Linkages between plant litter decomposition, litter quality, and vegetation responses to herbivores


§Author to whom correspondence should be addressed. E-mail:


  • 1There is increasing awareness that similar suites of plant traits may govern foliage palatability and litter decomposability, but whether there is an association between the response of vegetation to herbivory and litter decomposition rates across plant species remains unexplored.
  • 2We collected 141 samples of litter from 59 understorey and 18 canopy tree species from a total of 28 sites under natural forest throughout New Zealand. We assessed whether variables related to decomposition and quality of litter of the understorey species showed a statistical relationship with the response of vegetation density (assessed using a pole-intercept method) of the same species at the same locations to browsing by deer and goats. Decomposition and nutrient-loss data from litter were obtained using standardized laboratory bioassays.
  • 3There was a significant positive correlation between litter decomposition rate and the extent to which vegetation density was reduced by browsing mammals (r = 0·488, P < 0·001). Further, decomposition rate and vegetation response to herbivory were both correlated with several of the same litter quality variables.
  • 4The proportion of total initial phosphorus and nitrogen released from litter during decomposition was correlated with litter decomposition rate, but not with vegetation density response to browsing. This suggests that effects of browsers on vegetation composition are more likely to influence ecosystem carbon flow than nitrogen or phosphorus flow.
  • 5Litter-mixing experiments showed that good quality litters produced by plant species reduced by browsers tended to promote the decomposition of other litters. Meanwhile, poor quality litters from species promoted by browsers tended to decompose more rapidly when mixed with other litter types than when by themselves. However, these effects were weak and likely to be less important than the more direct effects of browsing mammals on vegetation composition.
  • 6The relationships between litter decomposition and effect of herbivory on vegetation density were driven primarily by differences among the main plant functional groups, which showed the same decreasing rank order for both variables: large-leaved dicots, small-leaved dicots, Nothofagus, ferns, and monocots.
  • 7The implications of these results for understanding how herbivores affect the decomposer subsystem are considered. Because the results of this work are only partially consistent with those of an earlier study on how browsers affect decomposer organisms and processes, conducted at the same 28 field sites, other mechanisms through which browser effects are manifested below-ground must often override that investigated in this study.


Most of the carbon that plants fix and allocate to their above-ground tissues is eventually consumed by herbivores or decomposer organisms. The nature of direct linkages between plants and above-ground herbivores, and those between plants and decomposers, has been studied extensively. There is an increasing recognition that important plant-mediated interactions also occur indirectly between above-ground herbivores and decomposers. Decomposer activity influences plant nutrient acquisition, and this in turn affects the population growth of invertebrate herbivores (Scheu, Theenhaus & Jones 1999). Further, foliar herbivory can exert a variety of effects on decomposer organisms and processes by influencing the quantity and quality of resources that become available to decomposers (reviewed by Bardgett, Wardle & Yeates 1998b; Wardle 2002).

Plant species that are preferentially browsed by herbivores differ from those that are not with regard to key ecophysiological traits (Coley, Bryant & Chapin 1985; Grime 2001). These traits relate to the quality of the foliage and litter produced by the plant (Coley et al. 1985; Grime & Anderson 1986). If this is true, then foliage palatability should be positively correlated with litter decomposability across species. Evidence for this comes from invertebrate herbivore feeding tests and litter decomposition bioassays (Cornelissen et al. 1999; Grime et al. 1996; Wardle et al. 1998).

Herbivores influence plant community structure, and may promote domination by either unpalatable or palatable species, depending upon whether the former are suppressed (Bryant 1987; Pastor et al. 1988) or promoted (Brown, Jepsen & Gibson 1988; McNaughton 1979) by herbivory. If foliage palatability, litter quality and decomposability are correlated across species, herbivore-induced changes in the relative balance of palatable and unpalatable plant species should be matched by changes in the net quality of litter produced, and therefore in decomposer activity. Pastor et al. (1988, 1993) found browsing by moose (Alces alces L.) in boreal forest to induce replacement of deciduous palatable species by unpalatable spruce (Picea) species, and proposed that the corresponding decline in several indicators of soil quality was due to the poorer litter quality of spruce. However, no study has determined if this pattern is a general trend, or whether there are predictable or consistent relationships across plant species between decomposability of leaf litter and the response of density of above-ground vegetation to herbivory.

In the present study, we sought explicitly to test for relationships between measures of leaf litter quality, decomposability of leaf litter, and plant species responses to herbivory in field conditions. This was achieved by collecting litter from a range of plant species in several locations throughout New Zealand, and comparing measures of litter decomposability and quality with the response of vegetation density of the same plant species at the same location to long-term exclusion of browsing mammals. The ultimate goal was to determine whether there are predictable effects of foliar herbivory on the quality of litter produced, and to assess what the ecosystem-level implications of these might be.

