Intraspecific seed trait variations and competition: passive or adaptive response?

Authors

  • Cyrille Violle,

    Corresponding author
    1. CNRS, UMR 5175 Centre d’Ecologie Fonctionnelle et Evolutive, 1919, Route de Mende, 34293 Montpellier Cedex 5, France
    Search for more papers by this author
  • Helena Castro,

    1. CNRS, UMR 5175 Centre d’Ecologie Fonctionnelle et Evolutive, 1919, Route de Mende, 34293 Montpellier Cedex 5, France
    2. Centre for Functional Ecology, Department of Botany, University of Coimbra, 3000-456 Coimbra, Portugal
    Search for more papers by this author
  • Jean Richarte,

    1. Montpellier SupAgro, UMR 5175 Centre d’Ecologie Fonctionnelle et Evolutive, place Viala, 34060 Montpellier Cedex 2, France
    Search for more papers by this author
  • Marie-Laure Navas

    1. Montpellier SupAgro, UMR 5175 Centre d’Ecologie Fonctionnelle et Evolutive, place Viala, 34060 Montpellier Cedex 2, France
    Search for more papers by this author

*Correspondence author. E-mail: cyrille.violle@cefe.cnrs.fr

Summary

  • 1The phenotype of offspring depends on the abiotic and biotic environment in which the parents developed. However, the direct effects of competition experienced by parent plants on single-seed traits are poorly documented despite their impact on plant fitness.
  • 2We hypothesize that single-seed traits can differentially respond to the resource deficiencies of parent plants due to competition: seed quality may decrease as seed number does, magnifying the negative effects of competition for offspring (‘passive response’ hypothesis), or increase and then enhance offspring fitness to offset the reduction in offspring number (‘adaptive response’ hypothesis). Here we tested these hypotheses for four single-seed traits. We assessed the sensibility of their responses to changes in competition intensity due to species with different competitive effects and to contrasting soil nitrogen conditions.
  • 3In a common-garden experiment, four single-seed traits related to fitness – seed mass, seed nitrogen concentration (SNC), germinability and the timing of germination – were measured on a phytometer species transplanted in 14 different neighbours grown in monoculture with and without soil nitrogen limitation.
  • 4Under nitrogen-limiting conditions, the responses of SNC and of the timing of germination were passive and mainly related to the effects of neighbours on soil nitrogen availability, as shown by the increase in SNC with N-fixing neighbours. Within-individual seed mass variability decreased with increasing competition intensity, as an adaptive response to counterbalance the reduction in seed production. With nitrogen supplementation, competitors had no detectable effect on single-seed traits despite an overall increase in SNC and germination rate, confirming their nitrogen-dependent passive responses to competition. Germinability did not change among treatments.
  • 5The impact of competition on single-seed traits depends on both phytometer trait identity and resource modulation by neighbours. The passive response of seed chemical composition to competitors may magnify the competitive effects on offspring. By contrast, the adaptive response of seed size variability may offset these competitive effects. As a consequence, experiments looking at the fitness consequences of competition should not only consider the effects on fitness parameters of a target plant but also on the offspring.

Introduction

Competition is one of the main selection factors that drive evolution (Darwin 1859). Therefore, it is essential to understand how the effects of competitive pressures experienced by a plant (the ‘maternal competitive effects’) affect its fitness through the modification of offspring performance. It is well known that a maternal individual's environment affects not only the number but also the size and properties of its offspring (Roach & Wulff 1987; Sultan 1996). The experiments focusing on the effect of competition on reproduction show that the number of seeds per plant decreases with increasing competition (e.g. Weiner 1988), as a consequence of the positive allometric relationship between the number of seeds produced and plant biomass (e.g. Samson & Werk 1986). However, there are very few data documenting the impact of competition on single-seed traits such as seed mass, chemical composition and germination ability, despite major contributions to fitness. (i) Seed mass is an important fitness-related trait. Many studies have shown that seedlings from large-seeded species have higher rates of survival than seedlings from small-seeded species (Westoby et al. 2002; Moles & Westoby 2006). Indeed the larger, better provisioned seedlings associated with large seeds (Fenner & Kitajima 2000; Leishman et al. 2000) are favoured in water- and light-limited situations created by established vegetation (Gross 1984; Leishman 2001; Turnbull et al. 1999) or due to environmental hazards (Westoby, Leishman & Lord 1996). (ii) Similarly, higher seed nitrogen concentration (SNC) results in larger seedlings through a better provisioning (Parrish & Bazzaz 1985), allowing plants to survive under hazards in heterogeneous environments (Venable & Brown 1988; Leishman et al. 2000). (iii) Given that the timing of germination is a heritable trait under selective pressure (Geber & Griffen 2003), a recent meta-analysis (Verdú & Traveset 2005) demonstrated that early emergence has significant benefits to offspring fitness in temperate systems: among perennials, early emergents survive and grow more than later ones, while among annuals, they grow and reproduce more. Hence, in competitive plant communities, species with faster germinating seeds are favoured by positive priority effects. (iv) Seed germinability for annual species without dormancy is an indirect indicator of seed quality because only mature, well-provisioned seeds are able to germinate (Baskin & Baskin 1998).

