1The seed-to-seedling transition may be a critical stage in determining the dynamics of plant populations. Insects which kill seeds either before or after dispersal can influence the population dynamics of individual plant species, and ultimately, plant diversity and assemblage composition.
2We discuss the potential for insect seed predators to maintain diversity in plant assemblages and to structure their composition, with a particular focus on diverse tropical forest habitats. We suggest that our ability to understand the functional effects of insect seed predators is hampered by a shortage of unbiased information on (i) their responses to the density of prey seeds at different spatial scales, and (ii) their host plant specificity.
3Density-dependence and its implications may be best assessed using manipulative field experiments. Such approaches can reveal how insect seed predators respond behaviourally and demographically to the density of individual host species and multiple host species across a range of spatial scales.
4Host specificity and its implications may be best addressed through quantitative food web approaches previously applied largely to host–parasitoid interactions. Food webs will allow ecologists to assess the likely importance of indirect interactions such as apparent competition and apparent mutualism in structuring plant assemblages, and the functional consequences of adding or removing individual resource or consumer species.
5Fully quantifying the wider effects of seed predators will require studies that better integrate seed stage-specific demographic information, and which quantify the long-term effects of variations in seed predation rates for plant recruitment.
6Synthesis and applications. Compared to other functionally important insect groups such as pollinators, seed predators have received relatively little attention in the context of the functioning and sustainability of agro-ecosystems and the consequences of global environmental change for ecological communities. A fuller understanding of the ecology of insect seed predator–plant interactions will be valuable to conservation and management in a range of natural and agricultural systems. For example, seed predator community ecology is relevant to predicting the consequences of deliberately or unintentionally introducing novel resource or consumer species; the process of habitat recovery following local disturbances; and managing the effects of pest or beneficial seed predators in agricultural crops. Furthermore, patterns of insect seed predation on a larger scale are likely to be highly sensitive to global environmental change drivers such as climate change and systematic habitat modification and fragmentation, with implications for the structure and organization of ecological communities more widely.
Seed predators are animals that kill seeds, either by consuming them completely or by damaging them to the extent that germination and growth are impossible (Janzen 1970). While herbivores typically damage a relatively small proportion of plant biomass with variable effects on plant fitness, seed predators have direct and obvious effects on plant fitness and frequently strong effects on patterns of plant recruitment for individual species (Fenner & Thompson 2005), although their effects can be difficult to detect in nature (Harper 1977). At the plant assemblage or community level, the sum of the effects of seed predators on individual species, and interacting effects among plant species mediated by shared seed predators, can have a major effect on plant diversity, composition and dynamics (e.g. Veech 2000; DeMattia et al. 2006; Paine & Beck 2007).
Many seed predators attack seeds pre-dispersal (while they are still attached to the plant), where they represent an abundant and spatially aggregated resource (Crawley 1992). The consequences of pre-dispersal seed predation for a plant are similar to those of producing a smaller seed crop (Harper 1977). A second group of seed predators act post-dispersal (while seeds are on or in the soil). These seed predators may alter the shape of the surviving seed shadow in a manner determined by their searching strategy or efficiency. Harper (1977) suggests an alternative dichotomy for categorizing seed predators. His first category includes specialized, short-lived species (mostly insects) whose phenology is synchronized with the phenology of the host plant. The second category, into which most vertebrate seed predators fall, includes longer-lived, polyphagous species which feed opportunistically on seeds of different species at different times, either pre- or post-dispersal. For the most part, in this review we are concerned with the first category: insect seed predators (primarily beetles, flies and moths) which are commonly considered to be relatively specialized (e.g. Janzen 1970), at least compared to most vertebrate seed predators. As we discuss below, the two defining characteristics of this group, specialization and synchronization, as well as the restricted spatial range of their foraging (Hammond & Brown 1998) are key to understanding their functional importance.
In this study, which accompanies a special collection of articles on the role of seed predators in applied ecology (van Klinken & Flack 2008; Regnier et al. 2008; Westerman et al. 2008), we focus on the wider impact of insect seed predators on plant communities. We highlight several gaps in our knowledge about insect seed predator specificity, density-dependence and synchronization that would reward further study. We argue that, compared to other functionally important insect groups such as pollinators, seed predators have received relatively little attention in the context of the organization and functioning of natural and agro-ecosystems, and their interaction with global environmental change. We first deal briefly with demographic effects on individual species, which have been reviewed recently elsewhere (Maron & Crone 2006; Kolb, Ehrlén & Eriksson 2007). We then focus on the community-level effects of seed predators and in particular their potential for promoting and organizing plant diversity in diverse tropical forest ecosystems through the Janzen–Connell mechanism and related phenomena. We conclude by considering how the community-wide effects of seed predators might interact with anthropogenic impacts of habitat modification and climate change.
