Ecology and Evolution of Lek Mating Behavior in the Long-Tailed Hermit Hummingbird

ECOLOGY AND EVOLUTION OF LEK MATING BEHAVIOR IN THE LONG-TAILED HERMIT HUMMINGBIRD

F. GARY STILES Escuela de Biologla, Universidad de Costa Rica, Ciudad Universitaria, Costa Rica

AND

LARRY L. WOLF Departmentof Biology, SyracuseUniversity, Syracuse,New York 13210

ORNITHOLOGIC'•L MONOGRAPHS NO 27

PUBLISHED BY THE AMERICAN ORNITHOLOGISTS' UNION WASHINGTON, D.C. 1979 ECOLOGY AND EVOLUTION OF LEK MATING BEHAVIOR IN THE LONG-TAILED HERMIT HUMMINGBIRD ORNITHOLOGICAL MONOGRAPHS

This series,published by the AmericanOrnithologists' Union, has been estab- lishedfor major paperstoo long for inclusionin the Union's journal, The Auk. Publicationhas been made possiblethrough the generosityof Mrs. Carll Tucker and the Marcia Brady Tucker Foundation,Inc.

Correspondenceconcerning manuscripts for publicationin the seriesshould be addressedto the Editor, Dr. Mercedes S. Foster, Department of Zoology, University of California, Berkeley, California 94720.

Copies of OrnithologicalMonographs may be ordered from the Assistantto the Treasurer of the AOU, Glen E. Woolfenden,Department of Biology, Uni- versity of South Florida, Tampa, Florida 33620. (See price list on back and inside back cover.)

OrnithologicalMonographs, No. 27, viii + 78 pp. Editor of AOU Monographs,Mercedes S. Foster Editor of this number, John William Hardy

SpecialAssociate Editors of this issue,Mercedes S. Foster, Department of Zoology,University of California, Berkeley,and Barbara K. Snow, The Old Forge,Wingrave, Aylesbury, Bucks, England

Authors,F. Gary Stiles,Escuela de Biologia,Universidad de CostaRica, Ciudad Universitaria,Costa Rica, and Larry L. Wolf, Department of Biology, SyracuseUniversity, Syracuse, New York 13210

First received, 15 January 1976; accepted,9 September 1976; final revision completed, 15 March 1977

Issued September24, 1979

Price $8.50 prepaid ($7.50 to AOU members)

Library of CongressCatalogue Card Number 79-67267

Printed by the Allen Press,Inc., Lawrence, Kansas 66044

Copyright @ by American Ornithologists' Union, 1979 ECOLOGY AND EVOLUTION OF LEK MATING BEHAVIOR IN THE LONG-TAILED HERMIT HUMMINGBIRD

F. GARY STILES Escuela de Biologla, Universidadde Costa Rica, Ciudad Universitaria, Costa Rica

AND

LARRY L. WOLF Departmentof Biology, SyracuseUniversity, Syracuse,New York 13210

ORNITHOLOGICAL MONOGRAPHS NO. 27

PUBLISHED BY THE AMERICAN ORNITHOLOGISTS' UNION WASHINGTON, D.C. 1979

TABLE OF CONTENTS

INTRODUCTION ...... 1 MORPHOLOGY AND GENERAL BIOLOGY OF PHAETHORNIS SUPER½ILIOSUS .... 2

METHODS AND MATERIALS ...... 3 STVVY AREAS ...... 3 COLOR MARKING ...... 3 LEK OBSERVATIONS ...... 6 FORAGING DATA AND NECTAR SAMPLING ...... 6 COLLECTING AND OTHER OBSERVATIONS ...... 6

LEKS, TERRITORIES, AND LEK BEHAVIOR ...... 7 SEX OF LEK RESIDENTS ...... 7 CHARACTERISTICS OF LEK SITES ...... 7 SONG PERCHES AND TERRITORIES ...... 10 VOCALIZATIONS ...... 12 VISUAL DISPLAYS ...... 15 MATINGS, TRUE AND FALSE ...... 18

FORAGING PATTERNS ...... 21 FOOD PLANTS ...... 21 INTERSPECIES RELATIONS ...... 25 STRATEGIES OF FLOWER EXPLOITATION ...... 28 NECTAR VS. INSECTS ...... 30

TEMPORAL PATTERNING OF LEK ACTIVITY ...... 32 BREEDING AND LEKKING SEASONS ...... 32 DAILY PATTERN OF LEK ACTIVITY ...... 35

LEK DYNAMICS ...... 43 SEASONAL VARIATION IN LEK COMPOSITION ...... 43 SURVIVORSHIP AND RECRUITMENT OF LEK RESIDENTS ...... 49 INTERMALE RELATIONS ON THE LEK ...... 51 INTER- AND INTRALEK MOVEMENTS ...... 56

DISCUSSION ...... 58 LEK EVOLUTION IN PHAETHORNIS SUPERCILIOSUS ...... 58 COMPARISONS WITH OTHER HERMIT SPECIES ...... 64 EVOLUTION OF LEK BEHAVIOR IN HUMMINGBIRDS ...... 69

ACKNOWLEDGMENTS ...... 71

SUMMARY ...... 71

LITERATURE CITED ...... 74 LIST OF FIGURES

Figure 1. Map of La Selva property ...... 4 2. Monthly variation in temperature and rainfall at Finca La Selva ...... 5 3. History of spatial relations of three subleks in lek I from 1969 to 1972 ...... 8 4. Typical song perch ...... 9 5. Relation between territory size and distanceof clear visibility from central song perches ...... 11 6. Location of four song types on lek II in 1971 and 1972 ...... 14 7. Visual displaysof P. superciliosusgiven around songperches in lek territories .... 16 8. Number of speciesof food plants of P. superciliosusin good bloom in different months, according to use categories...... 24 9. Daily nectar production of eight plant speciesfrequently visited by P. superciliosus ...... 26 10. Cumulative use of food plants by P. superciliosusin relation to corolla morphology and use by other hummingbirds ...... 27 11. Records of distances at which individual male P. superciliosus foraged from their territories ...... 31 12. Estimates of the breeding seasonof P. superciliosus...... 33 13. Annual variation in calling activity among residentmales on lek Ic ...... 34 14. Daily variation in calling activity on leks II, Ia, and Ib for one day early and late in the 1970 lek season ...... 36 15. Variation in daily calling activity of five males resident on lek Ic on 29 March, 1971 ...... 37 16. Variation in daily calling by male RY on lek Ic, end of 1971 lek seasonand be- ginning of 1972 season...... 38 17. Frequencyof chaseactivity on areasobserved during all-day watchesat lek Ic .... 39 18. Seasonalvariation in daily calling activity from one territory on lek Ic ...... 41 19. Presenceand territorial occupancyof male P. superciliosuson lek I, February 1971 to April 1973 ...... 42 20. Locationsof lek Ia, "El Surfi," and lek Ic, "Surprise,"showing vegetation types, major landmarks, and distribution of major food plants ...... 43 21. Locations of territories of males on leks Ia and Ic in March, May, June, July, 1971 ...... 44 22. Locations of territories of males on leks Ia and Ic in December 1971 and Jan- uary-March, 1972 ...... 46 23. Locationsof territories of males on leks Ia and Ic in April, May, and August 1972 47 24. Locations of territories of males on leks Ia and Ic in January and March, 1973 48 25. Proportions of P. superciliosusmarked at one site that were recorded at other sites different distances away ...... 56 26. Hypothesizedrelationship of dominanceand/or fitnessto lek positionin leks of grouse and Ruff vs. leks of P. superciliosus...... 62

vi LIST OF TABLES

Table I. Mensural characteristicsof Phaethornis superciliosus...... 2 2. Numbers of territorial male Phaethornissuperciliosus on each of four leks during February-March in five seasons...... 8 3. Observed temporal relationships of displays in 85 behavioral interactions in Phaethornis superciliosuslek territories ...... 19 4. Morphology, period of good bloom, and hummingbird visitation of food plants of Phaethornis superciliosus ...... 22 5. Feeding records of Phaethornis superciliosusat La Selva ...... 23 6. Dominance relations of Phaethornis superciliosusat several food plants ...... 28 7. Dominance relations of Phaethornis superciliosusand other hermits at different food plants ...... 29 8. Observedflower visits by lek males and other individualsalong a 450-m transect between lek I and the field station ...... 32 9. Daily activity patternsof Phaethornissuperci#osus in different habitats...... 32 I0. Phaethornis superciliosusmist-netted at lek I, 1971-1973 ...... 40 I1. Survivorship of resident male Phaethornis superciliosuswithin and between lekking seasons: lek I, 1971-1972 ...... 50 12. Age structureof Phaethornissuperciliosus on lek I in 1971-1973 lekking seasons50 13. Subsequentstatus of young male Phaethornissuperciliosus first appearing on lek I in 1971 and 1972 ...... 52 14. Shifts in lek territories of resident male Phaethornis superciliosusin relation to previous status ...... 53 15. Morphological and behavioral parameters of male Phaethornissuperciliosus in relation to lek position ...... 55

vii INTRODUCTION

In a lek matingsystem, several males gather in a localizedarea in the breeding season,where they display to other males and to females, the latter displays sometimesleading to matings. Male courtshipassemblies occur in a variety of birds (e.g., Lack 1968), many frogs and toads (e.g., Emlen 1976), somemam- mals(Buechner 1961; Leuthold 1966; Bradbury, pers. comm.), and certaininsects (Wynne-Edwards 1962; Alexander 1975). In anurans and some insects, the concentrationsof malescompeting for matingprivileges occur where femalesare predictablylocalized in spacedue to ovipositionrequirements. Although these aggregationsmay have a complexsocial structure (Duellman 1967; Parker 1970; Campanellaand Wolf 1974; Emlen 1976), their evolutionaryhistory is distinct from that of certain birds and mammals,in which the male aggregationsare generatedwithin the social systemitself. The actual site of bird and mammal communaldisplay grounds, or leks, is traditionalwithin the population. Only a few of many potentialsites are used over long periods. In most of the species, at least the initial stagesof mating occur solely or principally on the leks. It generallyis assumedthat a numberof malesdisplaying together is more attractive or stimulatingto femalesthan would be the sum of an equal number of males displayingseparately (Robel 1967); thus a male could potentiallyfather more offspringby joiningan assemblythan by displayingor waitingfor femalessolitarily. In lek speciesan individual male, to maximize his chancesof mating, must be availablewhenever a receptivefemale arrives on the lek. The evolutionof leks thus probablyis limited to speciesin which malescan devotemost of their nonmaintenancetime and energyto advertisingand defendingareas and courting femalesas they arrive. This practicallyrequires the partial or total emancipation of the male from the reproductiveeffort beyondthe inseminationof the female (D. Snow 1962, 1963). The strong selectivepressure on males for effective aggressiveand courtshipcharacteristics, together with the emancipationof the male from reproduction,have producedexaggerated sexual dimorphism in many lek species(Sibley 1957; Wynne-Edwards1962). Hummingbirdsand at least some speciesof grouse, shorebirds,manakins, cotingas,and birds-of-paradisemeet the prerequisitesfor the evolution of leks (D. Snow 1963). Lek socialsystems apparently are not uncommonin hummingbirds, but seem particularlycharacteristic of the genusPhaethornis, the hermit hummingbirds.This is somewhatanomalous in that the speciesof this genusare rather dull colored,mostly without marked sexual dimorphism. Published accounts of Phaethornisdeal with general descriptionsof lek behavior (Nicholson 1931; Davis 1934, 1958; Skutch 1951, 1964a; Arp 1957; B. Snow 1974) and song typeswithin and betweenleks (D. Snow 1968; Wiley 1971). The patternof resourceexploitation of the populationrarely has been integrated into any analysisof lek behavior (but see Kruijt et al. 1972; Campanellaand Wolf 1974; Lill 1974, 1976; Pitelka et al. 1974). Hummingbirdsare knownto be highlydependent upon nectarresources for breeding(review in Stiles1973), andcompetition for nectarin hummingbirdsis well documented(e.g., Cody 1968; Wolf 1969, 1970; Stilesand Wolf 1970), yet thesefactors have not been con- sideredin discussionsof the evolutionof lek behaviorin hummingbirds.In this ORNITHOLOGICAL MONOGRAPHS NO. 27

TABLE 1 MENSURAL CHARACTERISTICS OF PHAETHORNIS SUPERCILIOSUS

Sample Sex N ,• SD Range

Weight (g) Known lek residents 6` 146 6.08 0.25 5.5-6.7 Collected birds x 6` 32 6.13 0.25 5.6-6.7 9 29 5.87 0.23 5.4-6.3 Bill length-•exposed culmen (mm) Known lek residents 6' 146 37.49 1.18 34.0--40.0 Collected birds 6' 32 37.68 1.01 35.5-39.5 9 29 36.10 1.23 33.0-38.5 Wing length--flattened wing (ram) Known lek residents 143 59.96 1.46 55.0-64.0 Collected birds 32 60.07 1.46 57.0-63.5 29 57.22 1.44 54.0-60.0 Tail leng•---central rectrices (mm) Known lek residents 139 66.82 2.11 61.5-74.0 Collected birds 31 66.61 2.38 62.0-71.5 27 66.54 2.26 62.0-70.5

Sexed by dissection. paperwe describein detailthe lek behaviorof the Long-tailedHermit (Phaethornis superciliosus)in Costa Rica and showhow its socialsystem is intimatelyrelated to patternsof flower exploitation;these are in turn stronglyaffected by inter- specificcompetition for nectar.

MORPHOLOGY AND GENERAL BIOLOGY OF PHAETHORNIS SUPERCILIOSUS Phaethornissuperciliosus is a medium-sizedhummingbird (6 - g) with a long, decurvedbill (37 mm) and generallydull-colored plumage. The underpartsare grayishbrown anteriorlyshading to buff on the abdomen. The dorsumvaries from dull bronzygreen to dull purplishbronze, the feathersof the lower back and rump havingbroad buffy edgings.The face is conspicuouslypatterned with a bully postocularstripe, a whitish-buffmalar stripe,and a buff stripedown the center of the throat; these contrast with the dark brown of the rest of the face and crown. The maxilla is black, the mandible mostly dull pinkish orange. The mouth lining is orange,and, togetherwith the facial stripes,produces a striking patternwhen the bill is open. The outer rectricesare grayishbrown basally,and brownishblack medially,with narrow buffy tips. The two central rectricesare greatlyelongated and conspicuouslywhite tipped. Plumagesof the sexesare identical;however, most individualscan be sexed by measurements(Table 1). The plumageof immaturesdiffers from that of adults chiefly in having much more conspicuousbuffy feather edgingson the back, crown, and tertials;no marked changein plumagecolor occurswith age. In addition,in the horny sheathof the maxilla immatureshave corrugationsthat graduallydisappear over a period of months (Ortiz-Crespo1972), making it possibleto determinethe approximateage in monthsfor manyyoung birds. (See Stilesand Wolf 1974 for detailsof the agingprocedure.) P. superciliosuslacks LONG-TAILED HERMIT HUMMINGBIRDS 3 a postjuvenalmolt; youngbirds first molt when aboutone year old. Adults have a singlecomplete annual molt, the timingof whichvaries greatly among individuals but is quite constantfrom year to year in the sameindividual (Stiles and Wolf 1974). The nestingof P. superciliosushas been studiedin greatestdetail by Skutch (1964a), and in generalappears similar to that of mostother hermits(cf. Ruschi 1950; Skutch 1951; D. Snow and B. Snow 1973; B. Snow 1974). The female buildsthe nest,incubates, and rearsthe youngby herself. Presumedmales appear very rarely at nest sitesand apparentlyplay no constructiverole in nesting(cf. B. Snow1974). Nestsare placedon the undersideof the tips of palm or cyclanth leaves,or stripsof Heliconialeaves; the nest has been described by Skutch(1964a). Nestsare foundmost often alongtrails or streams,usually well away from leks (Skutch 1964a; B. Snow 1974; presentstudy). Nestingsuccess probably is low, perhaps15-25%, as is the casein mostother hermits studied to date (Skutch 1966; B. Snow 1974). Phaethornissuperciliosus inhabits the humidtropical and lowersubtropical zones from southernMexico througheastern Bolivia and AmazonianBrazil (Meyer de Schauensee1966). Scatteredreports in the literature and our own observations indicatethat it is a lek speciesthroughout this extensiverange (Nicholson1931; B. Snow 1973b; Wolf and Stiles,pers. obs.).

METHODS AND MATERIALS

STUDY AREAS Most of this studywas done at Finca La Selva,in the Sarapiqullowlands of northeasternCosta Rica (10 ø 26'N, 84 ø 01'W; elevations 50-150 m). About 75% of the roughly 700 ha of the La Selva property is primary tropical wet forest, with sizeableareas of old cacao plantationsand secondgrowth on the northernand easternedges of the property (Fig. 1). Much of the land north, east,and westof La Selvahas been clearedwithin the last 15 years,but the forest to the southis still continuousup to the volcanoesof the CordilleraCentral. The forest at La Selva is describedin detail by Holdridge et al. (1971); the secondgrowth and light-gapvegetations are discussedby Stiles (1975) with particular referenceto Heliconia. The avifauna of La Selva has been studied in general terms by Slud (1960). Rainfall at La Selva averaged4,150 mm per year between 1960 and 1969. A rather poorly defined dry seasonextends from about late January through April in most years; monthsof heaviestrainfall usually are June-July and/or November-December.Temperature shows little seasonalvariation with monthly averagesof 22.8-24.9øC and a maximumrange of 14ø-37øC (Fig. 2). A few supplementaryobservations were made near Rinc6n, Peninsulade Osa, in southwestern Costa Rica. This is another lowland wet forest area of similar climateand vegetation(described by Holdridgeet al. 1971).

COLOR MARKING

Nearly all of the observationsreported here were made on color-marked individuals.Birds were capturedwith mist nets at leks or feedingareas. The 4 ORNITHOLOGICALMONOGRAPHS NO.27

Fig.1. Map ofthe La Selva property. Leksites are indicated byRoman numerals (I-IV); diagonallyaregiven byhatchedhorizontal areas dashes.are feeding Abbreviations sites; nest are,locations streams: areQSugiven = by Quebrada triangles; E1swampySurfi; QSareas = Arboleda;QuebradamajorE1Salto; feeding QSt areas:= Quebrada P = Point; E1Saltito;E • East QSbBoundary.= Quebrada CacaoS•balo;and second QA =growthQuebrada are shownby heavy stipple; pastures arelight stipple. backfeathers weremarked withspots ofmodel airplane paint, and aplastic tag withstripes ofcolored plastic tapewas affixed tothe leg. Each individual was givenaunique colorcombination. Furtherdetails ofthe marking method are givenbyStiles andWolf (1973). Thepaint waslost ateach annual molt,and tagsfellor broke offfrom time totime, sowe attempted torecapture individuals oneor more times each year toreplace theirmarkings. Attwo leks we netted atapproximately monthlyintervals during1971-73 tofollow indetail theprogress ofthe annual moltin lek males (Stiles andWolf 1974). Over the course ofthis LONG-TAILED HERMIT HUMMINGBIRDS 5

A 32- 1969-1971

• 20*"---.,.,'/" '""' -- --

I I I I I I I I I I I I B 1962-1971 80(

•400

20O

I I I I I I I I I I I I ItJ F M A MttJtt J J A S O N D MONTHS

Fig. 2. Monthly variation in temperatureand rainfall at Finca La Selva. A. Monthly mean maximum, mean, and mean minimum temperatures,reading from top to bottom. B. Monthly mean rainfall, plus and minusone standarddeviation. study, we marked over 500 individuals. From motion pictures and observations we detectedno adversebehavioral effects of our marking system. Displays of markedand unmarkedbirds were indistinguishable,and the presenceof markings on one individualdid not appearto affect the behaviorof individualsinteracting with it. In a few individuals,mostly near the beginningof the study, perching ability and possiblysurvival were impairedby an improperlyattached tag. Data from theseindividuals were excludedfrom the analyses. 6 ORNITHOLOGICAL MONOGRAPHS NO. 27

LEK OBSERVATIONS

Observations on behavior of lek birds were made with 7x and 10x binoculars at 3-20 m rangesby an observerusually seated and partly concealedby the under- growth. We found that after a short period of habituationthe birds ignoredthe observer,provided he remainednearly motionless.Detailed reports of behavior were made on the spot,often recordedby a secondperson while the first watched. We also made 16 mm motion picturesof behavioralinteractions that provided documentationof our field descriptionsand wereused to draw Figure 7. We made periodiccensuses of the leks during this study, attemptingto locate all frequentlyused song perchesof each male and to map the extent of each territory. We also made a/l-day singingrecords at approximatelymonthly inter- vals for a year at one lek and lessfrequently at two other leks. For thesedata the observerrecorded singing or absenceof singingby each individualof a group of 5-7 adjacentresidents (or for the entire lek when this comprised5 or fewer birds), duringeach minute of the day. All chasesoriginating in or passingthrough the sectorof the lek also were recorded. These data provide a quantitativeindex of lek activity throughthe day for different seasons.

FORAGING DATA AND NECTAR SAMPLING

We recordedforaging at flowersby Phaethornissuperciliosus in our regular censuswalks throughvarious forest and second-growthhabitats and wheneverwe walked certain trails passingthrough several habitats. Each visit of a bird to the flowersof a plant or clonewas treatedas a separatefeeding record. Thesedata give a representativepicture of flower use throughthe year by the P. superciliosus populationat La Selva. We carriedout regularnetting and observationprograms at certain major foragingareas during periodsof bloomingof important food plants,to documentuse of thesesites by lek residents. We mademonthly censuses of the floweringof the food plantsof P. superciliosus from 1971 through1974. The averageperiod of "goodbloom" (= 50% or better of maximum observedflowering during this period) was used to assessoverall flower availabilityto P. superciliosus(for furtherdetails see Stiles 1977, 1978a). We samplednectar of severalimportant food plants of P. superciliosuswith fine capillarytubes at regular intervalsthrough the day. Flowers were bagged to prevent removal of nectar by birds or insects. Data on nectar production reported here were from days with clear mornings,to minimize variations in productionrates due to weather.

COLLECTING AND OTHER OBSERVATIONS During 1971 and 1972, we maderegular trips to a localityabout 1.5 km from La Selva acrossthe Rio PuertoViejo. We collected63 specimensof P. superciliosusto documentgonad cycles and to obtain measurementsof individualsof known sex. Measurementsof malescollected from this populationwere virtually identical to those of known lek males at La Selva; hence the measurementsso obtainedshould apply equallywell to both sexesof the La Selvapopulation. In addition, nine actively calling lek residentswere collectedfrom nine different LONG-TAILED HERMIT HUMMINGBIRDS 7 territorieson a lek of P. superciliosusnear Rinc6n de Osa at about0800 on 8 August 1969. Thesebirds were collectedwithin a 10-min period to document the sex of lek residents. Specimenscollected during this studyare depositedin the AmericanMuseum of Natural History, the WesternFoundation of VertebrateZoology, and the Universidad de Costa Rica.

LEKS, TERRITORIES, AND LEK BEHAVIOR

SEX OF LEK RESIDENTS In the absenceof discernibleplumage dimorphism in P. superciliosus,sex of live birdswas determinedby externalmeasurement. Males averagedheavier and had longer bills and wings than did females (Table 1). Tail length was not sexuallydimorphic. Although there is overlapin all measurements,a combination of two or three enabledus to determinethe sex of most individualswith a high degreeof certainty. Measurementsof 146 marked individualsknown to be lek residentsagree precisely with measurementsof known malescollected at a nearby locality (Table 1). Eight known lek residentsat La Selva sexedby dissection were all males. Of the 146 lek residentsmeasured (Table 1), 125 (86%) were unequivocallymales on the basisof measurements.Only 5 (3%) would probably have been called females as all of their measurements fell near the low end of the observedrange for males;one of thesewas sexedas a male by dissection. Furthermore, nine lek residentscollected on a lek near Rinc6n de Osa were all found to be males upon dissection.Thus the evidenceoverwhelmingly indicates that in P. superciliosusall lek territory holders are males (see also Nicholson 1931). This also is the casein other speciesof Phaethornis(Arp 1957; B. Snow 1974, pers. obs.).