Materials and Methods

The Study System and Vegetation Responses to Herbivory

Wardle et al. (2001) reported the results of a study in which 30 fenced exclosure plots, mostly approximately 20 × 20 m and established by the former New Zealand Forest Service between 1961 and 1984, were sampled. These exclosures are located across most of New Zealand's major natural forest types, and were designed to exclude introduced browsing mammals [notably red deer (Cervus elaphus scoticus Lönnberg) and feral goats (Capra hircus L.)]. Measurements of various above- and below-ground variables, including vegetation density of each plant species, were made inside and outside each exclosure. Each exclosure location was sampled once, between 26 September 1997 and 26 May 1999. The present study focuses on 28 of these locations (Appendix 1).

For each location, we selected up to six plant species (Appendix 1) which were abundant in the understorey browse layer (0–2 m vegetation height) according to data presented by Wardle et al. (2001). For each species fresh leaf litter was collected, usually off the forest floor, or in some cases when still attached to the plant stem. Litter was also collected from up to three dominant canopy tree species at each location (Appendix 1). Litter for the understorey (browse layer) species was used to answer questions about relationships between herbivore effects and decomposition variables; that from the canopy species was collected for comparative purposes. This litter was air-dried and stored until further use.

For each location, the response of vegetation density of each of the selected browse layer species to herbivory by browsing mammals was determined using the densities of that species both inside and outside the exclosure plot (Wardle et al. 2001). These were was used to derive an index, V (Wardle 1995), which ranges from −1 to +1 and becomes increasingly negative and positive as the mammals have an increasingly positive or negative effect, respectively, on the browse layer vegetation density of that species (a value of 0 indicates no effect).

Litter Quality Measurements

Subsamples of each litter sample were analysed for the concentration of several chemical components. Concentrations of phenolics and condensed tannins were determined colorimetrically following extraction with 50% acetone for 16 h; the procedure for phenolic measurements follows Price & Butler (1977); and that for condensed tannin measurements follows Broadhurst & Jones (1978; see also Mole & Waterman 1994). Phenolic concentrations were determined as tannic acid equivalents and condensed tannins as catechin equivalents. Lignin, cellulose and fibre fractions were determined using the acid-detergent fibre–sulphuric acid procedure (Rowland & Roberts 1994). Briefly, acid-detergent fibre (hereafter called ‘fibre’) was determined as the fraction remaining after treatment of the litter with boiling acid detergent to hydrolyse protein. This fibre residue was then treated with 72% sulphuric acid to destroy the cellulose; cellulose is defined as the difference between the initial and remaining fractions. Finally, the residue was ignited at 550 °C to destroy all remaining organic matter, leaving only the inorganic (ash) fraction; acid-detergent lignin (hereafter called ‘lignin’) is defined as the difference in this residue between pre- and post-ignition. Total nitrogen (N) and phosphorus (P) were determined using automated colorimetric methods (Technicon Instruments 1977) after digestion. Values of ratios of lignin to N, lignin to P, and N to P were derived from the above measurements.

Litter Decomposition Measurements

The decomposability of each litter collected was assessed using a standardized laboratory bioassay (Wardle et al. 1998). For each litter type, three 9 cm diameter Petri dishes were each two-thirds filled with a standardized humus substrate (1·8% N, pH 3·9; collected from a Metrosideros umbellata Cav.–Weinmannia racemosa L. f. forest near Otira, New Zealand; 42°50′ S, 171°37′ E) and amended to 250% moisture (dry weight basis); a disc of nylon mesh with 1 mm holes was placed on the humus surface. Litter (1 g, oven-dried), cut into 5 mm fragments, was placed on the surface of the mesh of each Petri dish; the dish was then sealed with tape to minimize water loss and incubated for 92 days at ≈22 °C. In addition, to assess interactive litter effects among species from each location when mixed together, Petri dishes were set up in triplicate for each of the possible two-way mixtures of litter for all the species sampled from that location. These were set up as for the monospecific Petri dishes, but with each dish containing 0·5 g of each of the two litter types.

Upon harvest, all remaining (undecomposed) litter was removed from each Petri dish and rinsed. For the Petri dishes containing two species mixtures of litter, the remaining litter was visually sorted into the component species. All litter was then oven-dried (80 °C, 24 h), and its dry mass determined. For all the litters that were decomposed in monoculture, N and P concentrations of the remaining litter were determined as described above.

Litter decomposition rate was determined as the percentage mass lost during incubation. Loss of N and P from the litter was calculated as the total mass × nutrient concentration prior to incubation minus that after incubation. The proportion of total initial N and P lost from the litter during decomposition was calculated from these values. The percentage mass loss of each litter type in each two-way mixture was compared with its percentage mass loss in monoculture. These values were used to determine two data values for each litter type. First, the net effect of each litter type on the decomposition rate of all the other litter types from that location was calculated as the mean enhancement of all other litter types by that species in mixture relative to their decomposition in monoculture. Second, the net response of decomposition rate of each litter type to all the other litter types from the same location was calculated as the mean enhancement of decomposition rate of that species in all mixtures in which it occurred relative to its decomposition rate in monoculture.

