Journal of

The effects of plant diversity and insect herbivory on

Blackwell Publishing Ltd

Ecology 2006
94, 922–931 performance of individual plant species in experimental
grassland
CHRISTOPH SCHERBER, ALEXANDRU MILCU *†, STEPHAN PARTSCH *,
STEFAN SCHEU* and WOLFGANG W. WEISSER
Institute of Ecology, University of Jena, Dornburger Str. 159, D-07743 Jena, Germany, *Institute of Zoology,
Technische Universität Darmstadt, Schnittspahnstr. 3, D-64287 Darmstadt, Germany, and †NERC Centre for
Population Biology, Division of Biology, Imperial College, Silwood Park, Ascot SL5 7PY, UK

Summary
1 There is increasing evidence that components of biodiversity affect processes at the
ecosystem level; yet, the effects of biodiversity on the performance of individual organ-
isms or particular trophic interactions are largely unexplored.
2 We transplanted 10 individuals of Rumex acetosa into 82 experimental grassland
plots differing in plant species and functional group richness. Half of the plants received
an insecticide treatment to manipulate insect herbivory.
3 We measured the amount of herbivory, plant size, survival and reproductive param-
eters in 2003 and 2004.
4 Insect herbivores removed on average 4.3% (2003) and 5.1% (2004) of leaf area in
unsprayed plants. Spraying significantly reduced damage levels on average by approx-
imately 50%. Herbivory significantly decreased plant weight, leaf size and number, and
inflorescence length and size.
5 Plant height and inflorescence size of R. acetosa significantly decreased with an
increase in species diversity. Mortality was slightly higher in the species-poor mixtures.
Plant functional group diversity had little effect on plant performance. The presence of
legumes generally increased, while the presence of grasses generally decreased, mor-
phological parameters and fitness in R. acetosa.
6 Overall, the presence of particular plant functional groups was more important than
functional group or species richness per se, and insect herbivores had additive effects of
the same magnitude as the presence of particular plant functional groups.
7 Insect herbivory and plant functional identity, rather than species richness, deter-
mine the performance of individual plant species in temperate grasslands.
Key-words: biodiversity, ecosystem functioning, mixed effects models, multitrophic
interactions
Journal of Ecology (2006) 94, 922–931
doi: 10.1111/j.1365-2745.2006.01144.x

artificially assembled plant communities, there is now

Introduction
a wealth of studies showing that decreasing biodiver-
Changes in the biodiversity of an ecosystem can affect sity can have negative effects on processes measured
ecosystem properties. If components of biodiversity at the ecosystem level (Hooper et al. 2005; Spehn et al.
are lost, the overall state and functioning of the system 2005). In contrast, the effects of biodiversity on the per-
can be impaired (Loreau et al. 2001; Tilman et al. 2001). formance of individual organisms or particular trophic
For a number of experimental ecosystems, for example interactions are largely unexplored. The study presented
© 2006 The Authors
here tries to fill this gap, focusing on the performance
Journal compilation of one particular plant species in communities differing
© 2006 British Correspondence: Christoph Scherber (tel. + 49 3641949406; in plant species and functional richness, combined with
Ecological Society fax + 49 3641 949402; e-mail christoph.scherber@web.de). an experimental manipulation of insect herbivory.
923 While previous studies have mainly concentrated on Roscher et al. 2004 for details), and containing all
Herbivory and the relationship between plant diversity and diversity possible combinations of (number of plant species) ×
plant diversity of herbivorous insects (Siemann 1998; Haddad et al. (number of functional groups). Due to non-orthogonality
effects on 2001), there is only limited knowledge on the extent in the design, the correlation coefficients for log-species
Rumex acetosa of herbivory and its feedback on the performance of richness and number of functional groups, legumes and
individual plant species within a diversity gradient. grasses were 0.63, 0.35 and 0.38, respectively. Plots are
Even those studies that report significant relationships continuously weeded to maintain the target communities.
between plant species richness and insect herbivory
have come to opposing conclusions: while Root’s classic
 
study on herbivore load in Brassica oleracea L. (Root
1973) has resulted in the formulation of a ‘resource con- Rumex acetosa is a dioecious perennial with a rosette
centration’ hypothesis, more recent results obtained by growth habit, racemose flowers and hastate leaves that
Otway et al. (2005) have been interpreted as showing have characteristically extended basal lobes and long
‘resource dilution’. This, in turn, would mean that ‘host petioles (Clapham et al. 1987). Several monophagous
plants in high diversity mixtures’ should experience beetles of the families Apionidae and Curculionidae
‘greater herbivore pressure’ (Otway et al. 2005). have been described for R. acetosa (Böhme 2001). There
In this study, we use experimental grassland com- is extensive literature on the effects of selective herbivory,
munities (described in detail in Roscher et al. 2004) plant competition and fungal infection on other (mostly
differing in the number of plant species (1–60) and the weedy) Rumex species (e.g. Hatcher et al. 1994; Keary
number and identity of functional groups (one to four & Hatcher 2004). In contrast, R. acetosa has been studied
functional groups: grasses, legumes, small herbs, tall in much less detail so far.
herbs). We introduce a given plant species into the ex-
perimental mixtures, at a density of five individuals
  R . A C E T O S A
per subplot (‘phytometer’ approach sensu Gibson 2002).
    
