1207

American Journal of Botany 90(8): 1207–1214. 2003.

EFFECTS OF HERBIVORY AND ITS TIMING ACROSS

POPULATIONS OF TRILLIUM GRANDIFLORUM (LILIACEAE)1

TIFFANY M. KNIGHT2

Department of Biological Sciences, University of Pittsburgh, Pittsburgh, Pennsylvania 15260 USA

The goal of this study was to identify the degree to which the frequency and timing of herbivory by white-tailed deer (Odocoileusvirginianus) and subsequent plant response varied across 12 populations of the perennial herb Trillium grandiflorum. Effects of naturaland experimental herbivory on the stage and size of reproductive plants were measured. Both the frequency and timing of herbivoryvaried across T. grandiflorum populations. Reproductive plants were more likely to regress to nonreproductive stages in the nextgrowing season when (1) reproductive plants were consumed by deer (vs. intact reproductive plants); (2) reproductive plants wereconsumed early in the growing season (vs. reproductive plants consumed late in the growing season); (3) reproductive plants weresmaller in size. Clipped plants that remained reproductive were smaller in the following season than unclipped controls. Plant sizewas positively correlated with the number of ovules, suggesting that reductions in the growth rate of reproductive plants diminish theirfuture reproductive success. Populations with high levels of natural herbivory had a greater proportion of reproductive plants thatregressed to nonreproductive stages, probably because reproductive plants in these populations were smaller in size. However, theplant response to herbivory was similar across populations.

Key words: eastern North America; Liliaceae; Odocoileus virginianus; reproductive success; simulated herbivory; spatial varia-tion; timing of herbivory; tolerance; Trillium grandiflorum.

Herbivores generally have negative effects on plant fitness.However, the magnitude of these effects often vary (reviewedin Huntly, 1991; Stowe et al., 2000). Plants will differ in fit-ness responses if they experience herbivory at different inten-sities (e.g., Tolvanen et al., 2001; Hickman and Hartnett,2002), on different tissues (e.g., stem vs. leaf damage; Ehrlen,1995; Houle and Simard, 1996; Marquis, 1996), at differentlife stages (e.g., Warner and Cushman, 2002), or at differenttimes during the growing season (e.g., Maschinski and Whi-tham, 1989; Marquis, 1992; Garcia and Ehrlen, 2002). Evenwhen the same level of herbivory is imposed upon differentindividuals within a population, fitness may vary as a resultof genetic, maternal, or environmental differences amongplants (e.g., Weiner et al., 1997; Juenger and Bergelson, 1998;Agrawal, 1999; Hawkes and Sullivan, 2001; Ferraro and Oes-terheld, 2002).

Although many studies have examined the effects of her-bivory on plant fitness and the variation among individuals intheir responses to herbivory (see earlier citations), far fewerstudies have examined the extent to which plants in differentpopulations vary in their responses to herbivory (e.g., Dyer etal., 1991; Huhta et al., 2000; Loreti et al., 2001). However,the mechanisms that create variation in fitness responses at theindividual level can be expected to create variation in fitnessresponses to herbivory across populations.

First, many plant populations vary in the frequency and pat-tern of consumption by a single herbivore species as a resultof variation in its local abundance or behavior (reviewed inHuntly, 1991). If herbivores vary in their density and/or in

1 Manuscript received 9 January 2003; revision accepted 18 March 2003.The author thanks T.-L. Ashman, J. Chase, R. Collins, S. Kalisz, R. Relyea,

J. Steets and S. Tonsor for discussions and comments and J. Chase, J. Dunn,and J. Kauffman for help in the field. This research was supported by grantsfrom the McKinley and Darbarker research funds, Botany in Action (PhippsConservatory and Botanical Garden), and the National Science Foundation(DEB-0105000). This is Pymatuning Laboratory of Ecology Publication 141.

2 Current address: Department of Zoology, University of Florida, 111 Bar-tram Hall, P.O. Box 118525, Gainesville, Florida 32611-8525 USA (e-mail:tknight@zoo.ufl.edu).

their proportional consumption of a particular plant speciesacross plant populations, then the overall effect of that herbi-vore will vary spatially. Thus, while experiments that exploredthe effects of herbivory by removal (e.g., caging or insecti-cides) have had variable results across sites (e.g., Augustineand McNaughton, 1998; Russell et al., 2001), this variationcould either be due to differential responses of plants or dif-ferential consumption by herbivores.

Second, even if the magnitude of herbivory is the sameamong plant populations, variation among populations in itsseasonal timing can create variation in its average effects onplant fitness. While only a few researchers have examined theeffects of herbivory timing on plant response, they have oftenfound that plants consumed early in the growing season hadhigher fitness than those consumed late (Maschinski and Whi-tham, 1989; Gedge and Maun, 1992; Tiffin, 2000). In thesestudies, which were all on short-lived plant species (annuals,biennials), partial defoliation early in the season was less det-rimental because it allowed more time for regrowth beforereproduction. However, defoliation early in the season may bemore detrimental to herbaceous perennials, which are oftencompletely defoliated (Miller et al., 1992) and unable to re-grow in the same season (e.g., Augustine and Frelich, 1998),because early herbivory causes the plant to lose a larger por-tion of its growing season.

