This is a self-archived – parallel published version of an original article. This 
version may differ from the original in pagination and typographic details. 
When using please cite the original. 
 
 
This is a post-peer-review, pre-copyedit version of an article published in 
Journal Oecologia 
 
 The final authenticated version is available online at 
DOI https://doi.org/10.1007/s00442-011-2210-y 
 
 
CITATION Zvereva, E.L., Kozlov, M.V. Sources of variation in plant 
responses to belowground insect herbivory: a meta-analysis. 
Oecologia 169, 441–452 (2012). 
https://doi.org/10.1007/s00442-011-2210-y 
 
 
 
 
 
 1 
SOURCES OF VARIATION IN PLANT RESPONSES TO BELOWGROUND INSECT HERBIVORY: 
A META-ANALYSIS 
 
 
 
 
Elena L. Zvereva* and Mikhail V. Kozlov  
Section of Ecology, Department of Biology, University of Turku, FI-20014 Turku, Finland 
 
 
*Corresponding author: Elena L. Zvereva, Section of Ecology, Department of Biology, 
University of Turku, FI-20014 Turku, Finland 
E-mail: elezve@utu.fi 
Phone: +358 2 333 5492 
Fax: +358 2 333 6550 
 2 
 
Abstract 
 
Growing interest in belowground herbivory and the remarkable diversity of the accumulated 
information inspired us to quantitatively explore the variation in the outcomes of individual 
studies. We conducted a meta-analysis of 85 experimental studies reporting the effects of root-
feeding insect herbivores (36 species) on plants (75 species). On average, belowground 
herbivory led to a 36.3% loss of root biomass, which was accompanied by a reduction in 
aboveground growth (16.3%), photosynthesis (11.7%) and reproduction (15.5%). The effects of 
root herbivory on aboveground plant characteristics were significant in agricultural and 
biological control studies, but not in studies of natural systems. Experiments conducted in 
controlled environments yielded larger effects on plants than field experiments, and infestation 
experiments resulted in more severe effects than removal studies employing natural levels of 
herbivory. Simulated root herbivory led to greater aboveground growth reductions than similar 
root loss imposed by insect feeding. External root chewers caused stronger detrimental effects 
than sap feeders or root borers; specialist herbivores imposed milder adverse effects on plants 
than generalists. Woody plants suffered from root herbivory more than herbaceous plants, 
although root loss was similar in these two groups. Evergreen woody plants responded to root 
herbivory more strongly than deciduous woody plants, and grasses suffered from root herbivory 
more than herbs. Environmental factors, such as drought, poor nutrient supply, among-plant 
competition and aboveground herbivory, increased the adverse effects of root damage on plants 
in an additive manner. In general, plant tolerance to root herbivores is lower than tolerance to 
defoliating aboveground herbivores. 
 
Keywords plant growth, photosynthesis, reproduction, root, tolerance  
 3 
Introduction 
 
Plants allocate to belowground parts up to 90% of their biomass (Andersen 1987; Bazzaz et al. 
1987). Herbivores feeding on belowground plant parts (roots, rhizomes and storage organs) may 
substantially reduce the fitness and survival of plants, thus affecting not only the net primary 
productivity of ecosystems, but also relative abundance of plants, species diversity and the 
succession of natural vegetation (Brown and Gange 1990; van der Putten 2003; Stein et al. 
2010). Many root herbivores are important agricultural pests, and some species are used as 
biological control agents (Brown and Gange 1990; Blossey and Hunt-Joshi 2003).  
 The importance of studying root-feeding herbivores for understanding the dynamics of 
plant communities in both natural and managed ecosystems was underscored in a number of 
narrative reviews (Andersen 1987; Brown and Gange 1990; Hunter 2001, 2008; Blossey and 
Hunt-Joshi 2003). A common point of all of these reviews is that herbivore research has largely 
ignored belowground herbivory. The phrase “out of sight out of mind” reflects the past attitude 
of many ecologists (except for those studying agricultural pests) to belowground processes 
(Brown and Gange 1990; Hunter 2001). However, the most recent review (Blossey and Hunt-
Joshi 2003) demonstrated considerable growth in the number of publications on root herbivory, 
especially due to the development of biological control research.  
 As in the case of aboveground herbivory (Maschinski and Whitham 1989; Wise and 
Abrahamson 2005), the reported effects of root herbivores on plant performance constitute a 
continuum from positive to strongly negative (Blossey and Hunt-Joshi 2003). This diversity in 
responses, which may be related to the experimental methods used, study system, or 
environmental characteristics, emphasises the urgent need for identifying common trends and 
differences from the outcomes of individual studies. This task can only be accomplished through 
a meta-analysis that (in contrast to a narrative review) allows assessing the magnitude and 
 4 
significance of effects across studies and statistically estimating the influence of different study 
characteristics on the outcomes of the research (Gurevitch and Hedges 2001). Moreover, meta-
analysis reveals both research and publication biases that may considerably influence the 
conclusions of narrative reviews (Zvereva and Kozlov 2010). While a number of meta-analyses 
are devoted to plant responses to aboveground herbivory (Bigger and Marvier 1998; Hawkes and 
Sullivan 2001; Nykänen and Koricheva 2004; Zvereva et al. 2010), our study is the first to 
quantitatively explore the accumulated information on the impacts of root herbivory on plants by 
means of a meta-analysis. This numerical assessment is needed, in particular, for incorporation 
of the effects caused by root herbivory into biogeochemical ecosystem models (Wolf et al. 
2008). 
 We restricted our meta-analysis to experimental studies addressing the effects of 
belowground herbivory on root biomass and/or associated changes in plant performance. Our 
goal was to evaluate the average magnitudes of root damage effects on plant growth, 
photosynthesis and reproduction both alone and in combination with various environmental 
factors, such as drought, soil nutritional quality, among-plant competition and aboveground 
herbivory, and to reveal sources of variation in plant responses to root herbivory. In particular, 
we (1) compared the effects of simulated herbivory and insect feeding; (2) compared changes in 
the below- and aboveground biomass of damaged plants; and (3) explored variation associated 
with taxonomic affinity, feeding habits and the level of host plant specialisation of root 
herbivores, and with taxonomic affinity and life forms of damaged plants. We also investigated 
how the methodology of the primary studies may have influenced their results and searched for 
publication bias. Finally, we used the outcomes of earlier meta-analyses (Bigger and Marvier 
1998; Hawkes and Sullivan 2001; Nykänen and Koricheva 2004; Morris et al. 2007; Zvereva et 
al. 2010) to compare plant responses to root damage with the consequences of damage imposed 
by aboveground herbivores. 
 5 
Methods 
Data collection 
 
To be included in our meta-analysis, a study had to fit the following criteria: 
1. Plant responses to root damage were studied in the experiments (i.e., observational studies 
and correlative data were excluded);  
2. Root damage was imposed by either insect feeding or simulated herbivory (mechanical root 
damage); 
3. Root damage treatment had an appropriate control;  
4. Plant responses were measured in terms of growth (including biomass per unit area), 
reproduction or photosynthesis; 
5. Means, variances and sample sizes were reported for damaged and control plants, were 
available from the authors, or it was possible to estimate missing variances from test 
statistics. 
 We searched for publications that met these criteria in the ISI Web of Science database 
and on the internet using several keywords reflecting the mode of feeding (e.g., ‘root’, 
‘belowground’, ‘herbivor*’) or names of the most common root-feeding insects in combination 
with ‘plant growth’ or ‘damage’ or ‘photosynthesis’ and further examined the reference lists of 
all of the identified papers. The search was completed on November 15, 2010. 
 
