vol. 157, no. 3 the american naturalist march 2001 Terrestrial Trophic Cascades: How Much Do They Trickle? Juraj Halaj* and David H. Wise† Department of Entomology, University of Kentucky, Lexington, Kentucky 40546-0091 Submitted February 14, 2000; Accepted October 19, 2000 abstract: Although more consensus is now emerging on the magnitude and frequency of cascading trophic effects in aquatic communities, the debate over their terrestrial counterparts continues. We used meta-analysis to analyze field experiments on trophic cascades in terrestrial arthropod-dominated food webs to evaluate the overall magnitude of trophic cascades and conditions affecting their occurrence and strength. We found extensive support for the presence of trophic cascades in terrestrial communities. In the majority of experiments, predator removal led to increased densities of herbivorous insects and higher levels of plant damage. Cascades in which removing predators led to decreased herbivory also were detected but were less frequent and weaker, suggesting a predominantly threetrophic-level behavior of arthropod-dominated terrestrial food webs. Despite the clear evidence that cascades often decreased plant damage, residual effects of predation produced either no or only minimal changes in overall plant biomass. Agricultural systems and natural communities exhibited similarly strong effects of predation on herbivore abundance. However, resulting effects on plant damage and community-wide effects of trophic cascades on plant biomass usually were highly variable, and only in the managed agricultural systems did predators occasionally have strong indirect effects on plant biomass. Our meta-analysis suggests that the effects of trophic cascades on the biomass of primary producers are weaker in terrestrial than aquatic food webs. Keywords: food webs, generalist predators, herbivory, meta-analysis, primary producers, trophic cascades. Despite extensive theoretical and empirical work on indirect effects in ecological communities (Paine 1966; Holt 1977; Abrams 1987, 1995; Stamp and Bowers 1991; Holt and Lawton 1993; Billick and Case 1994), the significance of these interactions is only poorly understood (Strauss 1991; Wootton 1994; Abrams 1995; Menge 1995). One * E-mail: [email protected]. † Corresponding author; e-mail: [email protected]. Am. Nat. 2001. Vol. 157, pp. 262–281. q 2001 by The University of Chicago. 0003-0147/2001/15703-0002$03.00. All rights reserved. form of indirect effect includes trophic events triggered by predation. Direct effects of a predator (donor) on its prey (transmitter) translate into changes in the prey’s energy supply (receiver) in an interaction chain (sensu Wootton 1993, 1994). Effects of resource consumption thus “cascade” from top consumers to the base of the Eltonian energy pyramid via feeding links between inversely related trophic levels (Paine 1980; Carpenter et al. 1985). In recent decades, the theory of trophic cascades has been rigorously developed and tested in aquatic systems. For example, experimental studies showed that piscivorous fish frequently reduce the abundance of planktivorous fish species. Size-dependent predation of planktivores on zooplankton assemblages (Brooks and Dodson 1965; Carpenter et al. 1985) may consequently alter the grazing pressure on primary producers, algal phytoplankton (Hrbácek et al. 1961; Carpenter et al. 1985; Vanni and Findlay 1990). Trophic cascades appear to be a significant and widespread type of indirect interaction in freshwater communities (Power et al. 1985; Carpenter et al. 1987; Power 1990; Vanni and Findlay 1990; Brett and Goldman 1996). The possibility of trophic cascades in terrestrial communities has long been recognized. Early references to cascading top-down effects date back to the Chinese, who as early as 2,500 yr ago promoted populations of predators to alleviate crop damage (Begon et al. 1996). In 1752, Linnaeus pointed out the benefits of using predatory arthropods to control crop pests (NRC 1996). Nevertheless, the actual role of trophic cascades in shaping the structure of terrestrial systems has been debated vigorously. Proponents of “the green world hypothesis” (HSS hypothesis; Hairston et al. 1960; Slobodkin et al. 1967) argue for the paramount role of predation (“top-down” effects) exerting cascading influences on the rest of the community. The hypothesis predicts that food-limited predators suppress herbivore populations to densities at which plants experience negligible levels of herbivory. In its original form, the HSS hypothesis applied only to folivores and sap and bark feeders in natural climax communities; however, its predictions have been compared with patterns in a diversity of taxa and systems (Sih et al. 1985; Schoener 1989; Polis 1994). The reasoning behind the HSS hypothesis has been ex- Terrestrial Trophic Cascades 263 trapolated to more than or fewer than three trophic levels by incorporating effects of ecosystem productivity (EEH; Fretwell 1977; Oksanen et al. 1981; Oksanen and Oksanen 2000). According to EEH, strong effects of herbivory should be prevalent in both unproductive (two trophic levels, no predators) and highly productive environments with even numbers of trophic levels, whereas herbivores should exert much weaker control of primary production in moderately productive systems with odd numbers of trophic levels. This linear trophic model assumes the presence of only one homogeneous functional group (“guild”) at each trophic level (Oksanen et al. 1981; Oksanen and Oksanen 2000). However, Leibold (1989) and Abrams (1993) have shown that an increase in system heterogeneity (e.g., more than one guild at each trophic level, feeding of a guild across several trophic levels, or consumer’s diet breadth) may alter responses of the trophic chain to the input of nutrients and produce trophic effects inconsistent with EEH predictions. Alternatively, some argue against the prevalence of trophic cascades in terrestrial systems. Propagation of indirect effects requires a relatively simple environment, in which strong trophic links and recipient control operate through a few key consumers (Paine 1980; Strong 1992), which may rarely occur in terrestrial systems. The reticulate nature of most terrestrial food webs, intraguild predation, and widespread trophic-level omnivory imply that terrestrial webs are strongly nonlinear trophic structures with the potential to buffer cascading effects (Polis et al. 1989; Polis 1991; Hunter and Price 1992; Polis and Holt 1992; Strong 1992), causing an incipient trophic cascade to become disconnected along the trophic chain (Schoener 1993; Abrams et al. 1996). Only under unusual conditions (e.g., presence of trophic subsidies; Polis 1999) in terrestrial systems can consumers generate strong cascading effects (MacInnes and Kerbes 1987; Strong 1992; Polis 1994; Polis and Strong 1996); thus, in most terrestrial food webs, instead of rushing cascades, “trophic trickles” are expected to be the norm (Strong 1992). The arguments against strong trophic cascades thus imply a donor-controlled, bottom-up template for terrestrial systems (White 1978; Hunter and Price 1992; Strong 1992; Polis 1994; Polis and Strong 1996). It is still too early to make reliable predictions about which communities exhibit strong cascading effects. The occurrence and strength of trophic cascades likely vary along a community continuum rather than being solely based on the community’s aquatic or terrestrial quality (Strong 1992). The simple aquatic-terrestrial dichotomy does not always hold. Even in many aquatic trophic cascades, the residual effects of predation on phytoplankton are usually weak (Brett and Goldman 1996, 1997), and factors such as nutrient enrichment and recycling (Vanni and Findlay 1990; Brett and Goldman 1997) or antagonistic interactions among zooplankton species (McCann et al. 1998) may modify the strength of top-down control in aquatic systems. Effects of predation cascading to primary producers have been detected in many terrestrial systems. This is true in relatively simple systems such as agroecosystems (Wright et al. 1960; Coaker 1965; Risch and Carroll 1982; Riechert and Bishop 1990; Carter and Rypstra 1995), grasslands (Schmitz 1994; Chase 1996; Moran et al. 1996; Schmitz et al. 1997; Moran and Hurd 1998), and scrub communities (Spiller and Schoener 1990, 1994; Polis and Hurd 1996), as well as in some of the most diverse terrestrial ecosystems, including temperate forests (Altegrim 1989; Marquis and Whelan 1994), tropical rain forests (Dial 1992; Letourneau and Dyer 1998), and soil communities (Strong et al. 1996a, 1996b, 1999). Regardless of their position on the trophic-cascade argument, most terrestrial ecologists now agree that insufficient evidence precludes generalizations about the functioning of species-rich systems (Oksanen et al. 1981; Strong 1992; Chase 1996; Letourneau and Dyer 1998; Lawton 1999; Oksanen and Oksanen 2000). Under some conditions, terrestrial communities clearly show cascading trophic effects. But how widespread and how strong are these effects? The focus now should be on resolving what factors favor the occurrence and modulate the magnitude of trophic cascades in terrestrial systems. Meta-analysis of Terrestrial Trophic Cascades Our objective was a quantitative assessment of the occurrence and strength of trophic cascades in terrestrial grazing food webs. We focused our review on the grazing subsystem of terrestrial food webs as a result of a greater availability of experimental data on trophic cascades in these systems. Cascading trophic effects in terrestrial detrital food webs (e.g., Kajak and Jakubczyk 1977; Kajak 1997; Lawrence and Wise 2000) is a growing area of ecological research and warrants a future separate assessment. We used meta-analysis to summarize our results. In contrast to more conventional, vote-counting literature reviews, meta-analysis allows a quantitative summary of findings and identification of central tendencies in a collection of different studies (Hedges and Olkin 1985; Gurevitch et al. 1992). Despite an increasing application of meta-analysis to ecological questions (Järvinen 1991; Gurevitch et al. 1992; Tonhasca and Byrne 1994; Brett and Goldman 1996; Hartley and Hunter 1998; Koricheva et al. 1998), its use in ecology remains controversial. A high diversity of ecological interactions and methods of study require a cautious approach to generalization (Osenberg et al. 1999). However, if