Systematics and Evolution of the North American Merlins

SYSTEMATICS AND EVOLUTION OF THE NORTH AMERICAN MERLINS

STANLEY A. TEMPLE

Tar. Merlin or PigeonHawk (Falco columbarius)is a polytypic species of panborealdistribution. In North America it breedsin the taiga, Pacific coastalforest, and prairie-parklandbiomes and migratessouthward. Four North American subspeciesare recognizedby Peters (1931) and the A.O.U. Check-list (1957). F. c. columbariusLinnaeus and F. c. bendirei Swannoccur in the easternand westernportions of the taiga, respectively; F. c. suckleyi Ridgway occursin the humid Pacific coastal forest; and F. c. richardsoniiRidgway occursin the prairie-parklands. Some controversyexists over the validity of these forms, particularly F. c. bendirei and F. c. suckleyi (Bent, 1938; Hellmayr and Conover, 1949). Swarth (1935) questionedthe validity of F. c. bendirei on the grounds that it is not phenotypically distinguishablefrom F. c. columbarius. He also questionedF. c. suckleyi becauseits geographicdistri- bution was poorly defined and seemedto overlap that of F. c. bendirei. More recently several authors have tried to circumvent the problems of F. c. bendirei and suckleyi by synonymizingthe former with the latter (Dementiev and Gladkov, 1951) or with F. c. columbarius (Brown and Amadon, 1969; Stresemann,MS). The authors offer no justification for thesesynonomies. Despite these disagreements,little systematicwork has been done on the North American Merlins since Peters' (1927) review. In fact, the Merlin remains one of the least studied of the North American falcons in mostaspects of its biology. The objectivesof this study are to analyze the phenotypicvariation in the North American Merlins, to interpret the evolutionaryorigins of the observedvariations and geographicdistribu- tions, and from this information to reevaluate the present subspecific designations. M,•TERX•LS ^NI> MET•OI>S

The comparative aspectsof this study are based upon examination of 1,585 museum study skins from five populations of Merlins as delineated in Figure 1. The taiga, coastal forest, and prairie-parkland populations are separated on the basis of Pitelka's (1941) biome classification although his terminology is not used. The divisions of the taiga are made at the Continental Divide, a natural geographic division, and at the western edge of the Hudson Bay, the presently acceptedbound- ary between F. c. columbarius and bendireL Unless otherwise noted, only specimenstaken between 15 May and 15 September, and within the recognizedbreeding range of the Merlin, were used for comparisons. This was done on the assumption that these birds were taken on or near their breeding grounds (see Reilly, 1954, for a discussionof this criterion).

325 The Auk, 89: 325-338. April 1972 326 ST^•r• A. Tr•tr•.E [Auk, Vol. 89

Figure 1. Breeding ranges of study populations and subspeciesof North American Merlins. 1, eastern taiga (F. c. columbarius); 2, central taiga (F. c. bendirei) ; 3, western taiga (F. c. bendirei); 4, coastal forest (F. c. suckleyi); 5, prairie-parkland (F. c. richardsonii).

