Antagonistic interactions between competition and insect herbivory on plant growth

Authors

  • Joshua J. Haag,

    1. Department of Botany, University of Toronto, Toronto, Ontario, M5S 3B2, Canada, and
    Search for more papers by this author
    • *

      Present address: Department of Ecology and Evolutionary Biology, University of Michigan, Ann Arbor, MI, 48109–1048, USA.

  • Malcolm D. Coupe,

    1. Department of Botany, University of Toronto, Toronto, Ontario, M5S 3B2, Canada, and
    2. Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada
    Search for more papers by this author
  • James F. Cahill Jr

    Corresponding author
    1. Department of Botany, University of Toronto, Toronto, Ontario, M5S 3B2, Canada, and
    2. Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada
    Search for more papers by this author

‡ Correspondence: James Cahill (tel. 780 492 3792; fax 780 492 9234; e-mail jc.cahill@ualberta.ca ).

Summary

  • 1Although substantial theoretical work suggests that competition and herbivory should exhibit a wide variety of interactions in their effects on plant growth, empirical studies have shown that the predominant interactions are simply multiplicative.
  • 2To determine both the relative strengths of, and interactions between, competition and herbivory, we conducted a field experimental study in a native grassland using four focal species: Koeleria macrantha , Coreopsis tinctoria , Linum lewisii and Helianthus petiolaris .
  • 3The effects of competition and herbivory on plant growth, biomass allocation and survival varied among species. When effects were present, neighbouring vegetation reduced plant growth and survival, and insect herbivores decreased relative biomass allocation to roots and increased plant mortality.
  • 4Although reduced herbivory caused by insecticide application had little direct effects on plant biomass, it did interact with competition to affect growth for three of the four species. Herbivory reduced the strength of competition experienced by Coreopsis , but increased it in Linum . For both Coreopsis and Helianthus , the combined effects of herbivory and competition on plant growth were less than expected from a simple multiplicative response.
  • 5We suggest that interactions were found because we used an experimental design that modified insect densities on both neighbours and the focal plant, allowing for determination of both direct and indirect effects.
  • 6We suggest that the antagonistic interactions between competition and herbivory may have occurred because herbivores altered the competitive environment by harming the neighbouring plants, or because the presence of neighbours facilitated focal plant growth by distributing herbivore loads over the greater amount of plant biomass available.
  • 7Although as an isolated factor competition is more intense in this system than insect herbivory, herbivory can alter the strength of competitive interactions and thus its importance cannot be determined by measuring intensity alone. This result emphasizes that these two ecological processes are not truly discrete, and further suggests that studies that simultaneously manipulate multiple ecological processes are needed.

Introduction

Competition and herbivory influence plant growth and fecundity (Ehrlen 1995a; Maron 1998), stature and growth form (Louda 1984; Paige & Whitham 1987), population growth (Bullock et al. 1994; Ehrlen 1995b; Miriti et al. 2001), geographical range (Bruelheide & Scheidel 1999), and succession (Brown & Gange 1992). Although they are generally regarded as among the most important deterministic, extrinsic factors that influence plant growth and community structure (Harper 1977; Crawley 1990; Gurevitch et al. 2000), it remains unclear how competition and herbivory interact to affect growth in natural systems.

There is substantial theoretical work suggesting that competition and herbivory should exhibit complex interactions in their combined effects on plant growth (Herms & Mattson 1992; Herms 1999; Weis & Hochberg 2000; Grover 2002). Herms (1999) outlined a model of trade-offs between shoot growth rates, which could be important in determining competitive abilities when competition is for light (i.e. when competition is size-asymmetric; Harper 1977; Weiner & Thomas 1986), and herbivore resistance. Assuming that plants will allocate more biomass to tissues responsible for capturing a limiting resource (e.g. Tilman 1988; but see Cahill 2003), shifts in biomass allocation resulting from herbivory may substantially alter competitive ability (Rodriguez & Brown 1998). Although trade-offs between competitive ability and herbivore resistance have been demonstrated in some species (Briske & Anderson 1992; Rosenthal & Dirzo 1997; Siemann & Rogers 2001), interactions are most commonly found to be multiplicative (reviewed by Sheppard 1996). This means that each factor has the same proportional effect, regardless of the level of other factors (Notzold et al. 1998), and such multiplicative interactions are often used as the null model (Rees & Brown 1992; Reader & Bonser 1998; Cahill 1999). If this is the case, herbivory does not generally alter a plant's competitive ability and these two ecological processes are independent. However, it makes intuitive sense that if a focal plant's neighbour is eaten, then it will be less able to compete with that focal plant. Additionally, if herbivory is influenced by plant size, then any factor that alters plant size (e.g. competition) should also alter herbivory (Agrawal & Van Zandt 2003). It appears that most studies designed to determine the combined effects of herbivory and competition on focal plants have restricted the scale of their insect manipulations to the focal individual, and do not generally modify the herbivores on neighbouring plants (Reader 1992; Rees & Brown 1992; Reader & Bonser 1998). Hence, it is possible that in some cases, the nature of the interactions observed may be a result of artificially reducing the scale of influence of herbivores, from many plants to one. This is likely to be of most concern in systems in which generalist, rather than specialist, herbivores are most common.