meta-analysis is used predominantly to summarize findings and to generate new hypotheses for future 264 The American Naturalist testing, it can be a powerful tool of research synthesis in ecology (Gurevitch et al. 1992; Englund et al. 1999; Goldberg et al. 1999; Gurevitch and Hedges 1999; Palmer 1999). We used meta-analysis to address a set of a priori questions: What is the overall prevalence and magnitude of trophic cascades in terrestrial communities? Do effects vary among different categories of primary producers? Are the effects of trophic cascades stronger in managed than natural habitats? The recent publication of a meta-analysis of terrestrial trophic cascades by Schmitz et al. (2000) provides a unique opportunity to assess the generality of findings from two independent meta-analyses. Despite many similarities in findings and techniques of meta-analysis, our approach and that of Schmitz et al. (2000) differ in three major ways: First, we included experiments in agricultural crops, a category of studies explicitly excluded by Schmitz et al. (2000). By including crop systems, we were able to contrast managed and natural communities and to assess the magnitude of the trophic cascade across a sharp gradient of producer quality and habitat complexity. Second, Schmitz et al. (2000) included “natural” experiments in which predator densities differed between nonmanipulated areas. We excluded such observational studies because observed densities of herbivores and plant responses could have been caused by factors other than indirect effects of predation. Third, almost 60% of the tests in Schmitz et al. (2000) had ants as manipulated predators. Although these insect predators are a major component of terrestrial food webs, such a high proportion of ant-centered studies may have introduced a bias. Like Schmitz et al. (2000), we unavoidably examined arthropod-dominated food webs (see below), yet our data set provides a broader selection of predators. Methods and Analyses Literature Search and Data Selection Database. We broadly searched the ecological literature for studies documenting the occurrence of trophic cascades in natural and managed terrestrial communities using standard publication search engines (Agricola, Biological Abstracts, Biological and Agricultural Index, Science Citation Index). We also extensively consulted literature lists in published papers. To reduce unconscious bias from selective reporting in compiling meta-analysis data sets (Englund et al. 1999; Palmer 1999), we had no a priori criteria for the quality of the publication source, but individual publications were critically screened (see below). We included studies performed specifically to test the occurrence of trophic cascades (e.g., basic food web research literature) and also those where trophic cascade effects were implicit from research objectives (e.g., biocontrol research literature). Publication Selection Criteria. We included in our analysis only studies that satisfied the following criteria: First, the investigation was a field experiment in which predator densities were manipulated directly, for example, by removal or stocking, or indirectly by alteration of the habitat either to enhance or to reduce predator numbers. Second, the study included an appropriate control. For the purpose of comparing studies, we designated the control to be the treatment in which predator density was closest to the natural level in the study system (Gurevitch et al. 1992; Rosenberg et al. 1997), which did not always correspond to the treatment identified as the control by the investigators. Thus, predator-removal and predator-enhancement treatments were considered experimental groups. Third, measures of variance and number of replicates were reported or could be obtained from the author. Finally, the study measured cascading effects of predation on primary producers as either changes in plant damage (i.e., percentage of injury) or plant abundance (i.e., biomass or vegetation cover). We analyzed data on changes in herbivore densities if they were reported. We also included studies in which the effect on herbivores was not stated but in which it was implicit that observed changes in plant performance resulted from altered grazing intensity by herbivores as a result of predator manipulations. Nonindependence. Separate experiments in one published paper that were established consecutively over the season or during different years were treated as independent comparisons, as were comparisons for several species (groups) of organisms responding to the same treatment. These criteria follow consensus in conventional (Connell 1983; Sih et al. 1985) and meta-analytic reviews (Gurevitch et al. 1992; Koricheva et al. 1998). In a series of treatments in the same study, effects of the highest predator addition and the lowest predator reduction were compared with the same control, and these were treated as independent comparisons. The seasonal mean for control and experimental groups was used if the response variable was measured over time; otherwise, the largest contrast between control and experimental group was selected. Our selection guidelines clearly resulted in sets of nonindependent comparisons. It has been argued, however, that selective omission of some contrasts from a metaanalysis may introduce more bias than the use of a larger set of nonindependent comparisons (Gurevitch et al. 1992). We postulated that including several comparisons from the same study (or same author) and thus using their average effect in the analysis provided a better assessment of treatment effects than selecting a subset of comparisons Terrestrial Trophic Cascades 265 based on dubious selection criteria. If several response variables were measured to evaluate the same treatment impact (e.g., herbivore abundance expressed as either egg, larval, or adult densities), we selected one parameter, the one showing the greatest effect, to lower possible nonindependent bias of including multiple responses (Gurevitch and Hedges 1999). Publication Bias. A large percentage of comparisons in several categories came from work of only a few authors, which could have biased the independence of our overall data set. To evaluate the effect of including these groups of publications, we compared mean effect size of each group against the mean effect size of the rest of the main data set using the between-class homogeneity statistic Qb. Its significance was measured with a x2 test (Gurevitch et al. 1992; Rosenberg et al. 1997). The extent of selective reporting was further evaluated by examining the scatter plot of effect size and sample size (Light and Pillemer 1984; Palmer 1999). For the purpose of this assessment, an average of sample sizes was used in experiments with unequal sample sizes in control and experimental groups. In addition to the standard measure of effect size (d1), we also performed meta-analyses with response ratios (RR). The response ratio (ratio of means of experimental and control groups; RR p X e /X c) is a relatively new statistic to measure effect size in ecological meta-analysis. The recent description by Hedges et al. (1999) of the statistical properties of response ratios facilitates their wider application in ecology. Response ratios are generally analyzed using their natural logarithms (L p ln (RR)) to improve their statistical properties. To minimize the bias of using ln (RR) , it is recommended that either Ïne X e /SDe or Ïnc X c /SDc, whichever is smaller, be 13.0 (Hedges et al. 1999; X, n, and SD denote the mean, sample size, and standard error, respectively, of experimental, e, and control groups, c). Since over 83% of Ïnc X c /SDc and 82% of all Ïne X e /SDe values in our data set were 13.0, we felt confident that our database was suitable for meta-analysis with response ratios. By including ln (RR), we attempted to minimize the possible bias of using a unique measure of effect size (Osenberg et al. 1997). All meta-analyses were performed with MetaWin statistical software (Rosenberg et al. 1997). Statistical Analyses Results and Discussion We used biased-corrected Hedges’s d as a measure of effect size (d1), that is, the magnitude of the overall treatment effect (Hedges and Olkin 1985). The difference between responses of experimental and control groups was divided by their pooled standard deviation and corrected to remove bias as a result of small sample sizes (Hedges and Olkin 1985; Gurevitch et al. 1992). As a convention, the magnitude of the effect size is viewed as being small if control and treatment groups were 0.2 standard deviations apart (d p 0.2), medium if d p 0.5, large if d p 0.8, and very large if d 1 1.0 (Cohen 1969). The effect was judged statistically significant if the confidence intervals of the effect size excluded 0. Predator removal was expected to have a positive effect on the abundance of herbivores. The same suppression effect, however, would be revealed by a reduction in herbivore numbers in studies in which predators were experimentally enhanced. Thus, the effect size in predator-enhancement experiments was multiplied by 21 in order to pool effect size from both types of predatormanipulation studies (Gurevitch et al. 1992; Rosenberg et al. 1997). As a convention, our references to the effects of “predator removal” throughout the manuscript thus include both predator-removal and predator-enhancement comparisons. We used the mixed meta-analysis model to calculate the cumulative effect size for individual grouping categories as a result of its more realistic assumptions compared to more conventional fixed-effect models (Gurevitch and Hedges 1993; Rosenberg et al. 1997). We identified 57 experimental papers on terrestrial trophic cascades. Our data set included papers published between 1960 and 1999, with the majority coming from the 1980s and 1990s (fig. 1). Our literature search suggested that arthropod-centered investigations are currently the only category of studies providing a sufficient database for a meaningful overview. Schmitz et al. (2000) report a similar finding and identify only one observational paper (McLaren Figure 1: Trend in research publications on terrestrial trophic cascades based on our literature survey. 