Herbivores on a Dominant Understory Shrub Increase Local Plant Diversity in Rain Forest Communities

Ecology, 91(12), 2010, pp. 3707–3718 Ó 2010 by the Ecological Society of America

Herbivores on a dominant understory shrub increase local plant diversity in rain forest communities

1,4 2 1 3 LEE A. DYER, DEBORAH K. LETOURNEAU, GERARDO VEGA CHAVARRIA, AND DIEGO SALAZAR AMORETTI 1Department of Biology, University of Nevada, Reno, Nevada 89557 USA 2Department of Environmental Studies, University of California, Santa Cruz, California 95064 USA 3Department of Biology, University of Missouri, St. Louis, Missouri 63121 USA

Abstract. Indirect effects of trophic interactions on biodiversity can be large and common, even in complex communities. Previous experiments with dominant understory Piper shrubs in a Costa Rican rain forest revealed that increases in herbivore densities on these shrubs caused widespread seedling mortality as a result of herbivores moving from Piper to seedlings of many different plant genera. We tested components of the Janzen-Connell hypothesis by conducting focused studies on the effects of specialist and generalist Piper herbivores on local seedling diversity. Whereas specialist herbivores are predicted to increase mortality to neighboring seedlings that are closely related to the source plant, true generalists moving from source plants may cause density-dependent mortality of many species, and possibly increase richness if new species replace abundant species that have been thinned by herbivores. Therefore, we hypothesized that seedling richness would be greater in understory control plots created in patches of Piper that had normal densities of generalist herbivores compared to plots from which we removed generalist herbivores manually from all Piper shrubs. After 15 months, generalist-herbivore-removal plots had .40% fewer seedlings, .40% fewer species, and 40% greater seedling evenness, on average, than control plots with generalist herbivores intact. Using a complementary approach in unmanipulated plots in four forests, we used path analysis to test for a positive association between seedling diversity and herbivore damage on Piper species. In unmanipulated plots, for both generalist and specialist herbivores, our data were significant fits to the causal model that Piper herbivores decrease evenness and increase plant species richness, corroborating the experimental results. Because herbivores changed how individuals were apportioned among the species and families present (lower evenness), one interpretation of these associations between herbivores on Piper shrubs and local seedling richness is that high seedling mortality in dominant families allowed the colonization or survival of less common species. If interspecific or apparent competition allowed for a relative increase in species richness, then the Janzen-Connell hypothesis may extend its predictions to generalist seedling predators. We speculate that apparent competition may explain some of the deviations from neutral model predictions, especially at small scales. Key words: apparent competition; biodiversity; herbivory; Janzen-Connell hypothesis; seedling mortality; tropical rain forest.

INTRODUCTION 2008). The diet breadth of these predators has been referred to as ‘‘acute specificity’’ (Norghauer et al. 2010) The role of specialist natural enemies has been or ‘‘high levels of host specificity’’ (Harms et al. 2000). emphasized in the well-known Janzen-Connell hypoth- However, if we generalize the Janzen-Connell hypothesis esis (Janzen 1970, Connell 1971), which posits that high (e.g., Bever 2003), it simply predicts that seed or seedling species richness is maintained by density dependent enemies dispersing from source plants reduce the density mortality of seeds and seedlings as they are attacked by of their host plants along an impact gradient that host-specific predators, pathogens, or herbivores dis- diminishes with distance from the source. The resultant persing from the parent tree. Theoretical and empirical effects on the community will depend on the diet studies of this hypothesis assume that these enemies breadth of these enemies. There are numerous working exhibit a clear preference for plants that are better definitions and measures of diet breadth (e.g., Odegaard competitors or possibly species that tend to form a dense 2000, Odegaard et al. 2000, Irschick et al. 2005, layer of seedlings or saplings (reviewed by Carson et al. Lewinsohn et al. 2005, Novotny and Basset 2005, Dyer et al. 2007), but here we use a common taxonomic Manuscript received 2 September 2008; revised 19 April measure: the number of plant families consumed by an 2010; accepted 22 April 2010. Corresponding Editor: W. P. Carson. herbivore, with specialists consuming plants within one 4 E-mail: [email protected] family and generalists consuming plants in multiple 3707 3708 LEE A. DYER ET AL. Ecology, Vol. 91, No. 12 families (Bernays and Graham 1988, Dyer 1995, Singer richness in a neotropical rain forest after removing and Stireman 2005, Dyer et al. 2007). generalist herbivores from all Piper species in understory If the seed or seedling predators attracted or plots. Understory Piper vines and shrubs are abundant maintained by a source plant are taxonomic specialists, and diverse understory plants, usually ranked in the top then associated plants in the same species, genus or 10 species for high species richness and numerical family will experience density-dependent mortality from abundance in Neotropical forests (Burger 1971, Gentry shared herbivores. True generalist feeders that eat plants and Emmons 1987, Dyer and Palmer 2004, Marquis in all families are on the other end of the theoretical 2004). The plants in this genus host a diverse array of spectrum of diet breadth, and seed or seedling mortality specialist and generalist herbivores (Marquis 1991, caused by these generalists would be increased by Greig 1993, 2004, Letourneau 2004, Dyer et al. 2007: increases in the relative densities of their various host Appendix A), and thus are good candidates for apparent seeds or seedlings in the community (probability of competition with other understory plants through encounter). The likely effect of both specialist and shared herbivores (Strong et al. 1984). Using a generalist movement from the source plant, then, would complementary correlational approach, we also tested be to reduce dominance at the level of species, genus or the effects of generalist versus specialist herbivory on family, with more generalized herbivores affecting more plant species richness in unmanipulated understory taxa than specialists. Since seeds and seedlings are often patches in four forests. We quantified generalist clumped in distribution, both specialists and generalists (orthopteran and lepidopteran) and specialist (lepidop- from source plants may be important in increasing or teran and coleopteran) herbivory on Piper shrubs as maintaining high plant diversity (evenness and richness) potential source plants, and measured herbivory levels in the community. That is, factors that reduce abun- and diversity of neighboring understory plants. The dance and dominance can enhance conditions for the specialists were all restricted to feeding on Piper and recruitment of new species or the persistence of rarer many were monophagous, while the generalists all fed species. on multiple families, with some generalists feeding on Our consideration of the potential importance of plants in over 50 families (Appendix A; Dyer et al. 2007; generalist herbivores stems from previous studies, in L. A. Dyer, unpublished data). We also included light which we documented predictable changes in plant availability (canopy cover) as a predictor variable, species richness after experimental additions of an apex because small changes in light availability can result in predatory beetle that resides in Piper ant-shrubs biologically significant changes in Piper chemistry, (Letourneau and Dyer 1998, Letourneau et al. 2004). herbivory levels, and biodiversity (Denslow 1987, Dyer Indirect effects occurred through four trophic levels; the and Letourneau 1999, 2007). This study is unique in its predatory beetle (Coleoptera: Cleridae) reduced densi- combination of experimental and path analysis ap- ties of resident Pheidole ants (Hymenoptera: Formici- proaches to test for community-level diversity effects. It dae), which were subsequently less effective at reducing is, to our knowledge, the first attempt to test for Janzen- herbivory on the host plant Piper cenocladum C.DC. Connell effects: (1) with patchy understory shrubs as (Piperaceae) (Letourneau and Dyer 1998). These cas- concentrated source pools for enemies, (2) with gener- Piper cading effects that released herbivores on rippled alist seedling predators, and (3) invoking apparent out to the local understory plant community, through competition as a possible mechanism for determining increased herbivory levels and mortality on nearby Piper plants species richness. and on other plantlets (,20 cm; Dyer and Letourneau 1999, Letourneau et al. 2004, Letourneau and Dyer MATERIALS AND METHODS 2005). On poorer quality ultisols the richness of plantlets Study site did not change. However, the species richness of plantlets in understory communities on rich, alluvial We conducted the manipulative study at La Tirimbina soils was significantly lower than that of control plots Reserve, Costa Rica, between May 2004 and August with no beetles, intact ant colonies, and low herbivory. 2005 and collected field data between September 2005 We hypothesized that the ubiquitous Piper,when and December 2006 in several nearby lowland, tropical undefended, hosted relatively large numbers of specialist rain forest reserves in Heredia Province, Costa Rica: La and generalist herbivores that moved to other plants in Tirimbina Reserve (1082500700 N, 848703000 W, 345 ha, the understory, causing damage or mortality. Such 200 m elevation, 3.8 m annual precipitation), El Bejuco effects have been shown for movement of herbivores Field Station (1082600600 N, 8481200000 W, 30 ha, 150 m from mature plants to seedlings (e.g., Maeto and elevation, 3.6 m annual precipitation), Bijagual Ecolog- Fukuyama 1997, Wada et al. 2000), and form the basis ical Reserve (1081800900 N, 8480100500 W, 290 ha, 350 m of the Janzen-Connell hypothesis (Janzen 1970, Connell elevation, 5.5 m annual precipitation), and Braulio 1971). Carrillo National Park (108800600 N, 8385903000 W, 300 To test if the Janzen-Connell hypothesis extends to m elevation, 4.2 m annual precipitation). The mean generalist enemies and can explain changes in rain forest temperature at all sites is 258C and ranges between 208C seedling diversity, we measured understory plant species and 308C. December 2010 HERBIVORES AND LOCAL PLANT DIVERSITY 3709

