Consequences of the Timing of Seed Release of Erythronium Americanum (Liliaceae), a Deciduous Forest Myrmecochore Author(S): Scott Ruhren and Michele R

Consequences of the Timing of Seed Release of Erythronium americanum (Liliaceae), a Deciduous Forest Myrmecochore Author(s): Scott Ruhren and Michele R. Dudash Source: American Journal of Botany, Vol. 83, No. 5 (May, 1996), pp. 633-640 Published by: Botanical Society of America Stable URL: http://www.jstor.org/stable/2445923 Accessed: 10/06/2010 10:37

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=botsam.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Botanical Society of America is collaborating with JSTOR to digitize, preserve and extend access to American Journal of Botany.

http://www.jstor.org AmericanJournal of Botany 83(5): 633-640. 1996.

CONSEQUENCES OF THE TIMING OF SEED RELEASE OF ERYTHRONIUMAMERICANUM (LILIACEAE), A DECIDUOUS FOREST MYRMECOCHORE'

SCOTT RUHREN2,3 AND MICHELE R. DUDASH

Departmentof Botany,University of Maryland,College Park, Maryland 20742

Myrmecochoresare plants with seeds adapted for ant dispersal. This specialized dispersal syndromemay provide Ery- throniumamericanum seeds withprotection from predators within the easterndeciduous forests.To determinethe adaptive significanceof myrmecochoryin E. americanum,seed removal rates and seed predationin relationto seed release date and location along the Potomac River in Langley, Virginia,were examined. The numberof seeds removedfrom four exclosure treatmentswere monitoredtwo times in 1992 and three times in 1993 within floodplain and hilltop populations of E. americanum.Overall, seed removal was greatestfrom control depots, and E. americanumseeds were removedat nearlythe same rate frompredator-exclusion depots, indicatingthat removal fromopen depots is largely due to ant removal. Ants removed significantlymore seeds than predatorsin the first48 h of seed exposure and could potentiallyremove all E. americanumseeds before nightfall.Aphaenogaster rudis was identifiedas the primarydisperser of E. americanum.Seeds placed in depots afterthe naturalseed release period were discovered more quickly and removedby ants at a significantly higherrate than seeds released at the naturaldate. These resultssuggest that ant dispersal of E. americanumseeds reduces the likelihood of seed predation.

Key words: Ant dispersal;deciduous forests;Erythronium americanum; Jeffersonia diphylla; Liliaceae; myrmecochores; seed release.

