Seedling growth response to added nutrients depends on seed size in three woody genera


  • Per Milberg,

    1. School of Environmental Biology, Curtin University of Technology, GPO Box U1987, Perth, W.A. 6001, Australia
    Search for more papers by this author
    • * Present address and correspondence: Department of Biology, Linköping University, S-581 83 Linköping, Sweden (fax 46 13 28 26 11; e-mail

  • María A. Pérez-Fernández,

    1. School of Environmental Biology, Curtin University of Technology, GPO Box U1987, Perth, W.A. 6001, Australia
    Search for more papers by this author
  • Byron B. Lamont

    1. School of Environmental Biology, Curtin University of Technology, GPO Box U1987, Perth, W.A. 6001, Australia
    Search for more papers by this author


1 We tested whether seedlings of small-seeded species were more reliant on soil nutrients than large-seeded ones by growing 21 species from three woody genera (Eucalyptus, Hakea and Banksia) along a gradient of nutrient availability.

2 At very low nutrient availability, larger seeds produced larger seedlings. This was seen especially among the eucalypts, but the difference was eliminated at optimal soil nutrient levels. Hakea species with large seed mass, and all Banksia species, appeared unable to exploit additional soil nutrients for growth, whatever the level supplied.

3 Larger seeds tended to have proportionately higher contents of N, P and K and, under nutrient-poor conditions, supplied more of these to their seedlings, although at a diminishing rate.

4 We suggest that large-seededness could be an adaptation to the high-light, nutrient-impoverished habitats in which these species occur by providing the seedling with the mineral nutrients, rather than carbon-based metabolites, needed for maximizing initial root growth. Reaching reliable moisture before summer (drought avoidance) is an alternative strategy to physiological tolerance of drought.


Large seeds contain more mineral nutrients and carbon-based reserves than small seeds. Young seedlings from large seeds are therefore more likely to tolerate adverse conditions. Larger seeded species, for example, establish better in the shade, under drought, when buried beneath soil or litter and in established vegetation (reviewed by Westoby et al. 1996). Another situation where large seed mass might be an advantage is in nutrient-poor soils (Stock et al. 1990), but this has received little attention. The few experiments conducted have shown that large-seeded species tend to survive longer in the absence of soil nutrients (Jurado & Westoby 1992), and that their early growth may be less affected by the level of soil nutrients (Milberg & Lamont 1997).

In the experiment reported here, we explored the importance of seed mass on the growth of seedlings under a wide range of nutrient availability. We tested the hypothesis that the growth of seedlings from small-seeded species would be more affected by low levels of soil nutrients than would seedlings from large-seeded species. We studied three woody genera: Eucalyptus (Myrtaceae), which shows a very wide range of seed sizes and occurs on a range of soil types, and Banksia and Hakea (Proteaceae), which tend to occur on poorer soils and where seed mass, although still variable, does not include very small seeds. Six to eight species were selected from each genus to cover their variation in seed mass.

Materials and methods

Eight species from the genera Eucalyptus, Hakea and Banksia were selected from among those native to south-western Australia, to represent a range of seed masses. However, two Banksia species were later excluded due to low seed viability and one Hakea species was excluded because it became infected by a fungus that caused almost complete mortality of seedlings (the species included are shown in Table 1). All seeds used originated from wild populations, although some Hakea species were obtained through a commercial seed dealer (Nindethana Seed Service, Albany, Western Australia). Fruits of Eucalyptus and Hakea were air dried to release their seeds, while fruits of Banksia had to be burnt to open their follicles. Average seed mass was established by weighing 25 seeds individually, except for the small-seeded eucalypts for which mass was established by weighing several batches of 10 or 100 seeds. Samples of milled seeds were analysed for N, P and K by an autoanalyser procedure after digestion with sulphuric acid and hydrogen peroxide (Varley 1966). Salicylic acid was added to prevent loss of nitrate-nitrogen.

