Intraspecific Variation in Elepaio Foraging Behavior in Hawaiian Forests of Different Structure

The Auk 111(4):917-932, 1994

INTRASPECIFIC VARIATION IN ELEPAIO FORAGING BEHAVIOR IN HAWAIIAN FORESTS OF DIFFERENT STRUCTURE

ERIC A. VANDERWERF t Departmentof Zoology,University of Florida,Gainesville, Florida 32611, USA

ABSTRACT.--Istudied intraspecificvariation in foraging behavior of an endemic, insectivorous bird, the Elepaio (Chasiempissandwichensis), in two Hawaiian foreststhat differed in degreeof human modification.The undisturbedforest had a closedcanopy, a denseunderstory, and a groundcover of native plants. The disturbed forest had much lower tree and shrub densities,and a ground cover of alien grasses.Search-and-attack rates, proportions of attackmaneuvers, and proportionalsubstrate use differed betweenhabitats. Birds in disturbed habitat attackedprey two-thirds as often as birds in undisturbedhabitat, hopped lessfre- quently, and flew farther and more often. They also did lessperch-gleaning and chasing, did more flight-gleaningand hawking,used small branches and the groundless often, and usedleaves and the air more often than birds in undisturbedhabitat. Disturbedareas may be lower-qualityforaging habitat because they require more difficult foragingmethods. Age was associatedwith variation in search-and-attackrates and proportionsof attackmaneuvers, but sexwas not. SubadultElepaio attacked prey lessoften than adults,searched more slowly, and usedsimpler maneuvers more often, possibly to compensatefor their lower proficiency. Log-linearanalysis showed that attackmaneuver was relatedto substrateand to tree species. Birdsperch-gleaned more often on twigs and in ohia (Metrosiderospolymorpha), hung more often on bark and in koa(Acacia koa), and flight-gleanedmore often on leaves.Elepaio showed muchflexibility in foragingbehavior and usedmore-diverse attack maneuvers and substrates than relatedcontinental species, which may allow Elepaioto exploitdisturbed habitats suc- cessfully.Received 2 August1993, accepted 11 January1994.

STUDIESOF AVIAN foraging behavior tradi- may not be valid (Martin 1986, Grubb and tionally have focusedon niche partitioningand Woodrey 1990, Martin and Karr 1990). Recog- community structureand, hence, have empha- nition of the importance of intraspecificvaria- sized differencesamong species(e.g. MacAr- tion has led to the discoverythat foraging by thur 1958, Holmes et al. 1979, Szaro and Balda an individual bird is influenced by numerous 1979,Wiens and Rotenberry1981, Alatalo 1982). biotic and abiotic factors: age (reviewed in Speciesmay be separatedby one or more of Wunderle 1991); sex (Selander 1966, Peters and severalniche components, including plant spe- Grubb 1983, Petit et al. 1990);morphology (Fitz- cies,substrate, height, horizontal position,for- patrick 1985, Sherry 1985, Gustafsson 1988, aging technique,and foraging speed(reviewed Moermond 1990); intraspecific dominance in Airola and Barrett 1985, Martin 1986, Schoe- (Hogstad 1988,Grubb and Woodrey 1990);hab- ner 1986). itat structure (Maurer and Whitmore ! 981, Rob- This emphasison interspecificvariation often inson and Holmes 1982, Saboand Holmes 1983); obscureddifferences in foraging within a spe- food distribution and abundance (Holmes and cies (but see, for example, Partridge 1976, Schultz 1988);season and stageof breeding cy- Holmes et al. 1978, Gustafsson1988). Moreover, cle (Robinson1986, Hejl and Verner 1990,Sakai modelsof communitystructure based on niche and Noon 1990);weather (Grubb 1978);and time partitioning and speciesoverlap assumethat all of day (Verbeek 1972). individualsof a speciesforage identically, which Most studiesof intraspecificvariation in avian foraging have concentrated on one or two of the above factors, but a foraging bird re- sponds simultaneously to a complex and dy- • Presentaddress: Department of Zoology,Univer- namic set of stimuli. Ideally, one should ex- sity of Hawaii, EdmondsonHall, 2538 The Mall, Ho- amine all factors and their interactions at once, nolulu, Hawaii 96822, USA. but this is difficult and often impractical(Grubb 917 918 EmcA. V•Nr•ERWœR• [Auk,Vol. 111