Data selection and response variables 
 
We extracted data on biomass and/or the linear size of the entire plant, above- and belowground 
parts or individual organs (both vegetative and reproductive); when both fresh and dry mass 
were reported, dry mass was preferred. We also used vegetative and reproductive biomass 
 6 
production per unit area and the numbers of organs per plant (shoots, leaves/needles, roots, 
flowers, fruits, seeds). The included characteristics of photosynthesis were both the 
photosynthetic rate (measured as the net amount of carbon fixed per unit area of a leaf) and the 
leaf chlorophyll content (both absolute and relative values).  
 When a study reported several subsequent measurements of plants, we selected the final 
measurement for the analysis. When more than one level of herbivore infestation or simulated 
root damage was studied, we selected the effects caused by the lowest and the highest treatments 
to explore the dependence of plant responses on the severity of damage. Our estimations of the 
overall effects and exploration of the effects of methodology on the outcomes of the primary 
studies included both levels of damage. However, we excluded low-severity treatments when 
analysing variation related to the characteristics of herbivores and host plants.  
 We created an additional database for studies in which the effects of root herbivory on 
plants were explored both alone and in combination with other environmental factors, such as 
aboveground herbivory (both simulated and natural), among-plant competition (both inter- and 
intraspecific), soil moisture regime or soil fertility. This database was used to compare the 
combined effect with the individual effects of root herbivory and other stressors. 
 As a rule, we extracted the means, variances and sample sizes from publications or 
obtained these from the authors (one study: Riedell and Rees 1999). For several publications 
fitting criteria 1-4, we approximated the variances of the control and treatment groups using the 
reported means, sample sizes and Least Significant Difference values as described by Zvereva et 
al. (2010). 
 
Classificatory variables 
 
 7 
We classified all studies by research domain into the following categories: agricultural and 
forestry studies (agricultural studies hereafter), dealing with crop plants and plants used in 
greenery or reforestation, in association with their pest species; biological control studies 
(invasive plants in association with biocontrol agents); or basic ecology studies, usually dealing 
with natural systems (wild plants in association with their natural herbivores). 
 The variables related to experimental design included the experimental environment: 
outdoors (field or common-garden) vs. controlled (greenhouse, glasshouse, growing chamber, 
laboratory); the duration of the experiment: one-season vs. multiyear; and the method of root 
damage: mechanical vs. due to insect feeding. Studies manipulating herbivores were divided into 
two groups: experiments comparing artificially infested plants with non-infested controls 
(infestation studies hereafter) and those comparing naturally infested plants with plants from 
which herbivores were removed, usually by an insecticide treatment (removal studies hereafter). 
We also explored the effects of the intensity of damage by comparing the lowest and highest 
treatment levels (both mechanical and imposed by insect feeding).  
 Herbivores were classified according to their high-rank taxonomy (following the Fauna 
Europaea: www.faunaeur.org), feeding habits (external chewers, sap feeders and borers) and 
level of host plant specialisation. When herbivore specialisation was not mentioned in the case 
study, we classified herbivores feeding on plants of a single genus as specialists and herbivores 
feeding on plants from two or more genera as generalists. 
 Plants were classified by their basic life forms: herbaceous or woody. Within the woody 
plants we contrasted classes (Gymnospermae and Angiospermae) and life forms (deciduous and 
evergreen); within the herbaceous plants we compared herbs and grasses, as well as annual and 
perennial species (species that can be both annual and perennial were excluded from the latter 
analysis). 
 
 8 
Meta-analysis 
 
The Hedges’s d measure of the effect size (ES) was calculated as the difference between the 
means of the experimental and control data divided by the pooled standard deviation and 
weighted by sample size. The mean effect sizes for each treatment were calculated and compared 
using the MetaWin 2.0 program (Rosenberg et al. 2000). The treatment was considered to have a 
statistically significant effect if the 95% confidence interval (CI) of the mean effect size did not 
include zero (Gurevitch and Hedges 2001). All analyses were performed using random effects 
categorical models, assuming that the studies differed not only by sampling error, but also by a 
random component of the ESs (Rosenberg et al. 2000). For comparisons between groups of 
studies, we calculated between-group homogeneity (QB) and tested it against the χ2 distribution 
with N (the number of groups) minus one degrees of freedom (Gurevitch and Hedges 2001). 
Publication bias was examined by correlating effect size with sample size (Møller and Jennions 
2001). 
 
Additional analyses 
 
The proportion of growth reduction was calculated by averaging the ratio between the means of 
the experimental and control plants. For a subset of studies that simultaneously measured 
changes in both below- and aboveground biomass, we regressed the ESs calculated for 
aboveground parts against the ESs calculated for roots (SAS REG procedure; SAS Institute 
2009). The proportions of non-negative effect sizes in different databases were compared using a 
Kruskal-Wallis test (SAS NPAR1WAY procedure; SAS Institute 2009). 
 
Results 
 9 
 
Databases 
 
The main database used to examine the effects of root herbivory included 453 ESs extracted 
from 85 papers published between 1941 and 2010 (Supplementary material 1-2). These 
publications reported root losses (57 papers, 147 ESs), as well as changes in growth 
characteristics of aboveground plant parts (65 papers, 229 ESs) or the whole plant (13 papers, 19 
ESs), in various reproductive characteristics (21 papers, 41 ESs) and in measures related to 
photosynthesis: chlorophyll content (2 papers, 2 ESs) and photosynthetic rate (8 papers, 15 ESs). 
The production per unit area was measured in 3 studies for vegetative biomass (6 ESs) and in 3 
studies for seed/fruit yield (6 ESs).  
 The effects of root herbivory were documented for 75 plant species (33 woody and 42 
herbaceous) from 27 families. The amount of data obtained for woody plants was approximately 
one-third the amount of data for herbaceous plants (22 papers, 107 ESs and 62 papers, 345 ESs, 
respectively); one study (1 ES) did not discriminate between the roots of different forest plants. 
The best studied woody plants were citruses (Citrus spp.; 29 ESs, or 26.8% of all woody plant 
data). Among herbaceous species, the largest amount of information (80 ESs, or 23.1% of all 
herbaceous plant data) was obtained for corn (Zea mays).  
 The effects of insect feeding were explored more frequently than the effects of simulated 
herbivory (71 and 14 papers, respectively). Root-damaging insects included 36 species from 14 
families of 4 orders (Hemiptera, Coleoptera, Lepidoptera, Diptera), among which beetles were 
investigated most intensively (25 species, 336 ESs). Six studies (11 ESs) considered damage 
imposed by natural multispecies complexes of root-feeding insects. Most of the studied 
herbivores were agricultural or forestry pests (20 species, 320 ESs) or weed control agents (8 
species, 48 ESs). 
 10 
 The second database (Supplementary material 3), which was created to explore the 
combined effects of root herbivory and other environmental factors, included 256 ESs extracted 
from 27 papers that investigated the interactions of root herbivory with aboveground herbivory 
(12 papers), among-plant competition (5 papers), drought (4 papers) and soil fertilisation (6 
papers). 
 
