Plant genetic diversity yields increased plant productivity and herbivore performance


*Correspondence author. E-mail:


1. Plant genotypic diversity has important consequences for a variety of ecosystem processes, yet empirical evidence for its effects on productivity, one of the most fundamental of these processes, is lacking. In addition, the performance of insect herbivores in response to high genotypic diversity is unknown, despite previous work demonstrating differential herbivore performance among plant genotypes.

2. We manipulated genotypic diversity of the annual plant Arabidopsis thaliana in both the presence and absence of the generalist herbivore Trichoplusia ni under semi-natural growth conditions. We used nine genotypes (eight ecotypes and one mutant) of A. thaliana known to differ widely in functional traits. Productivity and insect biomass were measured in monocultures and mixtures of all nine genotypes grown at multiple fertilization levels and planting densities.

3. In both the absence and presence of herbivores, genotypic diversity increased plant productivity and survival. This effect was, for the most part, independent of fertility or density. Sampling or selection effects did not appear to be wholly responsible for these results as all genotypes were maintained in equal proportion and no single genotype became dominant for the duration of the experiment.

4. High diversity increased T. ni biomass and survival in all treatments. Insect biomass was positively, but not tightly, correlated to plant biomass, indicating that the higher herbivore performance observed in genotypic mixtures was only partially due to higher productivity.

5.Synthesis. Our data support the idea that even within a single plant species, genotypic diversity can exert strong influences on both the producer and herbivore communities. The exact mechanisms responsible for these effects and the relative importance of genotypic diversity in natural communities warrant further investigation.


In recent decades, there has been considerable interest in the relationship between biodiversity and ecosystem function. A large body of work (reviewed by Hooper et al. 2005) has commonly demonstrated the positive effects of plant diversity on ecosystem processes such as primary productivity (Jolliffe 1997,Hector et al. 1999; Tilman et al. 2001; Fargione et al. 2007), ecosystem stability (Loreau 2000), nutrient cycling (Naeem et al. 1994; Tilman, Wedin & Knops 1996) and resistance to exotic plant invasions (Knops et al. 1999). Plant genotypic diversity, even without changes in species diversity, may have important consequences for ecosystem function as well (reviewed by Hughes et al. 2008), but empirical studies documenting such effects have only recently been conducted. Here, we examine primary productivity at high and low genotypic diversity and determine whether concomitant changes in insect performance impact upon levels of herbivory and plant damage.

There are reasons to believe that plant genetic diversity may alter the functioning of several fundamental ecological processes. For example, different plant genotypes have been shown to vary in important functional traits, such as competitive ability (Cahill, Kembel & Gustafson 2005), forage quality for insect herbivores (Karley et al. 2008) and resistance to herbivory (Maddox & Root 1987; Wise 2007). Although the variation among genotypes may be smaller than that among species, it may nevertheless be large enough to impact ecosystem function, as has been shown for species diversity. Empirical evidence supporting this contention is growing and studies have shown plant genetic diversity to improve ecosystem resistance to exotic plant invasions (Crutsinger, Souza & Sanders 2008), enhance plant community resistance to extreme climatic events (Reusch et al. 2005) and grazing (Hughes & Stachowicz 2004), accelerate litter decomposition rates (Schweitzer et al. 2005), maintain long-term species diversity (Booth & Grime 2003; Vellend 2006), reduce plant disease severity (Zhu et al. 2000) and impact arthropod diversity and community composition (Wimp et al. 2005; Johnson, Lajeunesse & Agrawal 2006). However, few studies have been undertaken to demonstrate the existence of diversity–productivity relationships at the genotypic level (but see Hughes & Stachowicz 2004; Crutsinger et al. 2006), or the effects of such relationships on insect herbivores. Genotypic diversity may, as has been shown for interspecific diversity (Hooper & Vitousek 1997; Loreau & Hector 2001; Tilman et al. 2001; Hector et al. 2002; Cardinale et al. 2007; Tylianakis et al. 2008), potentially enhance primary productivity through additive (i.e. sampling effects) or non-additive mechanisms (i.e. complementarity, selection effects).