1979). A reasonablecompromise is to examine METHODS a subsetconsisting of severalfactors that may interact. For insectivorous forest birds, habitat Studysite.--I conductedwork from Februarythrough structureis perhapsthe mostimportant deter- July 1991at HakalauForest National Wildlife Refuge minant of foraging behavior (Robinson and on the islandof Hawaii. Thesite lies at approximately Holmes 1982), and several habitat variables con- 1,900 m elevation on the east side of Mauna Kea and stitute such a subset of related factors. is characterizedby sloping terrain, heavy rainfall (3 Habitat structure affects the abundance, dis- m/year), and daytimetemperatures rarely above 20øC. The natural vegetation type is montane rain forest. tribution, and perceptibilityof prey, and also For a detailed descriptionof the region, see Scottet which search-and-attack methods birds can em- al. (1986). ploy to captureprey (Fitzpatrick1980, Robinson Although the site is now protected, parts of the and Holmes 1982, Holmes and Schultz 1988). foresthave been extensivelymodified by human ac- Aspectsof habitat structure known to affect for- tivities during the last 100 years. Timber was har- agingof insectivorousforest birds include: plant vestedin someareas and large tractswere clearedfor species (Holmes and Robinson 1981, Franzfeb cattle ranching, resulting in a patchwork of highly 1983, Morrison et al. 1985); substrate (Jackson disturbed and relatively undisturbed forest. The 1979, Fitzpatrick 1980, Greenberg and Grad- boundary between these habitats is not sharp, and wohl 1980);structural characteristics of vege- even the "undisturbed" forest has been somewhat tation (Robinson and Holmes 1984, Whelan modified by the sameactivities. The undisturbed habitat consistedof closed-canopyforest with a moder- 1989); and foliage density (Maurer and Whir- ately denseunderstory and a groundcover primarily more 1981, Sabo and Holmes 1983). Valuable of native forbs and ferns. The disturbed habitat was insights on the effect of habitat structure on a shorter,open-canopy woodland with almostno un- foragingpatterns may be gainedby comparing derstoryand a ground cover of alien grassesintro- the behaviorof a single speciesin two habitats duced for cattle grazing. that differ in structure (Robinson and Holmes I selectedthree plots totaling 11.4ha in undisturbed 1982). Understanding effects of habitat struc- habitat and three totaling 10.6ha in disturbedhabitat ture on foragingmay be particularlyuseful in that representedextremes of the continuum from un- the conservationof specieswith restrictedhab- disturbedto disturbed.I choseplots in both habitats itat or specializedhabitat requirementsand in that were as close to each other as possible and at similar elevations(within 100 m) without including making managementdecisions that will alter areas that were intermediate in structure. habitat structure. Studyspecies.--The Elepaio is placedin a monotypic In Hawaii, human alteration of native habi- genus endemic to the Hawaiian Islands; its closest tats by physical disturbance, introduction of relativesare the monarchineflycatchers of Australasia alienspecies, and loss of nativespecies diversity and Oceania(Boles 1979, Pratt et al. 1987,Sibley and has been extensive (Kirch 1983, Stone and Scott Ahlquist 1990). Elepaio are common permanent res- 1985, Cuddihy and Stone 1990). Native Hawai- idents of both habitats at the study site, although ian birds appear to be especiallysensitive to population density is higher in undisturbedhabitat disturbance(Olson and James1984, Sakai 1988), (unpubl. data). They are monogamousand nonmi- and many specieshave restrictedranges or are gratory, remaining paired and territorial throughout the year (MacCaughey1919, Conant 1977, Berger 1981). less abundant in disturbed areas (Scott et al. Elepaio are insectivorous(Munro 1960, Conant 1977, 1986). The Elepaio (Chasiempissandwichensis) is VanderWerfunpubl. data),although nectarivoryhas an insectivorous bird endemic to the Hawaiian been reported (MacCaughey1919). Islandsthat inhabitsforests of varying structure I was able to sexand age Elepaioby plumage dif- and degree of human modification. ferences(MacCaughey 1919, Pratt 1980). Males have The goalsof my study were to: (1) compare black throat feathers narrowly tipped with white, foraging methodsused by Elepaio in undis- whereas females have broader white tips that often turbed and human-modified forests to under- makethe throat appearalmost completely white. Sub- standhow they forageand whether humandis- adult birdsof both sexesretain a uniformly drabgray- turbanceaffects their ability to exploit a habitat; brown plumage for at least one year. All subadult Elepaio I observedwere second-yearbirds and were (2) investigatediversity and degree of intra- no longer being fed by adults. specificvariation in Elepaioforaging behavior; I used four methodsto identify 48 individual Ele- (3) simultaneously examine effects of several paio in the study plots: a unique combination of col- habitat variableson foraging to evaluatepos- ored leg bands(n = 17); distinctiveplumage (n = 5); sible interactions between factors. being paired with a color-bandedor distinctively- October1994] ElepaioForaging Behavior 919 plumagedbird (n = 10);and known territory bound- ward hover-glean"and "strike" (Fitzpatrick 1980), aries (n = 16). Based on observations of banded birds, "hover," and "snatch" (Robinson and Holmes 1982, it was unusualto find a bird in a territory that was Recheret al. 1985,Sherry 1985). Hawk indicatesthat not its own, and suchan occurrencequickly provoked a bird flew out from a perch to captureprey that was an attackby the territory owner. If I observeda bird in the air. This is synonymouswith "sally" (Eckhardt at a territory boundary or if I was unsure whether a 1979),"aerial hawk" (Fitzpatrick 1980), and "flycatch" bird wasthe territory owner, I followed the bird until (Maurer and Whirmore 1981). Chase occurred when I couldconclusively determine its identity.I compiled a bird chasedprey that had been flushed, often in a data separatelyfor each individual. downwardflight. Other termsused for this maneuver Data collection.--For characterization of habitats, I include "flush-chase" (Robinson and Holmes 1982), useda stratifiedrandom design to select15 pointsin "pursue"(Sherry 1985),"tumble" (Root 1967),"flut- plots of each habitat and used the method of James ter-chase"and "flush-pursue"(Remsen and Robinson and Shugart (1970) as modified by Noon (1981) to 1990). quantify differencesin habitat structure.This method Attack maneuverswere directedat a variety of sub- is basedon 0.04-ha circlesin which the speciesand stratesthat I classifiedinto sevencategories: ground, diameter of all trees are recorded, and maximum can- trunk (including branches> 20 cm in diameter), large opy height is estimated.Shrub density, percent of branch(>5-20 cm), smallbranch (1-5 cm), twig (< 1 groundcoverin specifiedcategories, and foliageden- cm), leaf, and air. sity at various heights are measuredat 2-m intervals Analyses.--For continuous variables, including along north-southand east-westdiameters. search-and-attackrates, proportions of attack maneu- To evaluateforaging in Elepaio,I collecteddata vers,and proportionalsubstrate use, I calculatedav- fromMarch through July 1991 by regularlytraversing eragevalues for each individual from all sequences study plots and searchingfor birds. I followed each combined,and used individuals as independent data bird for aslong aspossible, recording the following points.Another commonlyused method is to calcu- informationduring eachobservation sequence: age, late ratesfor each sequenceand averagevalues from sex,and individualidentity of bird, hops,flights, dis- all sequences(e.g. Robinson and Holmes1982!. The tancesflown, prey-attacking maneuvers, substrates of method I used providesinformation on variability at attacks,tree species, height, and duration of sequence. the individual level and is not biased by short ob- I recordeddata on a portablecassette recorder and servationsequences, but doesnot allow examination transcribedthem later while usinga stopwatch.Most of temporal variation. sequenceswere lessthan 1 rain long, but a few were In the strict sense,treating individuals sampledat over 2 min. the same site as independent is a form of pseudore- I employed five variablesto evaluaterates at which plication (Hurlbert 1984). However, for a territorial Elepaiosearched for and attackedprey. Hop, flight, speciesin a heterogeneousenvironment, one would and attackintervals were the averagetimes between not expectall individuals to forage in the sameway hops,flights, and attacks,respectively. Flight distance becausethey haveaccess to different setsof resources was the averagelength of flights between perches. (Dodge et al. 1990).If the speciesis behaviorally plas- Attack radiuswas the averagedistance flown during tic, individuals may respond differently. The alter- attackmaneuvers that involvedflying. I usuallycould native of lureping observationsfrom all individuals not determine if an attack was successful,so attack at a site may be more statisticallyjustifiable, but may raterefers only to therate at whichprey was attacked, causeserious loss of biological information. Viewed not captured. from another perspective,I tested differencesin the Elepaioused attack maneuvers that I classifiedinto way Elepaioforage in differentterritories in two hab- five mutuallyexclusive categories defined below, with itats. equivalentterms from other authors.Perch-glean in- I performed six multivariate analysesof variance cludedall maneuversin which prey were taken from (MANOVA; Sokal and Rohlf 1981), one each with a solid substratewhile perched.This is equivalentto search-and-attack-ratevariables, proportions of attack "glean" (Eckhardt1979, Robinson and Holmes 1982), maneuvers,or proportional substrateuse as depen- "pluck" (Emlen 1977, Mountainspring1987), and dent variables,and habitat with either sex or age as "pick"(Reinsen 1985); it includesboth "simple perch- independent variables. I analyzed variation associ- ing" and "landing and perching" (Fitzpatrick 1980). ated with sex separatelyfrom age becausesubadult Hang occurredwhen a bird clung upside-downby birds couldnot be sexed.Values presented for habitat its feet to examine a substrate that could not be reached are those from analyseswith age, not sex, because while perching.Flight-glean I definedas taking prey sample sizes for analyses with sex were smaller. I from a solid substratewhile flying. I did not distin- required at least 20 attack maneuvers and 300 s of guish whether the maneuver was directed outward observationper individual for inclusionin analyses or upward from a perch,if the prey wason the upper (see Dodge et al. 1990, Martin and Karr 1990). or lower leaf surface, or if the bird hovered while In studies of proportions of attack maneuvers and flying. Therefore,this includes"upward" and "out- foraging substrates,authors disagreewhether it is 920 Emc A. VANDSRWEP,r [Auk, Vol. 111

TASLE1. Contingency table of Elepaio foraging ob- 1984) or a specifiednumber of observations(e.g. Pe- servationsclassified by habitat, tree species,sub- ters and Grubb 1983, Martin and Karr 1990). I used strate,and attackmaneuver. Cells containing a dash all observationsin a sequencebecause Elepaio often are impossible classificationsthat were treated as were difficult to locatein densefoliage and because structural zeros (see Methods). I wanted information about search-and-attack rates,