Overall effects of root herbivory 
 
Root herbivory caused significant decreases in all of the investigated plant performance 
characteristics (Fig. 1). The magnitudes of the effects on the size of aboveground plant parts 
(measured as biomass, height, stem diameter, or leaf area) and on the number of plant parts 
(stems, leaves, fruits, or seeds) (Fig. 1) were similar, allowing us to combine these characteristics 
in the further analyses.  
 An observed overall 36.3% loss of root biomass was accompanied by a reduction of 
aboveground growth by 16.3%, reproductive characteristics by 15.5% and photosynthetic 
parameters by 11.7%. For studies reporting losses in both below- and aboveground plant 
biomass, the decrease of aboveground size was proportional to root loss (dabove = -0.389 + 
0.213*dbelow ; R
2 = 0.078, F1, 102 = 8.60, P = 0.0042). 
 
Variations related to research domain 
 
The agricultural and weed control studies yielded stronger overall adverse effects than basic 
ecology studies (Fig. 2). This pattern was detected both within infestation experiments (d = -
0.88, n = 327, CI = -1.98…-0.77 and -0.43, n = 40, CI = -0.70…-0.16, respectively, QB = 8.99, df 
= 1, P = 0.003) and within simulation experiments (d = -1.44, n = 33, CI = -1.81…-1.06 and -
 11 
0.81, n = 30, CI = -1.14…-0.48, respectively, QB = 6.81, df = 1, P = 0.009), indicating that the 
overall difference (Fig. 2) is not explained by the preferred method of manipulation. 
Consequently, studies of managed systems reported greater effects than studies of natural 
systems (d = -0.99, n = 305, CI = -1.10…-0.88 and d = -0.58, n = 148, CI = -0.72…-0.44, 
respectively; QB = 19.8, df = 1, P = 0.00001). The decrease in the magnitude of the effect with 
increase of sample size found in agricultural (τB = 0.177, n = 305, P = 0.0000) and weed control 
(τB = 0.208, n = 48, P = 0.048) studies suggested that outcomes of applied studies may have been 
affected by publication bias. In contrast, the ESs reported in basic ecology studies were 
independent of sample size (τB = 0.063, n = 100, P = 0.35), indicating an absence of publication 
bias in this research domain. 
 
Variations related to methodology 
 
The method of manipulation influenced the outcomes of the experiments (Fig. 2). Although 
simulation treatments on average removed a smaller proportion of root biomass than herbivores 
(d = -0.91, n = 21, CI = -1.39…-0.42 and d = -1.45, n = 126, CI = -1.64…-1.25, respectively; QB 
= 4.48, df = 1, P = 0.03), they imposed stronger detrimental effects on the growth of 
aboveground plant parts than root damage by insects (d = -1.04, n = 37, CI = -1.31…-0.78 and d 
= -0.52, n = 250, CI = -0.62…-0.42, respectively; QB = 13.5, df = 1, P = 0.0002).  
 Root herbivory imposed stronger effects on plants in controlled environments compared to 
field experiments (Fig. 2), but this difference was observed for aboveground traits only (QB = 
4.35, df = 1, P = 0.04). Among the included field studies, infestation and removal experiments 
yielded similar effects (d = -0.53, n = 159, CI = -0.65…-0.41 and d = -0.44, n = 23, CI = -
0.78…-0.09, respectively; QB = 0.30, df = 1, P = 0.59), indicating that weaker effects observed in 
the field were not due to application of herbivore removal exclusively in the field studies. 
 12 
Moreover, infestation experiments considered separately resulted in stronger effects in the field 
than in the controlled environments (d = -0.55, n = 159, CI = -0.69…-0.41 and d = -1.06, n = 
208, CI = -1.19…-0.92, respectively; QB = 26.9, df = 1, P = 0.0000). Removal experiments 
demonstrated that herbivores naturally occurring in soil imposed significant adverse effects on 
crop plants in agricultural ecosystems (d = -1.36, n = 7, CI = -2.08…-0.64), but not in non-crop 
plants natural ecosystems (d = -0.09, n = 16, CI = -0.44…0.27; QB = 14.2, df = 1, P = 0.0002). A 
similar difference was observed for infestation experiments (QB = 8.69, df = 1, P = 0.003), 
although the effects were significant in both agricultural and natural systems (d = -0.91, n = 272, 
CI = -1.03…-0.80 and d = -0.59, n = 95, CI = -0.77…-0.40, respectively).  
 In studies reporting several levels of simulated damage or infestation rates, the highest 
damage levels caused larger effects than the lowest damage levels. However, the lowest damage 
levels still resulted in significant decrease in plant performance (Fig. 2) in the experiments using 
both simulated herbivory (d = -1.03, n = 16, CI = -1.51…-0.55) and infestation treatments (d = -
0.64, n = 112, CI = -0.81…-0.48). 
  Multiyear experiments with woody plants yielded smaller effects than single season 
experiments (d = -0.04, n = 10, CI = -0.67…-0.60 and d = -1.30, n = 86, CI = -1.51…-1.09 
respectively; QB = 17.5, df = 1, P = 0.00003). In contrast, long-term studies with perennial 
herbaceous plants yielded stronger adverse effects than short-term studies (d = -1.38, n = 11, CI 
= -1.95…-0.81 and d = -0.74, n = 109, CI = -0.92…-0.57 respectively; QB = 5.49, df = 1, P = 
0.02).  
 
Variations related to herbivore taxonomy and life history 
 
Root-feeding species differed in their impacts on plants (QB = 130.3, df = 31, P < 0.00001). This 
variation was not explained by high-rank taxonomic affinities: differences between the four 
 13 
investigated insect orders were not significant (Fig. 3a). However, within the beetles 
(Coleoptera) we found significant differences between families: the strongest adverse effects on 
plants were imposed by Curculionidae and Scarabaeidae, whereas the effects of Elateridae and 
Cerambycidae species were not significant (Fig. 3a). The two most damaging families, 
Curculionidae and Scarabeidae, did not differ in their effects on plants (QB = 0.21, df = 1, P = 
0.65). Among species considered in more than two papers, the largest effects on plants were 
imposed by Diaprepes abbreviatus (Coleoptera: Curculionidae) and Agapeta zoegana 
(Lepidoptera: Tortricidae) (Fig. 3a).  
 
Insect feeding guilds differed in their effects on plant performance: external chewers caused two 
times stronger detrimental effects than sap feeders or borers (Fig. 3b). However, this difference 
was mostly due to effects on root biomass, while the effects on aboveground parts were similar 
(QB = 0.44, df = 2, P = 0.80). The stronger overall effect of generalist herbivores (Fig. 3b) was 
due to removal of a significantly larger portion of the root biomass relative to specialised 
herbivores (QB = 10.6, df = 1, P = 0.001), while the effects of both groups on aboveground plant 
parts were similar (QB = 0.85, df = 1, P = 0.36). This difference in root damage between 
generalists and specialists was not due to variation between feeding guilds because within a guild 
of external chewers, generalists still caused greater root losses than specialists (QB = 16.4, df = 1, 
P = 0.00005).  
 
Variations related to plant taxonomy and life form 
 
Woody plants suffered from root herbivory more than herbaceous plants (Fig. 4), although the 
average root losses were similar in these two groups (QB = 2.28, df = 1, P = 0.13). This 
difference was mostly due to the effects on aboveground growth (QB = 7.06, df = 1, P = 0.008), 
 14 
whereas the effects of root herbivory on reproduction did not differ between woody and 
herbaceous plants (QB = 0.002, df = 1, P = 0.96). Within woody plants, gymnosperms did not 
differ from angiosperms (QB = 0.45, df = 1, P = 0.51), but evergreens responded to root 
herbivory more strongly than deciduous plants (Fig. 4). The latter difference resulted mostly 
from large effects reported for evergreen leaf-bearing plants (primarily Citrus species). Among 
deciduous woody plants, juveniles suffered from root herbivory more than mature individuals 
(QB = 5.07, df = 1, P = 0.02), while for evergreen woody plants, we found no difference in the 
responses among ontogenetic stages (QB = 1.40, df = 1, P = 0.24).  
 Within herbaceous plants, monocots (grasses) suffered from root herbivory more than 
dicots (herbs), and perennials did not differ from annuals in their responses to root feeders (Fig. 
4).  
 The detected patterns were not affected by inevitable differences between herbivores 
associated with specific groups of plants, as demonstrated by the analyses restricted to a specific 
group of herbivores (results not shown). 
 
