Transcript
Titratingthecostofplanttoxinsagainstpredators:determiningthetippingpointforforagingherbivores CarolynL.Nersesian 1 *,PeterB.Banks 2 andClareMcArthur 1 1 SchoolofBiologicalSciences,UniversityofSydney,Sydney,NSW2006,Australia;and 2 SchoolofBiological,Earth,EnvironmentalScience,UniversityofNewSouthWales,Sydney,NSW2052,Australia Summary 1. Foraging herbivores must deal with plant characteristics that inhibit feeding and they mustavoid being eaten. Principally, toxins limit food intake, while predation risk alters how long ani-mals are prepared to harvest resources. Each of these factors strongly affects how herbivores usefood patches, and both constraints can pose immediate proximate costs and long-term conse-quencestofitness. 2. Using a generalist mammalian herbivore, the common brushtail possum ( Trichosurus vulpecu-la ),ouraimwastoquantitativelycomparetheinfluenceofplanttoxinandpredationriskonforag-ingdecisions. 3. We performed a titration experiment by offering animals a choice between non-toxic food at arisky patch paired with food with one of five toxin concentrations at a safe patch. This allowed usto identify the tipping point, where the cost of toxin in the safe food patch was equivalent to theperceivedpredationriskinthealternativepatch. 4. At low toxin concentration, animals ate more from the safe than the risky patch. As toxin con-centration increasedatthesafe patch,intakeshifteduntilanimalsatemainlyfrom theriskypatch.This shift was associated with behavioural changes: animals spent more time and fed longer at theriskypatch,whilevigilanceincreasedatbothriskyandsafepatches. 5. Ourresultsdemonstratethatthevariationintoxinconcentration,whichoccursintraspecificallyamongplants,cancritically influencetherelativecost ofpredation risk onforaging.Weshowthatherbivores quantify, compare and balance these two different but proximate costs, altering theirforaging patterns in the process. This has potential ecological and evolutionary implications forthe production of plant defence compounds in relation to spatial variation in predation risk toherbivores. Key-words: behaviours, foraging, herbivore, interactions, patch use, plant secondary metabo-lites,planttoxins,predationrisk,risktitration,tippingpoint Introduction Herbivores face two key constraints when foraging. Theymust deal with plant characteristics that inhibit feeding andthey must avoid being eaten. Both constraints pose immedi-ate costs to foraging with potential long-term fitness conse-quences (Freeland & Janzen 1974; Lima & Dill 1990).Understanding the influences of plants and predators is thefocus of ecological research on plant–herbivore and preda-tor–prey interactions, respectively. Considering themtogether is critical forquantifying their neteffecton foragingherbivores.For many herbivores, consuming leaves of trees andshrubs means regulating intake of plant secondary metabo-lites, including toxins (Freeland & Janzen 1974). Ingestingtoxins entails a measurable physiological cost as they areabsorbed, metabolized and excreted (Freeland & Janzen1974;Foley&McArthur1994).Detoxificationisarate-limit-ingprocess(Foley&McArthur1994),sotheamountoffoodan animal can eat depends on how quickly it can detoxifyingested compounds. Thus, a common effect of toxins is tolimit daily intake. This has been demonstrated in a range of marsupial folivores: Trichosurus vulpecula and Pseudocheirus peregrinus (common brushtail and ringtail possums) (Marsh et al. 2003; Marsh, Wallis & Foley 2005), Phascolarctoscinereus (koalas) (Lawler et al. 1998),andin eutherian herbi-vores: Sciurus aberti (tassel-earedsquirrels)(Farentinos et al. *Correspondence author. E-mail:
