Physical dormancy in seeds: a game of hide and seek?

Authors


Author for correspondence:

Torbjørn Rage Paulsen

Tel: +47 55 58 26 71

Email: torbjorn.paulsen@bio.uib.no

Summary

  • Historically, ‘physical dormancy’, or ‘hard seededness’, where seeds are prevented from germinating by a water-impermeable seed coat, is viewed as a dormancy mechanism. However, upon water uptake, resumption of metabolism leads to the unavoidable release of volatile by-products, olfactory cues that are perceived by seed predators. Here, we examine the hypothesis that hard seeds are an anti-predator trait that evolved in response to powerful selection by small mammal seed predators.
  • Seeds of two legume species with dimorphic seeds (‘hard’ and ‘soft’), Robinia pseudoacacia and Vicia sativa, were offered to desert hamsters (Phodopus roborovskii) in a series of seed removal studies examining the differences in seed harvest between hard and soft seeds. Volatile compounds emitted by dry and imbibed soft seeds were identified by headspace gas chromatography–mass spectrometry (GC-MS).
  • Fourteen main volatile compounds were identified, and hamsters readily detected both buried imbibed seeds and an artificial ‘volatile cocktail’ that mimicked the scent of imbibed seeds, but could not detect buried hard or dry soft seeds.
  • We argue that physical dormancy has evolved to hide seeds from mammalian predators. This hypothesis also helps to explain some otherwise puzzling features of hard seeds and has implications for seed dispersal.

Introduction

Seed survival, dispersal and persistence in soil determine the composition and dynamics of plant communities, and represent a cornerstone of our understanding of plant biology. To regulate the timing of germination, most plant species produce seeds that are dormant at maturity, and ‘physiological dormancy’ is the most frequent type of dormancy (Baskin & Baskin, 1998). Physiological dormancy prevents germination of imbibed seeds until specific environmental cues have been perceived, acting as a reversible switch that enables plants to fine-tune the germination response of seeds to the environment. When in soil seed banks, these seeds may lose and re-acquire dormancy in a predictable seasonal cycle for decades.

Some species have ‘hard’ seeds, where germination is prevented by water-impermeable seed coats. ‘Hard seededness’, or ‘physical dormancy’, is present in a few plant families, of which the Leguminosae are the largest (Baskin & Baskin, 1998). Physical dormancy can be broken only once, although sensitivity to dormancy-breaking stimuli can cycle in some species (Baskin & Baskin, 2000). It has been suggested that hard seededness protects against microbial attack (Dalling et al., 2011), and extends seed longevity (Mohamed-Yasseen et al., 1994) and persistence in soil seed banks (Shen-Miller et al., 1995). Nevertheless, in the very extensive literature on hard seeds, hard seededness is interpreted solely as a form of dormancy, and this dormancy function has been assumed to be the reason for the evolution of hard seeds. However, in terms of germination regulation, physical dormancy seems to do nothing that physiological dormancy cannot do better, which raises the question of why seeds of some species have evolved physical rather than physiological dormancy (Baskin & Baskin, 1998).

