ELEVATIONAL CORRELATES OF SPECIATION AND INTRASPECIFIC GEOGRAPHIC VARIATION IN PLUMAGE IN ANDEAN FOREST BIRDS
GARY R. GRAVES • Departmentof BiologicalScience, Florida State University, Tallahassee, Florida 32306 USA
ASSTRACT.--Intraspecificvariation in plumage was used to test the null hypothesesthat geographicvariation in 280 speciesof elevationallyrestricted Andean forestbirds is inde- pendent of elevation and is not a function of patchy geographicdistribution. Both null hypotheseswere rejected. At most taxonomiclevels, geographicvariation in plumage was correlatedpositively with both the mean of its elevationaldistribution and the size of its geographicrange. Vertical amplitude of elevational distribution was not a significantpredictor of geographicvariation in plumage in most taxa. Independent of these elevational correlates,patchily distributed speciesshowed significantly more geographicvariation than continuouslydistributed species. Theseresults show that geographicvariation and presumablyongoing speciation phenom- ena are greaterat higher elevations.The decreasedspecies richness at high elevationsmay be attributable to a higher rate of extinction from catastrophicdisturbance as well as to ecologicalfactors that limit sympatryin newly formed species.Received 6 August1984, ac- cepted26 December1984.
IN this paper I examine the relationship be- of potential new species. Second, there are tween elevational distribution and possible adaptivepeaks; the morenumerous these peaks, speciationphenomena in Andean forestbirds. the greaterthe numberof potentialnew species. Two initial observationsare pertinent. First, on Predictions--(1)The frequencyof fragmented an elevationalgradient in Peru, Terborgh(1977) populations and incipient speciation increases found a negativecorrelation between elevation with elevation. (2) Intraspecific geographic and bird speciesrichness, canopy structure, and variation increases with elevation. (3) Extinc- number of plant strata. Second, geographic tion ratesof isolatedpopulations are greater at barriersto dispersaland gene flow increasewith higher elevations.(4) The range of speciesmor- elevation. Catastrophic vicariant events (e.g. phologiesand the diversity of foragingbehav- glaciation,landslides, vulcanism, forest frag- ior within ecological assemblagesdecreases mentationby climatic events)are more preva- with increasing elevation in correspondence lent in the physiographicallycomplex Andean with the decrease in resources and habitat com- highlands than in the adjacent foothills and plexity. Amazonian lowlands. I investigatedthe first two predictionsby ex- I proposea simple model of speciation(see amining geographicvariation in plumage char- Mayr 1963,Pielou 1979)that incorporatesthese actersin Andean birds. I began with two basic observations,predicts a pattern of geographic assumptions:bird speciationis largely allopat- variation correlated with elevation, and ac- ric, and plumage differentiation is related to countsfor the decreasein speciesrichness with speciation. I tested the null hypothesis that elevation. Two factors control the number of geographicvariation is independent of eleva- potential speciesalong the elevational gradi- tion and a subsidiary hypothesis that geo- ent. First, parent speciesare fragmentedby vi- graphic variation is not a function of patchy carlant events into a number of mutually iso- geographicdistribution. lated demes. The more fine-grained the fragmentationpattern, the greaterthe number STUDY AREA AND METHODS Studyarea.--Humid Andean forest extendsnearly uninterrupted from Venezuela southward along the • Present address: Division of Birds, National Mu- easternslope of the Andesto centralBolivia. My study seum of Natural History, SmithsonianInstitution, area included the portion of the easternslope of these Washington, D.C. 20560 USA. mountains from the Rio Marafi6n in northern Peru
556 The Auk 102:556-579. July 1985 July1985] SpeciationinAndean Birds 557
south to the Bolivian border (Fig. 1). This region was chosenbecause of the relativehomogeneity and even distribution of humid Andean forests (ONERN 1976); the linear distribution of Andean forest bordered by pdramoand punaon the west and lowland rain forest on the east, which facilitated mapping geographic ranges and locating contact zones and gaps in the distribution of montane bird species;and the avail- ability of specimensfrom a seriesof mostly unpub- lished post-1965 expeditions [Louisiana State Uni- versity Museum of Zoology (LSUMZ), Princeton University (specimensdeposited in the American Museumof Natural History, AMNH), and the Field Museum of Natural History (FMNH)] to the eastern slopeof the PeruvianAndes. Although the study area spansca. 10ø latitude, there is little regional differ- ence in forest physiognomycorrelated with latitude. Exceptfor outlying spursof the EasternCordillera, the undulating ribbon of humid Andean forest is backedby high-elevation pdramoor puna.The entire Fig. 1. Distribution of middle- to high-elevation elevational slope measured from the Amazonian humid Andean forest (ca. 1,200-3,400 m) along the lowlands (<500 m elevation) to timberline is rarely easternslope of the PeruvianAndes (adaptedfrom more than 30-50 km. The area and quality of pres- ONERN 1976). Note the deeply dissectedcontours ent-day montane forest at any contour interval is associatedwith river canyonsand timberline. Forest unknown becauseof heavy cloud cover and inacces- on the lower slopes(600-1,200 m) is continuousand sibility. Large-scalemaps indicate that the area more evenly distributed. between the 500-m and 1,500-mcontour lines is per- haps twice as large as the area in the 1,500-m to 4,000-m contour interval. Humid Andean forest above and genera suggestsa long occupancyin the region 1,500 m elevation often is restricted to narrow cor- and an autochthonousorigin of the Andean forest ridors or is patchy at the heads of several deeply avifauna.Species residing in other habitatshave dif- convoluted,dry intermontanevalleys (e.g. upper Rio ferent evolutionary histories(Haffer 1969, 1974). For Huallaga, upper Rio Apurimac, Rio Huari Huari). this reason,species not restricted to humid Andean Humid forest of the Andean foothills (600-1,200 m forest are excludedhere. Examplesof excluded hab- elevation) is continuouslydistributed. Precipitation itats and speciesare: (1) puna-pdramograssland (e.g. data from Andean forest in Peru are incomplete,but Asthenesfiammulata, Phrygilus unico!or); (2) Polylepis annual precipitationranges from 2 to 16 m (ONERN woodlands above timberline (e.g. Oreomanesfraseri, 1976, Simpson and Haffer 1978). A distinct wet sea- Xenodacnisparina); (3) intermontane deciduousforest son occursfrom November to May, but rainlesspe- (e.g. Pachyramphusalbogriseus, Tangara viridicollis); (4) riods of more than a few days are rare year-round. montane streams(e.g. Cinclusleucocephalus); (5) agri- Four distinct vegetation zones (lowland rain forest, cultural areasand scrub (e.g. Troglodytesaedon); and montane rain forest, cloud forest, and elfin forest), (6) lowland forest (primarily <500 m elevation;e.g. usually are present along the continuousphysical Piaya cayana). gradientfrom the Amazonian lowlands to timberline I alsoexcluded several species recorded from only (see Terborgh 1971). The term "Andean forest" (ca. one transect block (see Geographic range), repre- 600-3,500m elevation)refers collectively to montane sentedby insufficientspecimen material (e.g. Accip- rain forest, cloud forest, and elfin forest (sensuTer- iter collaris,Gallinago imperialis, Aegolius harrisii), or tax- borgh 1971). Excellent descriptionsof the humid for- onomically problematic (Scytalopus,Elaenia). The est of the Peruvian Andes can be found elsewhere remaining 280 speciesand superspeciesoccurring in (Tosi 1960; Terborgh 1971, 1977; Terborgh and Dud- the studyarea and includedin the analysesare listed ley 1973;Terborgh and Weske 1975;ONERN 1976). in the Appendix. The compositionof families follows Specieslist.--Of the 1,678species recorded from Peru Morony et al. (1975). (Parkeret al. 1982),about 320 exclusivelyinhabit An- Geographicrange.--To facilitate the measurementof dean forest on the Amazonian side of the continental geographicrange, the studyarea was partitioned into divide. The habitatfidelity of this heterogeneousas- 15 blocks (Fig. 2). This was an attempt to reduce the semblage(30 families)is supportedby literally thou- effectsof coarsesampling (Pielou 1979) without sac- sandsof observationsspanning 70 yr (e.g. Chapman rificingaccuracy by excessivesubdivision of the study 1917, 1926; Weske 1972). Although fossil evidence is area. Each block was delineated to include at least not yet available,the large number of endemicspecies one major collecting locale (>250 specimens), and 558 GARYR. GRAVES [Auk, Vol. 102 A B GEOGRAPHICRANGE C'D 200 km o 2 ;5 4 5 6 ? 8 9 I0 II 12 I;5 14 15 NUMBER OF BLOCKS