Characters which best demonstrated patterns of variation were used for making comparisons. The following techniqueswere used in analyzing specimens: Size characters.--Standard linear measurements were taken in accordance with the methods described by Baldwin et al. (1931): (1) wing length: the chord of the unflattened wing; (2) tail length; (3) tarsus length; (4) toe length: length of middle toe less talon; (5) culmen length from cere; (6) wing loading index: wing loading was approximated by dividing the cube root of the body weight (g) by the chord of the wing (cm). This wing loading index showed a high correlation with the actual wing loading value computed as body weight in grams per total wing area in cm2 (for 8 specimens,r ---- 0.877). All measurementsare in millimeters. Plumage characters.--Specimenswere scored on the following plumage characters: (1) crown streaking class: the relative width of the longitudinal black streaks on the crown (see Table 4 for descriptionsof classes); (2) number of tail bands: the number of light bands in the tail bordered both proximally and distally by dark bands; (3) barring on the anterior web of the outermost primary: whether present or absent; (4) barring on the posterior web of the outermost primary: whether the next to most distal bar extended from the feather shaft to the posterior feather margin or not. Colorimetric characters.--Certain colorimetric data were taken from randomly selected study skins collected between 31 August and 15 October. This temporal restriction was placed on specimensto reduce the adverseeffects of fading and wear on the recently molted plumage. Colorimetric readings were taken from the center of the interscapular region of the dorsum. All readings were made on a Bausch and Lomb Spectronic 20 equipped with a Color Analyzer Attachment modified after the suggestionsof Selanderet al. (1964). For each specimena curve of the spectral re- April 1972] North American Merlins 327 flectance over the wavelength range of 400 to 700 m/• was constructed. From this curve the trichromatic coefficients(xs, ys, zs) were determined using the 10-selected ordinate system (Hardy, 1936). The dominant wavelength, brightness,and excitation purity of the plumage color were computed from these coefficients. Four randomly selectedstudy skins were washed using the techniquesdescribed by Selander and Johnston (1967). None of the changes in colorimetric characters attributable to washingwere significant; therefore, washing of outwardly clean skins was not judged necessary. All calculationsand statistical analyseswere done with the computing facilities of Cornell University. Statistical methods follow Steele and Torrie (1960). In the text and tables a significant differenceis taken to mean the 5 percent level of probability (p • 0.05) by the indicatedstatistical test.

RESULTS

Sv.x ^•D Aov. V^•x^Txo• As Merlins showboth sexualand age dimorphism,these types of variation are consideredbefore geographic variation is discussed. Variation in size characters.--As typical in the order Falconiformes, sexualdimorphism in Merlins is reversedfrom that in many other birds; femalesare larger than males in most measurements.The degreeof sexualdimorphism in size charactersis similar in all populations. Age variation in size occursin Merlins as in many other birds. Fre- quently juvenile flight feathersare longer than thoseof the subsequent adult plumage,thus affordingyoung birds lighter wing loadingwhile they developtheir powersof flight. Age variation in wing and tail lengthsof Merlins is consistentwith this pattern. No significantage variation exists for tarsuslength, toe length,or culmenlength. Variationin plumagecharacters.--For a falcon,sexual dichromatism is markedin the Merlin, and this differenceis mostpronounced in the adult plumages.Friedmann (1950) describesthe plumagesof the North Amer- ican Merlins in detail. In brief, the adult male'sdorsum is blue-grayand the adult female's is brownish. The dorsum colorationof juvenile males and femalesis similar to that of the adult female; the adult female rump is moregrayish. Table I presentscolorimetric data on the dorsumcolora- tion of adult Merlins. The sexual dichromatism in adult plumages is clearly shownby the differencesin their dominantwavelengths. No significantsexual or age variation occurs for any of the other plumagecharacters examined, and specimenswere pooledas to sex and age when makinginterpopulation comparisons of thesecharacters.

Gv. oov.am•xc V^•i^xxo• Variation in size characters.--Analysisof variance for size characters showno significantgeographic variation for tarsuslength, toe length,or culmenlength, but wing and tail lengthsdo showsignificant patterns of 328 STANLEYA. TEgrVLE [Auk, Vol. 89

TABLE 1 INTERPOPULATIOIq VARIATION IN DORSUM COLOR Or NORTYI AMERICAN MERLINS x

Dominant No. of Brightness wavelength Purity Population specimens (%) (m/•) (%)

Adult males Eastern taiga 5 17.0 ñ 0.7 478.6 ñ 2.9 4.0 Central taiga 4 17.9 + 0.4 475.3 ñ 4.1 3.5 Westerntaiga 6 17.4 -+ 0.6 477.1 ñ 5.3 4.1 Coastal forest 4 15.7 ñ 0.3 476.0 ñ 1.6 4.0 Prairie-parkland 7 21.0 ñ 0.6 472.4 ñ 4.9 5.1