Demographic consequences of seed predation
Pre-dispersal insect seed predators frequently kill >90% of developing seeds (Crawley 1992; Fenner & Thompson 2005). However, high levels of mortality do not necessarily translate into significant effects on plant populations, because sources of mortality acting at later stages of plant development may be critical to limiting recruitment (Crawley 1992; Kolb et al. 2007). For example, when seeds are produced abundantly, seed predation may simply remove individuals that were doomed to die later through intraspecific competition or other density-dependent processes. Harper (1977) cites the example of a bruchid beetle introduced to control gorse Ulex europaeus in New Zealand. Although beetles killed 98% of gorse seeds, the remaining 2% were sufficient to maintain Ulex populations and allow their spread. Rather few studies have documented unambiguously the consequences of insect seed predators for plant recruitment. Marked effects of experimentally excluding insect seed predators on seedling establishment and population sizes in the subsequent generation have been demonstrated for annual plants (Louda 1982; Louda & Potvin 1995). Although Crawley (1992) considered that seed-limited recruitment was the exception rather than the rule, Turnbull et al. (2000) found evidence for seed limitation in 50% of the published studies they reviewed. Similarly, Kolb et al. (2007) found widespread, accumulating evidence for seed-limited recruitment, and inferred that pre-dispersal predation from insects is likely to affect plant population dynamics widely. Because of their importance for the dynamics of individual plant species, insect seed predators are of considerable interest in the context of controlling invasive plants and weeds (e.g. van Klinken & Flack 2008; Westerman et al. 2008). Partly for this reason, much of the existing evidence for the ecological effects of seed predators concerns annual plants rather than long-lived perennials such as trees, which we focus on in the remainder of this study. An additional reason for the shortage of information on trees is that the necessary experiments are challenging in terms of their spatial and temporal scales (Crawley 1992). More cohort studies that quantify the consequences of seed predation at different life stages for perennial plant demography are badly needed (e.g. Worthy, Law & Hulme 2006).
Seed predation and the structure of tree communities
tropical tree diversity: the janzen–connell effect
A single hectare of moist tropical forest can support more than 300 tree species (Valencia, Balslev & Pazy Mino 1994). A leading explanation for high plant diversity is the Janzen–Connell hypothesis (Janzen 1970; Connell 1971), which suggests that distance-dependent and density-dependent mortality by specialist enemies puts locally rare plant species at an advantage. While many recent studies have provided evidence that the Janzen–Connell effect plays a key role in the maintenance of plant diversity in the tropics (e.g. Gilbert, Hubbell & Foster 1994; Wills et al. 1997; Wills & Condit 1999; Harms et al. 2000; Peters 2003), the causes of this density-dependence remain less certain, and the causal link between density-dependent mortality and plant diversity remains untested. While significant and widespread density dependence has been documented for plants in a variety of age cohorts, the seed-to-seedling transition stage may be critical for determining patterns of diversity in larger-size classes (Harms et al. 2000; Lambers, Clark & Beckage 2002; Wright et al. 2005). Janzen suggested seed-predating insects as likely diversity-promoting agents in tropical forests (Janzen 1970), because of their specificity and their rapid aggregative and demographic responses to seed density. Vertebrate seed predators in contrast may be satiated and more likely to inflict positively density-dependent mortality over scales appropriate to the Janzen–Connell mechanism (Schupp 1992).
Knowing the range of species consumed by insect seed predators is key to understanding their likely impact on plant diversity. Under the Janzen–Connell hypothesis, seed predators or other plant natural enemies (e.g. pathogens: Freckleton & Lewis 2006) will have the greatest diversifying effects if they are host-specific, because this will enhance their ability to depress recruitment of locally abundant species. Not all seed predators are specialists, but it is difficult to be sure what the typical host plant range of seed predators is in tropical forests because few studies have systematically reared insects from seeds of a large number of plant species co-occurring at particular sites. There are a few notable exceptions (Janzen 1980; Nakagawa et al. 2003), but even these pioneering studies are somewhat restricted in terms of the taxonomic range of either the plants or the seed predators investigated.