Myrmecochory,the dispersal of seeds by ants, is a 1973; Oostermeijer,1989). Withinthe easterndeciduous widespread phenomenon documented in more than 70 forestsmany myrmecochoresrelease theirant-dispersed plant families (Beattie, 1985) and numeroushabitats in- seeds during a relativelyshort period. Some myrmeco- cluding deciduous forests (e.g., Holland, 1974; Culver chores exhibitoverlapping seed release phenologies dur- and Beattie, 1978; Muller, 1978, 1990), dunes (Ooster- ing thistime, with the less abundantseeds oftenpreferred meijer,1989), tropicalepiphyte communities (Kleinfeldt, by dispersersand predators(e.g., Smithet al., 1989). For 1978), and sclerophylous shrub (Bond and Slingsby, othermyrmecochores, however, there appears to be some 1984; Hughes and Westoby, 1990). Examples of myr- temporal separation of release dates (e.g., Beattie and mecochoryare most abundantin Australiaand South Af- Culver, 1981; Handel, Fisch, and Schatz, 1981; Heithaus, rica (Hughes and Westoby,1990; Westobyet al., 1991), 1986; Ruhren, 1994). For example, Erythroniumameri- but this syndromeis well representedin the easternde- canum Ker. (Gleason and Cronquist, 1991) is a typical ciduous forestsof NorthAmerica where myrmecochores ephemeralmyrmecochore exhibiting an elaiosome (Hol- may constituteup to 40% of the herbaceous layer (Han- land, 1974; Muller, 1978; Harder et al., 1985; Wein and del, Fisch, and Schatz, 1981). Pickett,1989) and a synchronousseed release period that Many of these myrmecochoresare co-occurringspring occurs afterViola spp. yet beforeseveral othersympatric ephemerals,herbaceous plants thatgrow rapidlyfollow- myrmecochoresin northernVirginia (e.g., Jeffersoniadi- ing snowmeltand begin to senesce as the canopy closes. phylla and Trillium sessile) (Lobstein and Rockwood, The prevalence of this syndromein the spring flora of 1993; Ruhren, 1994). This temporal separationof seed the easterndeciduous forestsmay be maintainedby the release among myrmecochoresmay maximize exposure paucity of lipid-richfoods (Carrol and Janzen, 1973). to seed-dispersingants. Therefore,seed-dispersing ants are attractedto the lipids It is theorizedthat myrmecochory may provide seeds contained in most elaiosomes (e.g., Carrol and Janzen with several benefits,including protection from seed I Manuscriptreceived 21 March 1995; revision accepted 2 October predators(e.g., Culver and Beattie, 1978; O'Dowd and 1995. Hay, 1980; Heithaus, 1981; Turnbulland Culver, 1983), The authorsthank C. DiSalvo and D. Sealy at USDOI, NPS for site avoidance of interspecificcompetition (e.g., Handel, access to TurkeyRun Park and cooperation;J. Wilkinsonand R. Iras- 1976, 1978), deposition in a nutrient-enrichedsafe site quin forfield assistance; taxonomistsat BeltsvilleAgricultural Research (Culver and Beattie, 1980, 1983) and protectionfrom fire Center,USDA for ant identityverification; D. Carr, C. B. Fenster,L. Galloway, S. N. Handel, E. Nagy, and K. C. Ruhrenfor commentson (e.g., Berg, 1975). Withinthe eastern deciduous forests the manuscript;and M. McIntosh, D. Denmann, D. Carr, and L. M. of North America, a drivingforce in this plant-animal Rosa for statisticaladvice. All analyses were run on the mainframeat interactionis seed predation.The elaiosomes thatattract the UMD ComputerScience Center.This studywas partiallyfunded by the ants also attractpredators including rodents, primarily a research award from the WashingtonBiologists' Field Club to S. Peromyscus leucopus (e.g., Heithaus, 1981), birds (Mi- Ruhren. 2 Currentaddress: Departmentof Biological Sciences, RutgersUni- lewski, 1982), and many arthropods(e.g., Bond and versity,Piscataway, New Jersey08855-1059. Slingsby, 1984; Anderson,1988). Most of the arthropod 3 Authorfor correspondence. seed predatorsare nocturnalground beetles (e.g., Ohara 633 634 AMERICAN JOURNAL OF BOTANY [Vol. 83 and Higashi, 1987; Higashi and Ito, 1991). Rodents are E. americanum -9 X 15 m mixed with Galium spp. and the myrme- assumed to be destructivebecause they do not cache cochores C. virginicaand A. canadense. seeds in the springand early summer(Barry, 1976). Hei- thaus (1981) discovered thatP. leucopus may cue in on Plant analysislseed collection-All ramets were counted and the the naturalseed depots formedunder dehiscing capsules floweringramets were marked with plastic stakes and monitoredto and frequentlydestroy predehiscent fruits. However, it determinedates of opening,onset of senescence, and cessation of fruit also was revealed that seeds are less detectable by P. and seed production.These individualswere checked in 1992 and 1993 leucopus afterants stripseeds of theirelaiosomes within to determinereproductive output. All capsules withinthe floodplainand theirnests. Thereforeseeds discovered and collected by hilltop were bagged in bridal veil, a fine mesh fabric,to preventpre- ants should exhibitreduced predation. dispersal predationand dispersal by ants and to facilitaterecovery of The goals of our study were to examine factors af- seeds. All seeds were counted and weighed. Seeds were pooled within fectingE. americanumseed dispersal and to evaluate the sites, and additional seeds were collected and weighed fromadjacent nonstudypopulations withinTRP. Seeds were refrigeratedfor several theoreticalbenefit of this dispersal mechanismin reduc- days to preventlipid degradationuntil sufficientseed was available to ing predationand its response to an artificiallyextended begin field manipulationsand observations.Seeds of this species that seed release phenology.First, we testedthis with a series had been refrigeratedfor 1 yr were detected and quickly removed by of treatmentsdesigned to exclude eitherants, rodents, or ants in 1993 (Ruhren,1994). both from seeds placed in two sites in both 1992 and 1993. Second, we extendedthe seed release period of E. Monitoringseed dispersal-In 1992 two transectswere placed 5 m americanumto