Phytometer performance measures both direct and
indirect effects of the resident plant and insect herbivore Rumex acetosa seeds were obtained from Rieger-
communities. Disentangling these interacting processes Hofmann GmbH, Blaufelden-Raboldshausen, Germany.
is the major aim of the experiments reported here. In early April 2003, R. acetosa seeds were pre-germinated
Rumex acetosa L., a north-temperate member of the on standard compost, and grown on a standard compost:
Polygonaceae, was selected as a phytometer species, perlite mixture (4:1) in a glasshouse with night:day
because: (i) its insect herbivore fauna is well-known; (ii) cycles of 10:14 hours (15:22 °C). Prior to final transplan-
leaf morphology allows quick and easy determinations tation, all plants were sorted into three size classes and
of foliar herbivory; and (iii) it is present in the species pool hardened for 7 days. In mid-June 2003, n = 820 plants
of the experimental mixtures. We address three main ques- were randomly selected and transplanted into the diver-
tions. (i) How do plant size, survival and reproduction sity plots. Each main plot was divided into two subplots
of R. acetosa change with plant species richness, number (split-plot design), whose positions were randomized
of functional groups, and presence of particular func- along the north-south axis of each main plot. Each of
tional groups? (ii) What is the effect of insect herbivory these two subplots received one row of plants, containing
on plant size, survival and reproduction of R. acetosa? one large, two medium and two small individuals, 50 cm
(iii) What is the relationship between insect herbivory apart. While one subplot (2 × 4 m) served as a control,
and plant diversity in R. acetosa? We hypothesize that: another subplot was used for insect exclusion (‘insecticide’)
(i) plant functional identity will be more important for and covered a larger area (5 × 5 m) to allow for efficient
performance of R. acetosa than species richness per se insect herbivore exclusion. The initial sizes (as deter-
(Scherber et al. 2006); and (ii) insect herbivory will mined by leaf number) of all plants were recorded and
increase with plant species richness (Otway et al. 2005). included as covariates into all analyses (Crawley 2002).

Materials and methods  

The ‘insecticide’ subplots on all 82 large plots were
    
sprayed with an aqueous solution (30 mL m–2) of an

above-ground, semi-systemic organothiophosphate insec-
Plant communities of increasing species richness (1, 2, ticide (Dimethoate, C5H12NO3PS2, BASF, Ludwigs-
4, 8, 16, 60 species) were established on former arable land hafen, Germany) at 4-weekly intervals between April and
near Jena (Germany) in 2002 from a pool of 60 grassland August 2003 and 2004. Dimethoate has been shown to be
© 2006 The Authors
plant species. Plant species were divided a priori into both effective in reducing insect herbivory and having
Journal compilation
© 2006 British
four functional groups (grasses, legumes, small and tall little direct effects on plants (Hector et al. 2004; Schädler
Ecological Society, herbs; Roscher et al. 2004). The overall design is a ran- et al. 2004). Subplots were sprayed using a backpack
Journal of Ecology, domized complete blocks design with 82 plots each 20 × sprayer (Birchmeier Senior 20 L) at 6 × 105 Pa operating
94, 922–931 20 m in size, systematically divided into four blocks (see pressure. As every R. acetosa individual received only
924 about 2 mL of solution per month, we decided to leave the Percentage herbivory was calculated as damaged area
C. Scherber et al. plants on the ‘control’ subplots unsprayed (as, for exam- divided by total area, multiplied by 100. Community
ple, in Keary & Hatcher 2004). To assess the effectiveness biomass was harvested between 27 May and 10 June
of insecticide treatments, we repeatedly quantified insect 2004 in two randomly placed 20 × 50 cm quadrates per
herbivory (as suggested by Siemann et al. 2004). subplot at 3 cm above ground, oven-dried at 70 °C for
48 hours and weighed. In addition, we measured relative
cover of every plant species for all mixtures, using an inte-
   2003
ger cover degree scale with two independent observers.
Initial size (number of fully unfolded leaf laminas) of
R. acetosa plants was determined 2 weeks after trans-
 
