SEARCH

SEARCH BY CITATION

Keywords:

  • elevation;
  • geographic variation;
  • Pedicularis;
  • population differentiation;
  • reproductive ecology;
  • seed mass;
  • size–number trade-off

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

1. We examined geographic variation in mean individual seed mass (MISM) among 38 populations within and across nine Pedicularis species in the eastern Tibetan Plateau, measuring the effects of one extrinsic factor (elevation) and two intrinsic factors (plant size and mean seed number per fruit).

2. Across all populations, elevation is negatively correlated with MISM; within eight of nine species, this pattern is also observed among populations. This relationship, however, is not significant when controlling for variation in plant size or seeds per fruit. High-elevation populations produce smaller plants with more seeds per fruit than low-elevation populations; controlling for these variables eliminates the negative correlation between elevation and MISM.

3. Across all populations, the predicted trade-off between MISM and seeds per fruit is consistently strong, with or without controlling for the effects of plant size. By contrast, the trade-off between MISM and total seasonal fecundity can be detected only when controlling for plant size.

4. Independent of plant size, populations that produce small seeds do not support individuals with particularly low reproductive yield (fecundity × MISM). Accordingly, high-elevation populations exhibit neither lower reproductive yield nor smaller seeds than expected given their lower biomass.

5.Synthesis. In Pedicularis, elevation, plant size and seeds per fruit are all correlated with MISM among populations across species. Elevation is less important, however, than intrinsic factors in determining the MISM of a population; the effect of elevation on MISM disappears when the effects of intrinsic factors are controlled statistically. The observed decline in MISM with increasing elevation is therefore partly mediated by the decline in plant size and partly by an increase in mean seed number per fruit with elevation. Altitudinal variation in MISM across populations or species has been described before, but this is the first study to control for the effect of intrinsic factors simultaneously. This result calls into question the conclusions of studies that have detected geographic variation in MISM without controlling for variation in intrinsic factors.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

Seed mass varies greatly at all ecological levels, including across species (Harper, Lovell & Moore 1970; Westoby, Jurado & Leishman 1992; Moles et al. 2005a; Vamosi, Mazer & Cornejo 2008; Queenborough et al. 2009), within species (Winn & Werner 1987; Stromberg and Patten 1990; Winn & Gross 1993), among conspecific individuals (Janzen 1977; Michaels et al. 1988; Sakai & Sakai 1995; Vaughton & Ramsey 1998), within individuals (Thompson 1984; Wolf et al. 1986; Wulff 1986; Winn 1991; Obeso 1993; Méndez 1997; Vaughton & Ramsey 1998) and within fruits (Stanton 1984; Stöcklin & Favre 1994; Méndez 1997). Given that seed mass is often heritable and has strong effects on processes that influence survivorship and fecundity (as reviewed in Westoby, Jurado & Leishman 1992; Westoby, Leishman & Lord 1996; Moles et al. 2005b), it has been the focus of many studies of the process and outcome of natural selection.

Factors proposed to drive the evolution of seed mass fall into two distinct categories: extrinsic environmental conditions and intrinsic reproductive and life-history constraints (Venable 1992; Westoby, Leishman & Lord 1996; Moles et al. 2005b). Evidence that extrinsic conditions influence the evolution of seed mass includes the observation that distinct habitats or growing conditions often support species that differ in mean individual seed mass (MISM). For example, species with relatively large seeds are often associated with environmental challenges such as high interspecific competition (Rees 1995; Jakobsson & Eriksson 2000; Leishman 2001), low light availability (Salisbury 1942; Mazer 1989; Kelly & Purvis 1993; Saverimuttu & Westoby 1996; Hewitt 1998; Hodkinson et al. 1998; Bond, Honig & Maze 1999; Eriksson, Friis & Lofgren 2000), limited water availability (Baker 1972; Leishman & Westoby 1994; Murray et al. 2004; but see also Mazer 1989; Wright & Westoby 1999), other limited resources (Jurado & Westoby 1992; Westoby, Leishman & Lord 1996), basic (ultra-mafic) soils (Tautenhahn et al. 2008) and low latitude (Moles & Westoby 2003).

Seed mass has also been found to be associated with elevation, but the patterns observed are not highly consistent and the underlying mechanisms have not been identified. For example, Baker (1972) reported a negative relationship between seed mass and elevation among populations within and among species in the Californian flora. He argued that the lower temperatures at higher altitudes may reduce photosynthetic rates, and shorter growing seasons may reduce the time for seed development and seed provisioning. Plastic developmental responses as well as genetically based adaptive responses to these environmental conditions could therefore result in lower MISM in high-elevation populations and taxa. Bu et al. (2007) similarly found a negative correlation between elevation and seed mass in a flora of the eastern Tibetan Plateau.