Plants and herbivores of the herbivore was unknown (less than 5% of all Understory Piper shrubs in Costa Rica are patchily individual externally feeding Piper folivores encountered distributed in primary and late-successional secondary at Tirimbina), it was removed. We did not record the forests receiving at least 2000 mm rainfall per year number of herbivores removed from plots, but at (Marquis 2004). Shrubs included in this study were P. Tirimbina, a snapshot of herbivore density on Piper biolleyi C.DC., P. cenocladum C.DC., P. colonense shrubs is quite variable and ranges from 0 to 21 C.DC., P. glabrescens C.DC., P. imperiale C.DC., P. specialists and 0 to 14 generalists per shrub at any given melanocladum C.DC., P. multiplinervium C.DC., P. time, with larval Lepidoptera representing over 90% of schiedeanum Steud., P. trigonum C.DC., and P. uros- herbivores discovered via visual searches (L. A. Dyer, tachyum Hemsl., all common throughout the Caribbean unpublished data). After 14 months, the number of lowlands. Specialist Lepidoptera feeding on understory morphospecies of seedlings was counted in every plot; Piper foliage were primarily geometrids (e.g., Eois no seedlings were removed from the plots. numida Druce, Eois apyraria Gueneé, Eois guapa Correlational changes in plant communities Schaus, Eois nympha Schaus, Epimecis sp.) and skippers 2 (e.g., Quadrus cerialis Stoll, Family Hesperiidae). Forty-five square understory plots (25 m )were Specialist beetles were primarily stem-boring weevils systematically selected such that they were evenly (Coleoptera: Curculionidae: Ambates spp.) and leaf- distributed spatially within each of the four forest sites, feeding flea beetles (Coleoptera: Chrysomelidae: Phy- with 30 plots at Tirimbina, and five each at the three simera spp.). Each of these taxa leaves recognizable other reserves. The greater number of plots at Tirimbina feeding damage on their host plants (Appendix A; see was a result of easier access to that forest. Each plot was also Plate 1), and the Eois spp. rarely move as larvae to established around the first Piper shrub observed after another plant (Dyer and Palmer 2004). Generalist walking a minimum distance of 30 m from a systemat- folivores include Lepidoptera in several families (Apa- ically designated point on a trail. Within each forest, the telodidae, Limacodidae, Saturniidae), leafcutter ants resultant understory plots were separated from each (Hymenoptera: Formicidae: Atta cephalotes L.) and other by at least 500 m, and the four forests were at least orthopterans (primarily Orthoptera: Tetigoniidae) (Le- 10 km (but no more than 50 km) from each other. tourneau et al. 2004). The generalist caterpillars on Patches in these four different, but nearby forests were Piper have very broad diet breadths and as a group have used to capture variation in plant and herbivore been found and reared on a total of 249 genera in 98 assemblages in the understory to ensure that results plant families (Dyer et al. 2007; L. A. Dyer, unpublished were not unique to a quirky assemblage of species or an data). Damage from leafcutter ants is less than 1% of all otherwise non-representative forest. plants surveyed for herbivory in previous experimental We quantified specialist and generalist damage on and observational studies in these understory habitats newly expanding leaves on each Piper shrub and on all (Dyer et al. 2004, Letourneau and Dyer 2005, Fincher et seedlings in each plot. The percent of each new leaf eaten al. 2008). by all generalist (lepidopterans, orthopterans, and leafcutter ants) and specialist folivores (lepidopterans and Experimental changes in plant communities coleopterans) was determined using a Plexiglas grid to In May 2004, 24 understory plots (25 m2) were count the number of square centimeters of leaf area randomly selected utilizing a grid on a map of eaten divided by the total leaf area (also in cm2). All Tirimbina, resulting in plots that were at least 200 m herbivory present was quantified and categorized as from another plot. All plots were at least 10 m from a specialist or generalist herbivory for the analyses. We trail and included at least five individual shrubs in the measured canopy cover using a hand-held spherical genus Piper. Twelve plots were randomly assigned to densiometer (Forest Densiometers, Bartlesville, Oklaho- two treatments: control without herbivore removal and ma, USA) at 1 m height, in the center of each plot, experimental with removal of generalist herbivores. All taking the average of measurements in each cardinal plots were visited an average of two times per month (26 direction. Comparisons between spherical densiometer total visits by the authors and Earthwatch volunteers) to and hemispherical photography measurements are search for herbivores, and for the removal plots, all strongly positively correlated (Bellow and Nair 2003, generalist herbivores found on Piper were removed from Paletto and Tosi 2009) but densiometer readings may be the plot. Herbivore removal was accomplished via hand more accurate for tall canopy estimates of PAR (Bellow collection and transport to an ambient laboratory, and Nair 2003). The inverse of canopy cover provides a where immature herbivores were reared, or to a freezer relative estimate of the mean light availability per plot where adult herbivores were preserved and subsequently (Fiala et al. 2006) and Piper biology is sensitive to curated. Specialist herbivores that feed on plants in the changes in this variable as measured by a spherical genus Piper, based on data from a 14-year caterpillar densiometer (Letourneau and Dyer 1998, Dyer and database (described in Dyer et al. 2007) and 14 years of Letourneau 1999, 2003, 2007, Letourneau et al. 2004). personal observation of Piper herbivores, were not Seedling abundance in each plot was counted as the removed from their Piper host plant. If the diet breadth number of angiosperms with cotyledons and fewer than 3710 LEE A. DYER ET AL. Ecology, Vol. 91, No. 12