Competition may strongly modify single-seed traits within species. Indeed, first, independently of the presence of competitors, the numerous experimental modifications of resources and microclimatic conditions have created significant and often huge maternal effects on seeds (e.g. temperature: Lacey 1996; light: Sultan 1996; photoperiod: Gutterman 1992; nutrient availability: Parrish & Bazzaz 1985; soil moisture conditions: Luzuriaga, Escudero & Pérez-Garcia 2006). Second, all these environmental factors, and especially resources, are affected by competition, with different magnitudes depending on the identity of competitors (e.g. Tilman 1982). As stated by Sultan (1996), if the effects of parental environment on seed traits simply mirror the resource deficiencies of the parent, plants under competition will produce fewer seeds as well as seeds with lower quality (e.g. small seeds, seeds with low nitrogen content) (‘passive response’ hypothesis). However, if parental plants respond to resource-poor environments by provisioning so as to maintain or enhance offspring quality (and hence the probability of successful establishment), the reduction in fitness due to decreased offspring number under competition may be partly offset (‘adaptive response’ hypothesis). In the few experimental studies analyzing the impact of competition on seed production within species (e.g. Brainard, Bellinde & DiTommaso 2005; Violle, Richarte & Navas 2006), competition reduced the number of seeds produced by an individual without clear change in seed mass, a trait known to be rather conservative (Harper, Lovell & Moore 1970). However, seed mass variation is often more pronounced within individuals than between individuals (e.g. Thompson 1984; Wulff 1986; Vaughton & Ramsey 1998; Susko & Lovett-Doust 2000), because an individual can provision all its seeds differentially, depending on the species and environmental conditions (Roach & Wulff 1987). Therefore we advocate that competition may have little impact on mean seed mass but may modify within-individual seed mass variability in relation to resource depletion. Under the adaptive hypothesis, this variability should be reduced by competition: individuals submitted to a high level of competition should provision the few produced seeds equally and maximize all seed masses so as to produce seedlings with higher competitive ability to face future competitive environments. Similarly, we assumed that variations in seed chemical composition and germination ability under competition should follow the adaptive hypothesis: if a high level of competition is a signal of greater competition in the future, it would be adaptive for mother plants to ‘equip’ the offspring with traits that increase establishment success and competitive ability (e.g. higher nutrient content, faster germinating seeds).

The first aim of this study was to test whether the response of seed traits to competition is explained by the ‘passive’ vs. ‘adaptive’ hypothesis. We studied the impact of competition suffered by mother plants on intraspecific variability in four single-seed traits that have high implications for offspring success: seed mass (mean- and within-individual variability), SNC, germinability and the timing of germination. The second aim was to test if the magnitude of passive or adaptive responses of single-seed traits changes when competition intensity varies: (i) among competitors with contrasting competitive effects, and (ii) within competitors grown under contrasting resource levels.

We performed a common-garden experiment in which an annual phytometer species without seed dormancy, Bromus madritensis (Fig. 1), was grown within 14 different herbaceous species (neighbours hereafter) grown in monoculture and displaying a wide range of standing biomass so as to mimic a large range of competition intensity (defined as the depressing effect of neighbours on phytometer biomass: cf. Fig. 2a). These neighbours created a gradient of environmental conditions under their cover (Violle, Lecoeur & Navas 2007). A ‘passive’ response of single-seed traits is assumed to simply follow changes in resource levels modulated by neighbours while an ‘adaptive’ response is assumed to improve seed quality with increasing competition intensity so as to counterbalance the negative effects on seed number. We set up the experiment under two contrasting nitrogen addition levels to assess the changes in single-seed traits when a soil resource (here nitrogen) becomes non-limiting. Competition is generally assumed to be more intense under high nutrient levels in relation to the increase in above-ground competition due to larger standing biomass (see e.g. Grime 1973). On the one hand, if the response of single-seed traits to competition is similar to that of parent plants, seed quality should be negatively affected by nitrogen addition. On the other hand, if the response of single-seed traits mostly depends on soil resource availability for seed provisioning, maternal competitive effects should not be detected in non-limiting conditions for soil nitrogen.

