- Top of page
Many factors influence seed size. Tradeoffs between seed size and seed number (Smith & Fretwell 1974; Eriksson & Jakobsson 1999), latitude (Lord et al. 1997), plant growth form and dispersal mode (Leishman & Westoby 1994) and phylogeny (Lord et al. 1995) are all thought to have had an impact on the evolution of seed size. Whatever the end result for individual species, seed mass within a forest type typically varies across several orders of magnitude (Westoby et al. 1992). Variation in seed size has often been interpreted in terms of per capita seedling survival. Large seeds produce larger, more robust seedlings, which typically cope better than those of smaller seeded species in the face of a variety of hazards, including competition, leaf litter, drought, deep shade and defoliation (see reviews by Foster 1986; Westoby et al. 1996; Leishman et al. 2000).
Three different mechanisms may account for the general relationship between seed size and seedling performance, which have been called the seedling size effect, the metabolic effect and the reserve effect (Westoby et al. 1996; Leishman et al. 2000). Under the seedling size effect, larger seeds produce larger seedlings, which may gain advantages simply through being large. For example, larger seedlings emerge from a greater soil depth than small seedlings (Bond et al. 1999), they are less affected by the smothering effects of leaf litter (Guzmán-Grajales & Walker 1991; Molofsky & Augspurger 1992; Reader 1993), they have deeper roots, thus improving access to soil moisture during drought (Metcalfe & Grubb 1997) and, in open habitats with steep light gradients near ground level, they may compete more effectively for light (Grime & Jeffrey 1965; Leishman & Westoby 1994). The metabolic effect follows on from the oft-reported tendency of larger seeded species to have lower potential RGR (Fenner 1983; Shipley & Peters 1990; Cornelissen et al. 1996; Wright & Westoby 1999; Bloor & Grubb 2003), which could indicate inherently lower metabolic rates. If so, then resources would be consumed more slowly, allowing longer survivorship of seedlings in the face of a variety of hazards.
However, neither effect can explain all situations where seedling performance is positively correlated with seed size (Westoby et al. 1996). For example, the size effect cannot account for greater survivorship of seedlings where light gradients near ground level are negligible (Leishman & Westoby 1994; Osunkoya et al. 1994; Saverimuttu & Westoby 1996), under conditions of nutrient deprivation (Lee & Fenner 1989; Jurado & Westoby 1992) or in simulated grazing experiments (Armstrong & Westoby 1993; Harms & Dalling 1997). Similarly, the metabolic effect could not explain the positive correlation between seed size and cotyledon-stage seedling survival in deep shape found by Saverimuttu & Westoby (1996) because, although RGR was negatively correlated with survivorship, the relationship arose secondarily via a common link with seed mass. Further, Bloor & Grubb (2003) found that although RGR was negatively correlated with seed mass among Australian rain forest seedlings grown in 0.8% sunlight, there was no relationship between RGR and seedling mortality.
A third possibility is that the general correlation between seed size and seedling performance occurs via the correlation between seed mass and the functional morphology of cotyledons (Kitajima 1996). Cotyledons vary from being laminar and primarily photosynthetic in function, to being globoid and predominantly concerned with storage and, in the extreme case where they are completely enclosed in the seed coat, entirely storage in function (e.g. Kitajima 1992; Garwood 1996). Several authors have noted a correlation between seed mass and cotyledon function, such that small seed masses are often associated with photosynthetic-type cotyledons, while storage-type cotyledons occur most frequently among larger seeded species (Ng 1978; Hladik & Miquel 1990; Garwood 1996; Kitajima 1996; also see Wright & Westoby 1999; Wright et al. 2000). The tendency for cotyledon function to shift towards storage with increasing seed mass underpins the ‘reserve effect’ initially proposed by Westoby et al. (1996), and later refined by Kidson & Westoby (2000) and Leishman et al. (2000), respectively, as the ‘large-seed-later-commitment’, and ‘large-seed-slower-deployment’ hypotheses (hereafter, the ‘reserve effect’ for brevity). Here, the seedlings of larger seeded species perform better because they have more mobilizable reserves available to them during times of carbon deficit (e.g. in deep shade below the compensation point) or for interim support while replacing photosynthetic tissue lost through herbivory or mechanical damage. The key concept is that not only do large seeds have absolutely more stored energy reserves than smaller seeds, but that a greater proportion of seed reserves in larger seeded species remains uncommitted during seedling deployment, and is thus held in reserve to provision seedlings that germinate in hazardous environments (Westoby et al. 1996; Kidson & Westoby 2000).
