Sources and consequences of seed mass variation in Banksia marginata (Proteaceae)

Authors

  • Glenda Vaughton,

    1. School of Botany, La Trobe University, Bundoora, Victoria 3083, Australia; and
    Search for more papers by this author
    • *Present address: Division of Botany, University of New England, Armidale, NSW 2351, Australia.

  • Mike Ramsey

    1. Division of Botany and Zoology, The Australian National University, Australian Capital Territory 0200, Australia
    Search for more papers by this author

Abstract

1 We examined the sources and consequences of seed mass variation in Banksia marginata occurring in fire-prone heath on nutrient-poor soils to determine factors influencing seed size and possible fitness benefits of large seeds.

2 Individual seed mass varied fivefold. Variation occurred among populations (29% of total), among years (10%) and among plants (plants, 6%; year × plants, 13%), but was most pronounced within plants (42%). Within plants, seed mass variation was greater within infructescences (35%) than among infructescences (7%).

3 Seed mass variation within infructescences was not related to whether follicles contained one or two seeds. Seed mass was also unaffected by ovule position within follicles and follicle position within infructescences.

4 Seed mass variation among infructescences and plants was related to the limited availability of nutrient resources during seed provisioning. Mean seed mass was negatively related to seed number per infructescence and per plant. When resources decreased late in the flowering season and after defoliation, seed mass declined by 7–10% and seed number by 31–45%. When resources increased after inflorescence removal, seed mass increased by 8% but seed number was unaffected. Plants thus had only a limited capacity to maintain seed mass by adjusting seed number when resources varied.

5 The N and P contents of seeds (mg seed−1) increased linearly with seed mass, indicating costs of producing larger seeds in terms of limited environmental nutrients.

6 Seedling size increased with seed mass, implying fitness benefits of larger seeds in terms of increased seedling establishment on nutrient-poor soils. Seed mass had little or no effect on seed germination, relative growth rates and root:shoot ratios.

7 Although stabilizing selection should eliminate seed mass variation occurring within plants, such variation persists because resource constraints limit the ability of plants to control individual seed size.

Introduction

Seed mass is considered one of the least plastic of plant characters (Harper et al. 1970; Fenner 1985). Nevertheless, seed mass variation is commonly observed in many plant species and is often pronounced within individual plants (Thompson 1984; Wolf et al. 1986; Wulff 1986a; Winn 1991; Obeso 1993; Stöcklin & Favre 1994; Vaughton & Ramsey 1997). Such variation is at odds with theoretical considerations that predict that plants should provision all seeds equally, and should alter seed number rather than seed mass if resource levels vary. The optimal seed mass is that which maximizes the return per unit investment; seeds smaller than the optimum will have low fitness, whereas those greater than the optimum will waste resources that could otherwise be used to provision more offspring (Smith & Fretwell 1974; Lloyd 1987; Haig & Westoby 1988).

Seed mass variation within plants is also inconsistent, with empirical evidence of strong directional selection favouring large seeds. The advantage of large seeds lies in their ability to provide energy and nutrients necessary for successful seedling establishment. Positive relationships between seed mass and the probability and speed of germination, and between seed mass and subsequent seedling size and survival, have been demonstrated in a range of species (Stanton 1984; Wulff 1986b; Winn 1988; Tripathi & Khan 1990; Moegenburg 1996; but see Mazer 1987). Because mortality at the germination and seedling stage is often high, variation in seed mass is likely to have strong effects on plant fitness.

One explanation for seed mass variation is that resources for seed provisioning are limited and plants are not able to provision all seeds to the optimum level. Evidence for resource constraints comes from studies demonstrating trade-offs between seed number and mass (Wolf et al. 1986; Lalonde & Roitberg 1989; Wolfe 1995), and studies showing that seed mass is influenced by experimental manipulation of resource availability (Wulff 1986a; Aˇgren 1989; Wolfe 1995; Vaughton & Ramsey 1997). Seasonal declines in seed mass and variation with position within fruits or plants have also been attributed to resource constraints (Wulff 1986a; Winn 1991; Vaughton & Ramsey 1997), although developmental constraints may also be important (McGinley et al. 1987; Wolfe 1995). Variation in seed mass in response to altered resource levels indicates that, in contrast to theory, plants have only a limited capacity to maintain seed mass by adjusting seed number if resources vary.

