Are grazing increaser species better tolerators than decreasers? An experimental assessment of defoliation tolerance in eight British grassland species



    Corresponding author
    1. Department of Biological Sciences, Silwood Park, Imperial College London, Ascot, Berks. SL5 7PY, UK
      *Present address and correspondence: Ek del-Val, CASEB, Facultad de Ciencias Biológicas, Pontificia Universidad Católica de Chile, casilla 114-D, Santiago, Chile (tel. +56 2 686 2610; fax: +56 2 686 2621; e-mail:
    Search for more papers by this author

    1. Department of Biological Sciences, Silwood Park, Imperial College London, Ascot, Berks. SL5 7PY, UK
    Search for more papers by this author

*Present address and correspondence: Ek del-Val, CASEB, Facultad de Ciencias Biológicas, Pontificia Universidad Católica de Chile, casilla 114-D, Santiago, Chile (tel. +56 2 686 2610; fax: +56 2 686 2621; e-mail:


  • 1The importance of tolerance to herbivory for plant survival has long been recognized, but capacity for regrowth following recurrent defoliation has rarely been studied.
  • 2A glasshouse experiment was conducted with eight grassland species. We chose both species favoured under herbivory (increasers) and those that become scarce (decreasers) and manipulated the ontogeny of the plant when first clipped, as well as frequency (up to eight times) and intensity of clipping.
  • 3Plant survival was high (88%) but both frequency and intensity of clipping increased plant mortality in all species investigated (P < 0.05). Immature plants showed the highest mortality (15.9%).
  • 4Plant biomass mirrored mortality with immature plants again the most affected. As expected, control plants were largest and plants experiencing 100% biomass removal smallest. Clipping frequency was also significant, but the effect was not linear and plants that were clipped more than twice were more dramatically affected.
  • 5Biomass compensation showed species-specific responses that were significantly related to an a priori definition of species status in the grassland: increaser species had significantly higher compensation ability after recurrent defoliation than decreaser species.
  • 6Tolerance to herbivory has been viewed as a marginal strategy for plant survival, but our results suggest that it plays an important role in explaining abundance and distribution of plant species in environments with recurrent defoliation.
  • 7Plants do not always respond linearly to environmental pressures (such as repeated defoliation) and cannot therefore be predicted by studies considering only the lower end of the possible intensities. Understanding the effects of environmental pressures on plant fitness requires the study of a broader range of intensities, at different ontogenic stages and consideration of possible response thresholds.


Herbivory is one of the major selective forces acting on plants (Crawley 1983; Huntly 1991) and several studies have shown that herbivore grazing has a significant effect on the relative abundance and composition of species in plant communities (McNaughton 1979; Belsky 1987; Milchunas et al. 1988). In grasslands, the abundance of certain plant species is favoured by grazing (increasers) whereas other species become scarce (decreasers) as a consequence of herbivory (Crawley 1990; Bullock et al. 2001; Vesk & Westoby 2001). These changes can be generally explained by four broad mechanisms: herbivore preference (plant nutritive quality and plant defence mechanisms), tolerance to herbivory (plant capacity to regrow), disturbance caused by herbivores in the environment and alteration of the nutrient cycle, as well as by interactions between the different mechanisms (Crawley 1983, 1990; Augustine & McNaughton 1998; De Mazancourt et al. 1999).

Herbivore preference, which is related to plant resistance and plant phenological escape characters, has received much attention, and several chemical compounds and physical defences have been identified that can be linked with a decrease in herbivore impact (Feeny 1976; Rosenthal & Cates 1976; Coley et al. 1985; Mole 1994).

Tolerance is known to be important in many systems but, because it has been described relatively recently and because it is difficult to separate the underlying mechanisms from those related to other environmental constraints (Strauss & Agrawal 1999), has been less well studied.

