Comparison of the Reproductive Biology of Two Neotropical Wrens in an Unpredictable Environment in Northeastern Colombia

The Auk 118(1):191-210, 2001

COMPARISON OF THE REPRODUCTIVE BIOLOGY OF TWO NEOTROPICAL WRENS IN AN UNPREDICTABLE ENVIRONMENT IN NORTHEASTERN COLOMBIA

JORGEA. AHUMADA • Departmentof Ecologyand EvolutionaryBiology, Princeton University, Princeton, New Jersey08544-1003, USA

ABSTRACT.--Buff-breasted(Thryothorus leucotis) and Rufous-and-white(T. rufalbus) wrens living in a dry forestin northeastColombia (Parque Nacional Natural Tayrona)are faced with a large year-to-yearuncertainty in the arrival time of the rainy season,as well as the amountof rain falling in the first six monthsof the year.Those factors are thoughtto be importantcues used by thosespecies in theirreproductive decisions. In thisstudy, I gathered data on severalreproductive parameters (clutch size, nestingsuccess, timing of breeding, renestingattempts) for both speciesduring two yearsof contrastingrainfall patterns.I collectedinformation on the foragingbehavior of bothspecies to identifytheir main food and to studyhow rainfall affectsthe dynamicsof thoseresources. Buff-breasted Wrens fed mostly in the understory,gleaning arthropods from upper and lowerleaf surfaces,dry branches, and aerial litter. Numbersof arthropodsin thosemicrohabitats depend strongly on the amountof rainfall;understory arthropod levels are low duringthe dry seasonand increase with the arrival of the rains.Buff-breasted Wrens timed their reproductionwith the arrival of the rains in both years,delaying the onsetof breedingsignificantly and continuingto breedduring the dry year(1994). Rufous-and-white Wrens spent a largeproportion of their time feedingon arthropodsin the leaf litter Number of arthropodsin the litter variedlittle betweendry and wet periods.Therefore, Rufous-and-white Wrens had a moreconstant food environmentdespite large differences in rainfallwithin andbetween the yearsof the study. Thatspecies started breeding earlier in the dry seasonand extended its breeding longer than Buff-breastedWrens. My observationssuggest that the evolutionof the reproductivestrat- egiesin thosespecies was mostly through the changeof behavioralparameters rather than physiologicalreproductive parameters such as changesin clutchsize, egg size,or number of broods.Received 14 June1999, accepted16 September2000.

MANY BIRDS LIVE in environments that are similar body size (Martin 1987, Paladino 1989). somewhatunpredictable as to when resources That imposesconstraints on the life-history are availablefor growth, maintenance,and re- characteristicsthat birds canevolve (Walsberg production.When should birds breed in an en- 1983).Thus, when the arrival time of food nec- vironmentthat is unpredictablein its favora- essaryfor reproductionis unpredictable,birds bility for reproduction?The very existenceof a are constrainedto breedwhenever pulses of re- distinctiveperiod of the year when mostindi- sources are above a certain threshold. This viduals of a bird speciesbreed--a breeding translatesinto a "tracking"strategy in which season--suggeststhat most of the time, the individuals shouldbe able to detectand pro- amountof foodavailable to themis justenough cess information on the amount of food avail- to sustain their basic metabolic needs. Because able to them and time their reproduction of their high turnoverrate of energyper unit of accordingly. body weightand their inability to allocatelarge Few studieshave specificallyexamined ef- amountsof energyto shortand long-termstor- fectsof the duration and intensity of resource age and growth,birds are more dependenton pulseson timing and durationof the breeding food for breeding than other endothermsof seasonof birds. That generally requires detailed knowledgeof the temporaldynamics of the resourcesthat birds requirefor reproduc- t Presentaddress: Laboratorio de Ecologfade Pob- laciones,Departamento de Biologla,Universidad Jav- tion aswell aslong-term data on breeding phe- eriana,Bogota, Colombia and Departmentof Botany, nology.One of the best data setsavailable for Universityof Georgia,Athens, Georgia 30602, USA. temperatezones comes from a long-termstudy E-mail:[email protected] in the Hubbard-Brook forest (Holmes et al.

191 192 JORGEA. AHUMADA [Auk, Vol. 118

1986). Most speciesin that forest dependon in birds--providing a natural setting to com- pulses of lepidopteran larvae for successful pare how differencesin their respectivefood breeding. Between outbreaks,birds usually dynamics affectsonset and duration of repro- havea lower reproductiveoutput and success, duction.In this study,I gatheredinformation suggestingthat mostspecies depend on brief on morphology,foraging behavior, and repro- but intensepulses of food whichconstrain the ductionof thosespecies and measuredvaria- time of breeding. tion in abundance of their main food: arthro- Suchlong-term study information is largely pods that occur on live plant tissuein the unavailablefor Neotropicalbirds. Thereis ev- understoryand in leaf litter. The two yearsof idencethat suggests that some Neotropical spe- my study were dramatically different in ciesexhibit a distinctbreeding season (see re- amount of rainfall and in the date of initiation view in Poulin et al. 1992), but examinationof of the rainy season.This offeredme a unique communitypatterns shows that other species opportunityto studythe reproductive strategy breedall year (Skutch1950, Miller 1963,Snow of bothspecies when faced with environmental and Snow 1963, 1964; Gradwohl and Green- variation in food supply. It also allowed me to berg 1990).Do the latter speciesrely on re- gain new insightinto someof the proximate sourcelevels that are more constantthrough mechanisms and ultimate causes that underlie time or that occur in long resourcepulses? the life historyof thosespecies. What types of food resourcesshow this dy- namicbehavior? Can speciesthat rely on these METHODS resourceschoose to breedin periodswhen nest STUDY AREA predationand othercauses of nestloss are low? Thereis partial evidencethat suggestingthat This studywas carried out in ParqueNacional Nat- Neotropicalbirds that utilize foods that are ural Tayrona(henceforth "Tayrona") located on the constantthrough time breed year round (Miller northeastCaribbean coast of Colombia(Fig. 1). The and Miller 1968).Good candidatesfor thosere- park consistsof a strip of land of 15,000ha of dry sourcesare arthropods in bark and rotten forestand scrubreaching from the coastalplains to wood (Pierpont1986), arthropods in leaf litter the SierraNevada of SantaMarta, and rising from (Levingsand Windsor1990, Poulin et al. 1992), sealevel to about400 m. This region of the country somecommunities of shrubunderstory arthro- receivesmoderate amounts of rainfall on a yearlyba- sis(<1,500 mm) and is highly seasonalas a resultof pods (Young 1994), or resourcesassociated the seasonalmovement of the thermal equatorwith with human-made habitats. In contrast, most two dry seasons(December to Marchand July to Au- shrubunderstory arthropods and fruits vary in gust) and two wet seasons(September to November abundanceseasonally (Janzen 1973a, b, 1980; and April to June) (Fig. 2A). Annual averagetem- Wolda 1978a,b, 1980,1990; Levings and Wind- peratureis around25øC with daily temperaturesos- sor 1990, Poulin et al. 1992, Blake and Loiselle cillating between 20 to 31øC (taken from Instituto 1992, Heideman 1989, Hilty 1980, Kinnaird Nacionalde Hidrologia,Meteorologia y Adecuaci6n 1992,Levey 1988,Loiselle and Blake 1991,Van de Tierras--HIMAT). That climaticpattern is influ- Schaiket al. 1993). encedby the presenceof the SierraNevada of Santa Marta which can affectthe extentof rainfall by con- Thryothorusleucotis and T. rufalbusare two densationof moisture along its northeastslope speciesof insectivorouswren, (Buff-breasted (Herrmann 1970). This has created an east-west and Rufous-and-whitewren, respectively)liv- moisturegradient along the park that determinesthe ing sympatricallyin the northeasterndry for- structureof the vegetation(Herrmann 1970, Hernan- estsof Colombia.They constitutea goodsys- dez-Camachoand Rodriguez-Guerrero1972). De- temfor studinghow temporal variation in food spite that generalannual pattern in rainfall season- abundanceaffects timing and intensity of re- ality, the amount of rain in a given year varies productionbecause (1) the dry forestthey live dramaticallyespecially in the first half of the year in has highly unpredictablerainfall not only (Fig. 2B).An analysisusing Colwell's index (Colwell 1974, Beissingerand Gibbs 1993) showsthat pre- fromyear-to-year, but alsowithin a year,which dictabilityof rainfall for Tayronais low [0.239]when might influenceinsect and arthropodabun- comparedto climaticstations within the sameregion dancein their habitat,and (2) the speciesdiffer but outside of the influence of the Sierra Nevada de in feedingheight and slightlyin body size-- Santa Marta (Ahumada 1995). I installed a rain factors that are known to influence food choice gauge and a maximum-minimum thermometerin January2001] Reproductionoftwo Thryothorus wrens 193

Sierra Nevada

5 km Venezuela

Caribbean Sea N Colombia Los Naranjos

Per6

Santa Mart Parque Nacional Natural Tw Road Rio/ Piedras

Sierra Nevada de Santa Marta

FIG.1. Geographicallocation of the study site in northeastColombia.