Finally, even if the frequency and pattern of herbivory isidentical among plant populations, the average effects of her-bivory may still differ for a variety of reasons. First, manyabiotic and biotic factors, which may vary among populations,can affect plant growth and survival in the face of herbivory.For example, resource availability can either positively or neg-atively influence the degree to which plants can tolerate her-bivory (reviewed in Hawkes and Sullivan, 2001; Ferraro andOesterheld, 2002). Second, the effects of herbivory may varyamong populations that differ in the frequency of past herbi-vore attacks. The average effects of herbivory may be exac-erbated in populations with previously high levels of herbivoryrelative to populations with previously low levels of herbivory

1208 [Vol. 90AMERICAN JOURNAL OF BOTANY

if past herbivore attacks have physiologically weakened plants(Ehrlen, 2000). Alternatively, the opposite response could oc-cur if after past herbivore attacks plants have either evolvedtraits that confer tolerance to herbivory (reviewed in Straussand Agrawal, 1999; Stowe et al., 2000) or induced defenses(reviewed in Karban et al., 1999).

Clearly, empirical estimation of the effects of herbivory andits variation among plant populations is necessary to accurate-ly assess the effects of an herbivore on any given plant species.In this study, I addressed the effects of herbivory by white-tailed deer on the perennial herb, Trillium grandiflorum (Lil-iaceae). Using both natural and experimentally manipulatedherbivory, I examined the patterns of among-population vari-ation in the average effects of herbivory and evaluated poten-tial mechanisms that contributed to those patterns. Specifically,I studied 12 natural populations and first asked to what degreethe frequency and timing of herbivore attacks naturally variedacross populations. Next, I determined to what extent herbiv-ory and the timing of herbivory affected the future stage ofreproductive plants.

In addition to exploring the natural pattern of herbivory andresponse to herbivory among these 12 populations, several is-sues necessitated experimental manipulation of herbivory.First, it is statistically difficult to assess the average effects ofherbivory in populations with naturally low levels of herbiv-ory. Second, the effects of herbivory on plant size are difficultto discern because the size of plants eaten and not eaten bydeer and the timing of consumption naturally vary. Thus, Iexperimentally manipulated herbivory and the timing of her-bivory in four populations, while controlling for plant stageand size. I compared the responses of plants that were natu-rally eaten by herbivores to those that were experimentallyclipped to determine how well clipping simulated natural her-bivory. Finally, to determine if decreases in plant size due toherbivory could have consequences for future reproductivesuccess, I collected fruits from reproductive plants to deter-mine whether smaller reproductive plants made fewer ovules.

MATERIALS AND METHODS

Study system—Trillium grandiflorum (Michx.) Salisb. (Lilaceae), a long-lived herbaceous perennial, grows in the understory of deciduous foreststhroughout eastern North America (Case and Case, 1997). Trillium grandiflo-rum individuals are nonclonal and persist underground in a dormant stateduring autumn and winter. In northwest Pennsylvania, the leaves appear aboveground in late April, before the forest canopy leafs out, and senesce in lateJuly.

Trillium grandiflorum populations consist of four easily distinguishedaboveground stages: seedling, one-leaf, three-leaf, and reproductive. Repro-ductive plants produce a single stem, a whorl of three leaves, and a singlewhite hermaphroditic flower that gives rise to a single fruit. Plants in thereproductive stage can remain in that stage or regress to the three-leaf (non-reproductive) stage in the next growing season. Both three-leaf and repro-ductive plants can transition into a dormant stage, in which no abovegroundstructures are made for one or more seasons (Hanzawa and Kalisz, 1993).

The 12 populations of T. grandiflorum monitored in this study are all indeciduous forests in northwest Pennsylvania, USA (a 50-km radius, for exactlocations of populations see Supplementary Data accompanying the onlineversion of this article). The populations are separated from each other by anaverage of 15 km. The overstory of all populations was dominated by sugarmaple (Acer saccarum), beech (Fagus sylvatica), and red oak (Quercus rub-ra), but the habitat quality ranged from highly fragmented (area 5 0.25 km2)to near pristine (area 5 25 km2). In addition, the T. grandiflorum populations

themselves varied in density, spatial extent, and stage structure (Knight,2003).

Trillium grandiflorum is a preferred food of white-tailed deer, Odocoileusvirginianus (Augustine and Frelich, 1998), a common and increasingly abun-dant native herbivore (Alverson et al., 1988; McCabe and McCabe, 1997).When deer consume these plants, they usually removing all of its leaf andflower tissue (i.e., complete defoliation). After consumption, plants do notregrow in the same season. However, complete defoliation does not usuallykill this or other forest perennial herbs, but instead reduces growth and re-production in subsequent growing seasons (Edwards, 1985; Whigham, 1990;Primack et al., 1994; Rooney and Waller, 2001). In these T. grandiflorumpopulations, seedlings, one-leaf, and small three-leaf (leaf length , 5 cm)plants were never consumed by deer. These smaller stage classes were fre-quent in these populations, and therefore their lack of consumption likelyrepresents deer preference for larger plants. Only large three-leaf (leaf length. 5 cm) and reproductive plants were consumed by deer, and reproductiveplants were consumed at a greater frequency (Knight, 2003). Thus, the re-mainder of this paper will concentrate solely on reproductive plants.

Natural herbivory—In April 1999, 5–27 1-m2 plots (sample size dependedon density of population) were randomly placed along transects through eachpopulation. In each population, at least 40 reproductive plants were perma-nently tagged within these plots. In general, the same plants were followedin 2000 and 2001, but when necessary, additional plants were tagged to main-tain adequate sample sizes.