Tolerance (defined as a fitness reaction norm of genotypes under a gradient of herbivory pressure; Tiffin & Rausher 1999) can be regarded as a capacity to compensate for tissue loss, and this regrowth is reflected in final biomass (McNaughton 1983; Augustine & McNaughton 1998). Intrinsic factors influencing tolerance include relative growth rate, reallocation of resources from roots to shoots, photosynthetic enhancement, plant architecture, the number of meristems available increase or decrease in root respiration, increase in nutrient uptake, associations with mycorrhizas and ontogeny (Chapin & Slack 1979; Bilbrough & Richards 1993; Tuomi et al. 1994; Marquis 1996; Strauss & Agrawal 1999). Extrinsic factors such as resource availability (Hochwender et al. 2000), timing of defoliation, amount of tissue removed and intensity of herbivory can also modify dramatically a plant's ability to respond after tissue removal (McNaughton 1983; Oesterheld & McNaughton 1991; Doak 1992). Maschinski & Whitham (1989) proposed a continuum of responses from under-compensation (when defoliated plants fail to achieve complete compensation) to over-compensation (defoliated plants are larger than undefoliated plants).

Perennial plants are very abundant in environments with constant herbivore pressure, and little is known about their capacity to compensate. Plant responses to intense repetitive herbivory or clipping are seldom studied (Tiffin 2000; but see Mullahey et al. 1991; Zhang & Romo 1994, 1995; Jacobs & Sheley 1995; Thornton & Millard 1996; McPherson & Williams 1998). Under constant, natural grazing, we would expect species’ tolerance thresholds to depend on defoliation frequency and defoliation intensity, in addition to any differences in herbivore preference. Responses after recurrent intense herbivory may therefore suggest which characters enable species to survive in grazed environments and thus assist their management.

Intense herbivore pressure (Crawley 1990; Bullock et al. 2001), rather than climate (Milchunas et al. 1988), favours the persistence of grasslands in Britain, and their species composition depends upon the intensity of grazing and type of herbivores.

In order to understand if tolerance to herbivory was related to species presence/abundance, we conducted a glasshouse experiment to test the regrowth capacity of eight British grassland species. Both increasers and decreasers under herbivory (sensuCrawley 1990) were included, i.e. species whose abundance is augmented at least 50% when herbivores are present or diminished by at least 50%, respectively (del-Val & Crawley 2004). Capacity to regrow (i.e. compensation/tolerance) was defined as the difference between the total standing biomass of clipped plants at the end of the experiment and that of unclipped plants (compensation is indicated by clipped plants attaining the same biomass as the controls). Our general hypothesis was that decreaser species would be unable to compensate for recurrent intense defoliation and would therefore be situated at the bottom of the continuum (sensuMaschinski & Whitham 1989), whereas increaser species would be able to regrow even if defoliated several times.

One intrinsic factor (ontogeny of the plant when first clipped) and two extrinsic factors (clipping frequency and clipping intensity) were manipulated in the experiment. Our experimental hypotheses were: (a) increaser species would have greater capacity for regrowth after defoliation than decreasers; (b) early ontogenic stages would be more vulnerable to defoliation than older ones, irrespective of species identity; (c) repeated defoliation would be more detrimental for a plant than a single event, and increaser species would be more tolerant than decreasers; and (d) removal of a large proportion of biomass removed (extreme defoliation) would be more detrimental than minor defoliation.

Materials and methods

Eight common British grassland species were used, four herbivore increasers: Senecio jacobaea, Trifolium repens, Rumex acetosella and Holcus lanatus; and four herbivore decreasers: Vicia sativa ssp. nigra, Lathyrus pratensis, Festuca rubra ssp. rubra and Achillea millefolium. These species grow at different densities in grazed grasslands where rabbits are very abundant and the intensity of herbivory is high throughout the year, and they are also known to respond differently to a single defoliation event (del-Val & Crawley 2004). Molluscs and herbivorous insects are also present in these grasslands and are known to attack the study species.

We used a factorial design with four variables: Species (n = 8), Ontogenic stage when first clipped (cotyledon, first-true leaf, immature (five-leaf) or mature (flowering for T. repens, the only species that reproduced during the experiment, 3 months old for other species)), Clipping intensity (50%, 75% or 100% of leaf area removed at each clipping) and Clipping frequency (one, two, four or eight occasions, although mature plants could not be clipped more than twice within the course of the experiment). A control (unclipped) treatment was also established for each species.

The experiment was performed in a 20 °C controlled temperature glasshouse with 12-h light/12-h dark cycle. Three individuals were used per species per treatment (n = 129 plants per species) except for V. sativa ssp. nigra and L. pratensis (two plants per combination, n = 86) for a total of 946 individual plants.