In the Proteaceae, selection for the ability to establish on nutrient-poor soils is proposed as an important factor determining trade-offs between seed number and mass (Lamont et al. 1985; Esler et al. 1989; Stock et al. 1989; Stock et al. 1990; Mustart & Cowling 1992). Most Proteaceae are sclerophyllous shrubs that occur in fire-prone vegetation on nutrient-poor soils in southern Australia and South Africa. These plants often produce large seeds with high concentrations of N and P, nutrients that are in short supply in the environment (Kuo et al. 1982; Pate et al. 1986; Stock et al. 1990; Witkowski & Lamont 1996a). Most seedling establishment occurs after fire, and seedlings may utilize seed N and P reserves to complement the high levels of other nutrients that are present in the immediate postfire environment (Stock et al. 1990). Despite the presumed importance of seed mass for seedling establishment in the Proteaceae, however, little information is available on the extent and sources of seed size variation within species, and the consequences of such variation on seed nutrient reserves and seedling growth.

In this study, we examined the sources and consequences of seed mass variation in the shrub Banksia marginata Cav. (Proteaceae). First, we assessed the extent of natural variation in seed mass within and among plants in two populations for 2 years. Secondly, we examined how seed mass varies with seed number within and among plants, seed position within follicles and infructescences, and flowering time. Thirdly, we conducted defoliation and inflorescence removal experiments to determine the capacity of plants to alter seed number, rather than seed mass, in response to changes in resource availability. Finally, we determined the effect of seed mass on the seed nutrient content and seedling development.

Methods

STUDY SITE AND SPECIES

Banksia marginata is widely distributed along the coast and ranges of south-eastern Australia. We examined B. marginata within Gibraltar Range National Park (29°36′S, 152°16′E, 1000 m a.s.l.) in north-eastern New South Wales, where it is at the northern limit of its range (Harden 1991). In this area, populations occur in fire-prone sedge-heath swamps on acidic, granitic soils. Unlike most B. marginata populations, plants at Gibraltar Range are obligate seeders and rely on seeds for recovery after fire. Adult plants are single-stemmed, are killed by fire, and the roots of excavated young plants show no evidence of suckering (M. Ramsey, personal observation). Flowering occurs during autumn and winter (April to August). Inflorescences have up to 1000 flowers that open sequentially over several weeks. Nectar-feeding birds are the major pollinators and plants are self-compatible (M. Ramsey, unpublished data). Inflorescences develop into infructescences with woody fruits (follicles) about 6 months after the completion of flowering. Follicles have either one (20%) or two seeds (80%). Unlike most B. marginata populations (Harden 1991), follicles at this site are strongly serotinous: high temperatures during bushfires cause follicles to open and seeds with membranous wings are released.

Natural variation in seed mass, seed nutrient content and seedling growth was examined using seeds from two similarly sized replicate populations (SC and WT) located about 10 km apart. All other work was conducted in one of the populations (SC). Both populations were last burnt in 1989 and plants were therefore about 7 years old at the time of the study (1996).

NATURAL VARIATION

We assessed natural variation in seed mass by harvesting six infructescences, three each from 1994 and 1995, from each of 10 plants in each population. Follicles were opened by heating with a blow torch and seeds were extracted using tweezers. Fifteen seeds from along the length of each infructescence were weighed individually to the nearest 0.1 mg after removing the wing. Seeds damaged by insects were discarded.

We examined individual variation in seed mass using a partially hierarchical anova (Model II) with year and population as crossed factors, plants nested within populations and infructescences nested in plants. F-ratios for populations and plants within populations could not be constructed in the usual manner because the proper denominator for these tests had expected values that differed from the individual mean squares. Quasi-F ratios were therefore calculated using denominators that were obtained by adding and subtracting certain mean squares to obtain a composite mean square with the required expected value (Winer et al. 1991). Approximate degrees of freedom were also calculated.

SEED NUMBER

We examined whether seed mass differed between follicles that produced one seed or two seeds. From one or two infructescences on each of 15 plants, we extracted seeds from 10–20 one-seeded and two-seeded follicles and weighed 10 seeds from each group. Seed mass from one-seeded and two-seeded follicles was compared using a two-way anova (Model III). To examine trade-offs between seed number and mass at the infructescence level, we harvested three to four cones plant−1 from 25 similar sized plants. The 80 infructescences were chosen to span the natural range in size. The number of follicles on each infructescence was counted and 15 seeds from the mid-portion of each infructescence were weighed. Seed number per infructescence was estimated as the product of the number of follicles per infructescence and the mean number of seeds per follicle (seeds follicle−1 = 1.78; n = 400). The relationship between seed number per infructescence and mean seed mass was assessed by least-squares linear regression.

To examine trade-offs at the plant level, we selected 20 plants that produced between two and 42 infructescences. Five infructescences from each plant, or all infructescences if less than five infructescences were produced, were harvested. The number of follicles on each infructescence was counted and 15 seeds from each were weighed. Seed number per plant was estimated as the product of the number of infructescences per plant, the number of follicles per infructescence and the mean number of seeds per follicle. Seed production per follicle does not differ between plants with many and few infructescences (P > 0.05). The relationship between seed number per plant and mean seed size was assessed by least-squares linear regression.