Identifying the effects of biodiversity on ecosystem processes requires measurement of responses at multiple trophic levels, rather than primary productivity only, and manipulation of more than one trophic group (Duffy et al. 2007). There is increasing evidence that plant genotypes, even within a single species, can exert differential effects on insect herbivore performance and community structure (Service 1984; Cronin & Abrahamson 2001; Wimp et al. 2005). The effects of multiple plant genotypes grown in mixture on herbivore performance, however, are unknown. In particular, it is unclear whether herbivore performance can be wholly predicted by genotype performance, or whether genetic mixtures may influence herbivores in a non-additive manner. For instance, insect herbivores may achieve higher growth rates when grown on genetically mixed, as opposed to single-food, diets (Bernays et al. 1994; DeMott 1998; Mody, Unsicker & Linsenmair 2007).

Changes in insect abundance or performance, whether positive or negative, can potentially ‘cascade’ back to affect the plant community through higher or lower biomass consumption. For instance, Mulder et al. (1999) found a positive relationship between plant species diversity and invertebrate herbivore damage, and notable changes in the diversity–productivity relationship upon exclusion of invertebrates from experimental plots. Although the functional trait differences among plant genotypes may be minor compared with those among species, they may nevertheless be substantial enough to cascade beyond one trophic level (Johnson 2008).

In this study, we examined how genetic diversity of the model plant Arabidopsis thaliana (L.) Heynh. influenced primary productivity and insect herbivory by the generalist herbivore Trichoplusia ni (Hübner). The experiment was conducted on rooftop facilities in the Department of Biological Sciences at the University of Alberta, Edmonton, Canada. Nine genotypes (ecotypes) of A. thaliana were grown in monocultures or mixtures of all nine genotypes. These non-naturally co-occurring genotypes were chosen to maximize differences in important functional traits (J. F. Cahill, unpubl. data) and allow us to detect genotypic effects. Both diversity treatments were replicated at two levels of each of fertilization, density and herbivory in a fully factorial design, for a total of eight treatment combinations at each level of diversity. In total, 520 pots and 11 700 individual plants were used. By performing our study under controlled conditions and restricting biota to only one plant and one herbivore species, we were able to test if, (i) genotypic richness impacts primary productivity, (ii) genotypic richness impacts herbivore performance, and (iii) richness and herbivory interact to differentially affect the plant community.

Materials and methods

Study system

The model plant A. thaliana is a fast-growing, weedy annual with a worldwide distribution. Because of its ability to self-fertilize and prolific seed production, identical individuals can be grown to serve as replicates. We selected eight ecotypes and one mutant differing substantially in low-nutrient stress tolerance, palatability and competitive ability (J. F. Cahill, unpubl. data). Our goal in using these specific genotypes was to maximize the potential functional variation within a population rather than replicate the genetic diversity found in any particular natural population. In doing so, we are providing a test of whether diversity–function relationships could occur, with future studies needed to determine their relative strengths within natural populations or in other species.

The effects of diversity on insect herbivores and on herbivore damage to producers were tested using the cabbage looper T. ni. This species was chosen because it is an important agricultural pest throughout North America, feeding on many species of crucifers, including A. thaliana. In addition, the species is routinely bred in the laboratory, allowing for relative uniformity among individuals in terms of diet and growth rate.

Experimental design

The experiment was conducted on rooftop facilities at the Department of Biological Sciences, University of Alberta, Edmonton, Canada. Plants were grown in genetic monocultures or mixtures under all possible treatment combinations (eight in total) of fertilization, density and herbivory in a fully factorial experiment. Seedlings were allowed to grow for 1 week before they were transplanted into the appropriate treatment pots, and insects were added to herbivory treatments 4 weeks after transplantation. Harvesting of plant biomass took place c. 7 weeks after seed germination.

Diversity and density treatments

Individuals of A. thaliana were grown in monoculture (1 genotype per pot) or in mixture (9 genotypes per pot) at a density of either 9 (low density) or 36 (high density) individuals per 10-cm2 pot. In monocultures, all 9 or 36 individuals were of the same genetic line, while mixtures contained either 1 or 4 individuals of each line.

Fertilization treatment

Half of the pots in each diversity × density treatment combination received 16 mg m−2 slow-release NPK 4 days after transplantation into flats.

Herbivory treatment

Cohorts of third instar T. ni larvae (mean weight 5.54 ± 0.08 mg insect−1) were added to half of all diversity × density × fertilization treatment combinations. Five larvae were added to each pot subjected to the herbivory treatment c. 5 weeks after seed germination, before most plants had bolted. Insects were allowed to move freely among the plants within a pot.