By contrast, a positive relationship between seed mass and elevation was found between (but not within) congeneric lowland and alpine species (Pluess, Schutz & Stöcklin 2005), across species (Landolt 1967; Alexander et al. 2009) and within species (Mariko et al. 1993; Oyama 1993; Holm 1994; Lord 1994; Piano, Pecetti & Carroni 1996; Ayana & Bekele 2000; Boulli, Baaziz & M’Hirit 2001; Blionis & Vokou 2002). Pluess, Schutz & Stöcklin (2005) argued that natural selection should favour the production of larger seeds in species at higher elevations because larger seeds exhibit superior survivorship in stressful environments, thereby accounting for the pattern that they observed. The ecological sorting of species across elevations could also generate this pattern, independently of natural selection directly on seed mass; indeed, the fact that Pluess found no relationship between seed mass and elevation within species argues against a strong role for in situ natural selection.

The explanations proposed by Baker (1972) and Pluess, Schutz & Stöcklin (2005) are logical, but seed mass is also associated with a suite of intrinsic factors, such as vegetative biomass (plant size), flowering time, fruit size and seed number per fruit (reviewed in Primack 1987; Bolmgren & Cowan 2008; and references therein), any of which may covary with elevation and mask or confound the direct effects of elevation. Accordingly, several investigators have observed a positive interspecific relationship between plant size and seed mass (Mazer 1989; Thompson & Rabinowitz 1989; Rees 1996; Totland & Birks 1996; Moles & Westoby 2004, 2006; Grubb, Metcalfe & Coomes 2005; Vamosi, Mazer & Cornejo 2008; Queenborough et al. 2009). The increase in seed mass with plant size has been proposed to reflect adaptive responses to competitive interactions among seedlings, architectural constraints or dispersal requirements (Salisbury 1974; Tilman 1988; Thompson & Rabinowitz 1989; Moles & Westoby 2004, 2006; Grubb, Metcalfe & Coomes 2005; Moles et al. 2005b). Alternatively, differences in plant size and seed mass among populations or taxa may be due to plastic responses of both traits to local environmental conditions.

The mean seed mass of a population or taxon frequently may be correlated with plant size, but it is also considered to reflect the evolutionary outcome of a size–number compromise (Smith & Fretwell 1974; Westoby, Jurado & Leishman 1992). If the numerical advantage exhibited by small-seeded species is balanced by the survival advantage of large-seeded species during seedling establishment, then mean seed mass may often be selectively neutral. Moles (2004) has argued, however, that the greater number of seeds produced by small-seeded species may be only partly balanced by the higher rate of seedling survival associated with large seeds. In this case, the evolution of seed mass (and its association with adult plant size) must depend on its developmental and genetic correlations with a wide range of life-history traits, many of which may be correlated with plant size (Moles & Westoby 2004, 2006; Rees & Venable 2007; Falster, Moles & Westoby 2008; Venable & Rees 2009). Given that both plant size and mean seed number (per fruit or per individual) may affect MISM (Venable 1992; Jakobsson & Eriksson 2000; Aarssen & Jordan 2001), studies aiming to detect environmental conditions that are associated directly with evolutionary or environmentally induced changes in mean seed size within or among species should control for variation in both of these intrinsic factors.

In this study, we examined the effect of elevation on MISM among 38 populations representing nine species of Pedicularis in the eastern Tibetan Plateau, while controlling statistically for both mean plant size and mean seed number per fruit. Focusing on a single genus (as opposed to a highly taxonomically diverse community) allowed us to concentrate on ecologically and morphologically similar taxa while investigating sources of variation in seed mass. Here, we address the following primary questions:

  • 1
     Across populations, is MISM associated with elevation, and is this relationship independent of variation in plant size and seed number per fruit?
  • 2
     Is there a consistent trade-off between mean seed number per fruit and MISM across populations, independent of the effects of elevation and plant size on individual seed mass?
  • 3
     Is mean seed number per fruit associated with elevation independent of plant size, thereby directly or indirectly contributing to the decline in MISM with increasing elevation?
  • 4
     Across populations within and among taxa, which factor is most directly associated with MISM: elevation, plant size or mean seed number per fruit?

Altitudinal variation in MISM across populations or species has been described before, but this is the first study to control simultaneously for the effect of intrinsic factors. We found that the relationship between MISM and elevation is strongly influenced by the effect of plant size on MISM and by the trade-off between seed mass and seed number per fruit. Failure to control statistically for variation in these intrinsic traits can lead to misleading correlations between extrinsic factors (such as elevation) and mean seed mass.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

Study sites and sampling methods

Our investigations were conducted in the eastern Tibetan Plateau, considered to be the geographic centre of the origin and evolution of Pedicularis Linn. (Li 1951). In this region, Pedicularis species are distributed across a wide range of elevations, from 1600 to 5000 m a.s.l, but they share the same breeding system (self-compatible) and primary pollinators (bumblebees) (Wang & Li 1998; Macior, Tang & Zhang 2001). In this study, we focus on nine species (38 populations; P. chinensis, P. kansuensis and P. semitorta are biennial, the rest are perennial) that are widely distributed in Gansu province (China); each species is represented by at least three populations sampled in 2005 from different elevations located in the eastern Tibetan Plateau (population and species information are summarized in Appendix S1 in Supporting Information).