Adult females Eastern taiga 7 24.1 -+ 0.4 581.3 ñ 4.3 1.5 Central taiga 6 24.9 ñ 0.7 582.5 ñ 3.6 1.2 Western taiga 6 25.8 ñ 0.6 584.4 ñ 2.9 1.5 Coastal forest 5 23.4 ñ 0.3 575.7 ñ 2.1 1.7 Prairie-parkland 6 29.8 -+ 0.4 590.6 ñ 3.2 5.0

Values are the mean ñ SE. geographic variation. Linear product-moment correlation coefficients of the means of each population indicate that variations in wing and tail lengthsare also highly concordant.Wing and tail length data are given in Figure 2. Data for juveniles are not shownas interpopulationrelation- shipsin juvenilesare the sameas in adults. Geographicvariation was found in wing loading values (see Table 2). The prairie-parklandpopulation is significantlydifferent from all other populations,having the lightest wing loading. The coastalforest population is also significantlydifferent from all other populations,having the heaviest wing loading. None of the taiga populationsis significantly different, and all are intermediate between the other two populations. Variation in plumage characters.--Merlins show considerablegeo- graphic variation in plumagecharacters, and, consequently,plumage dif- ferences are the main criteria used to distinguish the North American subspecies. The plumagecoloration of North Americanpopulations generally varies accordingto Gloger's ecogeographicalrule. For colorimetric data see Table 1. Little interpopulation variation occurs in the dominant wavelength or excitationpurity of the plumagecolor for any sex or age class, suggestingthat the pigmentationresponsible for plumage color in Merlins tends to be conservative. Interpopulation variation in color brightness is significant;Merlins from the prairie-parklandpopulation are the lightest in color, Merlins from the coastalforest the darkest,and Merlins from April 1972] North A,nerlcanMerlins 329

• 200 .•E195

• 19o z 210•_ • 185 -- 205•- _z 180 • 175 pp CF WT CT ET pp CF WT CT ET

-- 125 E -- J:120 I

,,Z,I10 120 __

• •o5_ 115 --

ioo ADULTMALES ] I10 ADULTFEMALES J

Figure 2. Interpopulationvariation in wing and tail lengthsof North American Merlins. Horizontal lines indicate means; vertical lines show observedrange; open rectanglesmark q- 1 SD andsolid black mark q- 95 percentconfidence intervals of means;adjoining numerals are samplesizes. Population abbreviations: PP, prairie- parkland;CF, coastalforest; WT, westerntaiga; CT, centraltaiga; ET, easterntarga. the taigaintermediate. Both the prairie-parklandand coastal forest popu- lationsare significantlydifferent from all otherpopulations, but within the taigano significantvariation occurs (significance determined by t- tests). A considerableamount of geographicvariation is presentin the number of tail bands(Table 3, significancedetermined by X• tests). The prairie- parklandpopulation is significantlydifferent from all otherpopulations in that these Merlins have five tail bands. The coastal forest population is alsosignificantly different from all otherpopulations, and this differ-

TABLE 2 IlffTERPOPULATIOIffVARIATIOIff Ilff WIIffG LOADIIffG O1e NORTH AI•ERICAIff MERLIIffS•

Wing loadingindexes for indicatedpopulation '• PP ET CT WT CF

Adult males 0.19 0.23 0.23 0.25 0.28 Adult females 0.20 0.23 0.24 0.25 o.30

x Each value is the mean index (cube root of body weight in g •- chord of wing in cm) for 10 specimens;any two meansnot subtendedby the sameline are significantlydifferent by Duncan's multiple range test. 2 Populationsabbreviated as in Figure 2. 330 STANLEYA. TE•XPLE [Auk, Vol. 89

