Extending the Resource Concentration Hypothesis to Plant Communities: Effects of Litter and Herbivores Author(S): Zachary T

Extending the Resource Concentration Hypothesis to Plant Communities: Effects of Litter and Herbivores Author(s): Zachary T. Long, Charles L. Mohler, Walter P. Carson Source: Ecology, Vol. 84, No. 3 (Mar., 2003), pp. 652-665 Published by: Ecological Society of America Stable URL: http://www.jstor.org/stable/3107860 . Accessed: 04/03/2011 10:46

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at . http://www.jstor.org/action/showPublisher?publisherCode=esa. .

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Ecological Society of America is collaborating with JSTOR to digitize, preserve and extend access to Ecology.

http://www.jstor.org Ecology, 84(3), 2003, pp. 652-665 X 2003 by the Ecological Society of America

EXTENDING THE RESOURCE CONCENTRATION HYPOTHESIS TO PLANT COMMUNITIES: EFFECTS OF LITTER AND HERBIVORES

ZACHARY T. LONG,"3 CHARLES L. MOHLER,24 AND WALTER P. CARSON'

'Department of Biological Sciences, University of Pittsburgh, Pittsburgh, Pennsylvania 15260 USA 2Department of Ecology and Evolutionary Biology, Corson Hall, Cornell University, Ithaca, New York 14853 USA

Abstract. We extend the resource concentration hypothesis (herbivorous insects are more likely to find and stay in more dense and less diverse patches of their host plants) to plant communities. Specifically, whenever superior plant competitors spread to form dense stands, they will be found and attacked by their specialist insect enemies. This will decrease host plant abundance, causing a reduction in standing crop biomass, which will indirectly increase subordinate competitors and plant species richness. In this study, we found that a native, specialist chrysomelid beetle (Trirhabda virgata) in an old-field community decreased total standing crop biomass, leading to an increase in plant species richness. This reduction in biomass was due solely to a reduction in the biomass of the beetle's host plant, meadow goldenrod (Solidago altissima), which was the dominant plant species in this community. Our results demonstrate that when a superior competitor increases in density, the per-stem impact of herbivores increases due to a buildup of these herbivores in high- density host patches. Specifically, we found that as goldenrod density increased, the per- stem abundance of Trirhabda virgata also increased. In turn, species richness increased as the negative effect of insects on goldenrod biomass increased. Recent research suggests that litter accumulation could negate or cancel the effect of herbivorous insects on plant communities because litter accumulation increases with standing crop biomass, causing a decline in species richness. The litter accumulation hypothesis states that, in productive communities, the increase in the abundance of the superior competitor will lead to a dense accumulation of plant litter, causing a decline in species richness. Consistent with this hypothesis, we found that as the biomass of the dominant plant species increased, litter mass also increased. In turn, species richness decreased as the negative effects of litter on stem density increased. Interestingly, the effect of litter on stem density depended on whether insects were present. Our results suggest the potential for a general rule: specialist insect herbivores will function as classic keystone species in plant communities whenever host species form large, dense aggregations if host plants are dominant species. Key words: goldenrod; herbivory; insects; litter; old-field community; resource concentration hypothesis;Solidago altissima;Trirhabda virgata.

INTRODUCTION fects of herbivorous insects on plant communities. Un- Examples of herbivorous insects regulating terres- fortunately, few empirical data directly test this asser- trial plant communities are scarce, possibly because tion (Crawley 1997). Recent evidence suggests that predators and parasites usually keep insect abundance there are at least three conditions that strengthen the low, or because plants are well defended, or both links between insects and plants. These conditions fre- (Strong et al. 1984, Strong 1992, Hairston and Hairston quently occur during outbreaks (Carson and Root 1993, Schmitz 1994, Carter and Rypstra 1995, Polis 2000), whenever insects can muster an effective anti- and Strong 1996, Price 1997, Uriarte and Schmitz predator defense (Carson and Root 1999), and poten- 1998). Polis and Strong (1996) argued that the retic- tially when there are strong feedbacks between host ulate and diffuse nature of terrestrial food webs weaken plant density and per-stem rates of damage by insects trophic links and typically prevent strong top-down ef- (Root 1973, Andow 1991, Carson and Root 2000). The resource concentration hypothesis (Root 1973) Manuscript received 27 September 2001; revised 4 August describes a viable insect-plant feedback mechanism 2002; accepted 7 August 2002. Corresponding Editor: D. A. that may typically operate in plant communities. To Spiller. 3Present address: Department of Ecology, Evolution, and date, this mechanism has not been applied at the scale Natural Resources, Cook College, Rutgers University, 14 of entire plant communities. The resource concentra- College Farm Road, New Brunswick, New Jersey 08901-8551 tion hypothesis states that "herbivores are more likely USA. E-mail: [email protected] to find and remain on hosts that are growing in dense 4Present address: Department of Crop and Soil Sciences, 907 Bradfield Hall, Cornell University, Ithaca, New York or nearly pure stands" (Root 1973). Numerous popu- 14853 USA. lation-level studies lend strong support to this hypoth- 652 March 2003 HERBIVORY, LITTER, AND PLANT DIVERSITY 653

Insect | Resource } herbivores Dominant Plant species concentration: increase in plant richness is ceonnraton inrdensehot biomass is indirectly

F 1 patch a h reducedc pe increased

|Dominant plant|| species increases in abundance

R . , t Al L~~~~~~~ittern ~ ~ Totalstem | |Plant species| Titter ~~~~~accumulaton I HI density A Jrichness accumulation: ||increase WI| declines | | declines |

FIG. 1. The resourceconcentration model andthe litteraccumulation hypothesis make contrasting predictions. The resource concentration model predicts that an increase in host plant density leads to an increase in specialist insect herbivores and a subsequent decrease in host plant biomass. Insects indirectly increase plant species richness by reducing host plant biomass if the host plant is a dominant competitor. The litter accumulation hypothesis predicts that plant litter accumulation causes a decrease in total stem density, leading to a decrease in species richness.

esis (e.g., Root 1973, Risch et al. 1983, Andow 1991, herbivory, litter accumulation, and plant communities. Coll and Bottrell 1994, Schellhorn and Sork 1997). The goldenrod Solidago altissima is the dominant com- Furthermore, Bach (1980) found that the increased petitor in numerous old fields throughout the north- abundance of insects in host plant monocultures de- eastern United States (Bazazz 1996). This species often creased the growth of the host plant. Oddly, the logical forms dense, nearly monospecific stands with corre- extension of this hypothesis to entire communities has spondingly high net primary production and standing never been made. Here, for the first time, we extend crop biomass (Carson and Barrett 1988, Bazazz 1996, the resource concentration hypothesis to plant com- Carson and Root 2000); these stands result in communities. Specifically, if specialist herbivorous insects munities of low diversity (Bazazz 1996, Carson and are more likely to reduce the vigor and abundance of Root 2000). We tested if the density of the dominant host plants that form dense aggregations, and host plant species determined the abundance of herbivorous plants are dominant species in the community, then insects and their effect on plant diversity. We also test- specialist herbivorous insects could act as keystone ed whether dominant or subordinate plants primarily species, reducing the total standing crop biomass and determined litter mass and the effects of litter on plant increasing the diversity and fecundity of subordinate species richness. The factorial design of the experiment species. allowed us to integrate insect and litter effects. Litter accumulation, however, can have the opposite effect of herbivorous insects in plant communities. In- METHODS creases in plant community biomass are often associ- Study site ated with higher levels of litter accumulation (Carson and Barrett 1988, Tilman 1993, Foster and Gross 1997), We conducted this experiment in an abandoned ag- and a heavy accumulation of litter typically reduces ricultural field near Ithaca, New York, USA, (42?25' species richness (Carson and Peterson 1990, Foster and N, 76?31' W). The climate is humid continental, with Gross 1997, 1998, Carson and Root 2000). Although mean annual rainfall of 880 mm, relatively uniform all species present contribute to litter accumulation, the precipitation throughout the year, and an average frost- dominant plant should contribute more to litter accu- free growing season 146 d long from early May to early mulation than do other species in old fields. We there- October (Neeley 1961). fore tested the litter accumulation hypothesis, which The field was sprayed twice with an herbicide states that, in productive communities, an increase in (Roundup, Monsanto Corporation, St. Louis, Missouri, the abundance of the superior competitor will lead to USA) once in September and once in October of 1988, a dense accumulation of plant litter, causing a decline and plowed and disked in November. A small number in species richness (Fig. 1; Carson and Peterson 1990, of perennials survived this preparation and were killed Foster 1999, Suding and Goldberg 1999). with hand applications of the herbicide in the following We conducted a four-year insect and litter removal spring. Consequently, this experiment started in the experiment in a goldenrod-dominated old field in cen- first year of old-field succession, from bare ground with tral New York to understand relationships among insect species that established primarily from seed. 654 ZACHARY T. LONG ET AL. Ecology, Vol. 84, No. 3