[email protected]@bio.ulaval.ca JournalofAnimalEcology 2011, 80 ,753–760 doi:10.1111/j.1365-2656.2011.01822.x 2011TheAuthors.JournalofAnimalEcology 2011BritishEcologicalSociety 1981), Lepus americanus (snowshoe hares) (Bryant et al. 1989), lambs (Dziba & Provenza 2008) and deer (Duncan,Hartley&Iason1994).Toxinscanalsoreducegrowthrateasseen with insects, such as grasshoppers, Schistocerca ameri-cana (Bernays et al. 1994).The process of detoxification requires that herbivores stopeatingoncetheyreachaphysiologicalconstraint.Herbivoresmaymitigatetoxiceffectsbehaviourallybyslowingtheirrateoffeedingand ⁄ orspendinglesstimefeedingperfeedingbout(Wiggins et al. 2003; Sorensen, Heward & Dearing 2005).They may also mitigate these effects through diet mixing,consuming a variety of plant species that differ in toxin pro-file and are detoxified through alternative metabolic path-ways (Dearing & Cork 1999; Wiggins, McArthur & Davies2006). The link between increased intake and use of differentmetabolic pathways was recently demonstrated empiricallyinbrushtailpossums(Marsh,Wallis&Foley2005).Notably,the effects of increasing concentration of a single toxin onforaging behaviours are similar to that of reducing the diver-sity of plant toxins ⁄ species (Wiggins, McArthur & Davies2006).Along with toxic constraints, herbivores must avoidbecoming food themselves. Predation risk varies spatially inthelandscapebasedonhuntingbehaviours,densityofpreda-tors, as well as the structural complexity of vegetation andopportunities for escape or refugia (Kotler et al. , 1993,Brown, 1999). Therefore, risk of predation has the potentialto impose significant non-lethal effects on prey fitness andforaging patterns (Lima & Dill 1990; Preisser, Bolnick & Be-nard2005).Patch use by herbivores under predation risk varies inresponsetoindirectcues,suchaslight(Kotler1984)andcov-er ⁄ open microhabitats (Kotler, Brown & Knight 1999; Al-tendorf et al. 2001),andtodirectcuessuchasthepresenceof predators (Kotler et al. , 2004) or their odours (Shrader et al. 2008, Apfelbach et al. , 2005). Results consistently demon-strate that foragers prefer food patches that they considersafe but they can be forced to use risky patches whenresources are sufficiently depleted and intake is constrainedby harvesting costs in safe patches (Lima & Dill 1990;Hughes&Ward1993).The relative effect of plants and predators on herbivores,at the population level, has long been debated (Murdoch,1966, Hairston et al. , 1960, Fretwell, 1987). Interestingly, atthe individual level, theoretical models have often dealt withthe two separately, focusing on the influence of food qualityonforaging(Macarthur&Pianka1966),intermsofnutritionand plant secondary metabolites (as toxins and digestibilityreducers) (Freeland & Janzen 1974, Pulliam, 1975, Rauben-heimer and Simpson, 1997) or on the influence of predatorsonforagingbyprey(Charnov,1976,Brown,1988).Many empirical studies have examined tri-trophic effectsfrom predators and plant defences on herbivore foraging,focusing on interactive consequences associated with forag-ing trade-offs, in relation to plant nutrient content and ⁄ orplant species identity (Price et al. 1980; Leibold 1989; Duffy& Hay 1994; van der Stap et al. 2007). Several have recentlyrevealed the importance of plant secondary metabolites inthese interactive influences on foraging behaviour of variousmammalian species, using dichotomous comparisons for thepresence ⁄ absence of toxins or digestibility reducers and(generally) risky vs. safe habitats (Schmidt 2000; Fedriani &Boulay 2006; Hochman & Kotler 2006; Shrader et al. 2008).These studies demonstrate that herbivores