schmitt, 1996) and to contribute to adaptive differentiation between populations in different environments (Kindell et al., 1996; Sambatti & Rice, 2006).

The percentage cover of D. teres, of all other herbaceous plant species and of canopy from woody vegetation along 15 randomly located 1-m transects at the site of each population included in the study was estimated in June of 2004. The proportion of the transect tape covered by leaves of a given type of vegetation was converted to percentage cover of that type of vegetation (Barbour et al., 1987). At the same 15 locations, soil samples were taken and used gravitational analysis (Cox, 1978) was used to estimate the percentages of the soil made up by sand, silt and clay. Although environmental variables were measured in only one year, neither canopy cover nor soil texture would be expected to vary among years, and the rank order of per cent cover of D. teres and other herbaceous species would be expected to remain constant even though absolute amounts might differ. Over the 3-yr of the study there were no apparent changes in the relative density of D. teres, other herbaceous plants, or woody species among the planting sites (J. Hereford pers. obs.).

Tests for local adaptation

To determine the pattern of fitness differentiation among our populations, a six-by-six reciprocal transplant experiment was performed among the populations in each of two years. In winter of 2000 seeds from 25–45 individual plants in each population were collected. In March of 2001, one seedling from each of the field maternal genotypes (between 16 and 41 seedlings per population depending on germination) was planted in the glasshouse, and allowed to mature and produce seeds in a common maternal environment. The positions of plants in the glasshouse were periodically haphazardly shuffled to minimize edge and micro-environmental effects. In March of 2002, seeds produced in the glasshouse were germinated in sand, and seedlings were transferred to flats filled with potting soil as they emerged. Approximately 20 seedlings from each source population, each representing a different original maternal family, were planted at the native site of each of the six populations between April and May 2002, at the time of natural seedling emergence for a total of 749 seedlings. All seedlings transplanted to a given site were planted on the same day and were of similar size (two to four true leaves). Seedlings were assigned planting locations haphazardly and were transplanted directly into the soil, with minimal disturbance to natural vegetation. In December of 2002, when most plants had senesced at all sites, the total number of fruits produced by each individual were counted as an estimate of fitness. Plants that died before producing fruit were assigned a fitness of zero.

We repeated the reciprocal transplant in 2003 for the same set of populations as in 2002 with some modifications of the design. In winter of 2000 seeds were collected from additional maternal plants in each population as part of another experiment. Between seven and nine individuals per population of this collection were grown from seed in the glasshouse and allowed to self-fertilize and set seeds. The progeny from this generation were allowed to self-fertilize in the glasshouse for a second generation. These seeds were the source for the second reciprocal transplant. The seeds were germinated in February and March of 2003 and transplanted up to 10 seedlings (an average of seven) from each maternal family into each field planting site between March and May for a total of 2009 seedlings. At each planting site, 8–10 blocks accommodating up to 36 transplants each were established and seedlings were allocated to blocks such that plants from different source populations and families within source populations were equally likely to be planted on any given planting day and in any given block. For each seedling, the date of planting and the initial size, estimated from the size and number of leaves present on the day of planting, were recorded.

Data analysis

To determine whether the native environments (planting sites) of populations from the same habitat type were more similar to each other than to those of populations from different habitats, the means of environmental variables were converted into coordinate data and Euclidean distances between all pairs of native environments were calculated (cf Montalvo & Ellstrand, 2001). Environmental distances were calculated based on percentage cover of conspecifics, herbaceous cover, canopy cover and percentage sand in the soil analysis. Percentages of clay and silt were not included in this analysis because they are necessarily correlated with percentage of sand.

Local adaptation was tested for by comparing the fruit production of native and nonnative transplants in each planting site. Fruit set was not normally distributed in either year, and neither natural-log nor square-root transformation succeeded in making the distributions normal. In both years, the variance in fruit set was much larger than the mean, suggesting a negative binomial distribution (Bliss & Fisher, 1953; White & Bennetts, 1996; Quinn & Keough, 2002). Consequently, all the analyses of fruit set were performed under a generalized linear model (GLiM) framework (McCullagh & Nelder, 1989), with a natural-log link function and negative binomial distribution function. All tests were also carried out assuming a Poisson distributed response, including a correction for overdispersion (McCullagh & Nelder, 1989). There were no qualitative differences in the outcomes of the two types of analyses. All GLiM were performed with SAS procedure GENMOD (SAS Institute, 1999).