All tagged plants were censused biweekly for evidence of deer herbivory.Deer herbivory was easily distinguished by a direct cut on the stem. Plantsconsumed while in bloom (3-wk flowering period in the spring) were classi-fied as eaten early, while plants consumed after bloom but before fruit dropwere classified as eaten late. Small mammal herbivory, in which all threeleaves were defoliated but the stem was left intact, was rare (,1%) in thesepopulations, and plants subjected to this type of herbivory were excluded fromall analyses.

Each tagged plant was scored for stage in April of the following two years(2000 and 2001) when the plants first emerged and before any deer herbivory.Reproductive plants that did not emerge in the following years were assumedto be dormant and not dead. Of the 547 reproductive plants that were moni-tored in 1999, 14 did not re-emerge in 2000 and were scored as dormant. Ofthose 14, 11 reemerged in 2001 and the remaining three were assumed to stillbe dormant. I calculated the frequency in which reproductive plants remainedreproductive, regressed to the three-leaf stage, or became dormant in the nextyear for 1999–2000 and 2000–2001.

I used x2 analysis to determine if (1) the frequency herbivory varied acrosspopulations and (2) the stage of reproductive plants in the next growing seasonvaried among plants consumed and not consumed by deer. A separate x2

analysis was performed for the 1999–2000 and the 2000–2001 data. Theobserved frequency of reproductive plants that became dormant in the nextgrowing season was low (e.g., only 14 out of 547 reproductive plants in 1999became dormant). For this reason, I combined the two nonreproductive stages(dormant, three-leaf) for this analysis.

I used x2 analysis to determine if (1) the timing of herbivory varied acrosspopulations and (2) the stage of reproductive plants in the next growing seasonvaried among plants consumed early and late in the season. In these analyses,only reproductive plants that were consumed by deer were considered. Onlythe 1999–2000 data were used because almost no plants were eaten late inthe season in 2000. Populations with less than five plants consumed by deerwere excluded from the analysis testing if the timing of herbivory variedacross populations. Therefore, only five populations were included in thisanalysis. As described earlier, the two nonreproductive stages (dormant, three-leaf) were combined for this analysis. All x2 analyses were done using SYS-TAT 9.0 (1999).

Experimental herbivory—I experimentally manipulated plants in four pop-ulations, two with low ambient levels of herbivory, DC and RH, and two withhigh ambient levels of herbivory, TW and WH. At each population in 2000,I chose 30 triplets of reproductive plants that were not a part of the previous

August 2003] 1209KNIGHT—HERBIVORY ACROSS POPULATIONS OF TRILLIUM GRANDIFLORUM

Fig. 1. Percentage of plants of Trillium grandiflorum consumed by deerin different populations. Hatched 5 1999, black 5 2000. Each population isabbreviated with a two-letter code (for full names see Supplementary Dataaccompanying the online version of this paper). Populations vary in the per-centage of plants consumed by deer (in 1999, x2 5 149.07, df 5 11, P ,0.001; in 2000, x2 5 166.46, df 5 11, P , 0.001).

Fig. 2. Percentage of reproductive plants of Trillium grandiflorum eatenby deer early (white portion of the bar) and late (black portion of the bar) inthe season. Only populations in which herbivory was observed in at least fiveplants were included. Bars left of the dotted line 5 1999; bars right of thedotted line 5 2000. Populations vary in the timing of herbivory in 1999 (x2

5 65.19, df 5 4, P , 0.001), but not 2000. Abbreviations of populationsspelled out in Supplementary Data.

censuses. I chose each set of three based on similarity in size (within 1 cmin leaf length) and proximity (within 1 m) to each other. Each triplet wasseparated from other triplets by at least 3 m. Within each triplet, I randomlyassigned one of three herbivory treatments to each plant; clipped early (1May, while the plants were in bloom), clipped late (30 June, just prior to fruitdrop), or control (unmanipulated). I clipped plants with scissors 5 cm fromthe base of the stem, which mimics deer herbivory. In 2001, I scored thestage of each plant. I measured the leaf length of all plants on 1 May in 2000and 2001. I calculated the relative growth rate of plants between years as:RGR 5 (leaf length in 2001 2 leaf length in 2000)/(leaf length in 2000).Leaf length has been shown to be highly correlated with other size-relatedtraits of T. grandiflorum, such as leaf area, plant height, and stem diameter,and should thus be a good indicator of overall plant size (Rooney and Waller,2001).

To determine if experimental clipping effectively simulated natural deerherbivory, I used x2 analysis to test if the stage of reproductive plants in thenext year (reproductive, nonreproductive) differed between the two types ofherbivory (natural, experimental). In this analysis, only plants natural eatenearly and clipped early in TW and WH were included. Late herbivory wasnot considered because natural herbivory late in the season was rare in 2000.Further, plants in DC and RH were not included in this analysis becausenatural herbivory was rare in these populations.

To determine if plants in different populations responded differently to theclipping treatments (i.e., a three-way interaction between population, stagetransitions, and clipping), I used hierarchical log-linear analysis (Sokal andRohlf, 1995). Because the observed frequency of reproductive plants thattransitioned into dormancy was low (6 of 341 experimental plants), the twononreproductive stages (dormant, three-leaf) were combined.