CHARACTERISTICS OF LEK SITES

We found four leks of P. superciliosusduring this study. They were scattered over the La Selvaproperty, the averagedistance between them being about 1 km (Fig. 1). We feel sure this is a completeinventory of the leks on the northern two-thirdsof the property,which we searchedintensively in February, March, and April 1972. However, we stronglysuspect that a fifth lek existsnear the southwesterncorner of the property. The densityof leks in the La Selva forest is thus about one per km2, a surprisinglylow figure becauseP. superciliosusis one of the commonesthummingbirds of La Selva. When discoveredin June 1969, lek I was comprisedof two semi-isolated"subleks" about 150 m apart. The interveningarea, which containedcacao plantations and a 1.5-ha bananapatch that had been clearedin March 1968 (R. Chavarr•a, pers.comm.), was not suitablelek habitatin 1969. Prior to the clearingof the banana patch in 1968, the two subleksprobably were much closertogether, if not contiguous(Fig. 3). The larger"El Sur•" sublek(Ia) waslocated along the QuebradaE1 Sur•; in June1969 it wascomprised of 8 birds,and graduallydeclined thereafter (Table 2). The smaller "Banana Patch" sector (lek Ib) contained three birds during most of 1969 and 1970, with two birds being added along the edge of the banana patch toward lek Ia by July 1970. By February 1971 8 ORNITHOLOGICAL MONOGRAPHS NO. 27

o •o •o

.3facIt 197t •farch 19•2

Fig. 3. Historyof spatialrelations of threesubleks in lek I from 1969to 1972. From two ratherisolated sections in 1969the lek hasgradually changed to onecontinuous group of birds asintervening areas undergo secondary plant succession. Vertical hatching = forest;horizontal hatching= cacao;stipple = densethickets, the heaviness of thestipple corresponding to the density of the vegetation. the thirdor "Surprise"sector (Ic) hadbeen established about midway between thefirst two subleks.Lek lb concurrentlydeclined to a singlebird, which shifted to lekIc, leavinglb empty.Sectors Ia andIc graduallymerged into a singlelarge lekby mid-1972 (Fig. 3) asundergrowth grew up in theintervening area. LeksII (discovered1969) andIII (discovered1972) bothlie alongthe Que-

TABLE 2 NUMBERSOF TERRITORIALMALE PH.SETHORNI$$UPERCILIOSU$ ON EACH OF FOUR LEKS DURING FEBRUARY--MARCHIN FIVE SEASONS

1969 1970 1971 1972 1973 Lek numberIa 9 8 3 3 - Ib 3 5 (1) - - Ic - - 13 15 18 Total lek I 12 13 16 18 18 II 14x 19 23 19 22 III NT • NT NT 8• 16 IV NT NT NT 6 5

Incomplete census. NT = not taken. LONG-TAILED HFRMIT HUMMINGBIRDS 9

Fig. 4. A typical song perch (male GR on lek Ic). Note the open space around the song perch itself which is a slender, fairly horizontal twig or vine. The back paint and the tip of the tag are also visible on this marked male. See Figure 21 for the locations of this male in relation to the rest of the lek. brada El Salto, about I km apart (Fig. I ). The former contains20-25 birds, the latter about 15 birds (Table 2). Lek IV is small (Table 2) and is located on the westboundary of La Selva approximatcly600 m from lek I (Fig. 1). Three of the four leks are locatedat least partly along streams,which represent the largest and longest-livedgaps in the forest canopy at La Selva (Stiles 1975). This is importantfor three reasons: ( 1) Where light reachesthe forest floor dense thicketscan grow. (2) Streambanksprovide habitat for severalimportant food plantsof P. superciliosus.(3) The streamsare convenientflight pathsfor residents and visitors. All four leks are locatedin densethickets (Fig. 4), along a man- made edge (lek IV), a stream (leks [I and III), or a combinationof these (lek I). The rise of lek [c probablyrcsulted from invasionof the banana patchby second growthand the growthof densethickets in the adjacentforest and cacao (Fig. 3). The thicketsin which lek IV is locatedalmost certainly date from the clearing of the adjoiningtract of land and the resultant penetrationof the sun inside the forestalong the edge. The mostlikely site for thislek prior to 1971 wasjust west of La Selva, possiblyalong the Quebrada El Surfi. The two leksin undisturbedhabitat, I[ and 11I, are situatednear openswamps 10 ORNITHOLOGICAL MONOGRAPHS NO. 27

alongthe QuebradaEl Salto. Heliconiapogonantha, the most importantfood plant of P. superciliosus,is commonin suchswamps; this suggeststhat another importantrequisite of a lek siteis proximityof goodforaging areas. The situation of lek I appearsto supportthis inference,although habitat disturbance is a com- plicatingfactor. A large openswamp with muchHeliconia is some350-400 m from the lek up the QuebradaArbolera; the banks of both this creek and the QuebradaE1 Surfi support good stands of severalspecies of Heliconia(Stiles 1975). Moreover,lek I has increasedin size in recentyears as thick secondgrowth con- tainingmuch H. pogonanthahas grownup in the bananapatch. The locationof lek IV is probablyatypical. However, H. pogonanthawas commonalong the Quebrada E1 Sur• west of La Selva in 1968, and this area, now cleared, may then have been an importantforaging area for the birds of lek IV.

SONG PERCHES AND TERRITORIES

The foci of aggressiveand courtshipactivities on a lek are the songperches of the residentmales (Fig. 4). A male on the lek spentmost of his time on his songperches, singing, displaying, preening, or sittingquietly. Thesesong perches variedfrom 1 to 7 m aboveground, but mostwere 1.5-2 m. A male often used a high, relativelyexposed perch duringthe dim light of dawn and then moved to lower, more shadedperches as the sun rose. The percheswere typicallybare, slendertwigs or vinesof 1 cm diameteror less,horizontal or nearly so. Usually the percheswere locatedin a small clear space0.5-1 m in diameter,amidst thick vegetation,allowing the residenta goodview of his immediatevicinity and provid- ing room for the variousdisplays that involveflying near the perch. Males often usedthe sameperches for weeksor months.Moreover, they would nearly always singfrom preciselythe samespot on the twigs,and usuallyfacing the samedirection. Other males (and females7) hovered around these perchesas they flew through the lek, evenwhen the residentwas absent. It is probablethat most or all of the regularsong perches of all other residentsare known to each male on a lek or lek sector. The territoryof a lek maletypically consisted of two to five songperches and the area aroundand betweenthem. There was no fixed boundary;the airspace of a territory was defendedmuch less consistentlythan the songperches themselves. Someexpansion and contractionof territoriesoccurred on a day-to-day basis,reflecting the dynamicsof a male'sinteractions with his neighbors.These changesin areasalways involved the takingup or abandoningof songperches on the peripheryof the territory; males did not defend areas that containedno songperches. Other males sometimes were permitted to fly throughthe territory by the resident,but neverto singor to use a songperch unless the residentwas absent.At times,a neighboringresident chased an intruderfrom a songperch in the absenceof the residentof the territoryin question. Territory sizevaried from 20 m2 to almost500 m2 and was closelycorrelated with the opennessof the vegetationas perceivedfrom the songperches (Fig. 5). The smallestterritories were in the very densethickets adjoining breaks in the forest canopy such as streams,treefalls, open swamps,and edges. Territories tendedto be closertogether in suchsituations, leading to a considerablyhigher LONG-TAILED HERMIT HUMMINGBIRDS 11

500'

r = .875 400' p(,001

200- 100'

oo ß o o io is VISIBILITYOF SONGPERCHES Fig. 5. Relationbetween territory size and distanceof clear visibilityfrom centralsong perches.Visibility distance is the averagemaximum distance at whichthe paint and tag mark- ingsof the bird on its perchwere visibleto an observerfrom severalangles. densityof individuals.The largestterritories were in relativelyopen forest under- storyor cacao,usually towards lek peripheries.Openness of the vegetationinflu- encesthe abilityof a residentto detectintruders and defendhis songperches• and may set the minimumspacing between territories. Two adjacentresidents on their songperches were rarely in visualcontact. The few exceptionsusually involvedthe temporaryterritorial shifts mentioned above and were accompanied by increasesin aggressiveinteractions, notably chases. Constant visual contact amongseveral residents may increaseaggressive activity on the lek. This could have at leasttwo unfavorableconsequences: (1) a high degreeof interference with matingattempts, and (2) an extra energydrain on the males. Lek territories of P. superciliosusseldom contained flowers that were used as nectar sourcesby territory owners. Nearly all individualswe observeddid all of their feeding away from the lek. A few males had suitable flowers on their territoriesthat they visitedbut did not defendconsistently against other hermits. In at leasttwo cases,flowers on a male'sterritory were used mostly by hummingbirdsof otherspecies. One lek male abandonedpart of his territoryas a clump of Heliconiaimbricata came into bloomthere and was occupiedby a territorial male Thalurania furcata. The Thalurania chased the Phaethornis whenever the latter attemptedto use a songperch located40 cm from an H. imbricatainflor- escence. 12 ORNITHOLOGICAL MONOGRAPHS NO. 27

Phaethornissuperciliosus also eat smallinsects and spiders,obtained by foliage gleaning.Some males gleaned occasionally in their territories, but never frequently or regularly. It seemsevident that territorieswere not beingdefended for their food value,but were solelydisplay grounds that the malesleft to obtain virtually all their food. In all essentialsthe functional characteristicsof P. superciliosuster- ritoriesagree with thoseof the 'classical'lek territoriesreported for a varietyof birds and other animals.

VOCALIZATIONS

We recordeda variety of vocalizationsin P. superciliosus;most were heard chieflyor only on the lek. We madeno attemptto studyvocalizations of juveniles or femalesat the nest. Hencethe followingcatalogue of vocalizationsof P. superciliosusprobably is not complete. Flight call.•The flight call is a squeaky,whistled sweet or tseet. This wasthe vocalizationmost commonly heard away from the lek. It was givenby a flying bird in a variety of circumstances:upon arriving at flowers,upon sightinga conspecific,upon seeingthe observeror a predator,or sometimeswith no obvious externalstimulus. The flight call appearsto announcea bird's presence,and it probablyis analogousto the chip notesof certainNorth Americanhummingbirds in beinga generalcontact and spacingnote (Stiles 1971b). Aggressivesqueals and chatter.•Theseconsist of oneto four high,thin squeals, endingin a rapid ascendingor descendingchatter. Each successivesqueal is lowerin pitchthan the precedingone; sometimes the chatteris omitted,or it may be followedby one or two ascendingsqueals. This call usuallywas givenby a male perchedeither on his territoryor by a flower clump,upon the approachof anotherhermit. We also once heard a female at her nest give this or a very similarcall whenanother hummingbird flew by. The squealand chatterappear to serveas a warningor threat, signifyingthe readinessof the callingbird to attack the intruder. Should the latter not leave, a chase would often follow. Sometimessqueals and/or chatterswere givenduring the chase,apparently by the attackingbird (presumablythe residentduring a chaseon the lek). Chasenotes.•These are a varietyof staccatopip or chip notesgiven during a chase. When the chaseoccurred on the lek and involved two P. superciliosus, it was impossibleto tell whetherchase notes were beinggiven by the chaser, the bird chased,or both. However,when a P. superciliosuswas chasedby another hummingbird(e.g., from flowersdefended by the latter), it often gave chase notes,suggesting that in lek encountersthe bird chasedusually gives these notes. Lonebirds flying rapidly through the lek, not involvedin any chase,gave similar notes.The functionof the chasenotes remains unclear at present,and they do not seemto be strictly confinedto chases. The precedingvocalizations were heard both in and away from leks, and probablywere givenby both sexes.Those that followwere heardmainly on or nearleks, and appearedto be givenchiefly or only by males.The songwas the mostcharacteristic lek vocalization,but severalothers, including squeals, chatters, and chasenotes, were given in aggressiveencounters. Song.--The songof this speciesconsists of a singlephrase, repeated monot- LONG-TAILED HERMIT HUMMINGBIRDS 13

onouslyfor periodsof up to 30 min. Singingoccupied most of a male'stime on territoryand was his major meansof announcinghis presence.A male often spentup to 50% of the daylighthours singing during the earlypart of the lekking season. A malesinging on his territory in the absenceof conspecificsusually gave 60-70 songsper min. When an intruderapproached, this increasedto 90-110 songs per min, and the songsbecame louder and shriller.Neighboring residents often countersang,alternating their songswith nearlyperfect regularity. The singing rateof eachsuch male usually dropped by abouthalf, so that the numberof songs per min by the two malescombined was only slightlygreater than the number givenby a male singingalone. Nonresidentmales attempting to invadeor gain controlof a lek territory also sang,but at least initially their songswere soft and very irregularin tone and rhythm,particularly when the invaderwas a youngbird. If the nonresidentwent unchallengedfor any lengthof time,his songsbecame louder, more uniform, and regular.Thus the rate,volume, uniformity, and rhythmof the songsprovided a fairly goodindex of the stimulusinput, aggressivestate, and territorialstatus of the singingmale. We found a numberof distinctsong types in the P. superciliosuspopulation of La Selva. Someidea of the distinctnessof thesedialects can be gainedfrom some of our transliterations: chink, churk, kaching, chrrik, shree, beeyurr, and kisink. The numberof songtypes represented on a lek was relatedto lek size. Lek II, with 20 or moreresidents in mostyears (Table 1), had 4 distinctsong types. Lek Ia (5-8 residents)and Ib (3) had 1 songtype each,while Ic, formed in part fromthe amalgamation of these2 leksand containing approximately 15 birds, had2. In 1973lek III had 16 birdsand 3 songtypes, lek IV 5 and 1, respectively. Within a lek the individualssinging a particularsong type usuallyconstituted a geographicallydefined subunit of the lek. Thesesubunits maintain their integrity from year to year in spiteof a considerableturnover of adult males (Fig. 6). On large leks with severalsubunits, members of one subunitsometimes could not hear membersof another,although intervening subunits normally connected the two. Similargroupings of songtypes occur within leks of P. guy and P. 1onguemareus (D. Snow 1968; Wiley 1971; B. Snow 1974). Further detailsregarding song dialectsin P. superciliosusare presentedelsewhere (Wolf and StilesMS). For the present,suffice it to say that individualsappear to learn their songsin their first year and singthe samesong thereafter. Chip-churr.•This call consistsof one or severalstaccato notes, often on different pitches,followed by a nasal, rolling churr or chirr. Sometimesthe chip notes were omitted,and severalchurrs were given consecutively.The chip notes are quite similarto chasenotes; however, the churr was never heard in a chase. The chip-churrusually was given by a maleon his songperch in a varietyof situations, e.g.,when a conspecificor a chasecame through the territory,when a potentially dangerousanimal (squirrel,weasel, human being, etc.) was near, or preceding flightin the apparentabsence of otherhummingbirds or animals.The chip-churr appearsto be a generalizedpreflight note signifying the readinessof the calling bird to takeflight, for whateverreason. When a singingmale was about to leave 14 ORNITHOLOGICAL MONOGRAPHS NO. 27

'lOrn' LONG-TAILED HERMIT HUMMINGBIRDS 15

histerritory in the absenceof any intruderor challenger,he oftenmixed chip- churrswith his songsand frequentlystretched his wingsand/or tail and fluffed hisfeathers. The chip-churrappears to lackthe aggressivecharacter of the other preflightnotes, the squealand chatter. Bill-pop.--A dry, snappingsound resembling a sharp,explosive kiss is called the bill-pop.It is producedsimultaneously with a suddensnapping open of the bill (the gapedisplay, described below). This soundwas given by residentmales in aggressiveencounters at song perches, but was occasionally given by an intruding maleas well. The socialsignificance of this call is discussedmore fully in con- nectionwith the gape display.

VISUAL DISPLAYS

In the absenceof an intruder,a singingmale sat uprighton the perchwith the wingtipsheld belowthe tail (Fig. 7A). Usuallythe head was movedthrough a small horizontalarc while the tail was waggedthrough a small vertical arc (less than 30ø). Motion picturesindicate that neitherhead nor tail movementswere timed to coincidewith singing,and that one or both couldbe absent. The throat featherswere raisedslightly giving the throat a raggedappearance and possibly accentuatingthe patternof light and dark stripes. Whenan intruderflew into or pasta territory,the residentbegan singing louder and faster. Simultaneouslythe tail was spreadwith the two long central rectrices separatedfor most of their length,and was waggedvigorously through a vertical arc of up to 90ø. At the apexof a wag,the tail formednearly a right anglewith the body. The throatfeathers were very noticeablyfluffed, and the crownfeathers were often raised,accentuating the facial pattern (Fig. 7B). The followingdisplays (Fig. 7C-7I) occurredchiefly at the songperches of the residentmales. Certain displaysor componentswere given by lone birds, probablyyoung males, at some distancefrom the lek. In the float display,the displayingbird flies slowly back and forth over a distanceof 7-15 cm in front of a perchedbird. The flying bird movesparallel to the perchat the samelevel as, and facingthe sittingbird. As it traversesthis path, the bird pivots its body throughan angle of approximately75 ø to either side of the vertical. The bill is held at an angle well above the horizontal and is usuallywide open,presenting the orangemouth lining and lower mandibleas well as the throat patternto the perchingbird. The latter usuallypresents its bright orangegape to the hoveringbird. The perchedbird "tracks" the movementsof the hoveringbird, alwayskeeping its bill pointed at the latter. This displaywas given most often by the residentmale, but also was given by the intruder, particularly when it was another resident of the same lek. Gapeand bill-pop(Fig. 7D).--This displaywas givenmost often by the ter- ritoryresident but alsoby the intruder,especially if it belongedto the samelek.

Fig. 6. Locationof four songtypes on lek II in 1971(above) and 1972 (below). Each songtype (a-d) is indicatedby a differentsymbolic outline of the territoryof each resident (line,heavy dash, etc.). Despitethe turnoverof someresident individuals between years, the song type locations remained constant. 16 ORNITHOLOGICAL MONOGRAPHS NO. 27

• ,,,'/ \•,•'

Fig. 7. Visual displaysof P. superciliosusgiven around songperches in lek territories. A, postureof residentmale singingin the absenceof an intruder--thenormal posture; B, resident male singingwith vigoroustail waggingas an intruderapproaches; C, float; D, gape and bill- pop; E, F, perchexchange sequence (F is a continuationof E as the flying bird settlestoward the back of the perchedbird); G, side-by-side;H, copulation.See text for descriptionsof the displays. LONG-TAILED HERMIT HUMMINGBIRDS 17

The displayingbird hovers 15-20 cm from and facing the perchedbird. The bird closesits bill and raisesit to almostvertical as it startsa short,rapid dart in a shallowarc towardsthe perchedbird. At the top of the arc the bird snaps its bill open and downward,flashing the bright orangegape and simultaneously producingthe poppingsound described earlier. The rami of the mandibleappear to be spreadslightly as the bill is opened,but much less conspicuouslythan in the correspondingrock displayof P. guy (B. Snow 1974). As the bird reaches the closestpoint of the arc to the perchedbird (ca. 10 cm away), the head is lowered and the bill dosed. Severalof these displaysoften were given in rapid succession,the displayerbacking down and away to regain its original starting position. The gape and bill-pop was seen most often in aggressiveencounters and frequentlypreceded or followed a perch exchange. Perch exchange(Fig. 7E-F).-•The perch exchangeis usually initiated by the residentand is often given reciprocallyby residentand intruder. After starting in a manner similar to a gape and bill-pop, the hoveringbird circlesover the sittingbird to approachit from above and behind. The circlingbird has its bill open and pointed toward the perchedbird; the throat feathersusually are fluffed. The circlingbird settlesdown as thoughto land on the back of the perchedbird, which flies from the perch, and the circlingbird lands in its place. The formerly perchedbird is now hoveringin front of the perch,the positionsof the two birds havingbeen reversed. This perchexchange sequence was often repeatedas many as 10 timesin rapid succession;between perch exchanges the hoveringbird might gape and bill-pop. Side-by-side(Fig. 7G).-•The hoveringbird alights12-15 cm from the perching bird, usuallyfacing the oppositedirection, or both birds land togetherfollowing a hover-up. On the perch, each bird moves its head through a horizontal arc of approximately60 ø toward and away from the other. The bills are wide open and held at an angle of 30 ø or so above the horizontal. The facial feathers are fluffed accentuatingthe pattern of dark and light stripes. The tails are usually fully spreadand waggedvigorously through a vertical arc of 60-90 ø, and often from sideto. side as well. The wingsare vibratedor flutteredbelow the body, and the back feathersmay be raised. The high intensityside-by-side display appears in both aggressiveand sexual contentsand is seenless often than the preceding displays. Hover-up.-•Resident and intruder hover 15-25 cm apart, facing each other, as they rise togetherfor a meter or more (dependingpartly on the configuration of the vegetation). The birds hold their bodiesvertically; their bills are horizontal and usually closed. The postureresembles that of the bird that has just bill- poppedin Figure 7D; in fact, one bird, usually (always?) the resident,may gape and bill-pop during the hover-up. Immediately following this display the resident usuallydarts slightlydownward toward the intruder,chasing it away and pursuing it. Occasionallythe birds hover up and down several times, sometimesfinally perchingin a side-by-side.The hover-upappears to be a high intensityaggressive interaction without sexual significance. Chase.-•Althoughnot a display,chases are consideredhere becausethey •re by far the commonestconclusion to any behavioralinteraction between resident andintruder. The intruderflies rapidly from the territory,pursued by the resident 18 ORNITHOLOGICAL MONOGRAPHS NO. 27 for a few meters or, more often, for some distance. The chasingbirds often flew back and forth around the lek, traversingthe territoriesof other residentswho sometimesjoined in. Three- and four-bird chasesusually originated as two-bird encounters.Very rarely did more than one intruder appear simultaneouslyin the territory of a resident. It was often difficult to determine how a chase was initiated. When the chase startedfrom a hover-upor whenthe intruderhovered while the residentperched, the resident'sattack definitelystarted the chase. In many casesthe two birds appearedto start off simultaneously,but perhapsthere were subtlebehavioral signalsthat we were unable to identify. In addition to chasenotes, we often heard a few song notes during a chase, particularlywhen both participantswere lek residents.These songsseemed to occurmost frequently when the chasepassed through the territoryof one resident or the other. Probably,a residentsang as the chasepassed through his territory, but we were able to verify this only in a few caseswhere the two residentssang different dialects. It was usually impossibleto identify the singer, or even to ascertainwhether the same bird was consistentlythe chaser. Phaethornissuperciliosus appears to exhibitno clearly defined chain or sequenceof displaysalthough certain ones are sequentiallyassociated (Table 3). For instance, a perchexchange tends to be followedby anotherperch exchange or by gapeand bill-popby the resident,depending upon which bird initiatedthe perchexchange. The courseof an interactionoften was determinedlargely by whetherthe intruder hoveredbefore the residentor perchedupon enteringthe territory. This in turn may have dependedon the former's sex, age, and territorial status. Intruders that were residentsof the samelek seemedmuch more likely to perform displays suchas the float or gapeand bill-popupon enteringthe territory. We did not definitelyobserve females performing these displays. The differencein positionof certaindisplays in a sequence,depending on whetherthey were givenby the intruderor the resident,may reflect changesin relativeaggressiveness of the two birds. The aggressivenessof an intrudingmale who waschallenging or testingthe residentwas highest at the start. During the succeedinginteraction the aggressivenessof the intruderdeclined, culminating in a chase. The intrudershould be least subordinateto the residentduring the initial challenge,and mostsubordinate when he is chased.In a few casesthe intruder, at the initial challenge,dominated the resident,who eventuallylost control of part or all of his territory. This occurredmost often when the residentwas a youngmale that had just gaineda territoryor was still not in full controlof one. Usuallythe dominantintruders were neighboringmales (either adultsor older, establishedyoung males), who then added the song perch (es) to theirown territories.

