Seed Handling Ability, Bill Structure, and the Cost of Specialization for Crossbills

SEED HANDLING ABILITY, BILL STRUCTURE, AND THE COST OF SPECIALIZATION FOR CROSSBILLS

CRAIG W. BENKMAN • Departmentof BiologicalSciences, State University of New Yorkat Albany, 1400Washington Avenue, Albany, New York12222 USA

AssT•cT.--Crossbills(Loxia) are specialized to extractand handle seeds from conifercones. I evaluatedthe ability of crossbillsto handlenonconifer seeds by comparingseed handling efficiencieswith other carduelinefinches. For all seedsizes, crossbills required seed encounter ratesor seedabundances 2-3 times greaterthan other speciesto meet their daily energy requirements.Consequently, crossbills may suffer high mortality during conifer cone failures. Crossbillsare inefficientat meetingtheir energydemands on nonconiferseeds because of their narrowmandibles, lowered horny palate,and large body size.The narrow mandibles enablecrossbills to efficientlyextract seeds from conifer cones and the loweredpalate enables them to handle small seedsrapidly. Crossbillshave evolvedlarger bills, and associated musculatureand bodymass, to providethe power necessaryto separatecone scales. Some of the increasein bodymass, however, may counterbalance the largebills to improvepredator evasion. Received27 October1987, accepted17 June1988.

THE carduelinefinches (subfamily Cardueli- 1976) and Eurasia(Newton 1970, 1972),and at nae) are highly specializedseedeating birds this time many crossbillsfeed on nonconifer (Newton 1967, 1972). Most consume seeds, seed(see Newton 1970,1972). These periods of mainlyfrom dicotyledonousplants, throughout coniferseed shortage often coincide with fail- the year (Martin et al. 1951;Newton 1967,1972; ures of other boreal tree seed crops (Bock and Austin 1968). The crossbills(Loxia) eat conifer Lepthien! 976).Thus, competition for available seedsto the near exclusion of other seeds(New- seedamong cardueline finches would be inten- ton 1967, 1972; Austin 1968; Nethersole- sified. Thompson1975; Benkman 1987a). I comparedthe seed-handlingrates, on grad- Crossbillsare clearly more efficientat utiliz- ed seed sizes, of 3 noncrossbill speciesof car- ing seedsin closedconifer conesthan other duelineswith the 2 speciesof crossbillsin North cardueline finches(Newton 1967, 1972). Many America. These data are used to determine the characteristicsof the crossbill'sfeeding appa- relative seedhandling abilities of crossbillsin ratus are adaptedfor exploiting conifer seeds comparisonto othercarduelines and thosefea- (Benkman1987b). Experimental ablation of the tures that make crossbills inefficient when feed- crossedportion of the crossbill'smandibles ing on nonconiferseeds. The dataon noncross- demonstratedthat this portion was essentialfor bill carduelines are compared to those on crossbills to extract seeds from closed conifer emberizinesparrows, elsewhere (Benkman and cones(Benkman 1988). Further, no carduelines PullJam 1988). other than crossbills have been found to extract

seedfrom tightly closedconifer cones(Smith METHODS and Balda 1979; pers. obs.). Crossbillsmay be limited to foraging on co- Red Crossbills (L. curvirostra),American Goldfinches nifer seedbecause they are relativelyinefficient (Carduelistristis), House Finches (Carpodacus mexicanus), and at handlingother seed types (e.g. Roberts 1936). Evening Grosbeaks(Coccothraustes vespertinus) were The reducedefficiency of crossbillsis of partic- capturedin Albany Co., New York. White-winged ular interest becauseabout every 3-4 years co- Crossbills(L. leucoptera),were caught in Laurentides nifer seed crops fail over much of the boreal Reserve,Quebec. Several days prior to and during the experiments,all specieswere fed commercialsun- forests of North America (Bock and Lepthien flower (Helianthusspp.) and niger thistle (Guizotia abyssinica)seeds, and vitamin-enriched water. For- • Presentaddress: Department of Biology,Prince- aging experimentswere conductedon the crossbills ton University,Princeton, New Jersey08544-1003 after a minimum of 5 monthsin captivity.The other USA. specieswere completedwithin 6 daysof capture.Even

715 The Auk 105: 715-719. October 1988 716 CRA•CW. BENKMAN [Auk,Vol. 105

TABLE1. Bodymasses and bill dimensionsof the 5 speciesof carduelinefinches used in the experiments. Sample size for each bill measurementis 10 birds.