The host-specificity of tropical forest pre-dispersal seed predators is particularly poorly known; indeed, the existing literature on seed predation for most tropical forest systems largely overlooks pre-dispersal seed predation, for the very good reason that seeds in the rainforest canopy are generally inaccessible for study. The pre-dispersal fate of seeds has often been inferred from data collected from seed traps on the ground at a time when many important seed predators may already have vacated the seeds (Nakagawa et al. 2005). Unless seeds are sampled throughout different stages of seed development, it is difficult to assess the degree of seed predator host-specificity and the degree of overlap in pre-dispersal seed predator assemblages associated with different tree species. For this reason, the high specificity of tropical seed predators remains a largely untested assumption. Studies quantifying both pre- and post-dispersal seed predation by insects across a large number of tropical tree species co-occurring locally are of high priority.
The effect of insect seed predators on community composition and diversity will also critically depend on the ways in which they respond to the density of their resources. If insect seed predators are to enhance the tree species diversity of tropical forests they must cause density- or distance-dependent seed mortality. Few studies have critically assessed the responses of insects to different seed densities, and most attempts to assess insect seed predator responses to host densities are based on observational data. Exclusion experiments provide a powerful way to identify the causal factors driving density-dependent seed mortality (Freckleton & Lewis 2006). Although exclusion experiments are routinely conducted to assess the role of post-dispersal vertebrate seed predators in tropical forests, few studies have examined patterns of plant recruitment following the exclusion of insect seed predators. While excluding pre-dispersal seed predators from the canopy may be challenging, post-dispersal insect seed predators are more amenable to experimental manipulation (e.g. Wolffsohn 1961).
Our understanding of the effects of insect seed predators on patterns of plant recruitment (and ultimately diversity) is further complicated by the fact that very little is known about the spatial scales across which insects respond demographically and behaviourally to the density of their resources. As a result, the spatial scales across which mechanisms such as satiation and density-dependent seed predation by insects typically operate are essentially unknown. It seems likely that the relevant spatial scale at which these effects operate will differ markedly for pre- and post-dispersal seed predators. Post-dispersal seed predators will respond to the local density of seeds on the forest floor and the distance from the parent tree, as envisaged by Janzen (1970). In contrast, density and distance responses of pre-dispersal insect predators, if they occur, will be on the scale of the seed crops of individual trees, or their isolation from adult conspecifics. These are further topics which require critical experimental investigation.
apparent competition and other indirect effects
The range of species consumed by insect seed predators is also key to understanding their likely influence on the local composition of ecological communities, because of their potential to link the dynamics of plant species that do not compete or otherwise interact directly. For example, where two plant species share one or more natural enemy species but do not compete directly for resources, changes in the abundance of one may influence the abundance of the other, through aggregative or demographic changes in the abundance of the shared natural enemy or enemies. This form of enemy-mediated indirect interaction is known as apparent competition (Holt 1977; Holt & Lawton 1993). Insect seed-predator mediated apparent competition between plants has not been investigated in tropical forests, but could have major effects on the wider structure of these communities. For example, shared seed predators may reduce the likelihood of congeneric species (which are more likely to share seed predator species) recruiting in close proximity. Such indirect interactions between plant species are amenable to experimental investigations that separate direct and indirect interactions through manipulation of species’ abundances (Chaneton & Bonsall 2000; Morris, Lewis & Godfray 2004). Experiments and analyses involving vertebrate seed predators suggest that such effects may be widespread (e.g. Veech 2000), but data are lacking for insect seed predators.
Quantitative food webs (e.g. Fig. 1) could provide a powerful tool for predicting indirect interactions between plant species, and more generally in understanding the consequences of adding or removing species in ecological communities. These webs have previously been applied in particular to host–parasitoid interactions (e.g. Memmott, Godfray & Gauld 1994; Müller et al. 1999, Lewis et al. 2002; Tylianakis, Tscharntke & Lewis 2007), and have not yet been constructed for insect seed predators (but see Prado & Lewinsohn 2004). Host–parasitoid networks are directly analogous to networks of interactions among seeds and their insect predators in that there is typically a one-to-one correspondence between mortality of one individual of the prey species and complete development to maturity of one or more individuals of the consumer species. These webs allow us to make predictions about pairwise indirect effects among species which are amenable to experimental testing in the field (e.g. Morris et al. 2004; Morris et al. 2005). At one extreme (Fig. 1a), if seed predators are host-specific, then the seed-predator food web will consist of a series of ‘compartments’ comprising a single tree species and one or more host-specific seed predators. The dynamics of individual compartments are unlikely to be closely linked in space or time; such a situation maximizes the potential for density and distance-dependence for individual species, but cannot generate apparent competition between tree species mediated by shared seed predators. At the other extreme (Fig. 1b), if seed predators are opportunistic generalists, they will cause uniform and consistent mortality across all potential hosts. In this scenario, density- and distance-dependent seed predator attack is less likely, and the identity of a plant's neighbours will again be irrelevant in determining the amount of mortality it suffers from seed predators. A third possibility (Fig. 1c) is that the food web is divided into a series of multi-species compartments, each containing two or more tree species (in the case of Fig. 1c, two, two and three) and their associated oligophagous seed predators. In reality, patterns of specificity are likely to be a complex mixture of these simplified scenarios (Fig. 1d), with a variable contingent of specialist seed predators, many oligophagous species restricted in their diet by plant phylogeny or functional traits such as seed size, and a minority of generalists.