Fig. 3. Histogram of geographic range values for all species(n = 280). Speciesoccurring in only one H J block were omitted. I M studyarea (Fig. 3). This resultsfrom estimatingrange from a segmentof the Andes that restrictsthe upper value to 15 (equivalent to 15 or more) transectblocks. Geographicrange values indicate only the number Fig. 2. Division of the study area into 15 transect blocks. of blocksin the study area in which a speciesis be- lieved to occur.Range 15 valuesimply that a species is found throughoutthe study area;values less than 15 can have continuous or discontinuous distribu- representativecollecting stationsabove 2,500 m and tions.The true distributionof rangeof Andeanmon- below2,000 m elevation.The principalcollecting sta- tane-forestspecies probably would be more normally tions in eachtransect block are listed in Graves(1983). distributed,with fewer extremely small or large A full complementof gradient habitatsfrom lowland ranges.Species with very small continuousranges forestto timberlinewas presentin eachtransect block. are unlikely to exhibit marked geographicvariation. There was no significantcorrelation between the area Other factorsbeing equal, geographicrange is cor- of montane forest in a block and either the number related with the potentialfor geographiccharacter of recorded speciesor specimenscollected within a variation. However, species with small but geo- block. The boundaries between blocks were estab- graphically distinct populations often are highly lished, whenever possible,at discontinuitiesor nar- variable or polytypic. Therefore, the presenceand row corridors in montane forest distribution. Certain locationof all putativedistribution gaps were noted. transectblocks (e.g. Cordillera Colfin, Block A) en- Elevationaldistributions.--The most complete distri- closeisolated populations of high-elevationspecies. butional survey of bird speciesalong an elevational Geographic range was based primarily on speci- transect in the Andes was conducted in the Cordille- men label data in the LSUMZ, AMNH, and FMNH, ra Vilcabamba,Department of Cuzco,Peru (Terborgh and on publishedrecords (e.g. Zimmer 1931-1955). and Weske 1969, 1975;Terborgh 1971, 1977;Weske A specieswas consideredto occurin a transectblock 1972). Despite its thoroughness,the recordeddistri- if it wasrepresented by a specimenor, in somecases, butional limits, relative speciesabundance, and de- by sight recordsof easily recognizedspecies (e.g. greeof overlapor segregationof congenersalong a Oroaetusisidori). Terminal range boundaries were transectincorporate several possiblesources of error confirmedby publishedrecords or specimens.Species that must be recognized in comparativeanalyses. were consideredto be patchily distributed and ab- These are summarizedbriefly and apply to all An- sent from a transectblock only if they were not re- dean transects.First, irregularities in local topogra- corded by intensive post-1970fieldwork in suitable phy (e.g. landslides,canyons, rock faces)often pre- habitatfor the species(e.g. Atlapetes rufinucha in Block vent adequate determination of elevational limits E). Otherwise, rangeswere extrapolatedto include along portionsof the transect.Experienced "trail cut- intermediatetransect blocks. If one judgesby the dis- ters" follow the path of leastresistance and purpose- covery of many secretive,local, and sibling species fully avoid impenetrablehabitat formations. In many (e.g. O'Neill and Graves1977, Weske and Terborgh cases,mist nets unavoidablyare placedon trails, and 1981),systematic collecting has been distributeduni- observations are restricted to within a few meters of formly along the easternslope of the Peruvian An- them. Speciesthat occurin denseor patchily distrib- des. uted habitatsare missedfor this reason.Field camps The calculatedgeographic ranges have possible and mist-netting efforts usually are spacedat inter- values from 1 to 15; however, the distribution is valsalong the elevationaltransect [e.g. 9 campsalong highly skewed toward maximum coverageof the the 40-km Vilcabamba trail (ca. 3,000-m elevational July1985] SpeciationinAndean Birds 559 range), Terborgh 1977]. Resolutionof elevational limits between field camps is difficult, stemming in uJ part from the burden of maintaining and checking Q. mist nets up to severalhundred metersabove or be- low camp.Inclement weather, particularly prevalent • ;5 above 1,500 m elevation, often fouls mist nets, hin- dersspecimen preparation, and reducesbird activity •"50t during the time allotted for samplingat a particular elevation [e.g. LSUMZ field parties (1977-1980) re- MIDDLE OF INTERVAL cordedmoderate to heavy rains and high winds on Fig. 4. The cumulativedistribution of amplitude 79 of 83 campdays in the Divisoriade Huancabamba, (openbars) and mean elevation(shaded bars) values Departmentsof Cajamarcaand Piura, Peru]. for all species (n = 280). Elevation intervals are Although many distributional limits are docu- 200 m. mented by voucher specimens,observer experience is an importantsource of samplingerror. Localmon- tane-forestavifaunas cannot be evaluatedfully in a neutral random mutationsor through pleiotropic ef- few weeks of fieldwork. Distinguishing between fects. breeding residentsand elevational wanderers is dif- For this reason,I took an operationalapproach to ficult. Despitemany confoundingvariables, the ele- comparativesystematics and classified and compared vational limits of humid Andean forest birds are rel- speciesin termsof presence(continuous and discon- atively consistentalong the easternslope of Peru. tinuous variation) and absenceof geographicvaria- However, becausethe type of data (sight record or tion in plumage(Graves 1983). In this studyI treated specimen)and the numberof observationsdiffer from plumagevariation among populations rather than speciesto species,confidence bands of elevational variation within single geographiclocalities. Thus, limits cannot be reliably determined for between- speciesoccurring in only one transectblock were speciescomparisons. To minimize error in this study, excluded.All plumagesurfaces on museumstudy I used the maximum and minimum elevations re- skinswere scrutinized,including some not normally corded for each speciesalong the eastern slope of visible (e.g. interscapularpatches). Specimens in un- Peru. These were converted to elevational means and worn adult plumage were examined in geographi- verticalamplitudes (Fig. 4, Appendix).I compiledel- cally arrangedseries under naturallight. Color de- evationalranges from severalsources that used cali- terminations(Ridgway 1912)and Kodachromeslides brated altimeters:(1) my own fieldwork, July-No- of unique specimenswere taken in certainmuseums vember1976, May-November 1977, June-July 1978; for later comparisonwith standardspecimen series. (2) unpublishedfield notesof TheodoreA. ParkerIII, Specimenseries were examined at the LSUMZ, ThomasS. Schulenberg,Mark B. Robbins,J. W. Eley, AMNH, and FMNH. Faded or "foxed" specimens and John P. O'Neill; and (3) post-1965elevational were examinedfor plumagepattern but not for color data on museum labels in LSUMZ, AMNH, and comparison.My visionis normalwith respectto red- FMNH. green color-blindness(Ichikawa et al. 1978). Museumstudies.--The relationship of sizeand shape Classificationof plumage differentiation.--Initially, all charactersto speciation and race formation is un- populationsamples of a givenspecies were examined clear. Preliminary studiesof Andean montane-forest for geographicvariation in plumage.Those species birdsreveal a chaoticpattern in geographicsize vari- not displayingbetween-population variation within ation at the communitylevel unlike the smoothpar- the studyarea in eitherplumage color or pattern(i.e. allel clines observedin Northern Hemisphere stud- uniform along the easternslope of the PeruvianAn- ies (Remsenin prep., Gravesin prep.).Consequently, des)were classifiedas having "no variation."If vari- size and shapecharacters are probablyof lessvalue ation was observed, a second diagnostic procedure in the study of comparativespeciation than discrete was conducted. To determine whether variation was plumage characters,especially for poorly known continuousor discontinuous,each plumage character tropical avifaunas. or charactercomplex was evaluatedindependently. In contrastto size and shape,plumage pattern and For specieswith largegeographic distributions with- colorare relativelyconservative. In somecases there in the studyarea, variable characters were compared is no apparent correlation of environmental clines primarily with serially adjacent Peruvian popula- with color pattern (Graves 1982). Predation, sexual tions. WidespreadAndean specieswith limited dis- selection(see review by Bakerand Parker1979), in- tributionsin the PeruvianAndes were comparedwith traspecificcompetition (Rohwer 1982), interspecific populationsnorth (Ecuador,Colombia) or south(Bo- interactions(Moynihan 1963, 1968; Cody 1974), and livia) of the studyarea. However, classification here environmental abrasion (Burtt 1981) influence the wasbased only on variationwithin the studyarea. evolution of plumage pattern and color in birds. Characterstates that appearedto vary continuously Noncryptic plumagesalso may arise by selectively between populationsin serieswere noted. This pro- 560 GARYR. GRAVES [Auk,Vol. 102