All species were grown from seeds purchased from Emorsgate Seeds (Kings Lynn, UK) except S. jacobaea, which was harvested from Silwood Park, UK (GR 241/944691). Vicia sativa ssp. nigra and L. pratensis seeds were scarified with sand paper to improve germination. All seeds were germinated in compost trays and then transplanted into individual 0.25-L plastic pots containing a 50% sand, 20% loam and 30% peat mix. Pots were arranged randomly and rearranged every other week. Plants were watered every other day.

All treatments were randomly assigned. Scissors were used to remove complete leaves until the required proportion of the total leaf area of the given plant had been clipped. When plants were clipped more than once, the time lag between clippings was 1 week. The experiment was set up in July 2001 and harvested 5 months later (end of November 2001). The recovery period permitted therefore depended on clipping frequency and ontogenic stage, ranging from 100 days for plants clipped once at cotyledon stage to 30 days for plants clipped eight times at cotyledon and one-leaf stage, and also for mature plants clipped twice.

At the end of the experiment, plant mortality was recorded and total standing biomass (above- and below-ground biomass) was measured as well as root : shoot ratios for all the surviving plants. Plant regrowth was calculated as final biomass for a given treatment minus that of the control. Plants were harvested, roots washed and all plant tissue was dried in an oven (80 °C) for 48 h and then weighed.

statistical analysis

Plant survival was analysed with a generalized linear model with a binary response variable (dead or alive) and binomial errors. Of the four explanatory variables used, clipping intensity and clipping frequency were treated as continuous (1 d.f. each), and species (7 d.f.) and ontogenic stage (4 d.f.) were defined as categorical. Total standing biomass (biomass at the end of the experiment), above-ground biomass and root biomass were analysed with anovas using the same explanatory variables. For these analyses biomass data were log-transformed to achieve a constant variance and dead plants were excluded. The number of days between a plant's last clipping and its harvest date (i.e. its recovery period) was included as a covariate (continuous) in the model. All terms were included in the model as fixed variables. Because of the nonlinear behaviour of biomass in relation to clipping intensity, both variables (explanatory and response) were log-transformed and a piecewise regression was carried out in order to determine a threshold above which compensation was not complete (Crawley 2002). The power of using a piecewise regression instead of a nonlinear model is that it enables one to find the exact threshold where the response variable changes. Root : shoot ratios were also log-transformed and analysed using the same explanatory variables as above. After fitting the full model, any non-significant interaction term (P > 0.05) was removed stepwise and the model rerun to check that the removal was not significant. The resulting simplified model contained only main effects and significant interaction terms. In order to test our hypothesis that increaser species would be better able to compensate after herbivory we performed a second set of analyses including ‘status’ (increaser/decreaser) in the models and removing ‘species’. The response variables (total standing biomass, above-ground biomass, root biomass and root : shoot ratio) were all log-transformed for these analyses; plant survival was analysed using binomial errors. All analyses were performed in S-PLUS 2000 (MathSoft, Inc.).



Species showed similar mortalities (there were no significant species differences in death rate; Fig. 1a, Table 1). No unclipped control plant of any species died but clipping significantly increased plant mortality, especially for plants with 100% removal (Fig. 1d,e, Table 1). Plant survival was significantly affected by the recovery period (Fig. 1b) and the ontogenic stage at which plants were first clipped (Fig. 1c). Mortality was low in mature plants but one-leaf (14.8%) and immature stages were more likely to die (c. 15% vs. 1.5%). Percentage biomass removal treatments also had a significant effect and, as expected, plants with 100% biomass removed showed the highest mortality (30.4%). Clipping frequency also increased mortality (Fig. 1e). There was a significant interaction between ontogenic stage and clipping intensity (Table 1): mature and one-leaf stage plants only showed mortality at 100% defoliation whereas some cotyledon and immature stage plants also died at 50% and 75% defoliation.

Figure 1.