Data Analysis

For data analysis, each separate litter collection was considered as a separate data point. This yielded 98 points representing 51 different species in the understorey (browse) layer and 43 points representing 18 species in the tree canopy layer. Confounding problems due to non-independence among samples are unlikely because, for the response variables measured, variability among species within locations was frequently as great as among locations (indicative of considerable phenotypic plasticity among species), and variation across locations within species was often as great as among species. All variables were tested for normality and transformed if necessary. Correlation analysis was used to evaluate relationships for the browse layer species between decomposition variables, litter quality variables, and vegetation density response to browsing. Backward stepwise multiple regression was employed to determine, for the browse layer species, which combinations of litter quality variables served as best predictors of decomposition variables and the response of vegetation density to herbivory. One-way anova was used to test whether there were differences in litter quality and decomposition variables among plant species with the greatest density outside and inside the exclosures.

The data were also used to compare different plant functional groups. All plant species were assigned to one of six groups (Appendix 1): ferns; gymnosperms; monocots; large-leaved dicots, leaf lamina width >6 mm (excluding Nothofagus); small-leaved dicots, leaf lamina width <6 mm; and members of Nothofagus. Large-leaved dicots were distinguished from small-leaved dicots because many of the small-leaved species have different growth forms (such as divarication) to the large-leaved species, and have been proposed as having a stronger resistance to browsing (Greenwood & Atkinson 1977; McQueen 2000; but see Howell, Kelly & Turnbull 2002). Nothofagus was distinguished from other large-leaved dicots because Nothofagus represents the only large-leaved ectomycorrhizal genus, and Nothofagus species form highly competitive, low-diversity (often monospecific) stands relative to those dominated by other dicotyledonous species (Leathwick, Burns & Clarkson 1998). The gymnosperm group was excluded from analysis on the basis of too few data. One-way anova was used to assess whether decomposition variables, litter quality variables and vegetation density response to browsing differed significantly among the five remaining groups. In addition, correlation and stepwise correlation analyses were employed as described above, but to explore relationships between decomposition variables, vegetation response to browsing and litter quality variables within each of the five groups.


When all understorey species × site combinations were considered, there was a highly statistically significant relationship between the rate of decomposition and the response of vegetation density to browsing for the same plant species (Fig. 1a). These two variables were correlated with several of the same resource quality properties – concentrations of condensed tannins, fibre and lignin, and ratios of lignin to N and of lignin to P (Table 1). The variables in the multiple regression relationship that explained the most variation for the response of vegetation density to herbivory, concentration of phenolics and of fibre, were also two of the four variables that occurred in the regression relationship which best predicted litter decomposition rate (Table 2).

Figure 1.

Relationships between relative litter decomposition rates for a range of plant species, and the relative responses of these species to herbivory by browsing mammals. Plant responses to herbivores are measured using the metric V (Wardle 1995; Wardle et al. 2001) which ranges from −1 to +1 and becomes increasingly negative and positive as vegetation density in the browse layer is promoted and reduced, respectively, by herbivores (a value of 0 indicates no effect). The data set is based on 98 points (each based on decomposition rates of a given monospecific litter collection, and vegetation density inside vs outside a fenced exclosure plot for the same plant species and location as that from which the litter was collected), representing 51 plant species and collected from 28 locations. (a) Individual data points. × = Ferns; ▵ = monocots; □ = dicots with leaf lamina width >6 mm; ○ = dicots with leaf lamina width <6 mm; ▿ = Nothofagus spp. R = 0·488, P < 0·001, df = 96. (b) Means and 95% confidence intervals for the estimate of plant group means for data in (a). The five groups differ significantly at P < 0·001 with regard to both decomposition rate and vegetation response to herbivory (one-way anova).

Table 1.  Pearson's correlation coefficients between concentrations of chemical components of litter, and variables relating to litter decomposition and response of vegetation density to mammal browsing, for a range of plant species growing in the vegetation browse layer (0–2 m height)
 PhenolicsCondensed tanninsFibreCelluloseLigninNitrogen (N)Phosphorus (P)Lignin to N ratioLignin to P ratioN to P ratio
  • Data from 98 litter collections, each representing a unique species × location combination, from a total of 51 species and 28 locations.

  • ,

  • *


  • **


  • ***

    = correlation coefficient significantly different to 0 at P = 0·100, 0·050, 0·010 and 0·001, respectively. All values that are listed as significant at P = 0·001 remain significant at P = 0·05 when a Bonferroni correction is applied.

  • Log-transformed variable.

  • §

    Proportion of total initial N or P lost from litter during decomposition.

  • Mean effect of litter of species on decomposition rates of litter from all other species at that site, and mean response of decomposition rate of that species to all other species from that site, in litter-mixing bioassays.

  • ††

    Value of index V which is related to the relative density of vegetation of a given plant species inside vs outside fenced browsing mammal exclosure plot.