Maneuver which may be biased if only one observation is used (Fitzpatrick 1980). Tree Perch- Flight- I calculateddiversity of foraging maneuvers and species Substrate glean glean Hang Aerial substratesusing the Shannon-Weaver index, Undisturbed habitat H' = -SP,(log P,), (1) Ohia Bark 28 3 19 -- Twig 118 25 19 -- where P, is the proportional use of categoryi (Shan- Leaf 19 21 1 -- non and Weaver 1949). To facilitate comparison of Ground .... diversity indexes with studiesthat used different cat- Air ------22 egoriesor natural logarithms,I lumped categoriesnot Koa Bark 5 0 4 -- recognizedby other authorsor by me and scaleddi- Twig 4 2 1 -- versity as the equitability index, Leaf 0 2 0 -- Ground .... l' = H'/H'm•, (2) Air ------1 where H'• is the casewhere use of all categoriesis Other Bark 4 0 0 -- equal (Pielou 1966). Twig 15 6 0 -- Leaf 2 7 0 -- I investigated interaction of categorical variables Ground 10 0 -- -- using log-linear analysisof a multidimensional con- Air ------I tingency table (Fienberg 1981). This type of analysis is advantageousbecause it allows simultaneousex- Disturbed habitat aminationof all interactionsbetween categorical vari- Ohia Bark 16 8 11 -- ables (Schoener1970). In log-linear analyses,obser- Twig 85 73 17 -- Leaf 18 58 0 -- vations must be independent becausethe purposeis Ground .... to look for dependencyamong variables.Therefore, Air ------38 I used only the initial observation from each se- Koa Bark 0 I 2 -- quence.I cross-classifiedeach of 691 foraging obser- Twig 5 0 2 -- vations using four categorical variables; habitat, at- Leaf 0 4 0 -- tackmaneuver, substrate, and tree species.In practice, Ground .... it was necessaryto collapsethe data by combining Air ------2 some categoriesto reduce the number of empty or Other Bark 3 2 0 -- sparsecells, and becausea larger numberof categories Twig 4 I 0 -- resulted in fewer degreesof freedom, making it im- Leaf 0 0 0 -- possibleto evaluate modelswith many parameters.I Ground 2 0 -- -- combined "hawk" and "chase" as "aerial" maneuvers, Air ------I and all but the two mostcommon tree speciesas "other." I lumped "trunk" and "large branch" substrates as "bark" and grouped "small branch" with "twig." better to use all observationsin a sequenceor only The latter two lumpings probably did not causeloss the initial observation or a certain subset of each se- of information becausesubstrate interacted only with quence.Using sequentialobservations allows faster maneuver in the final model and these pairs of sub- data collectionand more information can be gathered strate categoriesmay have similar effectson maneu- from each bird, which may be important for species ver (Bishop 1971). The results was a 2 x 3 x 4 x 5 that are rare or hard to locate (Morrison 1984, Recher contingencytable (Table 1). and Gebski 1990). Nevertheless, sequential observa- Table I is "incomplete" becausesome cross-classi- tions may not be independentand, thus,may violate ficationsare not possible(Fienberg 1972). For exam- assumptionsof many statisticaltests (Morrison 1984, ple, "aerial" could never be classifiedwith "leaf" be- Bell et al. 1990,Hejl et al. 1990).Alternatively, initial causeby definition "aerial" is directed at "air." I treat- observationsmay be biased toward conspicuousma- ed cells representing impossible classificationsas neuvers, particularly for birds foraging in dense fo- structural zeroes. Expectedcell frequenciesare not liage where observationsequences typically are short calculatedfor structuralzeroes. They do not contrib- (Holmes et al. 1979,Bradley 1985).To avoid oversam- ute to goodness-of-fitstatistics, and degreesof free- pling from a few individuals, some authors recom- dom are decrementedby one for each(Fienberg 1972). mend truncating sequentialobservations after either The objectiveof multidimensional contingency-ta- a certainlength of time (e.g. Wagner 1981,Morrison ble analysisis to find the simplest model that ade- October1994] ElepaioForaging Behavior 921 quatelyexplains the data.I determined if a model was m), foliage densitywas higher in all height cat- adequate with a goodness-of-fittest based on the egories,and tree crown shapediffered between G-statistic(Sokal and Rohlf 1981). I found expected habitats (Fig. 1). Trees in undisturbed habitat cell frequenciesusing an iterative fitting algorithm had foliage concentratedtoward their tops, (Fienberg 1970) calculatedwith SYSTAT (Wilkinson whereas trees in disturbed habitat had rounded 1987).I rejectedany model where the probability of crowns with more foliage distributed toward fit was lessthan 0.05. To find the simplest adequate their middle. Ground cover in the undisturbed model,I useda stepwiseprocedure outlined by Good- man (1971). I startedby evaluating modelsthat con- habitat was primarily native forbs and ferns tained all interactionsof a given order of complexity (65%),with smallpatches of exoticgrasses (14%), (full models). The final model could be reached by particularly Kikuyu grass(Pennisetum clandestin- forward or backward selection from a full model; I um), velvet grass (Holcus lanatus), meadow tried both methods. Forward selection started with ricegrass(Microlaena stipoides), and sweet ver- the highest-orderfull model that did not fit the data nalgrass(Anthoxanthum odoratum). In disturbed and involved adding the most-significantterms of the habitat, these exotic grasseswere much more next-highestorder that did not causethe overall sig- common than native forbs (87 vs. 7.7%). Both nificance to be lower than 0.05. Backward selection involveddeleting the least-significantterms from the habitatsalso suffered ground-cover damage by lowest-order full model that did fit the data. ! cal- feral pigs and had areas of bare rock or soil culatedsignificance of a term using a conditionallog- without groundcover. likelihood ratio test(Fienberg 1981).Any term whose Variationin foragingassociated with habitat,age, conditional log-likelihood ratio had a probability of andsex.--Habitat was a significantfactor in ex- less than 0.05 was a significant component of the plaining variation in search-and-attack rates model. The final model was reached when no terms (MANOVA, n = 46, Fs,39= 25.14, P < 0.0001). could be added or deleted. Four of five search-and-attack variables differed I used the following notation to refer to variables between habitats(Table 2). On average,birds used in models: (H) habitat; (M) maneuver; (S) sub- in disturbedhabitat hopped less frequently, flew strate;and (T) tree species.Interactions between variables were representedby terms with symbolsfor fartherand morefrequently, and attackedprey two-thirds as often as birds in undisturbed hab- each variable (e.g. HM representsan interaction between H and M). Higher-order interactionsimplicitly itat. Only the attackradius was the samein both contained all lower-order interactions that were com- habitats.Age also was an overall factor in de- binations of the same variables (e.g. HMS includes termining search-and-attackrates (MANOVA, HM, HS, MS, H, M, and S). n = 46, F5,3•= 3.68, P < 0.005), but only two variablesdiffered betweenbirds of the two age RESULTS categories(Table 2). Subadultbirds hopped less frequentlyand attacked prey 25% less often than Habitat quantification.--Ohia (Metrosideros did adults. polymorpha)was the dominant tree speciesin Habitat and age also were significant factors each habitat (79.5% of total basal area in undis- in explainingoverall variation in proportionof turbed habitat, 74.3% in disturbed), and koa attack maneuvers (MANOVA, n = 46, Fs,3•= (Acacia koa) was also common (15.0% in undis- 25.80and 2.87,P < 0.001and 0.02,respectively). turbed, 25.2% in disturbed). Other trees were In comparisonsof individual attack maneuvers, sparselydistributed and were more commonin birdsin disturbedhabitat did lessperch-glean- undisturbed habitat: kolea (Myrsinelessertiana; ing and chasing,and more flight-gleaningand 2.1 vs. 0.4%), olapa (Cheirodendrontrigynum; 0.4 hawking. Subadultbirds did moreperch-glean- vs. < 0.1%), kawau (Ilex anomola;0.5 vs. < 0.1%), ing, and lesshanging and hawking (Table 2). pilo (Coprosmasp.; 0.2 vs. <0.1%), and standing Habitat was a significant predictor of sub- deadtrees of all species(2.2 vs. 0.1%).Common strate use, but age was not, although the sig- understoryplants in both habitatsincluded ak- nificance levels of these factors were not that ala (Rubussandwicensis), ohelo (Vacciniumcaly- different (MANOVA, n = 46, F7,•7= 2.37 and cinum),pukiawe (Stypheliatameiameiae), black- 2.11, P = 0.04 and 0.06, respectively).Small berry (Rubusargutus), and saplingsof tree spe- branchesand the ground were used more in cies listed above. Tree and shrub densities were undisturbed habitat, while leaves and the air higher in undisturbedhabitat (868vs. 198trees/ were used more in disturbedhabitat. Only leaf ha, and 1,633vs.150 shrubs/ha). Canopy height usediffered betweenages, with subadultsusing was higher in undisturbedhabitat (18.1 vs. 13.6 them more often (Table 2). 922 EPaCA. VANDERWERr [Auk,Vol. 111

