1The seedlings of larger-seeded species typically perform better than those of smaller-seeded species in the face of a variety of hazards. One mechanism that might explain this general pattern is the reserve effect, where larger-seeded species commit a smaller proportion of their initial seed reserves to seedling construction, keeping a greater proportion in reserve to sustain seedlings during times of resource deficit or imposed hazards.
2This study tested two hypotheses: (1) that seedlings of larger-seeded rainforest species can resprout more often, following simulated herbivory, than those of smaller-seeded species; and (2) that any such correlation is consistent with the reserve effect.
3To test these hypotheses we grew the seedlings of 15 rainforest species with hypogeal storage cotyledons in a shade house, excised the initial shoot, and sequentially clipped each resprout. Previous work had demonstrated a reserve effect among these species.
4The first hypothesis was proved true: seedlings of larger-seeded species resprouted significantly more often than those of smaller-seeded species.
5The second hypothesis was proved false: although large-seeded species resprouted more often, the total mass invested in resprouts scaled isometrically with seed mass across species. The reserve effect predicted that total resprout mass should have scaled more than proportionately with seed mass.
6This study adds to a growing body of work demonstrating that the seedlings of larger-seeded species often perform better than those of their smaller-seeded counterparts. However, the results presented here are at odds with the view that the reserve effect might be a common mechanism explaining this pattern.
One advantage of large seed size may be the ability of seedlings to tolerate and recover from herbivory (Foster 1986). The damage caused by herbivores can range from the removal of only a minor proportion of a seedling's photosynthetic surfaces, to complete excision of the shoot and total photosynthetic incapacitation. Those seedlings that can recover from such damage by constructing new shoots are obviously at a selective advantage over those that cannot. Several large-seeded species from both temperate (Sonesson 1994; Anderson & Frost 1996) and tropical regions (Dalling, Harms & Aizprúa 1997; Dalling & Harms 1999; Edwards & Gadek 2002) can resprout after experimental excision of the shoot. Further, there is some evidence that seedlings from larger-seeded species tolerate herbivory better than those of smaller-seeded species (Armstrong & Westoby 1993; Harms & Dalling 1997).
Several mechanistic hypotheses have been suggested to explain better seedling performance in large-seeded species in the face of herbivory and other hazards (Westoby et al. 1996; Leishman et al. 2000). One mechanism in particular, the ‘reserve effect’, has been identified by Westoby et al. (1996) and Leishman et al. (2000) as a possible single, unifying mechanism (also called the large-seed later-commitment hypothesis by Kidson & Westoby 2000, or the large-seed slower-deployment hypothesis by Leishman et al. 2000). Under this hypothesis, smaller-seeded species invest relatively more, and larger-seeded species relatively less, of their initial seed mass in constructing the initial shoot and root during the seed-dependent phase of seedling deployment. The reserve effect can be based either on seed mass-related shifts in cotyledon function and morphology (Hladik & Miquel 1990; Garwood 1996; Kitajima 1996; Ganade & Westoby 1999; Kidson & Westoby 2000), or on seed mass-related shifts in biomass allocation during seedling deployment within a single cotyledon type (Green & Juniper 2004). Other things being equal, both forms of the reserve effect suggest that the shoots and roots of larger-seeded species should be better provisioned with stored resources to deal with externally imposed resource deficits or hazards, compared with those of smaller-seeded species. However, the functional significance of the reserve effect has never been assessed.
In this shade-house study we examined the relationship between seed mass and the capacity of seedlings to recover from simulated herbivory, and whether any correlation might be explained by the reserve effect. We used a suite of Australian rainforest species with seed reserve masses spanning two orders of magnitude. Seedling morphology and cotyledon function are correlated with seed mass (Hladik & Miquel 1990; Garwood 1996; Kitajima 1996) such that small seed mass is often associated with epigeal, photosynthetic-type cotyledons, while hypogeal, storage-type cotyledons occur most frequently among large-seeded species. To avoid confounding any correlation between seed mass and the ability to tolerate herbivory with coincident differences in seedling morphology, we restricted our study to just those species with hypogeal storage cotyledons.