Decomposition rate0·000−0·274*−0·760***−0·525***−0·464***0·201*0·334**−0·379***−0·573***−0·434***
Relative N release§−0·145−0·157−0·0690·114−0·1970·600***0·000−0·590***0·1260·395***
Relative P release§0·1580·092−0·173−0·148−0·1540·0360·639***−0·141−0·550**−0·419***
Litter mixing effect−0·0000·0550·243*0·0510·250*0·1730·0380·0000·061−0·135
Litter mixing response0·172−0·133−0·633***−0·413***−0·342***0·0800·136−0·380***−0·251*−0·380***
Response to browsing††−0·130−0·319**−0·474***−0·196−0·314**0·1150·191−0·217*−0·308**−0·152
Table 2.  Stepwise multiple regression relationships relating variables for decomposition and response of vegetation density to mammal browsing to concentrations of chemical components of litter
Dependent variableIndependent variables in relationship*R2
  1. The data set used, and dependent and independent variables, are the same as for Table 1. Only those terms that explain a statistically significant proportion of the variation of the response variable at P = 0·05 are included. P < 0.001 for each regression.

  2. (+) and (–) after each variable indicate a positive or negative relationship between the independent and dependent variable.

  3. *Abbreviations: PH = phenolics; FIB = fibre; LIG = lignin; N = nitrogen; P = phosphorus; LIGN = lignin to N ratio; LIGP = lignin to P ratio; NP = N to P ratio.

Decomposition rateln(PH) (–); FIB (–); LIG (–); NP (–)0·693
Relative nitrogen releaseN (+); ln(LIGN) (–); NP (+)0·540
Relative phosphorus releaseln(PHOS) (+)0·418
Litter-mixing effectFIB (+); LIG (+); ln(P) (–); ln(LIGP) (–)0·259
Litter-mixing responseFIB (–); NP (–)0·421
Response to browsingln(PH) (–); FIB (–)0·297

When the data were analysed so that each species represented a different data point (with the data from each species averaged across all locations in which it appeared), the nature of patterns identified remained unchanged – there was still a significant relationship between herbivore effect on vegetation and litter decomposition rate (r = 0·416, P = 0·002, n = 51). With this analysis, decomposition rate and herbivore effect both remained significantly correlated at P = 0·05 with concentrations of condensed tannins and fibre, and with the lignin to N ratio; decomposition rate was also significantly related to lignin concentration and with the lignin to P ratio. When the data were analysed so that each site was represented by a different data point (data were averaged for all species within each location), the relationship between herbivore effect and decomposition was still significant (r = 0·173, P = 0·027, n = 28), meaning that sites that showed stronger overall negative effects of herbivores also, on average, produced litter that decomposed more rapidly.

Despite vegetation response to browsing being correlated with litter decomposition rate, there was no relationship between vegetation browsing response and the proportion of either total N or total P lost from litter during decomposition [r = 0·077 (P = 0·219) and 0·114 (P = 0·130), respectively]. Unlike decomposition rate and vegetation response to herbivory, litter N and P loss rates were not strongly correlated with concentrations of condensed tannins, lignin or fibre; instead, the best predictors were the initial concentrations of the N and P being lost and the ratios of these nutrients to lignin and each other (Tables 1 and 2).

With regard to the litter-mixing measurements, the effect of a given litter on decomposition rates of all the other litters collected from the same site was weakly positively related to the decomposition rate of its own litter (r = 0·284, P = 0·003), as well as the response of vegetation density of that species to browsing (r = 0·137, P = 0·099). This effect was weakly significantly correlated with litter fibre and lignin concentrations (Table 1), and was best predicted by a four-variable multiple regression that included these variables (Table 2). The net response of decomposition rate of a given litter to being mixed with other litters from the same site was negatively correlated with both the rate of decomposition of that litter (r = −0·456, P < 0·001) and response of vegetation density of that species to browsing (r = −0·195, P = 0·034). This variable was significantly correlated with most of the same litter quality attributes that were correlated with litter decomposition rate: concentrations of fibre, cellulose and lignin, and ratios of lignin to N and N to P.

When the data were analysed by placing all litter types in three categories (litter from understorey species with a greater vegetation density inside the exclosure and outside the exclosure, and litter from canopy tree species), there were large differences with regard to several litter quality variables (Table 3). Those species with greater density outside the exclosure produced litter with greater concentrations of phenolics, condensed tannins, fibre and lignin, and greater ratios of lignin to N and lignin to P, than those with a greater density inside the exclosure. However, litter from canopy tree species had higher values for most of these variables than litter from plants in either of the two understorey categories. Understorey species that were reduced by browsing produced litter that decomposed more rapidly than that produced by plants in the other two categories. However, there were no significant differences among categories for the proportion of total N or P lost from the litter, or for variables based on litter-mixing effects (Table 3). When differences occurred for a given variable between plant species favoured inside and outside exclosures, these differences generally became larger when data analysis was restricted to only those plant species for which the exclosure effect was significant at P = 0·05 (data not presented).

Table 3.  Litter quality and decomposition variables for litter of forest understorey plant species reduced by browsing mammals (vegetation density higher inside exclosure); understorey species promoted by browsing mammals (vegetation density greater outside exclosure); and canopy tree species
Litter variableReduced species (R)*Enhanced species (E)Canopy speciesP (all data)P (R vs E)
  • Data analysed using one-way anova. Litter variables, and the data for the understorey species, are as for Table 1.

  • *

    Reduced = understorey species 

  • ×

     location combinations for which vegetation density of species was greater inside than outside fenced exclosure plot; enhanced = combinations for which density was greater outside than inside exclosure.