Sex did not have a significant overall effect I• Undisturbed on search-and-attack rates (MANOVA, n = 42, 20 - ß Disturbed Fs,•5= 1.99, P = 0.10), proportionsof attack maneuvers (n = 41, F5,34= 1.13, P = 0.36), or proportional substrateuse (n = 41, F7,•2= 1.97, P = 15. 0.09). Only 3 of 17 foraging variablesdiffered between sexes(Table 2). Males flew less often, hung more often, and usedlarge branchesmore often than females. • 10' Diversityof foraging behavior and individual vari-

ation.--Diversity of attack maneuvers for the 5- speciesas a whole measured as H'/H'm• was 0.86, and diversity of substrateuse for the species was 0.67. There was considerable individ- 0 . ual variation in proportions of attack maneu- 0 2,0 4'0 6'0 8,0 vers and substratesused, and their relative pro- Foliage denalty (%) portions varied among individuals, even within a habitat (Figs. 2 and 3). Diversity of attack Fig. 1. Foliage-heightprofile showingpercent area maneuvers for individual birds measured as H' / in each height category covered by leaves in each habitat. H'm• ranged from 0.48 to 0.98 (œ= 0.77 + SD of 0.12) and was greater in disturbed habitat (n = 46, t = 2.04, P = 0.048). Diversity of substrate because "other" trees were used almost four usefor individual birds rangedfrom 0.29 to 0.91 times more often in undisturbed habitat (13.0 (œ = 0.57 + 0.14) and did not differ between vs. 3.7%;Table 1). This is not surprisingsince habitats (n = 46, t = 1.39, P = 0.11). Differences speciesother than ohia and koa were rare in in proportions of attack maneuvers between disturbed habitat. The interaction of maneuver habitats found in previous analyseswere cor- and substrate (MS) indicates an association of roborated by data from two birds for which I certain maneuvers with particular substrates. had sufficientobservations in both habitats(Fig. Specifically,birds on bark were more likely to 2; WABG and GAWB). In both individuals perch- perch-glean (53%) or hang (34%) than flight- gleaning was the most common maneuver in glean (13%). While on twigs, birds were more undisturbed habitat, and flight-gleaning the likely to perch-glean (61%) than flight-glean most common in disturbed habitat. Both birds (10%). Conversely, flight-gleaning was more also chased in undisturbed habitat but did not commonthan perch-gleaning on leaves(70 vs. hawk, and hawked but did not chase in dis- 30%). Certain maneuvers were also used more turbed habitat. frequentlyon differenttree species(MT). Perch- Interactionof categoricalvariables using log-linear glean was used more on ohia (47%)than on koa analysis.--HM + HT + MS + MT was the sim- (40%),and hang was more commonon koa (26%) plest model that adequatelyexplained the ob- than on ohia (11%). served data (Table 3). The full second-order model did not quite fit the data, so forward selection started with the full second-order DISCUSSION model. Backward selectionwas not possiblebe- cause the full third-order model contained so Effectsof habitat.--Elepaioin disturbedhabitat many parametersthat degreesof freedom fell attackedprey only two-thirds as often as those to zero. There were significantinteractions be- in undisturbed habitat, suggestingthat prey are tween eachpair of variablesin the final model, less abundant in disturbed habitat, harder to and no third-order interactions were necessary search for and detect, or both. Data from Peck to explain the data. (1993) indicate that arthropod biomassper dry ! interpret the model in the following way: mass of vegetation actually is higher in dis- The presence of the HM term indicates that turbed habitat. Search patterns of Elepaio sug- attack maneuver depended on habitat, as found gest that the secondexplanation is true. Birds in the previousanalysis. The HT term indicates in disturbed habitat flew farther and more often tree-speciesuse was related to habitat,primarily while searchingfor prey, which presumablyre- October1994] ElepaioForaging Behavior 923

• +1+1+1 +1+1 •+1 +1+1+1+1 • +1+1+1+1+1 +1+1

•+1+1 +1+1+1 • +1 +1+1+1 +1+1 +1

•• • .....

• +1+1+1+1 +1 • +1+1 +1 +1+1

• +1 +1+1+1 +1 924 ERIC g. VANDERWERF [Auk, Vol. 111

Undisturbed Habitat

1.0-

0.8

0.6 [] Chase [] Hawk [] Hang 0.4 [] Flight-glean ß Perch-glean

0.2

0.0

S S * S *

Disturbed Habitat

1.0-

0.8-

0.8'

0.4-

0.2

0.0

S S * * Individual

Fig. 2. Individual variation in proportions of attack maneuversin different habitats. Symbols for each individual are color-band codes or "UB" for unbanded birds. Individuals found in both habitats marked with asterisk (*); subadults marked with "S." quired more time and energy. The more-open with my earlier findings (VanderWerf 1993)in structureand lower foliage density of disturbed which Elepaio selectedforaging sites at a fine habitat may have causedfewer perchesto be scale based on habitat cues within 0.75 m. The within hopping range of each other, necessi- combination of lower foliage density, longer tating moreand longer flightsbetween perches. distancesbetween perches, and an upper limit Moreover, the attack radius did not differ be- to attackradius likely causedbirds in disturbed tween habitats,suggesting there may be an ef- habitat to search less substrate from a given fective prey-detectiondistance determined by perch (Robinsonand Holmes 1984). Because perceptual ability, beyond which searchingis birds in disturbed habitat searched less sub- inefficient (Fitzpatrick 1981). This is consistent strate from each perch and spent more time October1994] ElepaioForaging Behavior 925

Undisturbed Habitat

• [] Ground

• [] Leaf • [] Twig

•. []Large branch o. [] Trunk

S S * S *

Disturbed Habitat

1.0

= 0.8

.o 0.6

• 0.4

• 0.2

S S * * Individual Fiõ. 3. •di•idua] •ariado• i• proportionalsubstrate use i• diEerectbabitats. Symbols •i•e• •or eacb bird are cotor-ba•d codes, or "UB" •o• u•ba•ded birds. •di•iduats •ou•d i• botb babitats marAed •itb astedsA (•); subadults marAed •itb "S."

TABLE3. Stepsin reachingfinal log-linearmodel by forwardselection (H = habitat,M = maneuver,S = substrate,T = treespecies). Terms with symbols for more than one variable represent interactions between thosevariables. P > 0.05indicates adequate fit to observeddata. Asterisk (*) indicatesmodel whose conditionallog-likelihood ratio not significant (P < 0.05),meaning model is notstatistically different from previousone, and term added or deletednot necessaryto explainobserved data.