In total, 80 unique treatment combinations were obtained: eight mixture treatments (2 density × 2 fertilization × 2 herbivory = 8) and 72 monoculture treatments (9 genotypes × 2 density × 2 fertilization × 2 herbivory = 72). Monocultures were replicated five times for each treatment combination, for a total of 72 treatments × 5 replicates = 360 pots (8100 individuals). All mixture treatment combinations were replicated 20 times, for a total of 8 treatments × 20 replicates = 160 pots (3600 individuals). In total, 11 700 individual plants were used.

Growth conditions

The experiment was conducted in outdoor facilities under semi-natural growth conditions, where treatment pots were exposed to wind, rain, hail and natural light. Five replicate 80 × 220 cm blocks, 90 cm apart, were established to control for environmental conditions. Each block contained 104 pots, corresponding to one of each of the nine genetic monoculture × fertilization × density × herbivory treatments (72 pots) and four replicates of each genetic mixture × fertilization × density × herbivory treatment (32 pots).

Seeds were obtained from the Arabidopsis Biological Resource Center in Columbus, Ohio, USA ( A seed increase was conducted on lines under uniform growing conditions. The resulting offspring were then used in the study. Seeds were placed on moist filter paper and placed in a refrigerator for 48 h prior to being transplanted into the experimental pots at the beginning of the growing season (May). We used a 3 : 1 sand : soil mix in all pots, ensuring the low-fertilization treatment was, in fact, low in nutrients. Soon after planting the seeds, we added fertilizer [16 mg m−2 Osmocote slow-release 14 : 14 : 14 fertilizer (The Scotts Company, Marysville, OH, USA)] to the fertilized treatments. Pots were weeded of volunteer plants after c. 1 week of growth, such that only one seedling of a given genotype remained in each planting location.


Insects were confined to pots for 7 days before being removed for enumeration (number alive, dead, and missing) and weighing (fresh weight). Only insects found alive at time of harvest were weighed and included in biomass analyses. Upon removal of larvae from herbivory treatments, all plants in the experiment were allowed to grow for 1 week before each plant was clipped at ground level and weighed (dry weight). As different ecotypes of A. thaliana flower at different times (Pigliucci 2003), it was not possible to measure seed set without introducing error related to growing time. However, because seed production is highly correlated with above-ground biomass in this species (Cahill, Kembel & Gustafson 2005), we assumed that biomass would serve as a surrogate measure of overall plant fitness.

Statistical analyses

All analyses were performed using Linear Mixed Model procedures in spss v. 14.0 (SPSS Inc. 2005). Plant and herbivore biomass were square-root transformed prior to analysis to meet assumptions of normality and homogeneity of variance. Diversity, density, fertilization and herbivory were treated as fixed factors and blocks were included as random effects in all statistical models. These analyses yielded similar results for individual plant biomass as for combined pot biomass, thus analyses for individual biomass are not presented. Individual herbivore biomass was estimated by dividing pot herbivore biomass by the number of insects weighed, as raw measurements of individuals were not available. Direct comparisons between treatments (reported as per cent difference) were made using least squares means, thus they were based on square-root transformed rather than raw data.

The effects of the experimental treatments on both plant and herbivore survival were determined using a generalized linear mixed model (Proc Glimmix) in SAS 9.2 (SAS Institute Inc. 2008). In the analysis of herbivore survival, the number of living herbivores served as the response variable (Poisson error distribution; missing insects were treated as dead), block served as a random effect, and fertilization, density and genetic diversity served as the fixed effects. In the analysis of plant survival, the number of living plants found at the end of the experiment served as the response variable (Poisson error distribution), block served as a random effect, and fertilization, density, herbivore presence and genetic diversity served as the fixed effects.


Plant performance in the absence of T. ni

In the absence of any herbivores, plant genetic diversity significantly increased whole-pot plant biomass by 17% relative to genetic monocultures (Fig. 1a,b; Table 1). High plant density and fertilization also increased plant biomass (Fig. 1a,b; Table 1), such that maximum biomass occurred in high density, fertilized mixtures (Fig. 1a,b). No significant two- or three-way interactions were observed (Table 1).

Figure 1.

 Above-ground plant biomass (mean ± 1 SE) per pot (9 plants in low density or 36 plants in high density) as a function of diversity, density and fertility in the absence and presence of herbivores.