Here we examine the alternative hypothesis that hard seededness is largely a by-product of powerful selection by small mammal predators that detect seeds primarily by olfaction. It is well documented that scatter hoarding rodents harvest seeds persistently, and daily removal rates of artificial caches of, for example, black oil sunflower (Helianthus annuus), white millet (Panicum miliaceum), acorn squash (Curcurbita pepo), antelope bitterbrush (Purshia tridentate) and Jeffrey pine (Pinus jeffreyi) in a natural environment were 37.8, 16.2, 9.5, 8.2 and 7.5% under dry conditions and 46.8, 39.5, 36.1, 34.6 and 27.0% under wet conditions, respectively (Hollander et al., 2012). This assumption is in agreement with the suggestion by Vander Wall that seeds dispersed by scatter hoarding rodents are under strong selection pressure to reduce their olfactory signal and that impermeable seed coats may play a role in preventing the release of compounds that can be smelled by granivorous seed predators (Vander Wall, 2003; Hollander et al., 2012), based on the observation that rodents find it difficult to detect buried ‘dry’, metabolically inactive, seeds (Howard & Cole, 1967; Johnson & Jorgensen, 1981; Vander Wall, 1993a, 1995, 1998; Jorgensen, 2001). Such seeds emit only a few volatiles, at low concentrations (Jorgensen, 2001), produced via various chemical reactions, including lipid peroxidation, alcoholic fermentation and Maillard reactions (Colville et al., 2012). When seed metabolism is reactivated upon imbibition, seed storage reserves are hydrolysed to provide the energy and building blocks for the synthesis of cellular components required for growth, and this coincides with greatly increased emission of volatile compounds. For example, ethanol and acetaldehyde may derive from glycolysis (Stotzky & Schenck, 1976; Woodstock & Taylorson, 1981; Lee et al., 2001). Therefore, we examined whether water impermeability has evolved to prevent the production and release of volatile compounds. We tested whether water-impermeable hard seeds escape predation more frequently compared with permeable soft seeds, identified volatile compounds released by both soft and hard seeds, and finally discuss potential implications of hard seededness for seed dispersal and thus evolutionary success.

Materials and Methods

To experimentally unravel the effects of seed coat permeability, and thus smell, on seed predation, Vicia sativa L. and Robinia pseudoacacia L. were chosen, because they naturally produce a mixture of hard and soft seeds (‘dimorphic’ for seed coat). Desert hamsters (Phodopus roborovskii Satunin) were chosen as model seed predators, because they are larder hoarders and harvest seeds persistently, and are easy to keep.

Seed material and treatments

Seeds of V. sativa and R. pseudoacacia were obtained in 2001 from Herbiseed (Twyford, UK) and Corpo Forestale dello Stato Centro Nazionale per lo Studio e la Conservazione della Biodiversità Forestale (Peri-Vr, Italy), respectively. Seeds were stored at 15% relative humidity (RH) and 15°C to limit deterioration caused by ageing, and no decline in viability occurred during the storage period. To sort hard from ‘naturally soft’ seeds, dimorphic seed lots were spread in a mono-layer on filter paper moistened with sterile dH2O in plastic boxes, and imbibed for 8 h at 10°C. Soft seeds swelled (imbibed), whereas hard seeds did not, allowing their separation. Sorted seeds were rinsed with sterile dH2O, and then dried at 15°C and 15% RH. ‘Scarified seeds’ were produced by puncturing individual hard seeds with a needle, followed by 24 h of imbibition at 4°C before drying in a desiccator for 96 h. To test whether multiple imbibitions affect the detectability of seeds, punctured hard seeds were imbibed and dried again. Germination tests were conducted on 1% water agar at 20°C with a 8 h : 16 h, light : dark cycle with three replicate samples of 20 seeds. Seed moisture content (MC) was measured and expressed on a fresh weight (FW) basis. Dry weight (DW) was determined after heating to 103°C for 17 h and MC calculated: MC = (FW – DW)/FW × 100. Three replicate samples of 100–150 seeds were used.