evaluate the effectof timingon seed dis- apartin boththe floodplainand hilltopstudy sites to determineE. amer- persal and predationrates. Third,we compared removal icanum seed removal rates by ants and predators.The linear transects rates of E. americanum to Jeffersonia diphylla, a sym- contained seven replicates. Each replicate contained four exclusion patricmyrmecochore, to determineif any preferencewas treatmentsrandomly placed. The depots, constructedof 5-cm diameter exhibitedby ants and predators.Finally, to explore the polyvinylchloride (PVC) pipe sectionsdriven into the soil, received six nutrient-enrichedsafe site hypothesis,we analyzed ant seeds and were evenly placed 1 m apart in blocks thatwere spaced 1 nest and controlsoil to determineif E. americanumseeds m apartfrom adjacent blocks. This arrangementsimulated natural seed were presentin ant nests and whetherant nest soil pro- densities.Within the floodplainand the hilltop study sites, a mean of vided an enhanced germinationsite. 5.4 and 7.5 seeds were produced,respectively, per fruit.The PVC pipe was painted brown to blend in with the soil and leaf litter.Approxi- MATERIALS AND METHODS mately 2 cm of each pipe were left unburied,and this surface was treatedto provide one of four simulations:(1) untreatedpipe allowing Study organism-Erythronium americanum Ker. (Liliaceae) is a access by ants and predators,(2) pipe with a Tanglefoot?(The Tangle- common springephemeral herb that occurs in easternNorth America foot Company, Grand Rapids, MI) barrierallowing mammalian seed from Nova Scotia west to Minnesota and south to northernFlorida predatoraccess only, (3) pipe with 0.6-cm wire mesh cover allowing (Holland, 1974). This perennialherb exhibits rapid abovegroundgrowth ant access, and (4) pipe withTanglefoot? and wire mesh providingtotal duringApril and May in northernVirginia, with two-leafed ramets pro- exclusion. Transectswere designatedfor natural and delayed seed place- ducing a single flower.After a highlysynchronized anthesis, fruits are ment.Therefore, all treatmentswere replicatedwithin sites and times. formedas the abovegroundportion of the plant begins to senesce. At Beginning 3 June 1992 (naturalseed release period), the firstseries the end of May, the capsules begin to open on the groundand the hard, of exclosure treatmentswere initiatedin the floodplainand hilltopsites. crescent-shapedseeds lay exposed. Each seed is equipped withan elaio- To test for an effectof seed release timingon seed removal rates,the some that attractsant seed dispersersas well as seed predators.When initialtransect was followed 5 d later(day 6) by anothertransect within ants bring myrmecochorousseeds into their nests, they remove the the floodplainand hilltop sites. In both periods seeds were placed in elaiosome and discard the intactseed in undergroundburrows or on the depots at 0800 and each depot was checked approximatelyevery half mound (e.g., Culver and Beattie, 1978). Erythroniumamericanum elaio- hour for8 h daily for 1 wk followingrelease. Afterthis intensivemon- somes appear to degrade within 1-2 wk when exposed to heat and itoringperiod, observations were reduced as removal activitydeclined. humidity(Ruhren, 1994). Like othermyrmecochores examined by Lob- Overall, exclusion depots were monitoredfor -3 mo. Overnightseed stein and Rockwood (1993) (Asarum canadense, Jeffersoniadiphylla, removal fromexclusion-free depots was assumed to be by rodents(pri- Dicentra cucullaria, and Viola striata), E. americanum seeds do not marilyP. leucopus), seed-eatingarthropods, as well as nocturnallyfor- require scarificationby ants to enhance germination(Ruhren, 1994). aging ants. In 1993, the transectswere placed in the same locations in the flood- Studysite-All E. americanumseeds were collected fromnative pop- plain and hilltop sites and monitoredfor the same period of time be- ulationsat TurkeyRun Park (TRP) in Langley,Fairfax County, Virginia, ginningon 2 June. Treatmentswere rerandomizedwithin transects in -8 km northwestof Washington,DC. TRP is a mixed deciduous forest each block. Anothertransect was added to simulatea 15-d delay of seed habitatsituated along the Potomac River.Transects were placed in two release (day 16) in each site. Thus, therewere threeseed release dates habitatswithin TRP: the riverfloodplain and a dry upland hilltop site in 1993. Additionally,six E. americanumseeds were placed on the soil 30 m above the Potomac River, -300 m fromthe floodplain.Both sites surface withineach block of exclusion treatmentsto simulate natural were chosen because they containeddense patches of E. americanum, seed placementand testfor any pipe effect.In 1993 only,all seeds were representeddistinct habitats, and were relativelyundisturbed by human markedwith a small dot of enamel paint,color coded to correspondto activity. the threerelease dates. This was done in an attemptto determinethe In additionto E. americanum,other spring ephemeral sympatric herbs fate of seeds removed during the study. Because seed trackingwas at the floodplainincluded the myrmecochoresTrillium sessile, Asarum found to be ineffectivein the dense leaf litterat TRP, ant nests were canadense, and Claytonia virginica. The E. americanumpatch within markedduring the monitoringphase and subsequentlyexhumed at the the floodplainwas -12 X 27 m. This patch was dissected by an inter- end of the 1993 season. Ant nests were not markedduring the natural mittentstream that flowed into the riverduring heavy rainfall. E. americanumseed release period because of time constraints.While The hilltopcommunity was characterizedby a dense shrublayer of tabulatingseed removalrates, we monitoredarthropod visitors and sev- Asimina triloba saplings. The herb layer was composed of a patch of eral were capturedfor identification. Also, ant densityand activitywere May 1996] RUHREN AND DUDASH-DISPERSAL OF ERYTHRONIUM 635