266 The American Naturalist and Peterson 1994) and one experimental study (Gutiérrez et al. 1997) in which indirect effects of predation on primary production via mammalian herbivores were addressed. Other experiments that measured effects of mammalian grazers on vegetation did not simultaneously provide information on the indirect effects of predators on trophic dynamics in these systems (e.g., Tansley and Adamson 1925; Collins et al. 1998). Thus, our review focuses exclusively on studies of arthropod-dominated webs. Studies with vertebrates were included if these groups were experimentally manipulated as predators of arthropods in the food web (e.g., birds, lizards). Our trophic groups correspond to Schoener’s (1989) categories of trophic elements, including small phytophagous herbivores (folivorous arthropods), small carnivores (predaceous arthropods), and medium carnivores (vertebrate predators feeding on arthropods). A conceptual justification for our focus on arthropods also stems from the fact that these taxa constitute the most abundant and diverse group of terrestrial animals (Southwood 1978; Erwin 1982; Wilson 1987; Samways 1994) and are significant components of most terrestrial food webs (Polis 1991; Schoenly et al. 1991). Forty studies satisfied our selection criteria and made up the final database, yielding 299 individual comparisons (“experiments”) evaluating effects of predator manipulation on herbivore density, plant damage, and plant biomass (table 1). In the majority of experiments (83%), predator densities were lower in the experimental treatment than in the control (ambient levels). Predator-enhancement manipulations were found exclusively in crop systems. The impact on the analysis of including predator-enhancement experiments is evaluated and discussed below. We excluded studies that used an unreplicated design, did not report appropriate measures of variance, or employed laboratory tests. Almost 60% of the studies included effects of predator manipulation on plant damage, and 64% used changes in plant biomass as an indicator of trophic cascades. More than one-third of all studies did not include any data on herbivore responses to predator manipulations, and only five studies assessed predation effects on all three response variables (table 1; Wright et al. 1960; Coaker 1965; Mowat and Martin 1981; Zehnder and Hough-Goldstein 1990; Marquis and Whelan 1994). Publication Bias Over 40% of the experiments measuring predator effects on herbivore density originated from the work of three groups of authors: Altegrim (1989; 10% of comparisons), Schmitz (1993, 1994, 1998; Schmitz et al. 1997; 12%) and papers and unpublished data sets of Snyder, Tuntibunpakul, and Wise (21%). Papers of Spiller and Schoener (1990, 1994) provided the single largest set of comparisons on plant damage (20%), and almost half of the comparisons on plant biomass effects came from the work of Wise and his colleagues (22%) and Schmitz and his colleagues (25%). However, we found no evidence that the mean effect sizes (d1) of these individual groups and the rest of the data set in each response category were statistically different (herbivore density—Altegrim vs. others: [d 5 95% CI]1.10 5 0.55 vs. 0.73 5 0.18, Qb p 1.59, df p 1, P p .21; Wise et al. vs. others: 0.66 5 0.42 vs. 0.79 5 0.18, Qb p 0.31, df p 1, P p .58; Schmitz et al. vs. others: 0.53 5 0.44 vs. 0.80 5 0.18, Qb p 1.27, df p 1, P p .26; plant biomass—Wise et al. vs. others: 20.15 5 0.41 vs. 20.33 5 0.19, Qb p 0.64, df p 1, P p .42; Schmitz et al. vs. others: 20.10 5 0.30 vs. 20.39 5 0.20, Qb p 2.49, df p 1, P p .11). The exception was Spiller and Schoener versus others in the plant damage category, where the difference was marginally significant (0.29 5 0.67 vs. 1.02 5 0.31, Qb p 3.77, df p 1, P p .052). Thus, the extent of any bias in our selection of data set as a result of including groups of nonindependent comparisons originating from a unique research group likely was minimal. Plotting the effect size against sample size in all experiments (n p 299) produced a predicted funnel-shaped pattern showing a greater variation in effect size in experiments with smaller sample sizes because of an increased sampling error (Light and Pillemer 1984; Palmer 1999). The absence of selective reporting is often indicated by a symmetrical shape of the scatter funnel (Light and Pillemer 1984; Palmer 1999). The shape of the scatter funnel was reasonably symmetric in our study (fig. 2), and thus, we were confident that the bias in estimating average effect size as a result of selective reporting was low. Overall Magnitude of Terrestrial Trophic Cascades Predator-Prey Link. Predators must exert sufficient pressure initially to trigger a chain of trophic interactions (i.e., to become “functionally significant”; Fretwell 1977; Polis 1994). Prey defense mechanisms (Owen 1980; Witz 1990), effects of habitat refugia (Pimentel 1961), and dilution of predation pressure because of widespread omnivory (Polis et al. 1989; Polis 1991) may limit predator effectiveness, reducing top-down forces to an “incidental” mortality factor (Errington 1956). When all studies were considered together, lowering predation pressure had a large positive effect on the abundance of terrestrial plant-feeding arthropods ([d1 5 95% CI] 0.77 5 0.16; fig. 3). Overall, our findings support the generalization that predation can exert a significant pressure on herbivore populations in a wide range of terrestrial communities (Sih et al. 1985; Henttonen et al. 1987; Floyd 1996; Polis 1999). Table 1: Source data for the analysis of terrestrial trophic cascades No. of comparisons per reference measuring effects of predator manipulation Reference Predatorsa Herbivores Plants Producer Herbivore Plant Plant category density damage biomass Data source Altegrim 1989 Beckerman et al. 1997 Bock et al. 1992 Brust 1994 Brust et al. 1985 Carter and Rypstra 1995 Chase 1996 Clark et al. 1994 Coaker 1965 Dial 1992 Birds Spiders Birds GAPC GAPC Spiders Spiders GAPC GAPC Spiders, lizards Complex Grasshoppers Grasshoppers CPB LL Complex Grasshoppers LL DL Complex Bilberry shrubs Grasses, forbs Grasses, forbs Potato Corn Soybeans Grasses, forbs Corn Cauliflower, cabbage Tabonuco trees Woodland Grassland Grassland Crop Crop Crop Grassland Crop Crop Woodland 14 2 9 … … … … … 8 4 5 4 … 2 6 5 … 1 2 1 … … 6 2 … … 4 … 10 … Dyer and Letourneau 1999 Fowler et al. 1991 Gómez and Zamora 1994 Gould and Jeanne 1984 Helenius 1990 Letourneau and Dyer 1998 Mansour and Whitcomb 1986 Mansour et al. 1985 Marquis and Whelan 1994 Moran and Hurd 1998 Moran et al. 1996 Mowat and Martin 1981 Pacala and Roughgarden 1984 Polis and Hurd 1996 Riechert and Bishop 1990 Ants, clerid beetles Birds Parasitoid wasps Polistes wasps GAPC Ants, clerid beetles Spiders Spiders Birds Mantids Mantids GAPC Spiders, lizards Spiders GAPC Complex Grasshoppers Weevils LL Aphids Complex Scale insects LL Complex Complex Complex DL Complex Complex Complex Woodland Grassland Woodland Crop Crop Woodland Crop Crop Woodland Grassland Grassland Crop Woodland Woodland Crop … 10 1 … 4 … 1 1 2 4 3 1 1 … 11 1 … … … … 4 1 1 2 … … 1 1 1 7 1 1 1 2 5 4 … … 2 4 1 1 … … … Risch and Carroll 1982 Schmitz 1993 Schmitz 1994 Schmitz 1998 Schmitz et al. 1997 Snyder and Wise 1999 Ants Spiders Spiders Spiders Spiders Carabids, spiders LL Grasshoppers Grasshoppers Grasshoppers Grasshoppers Complex Crop Grassland Grassland Grassland Grassland Crop 2 2 2 10 2 6 1 … … … … … … 2 4 14 4 10 Table 1 Figure 6 Table 3; figure 3 Table 3; figures 4, 5 Figures 5, 6 Tables 1, 2; figure 4 Snyder and Wise, in press Carabids, spiders Complex Ant plants Grasses, forbs Crucifer shrubs Cabbage Oats Ant plants Citrus trees Avocado trees White oak Grasses, forbs Grasses, forbs Cauliflower Complex Atriplex shrubs Broccoli, Brussels sprouts, cabbage, potato, radish, tomato Squash Grasses, forbs Grasses, forbs Grasses, forbs Grasses, forbs Beans, cabbage, cucumber, eggplant, squash Cucumber, squash Figures 1, 2 Table 1; figure 2 Table 1; figure 6 Text; table 2 Figure 1 Figure 3 Figure 3 Figure 1b Tables 2, 3, 8 Figures 3.11, 3.15, 3.19, 3.21 Figure 3 Text; table 3 Text Table 3 Figures 2, 3 Figure 5 Table 2 Table II Figure 2 Figures 4a, 6 Figures 4b, 5 Table II Text; figure 1 Table 26.5 Table 4; figures 2, 3 6 … 6 Crop Figures 5, 6, 7 Table 1 (Continued ) No. of comparisons per reference measuring effects of predator manipulation Reference Predatorsa Herbivores Plants Producer Herbivore Plant Plant category density damage biomass Spiller and Schoener 1990 Spiller and Schoener 1994 Stoner 1993 Tuntibunpakul 1999 Spiders, lizards Spiders, lizards GAPC GAPC Complex Complex CPB Complex Sea grapes Sea grapes Potato Cucumber, potato, soybeans Woodland Woodland Crop Crop … 6 … 16 8 6 1 3 … … 1 5 Warrington and Whittaker 1985 Whittaker and Warrington 1985 Wright et al. 1960 Zehnder and Hough-Goldstein 1990 Ants Ants GAPC GAPC Complex Complex DL CPB Sycamore trees Sycamore trees Cauliflower Potato Woodland Woodland Crop Crop … … 2 3 3 … 3 1 … 1 3 1 Data source Figure 2 Figures 6, 8 Table 1; figure 6 Figures 6, 7, 21, 23, 24, 27, 29, 30, 39, 40, 41, 42 Table 4 Table 8 Tables 1, 2, 3 Tables 2, 4, 7, 8 Note: GAPC p Ground arthropod predator complex (includes studies where a diverse assemblage of predators was manipulated simultaneously; e.g., spiders, harvestmen, predaceous beetles, ants); CPB p Colorado potato beetles; LL p Lepidoptera larvae; DL p Diptera larvae. a Manipulated trophic level. This category includes groups that we considered to be most likely impacted by manipulations. Our designations did not always correspond to the one suggested by the author. Terrestrial Trophic Cascades 269 Figure 2: Relationship between the magnitude of effect size (Hedges d1) and the experimental sample size. Each point represents a unique comparison. Herbivore-Plant Link. To maintain cascading effects throughout the community, predators must be trophically coupled with potentially strong groups of herbivores, that is, guilds capable of inflicting strong grazing pressure on primary producers (Polis 1994, 1999). Effective plant defense mechanisms (Murdoch 1966; Feeny 1970; Coley et al. 1985) and low nutritional quality of food (Sinclair 1975; White 1978; Batzli 1986) in some systems, as well as the influence of the abiotic environment (Davidson and Andrewartha 1948; Ehrlich and Birch 1967), may reduce herbivore effectiveness, rendering them a donor-controlled entity (Polis 1999; but see Chase 1996). Changes in herbivore populations, nevertheless, cascaded through the trophic chain, increasing the incidence of plant injury (d1 p 0.88 5 0.25; fig. 3). Although increased populations of herbivores increased plant damage, the overall reduction in primary