  • Mixing effect = mean percentage enhancement by a given litter on the decomposition rates of litter from all other species collected from the same site when in mixture relative to monoculture; mixing response = mean percentage enhancement of decomposition rate of a given litter by all the other litters collected from the same site when in mixture relative to monoculture.

  • All P values <0·004 remain significant at P = 0·05 when a Bonferroni correction is applied.

Phenolics (%)1·412·614·15<0·001<0·001
Condensed tannins (%)0·2120·8621·557<0·001<0·001
Fibre (%)42·952·852·8<0·0010·001
Cellulose (%)21·023·120·90·2290·180
Lignin (%)20·025·330·5<0·0010·016
Nitrogen (N) (%)1·311·261·160·3920·616
Phosphorus (P) (%)0·1770·1190·1080·2050·174
Lignin to N ratio20·625·136·5<0·0010·147
Lignin to P ratio254·7377·1653·6<0·0010·028
N to P ratio14·115·517·50·0250·447
Decomposition rate (% day−1)31·819·519·60·003< 0·001
Release of N from litter (% day−1)50·248·642·00·2430·751
Release of P from litter (% day−1)32·928·823·60·2210·463
Litter-mixing effect (%)8·573·735·260·4270·217
Litter-mixing response (%)4·609·763·950·4630·308
Number of observations (n)593943  

When plant species were assigned to functional categories, there were highly significant differences among groups with regard to both decomposition rate (F4,93 = 15·6, P < 0·001) and vegetation density response to browsing (F4,93 = 5·42, P < 0·001) (Fig. 1b). There was a distinct relationship at the functional group level between litter decomposition rate and response of vegetation density to browsing, with large-leaved dicots having the highest values for both variables, and ferns and monocots having the lowest (Fig. 1b). These results were reasonably consistent with differences among functional groups with regard to litter quality; functional groups differed overall at P = 0·05 for all litter quality variables except P concentration. Small-leaved dicots and Nothofagus had the highest concentrations of condensed tannins and phenols; ferns and monocots had the highest concentrations of cellulose and fibre; and ferns and Nothofagus had the highest concentrations of lignin (Table 4). Litter nutrient concentrations were generally highest for the ferns and the large-leaved and small-leaved dicots (Table 4). Functional groups did not differ significantly overall with regard to the proportional loss during decomposition of either N (F4,93 = 1·41, P = 0·237) or P (F4,93 = 0·88, P = 0·447). However, in the litter-mixing experiment, the net effect of a given litter on all other litters collected from the same site differed among functional groups (F4,93 = 3·00, P = 0·022), with large-leaved dicots having the greatest effect and ferns and Nothofagus having the least. The net response of a given litter to all other litter types from the same site was strongly influenced by functional group effects (F4,93 = 7·61, P < 0·001); this response was greatest for the ferns and monocots.

Table 4.  Litter quality variables (mean ± SD) of all subcanopy litter collections for each of five plant groupings
Litter variablePlant groupingP*
FernsMonocotsDicots (large leaf)Dicots (small leaf)Nothofagus
  • *

    Derived using one-way anova among the five plant types.

  • Large leaf = leaf lamina width >6 mm; small leaf = leaf lamina width <6 mm.

Phenolics (%) 1·65 ± 2·41 0·77 ± 1·45 1·56 ± 1·26 3·78 ± 2·91 2·23 ± 0·73<0·001
Condensed tannins (%)0·323 ± 0·3250·181 ± 0·1830·315 ± 0·531 1·31 ± 1·8870·522 ± 0·3300·025
Fibre (%) 60·3 ± 14·8 57·3 ± 5·2 37·9 ± 9·6 40·2 ± 14·2 52·6 ± 8·9<0·001
Cellulose (%) 25·4 ± 4·2 32·6 ± 7·3 18·3 ± 5·8 19·2 ± 7·0 19·9 ± 2·2<0·001
Lignin (%) 29·2 ± 9·2 13·9 ± 5·3 19·3 ± 9·1 20·8 ± 11·7 30·8 ± 9·3<0·001
Nitrogen (N) (%) 1·36 ± 0·42 0·90 ± 0·39 1·26 ± 0·50 1·57 ± 0·48 1·20 ± 0·520·010
Phosphorus (P) (%)0·134 ± 0·1230·048 ± 0·0290·194 ± 0·2440·190 ± 0·2900·084 ± 0·1810·115
Lignin to N ratio 27·0 ± 22·1 21·4 ± 18·9 21·8 ± 21·3 15·2 ± 10·4 28·4 ± 8·20·040
Lignin to P ratio  363 ± 261  509 ± 489  227 ± 342  239 ± 204 362 ± 600·007
N to P ratio 16·1 ± 8·8 22·1 ± 7·7 12·1 ± 8·5 14·7 ± 9·6 13·7 ± 4·10·013
Number of observations (N)2011401610 

Within functional groups, only ferns and Nothofagus demonstrated statistically significant relationships between decomposition rate and vegetation density response to browsing (Table 5). Litter decomposition rates for all functional groups except monocots, and vegetation density responses to browsing for all groups except ferns and Nothofagus, could be related to at least some litter quality variables in multiple regression relationships (Table 5). Despite this, within functional groups, the variables that best explained decomposition rates in these relationships were never the same as those that best explained vegetation density responses to browsing (Table 5).