Differences between the design of the 2002 and 2003 transplant experiments necessitated separate analyses of the two data sets and precluded a clean test of differences in the pattern of adaptation between years. The primary difference in design was that seedlings planted in 2003 comprised selfed maternal sibships, and were produced by parents that had experienced an additional generation of self-fertilization in the glasshouse. In addition, the seedlings were planted into spatial blocks in 2003 only, and we were able to include data on initial seedling size and planting date as covariates in the analysis.

For the 2002 data set, the effects of planting site, source population, and planting site by source population interaction were quantified. Local adaptation between populations would be indicated by a significant planting site–source population interaction, and a pattern of greater fitness of native than nonnative populations. For analysis of the 2003 transplant data, the effects of several additional variables were quantified. An effect of family nested within source population was included to account for the effects of genetic variation within populations, and the effect of block nested within planting site to account for spatial environmental variation within planting sites. Models that included family and block as random effects failed to converge on a likelihood estimate, so in order to test the effects of site and population, family and block were treated as fixed effects. These models will not provide estimates of the effect sizes of family or block, but they control for variation in these factors in our tests of the effects of primary interest, planting site and source population (Allison, 2005, p. 6). The 2003 analysis also included the effects of seedling size at planting and date of planting as covariates.

To reveal patterns of local adaptation between pairs of populations, the fruit set of the native transplants at each planting site was compared with that of transplants from each foreign source population in GLiM. If a native population set more fruits than a foreign population, the native population was considered locally adapted relative to that foreign population. Separate analyses were performed for each year, and the effects of family nested within population, experimental block, initial size at planting and date of planting were included in the comparisons of populations for the 2003 analyses. We used the Bonferroni method to adjust the critical values for these tests for the number of comparisons at each planting site within in each year (Sokal & Rohlf, 1995, p. 240).

To test the hypothesis that local adaptation between populations from different habitats was more likely than between populations from the same habitat, the frequency of local adaptation in comparisons of populations from the same habitat type was compared with the frequency in comparisons of populations from different habitats. Each comparison of the fitness of native genotypes with the fitness of genotypes from a foreign population constitutes an observation in this analysis. Because the hypothesis tested in this analysis concerns the frequency of local adaptation rather than the case for local adaptation for each site, a comparison was scored as showing local adaptation if the native population had significantly greater fruit set at P < 0.05 (i.e. we did not correct for multiple comparisons). The Mantel-Haenszel χ2 test (Sokal & Rohlf, 1995, p. 760), implemented with SAS procedure FREQ (SAS Institute, 1999) was used to perform the analysis.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Environmental distance

The percentage cover of D. teres, percentage cover of other herbs, percentage canopy cover and soil composition varied considerably among planting sites (Table 1). As expected, the distinguishing features of the Dunes planting sites were the high sand content of the soil and the complete absence of canopy cover. The Inland habitat had much less sandy soil and greater canopy cover. The Sandhills sites were intermediate, on average, in both sand content and canopy cover, although the two Sandhills sites differed from each other in that one site had no canopy and the other had the greatest canopy cover of any site.

Table 1.  Mean percentage cover of Diodia teres, other herbs and canopy, and proportion of the soil composed of sand and clay at the sites of the six study populations
Planting site%D. teres% Herbs% CanopySandClay
  1. Values are means (standard deviation); data in italic type indicate habitat means (SD). The proportion silt is not presented because it is the remaining proportion after sand and clay have been quantified.