the park guard stationlocated at Tayronaand data (Desmoncusorthacanthos), and shrubs.Several species were collecteddaily from August 1992to June1993. of birds inhabit the secondaryforest at LosNaranjos The two years(1993 and 1994)in whichreproductive and mostof themare common.Among the dominant information of the wrens was collected differed in speciesare Buff-breastedWrens (Thryothorusleuco- amountof rainfall (Fig. 2C). That differencewas par- tis), White-bellied Antbirds (Myrmeciza longipes), ticularlypronounced for the first five monthsof the Crimson-backedTanagers (Ramphocelusdimidiatus), year; 1993 received300 mm more rain than 1994. and Lance-tailedManakins (Chiroxiphialanceolata). A Thosemonths are the most crucial for the breeding more comprehensiveand completelist of the birds of mostbird speciesin the area includingthe wrens of the area can be found in Ahumada (1995). (J.Ahumada pers. obs.). The differencein total rainfall between 1993 and 1994 was on the order of 200 STUDY SPECIES mm. Also, and maybemore significantfor the wrens, therewere differencesin the arrival of the rainy sea- The focalspecies of this studybelong to the family son in both years. In 1993, rains started by day 50 Troglodytidae (Wrens), genus Thryothoruswhich and accumulatedmuch more rapidly in a shortpe- means"the reedleaper" (Jobling1991). Most species riod in the third quarter of the year. In 1994,rains in the genusare mediumsized (11.5 to 15 cm in body started later, and accumulation was more constant length) and are usually found in pairs that defenda throughoutthe year. territory all year round. Individuals are often incon- My study site is locatedon Estaci6n"Los Naran- spicuousand difficult to seewhile they foragebut jos" (approx. 11ø17'N,73ø9'W), a 2 X 5 km strip of they are frequently heard. The pair constructsa forestlocated on the northeasterncoastline of Tayrona dome-shapednest with a sideentrance. (Fig. 1). I establisheda 25 ha studyplot in the eastern The Rufous-and-whiteWren (14 to 15 cm, body sideof LosNaranjos bordering the bank of RfoPiedras. length) is found throughoutthe Pacificslope of Cen- A 20 X 20 m grid was createdby placing20 cm plastic tral America from south Mexico to Panama and in stakesin thenodes of thegrid, which were located with northeastern South America. It lives in deciduous the aid of a hand-heldcompass and measuringtape. woodland, gallery forest, and forest borders up to Most data collectionwas concentratedon this plot, 1,500m (Hilty and Brown 1986).The songof Rufous- althoughsome information was drawn from a larger and-whiteWrens is very characteristic,consisting of area of about2 km2 surroundingit. The 25 ha study easily localizable pure tones and intermediate-fre- plot is dominatedby secondary(54.5%) and primary quencywhistles. Males sing the most,although oc- forest(12%). The remainingvegetation is composed casionallyduets are also heard (J. Ahumada pers. of a mixture of coastalvegetation, abandoned coco- obs.).The only publishedinformation about its re- nut plantations,and smallpatches of grass.The un- productivebiology and behaviorcomes from occa- derstoryis thickand composedof vines,spiny palms sional observationsof Skutch(1960) and a study by 194 JORGEA. AHUMADA [Auk, Vol. 118

400 productionand behaviorof this species.Birds live in 3501 pairs but also occasionallyforage in small groupsof four to five individuals,probably family groups(Hil- 250 3øø1 ty and Brown 1986;J. Ahumada pers. obs.).As the •5o lOO Rufous-and-whiteWrens, they also defend territo- 5 ries and duet frequently(J. Ahumada pers. obs.).

i 2 3 4 5 6 7 8 9 lO 11 12

Month COLLECTION OF FORAGING AND DIET DATA

I collecteddata on foragingbehavior of the wrens 9OO mostly during the secondfield season(March-Sep- 8OO tember1994) and most intensivelyduring the rainy 600 500 season.I locatedindividuals or pairsof eachspecies 400 by songand followedthem as long as possible.All 300 200 10 data were collectedby recordinginformation in a mi- crocassettetape recorder.As soon as I spotted a for- 8O 81 82 83 84 85 87 88 89 9O 91 92 93 94

Year agingindividual, I estimatedits heightabove ground and recordedthe type of microhabitat(forest floor, trunk, understoryshrub, vine tangle).I only record- • 300 ed this information the first time that an individual • 250 = 2oo was spotted to assure independencebetween sam- •-• •' 150100 ples.I then followedthe individualcontinuously, re- .•_•, so eo• cordingthe numberof captureattempts made and the number of patch changes.I identifieda patch • -lOO •, 450 changeas a shortflight of more than 2 to 3 m away i 2 3 4 5 6 7 8 9 10 11 12