TABLE 1. Results of three-wayANOVA showingthe effectof site, timing,and exclusion treatmentsand interactionson Erythroniumamericanum cumulativeseed removal for the first48 h in 1992 and 1993.

Year 1992 1993 Source df MS F P df MS F P Site (S) 1 0.003 0.01 0.92 1 0.44 0.95 0.34 Temporal (T) 1 6.45 19.13 0.0001 2 5.04 10.91 0.0001 Exclusion Treatment(ET) 3 16.06 47.62 0.0001 3 22.12 47.91 0.0001 S X T 1 0.37 1.11 0.29 2 0.45 0.98 0.38 S X ET 3 0.13 0.40 0.75 3 1.48 3.20 0.03 T X ET 3 1.29 3.84 0.01 3 0.93 2.00 0.07 S X T X ET 3 0.50 1.49 0.22 3 0.72 1.55 0.16 Error 96 0.34 144 0.46 monitoredwith tuna baits and pitfalltraps randomly placed withinboth RESULTS sites. The cumulativeamounts of seed removedfrom each exclusion treat- Population monitoring-The floodplain population ment were compared for the first48 h of seed removal. These totals was composed of -1 100 and 2400 rametsin 1992 and were analyzed assigningexclusion treatment,seed release date, and site in 1993, respectively.Thirty (7.7%) and 80 (11%) flow- as fixedeffects in the analysis of variance (ANOVA) using SAS (1989). ering individuals were counted in 1992 and 1993, re- Cumulative seed removal was square-roottransformed to meet the as- spectively.Flowers began to open on 30 March 1992, sumptionsof ANOVA. Tukey posthoc comparisons identifiedsignifi- and 6 April 1993, with seeds naturallyreleasing between cant differencesamong treatments.Interpretations were based on the 27 May and 2 June 1992, and 29 May and 3 June 1993. first48 h of seed exposure in exclusion depots because additionalseeds Significantlymore rainfalloccurred during the flowering placed on the soil in 1993 were frequentlyremoved within 8 h and period of E. americanum in 1993 than in 1992 (27 mm usually within48 h. Also long-termmonitoring revealed no significant comparedto 13 mm), causing heavy floodingat this site. change in removal rates after48 h (Ruhren, 1994). With the addition Approximatelythe same number(3 660) of rametswas of a thirdtransect and additional seed placed on the soil surface in at the hilltop in both years. Three percentof the ramets 1993, data fromthe 2 yr were analyzed separately. floweredin 1992 compared to 2% in 1993 at this site. Erythronium americanum flowersbegan opening at the Sympatricmyrmecochores-In 1992, withinan E. americanumpop- hilltop on 3 April 1992, and 6 April 1993. Seeds were ulation -200 m downstreamfrom the floodplain,we testedfor ant seed released between 31 and 2 June and 30 preferencebetween E. americanumand Jeffersoniadiphylla during the May 1992, May naturaldispersal period of J. diphylla,which occurred -7 d afterthe and 3 June 1993. naturalseed dispersal phase of E. americanum.This other myrmecochore occurred sympatricallywith several patches of E. americanum Seed dispersal vs. predation-Seed removalfrom both withinTRP. Because of limitedseed supply,only two treatmentsrep- sites in 1992 and 1993 was rapid in the initialphase (z48 licated seven times were used: rodentexclusion and our controlwith h) of seed exposure,followed by a near cessation of ac- no exclusion. Six seeds of eitherspecies were placed in separatedepots tivityfor the remainderof the observations. The seed withina single transect.Both treatmentsof each species were placed supply was never entirelydepleted withinany transects withineach square metre.Cumulative amounts of seed removal were or treatmentsby August 1992 and 1993 (Ruhren,1994). analyzed assigningspecies and exclusion treatmentas fixedeffects with However, all of the E. americanum seeds placed on the ANOVA using SAS (1989). This experimentwas not replicatedin 1993 soil surfacein 1993 were removedwithin 48 h, indicating because of insufficientseed supply followingthe addition of the 15-d the potentiallyhigh naturalremoval rate. delay release period. In 1992 and 1993 no significantdifference in removal was observed between the floodplainand hilltopfor the Soil analyses-To determineif E. americanumseed were deposited first48 h (Table 1). Consequently,the sites were pooled in ant nests,soil fromnests and referencesoil were separatedusing the within 1992. However, in 1993 there was a significant chemical extractionand flotationmethod described by Malone (1967). interactionbetween site and exclusion treatment(Table Organic debris was thendried and examined undera lightmicroscope. 1); consequently,the two sites were not pooled in the The method was tested by spiking six soil samples with previously analysis. collected E. americanumand Dicentra cucullaria seeds. We recovered Based on the comparisonof seed removal rates from 96% of the E. americanumand 98% of the D. cucullaria spiked seeds. the fourexclusion treatments,ants removed significantly To test the nest-enrichmenthypothesis (Culver and Beattie, 1980, more seeds in 1983) soil collected from 12 ant nests and 12 randomlylocated refer- (z25%) thanpredators (1%) the first48 h ence locations in 1992 was analyzed for nutrientsand chemical char- of the 1992 monitoring(Table 1; Fig. 1). Ants were the acteristics(Mg, P205, K20, Cu, Mn, NO3, S04, Zn, and pH) and physical only organismobserved removingE. americanum seeds. characteristics(sand, silt, clay, and organic matter).All ant nests sam- In 1993, a period of rapid discovery and removal in the pled were in the soil. The Universityof Maryland Soil TestingLabo- first48 h occurredonce again in the floodplainand hilltop ratoryconducted these analyses. The effectsof soil type (ant nest or (Fig. 2a, b). reference)and site (floodplainor hilltop) were analyzed using MAN- It appears thatremoval by predatorsoccurred primarily OVA of log-transformeddata on SAS (1989). One-way ANOVAs for at night in both years. Seven chewed E. americanum each soil characteristicwere performedto determinevariation in indi- seeds and emptyseed coats were discoveredin threede- vidual soil components. pots withinthe floodplain and hilltop transectson four 636 AMERICAN JOURNAL OF BOTANY [Vol. 83

5 O A. Floodplain 50 0 0 I)3g x~~~~~~~~~~ 40x

W 30 LX 30.

20.

5 10 20 30 -5 0 15 25 35 404550 - -10 Hours Elapsed -10 0 10 20 30 40 50

o A B. Hilltop Predators& Ants Ants 50 O Predators + Total Exclusion g 40. X Fig. 1. Comparisonof naturalseed removalrates from all exclusion treatments,pooled across floodplainand hilltopsites for the first48 h during 1992 (N = 84 seeds/exclusiontreatment). Symbols representE. 30. americanum seed removal by suspected organisms.Each point repre- sents the percentageof seeds removed from 14 replicates. Exclusion treatmentsfollowed by the same letterare not significantlydifferent at 20. X the P < 0.05 level. 10. morningsin 1992 and 1993. We also occasionally detected evidence of burrowingaround the depots in the 0 __ morningwhen arrivingat TRP. During the day, few if

any vertebrateswere observed in and around the tran- -10" . sects. Chipmunks (Tamias striatus) sometimes investi- -10 0 10 20 30 40 50 gated the floodplain area, but they normally confined their activities to the rocky slope above the floodplain Hours Elapsed transectsand were never seen at the depots during the 0 Predators& Ants A Ants day.