Treatment main effects on plant survival. (a) Species (am: Achillea millefolium, fr: Festuca rubra ssp. rubra, hl: Holcus lanatus, lp: Lathyrus pratensis, rx: Rumex acetosella, sj: Senecio jacobaea, tr: Trifolium repens, vs: Vicia sativa ssp. nigra); χ2 = 9.95, d.f. = 7, n = 129 except in lp and vs where n = 86, species effect was not significant. (b) Recovery period (days), χ2 = 6.93, d.f. = 1, P = 0.08. (c) Ontogenic stage, χ2 = 59.1, d.f. = 4, P << 0.001. (d) Clipping intensity, χ2 = 272.25, d.f. = 1, P << 0.001. (e) Clipping frequency, χ2 = 80.29, d.f. = 1, P << 0.001. (f) Ontogenic stage–status interaction, χ2 = 10.85, d.f. = 4, P = 0.03 Numbers in parentheses represent sample size (n).

Table 1.  Survival analysis showing all main effects (treatments applied) and the only significant interaction. Data were analysed as a binary response (dead or alive) with binomial error
Recovery period1944  6.93   0.008
Species7937  9.95   0.19
Ontogenic stage4933 59.1<< 0.001
Clipping intensity1932272.25<< 0.001
Clipping frequency1931 80.29<< 0.001
Ontogeny × intensity3928  8.1   0.04

total standing biomass

Species had different total standing biomass at the end of the experiment (Table 2). Ontogenic stage was also an important factor for plant regrowth, with the immature stage most affected (Fig. 2a). As expected, percentage biomass removed was significant: control plants were the largest and 100% biomass removed plants the smallest (Fig. 2b). Clipping frequency (i.e. the number of times biomass was removed) was also significant, but the effect was not linear. Piecewise regression showed that plants suffered a disproportionately greater biomass loss above a threshold of twice-clipped (Fig. 3a). Although clipping frequency was significant for all species, some (R. acetosella and T. repens) were less affected than others (A. millefolium, S. jacobaea and F. rubra ssp. rubra) (species × clipping frequency P << 0.001, steeper slopes in Fig. 3b). To exclude the number of days elapsed since last clipping as a cause for the differences between treatments, this variable was incorporated into the models. Not surprisingly, this recovery period had a significant effect on plant biomass, and plants were larger when the time elapsed was greater. Nevertheless, the remaining variables explained a significant amount of the variation and the significant effects of clipping frequency, intensity and ontogeny can therefore be regarded as independent of the recovery period. There was also a significant interaction between clipping intensity and clipping frequency, with plants disproportionately affected when high percentages of biomass were clipped four and eight times (Fig. 4a). Ontogenic stage at first defoliation and clipping intensity also showed a significant interaction (Fig. 4b); cotyledon and immature plants were more vulnerable than mature or one-leaf stages at the highest defoliation rates (mature plants could not be clipped more than twice before the experiment terminated and were therefore excluded from these analyses). One three-way interaction was also significant, with the immature ontogenic stages studied being the most affected at the highest levels of clipping intensity and clipping frequency.

Table 2.  Analysis of variance for the different experimental treatments on total standing biomass, above-ground biomass, root biomass and root : shoot ratio. Only the main effects, the covariate (recovery period) and significant interactions are shown. For the minimal model the response variables were log-transformed. ∼ indicates interactions taken out from the model because they were not significant
SourceTotal standing biomassAbove-ground biomassRoot biomassRoot : shoot ratio
Recovery period  1300.53<< 0.001  1685.4<< 0.001  1149.82<< 0.001  1356.99<< 0.001
Species  7 60.48<< 0.001  7 70.21<< 0.001  7106.69<< 0.001  7180.41<< 0.001
Ontogenic stage  4121.96<< 0.001  4 39.58<< 0.001  4142.19<< 0.001  4 51.96<< 0.001
Clipping intensity  1229.49<< 0.001  1234.24<< 0.001  1200.87<< 0.001  1  9.65   0.002
Clipping frequency  1114.24<< 0.001  1111.68<< 0.001  1122.31<< 0.001  1 13.02<< 0.001
Species × ontogeny 28  1.92   0.003 28  2.28 < 0.001 28  2.31 < 0.001 28  4.34<< 0.001
Species × clipping frequency  7  3.95  < 0.001  7  2.43   0.02  7  7.39<< 0.001  7  2.59   0.01
Species × clipping intensity  7  2.31   0.02  7  1.22   0.3
Ontogeny × clipping intensity  3  5.43   0.001  3  4.63   0.003  3 15.12<< 0.001
Ontogeny × clipping frequency  3  2.62   0.05  3  7.29   0.001  3 16.37<< 0.001
Clipping intensity × clipping frequency  1 48.85<< 0.001  1 20.93   0.001  1 20.77<< 0.001  1  9.44   0.002
Species × intensity × frequency  7  2.03   0.05  7  1.66   0.12
Species × ontogeny × frequency 21  2.44<< 0.001
Species × ontogeny × intensity 21  3.36<< 0.001
Ontogeny × frequency × intensity  3  3.63  0.01  3  2.81   0.04
Species × ontogeny × frequency × intensity 21  2.81<< 0.001
Error773  765  779  719  
Figure 2.