five leaves. This working definition of ‘‘seedling’’ was utilized to examine the associations between seedling utilized because this is an ecologically important stage herbivory and seedling abundance of the four (out of a (i.e., before phase changes occur; Karban and Thaler total of 66 families present) numerically dominant 1999) and created consistency in our counts. Seedling families or subfamilies in the plots: Rubiaceae, Areca- diversity was quantified as species richness (S) as well as ceae, Mimosoidea, and Piperaceae. All models were equitability (ED), which is a standard measure of constructed based on our predictions; the outcome evenness (Rosenzweig 1995, Magurran 2004) ED ¼ 1/D either supports the prediction or not. Goodness of fit 3 1/S, where 1/D is Simpson’s reciprocal index, D ¼ tests (distributed as chi squared) were used to estimate S Ri¼1 ni(ni 1)/N(N 1), N ¼ total number of plants in a the fit of the models to the data. P values of greater than plot, and ni ¼ total number of species i in a plot. ED 0.05 are considered to indicate a good fit of the model to varies from 0 to 1, with larger values indicating greater the data (Ullman 1996). evenness among species. The variables S and ED were In addition to controlling for seedling abundance via not collinear for any of our data. path analysis, we also utilized simple individual-based rarefaction to estimate richness for all experimental and Statistical analyses correlational plots. We used SAS (9.2) statistical For the experiment, we used multivariate analysis of software to simulate 1000 random draws of 24 variance (MANOVA) and profile analysis to compare individuals for the experimental plots and 23 individuals richness, evenness, and abundance between the control for the correlational plots (based on the plots with the and the herbivore manipulation (using the GLM lowest abundances) and to calculate mean richness (SR) Procedure in SAS 9.2; SAS Institute 2008). Residuals and 95% confidence intervals. We compared the effects were normal (using the criterion of Kolomogorov- of experimental manipulations on SR with ANOVA. Smirnoff, P . 0.01) for all variables so no transforma- For the correlational data, we used SR in the specialist tions were applied to the response variables. To examine and generalist herbivory path models without including patterns of plant community changes due to our abundance (which is the same in all plots for rarefaction manipulations, we calculated Jaccard’s index of similar- data) or evenness. ity (using the distance procedure in SAS 9.2) for all combinations of plots. Comparisons of these indices RESULTS among all pairs of control plots versus all control- Constant removal of generalist herbivores from Piper experimental pairs were used to test the hypothesis that shrubs over 15 months caused a significant decrease in plant communities that developed in control plots were seedling abundance (from a mean of 95.8 to 66.3 more similar to each other than they were to those that seedlings per plot), decreased richness (from a mean of developed in herbivore-removal plots. 28.6 to 20.3 species per plot), and increased seedling For the correlational plot data, we used the SAS Calis evenness (from 0.05 to 0.07; MANOVA Wilk’s lambda (covariance analysis of linear structural equations) F1,22 ¼ 6, P ¼ 0.02, Fig. 1). Profile analysis revealed a procedure to construct a specialist and two generalist stronger effect of herbivore removal on abundance herbivory models to test specific causal hypotheses versus richness and evenness (Wilk’s lambda F1,22 ¼ 6, P about associations between canopy cover, herbivory, ¼ 0.02, Fig. 1). The decrease in richness was also seedling herbivory, seedling abundance, and seedling apparent when abundance was taken into account with diversity (evenness and richness). Two models were rarefaction, with SR significantly lower in the herbivore included for generalist herbivores because we could not removal plots (decreasing from 13.0 to 10.8, F1,22 ¼ 4.7, include seedling richness and evenness in the same model P ¼ 0.04). All combinations of experimental plots due to insufficient sample size and multicollinearity; yielded low Jaccard’s similarity values (mean ¼ 0.21 6 evenness was not part of a priori hypotheses for effects 0.01 for all comparisons between plots). The control vs. of specialist herbivores on seedlings. The general causal experimental plot comparisons yielded the lowest model includes effects of light on Piper herbivory and similarity values (0.19 6 0.01 compared to 0.23 6 0.01 seedling diversity, effects of Piper herbivory on seedling for the control–control comparisons). herbivory, and effects of seedling herbivory on seedling A total of 1945 seedlings (trees, shrubs, and forbs) in diversity (the direction of effects, positive or negative, the 24 experimental plots were distributed among 106 depended on context and was determined by the species in 58 families and 23 individual seedlings that empirical data). Adding variables with known effects, were classified as 12 different morphospecies in undes- like light, puts the effect sizes of the other variables into ignated families. Both control and experimental plots at appropriate perspective, and enhances the construction La Tirimbina had roughly twice the number of seedlings of directed causal hypotheses. We ran a multiple (24–133 per 25-m2 area sampled) as the unmanipulated correlation analysis to explore relationships between plots used for path analysis (12–70), and the control and herbivory and seedling diversity. experimental plots also had a relatively high average We constructed additional models to explore mecha- species richness (24.5) and low evenness (0.06) per plot. nisms that may underlie the results of the general models A total of 3703 seedlings in the 45 unmanipulated described above. Four separate path analyses were understory plots were identified as 210 species belonging December 2010 HERBIVORES AND LOCAL PLANT DIVERSITY 3711