Table 1.  Seed mass, seed concentrations of N, P and K, and the occurrence and distribution in relation to soil type and rainfall of the species included in the experiment

Seed mass±SD
  1. * Estimated from batches of 10 or 100 seeds.

Eucalyptus rudis Endl. 0.21* in forest and woodland at edge of wetlands; deep humic sand; 350–1200 mm
accedens W. Fitz. 0.79* or co-dominant in woodland; clay-loam to laterite over clay; 450–1200 mm
brachyphylla Gardner 1.04* mallee on granite outcrops; 350 mm
conferruminata Carr & Carr  1.39* in woodland, exposed granite; 600–1000 mm
macrocarpa Hook. 5.0±0.9 in scrub-heath; deep acid sand to sand over laterite; 350–900 mm
tetragona (R.Br.) F. Muell. 6.5±2.0 in scrub-heath; deep acid sand; 300–600 mm
erythrocorys F. Muell. 12.8±3.3 in low woodland; sand over limestone; 300–550 mm
calophylla Lindley 78.9±19.6 in open to tall forest; deep acid sands to laterite over clay; 600–1400 mm
Hakea costata Meissner 6.9±0.91.5 m tall; scattered in low heath to scrub-heath; deep acid sand; 500–600 mm
undulata R.Br. 12.6±1.912.31.70.92 m; thickets in woodland; sandy to gravel clay over laterite; 600–1100 mm
stenocarpa R.Br. 14.1±2.311.61.60.91 m; scattered in scrub-heath to woodland; sand to sandy laterite; 500–1000 mm
recurva Meissner 18.4±3.7 m; scattered in scrub-heath to low woodland; sandy laterite; 300–450 mm
obliqua R.Br. 27.5±5.6 m; scattered in scrub-heath to open low woodland; sand; 300–450 mm
crassifolia Meissner129.3±22.2 m; scrub-heath; sand; 400–600 mm
platysperma Hook.630.2±68.710. m; scrub-heath to open woodland; sand to sandy laterite; 350–500 mm
Banksia prionotes Lindlay 25.0± in woodland; deep sands or shallow sand over limestone; 400–850 mm
hookeriana Meissner 46.5± in scrub-heath; deep acid sands; 450–550 mm
menziesii R.Br. 87.9±11.412.31.20.7Co-dominant in scrub-heath to low woodland; deep acid sands; 350–900 mm
attenuata R.Br.102.7±15.410.81.10.8Co-dominant in scrub-heath to woodland; deep acid sand; 300–900 mm
burdettii E.G. Baker103.4±12.3 dominant in scrub-heath; deep acid yellow sand; 500–700 mm
speciosa R.Br.113.7± in coastal low woodland; deep acid sand; 400–500 mm

Seeds were germinated on moistened filter paper in Petri dishes. Germinants were transferred to black plastic tubes (40 mm diameter, 160 mm long) containing a nutrient-poor sand collected near Perth. The eucalypts, which germinated quickly, were transplanted on a single day. Most of the Hakea and Banksia species, which germinated slowly and over an extended time period, were transferred successively during a 2-week period.

Six nutrient-availability treatments were created by placing 0, 2, 4, 8, 16 or 32 granules of a slow-release fertiliser (Osmocote Mini 2–3 months, Grace-Sierra International, Heerlen, the Netherlands) at 5 mm depth in the sand prior to planting of the germinants (Table 2). There were five replicates per nutrient treatment except in a few cases where there were three or four due to seed shortage and/or low germination.