planting on 9 July 2003. All other measurements were
performed between 7 and 18 August 2003. For each R. We used linear mixed-effects analysis of covariance
acetosa plant, we measured: (i) L, the absolute leaf area models (Pinheiro & Bates 2000; Crawley 2002) imple-
damaged by insect herbivores per plant (in mm2); (ii) mented in S-Plus 6.1.2 Professional for Windows
SS, SL, the sizes of the smallest and largest leaves (in cm2); (Copyright 2002, Insightful Corp., Seattle, USA) for all
(iii) T, the total number of leaves; and (iv) plant dry weight analyses. Proportion and mortality data were arcsine-
(in g). The sizes of the smallest and largest leaves were square root transformed and count data were square-
estimated by comparing leaves with a set of standard root or log-transformed to account for heteroscedasticity
ellipsoid paper templates of known area. Absolute leaf and non-normality of errors. Standard errors for geo-
area damaged by insect herbivores was estimated using metric means and for means calculated from square-root
a 1-mm2 grid. Leaves clipped by vertebrate herbivores or transformed data were derived using bootstrap
severely damaged by molluscs were excluded. Percentage resampling with 1000 replications. Because of intrinsic
herbivory, p, was calculated, after adjustment of units, as aliasing in the design (sensu McCullagh & Nelder
1989), construction of the full model follows two main
2L
p= × 100 [%] eqn 1 principles: (i) we fit covariates first; and (ii) the sequence
T(SS + SL )
of terms directly relates to the scientific hypotheses of
Plants were harvested 3 cm above soil surface between interest. As insecticide treatments were applied to subplots
1 and 3 September 2003, oven-dried at 70 °C for 48 hours within plots, models contained random effects at two
and weighed. Community leaf area index (LAI) per sub- levels. Blocks were entered as a fixed rather than random
plot was measured between 12 and 13 August 2004, using effect (a view that is supported, for example, by Piepho
a LAI-2000 Plant Canopy Analyser (Li-Cor BioSciences, et al. 2004), because: (i) functional group treatments
Lincoln, USA). Each LAI measurement consisted of a were unequally represented within blocks; and (ii)
reference value taken above the canopy, and five meas- block positions were not randomized, and blocks cannot
urements of light interception 5 cm above soil surface. be considered random samples from an infinite popula-
Measurements were not adjusted for leaf angles. tion. The fixed-effects structure of the maximal model fit
by maximum likelihood was:
[y ∼ initial number of leaves + block + community
   2004
biomass + insecticide + number of functional groups +
Between 5 and 7 April 2004, we counted the number of log2 (sown number of plant species + 1) + grass presence +
surviving individuals since transplantation. Reproduc- legume presence + insecticide:initial number of leaves
tive parameters were only measured in 2004, as plants + insecticide:block + insecticide:log species richness +
did not flower in 2003. We use morphological parameters number of functional groups:log species richness + grass
of the inflorescences as a measure of potential plant fitness presence:legume presence], where ‘:’ indicates inter-
(Conn & Blum 1981). Morphological measurements in actions (Chambers & Hastie 1992).
R. acetosa were performed between 17 and 18 May 2004. For 2003, we used community leaf area index instead
For each plant individual, we measured: (i) the absolute of community biomass. The significance of terms was
leaf area damaged by insect herbivores on a randomly assessed using conditional F-tests. Variance functions
chosen leaf; (ii) the size of that leaf; (iii) the maximum were used to model heteroscedasticity in the within-group
length of the main inflorescence axis using a metering errors (Pinheiro & Bates 2000). We simplified the maxi-
rule; (iv) the number of first-order branches on the inflo- mal model by sequentially deleting non-significant terms
rescence axis; and (v) the number of main inflorescence (starting with highest-order interactions) and comparing
axes. For herbivory and leaf area measurements, one leaf each model with its predecessor using Akaike informa-
per plant was selected by hypothetically constructing a tion criterion (AIC, Burnham & Anderson 1998) and
cylinder around each plant and dividing it into eight likelihood ratio tests, until minimal adequate models
© 2006 The Authors
cylinder sections, of which we selected one at random were retrieved. For graphical representation of data, we
Journal compilation
© 2006 British
and picked one leaf from it. Total and damaged leaf areas use trellis displays (Becker & Cleveland 1996) of the Lattice
Ecological Society, were measured to the nearest mm2 using clear plastic graphics package (version 0.12–9) in R 2.2.0 (R Devel-
Journal of Ecology, sheets with a mm2 grid. Leaves clipped by vertebrate her- opment Core Team, 2005). Observations are divided into
94, 922–931 bivores or severely damaged by molluscs were excluded. distinct groups according to several grouping factors.
925
Table 1 Summary of linear mixed-effects models for August 2003 data. For each of the six response variables (columns), the details of the minimal adequate
models are listed
Herbivory and in the rows, with explanatory variables (first column) retained in the models, their corresponding F- and P-values, denominator degrees
of freedom,
plant plus additional information (bottom rows). For the purpose of clarity, explanatory variables are ordered (i) by their denominator d.f. and (ii)
diversity
by the sequence in which they were entered into the maximal models. Bold font indicates significant P-values
effects on
Rumex2003
August acetosa Percentage herbivorya Number of leavesb Plant dry weight (g)b
Numerator Denominator
Source d.f. d.f. F-value P-value F-value P-value F-value P-value

Intercept 1 Plot 355.345 <0.0001 3962.650 < 0.0001 319.9612 <0.0001

Block 3 Plot Excluded Excluded 2.600 0.0586 Excluded Excluded
Number of plant species (S) 1 Plot 4.8713 0.0302 Excluded Excluded Excluded Excluded
Grasses 1 Plot 5.6628 0.0197 2.915 0.0920 4.8652 0.0303
Legumes 1 Plot Excluded Excluded 7.093 0.0095 5.8568 0.0178
Grasses:Legumes 1 Plot Excluded Excluded 3.404 0.0691 Excluded Excluded
Initial number of leaves 1 Subplot Excluded Excluded 31.425 <0.0001 22.1559 <0.0001
Insecticide treatment 1 Subplot 8.572 0.0044 21.082 <0.0001 5.1809 0.0259
Community leaf area index 1 Subplot Excluded Excluded 1.645 0.2035 Excluded Excluded
Number of observations 164 160 154
Number of groups (plots) 82 80 81
Number of d.f. used up in minimal model 4 10 5
Denominator d.f. (plot level) 81 77 78
Denominator d.f. (subplot level) 79 73 71
AIC of maximal model − 258.5666 121.0696 116.7885
AIC of minimal model − 290.4181 105.5165 97.03667

Annotations: aarcsine-square root; blog transformed; d.f., degrees of freedom; AIC, Akaike information criterion; ‘excluded’ indicates terms excluded
during model simplification.