Month from the originalforaging position followed by more foraging in the new location.I discardedall focal FIG.2. (A) Monthlyaverage rainfall (standard er- samplesof less than 2 min (maximum 3 min). Al- ror in whiskers)from 1980to 1994in ParqueTayrona thoughthe numberof captureattempts has some (data obtained from Instituto de Hidrologfa, Meteo- limitations as an estimatorof foraging efficiency,I rologia y Adecuaci6nde Tierras,HIMAT). (B) Yearto used it insteadof the proportionof successfulcap- year variationin rainfall from the first six monthsof tures becauseit was very difficult to observesuccess the year from 1980 to 1994 in ParqueTayrona. (C) or lack thereofdirectly because of the high speedof Differencein monthly rainfall (1993minus 1994)in thoseevents. I alsorecorded aggressive interactions Parque Tayrona. in the form of chaseswhenever they occurred.Be- causethere were only four territoriesof Rufous-and- white Wrenswithin theplot, datawere also collected Winnett-Murray (1986). Birds at MonteverdeNation- from an additionalfive to six pairs outsideit. All the al Park,Costa Rica, breed once a yearby constructing foraging information on Buff-breasted Wrens was an invertedbow-shaped nest and laying a clutchof collectedfrom pairs whose territories were within two to four eggs.As do mostwrens in this genus,it the plot (51 pairs). lives in pairs that defenda territory all year round. Detailed analysisof the diet compositionof both Individualsalso construct several dummy neststhat wrenswould have disturbedtheir reproductivebe- are not usedfor reproduction;some species use these havior,which wasthe primary interestof my study. nests as dormitories but their function is not clear in Therefore,to estimateindrectly what the birdswere others. eating,I observedthe main substratesthey used and The Buff-breastedWren (11 to 12 cm) is a smaller then sampled intensivelythe arthropodsin these and more widespread species.Its range extends substrates(next section).I only checkedthe stomach from southern Panama into South America east of contents of two Buff-breasted Wrens and three Ru- the Andes to northern Bolivia, central and northern fous-and-white Wrens and recorded the items Brazil and Caribbeanlowlands, and upper Amazon broughtto the nestby a pair of Buff-breastedWrens. basins of Colombia and Venezuela. It is usually Individuals outsideof the study plot were mist-net- foundin a largervariety of habitatsthan the Rufous- ted in February 1993. The method used was after and-white Wrens; deciduous woodland borders, Moody (1970). After capture, a 2-3 mm diameter overgrown clearings, mangroves, Amazonian var- plastictube coatedwith vaselinewas introducedin z6a, and river bordersup to 950 m abovesea level. It the mouth and forcedthrough the esophagusinto is also found in several of the Pearl islands in Panama the gizzard. Oncethe tubing was in place,a luke- (Rey, Vfveros, Puercos,and Carlas) (Ridgely et al. warm saline solutionwas injectedapplying slight 1989).There is no publishedinformation on the re- pressureto forceout the stomachand gut contents January2001] Reproductionof two Thryothorus wrens 195 via the cloacaor mouth.The gut contentswere col- a wren would find during 15 min in a cylinder 2 m lectedin a plasticcup andpreserved in 35%ethyl al- high with a baseof 4 m in diameter.This methodwas coholfor examinationin the laboratory. preferredover methods like sweepnetting, bagging, A pair of Buff-breastedWrens was observedfeed- malaisetraps, and light trapsbecause those usually ing its nestlingsduring three consecutivedays in sampleflying arthropods,which were rarely pur- May 1993. I observedparents bringing food to the suedby wrens.Some methods fail to givean appro-- nestlingswith binocularsand identifiedand record- priate measureof the effort requiredto find a given ed prey itemsto orderor suborderwhen possible. amountof arthropodsper unit of time or volumeof forestsampled. The methodof standardizedcounts ESTIMATION OF TERRITORY SIZE used here has been used before in avian studies, in- cluding previousstudies of wrens (Winneff-Murray With the help of an assistant,I mappedthe terri- 1986). toriesof all pairsof bothwren species within theplot. I searchedfor arthropodson upper and lowerleaf Thiswas done systematically by walkingthrough the surfaces,in rolled-up dry leaves,on the surfaceof plot playingback the recordedsong of a pair or in- dead and living twigs, and on the surfaceof flowers dividual of eachspecies. I estimatedthe sizeand lim- and fruits. I also countedlive arthropodsthat were itsof theterritory by movingthe playback to reference hangingfrom or standingin silk (e.g.spiders). When- pointsand observing the behavior of theresident pair ever I encounteredan arthropod,I recordedthe fol- and neighboringpairs. Severalindividuals of both lowing information:taxonomic order, substrate where species(12 Buff-breastedand 9 Rufous-and-white it was found (see above),life stage(adult or larvae), wrens) were color-bandedso it was possibleto con- and size.Only the lengthof thebody (head-tip to ab- firm the permanentnature of the territories,at least domen-tip)was used to estimatesize. I classifiedar- duringthe studyperiod. thropodsin five differentsize classes:1 (2-5 mm), 2 (6-10 mm), 3 (11-15 mm), 4 (16-20 mm), and 5 (>20 TEMPORAL VARIATION IN ABUNDANCE AND BIOMASS mm).Ants, isopods, dead arthropods, and arthropods OF ARTHROPODS with aposematiccoloration were excludedfrom the counts. I considered more than five individuals of the I collected information on the abundance of arthro- samearthropod species aggregated in a givensub- pods during two seasons:from September1992 to strate as a "cluster." For clusters, I recorded the same June1993 and from Marchto August1994. The type informationas for individual arthropods.However, of information collected and the methods used were clusters were considered separately in the data different in those two seasons. The first field season analysis. constitutedthe samplingof arthropodsthrough 10 The arthropodcounts made during the first field monthsat biweeklyintervals. This regularsampling seasondid not take into accountlitter-dwelling ar- precludeda large samplesize during eachsampling thropods,which were found to be extremelyimpor- event (10 stations),but was necessaryto look at the tant for the diet of Rufous-and-white Wrens. There- relationshipbetween arthropod numbers and rainfall fore,during the secondfield season(March through overseveral months. During the secondfield season, August1994) the methodologywas modified, both to intensivesampling was done throughout the plot (175 include litter-dwelling arthropodsand to increase stations)at eachsampling event, but only threesam- number of sampling stationsfor both types of ar- pling eventswere completedat two-monthintervals thropodsas explained before. Litter arthropodswere (firstdry season,beginning of thewet season,and sec- sampledby collectingsamples of litter at 175stations ond dry season).Those were intended mostly to in- (correspondingto previouslyplaced plastic stakes) vestigatefactors that affectthe spatialdistribution of distributed in a regular lattice 40 m apart in the arthropodsin the forestin boththe understoryand studyplot. At eachstation, two samplesof leaflitter forestlitter. Most data in this paper are from the sec- were collectedby placing a woodenbox (33 x 28 cm) ond field season; I refer to data from the first field sea- upside-downover the groundand slidinga pieceof sonwhen appropriate. plywoodunderneath, trapping the litter in the box. During the first season,I only evaluatedthe abun- The exact locationof the box was determinedby danceof foliage-dwellingarthropods by performing throwingit at randomwithin a 2 m radiusof the lo- counts in 10 randomly selectedstations within the cationof the plasticstake. The litter of bothsamples study plot. The stationswere locatedinside territo- at eachstation was combinedand placedin a zip- ries of wrens of both species.The countswere done lockbag. A total of 20 to 25 stationswere sampled in at bimonthlyintervals during morninghours with a day and their litter contentswere taken to the lab the help of an assistant.The rationalebehind the to separatethe arthropodsthe sameday. Litter sam- methodwas to imitate closelythe behaviorof a wren ples were emptied into metal trays and the arthro- basedon my best knowledgeof the substratesthat podswere separated by hand and collected in plastic thosebirds were seen to inspectwhen foraging. The bottles containing75% ethyl alcohol.Arthropods methodattempts to measurenumbers of arthropods smallerthan 1 mm, ants,isopods, and arthropods 196 JORGEA. AHUMADA [Auk, Vol. 118 with aposematiccolorations were excluded from the REPRODUCTIVE BIOLOGY countsbecause these are rarely consumedby insectivorous birds. Information on the size class, taxo- I searchedintensively for nestsof both species nomicorder, and life stagewas gathered for eachin- from August 1992 to July 1993 and March-August dividual arthropodfound in the sample.The whole 1994.My searchesfor Buff-breastedWren nests were plot wassampled three times during the season:first mostlyrestricted to the studyplot. However,due to dry season(March), first rainy season(May), and their lower density,I surveyeda large area for Ru- seconddry season(August). Each samplingeffort fous-and-whiteWren nests(about 2 km). I observed took about 10-12 days. In March, only 135 stations thebehavior of individualpairs of bothspecies to de- were sampledbecause of the following. The first termine their reproductiveactivity. This was relatively easy to do becausemembers of a pair usually litter samples were collected and, once bagged, perform their activitiestogether, so the prolonged sprayedwith insecticideto kill the arthropodsand absenceof one them (mostly the female) usually to facilitatetheir separationfrom the litter in the lab- meant that therewas nestingactivity of somekind oratory.However, it tookme andmy assistantsmuch (incubation,nestling feeding). I then carriedout an more time to separatearthropods from insecticide- intensive search for the nest. Pairs that I observed sprayedsamples than from insecticide-freesamples. building nestswere followedlater to checkif they In the latter, the arthropodswere detectedeasily had eggs. while trying to escapefrom the metal tray whereas Once I found an active nest, I checked number of in the former,we missedmany arthropodsbecause eggs every two days or as frequently as possible. of their immobility.Therefore, I abandonedinsecti- With thehelp of an assistant,I visitednests that were cide use and discarded the data from insecticide- higher than 2-3 m using an aluminumladder that sprayedsamples. could be extendedup to 12 m securedby ropes.In Countsof foliage-dwellingarthropods were simul- the 1994season, I measuredthe widths and lengths taneouslydone with the collectionof litter samplesat of eggsfrom severalclutches of both speciesto the each station. The methods followed were the same as nearest0.1 mm. I observedand recordedwhen eggs, usedin thefirst field season, with theexception of the nestlings,or fledglingsdisappeared or when addi- durationof eachcount. Due to the large numberof tional eggs were added by parasites.Sometimes it stationssampled and the needto samplethe plot in was easyto determinecauses of nestfailure (infertile no morethan 10to 15 days,counts were reduced from eggs,signs of predators,brood parasites).A nestwas 15 to 10 min in eachstation. To make comparisons considered successfulif at least one nestling was amongfield seasons, the numberof arthropodsfound fledged.Based on neststhat were followedin their in eachstation during the first field seasonwas ex- entirety,I estimatedthe lengthof the incubationand pressedas an averageper minuteand then multiplied nestlingperiods to be 14-15 daysfor eachspecies. I by 10. BecauseI collectedinformation on the size of used that informationto estimatethe startingdates eacharthropod counted, it was easyto calculatear- of clutchesthat were foundin the laying or nestling thropodbiomass by usingpublished within-order re- stage. gressionequations of dry biomassand body length (Rogerset al. 1977,Sample et al. 1993).I usedthe me- RESULTS dian size classas an estimateof the lengthof an arthropod:class 1 (3 mm), 2 (8 mm), 3 (13 mm), 4 (18 DIET mm), and 5 (25 ram). If the regressionequation for a givenorder was lacking, the following general regres- Both speciesof wrens are typical foliage sion for all insectswas applied (Rogerset al. 1976): gleanersthat spendmost of their time looking W = 0.0305L 262 for arthropodsin the undersidesand tops of understoryshrub leaves and tangles,in rolled- whereW is the weightof theinsect in milligramsand up dry leaves,on the surfaceof branchesand L its body lengthin millimeters. in leavesin the litter (seeAhumada 1995for de- To make temporal comparisonsin abundanceand tails on microhabitatuse). Inspectionof those biomassfor both foliageand litter-dwellingarthropods, it was necessaryto determinethe degreeof substratesshowed that they harbored mostly spatial autocorrelationfor the data. Becausethe spiders,coleopterans, orthopterans, homopter- countsand litter sampleswere takenat regular40 m ans, and to a lesserextent lepidopteran larvae intervals,the degreeof spatialassociation had to be (leaf surfaces:32% spiders, 18% coleopterans, determinedbefore considering each sampling point 13% homopterans,10% orthopterans,4% lep. as independent(Legendre 1993). Details of how the larvae, 23% others; undersides of leaves: 59% degreeof spatial autocorrelationwas calculatedand spiders, 19% coleopterans,9% homopterans, the main resultsare presentedin Appendix 1. 4% lepidopteranlarvae, 11% others;rolled-up January2001] Reproductionoftwo Thryothorus wrens 197

A u•90 65% of the samplesoccurring on the ground. That verticalsegregation also resulted in Buff- •E60 ßBuff-Breasted breasted Wrens foraging in moremicrohabitats 50 []Rufous-and-Wh•te ø40 than Rufous-and-whiteWrens. A comparison •3o of the total numberof captureattempts for both speciesshowed that therewas no differencebe- =•o I L I•_ m•_,m_ tweenthe species(Sign Test, Z = 1.23,P = 0.21, 0 2 4 6 8 10 12 14 16 Foraging height (m) n = 139). However,when captureattempts are B dividedaccording to microhabitatfor eachspecies (Fig. 3B), it is clear that Buff-breasted Wrenswere equallygood at capturingarthro- i ßVine pods(or at leastattempting to capturethem) in all the microhabitatsthey forage in, whereas '• 0.6 TI []Leaf litterRufous-and-white Wrens were equally goodto .• 0.4 Buff-breastedWrens only in the forest floor • 0.2 Z O, , (Mann-Whitney tests: Understory branch-- Buff-breasted Rufous-and-white Buff-breasted vs. Rufous-and-white, Z = Species -2.78, P < 0.001, n = 53; Floor, Z = -0.75, P FIC. 3. (A) Distributionof foragingheights for = 0.44, n = 46). both speciesof wrens. (B) Averagenumber of cap- While observingboth species, I noted that all ture attemptsper 2 rain observationbouts (standard aggressiveinteractions were directed at Ru- errorsin whiskers)for bothspecies of wrensin three different microhabitats. fous-and-whiteWrens which were always displaced. Out of a total of 10 aggressions,7 occurred while they were foragingabove 1 m in dry leaves:43% spiders,20% orthopterans, 20% heightin eithera understorybranch (n = 5) or coleopterans,2% lepidopteranlarvae, 15% oth- vine tangle (n = 2). In five cases,Buff-breasted ers; leaf-litter: 33% spiders,16% coleopterans, Wrensaggressively chased Rufous-and-white 19%pseudoscorpionids, 9% orthopterans,23% Wrens and in two cases White-bellied Antbirds others). Those results indicate that those taxa chased Rufous-and-white Wrens. wereprobably the mostimportant in thewren's diet, althoughpseudoscorpionids seemed also importantfor Rufous-and-whiteWrens (see be- TERRITORY SIZE AND DENSITY low and Ahumada 1995). Despitepartial digestion,the stomachcon- From those observations in habitat use of tents confirmed this. The stomach contents of both species,one would expectthat the two di- two Buff-breasted Wrens and three Rufous- mensional projectionsof Rufous-and-white and-white Wrens showedremains of orthop- Wrenterritories have to be largerthan thoseof terans,spiders, and coleopterans.Additionally, Buff-breastedWrens because their main forag- observationsof a pair of Buff-breastedWrens ing microhabitat(the forestfloor) is basically that broughtfood to a nest by stoppingin a two dimensional in nature. On the other hand, nearbybranch first, show that mostof theitems the territoryof a Buff-breastedWren encloses a consistedof spiders(13/30), lepidopteralarvae volume of forestranging from 0 to about 15 m. (7 / 30) coleopterans(4 / 30), adult butterfly (1 / 30), hemipterans(2/30), orthopteran(1/30), Indeed,a comparisonof the sizeof the two di- and smalldragonflies (2/30). mensionalprojections of the territoriesof both speciesshowed a differencein aboutan order