Interactions with environmental factors 
 
The effects of root herbivory on plants were of the same magnitudes on average as the effects of 
aboveground herbivory, competition, drought and nutrient stress (P = 0.27-0.76). However, root 
damage resulted in two times stronger adverse effects, on average (Fig. 5), when it was 
combined with aboveground herbivory (QB = 7.17, df = 1, P = 0.007), increased competition (QB 
= 6.60, df = 1, P = 0.06), drought (QB = 6.34, df = 1, P = 0.01), or low nutrient supply (QB = 
9.67, df = 1, P = 0.002). 
 
Discussion 
 15 
 
Trends in root herbivory research: plant perspective 
 
The number of studies reporting quantitative estimates of the effects of root herbivory on plants 
increased rapidly in the early 1990s: 83% of the studies included in our meta-analysis were 
published from 1991 to 2010. However, our analysis supports the opinion of Blossey and Hunt-
Joshi (2003) that the number of publications addressing root herbivores in recent years is 
growing very slowly: our database includes 36 studies published from 2001-2010, which is 
nearly equal to the 35 studies published from 1991-2000. Moreover, Blossey and Hunt-Joshi 
(2003) estimated that only approximately 10% of root herbivory studies have been conducted in 
the research domain of basic ecology, and our meta-analysis demonstrated no increase in the 
interest of basic ecologists on the effects of root herbivory on plant performance (15 studies were 
published from 1991-2000 and 15 from 2001-2010). Thus, the amount of ‘basic’ root herbivory 
research addressing changes in plant growth and productivity is increasing much more slowly 
than was expected two decades ago (Brown and Gange 1990). This may be at least partly related 
to considerable difficulties in the manipulation and observation of belowground objects. 
However, the number of studies exploring root feeders as potential biological control agents is 
growing continuously, and root herbivores are increasingly being used to control invasive plant 
species (Blossey and Hunt-Joshi 2003). Another rapidly developing research field addresses 
interactions between above- and belowground herbivores (van Dam & Heil 2011). 
 
General patterns and underlying mechanisms 
 
Our results confirm the conclusion of numerous studies (reviewed by Andersen 1987; Brown and 
Gange 1990; Hunter 2001, 2008; Blossey and Hunt-Joshi 2003) that belowground damage 
 16 
causes considerable detrimental effects on aboveground plant traits, and our estimate of the 
magnitude of this effect (Fig. 1) is similar to the estimate by Morris et al. (2007) (root feeders: d 
= -0.55, n = 10). Detrimental effects result from several processes: (i) decreases in water and 
nutrition uptake due to direct reductions in root biomass (mostly when fine roots are consumed) 
or disruption of water and nutrient flows (mostly when the main root is damaged); (ii) 
consumption of resources stored belowground; (iii) decreases in photosynthetic rates due to 
water deficits imposed by root damage; (iv) translocation of assimilates to roots for root 
regrowth; and (v) direct costs of inducible defences (for more details, consult Supplementary 
material 4). 
 The processes described above lead to the limitation of all types of resources, and 
therefore the effects of root herbivory may be very severe, leading not only to growth 
retardation, but even to plant death (Gange et al. 1991; Maron 2001). In agreement with this 
conclusion, a number of studies found that the detrimental impacts of root herbivory on plants 
may exceed the effects of aboveground herbivores (Brown and Gange 1989; Maron 1998), 
although the opposite pattern has been also reported (e.g., Houle and Simard 1996). The 
magnitude of the aboveground growth responses of woody plants to root herbivory estimated in 
our meta-analysis (d = -0.98) is very similar to the effects caused by defoliators (d = -1.00: 
Nykänen and Koricheva 2004) and sap feeders (d = -1.06: Zvereva et al. 2010). Moreover, in 
both meta-analysis by Morris et al. (2007) and in our meta-analysis of studies addressing the 
combined effects of above- and belowground herbivory on plants (Fig. 5), the magnitudes of the 
effects did not differ between above- and belowground herbivores acting alone. Thus, our results 
seem not to support the opinion (Ingham and Detling 1986; Brown and Gange 1990; Móron-Ríos 
et al. 1997) that root herbivores impose generally stronger impacts on plant fitness than 
aboveground herbivores. 
 17 
 However, although the overall reduction in plant fitness found in three meta-analyses 
mentioned above was of approximately same magnitude, the mechanisms underlying the impacts 
of belowground herbivores on plants differ considerably from those of their aboveground 
counterparts. The proportion of non-negative effect sizes (i.e., those indicating exact 
compensation or overcompensation) in response to defoliation found in the meta-analysis by 
Hawkes and Sullivan (2001) was twice as high as in response to root herbivory (34.6 and 17.2 
%, respectively; χ2= 13.0, df = 1, P = 0.0003). Furthermore, while the meta-analysis by Nykänen 
and Koricheva (2004) demonstrated that defoliation generally causes an increase in 
photosynthesis, we found that root damage leads to a significant reduction of photosynthesis 
(Fig. 1). The responses of plants to root feeders are more similar to plant responses to sap feeders 
(Zvereva et al. 2010), which induce compensatory growth even less frequently than root feeders 
(8.1 and 17.2 %, respectively; χ2= 15.9, df = 1, P < 0.0001). We suggest that this similarity is 
explained by a reduction of photosynthesis in response to both root feeders (Fig. 1) and 
aboveground sap feeders (Zvereva et al. 2010). Although the suppression of photosynthesis is 
underpinned by different mechanisms (discussed for sap feeders by Zvereva et al. 2010), it 
hampers the compensatory growth of plants in both cases. Moreover, while aboveground 
herbivory induces defences mostly in leaves, root damage elicits equally strong responses in both 
leaves and roots (Kaplan et al. 2008), which may result in higher fitness costs. 
Notably, in our meta-analysis, only 7% of the ESs calculated for root biomass were non-
negative, indicating that the accelerated root growth frequently observed in response to 
belowground herbivory (Andersen 1987; Brown and Gange 1990) is generally insufficient to 
compensate for root losses. However, it is possible that this rarity of compensation is explained 
by measuring plants immediately after the termination of herbivore feeding, whereas most 
intensive root regrowth may occur later. This hypothesis is supported by the observation of a 
 18 
steady recovery of CO2 assimilation rates in plants that have been monitored after termination of 
root damage (Riedell and Rees 1999).  
 