Bill size (ram) Bodymass (g) Depth Width Length Species • • _ SD • + SD • + SD A. Goldfinch 13.0 6.4 + 0.1 5.0 + 0.1 8.1 _+ 0.3 House Finch 21.0 8.2 + 0.2 7.2 + 0.2 8.6 + 0.4 WW Crossbill 26.4 7.8 + 0.3 5.6 + 0.3 13.8 + 0.6 Red Crossbill 35.9 9.3 _ 0.2 7.4 + 0.2 15.3 + 1.0 E. Grosbeak 55.0 14.7 + 0.5 13.8 + 0.5 15.4 + 0.7

thoughthe crossbillswere housedfor longerperiods of water.When I wasready to recordhandling times, than the other species,they appearedto remain in the lights were turned off, the door separating the excellentphysical condition. chamberswas slid open, and then the lights were I used 8 size classes of commercial sunflower and turned on. Usually the bird would hop onto the perch niger thistle seeds. Average individual seed wet in the observationchamber to feed.After feeding,the weights ranged from 2.3-117.5 mg. Sunflowerand light on the side the bird was on was turned off, and thistle were chosen becausethey are composites the bird returned to the other chamber. The door was (Compositae),and are an important natural food for closed.All seedremains were removedwith forceps many carduelines (Martin et al. 1951; Newton 1967, and weighed. 1972;Austin 1968). All 5 speciesreadily ate these Initially seed hulls and kernel remains were seedsand visitedfeeders, though crossbills do soonly weighed separately,but it soon becameevident that infrequently.The shapeof thistleand sunflowerseeds seed hulls were not consumed in any measurable differsfrom that of conifers,but the relationshipsbe- amount. I subtracted total seed remains from initial tween massof seedcovering (y; rag) and kernel mass weight to obtain the averageweight of kernel con- (x; mg) were nearly identical for the seedsused in sumedper seed.Wet weightswere usedbecause dry the experiments(y = 0.46x•-%r 2 = 0.94, n = 8, P < weights were over 97% of wet weight. Variation in 0.01)and coniferseeds in the Northeast(y = 0.47x•.% seedsize classesbetween finch speciesresulted from r 2 = 0.93, n = 10, P < 0.01). Seed size classes were seedsselectively consumed. This was especiallytrue initially sortedwith a sieve. Then seedsmost similar for seedsthat were relatively difficult to handle. in size and shapewithin size classesand most similar Handling time beganwhen a seedwas picked up in shape among size classeswere individually se- and continueduntil no part of the seedwas being lected and examined for cracks in the seed hull. I mandibulated.Time spent handling seedsthat were usedonly seedswith undamagedand uncrackedhulls. droppedwhole wasexcluded from analyses.I watched Seedswere weighed in groups of ten to the nearest birds through a one-way window and recordedhan- 10mg for sunflowerseeds and 0.1 mg for thistleseeds, dling times on an Apple II computer,programmed and reexamined for consistencyin size and shape. to time and recordevents to 0.1 s. Data were gathered Variance in seed shapeand quality was minimized on 3-5 individuals of each species.Each datum pre- so that seed handling differenceswould reflect seed sentedrepresents the mean for onebird handling on size differences. Two additional size classes of sun- average 15.1 seeds(SD = 5.7, range = 3-31). flower seedswere given to the goldfinchesand White- I estimated the rate at which seeds must be en- winged Crossbillsto provide finer testing of their countered (k), where k = 1/T and seedhandling abilities. 3.6 x 104s(S) During the tests,10 preweighedseeds of the same T- h. DEE size were placed in a partially covered plastic tray attachedto a perch 40 cm abovethe floor of a (60 cm)3 I assumed 3.6 x 104 s (10 h) to be the maximum time wooden chamber.Because carduelines normally for- availableto forageduring a midwinter day. Midwin- agewithin vegetationabove the ground(Newton 1967, ter was chosen because this is when diets consist most- 1972), the birds were required to handle seedswhile ly of seeds(Martin et al. 1951) and winter is a period perched.The food tray was designedso any kernel when food is mostlimiting for carduelines(Newton (embryo and endosperm)dropped by the birds fell 1967,Benkman 1987a). S is the productof mass(rag) to the floorwhere recoverywas prevented by a screen. of kernel consumedper seed,specific energy value A secondchamber was connectedby a sliding door (J/rag) of the kernel, and assimilationefficiency. I at the level of the perch. Birdswere held overnight measured mass of kernel consumed, and I assumed without food for >16 h prior to the seed-handling seedkernels commonly consumed by fincheshave 23 experiments.Each chamber had one perch and a dish J/rag (Grodzinskiand Sawicka-Kapusta1970) and as- October1988] Costof Specialization 717

