Abundance and flowering success patterns in a short-term grazed grassland: early evidence of facilitation
B. BOSSUYT, B. DE FRÉ and M. HOFFMANN
Terrestrial Ecology Unit, Department of Biology, University of Ghent, Ledeganckstraat 35, B-9000 Gent, Belgium
Summary
1
Associational resistance is a grazing avoidance mechanism resulting from herbivoresbeing less inclined to eat palatable plant species when these species grow in association withunpalatable species. Such facilitative interactions between plant species may have importantconsequences for patterns of relative abundance and reproduction in plant communities.
2
We studied the relationship between abundance and flowering success of palatablespecies and the cover of three unpalatable species (
Senecio jacobaea
,
Iris pseudacorus
and
Lysimachia vulgaris
) in a grassland where moderately intense grazing (0.17–0.42grazers ha
−
1
) had been ongoing for 3 years. Plots were selected so that there were nosystematic differences in microenvironmental conditions associated with the level ofcover of unpalatable species.
3
Several palatable species had a higher frequency, cover and/or flowering success whenthey grew in the neighbourhood of an unpalatable species.
4
Several species (both palatable and unpalatable) were significantly taller in thevicinity of a large unpalatable species, probably due to the combined effects of grazingavoidance and increased light competition.
5
These facilitative effects, however, have not yet resulted in a higher local species richnessor plant community evenness. It is nevertheless likely that more pronounced effects will beseen if grazing is continued, because facilitative interactions between plant species inducedby grazing have already led to a shift in patterns of abundance and flowering success.
One of the major challenges in ecology is the searchfor factors that explain patterns of species coexistencein a community. The main focus has historically beenon the negative effect that one species may exert onanother (competitive interactions) but there is growingawareness that facilitation may also be involved andthat its influence on population- and community-levelvariables is at least as important as other factors(Callaway & Walker 1997; Bruno
et al
. 2003).Grazing avoidance mechanisms, based on attributes
external to the plant (reviewed by Milchunas & Noy-Meir 2002), illustrate the importance of facilitationin structuring plant communities. A plant individualgrowing in the neighbourhood of large and spiny shrubsmay be protected against herbivory because herbivores
are physically hampered from reaching the targetspecies. Several studies have provided evidence forthese biotic refuges: establishment of tree seedlings wasfound to be more successful under spiny shrubs than inopen areas (Callaway 1992; Gómez
et al
. 2001; Bakker
et al
. 2004) and Rebollo
et al
. (2002) concluded that
Opuntia
cacti were a refuge for the grassland speciesmost sensitive to grazing, leading to higher plant diver-sity within cactus clumps. Most authors conclude,however, that the observed effect is the combined resultof protection against herbivory and ameliorated micro-environmental conditions below the shrub, such as ahigher moisture content (Gómez
et al
. 2001; Rebollo
et al
. 2002) or increased shading (Callaway 1992).The probability of herbivore damage can also be
reduced by growing in association with plant speciesthat influence the behaviour of herbivores, rather thanphysically hampering them, as illustrated for insectherbivory by Holmes & Jepson-Innes (1989) andKarban (1997). Associational resistance is a facilitation
mechanism based on the relative palatability of a plantspecies in comparison with the surrounding vegetationmatrix, and hence on herbivore foraging decisions(Milchunas & Noy-Meir 2002).
The degree of protection of a palatable species pro-vided by association with an unpalatable species dependson the level at which the herbivore makes foraging deci-sions: stand, community or bite scale (Hjälten
et al
.1993; Milchunas & Noy-Meir 2002). If the herbivoredecides on the bite scale, plant individuals are only pro-tected if they grow in the close vicinity of an individualof an unpalatable species. Associational resistance isthen the result of a complex interplay between negativeand positive interactions, as species simultaneouslycompete with and facilitate each other (Callaway &Walker 1997; Holmgren
et al
. 1997; Bruno
et al
. 2003).Whether a species will, ultimately, be positively or neg-atively affected by growing in association with anunpalatable species depends on the relative character-istics of both interacting species, such as competitiveability, light requirements and grazing tolerance, andon the site conditions, such as grazing pressure and pro-ductivity (Callaway & Walker 1997; Rebollo
et al
. 2005).Evidence for associational resistance as a protective
mechanism against large herbivores has been scarce.Callaway
et al
. (2000) found strong, species-specific,positive effects of unpalatable invaders on the abundanceand sexual reproduction of many subalpine meadowspecies. Their results indicated that the unpalatableweeds may preserve plant diversity in areas subject toheavy long-term grazing.