In 2005, we sampled wild populations at the fruiting stage (all flower production had ceased). In each population, we collected 15–37 individuals (see Appendix S1 for sample sizes) by severing the stem at the soil surface. Each plant was placed in a paper envelope until processed. From each infructescence on every plant, we removed 3–5 mature but unopened fruits. To reduce variation among individuals due to potential effects of fruit position on seed mass, we chose fruits at basal, middle and distal positions on each sampled infructescence. These fruits were dissected and the seeds removed and examined. Fully developed seeds were counted and oven-dried for a minimum of 48 h at 60 °C. We weighed the dry mass of all the fully developed seeds in a given fruit to the nearest 0.1 mg. From each fruit, the MISM was then estimated by dividing the total seed mass by the number of seeds. The mean value for all of the fruits sampled per plant was then used to estimate the MISM for an individual plant.

The total number of seeds produced per plant (i.e. fecundity in 2005) was estimated for each individual as the mean number of seeds per sampled fruit multiplied by the total number of fruits that the individual produced. Reproductive yield of each plant was estimated as total seed mass per plant (mean total seed mass per fruit multiplied by fruit number). Plant size was estimated by weighing to the nearest 0.1 mg the above-ground parts of each sampled individual, including stem, fruit stalks and leaves (i.e. vegetative mass). Below, we use ‘plant size’ to refer to vegetative mass. Population means for each trait were estimated based on phenotypic values estimated or measured on all individual plants.

Statistical analyses

All analyses were conducted using population mean phenotypes for the reproductive and vegetative traits described above. Because the published phylogenies for this large genus (> 400 species) are not consistent (Li 1951; Tsoong 1955), and because we sampled relatively few taxa (but multiple populations per taxon), we used an analysis of variance approach rather than a phylogenetic approach (e.g. Vamosi, Mazer & Cornejo 2008; Queenborough et al. 2009) to examine the phenotypic associations between elevation and our focal traits and to compare the relative importance of each factor in determining mean seed mass. The sampling of multiple populations per species allowed us to partition variance in MISM into components due to elevation, species identity, mean plant size (estimated as the above-ground stem biomass) and mean seed number per fruit.

For three reasons, we focused on the effects of seed number per fruit instead of on individual fecundity (fruit number per plant × mean seed number per fruit). First, seed number per fruit differs significantly among species and populations (within species), but is unaffected by plant size after species differences are taken into account (nested mixed ancova: effect of species: F8,28 = 12.63; P < 0.0001; effect of ln plant size: F1,1 = 0.49; P > 0.4912). Secondly, individual fecundity is much more strongly correlated with plant size and likely to be highly plastic (nested mixed ancova: effect of species: F8,28 = 8.89; P < 0.0001; effect of ln plant size: F1,1 = 35.12; P < 0.0001). Thirdly, among population means, seed number per fruit is independent of fecundity and may therefore influence MISM independent of variation in the latter (regression of ln fecundity on ln seeds per fruit: y = 0.165x + 5.192; r = 0.095; P > 0.5692; n = 38).

Given that, within species, only one population was sampled at a given elevation, we could not test simultaneously for variation in our focal traits due to both population identity and elevation. The inclusion of multiple populations per species, however, provided replication that enabled us to factor out the variation due to species membership when detecting other sources of variation in MISM in the analyses of covariance described below.

Using population means, linear regressions were conducted to examine the bivariate relationships among MISM, mean seed number per fruit, plant size and elevation. We sought significant associations in two ways. First, we determined whether the slope of the regression (with all populations included) was significantly different from zero. Secondly, for each bivariate relationship, we examined each species separately to determine whether a majority of them exhibited a positive (or negative) slope. Based on the discrete binomial distribution, if more than seven of nine species exhibit the same trend, this comprises a statistically significant majority (P < 0.0176).

When examining the relationship between two focal traits, we controlled for variation in other variables by using residuals. First, the residuals of seed mass on plant size and the residuals of seed mass on seed number per fruit were used in bivariate regressions to examine the relationship between seed mass and elevation, independent of variation in plant size or seed number per fruit.

Secondly, the residuals of mean plant size on population elevation were used in bivariate regressions to examine the relationship between seed mass and plant size independently of elevation. To control for the effect of seed number per fruit on MISM, we used the residuals of seed mass on seed number fruit to examine the more direct relationship between seed mass and plant size.

Thirdly, the residuals of seed number per fruit on plant size and on elevation were used to examine the relationship between seed mass and seeds per fruit independently of plant size or elevation. We also examined the relationship between mean seed number per fruit and population elevation when controlling for variation in plant size; and the relationship between seed number per fruit and plant size when controlling for population elevation.