The existence of interactions has important implications for the measurement of the intensity and relative importance of competition and herbivory on plant growth. Welden & Slauson's (1986, p. 26) definition of the intensity of competition as the ‘amount of strain competition induces in an organism’, can be adapted to apply to any ecological factor (e.g. herbivory). A variety of ways of determining intensity have been proposed, which can be applied to either gradients of a given factor, thereby measuring the intensity of competition relative to neighbour abundance (e.g. Goldberg & Werner 1983), or, as here, to removal experiments that measure the absolute intensity of competition (Welden & Slauson 1986). Welden & Slauson (1986, p. 27) further define the importance of competition as ‘the relative degree to which competition contributes to the overall decrease in growth rate, metabolism, fecundity, survival, or fitness of that organism below its optimal condition’ (which can again be adapted to describe the intensity of herbivory). Weldon and Slauson acknowledge the difficulties in determining the optimal state for an organism, but argue that comparing the intensity and importance of different processes does not require the actual optimal state to be known, but rather that we need only choose an appropriate measure against which to compare values.

While intensity is a physiological concept, the concept of importance is ecological, dependent upon all other extrinsic processes that simultaneously act on an organism. The need to differentiate between the intensity and importance of multiple ecological factors affecting community organization is widely recognized (Welden & Slauson 1986; Sammul et al. 2000), but appears to have had little practical effect in plant ecology. The limited number of studies that have taken an approach that allows for comparisons of the importance and intensity of competition and herbivory generally conclude that competition is the more important factor of the two (Milchunas et al. 1992; Rees & Brown 1992; Reader & Bonser 1998; Willis et al. 1998; Dormann et al. 2000). However, the importance of a factor is a measure of that factor's effect on fitness, relative to the average effect of all other factors. Therefore, if competition varies as a result of herbivory (or vice versa), the nature of such an interaction will necessarily be excluded from the examination of their relative importance. In other words, the concept of even being able to quantify relative importance is itself dependent upon the assumption of multiplicative (or additive) interactions among ecological processes.

We determined the direct impact of competition and insect herbivory on the growth and biomass allocation of four native grassland species, thereby providing a measure of their intensity. We also compare the relative intensities of these factors, allowing estimation of the importance of these ecological processes. We then determine the nature of any interactions present, and whether the results from our system differ from the null multiplicative model implicit in most examinations of intensity and importance.

Methods

site description and focal species

We conducted this study from May to August 2001 in a 50-ha field within the Kinsella Beef Cattle Research Ranch, in Alberta, Canada. The ranch is approximately 3200 ha in size and is located in the aspen parkland ecoregion (Barbour & Billings 2000). The field used has never been seeded or tilled, and had been grazed annually in the fall by cattle until 2000, at which time grazing in this field was ceased. Annual precipitation during the study year was 57% below the 27-year average.

Vegetation in the field is primarily graminoid (72% by biomass), dominated by Festuca hallii (= F. campestris), Stipa curtiseta and Koeleria macrantha. The majority of plant species, however, are forbs (74% of species), with common species including Galium boreale, Comandra umbellata and Solidago missouriensis (Coupe 2003). Woody plants are present (e.g. Rosa arkansana), though they contribute little to standing biomass or diversity. Over 80% of the phytomass in this system is below-ground, mostly in the upper 20 cm of soil (Coupe 2003).

As the effects of competition on plant growth are often species specific (Wilson & Tilman 1995) we used four native species as focal plants in our study: Linum lewisii Pursh, Koeleria macrantha Ledeb., Helianthus petiolaris Boivin and Coreopsis tinctoria Nutt. For simplicity in this paper, we will refer to each species by its genus name. Linum (Linaceae; Wild Blue Flax) is a perennial forb found in the grassland/forest transition zones at Kinsella. Koeleria (Gramineae; June Grass) is a native tufted perennial C3 grass common throughout the study site. Helianthus (Compositae; Annual Sunflower) and Coreopsis (Compositae; Common Tickseed) are annual forbs found in prairie habitats in southern Alberta, although neither occurs at Kinsella. Coley et al. (1985) suggest that species will vary in palatability and levels of defence against herbivores as a function of life-history traits, so we chose species having a wide range of life histories and growth forms, with both annuals and perennials included. As neither of the annual species occurs at Kinsella, we can be reasonably certain that neither would have a suite of specialist herbivores, and that all damage sustained through herbivory would be a result of generalists. However, any consideration of life-history traits or specialist and generalist herbivores in our study is confounded by considerations of endemic and non-endemic genotypes, and should be viewed with caution. All of these species have the advantage of being relatively resistant to transplant shock, thereby decreasing initial mortality. Nomenclature and species descriptions follow Moss (1994).

experimental design

Ten 15 × 19 m blocks were established in May 2001, distributed over 20 ha, with a minimum of 25 m separating blocks. Each block consisted of four 6 × 8 m plots, with a 1 m mown pathway surrounding each plot. A 1-m-wide buffer was established around the perimeter of each plot to minimize edge effects on the focal plants, and to ensure that any confounding visitation effects (Cahill et al. 2001; Hik et al. 2003) would be both reduced and comparable for all plots. Our initial aim was to examine the effects of both above- and below-ground insect herbivory, but drought conditions prevented the dissolution of the granular soil insecticide used. As a result, the below-ground treatment had no detectable effect (Coupe 2003), and only the two plots randomly assigned as control or sprayed with foliar insecticide are considered here.