Our objective was to assess the short-term effects ofgrazing on facilitative interactions in a dune grasslandplant community. We compared the abundance, flower-ing success and plant height of individual plant speciesand the characteristics of the plant community betweenecologically similar microsites with low or high cover ofan unpalatable species. Large grazers had been intro-duced to the study area for conservation purposes only3 years before the start of this study.
Before we conclude that any observed differences aredue to associational resistance, we need to exclude sys-tematic differences in microenvironmental conditionsrelated to the cover of the unpalatable species. Thesedifferences may occur if unpalatable species reach ahigh level of cover on sites with specific microenviron-mental conditions or, alternatively, if high cover of anunpalatable species alters microenvironmental condi-tions, e.g. by increasing shading or litter production.We kept these microenvironmental differences to aminimum by selecting paired plots separated by a max-imum of a few metres (and, in most cases, adjacent) andwithout observable differences in characteristics, suchas soil texture, microrelief or groundwater level. Weexcluded potential systematic differences in speciescomposition by using multivariate analysis. Moreover,if a facilitation effect can be demonstrated specificallyfor those species that can be considered as most preferredby the grazers in the area, associational resistance is
strongly implicated as a more important structuringfactor than microenvironmental conditions, which wouldaffect palatable and unpalatable species equally.
Because the facilitating effect may be species specific,we studied three unpalatable species (
Senecio jacobaea
L.,
Iris pseudacorus
L. and
Lysimachia vulgaris
L.) thatare known to be avoided by large herbivores becauseof toxic components (Van Genderen
et al
. 1996). Theyhave no spines or other adaptations that could physi-cally hamper large grazers and an observed effect onneighbouring species will hence be the result of therelative palatability of the species involved.
We predicted that palatable species will have a higherabundance, a higher number of inflorescences and agreater plant height in the immediate vicinity of anunpalatable species as a result of protection againstgrazing. As a consequence, community composition andstructure (species richness and evenness) was expectedto be different between microsites with high and lowcover of an unpalatable species.
Materials and methods
The study area is situated on the western part of the Bel-gian coast, in the nature reserve ‘De Westhoek’. Annualrainfall is 700 mm and the soils are sandy and rich incalcium. From 1997, ponies and Scottish Highland cattlehave grazed year round in a fenced part (60 ha) of thereserve. The area had previously been ungrazed since1960, when agricultural practices in the region wereabandoned to a large extent. The nature reserve con-sists of a landscape mosaic of open marram dune, dryopen and moist closed-canopy dune grassland, shrubvegetation, mainly of
Hippophae rhamnoides
and
Ligus-trum vulgare
, and an afforested area (Lamoot
et al
. 2005).Because of yearly fluctuations in the number of foalsand calves, the number of ponies and cattle varied be-tween 8 and 21 and between 2 and 4, respectively. Duringthe observation period animal biomass was
c
. 85 kg ha
−
1
.Compared with the biomass in near-natural grazing sys-tems in temperate regions [which, according to Wallisde Vries (1998) range between 8 and 67 kg ha
−
1
], thisgrazing density can be described as moderately high.
Vegetation and plant species data were collected insummer 2000, 3 years after the introduction of grazers.In order to obtain an estimate of species availability, ageneral vegetation survey was made, consisting of 352
×
2 m plots, randomly distributed over the study area.In each plot, we estimated the canopy cover of each occur-ring plant species to the nearest 10%, except for the low-est class which was divided into 1%, 3% and 5%. Therelative abundance of each species in each plot was cal-culated by dividing the cover of each species by the totalcover in the plot. The average of this relative abundance,
including zero values, over the 35 plots of the vegetationsurvey for each species was considered as a measure ofits relative availability for the grazers in the study area.