FIG. 1. Influence of generalist herbivore removal from Piper spp. shrubs on local seedling abundance (number of individuals 2 2 per 25-m plot), richness (number of species per 25-m plot), and evenness (equitability, ED; see Materials and methods: Correlational changes in plant communities for calculation). to 62 plant families, with an additional 11 unknown per plot), specialist beetles (8.8% 6 2.5%), generalist species with family designations undetermined (Appen- caterpillars (32.6% 6 6.2%) and generalist orthopterans dix B). Herbivore surveys on two thirds of the plant (1.0% 6 1.1%) (see Appendix 1 for images of genera represented by seedlings in these plots have characteristic herbivore damage). There were no signif- shown overlap of lepidopteran species with Piper in icant differences among forests for richness, abundance, every case (Dyer and Gentry 2002, Dyer et al. 2007; evenness, and herbivory per plot (MANOVA, Wilk’s L. A. Dyer, unpublished data). That is, 100% of the lambda, P . 0.05). seedling genera examined share herbivores with the The amount of canopy cover varied from 85.8% to plants from which herbivores were removed. This is not 93% in the plots (Table 1). These changes in light to say that individual lepidopteran larvae show no availability explained a substantial amount of variation preferences among plants, but this high level of overlap in specialist and generalist herbivory levels on newly at the community level, establishes the potential for emerging leaves of Piper shrubs. All of the specified generalist Piper herbivores to feed on neighboring models were not significantly different from the data (P seedlings. The four most common plant families and . 0.1 for all models), providing support for the subfamily accounted for an average of 54.5% 6 4.2% proposed causal relationships that explain the significant (mean 6 SE) of all individual seedlings per forest correlation between generalist herbivory and seedling understory plot. The subfamily Mimosoideae accounted richness (Fig. 2). Path analyses showed that decreased for 18.8% 6 2.5% of individuals per plot, Arecaceae for light was associated with higher levels of new leaf 17.4% 6 2.4%, Piperaceae for 12.7% 6 1.8%, and damage on Piper shrubs by generalist (mostly orthop- Annonaceae for 7.2% 6 1.1%. Average herbivory on terans and generalist lepidopterans) herbivores, which in new leaves of all Piper shrubs in the plot was similar for turn increased herbivory on seedlings, resulting in a specialists and generalists (Table 1); over 92% of the decrease in seedling evenness and an increase in seedling herbivory was caused by the following insects: specialist abundance in these patches of Piper shrubs (Fig. 3A). caterpillars (49.7% 6 9.9% of total recorded herbivory There was also a strong direct positive effect of increased

TABLE 1. Mean (6SE) canopy cover, abundance of all seedlings in plots (including Piper spp.), richness of all seedlings in plots (including Piper spp.), and percent damage of new leaves of Piper shrubs by generalist and specialist herbivores.

Total number of seedling Damage (%)

Variable N Canopy cover (%) Individuals SpeciesSeedling evenness (ED) Generalist Specialist All sites 45 92.1 6 0.2 84.0 6 7.1 22.5 6 1.3 0.06 6 0.009 10.8 6 1.8 9.9 6 1.4 Bijagual 5 92.1 6 0.3 71.5 6 13.9 19.0 6 1.9 0.05 6 0.01 6.2 6 1.2 9.4 6 2.6 Braulio Carrillo 5 89.8 6 1.1 47.0 6 2.2 20.6 6 1.6 0.14 6 0.02 2.4 6 1.2 4.6 6 1.8 El Bejuco 5 92.3 6 0.1 80.2 6 13.2 19.6 6 1.6 0.03 6 0.008 4.4 6 1.7 3.6 6 0.8 Tirimbina 30 92.4 6 0.07 92.8 6 9.8 23.9 6 1.9 0.05 6 0.01 14.2 6 2.4 12.0 6 2.4

Note: N ¼ number of plots. See Materials and methods: Correlational changes in plant communities for calculation of ED. 3712 LEE A. DYER ET AL. Ecology, Vol. 91, No. 12

light affects abundance which in turn affects evenness, does not ﬁt the data (v2 ¼ 12.5, df ¼ 4, P ¼ 0.01). Path analysis for generalist herbivory using species richness instead of evenness shows that the increases in seedling abundance associated with higher herbivory on seedlings and lower evenness of seedlings leads to greater species richness. That is, increases in herbivory on Piper shrubs were associated with decreased evenness and increased abundance and richness of seedlings in the plot (Fig. 3B), and indirect effects are similar in magnitude to direct effects of light on these factors (based on the size

of the path coefficients). Path analyses using SR instead of richness yielded identical relationships between SR, light, and herbivory (Fig. 3C), but because rarefied S FIG. 2. Uncorrected relationship between generalist herbivory on Piper shrubs and seedling species richness in patches of discounts abundance it does not provide the same level rain forest understory across four forests (from multiple of insight about relationships between the different regression and path analysis; R ¼ 0.38, P ¼ 0.01, N ¼ 44 plots). components of diversity (richness, evenness, abundance). The relationships among generalist herbivore light on seedling evenness that was equal in magnitude pressure, seedling abundance, evenness and richness to the direct effects of seedling herbivory (Fig. 3A). were consistent with our results from the herbivore Reversing the causal arrows in the path analysis in Fig. removal experiment. 3A, such that number and evenness of seedlings cause Herbivore damage by specialists was dominated by greater seedling herbivory, yields a path analysis that Eois spp. (85.0% of all specialist herbivory), which cause does not statistically fit the data (v2 ¼ 14.3, df ¼ 4, P ¼ a significant reduction in biomass, but do not move 0.006); similarly, a modified model from Fig. 3A where frequently as larvae to other Piper individuals. Never-

2 FIG. 3. (A) Path analysis results (model fit v ¼ 1.0, df ¼ 4, P ¼ 0.91) for the generalist herbivory model, examining effects of light and herbivory on evenness of all seedlings (plants with cotyledons and fewer than five true leaves) in Piper shrub patch plots. The alternative hypothesis that seedling evenness and abundance caused higher seedling herbivory (i.e., switching the direction of those two arrows) was not a significant fit to the data (v2 ¼ 14.3, df ¼ 4, P ¼ 0.006). (B) Path analysis results (model fit v2 ¼ 5.4, df ¼ 4, P ¼ 0.25) for the generalist herbivory model, using species richness (S) of all seedlings in Piper shrub patch plots. (C) Path 2 analysis results (model fit v ¼ 5.4, df ¼ 3, P ¼ 0.14) using rarefaction seedling species richness (SR). Mean generalist herbivory (%) was measured on established Piper shrubs within a plot. Positive path coefficients are illustrated with arrow heads, while closed circles indicate a negative effect. December 2010 HERBIVORES AND LOCAL PLANT DIVERSITY 3713

2 FIG. 4. Path analysis results (model fit v ¼ 5.1, df ¼ 4, P ¼ 0.28) for the specialist herbivory model, examining effects of light and herbivory on seedling richness (all plants with cotyledons and fewer than five leaves) in Piper shrub patch plots. Models replacing the abundance–richness correlation with causal (i.e., single-headed) arrows were not significant fits to the data. Mean specialist herbivory (%) was measured on established Piper shrubs within a plot. Positive path coefficients are illustrated with arrow heads, while closed circles indicate a negative effect. theless, specialist herbivory on Piper shrubs was still evenness. This scenario requires that some common associated with a significant increase in seedling species were killed disproportionately while other, more herbivory, richness and abundance, but the direct effects tolerant or spatially distant common species increased on seedling herbivory and indirect effects on richness slightly in numbers (see Alexander and Holt 1998), and abundance were substantially smaller than the resulting in fewer seedlings overall within the dominant direct effects of light availability (Fig. 4). As with families and lower plot evenness. If sustained herbivory generalist herbivory, the path analysis using SR sup- disproportionately killed dominant species, then indi- ported the same hypotheses about relationships between viduals of new species might have colonized more richness, light, and herbivory (model fit v2 ¼ 6.8, df ¼ 3, successfully. Exploring this potential scenario, we found P ¼ 0.08). that the paths of increased seedling richness, caused by We used path analyses to explore the hypothesis that increased overall levels of generalist herbivory and lower seedling mortality among particular, rather abundant evenness, were significant fits for the four dominant species could allow other common species and new rarer families or subfamilies in the plots (Fig. 5A–D). species to colonize or persist, thus increasing seedling Decreases in evenness within three of these families richness and abundance while decreasing seedling were associated with increases in total seedling richness