MATINGS, TRUE AND FALSE In copulationone P. superciliosusmounts the otherby alightingon its back followinga gapeand bill-pop,a float, or side-by-side(Fig. 7H). The mounted bird leansforward, its head held low and its back hunched,often flutteringits wingsbelow the body. The mountingbird graspsthe backfeathers of the other LONG-TAILED HERMIT HUMMINGBIRDS 19

I I I I I I I I I •'• I

I q"• I I cq I I I I I 20 ORNITHOLOGICAL MONOGRAPHS NO. 27 in its feet and vibratesits wings,presumably for balance.Both birdstwist their tails to facilitatecloacal contact, and both vibratetheir tails. On one occasion when the mountedbird was a female apparentcloacal contact lasted 3-5 sec; 2 other copulationsbetween a male and an unsexedbird were of similar duration. In two homosexualmountings cloacal contactwas much briefer if it occurred, andonly the superiorbird twistedits tail. The mountingand copulation positions in P. superciliosusare similarto thosein other hummingbirdspecies (B. Snow 1974; Wolf 1975a; Stilesand Ortiz-CrespoMS). We rarelysaw P. superciliosusmate. We witnessedonly six probablecopulations andtwo knownhomosexual mountings in severalhundred hours of watchingat lek territories.The copulationswere immediately preceded by a float (twice),a perch exchangewith the residentgaping and bill-popping(twice), and a side-by-side (twice). With one exceptionwe have not seena femaleperform any displayto signalher sexto the resident.The exceptionoccurred during a side-by-sidewhen the intrudershuffled along the perchtowards the residentmale. The latter then performeda float as the former continuedto fluff its feathersand wag its tail as in the side-by-side;the male mountedthis bird and copulationfollowed. Similarly,we sawno specialmale precopulatory display distinct from the aggressivedisplays occurring in male-maleencounters. At least someof thesedisplays, e.g., the perch exchange,are performed by the female as well. We never saw a femalegive the float or gape and bill-pop. The generalcopulation sequence thus differedrather little from that of manymale-male aggressive encounters until the actual mounting. We unfortunatelywere unable to film a copulationsequence, and thus may have missedsubtle behavioral differences from a male-male encounter. However,the mountingsfollowing perch exchanges seemed to occursimply because the sittingbird remainedon the perch and allowedthe circlingbird to land on its back; the latter then immediatelyattempted copulation. A female'smajor means of signallingher sex and receptivity may be just to stay perched during such a sequence;were she to fly, she probably would be chased. Conversely,a male P. superciliosusmay treat as a female, and attempt to copulatewith, any bird that does not fly as he circlesbehind it. The two homosexualmountings support this interpretation. As we marked our first P. superciliosus,we inadvertentlypulled out most of its tail. When released, this bird (YY) flew off approximately6 m and preenedand fluffed its plumage. Almost immediatelyanother P. superciliosusfloated in front of YY, who seemed to ignoreit. The secondbird circledover and behindYY and landedon its back in what appearedto be an attemptedcopulation, at which point YY flew off. YY turned out to be a residentmale on the lek (Ia) with a territory only 5 m from the site of the attemptedcopulation. However, becauseYY did not fly duringwhat appearedto be a perchexchange, he was treatedas a femaleby the other bird. The secondhomosexual mounting involved a markedresident of lek Ic (ROY) and a marked bird that turned out to be a nonresidentmale (BOP). BOP failed to fly during a perch exchange,and ROY alightedon its back and attempted copulation. After a secondor two, BOP shookloose and flew off alone. Males frequentlyattempted copulations with dead leavesabout hermit size and suspendedby a spiderweb. This behavior was most often shownby adult males LONG-TAILED HERMIT HUMMINGBIRDS 21 on their territories,but we occasionallysaw intruderscalling from a territory in the resident'sabsence engage in leaf copulations.We alsosaw this behaviorby severalindividuals well away from any lek, onceby a young,nonresident male. In a typicalleaf copulationthe maleflies slowly toward the leaf, whichis usually 2-3 m from the perchand somewhatabove it. He then gapesand bill-pops severaltimes at the leaf, andmay float beforehovering in a verticalposition with his feet graspingthe leaf and his abdomenin contactwith it. At contact,the abdomenand tail are vibratedin a manner similar to true copulation. Following the leaf copulation,the male usuallyflies backward, gapes and bill-popsagain, and fliesoff; occasionallyhe repeatsthe performanceup to five timesbefore leaving. Checks of several leaves revealed no trace of fluid. Similar behavior, called "leaf mating"or "falsemating," has been reported in severalother species of hermit hummingbirds(Arp 1957; B. Snow 1973a, b, 1974; pers.obs.). All true matings and homosexualmountings that we saw occurred on leks, usually within the territory of the male. Four copulationsequences and one homosexualmounting occurred at song perchesof the mountingmale. In one instance,the presumedfemale flew into the territoryand perchedabout 3 m from the songperch of the male, who approachedand copulatedwith her there. In the one copulationthat occurredoutside of a male's territory, the initial part of the sequenceoccurred within the territory;it wasfollowed by a chaseof approxi- mately20 m, at whichpoint the femaleperched, and the final displaysand mating took place. The one homosexualmounting occurring outside a territory, that involvingYY (see above), may have involveda neighboringmale that left his territoryon seeingYY perched.These observations tend to confirmthe importance of lek territoriesas matingstations. All matingsoccurred between 0730 and 0930 althoughour observationson lek territoriescovered the entire day. The small numberof matingsobserved could resultfrom three causes: (1) the numberof matingsis, in fact, low; (2) many matingsoccur away from the lek; and (3) the densevegetation blocked our view of some matings. Probably all three factors contribute.We presentevidence later that duringthe main lekking seasononly aboutone femalevisits a 15-malelek each day to mate. Thus, an observerat a randomlychosen territory would have, at best, a 6.7% chanceof seeingmating duringa day'sobservations (assuming he couldsee all partsof the territory,which is not always the case). One mating away from a male's territory began as a chasefrom within the territory;this may be advantageousto the male by reducing interferencefrom otherlek residentsduring mating. Given the rarity with which femalesappear on the lek, a malemay stand a betterchance of matingif he chases a potentialfemale than if he letsher go and waitsfor anotherto appear. Similarly, nonmatingchases should have little effect on male presencewhen the next receptivefemale arrives,a very rare event.

FORAGING PATTERNS

FOOD PLANTS

We have recordsof P. superciliosusforaging at the flowersof 21 speciesof plants(Table 4). Doubtlessother speciesare visitedoccasionally, but thesedata shouldgive a fairly completepicture of flower use by the La Selva population. 22 ORNITHOLOGICAL MONOGRAPHS NO. 27

TABLE 4 MORPHOLOGY•PERIOD OF GOOD BLOOMs AND HUMMINGBIRD VISITATION OF FOOD PLANTS OF PHAETHORNIS $UPERCILIOSUS

Proportion of total observed visits by

Mean effective Daily corolla Months Non- nectar length Corolla in good territorial Territorial production Food plant (mm) x shapeg blooms Hermits nonhermits nonhermits (•zl/fiower)

Heliconia pogonantha (Musaceae) 33 C Dec-May* 0.69 0.10 0.21 115 H. wagneriana (Musaceae) 48 C Feb-Apr 0.98 0.02 - 90 H. mariae (Musaceae) 22 SC Feb-Jun* 0.30 0.51 0.19 45 H. imbricata (Musaceae) 21 SC Jun-Sep 0.15 0.26 0.59 85 H. latispatha (Musaceae) 32 S Jun-Sep 0.09 0.32 0.59 75 H-3' (Musaceae) 41 C Aug-Oct* 0.94 0.06 - 70 H-18 • (Musaceae) 44 C Jul-Sep 0.99 0.01 - 65 H-16 • (Musaceae) 40 SC Jul-Aug 1.00 - - 70 H-17' (Musaceae) 28 S Aug-Oct 0.26 0.29 0.45 110 Calathea Iutea (Marantaceae) 33 C Feb-May 1.00 - - 20? Costus ruber (Zingiberaceae) 35 C May-Jul 0.97 0.03 - 85 C. malortieana (Zingiberaceae) 40 C Aug-Nov 0.96 0.04 - 75 Aechmea magdaIenae (Bromeliaceae) 26 SC Jul-Aug 0.89 0.11 - -- Passifiora vitifolia (Passifloraceae) 30 (S) Feb-Mar 1.00 - - 85+ Gurania Ievyana (Curcurbitaceae) 17 S Feb-Mar 0.11 0.69 0.20 -- MaIvaviscus arborea (Malvaceae) 38 S Feb-Apr, Oct* 0.87 0.13 - 80 Pentagonia donnelIsmithii (Rubiaceae) 34 S Mar-Jun 0.92 0.08 - 30 Jacobinia umbrosa (Acanthaceae) 28 C Feb-Mar 0.57 0.40 0.03 70 A pheIandrastorkii (Acanthaceae) 36 C Aug-Oct 0.91 0.09 - 55 CoIumnea nicaraguensis ( Gesneriaceae) 25 SC Apr-Jun, Oct-Nov 0.08 0.49 0.43 -- AIlopIectus coriaceus (Gesneriaceae) 25 SC Sep-Nov 0.24 0.57 0.19 --

x Minimum distance from the mouth of the corolla tube to the nectar chamber. g C ---- curved, SC ---- slightly curved, S ---- straight. For Passi]lora the tube of the corolla is broad and essen- tially straight, but arrangement of the stamens effectively prevents any but a long-billed hummer from entering the flower. s * ----flower that blooms year round. • Taxon numbers of presently unclescribedHeliconia species, as used in a revision of the genus (Daniels and Stiles 1979; see also Stiles 1975). [See note p. 78.--Ed.] LONG-TAILED HERMIT HUMMINGBIRDS 23

TABLE 5 FEEDING RECORDS OF PHAETHORNIS $UPERCILIOSUS AT LA SELVA x

Food plant J F M A M J J A S O N D Total Proportion Heliconia pogonantha 25 40 39 24 16 16 35 12 7 6 17 18 265 0.27 H. wagneriana 2 16 28 20 2 68 0.07 H. mariae 1 - I 4 4 2 I 13 0.01 H. imbricata 2 21 31 21 9 2 86 0.09 H. latispatha I - 5 4 2 - 12 0.01 H-3 6 7 5 3 4 5 16 13 12 7 3 4 85 0.09 H-18 1 29 24 13 7 74 0.08 H-16 1 4 II 2 18 0.02 H-17 4 7 6 17 0.02 Calathea lutea - I I 2 4 0.004 Costus ruber 3 12 14 15 7 I 2 54 0.06 C. malortieana 6 1 2 2 - 4 12 8 I0 13 II 4 74 0.08 Aechmea magdalenae 2 5 2 5 14 0.01 Passiflora vitiJolia I 9 8 4 I 3 27 0.04 Gurania levyana - 2 4 2 1 9 0.01 Pentagonia donnell- smithii 2 2 4 6 4 3 I - 2 1 2 I 28 0.03 Malvaviscus arborea 4 5 7 6 6 1 2 1 3 2 2 3 42 0.04 Jacobinia umbrosa 5 11 13 6 2 39 0.04 Aphelandra storkii 2 10 12 4 1 - 29 0.03 Columnea nicaraguensis 1 1 2 3 1 - 8 0.01 Alloplectus coriaceus I 1 1 2 - I - 5 0.005 No. feeding observations 52 95 116 93 53 76 151 III 89 53 38 44 971

No. food plants visited 10 11 13 14 9 12 14 13 15 I1 8 7

x Obtained during census walks and general observations, 1969-1973; not including observations at specific flowers.

Of the 21 species,14 are visitedprimarily by hermithummingbirds, which in most casescomprise more than 90% of the visitors(Table 4). Feedingrecords of P. superciliosusat these21 plant speciesare givenin Table 5. These data can be used to divide the plant speciesinto four use categories,as follows: a speciesthat contributes(1) lessthan 10% of foragingrecords in all months,and lessthan 2.5% overall; (2) at least 10% of foragingrecords in 1 or more months,and/or 2.5% overall; (3) at least 20'% of foragingrecords in 1 or more monthsand/or 5% overall; (4) at least 40% of foragingrecords in 1 or more months and/or 10% overall. Heliconiapogonantha is the only speciesin category4 and was by far the most importantfood plant of P. superciliosusat La Selva. It had the highestnectar productionof any heavilyused flower (Table 4) and was the most frequently visitedfor 8 months. It alsohad the longestperiod of good bloom of any major foodplant (Decemberto May or Junein mostyears), duelargely to its production of large,long-lived inflorescences (Stiles 1975 ). Use category3 includesfour other speciesof Heliconia and two of Costus. Heliconia H-3 was of secondaryimportance in any one month, but bloomed and was visitedyear round. The other specieshad short,well-defined periods of good 24 ORNITHOLOGICAL MONOGRAPHS NO. 27

USE CATEGORIES

4 3 2 1 I 2

RELATIVE FLOWER4 7 8 15 16 13 11 10 15 19 17 9 ABUNDANCE

Fig. 8. Number of speciesof food plants of P. superciliosusin good bloom in different months, according to use categories(see text). The relative flower abundance is obtained by multiplying the number of each use category by the number of plant speciesin good bloom in that category (e.g., for February: 4X lq-3 X 1 q-2 X 3 q- 1 X 2= 15). bloomduring which they were visitedfrequently, and long periodsof little or no bloomingor visitation (Tables 4, 5). H. wagnerianawas important during the dry season,H. imbricata,H-18, and Costusruber during the early wet season, and C. malortieanaduring the mid- to late wet season.Thus, all of the major food plants of P. superciliosusat La Selva were large herbaceousmonocots in only two genera. Category2 includesspecies that are visited regularlybut not frequently. These plants generally are uncommon or very locally distributed and tend to have relativelyfew flowersper plant and/or little nectarper flower. H-17 is a nectar- rich speciesoften visited and defendedby nonhermits,which reduced hermit usage. The flowersin category1 were visited only occasionallyby most P. superciliosus,although certain individualsvisited these speciesregularly. In contrast to categories3 and 4, categories1 and 2 containa considerabletaxonomic diversity with representativesof 9 families, including 3 of monocotsand 6 of dicots. Malvaviscuswas the only commonspecies in category1 or 2 with a long blooming season. Some individualscould be found in bloom every month, but the population rarely floweredsynchronously. The other minor food plantshad relatively short,well-defined flowering seasons, either in the dry season(6 species)or the early to mid-wet season(6 species)or both (Columneanicaraguensis). Heliconia mariae is known from only two clumps at La Selva; it bloomed year round, but with a strongpeak in the late dry and early wet seasons(Stiles 1975, 1978a). The usecategories can be usedto constructa semiquantitativeweighting system that allows us to estimate the overall flower abundance and how this changes seasonally(Fig. 8). There were definitepeaks of flower availabilityin the mid- dry season(February throughMarch or April) and the early wet season(July throughSeptember). A shortbut very pronouncedlean seasonoccurred between late October and early December in most years. The only food plant of P. superciliosusin category3 or 4 in good bloom during this period was Costus malortieana,which was locally common in second-growthand overgrowncacao LONG-TAILED HERMIT HUMMINGBIRDS 25 in the northernquarter of the propertybut very rare elsewhere.Thus, the P. superciliosuspopulations of the southernpart of the propertyeither had to move into the northernpart or subsistmostly upon insects.We have severalrecords of changesof feedingareas by individualsat thisseason. For example,BWY (a femaleby measurements)was consistentlyrecaptured or observedduring the monthsDecember-August of 1971 and 1972 only alongthe eastboundary where Heliconiapogonantha was abundant;but during September-November,this bird was caughtor observedonly in large Costuspatches north and west of lek I. SimilarlyGBP, a residentmale on lek II, was recapturedin the Costuspatches in November 1972; other feeding recordsfor this bird are for the east boundary. The periodof reducednectar availability ended in early to mid-Decemberwhen Heliconia pogonanthaattained good bloom. However, the lengthsof the lean seasonand the bloomingseasons of most food plants can vary by a month or more in different years (Stiles 1978a). Data on the daily patternof nectarproduction are availablefor all the Heliconia species(Fig. 9A, Stiles 1975) and severalother food plants (Fig. 9B). Nectar productionwas highestin the early morningand declinedto low valuesby early afternoon.Nectar harvestedby hermitsin the early morningwas replaced,but by middaynectar was a nonrenewableresource, at leastuntil the next day. Nectar harvestedin the afternoonconsisted mainly of that not taken duringthe morning. Flowersof all thesespecies lasted only one day (Stiles 1975, pers. obs.) neces- sitatingsome day-to-day shifts in foraging. The foragingpatterns (and probably other activitiesas well) of P. superciliosusmust reflect hourly, daily, and seasonal variationsin nectar availability. The weightingsystem based on use categoriescan also be used to evaluate certaincharacteristics of the food plants (Fig. 10). ApparentlyP. superciliosus visitsmainly plants whose flowers have long, curved corollas and that are relatively little used by territorial nonhermits. B. Snow and D. Snow (1972) and Stiles (1975) showedthat the hummingbirdsin two lowlandtropical areas visited flowers falling into two groupsbased on corolla morphology.Species with long, curved corollaswere visitedmainly by hermits(genera Eutoxeres, Phaethornis, Threnetes, Glaucis) that have curvedbills 30 mm or more in length. Most nonhermithum- mingbirdshave straightbills 25 mm or less in length, and tend to visit flowers with short, straightcorollas. Among the major food plants of P. superciliosus, only Heliconiaimbricata was visitedprimarily by nonhermits.Nonhermits were the major visitorsto severalfood plantsin categories1 and 2.

INTERSPECIES RELATIONS

An important behavioral difference between nonhermits and hermits is that the former (mostlymales) frequentlyhold flowered-centeredterritories while the latter do so very rarely and inconsistently(B. Snow and D. Snow 1972; Linhart 1973; Stiles 1975). The territoriesof nonhermitsat flowers are primarily for feeding,although they may servesecondarily as mating sites (pers. obs.; see also Pitelka 1942). The criticalresource being defended is the flowers,and they are defendedagainst all otherhummingbirds, regardless of speciesor sex. At La Selva, the nonhermitsthat held territoriesat speciesof plantsused by P. superciliosus 26 ORNITHOLOGICAL MONOGRAPHS NO. 27

25 %

I ' I ' I ' I ] I

-- COSTUSMALORTIEANA .... MALVAVISCUS ARBOREA JACOBINIAUMBROSA •'N,, -----PENTAG.• .•IADONNELLSMITHII \'%'\\\

o&' 1000 Joo' ,oo' 1600i TI•E OF DAY

Fig. 9. Daily nectar productionof eight plant speciesfrequently visited by P. superciliosus. A, four speciesin the genusHeliconia; B, four non-Heliconia species. [See note p. 78.--Ed.] includedmales of Chalyburaurochrysia (7 g), Thalurania[urcata (4.5 g), occa- sionallyFlorisuga mellivora (6.5 g), and Amazilia amabilis(4.2 g), and possibly both sexesof Amazilia tzacatl (5 g). Femalesof thesebird speciesusually were nonterritorial and often visited different flowers from males, or poached nectar from defendedflowers. These females,plus individualsof severalsmaller species of hummingbirdthat occasionallyvisited the food plantsof P. superciliosus(e.g., Klais guirneti,3 g) are groupedas "nonterritorialnonhermits" in the following discussion. When two hummingbirdsmeet at a clump of flowers,they almostalways interact aggressively;normally one bird forcesthe otherto moveaway or leavethe flowers LONG-TAILED HERMIT HUMMINGBIRDS 27

Corolla Length Corolla shape % Visitation, Territorial nonhermits

Fig. 10. Cumulative use categories of food plants in relation to: A. corolla length, B. corolla shape, and C. percent of visits to a food plant by territorial nonhermits. Note changes in the ordinal scale. entirely. Tables 6 and 7 presentthe outcomesof such aggressiveinteractions involvingP. superciliosusat different flowers. P. superciliosusalmost invariably lost in any suchaggressive interaction if the other bird was a nonhermit,regardless of its species,sex, or weight or the food plant at which the interactionoccurred (Table 6). Territorial nonhermitsalways drove P. superciliosusaway from the flowers; nonterritorialnonhermits almost always forced P. superciliosusto give way, but rarely actually chasedthem. Most interactionswith territorial nonhermits occurred at Heliconia pogonantha and H. imbricata. The latter is adaptedfor pollinationby nonhermits(Stiles 1975), and large clumpsregularly were defendedby male Chalybura,Thalurania, and A. tzacatl. P. superciliosustended to visit mostlysmall, undefendeddumps of H. imbricata,or tried to feed at the edgesof larger, defendedclumps. The flowering phenologyand corollamorphology of H. pogonantha,in contrast,are specialized for hermit pollination. However, the largest of the nonhermits,Chalybura, has a bill just long enough(22-25 cm) to negotiatethe corolla of H. pogonantha althoughit could not forage very efficientlyat this flower (pers. obs.). Where severallarge clumpsof H. pogonanthagrew closetogether, male Chalyburafre- quentlywere territorial. This occurredmainly in large areasof secondgrowth and largely resultedfrom human disturbance.In undisturbedhabitat, as along forest streams,most clumpsof H. pogonanthawere too smal! to be defendedby Chalybura, and far fewer interactions occurred (Stiles 1975). If the nonhermit was at all consistent in its territorial defense the amount of nectar a hermit could obtain at the defendedclump might be too low to justify the time and energy expendituresto visit it. The result would be to effectively remove large clumps from the floral food resourcesavailable to P. superciliosus. By contrast,dominance interactions appeared to play at most a minor role in 28 ORNITHOLOGICAL MONOGRAPHS NO. 27

TABLE 6 DOMINANCE RELATIONS OF PHAETHORNIS SUPERCILIOSUS AT SEVERAL FOOD PLANTS

Encounters with Encounters with Encounters with territorial other hermits nonterritorial nonhermits nonhermits Hrs. of Food plant observation Won x Lost Won Lost Won Lost Heliconia pogonantha 68 7 10 1 7 0 46 H. wagneriana 26 3 10 0 0 0 0 H. latispatha 29 0 0 0 1 0 7 H. imbricata 49 1 1 0 8 0 59 H. mariae 8 1 0 0 1 0 3 H-3 14 4 1 0 0 0 0 H-18 27 4 0 0 1 0 0 H-16 10 1 0 0 0 0 0 H-17 23 0 0 0 3 0 12 Costus tuber 7 1 0 0 0 0 0 C. malortieana 19 5 4 1 6 0 0 Malvaviscus arborea 5 0 0 0 1 0 0 Jacobinia umbrosa 10 3 1 0 2 0 1 Totals 30 27 2 30 0 128

x Won indicates that P. superciliosus was dominant in the encounter; lost means that P. superciliosus was subordinate. determiningfeeding patterns at flowersvisited mostly or exclusivelyby hermits. Interactionsbetween hermits were much lessfrequent than thoseinvolving nonhermits,in part becausehermits were much less aggressiveeven when two birds arrivedsimultaneously at a clump. Weight playeda role in theseinteractions as P. superciliosuswas dominatedby Eutoxeres(10-12 g) but dominatedP. longuemareus(2.5 g). No consistentpattern of dominancerelations was evident betweenP. superciliosus,Threnetes ruckeri (5.6 g), and Glaucis aenea (5.3 g) (Table 7). The former tended to be dominant at H. pogonantha and H-18, and subordinateto Glaucisat H. wagneriana.This was perhapsdue to habitat because Glaucisoften congregatedand nestedin the open secondgrowth areaswhere H. wagnerianagrew. This speciesand its closerelative in Trinidad, G. lu'rsuta,defend the nestingarea extremely vigorously, often attacking other hummingbirds at some distancefrom the nestitself (Skutch 1964a; B. Snow 1973a; Stiles,pets. obs.).

STRATEGIES OF FLOWER EXPLOITATION

Dependingon the spatial distributionof flowers,several alternative strategies of nectarharvesting are availableto hummingbirds(Janzen 1971; Hainsworth andWolf 1972a; Linhart 1973; Stiles1975). Territorialityrequires a rich, localized nectarsupply so that the costsof defenseare outweighedby the advantagesof exclusiveaccess (Brown 1964; Wolf 1969; Gill and Wolf 1975; Wolf et al. 1975). Commutingoccurs when a bird defendsa nonfood-centeredterritory at one site and travelsregularly to anothersite to feed (Stiles1973). There may be no single feedingarea but rathera seriesof small,scattered feeding sites that are arranged alonga moreor lessregular route or "trapline." Thesestrategies are not mutually exclusivebut representpoints along a continuumfrom clumpedto dispersednectar sources,from entirelywithin to entirely outsidethe territory. LONG-TAILED HERMIT HUMMINGBIRDS 29

TABLE 7 DOMINANCE RELATIONS OF PHAETHORNIS SUPERCILIOSUS AND OTHER HERMITS AT DIFFERENT FOOD PLANTS

Encounters with Encounters with Encounters with Encounters with Phaethorni$ Eutoxeres aquila Threnete$ ruckeri Glaucis hirsuta longuemareus Food plant Won x Lost Won Lost Won Lost Won Lost

Heliconia pogonantha 0 7 2 1 5 2 - - H. wagneriana - - 1 0 2 10 - - H-3 .... 1 0 3 17 H-18 - - 1 0 3 0 - - Costus rnalortieana - - 3 3 0 1 2 0 Jacobinia umbrosa .... 0 I 3 0 Others - - 0 0 1 1 2 0

• As in footnote 1, Table 6.