Timingeffects The timingof seed release had a high- O Predators ' Total Exclusion ly significantinfluence (Table 1; Fig. 3) on seed removal rateswithin the first48 h of seed exposurein 1992. Seeds Fig. 2. Comparisonof naturalseed removalrates from all exclusion released aftera 5-d delay were removed -40% morerap- treatmentswithin the transectsof (A) floodplainand (B) hilltop sites idly by ants than seed released at the naturaldate. The forthe first48 h during1993 (N = 42 seeds/exclusiontreatment). Sym- bols representE. americanum seed removal by suspected organisms. exclusion treatmentssignificantly interacted with timing Each point representsthe percentageof seed removal fromseven rep- of seed release date in 1992 (Table 1). licates. Exclusion treatmentsfollowed by the same letterare not sig- Variable resultswere observed between sites in 1993 nificantlydifferent at the P < 0.05 level. (Fig. 4a, b). Seeds released 5 d laterwere again removed significantlymore quickly (-60% and -15% increases at the floodplainand hilltop,respectively) within the first E. americanumseeds in 1992. Seventy percentof these 48 h of exposure (Table 1). Seeds released 15 d later did seeds were removed after 8 h (Fig. 5), compared to not resultin any significantincrease beyond the already -21 % removal of E. americanumseeds. Neitherspecies' significant5-d delay period in the floodplain.However a seeds were removed completelyfrom the depots. There 15-d delay did enhance seed removal by -40% in the was no significantdifference between 48-h seed removal hilltop,perhaps because the hilltopwas not as wet as the fromeither exclusion treatment. floodplain.Overall, seed removal plateaued as observed in 1992. Soil and seed separation No E. americanum seeds were detected in any of the Aphaenogaster rudis nests Sympatricmyrmecochores Significantlymore J. di- and referencesoil samples inspected in 1993. However, phylla seeds were removed by ants in the first24 h of Dicentra canadensis seeds, also known to be dispersed seed exposure (ANOVA, F = 61.63, P < 0.0001) than by ants in earlyMay, were discoveredin both soil sample May 1996] RUHREN AND DUDASH-DISPERSAL OF ERYTHRONIUM 637

A. Floodplain 100

60 75' 8 O sa X

0 -r 1: 4 O.

$ 2G __~~~~__ Y

2 0 __ __

-5 0 5 10 15 20 25 30 3540 45 50 > . O -10,o 0 10 20 30 40o 50 Hours Elapsed :) B. Hilltop 10 0 Natural a 5-Day Delay Fig. 3. Comparison of seed removal by ants afterdifferent seed 28 z release dates pooled across sites for the first48 h of 1992 (N = 84 94y seeds/exclusiontreatment). Each pointrepresents the percentageof seed removal from 14 replicates.Plots followed by the same letterare not significantlydifferent at the P < 0.05 level. types (63 in nests and 19 in referencesoil). These seeds 60240 l occurred significantlymore frequentlyin ant nest soil fromboth the floodplain(X2 = 7.35, P < 0.01, df = 1) and the hilltop (X2 = 10.39, P < 0.01, df = 1) than in referencesoil. Soil analyzed in 1992 revealed no significanttrend in nutrientcontent of ant nests accordingto Wilks' Lambda -10 0 10 20 30 40 50 test of the MANOVA, (F1310 = 2.026, P > 0.10) except organic matter(F122 = 4.73, P < 0.05). However, nest Hours Elapsed soil was found to be significantlygreater than reference o Natural ? 5-Day Delay A 15-Day soil accordingto the ANOVA of separatevariables (F3912 Delay = 5.734, P < 0.0009): a mean of 155.3 partsper million Fig. 4. Comparison of seed removal by ants afternatural, 5 d and (ppm) K20 in the nest soil vs. 124.2 ppm in the reference 15 d delay seed release dates in (A) floodplainand (B) hilltop for the soil (F3,16 = 6.97, P = 0.0007); 20.9 ppm Zn (nest), 17.7 first48 h of 1993 (N = 42 seeds/dateof release). Each pointrepresents ppm Zn (reference) (F316 = 23.66, P = 0.0001); and the percentage of seed removal from seven replicates. Seed release 19.2% clay (nest) and 17.7% clay in the reference(F3 16 times followed by the same letterare not significantlydifferent at the 12.74, P = 0.0002). P < 0.05 level.

Ant activity-Aphaenogasterrudis was assumed to be were observed on E. americanum seeds in the hilltopin the primarydisperser of E. americanumbecause it was both years. Although theirvisits to the exclusion treat- the most frequentant species observed at exclusion de- ments were infrequent,they appeared to be eating the pots, baits,and pitfalltraps within the floodplainand hill- elaiosomes in situ. top in 1992 and 1993 (Ruhren,1994). Out of all observed ant visits to seed depots in 1992, A. rudis were 77% of DISCUSSION the visitorswithin the floodplain,94% withinthe hilltop, and 69% and 65%, respectively,in 1993. The remainder Ants were the primaryremovers of E. americanum of the ant visitors were primarilyPrenolepis imparis; seeds and removed significantlymore E. americanum however,they never were seen removingseeds. In fact, seeds thanpredators in the first48 h of seed release. Seed several P. imparis were found eating the elaiosomes removal rates did not vary by site (floodplainor hilltop) withinthe pipe exclosures. Lasius alienus and Campon- but increased significantlywhen E. americanum seed re- otus pennsylvanicuswere less frequentlyobserved at the lease was delayed for 5 d past the natural seed release seed depots, baits, and pitfall traps, and they were not date in both 1992 and 1993. A furtherdelay of 15 d in seen removingseeds. 1993 continued to enhance significantlyseed removal Several harvestmen (Leiobunum bimaculatum) also rates by ants in only the hilltop site. In both 1992 and 638 AMERICAN JOURNALOF BOTANY [Vol. 83

