Community Organization in Hummingbirds: Relationships Between Morphology and Ecology

COMMUNITY ORGANIZATION IN HUMMINGBIRDS: RELATIONSHIPS BETWEEN MORPHOLOGY AND ECOLOGY

JAMESH. BROWNAND MICHAEL A. BOWERS Departmentof Ecologyand Evolutionary Biology, University of Arizona, Tucson, Arizona 85721 USA

AssTRACT.--Hummingbirdspecies inhabiting restrictedgeographic regions exhibit morphologicalpatterns that differ significantlyfrom thosepredicted by null modelsin which speciesare selectedat random from appropriatespecies pools. Temperate North American hummingbirds are convergent:more similar in bill length, body weight, and wing length than predictedby severalnull models.These temperate species also are more similar to each otherthan they are to moreclosely related (congeneric) species from subtropicaland tropical habitats.Hummingbirds of the Greater and LesserAntilles show a nonrandom distribution of speciesamong islands:all islandsinhabited by hummingbirdshave at least two species, and thesefall into two distinctly different size categories.Allometric scalingof bill length with respectto body massis distinctivein Antillean hummingbirds;bill length increases more rapidly with increasingbody weight in West Indian hummingbirdsthan in random samplesof hummingbirdsof the world or in other birds. These morphologicalpatterns appear to reflect two ecologicalprocesses: interspecific competition among hummingbirds and mutualisticcoevolution with flowers.Hummingbird speciesof similar morphologyuse similar floral resourcesbut rarely coexistin the samelocal areas.Species of divergent mor- phologyexploit differentfood resourcesand frequentlycoexist locally. Length of the bill, which influencesaccess to different kinds of flowers,is particularlyimportant in the organization of these simple hummingbird associations.Received 28 June1983, accepted 5 July 1984.

EVOLUTIONARYecology has a long tradition lems can be overcome in some casesby reana- of inferring dynamicprocesses of ecologicalin- lyzing information already available for cer- teraction from static patterns in the morphol- tain groups of organisms. Although these ogies and geographic distributions of closely analysesare no substitute for careful experi- related species(e.g. Lack 1947,Brown and Wil- mental and observationalfield studies,they may son 1956, Hutchinson 1959, MacArthur 1972). clarify issuesand focus future empirical work. Such patterns have played a particularly im- Hummingbirds, the nonpasserineavian fam- portantrole in the developmentof currentideas ily Trochilidae,provide an excellentsystem for about the role of competition and other inter- addressing these issues. This presumably specificinteractions in the organization of nat- monophyleticgroup, containing more than 100 ural communities(e.g. MacArthur 1972, Cody genera and 300 species,is widely distributed in and Diamond 1975). Recently, however, these the New World, attaining its greatest diversity ideas have been criticized for two reasons. First, in tropical latitudes. Hummingbirds are spe- manyof the purportedpatterns have not been cialized for feeding on floral nectar and share rigorously tested against appropriate null a suite of characteristicsincluding long, slen- models that assumethey could result from ran- der bills, tiny feet, and wings adapted for hov- dom processesrather than deterministic bio- ering flight. Becausethey are easily observed logical mechanisms(e.g. Strong et al. 1979, and their food resources can be monitored ac- Connor and Simberloff 1979, Simberloff and curately,hummingbirds have been the subjects Boecklen 1981). Second, the relationship be- of many behavioral and ecologicalstudies. tween morphologyand ecologyoften is simply Several investigators have suggestedthat assumed or inferred from indirect evidence in- speciesin particular geographicareas exhibit stead of being documented rigorously by ob- morphological patterns that reflect evolution- servation and experiment (e.g. Lack 1947, ary adaptationsboth for competitive interac- Hutchinson 1959, Findley 1973, Cody 1974, tions with other hummingbirds and for mutu- Brown 1975, Karr and James1975). These prob- alistic interactionswith bird-pollinated flowers

251 The Auk 102:251-269. April 1985 252 BI•OWNAND BOWEI•S [Auk, Vol. 102

(e.g. Grant and Grant 1968, Kodric-Brown and clusive use of local areas by individuals and Brown 1978, and Brown and Kodric-Brown 1979 species,and its outcome depends largely on for temperate North America; Lack 1973, 1976, habitat(including flower density)and wing disc and Kodric-Brown et al. 1984 for Caribbean is- loading (the ratio of body weight to the area lands; Snow and Snow 1972, Colwell 1973, swept out by the wings in flight; Feinsinger Feinsinger 1976, and Stiles 1975 for tropical and Chaplin 1975, Kodric-Brown and Brown America; and Feinsinger and Chaplin 1975, 1978,Feinsinger et al. 1979). Feinsingerand Colwell 1978,and Feinsingeret The claim that temperatehummingbirds are al. 1979 for hummingbirds in general). Al- convergentin morphologyhas not been tested though these interactionshave been studied rigorouslyagainst a null hypothesis.If the claim carefully in certain instances,there has been is correct,then temperatehummingbirds should little attempt to test the purported patterns of be more similar to each other than a compara- morphologicalorganization against alternative ble number of speciesdrawn at random from models that assume no interspecific interac- an appropriatespecies pool. We testedthis hy- tions. We performed such testsfor the species pothesisby comparingmeasurements of body inhabiting temperate North America and the weight, bill (exposedculmen) length, and wing Antillean islands, and the results are discussed length for the temperate specieswith sets of here in terms of what is known about the pro- eightspecies randomly drawn from a largepool cessesof ecologicalinteraction. Our analysisis of speciesof hummingbirds of the world for basedon three quantitativemorphological traits which measurementswere available (see Ap- that are known to be important in humming- pendix). We conductedthree tests of the null bird ecology:culmen (bill) length, which influ- hypothesisthat the temperatehummingbirds ences the kinds of flowers from which the birds are no more similar in size than randomly as- can forage;body weight, which indicatestotal sembled associationsof species. energy requirements;and wing length, which First, we used computer simulation tech- together with body weight affectsthe aerody- niques to draw 8 speciesat random 1,000 dif- namicsof flight (Greenewalt 1975). ferent times from the pool of 123 speciesfor which all measurements were available. For TEMPERATE NORTH AMERICAN eachset of eight specieswe computedthe vari- ASSOCIATIONS ance for each of the three measurements, and Eight species,belonging to four genera,have we recorded the number of cases in which the breeding rangesthat extend well into temper- variancefor a null communityexceeded that of ate North America. Several other species,not the observed temperate assemblage.For all consideredfurther here, are primarily subtrop- three characters, variance in the random com- ical in distribution and reach their northern munities was greater than in the real commu- limits just across the Mexican-United States nity for all 1,000draws (Table 1). Clearly, tem- border. The collective breeding range of these perate hummingbirdsare much more similar temperate speciesextends from the deserts of in size and shape than a random subsetof the the southwestern United States and northern hummingbirds of the world (Fig. 1). Mexico to southern Alaska and from Florida to A random draw from the global speciespool southeasternCanada. All of these speciesare may not be an appropriatemethod for evalu- migratory, wintering in Mexico and Central ating certainnull hypotheses,however. For one America. thing, the temperatespecies probably are more Several investigatorshave noted that the closelyrelated to eachother than to certain ex- north temperatehummingbirds are extremely clusively tropical groups.Thus, the similarity similar in sizeand shape.Grant and Grant (1968; could be attributed to failure to diverge from a see also Brown and Kodric-Brown 1979) sug- recent common ancestor rather than to conver- gested that this similarity representsconver- gencein responseto similar ecologicalcondi- gence among speciesto utilize a common set tions. Biogeographicand taxonomicrelation- of widely distributedflower species,which also shipssuggest that temperateNorth American have converged in size, shape, and nectar re- hummingbirdsderived from tropical montane wards.Hummingbirds compete for thesefloral and subtropicalancestors that colonized from resourcesprimarily by aggressionand territo- Mexico. The morphology of temperate hum- riality. This interferenceusually results in ex- mingbirdsalso might be constrainedby the de- April1985] HummingbirdMorphology andEcology 253

TABLE1. Resultsof simulationstesting the null hypothesisthat the eight speciesof temperateNorth Amer- ican hummingbirdsare more similar than expectedfrom a random draw of eight speciessampled from the world speciespool. Simulationscompared the actualvariances of morphologicalcharacters of temperate speciesagainst variances of the entire world speciespool, the speciespool that inhabitscontinental North America north of the Isthmus of Tehuantepec(biogeographic constraints), and speciespool with wing lengthsand body weights within the range observedfor temperatespecies (morphological constraints).