SEED POSITION

We determined the effect of ovule position within follicles on seed number and mass, by examining follicles from the mid-portion of 20 infructescences from different plants. At flowering each follicle contains two ovules, one above (apical) and one below (basal) a woody separator. For seed number, we counted the number of seeds produced (one or zero) by apical ovules and basal ovules for 10 follicles. For seed mass, we weighed seeds from the apical and basal ovule positions for 10 follicles that each contained two seeds per infructescence. To examine the effect of ovule position on seed number, we used a G-test of independence. The effect of ovule position on seed mass was determined using a partially hierarchical anova (Model III). Position was a crossed factor and follicles were nested within infructescences. Because each follicle contained one apical and one basal seed, there was no replication for follicles within infructescences and the position × follicle interaction could not be tested.

To determine the effect of follicle position within infructescences on seed number and mass, we examined follicles from the apical and basal positions on each of 20 infructescences from different plants. We divided infructescences into thirds and selected follicles from the upper (apical) and lower (basal) thirds. For seed number, we counted the number of seeds produced per follicle (two, one or zero) for 10 follicles from each position. For seed mass, we weighed 10 seeds from each position from each infructescence. To examine the effect of follicle position on seed number, we used a G-test of independence. Seed mass in the different positions was compared using a two-way anova (Model III).

FLOWERING TIME

To determine the effect of flowering time on seed number and mass, we tagged three early and three late flowering inflorescences on each of 30 plants. Peak flowering was about 11 July and early flowering was therefore considered to be before 4 July and late-flowering after 18 July. The size and position of early and late flowering inflorescences within plants was similar. The number of inflorescences setting fruit in each group was determined and compared with a G-test of independence. For a subsample of plants on which all six inflorescences produced fruits (n = 14), the number of seeds per infructescence was counted. From the mid-portion of each infructescence 15 seeds were weighed. The effect of flowering time on seed number per infructescence was assessed with a two-way anova (Model III). We determined the effect of flowering time on seed mass, using a partially hierarchical anova (Model III). Flowering time was a crossed factor and infructescences were nested within plants.

DEFOLIATION

We determined the effect of defoliation on seed number and mass by selecting six inflorescences on each of 33 plants. Inflorescences were of a similar size and stage of development. On each plant, inflorescences were randomly assigned to either a defoliation treatment or an undefoliated control. For the defoliated inflorescences, all subadjacent leaves were removed at the completion of flowering. Leaves did not regrow during the experiment. The number of inflorescences setting fruits in each treatment was determined and compared with a G-test of independence. For a subsample of plants on which all six inflorescences produced fruits (n = 14), the number of seeds per infructescence was determined and 15 seeds were weighed. The effect of defoliation on seed number per infructescence was assessed with a two-way anova (Model III). To determine the effect of defoliation on seed mass, we used a partially hierarchical anova (Model III). Defoliation was a crossed factor and infructescences were nested within plants.

INFLORESCENCE REMOVAL

We paired 20 plants on the basis of size and number of inflorescence buds. One plant from each pair was randomly assigned to a removal treatment and the other to an unmanipulated control. On removal plants, all inflorescences except one were removed prior to flowering. On control plants, one inflorescence bud of similar size and stage of development as that remaining on the paired removal plant was tagged for assessment of seed number and mass. The number of removal and control inflorescences setting fruit was determined and compared with a G-test of independence. For a subsample of plants (n = 16) where both removal and control infructescences were available, seed number per infructescence was determined and 15 seeds were weighed. Seed number on removal and control infructescences was compared with a random block anova, with plants as blocks. Seed mass in the two groups was compared with a two-way anova (Model III).

NUTRIENT ANALYSES

For N and P analyses, we used 30 individually weighed seeds from each population (six seeds from five plants population−1) that spanned the natural size range (total n = 60). It was not possible to assess N and P from the same seeds and different seeds were used for each analysis. For N, we digested seeds with 5 ml H2SO4 (98%) in the presence of a Kjeldahl catalyst tablet in calibrated Kjeldahl tubes. Tubes were heated for 2 h at 390 °C in a programmable heating block. After cooling, total N content of each seed was measured by steam distillation using a Tecator Kjeltec 1035 nitrogen analyser.

For P, we digested seeds with 1.2 ml of HNO3 (69%):HCLO4 (70%) (4:1, v/v) in calibrated borosilicate sample tubes. Tubes were left for 18 h at 18 °C before heating in a heating block that was programmed to start at 45 °C for 3 h with progressive increments over 12 h finishing at 230 °C for 30 min. After cooling, the digests were made up to 5 ml with distilled water, and total P in each seed was determined by inductively coupled plasma-optical emission spectrometry (ICP-OES) at the primary emission wavelength for P of 177.495 nm. A standard curve for each run was created using KH2PO4 in 1% HNO3.