Table 2.  N, P and K addition (mg per pot) in the various nutrient treatments. The amount of sand used per pot contained 0.4 mg available N and 0.05 mg extractable P
Number of slow-releasefertilizer granules addedNPK
 0 (sand only)
 1 0.720.1 0.4
 2 1.40.2 0.8
 4 2.90.4 1.6
 8 5.80.8 3.2
16121.7 6.4

The tubes were randomized within genera in racks carrying 50 tubes, and placed in an air-conditioned glasshouse in Perth, Western Australia, during summer (December 1996 to March 1997). They were watered daily with rainwater. Daily average maximum temperature±SD during the experiment was 27±3 °C, and the daily minimum temperatures were normally above 20 °C. The glasshouse was covered with green shade cloth that reduced the photon fluence rate in the greenhouse to 25% of that outdoors.

Plants were harvested after 9 weeks growth. The root, cotyledons and the rest of the shoot at harvest were separated and dried at 75–80 °C for 48 h before weighing. Biomass data from the various treatments were used in regression analyses.

In a parallel experiment, seedlings of the 15 species for which seed were readily available were grown without nutrient addition for 2 weeks. Cotyledons from these seedlings were harvested, dried and weighed.

Bulked samples of cotyledons from both the 2- and 9-week harvests were analysed for N, P and K (as described above for seeds) to investigate changes in nutrient content of cotyledons during growth. Some species were removed from the analysis because of inadequate sample sizes.

The relative growth rates (RGR) of the seedlings were calculated for two of the nutrient levels (0 and 16 granules) using the formula:

(Ln W- Ln S)/t(1)

where W = average mass of seedlings including cotyledons at harvest, S = average dry mass of seed, and t = number of days.


The relationships between seed mass and various variables were analysed with linear regressions. Individual regressions were calculated for each genus, in most cases after log-transformation. If the confidence intervals for slope and intercept overlapped, the genera were considered not to differ and the data were lumped and a new regression calculated to produce a P-value for the slope of the overall regression.



The contents of N, P and K per seed closely reflected seed mass (Fig. 1) and there was no indication that the three genera differed in their relationship between nutrient content and seed mass (the 95% confidence intervals for slope and intercept overlapped).

Figure 1.

Total amount of N, P and K in seeds of Eucalyptus, Hakea and Banksia with different seed masses.

The concentration of N differed between genera (anova; F = 26.704(2,17), P < 0.0001), with Banksia and Hakea having 10.9% and 9.9% nitrogen, respectively, while Eucalyptus had only 5.3%. This was also reflected in the overall power function relationship between seed mass and N content (Fig. 1) with the exponent, ±95% confidence interval, being 1.128± 0.073 (i.e. >1). P concentrations, which varied from 1.0 to 1.6%, did not differ between genera (F = 3.481(2,17), P > 0.05). K concentrations, however, did differ (F = 6.805(2,17), P < 0.01), with Hakea and Eucalyptus having higher concentrations in the seed (0.92% and 0.93%, respectively) than Banksia (0.66%). Seed mass showed a linear relationship (i.e. exponent = 1) with both P and K (0.994±0.078 and 0.957±0.052, respectively).


There was substantial mortality in all genera at the highest nutrient level (Fig. 2). Otherwise, mortality was most severe among the eucalypts.

Figure 2.

Percentage survivors of the three genera after 9 weeks growth along a nutrient gradient created by adding granules of slow-release fertilizer.


When grown for 9 weeks without added nutrients, the biomass of seedlings of Eucalyptus and Hakea (but not of Banksia) increased with increasing seed mass (Fig. 3a). Seed mass did not predict biomass of Eucalyptus or Banksia seedlings in the 16-granule treatment but continued to do so for hakeas (Fig. 3b).

Figure 3.

Seedling mass versus seed mass after 9 weeks growth in sand (a) without added nutrients, and (b) containing 16 granules of slow-release fertilizer. Equations refer to the solid lines which indicate significant relationships between seed and seedling mass, and broken lines indicate non-significant relationships (P > 0.05).

Individual species differed in their responses to nutrient availability. In Eucalyptus, the small-seeded species responded strongly while the largest seeded showed almost no response to added nutrients (Fig. 4). In Hakea, several of the small-seeded species showed a positive response to the increasing nutrient gradient, while the biomass of the largest seeded species was unaffected or slightly reduced by extra nutrients (Fig. 4). The Banksia species showed almost negligible responses to the added nutrients (Fig. 4).