Using linear regression, we examined the relationship between mean plant size and population elevation. In addition, we tested for a trade-off between seed number per plant and individual seed mass with and without controlling for variation in plant size. The bivariate relationship between total reproductive yield (total seed mass per plant) and MISM was also examined with and without controlling for variation in plant size. Variables were ln-transformed to improve normality.

To examine the effect of elevation, species membership, plant size and seeds per fruit on MISM, we performed ancova in which species membership was treated as a fixed effect and population elevation, plant size and seeds per fruit were included in the model as covariates. We compared this model to three less complete ancova models, examining the effects on MISM of: (i) species membership, elevation and plant size; (ii) species membership, plant size and seeds per fruit; (iii) species membership, elevation and seeds per fruit. We also conducted an ancova that included mean individual fecundity (total fruit number × mean seed number per fruit) to compare its effect on MISM to that of seed number per fruit. By comparing the effects of each factor across models, we could identify those factors that had significant effects on seed mass independently of the others. Within each model, type III sums of squares were used to test for significant effects of each of the included factors independent of the others.

We also conducted an ancova to detect the effects of species, elevation and plant size on the mean number of seeds per fruit. All of the analyses were analysed using jmp 7.0 (SAS INSTITUTE 2007).

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

Sources of variation in seed mass

The ancova detected that, relative to the other factors included in the model, species membership accounts for most of the variance in seed mass (Table 1). Species range in seed mass from 0.21 mg (SE = 0.0153) in P. polyodonta to 1.46 mg (SE = 0.167) in P. chinensis. Seed number per fruit accounts for a greater proportion of the variance in MISM than elevation or plant size, reflecting the trade-off between mean seed number per fruit and MISM. Elevation accounted for a significant proportion of the variance in seed mass only when either seed number per fruit or plant size was excluded from the model (Table 1: models 2 and 4). Independent of the other factors, plant size (above-ground vegetative mass) accounts for the least variation in seed mass (Table 1), and its effects were not statistically significant when elevation was included in the model (models 1, 2 and 5 vs. model 3). By contrast, the mean number of seeds per fruit had significant effects on MISM regardless of which other factors were included (models 1, 3, 4 and 5). Total seed number per plant did not have a significant effect on MISM independent of variation in mean seed number per fruit and mean plant size, although the effects of mean seed number per fruit on seed mass remain highly significant (model 5).

Table 1. ancovas to test the sources of variance in mean individual seed mass. Plant size was estimated as vegetative mass. Boldfaced P values are statistically significant at the < 0.05 level
Fit modelSourced.f.Sum of squaresF ratioP
Model 1 R= 0.94Species83.7925.78<0.0001
Elevation10.052.720.1111
ln plant size10.042.390.1340
ln seed number per fruit10.148.090.0086
Model118.0739.81<0.0001
Error260.48  
Total378.55  
Model 2 R= 0.93Species85.5029.52<0.0001
Elevation10.1807.780.0096
ln plant size10.0301.360.2535
Model107.92034.04<0.0001
Error270.63  
Total378.55  
Model 3 R= 0.94Species83.8124.26<0.0001
ln plant size10.126.220.0191
ln seed number per fruit10.2814.290.0008
Model108.0240.92<0.0001
Error270.53  
Total378.55  
Model 4 R= 0.94Species83.9325.37<0.0001
Elevation10.136.600.0160
ln seed number per fruit10.147.060.0131
Model108.0341.42<0.0001
Error270.52  
Total378.55  
Model 5 R= 0.94Species81.8812.26<0.0001
Elevation10.042.040.1653
ln plant size10.021.120.3009
ln seed number per fruit10.126.310.0189
ln seed number per plant10.0010.060.8117
Model128.0735.18 
Error250.48  
Total378.55  

Seed mass vs. elevation, plant size and seed number per fruit

The bivariate and residual analyses of population means enabled us to detect the sign and magnitude of the correlations between variables among populations within and across species. Across all populations, those at high elevations produced smaller seeds than those at low elevations (Fig. 1a). This pattern was also observed within eight of the nine species (a higher proportion than expected by chance; P < 0.0176). This relationship, however, is not a direct one; it is mediated by the effects of plant size and seed number per fruit on seed mass, both of which also covary with elevation. When the effect of plant size or seed number per fruit on seed mass was controlled by using the residuals of seed mass on plant size or on seed number per fruit, the relationship did not persist across populations, and species differed in the sign of the slope (Fig. 1b,c).