?Erythronium americanum DJeffersonia diphylla may providean enhancedgermination site while reducing predation.Also, A. rudis colonies are normallynomadic, reducingthe likelihood thatE. americanumseed collec- 100 tions will become too dense. This habit of relocating nests often after20 d (Culver and Beattie, 1978) was 80 observed in the floodplainand hilltop and may explain the absence of E. americanum seeds in exhumed ant ~U60. nests. Also, nests were marked afterthe naturalseed re- 0 lease period of E. americanum because of time con- 5 40k straints. The timingof seed release exerted a major influence

20. on E. americanum seed removal rates in this study. If indeed dispersalby ants confersan advantageto E. americanum seeds, it may be inferredthat extendingthe re- strictedseed release period could be favored by natural -10 0 10 20 30 40 50 selectionif it increases the period of exposure to ants. In 1992 and 1993, E. americanumseeds released 5 d later Hours Elapsed were removed significantlyfaster than those released at the naturaltime. A furtherdelay of 5 d may continueto Fig. 5. Comparisonof seed removalby ants of E. americanumand J. diphyllaseeds forthe first48 h of 1993 (N = 42 seeds/species).Each enhance seed removal in some sites. This initialincrease point representsthe percentageof seed removal fromseven replicates. in seed removal may be due to ant learning,an increase There was an initial significantdifference in seed removal at 8 h that in ant colony size, or changes in ant activities.Thus, after disappeared withinthe overall analysis at 48 h. an initial period of acclimation and learning at the two sites, ant activityappears to have stabilized at the floodplain, perhaps resultingin avoidance of seeds, leaving 1993 therewas a slowing of ant activityafter 48 h re- some E. americanumseeds in the exclusion depots. Dis- gardless of site,timing, dispersal, or predation.Estimates persers did at firstappear to be slow to respond in this of seed removalby ants actuallyare conservativebecause system,but this was because of an inhibitorytreatment some overnightseed removal, attributedto predationin effect,preventing the rapid and complete removal ob- this study,may be the resultof nocturnallyforaging ants served withE. americanumseeds placed on the soil. Fur- (e.g., Lynch, Balinsky,and Vail, 1980; Post and Jeanne, thermore,ants may preferless common sympatricseeds 1982). The results suggest an initiallylow risk of pre- withindensely seeded plots (Smith et al., 1989). Smith dation followed by nearlycomplete seed removal outside et al. (1989) revealed thatthe probabilityof seed removal the depots. often depends on density and species composition of Based on field observationsand ant trapping,there is available seeds. no shortageof ants duringthe seed dispersalphase of E. The slightseparation of the dispersalphase of the var- americanum. Seeds left in the depots, therefore,could ious sympatricmyrmecochores present at these sites may indicatesome satiationof the ants as well as an inhibitory be favored by natural selection to maintain as well as effectof the artificialdepots. Erythroniumamericanum maximize the rate of dispersalof these ant-seed relation- appears to be visited by a guild of generalistscavenging ships. For example, withinboth the floodplainand hill- ants thatprobably do not exhibithost specificity,a fact top, dense patches of ant-dispersedA. canadense and supportedby previous researchon othermyrmecochores Trilliumsessile often were interspersedwith E. ameri- (Beattie and Culver, 1981; Lobstein and Rockwood, canum. Asarum canadense seeds are naturallyreleased 1993). Aphaenogaster rudis is the most frequentvisitor 2-3 wk afterE. americanum,while T. sessile capsules and most likely disperserof E. americanumwithin TRP. dehisce in mid-Julyin northernVirginia. Anothermyr- This species, also identifiedas the disperserof E. amer- mecochore,J. diphylla,releases its seeds in mid-June-1 icanum by Wein and Pickett(1989), is a common ant in wk afterE. americanum at TRP. When placed in treat- Virginiaand Maryland(Lynch, Balinsky and Vail, 1980). mentdepots, J. diphylla seeds were rapidlydetected and Aphaenogaster rudis disperses several other spring were removed by A. rudis significantlyfaster in the first ephemerals including Viola spp. (Culver and Beattie, 8 h than seeds fromE. americanumbut at 48 h no dif- 1978) and Asarum canadense (Heithaus, 1986). Although ference was detected. The cause of this preferencehas A. rudis is behaviorallysubordinate to the less frequent not been quantifiedbut may be influencedby variation ants, L. alienus, P. imparis, and C. pennsylvannicus in elaiosome chemical composition(e.g., Carrol and Jan- (Lynch, Balinsky, and Vail, 1980; Fellers, 1987, 1989), zen, 1973; Marshall, Beattie, and Bollenbacher, 1979; they frequentlycan exploit food resources because they Brew, O'Dowd and Rae 1989). Lanza, Schmitt, and tend to discover food and recruitworkers much more Awad (1992) discovered that the variationin fattyacid quickly than otherant species (Fellers, 1987). This may compositionof elaiosomes of threeTrillium species elic- explain why E. americanumseeds placed on the soil sur- ited differentbehavior in seed-dispersingants. Also, ants face were almost entirelyremoved within8 h. could have been satiated with the more abundant E. Anotherhabit of A. rudis thatmay be beneficialfor E. americanumseeds withina populationof E. americanum. americanumis its shallow, temporarynests under fallen Still, thisoverlap of E. americanumand J. diphyllaseeds leaves and next to decaying logs within the floodplain does not occur naturally.Actually, E. americanumseeds and hilltop sites. According to soil analyses, these nests presentedwith J. diphylla seeds were still removed at May 1996] RUHREN AND DUDASH-DISPERSAL OF ERYTHRONIUM 639 rates comparable to the unmixed seed exclosures at the The Argentineant (Iridomyrmexhumilis) and myrmecochorous floodplainand hilltopsites (-40%). Based on our obser- Proteaceae. Ecology 65: 1031-1037. vations of naturalpopulations, E. americanumseeds will BREW,C. R., D. J. O'DowD, AND I. D. RAE. 1989. Seed dispersal by ants: behavior-releasingcompounds in elaiosomes. Oecologia 80: be removed by the time J. diphylla capsules dehisce at 490-497. both sites. CARROL,C. R., AND D. H. JANZEN. 1973. Ecology of foragingby ants. This abundance of myrmecochoreswithin TRP is not Annual Review of Ecology and Systematics4: 231-257. unique. In many other eastern deciduous forests (e.g., CULVER, D. C., AND A. J. BEATrIE. 