Conditional log- Model G2 df P likelihood

H+M+S+T 248.37 53 <0.0001 -- HM+HS +HT+MS +MT +ST 31.96 18 0.022 -- HST+HM+MS +MT 20.54 10 0.025 11.42' HM + HT + MS + MT + ST 32.60 22 0.068 0.64* HM + HT + MS + MT 41.35 30 0.081 8.75* 926 ERICA. VANDERWERF [Auk, Vol. 111 moving between perches,they may have en- Martin and Karr (1990) found that proportions counteredfewer prey. of attack maneuversused by specieschanged Differences in habitat structure also affected seasonally,but the relative rank of each ma- proportions of attack maneuvers. More fre- neuverdid not change.They suggestedspecies quent use of perch-gleaning in undisturbed were plastic only within certain limits deter- habitatmay be explainedby higherfoliage den- mined by their evolutionaryhistories. Because sity. In disturbedhabitat, lower foliage density attackmaneuvers used by a speciesmay be con- and more-distantperches may have resultedin strained by morphology, birds may be more more prey being out of reach, causingbirds to likely to changeother aspectsof foraging,such use flight-gleaning more often (Maurer and as height or plant species(Hutto 1981, Fitzpat- Whitmore 1981, Sabo and Holmes 1983). Birds rick 1985, Sherry 1985, Martin and Karr 1990, in disturbedhabitat alsoattacked more prey in Moermond 1990). This was not true for the Ele- the air by hawking, perhapsbecause the open paio; severalaspects of foraging behavior, in- structureof disturbedhabitat madeflying prey cludingattack maneuver, were plastic.Relative more visible (Seidel and Whitmore 1982). The proportionsof attackmaneuvers and substrates chasemaneuver was used to captureprey that used differed between habitats for the species flushed and attempted to escape,such as ho- as a whole, varied among individuals within a mopterans(Sherry 1985) and geometridmoths habitat, and even changed within individuals (pets. obs.).The fact that chasewas used lessin whose territories overlapped both habitats. disturbedhabitat suggests that prey of this type The differencesin degreeof flexibility found was lessabundant or harder to flush (or perhaps in my study comparedto the previousstudies easier to capture) in disturbed habitat. may reflect that bird specieson small islands Flight-gleaningand hawking presumablyare with relatively depauperate avifaunas often more energeticallyexpensive methods of cap- show much variation in foraging methods(Se- turing prey than perch-gleaning,and birds may lander 1966). Morse (1971, 1977) found that spe- employthese techniques to increasefood intake ciesof wood-warblersthat foraged opportunis- (Morse 1973, Bennett 1980). Martin and Karr ticallywere morelikely to inhabitsmall islands. (1990) found shifts in proportions of attack ma- Abbott et al. (1977) found that diet breadth of neuversat different seasonsand hypothesized Galapagosground finches (Geospizaspp.) was that changesin foraging patterns toward en- correlated with number of islands occupied. ergeticallyexpensive maneuvers reflected pe- Feinsingerand Swarm(1982) showed that dur- riods of demanding environmental conditions ing periodsof food scarcitya speciesof hum- and food limitation. Increaseduse of energet- mingbird had a broader feeding niche on To- ically expensivemaneuvers by Elepaio in dis- bago,where there were only two other nectar- turbedhabitat likewise may indicate greater food ivores, than on Trinidad, where there were limitation. Disturbed areasat the study site may many. The Bonin IslandsHoneyeater (Apalop- be lower-qualityhabitat for Elepaiobecause they teronfamiliare) also has been reported to have provide a more-difficultforaging environment, diverseforaging methodscompared to related which could account for the lower population speciesin Japan (Higuchi et al. 1984, pers. densityof Elepaioin disturbedhabitat (unpubl. comm.). An unusual example was reported by data). Werner and Sherry (1987) for the CocosIsland In other studiescomparing foraging between Finch (Pinaroloxiasinornata), in which individ- habitats, Szaro and Balda (1979) found that for- uals were highly specializedeven though the aging methods (maneuvers) of speciescom- speciesas a whole showeda wide range of be- prising the bird communityin ponderosapine haviors. (Pinusponderosa) forests did not change with The Elepaio alsoappears to use more-diverse forest structure,and that only 5 of 15 species foragingmaneuvers and substratesthan related changedproportions of maneuvers.Maurer and speciesfound on continentsand large islands. Whitmore (1981) found that of five species,only After conversionto categoriesequivalent to AmericanRedstarts (Setophaga ruticilla) changed those in my study and expressedas H'/H'm•, foragingmaneuvers between forests with dif- data from Croxall (1974) for 14 speciesin three ferent structure, and that other birds instead genera of monarchineflycatchers from New changedtree species,substrates, or height. Guinea showedthat diversity of foragingma- Diversityand variationin foragingmethods.-- neuvers ranged from 0.16 to 0.69 (œ= 0.47 + October1994] ElepaioForaging Behavior 927

0.16). Data from Bell (1984) showed that, for perch-gleaning.Subadults may use simpler and eight of the samespecies in New Guinea, di- energetically less expensivemaneuvers more versityof foragingmaneuvers ranged from 0.05 oftento compensatefor their lowerproficiency to 0.84 (• = 0.47 _+0.32) and diversityof sub- (Davies and Green 1976). stratesused from 0.31 to 0.79 (• = 0.53 _+0.16). Interactionof variablesin log-linearmodel.--Birds Data from Recher et al. (1985) for two fantails are thoughtto usedifferent maneuvers to cap- (Rhipidura)and a monarch (Monarcha)in Aus- ture differenttypes of prey (Rabenold1978). In tralia showedforaging maneuverdiversity my study, interaction of maneuver and sub- ranging from 0.48 to 0.68 (œ= 0.57 _+0.10) and stratein the log-linearmodel indicates Elepaio substratediversity from 0.47 to 0.67 (œ= 0.58 alsouse different maneuvers to captureprey on + 0.10).Thus, the Elepaioappears to usemore- differentsubstrates. Perch-gleaning was more diverse foraging maneuversthan its relatives common on bark and twigs, perhaps because on largeislands and continents,and hasamong thesesubstrates could supportthe massof a the highest diversitiesof substrateuse. The di- perchedbird, allowing birds the option of using versity of foraging maneuversused by Elepaio the easiestmaneuver. Most leavesprobably (0.86) is similar to the 0.88 reportedfor the Co- couldnot support a perchedbird, and may have cos Island Flycatcher (Nesotriccusridgwayii; been out of reachof birds perchedon bark or Sherry 1985), which also inhabits a small oce- twigs, therefore explaining why leaveswere anic island. usedmore in conjunctionwith flight-gleaning. Effectsof age.--As found in numerousstudies Hang wasused more on bark,possibly because of age-specificforaging (see Wunderle 1991), the diameterof trunksand largebranches was subadultElepaio had lowerforaging proficien- too largeto allow a bird to reachall partsof the cy than adults. The interval between attacks on substratewhile perched. prey was 25%longer in subadults.Subadult Ele- Structuraldifferences between tree species af- paio also hoppedless frequently than adults, fectwhich maneuvers birds can use to capture suggestingthe lower foragingproficiency of prey (Robinson and Holmes 1984). More fre- subadultswas causedby a slowersearch rate. quent use of perch-gleaningon ohia may be Giventhat flight intervaland flight distanceof explainedby its relatively densecrown with subadultElepaio were no different than those many twigs and small leaveson short petioles of adults,subadults do not perceiveas many (Jackson1979). Birds easily could reach many prey items,search incompletely before flying leaveswhile perchedon twigsor smallbranch- to a new perch, or both. Similar results were es.In contrast,the primaryphotosynthetic sur- obtained by Richardsonand Verbeek (1987), facesof koa are phyllodes(elongate, flattened who found that juvenile Northwestern Crows petioles)that are too thin to supportthe mass (Corvuscaurinus) had lower success,searched of a perchedbird and may be too distantfor moreslowly, and spentmore time searching birdsto reachwhile perched.Birds on koawere than adults.In contrast,Gochfeld and Burger forcedto useflight-gleaning more often. Hang (1984) found that juvenile American Robins mayhave been used more on koabecause they (Turdusmigratorius) searched faster than adults, typicallyreach much larger sizes than ohia,and butstill hadlower success. Subadult birds may their thicker trunks and branchesrequired notbe asskilled at perceivingprey or maynot hangingmore often. This effectof vegetation havedeveloped an appropriatesearch image structureon foragingbehavior of insectivorous (Wunderle1991). The increasein foragingef- birdsis analogous to fruit-accessibility and pref- ficiencywith agecould be due to learningor erencepatterns of frugivorousbirds (Denslow to eliminationof inefficientbirds from the pop- and Moermond1982, Levey et al. 1984). ulation (Groves 1978). Similarcomparisons can be made among sev- In the ontogenyof foragingbehavior, ma- eral studiesof foragingmethods of insectivo- neuversthat requiremore coordination, partic- rousbird communitiesat sitesdominated by ularlythose involving flight, often develop lat- tree specieswith differingvegetation structure. er (Davies and Green 1976, Moreno 1984, Brei- Airola and Barrett (1985) and Alatalo (1982) twisch et al. 1987). Both attack maneuvers used foundthat perch-gleaningwas more common lessby subadult Elepaio than by adults (i.e. hang than flight-gleaningat their studysites, which andhawk) are relatively complex and presum- were dominatedby coniferoustrees that had ably require more skill and coordination than foliage on short petiolesdistributed continu- 928 ERICA. VANDERWERF [Auk, Vol. 111 ouslyalong twigs. Robinsonand Holmes (1982) (Peck 1993), which cannot explain the use of found that flight-gleaning was more common, more difficult foraging methods in disturbed and most trees at their site were deciduous and habitat. had foliage on long petioles concentrated at Summaryand conclusions.--Open-canopydis- branch tips. At an Australian site consistingof turbed habitat may be of lower quality for the Eucalyptus spp. trees with widely spaced Elepaiobecause searching for and capturingprey branches and leaves, search-and-attackflights is more difficult. Birds in disturbed habitat flew were longer than in the above studies(Holmes farther and more often while searching,used and Recher 1986). Overall foraging methodsat energeticallyexpensive attack maneuversmore each site were related to physical structure of often, and encounteredprey less often. Sub- the dominant tree species.I found similar pat- adult Elepaio had lower foraging proficiency terns of attack maneuvers for the Elepaio be- that may be explainedby a slower searchrate. tween tree species at one site. Subadultsmay compensatefor their lower pro- There are several possiblealternative expla- ficiency by using simpler, less-expensivema- nationsfor the differencesin foraging between neuvers more often. Few differences in forag- habitats.The risk of predation may affect how ing were found between sexes.Foraging be- long birds forage or which foraging sitesthey havior of the Elepaio was very plasticand was choose(Lima 1985). Suhonen (1993) found that more diverse than that reported in the literature birds foragedin more protectedsites toward the for related specieson continents and large is- centers of trees in a year when predation risk lands. Proportions of attack maneuversand sub- was higher. At my study site, one might expect strates used differed between habitats, varied the risk of predation to be higher in the more among individuals within a habitat, and even open disturbed habitat becausebirds would be changedwithin individuals between habitats. more visible to predators. However, the ma- Elepaio were more likely to use specificma- neuvers (hawk and flight-glean) and substrates neuvers on different tree speciesand on differ- (air and leaves) used more in disturbed habitat ent substrates,perhaps because of the physical would seem to increasevulnerability to pred- structure of these surfaces. ators.Unless the benefitsof increasedprey cap- Habitat alterationby humansat Hakalau For- ture outweigh the predation risk, it is unlikely est National Wildlife Refuge may have made that predation risk is responsible.The primary some areas more demanding as foraging sites predator on the Elepaio at my site may be the for Elepaio, but their behavioral flexibility and Hawaiian Hawk (Buteosolitarius). I frequently diverserepertoire of foragingtechniques allow observedhawks hunting in both habitats,and Elepaio to successfullyexploit arthropod re- I observeda hawk prey on a fledgling Elepaio sourcesin disturbedhabitat. Population density once (in disturbed habitat). is slightly lower in disturbedhabitat, and it is Interspecificcompetition also can alter for- possiblethat habitat alteration affectsthe Ele- aging behavior(Morse 1967,Alatalo 1981,Ca- paio in other ways. In the future I hope to test rothers 1986). Some competition may occur be- whether disturbance influences territory size, tween the Elepaioand otherinsectivorous birds, population structure, time budgets, and ptil- particularlythe introducedJapanese White-eye ochronology. (Zosteropsjaponicus; Mountainspring and Scott