Natural history of old fields D- The use of old fields as a model system has greatly contributed to our understanding of many ecological questions (e.g., Tilman 1982, Goldberg and Barton 1992, Bazazz 1996). Goldenrods typically dominate old fields throughout the eastern and midwestern Unit- ed States early in succession, and can persist as dominant species for decades (Mellinger and McNaughton 1975, Werner et al. 1980, Carson and Barrett 1988, Vankat-and Snyder 1991, Bazzaz 1996). Herbivorous insects have been shown to reduce the growth and re- U- production of goldenrod species (McCrea et al. 1985, Cain et al. 1991, Meyer 1993, Meyer and Root 1993, Brown 1994, Root 1996, Carson and Root 1999, 2000), and goldenrods typically suffer more damage than other old-field plant species (Root 1996, Carson and Root 1999, 2000). This study focuses on meadow goldenrod, Solidago altissima (or S. canadensis; see Kartesz 1994), a widespread, perennial, native, herbaceous species that supports >42 insect species with immature stages that specialize on goldenrod (Werner et al. 1980, Root and Cappuccino 1992). Two of these insects (Tri- rhabda virgata and Microrhopala vittata) are specialist chrysomelid beetles that are known to outbreak and FIG. 2. Schematic of the experimental setup in a goldenrod-dominatedold field in Ithaca,New York,USA. Twenty periodically reach high abundance (Root and Cappuc- 5 X 5 m plots were divided into 10 pairs. One plot in each cino 1992). pair received insecticide treatment(large hatched squares), and one plot in each pair was left untreated as a control (large Relationship to previous research open squares). Smaller I 1 m litter removal plots (slash pattern), manipulated litter controls (small squares), and pro- This study differed from the related, previous work cedural controls (litter removed and replaced,wavy pattern) of Carson and Root (1999) because it focused on a were randomlynested withinboth the largerinsecticide treat- goldenrod specialist (the leaf-feeding insect Trirhabda ed and insecticide controlplots. This design createdsix treatment combinations: insects and litter absent; insects absent virgata), rather than a generalist herbivorous insect and unamanipulated litter; insects absent and litter removed (Philaenus spumarius). Also, vegetation was sampled and replaced; insects present and litter absent; insects present in the fourth, rather than the third, year of succession and unmanipulated litter; insects present and litter removed (as in Carson and Root 1999). Carson and Root (2000) and replaced. examined the long-term effects (over 10 yr) of insect removal on midsuccessional old fields. Here we ex- [2000]). We mixed Fenvalerate with water and sprayed amined effects during the establishment phase of suc- it on the 10 randomly chosen plots at the recommended cession, specifically the first four years of succession. concentration of 300 g Fenvalerate/ha (Root 1996). We By using methodologies similar to these studies, yet sprayed only during calm periods early in the morning varying year, age of field, and the dominant herbivorous or late in the evening (to avoid pesticide drift) at 2-wk insect, we hoped that our results would lead to the most intervals from late April to mid-September through the robust, general conclusions. first four years of succession. We sprayed the untreated control plots with an equivalent amount of water. This Design application rate significantly reduced both insect her- Insecticide application.-The experiment used a bivore loads and levels of insect plant damage on the two-factor split-plot design with two levels of insect common forbs and grasses (Root 1996, Carson and abundance (present or removed) as the main plots and Root 1999, 2000). In the untreated control plots, sub- two levels of litter (present or removed) as the subplots stantial insect damage was found only on S. altissima (Fig. 2). We divided the field into 20 5 X 5 m main and two uncommon understory forbs (Carson and Root plots, arranged in 5 X 4 array and separated by 2-m 1999). buffer strips. We randomly selected one plot from pairs Litter.-We manipulated litter in three i X 1 m sub- of adjacent main plots for application of the insecticide plots haphazardly nested within each of the 5 X 5 m Fenvalerate. Fenvalerate is a broad-spectrum synthetic main plots. We had three types of litter treatments: litter pyrethroid insecticide with no substantive side effects removal, procedural control, and an unmanipulated on the plants (for extensive details regarding the use control. We removed all litter in the litter removal sub- of this pesticide, see Root [1996] and Carson and Root plots by hand in early spring without disturbing the March2003 HERBIVORY,LITTER, AND PLANT DIVERSITY 655 soil surface. In the procedural control subplots, we re- analyses, we used the density of goldenrod in 1991, moved and then replaced all litter to determine if our rather than in 1992, because the specialist insects that procedure for removing the litter had any additional typically attack this species in the spring of any given influence beyond the intended effects of removing lit- year (e.g., Trirhabda beetles) were oviposited on these ter. The control subplots were left undisturbed. We re- goldenrods in the fall of the previous year and over- moved litter on 16-18 April 1990, 7-9 April 1991, and winter as eggs at the base of their host (Herzig 1995, 10-12 April 1992. We did not remove litter in the Herzig and Root 1996). Thus, these beetles were re- spring of 1989 because succession was starting from sponding to the 1991 density of these goldenrods, not tilled soil devoid of litter. their density in 1992. Measurements of herbivore abundance.-We quan- Statistical analyses and tests of hypotheses tified damage on 10 common old-field plant species during the first three years of this study and also vi- Species richness and abundance.-We tested for sually surveyed the unsprayed control plots for damage overall effects of insect suppression, litter manipulation on locally rare species (Carson and Root 1999). Our (unmanipulated control, procedural control, removed), damage surveys uncovered extensive damage on two and their interaction on five response variables includ- locally uncommon (<1% of total biomass) forbs (Plan- ing plant species richness, total plant biomass, total tago major and Rumex crispus); however, we never plant density, and biomass and density of goldenrod found damage >3% on any common species other than with a split-plot ANOVA (PROC GLM in SAS/STAT on the dominant species, S. altissima. Thus, we cen- v. 6.12; SAS Institute 1997). Insecticide treatment was sused herbivorous insects on S. altissima, in plots with- completely randomized with respect to each pair of out insecticide in early June 1992, because only this plots, and litter removal was nested within each plot. species was subject to heavy attack (for similar results Testing predictions of the resource concentration in two other nearby old fields, see Carson and Root model.