production (plant biomass) was small to medium but statistically significant (d1 p 20.32 5 0.19; fig. 3). Significant cascading trends were also found when data were analyzed using log response ratios. On average, lowering predation pressure increased herbivore abundance; that is, X e (experimental treatment mean) p 150.4% of the control treatment, X c (95% CI p 139.1%–162.6%), and increased plant damage, X e p 169.4% of X c (151.5%–189.4%). Effects of predation on plant biomass, however, were weaker; X e p 0.89% of control (0.82%–0.97%). Cascading trophic effects are expected to be buffered at each trophic-level transition as a result of system complexity and time lags in responses of individual consumer groups (Carpenter et al. 1985; Strong 1992; Polis 1994). Our data are not consistent with this prediction, since the responses of herbivores and primary producers measured at the level of plant damage were similar in magnitude (fig. 3). How- ever, the degree to which trophic cascades were dampened at the level of primary producers could have been confounded in our analysis by the absence of data on herbivore abundance in 31% of studies. Thus, we further analyzed a subset of comparisons where effects of predation on herbivory and at least one of the plant performance parameters were measured simultaneously. We found significant positive relationships between the magnitude of response of herbivores and the amount of plant damage with either estimator of effect size (fig. 4), which further supports the hypothesis of trophic cascades. A slope of 1.20 for the regression, however, suggests a slight heightening of the cascade at the level of plant damage. Removal of three outliers from the data set (a four-level cascade with lizards and spiders: Pacala and Roughgarden 1984; small-sized branch enclosures: Mansour and Whitcomb 1986; Mansour et al. 1985) preserved the statistical significance of this relationship in figure 4A but lowered the model’s predictive power (n p 31, R 2 p 0.31, F p 13.18, P ! .01) and suggested a dampening effect (slope p 0.85). Analysis with ln (RR) showed a similar dampening of the trophic cascade (slope p 0.89; fig. 4B). Absence of a stronger dampening of the trophic cascade at the level of plant herbivory may have at least two explanations. For example, behaviorally mediated cascades, where the presence of predators may inhibit feeding of herbivores, can significantly alter levels of herbivory without changes in herbivore densities (Schmitz 1994; Beckerman et al. 1997; Moran and Hurd 1998; Snyder and Wise 2000). Alternatively, densities of irrelevant herbivore species may have been assessed in experiments. Since the overwhelming majority of predators in the surveyed experiments were generalist predators, it is possible that their Figure 3: Overall mean effect size (595% CI) of lowering predation pressure on the abundance of herbivores, extent of plant damage, and changes in plant biomass across all experiments. Numbers above or below bar graphs indicate the number of comparisons within each category. 270 The American Naturalist level needed to impede its production significantly, a result of diffused herbivory in the system. Thus, this “heterogeneity” (Hunter and Price 1992) or “differentiation” (Strong 1992) in consumer feeding may have modified the strength of the trophic cascade at both the predator-herbivore as well as the herbivore-plant trophic link. Variations in the Magnitude and Length of Chains of Trophic Cascades Figure 4: Association between the magnitude of response of herbivores and primary producers to experimental lowering of predation pressure, estimated with (A) Hedges d1 (plant damage: n p 34, R2 p 0.68, F p 69.33, P ! .001; plant biomass: n p 68, R2 p 0.01, F p 0.65, P p .423) and (B) log response ratios (plant damage: n p 33, R2 p 0.33, F p 15.34, P ! .001; plant biomass: n p 68, R2 p 0.05, F p 3.76, P p .06). M1 p Mansour et al. 1985; M2 p Mansour and Whitcomb 1986; P p Pacala and Roughgarden 1984. effects on plant damage may have been partially channeled via groups of herbivores that were not the focus of an investigation. The relationship between the change in herbivore density and change in plant biomass was weaker and only marginally significant when analyzed with log response ratios (fig. 4B). A strong dampening effect was suggested at this level (d1 slope p 20.07, ln (RR) slope p 20.34). The use of log ratios slightly improved the fit of the model because of a logarithmic transformation of the relationship. Thus, although clear trophic cascades were detectable for plant damage from herbivory, further effects were strongly buffered, producing either no or only minimal reductions in the overall plant community biomass. These results are similar to those of Schmitz et al. (2000), who found much stronger effects of predation on plant damage than plant biomass. This suggests that plants had the ability to tolerate significant levels of herbivory without suffering highly adverse effects on primary production. Likewise, it is plausible that, if the plant community is sufficiently diverse, the amount of herbivory inflicted on a particular plant species may not reach the critical threshold A more detailed inspection of the data revealed several intriguing results. A total of 77% of the experiments showed positive responses of herbivores to predator removal (fig. 5A). The overall magnitude of this positive response was very large and highly significant (d1 p 1.10 5 0.18). In contrast, herbivore populations declined following predator removal in only about 20% of cases, and the absolute magnitude of this negative effect was almost 3.5 times lower (d1 p 20.32 5 0.19). Plant damage mirrored the extent of herbivore responses to predator removal. Predator removal increased plant damage in 76% of cases, and this effect was very large (d1 p 1.38 5 0.27). Negative effects of predator removal on plant damage were again less frequent (20%) and weaker (d1 p 21.03 5 0.31; fig. 5B). Removing predators most frequently reduced plant biomass (63%), and overall this effect was large (d1 p 20.73 5 0.15). Reverse trophic responses were weaker (d1 p 0.37 5 0.16) and less common (34%; fig. 5C). Analysis with ln (RR) produced very similar results (fig. 5D–5F ). According to theory, removal of top predators in food webs with an odd number of trophic levels should boost herbivory and reduce plant biomass (Hairston et al. 1960; Fretwell 1977; Oksanen et al. 1981). A reverse scenario should be prevalent in systems characterized by four (an even number) of trophic levels. Our analysis suggests that the overwhelming majority of terrestrial systems we reviewed behaved as odd-trophic-level systems. Increases in the number of trophic links (i.e., interactions across more than three trophic levels) appeared to reduce the frequency of trophic cascades and to attenuate their magnitude. Omnivory and Terrestrial Trophic Cascades Widespread trophic-level omnivory is cited as one factor causing trophic cascades to dissipate in terrestrial food webs (Polis 1991, 1994; Polis and Strong 1996). It is interesting that all but two experiments (parasitoids: Gómez and Zamora 1994) in our data set included generalist predators (table 1). However, solely using dietary data to predict the strength of trophic links and consequent cascading effects may not be adequate. A maze of seemingly complex trophic links may make it difficult to recognize community Terrestrial Trophic Cascades 271 Figure 5: Frequency distribution of the magnitude and direction of predator-removal effects on herbivore densities, amount of plant damage, and amount of plant biomass using Hedges d1 (A–C) and log response ratios (ln(RR); D–F ) as estimates of effect size. White and black bars denote negative and positive values of effect size, respectively. subsystems (modules; sensu Paine 1980; Holt 1996) that exhibit strong internal ties. Experimental manipulations may be a more appropriate approach to identify major interaction pathways in communities (Paine 1980; Wise 1984, 1993; Hairston 1994). For example, although praying mantids are generalist arthropod predators that consume a wide range of herbivore taxa (Reitze and Nentwig 1991) as well as other generalist predators (e.g., spiders: Moran et al. 1996; Moran and Hurd 1998), experiments show that mantids can evoke significant top-down effects and boost plant biomass (Moran et al. 1996; Moran and Hurd 1998; table 2). This result reveals a mantid-herbivore-plant module embedded in a matrix of complex feeding links in the successional grassland community (Moran and 272 The American Naturalist Table 2: Variation in the effect of predation by major predatory groups on herbivore density, plant damage, and biomass Response variable and predator group Herbivore density: GAPC Mantids Spiders Birds Lizards Arthropods Vertebrates Plant damage: GAPC Mantids Spiders Birds Lizards Arthropods Vertebrates Plant biomass: GAPC Mantids Spiders Birds Lizards Arthropods Vertebrates No. of comparisons Effect size (d1) 57 7 25 35 8 90 43 95% CI Lower Upper .85 .81 .61 .68 .85 .79 .71 .59 .39 .22 .42 .07 .57 .47 1.11 1.23 1.00 .93 1.61 1.01 .95 31 … 15 7 13 51 20 1.34 … .61 1.12 .54a .92 .81 1.00 … .01 .54 2.10 .61 .40 1.68 … 1.21 1.69 1.17 1.22 1.22 43 5 30 9 … 86 9 2.53 2.62 2.17a 2.23a … 2.33 2.23a 2.85 21.10 2.50 2.68 … 2.55 2.70 2.20 2.14 .17 .21 … 2.11 .23 Note: GAPC p ground arthropod predator complex. a Statistically not significant (95% CI includes 0). Hurd 1998). However, Strong et al. (1999) argue that these modules or “vignettes” are vulnerable trophic connections that may collapse upon influence from the environment in which they are nested. A high potential for intraguild predation among predators or competition among herbivores and among plants may modulate the intensity of interactions within vignettes (Strong et al. 1999). The multichannel omnivory model (Polis 1994; Polis and Hurd 1996; Polis and Strong 1996) further predicts that the ability of predators to feed and to survive on alternative food resources may modify the extent to which generalist feeders control their resources. Thus, rather than diluting trophic cascades, acquiring resources from a multitude of sources may in fact facilitate their occurrence in complex terrestrial food webs (Polis and Strong 1996; Polis 1999). This point appears to be supported in our review by a number of examples in which manipulated predator species may derive substantial nutritional provisions from