ECOGEOGRAPHIC VARIATION IN THE AMERICAN ROBIN (TURDUS MIGRATORIUS) JOHN W. ALDRICHa'3 AND FRANCESC. JAMES2 'Departmentof VertebrateZoology, National Museum of Natural History,Smithsonian Institution, Washington,D.C. 20560 USA, and 2Departmentof BiologicalScience, Florida State University, Tallahassee, Florida 32306 USA AnSTRACT.--Fromanalysis of externalmeasurements of 949 museumstudy skins of Amer- ican Robins(Turdus migratorius) taken in the breeding seasonthroughout North America,we estimatedthe extentof geographicvariation in size,shape, and colorof plumageon the basis of 39 ecoregions.After we defined seven univariate size and six scale-freeshape variables, we studiedtheir variationand the extentof their allometriccovariation. Because wing length is highlycorrelated with sixother reasonable measures of generalsize but is simpleand easy to visualize,we usedwing length as a measureof generalsize. The smallestrobins occur in the warm, humid easternUnited States,and the largestoccur in the high, cool, and dry Rocky Mountains,northern Great Plains,and northern deserts. Wing shape changesallometrically with size, becoming more pointed in larger birds, but wing shapein the Arctic is more pointed than would be predictedby this trend. Another exceptionto the allometriccovariation of size and shapeis that leg length as a proportion of generalsize is smallin Mexicoand large in the cool,humid forestsalong the Pacificcoast. Robinsin the coolmaritime forests of Newfoundlandhave unusually long tarsi. This variation in leg length is accompaniedby variation from pale plumagein Mexico,especially in Baja California, to dark plumage in Newfoundland. Bill length covarieswith general size, except that robinswith the longestbills occurin the isolatedpopulation in the mountainsof southern BajaCalifornia. Univariatecorrelations among size and shapevariables are usefulfor studyof the complex nature of geographicvariation in the robin, but graphicpresentations of the resultsin maps and bivariate graphsare even more informative.They allow comparisonsbetween specific aspectsof morphometricvariation, and they allow the identificationof extremecases. To understandthe relationship among geographicvariation in robins, climatic variation, and altitude,we analyzedcorrelations among these variables. The eight climatevariables are meannoon dry- and wet-bulb temperatureand relativeand absolutehumidity for April and for July.In the United States,the generalsize of robinsvaries not with dry-bulb temperature, as predictedby Mayr's definition of Bergmann'sRule, but with variablesthat are a function of both dry-bulb temperatureand the evaporativepower of the air. Multiple regression analysisusing climateand altitude aspredictors of the generalsize of robinsshows that July absolutehumidity alone has exceptionallyhigh predictive power. When general size is viewed on a psychrometricchart in which ecoregionsare locatedon the basisof their July climate, the relationshipsare clarified. The plumage of the robin is darkest not in the most humid regions, which are in the southeasternUnited States,as would be predictedby the mostcommon definition of Gloger's Rule, but in cool areaswhere the percent saturationof the air is high. Thesetrends of geographicvariation in size and color are concordantwith thosereported in many other species.Whether parallelismamong unrelated speciesin their patternsof geographicvariation and its high correspondencewith gradientsin the physicalenvironment are evidenceof naturalselection, either by physicalfactors or indirectlyby their covariates, is unknown. Received25 September1989, accepted 23 August1990. OURPRIMARY objective in this paper is to de- and color of the plumageof the AmericanRobin scribemajor geographic patterns in size,shape, (Turdusmigratorius) for all of North America. Size variables and shape variables are defined a priori, so their covariation can be assessedand 3 Presentaddress: 5666 EastHampton, Tucson,Ar- exceptions noted. Becausewe are interested in izona 85712 USA. ecologicalcorrelates of the variation,we organ- 23O The Auk 108: 230-249. April 1991 April 1991] EcogeographicVariation 231 ize the data by ecoregions.We demonstratecli- tion. In birds these include (1) the various matic variation within the United States on a covariatesof geographic variation in the phys- psychrometricchart that gives the relationships ical environment (e.g. microhabitat, diet), (2) between measuresof temperatureand humid- the influenceof crypticcolor on survival (Bow- ity. Then we show the pattern of size variation ers 1960), (3) character displacement (Grant in robinsin this space.We discussthe generality 1987), (4) isolation and other historical factors of correlationsbetween the physical environ- (Endler 1982, Cracraft 1985, Lande 