1978. Myrmecochoryin Viola: Handel, Fisch, and Schatz, 1981), the potentialfor com- dynamics of seed-ant interactionsin some West Virginia species. petition among myrmecochoresseems apparent. How- Journal of Ecology 66: 53-72. , AND . 1980. The fateof Viola seeds dispersedby ants. ever, most of these species, including E. americanum, AmericanJournal of Botany 67: 710-714. staggertheir seed release throughoutthe springand early , AND . 1983. Effectsof ant mounds on soil chemistry summer.Because ant dispersersmay be the only mech- and vegetationpatterns in a Colorado montane meadow. Ecology anism of escaping predationand seed predationappears 64: 485-492. to be rapid on seed remainingon the soil surface,mech- FELLERS,J. H. 1987. Interferenceand exploitationin a guild of wood- to land ants. Ecology 68: 1466-1478. anisms that maximize exposure effectivedispersers . 1989. Daily and seasonal activityin woodland ants. Oecologia should be selected. Perhaps, as suggested in numerous 78: 69-76. sympatricpollination studies (e.g., Schemske 1981), tem- GLEASON,H. A. AND A. CRONQUIST. 1991. Manual of vascular plants poral partitioningand sharingcould evolve in this seed of northeasternUnited States and adjacent Canada, 2d ed. New dispersal scenario. Among the myrmecochoresof the York Botanical Garden, Bronx, NY. eastern deciduous forest,seed predation appears to be HANDEL,S. N. 1976. Dispersal ecology of Carex pedunculata (Cyper- in aceae), a new NorthAmerican myrmecochore.American Journal very intense,and ant-dispersedseeds this systemap- of Botany 63: 1971-1979. pear to be released over a period of several months,pos- . 1978. The competitiverelationship of threewoodland sedges sibly reducingcompetition for seed dispersers. and its bearing on the evolution of ant dispersal of Carex pedun- Although total seed depletion did not occur in either culata. Evolution 32: 151-163. year fromany of the transects,a resultsupporting Wein ., S. B. FISCH, AND G. E. SCHATZ.1981. Ants dispersea majority and Pickett's(1989) conclusionthat E. americanumseeds of herbsin the mesic forestcommunity in New York State. Bulletin may be dispersed at a slower rate than other myrmeco- of the TorreyBotanical Club 108: 430-437. HARDER, L. D., J. D. THOMSON, M. B. CRUZAN, AND R. S. UNNASCH. chores, this may not representthe naturalseed removal 1985. Sexual reproductionand variationin floralmorphology in rates.In thisstudy, incomplete removal of E. americanum an ephemeralvernal lily,Erythronium americanum. Oecologia 67: seeds occurs most likelybecause of two factors.First, an 286-291. inhibitoryeffect of the pipes was detected when seeds HEITHAUS, E. R. 1981. Seed predationby rodentson threeant-dispersed were placed on the soil, outside of treatmentpipes. This plants. Ecology 62: 136-145. motivatedour emphasis on the first48 h of seed removal . 1986. Seed dispersalmutualism and the populationdensity of in Asarum canadense, an ant-dispersedplant. In A. Estrada,and T H. the final analyses. Second, seeds introducedinto the Fleming [eds.], Frugivoresand seed dispersal. Dr W. Junk,Dor- E. americanumpopulations were the predominantmyr- drecht. mecochore seeds present and may have increased the HIGASHI, S., AND F ITO. 1991. Ground beetles and seed dispersersof densityenough to satiate the fattyacid requirementsof the myrmecochorousplant Trilliumtschonoskii (Trilliacae). In C. foragingant colonies. R. Huxley, and D. F Cutler [eds.], Ant-plantinteractions. Oxford In this system,predation and timingseem to be major UniversityPress, Oxford. ecological forces on this ant-plantmutualism. Predators HOLLAND, P. G. 1974. The growthbehavior, ecology and geographyof Erythroniumamericanum in northeastNorth America. Canadian such as nocturnalground beetles (Ohara and Higashi, Journal of Botany 52: 1765-1772. 1987; Higashi and Ito, 1991), or more commonlyP. leu- HUGHES, L., AND M. WESTOBY. 1990. Removal rates of seeds adapted copus (Heithaus, 1981) could remove any E. americanum for dispersalby ants. Ecology 71: 138-148. seeds remainingafter the diurnal ant foraging.Conse- KLEINFELDT, S. E. 1978. Ant-gardens:the interactionof Codonanthe quently,the lack of specificityin this ant-plantinterac- crassifolia (Gesneriaceae) and Crematogasterlongispina (Formic- tion is favoredby naturalselection to maximize protec- idae). Ecology 59: 449-456. in LANZA, J.,M. A. SCHMITT, AND A. B. AWAD. 1992. Comparativechem- tionfrom nocturnal seed predation a guild of sympatric istryof elaiosomes of threespecies of Trillium.Journal of Chem- myrmecochores.This complementsother forces that dis- ical Ecology 18: 209-221. courage specialization,such as maximizingthe effectof LOBSTEIN, M. B., AND L. L. ROCKWOOD. 1993. Influenceof elaiosome the dispersalguild and increasingdispersion of seeds into removalon germinationin fiveant-dispersed plant species. Virginia the habitat. Journalof Science 44: 59-72. LYNCH, J. F, E. C. BALINSKY, AND S. G. VAIL. 1980. Foragingpatterns LITERATURE CITED in threesympatric forest ant species, Prenolepis imparis,Paratre- china melanderiand Aphaenogasterrudis (Hymenoptera:Formic- ANDERSON, A. N. 1988. Insect seed predatorsmay cause far greater idae). Ecological Entomology5: 353-371. losses than theyappear to. Oikos 52: 337-340. MALONE, C. R. 1967. A rapid methodfor enumeration of viable seeds BARRY, W. J. 1976. Environmentaleffects on food hoardingin deer- in soil. Weeds 15: 381-382. mice (Peromyscus).Journal of Mammalogy57: 731-746. MARSHALL, D. L., A. J. BEATTIE, AND W. E. BOLLENBACHER. 1979. Ev- BEATrIE, A. J. 1985. The evolutionaryecology of ant-plantmutual- idence for diglyceridesas attractantsin an ant-seed interaction. isms. Cambridge UniversityPress, Cambridge. Journal of Chemical Ecology 5: 335-344. , AND D. C. CULVER. 1981. The guild of myrmecochoresin the MILEWSKI, A. V. 1982. The occurrenceof seeds and fruitstaken by herbaceous floraof West Virginiaforests. Ecology 62: 107-115. ants versus birds in MediterraneanAustralia and southernAfrica, BERG, R. Y. 1975. Myrmecochorousplants in Australia and theirdis- in relationto the availabilityof soil potassium.Journal of Bioge- persal by ants. AustralianJournal of Botany 23: 475-508. ography9: 505-516. BOND, W., AND P. SLINGSBY. 1984. Collapse of an ant-plantmutualism: MULLER,R. N. 1978. The phenology,growth and ecosystemdynamics 640 AMERICAN JOURNAL OF BOTANY [Vol. 83

