Dyer and Floyd

Determinants of predation on phytophagous insects:

the importance of diet breadth.

1Lee A. Dyer and 2Ted Floyd

1Proofs and other communications should be addressed to:

Environmental, Population, and Organismic Biology

Campus Box 334

University of Colorado

Boulder, CO 80309-0334

EMAIL: dyer@rintintin.colorado.edu

FAX: (303) 492-8699

2Current address:

Pesticide Research Laboratory

Pennsylvania State University

University Park, PA 16802

Summary

To evaluate the role of predation in the evolution of diet specialization and to determine the effectiveness of various larval defenses, we offered lepidopteran larvae to colonies of the tropical ant Paraponera clavata. We recorded behavioral and physical characteristics of prey items and used log-linear models to analyze their importance as deterrents to predation by P. clavata. The most important determinant of probability of prey rejection by P. clavata was a prey's diet breadth; specialists were rejected by the ants significantly more than generalists. Other less important, but significant, predictors of prey rejection included ontogeny, morphology and chemistry. Late instar caterpillars were rejected more frequently than early instars, hairy caterpillars were rejected more frequently than caterpillars with other morphologies, and one caterpillar species with an unpalatable extract was rejected more frequently than two species with palatable extracts.

Key Words

diet specialization, host plant chemistry, Lepidoptera, Paraponera clavata, predation.

Introduction

Host ranges of most phytophagous insects are limited to a small proportion of potentially available plant taxa (Futuyma 1976, Jaenike 1978, Holloway and Herbert 1979, Barbosa 1988, Futuyma and Moreno 1988, Moran 1988, Weiss and Berenbaum 1989, Berenbaum 1990). Such specialization is typically viewed as a consequence of either the top-down effect of natural enemies or the bottom-up effect of host plant chemistry (see Strong 1988).

Bottom-up models rely on the assumption that host plant availability and apparency are the most important determinants of host range (Ehrlich and Raven 1964, Levins and MacArthur 1969, Feeny 1976, Rhoades and Cates 1976, Cates 1980, Berenbaum 1990). The top-down view holds that specialists are better able than generalists to utilize plant chemical or physical characteristics as defenses against enemies (Brower 1958, Atsatt 1981, Bernays 1988, Bernays and Cornelius 1989, Bernays and Graham 1989, Jones et al. 1989). For example, physical characteristics such as trichomes or a cryptic background have been shown to be important in providing refugia from predation and parasitism (Brower 1984, Gross and Price 1988, Hay et al. 1989, Kareiva and Sahakian 1990). If host plants offer only these physical safeguards to specialist herbivores, then predators should show no preference between generalists and specialists offered to them in the absence of host-plants. Bernays (1988), however, has shown that certain hymenopteran predators prefer generalist caterpillars over specialists in the absence of host-plants. These results suggest that specialists have gained some physiological benefit from their host plants or that they have sequestered plant secondary compounds, which, in general, have been shown to be important in deterring predators (Eisner 1970, Duffey 1980, Price et al. 1980, Brower 1984, Bowers and Larin 1989, Bowers 1990).

To test the assumption that diet breadth constitutes a major determinant of predation, we studied predation on lepidopteran larvae by the ponerine ant Paraponera clavata Emery (Hymenoptera: Formicidae). For three of the caterpillar species in our experiments, we also tested the assumption that plant chemistry is the mechanism which makes diet breadth an important determinant of predation. For example, herbivores that are more specialized may have evolved specific physiological adaptations that enable them to recycle plant defensive chemicals and accumulate them over time (see Bowers 1992). Alternatively, reduction in diet breadth may lead to greater casual sequestration of chemicals, imparting an effective chemical defense (Jones et al. 1989, Blum et al. 1990). We did not design our experiments to differentiate between these alternatives, but by offering caterpillars to ants in the absence of host plants we were able to eliminate physical benefits (such as trichomes) derived from the host plant.