At the western edge of each plot inside the buffer zone, a 1 × 4 m strip was marked, and subdivided into 16 subplots, each 0.5 × 0.5 m. We randomly assigned the eight treatment combinations (four species × two neighbour removal) to these subplots, such that there were two replicates of each species × neighbour combination within each plot. In total there were 10 blocks × 2 insecticide treatments × 4 species × 2 neighbour treatments × 2 replicates = 320 subplots.

insecticide application and neighbour removal

Insect densities were reduced through the application of 4.2 mL of liquid chlorpyrifos (Lorsban 4E) in a water emulsion (2.80 mL L−1). The insecticide was applied evenly over the entire plot, covering both neighbours and focal individuals. This foliar insecticide was applied with a backpack sprayer approximately every 2 weeks, although the exact intervals varied to avoid rainfall events. This application rate is within that recommended within an agricultural setting (Johnson 1998) for control of adult grasshoppers, the dominant arthropod herbivores at the site (M.D. Coupe & J.J. Haag, personal observation). The control plots were sprayed with an equal volume of water (equivalent to 0.02 mm rainfall per application and 0.1 mm over the entire experiment).

The relative effects of the insecticide treatment on herbivore densities were quantified by placing 11 × 14 cm acetate cards attached to stakes and coated with Tangle-Trap in every subplot in five randomly chosen blocks on 26 July. After 3 days, the traps were removed, and the number of orthopterans, lepidopterans and homopterans, the most abundant orders containing folivorous/herbivorous insects, were counted on each card. This provided a snapshot of the effects of insecticide application and neighbour removal on herbivore abundances, and was not intended to record changes in insect community structure. Insecticide application reduced insect abundances (F1,28 = 17.90, P < 0.001; Fig. 1b), with target species and neighbour removal having no effect on abundance (all P > 0.15). Although insect abundances and community composition do exhibit seasonality, and it is highly probable that insects moved between neighbour removal treatments, the lack of a difference in abundances between neighbour removal treatments near the peak of insect abundance (and presumably close to maximal herbivory) indicates that this factor was not important in this study.

Figure 1.

(a) Light penetration and (b) insect abundance as a function of focal species identity, neighbour removal and insecticide application. Light penetration is the proportion of available light reaching the soil surface. Insect abundance is the sum of the number of individuals in the orders Orthoptera, Homoptera and Lepidoptera collected on sticky traps. NN corresponds to focal plants surrounded by ‘no neighbours’ and AN corresponds to focal plants surrounded by ‘all neighbours’. The ‘control’ represents unsprayed plots, and ‘sprayed’ indicates plots were sprayed with insecticide.

Each subplot was assigned to a no neighbour (NN) or all neighbour (AN) treatment. On 10 May 2001, we sprayed the NN subplots with an aqueous solution of glyphosate (Roundup®), a non-selective, systemic herbicide. To reduce the likelihood of the herbicide moving from a treated plot into an untreated plot, we (i) cut the perimeter of each subplot to a depth of 12 cm 2 days prior to the herbicide application, and (ii) placed a 45 cm tall, open-ended plastic box on top of the subplots prior to spraying, reducing the potential for drift. This box was removed following herbicide spraying. The NN plots were weeded at regular intervals over the course of the growing season. There was no evidence that the herbicide moved laterally to adjacent plots, as there were no visually apparent differences in neighbour plant density at the edges of AN subplots relative to their centres, regardless of the neighbour treatment applied to the adjacent subplot (J.J. Haag, personal observation).

transplanting and plant measures

Seed of Koeleria, Helianthus and Coreopsis was obtained from Bedrock Seedbanks (Edmonton, Alberta), with parent stock from the surrounding area. Seed of Linum was collected from plants raised outside at the University of Alberta Phytotron. Helianthus, Coreopsis and Linum were sown on 19 April 2001, and Koeleria was sown 5 May 2001. All species were sown into individual cells of potting flats (72 cells per flat) containing Metromix 220®, and placed in a growth room. Germination of all species occurred within 7 days, and all flats were placed outside at the field site 2 weeks prior to transplanting. We transplanted the seedlings into the subplots on 28 May 2001, by forming a 2 cm diameter hole in the centre of each subplot, into which we placed a single plug from the flat containing a few (one to three) individuals of the assigned species. Transplants were watered with approximately 100 mL of water per day for 7 days, at which point we thinned the transplants to one individual per subplot.