•o• 80. , W-WOROSSaLL•DCmSSaLL ---•1It ß W-WCROSSBILLRED CROSSBILL n- .o *.c-ou•.c. • to•[O &= A.HOUSEFNCH GOLDFINCH z • • o ß Z .1 . ;0. ';

-I A' O• ß byFig.5species 1.The ofcarduelineamount offinches.seed kernelData consumed/seedforeach species •) .01 •' •,•3 n_ •ß ß • ßß areindicated byunique symbols thatare connected • • a3n(• o point is usually the mean for 20 seeds.For each seed size the different points representdifferent individuals,but data for the different seed sizes are from the .001 bysamesecond individuals.orthird Variancesorder polynomial are not givencurves. because Each UJ•• 0 20 40 60 80 100 120 seedremains were weighed in groups,usually of 10. SEED MASS (mg) similation efficiencies of 75% (Willson and Harmeson Fig. 2. The estimated necessaryseed encounter 1973).DEE equals the total daily energy expenditure rates for 5 speciesof cardueline finches to meet their (kJ) (Walsberg1983, eq 8). The measuredhandling estimateddaily energy requirements(see Methods). time in seconds is h. I useddigital calipersto measurebill length, depth, the goldfinch for all but the small (<5 mg) this- and width, and a Sartorius or Pesola scale to measure tle seeds. body mass. Bill measurementswere made on study Time spent handling the 2 sizes of thistle skinsof 10 individuals of eachspecies. Bill depth was the greatestmeasured depth at the baseof the bill. seeds (<5 mg) differed significantly among Bill length was measuredfrom the anterior end of species(ANOVA, F4,•6= 10.81, P = 0.0002;based the naresto the tip of the upper mandible. Bill width on the means of individual birds) and both was measuredat the anterior end of the nares.Body speciesof crossbillsrequired lesstime to handle masswas measuredon the captive birds usually with- thistle seeds than the other species(P < 0.05, in an hour of capture. Multiple Rangetest). Seed size had no effecton handling time (ANOVA, F•,,6= 3.26, P = 0.09). Similar analyseswere not conductedon the RESULTS larger sunflower seedsbecause there was tre- mendousvariation among species in the amount The amountof kernelconsumed per seedini- of kernel dropped (Fig. 1). tially increasedas seed mass increased but then To evaluate the seed handling efficiencies,I generallydeclined for larger sizedseeds (Fig. calculatedthe necessaryseed encounter rates 1). The massof kernel per seedapproximated for each species(see Methods). Seed sizes and that consumedby grosbeaks.For all species, minimum encounterrates by which eachspecies both the seedsize giving eachbird speciesthe could meet its energy demandsdiffered (Fig. maximumamount of kernel consumedper seed 2). White-winged Crossbillsrequired seed en- and the maximum seed size handled dimin- counter rates 2-3 times higher and Red Cross- ishedas the bodyand bill sizedecreased (Fig. bills required seedencounter rates about 2 times 1, Table 1). Relative to their bill dimensions higher than the mostefficient species at all seed (Table1), crossbillsconsumed proportionately sizes. Goldfinches could exist at lower encoun- less kernel than did the other finches for all but ter ratesof small seeds(2-23 mg) than the other the smallestseed sizes (Fig. 1). For example, species.House Finchesand grosbeakseach had White-wingedCrossbills have largerbills than a rangeof seedsizes for which they could exist goldfinches,yet theyconsumed less kernel than at lower seed encounter rates. 718 CROatW. BEN• [Auk,Vol. 105