Number of random draws with variances greater than observedvariance World Variancein 8 North species Biogeographic Morphological American species pool constraints constraints Culmen length (mm) 1.75 1,000 996 1,000 Wing length (mm) 15.21 1,000 999 248 Weight (g) 0.203 1,000 999 523

mands of the fluctuating temperate climate or had body weights and wing lengths within the by the necessityto migrate from the temperate range (2.66-4.25 g and 39.8-49.7 mm) for tem- zones to winter in tropical and subtropical re- peratespecies. Note that both the biogeograph- gions and survive in these habitats (see Fein- ically and the morphologically constrained singer 1980). Such constraintswould indicate species pools are also taxonomically con- that ecologicalfactors influence the morphol- strained, becausethey contain only 18 and 17 ogy and community organizationof humming- genera, respectively, of the 70 genera in the birds, but would suggestthat other conditions world speciespool for which all measurements probablyare more important than biotic inter- are available for at least one species. actions with flowers or with other humming- When 1,000 draws of 8 specieseach are taken bird specieswithin the temperatezone of North from either the biogeographicallyor morpho- America. logically constrainedpools, the variation in the We then testedtwo null hypothesesthat im- relevant traits is much greater than for the tem- poseadditional constraintson the speciespool. perate hummingbirds (Table 1). This is true for Unfortunately there is no modern, generally all three charactersin the test with biogeo- accepted classificationof the family Trochili- graphicconstraints, but of courseonly for cul- dae that attempts to reconstructphylogenetic men lengths in the test with morphological relationshipsamong the genera, so we did not constraints(since all speciesin the pool were feel confident restricting our species pool to within the range of variation of the temperate some taxonomicsubgroup within the family. speciesin both body weight and wing length). Instead we drew random communities from a The hypothesis that temperate humming- speciespool limited by either biogeographicor birds are similar, not as a result of evolutionary morphologicalconsraints. In a secondtest we convergencebut simply becausethey have not limited the speciespool to those 29 species(for diverged substantiallyfrom a common ances- which we have measurementsout of a possible tor, also can be evaluatedby comparingvaria- 38 species) that inhabit continental North tion within and between genera. Two of the America north of the Isthmus of Tehuantepec four genera of temperate hummingbirds have in southern Mexico (Peterson and Chalif 1973). other specieswith exclusively tropical distri- Becausemany biogeographersrecognize the butions. If temperate hummingbirds have not regionbetween the Isthmusof Tehuantepecand diverged significantly from a common ances- the United States-Mexican border as the area tor, these tropical congeners should also be of subtropicaltransition between tropical South similar in size and shape to their temperate rel- and Central America and temperate North atives. Clearly this is not the case (Table 2). America (Brown and Gibson 1983), this proce- Tropical speciesof the genera Selasphorusand dure defined a speciespool consistingof 8 tem- Calypteare uniformly smaller than their temperate North American speciesand 21 subtrop- perate congeners. The temperate representa- ical Mexican species.In a third test we limited tives of these genera are similar to each other the speciespool to only those 28 speciesthat and to the other two monotypicgenera of tern- 254 BROWNAND BOWERS [Auk,Vol. 102

TEMPERATE NORTH AMERICA

• GREATERAND LESSER ANTILLES

WORLD

LOG BOOYWEIGHT (g) LOG CULMEN LENGTH (f'om) LOG WING LENGTH(ram) Fig. 1. Frequencydistributions of body weights(g), culmen lengths(mm), and wing lengths(mm) for speciesof hummingbirdsfrom temperateNorth America, the Greater and LesserAntilles, and the world. Note the smallvariation among temperate species and the bimodalpattern for Antillean speciescompared to the world speciespool.

peratehummingbirds, especially in bill length. era of hummingbirds(Bond 1971).All but 2 of The greatersimilarity among temperatespecies these speciesare endemic to the Greater and than among congenericspecies provides strong Lesser Antilles. Other speciesnot considered evidence that the temperate hummingbirds in the presentanalysis occur on islandsaround have acquired their strikingly similar mor- the periphery of the CaribbeanSea (the Baha- phologiesby evolutionaryconvergence. mas, Old Providence, St. Andrew, Trinidad, and These analysesdemonstrate that temperate Tobago),but theseislands differ conspicuously North Americanhummingbirds are much more in habitat and geologicalhistory from the An- similarin morphology,especially in bill length, tilles. As noted by Lack (1973, 1976), the Antil- than expectedon the basisof several different lean hummingbirds appear to be distributed null hypotheses.Because bill length is of pri- among the 36 islands in nonrandom patterns. mary importance in determining the kinds of Although Mona, a relatively large island be- flowersfrom which hummingbirdscan forage tween Puerto Rico and Hispaniola, has no (e.g. Snow and Snow 1972,Colwell 1973,Stiles hummingbirds, the remaining 35 have 2-5 1975, Feinsinger and Colwell 1978, Kodric- species(Table 3). Lack noted that speciesoc- Brown et al. 1984), these results also provide curring together on the same island tended strong support for the idea (Grant and Grant either to differ substantiallyin body size and 1968, Brown and Kodric-Brown 1979) that the overlap widely in local distribution or to be birds have convergedto use the samekinds of relatively similar in size but occur in different flowers. It seemshighly unlikely that physio- habitats,usually at different elevations. Inten- logical adaptations to climate or migration sive work on Puerto Rico and more casual ob- would result in more precise convergencein servationson other islandssuggest that Antil- bill length than in either body size or wing lean hummingbirds have morphologiesthat length. reflect both interspecificcompetition with other hummingbirds and coevolutionwith mutu- ANTILLEAN COMMUNITIES alistic flowers (Kodric-Brown et al. 1984). The situation is similar to that in temperate North The tropical islandsof the Greaterand Lesser America, exceptthat in the Antilles there are Antilles are inhabited by 14 speciesand 9 gert- two size categoriesof hummingbirds that are April1985] HummingbirdMorphology andEcology 255

TABLE2. Mean (œ),range, and coefficientof variation (CV), with standard deviation (Soy)' in parentheses, for temperatehummingbirds and temperate,tropical, and combinedspecies of Selasphorusand Calypte.

Temperate TemperateSelasphorus Tropical Selasphorus All Selasphorus hummingbirds and Calypte and Calypte and Calypte (8 species,4 genera) (5 species) (6 species) (11 species) Culmen length (mm) œ 17.66 17.61 11.77 14.72 Range 15.54-19.73 16.57-19.06 10.59-12.90 10.59-19.06 CV 0.0747 (+0.0188) 0.0521 (+0.0165) 0.0790 (+0.0228) 0.2201 (+0.0469) Wing length (mm) œ 43.92 45.15 37.86 41.17 Range 39.77-49.68 39.77-49.68 31.61-41.57 31.61-49.68 CV 0.0887 (+0.0222) 0.0975 (+0.0308) 0.1033 (+0.0298) 0.1329 (+0.0283) Weight (g) œ 3.29 3.46 2.17 2.98 Range 2.66-4.25 3.08-4.25 1.90-2.47 1.90-4.25 CV 0.1370 (+0.0343) 0.1349 (+0.0427) 0.1315 (+0.0379) 0.2578 (+0.0549) a S• calculated from Sokal and Rohlf (1969). highly specificpollinators of two distinct sets approximately20-50% of the null assemblages of hummingbird-pollinatedplant species:small, were more different than the actual Antillean short-billed hummingbirds can forage only specieswhen both were comparedto the entire from short-tubed flowers that have low rates of speciespool (Table 4). This is hardly convinc- nectar secretion, whereas large, long-billed ing evidence for the nonrandom assemblyof birds feed almostexclusively from long-tubed Antillean communities, but note that fewer flowers that secrete more copious nectar. simulated distributions exceeded the observed The relationship between Antillean hum- differencefor culmen length (102 out of 500; mingbird morphologyand ecologycan be ex- P = 0.2) than for the other two measurements. amined more rigorouslyby comparingthe ob- Comparisonwith a randomizedAntillean species served morphologies and distributions of pool.--We next testedwhether the distribution specieswith thoseexpected on the basisof null of speciesmorphologies differed from that ex- models that assume that insular communities pectedif the 14 specieswere redistributedran- are assembledat random, uninfluenced by in- domly among islands in the frequenciesthat terspecificinteractions. they actually occur.There are 35 islandsinhab- Comparisonwith the global species pooL--We first ited by hummingbirds, but the same combi- tested whether the morphologiesof Antillean nationsof speciesoccur together on severaldif- hummingbirds differ significantly from a ran- ferent islands(Table 3). Thus, it is inappropriate dom sample drawn from the entire pool of for most tests to consider each island as an in- hummingbirds of the world. For each of the dependent sample.To be rigorousand conser- three measurements(body weight, culmen vative we usedonly the nine different combi- length, and wing length) we drew at random nations to obtain the observed distributions for from the pool 14 speciesfor which the relevant each measurement.We generated an expected data were available. This was repeated 500 distribution by drawing nine different combi- times. We quantified the maximum difference nations of speciesat random from the pool of between the actual Antillean distribution and 14 Antillean species(Table 3). This draw was that for the entire world speciespool, and then constrainedto have the samenumber of species determined the number of random assem- in each combination as the real combinations blagesthat were more different than this when (3 two-species,4 three-species,1 four-species, also compared to the world speciespool. We and 1 five-speciescombination). This proce- measured the difference using the D-statistic dure was repeated750 times to generatean ex- from the Kolmogorov-Smirnovtwo-sample test, pected frequency distribution for each mor- which is the maximum cumulative deviation phologicaltrait. Becausethe birds in both the over correspondingintervals between two dis- real and artificial assemblagesdiffer in absolute tributions(Siegel 1956). Depending on the trait, size, and we wanted to evaluatethe hypothesis 256 BROWNAND BOWERS [Auk,Vol. 102

T^BLE3. The distribution of hummingbird specieson 39 islandsof the Greater and LesserAntilles. From Bond (1971) and Lack (1973).