We destructively harvested all target individuals on 6–7 August, 10 weeks after they were transplanted. Harvest consisted of recording any evidence of leaf damage caused by invertebrates, clipping the shoot at ground level, and extracting the root systems from the soil. Roots were washed to remove soil, and both roots and shoots were dried at 70 °C and weighed.

light and water measures

We measured soil moisture and light availability in all subplots to indicate their potential response to neighbour removal and insecticide application. Volumetric water content was measured on 4 August 2001 using the HydroSense Soil Water Measurement System (Campbell Scientific, Edmonton, Canada). PAR was measured above and below the leaf canopy in each subplot within 2 h of solar noon on 6 August 2001 using an AccuPar linear PAR ceptometer (PAR-80, Decagon Devices Inc., Pullman, USA).

competition and herbivory indices

The focal species varied widely in size, and thus the effects of neighbour removal and insecticide application on plant biomass are best presented as species-specific measures. The two response variables considered for each species were:

Herbivory response (HR) = CONTROLNN/SPRAYNN( eqn 1)
Competitive response (CR) = SPRAYAN/SPRAYNN( eqn 2)

where, CONTROLNN is the focal plant total biomass (roots + shoot) in an unsprayed plot when grown with no neighbours, SPRAYNN is focal plant biomass in a sprayed plot when grown with no neighbours, and SPRAYAN is focal plant biomass in a sprayed plot when grown with neighbours. These indices represent the proportion of control plant growth that is exhibited by plants growing at natural herbivore densities (equation 1), or in the presence of neighbouring vegetation (equation 2), such that CR or HR = 1 represents no effect of neighbours or insects on plant growth, values < 1 indicate a negative effect on plant growth (e.g. competition, herbivory), and values > 1 indicate a positive effect.

These indices are formulated based on the null model that competition and herbivory do not interact, and therefore that one factor does not alter the strength of the other. The effect of each factor is therefore tested in the absence of the other factor. It is, however, possible that the strength of competition may be influenced by the presence of herbivores, or that the competitive environment may alter the abundance, behaviour or effects of herbivores. These effects do not represent biases in our experimental treatments, but rather are indirect effects and ecological consequences of the nature of competition and herbivory in our study system. They also represent a possible means by which our results could deviate from the null model. Therefore, the presence and detection of such indirect effects would indicate that the simple null multiplicative model is not adequate for describing interactions between competition and herbivory.

All statistical analyses were performed on these indices following natural log transformation, making them equivalent to the total competitive response (Cahill 1999) or log response ratio (Hedges et al. 1999) used for measuring competitive intensity, and extended to the measurement of the intensity of herbivory. These indices are symmetrical to both competitive and facilitative effects, are suitable for measuring effects in removal experiments (neighbouring plants or herbivores), and the logarithmic transformation normalizes the sampling distribution of otherwise skewed data (Hedges et al. 1999; Weigelt & Joliffe 2003). The goal of these measures is to determine the independent strengths of herbivory and competition for each species in each block separately, and they were calculated from the average biomass of the two replicates within each plot (or on the biomass of the only surviving plant if one replicate died).

To determine whether herbivory and competition were independent in their effects on plant growth, we assumed a null model of a multiplicative interaction between these two processes, such that if competition reduced growth by 20% (CR = 0.8), and herbivory reduced growth by 60% (HR = 0.4), their combined effect (Total response = TR) would be CR × HR = 0.32 (plant growth reduced by 68%). This approach is similar to that used by Reader & Bonser (1998). Total response (TR) was calculated for each focal species in each block in two ways:

TRPredicted = CR × HR(eqn 3)
TRTrue = CONTROLAN/SPRAYNN(eqn 4)

where CR and HR are calculated as described in equations 1 and 2, CONTROLAN is focal plant biomass in the presence of neighbours in an unsprayed plot, and SPRAYNN is as previously described. If the combined effects of herbivory and competition are multiplicative, then TRPredicted = TRTrue. Values of TRPredicted > TRTrue indicate a synergistic interaction (one factor increases the strength of another, resulting in reduced plant performance), while values of TRPredicted < TRTrue indicate an antagonistic interaction between these two processes (one factor reduces the strength of the other, resulting in increased plant performance).

statistical analyses

The effect of species identity, insecticide application and neighbour removal on mortality and the occurrence of herbivory were determined using logistic regression. As species varied substantially in the amount of herbivory or mortality suffered, separate analyses were performed for each species. These models consisted of binomial response variables (alive/dead; damaged/undamaged) with neighbour removal and insecticide application as the independent variables.