For N and P, we assessed the effect of seed mass on seed nutrient content (mg seed−1) by least-squares linear regression. Preliminary analyses showed that there were no differences between populations (all slopes homogeneous; P > 0.05) and regressions were calculated using pooled data.

SEEDLING DEVELOPMENT

To determine the effect of seed mass on seedling development, we used 30 seeds from each of five plants in each population that spanned the natural size range (total n = 300). Seeds were placed on the soil surface of 64-cell seedling trays (one seed cell−1; cells 52 cm3) containing soil from the study site. Trays were placed in a glasshouse and watered regularly. Each week trays were relocated to avoid position effects. Seeds were inspected daily for 30 days to determine the extent and speed of germination. Seeds were considered germinated when the radicle emerged from the testa. When seedlings were approximately 30 days old the length (L) and width (W) of their fully expanded cotyledons were measured, and the area calculated as L × W. Stem height was measured when seedlings were approximately 30 and 65 days old, as the distance from the soil surface to the cotyledons. Relative growth rate was calculated as RGR = (ln height at 65 days – ln height at 30 days)/35 days. Seedlings were harvested when they were approximately 65 days old. Shoots (cotyledons + leaves + stem) and roots of each were washed, packaged separately in aluminium foil, dried at 80 °C for 7 days, weighed to the nearest 0.1 mg, and root:shoot ratios determined.

We assessed whether germination success was related to seed mass using a G-test of independence. Most seeds germinated, and to obtain reasonable numbers of ungerminated seeds we were restricted to using only three seed mass categories representing small, medium and large seeds from the natural seed distribution (4.0–6.0 mg, 6.1–9.0 mg and 9.1–12.0 mg). To assess relationships between seed mass and other seedling traits, we used least-squares linear regression. Preliminary analyses showed that there were no differences between populations (all slopes homogeneous; P > 0.05) and regressions were calculated using pooled data. Simple regression was used to examine the effect of seed mass (predictor variable) on days to germination, cotyledon size and RGR. Multiple regression was used to examine the effects of seed mass and seedling age (predictor variables) on stem height, root and shoot mass, and root:shoot ratios. Seedling age was omitted from the final model if it was not significant (P > 0.05).

DATA ANALYSES

Numbers of follicles and seeds, days to germination and seed mass were square-root transformed, and cotyledon size, stem height, mass of roots and shoots, and seedling age were ln transformed to improve normality and homogeneity of variances. RGR, root:shoot ratios, and seed nutrient contents and associated seed masses were not transformed. Plants, infructescences and follicles were considered as random factors in anovas. For partially hierarchical anovas, the appropriate F-ratios were determined using the protocol for estimating expected values of mean squares as outlined by Winer et al. (1991). Untransformed means ± SE are presented. Analyses were performed using Minitab 10Xtra (Minitab 1995).

Results

NATURAL VARIATION

Individual seed mass spanned a fivefold range, from 3.0 to 15.9 mg in the two populations over the 2 years (Fig. 1). The distribution of seed masses was significantly skewed to the right (skewness = 0.41 ± 0.06, t = 6.83, P < 0.001, n = 1800), indicating that there were more seeds with a mass less than the mean than would be expected for a normal distribution. The distribution also exhibited significant positive kurtosis (kurtosis = 0.81 ± 0.12, t = 6.75, P < 0.001, n = 1800), indicating that there were more seeds near the mean than expected for a normal distribution.

Figure 1.

Frequency distribution of individual seed mass in Banksia marginata. Seeds were from three infructescences from each of 10 plants in both 1994 and 1995 at the SC and WT study sites within Gibraltar Range National Park (n = 1800 seeds).

Seed mass varied significantly between years and populations, but their interaction was not significant (Table 1; SC 1994, 7.61 ± 0.07 mg; 1995, 6.84 ± 0.05 mg; WT 1994, 9.13 ± 0.08 mg; 1995, 8.19 ± 0.06 mg). There was a significant interaction between year and plants within populations, indicating that not all plants exhibited the same trend in both years. Mean seed mass per plant was significantly positively correlated between years (r = 0.55, n = 20, P < 0.05). Within populations, seed mass did not vary significantly among plants (Table 1). Within plants, significant variation occurred among infructescences (7%), although variation existing within all infructescences was fivefold greater and explained 35% of the total variance. Thus, 42% of the total variation in seed mass occurred within plants.

Table 1.  Model II partially hierarchical anova of the effects of year, population, plant (nested within populations) and infructescence (nested within plants) on seed mass in Banksia marginata. The percentage of variance explained by these factors and their interactions are given. The error term explains the variation in seed mass within infructescences. Data correspond to those presented in Fig. 1
Source of variationd.f.MSFP% variance
  1. †Calculations for quasi F-ratios and associated degrees of freedom follow Winer et al. (1991).