incorporate bothplantsecondarymetabolites(presenceorabsence)andpreda-tion risk in their foraging decisions. However, as toxins arenot simply present or absent, but occur at a range of concen-trations in the landscape, it is important to incorporate thisvariation,alongwithpredationrisk,intothepicture.Conceptually, we can combine the costs of toxins and pre-dation risk by asking the question: how much toxin equatesto how much ‘fear’ (Fig. 1)? This framework encompassesthe view that foraging decisions represent titrations of mar-ginalcostsandmarginalbenefits within,andmarginalvaluesbetween food patches (Kotler & Blaustein 1995). Therefore,as animals are predicted to prefer safe to risky patches, otherthings being equal, we expect that an increase in the concen-trationofasingletoxinatasafepatchshouldatsomepoint– which we term the tipping point – forceanimals to shift fromforaging more in the safe patch to more in a risky, but non-toxic patch. Our aim was to test this conceptual model bytitrating toxin concentration against predation risk, similarto the laboratory method used for quantitative chemicalanalysis to determine an unknown concentration of a knownreagent.We used common brushtail possums ( T. vulpecula ), as ourmodel herbivore. These are small (2–5 kg), nocturnal gener-alist marsupial herbivores that are predominantly arboreal,although they come to ground to move between trees andoccasionally to forage (MacLennan 1984). They feed mainlyon eucalypt leaves (Dearing & Cork 1999); however, groundvegetation, such as grasses, can comprise up to 25% of theirdiet (Kerle1984).They are foundwidely throughoutAustra- Fig. 1. Conceptualmodelrelatingfoodintake(orotheroutcomevar-iable)atasafe-toxicpatchvs.ariskynon-toxicpatch.Astheconcen-tration of toxin increases at the safe patch, animals are predicted toeatlessthere andeatmore atthe risky patchto compensate. Thetip-ping point – when the cost of fear and toxin is equivalent – occurswhentheintakeateachpatchisthesame. 754 C.L.Nersesian,P.B.Banks&C.McArthur 2011TheAuthors.JournalofAnimalEcology 2011BritishEcologicalSociety, JournalofAnimalEcology , 80 ,753–760 lia in both natural and human-modified environments. Weused a captive animal environment, rather than a field study,to precisely control environmental quality, choices andeffectsrelatedtoourtreatments. Materialsandmethods Previous research on the effects of toxins on common brushtail pos-sums provided a strong physiological and behavioural underpinningfor the study (Dearing & Cork 1999; Wiggins et al. 2003, 2006;Marsh, Wallis & Foley 2005). Further, we know that possums altertheir foraging and movement patterns strongly in response to preda-tionrisk,particularly fromfoxes, which causes them tomake greateruse of cover than open environments (McDonald-Madden et al. ,2000,Pickett et al. ,2005). HOUSING AND MAINTENANCE OF POSSUMS Eight adult brushtail possums ( T. vulpecula ), seven males and onefemale (body weight mean 2 Æ 48 kg ± 0 Æ 26 SD, range 2 Æ 2–2 Æ 65 kg),weretrappedandhousedatCowan FieldStation,UniversityofNewSouth Wales, Muogamarra National Park (New South Wales, Aus-tralia). Possums were housed individually in outside fully enclosedwiremeshareaseachcomprisedoftwoconnectedpens(doorsperma-nently propped open) [12 m (L) · 5 m (W) · 4 m (H)]. A nestbox,placed in the passage connecting the two pens, provided shelter, andlogs were provided throughout the cage, above the ground, for bothclimbingandperching,andasenvironmentalenrichment.Each possum had access to two food patches, one in the centre of eachpen.Afoodpatchconsistedofabowlplacedinsideaclearplas-tic container protecting it from rain with a hole