Seed volatile analysis

The total volatile abundance was determined for hard and soft seeds in open vials to closely mimic the conditions used in the hamster experiments. Two grams of dry whole seeds (100–150 seeds) were initially sealed in 20-ml gas chromatography (GC) headspace vials (Chromacol; Sigma Aldrich, Gillingham, UK) and stored at room temperature for 2 h before headspace analysis. Thereafter, they were incubated in 5 ml of dH2O for 18 h in open glass vials, then transferred to 20-ml sealed headspace vials. Volatile analysis was performed after 2 h using automated solid-phase microextraction (SPME) with a 75-μm Carboxen-polydimethylsiloxane (PDMS) fibre (Supelco, Bellefonte, PA, USA) and an extraction time of 30 min at 30°C followed by 5 min of desorption in the GC injector port at 240°C. The volatiles were separated using GC (Thermo Finnigan Trace GC Ultra; Thermo Fisher Scientific, Waltham, MA, USA) on a FAMEWAX column (30 m length, 0.25 mm internal diameter, 0.25 μm film thickness; Restek, Bellefonte, PA, USA) running a temperature program (3 min hold at 35°C, 3°C min−1 to 60°C, 10°C min−1 to 220°C and 1 min hold; helium carrier gas at constant flow rate of 1 ml min−1). The volatiles were detected using mass spectrometry (MS; Thermo Finnigan Trace DSQ; ionization energy 70 eV and scan frequency range m/z 10–350 per 0.5 s).

To identify the volatile compounds released by hard and soft seeds, including low-abundance volatile compounds which were undetectable in unsealed vials, 2 g of seeds were sealed in 20-ml headspace vials at room temperature. Following analysis of the headspace of dry seeds, 2 mL of dH2O was added and the vials re-sealed. The headspace within the vials was sampled 2 h after the addition of water and every 3 h thereafter until 20 h. The volatiles detected were identified from the NIST (National Institute of Standards and Technology, Gaithersburg, MD, USA) mass spectral database. The identities of all compounds shown in Supporting Information Table S1 were confirmed through comparison with analytical standards (Sigma Aldrich). The results after 20 h of imbibition are presented, which is the time after which imbibed seeds were offered to the hamsters.

Two ‘volatile cocktails’ (the ‘Vicia cocktail’ and ‘Robinia cocktail’) were created to mimic the volatile composition in the headspace above seeds imbibed for 20 h (Tables S1, S2). Initially, the cocktails were prepared by mixing analytical standards (Sigma Aldrich) in proportions based on the relative abundance (in terms of GC-MS peak area) of volatile compounds released by imbibed soft seeds. The cocktails were prepared in 10-ml glass vials on ice, and aliquots were transferred to sealed 20-ml headspace vials. The headspace above the cocktails was analysed using SPME-GC-MS, and the cocktail composition was methodically altered and the headspace re-analysed until the relative abundance of the individual volatiles closely matched that measured in the headspace above seeds imbibed for 20 h.

Seed predation experiments

Ten seed removal experiments were conducted: for each plant species, ‘mono caches’ of either 20 hard or 20 soft seeds were offered in a small pit on top of dry (Expt 1) or wet (Expt 2) gravel or buried beneath a c. 1-cm layer of dry (Expt 3) or wet (Expt 4) gravel. To mimic conditions following seed shed, 10 hard and 10 soft seeds placed alternately in a circle on top of dry (Expt 5) or wet (Expt 6) gravel were offered. As a more rigorous test of hard seed predator escape and also to mimic cache pilferage or recovery, ‘mixed caches’ of 10 hard and 10 soft seeds were buried beneath dry (Expt 7) or wet (Expt 8) gravel. To test the effects of multiple imbibition on seed detectability, 20 scarified hard seeds imbibed once were tested against 20 scarified hard seeds imbibed twice, buried beneath wet gravel (Expt 9). To test whether the hamsters were able to distinguish naturally soft from scarified hard seeds, 20 imbibed seeds of each type were offered buried beneath wet gravel (Expt 10). Seeds were only offered to hamsters once, and seeds recovered from Petri dishes were not re-used in subsequent experiments.