Year19.87255.620.04010.1
Population129.14 35.40†0.00229.1
Plant (population)181.46 1.63†0.1155.8
Infructescence (plant)400.266.850.0006.8
Year × population10.040.060.8130.0
Year × plant (population)180.6717.680.00013.0
Error17200.04

35.1

SEED NUMBER

Seeds from one-seeded follicles weighed 3.7% less than seeds from two-seeded follicles, but this difference was not significant (8.14 ± 0.16 vs. 8.46 ± 0.12; F1,14 = 2.17, P = 0.163). Seed mass varied significantly among plants (F14,270 = 9.37, P = 0.000), and the effect of seed number per follicle on seed mass differed among plants (seed number per follicle × plant interaction, F14,270 = 2.79, P = 0.001).

At the inflorescence level, a weak but significant negative relationship between seed mass and seed number explained 5% of the total variance in seed mass (Fig. 2a). At the whole plant level, a stronger negative relationship between seed mass and seed number explained 34% of the total variance in seed mass (Fig. 2b).

Figure 2.

Relationships between (a) seed number per infructescence and mean seed mass (mg), and (b) seed number per plant and mean seed mass (mg) in Banksia marginata. Least-square linear regressions were: (a) √y = 2.95–0.03√x, F1,78 = 3.86, P = 0.05, R2 = 0.05; and (b) √y = 3.30–0.01√x, F1,18 = 9.35, P < 0.01, R2 = 0.34.

SEED POSITION

Seed number was dependent on follicle position within infructescences (G = 10.41, P < 0.01, d.f. = 2). For apical and basal follicles, 71% and 84% produced two seeds, and 27% and 16% produced one seed, respectively. Only apical follicles had zero seeds (2%). Seed mass did not differ with follicle position but varied significantly among infructescences (Table 2). The position × infructescence interaction was not significant.

Table 2.  Model III two-way anova of the effects of follicle position (apical, basal), infructescence and their interaction on seed mass in Banksia marginata, and a Model III partially hierarchical anova of the effects of ovule position (apical, basal), infructescence, follicle (nested within infructescences) and the position × infructescence interaction on seed mass. For ovule position, there was no replication for follicles within infructescences since follicles contained one apical and one basal seed, and the position × follicle interaction could not be tested. Mean (± SE) seed mass for apical and basal positions for both analyses are also given
Source of variationd.f.MSFP
Follicle position
Position10.060.780.387
Infructescence190.9520.460.000
Position × infructescence190.071.530.071
Error3600.05

Ovule position
Position10.031.280.271
Infructescence190.6115.910.000
Follicle (infructescence)1800.042.600.000
Position × infructescence190.021.340.162
Error3990.01

Means ± SE (mg)Apical positionBasal position
Follicle position7.78 ± 0.127.64 ± 0.12
Ovule position8.59 ± 0.098.51 ± 0.10

Within follicles, seed number was dependent on ovule position (G = 6.24, d.f. = 1, P < 0.05). In the apical and basal positions, 86% and 94% of ovules produced one seed, respectively. Seed mass did not differ with position (Table 2). Seed mass varied significantly among infructescences and among follicles within infructescences. The position × infructescence interaction was not significant (Table 2).

FLOWERING TIME

The number of inflorescences setting fruit was dependent on flowering time (G = 13.95, P < 0.001, d.f. = 1). Only 78% of late-flowering inflorescences set fruit, whereas 97% of early flowering inflorescences set fruit. Flowering time also had significant effects on the number and mass of seeds per infructescence (Table 3). Late-flowering infructescences produced 13% fewer seeds and seeds were 7% lighter than early flowering infructescences. Seed number varied significantly among plants, but the flowering time × plant interaction was not significant (Table 3), indicating that flowering time affected all plants similarly. Seed mass varied significantly among and within plants. Interactions between flowering time and plants, and flowering time and infructescences, were also significant, indicating that factors other than flowering time affected seed mass (Table 3).