The analyses of plant biomass and root allocation were conducted using generalized linear mixed models (GLMM), using Proc MIXED in SAS 7.0 (SAS 1999). This procedure provides correct standard errors and appropriate significance tests for fixed effects in unbalanced designs containing random-effects terms (Saavedra & Douglass 2002). Our initial analyses of plant biomass were performed upon both shoot and total plant biomass. Although manipulating neighbour vegetation can result in differences in root extraction efficiencies (Cahill 2002), the qualitative results of the analyses of biomass and competition/herbivory indices did not change as a function of whether shoot or total biomass was used as the response variable. This supports Cahill's (2002) suggestion that measuring root biomass in competition studies may add little additional information. We chose to use total plant biomass as our measure of plant performance because it may be a better overall measure of plant performance for perennial species. The species, herbivore exclusion and neighbour removal treatment were entered into the model as fixed effects, and block, plot and any interaction terms containing either one of them as random effects. Proc MIXED uses a log likelihood function to account for error associated with random effects (Littel et al. 1996; SAS 1999). A backwards stepwise approach was taken for best accounting for the random effects (SAS 1999). After each analysis, the random term with the lowest covariance estimate was dropped from the model and the new model rerun. This process was continued until the removal of a random factor significantly reduced the fit of the model (Littel et al. 1996; SAS 1999). This process partitions all significant error variance due to the random terms, while maintaining the highest possible error degrees of freedom, and thus maximizing the statistical power of the model. The retained random terms for each response variable were: light penetration, block × insecticide × species; biomass, block × insecticide × neighbours and block × insecticide × neighbours × species; soil moisture, block × insecticide × neighbours, and block × insecticide × neighbours × species. Where significant terms were present, least-square means post hoc tests were conducted using the LSMEANS CONTRAST statement in Proc MIXED in SAS.

Relative allocation to root biomass was analysed using the general mixed model described above, with ln(root biomass) as the response variable, and ln(shoot biomass) as a covariate. A significant insecticide application effect would indicate that relative biomass allocation to roots varied as a function of insect removal. Using this approach in analysing patterns of biomass allocation helps differentiate functional shifts in biomass allocation (plasticity) from differences associated with changes in plant size (ontogeny) (Muller et al. 2000).

We tested for effects of competition and herbivory using the indices outlined above by determining if a given index value differed significantly from 1 using t-tests separately for each species. For the competitive index, we performed separate t-tests for each level of insecticide application; for the herbivory index, we performed separate t-tests for each level of neighbour removal. If an index value was significantly smaller than 1, then the factor being tested had a negative effect on plant biomass; values significantly larger than 1 would indicate a positive effect of a factor on plant biomass. Multiple-tests corrections were not performed on these results.

Testing for interactive effects of herbivory and competition on plant growth was done in several ways: (i) testing whether the neighbour removal × insecticide application (or species × neighbour × insecticide) interaction was significant in the GLMM of plant biomass; (ii) conducting a GLMM to determine whether insecticide application influenced the strength of competition (block was a random effect, species and insecticide treatment were fixed effects, and competitive response was the dependent variable); (iii) conducting a GLMM to determine whether neighbour removal altered the strength of herbivory (block was a random effect, species and neighbour removal were fixed effects, and herbivory response was the dependent variable), and (iv) conducting an additional GLMM comparing TRTRUE with TRPREDICTED. In this latter analysis, species and method of calculation (true or predicted; equations 3–4) were fixed effects, with block as a random variable. A significant method of calculation term would indicate either an antagonistic or synergistic interaction between herbivory and competition.

To satisfy the assumption of normality in anova, natural log transformations were performed on the raw values of all biomass variables (including root, shoot and total biomass, and all indices) prior to analysis. This results in the indices being equivalent to the log-response ratio recommended by Goldberg et al. (1999) and Cahill (1999) for removal studies.

Results

abiotic and community level effects

The proportion of available light (PAR) reaching the soil surface increased with neighbour removal (F1,9 = 401.41, P < 0.001; Fig. 1a). Neither target species identity (F3,62.1 = 1.84, P < 0.14) nor insecticide spraying (F1,61.9= 0.95, P < 0.33) affected light availability, but there was a significant insecticide–neighbour removal interaction (F1,228 = 4.53, P = 0.034), where spraying decreased light in the AN, but not NN, plots (Fig. 1a). Soil moisture did not vary as a function of target species identity (F3,107 = 2.05, P < 0.11), insecticide spraying (F1,18 = 0.26, P < 0.61) or neighbour removal (F1,9 = 0.87, P < 0.38), nor were any two- or three-way interactions significant (all P > 0.15).

herbivory, mortality and biomass response

The effects of insecticide application and neighbour removal on focal plant mortality and leaf damage by herbivores varied among species. (Fig. 2a, Table 1). Neighbour removal reduced mortality for Linum and Coreopsis, and insecticide application reduced mortality for Linum. Insecticide application reduced the proportion of Helianthus and Koeleria damaged, while neighbour removal increased the proportion of Helianthus individuals damaged (Fig. 2b). There were no significant two-way interactions (neighbour removal × insecticide application) for any species for either mortality or leaf damage (Table 1).

Figure 2.

Proportion of plants which (a) died or (b) were damaged by insects by the end of the experiment as a function of target plant identity, neighbour removal and insecticide treatments. Plants were scored as damaged if there was evidence of herbivory on any part of the plant body.