Figure 1.

An individual of the phytometer species, Bromus madritensis.

Figure 2.

Intraspecific variations in mother plant biomass (phytometer biomass) at reproductive stage (a) and seed traits (b–f) in response to the presence of the 14 studied neighbours in the N– treatment. Mean (panel b) and coefficient of variation (CV) (panel c) are presented for seed mass values (see text for more details). F- and P-values of anovas testing the effect of neighbour identity are indicated for all variables. NS: corresponds to non-significant differences. Bars with different letters correspond to significantly different values of traits among neighbours (P < 0·05). The dashed line indicates the trait value for phytometers grown without neighbour. The black bars indicate the neighbours whose effect is significantly different from the effect of ‘no-neighbour’ treatment. Neighbours are ordered by decreasing competitive effects, based on ranking of phytometer biomass (cf. panel a).

Methods

the competition experiment

Fourteen neighbour species (Table 1), selected from dominant species of a Mediterranean succession (Garnier et al. 2004), were grown in monoculture in a common-garden experiment. The experiment was conducted in the experimental field of the Centre d’Ecologie Fonctionnelle et Evolutive located in Montpellier, France (43°59′ N, 43°51′ E). The mean annual temperature was 14·9 °C and the total annual precipitation was 607 mm. The soil was a clay loam soil, with low organic matter (2·5%). At the beginning of the experiment, in October 2003, total C and N concentrations were 1·5% and 0·14%, respectively, and the pH was 7·8. The choice of herbaceous species belonging to different stages of a secondary succession allows the simulation of a large range of species competitive effects (Violle et al. 2007). Monocultures (1·20 × 1·20 m plots) were replicated four times at two levels of nitrogen supply, randomly located in the common garden. They were established in Fall 2003 by transplantation of seedlings or ramets in order to ensure a standard plant density (100 plants m−2). In Fall 2004, annuals and biennials were re-transplanted in the same plots, with the same spatial arrangement. Plots were regularly weeded by hand. Nitrogen supply was 250 kg ha−1 year−1 in the fertilized treatment (N+ level) with three applications per year. No fertilization was added in the N– level. The N– level was found to be growth-limiting, whereas the N+ level was non-limiting for plant growth (Kazakou et al. 2007).

Table 1.  List of the neighbours used and their characteristics. Life cycle: A, annual; B, biannual; P, perennial
NeighbourBotanical familyLife cycle
Arenaria serpyllifoliaCaryophyllaceaeA
Calamentha nepetaLabiataeP
Crepis foetidaCompositaeA
Daucus carotaUmbelliferaeB
Geranium rotundifoliumGeraniaceaeA
Inula conyzaCompositaeP
Medicago minimaFabaceaeA
Picris hieracioidesCompositaeB
Psoralea bituminosa FabaceaeP
Rubia peregrinaRubiaceaeP
Teucrium chamaedrysLabiataeP
Tordylium maximumUmbelliferaeB
Trifolium angustifoliumFabaceaeA
Veronica persicaScrophulariaceaeA

One individual of the annual grass Bromus madritensis, used as a phytometer, was introduced at the seedling stage in each four-replicated plot when annuals were re-established in November 2004, and in four-replicated weeded plots as control plots without competition (‘no neighbour’ plots hereafter). Phytometers were harvested in June 2005 when seeds were mature but before dispersal; total above-ground biomass (‘phytometer biomass’ hereafter) was measured after drying at 60 °C for at least 2 days. Seeds produced by each individual were sorted and weighed. The seed number produced by each phytometer was estimated by dividing seed biomass by the mean seed mass (further details to follow) for each individual.

measurement of single-seed traits

Seed mass

As a preliminary test, 100 seeds were selected per nitrogen treatment from a subsample of all seeds produced; they were individually weighed with and without accessory structures. Seed weights with and without accessories were strongly related (R2 = 0·938, P < 0·0001). An ancova, in which nitrogen level was the categorical variable and seed weight without accessory structures was the covariate, showed that nitrogen addition had no effect on this relationship (F = 0·12, P = 0·73). Therefore seeds were further weighed with accessory structures.