1985), (5) ment and character variation in the robin in the distance and chance (Schnell and Selander 1981), context of ecogeographicrules. and (6) nongenetically based developmental Bergmann'sRule, once an important explan- plasticity (James 1983, Slagsvoid and Lifjeld atory theory in evolutionary biology, has been 1985). However, if unrelated speciesthat differ losing support (Zink and Remsen 1986). This in life history are found to exhibit parallelism thesiswas that birds and mammalshave an op- in character variation, causes1-5 are less par- timal ratio of size to surface area that is a func- simoniousexplanations than is 6. The argument tion of their thermal balancewith the physical that either environmentally induced physio- environment. As evidenceBergmann noted that logical responsesor genetically basedadapta- larger membersof a genus, with their larger tions to either the physical environment or its ratios of size to surface area, usually live in correlatesare a direct causeis strengthened. colder climates.This argument has been debat- ed by physiologists (Scholander 1955, Ken- METHODS deigh 1969,McNab 1971,Calder and King 1974), but systematistshave often found substantial Morphometricdata and analysis.--Aldrichmeasured empirical support for Bergmann's rule at the 949 museumstudy skins of adult robins collectedin intraspecific level (Mayr 1970), often by re- Mexico, the United States, and Canada. The mea- gressingsize on latitude, altitude, or isophene surements, made to the nearest 0.1 ram, are as follows: (e.g. Packard 1967, Johnston 1969). wing length (WL), from the bend of the folded wing James(1968, 1970) and Power (1969) showed to the tip of the longest primary feather; wing tip that the relationshipbetween intraspecific size (WT), from the end of the longestsecondary of the folded wing to the tip of the longest primary; tail variation and climate is stronger if the effects length (TL), from the insertion of the two central of temperatureand humidity are consideredto- rectrices to the end of the longest rectrix; culmen gether. James(1970), who found extensivepar- (CU), bill length from the depressionbetween the allelism in geographicsize variation in 12 spe- cranium and the maxilla to the tip of the maxilla; cies of birds in the eastern and central United maxilla height (MA), depth of upper bill at the distal States,suggested that Bergmann'sphysiological end of the depressionaround the nostril;tarsus length argument might still be valid in this context. (TS), from the posterioraspect of the junction of the Small size would be an advantage in a warm, tibiotarsusand the tarsometatarsusto the junction humid environmentbecause a large ratio of re- with the middle toe (anterior aspect);and middle toe spiratory surfacearea to body size would facil- (TO), from the junction with the distal end of the tarsometatarsusto the proximal end of the claw. Mea- itate heat loss.Larger birds would be expected surementerror was not evaluatedand probably con- in higher, or cooler, or drier environments. A tributes to variation in the data. combination of high, dry, and cool conditions In addition to the seven measurements, which we would favor the largest size. Grinnell (1910), call size variables, we used six dimensionless(scale- Boyce(1979), and Murphy (1985) have pointed free)shape variables in the analysis.These were wing out that variancein climatic variablesis higher tip/wing length (WT/WL), tail length/wing length in cool and dry areas than it is in warm and (TL/WL), culmen/wing length (CU/WL), culmen/ humid areas.Rather than dismissas overly sim- maxillaheight (CU/MA), tarsus/winglength (TS/WL), plisticthe hypothesisthat intraspecificpatterns and tarsus/toe(TS/TO). All of theseshape variables of size variation in birds may be thermal ad- have potential functional significance:WT/WL for flight, CU/MA for manipulationof food, TS/TO for aptations(Zink and Remsen1986), we prefer to ground locomotionand foraging,and TL/WL, CU/ examine relationshipsbetween size variation WL, and TS/WL for tail length, bill length, and leg and climate in more detail. size as their proportionsof general size (as measured Our analysesfor the robin are explicitly cor- by wing length) for a bird that forageson the ground relative. Causesother than physiologicalad- in parklandhabitat. The shapevariables describe var- aptations could account for geographic varia- ious aspectsof the shapeof the bird, and their geo- 232 ALDRICHAND JAMES [Auk,Vol. 108 graphic variation is of at least as much relevance to For our comparisons,pooling by ecoregions,which fitness in different environments as are size variables. are sensitiveto ecologicaldifferentiation, is prefera- Our statisticalcomparisons use lognormal assump- ble to pooling by other methods.