All availableevidence indicates that P. superciliosusemploys only commuting and regular foragingroutes in flower visitation. We have observedconsistent territorial defense only at the lek (males) or at the nest (females), never in connection with flowers. Territorial nonhermits exclude P. superciliosusfrom most clumps of flowers large enough to be economicallydefensible. Moreover, flowersspecialized for hermit pollinationrarely provideenough nectar at a single time and place to make territorialityfeasible (Stiles 1975). Thus, some movement betweennectar sources is obligatory. With small and scatterednectar sources, foragingis most efficient if flower clumpsare visited in a regular order so that reward levelsare maximizedand the time and energyexpended in movingbetween clumpsare minimized(Schoener 1971; Wolf et al. 1975). Most of our evidence for use of this foragingstrategy by P. superciliosusis indirect. Due to the density of the vegetationand the distancesinvolved, we have been unable to trace any foragingroute in its entirety. However,we have observeda numberof the essential featuresof suchan exploitationsystem: (1) A givenclump of flowerswas visited by a given individualat regularintervals through the day, over a period of days. (2) The directionsof arrival at and departurefrom a givenclump remainedfairly constantover many visits. (3) Any givenclump was visitedby one or a very few individuals,with rather little overlap betweenindividuals in flowers used. If a hermit visiteda clump sufficientlylarge that all its nectar could not be harvested in a singlebout of continuousfeeding, it often remainedin the vicinity of the clump for as long as 15-30 min. During this time, it sometimeschattered at or even attackedother hermitsattempting to feed at the clump. This was the closest P. superciliosuscame to feedingterritoriality, and it occurredonly at clumpsnot visited or defendedby nonhermits. Certain food plants of P. superciliosusreached very high densitiesin second growthareas, and manylek malescommuted to theseareas to feed. The major feedingareas at La Selvaincluded (see Fig. 1): the Point (chieflyH. imbricata and H. wagneriana), the East Boundary (chiefly H. pogonantha), and the old cacaoand secondgrowth between lek I and the field station(H. imbricata,H. pogonantha, and Costus malortieana). Once it arrived at a feeding area, P. super- ciliosusoften visited one flower clump and then passedby severalothers before 30 ORNITHOLOGICAL MONOGRAPHS NO. 27 visitinga second.During regularcensuses in the East Boundaryarea on several daysin late 1971 and early 1972, we locatedfour H. pogonanthaclumps that were usedconsistently and probablyin sequenceby one P. superciliosus(ROW). Distancesbetween clumps were approximately24, 60, and 85 m, and intervening clumpswere not visited. ROW probablywas a female (by measurements),but similardata are availablefor a knownlek male, GPB. This bird alwaysemerged from the forestat the samepoint, visited one largeclump (often remaining5-10 min), flew directlyto visit anotherclump approximately80 m away, and then continuedon beyondthe censusarea. A mist net set acrossthe path would con- sistentlycatch the bird. The route remainedstable for over one year. The spatialdistribution of foragingflights of lek malesparalleled that of floral food resourcesaround the lek. The birds of lek I usually foraged within 250 m of the lek, mostlyto the north (Fig. 11A). A few birds alsocommuted to feeding areasalong the QuebradaArbolera or up the QuebradaE1 Surfi;at certainseasons a few individualscommuted over 1 km to the Point or the East Boundary. The East Boundary,though 0.5 km away, was the closestlarge feeding area to lek III (Fig. liB). Clumps of flowers (e.g., H-16, H-17, H-18) along a trail midway between leks I and II were observedintensively in August-September1972 and July- August1973 for visitationby markedbirds (nearly all maleson theseleks were marked). Of the 48 observedforaging bouts 46 were by unmarkedindividuals, probablymostly females. Six of 7 P. superciliosus(all unmarked) mist-netted alongthis trail in 1972were clearly females. During July 1973, the trail between lek I and the field station was used as a transect to determine the number of flower visitsmade by lek males at differentdistances from the lek. A majority of flower visitswas by knownlek I malesat distancesof up to 300 m from the lek; at distancesgreater than this,most visits were by unmarkedbirds and females (Table 8). These observationssuggest that residentmale P. superciliosusfed reasonably near the leks, while femalesand perhapsnonlek males used areassome distance away. This assumesthat the observedsex ratio of flower visitorsin an area is representativeof the sexratio of the populationpresent in that area. A bird will probablyinclude in its regularforaging itinerary those flower clumpsat which it harvestsenough nectar to make revisitationprofitable. These are the clumps that havenot beenvisited recently by anotherhummingbird. By visitinga clump and removingits nectar, a P. superciliosusmakes visitation unprofitable for a considerabletime thereafter,which shouldmake it unlikely that another bird will visit the clump regularly. In this way suchan exploitationsystem can produce exclusiveor preponderantuse of at least small clumpsby the sameindividual. Given the concentrationof malesabout leks, femalesmust forage mainly at some distancefrom leks to encountersufficient unvisited clumps to make an energetic profit. This may explainwhy most P. superciliosusnests we found were well away from leks (Fig. 1).

NECTAR rS. INSECTS

The diet of P. superciliosusincludes small arthropodsas well as nectar, but the relativeproportions of eachare difficultto determinedue to the greatdifferences LONG-TAILED HERMIT HUMMINGBIRDS 31

N=202

t) E

N:57

I00 200 300400 500 600 700 •00 900 10001100 1200 1300 140015001600 1700 METERSFROM LEK Fig. 11. Recordsof distancesat which individualmale P. superciliosusforaged from their territories. A. lek I; B. lek III. We excludedrecords from long-termobservations in major feedingareas. P •_ recordsfrom the Point feedingarea; E • recordsfrom the eastboundary. in foragingtechniques required. Examinations of stomachcontents are not helpful becauseof the extremelyrapid postmortemdigestion of nectar. The arthropod food of P. superciliosusconsists almost entirely of spidersgleaned from foliage, dead twigs,or websin the mid- to upper understoryof forest,or the canopyof youngsecond growth (Stilesand HespenheideMS). P. superciliosusfoliage-glean usuallywell away from lek territoriesor flowers,often alongstreams or at breaks in the forest canopy. Compared with nectar, spidersshould representa food source the absolute availabilityof which changesrelatively little through the day; indeed foliage- gleaningactivity was much more evenly distributedover the day than was flower visitation(Table 9). An overall reductionin activitythrough the day is indicated 32 ORNITHOLOGICAL MONOGRAPHS NO. 27

TABLE 8 OBSERVED FLOWER VISITS BY LEK MALES AND OTHER INDIVIDUALS ALONG A 450-M TRANSECT BETWEEN LEK I AND THE FIELD STATION

Visits by Distancefrom Known lek • • and unsexedbirds Unmarked Lek I ct ct marked at leks birds

0-150 m 18 7 2 150-300 m 9 1 6 300-450 m 5 4 11 450-1- m 1 0 3 Totals 33 12 22 by the declinein total encounters,but the proportionof birdsseen foliage-gleaning actuallyincreased through the day.

TEMPORAL PATTERNING OF LEK ACTIVITY

BREEDING AND LEKKING SEASONS Our limited sampleof 15 nestsof P. superciliosusfrom La Selvaindicates that eggswere laid betweenJanuary and Augustwith peaksin January-Marchand June (Fig. 12C). At a nearby locality, females with enlarged gonadswere collected betweenearly Decemberand September,but not in October or November (Fig. 12A). Adult maleson the other hand had enlargedtestes year round (Fig. 12B). Considerationof the timing of molt in the populationand the presenceof bill corrugationson young birds (Ortiz-Crespo 1972) allows more detailed and quantitativeevaluation of the nestingseason. We have shownpreviously that individualP. superciliosusmolt at almostexactly 12-monthintervals, commencing their first molt about 12 monthsafter hatching. For a bird in any stageof primary molt, we can estimatethe date on which molt started; this should be close to the

TABLE 9 DAILY ACTIVITY PATTERNS OF PHAETHORNIS SUPERCILIOSUS IN DIFFERENT HABITATS

Total no. Feeding Foliage Habitat Time of day contacts• at flowers gleaning

Foresff Early morning (0600-0900) 54 15 7 Midday (1000-1300) 27 7 5 Afternoon (1400-1700) 18 2 4 Second growth, with Heliconia pogonanthaa Early morning (0600-0900) 85 33 4 Midday (1000-1300) 72 24 5 Afternoon (1400-1700) 39 15 3

Total no. contacts ---- total number of times a Phaethornis was seen or heard on census. Ca. 20 census hr for each time of day, over all times of year. Ca. 15 census hr for each time of day, all times of year. LONG-TAILED HERMIT HUMMINGBIRDS 33

N MONTHS

Fig. 12. Estimatesof the breeding seasonof P. superciliosus.A. data for size of largest ovarian follicle, an index of breeding readinessof females. B. length of left testis; males with testes more than 2.5 mm long probably are capable of inseminating females. In A and B, closed circles are adults, open circles are juveniles, and crossesthrough a circle indicate a molting bird. C. estimateddates of egg-layingin nestsobserved at La Selva, assuminga 21-day incubation period. D. dates of egg-laying estimated from timing of molt of adults and the bill corrugation characteristicsof immatures. See text for details of this method. calendardate of hatching. We have also shownthat bill corrugationscan be used to ageyoung birds up to about6 months(Stiles and Wolf 1974). If we subtract three weeksfor the incubationperiod of P. superciliosus(Skutch 1964a), we obtain the approximatedate on which the egg from which the bird hatchedwas 34 ORNITHOLOGICAL MONOGRAPHS NO. 27

I I I i I I ! I SURPRISE LEK

>_rr 300?õ0 I 9

z 200 8

Iõ0 5

z I00

5 • õ0 4 5 3 • 0 F M A M J J A S 0 N D J 1971 1972 MONTH

Fig. 13. Annual variation in calling activity among residentmales on lek Ic (Surpriselek) during one nearly completeannual cycle of lek occupancy.Numbers besideeach point indicate number of males for which data were recorded. The dotted line indicates that the records for April are for only one half-day; the others are for full days. laid. As this method estimatesthe dates of egg-layingin successfulnests only, any markedseasonal variations in nestsuccess will biasthis estimateof the nesting season. We have estimatedthe nestingseason from data on molt timing and/or bill corrugationsfor 208 individuals.It agreesfairly well with estimatesfrom nest recordsand gonaddata (Fig. 12D). Apparentlynesting occurred in everymonth of the year but was very rare in September,October, and November.The main nestingseason was January-July,with a sustainedhigh level of nestingfrom Februaryonward. This representsthe averagenesting season over several years, but it mayvary considerably from yearto yearas birds of differentages contribute to the estimate. The best availablemeasure of seasonalvariation in lek activity is given by all- day singingrecords (Fig. 13). Thesedata were taken concurrently with gonad data (Fig. 12) fromMarch 1971through January 1972. The mainlekking season of P. superciliosusextended from Decemberthrough July, droppingoff sharplyin LONG-TAILED HERMIT HUMMINGBIRDS 35

August;very low levelsof lek activityoccurred from Septemberthrough Novem- ber. The data thusinclude most of one lekkingseason, the periodof low activity betweenseasons, and the start of the next season. Highest activity levels were recordedat the start of the lekking season,probably the period of territory estab- lishment. During most of the season,singing activity fluctuatedconsiderably, possiblyon a day-to-daybasis. Weather sometimesaffected the amount of lek activity considerably;during midday lek activitywas higher on cloudy than on sunnydays, but loweston rainy ones. In general,however, singing activity on the lek remainedhigh throughJuly. The observedlekking seasonof P. superciliosusthus coincidesquite closely with our estimatesof the nestingseason, allowing for possibleyear-to-year variation. The main differencebetween male and female reproductiveseasons is that lek activityreached high levelspractically as soonas the lekking seasonstarted, whereasnesting reached its peak one to two monthslater. Egg-layingand lekking declinedmore or less simultaneouslyin August. The seasonalpatterns of lek activityand nestingwere stronglycorrelated with bloomingseasons of importantfood plants (cf. Fig. 8). In particular,the start of the lekkingseason agreed closely with the onsetof intenseflowering by Heliconia pogonantha. Nesting peaked later as other important food plants such as H. wagnerianacame into good bloom and the overall level of nectar availability increasedconsiderably. This later peakis not unexpectedas the energeticcost of nesting doubtlessis greaterthan that of holdinga lek territory. While early territorial establishmentis advantageousto the male,early nesting may createsevere energetic problemsfor the female,including depletion of her own energeticreserves to a dangerouslylow level (Ward 1969; Fogden 1972; Wolf and Wolf 1976). The major declinein lekking and egg-layingoccurred when H. pogonanthawas no longerin goodbloom but othermajor food plants(e.g., H-16, H. imbricata,H-3, and Costusmalortieana) were at their peak or still increasing.Females laying eggsat this time would still have anothermonth of caring for nestlingsand fledglingsshould the nest succeed.Young fledgedlater than Septemberwould be faced with low and still decliningnectar resourcesand might well incur a reduced chance of survival. The molting seasonwas not clearly definedin P. superciliosus,and birds in molt couldbe foundat any time of year. The peak of molt for the population wasJune and Julybut individualsdiffered by as muchas 8 monthsin the timing of their 4-monthmolting seasons (in accordancewith the long breedingseason) (Stilesand Wolf 1974). Molting and breedingseasons of the populationover- lappedbroadly, but the timingof theseactivities was largelyindependent among individuals.However, females sometimes interrupted molt to nest, while many maleswere activeon the lek and moltingsimultaneously, another indication of the higherrelative costs of nestingover lekking (Stilesand Wolf 1974).

DAILY PATTERN OF LEK ACTIVITY

Levels of singingand territorial aggressionon a P. superciliosuslek varied in a fairly regularand predictablemanner during a day. Most activitieswere con- centratedin three main singingsessions in the early morning,late morning,and 36 ORNITHOLOGICAL MONOGRAPHS NO. 27

MARCH,1970 JULY,1970 EL SALTO

301 7/19 birds 7/10 birds 20i

0 EL SURA

30 6/6 birds 5/5 birds 20

BANANA PATCH

30 4/4 birds 4/4 birds 20

0 0500 0900 1300 1700 0500 0900 1300 1700 Time of Doy

Fig. 14. Daily variation in calling activity on lek II (El Salto), lek Ia (El Surfi), and lek Ib (Banana patch) for one day early and late in the 1970 lek season.Fractions indicate proportion of all birds on the lek included in the sample. afternoon.This dailypattern was discernible over most of the lekkingseason, and for leks of differentsizes (Fig. 14). Patternsof individualmales often were less clear-cut. Singingand other activitiesof each residentwere influencedmore directlyby songsand displaysof neighboringresidents and intruders.A resident's presenceor absenceat the lek was dictatedlargely by the length of his foraging flights away from the lek and, perhaps,day-to-day variationsin the flowering of clumpsincluded on his foraging route. Thus, the periodsof most intense activity varied somewhatamong different individualson the sameday (Fig. 15) or for the same individualamong days (Fig. 16). The earlymorning singing period began at dawn and usuallylasted 30-50 min. Residentsapparently did not rooston their territories,as in severalnocturnal visits to leks we never found a roostingP. superciliosus.Rather, the residentsflew from their rooststo the lek just as it beganto get light. For a few minutesprior LONG-TAILED HERMIT HUMMINGBIRDS 37

3O PR

3O OYG

= 20 o {... Io

-= o

ß, 30 YP --,i [ 2o >, Io

ß, o --,i

• 3o YRP •20 o I0 u o

3O RY

o 0500 0700 0900 I100 1300 1500 1700 Time of Doy

Fig. 15. Variation in daily calling activity patterns of five resident males on lek Ic on 29 March 1971. Locations of these males are shown in Figure 21. to the onset of singingone could hear flight calls of hermits in the still-dark understory.Most of the residentsused relatively high, exposedsong perches at thistime. Singingbegan at about0515-0520 duringthe shortdays of December throughJanuary or February,and around0505-0510 duringthe longerdays of May throughJuly. Cloudy,dark weather at dawncould delay the startof singing for several minutes. 38 ORNITHOLOGICAL MONOGRAPHS NO. 27

1971 RY 3O Morch 20

3O April 2O

1o • No Records--•

o r r '/

3o Moy

IO • o •,30 June ß-- 20

•, I0 g o

E 30 b. July .• 20 ':' I0

30 August 20

30 1972 Jonuory

0 0500 0700 0900 I100 1300 1500 1700 Time of Fig. 16. Variation in daily calling pattern of male RY on lek Ic throughthe end of the 1971 lek seasonin August and the start of the 1972 seasonin January.

Usuallyone or twobirds, not alwaysthe sameindividuals, initiated singing and after a few minutes,most or all lek residentsjoined in. The most intenseand synchronoussinging activity for the lek as a whole duringthe day occurredin the following10-15 rain (Figs. 14, 15). This dawnchorus declined as the light intensityincreased to the point where an observercould distinguishthe colorsof LONG-TAILED HERMIT HUMMINGBIRDS 39

Fig. 17. Frequencyof chaseactivity on areas observedduring all-day watchesat lek Ic. Chasesare scaledby total minutes of singingfor all birds observed;this correctsfor absences of males and variations in numbers of residents observed on each date. The data indicate both a seasonaland daily decline in chases. Data are plotted at the midpoints of 90-rain observation periods. back paint and tagsof the singingbirds. It wasnow apparentlylight enoughin the understoryto permit the birds to engagein high-speedchases through the densevegetation. A period of very intenseaggressive activity followed in which territorialinteractions and chasesreached peak levels(Fig. 17). Singingactivity continuedat a high level, and the lek becamea bedlamof songs,chatters, squeals, chasenotes, and bill-pops. A givenresident sang for a few minutesat mostbefore going off to join a chase, challenge,or display at a neighbor,or an intruder. The mist-netsamples (Table 10) indicatethat virtually all of this early-morning activity was by residentmales; females and nonresidentmales were only rarely capturedat the lek duringthese hours. Probablythis activityserved to redefine the ownershipof songperches and territories. Most residentmales almostsimultaneously left the lek between0545 and 0610 to forage,and for about 1 hour the lek was virtually deserted.Except for the few birds with suitableflowers on their territory, most residentsprobably had not fed beforethis, havingbeen continuouslypresent on the lek sincedawn. We doubt that residentsfed before arriving on the lek becauseof very low light intensities. Mist-netting and censusesin feeding areas away from leks revealed a burst of P. superciliosusactivity beginningabout 0600-0615. As malesreturned to the lek followingthis "breakfastbreak," the synchrony of activitywas reducedconsiderably. Throughout the rest of the day, eachresident alternatedperiods of singingand interactionon the lek (usually 10-30 min) with absencesof a few minutesto an hour or more. These absences,presumably for foraging,were longestand most frequent at midday and separatedlate morning and afternooncalling periodsof most residents(Fig. 15). 40 ORNITHOLOGICAL MONOGRAPHS NO. 27

TABLE 10 PHAETHORNI$$UPERCILIOSU$ MIST-NETTED AT LEK I, 1971-1973

Dec-Jan Feb-May Jun-July Aug-Nov Total

Adult males Resident territorial 29 114 33 2 178 Resident nonterritorial 0 0 6 14 20 Nonresident 10 7 1 2 20

Immature males Territorial 0 0 12 10 22 Nonterritorial 0 13 26 26 65

Adult females 4 12 3 4 23 Immature females 0 2 6 6 14

Total 43 148 87 64 342

The later morningsinging period usuallyshowed a well-definedpeak about 0800-0830, but fairly high singinglevels usuallycontinued until 0900 or later (Figs. 14, 15). Chasesare lessfrequent (Fig. 17), and the periodsof uninterrupted singingaveraged longer than in the early morning. Femalesapparently visited the lek most often during the late morning. Of the 19 femalesmist-netted on lek I duringthe lekkingseason (Table 10), 1 was caughtbetween 0515 and 0700, 10 between 0700 and 0900, 6 between 0900 and 1100, and 2 between 1100 and 1200 (nets were usuallytaken down at noon). All observedcopulations occurred between 0730 and 0930. The onset and intensityof the afternoonsinging period were quite variable, but the latter usuallypeaked around 1530-1600 (Figs. 15-17). Synchronybe- tween maleswas further reduced,and active territorial defense,including chases, was rare (Fig. 17). Thus, a residentoften engagedin 15-30 min of unmolested, continuoussinging after arrivingon the lek, and the peak levelsof songfor the lek as a whole could be comparableto those of the morning. Residentsalso spentconsiderable time sittingquietly on their songperches in the afternoon. Few femalesvisited the lek at this time, but later in the breedingseason young maleswere especiallyconspicuous on the lek in early to midafternoon. Activity ceasedgradually in the late afternoon. One by one, the residentsleft their territoriesand flew from the lek. By 1700 usuallyonly one or two birds werestill present,and all singinghad stoppedby approximately1715. A resident mightremain on his songperch for severalminutes after he stoppedsinging. At the time lek activity ceasedfor the day, it was considerablylighter than when singingstarted at dawn. The daily course of lek activity by P. superciliosusshowed certain definite relationshipswith the nectarsecretion patterns of major food plants (cf. Fig. 9). In particular the ratio of active territorial defense(numbers of chases) to territorial advertisement(bird-minutes of singing)declined through the day from a peak in the early morning,as did the nectar productioncurves (Fig. 17). The mostenergy was expendedin territorialactivity just prior to the period in which LONG-TAILED HERMIT HUMMINGBIRDS 41

YRP YO(I st yr. bird)

20 March IO

IO May

0 • 30

June -• I0

• 0

2O I July

August

30 3O

20 2O IO I December

0 0500 0900 1300 1700 0500 0900 1300 1700 Time of Day

Fig. 18. Seasonal variation in daily calling activity from one territory on lek Ic. When YRP left the lek in mid-July it was quickly replacedby YO, a young bird with a relatively short period of high activity at the end of the season. The territory was reoccupiedby YRP in December 1971 as the seasonstarted and YO, now marked as YRO, held an adjacent territory (see Figure 22). nectarsupplies were at their daily maximum,the "breakfastbreak." Conversely, in the later afternoonwhen little nectarwas available,little energywas expended in chasesas mostbirds engaged in longbouts of energeticallyeconomical singing. In addition,flowers of someHeliconia species were still too tightlyclosed at dawn 42 ORNITHOLOGICAL MONOGRAPHS NO. 27

MONTHS HATCHII•YEAR I 1971 1972

1969 1970 ? 1970 1969 1970 ß 1970 1970 1970 ? ? ? ? ? 1971 1971 1971 1971 1971 1971 1971 1971 1971 1971 1971 1971 1971 1971 1971 1972 1972 1972 1972 1972 1971 ? 1972 1972 1972 1972 1972

Fig. 19. Presence and territorial occupancy of male P. superciliosus on lek I, February 1971 to April 1973. A thick bar indicates territorial occupancy, a line indicates the male was present but not controlling a territory. to permit easy accessto the nectar by hummingbirds.H. imbricata flowers, for instance,did not open fully until 0545-0600, about the start of the "breakfast break." The daily pattern of lek activity also varied seasonally.At the peak of lek activity (Decemberand January), singingwas practicallycontinuous through the day, and the middaybreak was reducedor absent(Fig. 18). Through most of the lekking season (February through May or early June), the three singing periodswere quite distinct. As the end of the lekking seasonapproached, the residentsprogressively restricted their lek activity to the early morning (e.g., YRP in Fig. 18). Justbefore leaving the lek for the season,the residentsappeared and sangbriefly only in the early morning;for the rest of the day young males hatchedearlier in the breedingseason contributed most to the activity of the lek. LONG-TAILED HERMIT HUMMINGBIRDS 43

Fig. 20. Locationsof lek Ia, "El Surfi," and lek Ic, "Surprise,"showing vegetation types, major landmarks,and distributionof major food plants. Heavy stipple= forest; light stipple = cacao;cross-hatching = arboretum(open grassyunderstory, trees left standing);wavy lines = densethickets. Letters denoteclumps of Heliconia plants as follows: I • H. imbricata; R = H. pogonantha; S = H-3; T = H-18.