Within-species variation in seed mass is often underestimated (Harper 1977; Fenner 1992) because standardized protocols (e.g. Cornelissen et al. 2003) generally propose to weigh lots of a given number of seeds (often 100), rather than individual seeds (Thompson 1984). Here, for each phytometer individual, we weighed individually 100 mature seeds, randomly selected from the set of seeds produced (i.e. 7200 seeds per nitrogen level in total), with a precision balance (precision: 10−4 g). The sampling effort was chosen after using a preliminary sampling (200 seeds randomly selected among phytometers) which allowed us to calculate an initial standard deviation and mean, then to simulate the change in precision of measurement with the number of weighed seeds n:

image

Thus weighing 100 seeds individually gives a precision on individual seed mass < 2% (sampling effort and precision level calculated with α = 0·05: data not shown). Data were further pooled to get a unique value per phytometer plant when analyzing effects of nitrogen and competition treatments on mean seed mass.

Seed nitrogen concentration (SNC)

SNC was determined with an elemental analyser (Carlo Erba 2 Instruments, model EA 1108, Milan, Italy) on five seeds (without accessory structures) per phytometer plant, randomly selected among the pool of 100 seeds.

Germination-related traits

Twenty-five seeds per phytometer were randomly set to germinate in Petri dishes in germinative incubators (22 °C, 12 h day/12 h night). Germination was scored as soon as the radicle emerged, then germinated seeds were removed from the Petri dishes. The survey was performed every day for 2 weeks. The total number of seeds germinated throughout the experiment was recorded to calculate the germinability of each parent plant as the ratio between germinated seeds and the initial number of seeds in the Petri dish. Adjustment curves were fitted (exponential models with three parameters) on time-dynamics of cumulated germination percentages (R2 > 0·998). T50, the time when 50% of seeds germinated in each Petri dish, was extracted from these curves to assess germination rate.

data analysis

Seed mass, germinability, T50 and SNC were averaged per phytometer (four replicates per neighbour and per nitrogen level). The within-individual distribution of seed mass was analyzed by calculating a coefficient of variation per phytometer (CV = standard deviation/mean) using the 100 seeds weighed per phytometer. The effects of neighbour and nitrogen treatments (fixed effects) and their interaction on single-seed traits were tested with two-way anova and ancova with mean seed mass used as a covariate. One-way anova and ‘a posteriori’ Student–Newman–Keuls tests were further used by nitrogen treatment to test for differences in maternal biomass and single-seed traits between neighbours and between neighbours’ life-forms (annuals vs. perennials; biannuals were considered as annuals in the analysis because they flowered the first year). Pearson coefficients were calculated to analyze correlations among single-seed traits, and between them and maternal biomass (as an estimate of the competition intensity perceived by phytometers) to test the effect of competition intensity on seed traits. Statistical analyses were performed with SAS (version 8, SAS Institute Cary, NC).

Results

Neighbours strongly reduced phytometer biomass in N– (competitive effect), compared to a situation without neighbours, except under the cover of the Fabaceae Trifolium angustifolia (neutral effect) and Medicago minima (facilitative effect) (Fig. 2a). All neighbours significantly negatively affected phytometer biomass in N+ (competitive effect), compared to a situation without neighbours (Fig. 3a). Because total seed biomass per plant, as a surrogate for the number of seeds produced by phytometers, was positively correlated to phytometer biomass independently of nitrogen treatment (R2 = 0·97, P < 0·0001), the variation in seed number among neighbours tightly followed the variation in phytometer biomass (data not shown).

Figure 3.

Intraspecific variations in mother plant biomass (phytometer biomass) at reproductive stage (a) and seed traits (b–f) in response to the presence of the 14 studied neighbours in the N+ treatment. Mean (panel b) and coefficient of variation (CV) (panel c) are presented for seed mass values (see text for more details). F- and P-values of anovas testing the effect of neighbour identity are indicated for all variables. NS: corresponds to non-significant differences. Bars with different letters correspond to significantly different values of traits among neighbours (P < 0·05). The dashed line indicates the trait value for phytometers grown without neighbour. The black bars indicate the neighbours whose effect is significantly different from the effect of ‘no-neighbour’ treatment. Neighbours are ordered by decreasing competitive effects, based on ranking of phytometer biomass (cf. panel a).