We tested two hypotheses: (1) that, on average, larger-seeded species resprout more often following simulated herbivory (excision of the initial shoot) than smaller-seeded species; and (2) that this response is consistent with the reserve effect. For the latter hypothesis we considered the scaling relationship between seed reserve mass and the total mass invested in constructing resprouts. We have already shown that cotyledon mass at full seedling deployment (presumably the source of raw material for constructing resprouts) scales more than proportionately with initial seed mass in these species (Green & Juniper 2004). For this study we reasoned that, other things being equal, the ‘extra’ seed mass held in reserve by larger-seeded species during seedling deployment should allow them to invest a relatively greater proportion of their initial seed mass in constructing more resprouts. If the reserve effect is the mechanism explaining a positive correlation between seed mass and the number of resprouts, we expected that the total mass invested in resprouts should also scale more than proportionately with initial seed mass. On log–log axes the slope of this relationship should exceed 1.
Fifteen species from four families were included in this study (Table 1), with dry seed reserve masses spanning two orders of magnitude from 262 mg to 25 g. All species have storage-type cotyledons when germinated. Fruits were collected from trees growing in the arboretum at the CSIRO laboratory at Atherton, or from rainforest on the Atherton Tableland, between June and December 1999. Sufficient fruits were collected (from one or two trees per species) to prepare 30–40 diaspores (embryo + cotyledons + testa + endocarp, if present) of each species, by removing them from woody pods or scraping away fleshy exocarps and arils. The diaspores were weighed wet, planted individually into pots containing red basaltic soil typical of the Atherton Tableland (0·2–9·0 l volume depending on the size of the diaspore), and placed in a shade house transmitting ≈0·2% of full sunlight (three layers of 90% neutral-density shade cloth). The pots were watered regularly.
Table 1. Species used in this study. Quoted seed mass is an average derived from diaspores used to generate equations predicting dry seed mass (embryo + cotyledons) from wet diaspore mass (embryo + cotyledons + testa/endocarp; P < 0·001 for all regressions). Taxonomic nomenclature from Hyland et al. (2003)
Dry seed mass (mg)
Castanospermum australe A. Cunn. & C. Fraser ex Hook.
Pongamia pinnata (L.) Pierre
Beilschmiedia obtusifolia (Meisn.) F. Muell
Beilschmiedia recurva B. Hyland
Beilschmiedia volckii B. Hyland
Cryptocarya oblata F. M. Bailey
Endiandra dielsiana Teschner
Endiandra impressicosta C. K. Allen
Endiandra jonesii B. Hyland
Litsea leefeana (F. Muell.) Merr.
Castanospora alphandii (F. Muell.) F. Muell.
Diploglottis smithii S. T. Reynolds
Mischocarpus exangulatus (F. Muell.) Radlk.
Toechima erythrocarpum (F. Muell.) Radlk.
Chrysophyllum sp. (Mt Lewis AKI 1402)
Germinating seedlings were checked frequently, and the shoot was subjected to simulated herbivory just as each seedling's first set of true leaves became fully expanded; at this point we used scissors to remove the shoot immediately above the point where the cotyledons joined the stem. The seedlings were continually monitored, and any regrown, fully expanded shoots were repeatedly harvested until the cotyledons were exhausted and no longer contributed stored reserves to their seedlings. Cotyledon exhaustion was indicated by obvious decay with mould growth or tissue softening, yellowing or blackening of cotyledons, acceleration in the appearance of holes on the cotyledon surface, or a reduction in cotyledon size indicated by wrinkling of the testa. At this point the root and what was left of the cotyledons were also harvested. All plant parts were dried at 70 °C for several days, and weighed.