but this interaction was not significant. In addition,

Results
herbivory was significantly higher when grasses were
The results of the statistical analyses are summarized in present in the communities, than when grasses were
Tables 1 (2003) and 2 (2004). These tables also list non- absent (3.8 ± 0.6% vs. 2.9 ± 0.5%; Table 1).
significant terms, which were retained in the minimal Average herbivory levels in 2004 were slightly higher
adequate models during model simplification. than in 2003 (5.1 ± 0.5%). With increasing initial number
of leaves (2003), herbivory in 2004 increased significantly
(Table 2). Insecticide-treated plants showed significantly
 
lower herbivore damage (Table 2), while plant species
In 2003, natural levels of herbivory in R. acetosa averaged richness did not have an effect in 2004 (Fig. 1b). Both the
4.3 ± 0.7%. Insecticide treatment significantly reduced block effect, and the interaction between number of func-
herbivory to 2.4 ± 0.3% (Table 1). Herbivory decreased tional groups and plant species richness were signifi-
significantly with increasing plant species richness cant (Table 2). In contrast to 2003, the presence of grasses
(Table 1). Figure 1(a) shows that insecticide treatment or legumes did not have a significant effect on herbivory.
partly decoupled the diversity-herbivory relationship, To test whether R. acetosa density in the surrounding

© 2006 The Authors

Journal compilation
© 2006 British Fig. 1 Relationship between number of plant species and leaf area eaten (%) in R. acetosa in (a) 2003 and (b) 2004. Open circles
Ecological Society, and solid lines, control plants; open triangles and dashed lines, insecticide-treated plants. Lines show mean squares fits. r2 values
Journal of Ecology, for 2003 are 0.017 (insecticide) and 0.024 (control); r2 values for 2004 are 0.001 (insecticide) and 0.008 (control). Larger symbols
94, 922–931 show means.
94, 922–931
Journal of Ecology,
Ecological Society,
© 2006 British
Journal compilation
© 2006 The Authors

C. Scherber et al.
926
Table 2 Summary of linear mixed-effects models for May 2004 data. For each of the six response variables (columns), the details of the minimal adequate models are listed in the rows, with explanatory variables
(first column) retained in the models, their corresponding F- and P-values, denominator degrees of freedom, plus additional information (bottom rows). For the purpose of clarity, explanatory variables are ordered
(i) by their denominator d.f. and (ii) by the sequence in which they were entered into the maximal models. Bold font indicates significant P-values

May 2004 Percentage mortalitya Percentage herbivorya Plant height (cm) Number of branchesb Number of main axesc
Numerator Denominator
Source d.f. d.f. F-value P-value F-value P-value F-value P-value F-value P-value F-value P-value

Intercept 1 Plot 119.2787 <0.0001 523.2211 <0.0001 719.8119 <0.0001 630.9144 <0.0001 1201.156 <0.0001
Block 3 Plot Excluded Excluded 3.1178 0.0310 Excluded Excluded Excluded Excluded Excluded Excluded
Number of FG 1 Plot Excluded Excluded 0.8239 0.3670 Excluded Excluded 0.9978 0.3211 5.420 0.0226
Species richness (S) 1 Plot 6.4580 0.0130 1.5800 0.2127 2.7288 0.1027 9.3103 0.0032 8.281 0.0052
Grasses 1 Plot Excluded Excluded Excluded Excluded 14.4954 0.0003 38.5587 <0.0001 32.208 <0.0001
Legumes 1 Plot 4.3943 0.0393 Excluded Excluded 25.2185 <0.0001 22.6474 <0.0001 23.341 <0.0001
Number of FG:S 1 Plot Excluded Excluded 4.8256 0.0311 Excluded Excluded Excluded Excluded Excluded Excluded
Initial number of leaves 1 Subplot 5.9899 0.0166 4.6302 0.0346 7.2747 0.0087 14.9495 0.0002 24.673 <0.0001
Biomass (gm−2) 1 Subplot Excluded Excluded Excluded n.s. 0.8387 0.3628 0.2253 0.6365 1.915 0.1708
Insecticide treatment 1 Subplot Excluded Excluded 9.0667 0.0035 17.5489 0.0001 21.7716 <0.0001 19.080 <0.0001
Number of observations 164 159 NA NS 154
Number of groups (plots) 82 82 80 80 80
Number of d.f. used up in minimal model 5 9 7 8 8
Denominator d.f. (plot level) 80 75 76 75 75
Denominator d.f. (subplot level) 79 75 72 71 71
AIC of maximal model −0.0167 −234.6253 1417.33 414.926 148.9746
AIC of minimal model −4.9059 −274.8168 1391.193 395.9124 128.9068

Annotations: aarcsine-square root; blog; csquare-root transformed; FG, functional groups; d.f., degrees of freedom; AIC, Akaike information criterion; ‘excluded’ indicates terms excluded during model
simplification.
927
Herbivory and
plant diversity
effects on
Rumex acetosa

Fig. 3 Effects of plant species richness and legume presence on

mortality in R. acetosa. Boxes represent lower and upper
quartiles; black dots indicate the median. Whiskers indicate
observations within 1.5 times the interquartile range from the top
(bottom) of the boxes. Open circles show outliers.

Plants had significantly fewer leaves in the presence

rather than in the absence of grass species (6.1 ± 0.4 vs.
4.6 ± 0.2, Table 1), while the presence of legumes always
increased leaf number (6 ± 0.4 vs. 4.5 ± 0.2, Table 1).