FIG. 5. Path analysis results for generalist herbivory on seedlings using abundance and herbivory for all seedlings in Piper shrub patch plots and evenness for all seedlings in the (A) Annonaceae, (B) Arecaceae, (C) Mimosoideae (Fabaceae), and (D) Piperaceae. All models were a significant fit: (A) v2 ¼ 3.1, df ¼ 2, P ¼ 0.21; (B) v2 ¼ 0.28, df ¼ 2, P ¼ 0.87; (C) v2 ¼ 2.7, df ¼ 2, P ¼ 0.26; (D) v2 ¼ 1.1, df ¼ 3, P ¼ 0.77. Positive path coefficients are illustrated with arrow heads, while solid circles indicate a negative effect. 3714 LEE A. DYER ET AL. Ecology, Vol. 91, No. 12

(Fig. 5; this yields a negative path coefficient between Tracking seedling colonization, eventual mortality, evenness and richness), with the effects of family and its causes over time for many thousands of evenness ranked from strongest to weakest as follows: seedlings, and testing for competition among plants Annonaceae, Arecaceae, and Mimosoideae. Evenness was beyond the scope of this and previous studies, so the within the Piperaceae was not associated with changes in underlying mechanisms by which herbivores modified richness; evenness within the Arecaceae and Mimosoi- community parameters cannot be determined directly. deae did not explain all of the variation in richness due Instead, we tested several scenarios based on theory to to herbivory, which is why there was still a significant explain the outcome of our experiments and correla- herbivory to richness path coefficient (Fig. 5B–D). There tions, and offer other suggestions involving parameters was very strong support for the causal hypothesis that that were not measured. There are at least three ways to Annonaceae evenness positively affects total seedling explain how generalist herbivores could decrease seed- richness, with a path coefficient of 0.9 (Fig. 5A), and ling evenness and increase seedling abundance and generalist herbivory on Piper shrubs was not directly richness. All of these scenarios involve concentrated associated with seedling richness. These differences in Piper shrubs acting as insectaries for seedling predators, evenness within families and subsequent associations through movement of larvae, nymphs or adults to with overall richness were due to reduced seedling nearby seedlings of many species. This is a realistic numbers for these numerically dominant species: Ana- assumption, given that generalist herbivores often prefer xagorea crassipetala (Annonaceae), Asterogyne martiana young, expanding leaves of tropical forest plants (Coley (Arecaceae), Pentaclethra macroloba (Fabaceae: Mi- and Aide 1991), which are available on vulnerable mosoideae), and Piper urostachyum (Piperaceae), all of seedlings in the vicinity of adult plants. Furthermore, which experienced high levels of seedling herbivory. generalist herbivores feeding on Piper consume most of Furthermore, the generalist caterpillars that feed on the genera of plant seedlings that were present in our Piper have been recorded feeding on more of the plots. The alternative causal model, that generalist herbivory on seedlings drove generalist herbivory on Piperaceae and Annonaceae species in our plots (100% Piper shrubs, did not statistically fit the data. Effects of of plot seedling species in these two families are shared herbivores, shown in previous field studies (Dyer consumed by the generalists) than the species of and Letourneau 1999, Letourneau et al. 2004, Letour- Mimosoideae (33%) and Arecaceae (29%) (Dyer et al. neau and Dyer 2005, Marquis 2005) and forming the 2007; L. A. Dyer, unpublished data). basis for much theory (Holt 1977, Vandermeer 1980, DISCUSSION Orrock et al. 2010) are a likely mechanism for the outcomes we observed. However, none of our scenarios Forest understory diversity is dynamic, with local involve random seedling mortality, because that simple changes in abundance, richness, and evenness of juvenile process would result in a negative association between plants resulting from high rates of mortality and generalist herbivory and seedling abundance—the op- recruitment. Herbivores are an important component posite of what our results indicate. of seedling mortality in forest understories (Massey et al. One possible explanation for our results is that the 2005); for example, herbivores contributed to an average combination of generalist and specialist herbivores mortality of over 50% of all plants ,20 cm in height that feeding on Piper shrubs caused density dependent make up the community shared with Piper cenocladum mortality to all seedlings in the vicinity, such that shrubs (Letourneau et al. 2004, Letourneau and Dyer abundance and richness declined initially, followed by 2005). In the present study, experimental removal of greater than replacement numbers of new seedling generalist herbivores from naturally occurring Piper recruits. For species richness to increase along with shrubs resulted in surprisingly strong and predictable abundance, at least some of these recruits were in differences from control plots, given that these diverse unrepresented species. This process could result in the understory rain forest patches contain over 25 seedling somewhat lower evenness and the higher richness and species, on average, and overlap very little in composi- abundance seen in our studies if the relative distribution tion. An average 17% reduction in the similarity index of individual seedlings after herbivory was skewed by an between control and manipulated plots coupled with increase in singletons. A second scenario is similar to the extremely low error variances, suggests that the removal first, but assumes disproportionate mortality among of herbivores caused a shift in the plant community (e.g., plant species because of a spatial lottery model, similar to some of the significant shifts in plant preferential feeding by generalist herbivores, or differ- communities reported in Murdoch et al. 1972 and Hejda ential tolerance among seedlings. In this case, some et al. 2009, both of which had higher error variance). common species were thinned, allowing for new species This experiment and a path analysis on unmanipulated recruitment, but other common species persisted or patches showed that presence of generalist herbivores increased in numbers near Piper shrubs, again resulting and herbivore damage on Piper shrubs was positively in higher overall abundance and richness but lower associated with seedling richness and abundance and evenness. Both of these scenarios invoke apparent negatively associated with seedling evenness. competition among plants via shared herbivores, with December 2010 HERBIVORES AND LOCAL PLANT DIVERSITY 3715