of Erythroniumamericanum in the northernhardwood forest. Eco- SAS. 1989. SAS/STAT user's guide. Version 6. SAS Institute, Cary, logical Monographs 48: 1-20. NC. . 1990. Spatial interrelationshipsof deciduous forestherbs. Bul- SCHEMSKE, D. W. 1981. Floral convergence and pollinator sharing in letin of the TorreyBotany Club 117: 101-105. two bee-pollinated tropical herbs. Evolution 62: 946-954. OOSTERMEIJER,J. G. B. 1989. Myrmecochoryin Polygala vulgaris, SMITH, B. H., C. E. DEREVIRA, C. L. BRIDGMAN,AND J. J. WOIDA. 1989. Luzula campestris,and Viola curtisiiForster in a Dutch dune area. Frequency-dependent seed dispersal by ants of two deciduous for- Oecologia 78: 302-31 1. est herbs. Ecology 70: 1645-1648. O'DowD, D. J., AND M. E. HAY. 1980. Mutualism between harvester TURNBULL, C. L., AND D. C. CULVER. 1983. The timing of seed dis- ants and a desert ephemeral: seed escape fromrodents. Ecology persal in Viola nuttallii: attraction of dispersers and avoidance of 61: 531-540. predators.Oecologia 59: 360-365. OHARA,M., ANDS. HIGASHI.1987. Interferenceby groundbeetles with WEIN, G. R., AND S. T. A. PICKETT. 1989. Dispersal, establishment and, the dispersal by ants of seeds of Trilliumspp. (Liliaceae). Journal survivorshipof a cohort of Erythroniumamericanum. Bulletin of of Ecology 75: 1091-1098. the TorreyBotanical Club 116: 240-246. POST, D. C., ANDR. L. JEANNE. 1982. Rate of exploitationof arboreal WESTOBY, M., K. FRENCH, L. HUGHES, B. RICE, AND L. RODGERSON. baits by ants in an old-fieldhabitat in Wisconsin. American Mid- 1991. Why do more plant species use ants for dispersal on infertile land Naturalist 108: 88-95. compared with fertile soils? Australian Journal of Ecology 16: RUHREN, S. 1994. Seed dispersal dynamicsof Erythroniumamerican- 445-455. um, a deciduous forest,spring ephemeral myrmecochore. Master's thesis,University of Maryland. College Park, MD.