Inland 13.67 (3.75)15.63 (14.75)34.33 (45.39)0.63 (0.06)0.18 (0.07)
Inland 27.93 (9.34)33.27 (28.43)17.33 (32.40)0.55 (0.04)0.10 (0.04)
Mean5.80 (7.32)24.45 (23.99)25.83 (39.70)0.59 (0.06)0.17 (0.07)
Sandhills 10.13 (0.52)8.93 (9.98)0.00 (0.00)0.81 (0.03)0.06 (0.02)
Sandhills 27.20 (6.01)7.87 (11.11)51.67 (41.13)0.86 (0.02)0.06 (0.01)
Mean3.67 (0.52)8.40 (10.40)25.83 (38.82)0.84 (0.04)0.07 (0.02)
Dunes 10.93 (1.67)38.13 (29.63)0.00(0.00)0.97 (0.01)0.01 (0.01)
Dunes 22.40 (3.71)8.87 (13.81)0.00 (0.00)0.97 (0.01)0.01 (0.01)
Mean1.67 (2.93)23.50 (27.16)0.00 (0.00)0.97 (0.01)0.01 (0.01)

Our qualitative assignment of planting sites to habitats was not entirely supported by quantitative measurement of environmental variables. The Euclidean distances between sites suggest that Dunes and Inland represent distinct habitats and that the two Sandhills sites may not represent the same habitat. The environmental distances between the two Dunes sites and between the two Inland sites were smaller than those between either of them and any site from the alternative (Dunes or Inland) habitat (Table 2). The two Sandhills sites were more distant from each other than each was from one or more of the sites from other habitats, suggesting that they may not represent a single habitat distinct from Dunes and Inland.

Table 2.  Euclidian distances between all pairs of sites based on percentage cover of Diodia teres, other herbaceous plants, tree canopy and percentage sand in the soil
  1. Data in italic type indicate distances between sites from the same habitat type. IN, Inland; SA, sandhills; DU, dunes.

IN2 3.061.652.612.61
SA1  2.781.951.80
SA2   2.852.24
DU1    1.52

Variation in fruit production

The overall proportion of seedlings that survived to flowering was 0.70 in 2002 and 0.65 in 2003, with much lower survival at the Dunes 2 planting site (0.30 in 2002 and 0.16 in 2003). Mean fruit set of transplants ranged more than tenfold from 3.77–55.04, with a grand mean of 18.6 in 2002 (Table 3), and even more widely, from 0.42–65.76, with a mean of 15.64, in 2003 (Table 4). This range is typical for naturally occurring individuals across these sites, and transplants were not obviously different from the naturally occurring plants surrounding them at each site (data not shown). In both years, the range of means varied more widely among planting sites (fourfold to eightfold) than among source populations (c. twofold), and planting site and source population means varied more in 2003 than in 2002 (Tables 3 and 4).

Table 3.  Mean number of fruits produced by transplants from each source population of Diodia teres transplanted into each planting site in 2002
 Source population
IN1IN2SA1SA2DU1DU2Site mean
  1. Values are means (standard deviation, sample size); data in italic type indicate sets of four means from the same habitat.

  2. Row marginal totals are means (SD and sample sizes) for each source population across all planting sites, and column marginals are means for each planting site. Marginal means that share superscripts are not significantly different. *, Population means that differ significantly from the native population at a given planting site after Bonferroni correction (α of 0.05/5 = 0.01); t, differences that are significant without correction (α = 0.05). IN, Inland; SA, sandhills; DU, dunes.