We chose controlled laboratory experiments where background smells are minor, and the hamsters could harvest seeds without being threatened by carnivores. For all experiments, 14 Petri dishes were placed along two opposite walls in a wooden arena (75 × 75 × 30 cm; Fig. 1a) covered with absorbent paper (BenchGuard; OneMed, Lørenskog, Norway). Six of the Petri dishes (90 × 16.2 mm) contained seeds (either mono caches of hard or soft seeds (three dishes each; Expts 1–4) or mixed caches of both seed types (Expts 5–8)), and eight of the Petri dishes contained only gravel (gravel size 2–5 mm; 1 : 1 AT 207 Alpha (red-black): AT205 Merkur (black); Akvastabil, Haderslev, Denmark); the Petri dishes were positioned randomly in two rows of seven Petri dishes. Also present in the arena was a nest box (30 × 8 × 6 cm) filled with nesting material (Hamsternest; Zoobest, Stolberg, Germany), a running wheel, and a water bottle (Fig. 1a). The seeds were placed c. 1 cm from the edge of the Petri dishes, and the relative orientation of the cache was randomized in relation to the arena wall to prevent the hamsters from learning where in the dish the cache could be found. The crepuscular hamsters were introduced to the arena before they became active (17 : 00 h) and removed after they had gone to rest the following day (12 : 00 h). The dishes were recovered, and seeds missing from a dish were considered removed. The wet gravel was rinsed once between trials and re-used only with the same seed type and species. The hamsters were used in experiments every third night and fed complete hamster feed (preStige Hamster Crispy; Versele-Laga, Deinze, Belgium) ad libitum between experiments. Each experiment was replicated 12 times with different hamsters (six of each sex) in each run, for a total of 72 Petri dishes containing 1440 seeds and 96 Petri dishes containing only gravel. Gloves were always used when handling seeds, dishes and vials (Duncan et al., 2002).

Figure 1.

The arena. (a) The set-up of the wooden arena in which the hamsters were offered seed caches and which included a nest box filled with nesting material, a running wheel, and a water bottle. (b) The hamsters easily find and harvest buried seed caches using olfaction.

For Expt 11, Vicia and Robinia cocktails were prepared (Table S2) on ice immediately before the experiments; 900 μl of the volatile cocktails diluted 1 : 10 with dH2O was offered to the hamsters in unsealed 1-ml glass vials (Chromacol; Sigma-Aldrich Norway AS) placed horizontally at the bottom of Petri dishes and buried beneath gravel. Five hard Robinia seeds were placed on top of each buried vial to reward searching behaviour. Two vials of each volatile cocktail were tested individually against two vials containing ‘seed water’ (dH2O in which seeds (Robinia: 1 seed ml−1; Vicia: 2 seeds ml−1) had been soaked for 24 h) and two vials containing dH2O (negative control) as described for Expts 1–10. Only vials that were uncovered or dug up were considered as excavated. In Expt 12, the Robinia cocktail was tested against vials containing dH2O only.

Statistical analysis

Binary logistic regression was used to test for differences in seed removal (defined as the proportion of seeds harvested across all offered caches), conditional seed removal (defined as the proportion of seeds removed from mixed caches, given that the cache is harvested) and cache discovery (defined as the proportion of caches where seed harvest occurred) between hard and soft (or scarified) seeds, and between the discoveries of vials containing cocktails, seed water and dH2O. The proportion of seeds taken from each dish is the response variable (coded as success/failure), and, to account for individual differences in activity, hamsters were treated as random effect factors using a generalized linear mixed effects model (GLMM), applying the glmmPQL function from the MASS library of R (R Development Core Team, 2011). However, it is a condition for this model to have a certain level of variance, and if this was not met Pearson's chi-squared tests for 2 × 2 tables were used. Because of possible test bias resulting from low n and/or variance in the cache discovery and volatile cocktail experiments, we also performed Pearson's chi-squared tests for 2 × 2 tables, which did not change the conclusions of the GLMM tests. R (R Development Core Team, 2011) for PC v2.14.0 was used for all computations, with the critical alpha level (two-tailed) set at 0.05.