alternative detrital channels. Both anoles (Dial 1992; Spiller and Schoener 1996) and spiders (Pacala and Roughgarden 1984; Polis and Hurd 1996; Spiller and Schoener 1996) feed on detritovorous Diptera, and ar- thropod predators in the crop system of Wise and his colleagues (Snyder and Wise 1999, in press; Tuntibunpakul 1999) may be subsidized by feeding on Collembola (D. M. McNabb, J. Halaj, and D. H. Wise, unpublished data). Thus, both facets of omnivory, that is, the provision of energy subsidies coupled with a diffusion of feeding links, have the potential to cast the traditional linear model of the trophic cascade (Hairston et al. 1960) into a more reticulate version, in which the strength of top-down control may vary widely among consumers and is more difficult to predict. Clearly, our findings indicate that generalist predators have the ability to trigger significant cascading effects. These results, however, do not directly address the more challenging question of whether, and to what extent, this capacity to generate cascades is buffered by their multiresource feeding? Comparing the magnitude of cascading effects induced by generalist and specialist natural enemies would address this question. Although evidence exists that specialized natural enemies (e.g., insect parasitoids) can significantly suppress herbivore populations, the extent of this trophic effect is highly variable (Murdoch et al. 1985; Hawkins 1992; Hawkins and Gross 1992; Faeth 1994; Hawkins et al. 1999), and its consequences for plant fitness are largely unknown (Faeth 1994). Hawkins (1992) even points out that top-down control in many parasitoid-based food webs may be minimal or absent. Our review of cascades generated by generalist predators reveals that effects on herbivore densities and the level of herbivory are not necessarily reliable predictors of their impact on plant production. Thus, whether or not natural enemy specialists can trigger stronger cascading effects than generalists is open to speculation at the moment as a result of a lack of sufficient experimental evidence. Effects of Primary Producer Quality on Terrestrial Trophic Cascades We further divided studies into broad categories defined in terms of primary producers, rates of primary production, and habitat complexity. Crop systems were predominantly annual monocultures, with the exception of two studies that studied trophic cascades in perennial plantations (citrus: Mansour and Whitcomb 1986; avocado: Mansour et al. 1985). The grassland category included perennial grasses and forb vegetation of natural prairies and successional old-field systems. Shrubby and woody vegetation of scrublands and temperate and tropical forests were jointly characterized as woodlands. This categorization allowed us to make comparisons along a gradient of ecosystem productivity/habitat complexity, that is, crops versus grasslands versus woodlands, and to contrast managed (crops) versus natural (grasslands 1 woodlands) sys- Terrestrial Trophic Cascades 273 tems. Although this classification is very broad, a more detailed community resolution is not feasible with the available data set. The quality of primary producers plays a significant role in consumer interactions (Feeny 1970; Sinclair 1975; Fretwell 1977; White 1978; Oksanen et al. 1981). Our results showed that the magnitude of trophic cascades varied significantly among community types. Although effects of predation on herbivores were similar in all three categories, differences were found in effects on primary producers (table 3). Net effects of trophic cascades on plants were very strong in crop communities, showing characteristics of oddtrophic-level systems; in the majority of experiments, predator removal increased herbivory (d1 p 1.14 5 0.25) and reduced primary production (d1 p 20.98 5 0.34; fig. 6A). The reverse scenario was less frequent and statistically not significant (fig. 6A). This finding may not be surprising considering the vulnerability and homogeneity of crop systems. Effects of trophic cascades were less predictable in more complex grassland communities. Despite strong predation effects on herbivores, there was no overall effect on plant damage, and the effect on plant biomass was almost 60% lower compared to crops (table 3; fig. 6). It appears that the higher variability in plant nutritional quality, lower overall plant quality, and higher habitat diversity of grasslands compared to crop systems may dampen overall effects of changing herbivore densities on plant biomass in grasslands. Moran and Hurd (1998) report detectable effects of a trophic cascade triggered by mantid predation on grasses but not on forbs, which contain higher levels of secondary compounds. Reduction in the biomass of more palatable plant resources may often be offset by an increase in the biomass of less preferred species, thus reversing and blurring Table 3: Magnitude of terrestrial trophic cascades in different community types Response variable and plant community type Herbivore abundance: Crops Grasslands Woodlands Plant damage: Crops Grasslands Woodlands Plant biomass: Crops Grasslands Woodlands a No. of comparisons Effect size (d1) Lower Upper 61 44 28 .92 .51 .93 .66 .26 .56 1.19 .76 1.30 35 4 32 1.60 .15a .35 1.22 21.16 .02 1.99 1.45 .67 46 40 9 2.48 2.20 2.01a 2.77 2.38 2.72 2.19 2.03 .71 Statistically not significant (95% CI includes 0). 95% CI Figure 6: Overall mean effect sizes (595% CI) and frequency of threeand four-level trophic cascades in crops (A), grasslands (B), and woodlands (C). Numbers above bar graphs indicate the number of comparisons within each category. ns p statistically not significant (95% CI includes 0). the trophic-cascade effect (Schmitz 1994). Predator-mediated effects on herbivore foraging behavior in more complex habitats may produce similar results (Beckerman et al. 1997). For example, in the presence of spiders, grasshoppers switch from feeding on grasses to consumption of herb vegetation, which provides more protection from predators. As a result, predators generated two strong but opposing species-level cascades (sensu Polis 1999) on herbs (d1 p 20.99 5 0.66) and grasses (d1 p 1.30 5 0.89; fig. 6B), resulting in no net effect on plant damage (d1 p 0.15 5 1.30; table 3). It is also possible that the magnitude of trophic cascades in grasslands might have been overestimated be- 274 The American Naturalist Figure 7: Overall magnitude of trophic cascades in crop and noncrop communities. Numbers above bar graphs indicate the number of comparisons within each category; ns p statistically not significant (95% CI includes 0). cause most experiments in this system employed completely enclosed cages. For example, effects of spider predation on grasshoppers and plant damage and biomass (Schmitz 1993, 1994, 1998; Chase 1996; Beckerman et al. 1997; Schmitz et al. 1997) could have been magnified by the exclusion of alternative prey and natural enemies of spiders from experimental microcosms, or could have been related to artificially inflated rates of predator-prey encounters within confined habitats. Examples of significant initiation of trophic cascades were found even in diverse woodland communities. Overall effects, however, were low for plant damage and statistically insignificant for plant biomass (table 3; fig. 6C ). It is interesting that this pattern was produced by a mixture of strong odd- and even-trophic-level effects on herbivory and plant biomass consumption. For example, feeding by lizards on homopterans indirectly benefits sea grape plants by lowering the extent of tissue damage. At the same time, however, predation by these vertebrates on spiders that feed on gall midges increases gall damage to the same producer (Spiller and Schoener 1990, 1994). Thus, the overall indirect impact of lizard predation on plants was not significant (table 2). Strong three-trophic-level effects of bird and ant predators in temperate forests (Warrington and Whittaker 1985; Whittaker and Warrington 1985; Altegrim 1989; Marquis and Whelan 1994) contrast with four-level trophic cascades on Piper trees triggered by predaceous Tarsobaenus beetles in a tropical rain forest (Letourneau and Dyer 1998; Dyer and Letourneau 1999). Although in some systems indirect effects of predation on plant damage were observed over wider areas (e.g., Spiller and Schoener 1997; Dyer and Letourneau 1999), trophic cascades in these systems occurred only on single plant species in otherwise diverse systems, and consequences of herbivory to plant biomass production and fitness were unclear. Thus, community-wide effects of trophic cascades were low in woodlands. Contrasting crop with noncrop communities yielded an intriguing pattern. Although effects on herbivore abundance were similar in both types of systems, the amount of herbivory following the relaxation of predation rates was significantly higher in crop than noncrop systems as suggested by their nonoverlapping 95% confidence intervals (crop: d1 p 1.60 5 0.36, 95% CI; noncrop: d1 p 0.32 5 0.29; fig. 7). Crop systems also suffered significant reductions in plant biomass (d1 p 20.48 5 0.31), whereas net effects of trophic cascades at the level of primary production were not statistically detectable in noncrop communities (d1 p 20.20 5 0.23). Strong trophic-cascade effects in crop communities are partly because of predator-enhancement manipulations, which were not done in other types of communities. Although boosting predator numbers above their natural levels increased the magnitude of plant responses to trophic cascades (fig. 8), strong effects of predation on plant damage (d1 p 1.24 5 0.47) and biomass (d1 p 20.40 5 0.33) were found in crops even after excluding predator-enhancement comparisons from the analysis. The net result of cascading effects seems to be significantly stronger in crops than natural communities. Crop plants were the only primary producers where predator removals occasionally resulted in “runaway consumption” Figure 8: Overall magnitude of trophic cascades in predator-removal and predator-enhancement comparisons. Numbers above bar graphs indicate the number of comparisons within each category. Note that the absolute magnitude of effect size is compared here; the sign for predator enhancement experiments was reversed during analyses to enable pooling of effects (Rosenberg et al. 1997). Terrestrial Trophic Cascades 275 (sensu Strong 1992) of plants, producing “cleared substrate” effects (e.g., Wright et al. 1960, figs. 