in one side allowingthepossumfullaccess.Thebowlwasplacedonatraywith1–2 cmof water to exclude ants, which in turn was placed on a table [60 cm(L) · 60 cm (W) · 77 cm (H)] to exclude rodents. We used this set-up for each animal in the experiments described later. A pilot studyconfirmedno‘pen’preference.Abasaldietwaspreparedevery3 days.Thedietwas18%drymat-ter (% DM calculated from samples taken during the experiment)and consisted of (% fresh matter) 46% apple, 35% English spinach,10% carrot, 5% lucerne(ground through a 1-mmsieve)and4% rawsugar. Fresh diet constituents (apple, spinach and carrot) weregroundinafoodprocessor,combinedwithdryingredients(Lucerne,sugar) and mixed thoroughly. Possums were provided water and ad libitum basaldietnightlyandfastedduringtheday.Allpossumswerefed the basal diet 7–10 days before starting experiments. 1,8-cineolewasaddedtotreatmentsjustbeforeserving. EXPERIMENTAL TREATMENTS AND DESIGN We compared intake and behavioural responses of individual pos-sums using five paired treatments, comprised of a risky, non-toxicpatch tested against a safe patch with one of five concentrations of the toxin 1,8-cineole ( Eucalyptus oil), a monoterpene. As possumsconsume eucalypt leaves as a major component of their diet (Free-land& Janzen 1974), 1,8-cineole isa commontoxinin theirdiet.Thehighest 1,8-cineole concentration that we used (0 Æ 10 g gDM ) 1 )reduces food intake by possums by 50% under no choice condi-tions (Marsh et al. 2006). Although this concentration of 1,8-cineoleis unlikely to be found naturally, the effect of 0 Æ 10 g gDM ) 1 onintake is a realistic reflection of the effect of the suite of toxins inmany individualeucalypttrees,whichcanreduce leaf intake by mar-supial herbivores by 50% or more (O’Reilly-Wapstra et al. , 2004,Lawler et al. , 2000). Further, to simulate spatial variation in risk, wecreated safe and risky patches in enclosures. Safe patches wereenclosed (covered), whereas risky patches were exposed (open) withanolfactorycueoffoxurine ⁄ scatsandilluminated.The trial ran for a total of 10 days, consisting of six experimentalperiods(i.e.days),withonerest daybetween eachtominimizecarry-overeffects.Possums1–4(Block1)weretestedfrom25Septemberto5October2008,andpossums5–8(Block2)weretestedfrom18to28October2008.Treatmentsweresetup2 hbeforedark.Eachanimal wasofferedone ofthe five treatment Pairs perperiodin a crossover design to control for carryover effects (Ratkowsky,Evans & Alldredge 1993). This experimental design means that eachpossumreceivedthetreatmentsinadifferentorder.A Pair comprisedthechoiceoftwo Treatments :one(safe) Cover foodpatch(toxincon-centrationsdescribedlater)andone Risk foodpatch,eachplacedinaseparate interconnected pen separated by 4 m. The location of the Cover and Risk patcheswasswitchedbetweenpenseachperiod.The Cover patch was covered with a Hessian sack and dried euca-lyptusbranchestiedaroundaremovablemetalframeplacedoverthefeeding table. Possums were able to move fully under the shelter.The Risk patch was an exposed feeding table with no shelter, foxurine ⁄ scats and illuminated. For light, LED bicycle head lights(PowerBeam ,Model:LH-5N)werehung0 Æ 5 mabovetables.Whitelightwasusedtoapproximateilluminationconditionsfoundwithnat-uralfullmoonlight(Biebouw&Blumstein2003).FoxurineandscatsweresuppliedfrozenbyMarkSobierajaski(Nowra,NSW,Australia).Approximately 5–10 mL was poured onto a cotton pad in a smallplastic tray with 2–3 scats. Fox scats ⁄ urine were replaced each treat-ment night; urine was refrigerated and scats frozen until used (maxi-mum10 weeks).Asecondpilotstudyconfirmedthesafevs.riskyfoodpatch,definedbysignificantlygreaterfoodintakeintheformerunderchoice