Following the excision of the initial shoot, some seedlings produced concurrent, multiple resprouts. These were removed from the data set prior to analyses (Dalling & Harms 1999). Further, the harvested parts of some seedlings occasionally showed signs of decay at the time of final harvest and so, depending on the analysis, some of these seedlings were also removed from analyses. Because of variation in germination success, multiple but concurrent resprouts, and decay, the actual number of seedlings included in the analyses varied between species, but was never fewer than five for calculating species means, and never fewer than eight for intraspecific analyses.
To test the hypothesis that larger-seeded species were able to resprout more often than smaller-seeded ones, we plotted the mean number of resprouts for each species against its corresponding mean seed reserve mass (hereafter ‘seed mass’ for brevity). Mean dry seed mass was derived from a second set of 30–40 diaspores for each species, each of which was weighed wet, dried for several days at 70 °C, and reweighed after first peeling away the dried testa and/or endocarp, leaving only the embryo and cotyledons (none of the species in this study had endosperm). The simplest measure of mean dry seed mass for each species would be calculated using these weights. However, we found that for several species the mean wet mass of diaspores that actually germinated in the pots was significantly different from that included in the dried sample. Therefore we used species-specific regression equations relating the dry seed mass to wet diaspore mass to estimate the dry mass of individual seeds that eventually germinated in the pots, and used the mean of these estimated masses to calculate species means for analyses (Green & Juniper 2004). We used standard major axis (SMA) regression (Legendre 2000) for all cross-species analyses. The SMA log–log slopes were compared using routines found in the (s)matr package of Falster et al. (2003).
number of resprouts
The seedlings of larger-seeded species resprouted more often following simulated herbivory than seedlings of smaller-seeded species (Fig. 1a, r = 0·85, n = 15 species, P < 0·001). The mean number of resprouts varied between a minimum of 0·2 per seedling in Beilschmiedia obtusifolia (ranked fifth smallest for seed size of 15 species), to a maximum of 3·9 per seedling in Endiandra jonesii, the largest-seeded species. Although larger-seeded species produced absolutely more resprouts, this increase was not isometric (Fig. 1b); on a log–log plot of number of resprouts vs seed mass, the slope of the SMA regression (0·44) was markedly <1 (lower and upper 95% CI = 0·30 and 0·66, respectively). This indicates a declining capacity to resprout per unit of seed mass with increasing seed size. Using the fitted regression equation from Fig. 1(b), seeds of mass 200, 2000 and 20 000 mg resprouted an average of 0·3, 0·9 and 2·6 times, respectively. This is the equivalent of 1·5, 0·5 and 0·1 resprouts g−1 original seed mass, respectively. For example, seedlings of Litsea leefeana (seed mass 290 mg, ranked fourth smallest species for seed mass) produced the equivalent of 4·4 resprouts g−1 original seed mass, compared with just 0·08 resprouts g−1 original seed mass in Beilschmiedia volckii (seed mass = 18·3 g, ranked second largest).
mass of resprouts
As expected, the seedlings of larger-seeded species invested absolutely more in the total mass of all their resprouts (Fig. 1c; r = 0·93, n = 15 species, P < 0·001), but unlike the number of resprouts, the total mass of resprouts scaled isometrically with seed mass (Fig. 1d; SMA slope (1·20) not significantly different from 1 (lower and upper 95% CI 0·96 and 1·51, respectively). Very little of the initial seed mass was eventually invested in resprouts: this averaged just 5·6 ± 3·6% (mean ± SD) across species, and varied between a minimum of 0·6% (B. obtusifolia) and a maximum of 13·2% (Pongamia pinnata).
In the near absence of photosynthesis, resprouts are (presumably) constructed from resources mobilized from the storage cotyledons remaining after the initial shoot and root have been deployed. It is therefore equally valid to compare the number or mass of resprouts against the mass of cotyledons at the point of full seedling expansion. We could not measure cotyledon mass directly, so we derived instead species-specific equations that predicted log cotyledon mass from log seed mass, using data from Green & Juniper (2004). The predicted cotyledon mass for each species was reasonably accurate; the Pearson r for each equation varied between 0·54 and 0·90 (mean 0·79, n = 15 species).