Results

When incubated in water for 20 h, the MC of ‘naturally soft’ seeds (termed ‘soft’) rose from 7.3 to 61%, but ‘naturally hard’ seeds (‘hard’) did not imbibe. Hard seeds that were made permeable (‘scarified seeds’) also imbibed and were used as a control. All seed lots were highly viable, with a total germination of > 95% after 3 d.

Hamsters struggle to find buried hard seeds by olfaction

When ‘mono caches’ of either hard or soft seeds were offered to the hamsters on top of dry or wet gravel, both hard and soft seeds were detectable by sight and touch, and were harvested within hours (Fig. 1b). There was no preference for hard or soft Vicia seeds, but soft Robinia seeds were preferred to hard seeds (Expts 1 and 2; Fig. 2a,d; seed harvest probabilities and test statistics can be found in Table S3). When buried beneath dry gravel, and thus detectable only by olfaction, hamsters were largely unable to locate hard seeds of both Vicia and Robinia, and soft Vicia seeds, although they still found soft Robinia seeds (Expt 3; Fig. 2a,d; Table S3). However, when seeds were buried beneath wet gravel, imbibed soft seeds of both species were harvested, but impermeable seeds were not; the probability of both seed harvest and cache discovery of impermeable seeds was only 1/4 of that for permeable seeds in Vicia and 1/6 in Robinia (Expt 4; Fig. 2a,d; Table S3).

Figure 2.

Seed harvest. Hamsters readily harvest imbibed, soft seeds of Vicia sativa and Robinia pseudoacacia with increased production of volatile compounds, while hard seeds do not imbibe and escape predation. Seeds were offered either on top of or buried beneath dry gravel (red shading), or incubated in water for 20 h before being offered on top of or buried beneath wet gravel (blue shading). Soft seeds imbibed (blue), and hard seeds did not (red). Seeds were either offered in (a, d) mono caches of either 20 hard or 20 soft seeds or (b, e; left y-axis) mixed caches of 10 hard and 10 soft seeds. Seeds removed from all offered dishes (72) are shown for mono caches (a, d), and the conditional probability of seed removal is shown for buried mixed caches (b, e), representing only seeds removed from discovered caches (Vicia, 53 of 72; Robinia, 65 of 72). (c, f) Vials containing volatile cocktails (purple; right y-axis) along with vials containing water or seed water (Expt 11; V. sativa; c) or only water (Expt 12; R. pseudoacacia; f) were offered buried beneath wet gravel. Expts 9 and 10 are shown in Supporting Information Fig. S1. Box plots show the medians and upper and lower quartiles, whiskers indicate maximum and minimum values, and circles show outliers. Each experiment was replicated 12 times for a total of 1440 seeds and 168 Petri dishes, 72 of which contained seeds, and the rest of which contained only gravel (see the 'Materials and Methods' section). Experiment numbers are indicated at the bottom of the graphs (Table S3). Asterisks denote confidence levels for the comparison between hard and soft seed coats (a, b, d, e) or the comparison between volatile cocktail/seed water and dH2O (c, f) (***, P < 0.001; **, P < 0.01; *, P < 0.05).

As a more rigorous test of hard seed predator escape, ‘mixed caches’ of 10 hard and 10 soft seeds were offered on top of dry (Expt 5) or wet (Expt 6) gravel, mimicking the conditions of primary seed dispersal, when seeds are shed from the plant. Soft seeds were taken significantly more often than hard seeds (Fig. 2b,e; Table S3). When mixed caches, used to mimic cache recovery and pilferage, were buried beneath dry gravel there was no difference in conditional seed removal (see the 'Materials and Methods' section) for either species (Expt 7). When buried beneath wet gravel, however, soft seeds were again taken more frequently than hard seeds (Expt 8; Fig. 2b,e; Table S3). To test if multiple imbibition of soft seeds makes them more detectable to the hamsters, scarified seeds were imbibed either once or twice. Vicia seeds imbibed twice were harvested slightly more often than seeds imbibed once (Expt 9; Fig. S1; Table S3), whereas no difference was found for Robinia. We also examined if hamsters distinguish between imbibed soft and imbibed scarified hard seeds, and found no difference in seed harvest in Vicia but scarified hard Robinia seeds were taken more frequently (Expt 10; Fig. S1, Table S3).