Diet composition was determined from feeding obser-vations, using a method similar to the bite-count method(Hobbs
et al
. 1983; Vulink & Drost 1991; Cosyns 2004).Each month during the 1999 and 2001 fruiting seasons,we observed herbivore activities over 48 h, distributedevenly over 6-h morning (06:00–12:00 h), afternoon(12:00–18:00 h) and evening (18:00–24:00 h) sessions.Before starting a session, one animal was randomlychosen and followed for the next 6 h. Observations wereconducted within a 3-m range, and animal behaviour wasnot visually affected by the presence of the observer. Allplant species and plant parts seen bitten were recorded.The proportion of each plant species to the diet of thegrazers was calculated by dividing the number of bitesof that species by the total number of bites observed.
We compared the contribution of each plant speciesto the grazers’ diet with its relative availability in thearea on a rank basis, resulting in a division of the 89plant species that occurred in the study area into threegroups reflecting different palatability. For eachspecies, we calculated its rank for availability and fordiet composition, both values ranging from 1 to 89. Aspecies was considered as preferred or relatively moreeaten than available, and hence as ‘palatable’, either if ithad a diet contribution rank of more than 80, indicat-ing that a large proportion of the diet consists of thatspecies, or if its diet contribution rank was at least 20higher than its availability rank, which indicates thatit had a larger contribution to the diet than could beexpected based on its availability in the vegetation.When a species had a diet contribution rank that wasat least 20 lower than its availability rank, meaning thatit did not contribute largely to the diet despite itsrelatively high abundance in the vegetation, it was con-sidered avoided and therefore ‘unpalatable’. The remainingspecies were classified as ‘neutral’. This resulted in 16palatable, 27 unpalatable and 46 neutral species.
To test the effect of
Senecio jacobaea
, we selected 50pairs of 0.5
×
0.5 m plots, each pair consisting of oneplot with a high cover of
Senecio
(> 50%) and one plotwith a low cover (< 2%). Similarly, we selected 50 pairsof 0.5
×
0.5 m plots with a high and a low cover of
Iris pseudacorus
; all plots had a low cover of the twoother unpalatable species (< 5%) and the pairs wereselected in a way that differences in habitat conditions(microrelief, moisture content, soil characteristics)were minimal. The plots within each pair were sepa-rated by a maximum distance of a few metres, and allplots were situated in an area of 10 ha within the grazedstudy area of 60 ha. As the cover of the unpalatable spe-cies reaches at least 50% in the plots with a high cover,all other species in these plots were growing at a max-imum distance of 10 cm from the unpalatable species.To test for the effect of
Lysimachia vulgaris
, we selected100 0.5
×
0.5 m plots, with a cover of
Lysimachia
rang-ing from 0% to 100% and a low cover of
Senecio
and
Iris
. Here also, plots were selected so that habitat con-ditions were as similar as possible.
For each of the 300 0.5
×
0.5 m plots, we estimatedthe canopy cover of each occurring plant species asdescribed above, counted the number of individualflowers or inflorescences (hereafter called inflores-cences) and measured the maximal height of each plantspecies in the plot.
As an indirect check for a bias in habitat characteristicsbetween the plots with a high or a low cover of theunpalatable species, we executed a detrended corre-spondence analysis (DCA) on the three data sets sep-arately, excluding the unpalatable species in question,using the program PCOrd.
Characteristics of the plant community (total spe-cies richness, evenness and total number of inflores-cences) were compared between plots with a high and alow cover of
Senecio
or
Iris
with a Wilcoxon signedrank test for related samples and between plots withfour cover classes of
Lysimachia vulgaris
(0–20%, 20–50%, 50–70% and 70–100%) with a Kruskal–Wallistest with multiple comparisons. Owing to a lack of nor-mal distribution of the data, even after transformation,a non-parametric statistical approach was used.
Similarly, we tested for differences in frequency,absolute and relative cover in the plots where the spe-cies was present, plant height and number of inflores-cences of each individual species between plots with ahigh and a low cover of
Senecio
or
Iris
(chi-squared orWilcoxon signed rank test for related samples) andbetween plots with the four cover classes of
Lysimachia
(chi-squared or a Kruskal–Wallis test with multiplecomparisons). We also correlated these variables withthe cover of
Lysimachia
using a Spearman rank corre-lation coefficient. The relative cover of each species wascalculated by dividing the cover of the individual spe-cies by the total cover in the plot minus that of
Senecio
,
Iris
or
Lysimachia
, respectively.For each plot we also calculated the sum of the rel-
ative cover, and the number of inflorescences, of palat-able, neutral and unpalatable species. Differencesbetween these values between each pair of plots weretested with a Wilcoxon test for related samples andbetween the four cover classes of
Lysimachia
with aKruskal–Wallis test with multiple comparisons. Theywere also correlated with the cover of
Lysimachia
usinga Spearman rank correlation coefficient.
