INTRODUCTION
Ormosia is a circumtropical genus of legumes with woody pods that dehisce to reveal one to several hard, shiny, red, yellow, black or bicoloured seeds (Knaap-van Meeuwen Reference KNAAP-VAN MEEUWEN1962, Merrill & Chen Reference MERRILL and CHEN1943, Rudd Reference RUDD1965, van der Pijl Reference VAN DER PIJL1982). Although the fruits lack any fleshy pulp that could serve as a disperser reward, they seemingly are dispersed by two distinct groups of birds. Ridley (Reference RIDLEY1930) suggested that the seeds mimic fleshy ornithochorous fruits, deceiving arboreal birds (e.g. tanagers, cotingas, thrushes) into dispersing the seeds without a nutrient reward. In fact, the seeds are dispersed by arboreal birds, albeit at a low rate (Foster & DeLay Reference FOSTER and DELAY1998, van der Pijl Reference VAN DER PIJL1982). Once these seeds are deposited, seed dormancy delays their germination until the beginning of the rainy season when more regular rainfall creates moisture conditions favourable for seedling establishment.
Ridley (Reference RIDLEY1930, p. 430) also suggested that some birds might swallow these very hard seeds to act as stones in the gizzard, an idea investigated further by Peres & van Roosmalen (Reference PERES and VAN ROOSMALEN1996). They proposed that large, terrestrial birds (e.g. tinamous, curassows, trumpeters) consume Ormosia seeds from the ground (Galetti Reference GALETTI, Levey, Silva and Galetti2002, Peres & van Roosmalen Reference PERES and VAN ROOSMALEN1996) and use them as grit for grinding the softer seeds that they eat (cf. Beer & Tidyman Reference BEER and TIDYMAN1942). The surfaces of the Ormosia seeds, which have been found in the guts of several of these bird species (Galetti Reference GALETTI, Levey, Silva and Galetti2002, Peres & van Roosmalen Reference PERES and VAN ROOSMALEN1996), are likely abraded or scarified – seeds that are regurgitated or pass through the guts of arboreal frugivorous birds are abraded to a lesser degree, if at all (cf. Traveset et al. Reference TRAVESET, RODRÍGUEZ-PÉREZ and PÍAS2008). Abraded areas may allow oxygen and water to enter a seed more quickly than does an intact surface, leading to earlier germination (Baskin & Baskin Reference BASKIN and BASKIN1998, Traveset et al. Reference TRAVESET, RODRÍGUEZ-PÉREZ and PÍAS2008). Rapid germination could be advantageous if it were to lead to earlier seedling establishment, first because a seed would be exposed to predators for a shorter period, and second because a seedling with a longer growth period during its first year might be able to overtop smaller, later-establishing competitors and increase its light interception (Poorter & Markesteijn Reference POORTER and MARKESTEIJN2008). Shortening or eliminating the seed-dormancy period could also have negative consequences, however, if it were to shift germination to an earlier time that was less favourable for seedling establishment and thereby reduce plant fitness (Augspurger Reference AUGSPURGER1979).
The relative effectiveness (Schupp Reference SCHUPP1993) of arboreal and terrestrial bird dispersers of Ormosia macrocalyx and O. bopiensis was determined by comparing the production of established seedlings by intact and scarified seeds. Mature exposed seeds of these species appear at least by June in the dry season. Seed dispersal begins when the seeds become available and, depending on crop size, continues to November (Foster & DeLay Reference FOSTER and DELAY1998). The peak availability of ripe fruit overlaps the season of lowest overall fruit availability in the area (May–August; Terborgh Reference TERBORGH1983, Reference TERBORGH and Gentry1990), which may increase the probability of dispersal by arboreal birds. Because fruit availability and seed dispersal begin in the dry season and at least partly overlap the transition from the dry to rainy season, seed dormancy is likely important in delaying germination until rainfall is more regular and plentiful. I hypothesized that if scarified seeds of Ormosia (i.e. dispersed by terrestrial birds) were to germinate with the early, irregular rains of the dry-to-wet-season transition (Foster & DeLay Reference FOSTER and DELAY1998) when the soil is dry and difficult for radicles to penetrate, then (1) desiccation or damage by seed predators and pathogens would prevent many germinated seeds from becoming established seedlings, and (2) intact seeds (dispersed by arboreal birds) would produce more seedlings than would scarified seeds.