-We used linear regression to test if abundance [1999]). Herbivore loads measured in June are a reli- of Trirhabda per stem in 1992 varied with goldenrod able indicator of the degree of damage experienced by density in 1991 (the year when Trirhabda oviposited), plants over the season because loads in spring are typ- as predicted by the resource concentration model. In- ically higher than in the summer (Root 1996). Our sam- sect abundance data and goldenrod densities were col- pling method is explained and justified in detail else- lected in unmanipulated control plots. We calculated where (Root and Cappuccino 1992, Root 1996) and the effect of herbivorous insects as an index, based on here is reviewed briefly. The exact timing of the insect goldenrod biomass, for each of the 10 pairs of main census was linked to insect phenology, and occurred plots: in each year soon after the larvae of the chrysomelid beetle Trirhabda virgata (a dominant herbivore that Insect effect index = (H+L+ - H-L+) maximum of (H+L+ or H-L+) specializes on goldenrod) molted into the third instar. We selected five goldenrod stems in each insecticide- where H+L+ was the biomass of goldenrod in a subplot sprayed plot and eight goldenrod stems in each control with herbivorous insects and unmanipulated litter (plus by haphazardly pointing a meter stick to the ground sign indicates presence), and H-L+ was the biomass of while looking away and then locating the stem nearest goldenrod in the paired subplot without insects (i.e., the base of the meter stick. We then carefully searched insecticide-treated) and unmanipulated litter (sensu each stem for herbivores and measured stem length. Wilson and Keddy 1986, Bonser and Reader 1995, Chase et al. 2000). Dividing by the maximum value of Vegetation sampling H+L+ or H-L+ constrained this index between 1 and We sampled total aboveground plant biomass in one -1, where 1 indicated complete facilitation (herbivory randomly selected 0.125 m2 circular quadrat in each promotes the abundance of goldenrod), and -1 indi- subplot during the summer of 1992, four years after cated complete inhibition (herbivory decreases the the start of the manipulations. Species richness was abundance of goldenrod). We performed the identical calculated as the number of species present in the bio- calculation for procedural control plots, substituting mass samples of each subplot. Samples of this size goldenrod biomass in procedural control plots for bio- provided conclusive results in comparisons of biomass mass in unmanipulated control. Thus, for procedural and species richness in several companion studies control plots, the insect effect index was the difference (Root 1996, Carson and Root 1999, 2000). Biomass between goldenrod biomass in insecticide-sprayed and samples were dried at -80'C for 48 h and weighed to unsprayed procedural control plots, divided by the the nearest 0.1 g. In 1992, we measured stem density maximum of those two values. of all species in eight of the 10 pairs of main plots in We used linear regression to determine if inhibition randomly selected 20 X 20 cm sample plots placed of goldenrod by insects increased with increasing gold- ?20 cm from the edge of each 1 X 1 m subplot. We enrod density, as predicted by the resource concentra- also sampled the density of goldenrod in 1991 in ran- tion hypothesis. Separate analyses were performed for domly selected I X 1 m quadrats. For all statistical unmanipulated control and procedural control plots. In 656 ZACHARY T. LONG ET AL. Ecology, Vol. 84, No. 3 order to determine if the effect of herbivorous insects TABLE 1. Mean (+ 1 SE) litter mass removed in insecticide- led to changes in species richness, we regressed species sprayed and control plots in each year. richness on the insect effect index. Separate analyses Mass of litter removed were performed for unmanipulated control and pro- (g/m2) cedural control plots. Year Spray Control t P Testing predictions of the litter accumulation hy- 1990 247 + 26.7 247 + 24.6 0.34 0.97 pothesis.-The litter accumulation hypothesis states 1991 320.2 + 26.8 223 + 15.4 4.04 0.003 that litter mass increases with biomass of the dominant 1992 375.7 + 32.3 203.1 + 13.8 5.20 0.0006 plant species, and the effects of litter on stem density Note: We compared the means each year with a paired t and plant species richness should increase with litter test (df = 10; Bonferroni-adjusted critical value of P = mass. We used analyses of convariance (ANCOVA) to 0.017). test if litter mass varied with dominant or subordinate plant biomass, and if insects affected either of these We used ANCOVA to determine if increased litter relationships. We used the presence or absence of in- mass led to a greater inhibition of density, as measured sects as the categorical explanatory variable, and the by the litter effect index and the effects of insects. We continuous explanatory variable was either goldenrod used the presence or absence of insects as the cate- biomass or subordinate plant biomass. Litter mass was gorical explanatory variable, litter mass as the contin- the response variable. Separate analyses were per- uous explanatory variable, and the litter effect index formed for unmanipulated control and procedural con- as the response variable. Separate analyses were per- trol plots. formed for unmanipulated control and procedural con- We calculated the effect of litter, as an index based trol plots. We regressed species richness on the litter on total stem density, to test if the effect of litter in- effect index to determine if the effect of litter led to creased with litter mass. Because litter manipulations changes in species richness. Separate analyses were were nested within insecticide we cal- manipulations, performed for unmanipulated control and procedural for with and without culated the litter effect index plots control plots. herbivorous insects. By testing the effect with and without herbivorous insects, we could determine whether RESULTS the litter effect was different when insects were present Insect and litter abundance vs. when they were absent. For plots with insects present, Trirhabda virgata, a chrysomelid beetle that specializes on goldenrods, was the dominant herbivore (H-L+ - H+L-) litter effect index during the fourth year of succession (1992), accounting maximum of (H+L+ or H+L-) for >90% of the insects on goldenrod by abundance in the insecticide control plots. Abundance of Tri- and for insecticide-treated plots, rhabda virgata was 4.51 ? 1.31 individuals/stem in control plots. Damage surveys in control plots in a litter effect index = (H-L - H-L-) maximum of (H-L+ or H-L-) previous, related study demonstrated that insect damage >3% was restricted to Solidago altissima (Carson where H+L+ and H-L+ were defined as before; H+L- and Root 1999). The insecticide treatment was highly was total stem density in the subplot with insects and effective at reducing insect damage (see also Carson with litter removed; and H-L- was total stem density and Root 1999); indeed, we never found insects for- in the subplot without insects and unmanipulated litter aging on goldenrods in sprayed plots in our censuses. (sensu Foster and Gross 1997, Suding and Goldberg Litter mass did not differ between sprayed and control 1999). Dividing by the maximum value of H-L+ or plots in the first year of removal, but was greater in H-L- constrained this index between 1 and -1, where sprayed plots than in plots with insects in the last two 1 indicated complete facilitation (litter promotes total years of removal (Table 1). stem density), and -1 indicated complete inhibition (litter decreases total stem density). We also calculated The influence of herbivory and litter on species this index for procedural control plots. This calculation richness and plant abundance is identical to the previous one, except that stem density Plant species richness, density, and biomass.-Litter in procedural control plots replaced stem density in and herbivorous insects had opposite effects on plant unmanipulated litter plots. Density data was collected species richness. Litter significantly decreased plant in only eight of the 10 main plots. Therefore, we cal- species richness, whereas insect herbivory significantly culated 16 indices based on unmanipulated litter con- increased richness by -4-5 species (Fig. 3). Litter and trol plots (eight for insecticide-treated plots and eight herbivory affected richness independently (i.e., there for insecticide-free control plots) and 16 indices based was no significant interaction). on procedural control plots (eight for insecticide-treat- Litter accumulation significantly decreased total ed plots and eight for insecticide-free control plots). stem density, whereas herbivorous insects had no sig- March2003 HERBIVORY,LITTER, AND PLANT DIVERSITY 657