TABLE 3 INTERPOPULATION VARIATION IN TllE NU1ViBER OF TAIL BANDS OF NORTH AMERICAN MERLINS

Percent with indicated number of tail bands No. of Population specimensx 3 4 5 6

Eastern taiga 137 3 92 5 0 Central taiga 99 2 91 7 0 Western taiga 106 3 85 12 0 Coastal forest 104 20 80 0 0 Prairie-parkland 116 0 10 88 2 Specimenspooled with respectto sex and age. enceis attributableto the higher frequencyof birds with only three tail bands. Although no significantdifferences were found within the taiga populations,the number of tail bands decreasesclinally from west to east. The pattern of geographicvariation in crown streakingclosely parallels that of the number of tail bands (Table 4), and significantinterpopula- tion differencesexist (significancedetermined by X2 tests). The prairie- parkland populationis significantlydifferent from all other populations, having very thin black streakson the crownand a resultingpredominance of backgroundcolor. The coastal forest populationis also significantly different from all other populations,having broad black streaks that, in some individuals, all but obscure the backgroundcoloration. While no significant differencesexist within the taiga populations,the relative width of the blackstreaking increases clinally eastward. The patterns of primary barring show significantgeographic variation

TABLE 4 INTERPOPULATION VARIATION IN CROWN STREAKING OF NORTIt AMERICAN MERLINS

Percent in indicated crown streaking classz No. of Population specimensx 1 2 3 4 5

Eastern taiga 137 0 15 46 33 6 Central taiga 99 0 7 43 34 16 Western taiga 106 0 2 38 31 29 Coastal forest 103 50 41 9 0 0 Prairie-parkland 116 0 0 0 11 89

• Specimenspooled with respect to sex and age. 2 Crown streaking classes: 1, black much greater than background; 2, black greater than background; 3, black equals background; 4, black less than background; 5, black much less than background. April 1972] North American Merlins 331

TABLE 5 INTERPOPULATION VARIATION IN PRI/VIARY BARRING PATTERNS OF NORTIt AlV•ERICAN MERLINS

Percent with Percent with No. of anterior web posterior web Population specimensx barred incompletelybarred

Eastern taiga 137 13 2 Central taiga 99 14 1 Western taiga 106 41 5 Coastal forest 104 0 94 Prairie-parkland 116 96 0

Specimens pooled with respect to sex and age. and are almost populationspecific in two instances(see Table 5). The prairie-parklandpopulation is essentiallycharacterized by the presenceof barring on the anterior web of the primariesand is significantlydifferent from all other populations.No individualsin the costalforest population showthis character,and it occursin the taiga populationat a very low frequency. The frequencyof anterior primary barring in taiga Merlins does increase westward. Incompletebarring on the posteriorweb of the primary is characteristic of the coastalforest Merlins. Incompletebarring occursat a high fre- quencyin this populationand is absentor presentat a very low fre- quencyin all otherpopulations (see Table 5). Variation in migration patterns.--Salomonsen(1955) suggeststhat the migrationpatterns and winter rangesof birds may be of evolutionary significance.North American Merlin populationshave rather complex migratory habits that may reflect their evolutionaryhistories. The prairie-parklandpopulation undergoes a relativelyshort migration into the southernGreat Plains and is residentin parts of its breeding range. The coastalforest population is semiresidentin the climatically stablecoastal region, and only a smallportion of the populationmigrates southof the breedingrange. The taiga populations,on the other hand, are highlymigratory. The westernand centraltaiga Merlins migratepast the coastalforest and prairie-parklandpopulations in leapfrogfashion to winter in Central and SouthAmerica. The easterntaiga Merlins are also highly migratorywith many individualscrossing the Gulf of Mexico to winter in the CaribbeanArchipelago and SouthAmerica.