(a) Effects of clipping at different ontogenic stages on mean total standing biomass (averaged across species, ± 1 SE) and on root and above-ground biomass (sample size shown in parentheses). (b) Effect of different clipping intensities (measured as percentage leaf area removed per clipping treatment) on total standing biomass. Values shown are the log(final biomass) per plant for all species across all timings and ontogenic stages. The model is represented by the line: y = 1.12 − 0.01x; F1,790 = 754.08, R2 = 0.03, P << 0.001.

Figure 3.

Impact of clipping frequency (log scale) on total standing biomass. (a) Piecewise regression showing the existence of a threshold above more than two clippings. Points shown are final values of log(total standing biomass) across all timings and ontogenic stages. The two lines represent (1) minimal model for zero to twice clipped (y = 0.76 − 0.04x), (2) minimal model for greater than twice clipped (y = 1.23 − 1.06x). (b) Impact of clipping frequency per species. Upper panels show increaser species and lower panels show decreasers. Note the species slope differences. Points represent final values of log(total standing biomass) and the lines represent linear regressions fitted to clipping frequency for each species. Species abbreviations are as in Fig. 1. Species–clipping frequency interaction: F7,791 = 5.19, P << 0.001.

Figure 4.

Interaction effects between: (a) clipping intensity (percentage of leaf area removed) and clipping frequency (number of clipping episodes) on log(total standing biomass) – lines shown are the predictions from the model considering all ontogenic stages across species, clipping intensity–clipping frequency interaction: F1,790 = 73.79, P << 0.001; and (b) ontogenic stage and clipping intensity – lines shown consider all species and all frequencies, ontogenic stage–clipping intensity interaction: F3,790 = 5.95, P < 0.001.

above-ground biomass

Results from above-ground biomass were similar to those of total standing biomass (all main effects significant, Table 2). Species had different final above-ground biomass regardless of intensity and frequency of damage. Plants first clipped when mature were most severely affected (0.55 ± 0.06 g), whereas those first clipped at the cotyledon stage were the least affected (0.84 ± 0.06 g). As expected, repeated and intense clipping led to more damage (reduced final above-ground biomass). The number of days between final clipping and harvesting was also significant. Examination of the interaction terms revealed that species differed in their reaction to clipping frequency and clipping intensity. Lathyrus pratensis was less affected at the higher frequencies of clipping (four and eight times). There was an interaction between ontogenic stages and clipping frequency: immature and mature plants were more affected at high clipping intensities than at other ontogenic stages. The interaction between clipping intensity and frequency was also significant, with intense repeated defoliation leading to a smaller above-ground biomass.


Root biomass differed between species (Table 2). Ontogenic stage, clipping intensity and clipping frequency were also significant, with species and ontogenic stage explaining most of the variation (27.7% and 21.1%, respectively). The recovery period was also important for final root biomass. The interaction between species and clipping frequency was significant, with T. repens and R. acetosella being the least affected. Clipping frequency also interacted with species such that average root biomass decreased with repeated defoliation, but T. repens and R. acetosella did not change much at higher clipping frequencies. Ontogenic stage again interacted with clipping intensity: cotyledon and immature plants were more affected than mature and one-leaf stage plants, particularly at the highest levels of biomass removal. Clipping intensity and clipping frequency also showed a significant interaction, and plants clipped four and eight times at the highest percentages of biomass removal were disproportionately affected.

root : shoot ratio

Main effects on root : shoot ratios were dominated by differences between species and ontogenic stage (42.1% and 6.9% variation explained; Table 2, Fig. 5). Ratios were greater when plants were first clipped as mature individuals, except for T. repens and R. acetosella, in which root : shoot ratios were unaffected by clipping (Fig. 6a). The root : shoot ratio increased as the percentage of foliage removed from mature plants increased, but did not change for cotyledon, one-leaf and immature plants (Fig. 6b). Species’ root : shoot ratios responded differently to changes in clipping frequency, although, overall, cotyledon and one-leaf stage plants were not affected whereas immature plants had a greater root : shoot ratio when clipped repeatedly (Fig. 6c). Recovery period was also significant for the root : shoot ratio.

