Host specificity of insect herbivores in tropical forests

Vojtech Novotny, Yves Basset

Abstract

Studies of host specificity in tropical insect herbivores are evolving from a focus on insect distribution data obtained by canopy fogging and other mass collecting methods, to a focus on obtaining data on insect rearing and experimentally verified feeding patterns. We review this transition and identify persisting methodological problems. Replicated quantitative surveys of plant–herbivore food webs, based on sampling efforts of an order of magnitude greater than is customary at present, may be cost-effectively achieved by small research teams supported by local assistants. Survey designs that separate historical and ecological determinants of host specificity by studying herbivores feeding on the same plant species exposed to different environmental or experimental conditions are rare. Further, we advocate the use of host‐specificity measures based on plant phylogeny. Existing data suggest that a minority of species in herbivore communities feed on a single plant species when alternative congeneric hosts are available. Thus, host plant range limits tend to coincide with those of plant genera, rather than species or suprageneric taxa. Host specificity among tropical herbivore guilds decreases in the sequence: granivores>leaf-miners>fructivores>leaf-chewers=sap-suckers>xylophages>root-feeders, thus paralleling patterns observed in temperate forests. Differences in host specificity between temperate and tropical forests are difficult to assess since data on tropical herbivores originate from recent field studies, whereas their temperate counterparts derive from regional host species lists, assembled over many years. No major increase in host specificity from temperate to tropical communities is evident. This conclusion, together with the recent downward revisions of extremely high estimates of tropical species richness, suggest that tropical ecosystems may not be as biodiverse as previously thought.

Keywords:

1. Introduction

Plant–phytophage food webs represent more than 40% of global terrestrial biodiversity (Price 2002), most of it concentrated in tropical forests. Arguably, their diversity makes tropical forests a logical starting place for the study of host specificity, rather than the impoverished forests of northern temperate regions. Estimates of host specificity (or lack thereof) played an important role in the recent controversy surrounding global estimates of insect species richness, ranging from 30 million (Erwin 1982) to 4–6 million (Novotny et al. 2002b). Narrow insect host specificity was also invoked as the mechanism maintaining the high species richness of insects in tropical forests (Basset 1992). Further, the Janzen–Connell hypothesis assumed that specialized insect herbivores or other pests reduce recruitment near conspecific plants, thus demographically favouring rare species and maintaining high plant diversity in tropical forests (Wright 2002). Finally, host specificity of insect herbivores is also central to the study of indirect effects between different plant species, mediated by their shared herbivores (Godfray et al. 1999; Novotny et al. 2004b).

Janzen (1983, p.167) characterized his review of tropical food webs as ‘rich in extrapolation and conjecture, held in place by very few data points’. Later, Basset (1992) remarked on the scarcity of information on the host specificity of tropical herbivores, while a recent review (Godfray et al. 1999) discussed possible developments in this discipline. Recently, there has been progress in the study of plant–herbivore interactions as the emphasis of tropical research has graduated from the diversity of species to that of interspecific interactions. This can be illustrated by a shift in focus in two recent monographs on tropical insects: namely, Stork et al. (1997) and Basset et al. (2003). Studies on phytophages used to be reliant on insect distribution data, obtained by canopy fogging, light and Malaise trapping and other mass collecting methods (reviewed in Basset 2001a). This approach is being superseded by studies that rely on observation in situ, experimental verification of insect feeding, and insect rearing, which are reviewed here.

2. Measuring host specificity

Herbivores are often classified as mono-, oligo- and polyphagous, or as generalists and specialists. Several meanings have been attached to these terms by different authors (Symons & Beccaloni 1999), since no universal criterion that distinguishes specialists from generalists exists. They probably cannot be resurrected for any specific use; instead, they should be redefined whenever used. The terms species-, genus- and family specialist (Barone 1998) are used for herbivorous species feeding on a single plant species, genus and family. They represent a potentially useful set of host specificity definitions, used in the present review.