Identification of seed volatiles

No volatiles emitted from dry, hard seeds were detected by headspace GC-MS analysis. Hard seeds incubated in water emitted volatiles at very low levels only, mainly ethanol, and Robinia additionally produced 1-pentanol. Only three low-level volatiles were emitted by dry soft seeds; acetic acid and ethanol were emitted from Vicia and Robinia seeds, which additionally emitted acetone and methyl acetate, respectively (Table S1). By contrast, total volatile emission by soft seeds increased following 20 h of incubation in water, and was 9× and 12× higher than that of hard wet seeds of Vicia and Robinia, respectively (Fig. 3). Imbibed soft seeds emitted 19 volatile compounds, of which 1-propanol, 2,3-butanedione, 3-hydroxy-2-butanone, acetic acid, acetone, ethanol, ethyl acetate, methanol and methyl acetate were produced by both species (Fig. 4, Table S1).

Figure 3.

Total amount of volatiles detected. Volatile production by soft, water-permeable seeds of Vicia sativa and Robinia pseudoacacia greatly increases following imbibition in water. ‘Dry’ denotes volatiles detected in the headspace of dry seeds (red), and ‘Wet’ denotes those measured in the headspace of seeds incubated in water for 20 h (blue). Hard seeds did not imbibe during incubation in water, whereas soft seeds did. Data represent mean total peak area ± SE (= 3) of volatiles present in the headspace of unsealed vials detected using solid phase microextraction (SPME) gas chromatography–mass spectrometry (GC-MS).

Figure 4.

Relative abundance of volatiles from seeds and cocktails. Seed volatiles accumulated in the headspace of soft seeds imbibed for 20 h, and in the headspace of the artificial Vicia and Robinia cocktails offered to hamsters. The abundance of each volatile compound detected by headspace solid phase microextracion (SPME) gas chromatography–mass spectrometry (GC-MS) above soft seeds imbibed for 20 h is shown as a percentage of total peak area. Each value represents the mean of six replicates. Volatile cocktails were created and their composition adjusted until their headspace profile determined using SPME GC-MS approximated that of imbibed, soft seeds (see the 'Materials and Methods' section). The volatile composition depicted by the pie charts is based on the peak areas obtained using SPME GC-MS, and does not represent the molar composition of volatiles released by seeds, which was not determined. Alcohols, blue to purple; ketones, yellow to green; aldehyde, orange; acetates, pink; and other compounds, miscellaneous colours.

We then created a ‘Vicia cocktail’ and a ‘Robinia cocktail’ corresponding to the volatile compounds accumulated in the headspace of seeds imbibed for 20 h (Fig. 4; Table S2). These two cocktails were offered to the hamsters in glass vials buried beneath wet gravel together with vials containing ‘seed water’ as a positive control (dH2O in which seeds had been soaked for 24 h) and dH2O as a negative control (Fig. 2c,f). The hamsters dug up seed water of both species, but not the negative control (Expt 11; Table S3). Both the Vicia and the Robinia cocktails were found significantly more frequently than vials containing dH2O (Expts 11 and 12, respectively; Fig. 2c,f).