  

In 2003, plant weight was on average 0.95 ± 0.09 g, and
increased significantly with increasing initial number
of leaves (Table 1). Insecticide-treated plants had a sig-
nificantly higher dry weight than control plants (0.98 ±
0.12 g vs. 0.93 ± 0.14 g, Table 1). When legumes were
present in the communities, plants weighed significantly
more than in the absence of legumes; grasses had the
opposite effect (Table 1, Fig. 2a).
Fig. 2 Effects of particular functional groups on (a) plant dry
weight, (b) maximum inflorescence length, and (c) number of  
first-order inflorescence branches in R. acetosa. Solid (dashed)
lines indicate plots without (with) grasses. Lines connect Approximately 10 months after transplantation, there
means ± 1 SE. were 190 dead individuals out of the 820 original trans-
plants (= 23.2%). Mortality across plots followed a
community had an effect on herbivory in the trans- negative exponential distribution, with in the majority
planted R. acetosa individuals, we plotted phytometer of cases (n = 52) 0–1 dead plants per plot. There was a
herbivory against the relative cover (range: 0–20%) of significantly higher survival in plants with a higher
R. acetosa. Herbivory was fully independent of R. ace- initial leaf number in comparison with plants with a
tosa cover (linear regression; intercept 4.2 ± 0.36%, slope small initial size (Table 2). Plant mortality decreased
– 0.05 ± 0.10, overall P = 0.63, r2 = 0.001). significantly and linearly with increasing plant species
richness, while legume presence generally increased
mortality (Table 2). Figure 3 gives an overview of the
  
combined effects of legumes and plant species richness.
© 2006 The Authors
In 2003, plants had on average 5.3 ± 0.2 leaves. Initial
Journal compilation
number of leaves, fitted as a covariate, had a highly sig-
© 2006 British  
Ecological Society, nificant effect on leaf number in August 2003 (Table 1).
Journal of Ecology, Plants treated with insecticide had significantly more In 2004, the maximum length of the main inflorescence
94, 922–931 leaves than control plants (5.5 ± 0.3 vs. 5.1 ± 0.3, Table 1). axes was on average 57.0 ± 2.1 cm. Plants with a higher
928
C. Scherber et al.

Fig. 4 Relationships between plant species richness (x-axis) and number of main inflorescence axes ( y-axis: square root scale) in
R. acetosa. Open circles and solid lines show plots without legumes; open triangles and dashed lines show plots with legumes.
Lines show mean squares fits; r2 values are for these mean squares fits only.

initial size produced significantly longer main inflores- axes than control plants (Table 2). The number of main
cence axes than plants with a smaller initial size axes was significantly reduced in communities that
(Table 2). Insecticide treatment also had a significant contained more functional groups, or more plant
positive effect on axis length, leading to an increase species (Table 2), and there was a significant interac-
from 54.9 ± 2.1 cm to 59.0 ± 2.1 cm. Presence of tion between these two terms. Plants growing in com-
grasses significantly decreased plant height, while the munities that contained grasses had significantly fewer
presence of legumes had a significant positive effect main axes, while legume presence generally increased
(Table 2). Figure 2(b) shows an interaction plot with the number of main axes (Table 2). Figure 4 shows the
both legume and grass effects on plant height. combined effects of legumes, grasses, number of plant
species, and insecticide treatment, on the number of
main axes. It can be clearly seen that there are no signif-
  -  
icant interactions, and that legume presence and insec-
 
ticide treatments generally lead to a parallel shift in the
Rumex acetosa inflorescences had on average 17.4 ± 2.4 regression lines.
first-order branches. Plants with a greater initial size
had more branches than those with a smaller initial size
Discussion
(Table 2). Insecticide-treated plants had significantly
more first-order branches (17.5 ± 2.2) than control The results presented in this manuscript clearly dem-
plants (17.4 ± 4.3); see Table 2. We found a significant onstrate that different components of biodiversity are
effect of plant species richness: with increasing number of different importance for the overall performance of
of plant species, the number of first-order branches sig- individual plant species. While species richness per se
nificantly decreased (Table 2). When grasses were present, has only weak effects on herbivory, reproductive traits
the number of branches significantly decreased, while and the survival of R. acetosa, plant functional identity
legume presence generally increased branch number affects herbivory, mortality, and all the morphologi-
(Table 2). These combined effects of legumes and grasses cal traits measured in this study. Insect herbivory, in
are presented in Fig. 2(c). general, acts as an additional factor, independent of
diversity effects. Of course, these findings are specific
to the system we studied, but we nevertheless believe
© 2006 The Authors     
that generalizations are possible and necessary, not
Journal compilation
© 2006 British
Plants had on average 1.8 ± 0.2 main inflorescence axes. least because we think that these results can be seen as
Ecological Society, The initial number of leaves in 2003 had a significant a step towards a more mechanistic understanding of
Journal of Ecology, positive effect on the number of main axes (Table 2). processes acting in grassland ecosystems of differing
94, 922–931 Insecticide-treated plants had significantly more main diversities.
929 Previous experiments have mainly focused on the com- This view is also supported by a study on the biological
Herbivory and munity- and ecosystem-level consequences of biodiversity control of R. obtusifolius L. (Grossrieder & Keary 2004);
plant diversity decline (reviewed in Hooper et al. 2005). For example, these authors have found that leaf beetle grazing can
effects on several studies have focused on the relationship between lead to fewer and lighter seeds, i.e. the potential fitness
Rumex acetosa plant and arthropod diversity in experimental grass- of a Rumex species can be significantly altered by insect
lands (Knops et al. 1999; Koricheva et al. 2000; herbivores.
Haddad et al. 2001). As part of the BIODEPTH bio- Notably, insect herbivore effects were independent
diversity experiment, Mulder et al. (1999) published of plant diversity. Herbivory acted in an additive way,
one of the first studies on the relationship between insect without an indication of either resource concentration
herbivory and plant species diversity, combined with (Root 1973) or resource dilution (Otway et al. 2005)
an insecticide treatment. However, their study focused effects. The negative trend in the 2003 herbivory data
on community herbivory, and the BIODEPTH exper- set is mainly caused by two monoculture outliers. Two
imental design does not allow a full separation of legume notes on testing of the above-mentioned hypotheses need
and grass effects from other components of biodiversity, to be made. First, neither of the two hypotheses has been
because grasses were present in all multispecies mix- formulated in a way that allows a test by experimenta-
tures, and plots with 100% legumes were restricted to the tion; and secondly, our experimental design does not
one- and two-species mixtures. Giller & O’Donovan allow a deliberate manipulation of R. acetosa density.
(2002) demonstrated single-species herbivory (Trifolium Instead, our experiment tests whether diversity per se
pratense L) in relation to plant species richness from a changes insect herbivory at the level of a single plant
study conducted at the Irish BIODEPTH site. Recently, species, independent of whether taxonomically related
Otway et al. (2005) published a detailed analysis of insect individuals co-occur in the communities. Interestingly,
herbivore abundance at the Silwood Park BIODEPTH even when we incorporated R. acetosa cover into our
site, demonstrating that insect herbivore damage (if it analyses, phytometer herbivory proved fully independ-
translated linearly from insect herbivore load) in several ent of host plant density. Thus, we conclude that there
plant species, including R. acetosa, might be negatively is no indication of either diversity or host plant density
related to plant species richness. effects on herbivory in our data sets.
The study presented here tries to extend the work from
previous studies, using a design that enables a separation
   