Results
The biplots of the DCAx analyses indicated that dif-ferences in species composition over all pairs of plotswere much larger than differences between plots with ahigh and a low cover of the unpalatable species (Fig. 1).There were no significant systematic differences inDCA scores on the first axes between plots with a high
and a low cover of the unpalatable species (Mann–Whitney and Krukal–Wallis tests, all
P
> 0.05), exceptfor the scores on the first DCA axis of the plots with acover lower than 25% and a cover higher than 70% of
Lysimachia vulgaris
(Fig. 1c).We found no differences in species richness or evenness
between plots with a high and a low cover of
Senecio
or
Iris
, although plots with a cover of
Lysimachia
of morethan 70% had significantly lower total species richness(Table 1). The number of inflorescences was higher in theplots with a high cover of
Senecio
, regardless of whetherthe number of
Senecio
inflorescences was included, butthere were no significant differences in the number ofinflorescences for the other two unpalatable species.
One species (
Cerasium fontanum
) had a higher fre-quency and three species (
Crepis capillaris
,
Rubus caesius
and
Vicia cracca
) a higher absolute or relative cover in theclose vicinity of
Senecio
(Table 2). Two species (
Cerastiumfontanum
and
Holcus lanatus
) produced significantlymore inflorescences and several species were of greaterheight if they were associated with
Senecio.
Only onespecies (
Agrostis stolonifera
) was negatively influencedby a high cover of
Senecio
, which was expressed in a sig-nificantly lower absolute cover value, but again this wasno longer significant after Bonferroni corrections.
One species (
Agrostis stolonifera
) had a higher fre-quency and one species (
Juncus subnodulosus
) had ahigher absolute cover value in the plots with a high coverof
Iris
after Bonferroni corrections, but we countedmore inflorescences for only one species (
Holcus lanatus
)(Table 3). One species (
Trifolium repens
) had a signi-ficantly lower cover after Bonferroni corrections andtwo species (
Agrostis stolonifera
and
Trifolium repens
)produced fewer inflorescences in these plots. Severalspecies were taller in the neighbourhood of
Iris
.The frequency of two species increased and of three
species decreased with an increasing cover of
Lysimachia
(
P
< 0.05), but this remains significant after Bonfer-roni corrections only for the decreasing
Holcus lanatus
(Table 4). The absolute and relative cover of
Agrostis
Fig. 1 DCA scatter plots of the scores on the first two axesfor (a) plots with high and low cover of Senecio jacobaea,(b) plots with high and low cover of Iris pseudacorus, and(c) plots with four cover classes of Lysimachia vulgaris. Inevery case the species under consideration was left out of theanalysis. In all three cases, n = 100. The average scores on bothaxes for the cover classes are indicated with their 95%confidence interval.
Table 1
Mean and standard error (SE) of the number of species 0.25 m
–2
, community evenness and number of inflorescences 0.25 m
–2
(with and without takinginto account the unpalatable species concerned) for plots with a low and a high cover of
Senecio jacobaea
,
Iris pseudacorus
and
Lysimachia vulgaris
.Significant differences between classes are indicated with different superscript letters
Table 2 Frequency and average absolute cover, relative cover, number of inflorescences and plant height (cm) with standard error (SE) for species that show differences between plots with a high (n = 50) and a lowcover (n = 50) of Senecio jacobaea at a significance level of 0.05. Positive effects are shown in bold type. P values are indicated in italic type if they are still significant after Bonferroni corrections (significance levelof 0.002)
Cover by Senecio jacobaea
Frequency (%) Absolute cover (%) Relative cover (%)Number of inflorescences Height (cm)
Table 3 Frequency, average absolute cover, relative cover, number of inflorescences and plant height (cm) with standard error (SE) for species that show differences between plots with a high (n = 50) and a low cover(n = 50) of Iris pseudacorus at a significance level of 0.05. Positive effects are shown in bold type. P values are indicated in italics if they are still significant after Bonferroni corrections (significance level of 0.002)
Cover by Iris pseudacorus
Frequency (%) Absolute cover (%) Relative cover (%) Number of inflorescences Height (cm)
–1114Table 4 (a) Frequency, and average absolute and relative cover, and (b) average number of inflorescences and plant height (cm), with standard error for species showing differences in plots with four cover classes
of Lysimachia vulgaris (< 20%, 20–50%, 50–70%, 70%) at a significance level of 0.05, and correlation between absolute and relative cover and between number of inflorescences and plant height with the cover ofLysimachia. Positive effects are given in bold type. P values and correlation coefficients in italics are still significant after Bonferroni corrections (significance level of 0.002). Significant differences are indicated withdifferent superscript letters
(a) Cover by Lysimachia vulgaris (%)
Frequency (%) Absolute cover (%) Relative cover (%)
stolonifera and Rubus caesius was significantly posi-tively correlated with the cover of Lysimachia, whereasthe cover of Ranunculus repens was negatively correlatedThe number of inflorescences increased significantly forAgrostis stolonifera and Juncus subnodulosus, and severalspecies were taller when the cover of Lysimachia was high.