Table 1.   Plant population biomass and survival in all experimental pots in response to diversity, fertilization and density (fixed effects) when herbivory treatments were excluded or included. Block was included as a random factor in all analyses. Subscripts under F ratios indicate numerator and denominator degrees of freedom for each term in the mixed-model analysis
SourceHerbivory excludedHerbivory included
Fertilization71.897< 0.0013.290.07083.623< 0.001
Density15.788< 0.00139.78< 0.00115.464< 0.001
Herbivory14.61< 0.00125.973< 0.001
Diversity × Fertilization1.1900.2760.810.3700.0190.890
Diversity × Density0.4320.5122.100.1480.3900.532
Diversity × Herbivory0.110.7391.7460.187
Fertilization × Density1.4320.2331.260.2610.2530.615
Fertilization × Herbivory0.360.5493.9750.047
Density × Herbivory0.170.6841.6570.199
Diversity × Fertilization × Density0.1110.7400.120.7250.1460.703
Diversity × Fertilization × Herbivory0.160.6851.6760.196
Diversity × Density × Herbivory0.380.5372.2140.137
Fertilization × Density × Herbivory0.900.3444.3050.039
Diversity × Fertilization × Density × Herbivory0.580.4470.6700.413

Plant performance in the presence of T. ni

To assess how insect damage to plants differed among diversity, density and fertilization treatments, herbivory was added as a fixed factor to the analysis.

Plant survival

The proportion of plants surviving was significantly higher in genetic mixtures than in monocultures, and this diversity effect was consistent across fertilization, density and herbivory treatments as evidenced by the lack of any significant two- or three-way interactions (Table 1). Plant survival was not significantly different among fertilization treatments, but fewer plants survived in high compared with low density treatments and when plants were exposed to insect herbivores (Table 1).

Plant biomass

Herbivory significantly reduced plant biomass in all treatment combinations (Table 1): experiment-wide reduction in plant biomass in herbivory treatments compared with no herbivory treatments was 22%. While the addition of herbivores did not negate the positive influence of high diversity on plant biomass (Table 1), it did decrease the size of the effect: in the absence of herbivores, plant biomass was 17% higher in mixtures than in monocultures, but only 5% higher in their presence (Fig. 1). Total pot biomass was significantly higher in high than in low density treatments (16%), and in fertilized than in unfertilized treatments (50%; Table 1; Fig. 1c,d). However, plant biomass was also dependent upon a three-way interaction between herbivore presence, fertilization and density. Plant biomass reduction by herbivores was higher in fertilized treatments (33%) compared with unfertilized (12%) treatments at high planting density, whereas at low density, this effect was reversed: herbivore suppression of plants was greater at low (25%) than at high fertility (15%).

Herbivore performance

Herbivore survival

Insect survival was higher in the plant genetic mixtures compared with genetic monocultures and also higher when plants were fertilized (Table 2). A greater number of insects survived at high rather than low density, although this effect was only marginally significant. There were no significant two- or three-way interactions (Table 2).

Table 2.   Herbivore survival, total biomass (all individuals in a pot) and estimated individual biomass in response to diversity, fertilization and density (fixed effects). Block was included as a random factor in all analyses. Subscripts under F ratios indicate numerator and denominator degrees of freedom for each term in the mixed-model analysis
SourceSurvivalTotal insect biomassEstimated individual insect biomass
Diversity10.9120.00115.073< 0.0015.7530.017
Fertilization8.6630.00435.840< 0.00133.694< 0.001
Diversity × Fertilization0.5750.4492.4100.1222.2840.132
Diversity × Density0.3830.5371.6700.1982.5520.112
Fertilization × Density0.2090.6480.0290.8660.0320.859
Diversity × Fertilization × Density0.6130.4340.0010.9730.3170.574

Herbivore biomass

Across all density and fertilization treatments, total herbivore biomass was 19% higher when insects were reared on genetically diverse host populations than when reared on genetic monocultures (Table 2). Overall, total insect biomass increased by 31% with fertilization and by 11% when grown at high density (Table 2).

Estimated individual insect biomass was 7% higher in mixtures than in monocultures and 18% higher in fertilized than unfertilized treatments (Table 2). There was no significant effect of density, nor any significant interaction terms (Table 2).

Effect of genotype on herbivore performance

To examine whether herbivore performance varied with plant genotype, we included genotype (monocultures only) as a fixed factor in the statistical analysis. Total herbivore biomass varied significantly among the nine A. thaliana lines. The response to each genotype depended on fertilization (Fig. 3), although this genotype × fertilization interaction was only marginally significant (Table 3). In some genetic monocultures (i.e. line 4, Fig. 3), fertilization actually decreased insect biomass, despite the overall positive effect of fertilization on biomass in all treatment combinations (Table 3). The effects of density and all interactions remained non-significant.