of the effects of species richness from the effects of plant
functional diversity, and from plant functional identity, Before testing for species richness and other compo-
combined with a manipulation of insect herbivory. nents of plant diversity, community biomass or leaf area
index was fitted as a covariate in our models, because
all the additional variation that could then be attributed
 
to plant diversity was corrected for pure biomass (or
The rates of insect herbivory reported are consistent LAI) effects (sequential fitting of terms, sensu Schmid
with the literature. Carson & Root (1999) found a rate of et al. 2002). For the same reason, plant species richness
0.9 vs. 0.7% leaf area damage in control vs. insecticide- was always fitted following the number of functional
treated plants in a closely related Rumex species. Accord- groups. Our data suggest moderate but significant influ-
ing to Scheidel & Bruelheide (1999), other invertebrate ences of plant species richness on mortality and repro-
herbivores, in particular molluscs, seem to avoid R. ductive traits in R. acetosa. While survival was generally
acetosa in free-choice feeding trials with other grass- greater in species-rich mixtures, the number and size
land plants. We excluded leaves damaged by molluscs, of inflorescences generally decreased. We emphasize,
but these were only encountered very infrequently. It however, that this result should not be interpreted as
is also important to note that we tested for herbivory ‘invasion resistance’ (cf. Diemer & Schmid 2001). Rather,
effects in established plants, as seedlings may be more we conclude that because R. acetosa is a subdominant
severely affected (e.g. Keary & Hatcher 2004). Our component of Arrhenatherum grasslands, it generally
insecticide treatment did not have a significant effect on decreases its resource allocation to reproductive tissues
plant mortality; this finding is not surprising, as insect when growing in mixtures of increasing plant species
herbivores only rarely kill their hosts (Crawley 1983; richness. This interpretation is in accordance with recent
Crawley 1997). The visible damage recorded did trans- analyses from the Cedar Creek biodiversity experiment,
late into morphological differences between sprayed and where Lambers et al. (2004) found several forb species
unsprayed plants. The reduction of insect herbivory, with yield exponents smaller than −1, indicating an
using Dimethoate, increased plant dry weight, leaf overall negative effect of plant species richness on
number, inflorescence length and the number of inflore- individual plant species’ performance. In addition, a
© 2006 The Authors
scence main axes and branches. As plant dry weight work published by van Ruijven & Berendse (2003)
Journal compilation
© 2006 British
and inflorescence parameters are directly correlated with showed similar effects. However, in this study, the overall
Ecological Society, potential fitness (Ainsworth et al. 2005), we deduce that negative effect of increasing plant diversity was small in
Journal of Ecology, potential fitness of a transplant phytometer, R. acetosa, terms of its effect and was outperformed by the effects of
94, 922–931 is significantly negatively affected by insect herbivory. particular plant functional groups (see below), a finding
930 that has recently been reported at the community level WE2618/4–1 and FOR 456 to WWW. CS was supported
C. Scherber et al. (Scherber et al. 2006). by the Studienstiftung des Deutschen Volkes. All experi-
ments conducted comply with current German laws.