image

Figure 1.  Bivariate regressions among population means within and across species. (a) ln seed mass and elevation (slope = 0.00090, R= 0.24, P < 0.00019); (b) residuals of ln seed mass on ln plant size and elevation (slope = −0.00038, R= 0.09, ns); (c) residuals of ln seed mass on ln seed number per fruit on elevation (slope = −0.00017, R= 0.03, ns); (d) ln seed mass and ln plant size (slope = 0.23, R= 0.18, P < 0.0007); (e) ln seed mass and residuals of ln plant size on elevation (slope = 0.20, R= 0.11, P < 0.0445); (f) residuals of ln seed mass on ln seed number per fruit and ln plant size (slope = 0.09, R= 0.06, ns); (g) ln seed mass and ln seed number per fruit (slope = −0.69, R= 0.46, P < 0.0001); (h) ln seed mass and residuals of ln seed number per fruit on elevation (slope = −0.56, R= 0.33, P < 0.0002); (i) ln seed mass and residuals of ln seed number per fruit on ln plant size (slope = −0.60, R= 0.24, P < 0.0017). Solid lines indicate significant regression relationships among all of the populations (P < 0.05). Dashed lines represent the regression relationships among populations within each species. Within each plot is reported the fraction of species that exhibit the same qualitative slope (e.g. > 0 or < 0) and the outcome of a binomial test of whether the fraction differs significantly from 0.5. In the figures, Pedicularis species are coded as follows: 1 = P. cheilanthifolia; 2 = P. chinensis; 3 = P. kansuensis; 4 = P. lasiophrys; 5 = P. longiflora; 6 = P. polyodonta; 7 = P. rhinanthoides; 8 = P. semitorta; 9 = P. szetschuanica.

Download figure to PowerPoint

Across all populations, we observed a significant positive relationship between MISM and mean plant size (Fig. 1d), even when controlling for elevation (Fig. 1e). This pattern was also observed among population means within eight of nine species (Fig. 1d,e). This relationship, however, was not significant when controlling for the effect of seed number per fruit on MISM (Fig. 1f).

Across all populations, we observed a strong trade-off between MISM and mean seed number per fruit (Fig. 1g; eight of nine species exhibited the same relationship among population means), with or without controlling for elevation and plant size (Fig. 1g vs. 1h,i). The within-species patterns, however, became more variable when controlling for elevation and plant size (Fig. 1h,i).

Seed number per fruit vs. elevation and plant size

Populations at higher elevations produced more seeds per fruit than low-elevation populations (Fig. 2a), even when controlling for plant size (Fig. 2b). This pattern was also observed within seven of the nine species, although this is not a statistically significant majority. Corroborating these patterns, the ancova detected that seed number per fruit was significantly affected by species and elevation independently of plant size (effect of species: SS = 4.38, F8,1 = 11.90, P < 0.0001; effect of elevation: SS = 0.27, F1,1 = 5.79, P < 0.0233). This model detected no significant effect of plant size on seed number per fruit (SS = 0.01, F1,1 = 0.19, ≥ 0.6633).

image

Figure 2.  Bivariate regressions among population means within and across species. (a) ln seed number per fruit and elevation (slope = 0.00074, R= 0.29, P < 0.0005); (b) residuals of ln seed number per fruit on ln plant size and elevation (slope = 0.00073, R= 0.26, P < 0.0012); (c) ln seed number per fruit and ln plant size (slope = −0.20, R= 0.14, P < 0.0198); (d) ln seed number per fruit and residuals of ln plant size on elevation (slope = −0.09, R= 0.02, P < 0.3773). Solid lines indicate significant regression relationships among all of the populations (P < 0.05). Dashed lines represent the regression relationships among populations within each species. Within each plot is reported the fraction of species that exhibit the same qualitative slope (e.g. > 0 or < 0) and the outcome of a binomial test of whether the fraction differs significantly from 0.5. In the figures, species are coded as in Fig. 1.

Download figure to PowerPoint

Across all populations, mean plant size was negatively correlated with mean seed number per fruit (Fig. 2c), although this pattern is not observed consistently within species. In addition, elevation is inversely correlated with plant size (regression of ln plant biomass vs. elevation: y = −0.0013x + 3.75; r = −0.5; P < 0.0015; n = 38). Consequently, when controlling for the effects of elevation on plant size, seed number per fruit was independent of plant size (Fig. 2d).

Seed number–size trade-off at the whole plant level

At the whole-plant level, total seed number (1-year fecundity) was independent of MISM among populations (Fig. 3a). When variation in mean plant size was controlled statistically by using the residuals of total seed number per plant on plant size and the residuals of MISM on plant size, however, we detect a strong trade-off between 1-year fecundity and MISM (Fig. 3b). This trade-off was also observed among population means within seven of nine species.

image

Figure 3.  Bivariate regressions among population means within and across species. (a) ln seed number per plant and ln seed mass (slope = −0.37, R= 0.05, ns); (b) residuals of ln seed number per plant on ln plant size and residuals of ln seed mass on ln plant size (slope = −0.93, R= 0.60, P < 0.0001); (c) ln reproductive yield and ln seed mass (slope = 0.63, R= 0.13, P < 0.0280); (d) residuals of reproductive yield on ln plant size and residuals of ln seed mass on ln plant size (slope = −0.12, R= 0.02, ns). Solid lines indicate significant regression relationships among all of the populations (P < 0.05). Dashed lines represent the regression relationships among populations within each species. Within each plot is reported the fraction of species that exhibit the same qualitative slope (e.g. > 0 or < 0) and the outcome of a binomial test of whether the fraction differs significantly from 0.5. In the figures, species are coded as in Fig. 1.