To test these hypotheses I (1) determined the degree of dormancy and time of germination in intact and scarified seeds of O. macrocalyx and O. bopiensis, (2) compared seedling development and rates of seedling establishment under natural forest conditions and of O. macrocalyx, only, in a screen house, (3) compared mortality of scarified and intact dispersed seeds and developing seedlings, especially from desiccation, and (4) examined seedling characteristics in both species to determine if seedlings from early germinating seeds were larger and/or more robust than those from later germinating seeds.
METHODS
Study area and rainfall
The study was carried out at the Cocha Cashu Biological Station (11°53′19″S, 71°24'28″W; elevation 337 m) in Manu National Park, Madre de Díos, Peru. The station is located in mature floodplain forest, which is described in Terborgh (Reference TERBORGH1983, Reference TERBORGH and Gentry1990). Rainfall data from 1984 to 2001 are posted on the Cocha Cashu website (http://www.duke.edu/~manu/index.html). Records for most years are incomplete, with varying numbers of complete samples for given months (n = 8–15).
Rainfall averaged about 2400 mm annually, with most rain falling from mid-October to mid-April (Figure 1a). Lengths of rainy intervals (consecutive days with rain) showed little variation through the year, including during the transition period (Figure 1b). Dry intervals, in contrast, were considerably longer and more variable through the transition (Figure 1b). A wet or dry period that overlapped two months was included in the month with the most days in the sequence. The probabilities that a dry period ≥7 d would occur in each of the transition-period months (total dry periods ≥ 7 d in all samples of a particular month/total dry periods for that month) was considerable (Figure 1a).

Figure 1. Selected weather data (1984–2001) from the Cocha Cashu Biological Station, Madre de Díos, Peru. Rainfall = solid circles: upper line = mean monthly maximum, middle line = monthly mean ± SD, lower line = mean monthly minimum; days with rain = open circle: mean number of days with rain mo−1 (a). Weather events: mean number of consecutive days with rain per bout = solid diamonds; mean number of consecutive days without rain per bout = solid circles, upper line = maximum, lower line = mean; probability of a dry period ≥ 7 d occurring in each mo = open circles (b).
Study species
Ormosia macrocalyx Ducke occurs from southern Mexico to Peru and Brazil in moist and wet forests (Dillon Reference DILLON1980, Rudd Reference RUDD1965). Trees reach 50 m tall and produce large crops (15 000 seeds in a peak year). Seeds measure c. 10.1 × 9.5 × 7.5 mm, with a mass of c. 0.5 g (Foster & DeLay Reference FOSTER and DELAY1998). Ormosia bopiensis Pierce ex Macbr. is found in the Amazonian forests of Peru and Bolivia, where trees grow to about 20 m tall (Killeen et al. Reference KILLEEN, GARCIA and BECK1993, Rudd Reference RUDD1965). Its seeds are about the same size (10.3 × 9.7 × 7.6 mm; 0.52 g) as those of O. macrocalyx, but crops are much smaller – a few thousand seeds (Foster & DeLay Reference FOSTER and DELAY1998).
Forest experiments
Seed germination and seedling development, establishment, and mortality of scarified and intact seeds in their normal forest habitat under natural conditions were compared; in the absence of captive birds, artificially scarified seeds were used as surrogates for seeds scarified by terrestrial birds. Fresh seeds were collected from the ground under fruiting trees and randomly assigned to groups. Seeds in one group were lightly abraded with a file near the hilum; those in the other group were left intact. I assumed that the abrasion was equivalent to that occurring in the guts of terrestrial birds, although it is likely that natural abrasion is more extensive (pers. obs.).
Seeds were marked in permanent ink with an S (scarified) or an I (intact). One I and one S seed (chosen randomly) of O. macrocalyx were placed 10 cm apart on top the substrate at 25-m intervals along four forest transects. Fifty pairs of seeds each were ‘dispersed’ on 22 September, 8 and 22 October, 1997, and 28 September 1998, for a total of 200 pairs. Fifty pairs of similarly treated O. bopiensis seeds were set out on 28 September 1998. The status of each seed in all trials was checked every 2 (1997) or 4 (1998) d for the first 2 to 3 mo after dispersal and then at least once each week until mid-June 2000, with one exception. Seeds were not checked from early December 1997 to early June 1998. From June 2000 until the fate (dead or seedling established) of all seeds was determined, seeds were checked at 1–4-wk intervals. A fully developed seedling was defined as a plant with its root firmly in the ground and its first two foliage leaves fully expanded. Plants from the initiation of seed germination to full seedling establishment are referred to as ‘developing seedlings’.