Figure 5.

Root : shoot ratios in (a) different species – values shown are mean log(ratio) ± 1 SE of control plants (F7,724 = 154.85, P << 0.001), and (b) different ontogenic stages – values shown are mean log(ratio) ± 1 SE across species considering all clipping frequencies and different clipping intensities (F7,724 = 76.33, P << 0.001). Species abbreviations and sample sizes are as in Fig. 1.

Figure 6.

(a) Species root : shoot ratios with respect of ontogenic stage at first clipping. Values shown are mean log(ratio) ± 1 SE across all frequencies and intensities of defoliation. Ontogenic stage–species interaction: F28,724 = 3.36, P << 0.001. Increasers are represented by filled symbols and decreasers by open symbols. Species abbreviations and sample sizes are as in Fig. 1. (b) Changes in root : shoot ratio at different ontogenic stages with respect to clipping intensity (percentage of biomass removed; F3,724 = 12.62, P << 0.001). (c) Changes in root : shoot ratio with clipping frequency (number of clipping episodes; F3,724 = 16.41, P << 0.001); mature plants were not clipped four and eight times and therefore values for these plants are not shown. Values shown are mean log(ratio) ± 1 SE across species.


Species status (increaser/decreaser), as defined a priori, was not significant as a main effect for plant survival, but there was significant interaction between status and ontogenic stage (Table 3). Decreaser species had greater mortality at the immature stage (Fig. 1f). Total standing biomass was higher for herbivore increaser species (2.3 ± 0.09 g) than for decreasers (1.57 ± 0.07 g), and this was also true for above-ground biomass (0.93 ± 0.05 g vs. 0.52 ± 0.03 g), root biomass (1.37 ± 0.05 g vs. 1.05 ± 0.06 g) (Table 4) and root : shoot ratio (3.6 ± 0.64 vs. 2.79 ± 0.2) (Fig. 5a). The main difference between increasers and decreasers was the reaction towards clipping frequency, decreasers being less able to compensate at the highest clipping frequency (Fig. 7).

Table 3.  Survival analysis considering status (increaser/decreaser), showing all main effects and one significant interaction. Data were analysed as a binary response (dead or alive) with binomial error
Recovery period1944  6.93   0.008
Status1943  1.24   0.27
Ontogenic stage4939 54.5<< 0.001
Clipping intensity1938255.08<< 0.001
Clipping frequency1937 88.5<< 0.001
Status × ontogeny4933 10.85   0.03
Table 4.  Analysis of variance considering status (increaser/decreaser) for the different experimental treatments on total standing biomass, above-ground biomass, root biomass and root : shoot ratio. Only the main effects, the covariate (recovery period) and significant interactions are shown. For the minimal model the response variables were log-transformed, with interactions marked ∼ taken out because they were not significant
SourceTotal standing biomassAbove-ground biomassRoot biomassRoot : shoot ratio
Recovery period  1212.1<< 0.001  1483.47<< 0.001  177.68<< 0.0011123.08<< 0.001
Status  1 98.64<< 0.001  1139.94<< 0.001  146.87<< 0.0011  5.84   0.02
Ontogenic stage  4 85.51<< 0.001  4 37.91<< 0.001  466.52<< 0.0014  9.59<< 0.001
Clipping intensity  1146.54<< 0.001  1168.48<< 0.001  188.24<< 0.0011  7.51   0.006
Clipping frequency  1 91.68<< 0.001  1 39.03<< 0.001  193.12<< 0.0011  4.32   0.04
Status × ontogeny
Status × clipping frequency  1  5.06   0.02
Status × clipping intensity
Ontogeny × clipping intensity  3  3.95   0.008  3  5.54<< 0.001
Ontogeny × clipping frequency  3  5.5< 0.001  3  5.24   0.001
Clipping intensity × clipping frequency  1 33.17<< 0.001  1 15.11< 0.001  112.14< 0.001  1  5.55   0.02
Status × intensity × frequency
Status × ontogeny × frequency
Status × ontogeny × intensity
Ontogeny × frequency × intensity
Error819  820  823  817  
Figure 7.