conditions ( P < 0 Æ 0001). These results clearly demonstratedthatpossumsperceivedtheriskpatch(open ⁄ foxurine ⁄ scat ⁄ illumina-tion) as risky (C. L. Nersesian, P. B. Banks, C. McArthur, unpub-lishedpaper).For the five treatment Pairs , the basal diet at the Risk patch wastoxinfree. Oneoffiveconcentrationsof1,8-cineole(1,8-cineole, pur-ity 99%; Felton Grimwade & Bickford Pty Ltd, Oakleigh, Victoria,Australia) was added to the basal diet at the Cover patch:A = 0 Æ 00 g 1,8-cineole g DM ) 1 , B = 0 Æ 01 g 1,8-cineole g DM ) 1 ,C = 0 Æ 02 g 1,8-cineole g DM ) 1 , D = 0 Æ 05 g 1,8-cineole g DM ) 1 and E = 0 Æ 10 g 1,8-cineole g DM ) 1 (for Pairs A–E, respectively).Eachfoodbowlcontained500 g FW. FOOD INTAKE Intakewasanalysedbasedongramsdrymatterofthebasaldietcon-sumed per kg body mass (gDM kg BW ) 1 ). Control samples (50 g),takendailyforalltreatments,wereovendriedat60 C(72 h)todeter-mine %DM and used to convert fresh weight FW offered into DMoffered.Remainingdietwascollectedeachmorning,weighedandthendried.IntakewascalculatedasDMofferedminusDMremaining. BEHAVIOUR AT FOOD PATCHES Animals were filmed each night (16.00–06.30 h) with one camera ateach food patch. Variables calculated were the following: (i) totaltime at patch, (ii) total number of visits, (iii) total time feeding atpatch, (iv) time feeding per visit, (v) feeding rate (intake gDM ⁄ totalfeeding time), (vi) total time vigilant and (vii) proportional vigilance Titratingthecostofplanttoxinsagainstpredators 755 2011TheAuthors.JournalofAnimalEcology 2011BritishEcologicalSociety, JournalofAnimalEcology , 80 ,753–760 [total time vigilant ⁄ (total time vigilant + total time feeding atpatch)].Visit wasdefinedaswhenananimalvisitedand ate atafoodpatch. Vigilance was defined as any time the animal stopped feedingorchewingandscannedorlookedup. STATISTICAL ANALYSIS Intake and behavioural data were the dependent variables and therewere several possible ways that these data could be analysed giventhe experimental design. We chose to test them against the indepen-dent variables of Pair , Treatment(Pair) i.e. Treatment nested within Pair , Period , Carryover , Block as fixed effects and Animal as a ran-dom effect, usinga generallinearmixedmodel procedure ( proc mixedsas v9.2 Software; SAS Institute Inc., 2007). With this analysis, wetherefore considered Pair as a blocking factor, which enabled us todetermine whether the net result for any dependent variable, such asintake, was consistent among Pairs regardless of the Cover Treat-ment (indicated if the effect of Pair was not significant). The Treat-ment(Pair) effect then tested the hypothesis that the dependentvariable was affected by the particular treatments used. A significant Treatment(Pair) effect would indicate that there was a differencebetweenthe Cover and Risk treatmentwithinatleastoneofthe Pairs .If so, we followed this with five planned contrasts of the two treat-ments within each Pair, using a adjusted a -level of 0 Æ 025. Thus, ourfocus was on the comparison of treatments within each pair,although we visually inspected patterns across pairs as well. Theeffect of carryover, which represents the particular Pair provided inthe period previously, was never significant and so it was removedfrom all final models. For (i) intake in Cover , (ii) intake in Risk and(iii) the difference in intake between Cover and Risk , we also ran theanalysis with concentration of cineole as a continuous independentvariable totest for a significantlinearrelationship.Weexcluded Per-iod and Carryover from the final models as they were not significant.We did not repeat this analysis for the behavioural