1985). However, since similar changes in for- ACKNOWLEDGMENTS aging were found even within individualsbe- tween habitats,and speciescomposition of the I thank Dick Wass,refuge manager,for permission insectivorousbird guild was the samein both to work at Hakalau Forest National Wildlife Refuge, habitats,it seemsunlikely that the differences and Bill Andrade for accessto adjacentareas of Pua in foraging behavior between habitats were Akala ranch. LeonardFreed generouslyallowed me to useUniversity of Hawaii facilitiesat the studysite, causedby competition. and I thank him for his hospitality.Jack Jeffrey and Finally, arthropodabundance and diversity JaanLepson helped me identify plants.I thank my undoubtedlyaffect foraging behavior of insec- committee members Peter Feinsinger and Carmine tivorous birds (Holmes and Schultz 1988). At Lanciani,and especiallymy advisor,Doug Levey;they my study site, arthropodabundance is higher provided many helpful commentsand criticismsat in disturbed habitat and diversity of most ar- variousstages of the work. ScottFretz, Tom Grubb, thropod taxa does not differ between habitats SallieHejl, JobyRohrer, and Gary Schnellprovided October1994] ElepaioForaging Behavior 929 several useful suggestionson earlier drafts of the CONANt,S. 1977. The breeding biology of the Oahu manuscript.This study was supportedby a Grinter Elepaio.Wilson Bull. 89:193-210. GraduateFellowship from the University of Florida CROXAILL,J.P. 1974. Feedingbehaviour and ecology and a grant from the John D. and CatherineT. Mac- of New Guinea rainforest insectivorouspasser- Arthur Foundation (to L. Freed, R. Cann, and S. Co- ines. Ibis 119:114-146. nant). CVDDIHY, L. W., AND C. P. STONE. 1990. Alteration of nativeHawaiian vegetation: Effects of humans, their activities and introductions.Cooperative National Park ResourcesUnit and Univ. Hawaii, LITERATURE CITED Honolulu. DAVIES, N. B., ^•qD R. E. GP•E•q. 1976. The devel- ABBOTT, I., L. K. ABBOTT, AND P. R. GRANT. 1977. opment and ecologicalsignificance of feeding Comparativeecology of Galapagosground finch- techniquesin the Reed Warbler (Acrocephalusscir- es(GeosFiza Gould): Evaluation of the importance paceus).Anim. Behav. 24:213-229. of fioristic diversity and interspecificcompeti- Ds•qsrow, J. S., ^•qD T. C. MOERMOND. 1982. The tion. Ecol.Monogr. 47:151-184. effectof accessibilityon rates of food removal AIROLA,D. A., AND R. H. BARRETT.1985. Foraging from tropical shrubs: An experimental study. and habitatrelationships of insect-gleaningbirds in a Sierra Nevada mixed-conifer forest. Condor Oecologia54:170-176. DODGE, K. M., R. C. WHITMORE, AND E. J. HARNER. 87:205-216. 1990. Analyzing foraginguse versus availability ALATALO,R. V. 1981. Interspecificcompetition in usingregression techniques. Stud. Avian Biol. 13: tits Parusspp. and the GoldcrestRegulus regulus: 318-324. Foragingshifts in multispeciesflocks. Oikos 37: 335-344. ECKHARDT,R. C. 1979. The adaptive syndromesof two guilds of insectivorousbirds in the Colorado ALATALO,g. V. 1982. Multidimensional foraging Rocky Mountains. Ecol. Monogr. 49:129-149. niche organization of foliage gleaning birds in EMLEN,J.T. 1977. Land bird communities of Grand northern Finland. Ornis Scand. 13:56-71. BahamaIsland: The structureand dynamicsof BELL, G. W., S. J. HEJL, AND J. VERNER. 1990. Pro- an avifauna. Ornithol. Monogr. 24. portional use of substratesby foraging birds: FFa•SINGER,P., ,•htDL. A. SW,•RM. 1982. "Ecological Model considerationson first sightingsand sub- release,"seasonal variation in food supply, and sequent observations. Stud. Avian Biol. 13:161- 165. the hummingbirdAmazilia tobaci on Trinidad and Tobago.Ecology 63:1574-1587. BELLE,H.L. 1984. A bird communityof lowland rain- FIENBERG,S. E. 1970. An iterative procedurefor es- forest in New Guinea. 6. Foraging ecologyand timation in contingencytables. Ann. Math. Stat. community structureof the avifauna. Emu 84: 41:907-917. 142-158. FIENSERG,S.E. 1972. The analysisof incompletemul- BENNETT,S. E. 1980. Interspecificcompetition and ti-way contingencytables. Biometrics 28:177-202. the niche of the American Redstart (Setophaga FIENBERG,S.E. 1981. The analysisof cross-classified ruticilla)in wintering and breedingcommunities. categoricaldata, 2nd ed. MIT Press,Cambridge, Pages319-335 in Migrant birdsin the Neotropics: Massachusetts. Ecologybehavior, distribution and conservation FITZPATRICK,J.W. 1980. Foragingbehavior of Neo- (A. Keast and E.S. Morton, Eds.). Smithsonian tropicaltyrant flycatchers.Condor 82:43-57. Institution Press,Washington, D.C. BERGER,A.J. 1981. Hawaiian birdlife, 2nd ed. Univ. FITZPATRICK,J.W. 1981. Searchstrategies of tyrant Hawaii Press, Honolulu. flycatchers.Anim. Behav. 29:810-821. BISHOP,Y. M. M. 1971. Effectsof collapsingmulti- FITZPATRICK,J.W. 1985. Form, foraging, behavior, dimensionalcontingency tables. Biometrics27: and adaptive radiation in the Tyrannidae. Or- 545-562. nithol. Monogr. 36:447-470. BoI,I•S, W.E. 1979. The relationshipsof the Australo- FRANZRES,K. E. 1983. A comparisonof avian for- Papuan flycatchers.Emu 79:107-110. aging behavior in unloggedand loggedmixed- BRADI,E¾, D.W. 1985. The effectsof visibility bias coniferous forest. Wilson Bull. 95:60-76. on time-budgetestimates of niche breadthand GOCHFELD,M., ANDJ. BURGER.1984. Age differences overlap.Auk 102:493-499. in foragingbehavior of the AmericanRobin. Be- BP,En•vIscrt, R., M. DIAZ, ^•qDR. LEE. 1987. Foraging haviour 88:227-239. efficienciesand techniquesof juvenile and adult GOODMAN,L. A. 1971. The analysisof multidimen- Northern Mockingbirds(Mimus polyglottos). Be- sional contingencytables: Stepwise procedures haviour 101:225-235. and direct estimationmethods for building mod- CAROtI-IERS,J. H. 1986. Behavioraland ecological elsfor multiple classifications.Technometrics 13: correlates of interference competition among 33-61. someHawaiian Drepanidinae. Auk 103:564-574. GREENBERG,g., AND J. GRADWOHL.1980. Leaf surface 930 ERICA. VANDERWERF [Auk, Vol. 111