Figure 4.

Seedling mass after 9 weeks growth at various nutrient levels created by adding granules of slow-release fertilizer. Seed mass is given in mg under the species name.

The relative growth rate of seedlings grown without extra nutrients varied fourfold (0.014–0.057 day−1) between the different species, but with a clear trend in hakeas and banksias to decrease with increasing seed mass (Table 3). The RGR of eucalypts increased substantially when seedlings were grown with added nutrients, and a trend for RGR to decrease with increasing seed mass was apparent in this genus too (Table 3). Fertilization, however, caused no marked increase in RGR for hakeas and banksias.

Table 3.  Relative growth rate (RGR) of seedlings grown for 9 weeks at two nutrient levels (0 or 16 granules of a slow-release fertilizer)
 RGR (day−1)
SpeciesSeed mass
0 granules16 granules
Eucalyptus rudis 0.210.0570.121
accedens 0.790.0560.092
brachyphylla 1.00.0390.082
conferruminata 1.40.0480.086
macrocarpa 5.00.0220.043
tetragona 6.50.0350.061
erythrocorys 130.0380.055
calophylla 790.0310.037
Hakea costata 6.90.0410.049
undulata 130.0460.050
stenocarpa 140.0420.051
recurva 180.0340.041
obliqua 280.0340.036
Banksia prionotes 250.0390.032
hookeriana 460.0320.033
menziesii 880.0210.019

The highest nutrient level used was supra-optimal for growth of most eucalypts (Fig. 4) and caused considerable mortality in all three genera (Fig. 2). This treatment was therefore excluded from the following regression analyses. When the slope of the regression of log (number of granules added+1) on log (seedling mass) was plotted against seed mass (Fig. 5), there was a clear trend for an increasing response to nutrients with decreasing seed mass among the eucalypts (y = −0.488 log x+0.961; r2 = 0.919; P < 0.001). There was a similar but less clear pattern among the hakeas (y = −0.143 log x+0.309; r2 = 0.581; P < 0.05), while there was no trend apparent among the banksias (y = 0.0885 log x−0.163; r2 = 0.166; P = 0.4222). These regressions suggest that seeds of eucalypts and hakeas with masses larger than 93 and 145 mg, respectively, would not respond to added nutrients, although several hakeas and banksias with seeds less than this (e.g. 13 mg) also showed a lack of response.

Figure 5.

Relationship between seed mass and growth responses to the nutrient gradient expressed as the slopes fitted in Fig. 4 (bars indicate±SE). Solid lines indicate significant relationships (P < 0.05), the broken line a non-significant relationship (P > 0.05).


Analyses of variance revealed that the mass of the cotyledons of Eucalyptus increased between 2 and 9 weeks (F = 73.312(1,53), P < 0.0001), while for Hakea and Banksia cotyledon mass was unaltered (F = 3.797(1,28), P = 0.0614; and F = 3.744(1,23), P = 0.0654, respectively). There were significant interactions between ‘time’ and ‘species’ in Eucalyptus (F = 4.660(7,53), P < 0.001), mainly because the cotyledons in some species did not increase substantially in mass over time (data not shown).

From 2 to 9 weeks growth, the N content of the cotyldeons dropped in all species examined (n = 9; data not shown). For most species P and K contents dropped: exceptions were E. erythrocorys (increase in P and K) and E. macrocarpa (increase in P). The cotyledon mass of these two species increased 64% and 30% over time, respectively.

Seedling mass after 9 weeks growth without additional nutrients correlated well with the amount of N, P and K translocated out of the cotyledons (Fig. 6; regressions were not calculated per genus because of the low number of species included).

Figure 6.