1–3; Snyder and Wise, in press). Trophic cascades in crops thus may represent a very special form of community-level cascade in which a few (vulnerable) plant species dominate the system (Polis 1999). This result presents a fascinating analogy with cascading effects causing “clear water” phenomena in aquatic systems (Carpenter et al. 1985; Power 1990; Strong 1992). Polis (1994) argues that trophic cascades are more likely to occur in a low-disturbance environment, with low spatial and temporal heterogeneity. The stochastic disruption hypothesis (Schoener 1993) further predicts that chance events such as weather variability may disrupt indirect effects by eliminating critical species or trophic links from a food web. The occurrence of strong cascades in crop communities, which experience significant periodical disturbances, does not support the prediction that high-disturbance environments weaken trophic cascades. Hawkins et al. (1999) argue that top-down control is more frequently observed in managed than natural systems as a result of the simpler habitat and food web structure of managed systems. Thus, it appears that release of herbivores from predation pressure nested within a structurally and species-simple environment with highly susceptible plant resources may explain the greater propensity of crop systems to cascade. Even though Schmitz et al. (2000) do not include crop systems in their review, some of their findings support our conclusions. For example, removal of predators produced significantly stronger effects on primary producers in low- compared to high-diversity systems, and plant chemical defense mechanisms appeared to buffer top-down effects (Schmitz et al. 2000). We compared our results with those of Brett and Goldman (1996; table 1) to assess the evidence that the magnitude of trophic cascades differs between terrestrial and aquatic communities. The absolute magnitude of the trophic cascade at the level of zooplankton (“herbivores”) was d p 1.39 (95% CI; 0.72–2.06). Although the topdown pressure on herbivores in terrestrial food webs was lower (d p 0.77, 0.61–0.93), this difference was not statistically significant because of overlapping confidence intervals. Even the difference between aquatic systems and structurally simple crop habitats was statistically nonsignificant (d p 0.92, 0.66–1.19). On the other hand, the absolute response of aquatic primary producers (phytoplankton; d p 2.01, 1.30–2.72) was significantly stronger than that in terrestrial communities (terrestrial overall effect: d p 0.32, 0.13–0.51; crops: d p 0.48, 0.17–0.79). This quantitative comparison strongly suggests that the effect of trophic cascades on primary producers is significantly greater in aquatic than terrestrial systems. Conclusions Our results present clear experimental evidence for trophic cascades in terrestrial communities. Since in the overwhelming majority of studies these effects were induced by generalist predators, it appears that intraguild predation and trophic-level omnivory alone do not preclude the possibility of strong cascading indirect effects among terrestrial consumers. However, the paucity of experimental evidence on trophic cascades induced by specialized natural enemies prevents an assessment of the extent to which trophic cascades are buffered by the breadth of predator diet. Our review does provide evidence that trophic-level omnivory contributes to the dampening of trophic cascades induced by complexes of generalist predators, because in several studies community-wide consequences for primary production were reduced by counteractive effects of three- and four-level cascades. Net cancellation of effects generated by different trophic modules appeared to be less prevalent in agricultural systems, although even in crops such counteractive interactions can be strong (Snyder and Wise, in press). Although predation had large effects on herbivore densities in all systems, overall effects on plant damage and plant biomass were highly variable in natural systems. In addition, adverse effects of herbivory on primary production were strongly buffered at the plant level, with the exception of crop systems where predators occasionally produced substantial indirect effects on plant biomass. We suggest that such effects occur, despite a strong annual disturbance regime, as a result of a relatively simple habitat structure and the presence of a more homogeneous and vulnerable plant community in agricultural systems. Despite a relatively narrow overlap in databases (21%; 17/[40 1 41] papers), our meta-analysis and that of Schmitz et al. (2000) both found statistically significant evidence for trophic cascades in terrestrial systems. Both meta-analyses also revealed that trophic cascades are stronger at the level of plant damage than plant biomass. This agreement between two independent meta-analyses strengthens the generality of the conclusions regarding the prevalence and intensity of terrestrial trophic cascades. Our joint results should be interpreted with some caution, however. The availability of adequate experimental evidence only from arthropod-centered food webs biased our reviews and limits the scope of inferences. Our understanding of terrestrial systems in general would be much improved by including studies with vertebrate herbivores. Such an analysis currently is not feasible and we hereby encourage these types of studies. We disagree with the conclusion of Schmitz et al. (2000) that the strength and pattern of terrestrial trophic cascades are equivalent to those of aquatic communities. Our meta- 276 The American Naturalist analysis of the literature supports the view that terrestrial trophic cascades represent qualitatively different indirect interactions compared to their aquatic equivalents (Strong 1992; Polis 1999). With the exception of some crop systems, terrestrial cascades appear to be species-level phenomena. Although residual effects of predation can cascade to primary producers in both systems, net consequences for primary producers in aquatic and terrestrial systems are different. Whereas grazing of planktonic herbivores under no-predator regimes can substantially reduce the biomass of the algal phytoplankton and occasionally produce community-wide effects on primary producers, this type of runaway consumption seems to be strongly buffered by terrestrial producers, suggesting a predominately bottom-up template of community structure as suggested by White (1978) and reiterated repeatedly by others (e.g., Hunter and Price 1992; Strong 1992). Do indirect effects of predation in terrestrial communities flow torrentially down the trophic waterfall, or do they trickle? Our analysis provides a preliminary answer for food webs dominated by generalist arthropod predators. The terrestrial cascade starts strongly but becomes diverted to a trickle when it reaches the pool of primary production, with the exception of those crop systems in which the cascade continues on to cause not a ripple but a splash. Acknowledgments We thank K. L. Lawrence, D. M. MacNabb, J. Moya-Laraño, C. M. Rauter, and J. L. Williams for reading and commenting on the manuscript. J. Helenius and S. E. Riechert generously provided their unpublished data. We also appreciate thorough and stimulating reviews by J. Gurevitch, P. J. Morin, and D. R. Strong. This material is based upon work supported by a National Science Foundation Postdoctoral Fellowship in Biosciences Related to the Environment (DBI-9804167) awarded to J.H. Unpublished data on vegetable crops are from experiments supported, in part, by U.S. Environmental Protection Agency grant G71A0056 to D.H.W. This is publication 00-08-25 of the Kentucky Agricultural Experiment Station. Literature Cited Abrams, P. 1987. Indirect interactions between species that share a predator: varieties of indirect effects. Pages 38–54 in W. C. Kerfoot and A. Sih, eds. Predation: direct and indirect impacts on aquatic communities. University Press of New England, Hanover. Abrams, P. A. 1993. Effect of increased productivity on the abundances of trophic levels. American Naturalist 141: 351–371. ———. 1995. Implications of dynamically variable traits for identifying, classifying, and measuring direct and indirect effects in ecological communities. American Naturalist 146:112–134. Abrams, P. A., B. A. Menge, G. G. Mittelbach, D. A. Spiller, and P. Yodzis. 1996. The role of indirect effects in food webs. Pages 371–395 in G. A. Polis and K. O. Winemiller, eds. Food webs: integration of patterns and dynamics. Chapman & Hall, New York. Altegrim, O. 1989. Exclusion of birds from bilberry stands: impact on insect larval density and damage to the bilberry. Oecologia (Berlin) 79:136–139. Beckerman, A. P., M. Uriarte, and O. J. Schmitz. 1997. Experimental evidence for a behavior-mediated cascade in a terrestrial food chain. Proceedings of the National Academy of Sciences of the USA 94:10735–10738. Begon, M., J. L. Harper, and C. R. Townsend. 1996. Ecology: individuals, populations and communities. Blackwell Science, London. Batzli, G. O. 1986. Nutritional ecology of the California vole: effects of food quality on reproduction. Ecology 67:406–412. Billick, I., and T. J. Case. 1994. Higher order interactions in ecological communities: what are they and how can they be detected? Ecology 75:1529–1543. Bock, C. E., J. H. Bock, and M. C. Grant. 1992. Effects of bird predation on grasshopper densities in an Arizona grassland. Ecology 73:1706–1717. Brett, M. T., and C. R. Goldman. 1996. A meta-analysis of the freshwater trophic cascade. Proceedings of the National Academy of Sciences of the USA 93:7723–7726. ———. 1997. Consumer versus resource control in freshwater pelagic food webs. Science (Washington, D.C.) 275:384–386. Brooks, J. L., and S. I. Dodson. 1965. Predation, body size, and composition of plankton. Science (Washington, D.C.) 150:28–35. Brust, G. E. 1994. Natural enemies in straw-mulch reduce Colorado potato beetle populations and damage to potato. Biological Control 4:163–169. Brust, G. E., B. R. Stinner, and D. A. McCartney. 1985. Tillage and soil insecticide effects on predator–black cutworm (Lepidoptera: Noctuidae) interactions in corn agroecosystems. Journal of Economic Entomology 78: 1389–1392. Carpenter, S. R., J. F. Kitchell, and J. R. Hodgson. 1985. Cascading trophic interactions and lake productivity. BioScience 35:634–639. Carpenter, S. R., J. F. Kitchell, J. R. Hodgson, P. A. Cochran, J. J. Elser, M. M. Elser, D. M. Lodge, D. Kretchmer, X. He, and C. N. von Ende. 1987. Regulation of lake primary productivity by food web structure. Ecology 68: 1863–1876. Carter, P. E., and A. L. Rypstra. 