DISCUSSION Geographic variation.---Interpopulation variation in size of Merlins cannotbe readily explainedby traditional ecogeographicalrules. Behle (1942) and Pitelka (1951) noted that larks and jays that inhabit open 332 STANLEYA. TEMPLE [Auk, Vol. 89 habitats such as prairies or desertstend to have longer wings and tails than forms inhabiting denser vegetation. Presumably the longer wing and tail are adaptive becausethese birds are required routinely to fly greater distancesthan are birds in a more closedhabitat. The prairie- parkland Merlins certainly conformto this generality. Correlatedwith havinga longerwing and tail, theseMerlins alsohave light wing loading that wouldbe a definite selectiveadvantage for frequentlong flights. The coastalforest Merlins, in contrast,inhabit the densesthabitat of all five populations.Perhaps reflecting this difference,these Merlins have the heaviestwing loading. In view of the interpopulationdifferences in temperaturesencountered both in the summerand winter ranges,it is surprisingthat no significant differencesexist in tarsusor toe lengthsas predictedby Allen's rule. The brightnessof Merlin plumagecolor varies as predictedby Gloger's rule. The climatic conditionswithin the range of the prairie-parkland populationare the most arid of all five populations,and Merlins from this populationshow the lightest coloration. The climatic conditionsin the range of the coastal forest Merlins are the most humid, and these Merlins showthe darkestplumage coloration. The variation in certain plumage patterns also follows Gloger's rule. The crown streakingclasses are essentiallymeasures of the intensity of dark pigmentationon the crown. On the basisof crownstreaking, the lightest coloredMerlins are from the prairie-parklandpopulation while the darkest are from the coastal forests. The barring on the posteriorweb of the primariesis probablyanother measureof the intensity of dark pigmentation. The high frequencyof incompletebarring in the coastal forest populationindicates that the dark pigmentationin the primary is so intense that the light colored bars are actually reducedto spots. The interpopulationvariation in barring on the anterior web of the primariescannot be explainedadequately on the basisof ecogeographical correlations. The presenceof this character is probably an ancestral characteristic(sensu Mayr, 1969) and is treated in the discussionof Merlin evolution. Interpopulationdifferences in the numberof tail bandsappear to be a combinationof ancestralcharacteristics and strong environmentalcorrela- tions. The large number of tail bands in the prairie-parklandpopulation is presumedto be an ancestralcondition to be discussedunder evolutionary history. In the coastalforest population the reductionin the number of tail bands is largely attributable to color saturationin the same way as primarybarring. The dark pigmentationis so intensethat it obliterates the narrowerproximal bands in the tail. The intermediatecondition of April 1972] North American Merlins 333 the taiga populationsis best explainedby the proposedevolutionary history of thesepopulations. Historical interpretation of Merlin evolution.--Mayr (1946) and Ud- vardy (1958) discussthe origins of the North American avifauna and agree that almost all North American birds had either Palearctic or Neotropicalorigins. The strong affinities the North American Merlins seem to have with those of the Palearctic region indicate a Palearctic origin for the Merlin. Darlington (1957) suggeststhat dispersionin the genusFalco was rapid and complexand believesthat some groupsof the genus reached the Nearctic fairly recently. While the fossil record of Merlins is poor,North Americanfossils indicate that Merlins had reached the Nearctic at least by the late Pleistocene(Brodkorb, 1964). The habitat affinities of Palearctic Merlins are essentially identical to thosein the Nearctic region. As in North America, the largestportion of the PalearcticMerlins' rangeis taiga, but opensteppe and densecoastal forests are also included. Examination of 91 specimensrepresenting all the Palearctic forms showed that several plumage characteristicswere commonto all. These specimensshowed a large number of tail bands (rangingfrom 5 to 8), barring on the anterior websof the primaries,and relatively light coloration. If the North American merlins are compared phenotypicallywith these Palearctic forms, an unexpectedrelationship is suggested. Assuming that selection pressuresare similar in both the Palearcticand Nearctic taiga, it is surprisingthat Palearctic taiga Merlins are so different from the North America taiga Merlins yet so similar to the prairie-parklandMerlins. This evidencesuggests that there was gene flow between the Palearctic Merlins and the prairie-parkland population more recently than with the other North American populations. One way to account for this is to supposea double invasion of Merlins from the Palearctic, with the first stock becomingthe taiga and coastal forest populations and the more recentstock becoming the prairie-parklandpopulation. The glacial events of the late Pleistocene could have allowed such a situation to occur. A hypothetical seriesof steps in the evolution of North American Merlins basedupon glacialevents are shownin Figure 3. During periodsof glacial advance,a Bering land bridge connectedAsia with an unglaciatedrefugium in Alaska (Colinvaux, 1964). Asian Merlins could easily have reachedNorth America via this route (see Figure 3A). With the ameliorationof the climate during subsequentperiods of glacial retreats, these Merlins had the opportunity to spread acrossthe extensive taiga zoneof North America (see Figure 3B). During the latter periodsof glacial advance,these Merlins would have been forced into coniferousrefugia south of the ice along with the other 334 STANLEYA. TEMPLE [Auk, Vol. 89