The relative cover of palatable species was signifi-cantly higher in the plots with a high cover of Senecioor Iris than in the plots where the cover of these specieswas low (Fig. 2). This is compensated for by a lower
cover of neutral and unpalatable species, although thiswas only significant for the neutral species in the case ofa high cover of Iris. The relative cover of palatable spe-cies increased and of unpalatable species decreasedwith an increasing cover of Lysimachia (Fig. 3). Therelative cover of palatable species was higher in theplots where the cover of Lysimachia was greater than50% than in the plots where the cover was less than 20%(Kruskal–Wallis test, P = 0.003). The opposite wastrue for the unpalatable species (P = 0.029).
Plots with a high cover of Senecio had a significantlyhigher number of inflorescences of palatable species,and plots with a high cover of Iris had a lower number ofinflorescences of unpalatable species (Fig. 4). Palatablespecies produced significantly more inflorescenceswhen the cover of Lysimachia was high (Fig. 5).
Fig. 2 Mean relative cover (sum of the cover divided by the totalcover) of unpalatable, neutral and palatable species in plots witha high or a low cover of Senecio jacobaea and Iris pseudacorus.P values are given for the Wilcoxon signed rank test, testing fordifferences between plots with high or low cover of the unpalat-able species. Error bars indicate the standard error.
Fig. 3 Relative cover of palatable (Rs = 0.38, P = 0.0001) andunpalatable (Rs = −0.29, P = 0.003) species as a function ofcover by Lysimachia vulgaris in the plots.
Fig. 4 Mean number of inflorescences 0.25 m−2 of unpalatable,neutral and palatable species in plots with high or low cover ofSenecio jacobaea and Iris pseudacorus.
Fig. 5 Number of inflorescences 0.25 m−2 of palatable species as afunction of cover by Lysimachia vulgaris (Rs = 0.41, P < 0.0001).
There was clear evidence for a facilitation effect,expressed by a higher abundance and flowering successof several palatable species when they were associatedwith a high cover of an unpalatable species. The associ-ation resulted in a lower probability of herbivore damage,most likely because herbivores’ foraging decisions werebased on a tendency to avoid foraging on micrositeswith a high cover of an unpalatable species. The effect ofthis grazing avoidance had become significant for sev-eral palatable species (e.g. Agrostis stolonifera, Cerastiumfontanum, Holcus lanatus, Juncus subnodulosus, Rubuscaesius) by only 3 years after the start of year-roundgrazing. These species reached a higher abundance inassociation with an unpalatable species or/and pro-duced more inflorescences. Several species were alsohigher if they grew in close vicinity with an unpalatablespecies. Interactions between palatable and unpalatablespecies seemed to be species specific because there wereno species that were facilitated by all three unpalatablespecies.
We cannot directly exclude habitat differences as acause for the observed effects, but there are convincingindirect arguments indicating that habitat differencesbetween the two plots of each pair were not largeenough to cause observable differences in species abund-ance and flowering success. The multivariate analysisindicated that the species composition of the two plotswithin a pair was very similar (Fig. 1), with no system-atic differences related to the cover of the unpalatablespecies. Differences in species composition within thegroup of plots with a low cover or within the group ofplots with a high cover were larger than between plotswith a high and a low cover. Moreover, there were noobservable differences in habitat characteristics, suchas soil texture, microrelief or groundwater level,between paired plots.