1995. Top-down effects in Terrestrial Trophic Cascades 277 soybean agroecosystems: spider density affects herbivore damage. Oikos 72:433–439. Chase, J. M. 1996. Abiotic controls of trophic cascades in a simple grassland food chain. Oikos 77:495–506. Clark, M. S., J. M. Luna, N. D. Stone, and R. R. Youngman. 1994. Generalist predator consumption of armyworm (Lepidoptera: Noctuidae) and effect of predator removal on damage in no-till corn. Environmental Entomology 23:617–622. Coaker, T. H. 1965. Further experiments on the effect of beetle predators on the numbers of the cabbage root fly, Erioischia brassicae (Bouché), attacking brassica crops. Annals of Applied Biology 56:7–20. Cohen, J. 1969. Statistical power analysis for the behavioral sciences. Academic Press, New York. Coley, P. D., J. P. Bryant, and F. S. I. Chapin. 1985. Resource availability and plant antiherbivore defense. Science (Washington, D.C.) 230:895–899. Collins, S. L., A. K. Knapp, J. M. Briggs, J. M. Blair, and E. M. Steinauer. 1998. Modulation of diversity by grazing and mowing in native tallgrass prairie. Science (Washington, D.C.) 280:745–747. Connell, J. H. 1983. On the prevalence and relative importance of interspecific competition: evidence from field experiments. American Naturalist 122:661–696. Davidson, J., and H. G. Andrewartha. 1948. The influence of rainfall, evaporation and atmospheric temperature on fluctuations in the size of a natural population of Thrips imaginis (Thysanoptera). Journal of Animal Ecology 17: 200–222. Dial, R. 1992. A food web for a tropical rain forest: the canopy view from Anolis. Ph.D. diss. Stanford University, Stanford, Calif. Dyer, L. A., and D. K. Letourneau. 1999. Trophic cascade in a complex terrestrial community. Proceedings of the National Academy of Sciences of the USA 96:5072–5076. Ehrlich, P. R., and L. C. Birch. 1967. The “balance of nature” and “population control.” American Naturalist 101:97–107. Englund, G., O. Sarnelle, and S. D. Cooper. 1999. The importance of data-selection criteria: meta-analysis of stream predation experiments. Ecology 80:1132–1141. Errington, P. L. 1956. Factors limiting higher vertebrate populations. Science (Washington, D.C.) 124:304–307. Erwin, T. L. 1982. Tropical forests: their richness in Coleoptera and other arthropod species. Coleopterists Bulletin 36:74–75. Faeth, S. H. 1994. Induced plant responses: effects on parasitoids and other natural enemies of phytophagous insects. Pages 245–260 in B. A. Hawkins and W. Sheehan, eds. Parasitoid community ecology. Oxford University Press, Oxford. Feeny, P. P. 1970. Seasonal changes in oak leaf tannins and nutrients as a cause of spring feeding by winter month caterpillars. Ecology 51:565–581. Floyd, T. 1996. Top-down impacts on creosotebush herbivores in a spatially and temporally complex environment. Ecology 77:1544–1555. Fowler, A. C., R. L. Knight, T. L. George, and L. C. McEwen. 1991. Effects of avian predation on grasshopper populations in North Dakota grasslands. Ecology 72:1775–1781. Fretwell, S. D. 1977. The regulation of plant communities by the food chains exploiting them. Perspectives in Biology and Medicine 20:169–185. Goldberg, D. E., T. Rajaniemi, J. Gurevitch, and A. StewartOaten. 1999. Empirical approaches to quantifying interaction intensity: competition and facilitation along productivity gradients. Ecology 80:1118–1131. Gómez, J., and R. Zamora. 1994. Top-down effects in a tritrophic system: parasitoids enhance plant fitness. Ecology 75:1023–1030. Gould, W. P., and R. L. Jeanne. 1984. Polistes wasps (Hymenoptera: Vespidae) as control agents for lepidopterous cabbage pests. Environmental Entomology 13:150–156. Gurevitch, J., and L. V. Hedges. 1993. Meta-analysis: combining the results of independent experiments. Pages 378–398 in S. M. Scheiner and J. Gurevitch, eds. Design and analysis of ecological experiments. Chapman & Hall, New York. ———. 1999. Statistical issues in ecological meta-analysis. Ecology 80:1142–1149. Gurevitch, J., L. L. Morrow, A. Wallace, and J. S. Walsh. 1992. A meta-analysis of competition in field experiments. American Naturalist 140:539–572. Gutiérrez, J. R., P. L. Meserve, S. Herrera, L. C. Contreras, and F. M. Jaksic. 1997. Effects of small mammals and vertebrate predators on vegetation in the Chilean semiarid zone. Oecologia (Berlin) 109:398–406. Hairston, N. G. 1994. Ecological experiments: purpose, design, and execution. Cambridge University Press, Cambridge. Hairston, N. G., F. E. Smith, and L. B. Slobodkin. 1960. Community structure, population control, and competition. American Naturalist 94:421–425. Hartley, M. J., and M. L. Hunter. 1998. A meta-analysis of forest cover, edge effects, and artificial nest predation rates. Conservation Biology 12:465–469. Hawkins, B. A. 1992. Parasitoid-host food webs and donor control. Oikos 65:159–162. Hawkins, B. A., and P. Gross. 1992. Species richness and population limitation in insect parasitoid-host systems. American Naturalist 139:417–423. Hawkins, B. A., N. J. Mills, M. A. Jervis, and P. W. Price. 1999. Is the biological control of insects a natural phenomenon? Oikos 86:493–506. 278 The American Naturalist Hedges, L. V., and I. Olkin. 1985. Statistical methods for meta-analysis. Academic Press, Boston. Hedges, L. V., J. Gurevitch, and P. S. Curtis. 1999. The meta-analysis of response ratios in experimental ecology. Ecology 80:1150–1156. Helenius, J. 1990. Effect of epigeal predators on infestation by the aphid Rhopalosiphum padi and on grain yield of oats in monocrops and mixed intercrops. Entomologia Experimentalis et Applicata 54:225–236. Henttonen, H., T. Oksanen, A. Jortikka, and V. Haukisalmi. 1987. How much do weasels shape microtine cycles in the northern Fennoscandian taiga? Oikos 50: 353–365. Holt, R. D. 1977. Predation, apparent competition, and the structure of prey communities. Theoretical Population Biology 12:197–229. ———. 1996. Community modules. Pages 333–350 in M. Begon, A. Gange, and V. Brown, eds. Multitrophic interactions. Chapman & Hall, London. Holt, R. D., and J. H. Lawton. 1993. Apparent competition and enemy-free space in insect host-parasitoid communities. American Naturalist 142:623–645. Hrbácek, J. M., M. Dvoráková, V. Korı́nek, and L. Procházková. 1961. Demonstration of the effect of the fish stock on the species composition of zooplankton and the intensity of metabolism of the whole plankton association. Verhandlungen der Internationalen Vereinigung für Limnologie 14:192–195. Hunter, M. D., and P. W. Price. 1992. Playing chutes and ladders: heterogeneity and the relative roles of bottomup and top-down forces in natural communities. Ecology 73:724–732. Järvinen, A. 1991. A meta-analytic study of the effects of female age on laying-date and clutch-size in the great tit Parus major and the pied flycatcher Ficedula hypoleuca. Ibis 133:62–67. Kajak, A. 1997. Effects of epigeic macroarthropods on grass litter decomposition in mown meadow. Agriculture, Ecosystems, and Environment 64:53–63. Kajak, A., and H. Jakubczyk. 1977. Experimental studies on predation in the soil-litter interface. Ecological Bulletin 25:493–496. Koricheva, J., S. Larsson, and E. Haukioja. 1998. Insect performance on experimentally stressed woody plants: a meta-analysis. Annual Review of Entomology 43: 195–216. Lawrence, K. L., and D. H. Wise. 2000. Spider predation on forest-floor Collembola and evidence for indirect effects on decomposition. Pedobiologia 44:33–39. Lawton, J. H. 1999. Are there general laws in ecology? Oikos 84:177–192. Leibold, M. A. Resource edibility and the effects of pred- ators and productivity of the outcome of trophic interactions. American Naturalist 134:922–949. Letourneau, D. K., and L. A. Dyer. 1998. Experimental test in lowland tropical forest shows top-down effects through four trophic levels. Ecology 79:1678–1687. Light, R. J., and D. B. Pillemer. 1984. Summing up: the science of reviewing research. Harvard University Press, Cambridge, Mass. MacInnes, C. D., and R. H. Kerbes. 1987. Growth of the snow goose, Chen caerulescens, colony at McConnell River, Northwest Territories: 1940–1980. Canadian Field Naturalist 101:33–39. Mansour, F., and W. H. Whitcomb. 1986. The spiders of a citrus grove in Israel and their role as biocontrol agents of Ceroplastes floridensis (Homoptera: Coccidae). Entomophaga 31:269–276. Mansour, F., M. Wysoki, and W. H. Whitcomb. 1985. Spiders inhabiting avocado orchards and their role as natural enemies of Boarmia selenaria Schiff. (Lepidoptera: Geometridae) larvae in Israel. Acta Oecologica Oecologia Applicata 6:315–321. Marquis, R. J., and C. J. Whelan. 1994. Insectivorous birds increase growth of white oak through consumption of leaf-chewing insects. Ecology 75:2007–2014. McCann, K. S., A. Hastings, and D. R. Strong. 1998. Trophic cascades and trophic trickles in pelagic food webs. Proceedings of the Royal Society of London B, Biological Sciences 265:205–209. McLaren, B. E., and R. O. Peterson. 1994. Wolves, moose, and tree rings on Isle Royale. Science (Washington, D.C.) 266:1555–1558. Menge, B. A. 1995. Indirect effects in marine rocky intertidal interaction webs: patterns and importance. Ecological Monographs 65:21–74. Moran, M. D., and L. E. Hurd. 1998. A trophic cascade in a diverse arthropod community caused by a generalist arthropod predator. Oecologia (Berlin) 113:126–132. Moran, M. D., T. P. Rooney, and L. E. Hurd. 1996. Topdown cascade from a bitrophic predator in an old-field community. Ecology 77:2219–2227. Mowat, D. J., and S. J. Martin. 1981. The contribution of predatory beetles (Coleoptera: Carabidae and Staphylinidae) and seed-bed-applied insecticide to the control of cabbage root fly, Delia brassicae (Wied.), in transplanted cauliflowers. Horticultural Research 21:127–136. Murdoch, W. W. 1966. “Community structure, population control, and competition”—a critique. American Naturalist 100:219–226. Murdoch, W. W., J. Chesson, and P. L. Chesson. 1985. Biological control in theory and practice. American Naturalist 125:344–366. NRC (National Research Council). 1996. Ecologically based pest management: new solutions for a new century. Com- Terrestrial Trophic Cascades 279 mittee on Pest and Pathogen Control through Management of Biological Control Agents and Enhanced Cycles and Natural Process. National Academy Press, Washington, D.C. Oksanen, L., and T. Oksanen. 