Planting site
IN131.54 (42.69, 24)36.76 (30.42, 17)32.86 (37.28, 21)22.07 (26.79, 27)17.41 (23.07, 22)16.65 (21.35, 31)25.12A (31.10, 142)
IN210.37 (17.34, 16)6.44 (8.98, 18)8.05 (10.90, 21)5.27 (7.04, 22)3.78 (6.24, 23)3.77 (6.08, 31)5.88D (9.53, 131)
SA19.30* (5.21, 10)5.50* (2.88, 6)33.13 (34.89, 15)15.42t (16.72, 12)32.50 (18.12, 10)26.00 (18.10, 20)22.64C,B,A (22.54, 73)
SA217.10 (11.76, 10)24.08 (18.29, 12)20.72 (12.60, 25)22.09 (12.96, 31)13.86t (13.92, 21)11.90* (8.27, 31)17.87C (13.17, 130)
DU113.69t (10.90, 16)16.33t (9.19, 12)23.26 (11.40, 19)17.76t (14.29, 25)28.76 (15.30, 27)24.19 (13.34, 36)21.84B,A (13.83, 135)
DU27.00 (13.38, 12)6.83 (13.35, 12)5.50t (12.4, 22)13.20 (31.25, 25)55.04 (72.63, 26)15.12 (30.19, 41)19.33C,B,A (42.00, 138)
Population mean16.93A,B (26.13, 88)17.42A,B (21.14, 77)19.81A,B (24.04, 123)16.59B (20.81, 142)25.53A (39.07, 129)15.86B (19.97, 190)18.60 (25.97, 749)
Table 4.  Mean number of fruits produced by reciprocal transplants in 2003 at each planting site for each Diodia teres source population
 Source population
IN1IN2SA1SA2DU1DU2Site mean
  1. Values are means (standard deviation, sample size); the marginal sample size for source populations is not a total sample size, but the number of maternal families from that population included in the experiment.

  2. Data in italic type indicate sets of four means from the same habitat. Row marginal totals are means (SD and sample sizes) for each source population across all planting sites, and column marginals are means for each planting site. Marginal means that share superscripts are not significantly different. *, Population means that differ significantly from the native population at a given planting site after Bonferroni correction (α of 0.05/5 = 0.01); t, differences that are significant without correction (α = 0.05). IN, Inland; SA, sandhills; DU, dunes.

Planting site
IN122.90 (28.60, 67)30.07t (39.69, 46)65.76* (70.48, 66)28.08 (31.68, 24)35.51 (73.94, 57)19.84* (26.37, 56)35.02A (53.53, 316)
IN217.46 (29.08, 67)17.34 (14.73, 44)26.91 (29.35, 65)19.59 (21.41, 34)13.49t (19.39, 59)11.59t (14.67, 61)17.73B (23.33, 330)
SA110.42* (12.06, 66)16.70 t (16.49, 46)26.97 (30.45, 66)15.18 (16.72, 43)21.18 (26.57, 60)13.86t (14.37, 61)17.56B (21.40, 342)
SA23.47* (4.24, 68)7.09 (8.55, 47)12.11 (13.48, 64)8.97 (8.59, 44)6.01* (7.53, 70)4.64t (6.07, 56)6.95C (9.00, 349)
DU15.28* (3.36, 66)10.11* (7.10, 44)15.49t (15.35, 65)12.41* (8.13, 41)21.29 (11.41, 59)20.27 (13.17, 55)14.17B (12.18, 330)
DU21.75 (4.84, 63)0.42 (1.93, 43)6.99 (17.18, 68)4.03 (9.04, 40)7.86* (20.11, 72)1.23 (6.11, 56)4.09D (13.07, 342)
Population mean10.28D (19.21, 9)13.74A,B,C (21.30, 9)25.70A (40.13, 8)13.53A,B,C (17.64, 8)16.81B (34.34, 7)11.91D (16.53, 9)15.64 (27.94, 2009)

Adaptation to local site

Our analysis revealed evidence of some local adaptation in both years of the study (Table 5), but overall, home site genotype advantage in fruit production was infrequent. The native population produced significantly more fruits than a foreign population in 3 of 30 comparisons in 2002 (Table 3) and in 7 of 30 comparisons in 2003 (Table 4), after accounting for multiple tests. Evidence for home site advantage was restricted to four of the six populations, three of which (Sandhills 1, Sandhills 2 and Dunes 1), showed evidence of local adaptation in both years. Notably, we found no case in which the native population produced significantly more fruit than transplants from the other population of the same habitat type (Tables 3 and 4). In only one case was there a significant reciprocal home site advantage, between Dunes population 1 and Sandhills population 2 in 2003 (Table 4), and no population had a home site advantage with respect to all other foreign populations.

Table 5.  Generalized linear model of the effects of planting site, Diodia teres source population, and their interaction on fruit number in each of two years
Source of variation/yeardfχ2P-value
  1. The analysis for 2003 also includes the effects of block nested within planting site and family nested within population, and initial size and date of planting as covariates.