Table 5.  Stepwise multiple regression relationships relating litter decomposition rates and response of vegetation density to mammal browsing to concentrations of chemical components of litter, for each of the five plant groupings shown in Table 4
Plant groupingIndependent variables serving as predictors ofR2 (decomposition vs browsing)
Decomposition rateR2
Response to browsing
  • The data set and independent variables used are the same as for Table 1. Only those terms that explain a statistically significant proportion of the variation of the response variable at P = 0·05 are included.

  • NS, *, **, ***, indicate that R2 is not significantly different to 0, and significantly different to 0 at P = 0·05, 0·01 and 0·001, respectively.

  • Abbreviations: CT = condensed tannins; PH = phenolics; LIG = lignin; FIB = fibre; N = nitrogen; P = phosphorus; LIGN = lignin to N ratio; LIGP = lignin to P ratio; NP = N to P ratio.

  • Large-leaved = leaf lamina width >6·0 mm; small-leaved = leaf lamina width <6·0 mm.

  • Relationship between plant litter decomposition rate, and response of vegetation of the same species at the same location, to mammal browsing.

Fernsln(LIGN) (–); ln(LIGP) (–); ln(CT) (–)0·725***NoneNS0·281*
MonocotsNoneNSln(P) (+)0·556*0·188 (NS)
Dicots (large-leaved)ln(PH) (–); FIB (–)0·451***LIG (–)0·254**0·089 (NS)
Dicots (small-leaved)ln(CT) (–); FIB (–); ln(LIGN) (–)0·824***N (+)0·328*0·071 (NS)
NothofagusNP (–)0·515*NoneNS0·570**


The present study extends the findings of previous studies (Cornelissen et al. 1999; Grime et al. 1996; Wardle et al. 1998) by demonstrating an across-species relationship between plant litter decomposability and the response of vegetation density to herbivory. This provides support for the hypothesis of Pastor et al. (1988, 1993) that plant species which produce good quality litter are disadvantaged by browsing, and that herbivory by browsing mammals therefore switches the competitive balance towards plant species with poor litter quality. Such an outcome is possible only if palatable plant species are not promoted by browsing (cf. McNaughton 1979, 1985), and in our study there was a clear pattern of such plants being only negatively affected by herbivores. With regard to New Zealand forest understorey species, at least, it appears probable that decomposability of litter and vegetation responses to herbivory are governed by the same or similar suites of traits, and this is reinforced by Tables 1 and 2 which demonstrate that both response variables were related to concentrations of secondary metabolites and structural carbohydrates in plant-derived tissues.

While litter mass loss (indicative of C mineralization) showed a distinct relationship with vegetation density response to herbivory, loss rates of N and P from the litter did not. It is apparent from Table 1 that, whereas litter decomposition rate was related to both concentrations of C constituents (secondary metabolites and structural carbohydrates) and concentrations of major nutrients (N and P), release rates of N and P were related only to concentrations of nutrients, while the response of vegetation density to browsing was related only to concentrations of C constituents. This suggests that, across species, vegetation density response to herbivory is driven by some of the same factors that influence litter C mineralization, but not N or P mineralization. This is consistent with the suggestion that browsing mammals preferentially browse vegetation on the basis of C chemistry and not nutrient status (Bryant, Chapin & Coley 1983), and indicates that browsers therefore selectively remove plant species that promote C mineralization, but not necessarily that of major nutrients. This means that browser-induced vegetation changes are likely to have a stronger influence on the ecosystem dynamics of C than on the dynamics of N or P.

Interactive effects among litters from different plant species may be ecologically important (Blair, Parmelee & Beare 1990; McTiernan, Ineson & Coward 1997; Wardle, Bonner & Nicholson 1997). Seastedt (1984) proposed that litter of better quality should accelerate the decomposition of associated poor quality litter and, conversely, poor quality litter should retard decomposition of good quality litter. If so, litter from species reduced by browsing in our study should promote the decomposition of associated litters when in mixture, while litter from species encouraged by browsing should retard decomposition of associated litters. Our results provide some evidence for this, but the effects are weak. Litter from species with faster decomposition rates did promote greater decomposition of associated litters and, although there was a positive relationship across species between vegetation response to browsing and net effect of litter on associated litters, it was significant only at P = 0·10. Those plants that produced litter of the lowest quality were also those whose decomposition rates were most strongly promoted by associated species, although the relationship between vegetation response to herbivory and response of litter decomposition to mixing with other litters was only marginally significant at P = 0·05. These results suggest that browsing mammals may reduce vegetation density of plant species that not only decompose more rapidly but also accelerate the decomposition of associated, more recalcitrant, litter types. However, the indirect effect of browsing mammals on interspecific litter interactions is small, and likely to be of only secondary importance for overall decomposition rates when compared with the more direct effects of browsing mammals removing plant species that produce faster decomposing litter.