Relationship between nutrients (N, P, K) re-translocated from cotyledons between 2 and 9 weeks, and seedling mass after 9 weeks of growth in sand without nutrient addition. Two cases where P, and one where K, had increased during this time span were excluded.


When grown for 9 weeks without additional soil nutrients, total seedling mass for the 21 species in the three genera studied was a function of seed mass (Fig. 3a). This could have been a consequence of the differing contents of remobilizable mineral nutrients and/or carbon-based reserves in the seeds. Our study showed that, under nutrient-poor conditions, it is the high content of mineral nutrients (Fig. 1) and their re-translocation from the cotyledons during growth (Fig. 6) that is probably enabling large-seeded species to produce large seedlings. This is confirmed by the fact that the seedling mass of the small-seeded species, but not the large-seeded ones, was greatly increased by even small amounts of added nutrients (Fig. 4). The fact that starch storage is negligible in these genera (Buttrose & Lott 1978; Kuo et al. 1982; Pate et al. 1986) and that the species typically grow in high-light environments, are consistent with the main function of large seeds being to promote storage of remobilizable mineral nutrients, rather than to supply energy for growth.

Although seed mass tended to show a linear relationship with content of seed nutrients (Fig. 1; the power function exponent ≈ 1), its relationship with seedling mass was logarithmic (Fig. 3a; exponent <1). This could be explained by a high RGR in the small-seeded species (Table 3; Swanborough & Westoby 1996), meaning that they would pre-empt available cotyledon resources quicker than large-seeded species. It is also possible that the small containers used in this experiment retarded growth among the largest seedlings, e.g. by preventing normal root development or by causing temporary water deficits in these seedlings.

The response to nutrients varied greatly among the species. Most dramatic was the effect on the small-seeded eucalypts (Figs 4 and 5). In contrast, many of the hakeas and all banksias seemed unable to utilize the extra nutrients supplied during early growth, since their seedling mass was unaffected (Fig. 4). This supports previous studies with other Banksia species (Siddiqi et al. 1976; Groves & Keraitis 1976; Barrow 1977), but it remains to be established if this is a consequence of the inherent large-seededness of this genus (i.e. a small range of seed masses available) or of a different physiology from unrelated genera. In a study of six north American tree species grown for 4 months (Latham 1992), the growth of Carya tomentosa (seed mass 5000 mg) was inhibited by added nutrients, while that of Querus rubra (3600 mg) and Castanea dentata (3500 mg) was stimulated. The largest response to added nutrients occurred in the three smallest-seeded species (Fagus grandifolia, Nyssa sylvatica, Liriodendron tulipifera; 280, 140, 30 mg, respectively). This study gives some support to our findings that nutrient response of young seedlings is a function of seed mass. However, the extent may vary between genera (slopes in Fig. 5). It is also worth noting that in the long term, when seedlings of ‘non-responsive’ large-seeded species have exhausted their stored nutrients, they too should become responsive to added nutrients.

At the highest levels of added nutrients, mortality among all three genera was substantial (Fig. 2). Many related Australian heathland species that occur on nutrient-impoverished soils also respond negatively to fertilizers (Specht 1963). It is possible that the large-seeded species in the present study rely exclusively on seed-borne nutrients in their first growing season (Barrow 1977) and that this is an adaptation to conditions where soil nutrient supply is persistently unreliable, as in the heathlands of south-western Australia (Lamont 1994).

A consequence of these results is the prediction that species on nutrient-poor soils should have larger seeds. Indeed, Lee & Fenner (1989) showed that species in the grass genus Chionochloa from poorer soils tended to have larger seeds. However, three other attempts to test this prediction in a local flora, including many different genera, have provided conflicting results. Seed mass was smallest in a nutrient-poor site in a South American rain forest (Grubb & Coomes 1997) while no pattern was detectable between different soil types in another South American rain forest (Hammond & Brown 1995) or in eastern Australia (Westoby et al. 1990). It is important to note that, in the present study, the influence of seed mass on seedling growth decreased with increasing nutrient availability in the soil (Figs 3 and 4), indicating that it is only under extremely nutrient-impoverished situations that seed mass exerts a large influence on seedling performance.