Working in open habitats in the mountains of Brazil, Prado & Lewinsohn (2004) found a highly compartmentalized pattern, similar to Fig. 1c, for tephritid flies feeding within Asteraceae flower heads. However, the limited existing seed predator community data for tropical forests suggest a pattern more similar to that in Fig. 1d. Janzen (1980) reported high host specificity for beetles in Costa Rica, with 75% of the studied seed predators restricted to a single plant species; but he also found a minority of species with broader diets. Nakagawa et al. (2003), studying 51 species from 11 insect families associated with 24 dipterocarp trees in Sarawak, found a small number of specialists but many oligophagous species. Diet breadths were variable between years and lower in a year when few trees were fruiting, consistent with the hypothesis that insect seed predators switch onto alternative hosts in ‘famine’ years. Such patterns of host specificity in an environment where synchronous fruiting is thought to be a mechanism to satiate seed predators raise the intriguing possibility that predator satiation could occur within subsets of tree species sharing seed predators. A plant individual isolated from conspecifics or from heterospecifics which share oligophagous seed predators is unlikely to satiate them. We might then expect to see the counter-intuitive trend towards enhanced recruitment through predator escape where plants grow close to conspecifics and heterospecifics with similar seed predator faunas.
Human impacts on insect seed predation
Because of their potentially important role in maintaining diversity and structuring tropical forest communities, as discussed above, any human activity influencing seed predator abundances or identities may have far-reaching consequences for wider ecological assemblages. We next discuss how two forms of human impact considered as the most significant threats to diversity globally (Sala et al. 2000), habitat modification and climate change, are likely to affect interactions between insect seed predators and plants in tropical forests.
anthropogenic habitat disturbance and degradation
Forest degradation and fragmentation have profound consequences for the diversity, composition and structure of associated plant and animal communities. A large body of literature considers the effects that changes to vertebrate seed predator assemblages may have on the demography of individual plant species and the wider structure of these communities (e.g. Dirzo & Miranda 1990; Asquith, Wright & Clauss 1997; Holl & Lulow 1997; Forget, Rankin-De Merona & Julliot 2001; Beckage & Clark 2005; Paine & Beck 2007). Far fewer studies have investigated insect seed predation in the context of different forms of anthropogenic disturbance. Thus, while studies documenting significant effects of vertebrate seed predation on plant dynamics and community structure and diversity dominate the literature, the importance of insects may have been significantly underestimated, for at least five reasons.
First, there have simply been far fewer studies focusing on insect seed predation. Second, even where insect seed-predation is studied, it may be that many of the seeds ‘removed’ and therefore scored as predated by vertebrates are already infested with insect seed predators and therefore doomed in any case; careful experiments are needed to quantify the degree of complementarity between vertebrate and invertebrate predation and (by inference) the extent to which one of the groups might be able to compensate for the other in the event of major loss or change of biodiversity. Third, most studies are for palms and other large-seeded plants that provide particularly attractive resources to vertebrate seed predators. Fourth, some forms of anthropogenic disturbance such as hunting can have unexpected cascading effects on insect seed predators (Stoner et al. 2007) which may not be detected or may be erroneously ascribed to the direct effects of vertebrates in vertebrate-focused studies. Finally, for practical reasons, there have been very few studies of pre-dispersal seed predation, almost certainly leading to a huge underestimate of the seed loss inflicted on trees by insects and a failure to detect significant effects on population dynamics (Kolb et al. 2007).
We argue that the probable effects of habitat disturbance in altering patterns of seed predator are likely to be at least as great for insects as for vertebrates, for two reasons. First, insects in general are highly sensitive to changes in the abiotic environment such as altered light or humidity levels (Lewis & Basset 2007). Thus, even if the quantity of the resources available to them remains unaffected by logging, fragmentation or other anthropogenic disturbance, their abundances and impacts may respond dramatically. Second, the patterns of specificity and responses to resource density discussed above mean that the effects on individual insect species and assemblages are more likely to have widespread repercussions for plant assemblages. This applies equally whether existing species decline, go extinct, or become more abundant. While the consequences of anthropogenic changes such as logging or fragmentation for other functionally important insect groups such as pollinators has received widespread attention (reviewed by Steffan-Dewenter & Westphal 2008), there have been few seed predator studies (e.g. Galetti et al. 2006).