cesswas repeated for all taxonomicallyapplicable 1939).Species with "continuoushorizontally stepped charactersof each species. clines"were tentatively categorizedas having dis- I mappedrange boundaries, distributional gaps, and continuous variation if the zone of clinal continuum steepeningcharacter gradients. Careful attentionwas betweengeographically uniform populationswas less given to all zonesof contact.Without specificknowl- than 5% of the species'range within the study area. edge of the evolutionaryhistory of the populations In many cases,discontinuous variation was correlat- under examination,primary rs. secondarycontact is ed with the presenceof a geographicalbarrier or a impossibleto determinein a generaloverview (En- gap in the distribution of a species.The discontin- dlet 1977).However, uniform, geographically wide- uous-variationcategory included specieswith pop- spreadpopulations connected by relatively narrow ulations thought to be below the allospecificlevel zones of intergradationare probably in secondary ("discretevariation," IV) andsuperspecies having two contact(Mayr 1963).Species lacking discretelyvari- or moreallospecies ("allospecies," V) within the study ablecharacters and differing only through clines were area (e.g. Metallura, Schizoeaca).Single allospecific classifiedin the "continuousvariation" category. populationswithin the studyarea (e.g. Buthraupis au- Thesespecies were subdividedinto thosewith only reodorsalis)were examinedfor geographicvariation smooth clinal variation (= continuoussmooth clines; without regard to allospeciesoccurring outside the seeHuxley 1939)and thosewith clinescontaining study area. distinctsteps (=continuous obliquely steppedclines). In summary,the following categoricallevels (I < In this paper, I refer to these as "smooth clinal" (II) II 28OO ;'508.5 1176.$ 2140ß•' I057ß U.I 2ooo r• 150.8 • 101.5 82.0 • 96.2 129.9 •. 99.4 116.7•. 9,.'3.0 O_ 1200 171'*'ß 692 315 II 19.2• •2.2 •8.2 •7.8 28.8 •. 28.4 400 154.5 77ß 4•0 1200 2000 2800 3600 III MEAN ELEVATION Fig. 5. Plot of amplitude and mean elevation val- 274 ues (m) for all species(n = 280). IV 28.4 •. 25.8 no significant difference between them and sexuallymonochromatic species in geographic Fig. 6. PairwiseMann-Whitney U-testsfor differ- variation (plumage I-V) within the study area encesin meanelevation between species with differ- (Mann-Whitney U-test, P > 0.05). Additional- ent levels of plumage differentiation (I-V). Mann- Whitney U countsare given at the top of each cell. ly, there was no significantdifference between The mean ranks of elevation for each plumage com- the two groupseither in extent of geographic parisonare shownat the bottomof eachcell (column range within the study area or in mean eleva- ranks on left, row ranks on right). A triangle indi- tion and amplitude (Mann-Whitney U-test,P > catesthat the differenceis significant(two-tailed; one 0.05). This indicates that any relationship be- triangle = P < 0.10, two triangles = P < 0.05). tween plumagedifferentiation and elevation or geographic range is unlikely to be biased by possibledifferential assessmentof variation in the use of parametricstatistics that assumenor- sexually dichromatic and monochromatic mal distributions was ill advised. species. A nonparametric test of the data from the Elevational distributionand geographicvaria- total specieslist indicated there was a signifi- tion.--Two important componentsof elevation- cant difference in mean elevation between al distribution are the location (here estimated specieswith different levels (I-V) of plumage by mean elevation)and vertical amplitude of a differentiation (Kruskal-Wallis one-way test; species'distribution along an elevational gra- x 2 = 16.75, P = 0.002). My expectation that dient. Both variables may be correlatedwith plumage levels are orderedwith respectto el- geographicvariation in Andean birds. Species evational parameters was examined with the that occurat extreme elevations(i.e. with high Mann-Whitney U-test (Fig. 6). Mann-Whitney or low mean elevations) have restricted ampli- countsfrom 6 of the 10 cells were statistically tude values. Those with intermediate mean el- significant;only 1 of these differed unexpect- evations may have large or small vertical am- edly (II >III). The alternativehypothesis in the plitudes. Thus, mean elevation and amplitude form Ha: I < II T^I•LE1. Spearman'srank correlationcoefficients for w the relationshipbetween plumage level and mean © © ©©5 elevationat eachrange value usingdifferent num- bers of plumage levels. w IV ß ß 2 ß 4 2 21 Number of plumagelevels a 14 Range value n b 2 3 5 © 2 ©5© 25 15 146 0.24** 0.29*** 0.28*** IiiI 12 8 7 6 5 9 2 2 16 8 9 I0 79 13 18 0.50* 0.43* 0.43* 12 19 -0.25 -0.47* -0.46* 0 5 6 11 11 0.58* 0.57* 0.55* GEOGRAPHIC RANGE 10 19 0.04 0.04 0.04 7 13 0.05 0.05 -0.03 Fig. 7. Plot of the relationshipbetween level of 6 10 0.24 0.24 0.31 plumagedifferentiation and sizeof geographicrange 5 6 0.00 c 0.00 c 0.00 c 4 8 0.41 0.41 0.41 in transectblocks for all species(n = 280). Dots in- 3 8 0.00 c 0.00 c 0.00 c dicateone species,and numbersindicate tota! species. 2 13 0.39 0.39 0.39 Fisher's test X• = 45.15 X2 = 47.38 X2 = 46.02 Whitney U countsand J* is normally distrib- P < 0.003 P < 0.002 P < 0.002 uted. ß p < 0.05, ** P < 0.01, *** P < 0.001. The null hypothesisthat plumagelevels are Sample size. not ordered with respectto mean elevation can All plumagevalues were equal. be rejected(J* = -4.02, P < 0.00003).These re- suitsimply that plumagedifferentiation is pos- itively correlatedwith mean elevation;how- probabilitiesfrom the correlationcoefficients ever, they do not accountfor variation due to showedthat plumagedifferentiation of species the size of geographicrange (Fig. 7). All other was positively correlatedwith mean elevation factorsbeing equal, the larger a species'geo- (Table 1). These results indicate that the rela- graphicrange is, the higher the probabilityit tionshipwas robustto minor changesin plum- will be dissectedby somebarrier to gene flow. age classification.Of particularinterest are the In thisstudy, "equal range" values were judged highly significantcoefficients for speciesthat to be roughly proportionalwhen used as an occurredthroughout the study area (range= independentvariable. 15). Becausethe contrastresults were similar, 5 The outcomeof any testsmay be biasedby plumagecategories were usedin the remaining misclassificationof speciesin plumagegroups. tests.The constraintof "equal" rangesreduced For statistical tests, I assumed that plumage the number of speciescombinations available levels were ordered (e.g. plumage level IV for comparisonat lower rangevalues. I relaxed speciesare more geographicallyvariable than the range constraintsby reducingthe number plumagelevel III species).Ambiguities in in- of range incrementsfrom 15 to 7 (e.g. 10-11, terpretationof clinesor degreesof specificdis- 12-13, 14-15) and recalculated the correlation tinctiveness(IV vs. V) could blur the division coefficientsfrom the new range categories.The between adjacentplumage categories. combinedprobabilities were highly significant I examined the possibleeffects of misclassi- (Fisher'stest, X2 = 32.6, df = 12, P < 0.004) and ficationby makinga priori contrastsof 2 (I vs. again indicatethat the resultswere robustto II, III, IV, V), 3 (I vs. II, III, vs. IV, V), and 5 minor changesin data organization.In other plumagecategories. The relationshipbetween words,slight expansion of the geographicrange plumagelevel and mean elevationof species categoriesdid not substantiallyaffect the cor- with equalrange values then was assessed with relation between plumage differentiation and Spearman'srank correlationtest. I usedFish- mean elevation. However, none of the corre- er's test of combinedprobabilities (Sokal and lation coefficientsfor range classesbelow 15 Rohlf 1969) with a Chi-square distribution wassignificant at P = 0.05.A further relaxation [-2• In(p) = x2, df = 2n] to evaluatethe prob- of range constraintsyielded still lower signif- abilitiesacross range values. The combinedtail icance values. For this reason, conclusions based July1985] Speciationin Andean Birds 563 TABLE2. Spearman'srank correlationcoefficients for the relationshipbetween plumage level and mean elevationat eachrange value, omitting different levels of plumage.Sample sizes appear in the left-hand column under eachdata category.Missing data valuesindicate a samplesize lessthan 5. Range Plumagelevel omitted a value I II Ill IV V 15 67 0.25* 119 0.30*** 132 0.35*** 125 0.18' 141 0.21'* 13 8 -0.29 17 0.43* 18 0.43* 13 0.42 13 0.51' 12 10 -0.78** 14 -0.71'* 19 -0.46* 15 0.01 18 -0.40* 11 5 0.22 10 0.49 11 0.55* I0 0.42 10 0.62* 10 -- -- 17 0.05 19 0.04 19 0.04 19 0.04 7 -- -- 10 -0.41 12 0.25 13 -0.03 13 -0.03 6 -- -- 10 0.31 9 0.34 7 0.54 10 0.31 5 -- -- 6 0.00 b 6 0.00 b 6 0.00 b 6 0.00 b 4 -- -- 8 0.41 8 0.41 7 0.00 • 8 0.41 3 -- -- 8 0.00 • 8 0.00 • 8 0.00 • 8 0.00 • 2 -- -- 13 0.39 13 0.39 12 0.00 • 13 0.39 Fisher's test X2 = 10.38 X2 = 43.40 X2 = 53.49 X• = 32.47 X• = 41.66 P < 0.25 P < 0.005 P < 0.0005 P < 0.07 P < 0.009 * P < 0.05, ** P < 0.01, *** P < 0.001. All plumagevalues were identical. on analysesusing expandedrange classesare steppes),montane forest, and dry intermon- suspect.I assumedthat plumagecategories rep- tane basinsand a few genera from other re- resenteddivisions of a continuumof geograph- gions in South America. The sample was sub- ic variation (I < II TABLE3. Spearman'srank correlationcoefficients for the relationshipbetween plumage level and mean elevationat eachrange value for higher taxonomiccategories. Missing data valuesindicate a samplesize less than 5. Higher taxonomiccategories a Nonpasserines Passerines Suboscines Oscines Rangevalue nb r• n rs n r• n rs 15 42 0.24 104 0.32*** 51 0.31' 53 0.32* 13 6 0.06 12 0.59* 6 0.14 6 0.74* 12 -- -- 17 -0.46* I0 -0.65* 7 -0.31 11 -- -- 8 0.48 -- -- 5 0.35 10 -- -- 15 0.02 9 -0.41 6 0.53 7 -- -- 9 -0.64* 5 -0.89 -- -- 6 -- -- 7 0.71' .... 5 -- -- 6 0.00 ..... 4 -- -- 6 0.66 .... 3 -- -- 7 0.00 ..... 