These analyses were consistent with those above, comparing resprouting against seed mass. Although the mean number of resprouts per species increased with increasing cotyledon mass (Fig. 2a; r = 0·90, n = 15 species, P < 0·001), the relationship was allometric (Fig. 2b) with an SMA slope (0·38) on log–log axes significantly <1 (lower and upper CI = 0·26 and 0·55, respectively). The seedlings of species with large cotyledons invested absolutely more in the total mass of their resprouts (Fig. 2c; r = 0·98, n = 15 species, P < 0·001), but this scaled isometrically with cotyledon mass (Fig. 2d); SMA slope 1·03, not significantly different from 1 (lower and upper 95% CI 0·85 and 1·26, respectively). Again, comparatively little of the cotyledon mass remaining at full seedling expansion was eventually invested in resprouts; this averaged 11·8 ± 6·7% across species, and varied between a minimum of 1·86% (B. obtusifolia) and a maximum of 26·8% (P. pinnata). Even seedlings of E. jonesii, cotyledons of which averaged 20·2 g at the time of full seedling expansion, invested just 2·3 g (11·1%) in eventual resprouts.
The mass of the initial shoot, and first and second resprouts were all correlated with initial seed mass (Fig. 3; r = 0·76, 0·90 and 0·98, respectively, n = 12 species for which five or more seedlings resprouted a second time, P < 0·01). However, the mass of all stages scaled less than proportionately with seed mass (SMA slopeinitial shoot = 0·80, 95% CI = 0·66 and 0·97; SMA slope1stresprout = 0·77, 95% CI = 0·64 and 0·93; SMA slope2ndresprout = 0·88, 95% CI = 0·77 and 1·00). These slopes were not significantly different from one another [(s)matr test statistic = 1·570, P = 0·456].
The mass of resprouts tended to decline with successive resprouting events in most species (Fig. 4; y-intercepts in Fig. 3). The mass of the first resprout was significantly less than that of the initial shoot in 10 of 13 species (paired t-tests, P < 0·05), and among these species varied between <20% of initial shoot mass (B. obtusifolia and Beilschmiedia recurva) and >60% (Endiandra dielsiana and P. pinnata). Although variable, the ratio of the mass of the first resprout to the mass of the initial shoot was not correlated with the mass of cotyledons at full seedling expansion (r = 0·232, n = 13 species, P = 0·44). However, this ratio was correlated with the ratio of cotyledon mass to initial seed mass across all species (Fig. 5; r = 0·65, n = 13 species, P = 0·017), and the relationship was even stronger when only Lauraceae were considered (r = 0·87, n = 7 species, P = 0·011).
seed mass and tolerance of herbivory
In this study at least a few of the seedlings of all species resprouted at least once, and among the smaller-seeded species there was considerable variation in the number of times seedlings could resprout in response to simulated herbivory. As predicted, however, the average per capita number of resprouts was positively correlated with seed reserve mass across two orders of magnitude. It was not uncommon for seedlings of the larger-seeded species, such as Castanospermum australe, E. jonesii and B. volckii, to produce four or more sequential resprouts, and some seedlings resprouted six or seven times. These results are consistent with studies describing the multiple resprouting ability of other very large-seeded species (5 to >100 g fresh mass) with hypogeal storage cotyledons in temperate forests (Hoshizaki, Suzuki & Sasaki 1997); the neotropics (Dalling et al. 1997; Harms & Dalling 1997; Dalling & Harms 1999); and Australia (Edwards & Gadek 2002). Further, our study extends the range of seed masses for which resprouting has been reported down to species with fresh seed weights of around 0·5 g. Given the occurrence of this phenomenon in species whose seed masses vary over more than three orders of magnitude, and its occurrence in plant families as diverse as the Clusiaceae, Fabaceae, Hippocastanaceae, Idiospermaceae, Lauraceae, Lecythidaceae, Sapotaceae and Sapindaceae (Harms & Dalling 1997; Hoshizaki et al. 1997; Edwards & Gadek 2002; this study) plus the Corynocarpaceae (P. Green & P. Juniper, unpublished data), we expect the capacity to resprout among species with storage cotyledons to be much more widespread than reported so far.