P. clavata is a generalist predator (Wilson 1971), an opportunistic nectivore (Young 1977), and possibly an herbivore and scavenger (Janzen and Carroll 1983, Breed and Bennet 1985). Colonies of P. clavata are located at the bases of large trees in lowland tropical rain forests, and foraging individuals use these and adjacent trees to reach the canopy and surrounding areas where they forage (Young and Hermann 1980). We used 5 colonies of P. clavata and 36 species of lepidopteran prey to address the following questions. Are specialist herbivores better protected than generalists against predation by P. clavata? If specialists are better protected than generalists, which characteristics make them less attractive to predators?

Materials and Methods

We conducted all work at the La Selva Biological Reserve in Costa Rica during the period 1 - 30 January 1992. This site is located at 10ø 25' N 84ø 05' W on the Caribbean Slope of the country. Mean annual precipitation is 4000 mm, and the site is approximately 100 m above sea level. La Selva is comprised of primary and secondary lowland evergreen rain forest and is noteworthy for its great biodiversity (see Janzen 1983). Our experiments consisted of obtaining larval lepidopterans (hereafter, "caterpillars") and their host plants and observing consumption of caterpillars and host plant extracts and caterpillar extracts by P. clavata. We completed our work during La Selva's "dry" season (see Janzen 1983), during which caterpillars are comparatively scarce. We analyzed correlates of predation with log-linear models.

Caterpillars (n = 36 species) and their plant substrates were obtained and identified to the lowest taxon possible (Table 1). Caterpillars were either collected in the wild at La Selva or purchased from the Finca Mariposa, a commercial butterfly farm near Alajuela. Voucher specimens of many of the caterpillars and of some of the adults were deposited at the Instituto Nacional de Biodiversidad de Costa Rica (INBio) in San Pedro. Caterpillars were held in captivity for several days to verify that they actually were using their presumed host plants as food resources. Each caterpillar was evaluated for the following traits: coloration, diet breadth, developmental stage, morphology, and size. All of these characteristics, along with prey behavior, may act as important anti-predatory traits (reviewed by Edmunds 1974, Devries 1987, Evans and Schmidt 1990).

We used a taxonomic definition of specialization similar to definitions used by Bernays (1988) and others (Futuyma 1976, Price 1983); caterpillars known to feed on less than two families of plants (according to DeVries 1987), or caterpillars of unknown diet breadth that were found feeding on only one plant species, were classified arbitrarily as "specialists" (n = 183 individuals), while caterpillars found feeding on plants in greater than five families were classified as "generalists" (n = 82 individuals). The total number of host plant species (rather than families) for caterpillar species as well as for individual caterpillars may be better indicators of those caterpillars' diet breadth, but those data are not available for a majority of lepidopteran larvae in Costa Rica (Price 1991).

While using caterpillars with unknown diet breadths is problematic, our characterization of diet breadth is conservative and takes into account known diet-breadth patterns at La Selva. Our null hypothesis assumes that specialists are not protected more than generalists, so erroneously characterizing caterpillars as specialists would only contribute to accepting the null hypothesis. Furthermore, the chance of making such a mistake is quite low, since most herbivores at La Selva have very narrow diet breadths (Marquis and Braker 1991). For example, 75% of the butterfly species at La Selva with known host plant ranges are monophagous (only one host plant species), and over 95% feed on plants within one family (Marquis and Braker 1991).

The ontogeny category referred to the developmental stages of the caterpillars and included: "early" (first through third instar) or "late" (fourth instar and later). Most caterpillars were reared in the lab after being collected and were offered to the ants as later instars. We offered both early and late instars to the ants if we could collect or buy enough replicates, and we used that subset of the data for the models that included ontogeny (see Table 1).

Morphology categories were: "spiny," "hairy," or "glabrous." These categories were based on visual inspection of the caterpillars. "Spiny" caterpillars had spines harder than the cuticle and at least 2 mm long. Caterpillars that possessed hairs and spines were rated as "hairy" only if more than 50% of their cuticle was covered with hair that was at least 5 mm in length. "Glabrous" caterpillars had no hairs or spines.

