Based on their response to desiccation, seeds can be divided into two broad categories: orthodox and recalcitrant. Orthodox (hereafter termed desiccation-tolerant) seeds can be dried to low water contents (<7%) with little effect on viability (Roberts 1973). In contrast, recalcitrant (hereafter termed desiccation-sensitive) seeds are killed by drying to water contents as high as 20–30% (Pritchard 2004). Because desiccation-sensitive seeds progress towards germination when stored wet, they are difficult to store for anything other than the short term. Thus their use in reforestation and ex situ conservation programmes is problematic.
A range of studies have attempted to predict seed responses to desiccation from seed, plant and habitat variables (Tompsett 1984, 1987; Hong & Ellis 1997, 1998; Dickie & Pritchard 2002; Pritchard et al. 2004). Desiccation-sensitive seeds have been reported to be, on average, larger than desiccation-tolerant seeds, a feature that will reduce the rate of seed drying. For example, Dickie & Pritchard (2002) reported the mean seed mass of 205 desiccation-sensitive tree and shrub species to be 3958 mg compared with 329 mg for 839 desiccation-tolerant species. However, a potential problem with this analysis is that it treats species as independent when, in fact, closely related species share evolutionary history and therefore are not independent. This can result in spurious significances if a suite of closely related species share some unrelated traits (Harvey & Pagel 2000). Instead, branching events in the phylogenetic tree involving a change in seed storage behaviour, and the associated change in seed mass (or other traits), should be identified and used as independent data points. Only a single study appears to have investigated whether seed mass is associated with desiccation tolerance while correcting for phylogenetic dependence between taxa (Gleiser et al. 2004). This study compared the seed mass of just two desiccation-sensitive and 22 desiccation-tolerant Acer species and reported that the desiccation-sensitive species had larger seeds. However, as this study included only two desiccation-sensitive species, one of which has recently been re-evaluated and found to be desiccation-tolerant, this conclusion is tentative (Daws et al., in press).
In a study of 886 tree and shrub species, Tweddle et al. (2003) reported that desiccation-sensitive seeds are most common in tropical rainforests, where they contribute ≈47% of species and are infrequent in drier environments such as savanna (≈12% of species). Even within dry environments, species with desiccation-sensitive seeds can minimize the risk of seed desiccation by timing seed shed to the period of maximum rainfall (Pritchard et al. 2004). An alternative strategy has been reported for nine species in the genus Coffea, where the level of desiccation tolerance is related to the duration of the dry period after seed shed: species shed before a prolonged dry spell were more desiccation-tolerant than those shed prior to a short dry spell (Dussert et al. 2000).
It has been suggested that desiccation tolerance is the ancestral state in seeds and has subsequently been lost in species with desiccation-sensitive seeds (Farnsworth 2000; Oliver, Tuba & Mishler 2000; Dickie & Pritchard 2002). The ability to tolerate desiccation is clearly advantageous, and enables seed persistence both through time and in relatively arid environments (Pammenter & Berjak 2000). While having large, round seeds shed to coincide with the peak in annual rainfall may minimize the risk of desiccation for desiccation-sensitive seeds, it is not yet clear whether seed desiccation sensitivity is a neutral trait, or whether there are selection benefits associated with desiccation sensitivity (Pammenter & Berjak 2000).
One potential advantage of seed desiccation sensitivity may be rapid germination. Desiccation-sensitive seeds are shed at high water contents, are metabolically active, and in some cases are actively progressing towards germination (Berjak et al. 1984). Consequently, limited or no imbibition is required for germination to progress rapidly following dispersal. This proposition has been tested for a limited number of African dryland trees by Pritchard et al. (2004) who found that, at a constant temperature of 25 °C, three taxa with desiccation-sensitive seeds germinated more rapidly than six desiccation-tolerant taxa.
Mast fruiting and rapid germination of dipterocarp seeds (Curran & Webb 2000), the vast majority of which are desiccation-sensitive (Tompsett & Kemp 1996), is thought to result from selection pressure from vertebrate seed predators. Similarly, Pammenter & Berjak (2000) have proposed that for climax species in tropical forests, many of which have desiccation-sensitive seeds, the formation of a seedling bank will reduce seed predation by fungi. Consequently, rapid germination of desiccation-sensitive seeds may minimize both the risk of seed drying and the duration of exposure to predation. Consequently there may be reduced selection for investing resources in seed physical defences, in terms of both reducing predation risk and the mechanical restraint to germination. In the study by Pritchard et al. (2004), an average of 15% of the dispersal unit was endocarp/testa for three desiccation-sensitive species compared with 46% for seven desiccation-tolerant species. Pritchard et al. (2004) hypothesized that desiccation-tolerant species require greater defences because dispersal may occur in the dry season or during short dry spells, and seeds may be exposed to predation in the soil seed bank for extended periods. However, the wider applicability of the findings of Pritchard et al. (2004) is unclear, particularly given the small data set involved (10 species).
Seed mass can vary over 10 orders of magnitude, and for tropical tree seeds may be correlated with germination rate: for 179 Malaysian tree species Foster (1986) reported a significant positive correlation between seed mass and time to first germination. Seed mass may also be positively related to the proportion of seed resources allocated to physical defences (e.g. within a family; Fenner 1983; Schütz 2000). Consequently, to address more fully the implications of desiccation sensitivity for seed mass, germination rates and seed defences, analyses that account for phylogenetic relationships and/or seed mass are also required.
In this paper we examine these propositions for a data set of 225 tree and shrub taxa from a semi-deciduous tropical forest in Central Panamá, including 36 with desiccation-sensitive seeds. Specifically, we use cross-species analyses and account for phylogenetic relationships to address the following questions: (1) Are desiccation-sensitive seeds larger than desiccation-tolerant seeds? (2) Do desiccation-sensitive seeds germinate more rapidly than desiccation-tolerant seeds, and does this relationship hold when accounting for seed mass? (3) Do desiccation-sensitive seeds have a lesser investment in physical defences than desiccation-tolerant seeds, and is this true when accounting for seed mass? In addition, using cross-species analyses we test whether desiccation-sensitive seeds are dispersed when water is most available (when the risk of drying is minimal), while desiccation-tolerant seeds may be shed in wet or dry periods. The results are discussed in the context of the ecological costs/benefits of seed desiccation sensitivity.