2 -- -- II 0.40 I0 0.41 -- -- Fisher's test X2 = 7.08 X2 = 47.13 X2 = 15.32 X2 = 23.48 P < 0.136 P < 0.002 P < 0.230 P < 0.010 * P < 0.05, *** P < 0.001. Sample size. All plumage levels were identical. found no significantdifference between the two ally positivecorrelation (Fisher's test, x2 = 7.08, taxonomicgroups. This suggeststhat at present P < 0.136) between plumage level and mean there are no majordifferences in the amount of elevation (Table 3). Passerinespecies showed a differentiation between higher taxonomic cat- highly significant correlation between plum- egoriesof Andean birds. age differentiation and mean elevation. Anal- Speciesfrom nonpasserineorders (tinamous ysis of the two principal subdivisionsof the through woodpeckers)showed only a margin- Passeriformesgave somewhatdifferent results. TABLE4. Spearman'srank correlationcoefficients for the relationshipbetween plumage level and mean elevation for specifiedrange values at the family level. Fisher's test for Family Range values n• rsb individual families Trochilidae 15 18 0.13 X2 = 7.98, P < 0.24 10-13 5 0.71 2-9 6 -0.34 Furnariidae 15 11 0.57 * Formicariidae 15 8 0.35 X2 = 4.69, P < 0.59 10-13 6 -0.09 2-9 8 - 0.58 Cotingidae 2-9 6 0.27 Tyrannidae 15 25 0.38* X2= 12.52, P < 0.052 10-13 15 0.00 2-9 12 0.36 Troglodytidae 15 5 0.67 Turdidae 15 6 0.66 Parulidae 15 8 0.30 Emberizidae 15 30 0.29 x 2 = 14.46, P < 0.025 10-13 20 0.46* 2-9 15 0.00 Fisher's test over all families 62.05, P < 0.001 Samplesize. b,p < 0.05. July1985] SpeciationinAndean Birds 565 TROCHILIDAE n=18 2 FURNARIIDAE I0] FORMICARIIDAE SPECIES/FAMILY 20_] • TYRANNIDAE Fig. 8. Plot of P-valuesfrom Spearman'srank cot- relation coefficientsand samplesize from family-level 101 .._r•"".' - ' '.zz[, comparisons(see Table 4). hi Suboscinesshowed a weak positive correlation :::) 2o (P < 0.23) between plumage level and mean el- z 15 evation, while oscines showed a significant i0 positive correlation (P < 0.01). When species 5 0 with discontinuous(patchy) geographicranges 6 i0 14' '18' '22' 26 30 34 (see Appendix) were omitted from the analy- ELEVATION (xlO0 rn) ses,the significancelevels of correlation coef- ficients and the combined probabilities across Fig. 9. The elevational distribution of speciesoc- range values generally were higher for all curringthroughout the studyarea (range = 15 blocks) from the five most species-richfamilies in the Andes. higher categories.These results reinforce the The elevationalgradient is divided into 200-m inter- conclusion that plumage differentiation was vals. Syrupatticspecies occurring within a particular positively correlatedwith mean elevation. Sig- interval are subdivided by magnitude of plumage nificant values across higher taxonomic cate- variation for comparison;unmarked area = no geo- goriesstrongly suggestthat this relationship is graphicvariation (I) in plumage,hatched area = cli- not limited to a small but influential taxonomic nal variation (II, III), and black area = discontinuous subset. variation (IV, V). Family-levelcomparisons.--At the family level, only 8 of 30 families had 5 or more speciesfor any range value. Over 66% (186 of 280) of the had lessthan 10 speciesfrom every range class. total specieslist occursin the 5 most species- For higher taxonomic categories,the signifi- rich families (Trochilidae, Furnariidae, Formi- cance level of combined probabilities across cariidae,Tyrannidae, Emberizidae).To increase families was higher when patchily distributed the sample size, I relaxed the "equal range" species were omitted (Fisher's test, X2 = constraintand groupedspecies from eachfam- 62.84, df = 30, P < 0.001). ily into three range classes(i.e. range values2- The elevational distributionsof continuously 9, 10-13, 14-15). Only 1 additional family (Co- distributedspecies (range = 15) from the 5 most tingidae) with 5 or more speciesper range class species-richfamilies are shown in Fig. 9. With was added. The correlation coefficients for the the exceptionof the Formicariidae,the propor- relationship of plumage differentiation and tion of specieswith discontinuousplumage mean elevation acrossrange classesand fami- variation (IV and V) increased with elevation. lies were highly significant (Table 4; Fisher's Only a few specieswith discontinuousvaria- test,X 2 = 62.05, df = 34, P < 0.001),although the tion occurredbelow 1,100 m. Although species probabilities of 5 families (Trochilidae, Cotin- with no variation (I) occurred along the entire gidae,Troglodytidae, Turdidae, Parulidae)were elevational gradient in each family, most were only marginally unusual (0.07 < P < 0.28), and found at lower elevations. The elevational dis- 1 was not significant (Formicariidae,P = 0.59). tributionsof specieswith continuousplumage Levels of significancein this test at the family variation (II and III) were somewhat interme- level were inverselyproportional to samplesize diate between the no-variation and discontin- (Fig. 8). Five of the 6 nonsignificantfamilies uous-variation groups. 566 GARYR. GRAVES [Auk, Vol. 102 TABLE5. Positiveand negative scoresfrom comparisonsof congenericspecies pairs with equal geographic ranges.Figures in parenthesesrepresent the number of scoresin genera when range restrictionswere relaxed(to incrementsof two blocks,i.e. 2-3, 4-5, etc.). The P-value is the binomial probability (for n > 25, the large-sampleapproximation of Hollander and Wolfe 1973) of finding this many positivescores in n trials, omitting ties.P-values in parenthesesrepresent the binomial probabilitywhen ties are partitioned and added to positive and negative categories. Genus Positive Negative Tie Pionus 1 0 0 Coeligena 2 0 1 Metallura 1 0 0 Grailaria 2 1 0 (2) Pipreola 0 1 0 Mionectes 1 0 0 Ochthoeca 2 1 3 Catharus 1 0 0 Cacicus 0 1 0 Basileuterus 1 1 1 Diglossa 3 1 2 Tangara 12 6 4 Chlorospingus 1 1 1 Hemispingus 0 (4) 1 (2) 1 (2) Atlapetes 1 (2) 0 0 Total 28 14 13 P < 0.02 (<0.03) Minus Tangara 16 8 9 P < 0.08 (<0.09) Comparisonsof congenericspecies.--I used a 3 plumagelevels were usedand to 3,747 (35.4%) modified binomial test of variable comparison when 5 were used.If more categorieswere used, for pairwise combinationsof congenericspecies. the proportion of ties would continue to drop. For eachgenus, all pairwisecombinations were The cumulative scoresof pairwise combina- generatedwith the constraintthat speciespairs tions were used to examine the relationship be- have equalgeographic ranges (occur in the same tween plumage differentiation and elevation. number of transectblocks). For each speciespair If there were no elevational differencesamong I askedthe question:Does the specieswith the plumage categories,the pairwise scoresshould higher level of plumage differentiation also follow a binomial distribution with P = 0.5. For have the higher mean elevation?If the answer the binomial test,the probabilityof finding that was "yes," I scoreda "+ 1" for that speciespair; many successes(positive or negative cate- if "no," I scoreda "-1." When plumage levels gories) in a given number of trials was calcu- were identical (tied), a "0" was scored for that lated with the large sample approximation speciespair. The + l's and -l's were totaled (n > 25) of Hollander and Wolfe (1973). Nontie for each genus. scorescould be calculatedfor speciespairs from It is worth noting that "0" values probably only 15 (9.0%) of 166 genera included in this do not represent identical levels of plumage study.Only 1 genushad 5 or more nontie scores differentiation but rather an inability to distin- (Tangara,18). When range constraintswere re- guish finer levels of differentiation. For these laxed by reducing the number of range incre- data, the frequency of identical values was a ments from 15 to 7 (e.g. 12-13, 14-15), only 6 function of the number of plumage levels dis- additional nontie scoreswere obtained (5 pos- criminated. The more plumage levels recog- itive, 1 negative). A binomial test of the cu- nized, the lesslikely it was that a pair of species mulative scoresover all genera showed a sig- would be identical. For example,from the total nificant correlation between plumage list there were 146 specieswith ranges of 15. differentiation and mean elevation (Table 5). In summary,there was a significantcorrela- Ofthe (1•46) =10,585 pairwise combinations, tion between plumagedifferentiation and mean there were 5,292 plumageties (50%)when only elevation at the generic, familial, ordinal, and 2 plumage categorieswere distinguished.The avifaunal levels in Andean forest birds. number of ties decreased to 4,234 (40%) when Verticalamplitude and differentiation.--Apre- July 1985] Speciationin AndeanBirds 567 TABLE6. Spearman'srank correlationcoefficients for the relationshipbetween plumage level and vertical amplitudefor the total speciespool and for all specieswith continuousranges (omitting species with distributional gaps >-I transectblock). a Range values 2-9 10-13 14-15 Mean elevation (m) Species nb r• n r, n 1,850 Total 18 0.65'* 27 0.40* 59 - 0.06 Continuous 15 0.44 21 0.04 59 -0.06 ,850-2,500 Total 26 0.37* 18 -0.22 43 -0.04 Continuous 24 0.24 13 -0.29 43 -0.04 2,500 Total 22 0.48' 22 - 0.29 45 -0.06 Continuous 17 0.33 18 -0.05 45 -0.06 * p < 0.05, ** P < 0.01. Sample size. liminary test of the data on the relationship To examine this hypothesisI divided the el- between plumage differentiation and ampli- evationalgradient into three subequalsections: tude indicated that there was a significant dif- (1) <1,850 m, (2) 1,850-2,500 m, and (3) >2,500 ferencein amplitude between specieswith dif- m. The mean elevation measures for 104, 87, ferent levels (I-V) of plumage differentiation and 89 species,respectively, fall in the three (Kruskal-Wallisone-way test, X2= 14.74, P < gradient sections.The possiblevalues of geo- 0.005). The vertical amplitude of a speciesmay graphicrange were divided in a similar fashion be viewed as a rough indicator of the breadth (i.e. 2-9, 10-13, 14-15). The relationship be- of habitat occupiedin the face of environmen- tween plumage differentiation and amplitude tal gradients and interspecific competitors. of speciesin each elevation/geographic range However, becauseof differencesin slopeangle, cell was examined. Overall, specieswith small topographicheterogeneity, structural complex- geographicranges (values 2-9) showed a sig- ity of habitat, and available area between any nificant positive correlation between plumage two elevational contours, the amplitude mea- differentiation and amplitude acrossall eleva- surement may not be directly comparable. For tional subdivisions(Table 6). This may be an instance,an amplitude of 500 m centered at 750 artifact of data collection. The number of lo- m elevation may encompass2 or 3 times as cations where elevational data were recorded much area as