Coloration was also based on visual inspection of the caterpillars and included these categories: "brightly colored," "visually cryptic" or "other." Since P. clavata foragers seem to rely more on chemical than visual cues for locating prey (see Breed and Bennet 1985, Janzen and Carroll 1983), the coloration category was included to see if purported correlations between coloration and palatability (see Cott 1940) were valid for this invertebrate predator.

The size statistic was based on the length (in mm) of a caterpillar stretched to the point that slight pressure was felt on the cuticle. Size categories were: size ó 15 mm ("small"), 15 mm < size ó 38 mm ("medium"), and size > 38 mm ("large"). The size categories were pooled in the preceding manner (based on the mean and standard deviation of all the lengths) to ensure sufficiently large cell sizes for data analysis.

For all predation experiments, we used 5 P. clavata colonies at the bases of trees in the Holbridge Arboretum at the La Selva Biological Reserve. Most experiments were conducted during the period 1600-0100 because P. clavata is crepuscular and/or nocturnal. Artificial illumination of the foraging trails did not seem to affect foraging or attack behavior by P. clavata. Using forceps, we placed live caterpillars on the main foraging trails within 3 m of colony entrances. If a caterpillar walked away or dropped from the main foraging trail, we returned it to maximize the chance of encountering a predator. We observed all caterpillar-ant interactions until either the caterpillar was carried into the colony or at least ten ants had physically encountered and then rejected the caterpillar. In pilot studies, we found that caterpillars rejected by 10 ants were never taken by the ants. An interaction was considered a "rejection" if the ant clearly touched the caterpillar (either with antennae or mandibles) and then continued foraging, retreated, or initiated but did not continue an attack. Caterpillar species were presented evenly to the colonies; for species which we were unable to collect 5 or more specimens, assignment to a particular colony or colonies was randomized. Each colony was exposed to 53 individual caterpillars (at least 36 specialists per colony and at least 16 generalists per colony).

The response factors recorded were number of prey rejections and behavioral response of a caterpillar. Rejections were pooled as follows: no rejections, some rejections (1-6 rejections), or complete rejection (10 rejections). Caterpillar behavioral responses were ranked as follows: no behavioral defense; thrashing and attempting to attach to foraging substrate; thrashing and biting or vomiting. All experiments were recorded with a Canon H850 Camcorder to permit thorough analysis of the encounters.

Additionally, we made crude water and acetone extracts of 3 caterpillar species and their host plants (Table 2). Host plants were dried at 40ø C overnight, and 0.3 g of dried plant material was added to 10 ml solvent (boiling water or acetone). In having dried the host plants at 40ø C, our plant extracts may have lost volatile terpenes and phenolics originally present in the plant material. The mixture was agitated frequently by hand and evaporated to 5 ml. Next, 0.3 ml of this solution was added to a micro-centrifuge tube into which was added a 50% v/v sugar water solution. Live caterpillars were starved for 24 hours, ground with a scalpel, and 50 mg of ground, homogenized caterpillar was placed in 10 ml of solvent. Both water and acetone extracts were made for all caterpillars and plants. As with the plant extracts, 0.3 ml of this solution was added to a micro-centrifuge tube into which was added a 50% v/v sugar water solution.

Caterpillar and host plant extracts, as well as sugar water controls, were offered to P. clavata colonies in micro-centrifuge tubes. (Acetone controls, which were offered with the acetone extracts, contained sugar water plus 0.3 ml of acetone.) The tubes were weighed to ñ 1.0 mg before and after being attached for 1 hour to trees containing foraging trails. Paired choice tests were conducted 3 times each at all 5 colonies. We used paired t-tests to compare consumption (i.e., change in weight of the extract before and after a feeding trial) of extract and control, with a Bonferonni adjustment of critical à to reduce Type I error in multiple paired t-tests (SAS 1990). The results of this experiment were used to create another predictor variable called chemistry for our analysis of predation determinants, described above. Chemistry was a dichotomous variable having these levels: "deterrent extract" or "palatable extract."