  2. The denominator degrees of freedom were 713 for 2002 and 1870 for 2003.

Planting site 589.19< 0.0001
Source population 510.060.0737
Planting site × source population2590.15< 0.0001
Initial size at planting 1159.43< 0.0001
Date of planting 125.05< 0.0001
Block nested within planting site44362.40< 0.0001
Families nested within source population4784.240.0007
Planting site 5507.22< 0.0001
Source population 571.70< 0.0001
Planting site × source population25175.20< 0.0001

In two instances, both in 2003, a foreign population produced significantly more fruits than the native population. The Sandhills 1 population produced more than three times the number of fruits of the native population at the Inland 1 site, and the Dunes 1 population produced more than five times as many fruits as the Dunes 2 population at the Dunes 2 site (Table 4). The average fruit production of the Dunes 1 population was also more than three times the average for the Dunes 2 population at the Dunes 2 site in 2002, but this difference was not significant (Table 3).

Local adaptation was significantly more frequent in comparisons between populations from different habitats than between populations from the same habitat in 2003. The frequency of local adaptation when populations were from the same habitat was 0, and the frequency when the populations were from different habitats was 0.542 (Mantel–Haenszel χ2 = 5.54, df = 1, P < 0.02. For the 2002 analysis, local adaptation was again more frequent when populations were from different habitats (0.333 vs 0.167), but the difference was not significant (Mantel–Haenszel χ2 = 0.613, df = 1, P = 0.433. For these analyses of the frequency of local adaptation, the two Sandhills sites were treated as belonging to the same habitat even though measures of environmental distance suggest that their environments are dissimilar in some respects (Table 1). This decision was based on the strong similarity between the sites in soil composition, which was the variable that most consistently distinguished our sites (Table 1). We note that treating these potentially different environments as representing the same habitat biases against our conclusion that local adaptation is less likely for populations from similar habitats.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Our results support contributions of both adaptive and nonadaptive divergence to the pattern of fitness differentiation among the six study populations of D. teres. We found several instances of local adaptation between populations from different habitats, but no evidence of adaptive divergence between populations from the same habitat type. Overall, limited evidence of significant native population fitness advantage was found, and all such cases involved an advantage over a population from an alternate habitat. This pattern is consistent with the hypothesis that natural selection is more likely to produce a pattern of local adaptation when populations come from more strongly contrasting environments of different habitats, but that other processes can predominate when environments are more similar. Two examples of significant foreign genotype advantage constitute additional, strong evidence of the contribution of nonadaptive processes to fitness divergence.

The frequency of local adaptation was low overall, and both the frequency and pattern of home site advantage differed between the two years of the study (Tables 3 and 4). Differences in the experimental procedure and analysis for the two transplant experiments might account for some quantitative differences between years in the patterns of local adaptation observed, but temporal variation in the environment is a more likely explanation for qualitative differences. Larger sample sizes and the inclusion of block and maternal family as effects, and seedling initial size and planting date as covariates could have conferred additional statistical power for the 2003 analysis. Greater power could account for the larger number of statistically significant differences between population means in 2003 (Table 4) than in 2002 (Table 3), though the overall pattern was similar in that little evidence of significant home site advantage was found between populations from the same habitat in either year. Temporal variation in patterns of local adaptation is nearly ubiquitous in reciprocal-transplant studies conducted for more than 1 yr (Rice & Mack, 1991; van Tienderen & van der Toorn, 1991; Jordan, 1992; Galloway & Fenster, 2000), possibly reflecting fluctuation in environmental variables that affect differences in selective regimes. Differences in the temporal distribution of rainfall in 2002 and 2003 might account for the differences observed in the pattern of fitness divergence among populations and habitats, but we do not have data at a fine enough spatial scale to confirm this hypothesis.