Dates of six events in seedling development were recorded: (1) onset of seed swelling (beginning of germination), (2) germination (emergence of the radicle), (3) entry of the radicle into the ground, (4) emergence of the shoot (epicotyl), (5) full expansion of the first foliage leaves (seedling established), and (6) death. Mortality factors were recorded, with the ultimate rather than proximal cause of death noted, if they differed. Disappearance of seeds was attributed to rodent predation.
The height of each seedling when first established and the maximum width of each fully expanded leaf were measured; height and new leaves were also measured at the end of the first year from the date of dispersal. For seeds germinating more than 1 y after dispersal, the ‘first’ year was calculated from the date of dispersal, but in the year of germination.
Screen-house experiments
Screen-house experiments were used to determine the effects of dry periods (i.e. consecutive days without watering) of different lengths on germination and seedling development in S and I seeds of O. macrocalyx. Wet periods, i.e. watering for two consecutive days, were the same length throughout the experiment, reflecting natural conditions (Figure 1a). Pot locations were rotated within the screen-house weekly so that all seeds were exposed to the same environments.
The screen-house consisted of a wood frame covered with fine-mesh, green fibreglass screen. The roof was covered, in addition, with a layer of clear plastic that permitted light, but not rain, to enter. Temperatures in the house usually exceeded ambient by a few degrees Celsius. Black plastic pots (c. 216 cm3) were filled with a 1:1 mixture of fine river-sand and forest topsoil that had been screened through 1-cm2-mesh hardware cloth. One seed was placed on the soil in each pot. Plants were watered between 07h00 and 09h00. Drainage holes in the pots prevented the sand/silt mixture from becoming saturated or ‘muddy’, but the soil remained moist for several days.
Fifty pots with intact and 50 with scarified seeds were maintained for 56 d under one of six different watering regimes (Figure 2) representative of the variation in rainfall during August to November. Seeds under regime 1 were never watered and were exposed only to ambient soil moisture when the pots were filled. Watering regimes 2–5 were based on the average number of days with rain mo−1 and average lengths of wet periods and intervening dry periods, in days. Seeds under regime 6 were watered every day. The stage of development or death of each seed or seedling was recorded weekly.

Figure 2. Watering regimes used in screen-house experiments carried out for 8 wk to determine the effects of dry periods (i.e. consecutive days without watering) of different lengths on germination and seedling development in scarified and intact seeds of Ormosia macrocalyx. Under regime 1, plants were never watered; under regime 6, plants were watered every day. Plants watered = ///; no water = ——.
Data analyses
After a test for homogeneity of variances (F test), means were tested for differences using either a two-sample Student's t-test (equal variances) or Welch's t-test (unequal variances). The chi-square goodness-of-fit two-sample test and Kolmogorov–Smirnov (K–S) two-sample test were also used. Analyses were done with PAST Version 1.54 (Hammer et al. Reference HAMMER, HARPER and RYAN2001).
RESULTS
Germination and development
Development in Ormosia macrocalyx and O. bopiensis was essentially the same. A few days before germination the seeds began to swell and soften. The seed coat lost its shine and split, and the radical emerged, elongated, and entered the ground. The cotyledons separated, and the epicotyl emerged, producing two opposite foliage leaves. The length of the root and the period before it entered the ground varied with soil moisture. When the soil was dry, roots grew up to 10 cm long before entering the ground but usually were eaten or desiccated well before that. Damaged roots produced multiple secondary roots.