Interactions with environmental factors 
 
As concluded above, root herbivory imposes strong resource limitations on plants, and therefore, 
the effects of root herbivory on plant performance should be more severe in conditions of low 
resource availability (Wise and Abrahamson 2005). Our meta-analysis confirmed this prediction: 
the effects of root herbivory interact with those of drought and nutritional stress, among-plant 
competition and aboveground herbivory in an additive manner, such that the combined effects 
considerably (by more than two times) exceed the detrimental effects of both root herbivory 
alone and the individual effects of other stressors. Consequently, increased supplies of water and 
nutrients mitigate the adverse impacts of root herbivores on plant growth (Fig. 5; Ladd and 
Buriff 1979; Gange and Brown 1989). Plants growing at high densities can be water and nutrient 
stressed (Tilman 1989; Maron 2001), as well as experience light limitations (Agrawal 2004). 
Therefore the additive effects of among-plant competition may be based on the same 
mechanisms as effects of drought and nutritional stress. Moreover, plant defences can be 
compromised when individuals grow at a high density (Karban et al. 1989). 
 The discovery of a synergy between the effects of above- and belowground herbivory on 
plant performance is not surprising because simultaneous damage of different plant parts 
imposes severe limitations on all of the resources required for plant growth and reproduction. 
However, the effects of spatially separated herbivores on a common host plant may be modified 
by the impacts of these two groups of herbivores on each other. Root herbivores may favour the 
development of aboveground herbivores by increasing the nutritional value of foliage (Masters et 
al. 1993; Johnson et al. 2009); in turn, aboveground herbivory may enhance the development of 
 19 
root herbivores (Seastedt et al. 1988; Johnson et al. 2009). These positive interactions may 
increase herbivore pressure on plants, and the cumulative impact of herbivores attacking 
different plant tissues is sometimes used to accelerate and improve biological weed control 
(McEvoy et al. 1991; Malecki et al. 1993). However, the interactions between below- and 
aboveground herbivory, which recently attracted considerable attention from researches 
(reviewed by van Dam and Heil 2011), are complex and variable, and therefore, their 
consequences for plants are context specific. 
 
Study systems and research domains 
 
Our meta-analysis found no effects of root herbivory on plant performance in natural 
ecosystems. This result is in line with the opinion that at low densities, the effects of root 
herbivores on plants can be considered neutral (Blossey and Hunt-Joshi 2003 and references 
therein). Therefore, significance of adverse effects caused by low herbivore densities suggests 
that in most infestation experiments (85% of which are agricultural studies), the lowest of 
densities applied still exceed the levels of root herbivory in natural (i.e., unmanaged) ecosystems. 
 Applied (agricultural and biological control) studies reported significantly stronger 
reductions in plant growth and reproduction than basic ecology studies. This can be explained by 
several factors. First, applied studies are goal oriented (as defined by Blossey and Hunt-Joshi 
2003), and therefore, the selection of study objects is limited to most damaging species, which 
are either agricultural pests or biocontrol agents. Second, in applied studies, researchers tend to 
use extremely high infestation rates far exceeding the herbivory levels in natural systems. This 
conclusion is supported by the significantly lower effects detected in removal studies, which 
employ natural levels of herbivory (Fig. 2). Importantly, the same difference between infestation 
and removal studies was found in meta-analyses of the effects of aboveground herbivory on 
 20 
plants (Bigger and Marvier 1998; Zvereva et al. 2010). Third, the significant publication bias 
found in applied studies suggests that researchers working in this domain are less likely to 
publish studies yielding non-significant or unexpected effects than researchers working on basic 
ecological problems. In weed control studies, in particular, the failure to find the desired effect of 
a potential biocontrol agent may well justify abandonment of the results. Thus, due to an 
overrepresentation of applied studies in our database (over 80% of the ESs), the average negative 
effect of root herbivory on plant performance is likely to be considerably overestimated. 
However, overestimation of the magnitude of these effects does not question the suitability of 
these studies for the exploration of mechanisms or sources of variation in plant responses to root 
herbivory.  
 
Variations related to methodology 
 
Studies conducted in controlled environments (glasshouse, climatic chamber or the laboratory, 
with plants being grown in pots or in hydroponics) yield two times stronger effects than field 
studies. This difference between field and greenhouse studies was also found in a meta-analysis 
of the effects of sap-feeding herbivores on plant growth (Zvereva et al. 2010). More generally, 
laboratory studies, by eliminating a substantial portion of natural variability, are likely to 
produce biased (usually exaggerated) results (discussed by Kozlov et al. 2009). Because plant 
responses to herbivory are affected by numerous abiotic and biotic factors (Strauss and Agrawal 
1999), the lack of these multiple interactions in greenhouse experiments could have contributed 
to the observed differences. Thus, although studies in controlled environments are vital for the 
exploration of mechanisms behind observed phenomena, the magnitudes of the effects detected 
in these experiments cannot be directly extrapolated to field conditions. 
 21 
 We have found that the detrimental effects of simulated herbivory on aboveground traits 
are twice as strong as the effects of natural herbivory (Fig. 2). Our meta-analysis indicates that 
this difference is not due to stronger root damage in simulation treatments. However, it may be 
partly explained by different spatial and temporal patterns of root damage (Andersen 1987): 
while simulated herbivory usually removes a certain proportion of roots all at once, damage by 
insects accumulates gradually and is partially counterbalanced by root regrowth. Furthermore, 
many wound-induced defensive responses are elicited by compounds contained in insect oral 
secretions but do not develop following mechanical tissue damage (Koo and Howe 2009; Poveda 
et al. 2010). Thus, our results confirm conclusions by Andersen (1987) and Blossey and Hunt-
Joshi (2003) that artificial damage can hardly mimic damage caused by herbivores.  
 We conclude that most of the experimental studies addressing the effects of root herbivory 
on plants apply methodology (e.g., mechanical root damage, high infestation rates, controlled 
environments) that unavoidably leads to a considerable overestimation of the effects relative to 
those observed under natural levels of herbivory. Therefore, results of such experiments should 
only very cautiously be used to estimate biomass losses due to root herbivory in either natural or 
managed systems. The most realistic information can be obtained from field studies. 
 
Variations related to herbivores 
 
We discovered two sources of variation related to herbivore feeding habits. First, external 
chewers impose stronger effects on plants than sap feeders and root borers because they cause 
greater losses of root biomass. External chewers feed on fine roots, which are the most 
vulnerable and least defended roots (van Dam 2009), and at the same time, they are most critical 
for water and nutrient uptake (Eissenstat and Yanai 1997). In contrast, root borers feeding within 
the main root not only consume resources stored belowground but may also disrupt vascular 
 22 
connections between roots and aboveground plant parts. Thus, although insects from different 
feeding guilds affect plants through different mechanisms and remove a different proportion of 
root biomass on average, they cause similar impacts on aboveground plant parts.  
 Second, non-specialised (generalist) species impose considerably greater detrimental 
effects on plants than specialised herbivores. This finding supports the conclusions of a meta-
analysis of plant responses to sap-feeding aboveground herbivores (Zvereva et al. 2010). The 
concordance between these two ecologically different groups of herbivores indicates that weaker 
plant responses to specialised herbivores may appear to be a general phenomenon. The evolution 
of co-adaptations of herbivores with their hosts is likely to depend on the intimacy of the 
relationships between them, which is closer for specialist than for generalist feeders. Because 
some studies have demonstrated the existence of a trade-off between the evolution of plant 
defences against generalist and specialist herbivores (Johnson et al. 2009), our findings may 
indicate that specialised root feeders impose stronger selective pressure on host plants during 
their coevolutionary history than generalists. As a result, plants become adapted to tolerate 
damage by highly specialised pests better than damage by generalist herbivores. 
 The stronger adverse effects of generalists cannot be explained by the predominance of 
external chewers in this group, as among external chewers, the difference between generalists 
and specialists was also significant. However, this difference may result from the greater root 
losses caused by generalists. This, in turn, may be related to the large size of scarab beetles, 
which are the most numerous group of generalist external chewers. More generally, we believe 
that the differences between the effects on plant performance caused by different insect taxa 
(Fig. 3a) are explained by their life histories rather than a taxonomic affinity per se.  
 