FORAGING BEHAVIOR of magnitude(Fig. 4). Therewas alsoan order of magnitudedifference in the densityof both The two speciesclearly differed in their for- speciesin the plot. Whereas Buff-breasted aging height (Fig. 3A). Buff-breastedWrens Wrens were very common (51 pairs, density = foragedbetween 0 and 18 m, peakingslightly 4 individuals/ha), Rufous-and-white Wrens at 1 and 6 m. Rufous-and-white Wrens had a were rare (5 pairs, density = 0.4 individuals/ smallervertical foraging range (0 to 8 m) with ha). 198 JORGEA. AHUMADA [Auk,Vol. 118

In 1994,number of stationssampled was increased(from 10 to 175) at the costof decreas- ing the interval between successivesamples. However, the pattern was similar to that of 1992-1993;arthropod numbers in the understory increasedwith the arrival of the rainy season and then decreased as the rains subsid- ed in August (Fig. 5) (Wilcoxonmatched pairs test:ZMa,ch-•ay = 4.550, P < 0.001,Z•ay-û•st = 7.521,P < 0.001).Average biomass of arthro- FiG.4. Comparisonof the two-dimensionalpro- podsshowed a substantialdecrease during the jectionsof territoriesof both speciesof wrens in the seconddry seasonwhen comparedto the first study plot. Small-filled territories are from Buff- dry seasonand rainy season (t-test: t•aarch-ûgust = breasted Wrens. Large-transparent territories are 1.54,P < 0.05;t•4•y_ûg• t = 7.6, P < 0.05).There from Rufous-and-white Wrens. wasno differencein the biomassof arthropods betweenthe first dry seasonand the wet season

TEMPORAL VARIATION IN ARTHROPOD (t-test:t•arch_•y = --0.17, P = 0.86). ABUNDANCE AND BIOMASS Leaflitter.--In contrastto understoryarthro- pods, litter-dwelling ones were relatively in- Forestunderstory.--Biweekly counts of under- variant in both numbersand biomass.Despite storyarthropods showed a positivecorrelation the largechange in rainfallfrom Marchto May with rainfall from September1992 to June1993 and from May to August,there was no signif- (r = 0.259,P = 0.03,n = 18).During mostof the icant differencein numberof litter arthropods dry season(January-April), arthropod numbers amongthe three differentseasons (Fig. 5). Bio- were low and then steadilyincreased with the mass,however, was significantlylower during arrival of the rains in May and June. the seconddry seasonby - 3 g in average(t-

ß Understory

14 [] Leaf litter 40 o 12 [] Rain '•3 10 30 • Eo 25 • 20 • 6 15 •- .• 4 10 • • 2 5 n- o March May August FiG.5. Variationin the averagenumber of arthropodsfound during 175,10 min countsper monthin the understoryand 175superficial litter samples(0.18 m x) per month(whiskers denote standard errors) in three differentmonths in 1994:March (first dry seasonof the year),May (first wet seasonof the year),and August (seconddry seasonof the year).White barsdenote the amountof rainfall within 45 daysprevious to the first day of eachsampling period. The numberof arthropodsper samplein the understorywas significantly higher in May (Wilcoxonmatched pairs test, seetext). January2001] Reproductiono.ftwo Thryothorus wrens 199 test: tMa¾_Au•ust= 2.5, P = 0.01, tMarch_Au•ust= 1.76, Sevenout of the 15 taxa showeda significant P = 0.02) (seeAhumada 1995). changein abundancethrough the samplingpe- riod. However, only one taxon, Pseudoscor-

COMPOSITION OF THE ARTHROPOD pionida, showeda significantincrease during COMMUNITY the rainy season in May which persisted through August (Appendix 5). Interestingly, spiders,which were the mostabundant taxon, ForestUnderstory.--A total of 4,813 arthro- showedno changein abundancethrough time. pods was found by inspectingthe forestun- In general,leaf-litter taxa were more invariant derstory during the 1994 field season.More throughtime in both relativenumbers and bio- than half of the arthropodswere spiders,fol- mass comparedto arthropod taxa in the un- lowed by coleopterans,orthopterans, and ho- derstory.Some taxa that were highly variable mopterans (Appendix 2). Biomassof spiders in the understory(Araneae, Orthoptera, Het- alsoconstituted the largestproportion (72.7%) eroptera)showed no changein the leaf-litter. followedby orthopterans(6.9%) and coleopterans (4.7%) (seeAppendix 2). Some,but not all taxa (7 of 12)showed chang- REPRODUCTIVE BIOLOGY es in absolutenumbers through time. Appen- dix 3 showsthe resultsof pairwise compari- Both speciesof wrens constructglobular sons between the numbers of different taxa nests made of small sticks, plant fibers, dry leaves,and feathers.The nesthas the shapeof for different sampling regimes(dry, wet, and dry season).Only spidersand homopterans an invertedelbow with the bendingpoint over a branchor vine supportingthe entire weight showedan increasein numbersduring the wet of the nest and the entrancepointing down- season(May) comparedto the dry periodsbe- wards. Both male and female participate in fore and after the rains (March,August). Pair- nest construction. Rufous-and-white Wren wise comparisonsof the biomassof different nestswere larger than thoseof Buff-breasted taxa throughtime showedsimilar trends com- Wrens, measuring50 to 60 cm from the en- paredto absolutenumbers of arthropods.Spi- tranceto the edge of the incubatingchamber ders were significativelysmaller (same number (about 30 to 40 cm for Buff-breastedWrens). but smallerbiomass) at the beginningof the The two speciesshowed differences in their seconddry season(August). Other taxa such preferred nestingheights and the generallo- as coleopterans,orthopterans, and phasmids cation of their nests. Buff-breasted Wrens usu- showed similar trends to spiders. The data ally nestedlower in the forestwith over50% of showa patternin whichgroups that increased the nestsplaced between 1 to 2 m (n = 28).They in numberduring the rainy seasondemonstrat- alsonested in a variety of substratesincluding ed a decreasein averagebiomass either after shrubs,vine tangles,and spiny vine palms.Ru- (Araneae)or before(Homoptera) the rains.For fous-and-whiteWrens nested higher (up to 10 othergroups, changes in numberswere consis- m) and constructedmost of their nestshanging tent with changesin biomass;when their num- from Desmoncusspiny palm vines and occa- berswere high, so was their biomass,and vice sionally(2/12) on the top of small understory versa.Half of the taxa showedno changein trees. abundanceor biomassthrough time. Surpris- Buff-breasted Wrens constructed additional ingly, lepidopteranlarvae were among that dormitory neststhat were never used for re- group. production.Those nests were smaller and shal- Leaflitter.--A total of 2,420arthropods were lower than breedingnests and severalof them found in 485 leaf-litter samples collected in could be found within each of the territories of 1994.Again, spiders were the predominanttax- Buff-breasted Wrens. I confirmed their use as on followed by pseudoscorpionids,coleopter- dormitoriesby observingBuff-breasted Wren ans,and orthopterans (Appendix 4). Diplopods individuals enteringthem at dusk. In contrast, were predominantin biomassowing to the I never observed such behavior in Rufous-and- their large size, but spidersstill constituted white Wrens.They did havetwo to three nests 27.8% of the total biomass,followed by cole- at a giventime and they reusedold nestsby re- opterans,dictyopterans, and scorpionids. placingold twigs, repairingholes, and renew- 200 JORGEA. AHUMADA [Auk, Vol. 118

TABLE1. Summaryof reproductiveparameters for both speciesof wrensin two differentyears: wet (1993) and dry (1994). Mean clutchstart date is given in Julian date _+SD. BB = Buff-breastedWrens, RW = Rufous-and-whiteWrens. * = significantat P < 0.05.

Mean clutch Proportion Lossto Lossto Lossto Species Year No. of nests startdate successful predation parasites other BB 1993 16 105 + 38.7 0.31 0.54 0.27 0.18 1994 29 156 _+ 32.7* 0.38 0.55 0.28 0.17 RW 1993 9 115 + 32.9 0.33 0.50 0.33 0.16 1994 7 129 + 16.4 1 0 0 0

ing the lining of the incubatingchamber. Once ants),the signswere unequivocalthat preda- I observeda pair of Rufous-and-whiteWrens tion had occurred. constructinga nestfor two weekswhile simul- Broodparasitism by Striped Cuckoos(Tapera taneouslyrepairing and old nestthat had sev- naevia)and Shiny Cowbirds (Molothrusbonar- eral holes.A week later the femalelaid eggsin iensis)was also a common sourceof nest failure the old nestwhile leavingthe new nestintact. for both species.Cuckoos were more cormnon Femalesof both specieslay one egg per day in Los Naranjosthan cowbirds,and accounted to completea clutchof 2 to 3 eggs(Buff-breast- for seven of nine parasitized nests. Cuckoos ed Wrens: 2.8 + 0.38, n = 23; Rufous-and-white were sightedand heardmostly at LosNaranjos Wrens:2.7 __0.45, n = 11). Eggsof Rufous-and- at the beginningof Junein the middle of the white Wrenswere significantlywider and lon- rainy season.For the 1994 season,Buff-breast- ger than Buff-breastedWren eggs (Ahumada ed Wren pairs that bred beforethe arrival of 1995). The eggs of Rufous-and-whiteWrens cuckooshad a higherprobability of fledgingat were uniform ocean blue without speckles, leastone chick(0.63) than pairs that bred after whereasBuff-breasted wren eggswere cream the arrival of cuckoos(0.22). colored,speckled with brown, and blue,espe- Other causes of nest failure included infertile cially at the larger end. eggsand falling nests. In 1994,these accounted The incubationand nestlingperiods lasted for about 10% of nest failures in Buff-breasted 14 to15 dayseach for both species.After hatch- Wrens and a similar proportion for Rufous- ing, both parentsbring food to the nestlings, and-whiteWrens in 1993(11%). The proportion and the fledglingsremain six to eight weeks of successful nests and the contribution to nest with their parents.Second broods were infre- failure from predation, brood parasitism,and quent; during the secondfield season,I only other causeswas relatively similar in Buff- observed second broods in one out of 6 Rufous- breastedWrens for both the wet and dry year and-white Wren breedingpairs and in 2 out of (Table1). In contrast,Rufous-and-white Wrens 24 breeding Buff-breastedWren pairs. During had a muchhigher nesting success in the dry 1994,it waspossible to estimatethe proportion year comparedto the wet year. In 1993,pro- of the populationof both speciesthat attempt- portion of successfulnests and relativecontri- ed to breed(laid at leastone clutch of eggs).Of butions of nest failure were similar for both 9 pairs of Rufous-and-whiteWrens, 6 bred species. All Rufous-and-white Wren nests (0.66), and of 51 pairs of Buff-breastedWrens, fledged at least one nestlingin 1994. 24 bred (0.47). Timing of nestingand durationof the breeding Nestingfailure.--The main causesof nesting season.--Thespecies differed significantlyin failure for both specieswere predation and their mean date of clutch initiation for 1994 brood parasitism(Table 1). Nests that were (Fig. 6). In the previousyear, Buff-breasted preyed upon showedtypical signsof disrup- Wrensstarted breeding around the third week tion, such as holes in the side, deformation and of April with most of the clutchesstarting in distentionof the main entrance,and egg shell the first and secondweeks of May. However,in remainson the ground.Although the natureof 1994only a few pairshad startedto showsome the predators was not confirmed (exceptfor reproductiveactivity by the first week of May three nests which were taken over by army and most of the populationstarted laying by January2001] Reproductionof two Thryothorus wrens 201