Timing of development
Ormosia macrocalyx. All seeds (both S and I) for which fates are known (341) were either preyed on or began to germinate, so scarification had no apparent effect on the numbers of seeds germinating. It did, however, affect the distribution of germination times of the S and I seeds, which differed significantly (K–S, D83, 138 = 0.69, P < 0.001). S seeds germinated 11.8 ± 7.1 d (mean ± SD; range = 4–52, n = 138) after dispersal, significantly (t = 9.13, P < 0.001) sooner and over a shorter period than I seeds, which germinated 44.1 ± 31.8 d (range = 6–153, n = 83) after dispersal (Figure 3). This difference carried over to all subsequent developmental events, which also occurred significantly earlier for seedlings developing from S seeds than they did for those developing from I seeds. For example, S seedlings were established, on average, 46.6 ± 14.0 d (range = 29–90) after dispersal, significantly earlier (t = 6.31, P < 0.001) than I seedlings at 76.3 ± 32.8 d (range = 31–153). When timing of developmental events was standardized to germination date, S and I seeds did not differ significantly (all P > 0.05), with one exception. Roots from S seeds entered the ground 11.5 ± 13.5 d (range = 1–76, n = 88) after germination, significantly later (t = –3.12, P = 0.0023) than the roots of I seeds at 6.6 ± 4.6 d (range = 1–17, n = 55; Figure 3).

Figure 3. Timing of events in the development of seedlings from scarified and intact seeds of Ormosia macrocalyx: Ger = time of germination in days following dispersal; Rig = radicle in ground, Sht = epicotyl appears, Sdlg = initial two foliage leaves completely expanded, timing of latter three events in days following germination. Bars = mean numbers of days, vertical lines = +1 SD (** = P < 0.01, *** = P < 0.001).
Ormosia bopiensis. Scarification also had no apparent effect on the numbers of O. bopiensis seeds germinating, although differences between S and I seeds in timing of germination and developmental events after dispersal were even greater than in O. macrocalyx, and all were significant (all P < 0.002). These differences reflect the broad variation in the length of the pre-germination dormancy period of the I seeds which germinated an average of 220 ±175 d (range = 37–852, n = 28) after dispersal, significantly (t = 5.79, P < 0.001) later than the S seeds, which germinated an average of 30.8 ± 12.8 d (range = 21–80, n = 45) after dispersal. As with O. macrocalyx, the rate of development of O. bopiensis seedlings standardized to days after germination did not differ significantly between the S and I seeds (all P > 0.05), again with one exception. Seedlings from S seeds were established an average of 48.3 ± 11.1 d (range = 30–79, n = 20) after germination, significantly later (t = 12.4, P = 0.022) than seedlings from I seeds, at an average of 40.3 ± 9.11 d (range = 28–56, n = 15) after germination.
Seedling establishment, size and survival
Ormosia macrocalyx. The fates of all 200 S seeds of O. macrocalyx used in the forest experiment were determined; the fates of only 141 of the 200 I seeds are known because of the hiatus in observations from 9 December 1997 until 7 June 1998. The other 59 I seeds died or disappeared during this period, and I do not know if they first formed seedlings. Fifty-eight (41.1%) of the 141 I seeds produced seedlings, in contrast to only 51 (25.5%) of the 200 S seeds (Figure 4). Differences in the proportions of seedlings and dead plants in the two groups were significant (χ2 = 9.30, P = 0.002, df = 1, n = 141, 200). The number of seedlings remaining in each group (I = 25, S = 13) 1 y after dispersal (i.e. at the beginning of the second dry-to-wet-season transition) also differed significantly (χ2 = 10.5, P = 0.001, df = 1, n = 200, 200; Figure 4).

Figure 4. Numbers and survival of seedlings arising from scarified and intact seeds of Ormosia macrocalyx: Estab sdlg = percentage of seeds producing seedlings; Sdlg surv 1 y = percentage of seeds producing seedlings surviving at 1 y. Properties of seedlings when first established and at 1 y (Ht = height, Lf wid = leaf width); bars = means, vertical lines = +1 SD. (** = P < 0.01, *** = P < 0.001).
In addition, seedlings from I seeds averaged significantly taller (t = 2.98, P < 0.004; mean ± SD = 110 ± 31.6 mm, range = 31–194, n = 57) than the seedlings from S seeds (mean ± SD = 91.8 ± 29.5 mm, range = 34–168, n = 49) when the seedlings were first fully developed (Figure 4). The leaves also tended to be broader in I seedlings than in S seedlings, but not significantly so (t = 1.89, P = 0.062).