Some authors (e.g. Sala et al. 2000) have argued that anthropogenic climate change will have relatively little effect on tropical forests, compared to temperate ecosystems. However, Bazzaz (1988) suggests that tropical forests will be particularly sensitive to climate change, because phenological events such as fruiting and flowering are highly tuned to climatic conditions. Synchrony between interacting species is thus at risk of being broken, with implications for species’ dynamics and abundance (Buse & Good 1996; Sparks & Yates 1997). While Bazzaz focuses on the tightly co-evolved mutualistic interactions of pollination and seed dispersal (Memmott et al. 2007), relatively specialized antagonistic interspecific interactions such as insect seed predation should also be of concern. In temperate systems, there has been considerable research into whether altered phenologies may decouple the dynamics of interacting consumer and resource species (e.g. Watt & Woiwod 1999; Visser & Both 2006). A consensus is emerging that phenological responses will typically differ in plants, insects and vertebrates (Voigt et al. 2003; Parmesan & Yohe 2005), making disconnections across trophic levels likely for specialized consumer-resource interactions. However, little consideration has been given to equivalent tropical forest systems.
Many perennial plants produce seed crops that are highly variable from year to year, often going for multiple years without producing any seeds at all, and then synchronizing seed production with local conspecifics in an apparent attempt to satiate seed predators (Kelly 1994). Such ‘mast fruiting’ is typical of many trees in temperate ecosystems (e.g. beech Fagus sylvaticus in Europe) as well as certain herbaceous plants and also some tropical trees, notably members of the Dipterocapaceae. Seed predators may exert strong selective pressure against unsynchronized flowering. For example, Augspurger (1981) found that Hybanthus prunifolius shrubs manipulated to flower out of season in Panama suffered greatly increased seed predation.
Climate change has considerable potential to alter patterns of temporal synchronization among species, particularly for masting species where the cue for masting is climatic. For example, in New Zealand tussock grasses (Chionocloa spp.) unusually high summer temperatures result in mast fruiting in the subsequent year (McKone, Kelly & Lee 1998). Elevated temperatures under future climate change might therefore lead to more regular flowering, allowing specialized seed predators to build up large populations. Although this example is for a temperate ecosystem, equivalent changes may occur in tropical forests, for example, in some forests of South-east Asia where mast-fruiting Dipterocapaceae can dominate forest biomass. Masting is thought to be triggered by El Niño climate conditions (Curran et al. 1999), and in some cases, occurs synchronously across nearly all dipterocarp species, as well as in many other species in a phenomenon known as ‘general flowering’ (Curran et al. 1999). During intervening years, flowering and seed production are considerably lower and at least a third of species do not flower at all. Again, general flowering seems to be a mechanism to reduce seed predation through predator satiation (Janzen 1969; Sun et al. 2007). Given the likely importance of seed predation in determining dipterocarp recruitment, it is not surprising that seed predators are believed to have an important role in forest regeneration (Lyal & Curran 2003). Over the last two decades, El Niño droughts have become longer and more severe, perhaps linked to global climate change. We can only speculate as to the effects associated with changes in the general flowering regime will have on interactions between insect seed predators and their mast-fruiting prey.
Although we have concentrated on the particular case of insect seed predators of tropical forest trees, many of the issues we raised will be of equal relevance in other, applied contexts. For example, patterns of seed predator specificity and food web organization are crucial to understanding how novel plants or seed predators – such as those introduced as biological control agents – will impact on the wider ecological community through both direct and indirect effects (non-target effects; e.g. Rose et al. 2005). Similarly, the extent and form of seed predator density dependence is directly relevant for planning management strategies for agricultural weeds (Westerman et al. 2008). While an active research agenda seeks to understand the functional consequences of changes to insect pollinator abundance and diversity (reviewed by Kremen & Chaplin-Kramer, 2007), we believe that more information is badly needed on insect seed predators and their role in both natural and managed ecosystems. The contributions to this special section are an important step towards filling this gap.
We thank Rob Freckleton for inviting us to contribute this paper, and for comments that greatly improved the manuscript; and Chris Lyal for helpful discussions. Owen Lewis is supported by a Royal Society University Research Fellowship and Sofia Gripenberg by a grant from the Academy of Finland.