Species classified as unpalatable, which are bydefinition avoided by the grazers, even in plots wherethey are not protected by a high cover of Senecio, Iris orLysimachia, reached a larger height in association withthese three study species. This suggests that the observedincrease in height may be at least partially the result ofincreased competition for light, as all three study spe-cies are taller than the other species in the vegetation.Intraspecific variation in plant height is known to berelated to light quantity and quality, and can be seen asa form of adaptive plasticity allowing plants to toleratecompetition for light by neighbours (Schmitt & Wulff1993; Sekimura et al. 2000), with strong competitionforcing plants to invest more in height growth to enablethem to capture sufficient light.
Although there may be some differences in micro-environmental conditions, more specifically in lightavailability, these are unavoidable in studies involvinglarge unpalatable species. Moreover, the facilitativeeffects were observable even though there was strongerlight competition. Irrespective of the underlying cause,
the effect of vicinity to an unpalatable species on plantheight results in vegetation patches that are taller thantheir surroundings and an increase in the structuraldiversity of the grassland. The same height effect, andassociated increased structural diversity, was found forplant species growing within clumps of Opuntia cacti(Rebollo et al. 2002).
Facilitation and competition effects at the specieslevel resulted in differences in community composition.There was a higher contribution of palatable species inthe plots with a high cover of Senecio and Iris and anincreasing contribution of these species with increasingcover of Lysimachia. This pattern again indirectly sug-gests that the observed effects were not due to habitatdifferences, but were a consequence of grazing avoid-ance resulting from differences in relative palatability.The total number of inflorescences of all palatable spe-cies together was much higher in plots with a high coverof Senecio. It can reasonably be assumed that a higherflowering success of palatable species in the vicinity ofunpalatable species will ultimately lead to a higher seedproduction. The total number of inflorescences was notaffected, but the number of inflorescences of palatablespecies also increased with increasing cover of Lysi-machia, despite increased competition. Only Iris seemsto exert a higher competitive pressure, as higher coverhad a negative effect on the number of inflorescences ofspecies classified as unpalatable, whereas the number ofinflorescences of palatable species was not affected. Nospecies was facilitated by all three unpalatable studyspecies, indicating that the outcome of the balancebetween facilitative and competitive effects will dependon the relative characteristics of the palatable and theunpalatable species.
The interactions between plant species had no effecton plant community richness or evenness, except forthe lower species richness in plots with > 70% cover ofLysimachia. It is likely, however, that the latter resultsfrom a ‘packing effect’, as it can be assumed that thenumber of co-occurring species is limited by spacewhen Lysimachia cover is this high. The lack of a gen-eral effect on species richness is in contrast with thefindings of Callaway et al. (2000), who found cleardifferences in species richness that could be attributedto grazing avoidance. However, these authors studiedthe effect of long-term, seasonal grazing, whereas year-round grazing was initiated only 3 years prior to thestart of our study. Although effects on individualspecies can already be observed, changes in speciesrichness patterns probably need longer to becomeapparent. Differences in species richness will onlybecome observable if species that go locally extinct as aconsequence of grazing survive in association withunpalatable species, and such processes probably takemore than 3 years. Moreover, the grazing pressure inthe area may still be too low, or the occurring speciesmay be too tolerant of grazing, to result effectively inlocal extinctions. This is also indicated by the limitednumber of species for which differences in frequency
are found. The effect of associational resistance mayindeed become more pronounced with increased graz-ing intensity (Rebollo et al. 2005), but higher grazingpressure might also eliminate protection by unpalatablespecies if herbivores are forced to be less selective. Thedegree of facilitation may hence be maximal at moder-ate grazing pressure, and this point will depend on siteproductivity.
Despite the lack of a species richness effect, we foundclear patterns in abundance and flowering success as aresult of facilitation by unpalatable species. Moreover,because grazing only started recently, it is to be expectedthat the effects will become more pronounced in thefuture. This means that associational resistance is animportant plant community structuring factor in grass-lands, even under moderate grazing pressure.
Acknowledgements
We are grateful to D. G. Milchunas and an anonymousreferee for their constructive comments that helpedus to improve the manuscript, and to the Ministryof the Flemish Community (AMINAL, Departmentof Nature) for permission to work in their naturereserve.