2.6-

2.4

2.2

2.0

1.8 0.9 1.0 1.1 1.2 1.3 1.4

Fig. 3. Diagrammaticcross sections of the upper mandibleat the anterior edgeof the naresof (a) Eve- 2.7- ning Grosbeak,(b) House Finch, (c) American Gold- finch, (d) Red Crossbill,and (e) White-wingedCross- bill. The maxillary tomia and horny palate are designatedby mt and hp, respectively.The bar equals 0.5 cm.

DISCUSSION

Crossbillsrequire seedencounter rates 2-3 times higher than requiredby other cardue- 1.5 , - , , , lines. Because seed encounter rates decrease as 0.9 1.0 1.1 1.2 1.3 1.4 seedabundance declines, particularly at low seed densities(Schluter 1984, Benkman 1987a), other BODY MASS (In [cuberoot]) carduelinesout-compete crossbills by lowering Fig. 4. The relationshipsof bill depth (a) and bill seed abundancesbelow those necessary for width (b) (mm) to body mass(gm) for crossbills(solid crossbills to survive. When conifer cone crops circles)and noncrossbillcarduelines (open circles). fail over largeareas, crossbills probably suffer The fitted linear regressionline amongthe noncross- bills is shown. Data are from Table 1. high mortality(see Eriksson 1970). This will resultin strongnatural selection on crossbills for increasesin foraging efficiencyon thoseco- nifersthat produceseeds during conefailures, seed sizes handled was apparently a conse- but, asa consequence,further reducetheir abil- quenceof the Red Crossbillrequiring a pow- ity to survive on nonconifer seeds.This is the erful bite to obtain seed in closed, thick-scaled costof specializationfor crossbills. conifer cones (Benkman 1987b). Three characteristics of crossbills reduce their A second reason crossbills were less efficient efficiencyon nonconiferseeds. First, crossbills at meeting energy demandswhen foraging on have long narrow bills that enable them to be large seedsis becauseof differencesin the struc- slid efficiently between conifer cone scales ture of the horny palate.The noncrossbillfinch- (Benkman 1987b). Narrow bills, however, re- es have a high palate relative to the maxillary duce the efficiencywith which large seedscan tomia,whereas the crossbill'spalate is more flush be handled. For example,as bill width decreases with the maxillary tomia (Fig. 3; see also Zis- in either noncrossbill finches or crossbills, the wiler 1965). The low palate may facilitate han- consumedproportion of large-sizedseed ker- dling small seeds;crossbills usually consume nels decreased.This relationship probably re- seedsweighing less than 20 mg (Benkman1987a, suits from the ability of fincheswith different b). For a given bill width, noncrossbill finches sized bills to surround and secure the kernel have a larger buccal volume. Presumablythis when it is broken into piecesprior to swallow- will reducethe amountof kernel droppedwhile ing. kernels are being broken and swallowed.This Bill size was also correlated with the size of may explainwhy goldfinchesdropped less ker- the largest seedshandled. The wide range of nel than did White-winged Crossbills, even October1988] Costof Specialization 719 though goldfincheshave narrower bills than sparrowsand allies,part 1. U.S. Nat. Mus. Bull. White-winged Crossbills. 237., Washington, D.C. Crossbillshave larger bills and bodies than BENKM^N,C. W. 1987a. Food profitability and the foragingecology of crossbills.Ecol. Monogr. 