Table 3.  Model III two-way anova of the effects of flowering time (early, late), plant and their interaction on seed number per infructescence in Banksia marginata, and a Model III partially hierarchical anova of the effects of flowering time, plant, infructescence (nested within plants) and their interactions on seed mass. Mean (± SE) seed number and seed mass for early and late flowering infructescences are also given
Source of variationd.f.MSFP
Seed number
Flowering time15.368.470.012
Plant139.2011.710.000
Flowering time × plant130.630.810.652
Error560.79

Seed mass
Flowering time12.825.500.036
Plant131.219.850.000
Infructescence (plant)280.175.600.000
Flowering time × plant130.513.440.003
F. time×infructescence (plant)280.154.800.000
Error11760.03

Means ± SEEarlyLate
Seed number per infructescence 48.0±2.8 41.8±2.9
Seed mass (mg) 7.15±0.05 6.64±0.05

DEFOLIATION

The number of inflorescences setting fruit was dependent on defoliation (G = 36.71, P < 0.001, d.f. = 1). Only 75% of the defoliated inflorescences set fruit, whereas 100% of control inflorescences set fruit. Defoliation also had significant effects on the number and mass of seeds per infructescence (Table 4). Defoliated inflorescences produced 27% fewer seeds and seeds were 10% lighter than control infructescences. Seed number varied significantly among plants, but the defoliation × plant interaction was not significant (Table 4), indicating that defoliation affected all plants similarly. Seed mass varied among and within plants. The defoliation × plant interaction was not significant. The defoliation × infructescence interaction was significant, indicating that within plants other factors also influenced seed mass (Table 4).

Table 4.  Model III two-way anova of the effects of defoliation treatment (defoliated, undefoliated), plant and their interaction on seed number per infructescence in Banksia marginata, and a Model III partially hierarchical anova of the effects of defoliation, plant, infructescence (nested within plants) and their interactions on seed mass. Mean (± SE) seed number and mass for defoliated and undefoliated infructescences are also given
Source of variationd.f.MSFP
Seed number
Defoliation124.4737.560.000
Plant135.2111.620.000
Defoliation × plant130.651.450.165
Error560.45

Seed mass
Defoliation16.0218.630.001
Plant135.8919.500.000
Infructescence (plant)280.307.620.000
Defoliation × plant130.320.840.614
Def.×infructescence (plant)280.389.640.000
Error11760.04

Means ± SEDefoliatedUndefoliated
Seed number per infructescence38.1 ± 2.352.1 ± 2.2
Seed mass (mg) 7.20 ± 0.07 7.96 ± 0.08

INFLORESCENCE REMOVAL

The availability of resources for seed provisioning per plant should have been greater in the removal treatment than the control treatment because control plants produced almost seven times more infructescences (6.8 ± 1.2 vs. 1.0 ± 0.0). The number of inflorescences setting fruit was independent of treatment (G = 1.16, P > 0.05, d.f. = 1; control, 85%; removal, 95%). Infructescences on removal plants produced 10% more seeds than those on control plants, although this difference was not statistically significant (Table 5). Seed number per infructescence did not differ among plants. Seeds were significantly 8% heavier in the removal treatment than in the control treatment (Table 5). Seed mass varied among plants and the removal × plant interaction was significant, indicating that factors other than inflorescence removal also affected seed mass.

Table 5.  Random block anova of the effects of inflorescence removal (removal, control) and plant on seed number per infructescence in Banksia marginata, and a Model III two-way anova of the effects of removal, plant and their interaction on seed mass. Mean (± SE) seed number and mass in the removal and control treatments are also given
Source of variationd.f.MSFP
Seed number
Treatment10.831.150.301
Plant151.351.870.118
Error150.72

Seed mass
Treatment11.874.430.053
Plant150.7727.680.000
Treatment × plant150.4215.270.000
Error4480.03

Means ± SE (mg)RemovalControl
Seed number per infructescence49.1 ± 3.444.8 ± 3.3
Seed mass (mg) 8.53 ± 0.09 7.89 ± 0.09

NUTRIENT ANALYSES

The N and P content of seeds increased linearly with seed mass (Fig. 3). Seed mass explained 97% and 81% of the total variance in the N and P content of seeds, respectively. The mean N and P content was 0.82 ± 0.03 mg seed−1 and 0.066 ± 0.002 mg seed−1, respectively. The mean N and P concentration was 103.9 ± 0.7 mg g−1 dry mass and 8.1 ± 0.1 mg g−1 dry mass, respectively.

Figure 3.

Relationships between seed mass (mg) and (a) the N content (mg seed−1) and (b) the P content (mg seed−1) of seeds in Banksia marginata. There were no differences between populations (•, SC; ○, WT; P > 0.05) and data were pooled for analyses. Least-square linear regressions were: (a) y = 0.000 + 0.104x, F1,58 = 2138.09, P < 0.001, R2 = 0.97; and (b) y = 0.005 + 0.007x, F1,58 = 244.47, P < 0.001, R2 = 0.81.

SEEDLING DEVELOPMENT

Of 300 seeds, 292 (97%) germinated. Over 95% of seeds germinated in each of the three seed mass categories and seed viability was independent of seed mass (G = 1.33, P = 0.516, d.f. = 2). Seed mass also had no influence on days to germination and the relative growth rate of seedlings (Table 6). In contrast, cotyledon size, stem height, root mass and shoot mass increased with seed mass. Root:shoot ratios declined with seed mass, although the relationship was weak. Seedling age only affected root and shoot masses (Table 6).