Variations related to plants 
 
 23 
Our meta-analysis revealed the higher sensitivity of woody plants to root herbivory (Fig. 4). This 
pattern contrasts the results by Hawkes and Sullivan (2001) who did not find any differences in 
growth responses to aboveground herbivory between woody plants and herbs. Moreover, based 
on comparisons of the degree of physiological integration among modules, life span and biomass 
partitioning among plant organs, Haukioja and Koricheva (2000) argued that woody plants 
should be more tolerant to herbivory than herbaceous plants. Therefore, we suggest that our 
result (Fig. 4) was influenced by the predominance of juveniles in studied woody plants (83%), 
whereas herbaceous plants used in the experiments were generally mature (i.e., had reached 
reproductive age). Mature plants tolerate herbivory better than juvenile plants (Strauss and 
Agrawal 1999; Nykänen and Koricheva 2004; Zvereva et al. 2010, but see Barton and Koricheva 
2010) because of limitations on resource acquisition and stored reserves in younger plants, and 
this appears to also be true (at least in deciduous plants) in response to root herbivores, which 
further hamper resource uptake. The dependence of the effects of root herbivory on plant 
ontogenetic stages explains a seemingly paradoxical result, that is, the smaller negative effects 
discovered in long-term studies of woody plants compared to short-term studies. We suggest that 
this pattern may result from increasing tolerance to herbivory as seedlings grow. 
 Our meta-analysis demonstrated that the growth of evergreen plants is reduced by root 
herbivory more than the growth of deciduous plants; and a similar pattern was found in a meta-
analysis of woody plant responses to aboveground herbivory (Nykänen and Koricheva 2004). 
The difference between evergreen and deciduous woody plants may be associated with patterns 
of biomass allocation: the storage of a larger proportion of biomass in woody tissues allows 
deciduous species to recover from herbivore damage by mobilising stored energy for 
compensatory growth (Chapin et al. 1990; Strauss and Agrawal 1999).  
 Grasses suffer from both below- and aboveground herbivory more than herbs (Fig. 4; 
Hawkes and Sullivan 2001), which is in agreement with the results of Stein et al. (2010), who 
 24 
found that in natural grassland communities, grasses became more dominant when belowground 
herbivores were excluded by insecticide treatment. The stronger adverse effects of aboveground 
herbivory on grasses are usually attributed to different locations of meristems, which is apical for 
herbs and basal for grasses (Haukioja and Koricheva 2000; Hawkes and Sullivan 2001), but this 
difference cannot explain greater sensitivity of grasses to belowground herbivory. Stein et al. 
(2010) suggested that the higher susceptibility of grasses to belowground herbivory results from 
their more extensive root system, which is therefore more exposed to root herbivores. However, 
this explanation is not valid for the majority of experiments included in our meta-analysis 
because they were usually established with monocultures.  
 Additionally, Haukioja and Koricheva (2000) emphasised the fact that woody plants and 
grasses share a number of life history traits, which may determine similarities in their responses 
to herbivory. Consistent with this prediction, we found that grasses are more similar to woody 
plants than to herbs in the magnitude of their response to belowground herbivory (Fig. 4). 
However, because very little is known about the mechanisms of plant tolerance to root herbivory 
(Hunter 2001), it seems premature to attempt to identify life history traits responsible for this 
similarity.  
 
Data gaps, biases and research needs 
 
Our meta-analysis allows reconsidering some data gaps and research needs pointed out in earlier 
reviews (Andersen 1987; Brown and Gange 1990; Hunter 2001, 2008; Blossey and Hunt-Joshi 
2003). Unfortunately, not all of the publications describing the effects of root feeding on plants 
contained numerical information sufficient to calculate the effect size. In particular, this concerns 
studies reporting effects of root herbivory on plant survival: we had to exclude this parameter 
from the meta-analysis because all of the identified studies (except for one: Maron 2001) 
 25 
contained mean values without measures of variance, and only a few of studies reported numbers 
of dead and survived plants (which is needed to perform meta-analysis based on odds ratios). 
Furthermore, many authors tend to present only the results of statistical tests, often omitting 
primary data. The relatively scant reporting of summary statistics in recent publications is 
especially disappointing. Because this problem has been independently discovered in different 
fields of ecology, we strongly support the suggestion by Paillet et al. (2010) that scientific 
journals should ask for mean values to always be presented along with their corresponding 
variance and sample size. 
 Another discovered problem is reporting bias, which affects meta-analyses much more 
strongly than narrative reviews. Three of the studies included in our database (Saner and Müller-
Shärer 1994; Maron 2001; Wurst and van der Putten 2007) presented (on graphs or in tables) 
only those results that allowed rejecting the null hypothesis, while no quantitative data were 
provided for experiments yielding nonsignificant results. This selective reporting may result in 
an overestimation of the consequences of root feeding on plant performance. 
 The majority of studies included in our meta-analysis were conducted in controlled 
conditions. The striking difference observed in the outcomes of ‘indoor’ and ‘outdoor’ 
experiments (Fig. 2) highlights the need to place greater emphasis on studying the effects of root 
herbivory on plants growing in their natural environments. This especially concerns non-
agricultural plants, particularly mature trees, for which responses to root herbivory remain 
virtually unknown. Furthermore, the study design should allow the exploration of compensatory 
root regrowth after termination of herbivore feeding. 
 One of research shortfalls that became obvious in the course of our meta-analysis is the 
need for the exploration of plant responses to ‘normal’ levels of root herbivory in natural 
systems. As we concluded in a meta-analysis of the effects of pollution on insects (Zvereva and 
Kozlov 2010), field studies on herbivory are more likely to be conducted in sites where 
 26 
herbivore damage is apparent. Therefore, the levels of plant damage calculated from published 
data are expected to more frequently be attributed to exceptionally high population densities of 
root herbivores (i.e., outbreaks), rather than to background (‘normal’) herbivory. Long-term 
studies of the population dynamics of root feeders in natural ecosystems are needed to compare 
the consequences of outbreaks with the background damage imposed by herbivores at low 
population densities. 
Finally, the synergy found between the effects of root herbivory and several biotic 
(competition, aboveground herbivory) and abiotic (water and nutrient availability) factors, 
highlights the urgent need to explore the interactive effects of root herbivory and the primary 
drivers of global change (temperature, ambient CO2 and UV-B radiation) on plant performance. 
Accounting for the sources of variation in the outcomes of primary studies uncovered by present 
and earlier meta-analyses (Zvereva and Kozlov 2006; Zvereva et al. 2010) would allow 
designing these experiments in such a way that the obtained information can serve as the basis 
for realistic predictions of plant performance under future climate conditions.  
 