PLATE 1. Piper cenocladum C.DC. (Piperaceae) in an experimental understory plot established at La Tirimbina Reserve, Costa Rica. The low levels of herbivore damage on this plant are characteristic of damage caused by Eois apyraria Guenee (Geometridae) and can be easily distinguished from other types of herbivore damage. White spots on the leaves are fresh damage (the underside of the leaf has been scraped away by caterpillars, leaving a window of clear tissue), and holes in the leaves are older damage (the clear tissue eventually falls out). The density and richness of plants and seedlings in this plot are typical of lowland understory vegetation in the region. Photo credit: L. A. Dyer. herbivores from Piper causing lower densities of some least some damage from herbivores that feed on Piper seedling species; interspecific competition is also part of shrubs. In particular, the herbivores from Piper may both scenarios since a decrease in the abundance of have caused density dependent mortality to all vulner- some species allows for recruitment or increases in able seedlings, causing an initial decline in seedling abundance of other species. Interspecific or apparent abundance and richness, followed by recruitment of high competition among seedlings has been documented in numbers of singletons of different species of seedlings some studies (e.g., McCarthy-Neumann and Kobe 2010) that are less susceptible to the Piper herbivores. Based but not found in others (e.g., Svenning et al. 2008). A on results from specific path analyses, it appeared that third possibility is that latent factors are associated with seedlings in the Fabaceae and Arecaceae were more herbivores or their effects on seedlings, and are tolerant or less susceptible to apparent competition with responsible for the quantitative effects on seedling Piperaceae than were those in the Annonaceae and abundance, richness and evenness. The possibility of Piperaceae—effects of generalist herbivores from Piper unmeasured abiotic factors that operate on a larger scale had very strong negative associations with evenness of favoring certain herbivores that simultaneously affect seedlings for the latter two families. We did not record adults and seedlings is not supported by the results of the identity of the generalist herbivores experimentally our generalist herbivore removal experiment, because removed from Piper, but it is possible that generalists removals were restricted to adult Piper shrubs. These avoided seedlings in these families disproportionately shrubs represented only one genus among a diverse compared to the other dominant families. In fact, the array of species of established understory plants, yet generalist caterpillars that feed on Piper have been removal of their generalist herbivores had measurable recorded feeding on all of the Piperaceae and Annona- and predictable effects on local seedling diversity in ceae species in our plots and less than one third of the these forests. seedling species in the Mimosoideae and Arecaceae. Our data best support the second scenario, which These results imply that there can be density dependent relies on disproportional mortality among species effects of some herbivores on Piperaceae and Annona- despite the likelihood that most seedlings experience at ceae seedlings, which is in contrast to other studies that 3716 LEE A. DYER ET AL. Ecology, Vol. 91, No. 12 have tested predictions of the Janzen-Connell hypothesis 2003, Letourneau et al. 2004). Whereas patterns of and have demonstrated density dependent effects of alpha diversity in trees at the scale of hectares can be herbivores on plants in the Mimosoideae (e.g., Condit et predicted in many cases without taking into consider- al. 1992, Uriarte et al. 2004) and Arecaceae (Condit et ation differences among species, Hubbell’s (2001) al. 1992, 1994, Cintra 1997, John et al. 2002) but not the ecological-equivalence assumption may be violated at Piperaceae (reviewed by Carson et al. 2008) or smaller scales (Uriarte et al. 2005). Complex interac- Annonaceae (Condit et al. 1994, Uriarte et al. 2004). tions between herbivores, plants, and plant resources Such apparent inconsistencies require focused research are likely to be responsible for major deviations from on forest understory plants, which have been less studied predictions of species assemblages based on neutral than tree seedlings, despite the high diversity of models (e.g., Volkov et al. 2003) especially for patterns understory vegetation (Wright 2002). found on smaller scales (Comita and Hubbell 2009). Apparent competition has been demonstrated when Our results in rain forest understory communities plants act as a source of phytophagous insects that can highlight the importance of incorporating these inter- damage or kill neighboring plant species (Parker and actions in a general dynamic theory of biodiversity. Root 1981, Lau and Strauss 2005). Tests of such ACKNOWLEDGMENTS apparent competition are not always framed in a community-level context or followed through to docu- This study was funded by NSF grant DEB-0344250 and CHE-0718732, as well as grants from Earthwatch Institute, ment eventual changes in plant diversity. Although National Geographic, and Tulane University. We thank the herbivore-seedling interactions are commonly demon- owners of El Bejuco and La Tirimbina, the Costa Rica National strated in a wide range of ecosystems (Dirzo and Parks, and the Organization for Tropical Studies (OTS) for use Miranda 1991, Hanley 1998, Lindquist and Carroll of their facilities. Excellent technical assistance was provided by 2004, Buschmann et al. 2005, Massey et al. 2005, Hanley Earthwatch volunteers and H. Garcia Lopez. We are grateful to T. Massad and three anonymous reviewers for insights that and May 2006), the appropriate empirical data for vastly improved previous drafts of the manuscript. demonstrating the effects of herbivores on plant diversity require long-term studies and intimate knowl- LITERATURE CITED edge of a community’s natural history. There are several Alexander, H. M., and R. D. Holt. 1998. The interaction excellent examples of studies demonstrating that insect between plant competition and disease. Perspectives in Plant Ecology, Evolution and Systematics 1:206–220. herbivores have strong, long-term impacts on diversity Bellow, J. G., and P. K. R. Nair. 2003. Comparing common (Carson and Root 2000, Schmitz 2003, 2004, Marquis methods for assessing understory light availability in shaded- 2005). Nevertheless, herbivore–diversity relationships perennial agroforestry systems. Agricultural and Forest most often focus on the effects of vertebrate herbivores Meteorology 114:197–211. (Dirzo and Miranda 1991, Howe et al. 2002) or seed Bernays, E. A., and M. Graham. 1988. On the evolution of host specificity in phytophagous arthropods. Ecology 69:1153– predators (as tests of the Janzen-Connell hypothesis, 1160. reviewed by Wright 2002, Carson et al. 2008), especially Bever, J. D. 2003. Soil community feedback and the coexistence in tropical forests. The potential role of insects, and of competitors: conceptual frameworks and empirical tests. particularly larval Lepidoptera, has not been studied New Phytologist 157:465–473. commonly in this context (Norghauer et al. 2010), but it Burger, W. 1971. Flora Costaricensis. Fieldiana Botany 35:1– 227. is likely that caterpillars play a role in determining plant Buschmann, H., M. Keller, N. Porret, H. Dietz, and P. J. diversity given their prominence in diverse communities Edwards. 2005. The effect of slug grazing on vegetation (Novotny et al. 2002, 2004). We would expect the development and plant species diversity in an experimental diversity effects due to immature insects moving from grassland. Functional Ecology 19:291–298. Carson, W. P., J. T. Anderson, E. G. Leigh, Jr., and S. A. source plants to be on a smaller scale than that proposed Schnitzer. 2008. Challenges associated with testing and by Janzen and Connell, but the concept is unchanged. falsifying the Janzen–Connell hypothesis: a review and Our results, showing that the removal of generalist critique. Pages 210–241 in W. P. Carson and S. A. Schnitzer, insect herbivores (that feed widely on plants in many editors. Tropical forest community ecology. Blackwell families) from Piper leads to decreased seedling Publishing, Oxford, UK. Carson, W. P., and R. B. Root. 1999. Top-down effects of abundance, decreased richness, increased evenness, insect herbivores during early succession: influence on and a shift in community composition, are consistent biomass and plant dominance. Oecologia 121:260–272. with Janzen-Connell effects operating in this rain forest Carson, W. P., and R. B. Root. 2000. Herbivory and plant system. These results add a new dimension to studies species coexistence: community regulation by an outbreaking that have shown that specialized herbivores will phytophagous insect. Ecological Monographs 70:73–99. Cintra, R. 1997. A test of the Janzen–Connell model with two increase plant diversity via decreasing biomass of a common tree species in Amazonian forest. Journal of dominant plant species (e.g., Dirzo and Miranda 1991, Tropical Ecology 13:641–658. Carson and Root 1999, 2000, Long et al. 2003, Schmitz Coley, P. D., and T. M. Aide. 1991. Comparison of herbivory 2003). Such indirect, trophic effects on biodiversity and plant defenses in temperate and tropical broad-leaved forests. Pages 25–49 in P. W. Price, T. M. Lewinsohn, G. W. have been shown to be relatively large and surprisingly Fernandes, and W. W. Benson, editors. Plant–animal common in complex, tropical communities (e.g., interactions: evolutionary ecology in tropical and temperate Schoener and Spiller 1996, Dyer and Letourneau regions. John Wiley and Sons, New York, New York, USA. December 2010 HERBIVORES AND LOCAL PLANT DIVERSITY 3717