Differences in the average pretreatment size of plants among populationsmay cause differences among populations in the frequencies by which plantsremain in the reproductive stage (e.g., if smaller plants are more likely toregress to nonreproductive stages). I used ANOVA followed by Tukey’s HSDto test for main and pairwise differences between populations in pretreatmentsize. I used logistic regression to determine if smaller plants were more likelyto regress to nonreproductive stages (only plants in the control treatment wereincluded in this analysis).

I used two-way ANOVA to test for differences among populations, clippingtreatments, and their interaction on the relative growth rate (RGR). Withineach population, pairwise differences among clipping treatments in RGR weretested with Tukey’s HSD. I only included the RGR of plants that remainedreproductive throughout this experiment, and therefore changes in RGR areindependent from changes in stage. I used SYSTAT 9.0 (1999) for thesestatistical tests.

Relationship between plant size and number of ovules—Plants that makemore ovules have the potential to make more seeds and thus have greaterfemale reproductive success. Because pollen and resource limitation may in-terfere with the number of ovules that actually become mature seeds, thenumber of ovules indicates the potential female reproductive success of plants.To determine if larger plants had more potential reproductive success, in 2000I randomly chose approximately 25 reproductive plants that were not part ofeither the natural or the experimental herbivory studies from each of eightpopulations (DC, DH, EL, LR, RH, TW, WC, WH) (N 5 180 plants total). Imeasured their leaf length and collected their fruits. I counted the seeds andunfertilized ovules per fruit using a dissecting microscope. I used linear re-gression to determine the relationship between plant size (leaf length) and thetotal number of ovules (total number of ovules 5 number of seeds 1 numberof unfertilized ovules). The total number of ovules was log transformed. Be-cause the linear regression did not differ among populations, populations werepooled.

RESULTS

Natural herbivory—The frequency of herbivory varied dra-matically across these 12 T. grandiflorum populations (Fig. 1).The percentage of reproductive plants eaten by deer in a pop-ulation ranged from 0 to 52% in 1999 and 0 to 61% in 2000(Fig. 1).

The timing of herbivory on reproductive plants variedacross populations in 1999 (Fig. 2), but not in 2000. In 1999,of the reproductive plants eaten by deer, the percentage eatenearly in the season (rather than late in the season) ranged from5 to 100% among the five populations with herbivory (Fig.2). In 2000, nearly all of the herbivory occurred early in theseason (only one of the 123 consumed plants was eaten latein the season).

Herbivory affected the future stage of reproductive plants.In both 1999 and 2000, plants that were eaten by deer weremore likely to regress to nonreproductive stages and less likelyto remain in the reproductive stage in the next year than plantsthat were not eaten by deer (Fig. 3). The timing of herbivoryalso affected the future stage of reproductive plants (Fig. 3).Plants consumed early in the season were more likely to re-gress to nonreproductive stages and less likely to remain inthe reproductive stage in the next year than plants eaten latein the season.

1210 [Vol. 90AMERICAN JOURNAL OF BOTANY

Fig. 3. The percentage of not eaten (top two panels) and eaten (bottom two panels) reproductive plants of Trillium grandiflorum that remained in thereproductive stage, regressed to the three-leaf stage, and became dormant in the next year. The arrow from reproductive to three-leaf indicates the percentageof reproductive plants that regressed to the three-leaf stage; that from reproductive to dormant indicates the percentage of reproductive plants that becamedormant. The arrow looping back to the reproductive stage indicates the percentage of reproductive plants that remained reproductive. The stage transitions ofreproductive plants in 1999 (left panels) and 2000 (right panels) are shown. The stage transitions for reproductive plants eaten early and late in the season in1999 are shown in parentheses. Plants eaten by deer are more likely to regress to nonreproductive (three-leaf and dormant) stages (in 1999, x2 5 85.01, df 51, P , 0.001; in 2000, x2 5 86.53, df 5 1, P , 0.001). Plants eaten early in the season were more likely to regress to nonreproductive stages than those eatenlate (x2 5 58.35, df 5 1, P , 0.001).

TABLE 1. Results of a hierarchical log-linear analysis of the three-waycontingency table of population (N 5 4), clipping (N 5 3: early,late, control), and stage in the next year (N 5 2: reproductive,nonreproductive). A nonsignificant three-way interaction betweenpopulation, clipping, and stage indicates that reproductive Trilliumgrandiflorum responded similarly to clipping across populations. Asignificant interaction between clipping and stage indicates that thestage of plants in the next year varied across clipping treatments.Specifically, plants clipped early in the season (which simulatedearly-season herbivory by white-tailed deer) were more likely toregress to nonreproductive stages than those clipped late or not atall (control). A significant interaction between population and stageindicates that the stage of plants in the next year varied acrosspopulations.

Term df G2 P

Population 3 clipping 3 stageClipping 3 stagePopulation 3 stage

623

3.7080.3822.01

0.72,0.001,0.001

Experimental herbivory—Clipping plants effectively sim-ulated natural herbivory. Plants eaten early in the season bydeer had similar frequencies in the reproductive and three-leafstage classes in the next year as plants clipped early in bothTW (N 5 23 naturally eaten, N 5 27 clipped, x2 5 0.42, df5 1, P 5 0.52) and WH (N 5 33 naturally eaten, N 5 27clipped, x2 5 0.021, df 5 1, P 5 0.89).