We used logit analyses to study the relative importance of the predictor variables as determinants of ant response (rejections). Logit analysis is a type of log-linear model in which one of the variables is treated as a response factor, dependent on two or more predictor variables (Agresti 1990). We used the Newton-Raphson algorithm for parameter estimation and model testing to permit use of nonhierarchical models (SAS 1990). We used the maximum likelihood method for parameter estimation of linear models and Chi-square statistics for hypothesis testing (see SAS 1990). We used data from 265 feeding trials at all 5 colonies. Cells that contained "sampling zeroes" (sensu Bishop et al. 1975) were assigned values of 1 X 10-20, and cells that contained "structural zeroes" (sensu Bishop et al. 1975) were deleted (see SAS 1990).

We used species means to determine the rejection statistic and the predictor variables in the first three models (see Table 3). "Means" of strictly categorical variables (coloration, diet breadth, morphology) were simply the values specific to the species. The models for which we did not use species means dealt with subsets of the data that had balanced species replicates.

Results

We ran 6 logit models to test the relative importance of predictor variables of ant response (Table 3). We chose the models after looking at frequency tables for all combinations of variables (see Tabachnick and Fidel 1989); we ran six models because running one model with all the variables of interest would be uninterpretable (often containing biologically irrelevant interactions) and would require a prohibitively large sample size. The six models revealed that diet breadth, morphology, ontogeny, and chemistry were consistently important predictors of number of prey rejections.

Model 1 revealed significant associations between number of prey rejections and the following: diet breadth (X2 = 7.11, v = 2, P = 0.029) and morphology (X2 = 9.58, v = 4, P = 0.048). Specialists were rejected more often than were generalists (Fig. 1), and, in particular, the probability of complete rejection of a specialist was considerably higher than the probability of complete rejection of a generalist (from the logits, 65% probability of complete rejection of a specialist vs. 0.42% for a generalist). Hairy caterpillars were completely rejected more often than spiny or glabrous caterpillars, but all morphologies received at least some rejections (Fig. 2). It is interesting to note that for "glabrous" caterpillars, which seem to have no morphological defenses, there were more completely rejected species than spiny caterpillars and there were the same number of species receiving some rejections as hairy caterpillars.

Model 2 revealed significant associations between number of prey rejections and diet breadth (X2 = 9.88, v = 2, P = 0.0072). Again, specialists were more frequently rejected than were generalists (Fig. 1). Coloration categories were not significant predictors of rejection (X2 = 6.47, v = 4, P = 0.1665).

Diet breadth again predicted rejections in Model 3 (X2 = 7.46, v = 2, P = 0.0240), while size did not (X2 = 1.53, v = 4, P = 0.8209).

Models 4 and 5 included caterpillar species for which there were balanced replicates of different instars and balanced replicates at different colonies (Table 1). Model 4 examined the relationship between number of prey rejections, ontogeny, and caterpillar behavioral response. Here, we found that ontogeny was a significant predictor of prey rejections (X2 = 11.71, v = 2, P = 0.0029) while caterpillar behavior was not (X2 = 7.29, v = 4, P = 0.12). Late instar caterpillars were completely rejected more frequently than early instars (Fig. 3).

For Model 5 there were significant associations between ant response and the following: ontogeny (X2 = 7.24, v = 2, P = 0.0268) and colony (X2 = 17.9, v = 8, P = 0.022). Some colonies rejected caterpillars more readily than others. Again, later instars were rejected more frequently than early instars.