variables becausewehadnoexpectation,necessarily,ofalinearfit.For all statistical tests, residuals were checked for homoscedastic-ity and normality, and transformations performed where necessary.Timefeedingpervisitandfeedingrateweresquare-roottransformed.One missing value (data collection error) and its paired value, andone outlier and its paired value, were excluded for tests of total timefeeding,feedingrateandtimefeedingpervisit(GLMManalyseswithandwithoutthisoutliergavesimilarresults). Results Animalswereabletocompensatefor anychangeinintakeorbehaviourat one patchbya concomitant change atthe otherpatch( Pair effect P > 0 Æ 16forallvariables).However,thereweredifferingresponsesbetweenthecoverandriskypatchinsix of these variables, indicated by a significant Treatmentwithin Pair effect : intake ( F 5,70 = 4 Æ 85; P = 0 Æ 0007), totaltime at patch ( F 5,70 = 6 Æ 04; P = 0 Æ 0001), total time feedingat patch ( F 5,68 = 6 Æ 19; P £ 0 Æ 0001), time feeding per visit( F 5,68 = 3 Æ 30; P = 0 Æ 0100),totaltimevigilant( F 5,70 = 2 Æ 71; P = 0 Æ 0267) and proportional vigilance ( F 5,70 = 13 Æ 99; P £ 0 Æ 0001).At zero or low (0 Æ 01 g 1,8-cineole gDM ) 1 ) toxin concen-tration (pairs A–B), animals ate more from the cover thanthe risky patch (Fig. 2). This shifted with increasing toxinconcentration and at 0 Æ 05 g 1,8-cineole gDM ) 1 (pair D),intake was most similar between treatments. At the highesttoxin concentration, 0 Æ 10 g 1,8-cineole gDM ) 1 (pair E), ani-mals had switched and were eating more from the risky thanthe cover patch. Mean number of visits varied from 5 Æ 7 to12 Æ 7, but the Treatment within Pair effect was not significant( F 5,70 = 1 Æ 85; P = 0 Æ 1141). The effect of cineole concentra-tion, as a continuous variable, was significant and negativefor intake in the cover patch ( P = 0 Æ 0008, parameter esti-mate = ) 88 Æ 0),significantandpositiveforintakeintheriskypatch ( P = 0 Æ 0003, parameter estimate = +90 Æ 65) and sig-nificant and negative for the difference in intake betweencover and risky patches ( P = 0 Æ 0002, parameter esti-mate = ) 179 Æ 3).As with intake, less time was spent at the cover patch astoxinconcentrationincreased,andconcomitantly,increasingtime was spent at the risky patch (Fig. 3a). When toxin con-centration reached 0 Æ 05 g gDM ) 1 (pair D), possums spentsimilar time between patches, but by 0 Æ 10 g 1,8-cine-ole gDM ) 1 (pair E), they had switched, spending more timeat the risky than the cover patch. Similarly, as toxin concen-trationincreasedinthecoverpatch,animalsfedfor longer atthe risky patch and less at the cover patch, and the point atwhich intake between treatments was most similar occurredat0 Æ 05 g1,8-cineole gDM ) 1 (pairD)(Fig. 3b).Time feeding per visit was similar at the cover and riskypatch at zero to 0 Æ 05 g 1,8-cineole gDM ) 1 toxin concentra-tion(pairsA–D) butshifteduntilitwaslessatthe coverthanrisky patch by 0 Æ 10 g 1,8-cineole gDM ) 1 (pair E) (Fig. 3c).Mean feeding rate was consistently higher at the risky thanthe cover patch, although the Treatment within Pair effectwasnotsignificant( F 5,68 = 2 Æ 21 P = 0 Æ 0632)(Fig 3d).Total time vigilant was higher in the risky than the coverpatch, with the greatest difference at 0 Æ 10 g 1,8-cine-ole gDM ) 1 (pair E) (Fig. 4a). As toxin concentrationincreased in the cover patch, possums became progressivelymore vigilant there, until pair D where they were also morevigilant in the risky patch. Vigilance then dropped in the Fig. 2. Food intake as a function of increased toxin concentration atsafevs.riskypatches.Valuesareleast-squaresmeans(±SE).Alinearfunction has been fitted to the safe and risky values for illustrativepurposes. An asterisk above a pair indicates a significant differencebetweentreatments. 