A. ILLINOIAN B. SANGAMON

GLACIER

•MI Mit

C. WISCONSIN D, TWO CREEKS

E. LATE WISCONSIN F. POST PLEISTOCENE

=TUNDRA-TAIGA •=TAIGA AND CO,NIFEROUS REFUGIA :PRAIRIE

MI=FIRST ASIAN MERLINS M2=SECOND ASIAN MERLINS

Figure 3. A model sequenceof events in race formation of the North American Merlins during the indicatedperiod of the Pleistocene.Boundaries of continents, glaciers,and biomesare only approximate. April 1972] North AmericanMerlins 335 boreal fauna (Hult•n, 1937). At the sametime, a Bering land bridge again connectedAsia and the Alaskanrefugium, and a secondAsian Merlin stockwas able to reachNorth America (see Figure 3C). During the Wisconsinglaciation, both northeasternAsia and the Alaskan re- fugiumwere sparselyvegetated (Colinvaux, 1964), and presumablyany Asian Merlins that successfullyreached North America via the Bering land bridgewere adapted for this openhabitat. Severaltimes during the late stagesof the Wisconsinglaciation size- able ice-free corridors opened between the Cordilleran and Laurentian ice caps, thereby connectingthe Alaskan refugium with unglaciated regionssouth of the ice (Flint, 1957). The Merlinsin the Alaskanrefu- gium may have movedsouthward along these corridors as is suspected to haveoccurred with otheranimals (Haynes, 1966) (seeFigures 3D and E). Beingadapted for an openhabitat, theseMerlins wouldhave found an unexploitedniche available to them in the central prairie-parklands as no other small bird-eatinghawk had successfullyinvaded this habitat in North America. With the final retreat of the Wisconsinglaciers, the taiga expanded northward to dominatethe boreal regionsof North America and with it moved the taiga and coastal forest Merlin populations. The prairie- parklandsbecame more extensivein xeric post-Pleistocenetimes (Wells, 1970), and the Merlins in this populationexpanded their range accord- ingly (seeFigure 3F). While any suchmodel basedupon indirect evidenceis subjectto inde- terminableerror, the proposedmodel appears tenable for severalreasons. It explainsthe presenceof ancestralplumage patterns in the prairie- parkland population by virtue of its more recent connectionwith the Palearcticgene pool. It explainsthe colorsaturation seen in other North Americanpopulations as theseMerlins were subjectedto long periodsof mesicconditions in glacial refugia. The extremecolor saturation of the coastalforest population is probably the result of a long in situ evolution in the humid coastal forests. The high incidenceof ancestralcharacters in the westerntaiga populationcan alsobe explained,for after the glacial retreat there would have been gene flow betweenthe northward moving taiga Merlins and the Merlins remainingin the Alaskanrefugium. Taxonomicconsiderations.--From the foregoingdiscussion, several taxonomicconclusions can be made. Certainly phenotypicallyand probably phylogenetically,the prairie-parklandpopulation is distinctlysep- arate from the other North American populations. The name F. c. richardsoniiRidgway has been and shouldcontinue to be usedfor this population. The coastal forest populationis also rather distinct phenotypically, 336 ST^NLE¾A. Tr•trnr [Auk, Vol. 89 and the proposedevolution of this populationin the Pacificcoast region also suggestsa distinct phylogenetichistory. Phenotypically,Merlins breedingin the coastalforest biome are markedly different from those of the adjoiningtaiga, and an exclusiverange for this phenotypeis indicated. I would concludethat the coastal forest population is taxo- nomicallydistinct, and the name F. c. suckleyiRidgway should be applied to Merlins from this population. Taxonomicdecisions become more difficult to make within the taiga populations. Phenotypically no basis exists for separating the taiga populationsgeographically in the manner the present subspecificdesig- nationsindicate. The clinal nature of geographicvariation in thesepop- ulationsmakes it clear that any attemptsat dividing the taiga populations will be arbitrary and not reflect accurately either phenotypic or phylo- genetic differencesas discussedin this paper. I therefore suggestthat the present nomenclatureapplied to the taiga populationsbe revised as follows: F. c. bendirei Swann should be considereda synonym of F. c. columbariusLinnaeus, and the name F. c. columbariusLinnaeus should be applied to all North Americantaiga populationsof the Merlin as they have beendelineated in this study.