The hypothesis that local adaptation is more likely when populations are native to more divergent environments was supported by the greater frequency of home site advantage in comparisons of populations native to different habitats than between populations from the same habitat, though the difference in frequency was significant only for the 2003 data. The results of several studies that, like this one, have included a range of environmental differences among transplant sites are also consistent with greater evidence of local adaptation for populations from more divergent environments. For example, Geber & Eckhart (2005) reported local adaptation of two subspecies of Clarkia xantiana to their native ranges, which differed substantially in a variety of habitat features, but did not find evidence of local adaptation of populations within each relatively environmentally homogeneous range (Schemske, 1984; Angert & Schemske, 2005; Becker et al., 2006). Similarly, studies in which populations were chosen for reasons other than the contrast in their native environments are also those in which the hypothesis of local adaptation was not consistently supported (Antonovics & Primack, 1982; Rice & Mack, 1991; Galloway & Fenster, 2000; Baack & Stanton, 2005). These results, along with those of this study, suggest that limits to adaptation may be more common than would be expected based on reciprocal transplant studies designed to study adaptation to divergent environments.

One caveat regarding our conclusions of local adaptation is that because we transplanted seedlings rather than seeds, the magnitudes of local adaptation reported here could be underestimates if local adaptation is expressed at the seed-germination stage. Lack of information on this stage of the life cycle in the field could have weakened some patterns in our analyses, though we have no reason to believe that the consequences of this omission would affect our conclusions regarding the frequency or relative magnitude of fitness differences between populations from different habitats relative to those from different sites within habitats.

Cases of foreign genotype advantage, along with the absence of home site advantage in fruit production in comparisons of populations from the same habitat in our study, suggest that there is a level of environmental divergence at which stochastic processes, such as mutation and genetic drift can overwhelm differences in selection. Dunes population 1 produced greater than three times the number of fruits of Dunes population 2 at the Dunes 2 site in both years, clearly illustrating that the most fit genotypes are not necessarily those that currently occupy a site. Interestingly, the populations that produced genotypes that showed foreign advantage, Sandhills 1 and Dunes 1, were also among the three that showed evidence of home site adaptation in both years of the study, suggesting that these populations were inherently superior to the others. One possible explanation for overall superiority of some populations is that some deleterious alleles become fixed by drift, and populations vary in the size of this fixed genetic load (Kimura et al., 1963; Whitlock et al., 2000). Crosses between plant populations have sometimes revealed the existence of fixed genetic load (Ouborg & van Treuren, 1994; Paland & Schmid, 2003; Kawecki & Ebert, 2004), as well as variation among populations in the magnitude of load (Paland & Schmid, 2003). If some populations of D. teres carry smaller fixed loads they could have greater fruit production in multiple habitats for this reason alone. Galloway & Fenster (2000) suggested a similar scenario for foreign site advantage and lack of local adaptation, which they observed for populations of the annual plant, Chamaecrista fasiculata. An alternative possibility is that some populations carry a novel mutation that confers high fitness in a range of environments but that it has not yet reached all of the study populations (Wright, 1931; Peck et al., 1998). We cannot distinguish between these and other possible explanations with the current data.

Our choice of a study system comprising widely spaced populations of a self-compatible species with limited dispersal increased the likelihood that drift would contribute to differentiation. The contributions of selection, drift, mutation and gene flow to differentiation in fitness will vary among sets of populations as a function of mating system, dispersal ability, population size, physical distances among populations and other system-specific features. More studies that consider the roles of both selective and stochastic processes in a variety of systems are needed to elucidate the balance among these forces and will contribute to a more nuanced understanding of patterns of population differentiation and the processes that produce them.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