Download figure to PowerPoint

Populations in which individuals produced large seeds appeared to have high reproductive yield (fecundity × MISM), which is expected given that reproductive yield includes MISM in its calculation (Fig. 3c). However, this relationship disappears when controlling for the effects of plant size on reproductive yield and mean seed mass (Fig. 3d). Due to the trade-off between 1-year fecundity and mean seed mass, mean seed mass is independent of reproductive yield, independent of plant size.

Effect of elevation on reproductive yield

Populations at high elevations exhibited lower reproductive yield than those at low elevations (Fig. 4a) due to the negative effects of elevation on plant size. When we controlled for variation among populations in mean plant size by using the residuals of reproductive yield on plant size, we found no significant association between elevation and reproductive yield (Fig. 4b).

image

Figure 4.  Bivariate regressions among population means within and across species. (a) ln reproductive yield and elevation (slope = −0.0010, R= 0.16, P < 0.0116); (b) residuals of ln reproductive yield on ln plant size and elevation (slope = 0.0001, R= 0.01, ns). Solid lines indicate significant regression relationships among all of the populations (P < 0.05). Dashed lines represent the regression relationships among populations within each species. Within each plot is reported the fraction of species that exhibit the same qualitative slope (e.g. > 0 or < 0) and the outcome of a binomial test of whether the fraction differs significantly from 0.5. In the figures, species are coded as in Fig. 1.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

Correlative studies do not allow the unambiguous identification of causal mechanisms that generate strong patterns. Nevertheless, detecting geographic variation in a trait can be a first step towards proposing mechanisms that may account for it. For example, the relationship between MISM and elevation observed here can be used to evaluate – but not to distinguish among – the non-mutually exclusive hypotheses that shorter growing seasons or cooler temperatures at higher elevations will: induce the production of smaller seeds (through phenotypic plasticity); exclude large-seeded taxa (through ecological species sorting), or favour smaller-seeded genotypes (via natural selection). Each of these predictions can be tested experimentally in the field or under controlled conditions.

Our most striking results were, first, among all 38 populations sampled, although MISM declines with elevation (Fig. 1a), this pattern disappears when controlling statistically for either above-ground plant biomass or mean seed number per fruit (Fig. 1b,c). Secondly, mean seed number per fruit increases with elevation (with and without controlling for variation in plant size; Fig. 2a,b) and decreases with plant biomass (Fig. 2c). The relationship between seed number per fruit and plant biomass, however, becomes independent when controlling statistically for elevation (Fig. 2d). Thirdly, we observed strong trade-offs between MISM and mean seed number per fruit, both with and without controlling for elevation and plant biomass (Fig. 1g–i). The combination of these observations indicates that the decline in mean seed mass with increasing elevation is partly mediated by: (i) the decline in plant size with elevation (r = −0.50; P < 0.0015, a pattern that has been found in many taxa (Körner 2003) and (ii) an increase in mean seed number per fruit with elevation.

Had we not controlled for variation in plant size and seed number per fruit, we may have reached the misleading conclusion that MISM declines directly with increasing elevation, rather than being mediated by variation in seed number per fruit and plant size. This interpretation is consistent with the possibility that the elevation gradient in mean seed size is the result of natural selection operating simultaneously on plant size, seed mass and seed number per fruit. For example, the shorter growing season and harsh winters at high elevation simultaneously may favour stunted plants, (small) seeds with short development times, and the production of fewer flowers (and fruits) to minimize flowering time.

Sources of variation in seed mass

The effect of elevation on seed mass

In this study, eight of nine species exhibited a negative correlation among population means between elevation and seed mass. Either negative or positive relationships between seed mass and elevation observed in previous studies have yielded to different explanations. First, a decline in seed mass with elevation may be due to environmentally induced plastic responses to a decline in resource availability (e.g. lower temperatures at higher elevations may reduce photosynthetic rates, and shorter growing seasons may reduce the time for seed development and seed provisioning, thereby reducing mature seed mass; Baker 1972; Totland & Birks 1996). The production of smaller seeds at high altitude may also evolve by natural selection if the growing season is not sufficiently long for large-seeded genotypes to produce mature seeds. Secondly, Pluess, Schutz & Stöcklin (2005) suggested that natural selection may favour larger seeds (or larger-seeded species) at higher elevations because they confer higher establishment success and greater resistance of seedlings to harsh climatic conditions than small ones.