Grenada x x Carriacou x x Grenadines x x St. Vincent x x Barbados x x St. Lucia x x La Martinique x x Dominica x x Marie Galante x x La Guadeloupe x x D•sirade x x Montserrat x x Antigua x x Barbuda x x Nevis x x Saba x x St. Eustatius x x St. Christopher x x Anguilla x x St. Martin x x St. Bartholomew x x Bequia x x Anegada x x St. Croix x x St. Thomas x x Puerto Rico X X x x x Mona •ispaniola x x Ile de la Gon•ve x x Tortuga Island, Haiti x x Jamaica x x Cuba x x Isle of Pines x x Culebra Vieques St. John

that coexisting speciestend to differ in size a Kolmogorov-Smirnov test (Fig. 2). Differ- more than expectedon the basisof chance(Lack ences between observed and null are signifi- 1973,1976), we expressedthe differencesin size cant for culmen length (P < 0.05), marginally between every pair of speciesin all combina- significant for body weight (0.05 < P < 0.10), tions (both real and randomly assembled)as and insignificant for wing length (P > 0.10). the difference in the logarithms for each mea- Note that it is particularly with respect to bill surement. length that real communities represent non- The observed and null distributions were random combinationsof morphologies. Also, comparedand the significanceevaluated with the observed distributions for both culmen April 1985] HummingbirdMorphology and Ecology 257

TABLE 4. Simulation results that tested whether 3

morphological charactersof Antillean humming- .40- bird speciesdiffer significantly from a sample of 14 speciesdrawn at random from the world species .•0' pool .gO- Number Charac- Number greater •' .10- ter D-statistic less or equa! Z Wings 0.15 246 254 • Weights 0.18 364 136 O Culmens 0.20 398 102 • .40-

h .30-

.20- length and body weight differ conspicuously .10 from the null assemblagesin having a pro- nounced peak at a difference in logarithms of approximately 0.30, which correspondsto a .40- Wings body size ratio of about two. This is a conser- vative analysisin the sensethat it incorporates .30' all specieswithin islands.When they occuron .20- the same island, speciesof similar size often [] Null are segregated into different elevations and .10- habitats (Lack 1973, 1975; confirmed on Puerto Rico by Kodric-Brown et al. 1984). i•.06 .12 .18 .24 .50 .56 [].42 Observed.48 .54 Distributionof speciesamong islands.--We tested directly againsta null model Lack's(1973, 1976) DIFFERENCE Of LOGARITHMS suggestionthat the Antillean assemblagesare Fig. 2. Comparisons between observed distribu- nonrandom with respect to the number of tions of measurements for Greater and Lesser Antil- speciesper island. Becausethere is one island lean hummingbird speciesand those measurements (Mona) with no hummingbirdsand many with predicted from a null model that assumesthat com- two species(Table 3), Lack suggestedthat, in binations of speciesare assembledat random from particular,the absenceof islandswith only one the pool of 14 Anti!lean species.Measurements are specieswas unlikely to be due to chance. expressedas differencesin the logarithms for each We tested the observed distribution of num- pair of species. bers of speciesamong islandsagainst a Poisson distribution. This null model predicts the in- dependent distribution of rare events. It is ap- that can supportone hummingbird speciescan propriate here becausethe number of species be invaded successfullyby at least one addi- per island is small and apparently not influ- tional species.The precisereasons for this are enced significantly by island area or interis- obscure,but we suspectthey have to do with land distance(Lack 1973, Terborgh 1973). Re- the successin colonizationby hummingbird-pol- suits of this analysis indicate that the actual linated plant speciesas well as hummingbirds, distribution differs (P < 0.01) from the pre- and by the coevolutionof the hummingbird- dicted null distribution in having too few is- plant interactions within islands (see Kodric- lands with only one species,too many islands Brown et al. 1984). with two species,and too few islandswith three Allometryof billsand body size.--A conspicu- or more species(Fig. 3). This pattern is highly ousfeature of Fig. 2 is the pronouncedpeak in nonrandom (X2 = 27.26, P << 0.01; Fig. 3 shows observedmeasurements of both body weights the data plotted by number of speciesper is- and culmenlengths of coexistingcombinations land, but some adjacent categories can be of speciesthat occurs at a ratio of about 2.0, lumped to give sufficientlyhigh expectedval- whereasthe peak in wing length corresponds ues to meet the assumptionof the Chi-square to a ratio of about 1.2. This indicates unusual analysis).This pattern suggeststhat any island allometric scaling of body parts in Antillean 258 BROWNAND BOWERS [Auk,Vol. 102

ly twice as high as the exponents for either hi bills or wings (0.32-0.53) of any of the other Z20 hi birds (Fig. 4). Statistically,the exponentfor An- tillean hummingbirdbills can be readily dis- 015 tinguishedfrom the exponentsof all the other ] Observed o allometric relationships (P < 0.01), which are [] Expecled all much more similar to each other (Table 5). b. io o The bills of Antillean hummingbirdsare much more different in length than would be ex- z 5 pected on the basis of body weight. Also of interest, although it will come as no surprise c• to studentsof hummingbirds,is the great vari- b3 o 2 3 4 5 ation around the allometric relationshipsfor bills (r2 = 0.39) of hummingbirdsin the global NUMBER OF SPECIES PER ISLAND pool, comparedto comparablerelationships for Fig. 3. Comparison between the observed num- either Antillean hummingbirds, owls, or cor- ber of hummingbirdspecies among 36 islandsof the vids (all r2 > 0.79). This indicates that differ- Greater and LesserAntilles and that predicted by a encesin bill length (and shape) play such an Poisson distribution. importantrole in the ecologyof hummingbirds that selectionhas producedmorphologies of many speciesthat deviate greatly from pre- hummingbirds.Because mass and volume scale dicted allometric relationships(see also Fein- as the cube of linear measurements,body singer and Colwell 1978). weight ratiosof about 2.0 usually are associated The unusualallometric scaling of bill length with ratios of linear measurements of 1.2-1.3. suggestsa final and admittedly ad hoc test of This typical allometry is observedfor the re- the null hypothesisthat Antillean humming- lationshipbetween body weight (a volumetric birds representa random sample of the hum- measurement)and wing length (a linear mea- mingbirds of the world. We used Monte Carlo surement),but bill lengths of different-sized methodsto draw 14 speciesat random from the Antillean hummingbirds are more different 123 speciesfor which both culmen length and than would be expectedon the basisof body body weight data were available. Then we cal- weight ratios. culated the correlation coefficient for the 14 These relationshipscan be examined more pairs of log-transformeddata. This procedure rigorouslyby deriving the allometricequations was repeated 1,000 times. In 996 of 1,000 sets from regressionsfitted to the data.These equa- of speciesassembled randomly from the world tions take the form l = CW'" where l is a linear species pool the correlation coefficient was dimension(bill length or wing length in mm), lower than the value (0.79) calculated for the C is a fitted constant, W is a volumetric mea- real Antillean fauna, resulting in unequivocal surement (body weight in g), and rn is the ex- rejectionof the null hypothesis(P = 0.004). ponent relating the two variables(or the slope All Antillean hummingbirdsappear to con- of the regressionline fitted to log-transformed form to a special relationship between bill data). In Fig. 4 we have plotted fitted regres- length and body size that is very different from sion equations for log-transformed measure- comparablerelationships both for humming- ments of Antillean hummingbirds,hummingbirds in general and for other kinds of birds. birds of the world, North American corvids Why should speciesthat differ in body size be (crowsand jays),and North Americanowls. We even more different in bill length than expect- included the last two groups becausewe be- ed on the basisof the allometric relationship lieve they are representativeof the general al- in other hummingbirds, unless speciesof dif- 1ometricscaling of bill length and wing length ferent sizes tend to occur together and have in birds. Each of these two families exhibits a been selectedto diverge in bill length? Both wide range of body sizesbut doesnot use its Lack's qualitative observationsand our quan- bill for foraging in the samespecialized way titative analyses(Figs. 1-3) show that pairs of that hummingbirdsdo. The exponent(slope) speciesof different size usually occurtogether for Antillean hummingbird bills (0.79) is near- on the same islands.Although these distribu- April 1985] HummingbirdMorphology and Ecology 259

HUMMINGBIRDS OWLS and CORVIOS

z 1.55 2.0[ CORVID$ LLI ._1 i 4•

z oO ....