The relationship between decomposability and vegetation response to herbivory was driven to a large extent by differences between plant functional groups; it is significant that the five groups we identified showed identical ranking with regard to both variables (Fig. 1b). It is apparent that the favourability of tissues from large-leaved dicots to herbivores and decomposers is because of the low concentrations of secondary metabolites and structural carbohydrates in these tissues, while the unfavourability of fern, monocot and Nothofagus tissues is due largely to high concentrations of structural carbohydrates.

The fact that small-leaved dicots were significantly less adversely affected by browsing than were large-leaved dicots (F1,54 = 6·12, P = 0·016), but did not produce litter that decomposed significantly more slowly (F1,54 = 0·71, P = 0·403), would suggest that the plants in this group have a greater resistance to herbivory than would be expected based on the decomposition data. Many species in this group show a divaricating growth habit, and our data provide some evidence for the view that this growth form may serve as a structural defence against herbivores (Greenwood & Atkinson 1977; but see Howell et al. 2002; McGlone & Clarkson 1993). In this light it is relevant that this habit has been proposed as having arisen through selection pressure from browsing by the recently extinct moas (Aves: Dinornithiformes; Greenwood & Atkinson 1977).

Our results provide evidence that foliar herbivores have the capacity to retard below-ground processes through causing a switch from understorey community domination by plants that produce good quality litter (which decomposes rapidly) to those that produce poor quality litter. However, it is important to note that this is only one of several mechanisms by which herbivores can influence the decomposer subsystem. First, such a mechanism will operate only if herbivores adversely affect only palatable species, and not if palatable species are optimized by herbivory (Davidson 1993; McNaughton 1979). Second, herbivores can also affect decomposer processes and organisms (either positively or negatively) through the return of dung and urine to the soil (Bardgett et al. 1998a), inducing physiological changes and shifts in litter quality of browsed plants (Kielland, Bryant & Ruess 1997), and affecting ecosystem productivity (Belovsky & Slade 2000). These mechanisms all operate simultaneously, and it is the balance of the mechanisms that helps determine the magnitude and direction of overall effects of foliar herbivory on the decomposer subsystem. In the study of Wardle et al. (2001), which utilized the same sites as the present study, decomposer organisms and processes were adversely affected by browsing mammals on some sites, and positively affected on others. It is likely that diminished quality of litter input through herbivore-induced vegetation changes was a driving factor in many of those sites in which adverse effects of herbivory on decomposers was detected. However, this mechanism must have been overridden by other mechanisms for those sites in that study where herbivores had positive effects on decomposers.


We thank Gaye Rattray for technical assistance in setting up the decomposition bioassays, Kate Dorling (University of Western Australia) for the nutrient analyses, and the Palmerston North Chemistry Laboratory, Landcare Research for measurements of lignin, phenolics and related compounds. Marie-Charlotte Nilsson made helpful comments on an earlier version of the manuscript. This work was supported by the New Zealand Marsden Fund and the Foundation of Research, Science and Technology (New Zealand).


Appendix 1

Table 6.  Species for which litter was sampled from each of the 28 sites (site codes correspond to Table 1 of Wardle et al. 2001)
Site codePlant species in understoreyCanopy tree species
Greater inside exclosureGreater outside exclosure
  • Greater inside exclosure and greater outside exclosure = measured vegetation density of that species is higher inside the exclosure, and outside the exclosure, respectively.

  • Letters in square brackets represent plant type; codings: F = ferns; G = gymnosperms; L = large-leaved dicots (leaf lamina width >6 mm); M = monocots; N = Nothofagus; S = small-leaved dicots (leaf lamina width <6 mm).

  • ¶Statistical significance: NS, * and ** indicate that the difference in vegetation density inside vs outside the exclosure plot of that species is not significantly different, is different at P = 0·05, or is different at P = 0·01, respectively (derived from Wardle et al. 2001).