References
Bakker, E.S., Olff, H., Vandenbergh, C., De Mayer, K., Smit, R.,Gleichman, J.M. & Vera, F.W.M. (2004) Ecological anach-ronisms in the recruitment of temperate light demandingtree species in wooded pastures. Journal of Applied Ecology,41, 571–582.
Bruno, J.F., Stachowicz, J.J. & Bertness, M.D. (2003) Inclu-sion of facilitation into ecological theory. Trends in Ecologyand Evolution, 18, 119–125.
Callaway, R.M. (1992) Effects of shrubs on recruitment ofQuercus douglassii and Quercus lobata in California. Ecology,73, 2118–2128.
Callaway, R.M., Kikvidze, Z. & Kikodze, D. (2000) Facilita-tion by unpalatable weeds may conserve plant diversity inovergrazed meadows in the Caucasus Mountains. Oikos,89, 275–282.
Callaway, R.M. & Walker, L.R. (1997) Competition andfacilitation: a synthetic approach to interactions in plantcommunities. Ecology, 78, 1958–1965.
Cosyns, E. (2004) Ungulate seed dispersal – aspects of endo-
zoochory in a semi-natural landscape. PhD thesis, Universityof Ghent.
Gómez, J.M., Hódar, J.A., Zamora, R., Castr, J. & García, D.(2001) Ungulate damage on Scots pines in Mediterraneanenvironments: effects of association with shrubs. CanadianJournal of Botany, 79, 739–746.
Hjälten, J., Danell, K. & Lundberg, P. (1993) Herbivore avoid-ance by association – Vole and hare utilization of woodyplants. Oikos, 68, 125–131.
Hobbs, N.T., Bakker, D. & Gill, R.B. (1983) Comparativenutritional ecology of montane ungulates during winter.Journal of Wildlife Management, 47, 1–16.
Holmes, R.D. & Jepson-Innes, K. (1989) A neighborhood analy-sis of herbivory in Bouteloua gracilis. Ecology, 70, 971–976.
Holmgren, M., Scheffer, M. & Huston, M.A. (1997) Theinterplay of facilitation and competition in plant commu-nities. Ecology, 78, 1966–1975.
Karban, R. (1997) Neighbourhood affects a plant’s risk ofherbivory and subsequent success. Ecological Entomology,22, 433–439.
Lamoot, I., Meert, C. & Hoffmann, M. (2005) Habitat use ofponies and cattle foraging together in a coastal dune area.Biological Conservation, 122, 523–536.
Milchunas, D.G. & Noy-Meir, I. (2002) Grazing refuges,external avoidance of herbivory and plant diversity. Oikos,99, 113–130.
Rebollo, S., Milchunas, D.G. & Noy-Meir, I. (2005) Refugeeffects of a cactus in grazed short-grass steppe. Journal ofVegetation Science, 16, 85–92.
Rebollo, S., Milchunas, D.G., Noy-Meir, I. & Chapman, P.L.(2002) The role of a spiny plant refuge in structuring grazedshortgrass steppe plant communities. Oikos, 98, 53–64.
Schmitt, J. & Wulff, R. (1993) Light spectral quality, phytochromeand plant competition. Trends in Ecology and Evolution, 8,47–50.
Sekimura, T., Roose, T., Li, B., Maini, P.K., Suzuki, J. & Hara, T.(2000) The effects of population density on shoot morphologyof herbs in relation to light capture by leaves. EcologicalModelling, 128, 51–62.
Van Genderen, H., Schoonhoven, L.M. & Fuchs, A. (1996)Chemical Ecological Flora of The Netherlands and Belgium(in Dutch). Stichting uitgeverij KNNV, XII, Utrecht.
Vulink, T. & Drost, H. (1991) A causal analysis of diet com-position in free ranging cattle in reed dominated vegetation.Oecologia, 88, 167–172.
Wallis de Vries, M.F. (1998) Habitat qualitiy and the performanceof large herbivores. Grazing and Conservation Management(eds. Wallis de Vries, M.F., Bakker, J.P. & Van Wieren, S.E.),pp. 275–320. Kluwer Academic Publishers, Dordrecht.
Received 30 May 2005 revision accepted 12 July 2005 Handling Editor: Scott Collins