LEK DYNAMICS

SEASONAL VARIATION IN LEK COMPOSITION

BetweenFebruary 1971 and April 1973, we tried to keep all the residentsof lek I marked,and to mark new individualsas they appeared,by settingmist-nets in the lek onceevery 1-2 months(Fig. 19). Both beforeand after mist-netting, we ascertainedthe territorialstatus of as many individualsas possible.Because we captured most or all of the lek residents each month, we doubt that birds 44 ORNITHOLOGICAL MONOGRAPHS NO. 27

•¾•" ! YRP,:.•

'•'?•:-•,•.•,,, • •(•,4,,-._.,.o•o•',._ -.. / : I, \ \ ? • ',,, LONG-TAILED HERMIT HUMMINGBIRDS 45 learned to avoid the nets. Residentsand intrudersflew at the same heights,so we probablycaptured most or all of the intruderson a givenday, as well. In view of the difficultyof identifyingand sexingintruders visually, the mist-netdata (Table 10) probablyprovide the mostcomplete and unbiasedpossible sample of the individualspresent on lek I on thoseparticular days. Data from lek II (1970-72) andlek III (starting1973) are consistentwith the informationderived from lek I (Table 10). We thereforebelieve that the lek I data are representative of the seasonaldynamics of P. superciliosusleks at La Selva. The largestnumbers of adult maleswere presenton the lek at the start of each lekkingseason (all youngof the previousbreeding season could now be considered adult). In Decemberand January,70-80% of thesemales obtained territories, althoughsome failed to hold them (e.g., YPG in 1972, BRN in 1973; see Figs. 20-24). By February relativelyfew nonterritorialmales were about, mostlyper- sistentintruders that never establishedthemselves on the lek (e.g., OO in 1972). Where the unsuccessfulmales went is unknown,although a few becameresidents of anotherlek (e.g., WY, marked on lek I in February 1971, and a residentof lek II by April 1971). Once territories were defined and nonresidentmales excluded,the tempo of lek activity stabilizedat a slightlylower level until May or early June (Fig. 13). Changesin lek compositionduring most of this period were due to occasional deaths of old residents and additions of new residents at the peripheryof the lek. Theseperipheral territories often were unstablein space and time, and the birdsestablishing them were individualsthat appearedon the lek late in the seasonas well as thosethat failed to obtain territoriesearlier (e.g., unmarked resident(s) in March and May 1971, RG in 1972; Figs. 21, 22). A major changein lek compositionoccurred with the appearanceof young maleshatched earlier in the samebreeding season. A few appearedon the leks as early as March (possibly2-3 monthspostfledging); their numbersbegan to increasein mid- to late May (Table 10). When thesebirds first appeared,their songswere variable, unstructured, and of no definitesong type. They often sang from various points on the lek, partly as a result of recurringdisplacement by residents,especially during periods of low lek activity,as at midday. In periods of more intenseactivity, young birds sangfrom high perches,often 6-10 m above ground,where adult males seldomsat exceptearly in the morning. After a few weekson the lek, someyoung males establishedterritories on the lek periphery. However, relativelyfew young malesbecame territorial before mid- to late June. The reductionin activity of an adult male prior to departurefrom the lek probablyreflected his own physiologicalcondition rather than conditionson the lek as a whole. In late June someadults were still fully active,while othershad reducedactivity to varying degrees(Fig. 19). A male that left the lek between

Fig. 21. Locations of territories of males on leks Ia and Ic in March, May, June, July, 1971. The dashed lines represent the territory of each male (= the area containing all his song perches) on the given date. Parenthesesindicate a male that sang or interacted briefly at a particular site but was not consistentlyterritorial on that date. Specific identity, if known, of each male is indicated by his color code; unmarked males are indicated by a U. Note the tendencyof territories to occur in dense vegetation and not to be centered around food plants. 46 ORNITHOLOGICAL MONOGRAPHS NO. 27

Fig. 22. Locationsof territoriesof males on leks Ia and Ic in December 1971 and January- March, 1972. See legend of Figure 21 for descriptionof symbols. LONG-TAILED HERMIT HUMMINGBIRDS 47

// 48 ORNITHOLOGICAL MONOGRAPHS NO. 27

Fig. 24. Locations of territories of males on leks Ia and Ic in January and March, 1973. See legend of Figure 21 for descriptionof symbols.

1 June and 1 July was lesslikely to return the next seasonthan a male leaving after 1 July (Fisher'sexact probabilitytest; p = 0.022). The departureof adults sometimescoincided with the start of molt or the period of heavy body molt, but some males showedno relation between the intensitiesof lekking and molt (cf. Fig. 3 of Stiles and Wolf 1974). The vacated areas at the end of the lekking seasonoften were taken over by youngmales, and a sharpincrease in the number of young males holdingterritories occurred at this time (Table 10). Through July and early August lek compositionchanged rapidly so that by August most territorieswere held by birds of the year. When an adult left his territory and a youngmale movedin, the latter usuallysang at a much higher rate than the adult, at leastinitially (Fig. 18). A brief resurgenceof songcould occuron the lek as a whole as adultswere replacedby youngbirds (note the small peakin July, Fig. 13). By late Augustlek activityhad declinedprecipitously. All adultseither had left the lek or appearedinfrequently. The immaturemales also were leavingthe lek, and only a few still sangregularly enough from a localizedarea to be con- sideredterritorial. In Septembersinging was brief and irregular, and no birds were territorial. Males, usuallyimmature, entered the lek for a few minutes,sang briefly from severalterritories of the previousyear, and left. Only in the early morningwas more than one bird regularlypresent. In October and November, the lek was virtually deserted. One heard at most a songor two at dawn, then for the rest of the day only flight callsand, rarely,a chase,presumably involving LONG-TAILED HERMIT HUMMINGBIRDS 49

birdswhose foraging routes happened to passthrough or closeby the lek area. Activityon thelek at thistime probably was similar to that in any othercomparable habitat at La Selva. Male P. superciliosusreturned to the lek within a week or two at the start of the next lekkingseason (late Novemberor Decemberin most years;Fig. 19). On 1 December 1971 lek I was practicallydeserted, although a brief songwas heardalong the QuebradaE1 Surfiin midmorning.On 13-14 December13 birds were vigorouslyterritorial, and 7 intruderswere identified, 4 of which eventually became residents. Females apparently visited the leks relatively infrequently, as we suspected from the scarcityof observedcopulations. From Table 10 we calculatethat an averageof 1.5 adult femalesper day appearedon the lek during the lekking season,compared with 0.55 per day when the leks were not active. Thus, unless femalesnot ready to mate avoidedthe lek during the lekking season,only about one female per day visitedthe lek to mate. If a lek of 15 males serves10-30 females,each female must come to the lek on the averageof onceevery 2-4 weeks. This seemsa reasonableestimate because a successfulnesting attempt of P. super- ciliosusrequires about 2 months,but at least 75% of the nestsof this speciesfail and thus require lesstime (Skutch 1964a). The length of time expendedin an averagenesting attempt thus is probablyabout three weeks, as femalesprobably renestfairly soonafter a failure (D. Snow and B. Snow 1964, 1973; Stiles 1973).

SURVIVORSHIP AND RECRUITMENT OF LEK RESIDENTS

We examinetwo aspectsof survivorshipin P. superciliosus: (1) the turnover of lek males between years, and (2) the seasonalpattern of mortality of lek residentsincluding immature birds that establishterritories in their first season. Females and intruderswere impossibleto study in this regard becausethey were not localized and were recapturedonly rarely and inconsistently.Determining seasonalmortality patterns is mademore difficultbecause many malesdisappeared completelyfor 3-5 monthsbetween lekking seasons, precluding regular censuses. Therefore, one can only comparesurvivorship over the lekking seasonwith that between lekking seasons. Measuring survivorshipover the nonlekking season assumesthat survivingresidents return to the samelek at the start of a lekking season.We have no evidenceof establishedresidents switching leks nor have we ever encountereda male that was territorial one year and nonterritorial the next. Our most reliable survivorshipdata are from lek I between 1971 and 1973 (Table 2). Survivorshipwas significantlyhigher through the lekkingseason (ca. 90%) than it was betweenlekking seasons (ca. 55%) (X2= 8.69; p < 0.005). It is unlikelythat humaninterference produced this differencebecause such interference,in theform of monthlyrecapturing and handling, was much greater during the lekkingseason. Also, no bird so completelylost its markingsduring the nonlekking seasonthat it was unidentifiablewhen it returnedto the lek in December. We haverecapture data for a numberof individualsduring the nonlekkingseason. Every malethat failed to returnto the lek in Decemberapparently disappeared in the latter part of the nonlekkingseason. The data suggestthat up to 75% of 50 ORNITHOLOGICAL MONOGRAPHS NO. 27

TABLE 11 SURVIVORSHn' OF RESIDENT MALE P•ET•O•S SVPERC•L•OSVS WITHIN AND BETWEEN LEKKIN• SEASONS:LEK I, 1971--72

Adult males Immature males Died during Died during Date Surviving interval Surviving interval

Feb 71-Aug 71 12 1 - - Aug 71-Dec 71 7 5 4 3 Dec 71-Jul 72 16 2 - - Jul 72-Dec 72 8 8 5 3

the annual adult mortality occurredin October and November,the time of year when flowerswere scarcest(Fig. 8); food shortagemay play a major role in limiting the P. superciliosuspopulation at La Selva. Young males survived as well as older malesover the nonlekkingseason. This is much less surprisingin P. superciliosusthan it would be in a specieswhere adultshold food-containing territoriesduring the nonbreedingseason. Given the seasonaland year-to-year shifts in flower abundanceto which P. superciliosuswere exposed (Stiles 1975, 1978a), old and youngmales faced the seasonof food shortageon a far more equal footing than is often the case in territorial species. A potentialbias may occurin that we consideronly youngmales that establishedlek territories;these may have been the strongestand mostvigorous young birds, possiblywith above- averagesurvivorship for their age-group.In any caseit seemsevident that if a periodof high postfledgingmortality existed, it cameprior to the time whenyoung males establishedthemselves on the lek. Combining the between-seasonand within-seasonsurvivorship percentages, one findsan annualturnover rate of males on a P. superciliosuslek of about 50% (Table 11). If survivorshipafter the postfledgingperiod is independentof age and half the maleson a lek are replacedeach year, then a male P. superciliosusmust live, on the average,for only one to two seasons.In any one lekkingseason half of the residentswill be in their first full seasonon the lek; very few birds will be over 3 yearsold (Table 12). Considerablerecruitment must occureach year to maintain the leks at a relativelyconstant size (Table 2).

TABLE 12 AtE STRUCTURE OF PHAETHORNIS SUPERCILIOSUS ON LEK I IN 1971-1973 LEKKING SEASONS

No. residentsaged 1971 1972 1973 1 yr 6 8 8 2 yr 1 6 5 3 yr 1 - 2 4 yr - 1 - Unknown (probably older than 1 yr) 5 1 1 Total 13 16 16 LONG-TAILED HERMIT HUMMINGBIRDS 51

Our best data on recruitmentagain come from lek I between1970 and 1973, whenwe usuallywere able to mark youngmales as they appearedon the lek and to followtheir subsequentattempts to establishresidency. We have someknowl- edgeof the historiesof 32 residentsof lek I prior to the time they established residency.Only YN was an adult when he first appearedon the lek; all others were marked as innnaturesand becamelek residentsin their first full lekking seasons.The lek experiencegained by a youngmale duringhis first seasonmay influencehis chancesof subsequentlybecoming a resident,as is demonstrated by the historiesof the 40 youngmales that appearedon lek I duringthe 1971 and 1972 lekking seasons(Table 13). All survivingyoung males who held territoriesas iramatures,and most youngmales who sangor interacted,became residentson lek I in the next season. Of the 12 young males without known experienceon lek I, 5 becameresidents of leks III or IV and may have had lek experiencethere; at least two of thesewere territorialon lek III. Only 3 of the 12 becameresidents of lek I (Table 13). The timing of a youngmale's first appearanceon the lek can affect his chances of eventuallybecoming a residentby influencing(1) his survivorship,and (2) the amountof lek experiencehe gains as an immature. Relativelymore young males arriving on lek I in March-May obtainedterritories as iramaturesthan did thoseappearing in June-August,or in Septemberor later (3/7 = 43%, 7/26 = 27%, and 0/7 = 0%, respectively). However, survivorshipto the next breeding seasonshowed the oppositetrend in the three groups (57%, 65%, and 71%, respectively: see Table 13). Perhapsthe early birds had less energy reserved goinginto the lean seasonbecause of their longerperiods of strenuouslek activity. The overall result was that recruitment of lek residentswas greatestfrom the June-Augustgroup, both relativelyand absolutely(14/26 = 54% vs. 3/7 = 43% for both early and late groups). Birds that appearedon the leks late could gain territoriesthe next year largelybecause many of their older and more experienced fellow yearlingsfailed to survive. Any youngmale survivingthe lean seasonhad a good chance (20/26 = 77%; Table 13) to become a lek resident,and the few that failed to do so were amongthe less experiencedand presumablyyounger membersof their year~class.Unfortunately, we have little information on the survivorshipof nonresidentmales, but if a young male failed to become a lek residentin his first full lekking season,his chancesof doing so in a subsequent seasonapparently were low. Practicallyall recruitmentof new lek residentswas from the ranks of yeafling males (Table 11).

INTERMALE RELATIONS ON THE LEK Possibleindicators of the relativestatus of lek residentsinclude morphological or other attributesof the males,variations in their territorial behavior, or differencesin the qualityof their territories.Obtaining information on thesepoints is difficult.Differences between males are likely to be subtle,and the densevege- tationof the lek makesit impossibleto observeseveral territories simultaneously or to follow interactionsfor any distance.In other speciesof lek birds dominant malesdo mostof the matingon the lek (Hogan-Warburg1966; Lill 1966; Robel 1966;Wiley 1973). However,we wereunable to evaluatemating success because we so seldom saw copulation. 52 ORNITHOLOGICAL MONOGRAPHS NO. 27

TABLE 13 SUBSEQUENT STATUS OF YOUNG MALE PHAETHORNIS SUPERCILIOSUS FIRST A??EARING ON LEK I IN 1971 AN•) 1972

Became resident on: Never became resident Experience on lek I in first season Lek I Lek III or IV on any lek Died

Held territory 6 0 0 4 Sang and interacted but not territorial 5 1 2 4 Neither sang nor interacted 3 5 4 6 First appeared on lek March-May 2 1 1 3 June-August 10 4 3 9 September and later 2 1 2 2

Totals 14 6 6 14

Severallines of indirectevidence suggest that malesin centralterritories on a P. superciliosuslek tended to be dominantto males with peripheralterritories. If a vacancyoccurred in the centerof a lek duringthe main lekkingseason, the spacewas always filled by the shiftingand expansionof the territoriesof neighboring residents.Peripheral males might attempt to establisha territoryin the center should a central residentdie (e.g., RG when RY disappearedin March 1972; Fig. 22), but they were never successfulin our experience.By contrast,vacant peripheralterritories were filled, if at all, by formerlynonterritorial birds, and newlyresident males established themselves on the periphery.Also whilecentral territorieshad relativelystable locations and boundaries,peripheral territories oftenwere very unstable in spaceand time (e.g., YB in the 1972 lekkingseason; Figs. 22, 23). Whenyoung males began to settleon the lek in May and June,they always establishedterritories on the lek periphery,at leastuntil the centralresidents began leavingtheir territoriesat the end of the lekkingseason. We never saw a young male chasea residentadult and there was never any suggestionof severalyoung malesmoving into a territoryand wearingdown the defensesof an adult, as apparentlyhappens in someNorth American hummingbird species (Stiles 1973). A youngmale taking over the territoryof a residentwas also setting the stagefor a contestwith that residentfor possessionof the territory shouldboth return for the next lekkingseason. In 20 of 24 suchcases (Table 14) the old resident regainedhis territory at the outset;in a few casesthe residentmoved to a more centralterritory when another resident failed to return. The oneexception occurred in 1972,when RR apparentlyreturned late to the lek andWW andBY became establishedin most of his former territory. Over the next month RR forced WW andBY back and eventuallycontrolled practically the samearea he had in 1971 (Fig. 22). Youngmales establishing residency on the lek at the startof their first full lekkingseason were much less likely to regainthe sameterritory they had held at the endof the previousseason (Table 14) becausethey were almost invariably LONG-TAILED HERMIT HUMMINGBIRDS 53

TABLE 14 SHIFTS IN ELK TERRITORIES OF RESIDENT MALE PHAETHORNIS SUPERClLIOSUS IN RELATION TO PREVIOUS STATUS 1

Status of male in previous year Adult Young Young resident territorial nonterritorial

At start of lekking season in second year: Regains old territory 20 5 1 Takes more central site 3 1 1 Takes more peripheral site 1 7 3 During lekking seasonin secondyear: Shifts toward center 5 5 2 Shifts toward periphery 1 1 0 Data are with respectto site of most frequent singing and interacting in previous year. displacedif the old resident returned. If a returning resident did not reoccupy his old territory at the start of a lekking season,he was more likely to move to a more central spot (3 of 4 instances)while a young male was more likely to occupy a more peripheral site (7 of 8 instances). Later shifts of both old and young birds (following the disappearanceof establishedresidents) were nearly alwayscentripetal in direction. Eight of 18 young males that sang or interacted fairly persistentlyat a site but were never really territorial there obtained the same spotsthey had held as immaturesor more central ones,while 23 of 24 returningresidents did so, a highly significantdifference (X•= 10.73, p < .01). However,this dominanceof old residentsover youngones was not absoluteand related mainly to contestsfor specificterritories, where the incumbentresident might be expectedto have the advantage. In a number of instances,a resident reoccupiedhis old territory insteadof a more central site (presumablyavailable becauseanother resident failed to return), and a young male succeededto the ownershipof the centralsite. Usuallythe youngmale had occupiedthe site at the closeof the previouslekking season(e.g., RBW, Figs. 23, 24). A high degree of site attachmenton the part of residentsis indicated,and the amountof previous experienceon a territory (whetheras a full residentor a yearling) may be important in determininga male'scapacity to regain that territory in succeedingyears. The chief meansfor residentsto improve their positionon the lek (i.e., moving to a more central location) is apparentlyto shift their territories as more central residentsdisappear during the lek season,rather than taking up presumablybetter territories at the start of the season. Some individuals seemed more or less dominant or subordinate from the start. At the end of the lekking seasonthe last residentsto leave sometimesshifted to more centralterritories before finally departing.YB, as a youngmale in August 1971, movedinto an area formerlycontrolled by OYG and RR. The following year YB was excludedfrom this area as the old residentsreturned; he spentthe 1972 season as a peripheral resident between leks Ia and Ic (Figs. 21-23). At the end of the 1972 seasonYB again moved into this area following the departureof RR and BW (Fig. 23). In 1973 YB was againon the peripheryof 54 ORNITHOLOGICAL MONOGRAPHS NO. 27 lek Ic after losinga contestwith RBW for his apparentlypreferred spot. Other birds that remainedmore or lessperipheral for two full seasonswere WW and WGP, even thoughyoung males like YO and RBW were able to occupymore central positions.It seemsthat an individual'sdominance status often may be establishedduring his first year on the lek and may not changemarkedly thereafter. There is considerableevidence that male P. superciliosuspreferred to hold territoriesin areaswhere activitylevels were high, generallythe more central locations (see discussion). For instance, GR, the lone survivor of lek lb in February 1971, desertedthe territory he had held since 1969 and moved to a peripheralspot nearby on lek Ic. Lek Ia graduallypassed from a majorcenter of activity to a minorone through1971 and 1972. With the exceptionof BW and GYW, maleson Ia tendedto be thoseunable to hold a positionon Ic. GBY was a resident on Ic in February1971, but soonthereafter lost his territoryto RR. GBY then took overRR's territoryon Ia, but movedback to Ic thenext year when a peripheralresi- dent,GO, failed to returnfor the 1972 lekkingseason. During the 1972 season,two otherresidents of Ia, YB and WGP, shiftedtheir territoriestowards the peripheryof Ic; by May BW was essentiallyalone on Ia. Finally in June 1972 BW moved to Ic and took over RR's territory,RR presumablyhaving left for the season(Figs. 22, 23). Unfortunately,neither BW nor RR returnedfor the 1973 season.Lek Ia has remained deserted since BW left it. As we were unableto measuredominance of a male directly,we have assumed from the precedinginformation for thesehummingbirds and alsofrom information on otherlek species(e.g., Hogan-Warburg 1966; Wiley 1973) that spatialposition on the lek often is a reflection of the male's relative dominanceposition. In general,dominant males tend to be more centrallylocated than subordinates.For the hummingbirds,position of a territorywas defined by its spatiallocation relative to adjacentterritories. A central territory had commonboundaries with at least four other territoriesand at least 270 ø of its circumferenceadjoining other territories. An intermediateterritory had commonborders with two to four other territories.A peripheralterritory had commonboundaries with only 1-3 territories and 180ø or lessof its circumferenceadjoining other territories. For mostrelation- ships examinedthe spatial position of a male was taken as the most central territoryheld for two or more monthsin the first four monthsof the lekking season.However, for the relationbetween time spentsinging and positionwhile singing,we usedthe positionoccupied when the singingdata were recorded. We found no quality of the males that consistentlycorrelated with position. Bill lengthis correlatedwith matingsuccess in a sandpipersexually dimorphic in this characterbut not in plumage (Jehl 1970), but there was no significant tendencyfor longer-billedmale P. superciliosusto be more central (Table 15). Alsono significantcorrelation exists between body weight and position(Table 15), in accordwith the data on body weight and dominancefrom variouspasserine species(Smith 1976). Yet the nonterritorialcategory does contain a relatively higherproportion of small-billed,lightweight birds than the territorial categories. No color characterin P. superciliosusis particularlyvariable except back color, which shouldplay no role in displaysor dominanceinteractions because displaying birdsface one another(Fig. 7). The qualityof loudnessof a male'ssong may LONG-TAILED HERMIT HUMMINGBIRDS 55

TABLE 15 MORPHOLOGICAL AND BEHAVIORAL PARAMETERS ! OF MALE PHAETHORNIS SUPERCILIOSUS IN RELATION TO LEK POSITION

Position

Nonterr. or failed to hold Central Intermediate Peripheral territories Morphologicalparameters • (data from leks Ia, Ic) Sample size 10 16 10 8 Body weight (g) 6.09 ñ 0.27 6.14 ñ 0.25 6.08 ñ 0.14 5.99 ñ 0.25 (5.7-6.4) (5.6-6.5) (5.9-6.4) (5.7-6.3) Bill length (ram) 37.50 ñ 1.00 37.50 ñ 1.35 37.55 ñ 0.89 36.69ñ1.43 (36-39) (35-39.5) (36-39) (35-39) Behavioral parameters (data from leks lb, Ic, II) Sample size 12 12 6• Minutes singing 0730-0930 76.83 ñ 11.04 67.50 _ 16.02 41.67 _ 13.494 (49-87) (32-93) (23-58) Length of early morning foraging period (rain) 2 58.92 ñ 13.07 59.58 ñ 15.33 69.93 ñ 11.14 (32-82) (30-82) (38-93) Data given are mean, standarddeviation, and range in each case. No means are significantly different by Student's t. Includesnonterritorial or failed to hold territories category. Highly significantlydifferent from other meansin group by Student'st. be correlatedwith his dominance,and certainly intruders and young males sing differentlyfrom residents.However, among established residents the situation was complicatedby dialect differences,and severalcentral residentshad rather soft songs. Variationsin territorialbehavior included rate and especiallypersistence of singing,and differencesin the frequencywith whicha male invadedthe territories of his neighborsin relationto how oftenhe had to defendhis ownterritory against their invasions.If dominanceon P. superciliosusleks is translated,directly or indirectly,into mating success,then it is importantto compareand evaluate territorialbehavior in that intervalwhen females are mostlikely to appearon the lek, about0730-0930. Data from all-day singingrecords indicate that peripheral malessang significantly less persistently during this periodthan did more central males (Table 15). Peripheralmales tended to have shorter songbouts as well, at least in part becausethey often were challengedby more centralmales as soon as they beganto sing. Central malessang for longerperiods uninterrupted by invasionsof otherresidents. On severaloccasions we notedthat prior to losing all or part of their territoriesbirds were challengedpersistently by other residents. Defenseby the residentmale graduallydiminished as he ceasedto chasethe invader consistently,and eventuallypermitted the latter to sing from song percheshe had formerly controlled. Finally, the invader chasedthe residentfrom his own songperches, and the losersoon abandoned part or all of his holdings. Another factor that may affect dominanceis a male's capacityto remain on territory continuously,which is influencedby the amountof time he must spend feedingelsewhere. This in turn dependson the distanceto and quality of the flowerson his foragingroute. Becausea malemust return from his early-morning 56 ORNITHOLOGICAL MONOGRAPHS NO. 27

1.00' •.90'

.60..Di &A Introlek \

2 .10'

0 500 1000 1500 2000 2500 3000 DISlANCE

Fig. 25. Proportionsof P. •uperciliosusmarked at one site that were recordedat other sites different distancesaway. 1 : movementsbetween leks and foragingareas by adult males; 2: movementsbetween foraging areas by adult females;3 : movementsbetween leks by adult males;4 = movementsbetween leks by •mature males;5: movementsbetween leks by adult females. For intralek movements,squares: lek I; diamonds: lek II; solid fi•res = adult males; open figures: immature males. See Figures 3 and 6 for l•ations of the different sectionsof leks I and II, respectively. foragingperiod or "breakfastbreak" with enoughaccumulated food reservesto permitintense activity over the next 2-3 hours,the lengthof his breakfastbreak may be an indicatorof the qualityof his foragingroute. However,no significant correlationexists between lek positionand the lengthof the earlymorning foraging period (Table 15). Beingable to spendless time foragingand hencepresumably more time on the lek, apparentlydoes not increasethe probabilityof holdingand maintaininga central territory in P. superciliosus.