The ability of seedlings to survive and recover from herbivory is one of the least explored correlates of cross-species variation in seed mass. In one study (Armstrong & Westoby 1993), 95% of the mostly leaf-like cotyledons of a suite of predominantly small-seeded (<20 mg), temperate Australian species was removed just after the emergence of the epicotyl. Subsequent to clipping, seedling survivorship was significantly greater in the larger-seeded species in 14 of 16 species pairs, consistent with the idea that one advantage of large seed size is a greater ability to tolerate herbivory (Foster 1986; Westoby et al. 1996). However, there was no cross-species relationship between seedling survivorship and seed mass.
The only other cross-species study to consider the relationship between seed mass and the tolerance of seedlings to herbivory was probably confounded by seed mass-related variation in seedling morphology. Like ours, the study by Harms & Dalling (1997) assessed the herbivory hypothesis for large-seeded rainforest species. They grew seedlings of 13 species ranging in seed mass from very small (0·22 g wet mass) to very large (107·6 g wet mass). Of these, only the five largest-seeded species could resprout after the stem was clipped – none of the eight smaller-seeded species resprouted at all and they died soon after. Although this result appeared to support the notion that large seed size was adaptive for tolerating herbivory, Harms & Dalling (1997) noted that their interpretation was potentially confounded by coincident differences in functional seedling morphology – all the resprouting species in their study had hypogeal storage cotyledons, while all but one of the eight non-resprouting species had raised, epigeal cotyledons. In their experiment all seedlings were clipped 1 cm above the soil surface, and we suspect they clipped anatomically different parts of their seedlings – the epicotyls of hypogeal species (leaving the cotyledons intact), which happened to be the largest species in the analysis, and the hypocotyls of epigeal species (thereby removing the cotyledons), which happened to be the smallest. It is anatomically impossible for the seedlings of virtually any species to resprout from a severed hypocotyl, regardless of seed mass or its mode of germination. Examination of more than 2000 Australian dicot species yielded only one with hypocotylar buds, and none with below-ground buds (T. Clifford, unpublished data). In the field, we have never seen epigeal germinating rainforest seedlings resprout after the hypocotyl has been chewed through by grazers, but they can resprout if just the epicotyl is severed, leaving the cotyledons intact. For these reasons we regard the test of the large-seed/greater herbivory tolerance hypothesis of Harms & Dalling (1997) as equivocal. In our study we circumvented these difficulties by clipping all seedlings through the epicotyl, immediately above its point of attachment to the cotyledons.
evolution of greater resprouting capacity in larger-seeded species
The ability of seedlings to resprout has undoubted value for seedling survival, especially in tropical forests where rates of seedling herbivory by large browsing and grazing animals can be high (Osunkoya et al. 1992; Asquith, Wright & Clauss 1997; Green, O'Dowd & Lake 1997), and physical damage to shoots caused by falling canopy debris may be substantial (Clark & Clark 1989; Scariot 2000). In north Queensland rainforests the main contemporary vertebrate seedling herbivores are probably Red-legged Pademelons (Thylogale stigmata) and possibly native rats (Melomys cervinipes, Rattus fuscipes and Uromys cuadiamaculatus), but Chowchillas (Orthonyx spladingii) may also damage seedling shoots as they turn over leaf litter (Theimer & Gehring 1999). Given that a variety of large, ground-dwelling herbivores were present in Australian rainforest at least 25 million years ago (Archer, Hand & Godthelp 1991), it is plausible that herbivory on young seedling shoots has been a primary selective pressure in the evolution of the resprouting phenomenon in general. However, the greater ability of large-seeded species to resprout suggests differential herbivory on large vs small seedlings, but to our knowledge there are no data to support this.