Figure 3.

 Plant and insect biomass in monoculture replicates of each A. thaliana genotype at low and high density, and at each level of fertilization. (a, b) Above-ground plant biomass (mean ± 1 SE) in monocultures in the absence of herbivores. (c, d) Summed insect fresh biomass (mean ± 1 SE) in monocultures. For reference, horizontal lines represent mean insect biomass in monoculture when fertilized (solid) or unfertilized (dashed). Bracketed numbers are mean number of surviving insects per pot (pooled fertilization).

Table 3.   Total insect biomass in monoculture in response to genotype, fertilization and density treatments (fixed effects). Block was included as a random factor in all analyses. Subscripts under F ratios indicate numerator and denominator degrees of freedom for each term in the mixed-model analysis
SourceTotal insect biomass
Fertilization7.4131,113< 0.001
Genotype × Fertilization2.2558,1130.068
Genotype × Density0.7418,1130.996
Fertilization × Density0.0271,1130.942
Genotype × Fertilization × Density1.0248,1130.271


Primary productivity

Genotypic richness of A. thaliana strongly influenced plant survival and biomass both in the presence and absence of insect herbivores. Plant biomass was 17% higher in genetic mixtures than in monocultures in insect-exclusion treatments, 5% higher in mixtures in insect-inclusion treatments and 11% higher in mixtures when all treatments were analysed together.

Our results are comparable with those of Crutsinger et al. (2006), who found a 36% increase in productivity in 12-genotype mixtures of Solidago altissima in the field. This is higher than the effect size seen here and raises the possibility that productivity may increase with richness in a linear fashion. Similarly, Johnson, Lajeunesse & Agrawal (2006) found a 27% increase in fitness (fruit production) of Oenothera biennis in mixtures containing up to eight genotypes compared with monocultures. Because seed production is highly correlated with above-ground biomass of A. thaliana (Cahill, Kembel & Gustafson 2005), it is possible that growth at high genotypic diversity may ultimately correspond to increased fitness of this species. It must be noted that our results are based on a semi-natural, mesocosm experiment and thus are not entirely comparable to results obtained in the field by Crutsinger et al. (2006) and Johnson, Lajeunesse & Agrawal (2006). However, the demonstration of positive genotypic diversity–productivity relationships in both natural and artificial settings suggests that these findings are biologically real, and may have important consequences for community structure and function in natural ecosystems.

Unlike recent experiments manipulating plant genotypic diversity only (Hughes & Stachowicz 2004; Crutsinger et al. 2006; Johnson, Lajeunesse & Agrawal 2006), the study described here is novel in that it manipulates insect herbivore presence or absence at two levels of diversity. By controlling for the influence of herbivores, we were able to demonstrate that the positive effect of genotypic diversity on plant performance is through plant-level interactions, and not necessarily mediated by insects (i.e. through selective herbivory). This is in contrast to the results of Hughes & Stachowicz (2004), who detected differences between low- and high-diversity eelgrass communities only after a grazing disturbance. A trade-off between fast growth rates, as in A. thaliana, an annual, weedy species, and resistance to disturbance, as in eelgrass (Hughes & Stachowicz 2004), may explain this discrepancy. We emphasize the need for concurrent manipulations of genotypic diversity and arthropods in natural settings to elucidate the role of multiple trophic groups in structuring diversity–productivity relationships.

Although the design of this experiment did not allow us to explicitly test the mechanisms responsible for higher plant biomass and survival at high genotypic diversity, several possibilities must be considered. First, mixtures may contain, by chance, one or more highly productive genotypes which drive total productivity. This ‘sampling’ effect (Huston 1997), however, is not entirely applicable to this experiment: all nine A. thaliana genotypes were present in equal proportions (1/9), thus all genotypes present in monocultures were equally and consistently represented in mixtures.