400 350 300 1993 250 I - T.rufalbus • 200 '!': 150 ,, j•1994 100 50 , T. leucotis 0 I -50 0 20 40 60 80 100 120 140 160 180 200

Julian Date FIG.6. Comparisonofthe length of the breeding season and mean date of clutch initiation for both species ofwren in thetwo years of the study. The cumulative rainfall for the first 181 days of each year is plotted in they-axis (thin line, 1993; thick line, 1994). The thin horizontal bars show the length of the breeding season in 1993for Rufous-and-whiteWrens (RW) and Buff-breastedWrens (BB). These were determinedfrom the firstand last date when a clutchwas initiated. The thick horizontal bars show the length of thebreeding seasonfor 1994. The dark circles over each bar indicate the mean date of clutch initiation. Date 0 = 1January. the end of that month(Mann-Whitney U-test, microhabitats,one would expectthat individ- Z = -3.51, P = 0.00, n = 23). In contrast,Ru- ualsexperience a high year-to-yearvariability fous-and-whiteWrens started breeding at the in foodabundance in a placelike Tayronawere beginningof April in bothyears and were more rainfall is so variablebetween years. Repro- spreadout in their nestingduring the season ductiveactivity of the mainnest parasite in the with some pairs breeding even into June area (Striped Cuckoo)and predator activity (Mann-WhitneyU-test, Z = -0.85, P = 0.39,n also seemclosely dependent on the arrival of = 13).Buff-breasted Wrens delayed their repro- the first rainy season.What is thereproductive ductionfor at least a month in the dry year, strategyshown by Buff-breastedWrens in face whereas Rufous-and-white Wrens behaved of thisenvironmental variation in foodsupply similarlyin boththe dry andwet years(Fig. 6). and activity of nest parasitesand predators? My resultsindicate that Buff-breastedWren re- DISCUSSION productionis closelytied to arthropodabun- dance,which in turnseems to bedetermined by Theresults of thisstudy show that differenc- the arrivalof the rainy seasonwith a minimum es in rainfall can affectsignificantly the abun- cumulativerainfall of around50 mm (Fig. 6). danceand biomassof arthropodsavailable for AlthoughI could not comparearthropod Buff-breastedWrens in the understory.How- abundancebetween the wet and dry yearsof ever, rainfall does not seem to have such a clear my study,the within-yearcomparison in 1994 effect on the abundance and biomass of arthro- clearly suggeststhat arthropod abundancein pods in the leaf litter. BecauseBuff-breasted the understoryand rainfall are closelylinked. Wrensprefer to eat arthropodsin the understo- In both years, Buff-breastedWrens started ry andexhibit similar capture rates in different breedingonly when approximately50 mm of 202 JORGEA. AHUM^D^ [Auk,Vol. 118 rainfall had fallen. That amount of rain was breasted Wrens. Those chases occurred while attained about 30 days earlier in 1993 than Rufous-and-whiteWrens were foragingabove 1994,and Buff-breastedWrens corresponding- 1 m in the forestunderstory. Observations of ly startedclutches 50 daysearlier on averagein Rufous-and-white Wrens in Panama and Costa 1993.If changesin photoperiodwere used as Rica indicate that in these sites, the species an environmentalcue by the birds to start re- seemsto feedin the understorymore frequent- producing, one would expectno differencesin ly than at Tayrona (Winnet-Murray1986; T. averageclutch initiation date betweenyears. Robinsonpers. comm.).At leastin Panama,the Therefore, it seems reasonable to assume that density of Buff-breastedWrens is lower than in thosebirds are beingcued by rainfall whichin Tayrona(S. Gill pers.comm.), which supports turn is positively correlated with the abun- the ideathat the microhabitat"cornering" ex- danceof their main foodsource (arthropods in hibitedby Rufous-and-whiteWrens in Tayrona the understory).Clearly, the strategyof Buff- may be linked to a higher density of Buff- breastedWrens fits well with a typical tracker breasted Wrens there. wherethe birds are monitoring closely changes Theabundance of arthropodsin theleaf litter in rainfall, food abundance, or both, and then did not changesignificantly between wet and start their reproduction when a minimum dry seasonsas did the abundanceof understory amountin any of these,or both is attained.Ad- arthropods.My dataagree with resultsof other ditionally,there is an advantagefor birds that studies (Wolda 1990, Poulin et al. 1992), show- breedearlier in the seasonbecause they have a ing that the litter offersa more constantenvi- higherprobability of escapingparasitism and ronmentfor arthropodscompared to the un- predation.Overall, that strategyseems to be derstory,although this is not alwaysthe case (T. workingwell for the speciesbecause there were Robinson pers. comm.). Therefore, Rufous- no differencesin overall nesting successbe- and-whiteWrens seem to be experiencinga tween contrastingyears of rainfall. muchmore constant food environment despite The picture is rather different for Rufous- year-to-year variations in rainfall. If food abun- and-whiteWrens. These wrens are shyand fur- danceis influencingtiming of breedingin Ru- tive birds that feed on arthropodsmostly on fous-and-white Wrens as it seems to do with the ground.My foragingbehavior data indicate Buff-breastedWrens, one would not expectto that this was the only microhabitatwhere Ru- seea trackingstrategy in thosebirds, because fous-and-whiteWrens had a capturerate of ar- their food environment seems to be less vari- thropodssimilar to that of Buff-breastedWrens ablethrough time. That is directlyreflected in in theirforaging microhabitats. Because of their the breedingphenology: they had a longerpe- foraginghabits, the territoriesof Rufous-and- riod of breeding(100 daysin 1993and 70 days white Wrensspanned a larger area and their in 1994) and started breeding earlier in both densitieswere lower in comparisonwith Buff- years compared to Buff-breasted Wrens. Al- breasted Wrens. though the first Rufous-and-whiteWren nest The evolutionaryprocesses that led to this was detected almost three weeks later in 1994 foraging specialization in Rufous-and-white than in 1993, there was no difference in the av- Wrensare not the topic of this paper,but my erage clutchinitiation date of the population data in conjunctionwith observationsof the betweenthese two years.The nestingsuccess same species in Costa Rica (Winnet-Murray of Rufous-and-white Wrens was similar to that 1986) suggestthat present-daycompetition of Buff-breasted Wrens in 1993, but no Rufous- with Buff-breastedWrens and other understory and-whiteWren nestswere predatedor para- insectivoresmight have an important effect. sitized in 1994. That can be attributed to the de- My observationsof the aggressiveinteractions lay in the rainy seasonwhich affected the betweenthe two speciessupport the idea that arrival of nest parasitesand predatorsto the Rufous-and-whiteWrens are being displaced area. Therefore,despite its conservativestrat- to foragecloser to the groundby Buff-breasted egy,the Rufous-and-whiteWrens might expe- Wrensand other understory birds. In all theag- riencea higheryearly variance in nestingsuc- gressiveencounters that I witnessed,Rufous- cess than Buff-breastedWrens through and-white Wrens were always attackedand year-to-year variation in nest parasitism or chasedaway by otherspecies, especially Buff- predation. January2001] Reproductionof two Thryothorus wrens 203