The effect of clipping on the stage of plants in the next yeardid not vary among populations (a nonsignificant three-wayinteraction, Table 1). In all populations, early but not late clip-ping reduced the proportion of reproductive plants that re-mained reproductive (Fig. 4).

The frequency of plants in the reproductive and nonrepro-ductive stages in the next year differed across populations (Ta-ble 1), which most likely resulted from differences amongpopulations in pretreatment plant size (ANOVA: F3, 341 5 63.5,r2 5 0.36, P , 0.001) (Fig. 4). Specifically, plants in popu-lations with high ambient levels of herbivory, TW and WH,had significantly smaller pretreatment sizes than plants in pop-ulations with low ambient levels of herbivory, DC and RH(Tukey’s HSD P , 0.001). Smaller plants were more likely toregress to nonreproductive stages (logistic regression of plants

August 2003] 1211KNIGHT—HERBIVORY ACROSS POPULATIONS OF TRILLIUM GRANDIFLORUM

Fig. 4. The frequency of transitions into different stages in 2001 by reproductive plants of Trillium grandiflorum in each size class in 2000 after clippingat various times. Size class was based on the leaf length of the plant in 2000. Plants in the left, middle, and right panels were in the early, late, and controlclipping treatments, respectively (clipped in 2000). Shading within each bar indicates the proportion of reproductive plants from 2000 that remained reproductive(white), regressed to three-leaf (grey), or became dormant (checked) in 2001. Two populations (DC, RH) had low ambient levels of herbivory, and two (TW,WH) had high ambient levels of herbivory. Plants clipped early and smaller plants were more likely to regress to nonreproductive stages. See SupplementaryData for meaning of population abbreviations.

in the control treatment, P , 0.01). The average pretreatmentleaf length of control plants was 7.6 cm for plants that re-gressed to nonreproductive stages in the next year and 9.2 cmfor plants that remained reproductive.

Clipping decreased the future size of reproductive plants.Among the reproductive plants in 2000 that remained repro-ductive in 2001, the relative growth rates of plants differedsignificantly among clipping treatments (F2, 207 5 6.98, P 50.001), populations (ANOVA: F3, 207 5 5.09, P 5 0.002), andtheir interaction (F6, 207 5 2.87, P 5 0.01). While the effect ofclipping treatment on growth rate varied among populations,in general, clipped plants had reduced growth relative to un-clipped plants (Fig. 5).

Relationship between plant size and number of ovules—There was a positive relationship between leaf length (in mil-limeters) and the log total number of ovules (linear regression,y 5 0.017x 1 1.624, r2 5 0.37, P , 0.001) (Fig. 6).

DISCUSSION

The amount of herbivory on Trillium grandiflorum washighly variable across the 12 study populations (Fig. 1). Somepopulations experienced almost no herbivory, while others hadover 60% of their reproductive plants eaten in one or bothyears. White-tailed deer do not solely depend on forest plantssuch as T. grandiflorum for their existence (Augustine andMcNaughton, 1998), and thus, there are many potential factorsat a landscape level that could cause some populations to ex-perience higher levels of herbivory than others. These includethe proximity of the populations to busy roads; the availabilityof other edible forest species; the level of habitat fragmenta-tion; and the adjacent land-use practices (e.g., agriculture, hu-man habitation). These factors may also create variation in thetiming of herbivory across these populations (Fig. 2).

Because natural and experimental herbivory affected plantssimilarly (in terms of the stage in the next growing season), I

1212 [Vol. 90AMERICAN JOURNAL OF BOTANY

Fig. 5. The relative growth rate (61 SE) of reproductive plants of Trilliumgrandiflorum that remained reproductive in four populations (DC, RH, TW,WH; see Supplementary Data for meanings of abbreviations) after beingclipped early in the season, late in the season, or not at all (control). Differentletters within a population indicate significant (Tukey’s, P , 0.05) differencesbetween treatments. Only one reproductive plant clipped early at WH re-mained reproductive; this treatment was excluded from these analyses. Plantsthat were clipped (early and late) had lower growth rates than unclipped con-trol plants.

Fig. 6. The log total number of ovules (seeds 1 unfertilized ovules) perfruit and size (leaf length) of reproductive plants of Trillium grandiflorum in2000 under natural field conditions. Smaller plants have less potential repro-ductive success.

will discuss the observational and experimental results simul-taneously. Eaten reproductive plants were three times morelikely to regress to nonreproductive stages than uneaten plants(Figs. 3, 4). Herbivory also decreased the growth rate of thoseplants that did remain reproductive (Fig. 5) (see also Ander-son, 1994). Because smaller plants produce fewer ovules (Fig.6), plants consumed by deer will have lower potential repro-ductive output in the next growing season. These results areconsistent with other studies at the population level in whichincreases in the frequency of herbivory caused plants to de-crease in size and regress in stage (Rauscher and Feeney, 1980;Doak, 1992; Bastrenta et al., 1995; Ehrlen, 1995).

Early-season herbivory was more detrimental than late-sea-son herbivory. Because deer completely defoliate T. grandiflo-rum, plants eaten both early and late in the season have acomplete loss of female reproductive success for that season.However, in the following growing season, more of the plantsthat were consumed early in the season were nonreproductive,whereas plants that were eaten late in the season were just aslikely to be reproductive in the next season as those that werenot consumed at all.