INTER- AND INTRALEK MOVEMENTS

Movementsbetween leks by adult male P. superciliosuswere extremelyrare (Fig. 25). In 4 years,we recordedjust 2 of 68 residentsof leksI and II appearing on the other lek. OW, a resident on lek II, was captured on lek I at the end of the 1971 lekking season.WY, marked on lek I in February 1971, was a resident of lek II in April 1971. Only one of 33 known residentsof lek I, OY, was captured on lek III, and we recordedno movementsin the other directiondespite intensivenetting on and aroundlek I. In much lessextensive netting we recorded one residentof lek II on lek III (GBP), but movementbetween these two leks may well be slightlyhigher as both are locatedon the QuebradaE1 Salto, probably LONG-TAILED HERMIT HUMMINGBIRDS 57 a major flyway (seeFig. 1). We neversaw a knownresident of one lek interacting with residentsof another lek, and it is difficult to see what advantagehe would gain by doing so. Although there are few available data, there is some suggestionthat non- residentadult males, presumablybirds without a lek territory, moved between leks much more often. One suchmale, YR/OG, was recordedon leks Ia, lb, and II between1969 and 1970. In 1972, BOP was recapturedon leks I and III. These malesmay have tried to establishterritories on one lek after failing to do so on another,or theymay even have tried to establishon two lekssimultaneously. In either casethe long-distancemovements must be expensivein termsof time and energy. A nonresidentmaking a shift of approximately1 km betweenleks must either fly a longdistance to reachhis old foragingroute, or mustestablish a new foraging routewith each shift. This may explain,in part, why we so rarely recordeda nonresident male in more than one season. During 1972 and 1973 we obtaineddata on movementsbetween leks I and III. Sevenyoung males were markedor seenon one lek at the end of a lekking season and were found to be residentson the other lek the followingseason. In five of these cases,young malesmarked on one lek definitelysang or interactedat the end of the seasonon the other lek where they eventuallybecame residents. Thus, a period may exist at the end of the lekking seasonduring which young males move from lek to lek until they finally establishthemselves on one. Such move- mentswould be least expensiveat this time of year becausethe number of major food plantsin goodbloom was at or near its annualmaximum (Fig. 8, Table 4; Stiles 1975, 1978a). This behaviorprobably was important in givingyoung males a secondchance should they fail to gain a territory on a given lek. As noted earlier,a youngmale that obtaineda territoryat the end of a lekkingseason was virtuallyassured of residentstatus on the samelek in the next season.No young male ever gaineda territory on one lek, then becameresident on another. This period of juvenilemale dispersalat the end of the lekking seasondoubtless pro- motesgene flow and reducesthe inbreedingpotential of eachlek, sinceestablished adult males never changeleks (see Buechnerand Roth 1974). Females occupiedmainly the areas between leks for nesting and foraging (Table 9, Fig. 1). Thusit is not unlikelythat a femaleready for copulationmay have had to choosefrom two or more approximatelyequidistant leks, and she couldwell visitmore than one of them duringa breedingseason. Females might thereforebe seenor capturedon two or more leks relativelyfrequently. Because femalesso seldomvisited leks our data are rather limited. We recordedonly 4 of 44 femaleson more than one lek. However, a more appropriateindex may be the ratio of the number of times femaleswere resightedor recapturedon the sameversus different leks. This indicatesa higherdegree of movementas females were recapturedor resighted12 times on the lek where they were first seen or marked, and 4 times on different leks. Intralek movementsby resident males were consistentlymore frequent than interlekmovements. From 30-60% of the residentsof a givensector of a lek were seenor recapturedon anothersector of the samelek (Fig. 25). The proportions of birds movingbetween leks Ia, Ib, and Ic were similar to the proportionsof birdsmoving between sectors of a lek that are similardistances apart; this supports 58 ORNITHOLOGICAL MONOGRAPHS NO. 27

our conclusionthat all subdivisionsof lek I are part of the samelek unit. How- ever, the fractionof birds seenor capturedon more than one sectorof the same lek was surprisinglylow, giventhe distancesinvolved, which are much shorter thanthe averageforaging distance (Fig. 11). As expected,young males showed higher proportionsof such movementsthan did adults, about 60% vs. 40% for lek I (Fig. 25). Territorialshifts between lek sectorsoccurred rarely (e.g., RR, GBY, andBW discussedin the precedingsection). Also, some young males that establishedterritories on onesector at the endof a lekkingseason became residents of anothersector in the next year (e.g., YB). A complicatingeffect that is not discussedin detail here is that of songtype. Males on differentsectors of a lek oftenhad different songs, and young males, once they had acquireda song,tended to settlein the sectorwith maleshaving the samesong (Wolf and StilesMS). The amountof interlek movementby malescan bestbe evaluatedin the context of movementsof thesemales from leks to foragingareas or of femalesbetween leksand foraging areas (Fig. 25). The proportionof malesmoving over a given distancefrom one lek to anotherwas far smallerthan that movingthe samedistance but from a lek to a foragingarea. The sameheld for intralekmovements. The proportionof malesmoving to a foragingarea was far higherthan that movingto another sectorof the samelek the samedistance away. Femalesshowed rates of movement similarto thoseof lek malesto foragingareas and of youngmales between leks (Table 8, Fig. 25). The low rate of interlekmovement by adult malesis explicable in termsof time and energy. The rivals most likely to displacea territorial male are his neighborson the samelek. Residentsof other leks representno threat whateverto his territory,because interlek transfers by territorialmales virtually neveroccur. Residentsof other sectorsof the samelek are at mosta minor peril, as intersectortransfers are rare also. A residentmale gainslittle by flying to anothersector of the lek and interactingwith the residentsthere, and gainsnothing by goingto anotherlek. Interlek movementsrepresent a considerablewaste of time and energythat couldbe betteremployed in reinforcingthe male'sdominance positionwith respectto hismost frequent challengers and intruders,the neighboring residentsof his own lek. This offers a male his best chanceto expand or shift his territory, without leaving it unguardedvery long in an interaction.

DISCUSSION

LEK EVOLUTION IN PHAETHORNIS SUPERCILIOSUS Leks of P. superciliosusoccur in the densestpart of a very dense habitat, the understoryof disturbed,tropical wet forest. Residentson their territoriesare rarely in visual contact,an obviousand important differencefrom the residents of leks of certaingrouse and shorebirds.The form of the displaysof P. superciliosus,including their orientationabout a songperch, obviously has been influencedby the structureof the habitat. It is unlikely that these leks evolved as an antipredatordevice analogousto a flock in enabling more rapid detection of predators(Lack 1968; Trivers, pers. comm.). This presumedadvantage of leks shouldhold primarily in habitatssufficiently open so that severalresidents simul- taneouslycan observethe approachof a predator. However,singing from dense thicketsmay give some protectionto individual males. LONG-TAILED HERMIT HUMMINGBIRDS 59

Our limited evidenceindicates that the lek territories of male P. superciliosus are the sole sitesat which matingsequences are initiated (althoughthe actual matingcan occurelsewhere following a chasefrom the territory). Lek residents thereforeshould enjoy a strongmating advantage over nonresidents,either nonterritorialmales or malesthat displaysolitarily. Severallines of indirectevidence suggestthat male P. superciliosuscan obtain more matingson a lek than they can by displayingalone. First, malesapparently preferred to hold territoriesat or near centersof activity, as the casesof BW and GR mentionedabove clearly indicate.Second, if malescould achieve significant mating success by displaying solitarily,one would expectto find lone malesscattered throughout the forest. That suchmales did not occur (in spiteof the smallnumber of leks and the fact that somemales failed to obtainlek territories)supports the thesisthat malescan producemore offspringby joining a group. If this is so, then the lek system generatesa positivefeedback reinforcingthe tendencyfor males to competefor lek territoriesrather than to establishsolitary mating stations. What then limits lek size? Why do not all malesconverge on a singlelarge lek? The most likely limiting factor is the distributionof suitableflowers. This could make it energeticallyunfeasible for more than a certainnumber of malesto join any givenlek as they would have to fly too far to forage. Also, concentrated foragingof too many malescould create a wide "denectarizedzone" around a lek that wouldbe energeticallyexpensive for a femaleto crossto reachthe lek, and femalesmight accordinglychoose to visit smallerleks. Anotherpossible limitation on groupsize is that an incomingmale probably stands a betterchance of becom- ing a resident(and henceachieving copulations) of a smalllek than of a large one, especiallygiven the short averagelifespan of P. superciliosusthat makes movingup the dominancehierarchy with ageunlikely. Within the lek systemfemales should attempt to mate with the male(s) most likelyto providetheir offspring with the greatestchance to surviveand reproduce, andselection should favor those capable of makingsuch discriminations. A possible indicatorof a male'sfitness is his proficiencyin territorial defense•in effect, his dominance.Our data indicatethat peripheralterritories tended to be held by youngermales who presumablywere subordinateand wouldnot predictablycon- tribute as much to a female's fitness as the more dominant males on a lek which held the most centrallylocated territories. If mating successis correlatedwith dominancein male P. superciliosus,then selectionshould favor the tendencyto hold central territoriesas these are the most stronglycontested territories on the lek and their holdersshould be the most dominantmales present. The essentialfeature of a central territory is probablynot its locationper se, but rather that it is surroundedby other territories,with defendedboundaries on all sides. Central territoriesalso tendedto be in the densestvegetation on the lek. The net result of thesefactors is that centralterritories were usuallythe smallestand closesttogether. This, coupledwith the fact that centralmales sang more persistentlyduring the period of female visitation,created a gradientof activity or overall song volume from the center of the lek (or sectionthereof, in a large and dispersedlek like lek II) outwards.This gradientwas maintained partlythrough the aggressionof centralmales toward peripheral ones, resulting in lessuninterrupted singing during female visits to thelatter. 60 ORNITHOLOGICAL MONOGRAPHS NO. 27

We have no evidencethat femaleP. superciliosusremained on the lek for long periods,although they might make severalvisits before mating. This, plus the lack of any morphologicalcharacteristic correlated with position(= dominance) amongmales, potentially makes the activitygradient a female'smost reliablecue to a dominantmale. It may alsoindicate to a male the directionin whichto expand or shifthis territoryshould the opportunityarise. The proximategoal of intermale competitionis thusto hold a territoryas high in the activitygradient as possible. It is importantto note here that activitygradients need not be symmetricalin all directions;due to configurationof lek vegetationor topography,"central" males need not be in the geographiccenter of the lek. A large lek of P. superciliosus may have several"centers" (which often correspondto songdialect groupings), much as do leks of certain grouseand shorebirds(Hogan-Warburg 1966; Wiley 1973, etc.). Some evidencethat femalescue on activity centersin mate choice existsfor other lek speciesincluding the Ruff (Philomachuspugnax) and severalgrouse. Ruffs on large leks tend to keep satellitemales off their territories,while males on small leks allow satellitemales to co-occupytheir territories(Hogan-Warburg 1966; van Rhijn 1973). This couldbe a mechanismto increasethe activity levels on small leks, and maleson suchterritories get more matings. Similar levels of activity could be generatedon large leks simply by the larger number of males. In the Greater Prairie Chicken (Tympanuchuscupido) females apparently will not submitto mountingby maleswithout the stimulusof the high activitylevels of the centerof the lek (Robel 1967). For Black Grouse (Lyrurus tetrix), Kruijt and Hogan (1967) suggestthat densityof malesand activity levelsper unit area may be important componentsof female choice of a mate. In this regard, the accretionof youngmales on the peripheryof P. superciliosus leks late in the breedingseason may be advantageousto the residents.So long as the youngmales do not establishthemselves and becometoo active and vocal, their chancesof mating are slight. However, their very presenceadds to the activitylevels of the lek, whichmay enhanceits effectivenessin attractingfemales. As most youngmales were not permittedto establishstable territories until the adultsbegan to leave the lek, they were unlikely to attract femalesaway from the adultsbut rather enhancedominance/activity gradients in favor of the latter. The advantageto young males of such precociouslek activity doubtlesslay in the experiencegained that virtually guaranteedthe survivorsa positionon the lek in the next breedingseason (Table 13). Similar argumentshave been advanced for the role of youngmales on leks of sometetraonids (Robel 1967; Kruijt and Hogan 1967). Althoughthe phenomenonof dominance/activitygradients seems widespread in lek systems,an importantstructural difference may exist betweenleks of P. superciliosusand thoseof most grouseand Ruffs. In theselatter leks, one or a few dominantmales accountfor virtually all of the matings;subordinate residents havevery little chanceof matingexcept insofar as theycan moveup the dominance hierarchywith time. A nonterritorialmale is only slightlyworse off than a subordinate residentin terms of his immediatechances of mating, but the latter has a better chance in the future if he survives. The dominance/fitnessgradient on a P. superciliosuslek appearsfar lesssteep, LONG-TAILED HERMIT HUMMINGBIRDS 61

suchthat peripheralmales stand at leastsome chance of mating.The majorgap in fitnessis probablybetween territorial and nonterritorialmales, with less of a differencebetween dominant and subordinateresidents. This may explainwhy returningadult malesalmost invariably try to regaintheir old territoriesinstead of takingup a morecentral, vacant territory. Subsequentcentripetal shifts occur duringthe lek season,with the male usinghis territoryas a basefrom whichto occupyadjacent areas, but neverrisking loss of his old areabefore gaining control of the new. The dominancehierarchy on a P. superciliosuslek probablyis less rigidlystructured spatially than, say,a grouselek. This is a logicalconsequence of the highannual turnover of lek residentsso that malesrarely survive to move up the dominancehierarchy with age. The hypothesizeddominance/fitness structures of P. superciliosusand grouse leks are diagrammedin Figure26. A matingadvantage of the centerwill set up a geneticfeedback that will tend to drive a P. superciliosus-typelek systemto evolvetoward a more structured,grouse-type system unless other factorsinter- vene. Factorslimiting such a feedbackon P. superciliosusleks probablyinclude not only the high male turnover,but also a generallylower probabilityof mating due to the denselek vegetation,the rarity and unpredictabilityof femalevisits, andthe fact that all malesfrequently must leave the lek to feed. That the residents are not in visual contactmakes it less likely that a dominantmale will be able to interferewith a mating on a subordinate'sterritory. Dominanceon the lek is apparentlyunrelated to how long the malesare away foraging;hence, dominant maleshave as great a chanceof missinga femalevisit for this reasonas do subordinate males. Unlike certain other hummingbirdsthat join singingassemblies (Skutch 1958, 1967), male P. superciliosusdid not defend food sourceson or near their lek territories. Indeed, we think it best to specificallyexclude from the definitionof a lek those territorial systemsin which the territories serve other functions as well as that of mating. Spatial separationof defendedmating sitesand regular feeding areas seem to characterizeother bird speciesthat form leks (D. Snow 1962, 1963; Lill 1976). Defenseof feedingareas is precludedby the time and energy demandsof maintaininga lek territory. Food sourcesor foraging zones of lek speciesare usuallyfixed in space,enabling easy location and rapid visitation, but are generallynot energeticallyprofitable to defend,either because of uniform abundanceover wide areas (e.g., buds and leaves), rapid turnover and shifting abundance (fruit), or occurrencein small, isolated patches, as in the nectar suppliesused by P. superciliosus. The densevegetation on P. superciliosusleks doubtless places a strongselective premiumon effectiveauditory communication. The loud, persistentsong, often audible for 100 m in the forest understory,was the sole means of advertisement over any distanceby lek males. Visual displaysassumed primacy only at very closerange, within about 1-2 m. A numberof otherlek systemsin densevegetation rely on sound,either vocal or mechanical,for distancecommunication. Examples includethe HammerheadBat (Bradbury,pers. comm.), manakins (e.g., D. Snow 1962;Lill 1976), cotingas(Gilliard 1962;B. Snow1972), and otherhummingbirds (e.g., Davis 1958; Skutch1964a; B. Snow1974). Visualdisplays given by P. superciliosusappeared to differ little in male-male 62 ORNITHOLOGICAL MONOGRAPHS NO. 27

Fig. 26. Hypothesizedrelationship of dominanceand/or fitness to lek position in leks of grouseand Ruff vs. leks of P. superciliosus.In the former, subordinate,peripheral males have virtually no chanceof mating; the P. superciliosussystem is much looser, with any resident male havinga fair chanceof mating. "Center" refers to areasof high lek activity, not neces- sarily to the geographiccenter of the lek. and male-femaleencounters, perhaps reflecting the lack of sexualdimorphism in this species.Without morphological or auditorycues to the sex of an intruder,a residentmale shouldrespond aggressively to any conspecificentering his territory and establishhis dominance.An intrudingmale wouldthen leave,while a female could signalher readinessto mate by remainingperched and allowingthe male to mount. Becausethe costto a male of spermlost from an inappropriatecopu- lation is minimal, the brief and unpredictableappearances of females probably placea strongselective pressure on the male to attemptmating with anythingthat sits still long enoughfor him to mount. Hermits in general show little or no sexual dimorphism;even in the most dimorphicspecies, young males pass through a female-like plumage (B. Snow 1974). This situationis very unusualamong lek species,in most of which strong intermale competitionhas produceddistinctive, even bizarre male plumages. A frequentcorrelate of sexualdimorphism in polygynousor promiscuousspecies is sexual bimaturism (Wiley 1974), that is, different rates of sexual maturation in malesand females(usually slower in the former). The high annualmortality rate of male P. superciliosusprobably reduces drastically the potentialmating advan- tages of delayedmaturation and insteadselects for lek residencyat the earliest possibleage. Sexualdichromatism in hummingbirdsoften is relatedto purely aggressiveinteractions,frequently involving territoriality at flowers. In highly dimorphicspecies, femalesusually are nonterritorialand malesterritorial (Wolf 1969, 1975b; Stiles 1973). In several monomorphic,brightly colored species,both sexes hold territoriesduring at leastpart of the year (Wolf 1969, 1975b; Stilesand Wolf 1970). Monomorphicbut dull-coloredhermits lack territoriality at flowers. Iridescent colorsapparently are not important aggressivesignals in hermits, perhapsbecause LONG-TAILED HERMIT HUMMINGBIRDS 63 of low lightintensities on displayareas. Many nonhermitlek hummingbirdsare highlydichromatic, suggesting that coloris usefulfor aggressivesignalling in at leastsome of their displays. The dailytemporal pattern of lek activityin P. superciliosusis organized around at least three factors. First is the infrequentand unpredictableappearance of females,leading to strongselective pressures for malesto spendas muchtime as possibleon their territories,especially during the late morningwhen females are mostlikely to appear. Secondis the recurringnecessity for defenseof the lek territory and reassertionof dominanceposition on the lek. Most of this was compressedinto a short period of aggressiveactivity just after dawn, although nonresidentmales could intrude at any time. A residentshould leave his territory unguardedas little as possiblebecause an intruder that has time to sing and interacton the territory may be much more difficult to expel than one who has not. The third major factor is the availabilityof nectar,which followsa highly predictabledaily coursearound which all other activitiesmust be organizedif the mostfavorable balance between foraging and lek activityis to be attained. The timingof the breakfastbreak corresponded closely to that of maximumdaily productionand accumulationof nectar in importantfood flowers (Fig. 9; Stiles 1975 and unpubl. data). The burst of aggressiveactivity before this break probablyreflected the importanceof reestablishingdominance position on the lek as earlyas possible,and redefiningterritorial rights prior to the arrivalof females. Theoretically,a nonterritorialmale could try to establishhimself in a lek territory duringthe breakfast break when the resident was absent, but he wouldbe sacrificing thebest foraging time of theday to do soand would be at an energeticdisadvantage in the subsequentconfrontation with the returningresident. That femalesappeared on the leks mostlyin mid- to late morningprobably also relatesto nectar availability. It is at this time that they shouldhave the most energyreserves and be bestable to leavetheir foraging areas to engagein courtship.The lesssynchronized activitiesof residentsfrom midmorningon probablyreflected varying distances flown and ratesof energyintake on feedingbouts. Food supplydoes not appear to playa majorrole in determiningthe dailyschedule of leksin otherbird groups, doubtlessbecause no distinctdaily rhythmof availabilityexists for fruits,buds, etc. On the other hand, predationmay stronglyinfluence daily schedulesin at leastsome grouse leks (Hartzler 1974) but seemsto be of little or no importance in P. superciliosusleks. The seasonalpattern of lek activityin P. superciliosusseems closely tied to the bloomingseasons of certain food plants, especiallyseveral species of Heliconia. Food supplycan regulatethe lekking seasoneither directlythrough the males' energybudgets, or indirectlythrough the nesting-•hencemating--season of females. In P. superciliosusthe onsetof lekkingappears related to the first factor, the end of the lek seasonto the second.A similarlyclose relationship between breeding activityand bloomingof food plantsexists in severalother hummingbirdspecies (reviewin Stiles1973), includingP. guy,a lek form (B. Snow1974). The lekking seasonsof manakinsin Trinidad appearto correspondto periodsof fruit abundance,though the relationshipmay not be as preciseas that betweenflowers and 64 ORNITHOLOGICAL MONOGRAPHS NO. 27

hummingbirdleks (D. Snow 1962, 1963). Comparativelylittle quantitative informationis availableon the effectsof food supplyon lekkingseasons in other species;for mostof these,detailed observations on foragingpatterns have yet to be made.