- 30 _ _ between the effect of insects and Insects goldenrod density was No insects statistically significant (F1,8 = 6.90, P = 0.03, r2 = rA 25- 0.46). We have no biological reason to remove this point, and therefore do not exclude it from the analysis. 2020

*1 15 250 * 10 20 .Insects A >> 200 _ No insects 5- 150 0 Litter Procedural No litter control 100 FIG. 3. The influence of herbivorousinsects and plant H litter on mean (?1 SE) species richness. The presence of 250 herbivorousinsects increased species richness (F19g= 9.89, P = 0.012). The presence of plant litter decreased species richness (F236 = 12.24, P < 0.0001). There was no significant 0 interaction between insects and litter = 0.85, P = (F2,36 250 0.437). "0 B c 200 nificant effect on total stem density (Fig. 4A). Litter accumulation still decreased total stem density when E 150 goldenrod density was removed from the analysis (Fig. 4B), and did not affect goldenrod stem density (Fig. E 100 4C). This suggests that the reduction of stem density by litter had little to do with goldenrod density. The : 50 presence of litter increased aboveground plant biomass, whereas herbivorous insects significantly decreased total aboveground plant biomass (Fig. 5A). There was E 0 no significant interaction (Fig. 5A). The reduction of 250 total aboveground plant biomass by insect herbivory C was explained entirely by the reduction in goldenrod 200- biomass (Fig. 5B). There was no significant effect of either litter removal or insect suppression when gold- CA enrod biomass was removed from the analysis (Fig. 5B). Herbivory significantly decreased goldenrod biomass, but litter and the interaction between litter and C; 100 insects did not significantly affect goldenrod biomass (Fig. 5C). 13 50

Testing the resource concentration model Per-stem abundance of Trirhabda in 1992 increased Litter Procedural No litter control with goldenrod density, as measured in 1991 (Fig. 6A). We also found that the per-stem percentage of leaf tis- FiG. 4. (A) The influence of herbivorous insects and plant litter on mean (?+1 SE) total stem sue damaged increased with increased goldenrod den- density. Herbivorous insects did not affect total stem density (Fl7 = 0.43, P = 0.535). sity in 1991 (Fig. 6B; for the methods used to estimate Plant litter decreased total stem density (F228 = 18.87, P < leaf damage, see Carson and Root [1999, 2000]). 0.0001). The interaction between insects and plant litter was The negative impact of insects on the biomass of marginally significant (F228 = 2.51, P = 0.099). (B) Com- goldenrod did not increase with goldenrod density in parison of mean stem density with goldenrod stems was not included. Removing goldenrod from the analysis did not the unmanipulated litter control plots (Fig. 6C). How- change these results for insects (F1,7 = 0.93, P = 0.367) or ever, the point in the top right corner of the figure was litter (F228 = 20.84, P < 0.0001) but did change the signif- identified, statistically, as an outlier because its stu- icance of the interaction (F2,28 = 2.37, P = 0.112). (C) Gold- dentized residual exceeded the Bonferroni (alpha-split- enrod stem density was not affected by herbivorous insects, = ting) correction critical value for determining outliers litter, or the interaction between insects and litter (F1,8 0.82, P = 0.396 for effects of insects; F2,28 = 1.61, P = 0.218 for (Belsley et al. 1980, Kleinbaum et al. 1998). When this effects of litter; F2,28 = 1.65, P = 0.211 for the interaction point was excluded from the analysis, the relationship between insects and litter). 658 ZACHARY T. LONG ET AL. Ecology, Vol. 84, No. 3

300 Most importantly, species richness increased as the X; = | Insects A negative effect of insects on goldenrod became more 250 N o insects severe in the unmanipulated litter control plots (Fig. 2300 * 7A). Specifically, when the insect effect index was negative (inhibitive), species richness was much higher. 0 E 150 There was no relationship between the insect effect . 0 index and species richness in the procedural control 2001 plots (Fig. 7B).

> 50 Testing the litter accumulation hypothesis 0 Litter mass was accidentally not recorded for one procedural control plot. Therefore, we did not include 300 data from this plot in any analysis. Litter accumulation _ ~~~~~~~~B increased as goldenrod biomass increased in both un- ? 2 250 manipulated litter control plots and procedural control 200 "0 plots, and insects did not affect these relationships (Fig. 00E 200 8A, B). There was a marginally significant decrease in "0150 litter accumulation with subordinate plant biomass in litter control plots, and insects did not affect these relationships (Fig. 8C). Litter accumulation was not related to subordinate plant biomass or insects in pro- 0;'500 t cedural control plots (Fig. 8D). We predicted that the effect of litter on total stem density, measured as the litter effect index, would increase with litter mass, but 300 C we did not find this relationship in litter control plots (Fig. 9A). The litter effect index was significantly more 2 250 negative (more inhibitive) in the absence of insects in EE200 litter control plots; this indicates that insects alleviated the effect of litter on stem density (Fig. 9A). We did 0-G. 50 A h nlec fhrbvru net n ln find a significant relationship between the litter effect "00 index and litter mass in procedural control plots, and O 50 this relationship depended on the presence or absence 0c of insects (significant interaction; Fig. 9B). As in the litter control plots, insects independently alleviated the effect of litter on stem density (Fig. 9B). As predicted by the litter accumulation hypothesis, species richness decreased as the effect of litter became more inhibitory in both litter control and procedural control plots (Fig. litter on mean (+- I SE) total aboveground plant biomass. Her- 10). bivorous insects decreased aboveground plant biomass (F,,g = 7.69, P = 0.022). Plant litter marginally significantly in- DISCUSSION creased aboveground plant biomass (F,,36 = 2.95, P = 0.065), and the interaction between insects and plant litter 'was not The top-down effect of herbivorous insects = 0. P = The influence of significant (F2,36 16, 0.850). (B) Herbivorous insects decreased total aboveground herbivorous insects and litter on mean total above- plant and increased richness in this 4- ground plant biomass without goldenrod biomass. There was plant biomass species no significant difference in biomass between plots with and yr-old goldenrod-dominated old field. The decrease in without insects or litter when the goldenrod biomass was total plant biomass was accounted for entirely by the removed from the analysis (Flte = 0.58, P = 0.466 for insects; decrease in goldenrod biomass. These results are sim- = P = 0.205 for F236 = 1.17, P = 0.322 for F236 1a66, litter; ilar to those in herbaceous communities in England the interaction). (C) The influence of herbivorous insects and plant litter on mean (0.16SE) aboveground goldenrod biomass. (Henderson and Clements 1977, Brown et al. 1988, Herbivorous insects decreased aboveground goldenrod bio- Brown and Gange 1989, 1992) and North America mass (F,,g = 7.65, P = 0.022). Removing litter did not affect (McBrien et al. 1983, Carson and Root 1999, 2000), = = aboveground goldenrod biomass (F236 0.45, P 0.638), where herbivorous insects increased species richness and the interaction between insects and plant litter was not because they disproportionately and negatively affect- significant (Fom36= 1.a87, P = 0. 169) . ed the dominant native plant species. This adds to a growing list of studies that document clear examples of top-down effects of herbivorous insects on reducing standing crop biomass (Henderson and Clements 1977, Chase 1998, Carson and Root 1999, 2000). March2003 HERBIVORY,LITTER, AND PLANT DIVERSITY 659

12 These findings run counter to much theoretical and A empirical work that suggests that factors such as pred- 1 20 ators and plant defenses limit top-down effects of herbivorous insects on terrestrial plant communities 8 (Strong et al. 1984, Strong 1992, Hairston and Hairston 1994, Carter and Rypstra 1995, Polis > 6 1993, Schmitz and Strong 1996, Price 1997, Uriate and Schmitz 1998). 4 Strong (1992) argues that the reticulate trophic connections in terrestrial communities prevent runaway 0~~~~~ consumption and reduction of standing crop in terrestrial systems. Predators and plant defenses may limit 10 the effect of herbivorous insects on plant communities, 4-~~~~~ but may not always obviate top-down effects. We sug- 25 B gest that some conditions strengthen trophic links between insects and their host plants, and allow for top- ~20 down effects. We can identify at least three sets of 0.8 C conditions, none of which is mutually exclusive, and 15 which may occur frequently: (1) during insect out- E 0. 2 breaks, when outbreaks are common relative to the life r~~0. ~ 0. 0 A) Control

t 20

C10 *

0.8 C 25

-10.0 0 0 - 1.0 -0. 8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 Insect effect index