Westoby et al. (1996) suggested that the various benefits of large-seededness could have a single, underlying mechanism: the ‘reserve effect’. Thus, extra metabolic resources in large seeds could remain uncommitted and could then serve to support unpredictable levels of carbon deficits caused by drought, shade or herbivory. Under this scenario, large-seededness would only be beneficial in relation to hazards that are fitful and temporary, and where there is some chance of conditions improving after a while. However, all our study species occur under the Mediterranean climate of south-western Australia, where most seedling mortality occurs during the dry summer months (Lamont & Witkowski 1995; Richards & Lamont 1996). A successful seedling must therefore either be physiologically tolerant to drought (Frazer & Davis 1988; Lamont & Witkowski 1995; Richards & Lamont 1996), or produce a deep taproot reaching reliable moisture before the onset of summer. The roots of some Banksia seedlings can, for example, penetrate almost 2 m in their first season (Enright & Lamont 1992) but, since growth is nutrient-limited in this environment (Lamont 1994), such long roots might only be achievable with seeds that store substantial amounts of nutrients. Hence, in this system, large seed mass may not act as a reserve effect to increase tolerance to unpredictable, temporarily adverse conditions. Instead, it will routinely allow great elongation of the taproot in a non-hazardous and nutrient-impoverished situation that is not likely to change during the first growing season nor to vary between recruitment years. In contrast, smaller-seeded species could survive by being physiologically more tolerant of drought (Lamont & Witkowski 1995), or simply by surviving despite the greater risk of death because of higher seedling production (Richards & Lamont 1996). The co-occurrence of species with very different seed masses in vegetation on nutrient-poor soils (Westoby et al. 1990; Hammond & Brown 1995) is a clear indication that several combinations of seed mass and life-history traits are viable (Richards & Lamont 1996; Rees & Westoby 1997). Thus, we hypothesize that in poor soils subjected to severe seasonal drought and heat, seedlings from larger seeds are more likely to reach stored water (Milberg & Lamont 1997) and survive as ‘drought avoiders’, while those with smaller seeds are more likely to be ‘drought tolerators’.

If mineral nutrient content, rather than carbon content, is the main storage function of large-seeded species from nutrient-poor soils (Milberg & Lamont 1997), it is not surprising that many species from nutrient-impoverished soils have high concentrations of nutrients (Pate et al. 1986; Stock et al. 1990). The high nitrogen concentration in the Banksia and Hakea seeds of this study confirms that seeds of Proteaceae from poor soils in Australia and South Africa tend to have twice the concentration compared with non-Proteaceae (Kuo et al. 1982; Pate et al. 1986; Fenner 1986; Esler et al. 1989; Lee & Fenner 1989; Grubb & Coomes 1997; Vaughton & Ramsey 1998).

In conclusion, we have shown that within two woody families larger seeds contain more N, P and K, much of which is translocated to the young seedling, and produce larger seedlings (at a diminishing rate) in nutrient-impoverished soil. In contrast, smaller seeds produce seedlings that respond to nutrient addition, markedly in the case of Eucalyptus and less so in the case of hakeas, such that they may catch up with seedlings of the larger seeded species given sufficient nutrients, while the only response of seedlings of the largest seeded species may be a negative one. We suggest that seed nutrient content may need to be considered together with seedling drought tolerance in understanding the role of seed size in explaining the distribution of woody species on nutrient-impoverished soils subjected to severe summer drought.


PM and MAPF were supported by grants from the Swedish Council for Forestry and Agricultural Research and the Spanish Council for Education and Culture, respectively. We are grateful to Philip Groom for providing seeds of some of the species and to Anna Milberg for assistance in the field and laboratory.

Received 30 June 1997revision accepted 12 January 1998