When examining patterns of altitudinal variation in MISM across populations or species, however, previous investigators have not controlled for the effect of multiple intrinsic factors. Among the 38 Pedicularis populations studied here, we detected no significant effect of elevation on seed mass when controlling for plant size and seed number per fruit. Other studies of seed mass variation in relation to environmental conditions have reported similar results. For example, Mazer (1989) found that habitat explained a small proportion of inter-specific variation in seed mass among 648 species in the Indiana Dunes (USA) flora relative to the effects of taxonomic family, and life-form and life history. Similarly, there was little difference in mean seed mass among species occupying different soil types in Sydney, western New South Wales and Central Australia (Leishman, Westoby & Jurado 1995). Our results are consistent with Leishman, Westoby & Jurado (1995) perspective that differences in establishment conditions may not be the primary cause of geographic variation in individual seed mass even though seed mass is often observed to be positively related to the ability to establish (Gross & Werner 1982; Gross 1984; Winn 1988; Reader 1993; Parciak 2002). If we had not controlled for plant size and seed number per fruit, our results could have been erroneously interpreted to mean that elevation has a direct and negative effect on mean seed individual seed mass.

In this study, we cannot conclude that the decline in mean seed mass among populations with increasing elevation (Fig. 1a) is the adaptive outcome of natural selection, even though this pattern is observed within eight of nine species and across all populations. First, given that seed mass is independent of elevation (when controlling for plant size), we have no direct evidence that seed size has evolved among these Pedicularis populations in response to environmental differences between them. Secondly, although smaller seeds may in fact be adaptive at high elevations (e.g. if competitive interactions are weaker, or if the shorter growing season requires shorter development times, necessitating the production of smaller seeds), the strong trade-off between seed mass and seed number per fruit prevents the detection of a negative relationship between elevation and seed mass when controlling statistically for seed number per fruit. The reason for the higher number of seeds per fruit among high-elevation populations requires explanation; we do not know whether the increase is associated with greater ovule production per flower or whether it may be the result of better pollinator service. In short, we cannot determine whether the observed geographic variation in seed mass is the adaptive outcome of natural selection on this trait, results from phenotypic plasticity in response to local environmental conditions, is a correlated response to selection on plant size, or is a correlated result of phenotypic plasticity affecting plant size.

The effect of plant size on seed mass

Plant size is an important determinant of the amount of energy available for reproduction and seed development (Bazzaz, Ackerly & Reekie 2000). In models developed by Rees & Westoby (1997) and Falster, Moles & Westoby (2008), size dependent increases in reproductive allocation may increase population density, which favours competitively superior larger seeds. In these models, it is assumed that larger seeds have an initial establishment advantage, and that this initial size difference persists during growth such that plants derived from larger seeds have a competitive advantage later in life. However, Rees & Venable (2007) and Venable & Rees (2009) argue that, unless unrealistic assumptions are made, it is unlikely that small initial differences in seed size will influence the competitive ability of individuals many months or years after recruitment occur; they suggest that much of the observed increase in seed mass with adult plant size may be due to constraints correlated with the latter.

Given this logic, a positive relationship among species or populations between MISM and plant size is predicted. Indeed, positive relationships between seed mass and plant size have been reported in many previous studies (Levin 1974; Thompson & Rabinowitz 1989; Leishman, Westoby & Jurado 1995; Rees 1996; Moles & Westoby 2004; Rees & Venable 2007; Venable & Rees 2009). Several explanations for the positive relationship have been proposed. For example, the mechanical constraints hypothesis predicts that the fragile branches of small plants should only be able to bear relatively small seeds, whereas large plants should be robust enough to bear either large or small seeds (Grubb, Metcalfe & Coomes 2005); the result of this constraint is a positive correlation among species’ means between MISM and mean plant size (with variance in seed mass increasing with mean plant size). Secondly, according to Charnov’s life-history theory, large offspring are necessary to offset the low survivorship to adulthood that would otherwise be a consequence of the longer juvenile periods associated with large adult size (Moles & Westoby 2004; Moles et al. 2005a). Thirdly, factors such as the intensity of competition for light and the height requirements needed for the successful dispersal of large seeds may contribute to a positive relationship between seed mass and plant size (Tilman 1988; Venable 1992).

In the present study, we detected no significant relationship between seed mass and plant size independently of seed number per fruit (Fig. 1f). Three factors contribute to the independence between individual seed mass and plant size among Pedicularis populations. First, Pedicularis populations of large plants produce fruits with relatively few seeds per fruit (Fig. 2c); these seeds, in turn, are relatively large due to the size–number trade-off. The negative relationship between seed number per fruit and plant size generates an indirect positive relationship between seed mass and plant size (Fig. 1d). However, this positive relationship disappears when controlling statistically for variation in seed number per fruit. It is possible that some of the positive relationships between MISM and plant size that have been widely reported are similarly due to the effects of other intrinsic traits that confound the relationship between seed mass and plant size.