an "equal" amplitude centered at was roughly proportional to the extent of a 2,500 m elevation. In allopatric speciation, a species'geographic range (i.e. the amplitude narrow band of habitat is more easily frag- measurement tended to increase asymptotical- mented by vicariant events than a wider band. ly with geographicrange). As a point indicator, It follows that species with specific habitat mean elevation is less sensitive to the effects of needs met only in a narrow elevational zone extreme values. Specieswith very small contin- may be expectedto show more geographic uous geographic ranges (<5 transect blocks) variation. Generalist speciesor other birds that showed no geographic variation in plumage; require specifichabitats or foraging substrates however, some specieswith slightly larger that occurin a wide elevationalzone (e.g. bam- rangesshowed clinal variation (seeAppendix). boo) are lesslikely to vary geographically.An These specieswith wider amplitudes by virtue appropriatenull hypothesismight be that giv- of larger rangeswill influence the rank corre- en two specieswith equal geographic ranges lation tests (Table 6). Unfortunately, partition- and mean elevations, ceterisparibus, there is no ing these range categoriesdilutes the sample relationship between plumage differentiation size. Specieswith small, discontinuousranges and amplitude.An alternativehypothesis sug- often are differentiated. Isolated populations geststhat narrow "amplitudes" are more easily that have experienced some shift in niche ex- fragmented by barriers to gene flow; thus, pressed in vertical amplitude would increase specieswith narrower amplitudesat any given the amplitude measurementfor the species.This elevation should show more geographicvari- also produces a positive relationship between ation. plumage differentiation and amplitude. When 568 GARYR. GRAVES [Auk,Vol. 102 TABLE7. Spearman'srank correlation coefficientsfor the relationship between plumage level and vertical amplitude for higher taxonomiccategories? Missing data valuesindicate a samplesize lessthan 5. Range values 2-9 10-13 14-15 Mean elevation (m) Species nb rs n r• n < 1,850 Nonpasserines -- -- 5 0.35 12 0.04 Passerines 15 0.64'* 22 0.36' 47 -0.09 Suboscines 12 0.00 13 0.49* 29 0.24 Oscines -- -- 9 0.27 18 -0.50* 1,850-2,500 Nonpasserines 6 0.66 6 -0.45 16 -0.13 Passerines 20 0.26 12 -0.21 27 0.00 Suboscines 12 0.59' 8 0.08 11 - 0.04 Oscines 8 0.00 -- -- 16 - 0.02 >2,500 Nonpasserines 5 0.00 -- -- 14 -0.10 Passerines 17 0.50* 18 -0.24 31 -0.03 Suboscines 10 0.35 7 -0.41 12 0.11 Oscines 7 0.72* 11 -0.25 19 -0.09 * P < 0.05, ** P < 0.01. Sample size. thesespecies were omitted, the correlationcoef- ranks of the independent variables (Conover ficients were nonsignificant (Table 6) and and Iman 1981). showed a trend toward the predicted negative At the faunal level, the partial correlation correlation between differentiation and ampli- coefficientsfor mean elevationand geographic tude. Species with medium-size geographic range were highly significant (Table 8). Am- ranges(values 10-13) showeda significantpos- plitude generally was not a significant predic- itive correlationat low elevationsbut a nega- tor of plumage differentiation, and mean ele- tive relationship at medium and high eleva- vation was a significantpredictor of plumage tions. Species with large geographic ranges differentiation in non-Passeriformes and Pas~ showed a slightly negative, but nonsignificant, seriformes.However, a dichotomyoccurred be- correlation between plumage differentiation tween suboscinesand oscinesin the signifi- and amplitude at all elevations. canceof this variable and between geographic The significancelevels of the correlationcoef- range and amplitude as predictorsof plumage ficients from taxonomic subsets mirrored those differentiation. P-values of the partial correla- from the total speciespool (Table 7). Species tion coefficientsfor geographicrange and mean with small geographicranges showed a posi- elevationgenerally were smallerwhen species tive correlationbetween levels of plumagedif- with discontinuous(patchy) ranges were omit- ferentiation and amplitude, while no correla- ted from the analyses.Despite the small amount tion or a slightly negative correlationtypified of variance (total R2, adjustedfor degrees of specieswith relatively large geographicranges. freedom)in plumagedifferentiation explained The only significantly negative correlationwas by these parameters,it is surprising that the found in oscineswith large geographicranges simple measurementof elevational distribu- and low mean elevations. tion had any predictive power. Mean elevation,amplitude, and geographic range At the family level there were marked dif- as predictorsof plumagedifferentiation.--A non- ferencesin the relationshipbetween plumage parametric multiple regressionwas run, with differentiation and vertical amplitude, mean the ranks of mean elevation, geographicrange, elevation, and geographic range. Overall, and vertical amplitude as independent vari- plumage variation in hummingbirds(Trochil- ablesand plumagelevel asthe dependentvari- idae) showed little correlationwith any inde- able. Eachvariable was ranked separately,with pendent variable.This is due in part to several averageranks assignedto ties; the rank of the high-elevation taxa that showed no significant dependent variable was predicted using the variation in plumage in the study area (e.g. July1985] Speciationin Andean Birds 569 TABLE8. Significantcorrelations (all positive) from nonparametricmultiple regressionusing the ranks of independentvariables to predictthe ranksof the dependentvariable (plumage level) for varioustaxonomic assemblages. Independent variablesb Vertical Mean Geographic Taxon na amplitude elevation range Total R2c Total specieslist A 280 NS *** ** 6.9 B 255 NS *** *** 15.1 Nonpasserines A 71 NS * NS 0.4 B 64 NS * * 9.4 Passetines A 209 NS * * 5.1 B 191 NS *** *** 20.3 Suboscines A 114 * NS NS 11.7 B 104 NS NS ** 20.6 Oscines A 95 NS *** * 9.8 B 87 NS *** ** 22.7 Trochilidae A 29 NS NS NS -4.2 B 27 NS NS NS 6.5 Furnariidae A 18 NS * ** 28.4 B 17 NS * *** 46.7 Formicariidae A 22 * NS NS 24.8 B 19 NS NS NS 29.9 Tryannidae A 52 NS NS NS 13.1 B 48 NS NS * 20.6 Emberizidae A 65 NS ** * 6.7 B 58 NS ** ** 20.9 aSample size. A = continuouslyand discontinuouslydistributed species, B = only continuouslydistributed species. b * p < 0.05, ** P < 0.01, *** P < 0.001. cAdjusted for degreesof freedom. Pterophanes,Ensifera). Of the 5 largest families, graphic range (3 of 5) were significant predic- the Trochilidae is the most heterogeneous tors of plumage differentiation, whereas verti- (1.31 species/genus).This suggeststhat signif- cal amplitude was a nonsignificantpredictor in icance levels within families are related to some all 5 families. index of taxonomicdiversity. Although species: From these analysesI conclude that plumage genus ratioscorrelated significantly with num- differentiation was related to elevational dis- ber of species/family(omitting families with 3 tribution. This differentiation generally is cor- or fewer species;n = 15, R2 = 19.2%,P < 0.05), related positively with mean elevation. For most the significance of independent variables as taxa amplitude did not appear to be a statisti- predictorsof plumage differentiation in the 5 cally significantpredictor of differentiation. largest families was not correlated with the Patchinessand geographicvariation.--Few at- species:genus ratio. tempts have been made to analyze the role of The Furnariidae and Emberizidae conformed patchinessin differentiationof populationsand bestto the predictionsof the modelsat the fam- speciationin tropical avifaunas. Yet, distribu- ily level (i.e. differentiation was greater in high- tional patchinessis a commonly reported phe- elevationbirds with narrow amplitudes).When nomenon in tropical avifaunas(Diamond 1972, species with discontinuous ranges (1-7 per 1981). Most discussions are case-by-case ac- family) were omitted, the total R2 increasedand counts of range disjunctionsand the possible the partial correlation coefficientsfor ampli- cause-and-effectrelationship between various tude were closer to the predicted negative cor- ecologicaland historical factorsand patchiness. relation. In continuously distributed species, From a sample of 19 patchy and 16 nonpatchy mean elevation (in 2 of 5 families) and geo- Andean species(plus "8 ?patchy,4 ?nonpatchy, 570 GARYR. GRAVES [Auk,Vol. 102 PLUMAGE LEVEL significantlydifferent. This suggestseither that there is little difference in speciationrates be- tween patchy and nonpatchyspecies of pdramo and puna habitats or that some specieswere 75] < 1850 rn misclassifiedas patchy or nonpatchy. To test whether patchinessin montane forest speciesis correlatedwith plumage differentia- tion and presumably speciation, I scored the geographicrange of each speciesfor discontin- OJ uities or "gaps." On small-scale maps (1: 5,000,000) most Andean forest speciesappear 60 to have continuous ranges punctuated only by narrow river valleys or canyons.However, 25 1850-2500m (9%) specieshad range gaps that spanned one or more transect blocks and were considered to I be patchy (see Appendix). This definition of patchiness is particularly cautious because a specieswas consideredto be absent only if it has not been found in appropriatehabitats and O' elevationssince 1965.Several of these"patchy" species(e.g. Pipreolafrontalis) eventually may be found to have continuous distributions. It 4045] F-• shouldbe noted that mosttropical species (91% z 20' in this study)were not patchy,perhaps no more >2500m than Temperate Zone species. 