ACXbrOWI,EI)GMEIqXS

Parts of this study were supported by the New York Zoological Society and the Frank M. Chapman Memorial Fund of The American Museum of Natural History. Specimenswere made available for study through the courtesy of the following: D. Areadon, American Museum of Natural History; P. Guiguet, British Columbia Pro- vincial Museum; K. Parkes, Carnegie Museum; W. Godfrey, National Museum of Canada; M. Traylor, Field Museum of Natural History; R. Paynter, Museum of Comparative Zoology, Harvard University; J. Barlow, Royal Ontario Museum; J. Jehl, San Diego Museum of Natural History; N. Johnson, Museum of Vertebrate Zoology, University of California, Berkeley; R. Storer, University of Michigan; D. Warner, University of Minnesota; R. Mengel, University of Kansas; W. Behle, University of Utah; R. Zusi, United States National Museum. I am indebted to Clayton M. White and Dean Amadon for suggestionsand discussionduring this study, to Tom J. Cade for reviewing this manuscript,and to my wife Barbara for encouragementand assistancein preparing this manuscript.

SUMM^•¾ The North Americansubspecies of the Merlin have been subjectedto little systematicwork, and the validity of two formshas beenquestioned. Sex, age, and geographicvariations as well as the presumedphylogeny of eachform werestudied to evaluatethe validity of the subspecies. Sex and age variationin size charactersare similarin all populations. Sexualdichromatism is markedin adultsand slight in immatures.The immaturesof both sexesresemble the adult female. Little sex or age variation existsin plumagepatterns. April 1972] North American Merlins 337

Geographicvariation existsfor wing length and tail length, and these two charactersare highly concordant. No geographicvariation occurs in the other size charactersexamined. Size charactersdo not vary as predictedby classicalecogeographical rules. Wing and tail length seem to be best correlatedwith the density of the vegetationin the Merlin's range; Merlins from open country have the longestwings and tail. Plu- mage color varies accordingto Gloger'srule. Variations in plumage patterns are correlated with environmental conditions and also with the phylogenyof the population. Interpopulationdifferences in migratory patterns and winter rangesare complicated. A conjectural evolutionary history of the North American Merlins is constructed,and a double invasion of Merlins from Asia during the Pleistoceneis proposed.The first Asian Merlin stock becamethe taiga and coastal Merlins, and the second Asian Merlin stock became the prairie-parklandMerlins. Taxonomic suggestionsmade on the basis of the above evidenceare as follows: F. c. richardsoniiand F. c. suckleyi should remain as presently described.F. c. bendireishould be considereda synonymof F. c. columbarius,which shouldbe used for all of the taiga-inhabitingNorth American Merlins.