Secondly, the positive relationship reported in other studies may arise because the data sets include species representing a variety of growth forms. For example, woody plants, which produce relatively large seeds, are typically larger than shrubs and herbs (Mazer 1989; Leishman, Westoby & Jurado 1995). In this study, all of the species are herbaceous, so there may not be sufficient variation in either seed mass or plant size to detect a strong positive relationship. Thirdly, in previous studies, the positive relationships between seed mass and plant size were usually detected across species while within species the relationships are often non-significant (Moles & Westoby 2004). This suggests that different mechanisms may operate at different ecological levels (e.g. within vs. across genera). In the present study, we focused on the relationship between seed mass and plant size among populations and species within a single genus. It may simply be more common to find positive relationships among a taxonomically highly diverse group of taxa.

The trade-off between mean individual seed massand seed number per fruit vs. per individual

Trade-offs between seed mass and seed number apparently occur because of resource constraints during seed provisioning and may contribute to the maintenance of seed mass variation within populations and species (Vaughton & Ramsey 1998). Negative, positive and independent relationships between mean seed mass and fecundity have been found in other interspecific and inter-population studies (Westoby, Jurado & Leishman 1992; Vaughton & Ramsey 1998; Aarssen & Jordan 2001; Leishman 2001; Koenig et al. 2009). Given that variation in plant size may mask trade-offs that would otherwise exist between seed mass and number (Venable 1992; Aarssen & Jordan 2001), we controlled for variation in plant size when seeking evidence for this trade-off.

We found a strong trade-off among populations between MISM and seed number per fruit independent of variation due to elevation and plant size (Fig. 1h,i). Similarly, when the effect of plant size was statistically controlled, we observed a significant negative relationship between mean individual seed size and mean annual fecundity (Fig. 3b). By contrast, the simple bivariate relationship between MISM and mean fecundity per plant among populations revealed no trade-off because the correlation was inflated (towards a positive value) by variation in plant size (Fig. 3a). Across 15 monocarpic species of herbs, Aarssen & Jordan (2001) similarly found that the trade-off between seed size and seed number was detectable only when the effects of plant size were controlled statistically.

Mean seed number per fruit accounts for a greater proportion of the variance among populations in MISM than either elevation or plant size (Table 1). This result implies that the geographic variation in MISM among Pedicularis populations and taxa cannot be attributed to variation in plant size alone.

The effect of seed mass and elevation on reproductive yield

Among the population means sampled in this study, individual seed mass is not positively correlated with high reproductive yield (independent of plant size) in spite of the fact that individual seed mass is a strong determinant of reproductive yield. The absence of a relationship between mean seed size and total reproductive yield per plant reflects the strong trade-off between size-adjusted fecundity and size-adjusted MISM (Fig. 3b). Populations of plants with higher MISM than expected based on plant size (i.e. large residuals of seed mass on plant size) produce fewer seeds per plant than expected (Fig. 3b). As a result, among populations, mean individual seed size is independent of reproductive yield (Fig. 3d, controlling for plant size). If individual fitness is more closely correlated with reproductive yield than with either MISM or fecundity alone, then mean seed size is relatively neutral with respect to selection.

We detected no variation in reproductive yield across elevations when controlling for plant size (Fig. 4b); per unit of vegetative mass, high-elevation populations had the same mean reproductive yield as low-elevation populations. Although high-elevation sites may have shorter or cooler growing seasons that limit biomass accumulation (resulting in smaller plants) relative to low-elevation sites, we found no evidence that extrinsic conditions at high elevation environments cause less efficient allocation of vegetative resources to reproduction (through either selection or plastic responses).

Conclusion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

In Pedicularis, elevation, plant size and seeds per fruit are all correlated with individual seed mass when examining geographic variation among congeneric populations. However, elevation appears to play a less direct role than intrinsic factors in determining the MISM of a population, as the negative effect of elevation on seed mass disappears when the effects of intrinsic factors (plant size, seed number per fruit) are statistically controlled. Therefore, the observed geographic variation in seed mass cannot be ascribed confidently and directly to adaptation of seed development rates in response to cooler environments, shorter growing seasons, or other stressful conditions associated with elevation. When assessing evidence for adaptive geographic variation in seed size within and among plant species, variation in intrinsic factors such plant size and seed number per fruit must be taken into account to help prevent the confounding of these factors. Our results call into question the conclusions of studies that have detected geographic variation in MISM without controlling for variation in intrinsic factors. Finally, we detected no evidence that reproductive yield (annual fecundity × MISM), which may be most closely associated with fitness, was associated with MISM or affected by elevation, independently of plant size.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

We thank two anonymous referees and editors for useful comments and suggestions; Leah Dudley and Yanjiang Luo for helpful discussions on the manuscript; Longchong Zhang, Zhongling Yang, Guohua Wu and Juan Sun for assistance in both the field and lab; and Zhigang Zhao, Shiting Zhang, ManTang Wang, Xianhui Zhou and Kechang Niu for useful suggestions on earlier work. We also thank UCSB for support and accommodations during the preparation of the manuscript. This study was supported by Key Project of the Chinese National Science Foundation (40930533).

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

Appendix S1. Information of species, population, location, elevation sample size and population means of reproductive traits.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
JEC_1688_sm_appendixS1.doc88KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.