COMPARISONS WITH OTHER HERMIT SPECIES The subfamilyPhaethorninae, collectively called hermits,comprises 6 genera and31 species(Meyer de Schauensee1966). In the largegenus Phaethornis, the speciessuperciliosus, 1onguernareus, and guy form leks throughouttheir ranges; tuber formsleks in someparts of its rangeand not in others(Davis 1934; Oniki 1970), while some species(e.g., pretrei) apparentlydo not form leks at all (Ruschi 1950). Leks sometimesoccur in the genusThrenetes, but this varies even in the samespecies at the samelocality (Skutch 1964a; B. Snow 1973b). At La Selva, we found solitarymale T. ruckeri singingin some areas, and leks of three to six birds in others. In Glaucis, leks are not known. In Trinidad, G. hirsutamales hold individualterritories, within whichfemales nest (B. Snow 1973a; Snowand Snow 1973). At La Selvawe found no suggestionof lek behaviorin G. aenea. In Eutoxereswe observeda small lek (3-4 birds) at La Selva. The social systemsof the remaining speciesand genera of hermits apparently are unknown.Thus, lek socialsystems are common,but by no meansuniversal, among hermitsbut seemparticularly characteristic of the genusPhaethornis. Lek residentsin all lekkinghermits for whichsex has been determined invariably are males (Arp 1957; D. Snow 1968; presentstudy). Femalesvisit the leks for the purposeof mating. The final phaseof mating, copulation,may occur inside the male's lek territory or elsewhere,sometimes around the nest of the female (B. Snow 1974), following a chasestarting from the territory. There has been no suggestionof matingsinitiated away from the lek in any of thesespecies. In all speciesmost lek activity is betweenmales, and visitsby femalesare relatively rare. Different hermit speciesvary considerablyin the temporal continuityof leks. Leks of P. superciliosushave been found at the same site for at least 12 years in Guyana(Davis 1958) and8 yearsin thepresent study. Leks of P. 1onguernareus and P. guy alsomay occupythe samesite consistentlyfor severalyears (B. Snow 1974; D. Snow 1968; Wiley 1971; pets. obs.). On the other hand leks of Threnetes and Eutoxeres are much more variable. A lek of Eutoxeres at La Selva was active betweenDecember 1971 and February 1972 but not the following year. A 3-5 bird lek of Threnetesremained more or less constantin size and positionfor 3 years;in anothersite we found a solitarymale in someyears and a lek of up to six malesin other years. Severalcalling siteshave been used con- sistentlyby solitarymales for up to five years;turnover of individualswas not studied. Leks of severalspecies are locatednear flywaysand/or feedingareas. In Trinidad,leks of P. guy are found on steepslopes or ridgeswhere streamsare availableas flyways, and where a majorfood plant, Heliconia bihai, is plentiful. Sometimesleks evenare locatedin standsof Heliconia,which is exceptionalin P. superciliosus.However, no large, dominant,nonhermit species like Chalybura, LONG-TAILED HERMIT HUMMINGBIRDS 65 that might disruptleks by defendingHeliconia, occurs in Trinidad. In neither Trinidad nor CostaRica are P. longuemareusleks especiallynear flowers. This speciesvisits many small,largely insect-pollinatedflowers and also eats relatively largequantifies of insects(B. Snowand D. Snow1972, pers.obs.); specificforaging sitesmay be lessimportant than in the other lek hermits. Glaucisin Trinidad is a nonlekspecies territorial along streams in areasimportant for nestingand foraging (B. Snow 1973a). No hermit is known to consistentlydefend flowers. In malesof all hermits studiedso far, the main advertisingtechnique whether aloneor in a lek is to sing. In somelarger species of Phaethornis,like superciliosus and guy, the songis simpleand repeatedmonotonously. At the oppositeextreme is the complex,rather melodious song of P. longuemareus(Skutch 1951; D. Snow 1968; Wiley 1971). P. ruber has a songintermediate in length and complexity (Davis 1958; Oniki 1970; B. Snow 1973b). The large P. yaruqui also has a notably loud and complexsong (Ruschi 1961). Threnetesruckeri has a long relativelymelodic song (Skutch 1964a; pers. obs.), while T. leucurushas a simpler song,a three-notephrase rapidly repeated, in the mannerof the large speciesof Phaethornis(Snow 1973b). Eutoxereshas a simplesong consistingof a single, repeatednote, but the successivenotes differ in tone and pitch (pers. obs.). Al- thoughpossessing a songcomparable to that of P. ruber in complexity,Glaucis hirsutararely sings. Perhapsthe unusualsocial system and restrictedhabitat of the speciesin Trinidad make vocal communicationby males largely unnecessary (B. Snow 1973a). Local song types exist in P. longuemareus(D. Snow 1968; Wiley 1971), P. superciliosus(this study) and probablyin P. guy (B. Snow 1974). Song varies geographicallyin P. superciliosus(B. Snow 1973b), P. longuemareus(comparing songsin Costa Rica with those reportedby D. Snow 1968), and P. ruber (Davis 1958; Oniki 1970). In specieswith local song types considerablegeographic variationis expectedbecause the final form of the songis learnedby youngmales (B. Snow 1974; Wolf and Stiles MS). Members of Phaethornis, Threnetes, Glaucis, and Eutoxeres all have con- spicuoustail patterns,and characteristicallywag the tail up and down when' perched;singing males wag the tail very strongly. Someauthors (e.g., Slud 1964; Skutch1964a) suggestedthat the tail wag is coordinatedwith the rate of calling, but this was not so in P. superciliosus.The rate and intensityof the wag increase in all species,however, and the tail often is fannedwhen a conspecificapproaches or enters the singingmale's territory. All hermitsso far studiedhave sometype of hover displayin which the displaying bird hoversbefore a perchedbird (sometimesboth may hover), displaying its gapeand/or throat pattern,depending on the head and bill coloration.P. guy has a conspicuousgape, but the adult has no conspicuousface pattern; it has a specializedgape display. P. superciliosushas a fairly conspicuousgape and a strikingface patternand displaysboth of thesein the float. P. guy has no display strictlyanalogous to the float, but, as discussedbelow, the tock displaycontains elementsof both gape and bill-pop and float. Gape displaysare not known in Glaucis or Threnetes,the speciesof which have dull mouth linings; instead the bright cinnamon breast (Glaucis) or conspicuousthroat pattern (Threnetes) is displayedby the hoveringbird (B. Snow1973a, b). The largerPhaethornis species 66 ORNITHOLOGICAL MONOGRAPHS NO. 27

and Threneteshover with the long axisof the bodyvertical, but in Gla'ucisand the smallerPhaethornis the body axisis horizontal,and the tail is raised the so-called "boat"posture (Skutch 1951). A back-and-forthmovement is mosttypical, but P. longuemareusand P. rubermay turn completecircles in somesituations (Skutch 1951;Arp 1957;Davis 1958 ). In general,the aerialdisplays of thesesmall species are considerablymore elaboratethan thoseof the largerhermits, including addi- tionalsound effects (Davis 1958;B. Snow1973a). Interestingly,perched displays suchas side-by-sidehave not been reportedin the small hermits. The extensionof the tongueplays a role in the hoverdisplays of P. tuber (Davis 1958) and Glaucis(B. Snow1973a). In a Guyanapopulation of P. superciliosus, B. Snow(1973b) observedthat the bird visitinga territorialmale typicallyperched besidehim andtouched the insideof his displayedgape with its tongue.We never observedthis in the Costa Rican populationof P. superciliosus,although the displayotherwise seems similar to a side-by-side.Guyanan populationsof P. superciliosusalso have a specialflight call usedonly on the lek; we never heard sucha call in CostaRica. It is worth notingthat the conspecificityof the Central and SouthAmerican birds is in question(Meyer de Schauensee1966). Tock or pop soundsare producedby P. guy, P. supercilious,and rarely, by Glaucishirsuta (B. Snow 1973a). The associatedgape displayis more elaborate and spectacularin P. guy than in P. superciliosus,involving a more conspicuous spreadingof the mandibularrami; the gape is bright red in P. guy. The tock displayof P. guy has both side-to-sideand dartingmovements; the rock or pop is given at the endsof rapid sidewaysarcs. In additionto usingthe rock in all of the same situationsas P. superciliosususe the bill-pop, male guy also character- isticallytock-display when changingperches in their territories,or may hover and tock-displayrepeatedly over a perch. Tock or bill-pop soundsmay be produced by a suddenexpulsion of air from the glottisin conjunctionwith a snappingopen of the bill. B. Snow (1974) describedseveral casesof loss of voice in P. guy and suggestedthat the tock is producedvocally in that speciesbecause on losing its voice, a bird no longer could tock. Becauseair comingfrom the lungs or air sacs must passthrough the syrinx en route to the glottis, the two explanationsare not mutually exclusive. Perch exchangesequences are known for P. superciliosus,P. guy (B. Snow 1974), P. longuemareus(D. Snow 1968), and Threnetes leucurus (B. Snow 1973a); they probablyoccur in mosthermits as they seemto be a ritualizedform of precopulatorymounting that occursregardless of the sex of the sittingbird. Copulationin most or all speciesresembles that in P. superciliosus(Fig. 7H); Davis' (1958) report of aerial copulationin P. tuber has not been confirmedby other authors (B. Snow 1973a). Severalauthors attempt to discriminatebetween displays given to femalesand thosegiven to other malesby lek residents.In the sexuallydichromatic species (P. ruber, P. guy) this may be feasible,but it is complicatedby the fact that in at least the latter, young males are indistinguishablefrom femalesby plumage and approachthe lek like females(B. Snow 1974). Many of the supposedfemale visitsto the lek endingin chasesmay have been by yearlingmales that had not yet obtaineda lek territory. In view of the variabilityin displaysequences found in P. superciliosus(Table 3), there is no satisfactoryevidence that the displays LONG-TAILED HERMIT HUMMINGBIRDS 67 of a lek residentto a visitingmale or femaleare qualitativelydifferent, and evidence for even a quantitativedifference is weak (B. Snow 1974). The similarity may includeeven mountingand attemptedcopulation if a perchingmale sits still. Homosexualmounting and similarityof mating and male aggressivedisplays in P. superciliosusare consistentwith the limited information for other hummingbirds,including nonlek species (Stiles and Wolf 1970; Stiles1973; Wolf 1975a; Stilesand Ortiz-CrespoMS). The brief, rare, and unpredictableappearances of femaleson male territories (B. Snow1974; this study)are probablymajor factors in the frequencyof leaf copulationsor falsematings in mosthermits and many otherhummingbirds. Such behaviorhas been reported for variousspecies of Phaethornis(Arp 1957; D. Snow 1968;B. Snow1973a, 1974; presentstudy), Threnetes(B. Snow1973a), Glaucis (B. Snow1973b), andhas been seen in Eutoxeres(Stiles, pers. obs.). The analogy to masturbation in mammals is obvious. The only other study of lek behavior in hummingbirdsthat deals with such parametersas survival,recruitment, molt, and foragingof lek residentsis the work of B. Snow(1974) on Phaethornisguy in Trinidad. P. guy is a fairly closerelative of P. superciliosusbut showsmarked sexualdimorphism, and young males pass through several more or less intergradingplumage stagesbefore acquiringthe definitive plumage. B. Snow did not work with marked birds, but stated that "differencesin head markings,combined with differencesin calls (songs), made it possibleto identifya numberof individualswith a high degreeof certainty." Her interpretationof plumagesequences, which is crucial to her evaluationsof survivorshipand recruitment,is basedlargely on territorialoccupancy by birds of similar songtype in different years. In P. superciliosus,at least, it may not be safe to assumethat a bird singingfrom the sameterritory with the samesong asthe previousyears' resident is in fact the sameindividual. We havemany cases of individualturnover in which the songtype of the territoryholder remainsthe same(see Fig. 6, esp.BRG and WPB). Therefore,certain of Snow'sconclusions requireconfirmation from studieswith markedbirds. In general,temporal and spatial patterns of lek activityin P. guyresemble those of P. superciliosus.Territory size in P. guy averagedslightly larger, but from B. Snow'scomments regarding frequent visual contact of singingmales we infer that the vegetationon the guy leks was somewhatmore openthan that on our superciliosusleks. The lekking seasonof guy began in November,but less suddenly than that of superciliosus;only by mid-Decemberwere all males attendingthe lek regularly. Highestdisplay intensities in guy (numberof tocksper minute)were attained at the startof the lekkingseason and declinedslightly thereafter, as in superciliosus. Thedaily patterns of lek activityof individualmale guy fell intolate morningand afternoonperiods as in superciliosus,but no data are availablefor guy before 0630, so an early morningactivity period is conjectural.Most adult male guy regularlyattended the lek from late Decemberthrough May or June. P. guy and superciliosusdiffered greatly in the timing of the annualmolt. In guy the molting seasonfollowed the lekkingseason with little or no overlap;adult malesleft the lek for the year as they startedto molt. The slow molt and possiblyunique timing mechanismof superciliosus(Stiles and Wolf 1974) resultin extensivemolt-breeding 68 ORNITHOLOGICAL MONOGRAPHS NO. 27 overlapand correlatewith a somewhatlonger lekking seasonthan in guy (8-9 versus6' months). Most young guy malesfirst appearedon the lek in April or May and moved into central territoriesas the adultsleft; their behaviorclosely resembled that of young male superciliosus.At this time they were in fresh plumage, and Snow concludedthat they had just completeda molt and must, therefore,be in their secondseason. From our experiencewith superciliosus,including aging of young malesby bill corrugations,we think it more probablethat the fresh plumageof youngmale guy was the first-yearplumage, and that thesemales were only a few monthsold. B. Snowstated that theseyoung males went througha secondseason in this sameplumage. If our interpretationof their agesis correct,these young male guy wore their juvenalplumage for 15-18 monthsand molted annuallythere- after. Preciselysuch a molt schedulehas been documentedin P. idaliae (Ruschi 1967). In P. superciliosusthe juvenal plumagealso is worn for slightlylonger than any succeedingfeather coat (Stiles and Wolf 1974). B. Snow (1974) distinguishedthree other plumagesof youngmales, which she interpretedas differentyear classes.These classes appear to us to intergradeand may even representonly a single,variable second-yearplumage. Thus the adult male plumagein P. guy is attained somewherebetween the secondand fourth completeannual molt; this is the most elaborateplumage sequence yet known for the genusPhaethornis. Accordingto B. Snow,annual turnover of lek malesin P. guy was 10% or less, a far lower figure than we found for P. superciliosus.Part of this differencecould reflect inherentbiases in our respectivemethods. The use of territory location and song type to identify individualscertainly is unsafe in P. superciliosusand might well underestimateturnover in P. guy. On the other hand, our marking systemcould have reducedsurvivorship in P. superciliosus.However, most mortality occurredwhen we handledthe birdsleast, and we excludedfrom the analysis all individualsthat we felt mighthave been adverselyaffected by the tags. Con- ceivably,the tag couldweaken a bird just enoughto affect its survivorshipover the lean season,but such effectswere probablyvery minor at best. Each tag weighedless than 0.1 g, far less than a crop of P. superciliosuswhen filled with nectar (Hainsworthand Wolf 1972b), and the plasticwas too stiff and short to becometangled easily in vegetation. The tags may interfere with nesting by females (Waser and Calder 1975), but our data on survivorshiptreat only lek males. Even allowingfor thesepossible errors of estimation,we feel that a real demo- graphicdifference exists between P. guy on Trinidad and P. superciliosusat La Selva, reflectingdiffering rates of both reproductionand mortality. The longer breedingseason of P. superciliosusmakes possible more nesting attempts. Indeed, there is time for up to three successfulnestings in the breedingseason of P. superciliosus,two in the breedingseason of P. guy. Thus the annualreproductive output of the former could be as much as 50% greater than the latter, althoughthe differenceprobably is less due to the low nestingsuccess in both populations. Regardingmortality, at leastone important predator on adulthummingbirds (Stiles 1978b), the Tiny Hawk (Accipitersuperciliosus), is absentfrom Trinidad (ffrench 1973). However, most other potential predators, including snakes, are well representedon Trinidad (D. Snow and B. Snow 1964). More important than LONG-TAILED HERMIT HUMMINGBIRDS 69 differencesin predationbetween the two areasmay be differencesin the seasonal distributionof nectar resources.In Costa Rica the seasonof low nectar availability from late Octoberto early Decemberwas very severe,with no abundantfood plant availableover mostof La Selva. The relativelyshort hiatus between the blooming seasonof the two major food plants of P. guy on Trinidad was bridgedby the bloomingof a third species(B. Snow1974). To judgefrom the numberof feeding recordsfor the latter, it was fairly abundant(B. Snow and D. Snow 1972). With- out the high mortality during October and November, annual survivorshipin superciliosusat La Selva would approachthat claimed by B. Snow for P. guy on Trinidad. If one acceptsB. Snow'sestimates of turnover in P. guy and considersthe differencein reproductiveoutput of the two populationsto be 50% or less,then there mustbe a relativelygreater surplus of youngmales produced each year in P. guy that do not obtain lek territories. A possibleindication of this is the high frequencyof "female" visits to P. guy leks. Snow did not considerthe fate of suchmales, or of the demographicconsequences of such a low recruitmentrate. The presenceof numerousnonbreeding males may limit reproductionof the populationby depletingnectar supplies that otherwisecould be availableto females. It would be most useful to know the interval between successive clutches in females of the two populations,but suchdata are virtually impossibleto obtain in such a dense habitat.

EVOLUTION OF LEK BEHAVIOR IN HUMMINGBIRDS

TWOconditions seem necessary for the evolutionof lek behaviorin any organism: (1) the emancipationof the male from participationin the reproductiveeffort beyondinsemination and (2) sufficienttime beyond that requiredfor maintenance to control a lek territory (D. Snow 1962, 1963). A variety of animal species, includingvertebrates and invertebrates,meet the first condition.For birds and mammals,Orians (1969) suggestedecologic and demographicsituations in which to expectmale emancipation.Among hummingbirds,almost all speciesmeet the prerequisitesfor the evolutionof lek behavior,apparently related to the fact that the clutchis two eggsthroughout the family (Van Tyne and Berger 1959; Orians 1969). However, severalinstances of apparentlysecondary reestablishment of male aid in reproductionhave been discoveredrecently (Wolf and Stiles 1970; B. Snow 1973a). Male aid, howeverindirect, is probablyincompatible with the time and energydemands of a lek socialsystem. One thereforeshould not expect leks to occurin speciesin which the femalescan producemore offspringwith aid from the male than without it (cf. Orians 1969). Once emancipatedfrom the reproductiveeffort, male hummingbirdsare under selectionchiefly to increasemating frequency and reducemaintenance costs. Males of manyspecies combine feeding and matingby holdingflower-centered territories duringthe breedingseason. Generally, male hummingbirdsare dominantto females and controlthe bestfeeding areas, so femalesattempt to forageon male territories. Most matingsare initiatedon theseterritories as a resultof foragingby females (Wolf and Stiles 1970; Stiles 1973; Wolf 1975a). The number of flowers a male can control on such a territory affectsnot only his energybudget but probably 70 ORNITHOLOGICAL MONOGRAPHS NO. 27 his mating successas well; the flowers in effect constitute a secondarysexual character (Wolf and Stiles 1970; Stiles 1973). These mating and feedingterritories and lek territories,representing complete coincidenceand completeseparation of defendedmating sitesand feedingareas, obviouslyare but the ends of a continuum. Within a single speciessome males controlenough flowers on territoryto satisfyall of their energyrequirements, and othersthat hold fewer flowers are requiredto forage outsideof their territories for varying periods(Stiles 1971a, 1973). In fact, in Calypte anna of California, changesin the dispersionof flowers can producelocal social systemsthat run the entiregamut from food-centeredmating territories to essentiallyleks. In the latter case, males hold flowerlessterritories in chaparral vegetationand commute to Eucalyptus trees to feed (Stiles 1973). In view of time and energy,the optimumstrategy for a male wouldbe to hold a food-centeredmating territory rich enoughto support him for a day. This would enablehim to forageefficiently and to be presentcontinuously so as not to missvisits by females. Whether a male can hold such a territory dependson the abundanceand dispersionof suitable flowers, the number of competing hummingbirds,the male's dominancestatus among his conspecifics,and the positionof his speciesin the interspecificdominance hierarchy (Stiles and Wolf 1970). When it is consistentlyimpossible for males of a given speciesto hold such rich breeding-feedingterritories, selection may favor the evolutionof lek matingsystems. Certain advantagesof flower-centeredterritories with regardto the criteria available to femalesfor male choice are also offered by lek systems, and thesemay not be offeredby otheralternative strategies, such as malesdisplaying solitarily.These include the fixed,constant locations of male matingstations and an indexof male'sdominance status independent of his personalcharacteristics -•his positionin an activity gradient. Lek social systemsand food-centered territoriesshould thus representadaptive peaks, and one may expectthat more hummingbirdspecies of a given area show thesetypes of mating systemsthan other, intermediatetypes. At La Selvaat least, this predictionis upheld. At least 2 species(Chalybura urochrysia and probably Amazilia tzacatl) show food-centered matingterritories, while 6 ( Eutoxeresaquila, Phaethornis superciliosus and longuemareus,Threnetes ruckeri, Klais guimeti,and ,,I mazilia amabilis) form leks. Glaucis aeneamay hold individual territories (cf. B. Snow1973a). The matingsystems of the other commonresident species, Florisuga mellivora and Thaluraniafurcata, are unknown. The criticaldeterminant of the adaptivepeak toward which a species may evolveappears to be whetherbreeding males can controlrich feedingareas which then serve as mating stations. To understandthe evolution of lek behavior it is thus important to identify factors that may prevent males of a given speciesfrom holding food-centered matingterritories. Lek-forming Costa Rican hummingbirdsinclude (1) subordinatespecies in the communitythat are forcedfrom defensiblefood supplies by dominantspecies and (2) speciesthat havespecialized on a food supplywhose spatialand/or temporaldistributions make it nondefensible(Brown 1964; Gill and Wolf 1975; Wolf et al. 1975). Included in the latter category are very large, dominantspecies that useresources defended by subordinatespecies but for which sufficientlylarge concentrationsof flowersto justify territorialityseldom or never LONG-TAILED HERMIT HUMMINGBIRDS 71 occur. Speciesin this categoryinclude Phaeochroa cuvierii (Skutch 1964b; Wolf 1970, pers. obs.), Campylopterushemileucurus (pers. obs.) and, in Guyana, Topaza pella (Davis 1958). The lek-forminghermits, as a group,fit into the categoryof subordinatespecies (Stiles 1975, presentstudy). The two nonhermitsat La Selva that form leks, Klais guimeti and Amazilia amabilis, are small, subordinatespecies that are excludedfrom mostrich nectarsources (pers. obs.). Suchspecies are forcedto exploita wide rangeof food plants,many of whichgrow in nondefensiblespatial arrangementsor haveinsect-pollinated flowers that producevery little nectar(Wolf 1970; B. Snowand D. Snow 1972; Hainsworthand Wolf 1972a; Wolf et al. 1976). It is importantto note that the two categoriesabove are not exclusiveand that subordinatestatus with exclusionfrom rich nectarsources and specializationfor feedingat nondefensiblefood plantscan evolveconcurrently. This clearly has occurredto a high degreein the hermits,including P. superciliosus,and has led to the coevolutionof hermitsas route-foraging,nonterritorial pollinators, and a groupof flowersecologically and morphologically specialized for hermitpollination (B. Snowand D. Snow1972; Wolf et al. 1972; Linhart 1973; Stiles1975). Clearly the exploitationsystems of hummingbirdsare intimately related to their social systems;the two are tightlyintegrated in the ecologyof any hummingbirdspecies. We contendthat the evolutionof leks in P. superciliosusand other hummingbirds cannotbe adequatelyunderstood without detailedconsideration of selectiveforces arisingfrom the foragingstrategy, interspecies relations, molt, and resourcedis- tribution in time and space.

ACKNOWLEDGMENTS This studywas supportedby a Chapman-NaumbergPostdoctoral Fellowship from the American Museum of Natural History to Stiles, and National Science Foundationgrants GB-7611, 19200, 40108, and 28956X to Wolf. We are grateful to the CostaRican officeof the Organizationfor Tropical Studies(O.T.S.) especially J. Campabadal,for logisticalsupport at La Selva. Help in the field was providedby P. Campanella,F. R. Hainsworth,L. Hurd, L. F. Kiff, D. Lyon, S. M. Smith, and the participantsof severalO.T.S. courses.Weather data for La Selvawere providedby R. Chavarda and G. Hartshorn. Helpful discussions and commentson the manuscriptwere suppliedby J. Bradbury,P. J. Campanella, D. H. Janzen,P. Marler, G. H. Orians,S. M. Smith,D. and B. Snow,R. H. Wiley, and D. Windsor. To all of thesepeople and institutionswe expressour sincere thanks.