Conclusions 
 
Our meta-analysis of the effects of belowground herbivory on root losses and associated changes 
in plant performance revealed significant heterogeneity in the outcomes of experimental studies 
and identified numerous sources of this variation. Various environmental factors, such as 
drought, poor nutritional quality of the soil, among-plant competition and aboveground 
herbivory, increase the adverse effects of root damage on plants in an additive manner. The 
changes in aboveground plant characteristics are proportional to root losses. The magnitude of 
the effects on root biomass depends on the feeding habits of herbivores: external chewers cause 
two times greater losses compared to sap feeders or root borers; changes in aboveground plant 
 27 
characteristics do not depend on insect feeding guilds. Generalists impose greater adverse effects 
than specialists, presumably because plants are better adapted to tolerate damage by their 
specialist consumers. The analysis of compensatory responses suggested that plant tolerance to 
root herbivores is lower than tolerance to defoliating aboveground herbivores.  
 A substantial portion of the variation in the outcomes of primary studies is related to 
methodological issues (field or controlled environments, simulated or natural herbivory, 
infestation or removal experiments, the herbivore density applied in infestation studies, duration 
of experiments), and this variations seems to be greater than variation related to the 
characteristics of the study systems. Not surprisingly, applied studies considerably overestimate 
general effects due to both research and publication biases. The most critical research gap found 
in this analysis is the shortage of information on the effects of root herbivory on woody plants, 
especially on mature forest trees. Additionally, long-term studies employing non-destructive data 
collection methods are urgently needed to better understand the compensatory responses of 
plants to root herbivory.  
 
Acknowledgements 
 
We are grateful to Dr. W. Riedell for providing additional information, to A. 
Stekolshchikov for assistance in publication searches and to two anonymous reviewers 
for valuable comments to an earlier version of the manuscript. The study was supported 
by the Academy of Finland (project number 122133) and Turku University strategic 
research grant. 
 
 28 
References 
 
Agrawal AA (2004) Resistance and susceptibility of milkweed: Competition, root herbivory, and 
plant genetic variation. Ecology 85:2118-2133 
Andersen DC (1987) Belowground herbivory in natural communities - a review emphasizing 
fossorial animals. Q Rev Biol 62:261-286 
Barton KE, Koricheva J (2010) The ontogeny of plant defense and herbivory: characterizing 
general patterns using meta-analysis. Am Nat 175:481-493 
Bazzaz FA, Chiariello NR, Coley PD, Pitelka LF (1987) Allocating resources to reproduction 
and defense. Bioscience 37:58-67 
Bigger D, Marvier MA (1998) How different would a world without herbivory be? A search for 
generality in ecology. Integr Biol Issues News Rev 1:60-67 
Blossey B, Hunt-Joshi TR (2003) Belowground herbivory by insects: Influence on plants and 
aboveground herbivores. Ann Rev Entomol 48:521-547 
Brown VK, Gange AC (1989) Differential effects of above-ground and below-ground insect 
herbivory during early plant succession. Oikos 54:67-76 
Brown VK, Gange AC (1990) Insect herbivory below ground. Adv Ecol Res 20:1-58 
Chapin FS, Schulze ED, Mooney HA (1990) The ecology and economics of storage in plants. 
Ann Rev Ecol Syst 21:423-447 
Eissenstat DM, Yanai RD (1997) The ecology of root lifespan. Adv Ecol Res 27:1-60 
Gange AC, Brown VK (1989) Effects of root herbivory by an insect on a foliar feeding species, 
mediated through changes in the host plant. Oecologia 81:38-42 
Gange AC, Brown VK, Farmer LM (1991) Mechanisms of seedling mortality by subterranean 
insect herbivores. Oecologia 88:228-232 
 29 
Gurevitch J, Hedges LV (2001) Meta-analysis. Combining the results of independent 
experiments. In: Schneider SM, Gurevitch J (eds) Design and analysis of ecological 
experiments. Oxford University Press, Oxford, pp 347-369 
Haukioja E, Koricheva J (2000) Tolerance to herbivory in woody vs. herbaceous plants. Evol 
Ecol 14:551-562 
Hawkes CV, Sullivan JJ (2001) The impact of herbivory on plants in different resource 
conditions: A meta-analysis. Ecology 82:2045-2058 
Houle G, Simard G (1996) Additive effects of genotype, nutrient availability and type of tissue 
damage on the compensatory response of Salix planifolia ssp. planifolia to simulated 
herbivory. Oecologia 107:373-378 
Hunter MD (2001) Out of sight, out of mind: the impacts of root-feeding insects in natural and 
managed systems. Agric For Entomol 3:3-9 
Hunter MD (2008) Root herbivory in forest ecosystems. In: Johnson SN, Murray PJ (eds) Root 
feeders: an ecosystem perspective. CABI, Wallingford, pp 68-95 
Ingham RE, Detling JK (1986) Effects of defoliation and nematode consumption on growth and 
leaf gas exchange in Bouteloua curtipendula. Oikos 46:23-28 
Johnson SN, Hawes C, Karley AJ (2009) Reappraising the role of plant nutrients as mediators of 
interactions between root- and foliar-feeding insects. Funct Ecol 23:699-706 
Kaplan I, Halitschke R, Kessler A, Sardanelli S, Denno RF (2008) Constitutive and induced 
defenses to herbivory in above- and belowground plant tissues. Ecology 89:392-406  
Karban R, Brody AK, Schnathorst WC (1989) Crowding and a plant ability to defend itself 
against herbivores and diseases. Am Nat 134:749-760 
Koo AJK, Howe GA (2009) The wound hormone jasmonate. Phytochemistry 70:1571-1580 
Kozlov MV, Zvereva EL, Zverev VE (2009) Impacts of point polluters on terrestrial biota: 
Comparative analysis of 18 contaminated areas. Springer, Dordrecht 
 30 
Ladd TL, Buriff CR (1979) Japanese beetle (Coleoptera: Scarabaeidae) - influence of larval 
feeding on bluegrass yields at two levels of soil moisture. J Econ Entomol 72:311-314 
Malecki RA, Blossey B, Hight SD, Schroeder D, Kok LT, Coulson JR (1993) Biological control 
of purple loosestrife. Bioscience 43:680-686 
Maron JL (1998) Insect herbivory above- and belowground: Individual and joint effects on plant 
fitness. Ecology 79:1281-1293 
Maron JL (2001) Intraspecific competition and subterranean herbivory: individual and 
interactive effects on bush lupine. Oikos 92:178-186 
Maschinski J, Whitham TG (1989) The continuum of plant responses to herbivory - the influence 
of plant association, nutrient availability, and timing. Am Nat 134:1-19 
Masters GJ, Brown VK, Gange AC (1993) Plant mediated interactions between aboveground 
and belowground insect herbivores. Oikos 66:148-151 
McEvoy P, Cox C, Coombs E (1991) Successful biological control of ragwort, Senecio 
jacobaea, by introduced insects in Oregon. Ecol Appl 1:430-442 
Møller AP, Jennions MD (2001) Testing and adjusting for publication bias. Trends Ecol Evol 
16:580-586 
Morón-Ríos A, Dirzo R, Jaramillo VJ (1997) Defoliation and below-ground herbivory in the 
grass Muhlenbergia quadridentata: Effects on plant performance and on the root-feeder 
Phyllophaga sp. (Coleoptera, Melolonthidae). Oecologia 110:237-242 
Morris WF, Hufbauer RA, Agrawal AA, Bever JD, BorowiczVA, Gilbert GS, Maron JL, 
Mitchell CE, Parker IM, Power AG, Torchin ME, Vazquez DP (2007) Direct and 
interactive effects of enemies and mutualists on plant performance: A meta-analysis. 
Ecology 88:1021-1029 
Nykänen H, Koricheva J (2004) Damage-induced changes in woody plants and their effects on 
insect herbivore performance: a meta-analysis. Oikos 104:247-268 
 31 
Paillet Y, Berges L, Hjalten J, Odor P, Avon C, Bernhardt-Romermann M, Bijlsma RJ, De 
Bruyn L, Fuhr M, Grandin U, Kanka R, Lundin L, Luque S, Magura T, Matesanz S, 
Meszaros I, Sebastia MT, Schmidt W, Standovar T, Tothmeresz B, Uotila A, Valladares 
F, Vellak K, Virtanen R (2010) Biodiversity differences between managed and 
unmanaged forests: meta-analysis of species richness in Europe. Conserv Biol 24:101-112 
Poveda K, Jimenez MIG, Kessler A (2010) The enemy as ally: herbivore-induced increase in 
crop yield. Ecol Appl 20:1787-1793 
Riedell WE, Reese RN (1999) Maize morphology and shoot CO2 assimilation after root damage 
by western corn rootworm larvae. Crop Sci 39:1332-1340 
Rosenberg MS, Adams DC, Gurevitch J (2000) MetaWin: Statistical software for meta-analysis. 
Version 2.0. Sinauer, Sunderland, Massachusetts, USA 
Saner MA, Müller-Schärer H (1994) Impact of root mining by Eteobalea spp. on clonal growth 
and sexual reproduction of common toadflax, Linaria vulgaris Mill. Weed Res 34:199-
204 
SAS Institute (2009) SAS/Stat. User's guide, Version 9.2. SAS Institute, Cary, N.C. 
Seastedt TR, Ramundo RA, Hayes DC (1988) Maximization of densities of soil animals by 
foliage herbivory - empirical evidence, graphical and conceptual models. Oikos 51:243-
248 
Stein C, Unsicker SB, Kahmen A, Wagner M, Audorff V, Auge H, Prati D, Weisser WW (2010) 
Impact of invertebrate herbivory in grasslands depends on plant species diversity. Ecology 
91:1639-1650 
Strauss SY, Agrawal AA (1999) The ecology and evolution of plant tolerance to herbivory. 
Trends Ecol Evol 14:179-185 
Tilman D (1989) Competition, nutrient reduction and the competitive neighborhood of a 
bunchgrass. Funct Ecol 3:215-219 
 32 
van Dam NM (2009) Belowground herbivory and plant defenses. Ann Rev Ecol Evol Syst 
40:373-391 
van Dam NM, Heil M (2011) Multitrophic interactions below and above ground: en route to the 
next level. J Ecol 99:77-88 
van der Putten WH (2003) Plant defense belowground and spatiotemporal processes in natural 
vegetation. Ecology 84:2269-2280 
Wise MJ, Abrahamson WG (2005) Beyond the compensatory continuum: environmental 
resource levels and plant tolerance of herbivory. Oikos 109:417-428 
Wolf A, Kozlov MV, Callaghan TV (2008) Impact of non-outbreak insect damage on vegetation 
in northern Europe will be greater than expected during a changing climate. Clim Change 
87:91-106  
Wurst S, van der Putten WH (2007) Root herbivore identity matters in plant-mediated 
interactions between root and shoot herbivores. Basic Appl Ecol 8:491-499 
Zvereva EL, Kozlov MV (2006) Consequences of simultaneous elevation of carbon dioxide and 
temperature for plant-herbivore interactions: a metaanalysis. Global Change Biol 12:27-41 
Zvereva EL, Kozlov MV (2010) Responses of terrestrial arthropods to air pollution: a meta-
analysis. Environ Sci Pollut R 17:297-311 
Zvereva EL, Lanta V, Kozlov MV (2010) Effects of sap-feeding insect herbivores on growth and 
reproduction of woody plants: a meta-analysis of experimental studies. Oecologia 
163:949-960 
 33 
Figure legends 
 