Becausemy data come only from two con- ferences in nesting successor numbers of trastingyears of rainfall, I cannotbe complete- broods between good and bad years. Rufous- ly surethat the specieswould showthe same and-whiteWrens did have a longerbreeding patternsdescribed here in other years with season,and the only reproductiveparameter similar conditions,or that this is a generalpat- that was less variable than in Buff-breasted tern in highly variableenvironments and other Wrens was their timing of reproduction.But speciesof birds. However,the differencesex- they did not havefewer broods and their nest- hibited by thesewrens in their reproductive ing successchanged from oneyear of my study timingbetween these two particularyears, and to the next. theway thiswas linked to the dynamicsof their The reasonfor this discrepancybetween my food resources,is highly suggestiveof an predictionsand what the wrensshowed lies in underlying pattern that deserves further the assumptionthat most of the reproductive investigation. parametersthat I examinedwere plastic.This provednot to be the case.The wrenswere vir- CONSTRAINTS IN THE EVOLUTION OF LIFE tually identicalin all the reproductiveparam- HISTORIES iN VARIABLE ENVIRONMENTS eters that I examined: clutch size, number of fledglingsproduced, incubation time, and nest- The resultsof this paper are importantfor ling time. This is somehowexpected because examination within the framework of the evo- thesetwo speciesare closelyrelated. However, lution of life historiesin tropicalbirds. What due to a differencein foragingbehavior, the are the reproductivestrategies that birds can food environmentperceived by eachspecies is adopt in an environmentin which the condi- different, and that has a direct effect on the tim- tionsfor successfulreproduction fluctuate from ing of breeding. Especiallyin tropical areas, one year to the next?What do my resultstell clutchsize and otherreproductive parameters about constraints in the evolution of bird re- vary little for many bird species(but seeYoung productive strategies in variable environ- 1994), presumablybecause of high predation ments? pressure(Skutch 1950, Kuleza 1990).Therefore, At the outsetof this study,I predictedthe re- it is possiblethat for many tropicalbirds, the productive strategiesof both wrens based charactersthat are proneto changeare mostly mostly on the existinglife-history theory in behavioral becausereproductive parameters variable environments (Cohen 1966, MacAr- might be under strongstabilizing selection or thur 1968, Schaffer 1974, Horn 1978, Ruben- they might not be as plastic(i.e. do not have stein 1982). I expectedBuff-breasted Wrens to enoughadditive genetic variance) as behavior- be very plasticin their reproductiveparame- al characteristicsmight. That might limit the ters, dependingon the amount of food avail- number of characters or traits available for the able;to reproducedisproportionately better in evolutionof a given reproductivestrategy of good years;and to have a shortreproductive tropicalbirds. Becausemy study was doneover period and a high numberof broodsper year a short-term,I couldnot gatherinformation on On the otherhand, because of theirlarger body otherlife-history characters of the wrenssuch size,I expectedRufous-and-white Wrens to be as age-specificsurvival, age at first reproduc- lessvariable in theirreproductive output, being tion, number of reproductive attemptsin a life- able to withstandbad yearsbetter than Buff- time, and dispersal.Clearly, long-term studies breastedWrens and being more conservativein on thelife historiesof tropicalbirds are needed their reproductiveoutput in good years (bet- to further clarify which charactersare more hedgers).I predicted that Rufous-and-white constrainedto changein responseto a variable Wrenswould concentrate their reproductive ef- environment.If thoseideas are right, I predict fortsin few broodsand would havea long re- that in cases where no other behavioral choice productiveperiod. is available,the timing of reproductionof trop- My results confirmed that Buff-breasted ical birds would be very constrainedtemporal- Wrenshad a shorterreproductive period than ly in a similarway exhibitedby manytemper- Rufous-and-white Wrens. However, I did not ate bird species. find any evidencefor plasticityin reproductive Previous studies of birds have shown that parameters(except onset of breeding), or dif- competitionbetween species can have an effect 204 JORGEA. AHUMADA [Auk, Vol. 118 on community structure (Pierpont 1986), and their time tossinglitter leavesaside looking for habitat utilization (Orians and Willson 1964, fleeing arthropods.The Clay-coloredThrush Greene1989). This study suggeststhat com- (Turdusgrayi) in Panamaalso spends a substan- petitioncan have an effecton the reproductive tial amountof time feedingon groundarthro- strategy of the speciesinvolved. To pursue pods and breeds in the dry season(Morton theseideas further, it is necessaryto have a 1971). thoroughunderstanding of the naturalhistory I proposethat insectivoresthat feed in sub- of a group of speciesby simultaneouslycol- stratessuch as leaf litter, deadwood, or species lecting informationon the foragingbehavior, that follow antsmight experienceless seasonal the dynamicsof their food sources,and their variationsin their food supply than insecti- reproductivebehavior A goodexample of such voresthat feed on arthropodsthat inhabitliv- a study is representedby the long-term re- ing plant material.That fact allows the first searchon the Galapagosfinches (Grant 1986). group of insectivoresto extendtheir breeding The different speciesof Galapagosfinches seasonand perhaps to breedat timeswhen pre- show no differences in clutch size, incubation dation, brood parasitism,or other external time,nesting time, or shapeof the nest.Except causes of nest failure are minimal. To test those for the CactusFinch (Geospiza scandens), which ideasfurther, it is necessaryto collectdetailed bred earlier, all speciessynchronized their breeding informationin a group of closelyre- breedingseason with the rainy seasonwhen- latedspecies (e.g. within a familyor genus)that everit arrived.Some individuals of this species feedon differenttypes of arthropodswith dif- could afford to breed earlier because of the ferent temporal dynamics.For example,the studyof Winnett-Murray(1986) on the behav- availabilityof pollen and nectarfrom cactias a ior of four speciesof wrens in Costa Rica food sourceduring the dry season.The Gala- showedthat House Wrens (Troglodytesaedon) pagosIslands are, however,an extremelysea- sonal environment and the finches do not have that lived in openhabitats experience a more constantfood supply and havea longerbreed- many choicesas to when to breed. We are in ing period than Gray-breastedWood Wrens need of similar data setsfor tropical birds in a (Henicorhinaleucophrys) and Rufous-and-white varietyof environmentswith differentdegrees Wrenswhich are forestspecies. However, Plain of seasonalityand predictability. Wrens(Thryothorus modestus), which also live in open habitats,showed a similar breedingpe- RELATIONSHIP BETWEEN TIMING OF riod comparedto the forestwrens. More com- REPRODUCTION AND FORAGING BEHAVIOR IN parative informationof that kind will allow us NEOTROPICAL INSECTIVOROUS BIRDS to untanglethe confoundedeffects of phylog- eny and behavioron the reproductivestrate- Is it possibleto make generalizationsfor pre- giesof Neotropicalinsectivorous birds. dictingthe reproductive phenology of a species by knowingits diet?The two wrensthat I stud- ACKNOWLEDGMENTS ied differ in the placesthey forageand, prob- I want to especiallythank the supervisorsof this ably becauseof its more constantfood environ- study,Henry Horn and Andy Dobson.Their uncon- ment, Rufous-and-whiteWrens had a longer ditional support and ideas were crucial for its suc- breeding seasonand started reproducingear- cessful development.Francisco Troncoso, Aracelly lier than Buff-breasted Wrens. Can these rela- Caselies, Liz Adriana Serrano, Maritza Jaramillo, Di- tionship be extendedto other neotropicalin- ana P. Molina, and Pablo Stevenson were of invalu- sectivorous birds? ablehelp in the field. Kathy Winnett-Murray,Frank Joyce,and an anonymousreviewer gave invaluable Plain Xenops(Xenops minutus) living in Tay- commentsto earlier versionsof the manuscript.I rona,started breeding well beforethe rainsar- also want to thank FONDO FEN-COLOMBIA for rivedin Februaryand March (J. Ahumada pers. their logisticsupport. This study was funded par- obs.). Those birds feed exclusivelyon insects tially by PrincetonUniversity in the United States and other arthropodsthat live inside dead and COLCIENCIAS in Colombia. branchesand hanging vines. White-bellied Antbirdsalso bred during the dry seasonand LITERATURE CITED well into the rainy season(August-September) AHUMADA, J. A. 1995. The effects of environmental (J. Ahumada pers. obs.).They spendmost of variation on the reproduction,ecology and be- January2001] Reproductionof two Thryothorus wrens 205