     

References
In our opinion, plant functional identity is one of the
Ainsworth, C., Rahman, A., Parker, J. & Edwards, G. (2005)
most important factors influencing morphological Intersex inflorescences of Rumex acetosa demonstrate that
parameters and reproductive traits in R. acetosa phy- sex determination is unique to each flower. New Phytologist,
tometers. Mortality, inflorescence length, number of 165, 711–720.
first-order branches, number of main axes, herbivory, Becker, R.A. & Cleveland, W.S. (1996) The design and control
number of leaves, and plant dry weight, were all highly of trellis display. Journal of Computational and Statistical
Graphics, 5, 123 –155.
significantly modulated by the presence or absence of Böhme, J. (2001) Phytophage Käfer und ihre Wirtspflanzen in
particular functional groups of plants in the commu- Mitteleuropa. Bioform, Heroldsberg, Germany.
nities. While the presence of grasses generally had adverse Burnham, K.P. & Anderson, D.A. (1998) Model Selection
effects on most morphological parameters, presence of and Inference: a Practical Information-Theoretic Approach.
legumes enhanced the plant size of R. acetosa in almost Springer, Berlin.
Carson, W.P. & Root, R.B. (1999) Top-down effects of insect
every case. Many recently published studies present herbivores during early sucession: influence on biomass
similar findings, highlighting the overall importance of and plant dominance. Oecologia, 121, 260 – 272.
plant functional identity for the performance of indi- Chambers, J.M. & Hastie, T.J. (1992) Statistical Models in
vidual species and even whole plant communities (Mulder S. AT & T Bell Laboratories, Wadsworth & Brooks/Cole
et al. 2002; Spehn et al. 2002; Lambers et al. 2004; Advanced Books & Software, Pacific Grove, California.
Clapham, A.R., Tutin, T.G. & Moore, D.M. (1987) Flora
Petchey 2004; Scherber et al. 2006), with mostly nega- of the British Isles, 3rd edn. Cambridge University Press,
tive effects as a result of the presence of grass species, and Cambridge.
positive effects as a result of legume presence, on the Conn, J.S. & Blum, U. (1981) Differentiation between the sexes
performance of individual plant species. of Rumex hastatulus in net energy allocation, flowering and
height. Bulletin of the Torrey Botanical Club, 108, 446–455.
Crawley, M.J. (1983) Herbivory. Blackwell Scientific, Oxford.
Conclusions Crawley, M.J. (1997) Plant-herbivore dynamics. Plant Ecology
(ed. M.J. Crawley), pp. 401– 474. Blackwell Science, Oxford.
The results presented here clearly indicate that size, sur- Crawley, M.J. (2002) Statistical Computing. An Introduction
vival and reproduction of an individual plant species to Data Analysis Using S-Plus. John Wiley & Sons, Chichester.
are significantly influenced by: (i) the functional identity Diemer, M. & Schmid, B. (2001) Effects of biodiversity loss
and disturbance on the survival and performance of two
(rather than diversity) of other plant species in the sur- Ranunculus species with differing clonal architectures.
rounding plant community; and (ii) the degree of insect Ecography, 24, 59 – 67.
herbivory. Increasing plant species and functional diver- Gibson, D.J. (2002) Methods in Comparative Plant Population
sity leads to enhanced survival, but decreased potential Ecology. Oxford University Press, Oxford.
fecundity. Grass presence decreases individual plant Giller, P.S. & O’Donovan, G. (2002) Biodiversity and ecosys-
tem function: do species matter? Biology and Environment:
performance, while legume presence has the opposite Proceedings of the Royal Irish Academy, 102B, 129–139.
effect. Reduction of insect herbivory results in enhanced Grossrieder, M. & Keary, I.P. (2004) The potential for the
growth and potential fecundity of individual plant spe- biological control of Rumex obtusifolius and Rumex crispus
cies. Insect herbivore effects are additive and independent using insects in organic farming, with particular reference to
of plant diversity. Extrapolating from species-specific Switzerland. Biocontrol News and Information, 25, 65N–79N.
Haddad, N.M., Tilman, D., Haarstad, J., Ritchie, M.E. &
results to whole communities with multi-species inter- Knops, J.M.H. (2001) Contrasting effects of plant richness
actions is a major challenge for future studies on insect and composition on insect communities: a field experiment.
herbivory and plant diversity. We hope that these results American Naturalist, 158, 17 – 35.
can help to further elucidate some of the responsible Hatcher, P.E., Paul, N.D., Ayres, P.G. & Whittaker, J.B. (1994)
mechanisms. The effect of an insect herbivore and a rust fungus indi-
vidually, and combined in sequence, on the growth of
two Rumex species. New Phytologist, 128, 71 – 78.
Hector, A., Wilby, A., Latsch, O.G. & Brown, V.K. (2004)
Acknowledgements
Phyto-activity of biocides used to manipulate herbivory:
We thank Mick J. Crawley for his generous encourage- tests of three pesticides on fourteen plant species. Basic and
ment and advice on statistical analysis. We thank Sylvia Applied Ecology, 5, 313 – 320.
Hooper, D.U., Chapin, F.S., Ewel, J.J., Hector, A., Inchausti,
Creutzburg, Jule Specht, Nadine Mitschunas, Peter N. P., Lavorel, S. et al. (2005) Effects of biodiversity on eco-
Mwangi and Sybille B. Unsicker for their enthusiastic system functioning: a consensus of current knowledge.