o o.• ._1

2.52.6 o .:••o o

z t so

12'0

2.O o .a 1.6 I 8 2 0 2 2 2.4 2.6 2.8 3.0 3.2 LOG BODY WEIGHT (g) Fig.4. Allometricrelationships of culmenand wing lengthagainst body weight for Antilleanhumming- birds,hummingbirds of the world, and North Americanowls and corvids.Note that the slopefor Antillean hummingbirdculmens (solid circles) is nearlytwice that of eitherbills or wingsof any of the otherbirds. tions may reflect historical and contemporary Only large hummingbirds with long bills can biogeographicconstraints as well as contem- legitimately extract nectar from long-tubed porary biotic interactions,these species appear flowers,which producesufficient quantitites of to have adapted to coexistenceby diverging in nectar to make foraging rewarding for large body size and especiallyin bill length. This is birds. On the other hand, both large and small supportedby taxonomicbody size patterns.For birds can feed from short-tubed flowers, but example, Chlorostilbonricordii on Cuba and C. these produce such small quantitites of nectar swainsoniion Hispaniola, where they coexist that foragingis economicalonly for the smaller with smaller hummingbirds, are much larger species. Thus differences among coexisting than C. maugaeuson PuertoRico and other con- speciesin bill length and body size are func- genericspecies on the tropicalAmerican main- tionally interrelated and appear to reflect se- land (see Appendix). lection both to diverge in resourceutilization This interpretation alsois consistentwith the so as to reduceinterspecific competition and to detailed ecological observations of Kodric- coevolve with specific flower speciesso as to Brown et al. (1984) on Puerto Rican humming- provide effective pollination. birds. They confirmed Lack'sobservation that speciesof similarsize are segregatedby habitat DISCUSSION and elevation,whereas species of dissimilar'size are broadly and locally sympatric.These pairs Our analysesdemonstrate that specificmor- of coexistingspecies feed from almost com- phological traits characterize the humming- pletely nonoverlappingsets of flower species, birds from particulargeographic areas. Species and flower utilization is related to bill length. from temperate North America differ from 260 BROWNAND BOWERS [Auk, Vol. 102

TABLE5. Resultsof t-teststhat test the null hypothesisthat slopesbetween culmen length or wing length and body weight are statistically(P < 0.05) equal.All comparisonswere done on log-transformeddata (see Fig. 3).a

Culmens Wings Hum- Antillean Antillean North North ming- hum- North North humming- American American birds of ming- American American birds owls corvids the world birds owls corvids

Culmens Antillean hummingbirds -- North American owls 5.87** -- North American corvids 4.63** 1.25 Hummingbirds of the world 3.01'* 2.37* 1.27 Wings Antillean hummingbirds 4.37** 1.52 0.25 !.00 -- North American owls 4.75** 1.13 0.13 1.25 0.37 -- North American corvids 3.87** 2.01 0.87 0.36 0.50 1.16 Hummingbirdsof the world 4.63'* 1.26 0.13 1.25 0.25 0.13 1.05

a, p < 0.01, ** P < 0.001.

those inhabiting the tropical islands of the different size coexistlocally to form the basic Greater and Lesser Antilles, and each of these two-speciescommunity that is found on all is- groupsrepresents a selectedsubsample of the lands with hummingbirds (Lack 1973, 1976; hummingbirds of the world. These morpho- Kodric-Brown et al. 1984). logical patterns indicate that ecological pro- These results suggestthat it should be pos- cessesseverely limit the kinds of humming- sible to predict quite preciselythe morpholog- birds that comprise communities in different icalcharacteristics of the hummingbirdspecies • geographic areas. Ecological constraints, of occupying different habitats and geographic which the mostimportant is probablythe num- regions. Depending on the flowers and per- ber and kinds of floral resources,limit the range haps on other environmental conditions,only and combinationsof morphologicaltraits of the specieswith certain combinationsof morphol- speciesthat inhabit a region. Competitive in- ogies should be able to coexistto form stable teractionsbetween hummingbirds, on the oth- communities.We have begun to characterize er hand, require speciesthat coexist in local these traits for temperate North American and areasto differ in morphology sufficientlyto ex- Antillean communities.We expectother hum- ploit substantially nonoverlapping food re- mingbird associations exhibiting different sources. The eight hummingbird species in combinationsof morphologiesto inhabit other temperate North America utilize a large set of areas, especially tropical continental habitats convergentlysimilar floral resources(Grant and where several speciescoexist locally. Available Grant 1968, Brown and Kodric-Brown 1979). information suggeststhat tropical mainland Consequentlythey exhibit little variationin bill communitiesoften containmorphologically di- length and body size, and only one species verse hummingbird assemblages,with species normally is found within a local habitat (Ko- exhibiting different combinationsof body size, dric-Brown and Brown 1978, Brown and Ko- wing length, bill length, and bill shape(Snow dric-Brown 1979). The Antillean islands are in- and Snow 1972, Stiles 1975, Feinsinger 1976, habited by two body size and bill length Feinsinger and Colwell 1978). categoriesof hummingbirds that have coe- Feinsingerand Colwell (1978) alsohave been volved with two distinct sets of flower species impressed by the close correspondencebe- (Kodric-Brownet al. 1984). Hummingbirds of tween morphologyand communityecology of similar size tend not to occur on the same is- hummingbirds.Based largely on their own ex- land, and when they do are normally found in periencesin continental tropical America, they different habitats, whereas hummingbirds of have attemptedto deducethe rules that govern April 1985] HummingbirdMorphology and Ecology 261

the assemblyof speciesinto communitiesof in- cations.Recently there has been much debate creasingdiversity. Interestingly, they describe about whether ecologicalcommunities exhibit the basic two-speciestropical community as nonrandom patterns of organization that re- consisting of a territorialist and a low-reward flect ecologicalprocesses such as interspecific trapliner or generalist,but thesecategories do competition.Several authors (e.g. Connor and not accurately characterize the two-species Simberloff 1979, Strong et al. 1979, Simberloff communities of the Antillean islands, where and Boecklen1981) have pointed out correctly the main differencesbetween locally coexist- that apparentpatterns in the morphologiesand ing speciesare in body size, bill length, and distributionsof speciesseldom have been tested flower specialization.In fact, both large and to determinewhether they differ significantly small speciestend to be trapliners, at least in from appropriate null models that assumethat the Greater Antilles (Kodric-Brown et al. 1984). speciesassociate at random. However, here and Nevertheless,the ecologicalconstraints and elsewhere (Bowers and Brown 1982, Brown and opportunities that affect the occurrence of Bowers 1984) we have shown that communities hummingbirdsin a particularregion obviously of both rodents and hummingbirdsexhibit are reflectedin the morphologiesof coexisting nonrandom organization that apparently re- species,suggesting that it eventually may be flectsthe role of interspecificcompetitive and possibleto formulatecommunity assembly rules (in the caseof hummingbirds)mutualistic in- that are both preciseand general. teractions.In both casesthere already existeda One general result of almost all of our anal- largebackground of field studieson the ecol- ysesis that of the three morphologicalcharac- ogy of thesespecies that included analysesof teristicsstudied, bill length consistentlyshows the relationshipbetween morphological traits the clearest, most nonrandom patterns. This and processesof resourceexploitation and ag- may not seemsurprising in view of the obvious gressiveinterference. This facilitated the for- importance of bill size and shape in the for- mulation and testing of null hypothesesthat aging of hummingbirds.Nevertheless, it em- are not only statisticallyrigorous but alsobio- phasizes the roles of food resources, inter- logicallyappropriate. It is hardlysurprising that specificcompetition for food, and mutualistic a goodknowledge of the biologyof the organ- plant-pollinatorinteractions in determiningthe isms concerned is a valuable aid in construct- morphologicalattributes and ecologicalroles ing and testingrealistic null hypothesesand in of those species that coexist to form natural elucidatingthe ecologicalprocesses that influ- communities.Not only do these kinds of eco- ence community organization. logical interactionslimit the morphologicalat- In investigating the relationship between tributes of the speciesthat occurwithin certain morphologyand communityecology, it is often geographicareas, but they also influence the usefulto restrictthe analysisto closelyrelated evolution of these traits. In the case of hum- species.This is not to imply that only closely mingbirds, mutualistic interactions between related speciesinteract strongly to influence birds and flowers may be at least as important community organization.On the contrary, dis- ascompetition among hummingbird speciesin tantly related taxa often are involved in all the adaptiveradiation of the entire family Tro- classesof interspecificinteractions, including chilidae and the evolution of those subsets of competition.However, their morphologiesmay speciesthat inhabit the particulargeographic be so different that they provide little clue to regions.In fact, it is probablyunrealistic to try their specificecological roles. For example,on to assessthe relative importance of these two most Antillean islands a passefinebird, the kinds of interactionsbecause they are intimate- bananaquit(Coereba fiaveola), removes nectar ly interrelated;convergence or divergencein from long-tubedflowers and presumablycom- bill length and other aspectsof morphology petes significantlywith large hummingbirds affect the relative abilities of different species (Kodric-Brownet al. 1984). In part becauseof to use particular kinds of flowers, and this in its distant phylogeneticrelationship, its mor- turn influencesboth the extent of interspecific phology is adapted for a different manner of competitionand the efficiencyof pollination. foraging:it cutsthrough the floral tube while The relationship between morphology and perched,rather than probing through the flow- community ecology that we have documented er opening while hovering as hummingbirds for hummingbirdshas severalgeneral impli- do. Consequently,its morphologicalcharacter- 262 BI•OWNAND BOWERS [Auk, Vol. 102 istics provide little indication that it interacts ß & A. KODRIC-BROWN.1979. Convergence, much more closely with hummingbirds than competition and mimicry in a temperate com- any other Antillean passefine. munity of hummingbird-pollinated flowers. Closely related species often are so con- Ecology60: 1022-1035. BROWN, W. L., & E. O. WILSON. 1956. Character dis- strained by their common phylogenetic histo- placement. Syst. Zool. 5: 49-64. ries that they differ substantiallyin only a few CARPENTER,F. L. 1976. Ecologyand evolution of an characteristics.Often, especially in sympatric Andean hummingbird ( Oreotrochilusestella). species, these differences reflect divergent Berkeleyand LosAngeles, Univ. California Press. mechanisms of resource utilization in response COD¾,M.L. 1974. Competition and the structureof to interspecificcompetition. Thereforeß they tell bird communities. Princeton, New Jersey, more about the role of ecologicalinteractions Princeton Univ. Press. in causingcharacter displacement and adaptive , & J. M. DIAMOND. 1975. Ecology and the radiation within phyletic lineagesthan they do evolution of communities. Cambridge, Massa- about the organization of diverse communities chusetts, Harvard Univ. Press. containing many distantly related taxa. Anal- COLWELL,R. K. 1973. Competition and coexistence in a simple tropical community. Amer. Natur. ysis of morphologicalpatterns among closely 107: 737-760. related speciesis useful becauseit may suggest CONNOR,E. F.ß•: D. SIMBERLOFF.1979. The assembly mechanismsof ecologicalinteraction that affect of speciescommunities: chance or competition? communitystructure, but otherapproaches will Ecology60: 1132-1140. be necessaryto explore the consequencesof FEINSINGER,P. 1976. Organizationof a tropicalguild theseinteractions among distantly related taxa. of nectivorousbirds. Ecol. Monogr. 46: 257-291. 1980. Asynchronousmigration patternsand