Because S seedlings were fully developed an average of 46.6 d and I seedlings an average of 76.3 d after seed dispersal, the growth period of S seedlings averaged 30 d longer than that of I seedlings, during their first year. At the end of that year, however, the average heights of the S and I seedlings did not differ significantly (t = 0.359, P = 0.72). Nine (36%) of the 25 surviving I seedlings added 1 to 3 leaves, and six (46.2%) of the 13 S seedlings added 1–2 leaves between seedling establishment and the end of year one. However, neither the average numbers of leaves per I or S seedling (t = –0.211, P = 0.40), nor the sizes of the additional leaves (t = 0.659, P = 0.52) differed significantly.
Ormosia bopiensis. All of the O. bopiensis S seeds had either become seedlings or died 175 d after dispersal. It was not until day 1063 (nearly 3 y) after dispersal that the fates of all the I seeds were known, which reflects the significant difference (t = 7.20, P < 0.0001) in average day of germination between the S (mean ± SD = 30.8 ± 12.8 d, range = 21–80, n = 45) and I (mean ± SD = 220 ±175 d, range = 37–852, n = 28) seed groups. However, success in establishing seedlings did not differ between the two groups, each of which produced 20 (40%) seedlings. Likewise, 16 seedlings in each group were alive at the end of the year in which they germinated. Average heights, leaf widths, and numbers of leaves on seedlings also did not differ significantly when the seedlings were first established or at the end of 1 y (all P > 0.10).
Mortality
Mortality in O. macrocalyx occurred during all developmental stages from a variety of factors, including desiccation of developing seedlings after their radicles failed to enter the ground (Figure 5). Sample sizes were too small for analysis of mortality factors within developmental stages. However, among plants of all stages for which causes of death were determined, significantly fewer (χ2 = 5.86, P = 0.015, df = 1, n = 81, 144) I plants (n = 3) than S plants (n = 20) succumbed to desiccation. In O. bopiensis, deaths from desiccation did not differ significantly (χ2 = 0.763, P = 0.383, df = 1, n = 29, 30) between I (5 of 29) and S (8 of 30) plants with known causes of death (Figure 5).

Figure 5. Percentage of seedlings developing from intact and scarified seeds of Ormosia macrocalyx and O. bopiensis that died as a result of water stress after their radicles failed to enter the ground following germination (* = P < 0.05).
Screen-house experiments
Seeds that received no water (regime 1) did not germinate, but remained hard, shiny, and presumably viable (Figure 6). Regime 2 was the most irregular (Figure 2) and had the longest dry period (18 d). The lengths of the dry periods decreased in regimes 3 to 6, and the numbers of I seeds developing increased. Mortality in regimes 2 and 3 exceeded 90% but decreased markedly from regimes 4 to 6, in which the numbers of seeds germinating and developing into seedlings showed major increases. Nearly all S seeds germinated (Figure 6). The few that did not were among the drier watering regimes; none contained parasites, and all appeared viable. The number of fully developed seedlings was essentially the same for I and S seeds under regimes 2 through 5 when the experiment ended. However, only 0–5% of the S seeds in those treatments had not yet germinated, whereas 48–86% of the I seeds remained ungerminated, leaving a much greater number of I seeds that could potentially produce seedlings under increasingly wet natural conditions. The experiment was too short to register mortality from desiccation due to radicles not entering the soil.

Figure 6. Percentage distribution of intact and scarified seeds of Ormosia macrocalyx among stages of seedling development after being grown for 8 wk in a screen house under one of six different watering regimes. Regime 1 = no water; days with water increase from regimes 2 to 5; regime 6 = water every day (Figure 2). Seed = hard, shiny, ungerminated seed; Devel = developing seedling; Sdlg = a mature seedling with two fully expanded foliage leaves. For each watering regime, n = 50 intact and 50 scarified seeds.
DISCUSSION
Seed dispersal and seedling establishment
Both the forest and screen-house experiments showed that intact seeds in some Ormosia species – i.e. O. macrocalyx, but not O. bopiensis – produce significantly more seedlings than scarified seeds. These results are consistent with the hypothesis that arboreal birds are more effective seed dispersers than terrestrial birds (Schupp Reference SCHUPP1993), at least of some species of Ormosia. In part, this outcome can be explained by the significantly larger number of developing S seedlings that succumbed to desiccation when their radicles failed to enter the ground than of developing I seedlings. This difference appears to have been mediated through seed dormancy. Dormancy in the I seeds proceeded uninterrupted, allowing them to pass through the dry-to-wet-season transition without germinating and thus to avoid exposure to a period of erratic rainfall interspersed with extended dry spells. The O. macrocalyx S seeds, in contrast, germinated over a span of 48 d, all of which fell within the transition period. Also, the earlier germination and longer growth period did not confer any size advantage on the S seedlings relative to the later-germinating and establishing I seedlings, which were as large and well-developed as the S seedlings. Similar results have been found with other tropical (Blain & Kellman Reference BLAIN and KELLMAN1991) and temperate (Traveset et al. Reference TRAVESET, RODRÍGUEZ-PÉREZ and PÍAS2008) plant species.