Interaction effect between status (increaser/decreaser) and clipping frequency (number of clipping episodes) on total standing biomass (mean log(total standing biomass)) ± 1SE. Lines represent predictions from the model considering all ontogenic species across clipping intensities. Status–clipping frequency interaction: F1,819 = 5.06, P = 0.02.


The aim of this study was to assess species differences in regrowth following defoliation. As expected, plant mortality increased with frequency and intensity of defoliation. Species-specific responses in terms of biomass compensation were significantly related to an a priori definition of species status in the grassland: increaser species (S. jacobaea, T. repens, R. acetosella and H. lanatus) had significantly higher compensation ability after defoliation than decreaser species (F. rubra susbp. rubra, A. millefolium, V. sativa ssp. nigra and L. pratensis; Fig. 7). This contrasts with the suggestion from some studies that increasers are intolerant to defoliation, with their enhanced biomass due instead to avoidance by herbivores (because of their defences) and reduced competition from the more palatable, faster growing species preferred by herbivores (Pacala & Crawley 1992).

All species showed biomass compensation under low clipping frequency (after one or two defoliation events; Fig. 3a). Although the effect of clipping frequency was significant for all species, species response curves were different (i.e. there was a significant interaction between species × clipping frequency; Fig. 3b). Three increaser species (R. acetosella, T. repens and H. lanatus) showed higher compensation ability after repeated clipping (four episodes). Compensation following a single experimental defoliation cannot therefore be easily extrapolated to responses to recurrent herbivory in the field, suggesting that previous estimates of herbivore tolerance from experiments using a single defoliation should be treated with caution. The dramatic decrease in biomass of the most affected species (the legume V. sativa ssp. nigra.) is likely to result in a reduction of fitness.

We found a continuum, from complete compensation to under-compensation, in responses to biomass removal, as predicted by Maschinski & Whitham (1989). Most species were able to compensate fully for low levels of biomass removal, as shown in other studies (Belsky et al. 1993; Crawley 1997; McPherson & Williams 1998) and we therefore expected to find a threshold for compensation below which herbivory is not detrimental for plants.

Such thresholds are well known: Datura stramonium can only fully compensate for 10% defoliation (Fornoni & Nuñez-Farfán 2000) and Epilobium latifolium is able to regrow after low intensities of herbivory, even at high frequency, but cannot compensate for infrequent, but intense, biomass removal (Doak 1992). Vaccinium myrtillus can achieve complete compensation at 50% defoliation (Tolvanen et al. 1994) whereas Purshia tridentata can compensate after 100% defoliation (Bilbrough & Richards 1993). We did not, however, find a clear threshold for clipping intensity (i.e. the amount of biomass removed): the highest clipping intensities were more detrimental and showed higher variance. One outstanding feature is that all plants survived after two events of 100% defoliation with only small losses in final biomass (Fig. 4a).