S1Freycinetia baueriana Endl. [M] NSBlechnum fraseri (Cunn.) Luerssen [F] NSAgathis australis Salisb. [G]
   Beilschmeidia tarairi (A. Cunn.) [L]
S2Coprosma arborea Kirk [L] NS B. tarairi[L]
 F. baueriana[M]*  
 Geniostoma rupestre J. R. Forst. & G. Forst. [L] NS  
S3Coprosma rhamniodes A. Cunn [S] NSPseudopanax lessonii (D. C.) C. Koch [L] NSKunzea ericioides (A. Ritch) Joy Thomps [S]
 Melicytus ramiflorus J. R. Forst. & G. Forst. [L]*  
S4Gahnia pauciflora Kirk [M] NSA. australis[G]*A. australis[G]
  Astelia trinervia Kirk [M] NS 
S5M. ramiflorus[L]**Cyathea dealbata (Forst. F.) Swartz [F] NSKnightia excelsa R. Br. [L]M. ramiflorus[L]
S6Coprosma grandifolia Hook. F. [L]**C. dealbata[F]*Beilschmeidia tawa (A. Cunn)
   Benth. & Hook. F. ex. Kirk [L]
 Hymenophyllum demissum (Forst. f.) Schwarz [F]**Uncinia sp. [M] 
S7Coprosma lucida J. R. et. G. Forst. [L]**Cyathodes juniperina (J. R. Forst. & G. Forst.) [S]*K. excelsa[L]
 Astelia sp. [M]*Leucopogon fasciculata (G. Forst.) A. Rich. [S]*Nothofagus solandri var.
   cliffortioides (Hook. F.) Poole [N]
S8C. grandifolia[L]**Brachyglottis repanda J. R. Forst. & G. Forst.K. ericioides[S]
  [L] NS 
  Microlaena avenacea (Rauol) Hook. F. [M]K. excelsa[L]
  Uncinia uncinata (L. f.) Kük [M]** 
S9Asplenium bulbiferum Forst. F. [F]**M. avenacea[M]**B. tawa[L]
 Cyathea smithii Hook. F. [F] NS  
 Schleffera digitata (J. R. & G. Forst.) [L]**  
S11Brachyglottis rotundifolia J. R. Forst. & G. Forst. [L]*Blechnum montanum Chambers & Farrant [F]*B. rotundifolia[L]
 Carmichaelia arborea (G. Forst.) Druce [S]*  
 C. grandifolia[L] NS  
 Hebe stricta[Benth][L]*  
S12Carpodetus serratus J. R. Forst. & G. Forst. [L]**L. fasciculatus[S]*K. ericioides[S]
  Uncinia sp. [M]* 
S13Astelia sp. [M]**C. juniperina[S]*K. ericioides[S]
 C. lucida[L]**  
 Griselinia littoralis Raoul [L] NS  
S14Coprosma ciliata Hook. F. [S] NS N. solandri var. cliffortioides[N]
 Coprosma sp. [S]*  
 G. littoralis[L]**  
 N. solandri var. cliffortioides[N]*  
 Pseudowintera colorata (Raoul) Dandy [L] NS  
S16Blechnum discolor (Forst. F.) [F] NS Myrsine salicina Hew. Ex. Hook. F. [L]
 C. grandifolia[L]** Weinmannia racemosa L. f. [L]
 Pseudowintera axillaris (J. R. et G. Forst.) [L]*  
S17Coprosma microcarpa Hook. F. [S]**C. juniperinus[S] NSN. fusca[N]N. menziesii (Hook. F.) [N]
 G. littoralis[L]*  
 Nothofagus fusca (Hook. F.) Oerst [N] NS  
 Pimelia sp. [L] NS  
S18 Coprosma linariifolia Hook. F. [S] NSN. solandri var. solandri[N]
  G. littoralis[L] NS 
  Nothofagus solandri var. solandri (Oerst) [N] NS 
S19C. serratus[L] NSN. fusca[N] NSK. ericioides[S]N. fusca[N]
S20N. fusca[N] NS N. fusca[N]
 N. menziesii[N]* N. menziesii[N]
S21N. solandri var. cliffortioides[N] NSC. rhamnoides[S] NSN. solandri var. cliffortioides[N]
S22Coprosma foetidissima J. R. Forst. & G. Forst. [L]*Neomyrtus pedunculatus (Hook. F.) [S] NSLibrocedrus bidwillii Hook. f. [G]
 C. lucida[L]*  
 G. littoralis[L]**  
 Pseudopanax simplex (G. Forst.) Philipson [L]**  
 P. colorata[L] NS  
S23A. bulbiferum[F] NSCoprosma rotundifolia A. Cunn. [S] NSG. littoralis[L]
 S. digitata[L]**Polystichum vestitum (Forst. F.) [F]**L. bidwillii[G]
S24Pittosporum tenuifolium Sol. Ex. Gaertn. [L]**C. rotundifolia[L]*M. ramiflorus[L]
S25N. solandri var. cliffortioides[N] NSHymenophyllum multifidum (Forst. F.) [F] NSN. solandri var. cliffortioides[N]
  Hymenophyllum sanguinolentum (Forst. F.) [F] NS 
S26Blechnum fluviatile (R. Br.) Salomon [F] NSH. sanguilentum[F]**N. menziesii[N]
 C. serratus[L] NSN. menziesii[N]** 
 Coprosma parviflora Forst. f. [S]**N. solandri var. cliffortioides[N] NS 
  P. vestitum[F] NS 
S27Coprosma s. t. as in Eagle, A. [S]**B. discolor[F]*N. menziesii[N]
 G. littoralis[L]**Pseudowintera colorata[L]* 
 Pseudopanax sp. [L]*  
S28B. discolor[F] NSC. foetidissima[L] NSDacrydium cupressinum Lamb. [G]Metrosideros umbellata Cav. [L]W. racemosa[L]
S29Histiopteris incisa (Thunb.) J. Smith [F] NSB. discolor[F] NSB. rotundifolia[L]
  B. rotundifolia[L]*C. serrata[L]Dicksonia squarrosa (Forst. f.)
   Swartz [F]
S30C. foetidissima[L]**B. discolor[F] NSD. squarrosa[F]
 D. squarrosa[F]* M. umbellata[L]W. racemosa[L]