Table 1.  Results from logistic regressions designed to determine the effects of neighbour removal and insecticide application on mortality and leaf damage by insect herbivores for each species. In each analysis, a binomial response variable was used (alive vs. dead; damaged vs. undamaged). Neighbour removal, insecticide application, and the neighbour–insecticide interaction all had one degree of freedom. Bold values indicate significance at P  < 0.05
 LinumCoreopsisHelianthusKoeleria
Wald χ2P -value Wald χ2P -value Wald χ2P -value Wald χ2P -value
Mortality
Neighbour  6.84  0.01  4.36  0.04  0.37  0.54< 0.01> 0.90
Insecticide  9.03< 0.01  2.81  0.09  1.37  0.24< 0.01> 0.90
Neighbour × insecticide   0.23  0.63  0.31  0.58  0.02  0.89< 0.01> 0.90
Leaf damage
Neighbour< 0.01> 0.90< 0.01> 0.90  5.89  0.02  3.12  0.08
Insecticide< 0.01> 0.90< 0.01> 0.90 15.86< 0.01  6.46  0.01
Neighbour × insecticide < 0.01> 0.90< 0.01> 0.90< 0.01> 0.90  0.83  0.36

Target plant biomass varied among species and as a function of neighbour removal, with a significant species identity–neighbour removal interaction (Fig. 3a, Table 2). Insecticide application did not alter target plant biomass (Fig. 3a). There was a trend towards an interaction between competition and herbivory within species (Fig. 3a, Table 2), with post hoc tests indicating that the biomass of Coreopsis in the absence of neighbours was higher with herbivores excluded than with herbivores present.

Figure 3.

(a) Total plant biomass (root + shoot) and (b) ln root biomass per unit ln shoot biomass as a function of target plant identity, neighbour removal and insecticide treatments. Plant biomass is presented in a log scale due to the large variation among species. Within each species, different letters above each bar indicate differences in post hoc comparisons (least-square means contrasts, P  < 0.05).

Table 2.  Results of GLMMs analysing the effects of species identity, neighbour removal and insecticide application on target plant biomass (roots + shoots), relative biomass allocation to roots, and the competition and herbivory indices. In the analysis of root biomass, shoot biomass was used as a covariate and thus a significant effect of any factors indicates that the factor alters relative allocation to root biomass, rather than root biomass itself. Competitive response and herbivory response were calculated as described in equations 1 and 2 (see text). Degrees of freedom were calculated using Satterthwaite's method ( Littel et al. 1996 ). Bold values indicate significance at P  < 0.05
LevelTotal biomassRoot biomassCompetitive responseHerbivory response
F (d.f.1,d.f.2) P -value F (d.f.1,d.f.2) P -value F (d.f.1,d.f.2) P -value F (d.f.1,d.f.2) P -value
Ln(shoot)  –356.78 (1,210)< 0.001  –
Species70.75< 0.0017.8< 0.00129.8< 0.0012.40.091
(3,91.9) (3,208) (1,9.83) (3,24.5) 
Neighbour removal278.73< 0.0013.4  0.068  –1.70.221
(1,40.1) (1,113)   (1,11.2) 
Insecticide1.06  0.3104.97  0.0320.2  0.7
(1,40.1) (1,37.9) (1,9.83)   
Species × neighbour91.9< 0.0016.010.001  –1.40.279
(3,91.9) (3,209)   (3,26.2) 
Species × insecticide1.51  0.2180.4  0.7573.7  0.025
(3,91.9) (3,203) (3,24.2)   
Neighbour × insecticide0.97  0.3300.83  0.369
(1,40.1) (1,38.9)     
Spp. × neighbour × insect2.4  0.0730.6  0.616
(3,91.9) (3,203)     

Species identity, insecticide application and the species–neighbour removal interaction all influenced relative biomass allocation to roots (Table 2; Fig. 3b). Target plants of two of the species had lower root allocation when neighbours were present than when they were removed. Although this is the same pattern predicted if root extraction is less efficient when plants were grown with neighbours than when grown without neighbours (Cahill 2002), recent findings in a separate study at this site show that a large percentage of target plant roots (> 25%) must go uncollected before a significant artefact due to extraction efficiency is detected (Cahill 2003). Such a substantial loss of roots is unlikely, as the largest roots, which are the strongest and most reliably extracted, form the greatest proportion of root biomass (Smit et al. 2000). Relative root allocation increased with insecticide application (Table 2, Fig. 3b), and as this pattern is not influenced by differences in root system extraction efficiencies among sprayed and unsprayed plants, this result is likely to be of biological significance.

competition and herbivory indices

Competitive response (CR) was significantly less than 1 for all species in both insecticide treatments (P < 0.001, t-tests; Fig. 4a), indicating that neighbours reduced target plant growth. CR varied among species, with a significant species–insecticide application interaction (Table 2, Fig. 4a). Post hoc tests revealed spraying increased competition for Coreopsis, and reduced competition for L. lewisi (Fig. 4a). Herbivory response (HR) ratios were not different from 1 for any species × neighbour removal combination, except Coreopsis in NN treatment (P < 0.05, t-test; Fig. 4b), and did not vary among species or with neighbour removal (Table 2, Fig. 4b).