specializationsof birds and arthropodsin a Pan- behaviorof somemigratory western wood war- amanianforest. Oecologia 46:115-124. blers. Auk 98:765-777. GROVES,S. 1978. Age-relateddifferences in Ruddy JACKSON,J.A. 1979. Tree surfacesas foragingsub- Turnstoneforaging and aggressivebehavior. Auk stratesfor insectivorousbirds. Pages 69-94 in The 95:95-103. role of insectivorousbirds in forestecosystems GRUBB,T. C., JR. 1978. Weather dependent foraging (J. G. Dickson, R. N. Conner, R. F. Fleet, J. C. ratesof wintering woodlandbirds. Auk 95:370- Kroll, and J. A. Jackson, Eds.). Academic Press, 376. New York. GRUBB,T. C., JR. 1979. Factorscontrolling foraging JAMES,F. C., AND H. H. SHUGART.1970. A quanti- strategiesof insectivorousbirds. Pages119-135 tative method of habitat description.Audubon in The role of insectivorous birds in forest eco- Field Notes 24:727-736. systems(J. G. Dickson, R. N. Conner, R. R. Fleet, KIRCH,P.V. 1983. Man's role in modifyingtropical J. C. Kroll, and J. A. Jackson,Eds.). Academic and subtropicalPolynesian ecosystems. Archae- Press, New York. ology in Oceania 18:26-31. GRUBB,T. C., JR.,AND M. S. WOODREY.1990. Sex,age, LIMA,S.L. 1985. Maximizing feedingefficiency and intraspecificdominance status, and the useof food minimizing time exposedto predators:A tradeoff by birds wintering in temperate-deciduousand in the Black-cappedChickadee. Oecologia 66:60- cold-coniferous woodlands: A review. Stud. Avi- 67. an Biol. 13:270-279. LEVEY,D. J., T. C. MOERMOND,ANco J. S. DENSLOW. GUSTAFSSON,L. 1988. Foragingbehaviour of indi- 1984. Fruit choice in Neotropical birds: The ef- vidual Coal Tits, Parus ater, in relation to their fect of distancebetween fruits on preferencepat- age,sex, and morphology.Anita. Behav.36:696- terns. Ecology65:844-850. 704. MACARTHUR,R. H. 1958. Populationecology of some HEJL,S. J., ANDJ. VERNER.1990. Within seasonand warblers of northeastern coniferous forests. Ecol- yearlyvariations in avianforaging locations. Stud. ogy 39:599-619. Avian Biol. 13:202-209. MACCAUGHEY,V. 1919. The Hawaiian Elepaio.Auk HEJL, S. J., J. VERNER,AND G. W. BELL. 1990. Se- 36:22-35. quential versusinitial observationsin studiesof MARTIN,T.E. 1986. Competitionin breedingbirds: avian foraging.Stud. Avian Biol. 13:166-173. On the importanceof consideringprocesses at HIGUCHI, H., M. NAKANE, AND A. SUZUrd. 1984. Some the level of the individual. Curt. Ornithol. 4:181- noteson the ecologyand behaviorof the Bonin 210. IslandsHoneyeater Apalopteron familiare. Tori 33: MARTIN,T. E., ANDJ. R. KaRR. 1990. Behavioral plas- 67-73. ticity of foraging maneuversof migratory war- HOGSTAD,O. 1988. Rank-related resource access in biers:Multiple selectionperiods for niches?Stud. winter flocks of Willow Tit Parus montanus. Ornis Arian Biol. 13:353-359. Scand. 19:169-174. MAURER,B. A., ANDR. C. WHITMORE.1981. Foraging HOLMES,R. T., R. E. BONNEY,AND S. W. PACALA. 1979. of five bird speciesin two forestswith different Guild structure of the Hubbard Brook bird com- vegetationstructure. Wilson Bull. 93:478-490. munity:A multivariateapproach. Ecology 60:512- MOEm'aOND,T. C. 1990. A functional approach to 520. foraging:Morphology, behavior, and the capac- HOLMES,R. T.,AND H. F. RECHER. 1986. Search tactics ity to exploit. Stud. Avian Biol. 13:427-430. of insectivorousbirds foraging in an Australian MORENO,J. 1984. Parental care of fledged young, eucalyptforest. Auk 103:515-530. division of labor, and the development of for- HOLMES,R. T., A•D S. K. ROBINSON.1981. Tree spe- aging techniquesin the Northern Wheatear(Oe- cies preferencesin foraging insectivorousbirds nantheoenanthe L.). Auk 101:741-752. in a northern hardwoodsforest. Oecologia48: 31-35. MOP,mSON, M.L. 1984. Influence of samplesize and HOLMES,R. T., AND J. C. SCHULTZ. 1988. Food avail- sampling design on analysis of avian foraging behavior. Condor 86:146-150. ability for forestbirds: Effects of prey distribution and abundanceon bird foraging.Can. J. Zool. 66: MORRISON,M. L., I. C. TIMOSSI, K. A. WITH, AND P. N. 720-728. MANLEY.1985. Useoftreespeciesbyforestbirds HOLMES, R. T., T. W. SHERRY,AND S. E. BENNETT. 1978. during winter and summer.J. Wildl. Manage.49: Diurnal and individual variability in the forag- 1098-1102. ing behavior of American Redstarts(Setophaga MORSE,D. H. 1967. Competitive relationshipsbe- ruticilla).Oecologia 36:141-149. tween ParulaWarblers and other speciesduring HURLBERT,S. H. 1984. Pseudoreplication and the the breeding season.Auk 94:490-502. designof ecologicalexperiments. Ecol. Monogr. MORSE,D.H. 1971. The foragingof warblersisolated 54:187-211. on small islands.Ecology 52:216-228. Hutfro, R.L. 1981. Seasonalvariation in the foraging MORSE,D. H. 1973. The foraging of small popula- October1994] ElepaioForaging Behavior 931