Contrary to our original hypothesis that earlier ontogenic stages would be the most affected by defoliation, the most critical phase for all species (in terms of mortality, total standing biomass and root biomass) was the immature stage, when increasing levels of defoliation and clipping frequency translated into disproportionately greater damage. At this particular stage, plants have probably exhausted their seed reserves, but are not yet completely established and cannot therefore obtain all the nutrients required for compensation. In another study with repeated defoliation of the perennial Prosopis glandulosa, Weltzin et al. (1998) also found that old seedlings were relatively more vulnerable to defoliation than young seedlings. Other studies have found different responses depending on the ontogenic stage at defoliation: Ipomoea hederacea was able to compensate completely when mature leaves were damaged but not when cotyledons were damaged (Stinchcombe 2002); Gentianella campestris only compensated when clipped during intermediate stages (Lennartsson et al. 1998); Lupinus chamissonis compensated as a mature plant but not as juveniles (Warner & Cushman 2002); and Cucumis sativus preflowering plants were able to compensate for up to 80% defoliation whereas flowering plants were unable to do so (Thomson et al. 2003). As expected, our mature plants were the least affected: older leaves may be less valuable to the plant (Gold & Caldwell 1990) or the logistic growth shown by most plants may mean that mature individuals could be released from self-limitation by clipping and their growth rate could thus be enhanced (Weis et al. 2000). Although all ontogenic stages were more affected when a greater percentage of biomass was removed, cotyledon and immature plants were more damaged (steeper slope) than one-leaf and mature plants for all eight species (three-way interaction was not significant), reinforcing the proposition that the immature stage is particularly vulnerable. Other investigations that studied community responses to defoliation have also found that timing is a key issue for plant compensation. Biomass production in a northern wheat grass community and in Fescue prairie was less affected if grazing was delayed (Zhang & Romo 1994; Bogen et al. 2003), whereas plants growing in early grazed plots in Yellowstone National Park, USA, allocated more resources to reproduction (Anderson & Frank 2003). All the evidence suggests that the efficiency for regrowth is contingent upon which ontogenic stage species were eaten.

Tolerance to herbivory has been predicted to be an important strategy for plants with high growth rates (Hilbert et al. 1981), with large storage organs and an enhanced ability for resource reallocation, with numerous meristems or able to increase net photosynthetic rate after damage (Chapin & Slack 1979; Whitham et al. 1991; Trumble et al. 1993; Mole 1994; Tuomi et al. 1994; Strauss & Agrawal 1999; Tiffin 2000). In this study, R. acetosella, H. lanatus and S. jacobaea were able to reallocate resources from root to shoot, as they maintained their root : shoot ratio no matter how heavily their shoots were clipped. Perkins & Owens (2003) also found that grass seedlings are able to sustain repeated defoliation maintaining root : shoot ratios. Compensation by T. repens may have involved activation of dormant meristems released by the clipping of shoot biomass (Burdon 1983).

The response to repeated defoliation was associated with species status (increaser/decreaser). The highly tolerant species in terms of biomass regrowth and low mortality (R. acetosella, H. lanatus and T. repens) are able to persist even at high clipping intensity and repeated defoliation, and this capacity may be related to their persistence in heavily grazed areas. Senecio jacobaea, which has been previously shown to be tolerant (Islam & Crawley 1983; van der Meijden et al. 2000), was also able to compensate for high levels of defoliation (up to 75%) but could not compensate for repeated clipping, suggesting that it would not persist in an area if it was clipped/grazed several times (in the wild it is avoided by larger herbivores such as horses and cattle). The species categorized as herbivore decreasers (V. sativa ssp. nigra, L. pratensis, F. rubra ssp. rubra and A. millefolium) were the most affected, suggesting that their inability to regrow under repeated defoliation is correlated with the fact that they are significantly reduced when large herbivores are present in a system. We realize that tolerance to herbivory is not the only mechanism explaining plant presence/abundance in a particular system and that tolerance and defence mechanisms are not mutually exclusive (Mauricio et al. 1997; Almeida-Cortez et al. 1999; Tiffin & Rausher 1999; Gadd et al. 2001), but we suggest that it should always be considered as a possible contributory factor. Hendrickson & Berdahl (2002) also found that a grass sensitive to grazing was less able to compensate for repeated clipping than a grazing-tolerant species.

We wanted to associate species’ status in a grazed habitat with their capacity to regrow (i.e. correlate community and individual levels of organization), as increaser species have been shown to have higher compensation capacity (Brown & Allen 1989). Clipping frequency dramatically altered the regrowth abilities of British grassland species, suggesting that consideration of the effects of frequency and timing of defoliation is important for understanding plant responses to herbivory. Results from these widespread species may be applicable to grazed habitats in other countries. We suggest that the detrimental impact of repeated herbivory may not be directly extrapolated from results obtained from single defoliation experiments.


We thank C. de Mazancourt, R. Keane and M. Rees for many contributions to this study. We thank K. Boege, M. Bonsall, J. Fornoni, R. Norby and several anonymous referees for their comments on earlier drafts. Funding resources for EdV were provided by Conacyt; this manuscript was written with the support of CASEB (Fondecyt-Fondap 1501-0001), Pontificia Universidad Católica de Chile.