Plant–herbivore matrices from tropical forests are characterized by a large number of rare plant–herbivore interactions. For example, as many as 40–50% of trophic links can be established on single feeding individuals (Nakagawa et al. 2003; Novotny et al. 2004a). Thus, estimates of the number of host species used by herbivores are dependent on sample size and are unreliable even for a large sampling effort (Novotny et al. 2002a, 2004a). The host plant lists that indiscriminately include all host plant records become rapidly dominated by marginal hosts and errors committed by either herbivores or ecologists. A more robust definition of monophagous species, that is, species with at least 90% of individuals feeding on a single host species (Thomas 1990; Tavakilian et al. 1997; Novotny et al. 2002a) may be more legitimate.

Analyses of host specificity need to consider phylogenetic relationships of the host plants. Thus far, these have been approximated by supraspecific taxonomic ranks. Host range is then described as the number of host plant species and various supraspecific taxa (Symons & Beccaloni 1999). As information on plant phylogeny accumulates, this imprecise approach will become superseded by host specificity measures based on host plant phylogeny, defined by cladograms (Symons & Beccaloni 1999; Webb et al. 2002). New phylogenetic measures of host specificity, including the clade dispersion index (Symons & Beccaloni 1999) and the net relatedness index (Webb et al. 2002), quantify overall clustering of host taxa on a phylogenetic tree. Although promising, these indices have not been widely applied (Weiblen et al. in press). Phylogenetic constraints on host plant selection may be examined as a relationship of species turnover between herbivore communities and the phylogenetic distance of their host plant species (Novotny et al. 2002b).

Indices describing the distribution of herbivore individuals among host plant species represent an alternative approach to host specificity measurement, which ignores plant phylogeny but responds to herbivore abundance on various hosts and, as such, is relatively unaffected by rare host plant records. These include Lloyd's index (Novotny & Basset 1998) and Hurlbert's (1971) probability of interspecific encounter index (the latter was recently modified and used for food webs; e.g. Müller et al. 1999; Lewis et al. 2002; Morris et al. 2004). Local plant availability can also constrain host plant use, particularly in tropical forests where many genera and families are locally monotypic (e.g. Condit et al. 1996). Therefore, estimates of host specificity should always be considered with regard to the number of potential alternative hosts that are available.

Host specificity patterns of individuals in herbivore communities have received so little attention that even the simple but fundamental question of whether a herbivorous insect selected at random from tropical vegetation is likely to be a generalist or a specialist has yet to be answered (Novotny et al. 2004b).

Host specificity estimates depend on sound taxonomy of herbivores and plants. Modern revisions of tropical insect faunas often synonymize almost as many species as they describe (Scoble et al. 1995), while molecular research may reveal specialized cryptic species in what appeared to be a single generalist species (Berkov 2002). The analysis of cytochrome oxidase I (‘DNA bar coding’) may be particularly useful for species-level identifications in insects (Hebert et al. 2003).

Usually, studies of herbivores in tropical forests sample only part of the local flora—typically less than 100 species (Marquis 1991; Basset 1996; Barone 1998; Ødegaard 2000a; Novotny et al. 2002a). Often, the flora studied is structured according to plant phylogeny (e.g. congeneric species: Marquis 1991; allofamilial species: Basset 1996) or ecology (e.g. lianas versus trees: Ødegaard 2000a; early versus late successional species: Leps et al. 2001) and selected from locally common species. More extensive studies sampling a larger proportion of local vegetation (Janzen 1980, 2003; Tavakilian et al. 1997) are qualitative surveys that catalogue host–plant relationships, but do not monitor sampling effort and population densities of herbivores on individual plant species. A study that is both quantitative and comprehensive is still awaited.

The number of tropical multispecies studies of host specificity is limited (table 1). Regrettably, groundbreaking studies, rather than promoting analogous research in other ecosystems and geographical areas, tend to petrify into solitary textbook case studies. For instance, Janzen's (1980) pioneer survey of granivorous beetles from local vegetation still remains the only such comprehensive study of this herbivore guild from tropical forests.