Discussion

Hamsters were highly efficient at finding seeds when they were detectable by sight, touch or smell, but poor at finding hidden seeds of either test species when denied olfactory cues (Fig. 2). Hard seeds released very low levels of volatile compounds compared with imbibed soft seeds (Fig. 3). Metabolism in dry seeds is restricted because of the low molecular mobility in the cell. However, low levels of volatiles may be produced as a result of degradative processes such as lipid peroxidation and Maillard reactions (Colville et al., 2012). In addition, localized hydrated regions may allow metabolic reactions to occur at slow rates (Leubner-Metzger, 2005). We identified three volatiles in dry seeds of each species at low levels, just above the detection limit of the GC-MS for a large seed mass of 2 g. Ethanol, acetic acid, acetone and methyl acetate are typical products of lipid peroxidation, although they can also be produced metabolically through glycolysis. Many of the volatile compounds released by imbibing soft seeds are unavoidable by-products of metabolic reactions such as glycolysis (Colville et al., 2012). For example, pyruvate from glycolysis can be converted by pyruvate decarboxylase to acetaldehyde, then to ethanol by alcohol dehydrogenase or acetic acid by aldehyde dehydrogenase. Acetone may also arise from pyruvate decarboxylation. Other compounds, including many alcohols and ketones, may derive from lipid peroxidation, which occurs at low rates during seed development, in dry and in germinating seeds (Kranner et al., 2010). Aliphatic ketones have also been shown to attract mammal pollinators to the flowers of Cytinus visseri Burgoyne, the scent of which differs markedly from that of insect-pollinated plants (Johnson et al., 2011). This may reflect the evolved sensitivity of rodents to aliphatic volatile compounds, such as those produced by seeds in this study.

It is possible that soft seeds are preferred by hamsters for reasons unrelated to detectability; for example, imbibed seeds are larger than dry seeds, which could explain the higher removal rates of imbibed seeds in mixed caches offered on top of gravel. However, this cannot account for the 4- to 6-fold difference in detectability between buried caches of hard and imbibed soft seeds (Table S3), because smell is the only possible search mode for buried seeds that would differentiate between hard and soft seeds. Likewise, dry soft Robinia seeds emitted more volatiles than dry soft Vicia seeds, and were harvested at a higher rate. Furthermore, imbibed ‘scarified hard’ seeds were harvested at the same rate as, or at a higher rate than, imbibed soft seeds. Hence, the differences in seed harvest between seed types are attributable to olfactory detectability and not to seed preference.

Using volatile cocktails we demonstrated that the volatiles released by imbibed soft seeds are olfactory cues that can be used by hamsters to detect buried seed caches. In contrast to soft seeds, hard seeds do not imbibe unless the seed coat has been breached, and this prevents the release of volatile compounds, thereby making the seeds very difficult to find by olfaction when buried in wet or dry soil. The animal and plant species used in this study do not naturally overlap in distribution, but previous studies have demonstrated that dry soft seeds are hard to find even for sympatric seed predators (Howard & Cole, 1967; Johnson & Jorgensen, 1981; Vander Wall, 1993a, 1995, 1998; Jorgensen, 2001). Notably, the MC of soft seeds strongly influences their detectability. For example, neither yellow pine chipmunks (Tamias amoenus Allen) nor deer mice (Peromyscus maniculatus Wagner) could find buried antelope bitterbrush seeds with 6.8% MC, but at 8% MC seeds were readily found (Vander Wall, 1993a), demonstrating the challenge plants face for their seeds to remain undetectable by olfaction. Even high relative humidity (95%), assumed to produce dew and increase soil water vapour, was enough to make buried Jeffrey pine seeds olfactionally detectable for yellow pine chipmunks (Downs & Vander Wall, 2009), and nightly dew formation (Jacobs et al., 1999) will make soft seeds detectable by olfaction on a daily basis, and thus an easy prey for foraging rodents. This could explain why hard seeds are surprisingly frequent in hot deserts (Baskin & Baskin, 1998), where granivory by small rodents is very intense (Price & Joyner, 1997), exerting very strong selection pressure on seeds to evolve methods of avoiding detection (Vander Wall, 2003).