To date, we know of only two demonstrations of the reserve effect. While exploring the relationship between seed mass and etiolation in deep shade, Ganade & Westoby (1999) found that larger seeded species produced cotyledons of greater thickness and mass than expected on an isometric basis, and that cotyledons made an increasingly greater contribution to overall seedling mass as seed mass increased. In an explicit test of the hypothesis, Kidson & Westoby (2000) found that as cotyledon mass increased, cotyledon area and functional seedling mass increased, but at rates less than expected on an isometric basis. From this, they inferred that in larger seeded species, a relatively greater proportion of original seed mass was retained in the cotyledons to act as reserves for the newly expanded seedling.
In both these tests, the basis of support for the reserve effect was the tendency for cotyledon functional morphology to shift from photosynthesis to storage with increasing seed mass. However, there have been no studies to assess whether a reserve effect might operate independently of the correlation between seed mass and shifts in cotyledon function. In this study, we tested for such an effect along a gradient of seed mass for species whose cotyledons have a predominantly storage function (i.e. for species at one end of the range of cotyledon functional morphologies).
- Top of page
Under the reserve effect hypothesis, the seedlings of larger seeded species perform better than those of their smaller seeded counterparts because they have more metabolisable reserves available to them during times of carbon deficit. Central to the hypothesis is that not only do larger seeds have absolutely more stored resources than smaller seeds, but that a greater proportion of those resources should remain uncommitted during seedling deployment, and be held in reserve to provision the shoot and root after germination and seedling establishment in hazardous environments (Westoby et al. 1996; Kidson & Westoby 2000).
There was strong evidence for a cross-species reserve effect in rain forest species with storage-type cotyledons. First, cotyledon mass scaled more than proportionately with seed mass, with cotyledon mass increasing by 12% per order of magnitude of seed mass (Fig. 1a). Secondly, the slope of the scaling relationship between cotyledon and seed mass was significantly greater than that for shoot-plus-root mass and seed mass, indicating that the ratio of reserve (cotyledon) mass to functional seedling (shoot plus root) mass increased with increasing seed size. Both of these key results were repeated in analyses against seedling mass, which are arguably a more appropriate test of the reserve effect hypothesis (Westoby et al. 1996; Kidson & Westoby 2000). Cotyledons comprised 40% of total seedling mass for a seedling 0.02 g, but 84% of one weighing 20 g. The mass of the non-cotyledonous portion of seedlings scaled less than proportionately with seed mass, attributable to less than proportional scaling relationships in both shoots and roots. The slopes of these relationships were not significantly different, indicating no cross-species correlation between shoot-root ratios and seed mass. This is consistent with the results for the seedlings of a different suite of Australian rain forest species grown in very dim conditions (Osunkoya et al. 1994).
Two other studies have found direct evidence for a reserve effect among generally small-seeded (less than 100 mg) and mostly temperate species, in both cross-species and phylogenetic analyses. Like us, Ganade & Westoby (1999) found that cotyledons made a greater contribution to total seedling mass in larger seeded species (their Fig. 3b), and Kidson & Westoby (2000) showed that cotyledon mass increased more than proportionately with increasing seedling mass (their Fig. 2c). In both studies, there was a clear correlation between seed mass and cotyledon function morphology; larger seeded species had thicker cotyledons for their seed mass than expected on an isometric basis, so their finding of a reserve effect was based primarily on the tendency for cotyledon function to shift from photosynthesis to storage with increasing seed size (e.g. Ng 1978; Hladik & Miquel 1990; Garwood 1996; Kitajima 1996; also see Wright & Westoby 1999; Wright et al. 2000). On the contrary, our study demonstrated a reserve effect independent of shifts in cotyledon function morphology. We restricted our study to species with storage cotyledons, and so the mechanism underpinning the reserve effect was based specifically on seed mass-related patterns of the differential allocation of resources during seedling deployment.