We thank D. Houle and J. Travis for advice and comments at all stages of this study, and J. Travis for comments on earlier versions of the paper. Ruth Shaw and three anonymous reviewers provided valuable suggestions for revising the paper. The following managers and agencies generously allowed us to conduct research on their land: G. Seamon at The Nature Conservancy, S. Herman, C. Martin and T. Engstrom at Pebble Hill Plantation and Tall Timbers Research Station, D. Scott at Saint Marks National Wildlife Refuge and the Florida Department of Environmental Protection. This work was supported by NSF award DEB 9903878 to A.A.W. and NSF DDIG 0407968 to A.A.W. and J.H.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Allison P. 2005. Fixed effects regression methods for longitudinal data using SAS. Cary, NC, USA: SAS Press.
  • Angert AL, Schemske DW. 2005. The evolution of species’ distributions: reciprocal transplants across the elevation ranges of Mimulus cardinalis and M. lewisii. Evolution 59: 16711684.
  • Antonovics J, Primack RB. 1982. Experimental ecological genetics in Plantago. VI. The demography of seedling transplants of P. lanceolata. Journal of Ecology 70: 5575.
  • Baack EJ, Stanton ML. 2005. Ecological factors influencing tetraploid speciation in snow buttercups (Ranunculus adoneus): niche differentiation and tetraploid establishment. Evolution 59: 19361944.
  • Barbour MG, Burk JH, Pitts WD. 1987. Terrestrial plant ecology. Menlo Park, CA, USA: The Benjamin/Cummings Publishing Company Inc.
  • Becker U, Colling G, Dostal P, Jakobsen A, Matthies D. 2006. Local adaptation in the monocarpic perennial Carlina vulgaris at different spatial scales. Oecologia 150: 506518.
  • Bliss CI, Fisher RA. 1953. Fitting the negative binomial distribution to biological data. Biometrics 9: 176200.
  • Brady KU, Kruckeberg AR, Bradshaw HD. 2005. Evolutionary ecology of plant adaptation to serpentine soils. Annual Review of Ecology Evolution and Systematics 36: 243266.
  • Callahan HS, Pigliucci M. 2002. Shade-induced plasticity and its ecological significance in wild populations of Arabidopsis thaliana. Ecology 83: 19651980.
  • Cox GW. 1978. Laboratory manual of general ecology. Dubuque, IA, USA: WC Brown Company Publishers.
  • Crow JF, Kimura M. 1970. An introduction to population genetics theory. Minneapolis, MN, USA: Burgess Publishing Company.
  • Dudley SA. 1996. Differing selection on plant physiological traits in response to environmental water availability: a test of adaptive hypotheses. Evolution 50: 92102.
  • Dudley SA, Schmitt J. 1996. Testing the adaptive plasticity hypothesis: density-dependent selection on manipulated stem length in Impatiens capensis. American Naturalist 147: 445465.
  • Endler JA. 1977. Geographic variation, speciation, and clines. Princeton, NJ, USA: Princeton University Press.
  • Galloway LF, Fenster CB. 2000. Population differentiation in an annual legume: local adaptation. Evolution 54: 11731181.
  • Geber MA, Eckhart VM. 2005. Experimental studies of adaptation in Clarkia xantiana. II. fitness variation across a subspecies border. Evolution 59: 521531.
  • Geber MA, Griffen LR. 2003. Inheritance and natural selection on functional traits. International Journal of Plant Sciences 164: S21S42.
  • Hendry AP, Taylor EB. 2004. How much of the variation in adaptive divergence can be explained by gene flow? An evaluation using lake-stream stickleback pairs. Evolution 58: 23192331.
  • Huber H, Kane NC, Heschel MS, Von Wettberg EJ, Banta J, Leuck AM, Schmitt J. 2004. Frequency and microenvironmental pattern of selection on plastic shade-avoidance traits in a natural population of Impatiens capensis. American Naturalist 163: 548563.
  • Jordan N. 1992. Path-analysis of local adaptation in two ecotypes of the annual plant Diodia teres Walt (Rubiaceae). American Naturalist 140: 149165.
  • Kawecki TJ, Ebert D. 2004. Conceptual issues in local adaptation. Ecology Letters 7: 12251241.
  • Kearney TH, Peebles RH. 1964. Arizona flora. Los Angeles, CA, USA: University of California Press.
  • Kimura M, Maruyama T, Crow JF. 1963. The mutation load in small populations. Genetics 48: 13031312.