While several other studies have considered the effects ofherbivory timing, this is the first on a perennial species that iscompletely defoliated. Studies on short-lived annual and bi-ennial plants have found that plants eaten earlier had higherreproductive success than those consumed later in the season(e.g., Maschinski and Whitham, 1989; Gedge and Maun, 1992;Tiffin, 2000). These plants have few stored resources and re-produce only once. Thus, early-season herbivory allowed moretime for regrowth in that season prior to reproduction. Studieson these plants differ from T. grandiflorum in that completedefoliation does not occur and the response to herbivory ismeasured only in the same growing season in which the her-bivory occurs.

In long-lived plants, late-season herbivory should generallybe less detrimental than early-season herbivory, as was ob-served here. Two nonexclusive mechanisms contribute to thisresult. First, if plants cannot regrow leaves until the next sea-son, then a plant eaten early has less potential for photosyn-thesis in the current growing season than a plant eaten late.Second, plants that are limited by light from a canopy may domost of their photosynthesis early in the season, prior to theleaf canopy development (e.g., Routhier and Lapointe, 2002).In this scenario, plants eaten during this early period will losethis peak time for photosynthesis in a given season. Both ofthese mechanisms could occur in T. grandiflorum, because (1)defoliation by deer is complete and individuals cannot regrowfollowing herbivory until the next growing season, and (2)plants emerge in the early spring, prior to forest canopy clo-sure.

Other studies on perennial plants have shown that early-season herbivory was more detrimental than late (e.g., Mar-quis, 1992; Ehrlen, 1995; Garcia and Ehrlen, 2002). For ex-ample, although Ehrlen (1995) did not explicitly consider thetiming of herbivory on Lathyrus vernus, differences in timingof plant consumption by vertebrate and mollusk herbivoresmay have contributed to differences in their effects. Vertebrategrazers removed much more biomass than mollusks, but sur-prisingly, mollusks had more severe effects on survival,growth, and reproduction. One explanation for this result isthat mollusks removed meristematic tissue early in the season,which greatly inhibited plant growth for the remainder of theseason. Vertebrate grazers removed both vegetative and mer-istematic tissue much later in the season, after the plants hada longer opportunity to grow and acquire resources. Similarly,in an experimental study manipulating the timing of leaf re-moval in Primula veris, Garcia and Ehrlen (2002) found thatearly-season removal diminished current-year reproductionand future growth while late-season removal did not influenceany fitness components. They suggested that P. veris acquiresmuch of its resources for reproduction and storage early in the

August 2003] 1213KNIGHT—HERBIVORY ACROSS POPULATIONS OF TRILLIUM GRANDIFLORUM

growing season, and thus herbivory at this time is most det-rimental.

Although all populations responded similarly to natural andexperimental herbivory, reproductive plants in populationswith high ambient levels of herbivory were smaller in size andmore likely to regress to nonreproductive stages than repro-ductive plants in populations with low ambient levels of her-bivory (Fig. 4). The legacy of past herbivory attacks may beresponsible for the smaller size of these plants. In this scenario,I would suggest that there is nothing fundamentally differentabout the traits of the individuals across populations, or theirenvironments, aside from past herbivory, that causes them torespond differentially to herbivory. Alternatively, I cannot ruleout the slightly more complex scenario that unseen environ-mental or genetic differences among these populations causedthe initial differences in size of reproductive plants and thustheir future stage.

Herbivory and the timing of herbivory affect both currentand future reproduction of T. grandiflorum. Defoliation of T.grandiflorum removes all vegetative and reproductive struc-tures and as a result, consumed plants lose all of their femalereproductive success for the current growing season, regardlessof when they are eaten. However, because plants eaten earlyin the season are more likely to regress to nonreproductivestages than plants eaten late in the season, early-season her-bivory diminishes future reproduction more than late-seasonherbivory. In addition, even if plants are able to remain repro-ductive following a bout of herbivory, they suffer a reductionin size. Such decreases in the size of the reproductive plantsresults in a decrease in the number of ovules those plants willproduce, because plant size is positively correlated with thenumber of ovules (Fig. 6). Thus, herbivory affects the futurereproduction of a plant either by causing it to regress to anonreproductive stage (thereby having no reproduction in thefollowing season) or by reducing its number of ovules.

This documented herbivory-driven reduction in female re-productive output may be accompanied by a change in the sexstructure of the population. Recently, in another study on T.grandiflorum, Wright and Barrett (1999) demonstrated thatsmaller plants allocated significantly more resources to malefunction, while larger plants allocated significantly more re-sources to female function. In the current study, herbivory re-duced the overall size of reproductive plants and may causethe overall population to become ‘‘more male.’’

By increasing the probability that reproductive plants re-gress to nonreproductive stages, herbivory reduces the numberof reproductive plants in a population. If white-tailed deer areless likely to forage on T. grandiflorum when its density islow, then herbivory will be density dependent. However, pol-linators may also be less likely to forage on T. grandiflorumwhen its density is low. Trillium grandiflorum is an obligateoutcrossing species that relies on its insect pollinators (Kaliszet al., 1999; Irwin, 2000). In a concurrent study, I found thatlow-density populations are more pollen limited (Knight,2003). Thus, deer herbivory may cause Trillium grandiflorumpopulations to be inverse-density dependent (i.e., have Alleeeffects).