57: birds in the genus Carduelis,yet crossbillstyp- 251-267. ically eat smallerseeds (Newton 1967, 1972; 1987b. Crossbill foraging behavior, bill Benkman1987a, b). The large bills of crossbills structure,and patternsof food profitability.Wil- are requiredto extractseeds from conifercones. son Bull. 99: 351-368. Most of the variation in bill size and shape in- 1988. On the advantagesof crossedman- creasesthe efficiencyof extractingseeds (Benk- dibles:an experimentalapproach. Ibis 130:288- man 1987b). In contrast, the general bill char- 293. acteristicsof noncrossbill carduelines may be , & H. R. PULLL•M. 1988. The comparative adapted most for handling seedand not secur- feeding rates of North American sparrowsand ing them from the plant. Indeed, when seeds finches. Ecology 69: 1195-1199. BOCK,C. E., & L. W. LEPTH1EN.1976. Synchronous are readilyaccessible in openconifer cones, Car- eruptionsof boreal seed-eatingbirds. Am. Nat. duelis finches and crossbills have similar feed- 110: 559-571. ing rates (pers. obs.). ERIKSSON,K. 1970. Ecologyof the irruption and win- A large body mass,however, increasesmet- teringof FennoscandianRedpolls (Carduelis fiam- abolic costs and is the third reason crossbills are meacoil.). Ann. Zool. Fennici 7: 273-282. relatively inefficient on nonconifer seeds. GRODZINSKI,W., & K. $AWICKA-KAPUSTA. 1970. En- Moreover, crossbillshave proportionatelylarg- ergy valuesof tree-seedseaten by small mam- er bodies in relation to their bill depth and rnals. Oikos 21: 52-58. width than other carduelines(Fig. 4). Largebody MARTIN, A. C., H. S. ZIM, & A. L. NELSONß 1951ß massmay be important for predator evasion Americanwildlife and plants.New York, Dover. NETHERSOLE-THOMPSON,D. 1975. Pine Crossbills: a becausea large body and wings may counter- Scottish contribution. Berkhamsted, T. & A.D. balancea largebill sothat flight is lessawkward Poyser. (Schluter 1988). Crossbills eat small seeds, and NEWTON,I. 1967. The adaptiveradiation and feeding must spend relatively more time feeding, and ecologyof someBritish finches. Ibis 109:33-98. they often forageout on the distal end of limbs ß 1970. Irruptionsof crossbillsin Europe.Pp. exposedto predators.Hence, crossbillsmay be 337-357in Animal populationsin relationto their more vulnerable to predators and their rela- food resources(A. Watson, Ed.). Oxford, Black- tively large bodiescompared to the other car- well Scientific Publ. duelines(Fig. 4a) maybe importantfor predator --. 1972. Finches. London, Collins. ROBERTS,T. S. 1936. The birds of Minnesota, 2nd evasion. ed., vol 2. Minneapolis, Univ. Minnesota Press. SCHLtSTER,D. 1984. Feeding correlatesof breeding ACKNOWLEDGMENTS and socialorganization in two Gal•pagosfinches. Auk 101: 59-68. Throughoutthis research,materials and spacewere 1988. Characterdisplacement and the adap- generouslyprovided by H. R. Pulliam. NSF Grants tive divergenceof fincheson islandsand conti- DEB-13017and DEB-8206936to Dr. Pulliam provided nents. Am. Nat. 131: 799-824. financial support. I received a PostdoctoralFellow- SMITH,C. C., & R. P. BALDA.1979. Competitionamong shipin EnvironmentalBiology during the final stages insects,birds and mammals for conifer seeds.Am. of writing. Discussionswith Dr. Pulliam and C. Stai- Zool. 19: 1065-1083. cerwere particularlyvaluable during the early stages WALSBERG,G. E. 1983. Avian ecological energetics. of the research. A. H. Brush, N. K. Johnson, M. Rau- sher, D. Schluter, S. Sherrod, and an anonymous re- Pp. 161-220in Avian biology,vol 7. (D. S.Farher, J. R. King, and K. C. Parkes,Eds.). New York, viewer made helpful commentson the manuscript. Academic Press. M. LeCroyallowed accessto specimensat the Amer- WILSON,M. F., & J. C. HARMESON.1973. Seed pref- ican Museum of Natural History. D. Nash graciously erencesand digestiveefficiency of Cardinalsand drew Fig. 3. Song Sparrows.Condor 75: 225-234. ZISWILER, V. 1965. Zur kenntnis des samen6ffnens LITERATURE CITED und der struktur desh6rnernen gaumensbei k6r- nerfressendenoscines. Journal fir Ornithologie AUSTIN, O.L. 1968ß Life histories of North American 106: 1-48. cardinals,grosbeaks, buntings, towhees, finches,