Mean seed mass did not change between neighbours, regardless the N treatment (Table 2; Figs 2b and 3b). On the contrary, within-individual seed mass variation (seed mass CV) varied among neighbours, independently of nitrogen addition (Table 2; Figs 2c and 3c). The difference between neighbours was mainly due to differences in phytometer seed number (Fig. 4), partly resulting from a strong competitive effect of Inula conyza in N– (Fig. 2c): the within-individual variability in seed mass significantly increased with the number of seeds produced by the mother plant (Fig. 4). This result could not be caused by a sampling bias due to a wider seed mass distribution for individuals with a larger pool of seeds, because a preliminary test showed that the error on seed mass measurement reaches an asymptote at the chosen sampling effort (100 seeds per individual) (see Methods). It also did not change when mean seed mass was used as a covariate (not shown).

Table 2.  Results of anovas analyzing the effects of nitrogen addition, neighbour identity (‘no-neighbour’ treatment considered as a neighbour) and their interaction on mean single-seed traits and on the coefficient of variation (CV) for seed mass
Source of variationd.f.SSFP
Seed mass (mean)
 Nitrogen 1 0·006  0·180·67
 Neighbour14 0·05  1·40·20
 Nitrogen × Neighbour14 0·05  1·40·15
 Residual90 0·04  
Seed mass (CV)
 Nitrogen 1 0·0002  0·370·55
 Neighbour14 0·004  6·8< 0·0001
 Nitrogen × Neighbour14 0·001  1·50·11
 Residual90 0·0006  
Germinability
 Nitrogen 1 0·0001  0·050·83
 Neighbour14 0·0008  0·710·76
 Nitrogen × Neighbour14 0·0005  0·380·98
 Residual90 0·001  
T50
 Nitrogen 1 2·68207·4< 0·0001
 Neighbour14 0·13 10·3< 0·0001
 Nitrogen × Neighbour14 0·14 10·6< 0·0001
 Residual90 0·013  
SNC
 Nitrogen 119981< 0·0001
 Neighbour14 0·12  5·9< 0·0001
 Nitrogen × Neighbour14 0·13  6·5< 0·0001
 Residual90 0·020  
Figure 4.

Coefficient of variation of seed mass (seed mass CV) as a function of the number of seeds produced by phytometer in N– and N+. Each point represents the mean effects of a neighbour (four replicates) or of ‘no-neighbour’ treatment (four replicates) on phytometer seed mass CV. Dashed and full lines are the regression lines for N– (R2 = 0·78, P < 0·0001) and N+ (R2 = 0·81, P < 0·0001), respectively. Note the log x-scale. The slopes of the regression lines are not significantly different (see Table 2).

Germinability was near to 100% in all plots and did not differ from the ‘no-neighbour’ treatment (Figs 2d and 3d); no effect of neighbour identity, nitrogen treatment and their interaction was recorded for this trait (Table 2). Neighbours differentially affected the timing of germination (T50) between nitrogen treatments (Table 2). In N–, the timing of germination varied among neighbours (Fig. 2e) and, for five of them (Crepis foetida, Calamentha nepeta, I. conyza, Teucrium chamaedrys, Tordylium maximum), the germination is significantly delayed compared to phytometers grown without neighbours. Nitrogen addition had an overall significant effect (Table 2) by accelerating germination (1·55 and 1·25 days in average for T50 in N– and N+, respectively) but the timing of germination did not significantly differ between neighbours, and between neighbours and ‘no-neighbour’ treatments in N+ (Fig. 3e). Results were the same when seed mass was used as a covariate (not shown).

Neighbours differentially affected SNC between nitrogen treatments (Table 2). In N–, SNC significantly varied among neighbours (Fig. 2f). Mother plants grown with the three N-fixing Fabaceae monocultures (M. minima, Psoralea bituminosa and Trifolium Angustifolium) produced seeds with SNC significantly higher than for phytometers grown without neighbours. Phytometers grown with Arenaria serpyllifolia, Geranium rotundifolium and Veronica persica, three annual species with early flowering periods compared to others (C. Violle, unpublished data), produced seeds with SNC values that did not differ from seeds produced by phytometers grown without neighbours. Phytometers grown with other annual and perennial neighbours had seeds with SNC values significantly lower than in ‘no-neighbour’ situation. Nitrogen addition had an overall significant effect (Table 2) by increasing nitrogen concentration in seeds (compare y-axis for Figs 2f and 3f) but SNC did not significantly differ among neighbours, and between neighbours and ‘no-neighbour’ treatment in N+ (Fig. 3f). Results were identical when seed mass was used as a covariate (not shown).