0 2 4 6 8 10 12 14 16 18 20 22 25 B) Proceduralcontrol Goldenrod density in 1991 (no. stems/rn2) ( 20 - FIG. 6. (A) Per-stem abundance of Trirhabdain 1992 increased with goldenrod density measured in 1991 (F.1a= 6.90, P = 0.030, r2 = 0.46). (B) Per-stem percentage of leaf tissue 1= 15 - damaged in 1992 showed a marginally significant increase with goldenrod density measured in 1991 (F1,9 = 4.73, P = 0.07, r2 = 0.40). (C) Insect effects on goldenrod biomass did > 10 not increase with goldenrod density measured in 1991 (F1,9 = 0.01, P = 0.9 15, r2 < 0.01). Insect effects were based on goldenrod biomass and were constrained between values of 5- 1 and -1, with 1 indicating complete facilitation by insects and -1 indicating complete inhibition or consumption by 0 insects (see Methodsfor details). The point in the upper right -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 corner of the graph was statistically identified as an outlier. When this point was removed from the analyses, insect effects Insect effect index significantly increased with goldenrod density (F18 = 6.90, FIG. 7. Plant species richness increased as the effect of P = 0.03, r2 = 0.46; see Results for further details). insects on goldenrod became more inhibitive in the litter control plots (F19 = 4.88, P = 0.058, r2 = 0.38). Specifically, when the insect effect index was negative (inhibitive), species richness was much higher. This relationship was not observed in the procedural control plots (F19 = 0.94, P = 0.361, r2 = 0.10). Effects of insects on goldenrod are as defined in Fig. 6C. 660 ZACHARY T. LONG ET AL. Ecology, Vol. 84, No. 3

A) Control B) Proceduralcontrol 140 ~-.120 V 0100 V ~~80 VYV G0 b 20VNoiset 60

VV 40 V VV V 20 vv Vynsect

20 0 50010 100 150 200 250 300 350 0 50 100 150 200 250 300 350 BiomasswitoutGims ihuoldenrod bgm2gioldenro (ithmut(g/M2) iomass BiomasswitoutG oldenrodbims gioldenro(g/m(g/m2) 2)

inceasngincreasiC) Codnooldnro bimsCims npoeuaontrol pltD(18=D)tF 1 .5 ProeuaPro0.019),co.09)ndthsro aondthsrelainhpwsotfecdlainhpwsotfecd byhrioos asmrinlysgiianl1eraedwt4nrain0uodntnet F1 =14,P=0.4) C ite pln ims (oa ims wtotgleno ims)inlte otrlpos(19 .5 000,adtisrltosi wa o fete ite1aswsnt2eae0osbodntyhrioou net F19=20,P=0.6) D ln bims npoeurlcnrlpos F1 .7 =036,adti rltosi a ntafce yhebvru net (F(F181 01000 19, P .65 V

span of the plants (Carson and Root 2000); (2) when- patches of goldenrod (sensu Morrow et al. 1989, Herzig ever herbivorous insects can defend themselves from and Root 1996). The increase in these beetles in high- predators (Carson and Root 1999); and (3) when tro- density patches is not surprising because these beetles phic connections between herbivorous insects and their tend to preferentially colonize dense and undamaged host plants are modified by negative feedbacks that patches of their host plants (Morrow et al. 1989, Herzig occur when plant resources are highly concentrated and 1995, Herzig and Root 1996, Carson and Root 2000). herbivores can track those resources. Indeed, Trirhabda beetles can fly for many kilometers to colonize lush and undamaged goldenrods (Herzig model of plant The resource concentration and Root 1996). community regulation We did not observe the predicted negative relation- The resource concentration model predicts that the ship between the impact of insects and goldenrod den- per-stem abundance of herbivorous insects, and their sity (Fig. 6C). If we exclude the data point in the upper subsequent effect on their host plant, should increase right corner of the graph, based on its statistical iden- with host plant density. If the host plant is competi- tification as an outlier, then the relationship between tively dominant, then insects will have a top-down ef- the effect of insects and goldenrod density becomes fect on species richness. In this study, we found that significant (see Results for further details). We have no the per-stem abundance of the dominant herbivorous biologically based reason to exclude this point. How- insect, Trirhabda virgata, and the percentage of leaf ever, given that we found an increase in both Trirhabda damage increased with goldenrod (Solidago altissima) and the damage that they caused with increasing gold- density (Fig. 6B). The increase of this univoltine insect enrod density, it would certainly not be surprising if probably resulted from attraction to the high-density they ultimately cause a concomitant decrease in den- March2003 HERBIVORY,LITTER, AND PLANT DIVERSITY 661

A) Control a similar relationship between density and damage in a much older goldenrod-dominated old field a few ki- 0.0 V v Insects lometers away from the field considered in the present x s v V No insects study. Overall, these responses suggest that this is a Hi W).2 v V fairly robust result for goldenrod-dominated communities. . -0{.6 - c) y V Interestingly, two data points suggest that insects had strong facilitation effects on goldenrod biomass (insect - -0.8 effect index -1; Fig. 6C). One of these points is the XV V statistical outlier just discussed, and we have no biological explanation for this observation. The second occurred in a patch of low goldenrod density. The in- - -1.0-0.0 V dividuals of goldenrod in this patch experienced less 0 20 40 60 80 100 120 140 yV ~ ~ damage by insects. The slight damage experienced by Litter V mass (g/m2) these goldenrods may have caused changes in alloca- "0 ~~~~~VyV tion, or increased light to B) Proceduralcontrol availability underlying leaves, allowing for compensatory growth (e.g., Belsky 1986). -0.8- Most importantly, as the negative influence of the beetles on goldenrod biomass increased, plant species