Terrestrial birds could compensate for the lower recruitment from scarified seeds by dispersing substantially more seeds than arboreal birds or by depositing them in more favourable microhabitats (Janzen Reference JANZEN1970, Schupp Reference SCHUPP1993). I have no data on microhabitats, but can consider quantity of seeds dispersed. In earlier work we estimated that terrestrial birds dispersed c. 41 times as many seeds of O. macrocalyx as did arboreal birds (Foster & DeLay Reference FOSTER and DELAY1998). A difference of such magnitude would swamp any benefit accruing from I seeds giving rise to 1.6 times as many established seedlings as S seeds, as in this study. However, for our previous estimates we assumed that all seeds that disappeared from under a parent tree had been dispersed by terrestrial birds. We did not consider seed displacement or seed predation, which are both common in Ormosia (MSF unpubl. data). Likewise, extensive research on many plant species has shown that seed mortality is generally greatest (up to 100%) under a parent tree (Augspurger Reference AUGSPURGER1983a, Reference AUGSPURGERb; Clark & Clark Reference CLARK and CLARK1981, Reference CLARK and CLARK1984; Dirzo & Domínguez Reference DIRZO, DOMÍNGUEZ, Estrada and Fleming1986, Janzen Reference JANZEN1970). If our earlier data are re-analysed to account for seed displacement and predation based on the percentages recorded in this study, then terrestrial birds would have dispersed only 2.8 times as many seeds as arboreal birds (MSF unpubl. data). In addition, in more recent monitoring of O. macrocalyx seeds under a parent tree, the rate of seed disappearance was much lower than that recorded previously. The re-analysis and the between-year variation in seed fates under a parent tree suggest that a rate of seedling establishment for O. macrocalyx I seeds of c. 1.6 times that of S seeds may be sufficient for selection to favour characteristics, such as delayed seed detachment from a pod, that promote seed dispersal by arboreal rather than terrestrial birds. Selection might also favour traits, such as increased hardness of the seed coat, that prevent terrestrial birds from scarifying seeds sufficiently to increase the germination rate.
It is interesting that in O. bopiensis uninterrupted dormancy did not confer a benefit to I seeds over S seeds in terms of seedling establishment, even though the S seeds germinated significantly earlier than the I seeds. This difference from the O. macrocalyx seeds may reflect the longer dormancy period in I seeds of O. bopiensis. Germination of the first and last I seeds spanned a period of 815 d, whereas the span in S seeds was only 59 d. Although the longer span in I seeds likely increased the probability of individual seeds encountering conditions favourable for seedling establishment, it also increased the likelihood of germination during a subsequent dry season or transition period. Nine of the 50 I seeds in the forest did germinate during a subsequent dry season or transition period, and two of those died from desiccation when their roots failed to enter the ground. Comparative studies of other Ormosia species with dormancy periods similar in length to that of O. bopiensis (Galetti Reference GALETTI, Levey, Silva and Galetti2002, Peres & van Roosmalen Reference PERES and VAN ROOSMALEN1996, Uhl Reference UHL1987) or O. macrocalyx (Foster & DeLay Reference FOSTER and DELAY1998, Garwood Reference GARWOOD1986, Ng Reference NG1973, Parolin et al. Reference PAROLIN, FERREIRA and JUNK2003) may help to explain the different responses of the species.