Comita, L. S., and S. P. Hubbell. 2009. Local neighborhood Hanley, M. E., and O. C. May. 2006. Cotyledon damage at the and species’ shade tolerance influence survival in a diverse seedling stage affects growth and flowering potential in seedling bank. Ecology 90:328–334. mature plants. New Phytologist 169:243–250. Condit, R., S. P. Hubbell, and R. B. Foster. 1992. Recruitment Harms, K. E., S. J. Wright, O. Calderon, A. Hernandez, and near conspecific adults and the maintenance of tree and shrub E. A. Herre. 2000. Pervasive density-dependent recruitment diversity in a Neotropical forest. American Naturalist 140: enhances seedling diversity in a tropical forest. Nature 404: 261–286. 493–495. Condit, R., S. P. Hubbell, and R. B. Foster. 1994. Density- Hejda, M., P. Pysek, and V. Jarosik. 2009. Impact of invasive dependence in two understory tree species in a Neotropical plants on the species richness, diversity and composition of forest. Ecology 75:671–680. invaded communities. Journal of Ecology 97:393–403. Connell, J. H. 1971. On the role of natural enemies in Holt, R. D. 1977. Predation, apparent competition and the preventing competitive exclusion in some marine animals structure of prey communities. Theoretical Population and in rain forest trees. Pages 298–312 in P. J. den Boer and Biology 12:197–229. G. R. Gradwell, editors. Dynamics of populations. Centre Howe, H. F., J. S. Brown, and B. Zorn-Arnold. 2002. A rodent for Agricultural Publishing and Documentation, Wagenin- plague on prairie diversity. Ecology Letters 5:30–36. gen, The Netherlands. Hubbell, S. P. 2001. The unified neutral theory of biodiversity Denslow, J. S. 1987. Tropical rainforest gaps and tree species and biogeography. Princeton University Press, Princeton, diversity. Annual Review of Ecology and Systematics 18: New Jersey, USA. 437–451. Irschick, D., L. Dyer, and T. W. Sherry. 2005. Phylogenetic Dirzo, R., and A. Miranda. 1991. Altered patterns of herbivory methodologies for studying specialization. Oikos 110:404– and diversity in the forest understory: a case study of the 408. possible consequences of contemporary defaunation. Pages Janzen, D. H. 1970. Herbivores and number of tree species in 273–287 in P. W. Price, T. M. Lewinsohn, G. W. Fernandes, tropical forests. American Naturalist 104:501–528. and W. W. Benson, editors. Plant–animal interactions: John, R., H. S. Dattaraja, H. S. Suresh, and R. Sukumar. 2002. evolutionary ecology in tropical and temperate regions. John Density-dependence in common tree species in a tropical dry Wiley and Sons, New York, New York, USA. forest in Mudumalai, southern India. Journal of Vegetation Dyer, L. A. 1995. Tasty generalists and nasty specialists? A Science 13:45–56. comparative study of antipredator mechanisms in tropical Karban, R., and J. S. Thaler. 1999. Plant phase change and lepidopteran larvae. Ecology 76:1483–1496. resistance to herbivory. Ecology 80:510–517. Dyer, L. A., and G. L. Gentry. 2002. Caterpillars and Lau, J. A., and S. Y. Strauss. 2005. Insect herbivores drive parasitoids of a Costa Rican tropical wet forest. hhttp:// important indirect effects of exotic plants on native caterpillars.unr.edu/lsacatold/Families.htmi communities. Ecology 86:2990–2997. Dyer, L. A., G. L. Gentry, and M. A. Tobler. 2004. Fitness Letourneau, D. K. 2004. Mutualism, antiherbivore defense, and consequences of herbivory: impacts on asexual reproduction trophic cascades: Piper ant-plants as a mesocosm for of tropical rain forest understory plants. Biotropica 36:68–73. experimentation. Pages 5–32 in L. A. Dyer and A. D. N. Dyer, L. A., and D. Letourneau. 1999. Trophic cascades in a Palmer, editors. Piper: a model genus for studies of phyto- complex terrestrial community. Proceedings of the National chemistry, ecology, and evolution. Kluwer Academic/Ple- Academy of Sciences USA 96:5072–5076. num, New York, New York, USA. Dyer, L. A., and D. K. Letourneau. 2003. Top-down and Letourneau, D. K., and L. A. Dyer. 1998. Experimental test in bottom-up diversity cascades in detrital vs. living food webs. lowland tropical forest shows top-down effects through four Ecology Letters 6:60–68. trophic levels. Ecology 79:1678–1687. Dyer, L. A., and D. K. Letourneau. 2007. Determinants of Letourneau, D. K., and L. A. Dyer. 2005. Multi-trophic lichen diversity in a rainforest understory. Biotropica 39:525– interactions and biodiversity: beetles, ants, caterpillars, and 529. plants. Pages 366–385 in D. F. R. P. Burslem, M. A. Pinard, Dyer, L. A., and A. D. N. Palmer. 2004. Piper: a model genus and S. E. Hartley, editors. Biotic interactions in the tropics: for studies of phytochemistry, ecology, and evolution. their role in the maintenance of species diversity. Cambridge Kluwer Academic/Plenum, New York, New York, USA. University Press, Cambridge, UK. Dyer, L. A., et al. 2007. Host specificity of Lepidoptera in Letourneau, D. K., L. A. Dyer, and G. Vega Chavarria. 2004. tropical and temperate forests. Nature 448:696–699. Indirect effects of a top predator on a rain forest understory Fiala, A. C. S., S. L. Garman, and A. N. Gray. 2006. plant community. Ecology 85:2144–2152. Comparison of five canopy cover estimation techniques in Lewinsohn, T. M., V. Novotny, and Y. Basset. 2005. Insects on the western Oregon Cascades. Forest Ecology and Manage- plants: diversity of herbivore assemblages revisited. Annual ment 232:188–197. Review of Ecology, Evolution, and Systematics 36:597–620. Fincher, R. M., L. A. Dyer, C. D. Dodson, J. L. Richards, Lindquist, E. S., and C. R. Carroll. 2004. Differential seed and M. A. Tobler, J. Searcy, J. E. Mather, A. J. Reid, J. S. Rolig, seedling predation by crabs: impacts on tropical coastal and W. Pidcock. 2008. Inter- and intraspecific comparisons forest composition. Oecologia 141:661–671. of antiherbivore defenses in three species of rainforest Long, Z. T., C. L. Mohler, and W. P. Carson. 2003. Extending understory shrubs. Journal of Chemical Ecology 34:558–574. the resource concentration hypothesis to plant communities: Gentry, A. H., and L. H. Emmons. 1987. Geographical effects of litter and herbivores. Ecology 84:652–665. variation in fertility, phenology, and composition of the Maeto, K., and K. Fukuyama. 1997. Mature tree effect of Acer understory of neotropical forests. Biotropica 19:216–227. mono on seedling mortality due to insect herbivory. Greig, N. 1993. Predispersal seed predation on five Piper species Ecological Research 12:337–343. in a tropical rainforest. Oecologia 93:412–420. Magurran, A. E. 2004. Measuring biological diversity. Black- Greig, N. 2004. Introduction. Pages 1–4 in L. A. Dyer and well, Oxford, UK. A. D. N. Palmer, editors. Piper: a model genus for studies of Marquis, R. J. 1991. Herbivore fauna of Piper (Piperaceae) in a phytochemistry, ecology, and evolution. Kluwer Academic/ Costa Rican wet forest: diversity, specificity, and impact. Plenum, New York, New York, USA. Pages 179–208 in P. W. Price, T. M. Lewinsohn, G. W. Hanley, M. E. 1998. Seedling herbivory, community composi- Fernandes, and W. W. Benson, editors. Plant–animal tion and plant life history traits. Perspectives in Plant interactions: evolutionary ecology in tropical and temperate Ecology, Evolution and Systematics 1:191–205. regions. John Wiley and Sons, New York, New York, USA. 3718 LEE A. DYER ET AL. Ecology, Vol. 91, No. 12