Rather, our results may be driven by non-additive mechanisms, which occur when productivity in mixture cannot be predicted from each genotype’s performance in monoculture, as in this study. For instance, complementarity (i.e. niche partitioning or facilitation) has been frequently demonstrated at the species level (Tilman et al. 2001; Hector et al. 2002; Cardinale et al. 2007), and may also play a role in diversity–productivity relationships within populations. While it may be argued that the functional differences among genotypes are smaller than those among species, intraspecific variation is nevertheless large enough to alter ecological processes (Hughes et al. 2008). Given the global distribution of the genotypes employed in this experiment, and the large among-genotype variability in important functional traits such as competition (J. F. Cahill, unpubl. data), it is not unreasonable to suppose that niche partitioning or facilitation led to the high performance of genetic mixtures. Variation in resource uptake strategies may lead to reduced intraspecific competition and greater overall resource capture in genetic mixtures. Decreased ammonium concentrations in sediments in diverse eelgrass communities (Hughes & Stachowicz 2004) suggest that niche partitioning may occur at the genotypic scale in a similar manner as at the species scale.

One alternative, although not mutually exclusive, explanation is that of a ‘selection’ effect (Loreau & Hector 2001; Hector et al. 2002), which can occur if species or genotypes with particular functional traits (i.e. high productivity) come to dominate a mixture over time. The duration of this experiment (one generation) did not allow any genotype to become numerous or dominant via enhanced reproductive success, thus the above definition may not apply. If, however, mortality rates within a genotype are a function of diversity, selection effects may occur through greater survival of highly productive genotypes in mixture than in monoculture.

Recent work also sheds light on the importance of genetic identity of neighbours to plant performance in diverse species (Fridley, Grime & Bilton 2007) and genotypic assemblages (Cahill, Kembel & Gustafson 2005). Such ‘neighbour effects’ may occur, for example, if a genotype with low fitness in monoculture expresses higher growth or lower mortality in the presence of particular neighbour genotypes in mixture. Similarly, competitive asymmetry (Weiner 1990) between genotypes may lead to performance in mixture (i.e. high productivity) which cannot be predicted solely from performance in monoculture. The identification of phenotypic plasticity and trait differences associated with diversity effects has been previously omitted from many biodiversity studies and constitutes a promising direction for future research (Hughes et al. 2008).

Non-additive and additive mechanisms, outlined above, are not mutually exclusive and may work in combination to drive biodiversity–productivity relationships. However, a detailed exploration of the relative contribution of each of these mechanisms to our results, although warranted, is beyond the scope of this article.

Herbivore performance

Herbivore survival and biomass were both significantly higher in mixtures than in monocultures of A. thaliana (Fig. 2). As with primary productivity, the potential mechanisms responsible for this pattern fall into two broad categories. First, herbivore responses to varying plant diversity may be additive, where observed biomass and survival of insects in genotypic mixtures are wholly predictable from performance in monoculture. Under this scenario, high insect biomass in genotypic mixtures is simply a result of greater plant quantity (see Primary productivity section, above), and we would expect to see a tight correlation between herbivore and plant biomass. Several lines of evidence indicate, however, that insect performance was not entirely dependent on primary productivity. Comparison of plant biomass in herbivore-exclusion and herbivore-addition treatments (roughly approximating plant biomass ‘before’ and ‘after’ herbivory) reveals that insects were unlikely to be limited by available biomass: the reduction of plant biomass due to insect presence was only 18% in monocultures and 26% in mixtures. We also compared herbivore biomass with plant biomass in each treatment. Herbivore biomass reflected general patterns in primary productivity (Fig. 1a,b; Fig. 2), but it did not match these patterns perfectly, nor did it closely follow the productivity of each genotype separately (Fig. 3). Linear regression analysis (not shown) indicated that plant biomass measured in insect-exclusion treatments explained only 22% of the variation in total herbivore biomass (slope = 0.13919, SE = 0.043, F1,37 = 10.663, = 0.002, r2 = 0.224). Collectively, these results suggest that plant growth was an important, but not absolute, predictor of herbivore biomass.

Figure 2.

 Mean fresh biomass (± 1 SE) of all insects in an experimental pot in each diversity, density and fertility treatment combination.

Alternatively, non-additive mechanisms, such as greater food quality, may account for the disparities in insect performance observed between genotypic monocultures and mixtures. Trichoplusia ni is considered a generalist species, known to feed on a multitude of agricultural crops and weeds (Cameron, Isman & Upadhyaya 2007), and as such may benefit from a mixture of food items compared with only one or a few plant species. A mixture of food items differing in nutritional quality and toxin content may act in a complementary fashion to improve herbivore growth rate and fitness, as has been demonstrated in laboratory feeding experiments performed on daphnids (DeMott 1998) and polyphagous insects (Bernays et al. 1994; Mody, Unsicker & Linsenmair 2007). Pfisterer, Diemer & Schmid (2003) found a strong, positive relationship between plant species richness and grasshopper biomass in a natural grassland ecosystem, indicating that the benefits of dietary mixing are not limited to artificial settings.