Dealing with dry periods
The seed/seedling stage is one of the most vulnerable for tropical forest trees and can significantly influence the species composition of a tropical forest (Engelbrecht et al. Reference ENGELBRECHT, KURSAR and TYREE2005). Dispersed seeds are vulnerable to predation, pathogens and desiccation. Ormosia seeds are apparently desiccation resistant (Pritchard et al. Reference PRITCHARD, DAWS, FLETCHER, GAMÉNÉ, MSANGA and OMONDI2004). None was observed to desiccate, even after exposure to multiple dry seasons, and seeds remained viable for many years (Uhl Reference UHL1987, this study). This is expected for species in which seeds may remain on a tree for months and are dispersed over much of the dry season. The seeds are also protected by both physical and physiological defence mechanisms. They are extremely hard with a smooth surface that insects appear unable to penetrate (Foster & DeLay Reference FOSTER and DELAY1998). In addition the seed coats of many species contain poisonous alkaloids that are avoided by both rodent and insect seed predators (Foster & DeLay Reference FOSTER and DELAY1998, Guimarães et al. Reference GUIMARÃES, JOSÉ, GALETTI and TRIGO2003, Lloyd & Horning Reference LLOYD and HORNING1958, Moran et al. Reference MORAN, QUINN and BUTLER1956, Reference MORAN, QUINN and BUTLER1957, Quinn et al. Reference QUINN, BUTLER and MORAN1956, Rudd Reference RUDD1965).
The plants are most vulnerable to water stress in the period between the onset of germination and the end of seedling establishment, which lasts 4.5–5 wk in O. macrocalyx and 6–7 wk in O. bopiensis (MSF unpubl. data). During this period, the seed is relatively soft, the seed coat is sloughing, the root system is rudimentary, and the plant depends on seed reserves for sustenance. The most effective way for plants to minimize mortality in this period is to shorten its length by developing rapidly. A strong dormancy mechanism that prevents seeds from germinating until conditions most favourable for seedling establishment arise decreases the likelihood that once initiated, development will be interrupted and delayed by unfavourable environmental conditions. Augspurger (Reference AUGSPURGER1979) found that developing seedlings of the shrub Hybanthus prunifolius (Violaceae) with the shortest period between sowing (the seeds have no dormancy period) and reliable rains had the highest survival, an outcome consistent with the findings for Ormosia. In contrast, Blain & Kellman (Reference BLAIN and KELLMAN1991) found that fluctuating moisture levels delayed seedling establishment in their study species, but did not decrease it. They suggested that the lack of a rainfall effect indicated the presence of effective mechanisms for dealing with water fluctuations and intermittent dry periods. However, they compared only intact seeds under different water exposures, avoiding any potential effects of a truncated dormancy period.
Survival of established seedlings is beyond the scope of this paper, which focuses on the ability of a seed to become an established seedling. Seedling survival will be considered elsewhere. Nevertheless, an abundant literature on the topic indicates that seedlings employ various morphological and physiological mechanisms to minimize the effects of water stress (Poorter & Markesteijn Reference POORTER and MARKESTEIJN2008, Slot & Poorter Reference SLOT and POORTER2007).
CONCLUSIONS
The degree of seed modification and magnitude of change in seed germination rate that occur after seeds of a particular plant species pass through an avian gut differ with the species of frugivore (Traveset Reference TRAVESET1998). Nevertheless, it has been assumed that any increase in the rate of germination will be beneficial, enhancing plant fitness (Barnea et al. Reference BARNEA, YOM-TOV and FRIEDMAN1991, Verdú & Traveset Reference VERDÚ and TRAVESET2005, but see Tewksbury et al. Reference TEWKSBURY, LEVEY, HUIZINGA, HAAK and TRAVESET2008). The results presented here show that whether earlier germination enhances or harms plant fitness will depend on the environmental context in which it occurs. It is likely that each plant species has an optimal range of germination rates for any particular array of environmental conditions, a topic that requires additional research.
ACKNOWLEDGEMENTS
I thank the Depto. de Áreas Naturales Protegidas, Instituto Nacional de Recursos Naturales del Perú and Parque Nacional del Manu for permission to carry out the research. I am also deeply indebted to the many field assistants whose help made the study possible: Thadeigh Baggallay, Martina Chavez de Sanchez, Daniel Huaman, I. Nanjokaitis-Lewis, Tim Paine, Amy Porter, Wendy Schelsky, Katherine Summers and Fiona Wilkinson. V. Funk, L.-A. Hayek and F. Wilkinson helped in other ways. R. W. McDiarmid, C. Krafft, D. J. Twedt and two anonymous reviewers provided useful comments on early drafts of the manuscript.