Model 6 dealt with the 3 species for which we did extract experiments and examined the relationship between extract palatability and number of prey rejections. Papilio anchisiades (Papilionidae) and its host plant Citrus limon (Rutaceae) were the only species that had deterrent extracts (t = 8.20, v = 1, corrected P = 0.042; see Table 2). The model indicated that number of prey rejections was significantly associated with both chemistry (X2 = 20.49, v = 2, P < 0.0001) and ontogeny (X2 = 30.81, v = 4, P < 0.0001). There was no significant interaction between chemistry and instar, but analysis of the standardized parameter estimates indicates that chemistry was the more important predictor variable of number of prey rejections (Table 3). For example, late instars of Papilio anchisiades (which had a deterrent extract) were always completely rejected, whereas late instars of the two caterpillar species whose extracts were palatable were never completely rejected.

Predictor variables were ranked in importance by using standardized parameter estimates (Tabachnick and Fidel 1988). Whenever diet breadth was in a model, it was the most important predictor. Morphology, coloration, size, and behavior were all less important predictor variables than were diet breadth, chemistry, ontogeny and colony. Important predictor variables are summarized in Table 3.

Discussion

The tropical ant Paraponera clavata is a broad generalist predator on many invertebrates (Wilson 1971), including a wide variety of caterpillars. Even so, we demonstrate that prey choice by P. clavata is by no means random or unpredictable. An ant's "decision" to attack a caterpillar depends on many characteristics of the caterpillar. Our study indicated that diet breadth, extract palatability, ontogeny, morphology, and colony are highly correlated with probability of predation by P. clavata.

Diet breadth consistently represented the most important determinant of predation in our study. Specifically, specialists were victims of successful predation much less frequently than were generalists, a pattern which suggests the importance of top-down forces. Generalists and specialists were offered to colonies of P. clavata on foraging trails, thus diminishing any physical protection derived from the host plant (see Brower 1984, Gross and Price 1988, Hay et al. 1989, Kareiva and Sahakian 1990). For example, cryptic specialists were rejected frequently, despite the absence of the host plant's correct background, which points to some other (presumably chemical) benefit derived from the host plant. Alternatively, chemical defense in this situation could have arisen from de novo synthesis, rather than from direct acquisition of host plant chemistry. We did not test this alternative hypothesis of de novo synthesis.

Other predictor variables also point to the potentially important role of chemistry as a deterrent to predation. For example, the dramatic effect of extract palatability on probability of predation provides anecdotal support. In particular, the caterpillar Papilio anchisiades, from which unpalatable extracts were obtained and from whose host plant (Citrus limon) unpalatable extracts were likewise obtained, was often completely rejected by P. clavata and was never attacked without at least some rejections. The two species that had palatable extracts (both from the caterpillar and host plant), one of which was a specialist (Dione juno), were seldom rejected by P. clavata.

In addition, caterpillar ontogeny, another consistently important predictor of predation, suggests the importance of chemistry. Colonies of P. clavata routinely preferred earlier instars to later instars, and we suggest that increased accumulation of plant toxins in later instars of caterpillars may contribute to the negative relationship between caterpillar development and probability of ant predation (Figure 4). While it could be argued that ontogeny-related rejections may have reflected difficulty in subduing and transporting large caterpillars (independent of their chemistry), in the context of foraging theory (see Pyke 1984), we point out that species size was not a significant predictor of rejections, despite the wide variation in sizes between species (7 to 135 mm).

Lastly, the significant inverse relationship between caterpillar hairiness and probability of predation may also have chemical origins. Caterpillar hairs are frequently urticating and may contain plant secondary compounds (Kawamoto and Kumada 1984), and for this reason vertebrate and invertebrate predators and specialist parasitoids may avoid hairy prey or attempt to remove hairs from caterpillars before eating them (Edmunds 1974, Heinrich 1979, Heinrich and Collins 1983, Bernays 1988, Sheehan 1991). Alternatively, the deterrent value of caterpillar hairs may be due to their allergenic proteins, abrasive surfaces, or difficulty in handling, all of which are independent of host plant derived allelochemistry (Kawamoto and Kumada 1984).