Table 6.  Linear regressions of the effects of seed mass on seedling development in Banksia marginata. Simple linear regressions were used except for stem height, root mass, shoot mass and root:shoot ratios. For these traits the effects of seed mass and seedling age were examined using multiple regressions. If not significant age was excluded from the final regression (P > 0.05). The two populations did not differ and data were pooled for analyses
Traitsd.f.FPR2Regression equation
Days to germination1,2920.570.4500.002
Cotyledon size (mm2)1,288487.910.0000.629ln y = 0.45 √mass + 2.71
Height at c. 60 days (mm)1,28813.190.0000.044ln y = 0.08 √mass + 2.75
Relative growth rate (mg mg−1 day−1)1,2882.280.1320.008
Root mass(mg)2,28756.440.0000.282ln y = 0.47 √mass + 2.59 ln age – 10.60
Shoot mass (mg)2,287393.330.0000.732ln y = 0.65 √mass + 2.76 ln age – 10.10
Root:shoot ratio1,28814.400.0000.048y = 0.03 √mass −0.27

Discussion

Seed mass in B. marginata varied fivefold within two populations over 2 years. Under natural conditions variation in seed mass occurred among populations and years, but the most important source was within plants, which accounted for 42% of the total variation. Within plants, seed mass varied both among infructescences and within infructescences, although the latter was the most substantial. Variation in seed mass among plants was relatively minor (6%) and did not contribute significantly to total variation. However, significant variation among plants was detected in other analyses in the study (all P < 0.001), which indicates genetic and/or microenvironmental differences among plants. In other plant species, variation in seed mass has been documented among populations, among years, and among plants, but as found here variation within plants is often the most substantial (Thompson 1984; Winn 1991; Obeso 1993; Stöcklin & Favre 1994; Vaughton & Ramsey 1997). These findings indicate that seed mass variation is predominantly caused by processes occurring within plants.

Variation in seed mass within B. marginata infructescences was not related to differences in the mass of seeds from one-seeded and two-seeded follicles. The lack of differences in seed mass between follicles indicates that on a per seed basis follicles containing either one or two seeds were equivalent nutrient sinks (cf. Lamont & van Leeuwen 1988). Seed mass variation within infructescences was also not related to either the position of ovules within follicles or the position of follicles within infructescences. Other factors must therefore have been responsible for seed mass variation within infructescences. Seeds are often smaller following self-fertilization than cross-fertilization (Manasse & Stanton 1991; Wolfe 1995; but see Ramsey & Vaughton 1996). Because B. marginata is self-compatible (M. Ramsey, unpublished data), variation in the quality of pollen fertilizing ovules may have contributed to seed mass variation occurring within infructescences, as found in B. spinulosa (Vaughton & Ramsey 1997). In contrast, seed number was affected and fewer seeds were produced by apical compared with basal ovules and follicles. Developmental constraints may have contributed to seed number variation if ovule size or vascular development varies with position within follicles or infructescences (Vaughton & Ramsey 1997). In addition, the effect of follicle position on seed number may have been caused by resource constraints since apical follicles are further from sources of photosynthates and nutrients and may be competitively disadvantaged compared with basal follicles.

Resource constraints that occur during seed provisioning can cause trade-offs between seed number and mass, and contribute to seed mass variation within and among plants (Lalonde & Roitberg 1989). Among fruits within plants, trade-offs between seed number and mass may indicate the inability of plants to provision all seeds equally because resources are limited (Wolf et al. 1986; Lalonde & Roitberg 1989; Mehlman 1993). In B. marginata, a trade-off occurred between mean seed mass and seed number per infructescence, but was relatively weak, implying that this source of variation contributed relatively little to within plant variation in seed mass. A stronger negative relationship occurred between mean seed mass and seed number per plant. Seed production is positively related to plant size in B. marginata (r = 0.84, P < 0.001, n = 20), and the trade-off at the plant level indicates that larger plants with greater resource access could not overcome the demands associated with the production of more seeds. In other species, such trade-offs are often obscured by plant size differences or microsite differences between plants (Venable 1992; Mehlman 1993; Obeso 1993; Lamont et al. 1994; but see Wolfe 1995).