Figure 4.

(a) Competitive response as a function of insecticide application, (b) herbivory response as a function of neighbour removal, and (c) total response as function of the method of calculation. Total response was calculated either assuming a multiplicative effect between competition and herbivory (Predicted) or directly through experimentation (True). The horizontal line indicates a value of 1, which indicates no net effect of spraying or neighbour removal on plant biomass. Asterisks indicate a difference in post hoc tests contrasting the means within a given species (least-square means contrasts, P  < 0.05).

In the GLMM analysing whether the combined effects of competition and herbivory were a simple product of their individual effects, there was a significant species × method of calculation interaction (F3,23.6 = 3.92, P = 0.021; Fig. 4c). The two non-endemic annual species, Coreopsis and Helianthus, demonstrated an antagonistic interaction between competition and herbivory, with actual plant growth in the presence of neighbours and insects greater than predicted assuming a multiplicative effect between the two ecological factors (Fig. 4c), while the non-endemic perennials did not exhibit this interaction. Despite finding only a trend for a species × neighbour removal × insecticide application in the analysis of plant biomass (P = 0.073; Table 2), we find a significant species–method interaction here. Such interactions are more likely to be found in this latter analysis for several reasons: (i) there are fewer factors in the GLMM, and we are testing for a lower order interaction; (ii) these ratios standardize for differences in size across species, eliminating a main source of variation in the biomass GLMM; and (iii) calculating the ratios using the average response of the two replicates within each plot reduces variation within blocks.

Discussion

independent effects of competition and herbivory on biomass and mortality

Growth of all four species was more strongly limited by competition than insect herbivory (Figs 3 and 4), indicating competition to be the more intense process. There was, generally, no discernible direct effect of herbivory on plant biomass, suggesting that if competition and herbivory were independent, competition would also be the more important of the two in terms of reducing plant growth. Although herbivory did not directly alter plant biomass, it did increase mortality for Linum (Table 2, Fig. 2a). For this species therefore, conclusions about relative strengths and importance would differ if survival, rather than biomass, was used as the indicator of fitness. Which life stages matter most for future representation remains a matter of debate (Howard & Goldberg 2001; Aarssen & Keogh 2002).

The lack of a direct effect of herbivory on plant biomass could be due to a variety of mechanisms. Compensatory growth following insect herbivory can result in no net effect on individual plant growth (e.g. Rees & Brown 1992; Huhta et al. 2000). Hawkes & Sullivan (2001) found herbivory to affect forbs less negatively than graminoids, with compensation most prevalent in forbs under low resource conditions. All four species did exhibit insect damage during this drought year (Fig. 2b), indicating that compensation may at least be possible. Alternatively, herbivore densities may have been too low to cause biologically significant effects on plant biomass. There is some evidence that ambient herbivore densities (i.e. non-outbreak levels) may have limited effects on plant growth and production (Gibson et al. 1990; Rees & Brown 1992), although this idea has rarely been acknowledged (e.g. Coupe & Cahill 2003).

If plants do compensate for herbivory to shoots through increased shoot growth, one would expect to find a decrease in relative biomass allocation to roots, even if total plant biomass does not vary. Such a trend was found for three of the four species (Table 2, Fig. 3b). This result is interesting in that root competition is very strong in this system (Cahill 2003), and such a finding is directly counter to the commonly made assumption that plants will increase biomass allocation to those organs responsible for capturing limiting resources (e.g. Tilman 1988). However, recent findings suggest that relative biomass allocation to roots does not influence a plant's below-ground competitive ability, and the observed increase in allocation to shoots might not come at a cost during below-ground competition (Cahill 2003). It is also important to note that competition significantly lowered root allocation for two species, while Cahill (2003) found that there was no change in allocation resulting from root competition for 10 species in this field. One likely explanation is that despite the drought conditions and low soil resource levels, shoot competition played at least some role in the current study.

interactions between competition and herbivory

Herbivory and competition were not independent in their effects on plant biomass for most species tested. Although herbivory alone did not directly alter total biomass, it did reduce competition and indirectly increased growth of three of the four study species. Though the presence of neighbours did not influence the effects of herbivory for any species (Fig. 4b), the application of insecticide influenced the strength of competition experienced by Linum and Coreopsis (Fig. 4a). Additionally, Coreopsis and Helianthus exhibited antagonistic interactions between competition and herbivory (Fig. 4b). Such interactions indicate that measuring the independent effects of ecological factors (intensity) does not provide adequate information regarding the relative importance or influence of that factor on plant growth in a natural community.