In any case, the evolution of the phenomenon may have been driven less by pressures selecting for a greater resprouting response in large-seeded species, and more by pressures operating on small-seeded species selecting against multiple resprouting. One of these might be the size of the resprouts themselves. Small shoots are especially susceptible to smothering to by leaf litter (Guzmán-Grajales & Walker 1991; Molofsky & Augspurger 1992: Reader 1993), and we found that within species shoot mass tended to decline with each successive resprouting event (Fig. 4). There is probably little advantage in regrowing many ineffective, small shoots, so the best evolutionary response to shoot damage for smaller-seeded species may have been to invest relatively more of their cotyledonary reserves in constructing fewer, relatively large resprouts. The fact that the mass of the first and second resprouts scaled less than proportionately with seed mass (Fig. 3) supports this idea. Declining resprout mass is likely to be less of a problem for large-seeded species (e.g. E. jonesii, B. volckii, C. australe), second and third resprouts of which were comparable in mass to the initial shoot mass of several of the smallest species in this study.
tolerance of herbivory and the reserve effect
The results from this study are at odds with the idea of the reserve effect being a single, unifying mechanism accounting for better seedling performance in large-seeded species (cf. Westoby et al. 1996; Leishman et al. 2000). In another study (Green & Juniper 2004), we showed that the proportion of initial seed mass left as uncommitted reserves in storage cotyledons during seedling deployment scaled more than proportionately with seed mass, varying from 42 to 68% of initial seed mass for notional seed masses of 200 mg and 20 g, respectively (the approximate range of seed masses used in this study). In this study we reasoned that, other things being equal, the ‘extra’ seed mass held in reserve by larger-seeded species during seedling deployment should allow them to invest a relatively greater proportion of their initial seed mass in constructing more resprouts. In other words, if the reserve effect had functional significance for tolerating and recovering from herbivory, the total mass of resprouts should have scaled more than proportionately with seed mass.
This was not the case; across species the total mass of resprouts scaled isometrically with seed mass. Admittedly, the slope of this relationship (1·22) was almost significantly >1 (lower 95% CI 0·96), as predicted by the reserve effect. However, the removal of an outlier from the analysis (B. obtusifolia; Fig. 1d) makes it more certain that the relationship between total resprout mass and seed mass is truly isometric (revised slope = 1·09, 95% CI = 0·88 and 1·34).
One caveat to the preceding discussion is that the potential benefits of the reserve effect for larger-seeded species may have been overstated. There is some evidence of significantly negative correlations, or negative trends, between seed mass and the concentrations of several nutrients, lipids and energy (Fenner 1983; Choe et al. 1988; Grubb & Coombes 1997; Grubb & Burslem 1998; Grubb et al. 1998; Finkelstein & Grubb 2002), which would tend to negate the potential gains for large-seeded species indicated by their heavier-than-expected cotyledons. Even if true, the fact remains that large-seeded species in this study still resprouted more often than their smaller-seeded counterparts in the absence of a truly functional reserve effect. This is still inconsistent with the idea of the reserve effect being a single, unifying mechanism accounting for better seedling performance in larger-seeded species.
If the reserve effect did not explain the greater ability of larger-seeded species to resprout, what did? The evidence suggests the increase was achieved through a trade-off between the number and mass of resprouts – smaller-seeded species tended to make fewer, relatively more expensive (heavy) resprouts, while larger-seeded species tended to make more, relatively cheaper (lighter) resprouts. This trade-off involved the less-than-proportionate scaling relationship between initial shoot mass and seed mass, which was then maintained through successive resprouting events (Fig. 3). By deploying relatively fewer reserves to construct each successive resprout, large-seeded species presumably took longer to exhaust the limited cotyledonary reserves available for resprouting, allowing them to resprout more often.