Competition intensity, expressed by changes in phytometer biomass between neighbours, had a significant effect on the timing of germination and SNC in N– (Table 3): phytometer seeds germinated later and had lower SNC with increasing competition intensity. This was partly explained by the life-form of neighbours. (i) Seeds germinated earlier with annual, inferior competitor species (1·5 days in average for T50) than with perennial, superior competitor species (1·70 days in average for T50) (F = 7·34, P < 0·001). (ii) Phytometers grown with annual, inferior competitor species produced seeds with slightly higher SNC (19·7 mg g−1 in average) than phytometers grown with perennial, superior competitor species (18·2 mg g−1 in average) (F = 5·45, P = 0·02). Neighbours’ life-form had no effect on other mean single-seed traits. Competition intensity had no effect on mean single-seed traits in N+ (Table 3). Increasing competition intensity, in relation to a reduction in seed number (Fig. 4), strongly reduced within-individual seed mass variability (seed mass CV) in N– (r = –0·90, P < 0·0001), and N+ (r = –0·91, P < 0·0001).

Table 3.  Pearson coefficients of correlations between mean seed traits and maternal phytometer biomass within each nitrogen level (italic characters: N– level, bold characters: N+ level). **P < 0·01; *P < 0·05, NS, no significant. SNC, seed nitrogen concentration
 Maternal plant biomass (log)Seed mass (mean)GerminabilityT50SNC
Maternal plant biomass (log)1–0·42NS0·06NS0·27NS–0·09NS
Seed mass (mean)0·41NS1–0·32NS–0·32NS–0·39NS
Germinability0·01NS0·06NS10·03NS0·04NS
T50–0·64**–0·68**–0·18NS1–0·04NS
SNC0·46*0·69**–0·16NS–0·72**1

In N–, seed mass and SNC were positively correlated while T50 was negatively correlated with both mean seed mass and SNC (Table 3). In N+, no significant relationship was found between mean single-seed traits (Table 3).

Discussion

In this experiment, we imposed a strong competition intensity gradient to the annual phytometer under both limiting and non-limiting nitrogen conditions (40- and 9-fold variation between neighbours, respectively). This design allowed us to test the hypothesis of ‘passive’ vs. ‘adaptive’ response of single-seed traits of the target species to face the reduction in the number of seeds produced due to competition and how these responses varied with competition intensity.

Under limiting nitrogen conditions, competition strongly affected phytometer biomass and seed number, except under the cover of the Fabaceae annuals. However, competition did not affect mean seed mass, a trait poorly variable comparatively to seed number (Harper et al. 1970; Fenner 1985; Sultan 1996). The novelty of our study was to reveal the positive link between the variability in seed mass per phytometer and the number of seeds produced. Competition thus did not act on seed mass average but on seed mass variability, by reducing within-individual seed mass variability. Without competition, a larger variability in seed size increases the ability to establish in a wider range of unknown conditions. When competition is of high intensity (e.g. in presence of I. conyza), provisioning all the seeds equally allows them to maximize mass in response to current and future harsh competitive environments. Indeed, plants with larger seeds will produce larger seedlings, that is, seedlings with higher competitive ability (Gross 1984). Then, even without changes in mean seed mass, within-individual seed mass variability supports the adaptive response hypothesis under competition.

Phytometer SNC significantly differed from ‘no-neighbour’ situation under most neighbours. Interestingly, the direction of variation depends on the identity of neighbours. (i) SNC was significantly enhanced for phytometers grown with N-fixing legume neighbours, compared to phytometers grown without competition. This result suggests an impact of soil nitrogen availability on seed quality. Indeed the higher soil nitrogen availability under legumes is related to the decomposition of tissues with higher nitrogen concentrations (C. Fortunel, unpublished data). The magnitude of this response in presence of legumes may reveal a carry-over effect due to decomposition of tissues produced by legumes grown on the same plots the previous year. (ii) The neighbours which had no effect on phytometer SNC were early-flowering species, that is, species with early resource use. Then they may play the role of ‘no-neighbour’ plots for nitrogen uptake at the period of phytometer seed production and provisioning. (iii) SNC decreased when phytometers were grown with other neighbours, which may be due to significant nitrogen depletion under their cover in relation to plant activity. Altogether, these results suggest that changes in phytometer SNC among neighbours mirrored the availability of soil nitrogen under their cover, in relation to facilitation/competition for this resource, even if direct soil data are not available. As a consequence, we proposed that the variation in SNC in response to maternal competitive effects is in agreement with the passive response hypothesis. This can have important impacts for the fitness of plants submitted to competitive environments as plants with seeds with lower SNC produce seedlings with lower competitive ability (Parrish & Bazzaz 1985; Sultan 1996).