LITERATURE CITED AMERICANORNITttOLOGISTS' UI•rlOI•. 1957. Check-listof North Americanbirds, fifth ed. Baltimore, Amer. Ornithol. Union. BALDWIN,S. P., a. C. OBERIIOLSER,AND L. G. WORLEY. 1931. Measurementsof birds. Sci. Publ. ClevelandMus. Nat. Hist., No. 2. BE}rLE, W. H. 1942. Distribution and variation of the Horned Larks (Otocoris alpestris)of westernNorth America. Univ. California Publ. Zool., 46: 205-316. BEAT, A. C. 1938. Life historiesof North American birds of prey, part 2. U.S. Natl. Mus., Bull. No. 170. BRODXORB,P. 1864. Catalogue of fossil birds, part 2. Bull. Florida State Mus., 8: 195-335. B•ow•r, L., ^•rD D. A•rADO•r. 1969. Hawks, eaglesand falconsof the world. New York, McGraw-Hill. COLINVAUX,P. A. 1964. The environmentof the Bering land bridge. Ecol. Monogr., 34: 297-329. DARLINCTO•V,P. J. 1957. Zoogeography:the geographicaldistribution of animals. New York, Wiley. DE•VrE•rTIEV,G. P., ANDN. A. GL^DXOV. 1951. Birds of the Soviet Union, vol. 1. Moscow, Soviet Skaya Nanka. FZXNT,R.F. 1957. Glacial and Pleistocenegeology. New York, Wiley. F•iEmVrA•r•,H. 1950. The birds of North and Middle America,part 2. U.S. Natl. Mus., Bull. No. 50. HARDY,A. C. 1936. Handbook of colorimetry. Cambridge,Massachusetts, Tech- nology Press. HAYNES,C.V. 1966. Elephant hunting in North America. Sci. Amer., 214: 104-112. 338 STANLEYA. TE•rVLE [Auk, Vol. 89

HELL1VIAYR,C. E., AND B. CONOVER. 1949. Catalogue of the birds of the Americas. Field Mus. Nat. Hist., Zool. Ser., vol. 13, part 1, No. 4. HIJLTi;N, E. 1937. Outline of the history of arctic and boreal biota during the Quarternary period. Stockholm,Artiebolaget Thule. MAYR, E. 1946. History of the North American bird fauna. Wilson Bull., 58: 1-41. MAYR, E. 1969. Principles of systematiczoology. New York, McGraw-Hill. PETERS,J. L. 1927. The North American races of Falco columbarius. Bull. Essex County Ornithol. Club: 20-24. PETERS,J. L. 1931. Check-list of birds of the world, vol. 1. Cambridge, Massa- chusetts,Harvard Univ. Press. PtTELX^, F.A. 1941. Distribution of birds in relation to major biotic communities. Amer. Midi. Naturalist, 25: 113-137. PITELKA,F.A. 1951. Speciationand ecologicaldistribution in American jays of the genusAphelocoma. Univ. California Publ. Zool., 50: 195-464. REILLY, E. M. 1954. Bird records and the A.O.U. Check-list ranges. Auk, 71: 156-163. SALOMONSEN,F. 1955. The evolutionary significance of bird migration. Danske Biol. Medd., 22: 1-66. SELANDER,R. K., AND R. F. JOHNSTON. 1967. Evolution in the House Sparrow, interpopulationvariation in North America. Condor, 69: 217-258. SELANDER,R. K., R. F. JOHNSTON,AND T. H. HAMILTON. 1964. Colorimetric methods in ornithology. Condor, 66: 491-495. STEELE,R. G. D., ^ND J. H. TORmE. 1960. Principles and proceduresof statistics. New York, McGraw-Hill. SWA•T•, H. S. 1935. Comments on the northwestern races of several birds. Con- dor, 37: 201. UDVA•D¾,M. D.F. 1958. Ecological and distributional analysis of North American birds. Condor, 60: 50-66. WELLS,P.V. 1970. Post glacial vegetational history of the Great Plains. Science, 167: 1574-1582.

Laboratory of Ornithologyand Division of BiologicalSciences, Cornell University,Ithaca, New York 14850. Accepted26 March 19'71.