Although evidence for the benefits of dietary mixing has largely been restricted to studies examining plant species richness, similar principles may apply to genotypic richness. Host plant genotypes have been shown to differentially affect performance and fecundity of aphids on Solidago clones (Moran 1981; Maddox & Cappuccino 1986), aphids on Rudbeckia laciniata (Service 1984) and stem gallers on Solidago altissima (Cronin & Abrahamson 2001). Direct support for intraspecific dietary mixing was recently demonstrated by Mody, Unsicker & Linsenmair (2007), who found that the differences in food quality among host plant conspecifics were large enough to render mixed diets beneficial to lasiocampid caterpillars. In this study, genotype had a significant effect on herbivore biomass, suggesting that A. thaliana genotypes varied considerably in nutritional quality or palatability (Fig. 3). A reasonable explanation for the results of our study is that in mixture, genotypes acted in a complementary fashion to increase herbivore survival and biomass.

Interestingly, the relative palatability of each line changed as a result of fertilization, as evidenced by a marginally significant genotype × fertilization interaction in monocultures (Table 3). Some genotypes which were highly palatable relative to other genotypes at low fertilization were only somewhat palatable at high fertilization, and vice versa. In some cases (i.e. lines 2 and 4, Fig. 3d), fertilization actually had a negative effect on herbivore biomass, despite the strong, positive effect of fertilization in the study overall. The mechanism behind this interaction is unclear, and, unfortunately, impossible to determine from the data available.

Herbivores in natural systems have shown a variety of responses to manipulations of plant species diversity (Knops et al. 1999; Koricheva et al. 2000; Scherber et al. 2006a, 2006b, Unsicker et al. 2006). Studies examining plant intraspecific richness, however, have only recently been undertaken and field experiments have thus far shown either an increase (Crutsinger et al. 2008) or no change (Johnson, Lajeunesse & Agrawal 2006) of insect herbivore abundance. Our study is novel in that it demonstrates direct, positive effects of genotypic diversity on herbivore performance in semi-natural conditions, at multiple levels of resource availability and plant density. Further work is needed to determine whether this relationship is found for other insect species, and whether non-additive mechanisms, such as complementarity (i.e. dietary mixing), can account for these results.

Levels of herbivory

Although insect biomass was consistently higher in mixtures than in monocultures, levels of herbivory (as indicated by plant performance in the presence of T. ni) did not significantly differ between the two treatments. Similarly, despite greater insect biomass and survival at high than at low plant densities, levels of herbivory did not differ between these two treatments. Instead, herbivory was dependent upon an interaction between density and fertilization (Table 1): at high density, herbivory was higher in fertilized treatments, whereas at low density, herbivory was greater at low fertility (Fig. 1). Higher survival of insects in fertilized and high density plots (Table 2) may account for these results. The interaction lends some support to the idea that changes at the producer level can ‘cascade’ back and affect herbivory, although it is apparent that changes in plant and herbivore performance brought about by high diversity are not strong enough to produce similar patterns.


By conducting this experiment in semi-natural, controlled conditions, we were able to directly demonstrate higher plant productivity and survival at high genotypic diversity. The consistency of this result in both the presence and absence of insect herbivores, and at multiple density and fertility levels, suggests that this relationship is not necessarily mediated by, or dependent upon, environmental factors such as fertility or grazing disturbance. However, we did not attempt to elucidate how richness might affect the population dynamics of either producers or herbivores in natural populations and in field conditions. Nevertheless, the strength of our findings suggests that the effects seen here are biologically real and may have important consequences for biodiversity–ecosystem function relationships in both natural and agricultural systems. The mechanisms responsible for the effects seen here are not readily determined from our data, and warrant further investigation. Future studies may benefit from measurements of phenotypic traits and their plasticity in response to varying genotypic diversity to fully account for patterns seen in diversity–productivity relationships.


We thank all members of the Cahill laboratory for providing comments throughout the development of this project, and two anonymous referees for their comments and suggestions. Grayson Alabiso-Cahill assisted with the collection of data, and funding was provided by an NSERC Discovery Grant to J.F.C. and an undergraduate scholarship to A.M.K.