View this table:
Table 1

Host specificity of insect herbivores in tropical forest communities.

(Wherever possible, data for feeding guilds and insect orders (Coleoptera, Diptera, Hemiptera, Lepidoptera, Orthoptera and Phasmatodea, abbreviated by the first letter of their name) are separated. All analyses are based on larval (L) or adult (A) feeding, noted in the L/A column. Host specificity is estimated as the percentage of herbivore species feeding on a single host within a set of three or more plant species that are congeneric (Spp), from different confamilial genera (Gen), and from different families (Fam). Only datasets that included 10 or more herbivore species are considered. Results from multiple datasets originating from the same study are averaged. As far as possible, (A) confamilial genera and different families were each represented by only one plant species (compliance noted by 1 in the A column), (B) only herbivore species with 10 or more feeding records from the analysed set of plants were included (1 in the B column), and (C) all study plants were sampled with equal sampling effort (1 in the C column).)

3. Host specificity and host plant traits

The host specificity of herbivores feeding on a particular plant species may be regarded as a one of the latter's characteristics and analysed in the context of its other traits. Rather surprisingly, such analyses are rare, which contrasts with the attention paid to a similar variable: the number of herbivores feeding on different plant species (Lewinsohn 1991; Basset 1996; Basset & Novotny 1999; Kelly & Southwood 1999).

Temperate zone studies along successional series documented a positive relationship between plant predictability and host specificity of their herbivores, since ephemeral plants from early successional stages are colonized by polyphages (Brown & Southwood 1983). However, tropical succession often starts with pioneer trees, rather than with annual herbs. Already, these represent a permanent and predictable resource for insects with no advantage for polyphages (Leps et al. 2001). Polyvoltine insects in continuous tropical forests may be able to colonize even rare perennial plants. Thomas (1990) reported on rapid colonization of isolated plants in a rainforest, but further studies would need to examine plants isolated from their conspecifics by a distance larger than the 50 m used in his study. Shanahan et al. (2001) reported rapid colonization by fig wasps of a new volcanic island 55 km offshore.

Tree species from mature forests invest more in anti-herbivore defences than early successional species (Coley & Barone 1996). This is particularly so in mature leaves, which also suffer lower herbivore damage in late, rather than early, successional trees. Although these important patterns in anti-herbivore defence were observed 20 years ago (Coley 1983), it remains unclear how they are associated with host specificity patterns in herbivore communities from early and late successional trees. Young and mature leaves in tropical trees are so different, physically and chemically, that they represent distinct resources for insect herbivores. Further, there seem to be two anti-herbivore strategies, ‘defence’ and ‘escape’, employed by tropical trees to defend their young leaves (Kursar & Coley 2003). It is probable that herbivores respond by different levels of host specificity. In particular, the escape strategy may select for polyphagy in herbivores, but no data are available to confirm this assumption.

Plant growth form determines host specificity of herbivores in temperate areas (Ward et al. 1995), but most studies from the tropics concentrate only on shrubs and trees. Comparative studies of herbivores from other growth forms (e.g. herbaceous or epiphyte species) would be more difficult but also more rewarding, as demonstrated for example by a recent study of lianas (Ødegaard 2000a). Herbivore host specificity probably differs among plant lineages (cf. an example from temperate plants in Frenzel & Brandl 2001). Tropical woody vegetation provides an excellent, albeit neglected, opportunity for the analysis of the host specificity patterns in all major plant lineages, using plant species from the same growth form.

In addition to plant traits effecting ‘bottom–up’ influences on host specificity via secondary metabolites and other anti-herbivore defences, predators and parasitoids associated with each plant species may exercise ‘top–down’ control. Specialized herbivores appear to be less susceptible to predators in tropical forests (Dyer 1995), but tend to suffer higher parasitism rates (Gentry & Dyer 2002).