An important assumption in our demonstration of a cross-species reserve effect was that seedlings of all species were harvested at a comparable developmental stage. The ideal stage to harvest would have been the point at which the shoot and root were no longer strictly dependent on the seed reserves for energy and nitrogen, such that whatever was left of the cotyledons could have been considered as ‘held in reserve’. The time at which this happens varies between species and between resources within species, but can be identified experimentally (Kitajima 2002). In lieu of such labour-intensive experimentation, we harvested seedlings when their first sets of leaves were fully expanded. This stage of development was sometimes difficult to identify with precision. As such, an alternative explanation for our results is that seedlings were not in fact harvested at a comparable stage, and that larger seeded species might have been harvested ‘too early’, before their shoots and roots had been fully deployed. Two lines of evidence suggest this is highly unlikely. First, if larger seeded species were harvested too early, then it would be expected that the average number of leaves per seedling would be lower in these species than in smaller-seeded species. Within the Lauraceae and Sapindaceae, however, there was no obvious bias towards a lower number of leaves in larger seeded species (Table 1). Secondly, the relationship between total seedling mass and seed mass was isometric. After dispersal the seed-cum-seedling looses mass as a result of respiration, leaching and decay. At some point after germination, total mass begins to increase as the shoot and leaves are deployed and continued losses are offset through photosynthetic gains (e.g. Swanborough & Westoby 1996; Ichie et al. 2001). We harvested during this period, and the isometric relationship between seedling mass and seed mass indicated there was no consistent bias with respect to seed mass in the point at which seedlings were harvested. This argument rests on the assumption that the proportional decline in total mass of the seed-cum-seedling did not differ between small and large-seeded species. There was no such difference in 9 of 13 phylogenetically independent contrasts reported by Swanborough & Westoby (1996).
The reserve effect was not very common within species. Only seven of 22 species tested showed unequivocal evidence of the effect, while an eighth species showed the effect in analyses against seedling mass, but not seed mass. Computationally, the reason for this was the comparatively low degree of correlation, and consequently wide confidence limits about the slope, of the relationships between shoot-plus-root mass and both seed and seedling mass (compare Pearson r-values in Table 3), reducing the likelihood of significant differences between slopes. In the absence of a plausible explanation for these low correlation coefficients, we cannot rule out the possibility that the low incidence of intraspecific reserve effects might be more a methodological artifact than a real phenomenon.
the reserve effect and seedling performance
In their reviews of studies examining the relationship between seed mass and seedling performance, Westoby et al. (1996) and Leishman et al. (2000) both concluded that, with few exceptions, seedlings from larger seeded species perform better than those from smaller seeded species under a variety of hazards in cross-species or phylogenetically constrained analyses. They suggested that the reserve effect might be a general mechanism underpinning this general pattern, because others (the seedling size effect and the metabolic effect) could not account for all situations in which seedling performance was contingent on seed mass.
The reserve effect is too little studied for this idea to be critically examined. In one study, Saverimuttu & Westoby (1996) invoked the reserve effect to explain greater seedling longevity in larger seeded species, but they did not explicitly demonstrate a reserve effect among the species in their study. We have examined the cross-species correlation between a reserve affect and seedling performance, but found no evidence that the phenomenon explained better seedling performance in larger seeded species (Green & Juniper 2004). We grew the seedlings of 15 rain forest species with hypogeal storage cotyledons (in common with this study), clipped off the initial shoot, and sequentially clipped each resprout thereafter (cf. Dalling et al. 1997; Dalling & Harms 1999; Edwards & Gadek 2002). Although seedlings of larger seeded species resprouted significantly more often than those of their smaller seeded counterparts, the total mass invested in resprouts scaled isometrically with seed mass across two orders of magnitude. In other words, despite a reserve effect favouring larger seeded species, they performed no differently to smaller seeded species in terms of their biomass allocation responses to simulated herbivory. Future studies of the reserve effect may show that it underpins better seedling performance in larger seeded species in the face of other hazards, but it might not be the unifying mechanism suggested by Westoby et al. (1996) and Leishman et al. (2000).
One assumption implicit in our herbivory study was that a reserve effect measured in terms of cotyledonary biomass was the equivalent of assessing the effect more directly in terms of resources important to the deploying seedling. However, there is an accumulating body of evidence to indicate that significantly negative correlations, or negative trends, often exist between seed mass and the concentrations of several mineral nutrients, lipids and energy (e.g. Fenner 1983; Choe et al. 1988; Grubb & Coomes 1997; Grubb & Burslem 1998; Grubb et al. 1998; Finkelstein & Grubb 2002; and references therein). These studies raise the possibility that a putative reserve effect based on differential patterns of biomass allocation may, in fact, have little or no functional significance; in the present study, the potential benefits for the seedlings of larger seeded species indicated by their heavier than expected cotyledons could be offset by lower concentrations of key resources.