Marquis, R. J. 2004. Biogeography of Neotropical Piper. Pages Parker, M. A., and R. B. Root. 1981. Insect herbivores limit 78–94 in L. A. Dyer and A. D. N. Palmer, editors. Piper:a habitat distribution of a native composite, Machaeranthera model genus for studies of phytochemistry, ecology, and canescens. Ecology 62:1390–1392. evolution. Kluwer Academic/Plenum, New York, New York, Rosenzweig, M. L. 1995. Species diversity in space and time. USA. Cambridge University Press, New York, New York, USA. Marquis, R. J. 2005. Impacts of herbivores on tropical plant SAS Institute. 2008. SAS version 9.2. SAS Institute, Cary, diversity. Pages 328–346 in D. Burslem, M. Pinard, and S. North Carolina, USA. Hartley, editors. Biotic interactions in the tropics: their role Schmitz, O. J. 2003. Top predator control of plant biodiversity in the maintenance of species diversity. Cambridge Univer- and productivity in an old-field ecosystem. Ecology Letters 6: sity Press, Cambridge, UK. 156–163. Massey, F. P., M. C. Press, and S. E. Hartley. 2005. Have the Schmitz, O. J. 2004. Perturbation and abrupt shift in trophic impacts of insect herbivores on the growth of tropical tree control of biodiversity and productivity. Ecology Letters 7: seedlings been underestimated? Pages 347–365 in D. Burslem, 403–409. M. Pinard, and S. Hartley, editors. Biotic interactions in the Schoener, T. W., and D. A. Spiller. 1996. Devastation of prey tropics: their role in the maintenance of species diversity. diversity by experimentally introduced predators in the field. Cambridge University Press, Cambridge, UK. Nature 381:691–694. McCarthy-Neumann, S., and R. K. Kobe. 2010. Conspecific Singer, M. S., and J. O. Stireman. 2005. The tri-trophic niche plant–soil feedbacks reduce survivorship and growth of concept and adaptive radiation of phytophagous insects. tropical tree seedlings. Journal of Ecology 98:396–407. Ecology Letters 8:1247–1255. Murdoch, W. W., F. C. Evans, and C. H. Peterson. 1972. Strong, D. R., J. H. Lawton, and T. R. E. Southwood. 1984. Diversity and pattern in plants and insects. Ecology 53:819– Insects on plants: community patterns and mechanisms. 829. Harvard University Press, Cambridge, Massachusetts, USA. Norghauer, J. M., J. Grogan, J. R. Malcolm, and J. M. Felfili. Svenning, J. C., T. Fabbro, and S. J. Wright. 2008. Seedling 2010. Long-distance dispersal helps germinating mahogany interactions in a tropical forest in Panama. Oecologia 155: seedlings escape defoliation by a specialist caterpillar. 143–150. Oecologia 162:405–412. Ullman, J. B. 1996. Structural equation modeling. Pages 709– Novotny, V., and Y. Basset. 2005. Review: host specificity of 811 in B. G. Tabachnick and L. S. Fidell, editors. Using insect herbivores in tropical forests. Proceedings of the Royal Society B 272:1083–1090. multivariate statistics. Third edition. Harper Collins, New Novotny, V., Y. Basset, S. E. Miller, R. L. Kitching, M. York, New York, USA. Laidlaw, P. Drozd, and L. Cizek. 2004. Local species richness Uriarte, M., C. D. Canham, J. Thompson, J. K. Zimmerman, of leaf-chewing insects feeding on woody plants from one and N. Brokaw. 2005. Seedling recruitment in a hurricane- hectare of a lowland rainforest. Conservation Biology 18: driven tropical forest: light limitation, density-dependence 227–237. and the spatial distribution of parent trees. Journal of Novotny, V., Y. Basset, S. E. Miller, G. D. Weiblen, B. Bremer, Ecology 93:291–304. L. Cizek, and P. Drozd. 2002. Low host specificity of Uriarte, M., R. Condit, C. D. Canham, and S. P. Hubbell. herbivorous insects in a tropical forest. Nature 416:841–844. 2004. A spatially explicit model of sapling growth in a Odegaard, F. 2000. How many species of arthropods? Erwin’s tropical forest: does the identity of neighbours matter? estimate revised. Biological Journal of the Linnean Society Journal of Ecology 92:348–360. 71:583–597. Vandermeer, J. H. 1980. Indirect mutualism: variations on a Odegaard, F., O. H. Diserud, S. Engen, and K. Aagaard. 2000. theme by Stephen Levine. American Naturalist 116:441–448. The magnitude of local host specificity for phytophagous Volkov, I., J. R. Banavar, S. P. Hubbell, and A. Maritan. 2003. insects and its implications for estimates of global species Neutral theory and relative species abundance in ecology. richness. Conservation Biology 14:1182–1186. Nature 424:1035–1037. Orrock, J. L., R. D. Holt, and M. L. Baskett. 2010. Refuge- Wada, N., M. Murakami, and K. Yoshida. 2000. Effects of mediated apparent competition in plant–consumer interac- herbivore-bearing adult trees of the oak Quercus crispula on tions. Ecology Letters 13:11–20. the survival of their seedlings. Ecological Research 15:219– Paletto, A., and V. Tosi. 2009. Forest canopy cover and canopy 227. closure: comparison of assessment techniques. European Wright, S. J. 2002. Plant diversity in tropical forests: a review of Journal of Forest Research 128:265–272. mechanisms of species coexistence. Oecologia 130:1–14.

APPENDIX A Damage to Piper leaves: a summary and key (Ecological Archives E091-260-A1).

APPENDIX B Seedlings found in 45 lowland forest fragment plots (Ecological Archives E091-260-A2).