Fig. 1 Overall effects (mean Hedges’s d effect sizes) of root herbivory on below- and 
aboveground plant performance characteristics. Horizontal lines denote 95% confidence 
intervals; sample sizes are shown in brackets. The individual effect is significant if the 95% 
confidence interval does not include zero. Significant (P < 0.05) QB values indicate between-
group heterogeneity. 
 
Fig. 2 Effects of methodology on plant responses to root herbivory. For explanations, consult 
Fig. 1. 
 
Fig. 3 Effects of insect taxonomic characteristics (a), feeding guild and host plant specialisation 
(b) on plant responses to root herbivory. For explanations, consult Fig. 1. 
 
Fig. 4 Effects of plant characteristics on plant responses to root herbivory. For explanations, 
consult Fig. 1. 
 
Fig. 5 Individual and combined effects of root herbivory and environmental factors on plant 
performance. For explanations, consult Fig. 1; for significance of pairwise comparisons, consult 
text.
 34 
Effect size (ES)
0-0.5
Aboveground traits
Growth (225)
Aboveground traits (287)
Photosynthesis (17)
Sources of variation
(plant characteristics)
Root biomass (147)
-1
QB=1.67, df=2, P=0.43
Plant parts  
QB=53.1, df=1, P<0.00001
-1.5
Reproduction (41)
Size (210)
Number of organs (36)
Aboveground traits  
QB=0.31, df=1, P=0.58
Fig. 1
 
 
Effect size (ES)
0-0.5
Research domain
Method of manipulation
Agriculture (305)
Weed control (48)
Field (194)
Ecology (100)
Herbivore removal (23)
Infestation (367)
Simulated herbivory (63)
Sources of variation
(methodology)
Controlled (259)
-1
QB=21.6, df=2, P=0.0002
QB=7.22, df=2, P=0.03
Environment
QB=30.5, df=1, P<0.00001
-1.5
Low (128)
High (128)
Severity of treatment
QB=11.6, df=1, P=0.0007
Fig. 2
 
 35 
Effect size (ES)
0
Order
Species present in 
more than two papers
Diptera (13)
Coleoptera (227)
Cerambycidae (5)
Curculionidae (84)
Popillia japonica (24)
Phyllophaga spp. (10)
Sources of variation
(herbivore taxonomy)
Diaprepes abbreviatus (56)
-1
QB=7.41, df=3, P=0.06
QB=23.6, df=6, P=0.0006
Family of Coleoptera
QB=12.5, df=4, P=0.01
-2
Lepidoptera (19)
Hemiptera (10)
Scarabaeidae (80)
Chrysomelidae (46)
Elateridae (12)
Agapeta zoegana (8)
Sitona spp. (12)
Diabrotica virgifera (37)
Agriotes spp. (12)
Fig. 3a
 
Effect size (ES)
0-0.5
Feeding guild
External chewers (219)
Sap feeders (10)
Specialists (108)
Borers (41)
Sources of variation
(herbivore life history)
Generalists (161)
-1
QB=9.10, df=2, P=0.01
Specialisation
QB=21.8, df=1, P<0.00001
Fig. 3b
 
 36 
 
 
 
Effect size (ES)
0 1-1
Competition 
among plants
Nutrient supply
Root herbivory (20)
Root herbivory (34)
Competition (18)
Fertilisation (15)
Root herbivory+fertilisation (18)
Environmental factors
Aboveground herbivory (34)
-2
Aboveground 
herbivory
-3
Root herbivory+competition (20)
Root+aboveground herbivory (34)
Root herbivory (17)
Root herbivory (18)
Water regime  
Root herbivory+drought (17)
Drought (11)
-4
Fig. 5
 
Effect size (ES)
0 0.5-0.5
Life forms
Herbaceous plants
Woody (96)
Evergreen (70)
Herbaceous (228)
Herbs (94)
Grasses (123)
Sources of variation
(plant groups)
Deciduous (24)
-1
QB=9.15, df=1, P=0.002
QB=4.12, df=1, P=0.04
Woody plants  
QB=5.68, df=1, P=0.02
-1.5
Perennials (120)
Annuals (71)
Herbaceous plants
QB=0.56, df=1, P=0.45
Fig. 4