Future studies of the reserve effect would profit from integrating assessments of seed mass-related patterns of biomass allocation during seedling deployment, with chemical assays for resource concentrations in ungerminated seeds and post-germination cotyledons (or any other putative storage tissues). In this way, patterns of resource use during seedling deployment, and the degree of resource provisioning for the newly expanded seedling, could be quantified across a range of seed masses. Such analyses could include all potential resources that aid seedling performance, but in some species, cotyledons function primarily to provision the seedling with some, rather than all, kinds of resources (e.g. Milberg & Lamont 1997; Lamont & Groom 2002). One of the many challenges in assessing the reserve effect more fully will be to first identify which resources stored in the cotyledons at germination are the most limiting for seedling performance in specific habitats.
evolution of the reserve effect
It is widely accepted that historical selective pressures leading to the evolution of traits may not necessarily be the same as those for which the trait has contemporary utility (Gould & Lewontin 1979; Gould & Vrba 1982; Gould 1997). This, and the fact that the reserve effect (as we have described it here) and its function in present-day seedling performance have yet to be widely demonstrated, urges caution in speculating about the evolutionary origins of the phenomenon. Nevertheless, we suggest that the reserve effect may be the result of selective pressures operating at both ends of the seed mass continuum, to preserve a greater proportion of initial seed mass in the cotyledons of larger seeded species, but also to invest a larger proportion of initial seed mass in the shoots and roots of smaller seeded species. These selection pressures can be thought of as pushing the allometric slope of cotyledon mass vs. seed or seedling mass (Figs 1a and 2) up at the high end of the range of seed masses, and down at the low end.
The benefits conveyed by the reserve effect are finite, and would be expected to confer advantages only in relation to hazards that are temporary (Westoby et al. 1996). Herbivory can be regarded as a temporary hazard, and if the seedlings of larger seeded species are more apparent to herbivores and therefore at greater risk of discovery and damage, then the ‘extra’ reserves held in their cotyledons could offset the realized impacts of these greater risks. As explained above, however, the greater resprouting response exhibited by larger seeded species in a subset of those used in this study was not underpinned by the reserve effect (Green & Juniper 2004). Unlike herbivory, canopy shade would not typically be regarded as a temporary risk; deep shade at ground level is almost ubiquitous in tropical rain forest, and the probability of the light environment improving above any one individual seedling during the time that their storage cotyledons remain functional (typically weeks to months), must be extremely small and just as importantly, independent of seed mass. However, canopy shade can be regarded as being temporary in space. Irregularities in the canopy cause significant spatial variation in photon flux densities near ground level, even in the absence of canopy gaps. Although relatively well-lit microsites are relatively rare, they can have important consequences for seedling survival and growth (Nicotra et al. 1999; Montgomery & Chazdon 2002). Other things being equal, the proportion of seedlings establishing in well-lit microsites should be invariant across the spectrum of seed sizes. However, due to tradeoffs between seed size and number, the absolute number of seedlings establishing in these well-lit microsites will be lower in larger seeded species. Greater seedling survival is one mechanism posited to compensate for lower seed production in larger seeded species (Westoby et al. 2002), and the reserve effect could have evolved as one of a suite of mechanisms promoting greater seedling survival in larger seeded species under deep shade.
Leaf litter may have been an important selective pressure operating on smaller seeded species in the evolution of the reserve effect. Litter casts shade and is a physical barrier to the expansion of shoots and roots, and since they produce shorter, less robust shoots, smaller seeded species are disadvantaged by litter (Guzmán-Grajales & Walker 1991; Gulmon 1992; Molofsky & Augspurger 1992; Reader 1993; Seiwa & Kikuzawa 1996). Therefore, smaller seeded species may have been under a greater selective pressure to invest a greater proportion of their seed reserves to produce relatively larger and more robust seedling shoots, which would increase the probability of elevating their first leaves above the litter, and decrease the likelihood of later being smothered.
Manipulative experiments are needed to show that in the presence of the reserve effect, but not in its absence, the seedlings of larger seeded species perform better than those of smaller seeded species. In this respect, clipping cotyledons (e.g. Armstrong & Westoby 1993; Sonesson 1994; Dalling et al. 1997; Mack 1998; Dalling & Harms 1999; Meiners & Handel 2000) of large but not smaller seeded species may prove a useful approach. The reserve effect is an attractive mechanism to explain the widespread observation that the seedlings of larger seeded species outperform those of their smaller seeded counterparts, but its generality and functional significance, as well as the contribution of other explanations, (the seedling size and metabolic effects) have yet to be demonstrated.