ACKNOWLEDGMENTS coexistenceof tropical hummingbirds. Pp. 411- 419 in Migrant birds in the Neotropics(A. Keast We thank R. Zusi, R. K. Colwell, and J. B. Dunning and E. S. Morton, Eds.). Washington, D.C., for providing unpublished measurementsof hum- Smithsonian Institution Press. mingbirds.A. Kodric-Brown,G. S. Byersßand D. F. , & S. B. CHAPLIN. 1975. On the relationship Gori collaborated with J. H. Brown in field work in betweenwing disk loading and foraging strate- the WestIndies, which yielded valuableinsights into gy in hummingbirds.Amer. Natur. 109:217-224. the ecologyof Antillean hummingbirds.Brown's re- ß & R. K. COLWELL.1978. Community orga- searchwas supportedby NSF GrantsDEB 76-99499 nizationamong neotropical nectar-feeding birds. and DEB 80-21535. Amer. Zool. 18: 779-796. , --, J. TERBORGH,& S. B. CHAPLIN. 1979. LITERATURE CITED Elevationand the morphologyßflight energeticsß and foraging ecologyof tropicalhummingbirds. AMERICANORNITHOLOGISTS' UNION. 1983. Check-list Amer. Natur. 113: 481-497. of North American birds, 6th ed. Baltimore, FINDLE¾,J. S. 1973. Phenetic packing as a measure Amer. Ornithol. Union. of faunal diversity.Amer. Natur. 107:580-584. BOND,J. 1971. Birds of the West Indies. Bostonß GRANT,K. A., & V. GRANT. 1968. Hummingbirds Houghton Mifflin Co. and their flowers. New York, Columbia Univ. BOWERSßM. A., & J. H. BROWN.1982. Bodysize and Press. coexistence in desert rodents: chance or com- GREENEWALT,C. H. 1975. The flight of birds.Trans. munitystructure? Ecology 63: 391-400. Amer. Phil. Soc. 65: 1-67. BROWNßJ. H. 1975. Geographicalecology of desert HUTCHINSONßG.E. 1959. Homage to Santa Rosalia, rodents.Pp. 315-341 in Ecologyand evolutionof or why are thereso many kinds of animals.Amer. communities(M. L. Cody and J. M. Diamond, Natur. 93: 145-159. Eds.).Cambridge, Massachusetts, Harvard Univ. KARR,J. g., •: F. C. JAMES.1975. Ecomorphological Press. configurationsand convergent evolution. Pp. --, & g. A. BOWERS.1984. Patterns and pro- 258-291 in Ecologyand evolution of communi- cessesin three guilds of terrestrialvertebrates. ties (M. L. Cody and J. M. Diamond, Eds.).Cam- Pp. 282-296 in Ecologicalcommunities: concep- bridge, Massachusetts,Harvard Univ. Press. tual issuesand the evidence (D. Simberloff and KODRIC-BROWN,A., & J. H. BROWN. 1978. Influence D. R. StrongßEds.). PrincetonßNew Jersey, of economics,interspecific competition and sex- Princeton Univ. Press. ual dimorphism on territoriality of migrant Ru- --, & A. C. GInSON. 1983. Biogeography. St. fous Hummingbirds.Ecology 59: 285-296. Louis, MissourißMosby. --, --, G. S. BYERS,& D. F. GORI. 1984. Or- April 1985] HummingbirdMorphology and Ecology 263

ganization of a tropical island community of SNOW,B. K., & D. W. SNOW. 1972. Feeding niches hummingbirdsand flowers. Ecology65: 1358- of hummingbirdsin a Trinidad valley. J. Anita. 1368. Ecol. 41: 471-485. LACK, D. 1947. Darwin's finches. London, Cam- $OKAL,R. R., & F. J. ROHLF. 1969. Biometry: the bridge Univ. Press. principles and practiceof statisticsin biological ß 1973. The numbersof hummingbird species research. San Francisco, W. H. Freeman and Co. in the West Indies. Evolution 27: 326-337. STILES,F. G. 1973. Food supply and annual cycle of --. 1976. Island biology. Oxford, BlackwellSci- the Anna Hummingbird. Univ. California Publ. entific. ZooL 97: 1-1110. MACARTHUR,g. $. 1972. Geographicalecology. New 1975. Ecology,flowering phenology, and York, Harper and Row. hummingbird pollination of some Costa Rican PETERS,J. L. 1945. Check-list of birds of the world, Heliconiaspecies. Ecology 56: 285-301. vol. 5. Cambridge, Massachusetts,Harvard Univ. STRONG,D. g., JR., L. g. $ZYSKA,& D. $IMBERLOFF. Press. 1979. Tests of community-widecharacter dis- PETERSON,g. t., & E. L. CHALIF. 1973. A field guide placementagainst null hypotheses.Evolution 33: to Mexicanbirds. Boston,Massachusetts, Hough- 897-913. ton Mifflin Co. TER}•ORGH,J. 1973. Chance,habitat, and dispersalin RIDGEWAY,R. 1904. The birds of North and Middle the distribution of birds in the West Indies. Evo- America, part 3. Washington, D.C., U.S. Govern- lution 27: 338-349. ment Printing Office. WASER,N.M. 1978. Competitionfor pollination and --. 1914. The Birds of North and Middle Amer- sequentialflowering in two Coloradowild flow- ica, part 4. Washington, D.C., U.S. Government ers. Ecology59: 934-944. Printing Office. WETMORE,g. 1968. The birds of the Republic of SIEGEL,S. 1956. Nonparametricstatistics for the be- Panama,part 2. Washington, D.C., Smithsonian havioral sciences. New York, McGraw-Hill. Press. $IMBERLOFF,D., & W. BOECKLEN. 1981. Santa Rosalia reconsidered:size ratios and competition. Evo- lution 35: 1206-1228.