15' Only three patchy species[Chamaepetes gou- dotii, Haplophaediaaureliae (one of two gaps), IO. Ampelionrufaxilla] exhibited no variation in 5' plumage acrossgaps. For the remaining patchy species,the range gap separated distinctive O' populationsor allospecies. III IV V In contrastto Vuilleumier's data for high An- dean birds, patchy montane-forestspecies (n = Fig. 10. Histogram of plumage levels (l-V) for patchy and nonpatchyspecies with mean elevation 25) in this study had significantly more geo- values falling in three intervals, <1,850 m, 1,850- graphic variation than nonpatchy species(n = 2,500 m, and >2,500 m. Specieswith continuous 225; Mann-Whitney U-test, P < 0.00001; Fig. (nonpatchy) ranges are representedby open bars; 10). There was no significant difference be- specieswith discontinuous(patchy) ranges are rep- tween the groupsin mean elevation or ampli- resentedby hatchedbars. tude. Becausepatchy species(24 of 25 showedgeo- graphic variation) may unduly influence the and 1 uncertain species"),Vuilleumier (1980) resultsof rank correlationtests, all specieswith concludedthat patchilydistributed species have discontinuous (patchy) ranges were omitted more geographicalvariation than nonpatchily from Table 1 (5 plumage levels), and the Spear- distributed species. man's rank correlation coefficients were recal- I reanalyzed Vuilleumier's (1980) data. The culated for nonpatchy species.The correlation reported trend toward increased geographic between plumage level and mean elevation variation with patchinesswas not statistically from this test was significant (Fisher's test, significant (absent vs. present, lines 1 and 2, X2 = 38.58, df = 22, P < 0.018). X2 = 1.80, P > 0.10; lines 5 and 6, X2 = 1.58, P > Despite the lack of statisticaldifference be- 0.20). Likewise, the proportionsof patchy and tween the groups in elevational parameters, nonpatchy specieswith morphologically dif- patchy species may differ fundamentally in ferentiated isolates, secondary contact phe- their abilities to disperse, colonize, use re- nomena,or belonging to superspecieswere not sources, or compete. July1985] SpeciationinAndean Birds 571 Diamond (1981) suggestedfour explanations TABLE9. The observedand expected number of for patchinessthat may act singly or in com- range gapsfor transectblocks B-N. bination: (1) undetected patchinessof habitat, Total (2) historicaleffects, (3) immigration-extinction Rangegaps Rangegaps species/ equilibria, and (4) "lockouts" by established Block observed expected block populations of competitors. As in most avifau- B 0 2.62 241 has, it is not certain what factorsare responsi- C 2 5.09 237 ble for patchinessin Andean montane bird dis- D 5 6.28 242 tributions. In this case, however, three of E 7 7.13 239 F 4 8.22 237 Diamond's proposed explanations for patchi- G 3 8.75 240 nessdo seem unlikely. H 9 9.18 219 Most patchy speciesidentified in this study I I0 8.88 217 have wide elevational distributions, and all are J 8 8.15 224 known to occurin the array of habitatspresent K 13 8.18 199 L 13 7.10 196 in the unoccupied gaps. It is highly unlikely M I0 5.91 186 that undetectedhabitat patchinessis responsi- N 4 3.53 183 ble for large disjunctions(>800 kin) in geo- 19.68, df = 12, 0.05 < P < 0.10 graphic range. Immigration-extinction equilibria probably are not an important factor in the production of large, long-lived gaps in otherwise contin- The distribution of range gaps, however, was uous populations. Repeated collections from nonrandom. Moreover, there was a negative many of the range discontinuitiesconfirm the correlation between the number of species absenceof many specieswhose absencefrom with range gaps at any single transect block the region was first noted over 60 yr ago (Chap- and the total number of speciesrecorded from man 1921). This indicates that patchiness in that block. these casesalmost surely is not an artifact of Competition at the interspecific or "guild" incomplete sampling. Additionally, the size- level is unlikely to have produced the gaps. able proportion of specieswith well-differen- Evidence of "lockouts"by establishedpopula- tiated taxa separatedby gapsindicates a long- tions of competitors,such as checkerboarddis- term genetic separationof populations(Mayr tributions of congenersor geographicreplace- 1963). ment of "guild" members,was not present.The To determine if the pattern of range gaps competition model predicts an increase in was nonrandom, the expected distribution of "lockouts" with increasingspecies richness of gaps in blocks B-N was calculated(by N. Go- the community (Gilpin and Case1976). As not- reill) for each patchy species(n = 25). All pos- ed previously, the number of gapsper transect sibledistributions of gapswere enumeratedand block actuallydecreased with increasingspecies assumedto be equiprobable:(1) the expected richness. values over all blocksare equal to the number I concludethat the most likely factor influ- of blocks in the gap(s), (2) range-gap blocks encing the locationof gapsis the historicalal- must be contiguous,and (3) multiple gaps for teration of montane forest habitats during the speciescannot overlap. The expected number Pleistocene. The nonrandom distribution of over all specieswas summed to a cumulative gapsalong a uniform or imperceptibly grading expected number for each block (Table 9). The environment supportsa vicariant mode of iso- differencebetween the observedand expected lation and differentiation. Range gapsprobably distributionwas marginally significant(0.05 < were producedby the elimination of interme- P < 0.10). The immigration-extinction model diate populations by climatic shifts (Graves predictsa uniform distributionof range gaps 1982). along the easternAndean slope, proportional to regional speciesrichness. That is, the num- DISCUSSION ber of speciesabsent from a transectblock, but having populationsnorth and southof it in the Terborgh and Weske (1975) concludedthat study area, should be positively correlatedto direct and diffuse competition accounts for the number of speciesrecorded from the block. more than two-thirds of the distributional lira- 572 GARYR. GR•VF.S [Auk, Vol. 102 its of Andean birds whose elevational ranges acquiredreproductive isolating mechanisms in terminate between the lowlands and timber- allopatry may be lessable to reinvade the geo- line. Most evidence of direct and diffuse com- graphicrange of their sisterspecies because of petition is anecdotal in nature or derived from ecologicalfactors. As evidence of this, geo- "natural experiments."Many examplesof non- graphicallyreplacing allospecies (plumage level overlapping congeneric species cited as evi- V) are mainly high-elevation taxa (e.g. Metal- dence of competition on the Vilcabambatran- lura,Schizoeaca, Diglossa). The increasednumber sect(e.g. Coeligena,Thripadectes, Grailaria, Cacicus) of adaptive peaksat lower elevationsmay fa- overlap at some other Peruvian Andean loca- cilitate sympatry in newly formed species. tions. Lessfrequently mentioned are the doz- An alternate but related idea is that high- ensof casesin which congenersoverlap on the elevation populationsmay not be able to resist Andean elevational gradient. extinction from vicariant events long enough Nonetheless,competition remains one plau- to acquire reproductive isolating mechanisms sible factor in limiting elevational distribu- and ecological adaptationssufficient for sym- tions. If diffuse competition is a function of patry. Thus,higher extinctionrates and limited species richness, dispersing individuals may ecologicalopportunities may work in concert face increased competition when they move to depressspecies richness at high elevations. downslopeand passbelow a barrier extending Clearly, more autecologicaldata are needed downward from timberline. This would be es- before the causal mechanismsof bird specia- peciallyimportant if the colonizationtook more tion in the Andes can be understood. Here I than one generationto complete(e.g. breeding emphasizea widespreadpattern of geographic and molt may be difficult or impossible).Con- variation. I tentatively concludethat, although versely, low-elevation barriers could be cir- higher-elevationbirds at present may be spe- cumvented by speciespassing through contin- ciating more rapidly than low-elevation taxa, uous forest either above or below the barrier. the accumulation of speciesis limited by ex- Consequently,a low-elevation barrier may be tinction through disturbanceand the inability less effective in preventing dispersal than of newly formed speciesto reinvade the range high-elevationbarriers. If diffuse competition of sister species.Development of this model is operativealong the gradient, wandering in- must await additional information on climate, dividuals of low-elevation speciesshould be physiography,and the history of major forest encountered above their customary breeding disturbancealong the Andean elevational gra- elevation more often than vice versa. Data are dient. I predict that similar patterns of geo- not yet available to test this hypothesis. Pre- graphicvariation will be found in other terres- dation would not seem to be a relevant factor trial animals (e.g. small mammals, frogs, as diurnal bird-eating predatorsare presentat nonvolant insects) that are elevationally re- all elevations. stricted in humid Andean forest. ModeL--The concordant results from a vari- ety of statisticaltests indicate that plumage dif- ACKNOWLEDGMENTS ferentiation and ongoing speciationphenom- I extend my sincerestgratitude and appreciation ena are positively correlated with mean to the following individuals and institutions. My elevation and strongly suggestthat barriersto committee, F. C. James, D. Meeter, D. Simberloff, and gene flow are more effective or more prevalent J. Travis, offered advice and cogent criticism. J. V. at higher elevations in the Andes. This sup- Reinsen, Jr. carefully