SUMMARY

The ecologyand lek mating behaviorof the Long-tailedHermit, Phaethornis superciliosus,were studiedover a four-yearperiod at Finca La Selva, in the wet Caribbeanlowlands of Co.staRica. The sexesof this hummingbirdare alike in plumagebut differ in measurements;all lek residentsare males. Most of our observations were made on color-marked birds. Densityof P. superciliosusleks at La Selvawas aboutone per km2. Four leks had 5-6 to 20-25 malesduring the studyperiod. In undisturbedforest, leks were 72 ORNITHOLOGICAL MONOGRAPHS NO. 27 located in dense thickets along streams,which provided convenientflyways as well as habitat for importantfood plants. Lek territorieswere mating stations only,never containing enough flowers to affectthe male'senergy budget. A territory consistedof a male'ssong perches and the area betweenthem. The foci of lek activitywere the songperches. Territory size varied directlywith the opennessof the vegetation. Songwas the major meansof territorial advertisement;males on territory were rarely in visualcontact and spentmost of their time singing.Distinct song types or dialectsoccurred in P. superciliosus;males singing a given dialectgenerally occupieda geographicalsubunit of the lek, and the numberof subunitsvaried directlywith lek size. In spite of turnoverof individualmales, these subunits tendedto retain their integrityfrom year to year; maleson particularterritories sangthe samedialects as their predecessors. Visual displayswere givenonly when two birds were in closecontact, usually at a songperch. The conspicuousfacial stripes,orange mouth-lining and lower mandible,and perhaps the long,white-tipped tail werethe mostimportant morphological structuresused in visual displays,some of which were associatedwith distinctivesounds. Allowing for differencesin colorationof head and bill, the displaysof P. superciliosusare fairly similarto thoseof otherhermits. No regular sequenceof displayswas evidentin all interactions,but certaindisplays tended to precedeor follow otherswith someconsistency. The displaysand aggressivenessof an intrudingmale peakedwhen he first invadedanother's territory. During the ensuingexchange of displays,the resident becameprogressively more active, in mostcases finally driving the intruderfrom the territory. We found no differencesin displaysgiven to male and female intruders,and observedhomosexual as well as heterosexualcopulations. A female apparentlysignaled her sexby simplysitting still andallowing the maleto mount; a male treatedanything that sat still as a willing female. Mating sequenceswere alwaysinitiated in the males'territories, although copulation might take place elsewhere. Lek malesforaged mostly within 200-500 m of the lek, dependingon the distributionof suitableflowers. The remaining areas were used mainly by nonresidents,mostly females; all nestswere found well awayfrom establishedleks. P. superciliosusforaged mainly at flowerswith long, curvedcorollas that were usedinfrequently by nonhermithummingbirds that were dominantto hermits andusually displaced them at flowers.Individual P. superciliosusvisited a series of small, nondefensibleclumps of flowers,apparently along a regularforaging route. The mostimportant food plants of P. superciliosusat La Selvawere several speciesof Heliconia,especially H. pogonantha,and Costus. The lekkingseason of P. superciliosuscommenced abruptly in late November or December,most males returning to the lek withina few weeks.Onset of lekking wasclosely correlated with the onset of goodbloom by H. pogonantha.Lek activity peakedby January,and high levels were maintained through June or July. The nestingand lekking seasons corresponded closely, but nestingdid not peakuntil Februaryor March, as otherdry-season flowers began to bloomprofusely. A sharpdrop in lekkingand nestingoccurred in Augustwhile flowerswere still abundant,perhaps reflecting selection against late nestings.Flower availability LONG-TAILED HERMIT HUMMINGBIRDS 73 declinedprecipitously after late August,and the annualminimum of lekkingand nestingcorresponded with a periodof extremescarcity of flowersin Octoberand November. Duringmost of the breedingseason, daily peaksof lek activityoccurred in early morning,late morning,and midafternoon.Males arrivedon the lek at dawnbefore feeding,and sangand interactedvigorously for 30-45 min. An hour of concentrated, synchronousforaging followed. The late morningincluded intense singing and lessovert aggressiveactivity, and in the afternoonmales did little more than sing. This declinein active territorial aggressionthrough the day paralleledthe declinein nectarsecretion of major food plants,and the early morningforaging periodcorresponded to the time of maximumnectar availability. Most of the rare, brief visits to the lek by females occurredduring the late morningactivity period, when all observedcopulations also occurred. We estimate that only about one female per day visited a lek of 15 malesfor the purposeof mating,probably a resultof the long, asynchronousbreeding season. Males therefore had to be ready for monthsto mate at an instant'snotice, yet the number of matingsper male was relativelylow. This may explain the males' willingnessto mate with anythingcompliant and their frequent copulationswith dead leaves. Young maleshatched early in the breedingseason began to appear on the leks in March, but did not becomenumerous until May or June. At first theseyoung malessang hesitantly, with no fixed dialect and from any positionon the lek, and they were frequentlydisplaced by the residents.A few young.males established peripheralterritories by May or June, but most did not becometerritorial until July when adults began leaving the lek for the year. Residentswere dominant until they departed. By Augustmost territorial, singingmales on the lek were youngof the year. No singlefactor could account for the departureof the residents; molt played a role in some birds but not in others. The decline of flowers on a bird's foragingroute may also have been important;birds that had to find new feeding areas may have had less time to spendon the lek. The annualturnover of lek residentsin the La Selvapopulation of P. superciliosus wasabout 50%, far higherthan reportsfor otherneotropical birds, including P. guy in Trinidad. About 75% of the annual mortality occurred during October and Novemberwhen flowerswere scarce. Survivorshipof adult and first-year males was similar, perhapsreflecting the pronouncedyear-to-year variations in flower availabilitythat decreasethe valueof previousforaging experience to adults. Nearly all recruitmentof new residentswas from the ranks of yearlingmales. The more experiencein territorial behaviora young male had during his first season,the betterhis chancesof becominga residentthe followingyear. Young maleshatched late in the seasonand with little previouslek experiencealso had a chance to becomelek residentsdue to the high mortality of older yearlingsor adults. The mostdominant residents on a P. superciliosuslek occupiedterritories near the center,with subordinateindividuals towards the periphery. Central territories were the moststable in time and spaceand the moststrongly contested. Residents of the previousyear returning to the lek nearly always reoccupiedtheir old territoriesor vacantterritories more toward the centerof the lek. Young males returningat the start of their first full lekkingseason were able much lessoften to reoccupythe siteson whichthey were territorialor interactedas yearlingsbut 74 ORNITHOLOGICAL MONOGRAPHS NO. 27 usuallyhad to settle in more peripheralsites. Most movementtowards the lek centeroccurred during the lekkingseason as centralresidents died. However,the dominancestatus of some individualschanged little with age, even over several years. We foundno morphologicalcharacter that was consistentlycorrelated with dominance,although small, short-billedmales often failed to obtain territories. Becausethey were often in denservegetation with defendedboundaries on all sides,central territoriestended to be smallerand closertogether than peripheral territories.Central males sang more than peripheralmales during the late morning when femaleswere likely to appear,in part becauseperipheral males were per- secutedmore by centralmales. Thesefactors produced a gradientin activity and overallsong volume per unit areafrom the lek centertowards the periphery. Male P. superciliosustried to hold territorieswhere activity levels were high. If, as seemslikely, femalesprefer to matewith the mostdominant males, then their most reliablecue to a male'sstatus may be his positionin the lek activitygradient. In general,the dominance/fitnessgradient on P. superciliosusleks appearsto be muchless steep than that on leks of certaintetraonids and shorebirds.The proportionof all matingsaccounted for by central males is probablyless, and subordinatemales have some chance of mating. Unless other factors intervene, anymating advantage at the centershould set up geneticfeedback driving a hermit- typelek towardmore structured, grouse-type conditions. In P. superciliosusthese factorsinclude the high annualturnover of residentsand the generalrarity and unpredictabilityof mating. Nearlyall hummingbirdssatisfy the basicprerequisites for the evolutionof leks: emancipationof the male from the nestingeffort and a highly predictablefood source(flowers). The most energeticallyefficient mating strategy probably is to controla rich flower supplyon the matingterritory. Leks tend to evolvein hum- mingbirdswhen this is impossible,either because defensible flowers are controlled by more dominantspecies, or becausesufficiently rich, defensiblenectar sources do not occurin the species'habitat. As a group, the hermits,including P. superciliosus,fall into the first category.They have concurrentlyevolved a high degree of specializationfor exploitingnondefensible flowers, and manysuch flowers have coevolvedmorphological and ecologicalspecializations promoting hermit pollination. Leks and flower-centeredmating territoriesrepresent the ends of a continuumof matingtypes in hummingbirds.At La Selvamost species show one extrememating system or the other;relatively few speciesare intermediate.The two extremestrategies may well representadaptive peaks.

LITERATURE CITED

ALEXANtmR,R.D. 1975. Natural selection and specialized chorusing behavior in acoustical insects. In: Pimentel, D. (ed.). Insects, Science, and Society. Academic Press, New York. Pp. 35-77. ARe, W. 1957. Observacionessobre el comportamiento en grupo del Phaethornis longuemareus (Aves-Trochilidae). Bol. Soc. Venezolana Hist. Nat. 17: 156-168. BROWN, J. L. 1964. The evolution of diversity in avian territorial systems. Wilson Bull. 76: 160-169. BUECH•ER,H.K. 1961. Territorial behavior in Uganda kob. Science 133: 698-699. , AND H. D. ROTH. 1974. The lek system in Uganda kob antelope. Amer. Zool. 14: 145-162. LONG-TAILED HERMIT HUMMINGBIRDS 75

CAMPANELLA,P. J., ANDL. L. WOLF. 1974. Temporalleks as a matingsystem in a temperate zone dragonfly (Odonata: Anisoptera). I. Plathemislydia (Drury). Behaviour 51: 49-87. Co•)¾,M. L. 1968. Interspecificterritoriality among hummingbirdspecies. Condor 70: 270-271. DANIELS,G. S., ANDF. G. STILES. 1979. The Heliconia taxa of Costa Rica: keys and de- scriptions.Brenesia 15, suppl.1: 1-150. DAVIS, T. A. W. 1934. Notes on display in the hummingbirdsPhaethornis superciliosus (Linn.) and Pygmornis ruber (Linn.). Ibis 4: 732-738. 1958. The displaysand nestsof three forest hummingbirdsof British Guiana. Ibis 100: 31-39. DUELEMAN,W.E. 1967. Socialorganization in the mating calls of someneotropical anurans. Amer. Midl. Naturalist 77: 156-163. EMLEN, S.T. 1976. Lek organizationand mating strategiesin the bullfrog. Behav. Ecol. Sociobiol. 1: 283-313. FFI•ENCH,R. 1973. A guideto the birds of Trinidad and Tobago. Wynnewood,Pennsylvania, Livingston Publ. Co. FOG•)E•,M. P.L. 1972. The seasonalityand populationdynamics of equatorialforest birds in Sarawak. Ibis 114: 307-343. GILL, F. B., ANDL. L. WOLF. 1975. Economicsof feedingterritoriality in the golden-winged sunbird. Ecology 56: 333-345. GILLIARD,E.T. 1962. On the breedingbehavior of the Cock-of-the-Rock(Aves, Rupicola rupicola). Bull. Amer. Mus. Nat. Hist. 124: 31-68. HAINSWORTH,F. R., AN•)L. L. WOLF. 1972a. Energeticsof nectarextraction in a small,high altitude,tropical hummingbird, Selasphorus flammula. J. Comp.Physiol. 80: 377-387. , AN•)-- 1972b. Crop volume, nectar concentrationand hummingbird energetics. Comp. Biochem. Physiol.42A: 359-366. HARTZLER,J.E. 1974. Predationand the daily timing of SageGrouse leks. Auk 91: 532- 536. HOGAN-WAI•mmG,A.J. 1966. Socialbehavior of the Ruff, Philomachuspugnax (L.). Ardea 54: 109-229. HOLDRIDGE,L. R., W. C. GRENKE,W. H. HATHEWAY,T. LIANG,AND J. A. TOSI, JR. 1971. Forestenvironments in tropicallife zones: A pilot study. New York, PergamonPress. JANZEN,D.H. 1971. Euglossinebees as long-distancepollinators of tropicalplants. Science 171: 203-205. JEHL,J. R., JR. 1970. Sexualselection for size differencesin two speciesof sandpipers. Evolution 24: 311-319. KRUIJT,J.P., ANDJ. A. HOGAN. 1967. Social behavior on the lek in the black grouse, Lyrurus tetrix tetrix (L.). Ardea 55: 203-240. , C. J. DEVOs,AND I. BOSSEMA.1972. The arena systemof Black Grouse,Lyrurus tetrix tetrix (L.). Proc. 15th Intern. Ornithol. Congr.: 399-423. LACIC,D. 1968. Ecologicaladaptations for breedingin birds. London,Methuen and Co., Ltd. LEUTHOLD,W. 1966. Variationin territorialbehavior of Ugandakob ,4denotakob thomasi (Neumann 1869). Behaviour 27: 214-257. LILL, A. 1966. Someobservations on socialorganization and nonrandommating in captive Burmesered junglefowl (Gallus gallus spadiceus).Behaviour 26: 228-242. 1974. Sexualbehavior of the lek-formingWhite-bearded Manakin (Manacusmana- cus trinitatis Hartert). Z. Tierpsychol. 36: 1-36. 1976. Lek behaviorin the golden-headedmanakin, Pipra erythrocephalain Trinidad (West Indies). Adv. in Ethology,Suppl. 18 to J. Comp. Ethology. LINHtmT,Y.B. 1973. Ecologicaland behavioraldeterminants of pollen dispersalin hummingbird-pollinatedHeliconia. Amer. Naturalist 107: 511-523. MEYE• •)E SCHAOE•qS•,R. 1966. The speciesof birds of South America and their distribu- tions. Narberth, Pennsylvania,Livingston Press. NICHOLSON,E.M. 1931. Communaldisplay in hummingbirds.Ibis 1: 73-83. 76 ORNITHOLOGICAL MONOGRAPHS NO. 27

ONIKI, Y. 1970. Nesting behavior of reddish hermits (Phaethornis ruber) and occurrenceof wasp cells in nests. Auk 87: 720-728. ORGANS,G.H. 1969. On the evolution of mating systemsin birds and mammals. Amer. Naturalist 105: 589-603. ORT•Z-CRESPO,F. I. 1972. A new method to separate adult and immature hummingbirds. Auk 89: 851-857. PARKER,G.A. 1970. The reproductivebehavior and the nature of sexual selectionin Scato- phaga stercoraria L. (Diptera: Scatophagidae). I and II. J. Anim. Ecol. 39: 185-228. PrrELr•, F.A. 1942. Territoriality and related problemsin North American hummingbirds. Condor 44: 189-204. , R. T. HOLMES, AND S. F. MACLEAN,JR. 1974. Ecology and evolution of social organizationin Arctic sandpipers.Amer. Zool. 14: 185-204. ROBEL,R.J. 1966. Booming territory size and mating successof the Greater Prairie Chicken (Tympanuchuscupido pinnatus). Anim. Behav. 14: 328-331. 1967. Significance of booming grounds of Greater Prairie Chickens. Proc. Amer. Phil. Soc. 111: 109-114. RUSCHI, A. 1950. O territ6rio e as ftreas de alimenta•fio e de nidifica•fio de Anisopterus pretrei (DeLattre and Lesson), observadasatrav•s algumasgera•6es. Bol. Mus. Biol. Prof. Mello-Leitfio 8: 1-20. 1961. Algumas observacoessobre: Phaethornis yaruqui yaruqui (Bourcier); etc. Bol. Mus. Biol. Prof. Mello-Leitfio 27: 1-21. 1967. A plumagem e a muda em Phaethornis idaliae. Bol. Mus. Biol. Prof. Mello- Leitfio 33: 1-5. SCHoENER,T.W. 1971. Theory of feeding strategies.Ann. Rev. Ecol. Syst. 2: 369-404. SIm.EY, C.G. 1957. The evolution and taxonomic significanceof sexual dimorphism and hybridization in birds. Condor 59: 166-191. SKUTCH,A.F. 1951. Life history of Longuemare'sHermit Hummingbird. Ibis 93: 180-195. 1958. Life history of the Violet-headedHummingbird. Wilson Bull. 70: 5-19. 1964a. Life histories of hermit hummingbirds. Auk 81: 5-25. 1964b. Life history of the Scaly-breastedHummingbird. Condor 66: 186-198. 1966. A breedingbird censusand nestingsuccess in Central America. Ibis 108: 1-16. 1967. Life histories of Central American highland birds. Publ. Nuttall Ornithol. Club No. 7. Si•UD, P. 1960. The birds of Finca "La Selva," a tropical wet forest locality. Bull. Amer. Mus. Nat. Hist. 121: 49-148. 1964. The birds of Costa Rica: distribution and ecology. Bull. Amer. Mus. Nat. Hist. 128: 1-430. SMITrI, S. M. 1976. A field study of dominancehierarchies in Black-cappedChickadees. Auk 93: 95-107. SNow, B.K. 1972. A field studyof the calfbird Perissocephalustricolor. Ibis 114: 139-162. 1973a. Socialorganization of the hairy hermit Glaucishirsuta. Ardea 61: 94-105. 1973b. The behavior and ecologyof hermit hummingbirdsin the Kanaku Moun- tains, Guyana. Wilson Bull. 85: 163-177. 1974. Lek behaviourand breedingof Guy's Hermit Hummingbird Phaethornisguy. Ibis 116: 278-297. , ANDD. W. SNow. 1972. Feedingniches of hummingbirdsin a Trinidad valley. J. Anim. Ecol. 41: 471-485. SNow, D.W. 1962. A field studyof the Black and White Manakin, Manacusrnanacus, in Trinidad. Zoologica 47: 65-104. --.. 1963. The evolution of manakin displays. Proc. 13th Intern. Ornithol. Congr.: 553-561. 1968. The singingassemblies of little hermits. Living Bird 7: 47-55. , ANDB. K. SNOW. 1964. Breedingseasons and annualcycles of Trinidadland-birds. Zoologica 49: 1-39. LONG-TAILED HERMIT HUMMINGBIRDS 77

, •,NI>-- 1973. The breeding of the Hairy Hermit Glaucis hirsuta in Trinidad. Ardea 61: 106-122. STILES,F.G. 1971a. Time, energy, and territoriality of the Anna Hummingbird (Calypte anna). Science 171: 818-821. 1971b. On the field identification of California hummingbirds. California Birds 2: 41-54. 1973. Food supply and the annual cycle of the Anna Hummingbird. Univ. Cali- fornia Publ. Zool. 97: 1-109. 1975. Ecology, flowering phenology,and hummingbird pollination of some Costa Rican Heliconia species.Ecology 56: 285-302. 1978a. Temporal organizationof flowering among the hummingbirdfood plants of a tropical wet forest. Biotropica 10: 194-210. 1978b. Possible specialization for hummingbird-hunting in the Tiny Hawk. Auk 95: 550-553. , AND L. L. WOLF. 1970. Hummingbird territoriality at a tropical flowering tree. Auk 87: 467-491. , AND- 1973. Techniques for color-marking hummingbirds. Condor 75: 244- 245. , riND -- 1974. A possiblecircannual molt rhythm in a tropical hummingbird. Amer. Naturalist 108: 341-354. VnN RmJN, J. 1973. Behavioral dimorphism in male Ruffs, Philornachus pugnax. Behaviour 47: 153-229. VAN TYNE, J., AND A. J. BERGER. 1959. Fundamentals of ornithology. New York, John Wiley and Sons, Inc. WARt>,P. 1969. The annual cycle of the Yellow-vented Bulbul Pycnonotus goiavier in a humid equatorial environment. J. Zool. 157: 25-45. WASER,N.M., AND CALDER,W.a. 1975. Possible impairment of nest-building of hummingbirds by acetateleg tags. Condor 77: 361. WILEY, R.H. 1971. Song groups in a singing assemblyof Little Hermits. Condor 73: 28-35. 1973. Territorially and non-random mating in Sage Grouse, Centrocercus uro- phasianus. Anim. Behav. Monogr. 6: 87-169. 1974. Evolution of social organization and life-history patterns among grouse. Quart. Rev. Biol. 49: 201-227. WOLF,L.L. 1969. Female territoriality in a tropical hummingbird. Auk 86: 490-504. 1970. The influence of seasonalflowering on the biology of some tropical hummingbirds. Condor 72: 1-14. 1975a. "Prostitution" behavior in a tropical hummingbird. Condor 77: 140-144. 1975b. Female territoriality in the Purple-throated Carlb. Auk 92: 511-522. , F. R. HAINSWORTH,AND F. B. GILL. 1975. Foraging efficiencies and time budgets in nectar feeding birds. Ecology 56: 117-128. --, --, ANDF. G. STILES. 1972. Energetics of foraging: rate and efficiency of nectar extraction by hummingbirds. Science 176: 1351-1352. --., ANDF. G. STILES. 1970. Evolution of pair cooperationin a tropical hummingbird. Evolution 24: 759-774. , --, AND F. R. HAn•SWORTH. 1976. Ecological organization of a highland, tropical hummingbirdcommunity. J. Anim. Ecol. 45: 349-379. , ANDJ. S. WOLF. 1976. Mating system and reproductive strategy of malachite sun- birds. Condor 78: 27-39. WYNNE-EDWARDS,V.C. 1962. Animal dispersionin relation to socialbehaviour. Edinburgh, Oliver and Boyd. 78 ORNITHOLOGICAL MONOGRAPHS NO. 27

Heliconia speciesdesignated by taxon number in Tables 4-7, Figure 20, and elsewhere in the text are describedin a recentpublication (Daniels and Stiles 1979) that appearedtoo late for the names to be changed here. Names assignedare: H-3 •- H. mathiasii; H-16 =H. umbrophila; H-17 : H. sarapiquensis;H-18 =H. irrasa. H. sublata of Figure 9 = H. mathiasii.--Auth. No. 21. SocialOrganization and Behaviorof the AcornWoodpecker in Central Coastal California, by Michael H. MacRoberts and Barbara R. Mac- Roberts.1976. $7.50 ($6.00 to AOU members). No. 22. MaintenanceBehavior and Communicationin the BrownPelican, by RalphW. Schreiber.1977. Price $6.50 ($5.00 to AOU members). No. 23. SpeciesRelationships in the AvianGenus Aimophila, by Larry L. Wolf. viii + 220pp.. 17text figures+ 10plates, long-play phono disc album. 1977.Price $12.00($10.50 to AOU members). No. 24. Land Bird Communities of Grand Bahama Island: The Structure and Dynamicsof an Avifauna,by JohnT. Emlen. xi + 129pp., 38 text figures,appendix. 1977. Price $9.00 ($8.00 to AOU members). No. 25. Systematicsof SmallerAsian Night Birds Based on Voice,by JoeT. Mar- shall.viii + 58 pp., frontispiece,15 plates, phono disc supplement, ap- pendicesI-III. 1978.Price $7.00 ($6.00 to AOU members). No. 26. Ecologyand Behaviorof the PrairieWarbler Dendroica discolor, by Val Nolan,Jr. xx + 595pp., colorfrontispiece, 42 text figures, 8 appendices. 1978. Price $29.50. No. 27. Ecologyand Evolutionof Lek MatingBehavior in theLong-tailed Hermit Hummingbird,by F. GaryStiles and Larry L. Wolf.viii + 78 pp., 26 text figures. 1979. Price $8.50 ($7.50 to AOU members).

Likeall other AOU publications, Ornithological Monographs are shipped prepaid. Make checkspayable to "The AmericanOrnithologists' Union." For the convenienceof those whowish to maintaincomplete sets of OrnithologicalMonographs and to receivenew numbersimmediately upon issue, standing orders will beaccepted. Orderfrom: Glen E. Woolfenden,Assistant tothe Treasurer AOU, Department of Biology, Universityof SouthFlorida, Tampa, Florida 33620. ORNITHOLOGICAL MONOGRAPHS

No. A DistributionalStudy of the Birds of British Honduras,by StephenM. Russell. 1964. $7.00 ($5.50 to AOU members). A ComparativeStudy of SomeSocial Communication Patterns in the Pe- !ecaniformes,by GerardFrederick van Tets. 1965.$3.50 ($2.50 to AOU members). No. The Birdsof Kentucky,by Robert M. Mengel. 1965.$15.00 ($12.50 to AOU members). No. A ComparativeLife-history Study of Four Speciesof Woodpeckers,by Louise de Kiriline Lawrence. 1967. $6.00 ($4.50 to AOU members). No. Adaptationsfor Locomotionand Feedingin the Anhingaand the Double- crestedCormorant, by OscarT. Owre. 1967.$6.00 ($4.50 to AOU members). No. 7. A DistributionalSurvey of the Birds of Honduras, by Burr L. Monroe, Jr. 1968. $14.00 ($11.00 to AOU members). No. 9. Mating Systems,Sexual Dimorphism, and the Role of Male North Amer- ican PassefineBirds in the NestingCycle, by Jared Verner and Mary F. Willson. 1969. $4.00 ($3.00 to AOU members). No. 10. TheBehavior of SpottedAntbirds, by EdwinO. Willis. 1972.$9.00 ($7.50 to AOU members). No. 11. Behavior,Mimetic Songsand SongDialects, and Relationshipsof the Para- sitic Indigobirds(Vidua) of Africa, by Robert B. Payne. 1973.$12.50 ($10.00 to AOU members). No. 12. Intra-island Variation in the Mascarene White-eye Zosteropsborbonica, by Frank B. Gill. 1973.$3.50 ($2.50 to AOU members). No. 13. EvolutionaryTrends in the NeotropicalOvenbirds and Woodhewers,by Alan Feduccia. 1973. $3.50 ($2.50 to AOU members). No. 14. A Symposiumon the HouseSparrow (Passerdomesticus) and European Tree Sparrow(P. montanus)in North America,by S. CharlesKendeigh. 1973. $6.00 ($4.50 to AOU members). No. 15. FunctionalAnatomy and Adaptive Evolutionof the FeedingApparatus in the HawaiianHoneycreeper Genus Loxops (Drepanididae), by Lawrence P. Richards and Walter J. Bock. 1973. $9.00 ($7.50 to AOU members). No. 16. The Red-tailedTropicbird on Kure Atoll,by RobertR. Fleet. 1974.$5.50 ($4.50 to AOU members). No. 17. ComparativeBehavior of the AmericanAvocet and the Black-neckedStilt (Recurvirostridae),by Robert Bruce Hamilton. 1975. $7.50 ($6.00 to AOU members). No. 18. BreedingBiology and Behaviorof the Oldsquaw(Clangula hyemalis L.), by RobertM. Alison. 1975.$3.50 ($2.50to AOU members). No. 19. Bird Populationsof AspenForests in WesternNorth America,by J. A. DouglasFlack. 1976.$7.50 ($6.00 to AOU members). No. 20. SexualSize Dimorphism in Hawks and Owls of North America,by Noel F. R. Snyderand James W. Wiley. 1976.$7.00 ($6.00 to AOU members).

(Continued on inside back cover)