Control 30 A) V V 253 v I) V~ - -1.0 c20 0 20 40 60 80 100 120 140 B 15 V Littermass (g/m2) FIG. 9. (A) The litter effect index was not relatedto litter = 10 mass in litter control plots (Fl15 = 0.01, P = 0.937). Insects marginallysignificantly lessened the effect of litter on stem - 5 V Insects density in litter control plots (F115 = 4.08, P = 0.065). (B) The litter effect index was relatedto littermass in procedural V No insects 0 control plots (F114 = 11.65, P = 0.005). Insects lessened the -1.0 -0.8 -0.6 -0.4 -0.2 0.0 effect of litter on stem density (F114 = 11.97, P = 0.005) and increasedthe slope of the relationshipbetween the effect of Littereffect litter and litter mass (significant insect x litter mass interaction, F.1,4 = 5.12, P = 0.045). The litter effect index was B) Proceduralcontrol based on total stem density and was constrainedbetween 30 values of 1 and -1, with a 1 indicatingcomplete facilitation by litter and -1 indicating litter reduced stem density to 0 25 V (see Methodsfor details). Filled circles indicate the litter effect in the presence of insects; open circles indicate litter effect in the absence of insects. 20 V- VVV~~ .~15 -V sity. The decrease in density may take some time to occur, however. Carson and Root (2000) found a two- = 10 year lag between declines in density and damage caused by a related specialist chrysomelid beetle (Mi- 5- crorhopalavittata). Goldenrod is a widespread, native 0 perennial that is host to more than 40 specialist insects -1.0 -0.8 -0.6 -0.4 -0.2 0.0 (Root and Cappucino 1992). Two of these, Trirhabda Littereffect virgata and Microrhopala vittata, have now been shown to substantially reduce the biomass of this spe- FIG. 10. (A) Plant species richness showed a marginally cies, and both species cause greater damage in denser significantincrease as the inhibitiveeffects of litterdecreased in litter control-plots (F, 15 = 4.16, P = 0.060, r2 = 0.23). stands (McBrien et al. 1983, Carson and Root 2000). (B) Plant species richness significantly increased in proce- In a related study, Carson and Root (2000) also found duralcontrol plots (F, 14 = 5.13, P = 0.041, r2 = 0.28). 662 ZACHARY T. LONG ET AL. Ecology, Vol. 84, No. 3 richness increased in unmanipulated litter plots (Fig. Snyder 1991, Bazzaz 1996). A number of potential 7A). The increase in plant species richness that resulted answers exist. First, smaller patches within a field, or was probably caused by an increase in resources, par- individual clones, may indeed be decimated. Because ticularly light, that were freed up by the reduction in of the time lag between increases in plant density and the dominance and abundance of goldenrod (Carson their discovery by insects, less dense patches may in- and Pickett 1990, Brown 1994, Carson and Root 2000). crease in density concomitant with the decimation of Interestingly, there was no relationship between the other patches. Thus, observation at the field scale may effect of insects and species richness in procedural con- suggest persistence of the dominant plant. Second, host trol (litter removed and replaced) plots (Fig. 7B), even plants may become more resistant with subsequent at- though we did find that the presence of insects lowered tacks, or more resistant plants may spread. Third, it species richness in the procedural control plots (Fig. makes sense that if the herbivorous insects can track 3). This may have occurred because Trirhabda over- host plant density, eventually their predators should winter as eggs in the litter layer (Herzig 1995). Dis- track their density. When this occurs, predators may rupting, removing, and replacing litter prior to their prevent the decimation of host plant patches (Strong emergence in the spring may have removed eggs from 1992). However, it is important to stress again that the plots as litter was excavated from the plot, placed herbivorous insects need not decimate host plant pop- outside the plot, and then replaced back into the plot. ulations to regulate plant communities (e.g., Brown et It is possible that the procedure itself may have caused al. 1988, Brown and Gange 1989, 1992, Carson and egg mortality. Similarly, Facelli (1994) found that the Root 2000). Herbivorous insects need only reduce the presence of plant litter increased arthropod abundance, competitive impact of dominant host plants on sub- activity, and subsequent damage on tree seedlings. ordinate plants to allow coexistence of the subordi- Clearly, more research is required to sort out the com- nates. plex patterns by which litter, insects, and their inter- Generalist vs. specialist herbivorous insects.-We actions influence plant communities. suggest that keystone predation models based primarily The resource concentration model explains how in- on productivity (e.g., Armstrong 1979, Holt et al. 1994, sects could commonly play a major role in regulating Leibold 1996) may apply more to highly polyphagous plant communities by causing greater damage when species that may increase with net primary production host plants that are superior competitors spread and (Ritchie 2000), whereas the resource concentration become dominant. This regulation would only be model may apply more to specialist insects that respond "switched on" when dominant species become dense to host plant density. Generalist herbivores can respond and are then found by their insect enemies. Without to productivity. For example, Chase et al. (2000) found the mitigating influence of insects, these plant species an increase in the effect of large mammalian herbivores would depress the abundance of, or exclude, subordi- on plant communities over a broad increase in pro- nate plant species (sensu Carson and Root 2000). Reg- ductivity. Similarly, McNaughton et al. (1989) found ulation of plant communities by the resource concen- that biomass of herbivores and consumption by her- tration model could be a very general phenomenon. bivores increased with primary productivity. For in- The per capita abundance of specialist herbivorous insects, Carson and Root (1999) found that the highly sects is often greater in monocultures than polycultures polyphagous spittle bug (Philaneous spumarious) sig- (e.g., Root 1973, Risch et al. 1983, Andow 1991, Coll nificantly reduced goldenrod abundance at this same and Bottrell 1994, Schellhorn and Sork 1997) and this Ithaca field site and two others, but there was no re- increase in insect abundance can lead to an increase in lationship between plant density and spittle bug abun- per capita damage on host plants and reductions in plant dance (F19 = 1.44, P = 0.26, I?2 = 0.15). As just vigor (Bach 1980, Andow 1991, Ramert and Ekbom outlined, monophagous insects respond to host plant 1996, Carson and Root 2000). If native plants that are density, and their effects on plant communities are bet- superior competitors are also widespread, which is typ- ter predicted by the resource concentration model. It ical, then these plants will have a substantially richer is important to note that the resource concentration herbivorous insect fauna than inferior competitors that model and models based on productivity make similar are sparsely distributed (Strong et al. 1984). Conse- predictions when the density of the host plant varies quently, widespread, superior competitors may com- with productivity. This-is probably a common phenom- monly have specialist insects that track their abun- enon. Further research that separates effects of plant dance, thereby reducing the impact of these dominant density vs. biomass (productivity) and effects of spe- species on subordinate plant species within the com- cialist herbivores vs. generalist herbivores is required. munity. This begs the question of why dense patches of dom- Effects of litter accumulation inant host plants can persist for long periods of time Similar to findings of other studies, we found that without being decimated by herbivores (e.g., decades litter decreased species richness and total stem density for goldenrod: Mellinger and McNaughton 1975, Wer- (Carson and Peterson 1990, Facelli and Pickett 1991, ner et al. 1980, Carson and Barrett 1988, Vankat and Tilman 1993, Foster and Gross 1997, 1998, Foster March2003 HERBIVORY,LITTER, AND PLANT DIVERSITY 663