support at various stages of the experiments. The com- Ecological Monographs, 75, 3 – 35.
© 2006 The Authors Keary, I.P. & Hatcher, P.E. (2004) Combining competition
ments of two anonymous referees, Lindsay Haddon,
Journal compilation from Lolium perenne and an insect-fungus combination to
© 2006 British
Meran Owen, Jens Schumacher, Bernhard Schmid and
control Rumex obtusifolius seedlings. Weed Research, 44,
Ecological Society, Christiane Roscher on earlier versions of this manu- 33 – 41.
Journal of Ecology, script have been very helpful. This study was supported Knops, J.M.H., Tilman, D., Haddad, N.M., Naeem, S.,
94, 922–931 by the Deutsche Forschungsgemeinschaft through grants Mitchell, C.E., Haarstad, J. et al. (1999) Effects of plant species
931 richness on invasion dynamics, disease outbreaks, insect approach in a grassland community. Basic and Applied
Herbivory and abundances and diversity. Ecology Letters, 2, 286 –293. Ecology, 5, 107 –121.
Koricheva, J., Mulder, C.P.H., Schmid, B., Joshi, J. & Huss- van Ruijven, J. & Berendse, F. (2003) Positive effects of plant
plant diversity
Danell, K. (2000) Numerical responses of different trophic species diversity on productivity in the absence of legumes.
effects on groups of invertebrates to manipulations of plant diversity Ecology Letters, 6, 170 –175.
Rumex acetosa in grassland. Oecologia, 125, 271– 282. Schädler, M., Jung, G., Brandl, R. & Auge, H. (2004) Secondary
Lambers, J.H.R., Harpole, W.S., Tilman, D., Knops, J. & succession is influenced by belowground insect herbivory
Reich, P.B. (2004) Mechanisms responsible for the positive on a productive site. Oecologia, 138, 242 – 252.
diversity-productivity relationship in Minnesota grass- Scheidel, U. & Bruelheide, H. (1999) Selective slug graz-
lands. Ecology Letters, 7, 661– 668. ing on montane meadow plants. Journal of Ecology, 87,
Loreau, M., Naeem, S., Inchausti, P., Bengtsson, J., Grime, 828 –838.
J.P., Hector, A. et al. (2001) Biodiversity and ecosystem Scherber, C., Mwangi, P.N., Temperton, V.M., Roscher, C.,
functioning: current knowledge and future challenges. Schumacher, J., Schmid, B. et al. (2006) Effects of plant
Science, 294, 804 – 808. diversity on invertebrate herbivory in experimental grass-
McCullagh, P. & Nelder, J.A. (1989) Generalized Linear land. Oecologia, 147, 489 – 500.
Models, 2nd edn. Chapman & Hall, London. Schmid, B., Hector, A., Huston, M.A., Inchausti, P., Nijs,
Mulder, C.P.H., Jumpponen, A., Hogberg, P. & Huss-Danell, I., Leadley, P.W. et al. (2002) The design and analysis of
K. (2002) How plant diversity and legumes affect nitrogen biodiversity experiments. Biodiversity and Ecosystem
dynamics in experimental grassland communities. Oecolo- Functioning: Synthesis and Perspectives (eds M. Loreau,
gia, 133, 412 – 421. S. Naeem & P. Inchausti), pp. 61–75. Oxford University
Mulder, C.P.H., Koricheva, J., Huss-Danell, K., Högberg, P. Press, Oxford.
& Joshi, J. (1999) Insects affect relationships between plant Siemann, E. (1998) Experimental tests of effects of plant pro-
species richness and ecosystem processes. Ecology Letters, ductivity and diversity on grassland arthropod diversity.
2, 237 –246. Ecology, 79, 2057 –2070.
Otway, S., Hector, A. & Lawton, J.H. (2005) Resource Siemann, E., Carson, W.P., Rogers, W.E. & Weisser, W.W.
dilution effects on specialist insect herbivores in a grassland (2004) Reducing herbivory using insecticides. Insects and
biodiversity experiment. Journal of Animal Ecology, 74, Ecosystem Function (eds W.W. Weisser & E. Siemann), Eco-
234 – 240. logical Studies 173. Springer, Berlin.
Petchey, O.L. (2004) On the statistical significance of func- Spehn, E.M., Hector, A., Joshi, J., Scherer-Lorenzen, M.,
tional diversity effects. Functional Ecology, 18, 297– 303. Schmid, B., Bazeley-White, E. et al. (2005) Ecosystem
Piepho, H.P., Büchse, A. & Richter, C. (2004) A mixed modelling effects of biodiversity manipulations in European grass-
approach for randomized experiments with repeated measures. lands. Ecological Monographs, 75, 37 – 63.
Journal of Agronomy and Crop Science, 190, 230 –247. Spehn, E.M., Scherer-Lorenzen, M., Schmid, B., Hector, A.,
Pinheiro, J.C. & Bates, D.M. (2000) Mixed-Effects Models in Caldeira, M.C., Dimitrakopoulos, P.G. et al. (2002) The
S and S-PLUS. Springer, New York. role of legumes as a component of biodiversity in a cross-
R Development Core Team (2005) R : A Language and Environ- European study of grassland biomass nitrogen. OIKOS, 98,
ment for Statistical Computing. R Foundation for Statistical 205 –218.
Computing, Vienna, Austria, http://www.R-project.org. Tilman, D., Reich, P.B., Knops, J., Wedin, D., Mielke, T. &
Root, R.B. (1973) Organization of a plant–arthropod associ- Lehman, C. (2001) Diversity and productivity in a long-
ation in simple and diverse habitats: the fauna of collards term grassland experiment. Science, 294, 843–845.
(Brassica oleracea). Ecological Monographs, 43, 95 –124.
Roscher, C., Schumacher, J., Baade, J., Wilcke, W., Gleixner, Received 17 January 2006
G., Weisser, W.W. et al. (2004) The role of biodiversity for revision accepted 8 March 2006
element cycling and trophic interactions: an experimental Handling Editor: Roy Turkington

© 2006 The Authors

Journal compilation
© 2006 British
Ecological Society,
Journal of Ecology,
94, 922–931