In this respect it is noteworthy that a cross-species average of just 11·8% of the mass of cotyledons remaining at the time of full seedling expansion was eventually allocated to the construction of new resprouts. This is an underestimate, because a component of total cotyledonary mass must be immobilizable, structural tissue (as evidenced by the existence of ‘exhausted’ cotyledons). However, even after subtracting this component, a cross-species average of just 19·1 ± 9·6% (mean ± SD) of metabolizable cotyledonary reserves was eventually deployed for resprouts. The species used in this study seemed to be over-sized for the total mass they eventually invested in resprouts, much in the same way that manipulative experiments with hypogeal cotyledons (Sonesson 1994; Anderson & Frost 1996; Dalling et al. 1997; Mack 1998; Dalling & Harms 1999; Meiners & Handel 2000) have indicated that the seeds of many species seem to be much larger than necessary to construct a functional seedling. However, over the lifetime of seedlings in this study, an average of 57% of the metabolizable component of original seed mass was eventually used in maintenance respiration, or lost through leaching and decay (P. Green & P. Juniper, unpublished results). Given this, and the construction costs of deploying the initial shoot and root (Green & Juniper 2004), it is perhaps not surprising that a relatively small fraction of the mobilizable resources left in the cotyledons at initial seedling deployment was allocated to recovery from herbivory.
declining resprout mass and‘double bet-hedging’
Like other studies (Dalling et al. 1997; Dalling & Harms 1999; Edwards & Gadek 2002), we found that shoot mass tended to decline with each successive resprouting event. In fact, the mean mass of the first resprout was significantly less than that of the initial shoot in 10 of 13 species, and varied from >60% of initial shoot mass to <20%. This percentage was positively correlated with the ratio of cotyledon mass to initial seed mass, a trend that was quite strong for all species, but even more so for the lauraceous species. We interpret this pattern as a striking case of ‘double bet-hedging’. Long-persistent cotyledon reserves have previously been referred to as bet-hedging structures (Garwood 1996; Hoshizaki et al. 1997), and we argue that the species used in this study hedged their bets against future resource deficits in at least two ways. The first occurred during seedling deployment: larger-seeded species retained a relatively greater proportion of their initial seed reserves in storage cotyledons than did smaller-seeded species (Green & Juniper 2004). The second stage at which seedlings hedged their bets was when deploying the first resprout. Species that expended relatively more of their original seed mass on tissue construction and maintenance respiration during seedling deployment (a low cotyledon : seed ratio in Fig. 5) tended to invest relatively less in the construction of their first resprouts (a low first resprout : initial shoot ratio). The mechanism(s) explaining this previously undemonstrated pattern is/are unknown.
In this study, the seedlings of larger-seeded species resprouted more often in response to simulated herbivory than seedlings of smaller-seeded species, adding to a growing body of evidence demonstrating that, in the face of a variety of hazards, the seedlings of larger-seeded species perform better than those of their smaller-seeded counterparts (Westoby et al. 1996; Leishman et al. 2000). However, despite a greater proportional degree of resource provisioning in larger-seeded species at full seedling deployment (Green & Juniper 2004), the greater resprouting response in larger-seeded species was not dependent on this. The reserve effect may have functional significance under other circumstances, but the results of this study are at odds with the view that the reserve effect is a unifying mechanism explaining the better performance of larger-seeded species (Westoby et al. 1996; Leishman et al. 2000).
We thank P. Grubb, K. Harms, D. Metcalfe, A. Moles and M. Westoby for helpful discussions on the ideas presented in this paper. We also thank the Tropical Forest Research Centre, CSIRO Atherton, Australia, for the use of shade-house facilities. This study was supported by an NSF Long-term Ecological Research grant to J. H. Connell and C. A. Gehring (DEB 98–06310).