1999, Suding and Goldberg 1999). The litter accu- (Carson and Pickett 1990, Bazzaz 1996, Carson and mulation hypothesis explains these results. That is, in Root 2000). productive communities, increases in the biomass of Dense stands of goldenrod, therefore, can suppress the superior competitor can lead to a dense accumu- richness both through competitive effects and by cre- lation of plant litter, causing a decline in species rich- ating a dense litter layer. This suggests another feed- ness (Carson and Peterson 1990, Foster 1999, Suding back loop based on abundance of the dominant plant and Goldberg 1999). In line with this hypothesis, we species. Specifically, goldenrod creates a litter layer found that litter mass increased with the biomass of that may prevent the establishment of other species, the dominant plant species (Fig. 8). Although insects thereby decreasing competition. Additionally, litter can potentially increase litter accumulation by increas- may create better environmental conditions (i.e., in- ing abscission (Owen 1978, Risley and Crossley 1988), creased soil moisture) for the already present goldenrod or decrease litter through consumption (Whittaker and (Facelli and Pickett 1991). This competitive release and Woodwell 1969), we found that insects had no effect habitat amelioration can allow goldenrod to increase on the relationship between goldenrod biomass and lit- in size and abundance, which, in turn, may feed back ter mass. into the litter layer when goldenrod senesces in the fall. The litter accumulation hypothesis predicts that litter This positive feedback loop between the dominant will inhibit stem density more with increasing litter plant and litter directly opposes the negative feedback mass. We did not find a consistent relationship between loop between the dominant plant and specialist her- the litter effect index and litter mass. We- found that bivorous insects. Herbivorous insects decrease the size insects, not litter mass, determined the litter effect in- and abundance of goldenrod. This decreases the con- dex in control plots, whereas litter mass, insects, and tribution of goldenrod to litter and decreases the com- their interaction determined the litter effect index in petitive displacement of other species by goldenrod. procedural control plots (Fig. 9B). As previously ar- Indeed, we found that the presence of insects reduced the effects of litter on total gued, the removal and replacement of litter in proce- negative stem density (Fig. 9). Herbivorous can thin of dural control plots may have disturbed the overwin- insects stands goldenrods by damaging leaves, killing stems, decreasing vege- tering Trirhabda virgata. This would have allowed the tative reproduction, disrupting belowground clonal in- relationship between the litter effect index and litter tegration, reducing nitrogen uptake, and reducing final mass. In the control plots, however, herbivorous insects plant size (McCrea et al. 1985, Cain et al. 1991, Brown may have obviated this relationship. 1994, Carson and Root 1999, 2000). Insects decreased Increases in the biomass of goldenrod probably de- goldenrod biomass, which probably increased light lev- creased richness in two ways: (1) by increasing litter els in stands of goldenrod (e.g., Brown 1994, Carson mass and decreasing the establishment of other species, and Root 2000). Decreasing goldenrod biomass also and (2) through competitive effects on the subordinate decreased litter mass (Fig. 8). Thus, because the effects plants. This is similar to Foster (1999), who found that of litter depended on goldenrod biomass and insects the negative effects of competition and litter on the determined goldenrod biomass, insects ultimately in- establishment of two target species became more severe fluenced the effects of litter accumulation. with increased total aboveground plant biomass. We found that litter mass increased with increasing bio- Conclusions mass of goldenrod (Fig. 8) and with total aboveground We found that herbivorous insects increased species = = plant biomass (F118 = 20.86, P 0.0002, R2 0.54), richness and decreased standing crop, largely through but was not related to aboveground plant biomass with- the effects of insects on the dominant, native herba- out the biomass of goldenrod (Fig. 5). This was sur- ceous species: meadow goldenrod. These findings run prising, given the high biomass of subordinate species counter to the general belief that herbivorous insects in some plots. One explanation would be that litter from have only weak effects on terrestrial plant communi- other species in these plots decomposes more quickly. ties. Our results suggest that herbivores can typically However, we did not directly quantify the contributions regulate terrestrial plant communities. Numerous other that each species made to litter, nor did we quantify studies also document the direct importance of herbiv- the relative rates of decomposition of each species. In ory in terrestrial plant communities (e.g., Spiller and the absence of this information, and because of the Schoener 1990, Marquis and Whelan 1994, Chase strength of the relationship between goldenrod and lit- 1998, Schmitz et al. 2000) and important interactions ter mass, we conclude that the dominant plant deter- between herbivores and resource competition (e.g., mined the litter mass in this study. Also similar to the Hunter and Price 1992, Ritchie et al. 1998, Chase et results of Foster (1999), goldenrod probably decreased al. 2000), further strengthening this suggestion. Here, the establishment of other species through competition. we not only show that insects can have important ef- Specifically, dense stands of goldenrod can dramati- fects on terrestrial plant communities, but also identify cally decrease light levels reaching the soil surface, when herbivorous insects should regulate plant com- thereby reducing stem density and species richness munities. Herbivorous insects should regulate plant 664 ZACHARY T. LONG ET AL. Ecology, Vol. 84, No. 3 communities when they track host plant density, if their Brown, V. K., M. Jepson, and C. W. D. Gibson. 1988. Insect demonstrat- host plant is a competitively dominant species. In this herbivory: effects on early old-field succession ed by chemical exclusion methods. Oikos 52:293-302. study, damage to meadow goldenrod and the resulting Cain, M. L., W. P. Carson, and R. B. Root. 1991. Long term increase in plant species richness occurred because of suppression of inset herbivores decreases rhizome produc- the ability of the specialist herbivore, Trirhabda vir- tion and number in Solidago altissima. Oecologia 88:251- gata, to track goldenrod density. 257. Carson, W. P., and G. W. Barrett. 1988. Succession in old- of her- Litter accumulation had the opposite effect field plant communities: effects of contrasting types of nu- bivorous insects on plant species richness. As predicted trient enrichment. Ecology 69:984-994. by the litter accumulation hypothesis, litter mass in- Carson, W. P., and C. J. Peterson. 1990. The role of litter in creased with the biomass of the dominant plant species, an old-field plant community: impact of litter quantity in different seasons on plant species richness and abundance. meadow goldenrod. This led to a decrease in species Oecologia 85:8-13. richness. Insects, however, alleviated the effect of litter Carson, W. P., and S. T. A. Pickett. 1990. The role of resources by suppressing goldenrod biomass. Therefore, insects and disturbance in the organization of an old-field plant may have had a greater effect on the plant community community. Ecology 71:226-238. effects of than did litter, through their effects on the dominant Carson, W. P., and R. B. Root. 1999. Top-down insect herbivores during early succession: influence on bio- plant species. mass and plant dominance. Oecologia 121:260-272. Carson, W. P., and R. B. Root. 2000. Herbivory and plant ACKNOWLEDGMENTS species coexistence: community regulation by an outbreak- This work was supportedby NSF grantsBSR-8817961 and ing phytophagous insect. Ecological Monographs 70:73- BSR-95 27536 to R. B. Root. W. P. Carsonwas supportedby 99. NSF grantsDEB-95 27729 and DEB-95 15184 duringman- Carter, P. E., and A. L. Rypstra. 1995. Top-down effects in uscriptpreparation at the Universityof Pittsburgh.C. L. Moh- soybean agroecosystems: spider density affects herbivore ler was supported,in part,by Hatch funds (Regional Project damage. Oikos 72:433-439. NE-92, NY(C) 183458) from the Cornell AgriculturalEx- Chase, J. M. 1998. Central-place forager effects on food web perimentStation. Peter Morin'slab group providedvaluable dynamics and spatial pattern in northern California mead- discussionon an earlierversion of this work.We thank Jeremy ows. Ecology 79:1236-1245. Fox, Amy Long, TimonMcPhearson, Peter Morin, Scott Mei- Chase, J. M., M. A. Leibold, A. L. Downing, and J. B. Shurin. ners, Owen Petchey, Stefan Schnitzer,Henry Stevens, and 2000. The effects of productivity, herbivory, and plant spe- Scott Wilson for comments on various drafts of the manu- cies turnover in grassland food webs. Ecology 81:2485- script.We are particularlygrateful to R. B. Root for numerous 2497. insightful discussions and comments.We owe a huge thanks Coll, M., and D. G. Bottrell. 1994. Effects of nonhost plants to Greg Bartus, who assisted with the study duringthree key on an insect herbivore in diverse habitats. Ecology 75:723- field seasons, and to Chris Asaro, Sean Boerke, Amatt Fa- 731. 401- cenza, Amy Farstad,and Loden Mohler, who assisted with Crawley, M. J. 1997. Plant-herbivore dynamics. Pages plant sampling in 1992. 474 in M. J. Crawley, editor. Plant ecology. Blackwell Sci- entific, Cambridge, Massachusetts, USA. Facelli, J. M. 1994. Multiple indirect effects of plant litter LITERATURE CITED affect the establishment of woody seedlings in old fields. Andow, D. A. 1991. Vegetational diversity and arthropod Ecology 75:1727-1735. population response. Annual Review of Entomology 36: Facelli, J. M., and S. T. A. Pickett. 1991. Plant litter: its 56 1-586. dynamics and effects on plant community structure. Bo- Armstrong, R. A. 1979. Prey species replacement along a tanical Review 57:1-31. gradient of nutrient enrichment: a graphical approach. Ecol- Foster, B. L. 1999. Establishment, competition and the dis- ogy 60:76-84. tribution of native grasses among Michigan old-fields. Jour- Bach, C. E. 1980. Effect of plant density and diversity on nal of Ecology 87:476-489. the population dynamics of a specialist herbivore, the Foster, B. L., and K. L. Gross. 1997. Partitioning the effects striped cucumber beetle, Acalymma vittata (Fab.). Ecology of plant biomass and litter on Andropogon gerardi in old- 61:1515-1530. field vegetation. Ecology 78:2091-2104. Bazzaz, F A. 1996. Plants in a changing environment. Cam- Foster, B. L., and K. L. Gross. 1998. Species richness in a bridge University Press, New York, New York, USA. successional grassland: effects of nitrogen enrichment and Belsky, A. J. 1986. Does herbivory benefit plants? A review plant litter. Ecology 79:2593-2602. of the evidence. American Naturalist 127:870-892. Goldberg, D. E., and A. M. Barton. 1992. Patterns and con- Belsley, D. A., E. Kuh, and R. E. Welsch. 1980. Regression sequences of interspecific competition in natural commu- diagnostics: identifying influential data and sources of col- nities: a review of field experiments with plants. American linearity. John Wiley, New York, New York, USA. Naturalist 139:771-801. Bonser, S. P., and R. J. Reader. 1995. Plant competition and Hairston, N. G., Jr., and N. G. Hairston, Sr. 1993. Cause- herbivory in relation to vegetation biomass. Ecology 76: effect relationships in energy flow, trophic structure, and 2176-2183. interspecific interactions. American Naturalist 142:379- Brown, D. G. 1994. Resource availability and plant invasion 411. after folivory by leaf beetles. Ecology 75:1673-1683. Henderson, I. E, and R. 0. Clements. 1977. Grass growth in Brown, V. K., and A. C. Gange. 1989. Differential effects of different parts of England in relation to invertebrate num- above- and below-ground insect herbivory during early bers and pesticide treatment. Journal of the British Grass plant succession. Oikos 54:67-76. Society 32:89-98. Brown, V. K., and A. C. Gange. 1992. Secondary plant suc- Herzig, A., and R. B. Root. 1996. Colonization of host patch- cession: how is it modified by herbivory? Vegetatio 101: es following long-distance dispersal by a goldenrod beetle, 3-13. Trirhabda virgata. Ecological Entomology 21:344-351. March2003 HERBIVORY,LITTER, AND PLANT DIVERSITY 665