Interpreting differences in herbivorous communities feeding on different plant species is always difficult as they differ in multiple traits. Therefore, it is surprising how few studies have attempted to separate historical and ecological factors by comparative analyses of herbivore communities feeding on the same, or at least closely related, plant species exposed to well-defined differences in their environment or different experimental treatments. This approach has been used to study herbivore communities and their response to disturbance caused by selective logging (Basset et al. 2001), stratification between young and mature (Basset 2001b) and differences between living and dead (Ødegaard 2004) trees. However, the comparative approach has much greater potential, for example, for the studies of herbivore communities along successional, altitudinal or other environmental gradients (Novotny et al. in press). Replicated studies of herbivores feeding on identical or related plant species in various geographical areas can provide useful insights into the process of assembly of herbivore communities from regional species pools. Regrettably, a pioneering study on the cosmopolitan bracken fern (Lawton et al. 1993) has not been followed by similar studies concentrating on other widely distributed plant species or genera. Species-rich pan-tropical genera, such as Ficus, Piper or Psychotria, are promising candidates for such studies (Dyer & Palmer 2004).

4. Host specificity in herbivorous guilds and communities

Only a small proportion (27%) of species in herbivore communities feed on a single plant species when alternative congeneric hosts are available (table 1). By contrast, the proportions of species feeding on a single plant genus (48%) and family (58%) are higher and not statistically different from one another. Thus, limits of host plant range tend to coincide with those of plant genera, rather than those of families or higher taxonomic ranks. However, this analysis employed taxonomic ranks as a substitute for plant phylogeny and also ignored numerous methodological discrepancies among individual datasets and is therefore preliminary.

In theory, host specificity comparisons between different herbivore taxa and guilds are straightforward since all herbivores may be sampled simultaneously from identical study plants and sites. However, while some studies include herbivores from several insect orders (table 1), studies of more than one guild on the same hosts are lacking from tropical forests, probably because each guild requires a specific sampling method.

Concealed herbivore guilds are more specialized than those that feed externally on temperate vegetation (Mattson et al. 1988). In tropical forests, the average percentage of family specialists decreases among guilds in the sequence: granivores (99%)>leaf-miners (96%)>fructivores (83%)>leaf-chewers (56%)=sap-suckers (56%)>xylophages (24%)>root-feeders (10%) (see table 1). The proportion of genus specialists (but not species specialists) follows the same pattern. With the exception of xylophages, which feed on dead plant material, all endophytic guilds are thus more specialized than the external feeders. These conclusions remain tentative since replicated studies produced inconsistent host specificity estimates in several guilds: xylophagous beetles, leaf-chewing beetles, granivorous beetles and leaf-chewing Orthoptera (table 1). Some of these differences may have been caused by different methods of study, but they emphasize the need for replicated studies.

Although it is believed that larval host plant ranges are more restricted than those of adults, rigorous larval–adult comparisons are rare. This is a notable problem in herbivorous beetles where larvae and adults often belong to different guilds. The species richness (Basset 2001a) and host specificity (Ødegaard 2000b) of leaf-chewing beetles were studied extensively in tropical forests. However, adults predominate in this guild, while their root-feeding, xylophagous and granivorous larvae are mostly ignored.

Trends in the diversity of species along environmental gradients (e.g. latitudinal or altitudinal) are popular themes for study, whereas analogous trends in the diversity of interactions, including plant–herbivore trophic links, are not (Ollerton & Cranmer 2002). Although there is extensive literature comparing species richness in herbivorous communities at different altitudes (Brehm et al. 2003), host specificity comparisons between lowland and montane tropical forests are few (table 1) since lowland forests remain the most popular topic for study (Basset 2001a). Steep altitudinal gradients in temperature, rainfall and the abundance of ants (important as predators and mutualists of herbivores) across small geographical distances provide yet unexplored opportunities for the study of host specificity determinants in the tropics.