Germinability of target seeds remained unchanged among neighbours, even when compared to the situation without competition. This suggests that, for winter annual grass without dormancy, seeds were well-provisioned under all neighbour treatments. On the contrary, competition delayed the timing of germination under five neighbours. We explained this result by fine – although non-significant – changes in mean seed mass, a trait considered as a ‘cornerstone trait’ to explain maternal effects (Weiner et al. 1997). Indeed, larger seeds, resulting from higher SNC (Vaughton & Ramsey 1998), germinated faster, with subsequent larger seedlings (e.g. Harper et al. 1970; Stanton 1984; Schmid & Dolt 1994). As a consequence, and similarly to SNC variations, changes in the timing of germination may be a passive response due to indirect effects of reduction of soil nitrogen availability on seed chemical composition. This passive response is negative for plant fitness in natural competitive communities because a delayed germination leads to smaller seedlings with lower competitive ability (Verdú & Traveset 2005).

While competition ‘simply’ reduces the number of seeds produced, its impact on single-seed traits is more complex and depends on both identity and resource modulation by neighbours. Indeed, chemical composition and the timing of germination seem to mirror soil resource availabilities (passive response) along the competition intensity gradient and can then be strongly affected by resource competition while the reduction in seed mass variability with increasing competition intensity may partly offset the reduction in seed number (adaptive response) contrary to the well-known prediction of invariability for this trait. Analyzing the maternal competitive effects on seed traits is then important to understand the effects of competition on plant fitness (Sultan 1996) and can reveal complex evolutionary regulation processes. In that context, the findings for N-fixing legumes are particularly interesting. For instance, (i) for M. minima, an annual neighbour, there were facilitative effects which are expressed (and thus multiplied) in all fitness-relevant parameters: maternal seed production, offspring seed quality (nutrient content) and offspring germination timing. (ii) For P. bituminosa, a perennial neighbour, there were high competitive effects on maternal seed production (due to above-ground competition) that were partly offset by facilitative effects on offspring seed quality (higher nutrient content) and offspring germination timing (faster germinating seeds).

With nitrogen addition that makes nitrogen non-limiting for plant growth, the competition intensity gradient remained steep but did not affect mean single-seed traits anymore, probably because variations in seed traits due to neighbour activity were low comparatively to the amount of nitrogen provided by fertilization. These results confirmed our hypothesis that the differential responses of SNC and the timing of germination to neighbours grown without nitrogen supplementation were passive and mainly driven by soil nitrogen availability. Finally, we still found a strong positive relationship between seed number and within-individual seed mass variability in N+, which confirmed the competition intensity-dependence for the adaptive response of this variable, regardless of soil nitrogen availability.

In conclusion, our study suggests that the identification of maternal effects due to competition may be of importance for demographic and population studies. Classically, competition is considered as a density-dependent process which modulates population growth rates (Murray 1994). Here we showed that competition can modulate population dynamics at a fine spatial scale through the change in resource status of mother plants when resources are limiting. Thus, the recognition of this effect in demographical models, by considering species-specific effects of neighbours, should modify predictions of mid- and long-term population dynamics and community structure. More generally, the strong effects of parental nutrition on seed quality and timing of germination should be taken into account when developing hypotheses and models that predict the consequences of environmental effects on ecological and evolutionary processes (Rossiter 1996), because of tight links between seed traits and plant fitness (e.g. Parrish & Bazzaz 1985; Leishman et al. 2000).

Acknowledgements

We thank C. Podeur, C. Roumet, C. Collin and I. Hummel for their technical help. We acknowledge very valuable comments by Katja Tielbörger and one anonymous reviewer. This work was supported by the French National Program PNBC ‘Geotraits’; HC received a PhD grant from FCT (Portuguese Foundation for Science and Technology). This is a publication from the GDR 2574 ‘Utiliterres’ (CNRS, France). The experiment conforms with the law in France.

Ancillary