2000. The logic and realism of the hypothesis of exploitation ecosystems. American Naturalist 155:703–723. Oksanen, L., S. D. Fretwell, J. Arruda, and P. Niemelä. 1981. Exploitation ecosystem in gradients of primary productivity. American Naturalist 118:240–261. Osenberg, C. W., O. Sarnelle, and S. D. Cooper. 1997. Effect size in ecological experiments: the application of biological models to meta-analysis. American Naturalist 150:798–812. Osenberg, C. W., O. Sarnelle, and D. E. Goldberg. 1999. Meta-analysis in ecology: concepts, statistics, and applications. Ecology 80:1103–1104. Owen, D. F. 1980. Camouflage and mimicry. Oxford University Press, Oxford. Pacala, S., and J. Roughgarden. 1984. Control of arthropod abundance by Anolis lizards on St. Eustatius (Neth. Antilles). Oecologia (Berlin) 64:160–162. Paine, R. T. 1966. Food web complexity and species diversity. American Naturalist 100:65–75. ———. 1980. Food webs: linkage, interaction strength and community infrastructure. Journal of Animal Ecology 49:667–685. Palmer, A. R. 1999. Detecting publication bias in metaanalyses: a case study of fluctuating asymmetry and sexual selection. American Naturalist 154:220–233. Pimentel, D. 1961. An evaluation of insect resistance in broccoli, Brussels sprouts, cabbage, collards and kale. Journal of Economic Entomology 54:156–158. Polis, G. A. 1991. Complex trophic interactions in deserts: an empirical critique of food-web theory. American Naturalist 138:123–155. ———. 1994. Food webs, trophic cascades and community structure. Australian Journal of Ecology 19:121–136. ———. 1999. Why are parts of the world green? multiple factors control productivity and the distribution of biomass. Oikos 86:3–15. Polis, G. A., and R. D. Holt. 1992. Intraguild predation: the dynamics of complex trophic interactions. Trends in Ecology & Evolution 7:151–154. Polis, G. A., and S. D. Hurd. 1996. Allochthonous input across habitats, subsidized consumers, and apparent trophic cascades: examples from the ocean-land interface. Pages 275–285 in G. A. Polis and K. O. Winemiller, eds. Food webs: integration of patterns and dynamics. Chapman & Hall, New York. Polis, G. A., and D. R. Strong. 1996. Food web complexity and community dynamics. American Naturalist 147: 813–846. Polis, G. A., C. A. Myers, and R. D. Holt. 1989. The ecology and evolution of intraguild predation: potential competitors that eat each other. Annual Review of Ecology and Systematics 20:297–330. Power, M. E. 1990. Effects of fish in river food webs. Science (Washington, D.C.) 250:811–814. Power, M. E., W. J. Matthews, and A. J. Stewart. 1985. Grazing minnows, piscivorous bass, and stream algae: dynamics of a strong interaction. Ecology 66:1448–1456. Reitze, M., and W. Nentwig. 1991. Comparative investigations into the feeding ecology of six Mantodea species. Oecologia (Berlin) 86:568–574. Riechert, S. E., and L. Bishop. 1990. Prey control by an assemblage of generalist predators: spiders in garden test systems. Ecology 71:1441–1450. Risch, S. J., and C. R. Carroll. 1982. Effect of a keystone predaceous ant, Solenopsis geminata, on arthropods in a tropical agroecosystem. Ecology 63:1979–1983. Rosenberg, M. S., D. C. Adams, and J. Gurevitch. 1997. MetaWin: statistical software for meta-analysis with resampling tests. Version 1.0. Sinauer Associates, Sunderland, Mass. Samways, M. J. 1994. Insect conservation biology. Chapman & Hall, London. Schmitz, O. J. 1993. Trophic exploitation in grassland food chains: simple models and a field experiment. Oecologia (Berlin) 93:327–335. ———. 1994. Resource edibility and trophic exploitation in an old-field food web. Proceedings of the National Academy of Sciences of the USA 91:5364–5367. ———. 1998. Direct and indirect effects of predation and predation risk in old-field interaction webs. American Naturalist 151:327–342. Schmitz, O. J., A. P. Beckerman, and K. M. O’Brien. 1997. Behaviorally mediated trophic cascades: effects of predation risk on food web interactions. Ecology 78: 1388–1399. Schmitz, O. J., P. A. Hambäck, and A. P. Beckerman. 2000. Trophic cascades in terrestrial systems: a review of the effects of carnivore removals on plants. American Naturalist 155:141–153. Schoener, T. W. 1989. Food webs from the small to the large. Ecology 70:1559–1589. ———. 1993. On the relative importance of direct versus indirect effects in ecological communities. Pages 365–411 in H. Kawanabe, J. E. Cohen, and K. Iwasaki, eds. Mutualism and community organization: behavioral, theoretical and food web approaches. Oxford University Press, Oxford. Schoenly, K., R. A. Beaver, and T. A. Heumier. 1991. On the trophic relations of insects: a food-web approach. American Naturalist 137:597–638. Sih, A., P. Crowley, M. McPeek, J. Petranka, and K. Stroh- 280 The American Naturalist meier. 1985. Predation, competition and prey communities: a review of field experiments. Annual Review of Ecology and Systematics 16:269–311. Sinclair, A. R. E. 1975. The resource limitation of trophic levels in tropical grassland ecosystems. Journal of Animal Ecology 44:497–520. Slobodkin, L. B., F. E. Smith, and N. G. Hairstone. 1967. Regulation in terrestrial ecosystems, and the implied balance of nature. American Naturalist 101:109–124. Snyder, W. E., and D. H. Wise. 1999. Predator interference and the establishment of generalist predator populations for biocontrol. Biological Control 15:283–292. ———. 2000. Antipredator behavior of spotted cucumber beetles, Diabrotica undecimpunctata howardi (Coleoptera: Chrysomelidae), in response to predators that pose varying risks. Environmental Entomology 29:35–42. ———. In press. Contrasting trophic cascades generated by a community of generalist predators. Ecology. Southwood, T. R. E. 1978. The components of diversity. Symposia of the Royal Entomological Society of London 9:19–40. Spiller, D. A., and T. W. Schoener. 1990. A terrestrial field experiment showing the impact of eliminating top predators on foliage damage. Nature (London) 347:469–472. ———. 1994. Effects of top and intermediate predators in a terrestrial food web. Ecology 75:182–196. ———. 1996. Food-web dynamics on some small subtropical islands: effects of top and intermediate predators. Pages 160–169 in G. A. Polis and K. O. Winemiller, eds. Food webs: integration of patterns and dynamics. Chapman & Hall, New York. ———. 1997. Folivory on islands with and without insectivorous lizards: an eight-year study. Oikos 78:15–22. Stamp, N. E., and M. D. Bowers. 1991. Indirect effect on survivorship of caterpillars due to presence of invertebrate predators. Oecologia (Berlin) 88:325–330. Stoner, K. A. 1993. Effects of straw and leaf mulch and trickle irrigation on the abundance of Colorado potato beetles (Coleoptera: Chrysomelidae) on potato in Connecticut. Journal of Entomological Science 28:393–403. Strauss, S. Y. 1991. Indirect effects in community ecology: their definition, study and importance. Trends in Ecology & Evolution 6:206–210. Strong, D. R. 1992. Are trophic cascades all wet? differentiation and donor-control in speciose ecosystems. Ecology 73:747–754. Strong, D. R., H. K. Kaya, A. V. Whipple, A. L. Child, S. Kraig, M. Bondonno, K. Dyer, and J. L. Maron. 1996a. Entomopathogenic nematodes: natural enemies of rootfeeding caterpillars on bush lupine. Oecologia (Berlin) 108:167–173. Strong, D. R., J. L. Maron, and P. G. Connors. 1996b. Topdown from underground? the underappreciated influ- ence of subterranean food webs on above-ground ecology. Pages 170–175 in G. A. Polis and K. O. Winemiller, eds. Food webs: integration of patterns and dynamics. Chapman & Hall, New York. Strong, D. R., A. V. Whipple, A. L. Child, and B. Dennis. 1999. Model selection for a subterranean trophic cascade: root-feeding caterpillars and entomopathogenic nematodes. Ecology 80:2750–2761. Tansley, A. G., and R. S. Adamson. 1925. Studies of the vegetation of the English chalk. III. The chalk grasslands of the Hampshire-Sussex border. Journal of Ecology 13: 177–223. Tonhasca, A., and D. N. Byrne. 1994. The effects of crop diversification on herbivorous insects: a meta-analysis approach. Ecological Entomology 19:239–244. Tuntibunpakul, P. 1999. The impact of spiders and ground beetles on the abundance of insect pests and yields in soybean and vegetable agroecosystems. Ph.D. diss. University of Kentucky, Lexington. Vanni, M. J., and D. L. Findlay. 1990. Trophic cascades and phytoplankton community structure. Ecology 71: 921–937. Warrington, S., and J. B. Whittaker. 1985. An experimental field study of different levels of insect herbivory induced by Formica rufa predation on sycamore (Acer pseudoplatanus). I. Lepidoptera larvae. Journal of Applied Ecology 22:775–785. White, T. C. R. 1978. The importance of a relative shortage of food in animal ecology. Oecologia (Berlin) 33:71–86. Whittaker, J. B., and S. Warrington. 1985. An experimental field study of different levels of insect predation induced by Formica rufa predation on sycamore (Acer pseudoplatanus). III. Effects on tree growth. Journal of Applied Ecology 22:797–811. Wilson, E. O. 1987. The little things that run the world. Conservation Biology 1:344–346. Wise, D. H. 1984. The role of competition in spider communities: insights from field experiments with a model organism. Pages 54–63 in D. R. Strong, D. Simberloff, L. G. Abele, and A. B. Thistle, eds. Ecological communities: conceptual issues and the evidence. Princeton University Press, Princeton, N.J. ———. 1993. Spiders in ecological webs. Cambridge University Press, Cambridge. Witz, B. W. 1990. Antipredator mechanisms in arthropods: a twenty year literature survey. Florida Entomologist 73: 71–99. Wootton, J. T. 1993. Indirect effects and habitat use in an intertidal community: interaction chains and interaction modifications. American Naturalist 141:71–89. ———. 1994. The nature and consequences of indirect effects in ecological communities. Annual Review of Ecology and Systematics 25:443–466. Terrestrial Trophic Cascades 281 Wright, D. W., R. D. Hughes, and J. Worrall. 1960. The effect of certain predators on the numbers of cabbage root fly (Erioischia brassicae [Bouché]) and on the subsequent damage caused by the pest. Annals of Applied Biology 48:756–763. Zehnder, G. W., and J. Hough-Goldstein. 1990. Colorado potato beetle (Coleoptera: Chrysomelidae) population development and effects on yield of potatoes with and without straw mulch. Journal of Economic Entomology 83:1982–1987. Associate Editor: Peter J. Morin
© Copyright 2026 Paperzz