Whether there is a latitudinal gradient in host specificity between temperate and tropical forests is surprisingly hard to assess given a methodological dichotomy between tropical and temperate research. Almost all information on the host plants of tropical herbivores comes from local community studies on selected plant species (table 1), while even incomplete regional host plant lists are available for only a few taxa (e.g. Fiedler 1995; Flowers & Janzen 1997). By contrast, regional host plant lists for temperate herbivores, detailed particularly for Great Britain and central Europe, have been compiled for over two centuries, while quantitative community studies that include feeding experiments and rearing of larvae (e.g. Futuyma & Gould 1979) are rare. Incongruent methodology represents a serious problem as both host specificity estimates based on community and regional data have their own sets of biases (Ward 1988).

Given that reliable regional host species lists are unlikely to be forthcoming for tropical habitats, there is a need for renewed effort in collecting quantitative host specificity data for herbivorous insects from temperate vegetation. Ideally, such studies should produce quantitative food webs (Godfray et al. 1999) that measure the abundance or biomass of both plant and herbivore species as well as the intensity of their trophic interactions. It is important that food web studies are replicated as any tropical/temperate trends are only likely to emerge by analysing what may appear as unnecessary ‘superabundance’ of datasets.

There is no marked difference in host specificity between tropical and temperate butterflies (Fiedler 1995, 1998), although particular lineages may be less (Lycaenidae: Polyommatini; Fiedler 1995) or more specialized (Papilionidae: Scriber 1988) in the tropics than in temperate regions. Bark beetles (Coleoptera: Curculionidae: Scolytinae; Beaver 1979) and treehoppers (Hemiptera: Membracidae; Wood 1984) appear to be less specialized in the tropics than in temperate regions, while a community study of temperate caterpillars (Futuyma & Gould 1979) shows lower host specificity than analogous data from the tropics (Barone 1998; Novotny et al. 2002a). The limited evidence available indicates that there may be no major increase in host specificity from temperate to tropical herbivore communities. This conclusion, and the recent downward revisions of extremely high estimates of tropical species richness (Basset et al. 1996; Ødegaard 2000b; Novotny et al. 2002b), suggests that tropical ecosystems, although different from those temperate in many respects, may not be as species rich as thought previously.

Empirical study of host specificity patterns is only the first step towards the ultimate goal of explaining host plant use by evolutionary and ecological factors. The evolutionary dynamics of host plant selection range from cospeciation between plants and herbivores, through sequential speciation of herbivores following the evolution of their hosts, to multiple colonizations of unrelated plants with similar chemical or other traits (e.g. Becerra 1997; Weiblen 2004). These evolutionary processes generate regional species pools of herbivores that serve as sources for local communities. The composition of communities may be tested against an estimate derived from a random selection from the regional species pool. This approach, already applied to plants (Webb 2000), could be useful in the study of herbivorous communities.

5. Conclusions

The present review describes the rather unsatisfactory state of current research where the lack of empirical data prevents us from making pan-tropical generalizations on host specificity patterns and their response to the key environmental variables. Rich empirical data are also essential as a source of inspiration for new hypotheses (LaFrankie 1996). As already argued by Godfray et al. (1999), there is a serious need for comprehensive quantitative surveys of plant–herbivore food webs in tropical forests that include a sampling effort which is 1–2 orders of magnitude greater than has been customary to date (Novotny et al. 2002c). Such surveys require a transition from solitary researchers to teams of researchers where ecologists work in synergy with taxonomists, and are supported by local field assistants and parataxonomists (Basset et al. 2004).

Acknowledgments

We thank Héctor Barrios, Lukas Cizek, Lee Dyer, Owen Lewis, Frode Ødegaard and Rapo Pokon for providing their unpublished data. Scott Miller, Lee Dyer and Neil Springate commented on the manuscript. Funding from US National Science Foundation (DEB-02-11591), Czech Academy of Sciences (A6007106, Z5007907), Czech Ministry of Education (ME646), Czech Grant Agency (206/04/0725, 206/03/H034) and Darwin Initiative for the Survival of Species (162/10/030) contributed to our research on host specificity.

Footnotes

    • Received August 5, 2004.
    • Accepted November 23, 2004.

References

View Abstract