  • Kindell CE, Winn AA, Miller TE. 1996. The effects of surrounding vegetation and transplant age on the detection of local adaptation in the perennial grass Aristida stricta. Journal of Ecology 84: 745754.
  • Lande R. 1976. Natural selection and random genetic drift in phenotypic evolution. Evolution 30: 314334.
  • Lenormand T. 2002. Gene flow and the limits to natural selection. Trends in Ecology and Evolution 17: 183189.
  • Linhart YB, Grant MC. 1996. Evolutionary significance of local genetic differentiation in plants. Annual Review of Ecology and Systematics 27: 237277.
  • Lynch M, Blanchard J, Houle D, Kibota T, Schultz S, Vassilieva L, Willis J. 1999. Perspective: spontaneous deleterious mutation. Evolution 53: 645663.
  • McCullagh P, Nelder JA. 1989. Generalized linear models. London, UK: Chapman and Hall.
  • McKay JK, Latta RG. 2002. Adaptive population divergence: markers, QTL and traits. Trends in Ecology and Evolution 17: 285291.
  • Merila J, Crnokrak P. 2001. Comparison of genetic differentiation at marker loci and quantitative traits. Journal of Evolutionary Biology 14: 892903.
  • Montalvo AM, Ellstrand NC. 2001. Nonlocal transplantation and outbreeding depression in the subshrub Lotus scoparius (Fabaceae). American Journal of Botany 88: 258269.
  • Ouborg NJ, Van Treuren R. 1994. The significance of genetic erosion in the process of extinction. 4. Inbreeding load and heterosis in relation to population-size in the mint Salvia pratensis. Evolution 48: 9961008.
  • Paland S, Schmid B. 2003. Population size and the nature of genetic load in Gentianella germanica. Evolution 57: 22422251.
  • Peck SL, Ellner SP, Gould F. 1998. A spatially explicit stochastic model demonstrates the feasibility of Wright's Shifting Balance Theory. Evolution 52: 18341839.
  • Phillips PC. 1996. Waiting for a compensatory mutation: phase zero of the Shifting-Balance process. Genetical Research 67: 271283.
  • Quinn GP, Keough MJ. 2002. Experimental design and data analysis for biologists. Cambridge, MA, USA: Cambridge University Press.
  • Rice KJ, Mack RN. 1991. Ecological genetics of Bromus tectorum. III. The demography of reciprocally sown populations. Oecologia 88: 91101.
  • Sambatti JBM, Rice KJ. 2006. Local adaptation, patterns of selection, and gene flow in the Californian Serpentine Sunflower (Helianthus exilis). Evolution 60: 696710.
  • SAS Institute Inc. 1999. SAS/STAT user's guide, Version 8. Cary, NC, USA: SAS Institute Inc.
  • Schemske DW. 1984. Population structure and local selection in Impatiens pallida (Balsaminaceae), a selfing annual. Evolution 38: 817832.
  • Schluter D. 1995. Adaptive radiation in sticklebacks: trade-offs in feeding performance and growth. Ecology 76: 8290.
  • Schluter D. 2000. The ecology of adaptive radiation. New York, NY, USA: Oxford University Press.
  • Schmidt KP, Levin DA. 1985. The comparative demography of reciprocally sown populations of Phlox drummondii Hook. I. Survivorships, fecundities, and finite rates of increase. Evolution 39: 396404.
  • Slatkin M. 1985. Gene flow in natural populations. Annual Review of Ecology and Systematics 16: 393430.
  • Sokal RR, Rohlf FJ. 1995. Biometry. New York, NY, USA: WH Freeman and Company.
  • Van Tienderen PH, Van Der Toorn J. 1991. Genetic differentiation between populations of Plantago lanceolata. I. Local adaptation in three contrasting habitats. Journal of Ecology 79: 2742.
  • Turelli M, Gillespie JH, Lande R. 1988. Rate tests for selection on quantitative characters during macroevolution and microevolution. Evolution 42: 10851089.
  • Via S. 1991. The genetic structure of host plant adaptation in a spatial patchwork – demographic variability among reciprocally transplanted pea aphid clones. Evolution 45: 827852.
  • White GC, Bennetts RE. 1996. Analysis of frequency count data using the negative binomial distribution. Ecology 77: 25492557.
  • Whitlock MC, Ingvarsson PK, Hatfield T. 2000. Local drift load and the heterosis of interconnected populations. Heredity 84: 452457.
  • Wright S. 1931. Evolution in Mendelian populations. Genetics 16: 97159.