APPENDIXßWing length (mm), body weight (g), and APPENDIX. Continuedß culmenlength (ram) of hummingbirdsin the world speciespool. Taxonomyfollows the A.O.U. (1983; Char- North American species)and Peters (1945). Mea- Species acter œ SD n surementswere taken from Ridgeway (1904, 1914), Wetmore (1968), Stiles (1973), Carpenter (1976), Glaucishirsuta (A) Wing 59.22 2.10 76 Feinsinger (1976), Waser (1978), Brown and Ko- Weight 6.82 0.52 96 dric-Brown (1979), and Feinsingeret al. (1979). J. Culmen 32ß25 -- 73 H. Brown, R. K. Colwell, and R. Zusi contributed Threnetes Wing 59.13 0.80 20 unpublisheddata. Temperate (T) and Antillean (A) leucurus Weight 5.59 0.10 42 hummingbird speciesare designated,as are those Culmen 29.50 -- -- usedin the biogeographically(B) and morpholog- T. ruckeri Wing 56.85 -- 45 ically (M) constrainedtests of null hypothesesfor Weight 5.92 0.70 11 the temperatefauna. Culmen 30ß92 -- 45

Char- Phaethornis Wing ------yaruqui Weight 6.00 -- 1 Species acter œ SD n Culmen ------Doryferajohannae Wing 53.60 1.00 9 P. guy Wing 59.85 0.40 40 Weight 4.10 0.10 9 Weight 5.66 0.24 64 Culmen ------Culmen 42.50 -- 20 D. ludoviciae Wing 58.16 0.70 35 P. syrmatophorus Wing ------Weight 5.72 0.20 20 Weight 5.70 -- 2 Culmen 36.13 -- 16 Culmen ------Androdon Wing 65.37 -- 15 P. superciliosus Wing 59.74 1.34 268 aequatorialis Weight 8.00 -- 1 (B) Weight 6.04 0.27 277 Culmen 39.77 -- 15 Culmen 37.65 1.17 255 264 BROWNANt) BOWERS [Auk,Vol. 102

APPENDIX. Continued. APPENDIX. Continued.

Char- Char- Species acter œ SD n Species acter œ SD n

P. hispidus Wing 56.90 0.60 22 C. coruscans Wing 75.20 -- 2 Weight 4.86 0.12 26 Weight 7.55 0.49 17 Culmen ------Culmen 21.50 -- 1 P. anthophilus Wing 57.23 -- 23 C. serrirostris Wing ------Weight 4.60 -- 1 Weight 6.75 -- 45 Culmen 35.30 -- 23 Culmen ------P. ruber Wing ------Anthracothorax Wing 62.31 2.61 31 Weight 2.15 -- 4 viridigula(A) Weight 7.46 0.49 25 Culmen ------Culmen 24.53 1.70 31 P. griseogularis Wing • -- -- A. prevostii(B) Wing 65.45 -- 109 Weight 2.00 -- 1 Weight 6.71 0.22 11 Culmen ------Culmen 26.14 -- 109 P. longuemareus Wing 49.61 0.30 85 A. nigricollis Wing 66.18 -- 89 (M) Weight 2.97 0.38 58 Weight 7.00 0.55 51 Culmen 31.19 -- 79 Culmen 24.18 -- 88 Eutoxeresaquila Wing 71.45 -- 29 A. veraguensis Wing 66.75 -- 8 Weight 10.83 -- 3 Weight ------Culmen 27.09 -- 29 Culmen 24.57 -- 7 E. condamini Wing 68.10 0.80 22 A. dominicus(A) Wing 62.81 1.86 42 Weight 9.36 0.27 24 Weight 5.66 0.41 32 Culmen ------Culmen 24.13 0.75 42 Phaeochroa Wing 69.34 -- 51 A. mango(A) Wing 68.31 -- 21 cuvierii Weight 8.87 0.41 37 Weight 7.81 0.67 4 Culmen 23.17 -- 51 Culmen 26.03 -- 21 Campylopterus Wing ------Eulampisjugulaffs Wing 73.21 -- 45 curvipennis Weight 6.40 -- 7 (A) Weight 8.67 0.56 11 Culmen ------Culmen 23.59 -- 45 C. largipennis Wing 73.20 0.80 12 E. holosericeus Wing 60.00 2.04 92 Weight 8.32 0.21 29 (A) Weight 5.60 0.40 21 Culmen ------Culmen 22.74 2.20 91 Wing 72.56 -- 9 Chrysolampis Wing 54.57 -- 48 Weight ------mosquitus Weight 3.88 0.39 20 Culmen 26.20 -- 9 Culmen 12.39 -- 48 C. hemileucurus Wing 76.87 -- 122 Orthorhyncus Wing 47.34 1.09 54 (B) Weight 10.46 0.50 56 cristatus(A, M) Weight 2.71 0.18 23 Culmen 32.48 -- 123 Culmen 10.72 1.18 54 C. ensipennis Wing ------Klaisguimeti Wing 47.77 -- 64 Weight 9.67 -- 3 Weight 2.54 0.20 7 Culmen ------Culmen 13.24 -- 64 C. falcatus Wing ------Abeillia abeillei Wing ------Weight 7.50 -- 3 Weight 2.67 -- 3 Culmen ------Culmen ------Eupetomena Wing ------Stephanoxislalandi Wing ------macroara Weight 8.74 -- 29 Weight 4.03 -- 3 Culmen ------Culmen ------Florisuga Wing 67.41 -- 88 Lophornisornata Wing ------mellivora Weight 7.41 0.63 28 Weight 2.20 -- 8 Culmen 19.60 -- 88 Culmen ------Melanotrochilus Wing ------L. magnifica Wing ------f•,lSC•S Weight 8.05 -- 150 Weight 2.13 -- 3 Culmen ------Culmen ------Coh•ff delphinae Wing 70.55 -- 50 L. delattrei Wing 37.54 -- 25 Weight 6.66 0.67 9 Weight ------Culmen 17.71 -- 50 Culmen 10.89 -- 25 C. thalassinus(B) Wing 63.79 -- 65 L. helenae Wing ------Weight 5.91 0.34 84 Weight 2.60 -- 4 Culmen 23.20 -- 64 Culmen ------April1985] HummingbirdMorphology andEcology 265

APPENDIX. Continued. APPENDIX. Continued.

Char- Char- Species acter œ SD n Species acter •? SD n

L. adorabilis Wing ------Hylocharis Wing 49.30 -- 1 Weight 2.70 -- 4 xantusii(B, M) Weight 3.53 0.18 Culmen ------Culmen 17.00 -- 1 Popelairia Wing ------H. leucotis(B) Wing 57.00 -- 1 langsdorffi Weight 3.00 -- 3 Weight 3.50 0.27 210 Culmen ------Culmen 17.60 -- 1 Discosura Wing 40.60 -- 20 H. eliciae(M) Wing 48.72 -- 59 conversii(M) Weight 3.00 0.40 5 Weight 3.60 -- 13 Culmen 13.85 -- 20 Culmen 18.24 -- 59 D. longicauda Wing ------H. sapphirina Wing ------Weight 3.00 -- 1 Weight 4.00 -- 2 Culmen ------Culmen ------Chlorestesnotatus Wing ------H. cyanus Wing ------Weight 3.61 -- 14 Weight 3.47 -- 8 Culmen ------Culmen ------Chlorostilbonaure- Wing 43.33 -- 6 Goldmania Wing 50.28 -- 26 oventris(M) Weight 2.97 0.50 3 violiceps Weight 3.63 0.20 6 Culmen 13.40 -- 6 Culmen 18.73 -- 26 C. gibsoni Wing ------Goethalsiabella Wing 52.94 -- 5 Weight 2.72 -- 5 Weight ------Culmen ------Culmen 17.86 -- 5 C. ricordii(A) Wing 51.50 -- 22 Trochilus Wing 53.90 -- Weight 4.23 -- 1 polytmus(A) Weight 4.10 -- 1 Culmen 17.20 -- 22 Culmen 19.80 -- 1 C. swainsonii(A) Wing 54.51 -- 14 Leucochloris Wing ------Weight 4.85 -- 1 albicollis Weight 6.45 -- 26 Culmen 17.30 -- 14 Culmen ------C. maugaeus(A, M) Wing 47.53 -- 45 Polytmus Wing ------Weight 2.93 -- 45 guainumbi Weight 4.70 -- 3 Culmen 13.62 -- 45 Culmen ------Cyanophaia Wing 58.58 -- 26 Leucippus Wing ------bicolor(A) Weight 4.55 -- 8 chionogaster Weight 3.80 -- 7 Culmen 16.41 -- 22 Culmen ------Thalurania Wing 52.61 0.50 52 Talaphorus Wing 69.90 -- 1 furcata(A) Weight 4.35 0.24 72 tuczanowskii Weight 6.00 -- 1 Culmen 20.19 -- 36 Culmen 22.80 -- 1 T. watertonii Wing ------Amazilia Wing 49.00 -- 1 Weight 4.50 -- 2 candida(B, M) Weight 3.74 0.43 14 Culmen ------Culmen 17.60 -- 1 T. glaucopis Wing ------A. chionopectus Wing 54.40 2.20 3 Weight 5.32 -- 60 Weight 4.72 -- 62 Culmen ------Culmen ------T. lerchi Wing ------A. versicolor Wing ------Weight 2.00 -- 3 Weight 4.00 -- 1 Culmen ------Culmen ------Panterpeinsignis Wing 64.47 -- 68 A. fimbriata Wing 51.60 -- 1 Weight 5.69 0.39 47 Weight 4.78 0.48 9 Culmen 20.86 -- 68 Culmen 21.10 -- 1 Damophilajulie Wing 43.46 -- 51 A. lactea Wing 54.20 1.10 10 (M) Weight 3.31 0.15 51 Weight 4.55 0.10 Culmen 15.17 -- 51 Culmen ------Lepidopygacoeru- Wing 48.76 -- 21 A. amabilis Wing 52.45 -- 26 leogularis(M) Weight 4.20 0.30 3 Weight 4.32 0.55 18 Culmen 19.40 -- 20 Culmen 19.20 -- 26 L. goudoti Wing 50.00 -- 1 A. cyaneotiocta Wing 61.10 -- 1 Weight 3.95 0.30 4 Weight 5.84 0.63 13 Culmen 19.10 -- 1 Culmen 19.90 -- 1 266 BROWNAND BOWERS [Auk,Vol. 102

APPENDIX. Continued. APPENDIX. Continued.