Interactions between herbivory and competition can be viewed in two ways. If resources are limiting, interactions between herbivory and competition can occur as the result of trade-offs within the individual plant. Allocation of a resource to one function (e.g. plant defence) can limit performance in other functions (e.g. competitive ability). However, interactions between herbivory and competition can also be due to processes external to the individual plant, mediated through changes in the local neighbourhood (e.g. size structure and insect densities). Such ecologically mediated interactions between competition and herbivory are more complex than those based primarily on physiological trade-offs, because they include both the internal effects of competition and herbivory and their effects on the surrounding vegetation. Although herbivores may directly reduce an individual plant's growth, they could differentially damage the neighbouring plants, resulting in increased light (Fig. 1a) or soil resource availability and a net benefit to the focal plant (Carson & Root 2000). Few studies of competition and herbivory have demonstrated antagonistic interactions such as those we observed. Parmesan (2000) found an antagonistic interaction between insect herbivory and competition that resulted from herbivory only negatively affecting plants at low densities. This result was dependent upon the effect of herbivory varying across competitive environments. In the present study, there are two main hypotheses that could explain why an antagonistic interaction was observed: either the strength of herbivory varied across competitive environments (lower levels of damage in plots with all neighbours present), or competition varied as a result of herbivore removal (lower intensity of competition with herbivores present). The presence of neighbours resulted in a lower probability of a focal individual being damaged by insect herbivores (Fig. 2b), supporting the idea that neighbours may exert a protective influence. This could have come about because there was greater total phytomass in the plots with neighbours present, thus spreading the total herbivore load out over all plants, and thus facilitating focal plant growth by reducing the probability of the focal plant being attacked with neighbours present (as there were no significant differences in insect herbivore abundance between plots with neighbours and without; Fig. 1b). This is supported by other work showing modest changes in community structure and productivity resulting from insect removal, with no one species examined having a detectable response to herbivore exclusion (Coupe 2003). However, there was also reduced light availability in the sprayed plots (Fig. 1a), suggesting that herbivores may in fact influence resource availability for focal plants, providing evidence that herbivory can reduce competition. This potential reduction in competition could have occurred because neighbouring plants were preferentially attacked, but our hypothesis does not necessarily depend on such a response. Herbivory did not have a significant effect on focal plant biomass, but it is possible that the effect of herbivory, when looked at over the scale of the entire neighbourhood, did reduce plant growth enough to alter the competitive environment. In a study of the effects of herbivore exclusion on plant growth conducted at the same site (Coupe 2003), there was no effect of herbivory on the biomass of six grassland species, but herbivory did significantly reduce vegetation cover. However, determining which of these two hypotheses is correct is beyond the scope of the current study. Future studies could directly measure the effect of herbivores on neighbouring biomass, or manipulate herbivore densities to maintain equivalent per plant densities across plots with or without neighbours.

Most prior studies of competition and herbivory have used an experimental design that involves either the experimental addition of specialist herbivores to focal plants (e.g. Notzold et al. 1998), or the exclusion of herbivores immediately around focal plants (e.g. Reader 1992; Rees & Brown 1992; Reader & Bonser 1998). Such methods are ideally suited for examining selective pressures on plants and the physiological effects of competition and herbivory, but they tell us little about the indirect effects of competition and herbivory, or how they interact ecologically to affect the growth of individuals. As the scale of herbivore manipulation is limited to the focal plant, any competition–herbivory interactions would be due to the direct effects of herbivory and competition on the focal plant. It is therefore not surprising that the most commonly observed interactions between herbivory and competition are multiplicative and synergistic (Rees & Brown 1992; Sheppard 1996 and references therein; Notzold et al. 1998; Reader & Bonser 1998; Meiners & Handel 2000). Our study allowed for both direct and indirect effects to shape the interaction between competition and herbivory experienced by individual plants, and we believe this contributed to the different results found here. The relative lack of knowledge about the combined effects of competition and herbivory at the level of the plant neighbourhood may explain the difficulties in extrapolating from individual performance and population dynamics to the structure and behaviour of natural plant communities.

Conclusions

Most studies of the joint effects of competition and herbivory on plants have demonstrated either synergistic or multiplicative interactions between the two, and many individual-based models of plant growth agree with these experimental results. However, restricting the effects of these factors to the focal individuals alone neglects any changes that may result in either altered competitive environments as a result of herbivory, or altered effects of herbivory as a result of the competitive environment. Thus a wide range of potential interactions between the two may have been overlooked as a result of specific experimental designs. To make conclusions about the relative importance of competition and herbivory on plant growth in natural environments, we need to consider both the internal and external effects of processes on plants, which may result in strikingly different outcomes.

Acknowledgements

We gratefully acknowledge funding from an NSERC Discovery Grant (to JFC), NSERC USRA award (to JJH), and the Challenge Grants in Biodiversity Programme, supported by the Alberta Conservation Association (to JFC and MDC). Patrick James, Bryon Shore and Alex Rutherford provided field assistance, and Barry Irving provided permission and logistical support for work at Kinsella Ranch. Seed was provided by Bedrock Seed Bank. The quality of this manuscript was improved through comments by Anna Gosline, Nat Cleavitt, and four anonymous reviewers.

Ancillary