havior of two neotropicalwrens. Ph.D. disser- HORN, H. S. 1978. Optimal tacticsof reproduction tation, Princeton University, Princeton, New and life-history. Pages411-429 in Behavioural Jersey. Ecology: An Evolutionary Approach. (J. R. BEISSINGER,S. R., AND J.P. GIBBS.1993. Are variable Krebs and N. B. Davies, Eds.). Sinauer Associ- environments stochastic? A review of methods ates, Sunderland, Massachusetts. to quantifyenvironmental predictability. Pages JANZEN,D. H. 1973a.Sweep samplesof tropical fo- 133-146 in Adaptation in StochasticEnviron- liage insects:Description of study sites, with ments (J. Yoshimuraand C. W. Clark, Eds.). data on speciesabundances and size distribu- Springer-Verlag,Berlin. tions. Ecology54:659-686. BLAKE,J. G., AND B. A. LOISELLE.1992. Fruits in diets JANZEN,D. H. 1973b.Sweep samples of tropicalfo- of Neotropicalmigrant birds in CostaRica. Bio- liage insects: Effects of seasons,vegetation tropica 24:200-210. types, elevation, time of day, and insularity. CLIFF,A.D., ANDJ. K. ORD.1981. Spatial Processes: Ecology54:687-708. Models and Applications.Pion, London. JANZEN,D. H. 1980. Heterogeneityof potential food COHEN,D. 1966. Optimizing reproductionin a vari- abundancefor tropical small land birds. Pages able environment. American Naturalist 126:418- 545-552 in Migrant Birds in the Neotropics: 429. Ecology,Behavior, Distribution and Conserva- COLWELL,R. K. 1974. Predictability,constancy and tion (A. Keast and E. S. Morton, Eds.). Smith- contingency.Ecology 55:1148-1153. sonianInstitution Press,Washington, D.C. GRADWOHL,J., AND R. GREENBERG.1990. Temporada JOBLING,J. A. 1991. A Dictionary of ScientificBird de reproducci6nde tres pajaroshormigueros en Names.Oxford UniversityPress, Oxford. la Isla de BarroColorado. Pages 433-440 in Ecol- KINNAIRD,M. E 1992. Phenologyof floweringand ogfa de un BosqueTropical (E.G. J.Leigh, A. S. fruiting of an east African riverine forest eco- Rand, and D. M. Windsor, Eds.). SmithsonianIn- system. Biotropica 24:187-194. stitutionPress, Washington, D.C. KULEZA,G. 1990. An analysisof clutch size in New GRANT,P. R. 1986. Ecology and Evolution of Dar- World passerinebirds. Ibis 132:407-422. win's Finches. Princeton University Press, LEGENDRE,P. 1993. Spatial autocorrelation:Trouble Princeton, New Jersey. or new paradigm?Ecology 74:1659-1673. GREENE,E. 1989. Food resources,interspecific ag- LEVEY,D. J. 1988. Spatial and temporal variation in gression, and community organization in a Costa Rican fruit and fruit-eating bird abun- guild of insectivorousbirds. Ph.D. dissertation, dance.Ecological Monographs 58:251-269. PrincetonUniversity, Princeton, New Jersey. LEVINGS, S.C., AND D. M. WINDSOR. 1990. Fluctua- HEIDEMAN,P. D. 1989. Temporal and spatial varia- clonesde las poblacionesde artr6podosde ho- tion in the phenologyof floweringand fruiting in a tropical rainforest.Journal of Ecology77: jarasca.Pages 443-451 in Ecologiade un Bosque 1059-1079. Tropical (E.G. J. Leigh, A. S. Rand, and D. M. HERN,•NDEZ-CAMACHO,J., AND P. RODRiGUEZ-GUER- Windsor, Eds.). Smithsonian Institution Press, RERO.1972. Estudioeco16gico de la vegetaci6n Washington,D.C. del ParqueNacional Natural Tayrona.Divisi6n LOISELLE,B. A., ANDJ. G. BLAKE.1991. Temporal var- de Parques Nacionales y Vida, INDERENA, iation in birds and fruits along an elevational Colombia. gradientin CostaRica. Ecology 72:180-193. HERRMANN,R. 1970. Las causasde la sequfaclima- MACARTHUR,g. 1968.Selection for life tablesin pe- tica en la regi6n costanerade Santa Marta Co- riodic environments. American Naturalist 102: lombia. Revista de la Academia Colombiana de 381-383. CienciasExactas, Ffsicas y Naturales13:479-490. MARTIN,t. E. 1987.Food as a limiting on breeding HILTY, S. L. 1980. Relative abundance of north tem- birds:A life historyperspective. Annual Review perate zone breeding migrants in western Co- of Ecologyand Systematics18:453-487. lombiaand their impactat fruiting trees.Pages MILLER,A. H. 1963.Seasonal activity and ecologyof 265-271 in Migrant Birds in the Neotropics: the avifaunaof an Americanequatorial cloud Ecology,Behavior, Distribution and Conserva- forest. University of California Publicationsin tion (A. Keast and E. S. Morton, Eds.). Smith- Zoology 66:1-74. sonJanInstitution Press,Washington, D.C. MILLER, A. H., AND V. D. MILLER. 1968. The behav- HILTY, S. L., AND W. L. BROWN. 1986. A Guide to the ioral ecologyand breeding biology of the An- Birds of Colombia.Princeton University Press, deansparrow (Zonotrichia capensis). Caldasia 10: Princeton,New Jersey. 83-154. HOLMES, R. T., T. W. SHERRY, AND E W. STURGES. MOODY,D. T. 1970. A method for obtainingfood 1986. Bird communitydynamics in a temperate samplesfrom insectivorousbirds. Auk 87:579. deciduousforest: Long-term trends at Hubbard MORAN, P. A. P. 1950. Notes on continuous stochastic Brook.Ecological Monographs 56:201-220. phenomena.Biometrika 37:17-23. 206 JORGEA. AHUMADA [Auk, Vol. 118

MORTON,M. L. 1971. Nest predation affectingthe WOLDA,H. 1978a.Fluctuations in abundanceof trop- breeding seasonof the Clay-coloredRobin, a ical insects. American Naturalist 112:1017-1045. tropical songbird. Science171:920-921. WOLDA, H. 1978b. Seasonal fluctuations in rainfall, ORIANS,G. H., AND M. E WILLSON.1964. Interspe- food and abundanceof tropical insects.Journal cific territoriesof birds. Ecology45:735-745. of Animal Ecology47:369-381. PALADINO,F. V. 1989. Constraintsof bioenergeticson WOLDA,H. 1980.Seasonality of tropicalinsects. Jour- avian population dynamics.Physiological Zo- nal of Animal Ecology49:277-290. ology 62:410-428. WOLDA,H. 1990. Estacionalidadde los Hom6pteros PIERPONT,N. 1986.Interspecific aggression and the de la Isla de Barro Colorado.Pages 403-415 in ecologyof woodcreepers(Aves: Dendrocolapti- Ecologiade un BosqueTropical (E.G. J. Leigh, dae). Ph.D. dissertation,Princeton University, A. S. Rand, and D. M. Windsor,Eds.). Smithson- Princeton, New Jersey. ian InstitutionPress, Washington, D.C. POULIN,B., G. LEFEBVRE,AND R. MCNEIL. 1992.Trop- YOUNG,B. 1994. The effects of food, nest predation ical avian phenologyin relation to abundance and weatheron the timing of breedingin trop- and exploitationof food resources.Ecology 73: ical House Wrens. Condor 96:341-353. 2295-2309. Associate Editor: T. Martin RIDGELY,R., G. TUDOR,AND W. L. BROWN.1989. The Birdsof SouthAmerica, vol. 1. Universityof Tex- as Press, Austin. APPENDIX1. I calculatedthe spatialautocorrela- ROGERS, L. E., W. T. HINDS, AND R. L. BUSCHBOM. tion among arthropodsampling points within the 1976.A generalweight vs. length relationship plot using Moran'sindex (I) (Moran 1950): for insects.Annals of the EntomologicalSociety of America 69:387-389. N• • (x•-œ)(x;- œ) l-1 j 1 RUBENSTEIN,D. 1982. Risk, uncertaintyand evolu- I= tionary strategies.Pages 91-110 in Current Prob- lems in Sociobiology(C. King's College Socio- biology Group, Eds.). Cambridge University where N is the numberof data points in the lattice,f Press,Cambridge, United Kingdom. is themean for all datapoints, x, and x• are the values SAMPLE,B. E., R. J. COOPER,R. D. GREER,AND R. C. of two pointsthat are contiguous(at the appropriate WHITMORE. 1993. Estimation of insect biomass lag) and ZL, is the sum of the numberof links be- by length and width. AmericanMidland Natu- tween elementsthat are contiguousin the lattice ralist 129:234-240. (Cliff and Ord 1981). SCHAFFER,W. M. 1974. Optimal reproductiveeffort For regular, unweightedlattices, I behavessimi- in fluctuatingenviroments. American Naturalist larly to p (a correlationcoefficient) varying between 108:783-790. -1 and i (Cliff and Ord 1981). A value of -1 indi- SKUTCH,A. E 1950.The nestingseasons of Central catesnegative spatial autocorrelation (spiked surfac- American birds in relation to climate and food es), a value near 1 indicatespositive spatial autocor- availability. Ibis 92:185-222. relation (smooth surfaces), and a value near 0 SKUTCtt, A. E 1960. Life Histories of Central Ameri- indicatesno spatial autocorrelation(data points are can Birds, vol. 2. Pacific Coast Avifauna no. 34. independent from each other). With large sample SNOW,D. W., AND B. K. SNOW.1963. Breeding and sizes,I is distributednormally so it is easyto test its annual cycleof three Trinidad thrushes.Wilson departurefrom 0 statistically(ibid). The expected Bulletin 75:27-41. value of I for large samplesizes is:

SNOW,D. W., AND B. K. SNOW.1964. Breeding sea- 1 sonsand annual cyclesof Trinidad land-birds. E(I) - (N - 1) Zoologica49:1-39. VAN SCHAIK, C., J. W. TERBORGH,AND S. J. WRIGHT. and its expected variance is: 1993.The phenologyof tropicalforests: Adap- 4AN • - 8N(A + D) + 12A 2 tive significanceand consequencesfor primary E(I 2) = consumers.Annual Review of Ecologyand Sys- 4A2(N 2- 1) tematics 24:353-377. whereA = 1/2ZL, andD = 1/2ZL,(L,- 1) Ical- WALSBERG,G. 1983.Avian ecologicalenergetics. Pag- culatedI not only for the raw data but alsofor the es 161-220 in Avian Biology,vol. 7 (D. S. Farner difference between the numbers or biomass of ar- and J.R. King, Eds.).Academic Press, New York. thropodsin two monthsfor the threepairwise com- WINNETT-MURRAY,K. 1986. Variation in the behavior parisons:March-May, March-August, and May-Au- and food supply of four neotropical wrens. gust.This difference was further tested for departure Ph.D. dissertation, University of Florida, from 0 (Wilcoxonmatched pairs test) to seeif there Gainesville. was any changein numbersor biomassof arthro- January2001] Reproductionof two Thryothorus wrens 207

TABLEA1. Resultsfrom the spatialautocorrelation analysis for the arthropoddata collectedin 1994.Au- tocorrelationcoefficients (I) were calculatedfor the differencein eithernumber or biomassof arthropods betweentwo monthsin the understoryand the leaf litter and at two spatiallags: nearest neighbor (40 m) and secondnearest neighbor (80 m). The differencesbetween March and May and betweenMarch and August have a different expectedvalue becauseof the smallernumber of litter stationssampled in March (seethe Methodssection for details).

I (biomass) I (numbers) Microhabitat Difference between lag 40 m lag 80 m lag 40 m lag 80 m Understory March & May• 0.0380 - 0.0289 - 0.0585 - 0.0099 March & AugusP -0.0318 0.0086 -0.0133 0.0644* May & August• 0.0620 - 0.0249 0.0078 0.0301 Leaf-litter March & May2 - 0.0419 - 0.0092 0.0734 - 0.0704* May & August2 0.0099 - 0.0976* 0.0526 - 0.0305 March & AugusP 0.0443 0.0268 -0.0323 0.0458 • E(I) = -0.0057, g(I) = 0.0399, N = 175. 2 E(I) = -0.0074, g(I) = 0.0458, N = 135. * Significantat ct = 0.05.

pods from month to month. A program in C was semivariogramsfor all data setsto gain insight on written to perform the necessarycalculations. I was the degreeof spatialautocorrelation at largerspatial calculatedat two spatial lags:nearest neighbor (40 lags (Cliff and Ord 1981). m) and secondnearest neighbor (80 m). Significant There was no indicationof spatial autocorrelation departuresof I from 0 were checkedusing tablesof at the nearestneighbor and secondnearest neighbor the normal distribution. Additionally, I constructed in any of the three monthssampled neither for un-

APPENDIX2. Total number (first row in eachcell) and biomass(second row) of arthropodsfrom different taxafound in 525, 10-minunderstory counts. Data arebroken-up for eachof threesampling periods: March (dry season),May (beginningof the wet season),and August1994 (end of the wet season).Percentages of the total for eachcolumn are shownin parentheses.Biomass is expressedas milligramsof dry weight.