756 C.L.Nersesian,P.B.Banks&C.McArthur 2011TheAuthors.JournalofAnimalEcology 2011BritishEcologicalSociety, JournalofAnimalEcology , 80 ,753–760 cover patch at pair E. Proportional vigilance showed asimilar pattern, but the largest difference between the safeand risky patch occurred at zero, 0 Æ 01 and 0 Æ 02 g 1,8-cine-ole gDM ) 1 (pairsA–C)(Fig. 4b). Discussion INTEGRATING THE INFLUENCE OF TOXINS ANDPREDATION RISK WHILE FORAGING Possumsfacedthedilemmaofchoosingatoxicfoodinasafepatch vs. moving to a patch with a higher cost of predationrisk but no toxins in the food. Our results demonstrate thatpossums were able to quantify, compare and balance thesetwo different, proximate costs, altering their foraging pat-terns in the process. Using our experimental framework, wewere able to establish the point – the tipping point –at whichpossums equated the cost of the toxin 1,8-cineole with thecost of fear induced by the predation cue of fox odour ⁄ light ⁄ open at the risk patch. Under these conditions, this tippingpoint occurred around the 1,8-cineole concentration of 0 Æ 05 g gDM ) 1 (pair D). At this point, food intake, total timeat patch and time spent feeding at patch were most similarbetween safe and risk treatments. In contrast, animals eitherfavoured the safe patch at lower 1,8-cineole concentrations(pairs A–C) or the risky patch at the higher concentration(pair E). Other studies have compared plant toxins andpredation risk (Schmidt 2000; Fedriani & Boulay 2006;Hochman & Kotler 2006; Shrader et al. 2008; Kirmani,Banks&McArthur2010),butonlyasdichotomousvariables(i.e.presenceorabsenceoftoxin).Thecurrentstudyincorpo-rates the realistic variability in toxin concentration, foundamong individual plants, into the harvesting ⁄ foraging equa-tion. Feeding decisions are clearly dependent on toxin con-centration; consequently, the relative influence of plant toxinand predation risk on feeding decisions varies too. Possumsperceive both plant toxins and predation risk as costs to for-aging and are not willing to increase their exposure to preda-tionriskuntiltoxinconcentrationforcesthemtodoso.The progressive reduction in intake and time spent in thesafe patch as toxin concentration increased there, and theconcomitant switch to the risky patch, arose from reducedtimefeedingpervisit(i.e.feedingboutlength)and,toalesserextent, slower feeding rate. Both changes are consistent withprevious studies on effects of plant toxins on feeding behav-iours(Wiggins et al. 2003,Marsh et al. ,2007).Animalsatefasterinthe riskythanthe safepatchirrespec-tive of the concentration of toxin in the safe patch, at leastuntil the concentration was highest (pair E). Feeding rate forpossums was presumably optimal under safe, non-toxic con-ditions (i.e. cover patch, pair A), so eating faster (in the riskypatch)orslower(inthesafepatchwith hightoxinconcentra-tion, pair E) both represent a cost. Eating slowly is clearly acost if intake is reduced, but there are several reasons whyeating faster could also be a cost. Eating faster reduces food-processing time, and the concomitant increase in size of Fig. 3. Differencesinbehaviourasafunctionofincreasedtoxinconcentrationatsafevs.riskyfoodpatches:(a)totaltimeatpatch,(b)totaltimefeeding atpatch, (c) totaltime feeding per visit and (d) feeding rate. Values areleast-squares means(±SE).An asterisk abovea pair indicates asignificantdifferencebetweentreatments. Titratingthecostofplanttoxinsagainstpredators 757 2011TheAuthors.JournalofAnimalEcology 2011BritishEcologicalSociety, JournalofAnimalEcology , 80 ,753–760