Evidence against the importance of chemistry arises from an analysis of coloration characteristics. Visual crypsis and bright coloration, for example, are typically a consequence of chemistry, and one would expect that brightly colored caterpillars would be unpalatable and frequently rejected by predators, while cryptic caterpillars (away from their host plants) would be palatable and not rejected by predators (Cott 1940, Edmunds 1974, 1990, Guilford 1986, 1990). However, we found that caterpillar coloration did not affect predation.

Evidence against the importance of predation by P. clavata as a selective force arises from analysis of inter-colony variation. Some colonies were more likely to reject caterpillars than other colonies; such variation lessens the importance of this predator as a selective force. Since our work was conducted in the dry season, when caterpillars are relatively scarce, it is possible that some ant colonies were being less selective about which prey items to attack.

We were not able to use specific host plants as predictors of caterpillar rejection. Thus, while our data suggest that specialists derive some chemical benefit from their host plants that generalists do not, we are unable to distinguish whether this benefit results from the specialists eating more toxic plants or from generalists diluting toxins by eating multiple plant species. We are also unable to address the possibility that specialists are better physiologically equipped than generalists to sequester secondary compounds. We do, however, demonstrate that a broad suite of caterpillar characteristics, all centered around diet specialization, serves to deter predation by P. clavata.

Acknowledgments

We thank Lucinda McDade and Barbara Lewis of the Organization for Tropical Studies (OTS) for logistical support, Jenny Phillips of the Instituto Nacional de Biodiversidad de Costa Rica (INBio) for help in identifying caterpillars, Orlando Vargas of OTS for help in identifying host plants, and Bob Wright of Cal State Fullerton for assistance with field and lab work. Scott Cooper, Deane Bowers, Jack Schultz, and two anonymous reviewers made helpful comments on an earlier draft of this manuscript. This work was supported by: an OTS Post Course Award from the Pew Charitable Fund to L.D. and T.F.; a University of Colorado Dean's Small Grant Award to L.D.; travel awards from the Graduate School, the Entomology Department, and the Ecology Program at Pennsylvania State University to T.F.; and a National Science Foundation Graduate Fellowship to T.F.

References

Agresti, A (1990) Categorical data analysis. John Wiley and Sons, New York and Toronto

Atsatt, PR (1981) Lycaenid butterflies and ants: selection for enemy-free space. Am Nat 118:639-654

Barbosa, P (1988) Natural enemies and herbivore-plant interactions: influence of plant allelochemicals and host specificity. In: Barbosa, P and DK Letourneau (eds) Novel aspects of insect-plant interactions. Wiley, New York, pp 201-229

Berenbaum, MR (1990) Evolution of specialization in insect-Umbellifer associations. Annu Rev Entomol 35:319-343

Bernays, EA (1988) Host specificity in phytophagous insects: selection pressure from generalist predators. Entomol Exp Appl 49:131-140

Bernays, EA and ML Cornelius (1989) Generalist caterpillar prey are more palatable than specialists for the generalist predator Iridomyrmex humilis. Oecologia 79:427-430

Bernays, EA and M Graham (1989) On the evolution of host specificity in phytophagous arthropods. Ecology 69:1153-1160

Bishop, YMM, SE Fienberg, and PW Holland (1975) Discrete multivariate analysis: Theory and practice. MIT, Cambridge

Blum, MS, RF Severson, RF Arrendale, DW Whitman, P Escoubas, O Adeyeye, and CG Jones (1990) A generalist herbivore in a specialist mode: metabolic, sequestrative, and defensive consequences. J Chem Ecol 16:223-244

Bowers, MD (1992) The evolution of unpalatability and the cost of chemical defense in insects. In: Roitberg, BD and MB Isman (eds) Insect chemical ecology: an evolutionary approach. Chapman & Hall, New York, pp 216-244