Plants are expected to alter seed number rather than seed mass if resource levels vary (Smith & Fretwell 1974; Lloyd 1987; Haig & Westoby 1988). In B. marginata, however, both seed number and mass varied in response to altered resource levels. In the flowering time and defoliation experiments, total seed production declined by 31–45%, and seed mass declined by 7–10%. Over the flowering season, the availability of resources for seed provisioning probably decreased as plants committed resources to ovules that were fertilized earlier in the season (Lalonde & Roitberg 1989; Winn 1991). Similarly, defoliation would have reduced photosynthate production and depleted carbohydrate and nutrient reserves (Witkowski et al. 1994). In the inflorescence removal experiment, total seed production increased by 10% and seed mass increased by 8%, although only the latter was significant. Inflorescence removal would have eliminated potentially competing carbohydrate and nutrient sinks, thereby increasing resource availability for infructescences on removal plants compared with control plants (Stock et al. 1989). Pollen source effects may have also contributed to increases in seed mass on removal plants, if greater cross-fertilization occurred in this treatment (Vaughton & Ramsey 1997). These findings indicate that seed number was more readily altered than seed mass when resources were decreased, but not when they were increased. Significant changes in seed mass in response to altered resource levels is contrary to theoretical predictions.

The capacity of plants to maintain seed mass when resource levels vary within plants is likely to depend on the extent to which resources can be reallocated among flowers and inflorescences (Vaughton 1993). Plants may be unable to abort seeds in response to resource limitation once seeds are provisioned beyond a certain point. In addition, plants may be unable to recover fully and reallocate resources invested in aborted, immature seeds. Under such conditions it may be beneficial to provision seeds at a reduced level and this would contribute to seed mass variation (Lloyd 1987; Lalonde & Roitberg 1989). When resources are increased, it may be beneficial to increase provisioning to seeds if factors such as spatial constraints or high natural seed set limit increases in seed number. In B. marginata, increases in seed number in response to additional resources would have been limited by the available space on infructescences for more follicles to develop and the high level of natural infructescence production (89.4 ± 1.8%; n = 54). In B. spinulosa, where fewer infructescences are produced (13–50%), seed number increased in response to additional resources but seed mass was not affected (Vaughton 1991).

The N and P content (mg seed−1) of seeds increased linearly with seed mass. The soil availability of these nutrients would constrain seed production and provisioning, and impose a cost on producing larger seeds. The concentration of N and P in B. marginata seeds was in the same range as that reported for other Banksia species (N, 80.7–172.0 mg g−1 dry mass; P, 7.7–16.5 mg g−1 dry mass; Kuo et al. 1982; Groves et al. 1986; Pate et al. 1986; Stock et al. 1989; Witkowski & Lamont 1996b). The production of seeds rich in N and P in the Proteaceae is thought to be important in allowing seedling establishment on nutrient-poor soils (Lamont et al. 1985; Stock et al. 1990). Although seedlings establish in the immediate postfire environment when the availability of N and P is temporarily increased, these nutrients are nevertheless in short supply for seedlings. Thermal volatilization from plant and litter material during a fire decreases the availability of soil N, and low soil mobility of P makes uptake difficult for seedlings that have small underdeveloped root systems. High N and P content in seeds therefore should render seedlings less dependent on external supplies of these nutrients and allow increased growth and survival on nutrient-poor soils. Consistent with this expectation we found a positive relationship between seed mass and seedling size under glasshouse conditions using field-collected soil. In contrast, seed mass had little or no effect on seed germination, RGR or root:shoot ratios of seedlings. Field studies are now needed to determine whether these trends persist under natural conditions, and whether there are benefits of larger seed mass at later life-cycle stages.

The present study of seed mass variation in B. marginata supports the findings of other studies that have examined relationships between seed mass and seedling development in the Proteaceae. In contrast to our within-species approach, these other studies have focused on variation among species with different seed masses. Species with larger and more nutrient-rich seeds produced larger seedlings (Stock et al. 1990; Groom & Lamont 1996), although seed mass had little influence on other aspects of seedling development (Stock et al. 1990). In other plant families, within-species studies have found that seedling size often increases with seed mass, and that such differences in seedling size can affect fitness via the subsequent growth and survival of individuals (Stanton 1984; Wulff 1986b; Mazer 1987; Tripathi & Khan 1990; Moegenburg 1996).

On nutrient-poor soils, large B. marginata seeds should have higher fitness than small seeds. However, directional selection for large seeds would be hindered by N and P limitation and subsequent trade-offs that occur between seed number and mass. Such a selection regime would discriminate against individuals producing either very large or very small seeds, and would be expected to minimize seed mass variation within plants. The persistence of seed mass variation in B. marginata is likely to be related to resource constraints operating during seed provisioning, which limits the ability of plants to control the size of individual seeds.

Acknowledgements

We thank Corinna Orscheg for skilled technical help, Max Bartley and Sunarpi for nutrient analyses, and Stuart Cairns for statistical advice. This research was supported by an Australian Research Council (ARC) grant to GV. During the study GV and MR were recipients of a La Trobe University Research Fellowship and an ARC Postdoctoral Fellowship, respectively.

Received 11 August 1997revision accepted 8 December 1997

Ancillary