In the majority of species with impermeable seed coats, the embryo is nondormant (Baskin & Baskin, 1998). Once the seed coat is broken, the typical hard seed germinates rapidly over a wide range of temperatures in both light and darkness, thus minimizing the period when the seed is present in the soil in a ‘discoverable’ state. We argue that the fact that physical dormancy can only be broken once (Grime et al., 1981; Baskin & Baskin, 1998) raises questions about its efficiency as a dormancy trait. We propose that predator escape by crypsis is a more plausible explanation for the evolution of hard seeds than its commonly assumed role as a dormancy trait. The ‘crypsis hypothesis’ explains why physical dormancy is absent from lineages with predominantly small seeds (Leishman & Westoby, 1998; Šerá & Šerý, 2004; Moles et al., 2005), for which burial itself offers protection against predation (Hulme, 1998), whereas large seeds are attractive and detectable to rodents, even when buried. The crypsis hypothesis rather than the dormancy hypothesis also explains why large-seeded species often either have ‘recalcitrant’ (i.e. desiccation-sensitive) seeds, which do not form soil seed banks and germinate soon after shedding from the mother plant, or are hard-seeded (Baskin & Baskin, 1998). Here, recalcitrance and hard seededness seem to be alternative ways of dealing with the same problem, that is, predation (Pritchard et al., 2004; Daws et al., 2005), whereby wet environments seem to have predisposed plants to evolve recalcitrance, whereas hard seeds have arisen in plants of (at least seasonally) dry environments (Baskin & Baskin, 1998).

The seed coat of hard seeds may also function as a barrier to soil microorganisms (Dalling et al., 2011). However, this does not answer the question of why most hard-seeded plants produce dimorphic seeds (Meisert, 2002). Many plant species rely on scatter-hoarding rodents for secondary seed dispersal (Vander Wall, 1990), probably benefiting from relocation of seeds away from potentially hazardous environments, reduced sibling competition, and topsoil burial (5–40 mm depth) beneficial for seed germination and establishment (Howe & Smallwood, 1982; Vander Wall, 1993b). However, cached seeds are usually recovered and either eaten or re-cached by the cacher, or pilfered by neighbours (Vander Wall, 1998; Jansen et al., 2002; Vander Wall et al., 2003), and the seed must escape the hoarder in order to profit from the scatter hoarding (Jansen et al., 2002). This presents an evolutionary dilemma: seeds must be located by predators if they are to be dispersed, but some of them must later escape predation. The mixed cache experiments demonstrate that both hard and soft seeds will be found when shed from the plant onto the soil surface, and that the probability of predator escape during re-caching/pilfering is much higher for hard than for soft seeds, and suggests that hard seeds have an advantage by surviving secondary seed dispersal better than soft seeds. This suggests that the anti-predator trait of hard seeds also facilitates seed dispersal and explains why plants continue to produce some permeable soft seeds. The soft seeds are bound to experience high levels of predation, arguably the plant's payment for dispersal services, analogous to other plant payments like pollen for pollination or the nutritious pulp of fleshy fruits. In terms of reward, dimorphic seeds support the hypothesis that the primary evolutionary explanation for hard seeds is predator escape, and we argue that this explains why hard seeds have evolved independently at least six times (Baskin et al., 2000).

Our new interpretation of hard seededness calls for a renewed interest in their function and evolutionary history. The composition and dynamics of plant communities are strongly influenced by seed germination, survival, dispersal and persistence in soil, and seed traits are routinely interpreted in terms of these functions (Römermann et al., 2005). If we are to understand plant communities, it is essential that we critically examine the evidence that these traits do what we think they do. In the case of hard seeds, we may have been guilty of misinterpreting their primary function for at least a century.

Acknowledgements

The Royal Botanic Gardens, Kew receive grant-in-aid from Defra. We thank Knut Helge Jensen and Einar Heegaard for assistance with statistical tests, Sebastian König and Orlando de Lange for help with creating the volatile cocktails, and Steve Vander Wall, Steve Johnson and three unknown reviewers for their useful suggestions on how to improve the manuscript.

Ancillary