Bowers, MD (1990) Recycling plant natural products for insect defense. In: Evans, DL and JO Schmidt (eds) Insect defenses: adaptive mechanisms and strategies of prey and predators. University of New York Press, Albany, pp 353-387

Bowers, MD and Z Larin (1989) Acquired chemical defense in the lycaenid butterfly, Eumaeus atala. J Chem Ecol 15:133-146

Breed, MD and B Bennet (1985) Mass recruitment to nectar sources in Paraponera clavata: A field study. Insectes Sociaux 32:198-208

Brower, LP (1958) Bird predation and foodplant specificity in closely related procryptic insects. Am Nat 92:183-187

Brower, LP (1984) Chemical defense in butterflies. In: Vane-Wright, RI and PR Ackery (eds) The biology of butterflies: symposium of the Royal Entomology Society number 11. Academic, New York, pp 109-134

Cates, RG (1980) Feeding patterns of monophagous, oligophagous, and polyphagous insect herbivores: The effect of resource abundance and plant chemistry. Oecologia 46:22-31

Cott, HB (1940) Adaptive colouration in animals. Oxford University Press, New York

DeVries, PJ (1987) The butterflies of Costa Rica and their natural history. Princeton University Press, Princeton

Duffey, SS (1980) Sequestration of plant natural products by insects. Annu Rev Entomol 25:447-477

Edmunds, M (1974) Defence in Animals. Longman Group Limited, Essex

Edmunds, M (1990) The evolution of cryptic coloration. In: Evans, DL and JO Schmidt (eds) Insect defenses: adaptive mechanisms and strategies of prey and predators. University of New York Press, Albany, pp 3 - 23

Ehrlich, PR and PH Raven (1964) Butterflies and plants: a study in coevolution. Evolution 18:568-608

Eisner, T (1970) Chemical defense against predation in arthropods. In: Sondheimer, E and JB Simeone (eds) Chemical ecology. Academic Press, New York, pp 157-217

Evans, DL and JO Schmidt (eds) (1990) Insect defenses: adaptive mechanisms and strategies of prey and predators. University of New York Press, Albany

Feeny, P (1976) Plant apparency and chemical defense. In: Wallace, JW and RL Mansell (eds) Biochemical interactions between plants and insects. Plenum, New York, pp 1-40

Futuyma, DJ (1976) Food plant specialization and environmental predictability in Lepidoptera. Am Nat 110:285-292

Futuyma, DJ and B Moreno (1988) The evolution of ecological specialization. Annu Rev Ecol Syst 19:207-33

Gross, P and PW Price (1988) Plant influences on parasitism of two leafminers: a test of enemy-free space. Ecology 69:1506-1516

Guilford, T (1986) How do warning colors work? Conspicuousness may reduce recognition errors in experienced predators. Anim Behav 34:286-288

Guilford, T (1990) The evolution of aposematism. In: Evans, DL and JO Schmidt (eds) Insect defenses: adaptive mechanisms and strategies of prey and predators. University of New York Press, Albany, pp 23-63

Hay, ME, JR Pawlik, JE Duffy, and W Fenical (1989) Seaweed-herbivore-predator interactions: host-plant specialization reduces predation on small herbivores. Oecologia 81:418-427

Heinrich, B (1979) Foraging strategies of caterpillars. Oecologia 42:325-337

Heinrich, B and SL Collins (1983) Caterpillar leaf damage and the game of hide-and-seek with birds. Ecology 64:592-602

Holloway, JD and PDN Herbert (1979) Ecological and taxonomic trends in macrolepidopteran host plant selection. Biol J Linn Soc 11:229-251

Jaenike, J (1978) Resource predictability and niche breadth in the Drosophila quinaria species group. Evolution 32:676-678

Janzen, DH (ed) (1983) Costa Rican natural history. University of Chicago Press, Chicago