To determine how important biotic factors, such as herbiv-ory, will affect the persistence of a species, studies must con-sider how the magnitude and pattern of herbivory, as well asthe plant response to herbivory, vary across populations. Mystudy indicates that both the magnitude and timing of herbiv-ory vary across populations of T. grandiflorum and that both

significantly affect the future stage and size of reproductiveplants. Plants in all populations responded similarly to herbiv-ory. This suggests that the frequency and timing of herbivoryalone should indicate how plants in a population will be af-fected by herbivory and that more detailed knowledge of en-vironmental characteristics and plant traits may not be neces-sary for T. grandiflorum. Alternatively, environmental char-acteristics at larger spatial and temporal scales than those con-sidered in this study may influence the plant response toherbivory. Understanding the mechanisms that create differ-ences across T. grandiflorum populations in the average fre-quency by which reproductive plants remain reproductive andregress to a nonreproductive stage is more complex. One po-tential mechanism is that populations that have experiencedherbivory in the past may contain plants that are of smallersize and consequently more likely to regress to nonreproduc-tive stages.

LITERATURE CITED

AGRAWAL, A. A. 1999. Induced responses to herbivory in wild radish: effectson several herbivores and plant fitness. Ecology 80: 1713–1723.

ALVERSON, W. S., D. M. WALLER, AND S. L. SOLHEIM. 1988. Forests toodeer: edge effects in northern Wisconsin. Conservation Biology 2: 348–358.

ANDERSON, R. C. 1994. Height of white-flowered trillium (Trillium grandi-florum) as an index of deer browsing intensity. Ecological Applications4: 104–109.

AUGUSTINE, D. L., AND L. E. FRELICH. 1998. Effects of white-tailed deer onpopulations of an understory forb in fragmented deciduous forests. Con-servation Biology 12: 995–1004.

AUGUSTINE, D. L., AND S. J. MCNAUGTON. 1998. Ungulate effects on thefunctional species composition of plant communities: herbivore selectiv-ity and plant tolerance. Journal of Wildlife Management 62: 1165–1182.

BASTRENTA, B., J. LEBRETON, AND J. D. THOMPSON. 1995. Predicting de-mographic change in response to herbivory: a model of the effects ofgrazing and annual variation on the population dynamics of Anthyllisvulneraria. Journal of Ecology 83: 603–611.

CASE, F. W., JR., AND R. B. CASE. 1997. Trilliums. Timber Press, Portland,Oregon, USA.

DOAK, D. F. 1992. Lifetime impacts of herbivory for a perennial plant. Ecol-ogy 73: 2086–2099.

DYER, M. I., M. A. ACRA, G. M. WANG, D. C. COLEMAN, D. W. FRECKMAN,S. J. MCNAUGHTON, AND B. R. STRAIN. 1991. Source-sink carbon re-lations in 2 Panicum coloratum ecotypes in response to herbivory. Ecol-ogy 72: 1472–1483.

EDWARDS, J. 1985. Effects of herbivory by moose on flower and fruit pro-duction of Aralia nudicaulis. Journal of Ecology 73: 861–868.

EHRLEN, J. 1995. Demography of the perennial herb Lathyrus vernus. II.Herbivory and population dynamics. Journal of Ecology 83: 297–308.

EHRLEN, J. 2000. The dynamics of plant populations: does the history ofindividuals matter? Ecology 81: 1675–1684.

FERRARO, D. O., AND M. OESTERHELD. 2002. Effect of defoliation on grassgrowth. A quantitative review. Oikos 98: 125–133.

GARCIA, M. B., AND J. EHRLEN. 2002. Reproductive effort and herbivorytiming in a perennial herb: fitness components at the individual and pop-ulation levels. American Journal of Botany 89: 1295–1302.

GEDGE, K. E., AND M. A. MAUN. 1992. Effects of simulated herbivory ongrowth and reproduction in two beach annuals, Cakile edentula and Cor-ispermum hyssopifolium. Canadian Journal of Botany 70: 2467–2475.

HANZAWA, F. M., AND S. KALISZ. 1993. The relationship between age, size,and reproduction in Trillium grandiflorum (Lilaceae). American Journalof Botany 80: 405–410.

HAWKES, C. V., AND J. J. SULLIVAN. 2001. The impact of herbivory on plantsin different resource conditions: a meta-analysis. Ecology 82: 2045–2058.

HICKMAN, K. R., AND D. C. HARTNETT. 2002. Effects of grazing intensityon growth, reproduction, and abundance of three palatable forbs in Kan-sas tallgrass prairie. Plant Ecology 159: 23–33.

HOULE, G., AND G. SIMARD. 1996. Additive effects of genotype, nutrientavailability and type of tissue damage on the compensatory response of

1214 [Vol. 90AMERICAN JOURNAL OF BOTANY

Salix planifolia ssp. planifolia to simulated herbivory. Oecologia 107:373–378.

HUHTA, A. P., T. LENNARTSSON, J. TUOMI, P. RAUTIO, AND K. LAINE. 2000.Tolerance of Gentianella campestris in relation to damage intensity: aninterplay between apical dominance and herbivory. Evolutionary Ecology14: 373–392.

HUNTLY, N. 1991. Herbivores and the dynamics of communities and eco-systems. Annual Review of Ecology and Systematics 22: 477–503.