Herzig, A. L. 1995. Effects of population density on long- Risley, L. S., and D. A. Crossley. 1988. Herbivore-caused distance dispersal in the goldenrod beetle Trirhabda vir- greenfall in the southern Appalachians. Ecology 69:1118- gata. Ecology 76:2044-2054. 1127. Holt, R. D., J. P. Grover, and D. Tilman. 1994. Simple rules Ritchie, M. E. 2000. Nitrogen limitation and trophic vs. abi- for interspecific dominance in systems with exploitative otic influences on insect herbivores in a temperate grass- and apparent competition. American Naturalist 144:741- land. Ecology 81:1601-1612. 771. Ritchie, M. E., D. Tilman, and J. M. H. Knops. 1998. Her- Hunter, M. D., and P. W. Price. 1992. Playing chutes and bivore effects on plant and nitrogen dynamics in oak sa- ladders: heterogeneity and the relative roles of bottom-up vanna. Ecology 79:165-177. and top-down forces in natural communities. Ecology 73: Root, R. B. 1973. Organization of a plant-arthropod asso- 724-732. ciation in simple and diverse habitats: the fauna of collards Kartesz, J. T. 1994. A synonymized checklist of the vascular (Brassica oleracea). Ecological Monographs 43:95-124. flora of the United States, Canada, and Greenland. Timber, Root, R. B. 1996. Herbivore pressure on goldenrods (Soli- Portland, Oregon, USA. dago altissima): its variation and cumulative effects. Ecol- Kleinbaum, D. G., L. L. Kupper, K. E. Muller, and A. Nizam. ogy 77:1074-1087. 1998. Applied regression analysis and other multivariable Root, R. B., and N. Cappuccino. 1992. Patterns in population methods. Third edition. Duxbury Press, Pacific Grove, Cal- change and the organization of the insect community as- ifornia, USA. sociated with goldenrod, Solidago altisima. Ecological Leibold, M. A. 1996. A graphical model of keystone pred- Monographs 62:393-420. ators in food webs: trophic regulation of abundance, in- SAS Institute. 1997. SAS/STAT PROC GLM. Version 6.12. cidence, and diversity patterns in communities. American SAS Institute, Cary, North Carolina, USA. Naturalist 147:784-812. Schellhorn, N. A., and V. L. Sork. 1997. The impact of weed Marquis, R. J., and C. J. Whelan. 1994. Insectivorous birds diversity on insect population dynamics and crop yield in increase growth of white oak through consumption of leaf collard, Brassica oleraceae (Brassicaceae). (ecologia 111: chewing insects. Ecology 75:2007-2014. 233-240. McBrien, H., R. Harmson, and A. Crowder. 1983. A case of Schmitz, 0. J. 1994. Resource edibility and trophic exploi- insect grazing affecting plant succession. Ecology 64: tation in an old-field food web. Proceedings of the National 1035-1039. Academy of Sciences (USA) 91:5364-5367. McCrea, K., W. Abrahamson, and A. Weis. 1985. Goldenrod Schmitz, 0. J., P. A. Hamback, and A. P. Beckerman. 2000. ball gall effects on Solidago altissima 14C translocation and Trophic cascades in terrestrial systems: a review of the growth. Ecology 66: 1902-1907. effects of carnivore removals on plants. American Natu- McNaughton, S. J., M. Oesterheld, D. A. Frank, and K. J. ralist 155:141-153. Williams. 1989. Ecosystem-level patterns of priamary pro- Spiller, D. A., and T. W. Schoener. 1990. A terrestrial field ductivity and herbivory in terrestrial grasslands. Nature experiment showing the impact of eliminating top predators 341:142-144. on foliage damage. Nature 347:469-472. Mellinger, M. V., and S. J. McNaughton. 1975. Structure and Strong, D. R. 1992. Are trophic cascades all wet? Differ- function of successional vascular plant communities in cen- entiation and donor control in speciose ecosystems. Ecol- tral New York. Ecological Monographs 45:161-182. ogy 73:747-754. Meyer, G. A. 1993. A comparison of the impacts of leaf- and Strong, D. R., J. H. Lawton, and T. R. E. Southwood. 1984. sap-feeding insects on growth and allocation of goldenrod. Insects on plants: community patterns and mechanisms. Ecology 74:1101-1116. Harvard University Press, Cambridge, Massachusetts, Meyer, G. A., and R. B. Root. 1993. Effects of herbivorous USA. K. and D. E. insects and soil fertility on reproduction of goldenrod. Ecol- Suding, N., Goldberg. 1999. Variation in the effects of and litter on ogy 74:1117-1128. vegetation recruitment across productivity gradients. Journal of Ecology 87:436-449. Morrow, P. A., D. Tonkyn, and R. Goldburg. 1989. Patch Tilman, D. 1982. Resource competition and community colonization by Trirhabda canadensis (Coleoptera: Chry- structure. Princeton University Press, Princeton, New Jer- somelidae): effects of plant species composition and wind. sey, USA. Oecologia 81:43-50. Tilman, D. 1993. Species richness of experimental produc- Neely, J. A. 1961. Soil survey. U.S. Department of Agri- tivity gradients: how important is colonization limitation? culture Soil Conservation Service, Series 25. Tompkins Ecology 74:685-700. County, New York, USA. Uriarte, M., and 0. J. Schmitz. 1998. Trophic control across Owen, D. F 1978. The effect of a consumer, Phytomiza ilicis, a natural productivity gradient with sap feeding herbivores. on seasonal leaf fall in the holly, flex aquifolium. Oikos Oikos 82:552-560. 31:268-271. Vankat, J. L., and G. W. Snyder. 1991. Floristics of a chron- Polis, G. A., and D. R. Strong. 1996. Food web complexity osequence corresponding to old-field-deciduous forest suc- and community dynamics. American Naturalist 147:813- cession in southwestern Ohio. I. Undisturbed vegetation. 846. Bulletin of the Torrey Botanical Club 118:365-376. Price, P. W. 1997. Insect ecology. John Wiley, New York, Werner, P. A., I. K. Bradbury, and R. S. Gross. 1980. The New York, USA. biology of Canadian weeds. Solidago canadensis L. Ca- Ramert, B., and B. Ekbom. 1996. Intercropping as a man- nadian Journal of Plant Science 60:1393-1409. agement strategy against carrot rust fly (Diptera: Psilidae): Whittaker, R. H., and G. M. Woodwell. 1969. Structure, pro- a test of enemies and resource concentration hypotheses. duction, and diversity of the oak-pine forest at Brookhaven. Environmental Entomology 25:1092-1100. Journal of Ecology 57:157-174. Risch, S. J., M. Altieri, and D. Andow. 1983. Agroecosystem Wilson, S. D., and P. A. Keddy. 1986. Measuring diffuse diversity and pest control: data, tentative conclusions, and competition along an environmental gradient: results from new research directions. Environmental Entomology 12: a shoreline plant community. American Naturalist 127: 625-629. 862-869.