Janzen, DH and CR Carroll (1983) Paraponera clavata. In: Janzen, DH (ed) Costa Rican natural history. pp 752-753 University of Chicago Press, Chicago, pp 752-753

Jones, FM (1932) Insect colouration and the relative acceptability of insects to birds. Trans Roy Entomol Soc Lond 80:345-385

Jones (1934) Further experiments on colouration and relative acceptability of insects to birds. Trans Roy Entomol Soc Lond 82:443-453

Jones, CG, DW Whitman, SJ Compton, PJ Silk, and MS Blum (1989) Reduction in diet breadth results in sequestration of plant chemicals and increased efficacy of chemical defense in a generalist grasshopper. J Chem Ecol 15:1811-1822

Kareiva, P and R Sahakian (1990) Tritrophic effects of a simple architectural mutation in pea plants. Nature 345:433-434

Kawamoto, F and N Kumada (1984) Biology and venoms of Lepidoptera. In: Tu, AT (ed) Handbook of natural toxins. Marcel Dekker, New York, pp 270-351

Levins, R and RH MacArthur (1969) An hypothesis to explain the incidence of monophagy. Ecology 50:910-911

Marquis, RJ and HE Braker (1991) Plant-herbivore interactions at La Selva: diversity, specificity and impact. In: McDade, LM, KS Bawa, GS Hartshorn and HE Hespenheide (eds) La Selva: Ecology and natural history of a neotropical rain forest. University of Chicago, Chicago, in press

Moran, NA (1988) The evolution of host-plant alternation in aphids: evidence for specialization as a dead end. Am Nat 132:681-706

Price, PW (1991) Patterns in Communities Along Latitudinal Gradients. In: Price, PW, TM Lewinsohn, GW Fernandes and WW Benson (eds) Plant-animal interactions: evolutionary ecology in tropical and temperate regions. John Wiley & Sons, New York, pp 51-70.

Price, PW (1983) Hypotheses on organization and evolution in herbivorous insect communities. In: Denno, RF and MS McClure (eds) Variable plants and herbivores in natural and managed systems. Academic Press, New York and London, pp 559-598

Price, PW, CE Bouton, P Gross, BA McPheron, JN Thompson, and AE Weis (1980) Interactions among three trophic levels: influence of plants on interactions between insect herbivores and natural enemies. Annu Rev Ecol Syst 11:41-65

Pyke, GH (1984) Optimal foraging theory: a critical review. Annu Rev Ecol Syst 15:523-575

Rhoades D, Cates RG (1976) Toward a general theory of plant antiherbivore chemistry. In: Wallace, J and R Mansell (eds) Biochemical interactions between plants and insects. Recent advances in phytochemistry 10. Plenum Press, New York, London, pp 168-213

Ricklefs, RE and K O'Rourke (1975) Aspect diversity in moths: a temperate-tropical comparison. Evolution 29:313-324

SAS (1990) SAS User's Guide: Statistics. SAS Institute, Cary, NC

Sheehan, W (1991) Host range patterns of hymenopteran parasitoids of exophytic lepidopteran folivores. In: Bernays, EA (ed) Insect-plant interactions, volume III. CRC Press, Boca Raton, pp 209-248

Strong, DR (1988) Insect host range. Ecology 69:885

Tabachnick, BG and LS Fidell (1989) Using multivariate statistics. Harper & Row, New York

Weiss, AE and MR Berenbaum (1989) Herbivorous insects and green plants. In: Abrahamson, WG (ed) Plant-animal interactions. McGraw Hill, New York, pp 123-157

Wilson, EO (1971) The insect societies. Harvard University, Cambridge MA

Young, AM (1977) Notes on the foraging of the giant tropical ant Paraponera clavata (Formicidae:Ponerinae) on two plants in tropical wet forest. J Georgia Entomol Soc 12:41-51

Young, AM and HR Hermann (1980) Notes on the foraging of the giant tropical ant Paraponera clavata. J Kans Entomol Soc 53:32-55