Taxon All monthscombined March n = 175 May n = 175 August n = 175 Araneae 2,667 (55.4) 857 (54.0) 1,082 (55.9) 728 (56.6) 35,736 (72.7) 14,527 (74.1) 16,909 (78.7) 4,299 (53.4) Coleoptera 752 (15.6) 243 (15.3) 303 (15.6) 206 (16.0) 2,325 (4.7) 619 (3.2) 1,043 (4.8) 663 (8.2) Orthoptera 424 (8.8) 149 (9.4) 191 (9.9) 84 (6.5) 3,401 (6.9) 1,375 (6.7) 1,055 (4.9) 971 (12.1) Hemiptera Homoptera 346 (7.2) 88 (5.5) 153 (7.9) 105 (8.2) 452 (0.9) 84 (0.4) 137 (0.6) 231 (2.9) Heteroptera 107 (2.2) 46 (2.9) 31 (1.6) 30 (2.3) 1,495 (3.0) 770 (3.9) 346 (1.6) 379 (4.7) Lepidoptera larvae 143 (3.0) 43 (2.7) 51 (2.6) 49 (3.8) 501 (1.0) 133 (0.7) 182 (0.8) 186 (2.4) adults 43 (0.9) 11 (0.7) 16 (0.8) 16 (1.2) 358 (0.7) 68 (0.4) 125 (0.6) 165 (2.0) Diptera 92 (1.9) 31 (2.0) 32 (1.7) 29 (2.3) 123 (0.2) 45 (0.2) 23 (0.1) 55 (0.6) Dictyoptera 43 (0.9) 19 (1.2) 17 (0.9) 7 (0.5) 1,324 (2.7) 652 (3.3) 463 (2.1) 209 (2.6) Phasmida 25 (0.5) 11 (0.7) 13 (0.7) 1 (0.1) 2,025 (4.1) 984 (5.0) 1,034 (4.8) 7 (0.1) Dermaptera 8 (0.2) 4 (0.3) 3 (0.2) 1 (0.1) 17 (0.4) 2 (0.0) 8 (0.1) 7 (0.1) Other 163 (3.4) 87 (5.5) 45 (2.3) 34 (2.6) 1,432 (2.9) 371 (1.9) 190 (0.9) 871 (10.8) Total 4,813 1,589 1,937 1,287 49,135 19,607 21,485 8,050 208 JORGEA. AHUMADA [Auk,Vol. 118 derstoryarthropods nor litter arthropods(Table A1) APPENDIX3. Resultsof pairwise comparisons(Wil- in numbersor biomass.Although the spatialauto- coxonmatched pairs test) betweennumbers (first correlationfor a few grids showeda significantde- row in each taxon) and biomass(second row) of parturefrom 0, the strengthof the signalis sosmall understory arthropodsfrom different taxa betweenthe three different sampling periods: March (<0.1) that it can be ignored.Semivariograms indi- (first dry season),May (beginningof rainy season), cated that this lack of spatial autocorrelationat the August (seconddry season).Comparisons were first and secondlags was consistentat larger spatial done betweenMarch and May and May and Au- lags. Therefore,comparisons between months were gust (lasttwo columns).Months labeled with "=" carried out assumingindependence of samples. within each row do not differ significantly.A monthlabeled with "+" or ..... hasa significantly higheror lowervalue compared with othermonths in the same row.

gu- Taxon March May gust ZMar_MayZMay_Aug Araneae = + = 4.12' 6.01' = = - 0.15 6.91' Coleoptera = = - 1.87 3.18' = = - 0.97 3.51' Orthoptera = = - 1.47 4.97* = = - 1.21 2.72* Hemiptera Homoptera = + = 3.52* 2.37* - = = 2.38* 0.46 Heteroptera + = = 2.36* 0.46 + = = 2.36* 0.17 Lepidoptera larvae = = = 1.03 0.48 = = = 1.93 0.12 adults = = = 0.64 0.16 = = = 0.63 0.20 Diptera = = = 0.05 0.16 = = = 0.31 0.00 Dictyoptera = = = 0.25 1.77 = = = 0.74 1.56 Phasmida = = - 0.39 2.82* = = - 0.22 2.94* Dermaptera = = = 0.36 0.80 = = = 0.50 0.00 Other + = = 2.28* 1.47 + = = 2.03* 1.76

* = Significantat the P < 0.05 level. January2001] Reproductionoftwo Thryothorus wrens 209

APPENDIX4. Totalnumber (first row in eachcell) and biomass(second row) of arthropodsfrom different taxafound in 485 leaf-littersamples. Data are shownfor eachof threesampling periods: March (dry season),May (beginningof thewet season), and August 1994 (end of thewet season). Percentages of the totalfor eachcolumn are shown in parentheses.Biomass is expressedas milligrams of dry weight.

All months Taxon combined March n = 135 May n = 175 August n = 175 Araneae 791 (32.7) 228 (31.5) 270 (29.9) 293 (36.9) 3,213 (27.8) 1,582 (46.7) 626 (13.1) 1,005 (29.6) Coleoptera 385(15.9) 112(15.5) 160(17.7) 113 (14.2) 1,141 (9.9) 240 (7.1) 437 (9.2) 464 (13.7) Pseudoscorpionida 470 (19.4) 103(14.2) 195(21.6) 172 (21.7) 432 (3.7) 224 (6.6) 94 (2.0) 115 (3.4) Orthoptera 226(9.3) 59 (8.1) 94 (10.4) 73 (9.2) 521 (4.5) 200 (5.9) 184 (3.9) 137 (4.0) Dictyoptera 88 (3.6) 39 (5.4) 34 (3.8) 15 (1.9) 838 (7.2) 201 (5.9) 509 (10.7) 128 (3.8) Lepidoptera adults 3 (0.1) i (0.1) 2 (0.2) 0 (0.0) 21 (0.2) 0.4 (0.0) 20 (0.4) 0 (0.0) larvae 27 (1.1) 11 (1.5) 11 (1.2) 5 (0.6) 14 (0.1) 0.7 (0.0) 13 (0.2) 0.3 (0.0) Protura 58 (2.4) 37 (5.1) 12 (1.3) 9 (1.1) 49 (0.4) 22 (0.6) 7 (0.1) 21 (0.6) Hemiptera Homoptera 16 (0.7) 13 (1.8) 1 (0.1) 2 (0.3) 17 (0.1) 11 (0.3) 6 (0.1) 0.8 (0.0) Heteroptera 99 (4.1) 30 (4.1) 34 (3.8) 35 (4.4) 207 (1.8) 92 (2.7) 49 (1.0) 66 (1.9) Diptera 21 (0.9) 2 (0.3) 14 (1.6) 5 (0.6) 6 (0.1) 1 (0.0) 1 (0.0) 4 (0.1) Scorpionida 13 (0.5) 3 (0.4) 6 (0.7) 4 (0.5) 816 (7.1) 170 (5.0) 360 (7.6) 287 (8.4) Thysanura 78 (3.2) 11 (1.5) 22 (2.4) 45 (5.7) 500 (4.3) 29 (0.9) 30 (0.6) 441 (13.0) Chilopoda 15 (0.6) 11 (1.5) 4 (0.4) 0 (0.0) 113 (1.0) 7 (0.2) 106 (2.2) 0 (0.0) Diplopoda 60 (2.5) 35 (4.8) 20 (2.2) 5 (0.6) 3,357 (29.0) 591 (17.4) 2,114 (44.6) 652 (19.2) Other 70 (2.9) 29 (4.0) 24 (2.7) 27 (2.1) 285 (2.5) 18 (0.5) 189 (4.0) 78 (2.3) Total 2,420 724 903 793 11,532 3,388 4,744 3,399 210 JORGEA. AHUMADA [Auk, Vol. 118

APPENDIX5. Resultsof pairwise comparisons(Wil- coxonmatched pairs test) betweennumbers (first row in eachtaxon) and biomass(second row) of litter arthropodsper stationof a giventaxon between the three differentsampling periods: March (first dry season),May (beginningof rainy season),Au- gust (seconddry season).Comparisons were done betweenMarch and May and May and August (last two columns). Months labeled with "=" within each row do not differ significantly.A monthlabeled with" +" or "-" hasa significantly higheror lowervalue compared with othermonths in the same row.

Au- Taxon MarchMay gustZMar_May ZMay_Aug Araneae = = = 0.71 0.88 = = = 0.28 0.58 Coleoptera = = - 1.61 2.48* = + = 2.83* 2.35* Pseudoscor .... 2.14' 0.61 pionida + = = 2.09* 1.02 Orthoptera = = = 0.50 0.93 = = = 0.01 1.70 Dictyoptera = = - 1.43 2.02* = = - 0.78 2.85* LepidopteraI = = = 0.16 1.25 = = = 0.92 1.88 Protura + = = 2.71' 0.59 + = = 2.35* 0.02 Hemiptera Homoptera + = = 2.53* 0.53 + = = 2.04* 0.00 Heteroptera = = = 1.24 0.75 = = = 0.53 0.47 Diptera = = = 1.00 1.12 = = = 0.00 0.84 Scorpionida = = = 0.63 0.56 = = = 0.84 0.50 Thysanura = = + 0.35 1.98' = = + 0.18 2.53* Chilopoda = = -- 1.53 1.82 = = = 0.50 1.82 Diplopoda = = - 1.61 2.42* = = - 0.56 2.11' Other = = = 0.75 0.75 = = = 0.15 1.15

* = Significantat the P < 0.05 level. • This includesonly larvae.There were not enoughadults to carry out the analysis.