tions of Yellow Warblers and American Red- REMSEN,J.V.,JR. 1985. Communityorgamzationand starts.Ecology 54:346-355. ecologyof birds of high elevation humid forest MORSE,D.H. 1977. The occupationof small islands of the BolivianAndes. Ornithol. Monogr.36:733- by passefinebirds. Condor 77:399-412. 756. MOUNTAINSPRING,S.R. 1987. Ecology,behavior, and REMSEN,J. V., JR., AND S. K. ROBINSON. 1990. A clas- conservation of the Maul Parrotbill. Condor 89: sificationscheme for foraging behaviorsof birds 24-39. in terrestrial habitats. Stud. Avian Biol. 13:144- MOUNTAINSPRING,S. R., AND J. M. SCOTT. 1985. In- 160. terspecificcompetition among Hawaiian forest RICHARDSON,H., AND N. A.M. VERBEEK. 1987. Diet birds. Ecol. Monogr. 55:219-239. selection by yearling Northwestern Crows (Cor- MUNRO, G.C. 1960. Birds of Hawaii. Charles E. Tut- vus caurinus)feeding on littleneck clams (Vene- tle, Rutland, Vermont. rupisjaponica). Auk 104:263-269. NOON, B. R. 1981. Techniques for sampling arian ROBINSON,S.K. 1986. Three-speedforaging during habitats.Pages 42-52 in The use of multivariate the breeding cycle of Yellow-rumped Caciques statistics in studies of wildlife habitat (D. E. Ca- (Icterinae:Cacicus cela). Ecology 67:394-405. pen, Ed.). USDA ForestServ. Gert. Tech. Rep.RM- ROBINSON,S. K., ANDR. T. HOLMES.1982. Foraging 87. behaviorof forestbirds: The relationshipsamong OLSON, S. L., AND H. F. J•MES. 1984. The role of searchtactics, diet, and habitat structure.Ecology Polynesiansin the extinction of the avifauna of 63:1918-1931. the Hawaiian Islands. Pages768-780 in Quater- ROBINSON, S. K., AND R. T. HOLMES. 1984. Effects of nary extinctions (P.S. Martin and R. G. Klein, plant speciesand foliage structure on the for- Eds.). Univ. Arizona Press, Tucson. aging behavior of forest birds. Auk 101:672-684. PARTRIDGE,L. 1976. Individual differences in feed- ROOT,R.B. 1967. The niche exploitation pattern of ing efficienciesand feeding preferencesin cap- the Blue-grayGnatcatcher. Ecol. Monogr. 37:317- tive Great Tits. Anim. Behav. 24:230-240. 350. PEcK,R.W. 1993. Theinfluenceofarthropods, forest SABO,S. R., ANDR. T. HOLMES.1983. Foragingniches structure, and rainfall on insectivorous Hawaiian and the structure of forest bird communities in forest birds. Master's thesis, Univ. Hawaii, Ho- contrastingmontane habitats. Condor 85:121-138. nolulu. SAKAI,H. F. 1988. Avian responseto mechanical PETERS,W. D., AND T. C. GRUBB,JR. 1983. An exper- clearing of a native rainforest in Hawaii. Condor imental analysisof sex-specificforaging in the 90:339-348. Downy Woodpecker Picoidespubescens. Ecology SAKAI, H. F., AND B. R. NOON. 1990. Variation in the 64:1437-1443. foraging behaviors of two flycatchers:Associa- PETIT,L. J., D. R. PETIT,K. E. PETIT,AND W. J. FLEMING. tions with stageof the breeding cycle.Stud. Avi- 1990. Intersexualand temporal variation in for- an Biol. 13:237-244. aging ecologyof ProthonotaryWarblers during SCHOENER,T. W. 1970. Nonsynchronous spatial the breeding season.Auk 107:133-145. overlap of lizards in patchy habitats.Ecology 51: PmLOU,E.C. 1966. The measurementof diversity in 408-418. different typesof biologicalcollections. J. Theor. SCHOENER,T.W. 1986. Resourcepartitioning. Pages Biol. 13:131-144. 91-126 in Community ecology (J. Kikkawa and PRATT, H.D. 1980. Intra-island variation in the Ele- D. J. Anderson, Eds.). Blackwell Scientific, Mel- paio on the island of Hawaii. Condor 82:449-458. bourne. PRATT, H. D., P. L. BRUNER,AND D. G. BERRETT. 1987. ScoTT, J. M., S. MOUNTAINSPRING,F. L. RAMSEY,AND A field guide to the birds of Hawaii and the trop- C. B. KEPLER. 1986. Forest bird communities of ical Pacific. Princeton Univ. Press, Princeton, New the Hawaiian islands:Their dynamics,ecology, Jersey. and conservation. Stud. Avian Biol. 9. RABENOLD,K.N. 1978. Foragingstrategies, diversity, SEIDEL, G. E., AND R. C. WHITMORE. 1982. Effect of and seasonalityin bird communitiesof Appala- forest structureon American Redstartforaging chian spruce-fir forests. Ecol. Monogr. 48:397- behavior. Wilson Bull. 94:289-296. 424. SELANDER,R. K. 1966. Sexual dimorphism and dif- RECHER,H. F., ANDV. GEBSra.1990. Analysis of the ferential niche utilization in birds. Condor 68: foraging ecology of eucalypt forest birds: Se- 113-151. quential versus single point observations.Stud. SHANNON, C. E., AND W. W•VER. 1949. The math- Avian Biol. 13:174-180. ematicaltheory of communication.Univ. Illinois RECHER,H. F., R. T. HOLMES, M. SCHULZ, J. SHIELDS, Press, Urbana. AND R. KAVANAGH.1985. Foraging patterns of SHERRY,T.W. 1985. Adaptation to a novel environ- breeding birds in eucalypt forest and woodland ment: Food,feeding, and morphologyof the Co- of southeastern Australia. Aust. J. Ecol. 10:399- cosIsland Flycatcher.Ornithol. Monogr. 36:908- 421. 920. 932 EaxcA. VAN•)ERWERF [Auk, Vol. 111

SIBLEY,C. G., AND J. E. AHLQUIST.1990. Phylogeny WAGhlm•,J.L. 1981. Visibility and biasin avian for- and classificationof birds: A study in molecular aging data. Condor 83:263-264. evolution. Yale Univ. Press, New Haven, Con- WERNER,t. K., AND t. W. SHERRY.1987. Behavioral necticut. feedingspecialization in Pinaroloxiasinornata, the SOKAL,R. R., AND F. J. ROHtF. 1981. Biometry, 2rid "Darwin's finch" of Cocos Island, Costa Rica. Proc. ed. W. H. Freeman, New York. Nat. Acad. Sci. USA 84:5506-5510. STONE,C. P., AND J. M. SCOTT. 1985. Hawaii's ter- WHEL.•, C. J. 1989. Avian foliage structurepref- restrial ecosystems:Preservation and manage- erencesfor foraging and the effectof prey bio- ment. CooperativeNational Park ResourcesStud- mass. Anita. Behav. 38:839-846. ies Unit and Univ. Hawaii, Honolulu. WIENS,J. A., AND J. T. ROTENBERRY.1981. Habitat SUHONEN,J. 1993. Predation risk influences the use associationsand communitystructure of birds in of foraging sitesby tits. Ecology74:1197-1203. shrubsteppeenvironments. Ecol. Monogr. 51:21- SZARO,R. C., ANt)R. P. BALDA. 1979. Bird community 41. dynamicsin a ponderosapine forest.Stud. Avian WILKINSON,L. 1987. SYSTAT:The systemfor statis- Biol. 3. tics. SYSTAT, Inc., Evanston, Illinois. VANDERWERF,E.A. 1993. Scales of habitat selection WUNDERLE,J.M., JR. 1991. Age-specificforaging pro- by foraging Elepaio in undisturbed and human- ficiencyin birds. Curr. Ornithol. 8:273-324. altered forests in Hawaii. Condor 95:980-989. VERBEEIC,N. A.M. 1972. Daily and annual time bud- get of the Yellow-billedMagpie. Auk 89:567-582.