IRWIN, R. E. 2000. Morphological variation and female reproductive successin two sympatric Trillium species: evidence for phenotypic selection inTrillium erectum and Trillium grandiflorum (Liliaceae). American Jour-nal of Botany 87: 205–214.

JUENGER, T., AND J. BERGELSON. 1998. Pairwise versus diffuse natural se-lection and the multiple herbivores of scarlet gilia, Ipomopsis aggregata.Evolution 52: 1583–1592.

KALISZ, S., F. M. HANZAWA, S. J. TONSOR, D. A. THIEDE, AND S. VOIGT.1999. Ant-mediated seed dispersal alters pattern of relatedness in a pop-ulation of Trillium grandiflorum. Ecology 80: 2620–2634.

KARBAN, R., A. A. AGRAWAL, J. S. THALER, AND L. S. ADLER. 1999. In-duced plant responses and information content about risk of herbivory.Trends in Ecology and Evolution 14: 443–447.

KNIGHT, T. M. 2003. The effects of herbivory and pollen limitation on thepopulation dynamics of Trillium grandiflorum. Ph.D. dissertation, Uni-versity of Pittsburgh, Pittsburgh, Pennsylvania, USA.

LORETI, J., M. OESTERHELD, AND O. SALA. 2001. Lack of intraspecific var-iation in resistance to defoliation in a grass that evolved under lightgrazing pressure. Plant Ecology 157: 195–202.

MARQUIS, R. J. 1992. A bite is a bite is a bite? Constraints on response tofolivory in Piper arianum. Ecology 73: 143–152.

MARQUIS, R. J. 1996. Plant architecture, sectoriality and plant tolerance toherbivores. Vegetatio 127: 85–97.

MASCHINSKI, J., AND T. G. WHITHAM. 1989. The continuum of plant respons-es to herbivory: the influence of plant association, nutrient availability,and timing. American Naturalist 134: 1–19.

MCCABE, R. E., AND T. R. MCCABE. 1997. Recounting whitetails past. In W.J. McShea, H. B. Underwood, and J. H. Rappole [eds]. The science ofoverabundance: deer ecology and population management, 11–26. Smith-sonian Institution Press, Washington D.C., USA.

MILLER, S. G., S. P. BRATTON, AND J. HADIDIAN. 1992. Impacts of white-tailed deer on endangered and threatened vascular plants. Natural AreasJournal 12: 67–74.

PRIMACK, R. B., S. L MIAO, AND K. R. BECKER. 1994. Costs of reproduction

in the pink lady’s slipper orchid (Cypripedium acaule): defoliation, in-creased fruit production, and fire. American Journal of Botany 81: 1083–1090.

RAUSCHER, M. D., AND P. FEENEY. 1980. Herbivory, plant density, and plantreproductive success: the effect of Battus philenor on Aristolochia retic-ulata. Ecology 61: 905–917.

ROONEY, T. P., AND D. M. WALLER. 2001. How experimental defoliation andleaf height affect growth and reproduction in Trillium grandiflorum.Journal of the Torrey Botanical Society 128: 393–399.

ROUTHIER, M. C., AND L. LAPOINTE. 2002. Impact of tree leaf phenology ongrowth rates and reproduction in the spring flowering species Trilliumerectum (Liliaceae). American Journal of Botany 89: 500–505.

RUSSELL, F. L., D. B. ZIPPIN, AND N. L. FOWLER. 2001. Effects of white-tailed deer (Odocoileus virginianus) on plants, plant populations andcommunities: a review. American Midland Naturalist 146: 1–26.

SOKAL, R. R., AND F. J. ROHLF. 1995. Biometry. W. H. Freeman, New York,New York, USA.

STOWE, K. A., R. J. MARQUIS, C. J. HOCHWENDER, AND E. L. SIMMS. 2000.The evolutionary ecology of tolerance to consumer damage. Annual Re-view of Ecology and Systematics 31: 565–595.

STRAUSS, S. Y., AND A. A. AGRAWAL. 1999. The ecology and evolution ofplant tolerance to herbivory. Trends in Ecology and Evolution 14: 179–185.

SYSTAT. 1999. User’s guide version 9. SPSS Institute, Chicago, Ilinois,USA.

TIFFIN, P. 2000. Mechanisms of tolerance to herbivore damage: what do weknow? Evolutionary Ecology 14: 523–536.

TOLVANEN, A., J. SCHRODERUS, AND G. H. R. HENRY. 2001. Age- and stage-based bud demography of Salix arctica under contrasting muskox graz-ing pressure in the high Arctic. Evolutionary Ecology 15: 443–462.

WARNER, P. J., AND J. H. CUSHMAN. 2002. Influence of herbivores on aperennial plant: variation with life history stage and herbivore species.Oecologia 132: 77–85.

WEINER, J., S. MARTINEZ, H. MULLER-SCHARER, P. STOLL, AND B. SCHMID.1997. How important are environmental maternal effects in plants? Astudy with Centaurea maculosa. Journal of Ecology 85: 133–142.

WHIGHAM, D. F. 1990. The effect of experimental defoliation on the growthand reproduction of a woodland orchid, Tipularia discolor. CanadianJournal of Botany 68: 1812–1816.

WRIGHT, S. I., AND S. C. H. BARRETT. 1999. Size-dependent gender modi-fication in a hermaphroditic perennial herb. Proceedings of the RoyalSociety of London Series B 266: 225–232.