Char- Char- Species acter œ SD n Species acter œ SD n

A. franciae Wing 50.60 -- 1 Chalyburabuffonii Wing 66.65 -- 45 Weight 5.04 -- 5 Weight 6.20 0.40 6 Culmen 23.00 -- 1 Culmen 25.39 -- 45 A. leucogaster Wing ------C. melanorrhoa Wing 66.70 -- 20 Weight 4.90 -- 6 Weight ------Culmen ------Culmen 23.30 -- 20 A. cyanocephala Wing 61.10 -- 1 C. urochrysia Wing 66.70 -- 18 (B) Weight 5.84 0.63 13 Weight 7.00 -- 1 Culmen 19.90 -- 1 Culmen 25.54 -- 18 A. beryllina(B) Wing 56.50 -- 1 Aphantochroa Wing ------Weight 4.94 0.40 7 cirrochloris Weight 7.29 -- 30 Culmen 19.40 -- 1 Culmen ------A. saucerrottei Wing 53.90 -- 120 Lampornis Wing ------Weight 4.49 0.37 139 clemenciae Weight 7.96 -- 268 Culmen 23.16 -- 120 Culmen ------A. tobaci(M) Wing 47.00 -- 1 L. amethystinus Wing 67.23 -- 13 Weight 4.10 0.39 35 (B) Weight 6.73 0.87 26 Culmen 18.30 -- 1 Culmen 20.92 -- 13 A. viridigaster Wing ------L. viridipallens Wing ------Weight 6.70 -- 5 Weight 5.42 -- 11 Culmen ------Culmen ------A. edward Wing 52.40 -- 39 L. hemileucus Wing 61.69 -- 18 Weight 4.70 0.20 58 Weight 5.60 -- 1 Culmen 19.43 -- 39 Culmen 20.68 -- 18 A. rutila (B) Wing 58.60 -- 1 L. castaneoventris Wing 62.15 -- 20 Weight 4.71 0.54 11 Weight 5.64 0.27 40 Culmen 23.40 -- 1 Culmen 21.80 -- 20 A. yucatanensis Wing ------Adelomyia Wing 50.73 0.50 26 Weight 3.57 -- 7 melanogenys Weight 3.77 0.31 35 Culmen ------Culmen 16.10 -- 1 A. tzacatl(B) Wing 57.28 -- 26 Clytolaema Wing ------Weight 5.03 0.28 105 rubricauda Weight 7.32 -- 5 Culmen 22.56 -- 36 Culmen ------A. amazilia Wing 52.45 -- 26 Polyplancta Wing 58.80 0.80 8 Weight 4.32 0.55 18 aurescens Weight 6.30 0.10 8 Culmen 19.20 -- 26 Culmen ------A. violiceps(B) Wing 58.40 -- 1 Heliodoxa Wing ------Weight 5.87 0.40 7 rubinoides Weight 7.74 -- 8 Culmen 21.60 -- 1 Culmen ------Eupherusa Wing 62.25 -- 2 H. leadbeateri Wing 66.40 0.70 24 poliocerca Weight ------Weight 7.38 0.18 32 Culmen 18.00 -- 1 Culmen ------E. eximia(B) Wing 57.32 -- 120 H. jacula Wing 70.91 -- 47 Weight 4.27 0.30 126 Weight 8.30 0.40 10 Culmen 21.37 -- 114 Culmen 23.22 -- 47 E. nigriventris(M) Wing 48.05 -- 21 Eugenesfulgens Wing 72.61 -- 67 Weight 3.00 -- 1 (B) Weight 6.80 0.46 38 Culmen 15.43 -- 21 Culmen 30.68 -- 67 Elvirachionura (M) Wing 49.05 -- 45 Sternoclyta Wing ------Weight 2.99 0.21 16 cyanopectus Weight 9.06 -- 17 Culmen 15.75 -- 45 Culmen ------E. cupreiceps(M) Wing 46.58 -- 26 Topazapella Wing ------Weight 3.00 -- 13 Weight 12.12 -- 6 Culmen 15.12 -- 26 Culmen ------Microcheraalbo- Wing 40.77 -- 17 Oreotrochilus Wing 65.00 -- 1 coronata(M) Weight 2.71 0.20 8 melanogaster Weight 4.40 -- 2 Culmen 12.24 -- 16 Culmen 17.70 -- i April1985] HummingbirdMorphology and Ecology 267

APPENDIX. Continued. APPENDIX. Continued.

Char- Char- Species acter œ SD n Species acter g SD n

O. estella Wing 66.40 -- 1 H. viola Wing ------Weight 8.11 0.39 37 Weight 5.33 -- 3 Culmen 18.50 -- 1 Culmen ------Patagonagigas Wing 132.30 -- 1 H. squamigularis Wing ------Weight 20.24 1.29 10 Weight 5.00 -- 1 Culmen 36.30 -- 1 Culmen ------Aglaeactis Wing 80.10 -- 1 Eriocnemis Wing 58.90 -- 1 cupripennis Weight 7.55 0.59 23 vestitus Weight 5.17 -- 3 Culmen 20.00 -- 1 Culmen 19.10 -- 1 Lafresnaya Wing 63.60 -- 1 E. luciani Wing 67.35 0.15 4 lafresnayi Weight 4.50 -- 2 Weight 6.26 0.16 8 Culmen 25.60 -- 1 Culmen 21.40 -- 1 Pterophanes Wing 105.70 8.00 3 E. mosquera Wing ------cyanopterus Weight 11.17 0.50 6 Weight 5.80 -- 1 Culmen 30.50 -- 1 Culmen ------Coeligena Wing 70.76 0.60 27 Haplophaedia Wing 58.75 0.50 41 coeligena Weight 6.75 0.10 34 aureliae Weight 5.46 0.17 37 Culmen 26.00 -- 1 Culmen 20.32 -- 15 C. wilsoni Wing ------Ocreatus Wing 44.14 0.50 8 Weight 7.00 -- 1 underwoodii(M) Weight 3.03 0.14 19 Culmen ------Culmen 13.00 -- 1 C. torquata Wing 73.27 0.80 19 Lesbiavictoriae Wing 59.10 -- 1 Weight 7.13 0.20 19 Weight 5.10 -- 1 Culmen 31.20 -- 1 Culmen 15.00 -- 1 C. helianthea Wing 70.10 -- 1 L. nuna(M) Wing 49.70 -- 1 Weight 8.17 -- 3 Weight 3.75 -- 6 Culmen 29.40 -- 1 Culrnen 10.40 -- 1 C. lutetiae Wing 78.80 -- 1 Sappho Wing 57.50 -- 1 Weight 8.13 -- 3 sparganura Weight 5.88 -- 4 Culmen 33.00 -- 1 Culmen 19.20 -- 1 C. violifer Wing 74.78 1.60 6 Polyonymuscaroli Wing 56.80 -- 1 Weight 7.44 0.30 7 Weight 5.03 -- 4 Culmen 37.50 -- 1 Culmen 20.00 -- 1 C. iris Wing 77.90 -- 1 Metalluraphoebe Wing ------Weight 8.00 -- 2 Weight 5.35 -- 2 Culmen 28.00 -- 1 Culmen ------Ensiferaensifera Wing 77.20 -- 1 M. theresiae Wing 64.10 -- 1 Weight 11.83 0.50 3 Weight 5.00 -- 2 Culmen ------Culmen 11.80 -- 1 Sephanoides Wing ------M. eupogon Wing 59.60 0.80 10 sephanoides Weight 5.00 -- 1 Weight 4.60 0.10 10 Culmen ------Culmen ------Boissonneaua Wing 74.00 -- 1 M. williami Wing 55.60 -- 1 fiavescens Weight 8.05 0.90 8 Weight 4.56 0.20 5 Culmen 16.30 -- 1 Culmen 11.80 -- 1 B. matthewsii Wing ------M. tyrianthina Wing 55.24 0.11 10 Weight 6.65 -- 4 Weight 3.52 0.33 21 Culmen ------Culrnen 10.10 -- 1 B. jardini Wing ------Chalcostigma Wing 66.70 -- 1 Weight 8.20 -- 2 stanleyi Weight 5.84 -- 11 Culmen -- -- • Culmen 9.80 -- 1 Heliangelus Wing 61.20 0.50 26 Oxypogonguerinii Wing 61.40 -- 1 amethysticollis Weight 5.27 0.10 28 Weight 5.00 -- 3 Culmen ------Culmen 7.50 -- 1 H. exortis Wing ------Aglaiocercuskingi Wing 56.85 0.60 4 Weight 4.00 -- 1 Weight 4.76 0.58 17 Culmen ------Culmen 13.70 -- 1 268 BROWNAND BOWERS [Auk,VoL 102

APPENDIX. Continued. APPENDIX. Continued.