reviewed the data, and J.P. ports the first two predictionsof the proposed O'Neill, T. A. Parker III, K. C. Parkes, T. S. Schulen- model. The lower productivity, smaller re- berg, and L. L. Short answered inquiries about cer- sourcebase, and decreasingstature of vegeta- tain species.J. W. Eley, J.P. O'Neill, T. A. Parker III, tion at higher elevations probably limit the M. B. Robbins,and T. S. Schulenbergprovided access number of adaptivepeaks and thus the number to unpublishedfield notes.D. Meeter and N. Gotelli offered statistical advice. J. Del Rowe, A. and H. Koe- of coexisting species.Adaptation to a narrow nig, J.P. O'Neill, M. Plenge, M. B. and K. Robbins, climatic regime, habitat selection,strong terri- and G. del Solar offered hospitality and support in toriality, interspecificcompetition, and a psy- Peru, New York, and Baton Rouge. F. C. James, D. chologicalreluctance to crossalien habitatsmay Meeter, J.P. O'Neill, T. A. Parker III, K. C. Parkes, J. prevent somespecies from crossingeven small V. Reinsen,Jr., T. S. Schulenberg, D. Simberloff, and barriers. High-elevation populationsthat have J. Travis reviewed the manuscript and offered valu- July 1985] Speciationin AndeanBirds 573 able criticism.B. Felgenhauerassisted with the fig- published PhßD.dissertation, Tallahassee, Flori- ures. Field researchwas supportedby the Louisiana da State Univ. State University Museum of Zoology with the per- HAFFER,J. 1969. Speciation in Amazonian forest missionof the DireccionGeneral Forestal y de Flora, birds. Science 165: 131-137. Ministeriode Agricultura,Peru. Museum studies were 1974. Avian speciation in tropical South supportedby a grant from the Frank M. Chapman America. Publ. Nuttall Ornithol. Club No. 14. Memorial Fund of the American Museum of Natural HOLLANDERßM., & D. A. WOLFE. 1973. Nonpara- History and the Florida StateUniversity Department metric statisticalmethods. New York, Wiley. of BiologicalScience and ComputingCenter. The cu- HUXLEY,J.S. 1939. Clines: an auxiliary method in rators of the LouisianaState University Museum of taxonomy.Bijdr. Dierk. 27: 491-520. Zoology, American Museum of Natural History, and ][CHIKAWA,H., K. HUKAMI, S. TANABE,& G. KAWAKA- Field Museum of Natural History provided unlimit- MI. 1978. Standard pseudoisochromaticplates. ed access to collections. Tokyo, Igaku-ShoinLtd. This work would not have been possiblewithout MAYR,E. 1942. Systematicsand the origin of species. the love and supportof Sidney Graves. New York, Columbia Univ. Press. 1963. Animal speciesand evolution. Cam- bridge, Massachusetts,Harvard Univ. Press. LITERATURE CITED MORONY,J. j., JR.,W. J. BocK, & J. FARRAND,JR. 1975. Reference list of the birds of the world. New BAKER,g. g., & G. A. PARKERß 1979. The evolution York, Amer. Mus. Nat. Hist. of bird coloration.Phil. Trans. Royal Soc. Lon- don B 287: 63-130. MOYNIHAN,M. 1963. Inter-specific relations be- tween some Andean birds. Ibis 105: 327-339. BURTT,E. H., JR. 1981. The adaptivenessof animal colors. BioScience 31: 723-729. 1968. Social mimicry: character conver- CHAPMAN, F.M. 1917. The distribution of bird-life genceversus character displacement. Evolution 22: 315-331. in Colombia:a contributionto a biologicalsur- OFICINA INIATIONAL DE EVALUACION DE RECURSOS vey of South America. Bull. Amer. Mus. Nat. Hist. 36: 1-729. NATURALES.1976. Mapa eco16gicodel Peru:guia 1921. The distribution of birdlife in the explicative.Lima, ONERN. Urubamba valley of Peru. Bull. U.S. Natl. Mus. O'NEILL,J.P., & G. R. GRAVES.1977. A new genus 117: 1-138. and speciesof owl (Aves:Strigidae) from Peru. Auk 94.' 409-416. 1926. The distribution of bird-life in Ecua- PARKER,T. A., III, S. A. PARKERß& M. A. PLENGE. 1982. dor: a contributionto a study of the origin of An annotated checklist of Peruvian birds. Vet- Andean bird-life. Bull. Amer. Mus. Nat. Hist. 55: 1-784. million, South Dakota, Buteo Books. PETERS,J.L. 1979. Check-list of birds of the worldß CODY,M. L. 1974. Competition and the structureof vol. 8 (M. E. Traylot,Ed.). Cambridge, Massachu- bird communities. Princetonß New Jersey, Princeton Univ. Press. setts,Mus. Comp. Zool. CONOVER,W. J., & R. L. IMAN. 1981. Rank transfor- PIELOU,E.C. 1979. Biogeography.New YorkßWiley. RIDGWAY, R. 1912. Color standards and color no- mations as a bridge between parametric and nonparametric statistics. Amer. Statistician 35: menclature.Washington, D.C., published by the author. 124-129. DIAMONDßJ.M. 1972. Avifauna of the easternhigh- ROHWER, S. 1982. The evolution of reliable and un- lands of New Guinea. Publ. Nuttall Ornithol. reliablebadges of fighting ability. Amer. Zool. Club No. 12. 22: 531-546. ß 1981. Why are many tropical bird species SIMPSON,B., & J. HAFFER.1978. Speciationpatterns distributed patchily with respect to available in the Amazonian forest biota. Ann. Rev. Ecol. habitat?Proc. 17th Intern. Ornithol.Congr.: 968- Syst.9: 497-518. 973. SOKAL,R. R., & R. J. ROHLF. 1969. Biometry. The ENDLER,J. A. 1977. Geographic variation, specia- principal and practiceof statisticsin biological tion, and clines. PrincetonßNew Jersey,Prince- research. San Francisco, W. H. Freeman and Co. ton Univ. Press. TERBORGH,J. 1971. Distribution on environmental GILPIN,M. E., & T. J. CASE. 1976. Multiple domains gradients:theory and a preliminary interpreta- of attraction in competition communities. Na- tion of distributionalpatterns in the avifauna of ture 261: 40-42. the CordilleraVilcabamba, Peru. Ecology52: 23- GRAVES,G. R. 1982. Speciation in the Carbonated 40. Flower-Piercer(Diglossa carbonaria) complex of the 1977. Bird speciesdiversity on an Andean Andes. Condor 84: 1-14. elevationalgradient. Ecology 58: 1007-1019. ß 1983. Elevationalcorrelates of intraspecific ß& T. R. DUDLEY.1973. Biologicalexploration variation in plumage in Andean forestbirds. Un- of the northernCordillera Vilcabamba, Peru. Pp. 574 GARYR. GRAVES [Auk, Vol. 102 225-264 in Natl. Geogr. Soc. Res. Repts., 1966 birds: a progressreport. Acta 4th Congr. Latin projects. Zool. Vol. i: 239-255. ß& J. $. WESKE. 1969. Colonization of second- 1980. Ecologicalaspects of speciation in ary habitatsby Peruvian birds. Ecology50: 765- birdsßwith special reference to South American 782. birds. Pp. 101-148 in Ecologyand geneticsof an- imal speciation(O. A. Reig, Ed.). CaracasßVene- ß& . 1975. The role of competition in zuelaß Univ. Simon Bolivar. the distributionof Andeanbirds. Ecology 56: 562- WES•CE,J. S. 1972. The distribution of the avifauna 576. in the ApurimacValley of Peru with respectto TosI, J. A. 1960. Zonas de vida natural en el Peru: environmental gradients, habitat, and related memoria explicativasobre el mapa eco16gicodel species. Unpublished Ph.D. dissertationßNor- Peru. Instituto Interamericanode CienciasAgri- man, Univ. Oklahoma. colasde la OEA, Zona Andina, Proyecto39, Pro- ß & J. W. TERBORGH. 1981. Otus marshalii, a gramade Cooperaci6nT•chnica, Boletin Techni- new speciesof screech-owl from Peru. Auk 98: co No. 5. WashingtonßD.C.ß Organizaci6n de 1-7. Estados Americanos. ZIMMER, J.T. 1931-1955. Studies of Peruvian birdsß VU1LLEUMIER,F. 1972. Speciationin SouthAmerican Nos. 1-66. Amer. Mus. Novitates. APPENDIX. (A) Mean elevation;(B) vertical amplitude; (C) geographicrange (number of transectblocks); (D) range continuity [0 = continuous(unpatchy); i = discontinuous(patchy), one range gap; 2 = patchy, two range gaps];(E) 0 = sexuallymonochromaticß 1 = dichromatic;(F) level of plumagedifferentiation (I = no variationßII = smooth clinal, III= step clinal, IV = discretevariationß V = allospecies). A B C D E F Tinamidae (tinamous) Nothocercusjulius 2,987 725 7 0 0 II N. nigrocapillus 2ß500 1,000 15 0 0 II Accipitridae (hawks and eagles) Accipiterstriatus 2,330 2ß040 15 0 0 I Buteo leucorrhous 2,235 1ß170 11 0 0 I Oroaetus isidori 2ß630 1ß780 15 0 0 I Cracidae (guans and curassows) Penelopemontagnii 2ß680 1,380 15 0 0 IV Aburria aburri 1ß442 635 13 0 0 I Chamaepetesgoudotii 2ß262 1,475 8 i 0 IV Phasianidae(partridges and quail) Odontophorusballiviani 2ß470 1ß060 6 0 I I Columbidae(pigeons and doves) Columbafasciata 2ß375 2ß350 15 0 0 I Claravis mondetoura 1ß790 800 15 0 i I Geotrygonfrenata 2ß025 1ß750 15 0 0 I Psittacidae(parrots) Leptosittacabranickii 2ß875 950 12 0 0 I Bolborhynchuslineola 2ß480 1ß640 i0 0 0 I Hapalopsittacamelanotis 2ß535 1ß130 6 0 0 I Pionus sordidus 1,785 1,320 15 0 0 I P. tumultuosus(supersp.) 2ß735 1ß270 15 0 0 V Amazona mercenaria 2ß322 2ß395 15 0 0 I Strigidae (owls) Otusingens 1ß722 1,005 15 0 0 I O. marshalii 1ß985 510 4 0 0 I O. albogularis 2ß950 1ß140 15 0 0 I Glaucidiumjardinii 2ß825 1ß450 15 0 0 I Xenoglauxloweryi 2ß167 555 2 0 0 I Ciccabaalbitarsus 2ß262 1ß475 15 0 0 I July 1985] Speciationin AndeanBirds 575 APPENOIX. Continued. A B C D E F Caprimulgidae (nightjars) Uropsalissegmentata 2,810 1,120 15 0 1 I U. lyra 1,465 730 15 0 1 I Trochilidae (hummingbirds) Doryferaludoviciae 1,865 1,930 15 0 1 I Phaethornisguy 1,172 975 15 0 0 I P. syrmatophorus 1,835 330 2 0 0 I Cohbri thalassinus 2,262 1,475 15 0 0 I Adelomyiamelanogenys 2,462 2,325 15 0 0 IV Heliodoxa rubinoides 2,182 885 12 0 0 I H. leadbeateri 1,577 1,295 15 0 1 II H. branickii 1,160 820 i1 0 1 I Lafresnayalafresnayi 2,865 970 9 0 1 II Pterophanescyanopterus 3,267 665 15 0 1 I Coeligenacoeligena 1,877 1,495 15 0 0 II C. torquata 2,440 1,320 15 0 i IV C. violifer 2,925 1,350 15 0 1 IV Ensiferaensifera 2,922 1,355 15 0 1 I Boissonneauamatthewsii 2,245 1,170 13 0 0 I Heliangelusamethysticollis 2,810 1,420 15 0 i II Eriocnemisluciani 3,215 770 11 i i IV E. alinae 2,257 735 7 0 0 I Haplophaediaaureliae 1,755 890 6 2 0 IV Ocreatus underwoodii 1,875 1,500 15 0 i III Ramphomicronmicrorhynchum 2,975 750 13 0 i I Metallurraaeneocauda (supersp.) 