Char- Char- Species acter œ SD n Species acter œ SD n

A. emmae Wing 66.90 -- 1 Tilmaturadupontii Wing 34.33 -- 19 Weight 5.70 -- 1 (B) Weight 2.23 0.15 12 Culmen 14.40 -- 1 Culmen 13.11 -- 19 Oreonympha Wing ------Calothoraxlucifer Wing 39.28 -- 19 nobilis Weight 9.00 -- 1 (B) Weight 2.80 -- 3 Culmen ------Culmen 21.05 -- 19 Augastesscutatus Wing ------C. pulcher(B) Wing 37.08 -- 12 Weight 3.00 -- 1 Weight 2.93 0.30 3 Culmen ------Culmen 18.07 -- 12 A. lumachellus Wing ------Archilochuscolu- Wing 41.05 -- 26 Weight 4.00 -- 2 bris(T,B, M) Weight 3.19 0.30 35 Culmen ------Culmen 17.96 -- 27 Schistesgeoffroyi Wing 51.40 1.00 8 A. alexandri Wing 44.21 1.19 119 Weight 3.62 0.22 13 (T, B, M) Weight 3.20 0.19 76 Culmen ------Culmen 19.73 0.55 137 Heliothryxbarroti Wing 66.15 -- 76 Calliphlox Wing ------Weight 5.47 0.20 9 amethystina Weight 2.43 -- 24 Culmen 17.98 -- 76 Culmen ------H. aurita Wing 68.60 0.30 4 Mellisugaminima Wing 37.10 -- 17 Weight 5.43 0.30 7 (A) Weight 2.43 0.10 6 Culmen ------Culmen 10.45 -- 17 Heliactincornuta Wing ------M. helenae(A) Wing 31.61 -- 13 Weight 7.80 -- 4 Weight ------Culmen ------Culmen 10.76 -- 13 Loddigesia Wing 38.50 -- 1 Calypteanna Wing 49.56 0.97 71 mirabilis Weight 3.00 -- 1 (T, B, M) Weight 4.25 0.33 79 Culmen 14.70 -- 1 Culmen 19.06 0.95 71 Heliomaster Wing 68.00 -- 1 C. costae Wing 44.45 0.82 76 constantii(B) Weight 7.30 -- 2 (T, B, M) Weight 3.08 0.29 58 Culmen 41.00 -- 1 Culmen 17.71 0.56 107 H. longirostris(B) Wing 59.24 -- 16 Stellulacalliope Wing 40.31 0.80 71 Weight 6.55 0.68 8 (T,B,M) Weight 2.66 0.32 51 Culmen 35.18 -- 16 Culmen 15.54 0.70 71 H. squamosus Wing ------Atthis heloisa(B) Wing 35.12 -- 25 Weight 5.00 -- 1 Weight ------Culmen ------Culmen 11.77 -- 25 H. furcifer Wing ------Myrtis fanny Wing 40.80 -- 1 Weight 5.00 -- 1 Weight 2.30 -- 1 Culmen ------Culmen 19.20 -- 1 Rhodopisvesper Wing 49.30 -- 1 Acestrura Wing 39.55 -- 2 (M) Weight 3.88 -- 4 mulsanti Weight 3.75 -- 6 Culmen 17.40 -- 1 Culmen 16.20 -- 1 Thaumasturacora Wing 36.10 -- 1 A. heliodor Wing 32.40 -- 20 Weight 2.00 -- 1 Weight ------Culmen 12.80 -- 1 Culmen 15.00 -- 20 Calliphlox Wing 36.10 -- -- Selasphorus Wing 49.68 -- 58 evelynae Weight 2.40 -- 2 platycercus Weight 3.43 0.31 86 Culmen ------(T,B,M) Culmen 17.51 0.45 145 C. bryantae(M) Wing 41.50 -- 8 S. rufus(T, B, M) Wing 42.30 0.84 326 Weight 3.29 -- 21 Weight 3.36 0.31 68 Culmen 19.30 -- 8 Culmen 16.57 0.67 326 Philodicemitchellii Wing 37.80 -- 1 S. sasin(T, B, M) Wing 39.77 0.83 111 Weight 3.13 0.10 3 Weight 3.16 0.21 38 Culmen 15.00 -- 1 Culmen 17.20 0.60 111 Dorichaenicura Wing 35.26 -- 29 S.fiammula (M) Wing 41.57 -- 40 Weight 2.40 -- 1 Weight 2.47 -- 3 Culmen 20.41 -- 29 Culmen 12.90 -- 40 April1985] HummingbirdMorphology and Ecology 269

APPENDIX. Continued. APPENDIX. Continued.

Char- Char- Species acter œ SD n Species acter œ SD n

S. torridus Wing 40.50 -- 8 S. ardens Wing 40.17 -- 19 Weight ! .90 -- ! Weight ------Culmen !!.89 -- 8 Culmen !2.54 -- !9 S. simoni Wing 38.68 -- ! ! S. scintilla Wing 34.60 -- 43 Weight ------Weight 2.!5 0.!6 39 Culmen !0.59 -- !! Culmen !!.93 -- 43

[From "Sexual selectionand the nesting of birds," by J. A. Allen (!885 Auk 2: 129-!39):

"Mr. Wallace, and after him Mr. Dixon and others, imitation,memory, and hereditary habit, 'play the mi- in discusing[sic] the question How do young birds nor parts.' 'To credit birds,' he says,'with such mar- learn to build their first nest? claim that 'instinct' has vellous power as blind and infallible instinct in nothing to do with the matter,--that they learn by building their nestswould be to place them far be- observation and are guided by memory! Says Mr. yond man himself in intelligence,and allot to them Wallace: 'It has, however, been objectedthat obser- a faculty which is superhuman.... A bird's intellec- vation, imitation, or memory, can have nothing to do tual powers advance towards maturity much more with a bird's architecturalpowers, because the young quickly than in the human species.A young bird birds which in England are born in May or June,will three or four daysold is capableof considerablepow- proceedin the following April or May to build a ers of memory and observation,and during the time nests[sic] as perfect and as beautiful as that in which that elapsesin which it is in the nest it has ample it was hatched, although it could never have seen opportunityof gaining an insight into the architec- one built. But surely the young birds beforethey left ture peculiar to its species.It seesthe position of the the nesthad ampleopportunities of observingits form, nest, it notes the materials, and when it requires one its size,its position,the materialsof which it was con- for itself,is it sovery extraordinarythat, profitingby structed, and the manner in which those materials such experience, it builds one on the same plan? were arranged. Memory would retain these obser- Again, birds often return to the place of their birth vationstill the following spring,when the materials the following season,and possiblysee the old home would comein their way during daily searchfor food, many times ere they want one for themselves.This, and it seems highly probable that the older birds aided by the strong hereditary impulse to build a would begin building first, and that those born the nest similar to the one in which they were born, preceding summer would follow their example, inherited from their parents,aids them in their task.' learningfrom them how the foundationsof the nest This reasoning, I am free to confess,strikes me, to were laid and the materials put together. Again we say the least,as extraordinary!A degreeof mental have no right to assumethat young birds generally power, at leastof memoryand of imitation, is ascribed pair together,'etc. Mr. Dixon restatesthe casein much to young birds which is not only 'superhuman,'but the sameway. Alluding to 'blind instinct' as a factor of which there is neitherproof, nor even possibility in the case, he says: 'To credit the bird with such of proof. Mr. Dixon has the 'three or four daysold' instinct, which because it seems so self-evident is nestlingtaking note of and memorizing its surround- taken to be matter of fact, is to admit that it possesses ings before, in the caseof the higher Oscines,it has intellectualpowers infinitely superiorto thoseof man; thepower to evenopen its eyes!Yet with all this ascribed whilst the evidencethat can be gatheredon the sub- precosityand keennessof observation,and this won- ject all goesto show that its intellectual powers are derful powerof memoryand imitationin youngbirds, of preciselythe samekind asman's, but someof them, Mr. Dixon finds it neccessary[sic] to call in the aid of course,are infinitely inferior in degree, whilst of 'a strong hereditary impulse to build a nest similar othersare unquestionablysuperior.' He assumesthat to the one in which they were born,' which is more

(continuedon p. 322)