3,267 665 15 0 i V M. tyrianthina 3,022 1,155 15 0 1 I Chalcostigmaruficeps 2,675 1,350 15 0 1 I Opisthoproraeuryptera 3,142 415 4 0 0 I Aglaiocercuskingi 1,875 1,500 15 0 1 I Schistesgeoffroyi 1,675 1,100 15 0 i III Loddigesiamirabilis 2,345 490 3 0 1 I Acestrura mulsant 2,162 925 15 0 1 I Trogonidae(trogons) PharomachrusantisJanus 1,875 1,500 15 0 1 I P. auriceps 1,827 1,595 15 0 1 I Trogonpersonatus 2,438 1,825 15 0 1 III Momotidae (motmots) Momotusaequatorialis 1,707 855 15 0 0 I Bucconidae(puffbirds) Malacoptilafulvogularis 1,585 610 15 0 0 I Capitonidae (barbets) Eubuccoversicolor 1,575 850 15 0 1 IV Ramphastidae (toucans) Aulacorhynchusderbianus 1,350 1,600 15 0 0 I A. coeruleicinctus(supersp.) 2,142 965 10 1 0 V Andigenahypoglauca (supersp.) 2,750 1,200 13 1 0 V Ramphastosambiguus 1,325 400 7 0 0 I Picidae (woodpeckers) Picumnusdorbygnianus (supersp.) 1,630 1,010 13 1 1 V Piculus rivolii 2,445 1,110 13 1 1 IV Veniliornisdignus 1,807 655 i0 0 1 I V. nigriceps 2,880 1,280 15 0 1 IV Phloeoceastespollens 2,897 905 7 0 1 II P. haematogaster 1,637 1,025 10 0 1 I 576 GARYR. GRAVES [Auk, Vol. 102 APPENDIX. Continued. A B C D E F Dendrocolaptidae(woodcreepers) Dendrocinclatyrannina 1,877 1,135 12 0 0 II Xiphorhynchustriangularis 1,675 1,100 15 0 0 III Lepidocolaptesaffinis 2,150 1,250 15 0 0 I Campylorhamphuspucherani 2,532 805 13 0 0 I C. pusillus 1,902 465 2 0 0 I Furnariidae (ovenbirds) Synallaxisunirufa 2,620 1,460 10 0 0 I Hellmayreagularis • 3,063 575 7 0 0 I Cranioleuca curtata 1,565 875 15 0 0 II C. marcapatae 2,840 920 2 0 0 I C. albiceps 3,020 760 2 1 0 IV Schizoeacafuliginosa (supersp.) 3,070 960 15 0 0 V Thripophagaberlepschi 2,897 905 4 0 0 I Margarornissquamiger 2,800 1,600 15 0 0 IV Premnornisguttuligera 1,880 1,140 15 0 0 I Premnoplexbrunnescens 1,677 1,105 15 0 0 III Pseudocolaptesboissonneautii 2,462 1,875 15 0 0 II Syndactylarufosuperciliata 2,150 1,700 15 0 0 I S. subalaris 1,612 975 10 0 0 I Anabacerthia striaticollis 1,647 1,035 15 0 0 I Thripadectesscrutator 2,810 1,420 15 0 0 II T. holostictus 2,147 955 15 0 0 II T. melanorhynchus 1,432 615 13 0 0 I Lochmias nematura 1,542 1,715 15 0 0 I Formicariidae (antbirds) Thamnophilusunicolor 2,015 550 6 0 1 I t. caerule$cen$ 1,767 915 15 0 1 II M•rmotherula $chisticolor 1,647 1,045 13 0 1 I Herp$ilochmu$pileatu$ 1,680 740 6 0 1 I H. axillari$ 1,432 615 15 0 1 I Drymophilacaudata 1,917 1,055 15 0 1 I Terenuracallinota (supersp.) 1,563 875 9 1 1 V Pyriglenaleuconota 1,525 1,150 15 0 1 IV Chamaezacampanisona 1,230 860 10 0 0 II Formicariusrufipectus 1A70 540 13 0 0 I Grailariasquamigera 2,590 1,520 15 0 0 III G. guatimalensis 1,072 775 15 0 0 I G. carrikeri 2,502 245 4 0 0 I G. albigula 1,862 675 3 0 0 I G. hypoleuca(supersp.) 2,135 1,220 10 1 0 V G. rufula 3,125 950 15 0 0 II G. sp. n. 2,302 335 7 0 0 I G. quitends 3,300 100 2 0 0 I Grallariculafiavirostris 1,657 1,065 15 0 0 II G. ferrugineipectus 2,675 1,350 10 0 0 I G. ochraceifrons 1,945 110 2 0 1 I Conopophagacastaneiceps (supersp.) 1,565 870 13 1 1 V Rhinocryptidae(tapaculos) Myornis senilis 2,787 1,125 4 0 0 I Scytalopusmacropus 2,962 675 3 0 0 I Cotingidae (cotingas) Ampelionrufaxilla 2,112 225 8 1 0 I Doliornis sclateri 3,063 575 5 0 0 I Pipreolariefferii 2,147 955 6 0 1 III P. intermedia 2A65 930 12 0 1 II P. arcuata 2,872 1,295 15 0 1 II P. pulchra 1,845 640 9 0 1 I July1985] Speciationin Andean Birds 577 APPENDIX. Continued. A B C D E F P. frontalis 1,400 1,200 12 1 1 IV Ampelioidestschudii 1,752 1,255 15 0 1 I Lipauguscryptolophus 1,750 900 8 0 0 I Pyroderusscutatus 1,952 855 7 0 0 II Rupicolaperuviana 1,340 1,320 15 0 1 III Pachyramphusversicolor b 1,938 1,675 15 0 1 I Pipridae (manakins) Chloropipounicolor 1,813 1,375 15 0 1 II Masiuschrysopterus 1,432 615 2 0 1 I Pipracoeruleocapilla (supersp.) 1,400 1,000 14 0 1 V Tyrannidae (flycatchers) Phyllomyiassclateri 1,762 475 5 0 0 I P. plumbeiceps 1,840 320 11 0 0 I P. nigrocapillus Z590 1,520 10 0 0 I P. cinereiceps 1,722 1,195 15 0 0 I P. uropygialis Z787 1,925 12 0 0 I Zimmerius bolivianus Z070 1,520 10 0 0 I Z. viridifiavus 1,785 1,320 7 0 0 III Mecocerculuspoecilocercus Z395 1,310 12 0 0 I M. hellmayri 1,500 300 2 0 0 I M. calopterus 1,432 615 4 0 0 I M. minor Z152 945 5 0 0 I M. stictopterus •022 1,155 15 0 0 I Anairetesagraphia 3,095 910 12 0 0 II Mionectes striaticollis 1,985 Z770 15 0 0 II M. olivaceus 670 920 15 0 0 I Leptopogontaczanowskii 2,188 905 12 0 0 I L. superciliaris 1,342 1,315 15 0 0 III Phylloscartespoecilotis 1,805 490 15 0 0 I P. ophthalmicus 1,563 875 12 1 0 IV P. gualaquizae 1,800 400 2 0 0 I P. orbitalis 800 400 15 0 0 I P. ventralis 1,675 1,100 15 0 0 I Pseudotriccuspelzelni (supersp.) 1,662 1,125 12 1 0 V P. ruficeps 2,630 1,480 15 0 0 II Lophotriccuspileatus 1,460 1,180 15 0 0 IV Hemitriccusgranadensis Z500 1,000 13 1 0 IV H. rufigularis 1,050 600 10 0 0 I H. cinnamomeipectus 1,870 260 2 0 0 I Rhynchocyclusfulvipectus 1,550 900 15 0 0 I Platyrinchusmystaceus 1,563 875 15 0 0 III P. fiavigularis 1,630 220 10 0 0 I Myiotriccusornatus 1,397 1,205 15 0 0 I Myiobiusvillosus 1,397 1,205 15 0 0 I Myiophobusfiavicans 1,945 750 10 0 0 I M. phoenicomitra 1,313 375 2 0 0 I M. inornatus 1,710 580 9 0 0 I M. roraimae 1,425 950 15 0 0 I M. pulcher 1,900 600 3 0 0 I M. ochraceiventris 2,917 865 15 0 0 I Pyrrhomyiascinnamomea Z237 2,225 15 0 0 II Mitrephanesolivaceus 1,740 520 15 0 0 I Contopusfumigatus 1,875 1,500 15 0 0 III Ochthoeca cinnamomeiventris 2,675 1,350 15 0 0 II O. frontalis 2,982 1,075 15 0 0 IV O. pulchella 2,395 850 15 0 0 IV O. rufipectoralis 3,085 930 15 0 0 IV Myiotheretesfumigatus 2,960 1,120 12 0 0 I M. fuscorufus 2,960 1,120 6 1 0 IV 578 G^R¾R. GRAVES [Auk,Vol. 102 APPENDIX. Continued. A B C D E F Knipolegussignatus 2,215 470 7 0 1 I Myiarchuscephalotes 1,675 1,100 15 0 0 I Conopiascinchoneti 1,205 910 11 0 0 I Myiodynasteschrysocephalus 1,465 1,070 15 0 0 I Hirundinidae (swallows) Notiochelidonfiavipes 2,982 1,075 15 0 0 I Corvidae (jays) Cyanolycaviridicyana 2,637 1,125 15 0 0 IV Troglodytidae(wrens) Odontorchilusbranickii 1,563 875 15 0 0 II Cinnycerthiaperuana 2,540 1,660 15 0 0 IV Thryothoruseuophrys (supersp.) 2,540 1,420 6 1 0 V Troglodytessolstitialis 2,530 1,640 15 0 0 III Henicorhinaleucophrys 2,112 1,975 15 0 0 I H. leucoptera 2,045 310 4 0 0 I Cyphorhinusthoracicus 1,875 1,500 15 0 0 II Turdidae (thrushes) Myadestesralloides 2,012 1,225 15 0 0 I M. leucogenys 1,200 400 3 0 0 I Entomodestesleucotis 2,135 2,470 12 0 0 I Catharusfuscater 2,465 1,570 15 0 0 II C. dryas 1,563 875 15 0 0 I Platycichlaleucops 1,900 2,000 15 0 1 I Turdusserranus 2,210 2,620 15 0 1 I T. nigriceps 1,367 1,535 15 0 1 I Vireonidae (vireos) Vireogilvus 1,752 1,255 15 0 0 I Icteridae (oropendolasand caciques) Psarocoliusatrovirens 1,715 870 11 0 0 I Cacicusuropygialis 1,625 710 10 0 0 I C. leucorhamphus(supersp.) 2,580 1,540 15 0 0 V Amblycercusholosericeus 2,810 1,420 15 0 0 I Parulidae (wood warblers) Myioborusminiatus 1,537 1,275 15 0 0 I M. melanocephalus 2,670 1,360 15 0 0 III Basileuterusluteoviridis 2,810 1,420 15 0 0 IV B. signatus 2,565 1,470 11 0 0 II B. tristriatus 1,567 885 15 0 0 IV B. coronatus 2,005 1,390 15 0 0 II Conirostrum sitticolor 3,022 1,155 15 0 0 II C. albifrons 2,075 1,100 15 0 1 II Emberizidae (tanagersand finches) Diglossacaerulescens 1,967 1,315 15 0 0 I D. lafresnayii(supersp.) 3,050 1,100 15 0 0 V D. albilatera 2,650 1,300 10 0 1 I D. glauca 1,862 1,475 15 0 0 I D. cyanea 2,620 1,920 15 0 0 I Iridophanespulcherrima 1,637 1,025 12 0 0 I Nephelornisoneilli 3,170 360 3 0 0 I Chlorophoniacyanea 1,300 1,400 15 0 1 I C. pyrrhophrys 2,313 625 7 0 1 I Euphoniamusica 1,870 1,140 15 0 1 I E. mesochrysa 1,200 1,200 15 0 1 I Chlorochrysacalliparaea 1,563 875 15 0 1 IV Tangarapunctata 1,392 535 15 0 0 II T. arthus 1,495 790 15 0 0 II July 1985] Speciationin AndeanBirds 579 APPENDIX. Continued. A B C D E F T. xanthocephala 1,875 1,500 15 0 0 III T. chrysotis 1,535 930 15 0 0 I T. parzudakii 1,875 1,500 11 0 0 I T. cyanotis 1,500 750 11 0 0 I T. labradorides 1,890 420 2 0 0 I T. ruficervix 1,563 875 15 0 0 III T. nigroviridis 1,877 1,495 15 0 0 I T. vassorii 2,510 1,680 15 0 0 IV T. argyrofenges 1,810 1,020 6 1 1 IV Iridosornisanalis 1,815 1,370 15 0 0 I I. jelskii 3,175 350 13 0 0 II I. ruffvertex 2,760 1,520 10 0 0 I Anisognathusigniventris 2,930 1,340 15 0 0 IV A. lacrymosus 2,760 1,520 10 0 0 II A. fiavinuchus 1,785 950 12 1 0 IV Buthraupismontana 2,810 1,420 15 0 0 II B. aureodorsalis 3,170 360 3 0 0 I Dubusia castaneoventris 2,825 1,390 12 0 0 II D. taeniata 2,760 1,180 12 0 0 II Thraupiscyanocephala 2,262 1,475 15 0 0 I Calochaetes coccineus 1,587 925 10 0 0 I Pirangaleucoptera 1,432 615 15 0 1 I P. rubriceps 2,313 625 5 0 1 I Creurgopsverticalis 1,875 1,500 7 0 1 I C. dentata 1,800 600 6 0 1 I Trichothraupismelanops 1,688 1,075 13 0 0 I Sericossyphaalbocristata 2,340 1,195 7 0 0 I Chlorospingusophthalmicus 1,920 1,440 15 0 0 IV C. fiavigularis 1,160 820 15 0 0 I C. parvirostris 2,063 1,125 15 0 0 I C. canigularis 1,392 535 13 0 0 I Cnemoscopusrubrirostris 2,620 1,460 10 0 0 I Hemispingusatropileus (supersp.) 2,835 1,070 13 1 0 V H. parodii 3,355 330 3 0 0 I H. superciliaris(supersp.) 2,725 1,250 13 1 0 V H. frontalis 1,967 1,315 12 0 0 I H. melanotis 1,952 1,345 12 1 0 IV H. rufosuperciliaris 2,955 790 5 0 0 I H. xanthophthalmus 2,950 1,140 15 0 0 I H. trifasciatus 3,185 370 11 0 0 I Chlorornisriefferii 2,675 1,350 15 0 0 II Catamblyrhynchusdiadema 2,760 1,520 15 0 0 I Saltator cinctus 2,305 1,270 5 0 0 I Haplospizarustica 2,590 1,860 15 0 1 I Atlapetesrufinucha 2,645 1,810 8 2 0 IV A. tricolor 2,075 1,100 7 0 0 I A. schistaceus 2,975 450 4 1 0 IV A. brunneinucha 1,725 1,550 15 0 0 I A. torquatus 2,685 1,370 15 0 0 IV Lysuruscastaneiceps 1,215 930 13 0 0 I Carduelisolivacea c 1,813 1,375 15 0 1 I Formerly in Synallaxis. Placedin the Tyrannidae(Traylor in Peters1979). Carduelis(Fringillidae) placedin Emberizidaefor comparison.