Phalacrocorax [Stictocarbo] Kenyoni, Which Was De- Scribed by Siegel-Causey (1991) from the Aleutian Islands Using Midden Remains and Exist- Ing Skeletal Specimens

The Auk 117(2):308-320, 2000

A CRITICALEVALUATION OF KENYON'SSHAG (PHALACROCORAX[STICTOCARBO] KENYONI)

SIEVERT ROHWER,1'3 CHRISTOPHER E. FILARDI,1 KIMBERLY S. BOSTWICK,2 AND A. TOWNSENDPETERSON2 'Burke Museum and Department of Zoology, Box 353010, University of Washington, Seattle, Washington 98195, USA; and 2Natural History Museum and Biodiversity Research Center, University of Kansas, Lawrence,Kansas 66045, USA

ABSTRACT.-We examine the validity of Phalacrocorax [Stictocarbo] kenyoni, which was described by Siegel-Causey (1991) from the Aleutian Islands using midden remains and exist- ing skeletal specimens. We emphasize a morphometric evaluation of the taxon using 224 skeletal specimens of North Pacific cormorants, but we also evaluate the qualitative characters originally used to characterize R kenyoni. Principal components and discriminant function analyses of 14 skeletal characters failed to support the validity of the species. Similarly, all seven of the character states that Seigel-Causey described as unique to R kenyoni also were found in R pelagicus and P urile. Thus, the three type specimens of R kenyoni appear to be R pelagicus. Although we could not confirm the validity of P kenyoni, our morphometric analyses revealed that R pelagicus individuals from the central Aleutians are smaller than those from surrounding populations. Received 11 December 1998, accepted 20 July 1999.

SIEGEL-CAUSEY (1991) described a new spe- kenyoni, based on detailed examinations of a cies of cormorantfrom the Aleutian Islands;he more limited reference series of specimens. named it Stictocarbo (=Phalacrocorax) kenyoni, Rohwer and Filardi were responsible for the reflecting his prior studies of major clades in morphometric analyses, and Bostwick and Pe- the Phalacrocoracidae (Siegel-Causey 1988). terson performed the qualitative analyses. Siegel-Causey first discovered this bird in midden remains from Amchitka Island in the far METHODS western Aleutian Islands but later found three recent skeletal specimens that he ascribed to Hereafter,we refer to the three North Pacific cor- kenyoni and designated as types (Siegel-Causey morants as kenyoni, pelagicus, and urile for Kenyon's 1991, Siegel-Causey et al. 1991). No other spec- Shag (P kenyoni), Pelagic Cormorant (P pelagicus), imens of this species exist. The external ap- and Red-faced Cormorant(P urile),respectively. Ta- ble 1 provides general localities and dates of collec- pearance of R kenyoni remains undescribed, tion for the specimens used in our morphometric and the species was not accepted as valid by analyses. The three kenyonispecimens were from AOU (1998). AmchitkaIsland, Alaska;the 25 urilespecimens were To facilitate an evaluation of the validity of P from Prince William Sound, or from other Alaskan kenyoni, the University of Washington Burke localities to the west and north; localities for the 196 Museum (UWBM) salvaged every intact cor- pelagicusspecimens are plotted in Figure 1. About morant recovered from the beaches of Prince one-third of the 196 pelagicusspecimens were col- William Sound, Alaska, following the Exxon lected from May to August and thus are likely to be Valdezoil spill of 24 March 1989. All were pre- from breeding areas. Although many of the speci- served as skeletal specimens, and when car- mens were not yet of breeding age, young pelagicus casses were not too rotten, associated extended are known to summer near the colonies where they hatched (Palmer1962). wings and flat skins also were preserved. Our Qualitative characters.-Because Seigel-Causey's di- analyses focused primarily on a morphometric agnosis of kenyoniwas based on seven autapomorph- evaluation of P kenyoni using 224 skeletal spec- ic skeletal characters(Siegel-Causey 1991: appendix imens of cormorantsfrom the North Pacific.We 2) and its small size, we paid special attention to the also evaluated the qualitative characters that qualitative characters of skeletal morphology de- Siegel-Causey (1991) used to characterize R scribed in his Appendix 1. Our interpretationsof Sie- gel-Causey's characterswere based on standard avian anatomicalreferences (Howard 1929, George and 3E-mail: [email protected] Berger 1966, Baumel et al. 1979, Baumel and Witmer 308

This content downloaded from 129.237.46.100 on Thu, 18 Sep 2014 14:09:14 PM All use subject to JSTOR Terms and Conditions April 2000] Evaluation of Kenyon's Shag 309

TABLE 1. Specimens examined by species, locality, and season (breeding = 1 May to 31 August). Map symbols for pelagicus refer to Figure 1.

Breeding Nonbreeding Locality Map symbol season season R kenyoni Aleutian Islands 0 3 R urile Alaska 15 10 R pelagicus Valdez and Southeast Alaska V 29 124 Aleutian Islands A 17 1 Pribilofs and northern Alaska p 0 3 British Columbia and Washington W 3 3 California C 2 6 Russia R 7 0 Japan J 1 0 Totals 74 150

1993) and on careful inspection of specimens, in- 48557, 48617) and six female (UWBM 14484, 38795, cluding one of the paratypes (UWBM 18613). For the 42012, 43071, 44195, 44402) pelagicus, and six male purpose of evaluating the validity of each character (UWBM 48619, 50610, 52084, 52090, 52091, 52094) in diagnosing kenyoni, we assembled a reference se- and six female (UWBM 50611, 50612, 52086, 52089, ries of six male (UWBM 22442, 44193, 44194, 44403, 52093, 52107) urile, the other two "Stictocarbo" shags

120? 150i 180 150? 120?

ARCTIC OCEAN LOCALITIES

V Valdez & Southeast Alaska A Aleutian Islands P Pribilofs & Northern Alaska W British Columbia & Washington C California R RussiaRUSALSK sK J JapanRUSAA

60 60 P(3) V(153) CANADA

(7) ~~~A(18) W(6) CHINA 1

40'- ~~~~~~~~~~~~~~~~~~~~~~~~U.S.A.40

C(8) JM1

PACIFIC OCEAN km 0 500 1000 20 -, , , , _ , - 20O 120? 150? 180c 150c 120O FIG. 1. North Pacific localities for the 196 pelagicus specimens used in the morphometric analyses. Symbols for the locality names are followed by sample sizes in parentheses.

This content downloaded from 129.237.46.100 on Thu, 18 Sep 2014 14:09:14 PM All use subject to JSTOR Terms and Conditions 310 ROHWER ET AL. [Auk, Vol. 117

TABLE 2. Description of the 14 skeletal charactersused in the morphometricanalyses. Terminologyused in the characterdescriptions follows Baumel and Witmer (1993).

Character Description Cranium depth Distance from center of external base of cranium to highest point of calvaria (skull cap) Cranium length From craniofacialflexion zone (hinge) to dorsal rim of foramen mag- num Maxilla length From craniofacialflexion zone (hinge) to rostrum of maxilla (tip of premaxilla) Coracoid length From acrocoracoidto lateral process of sternal end of coracoid Humerus length From head of humerus to ventral condyle at its distal end Ulna length From olecranon to ventral condyle at its dorsal end Carpometacarpuslength From carpal trochlea to distal end of major metacarpal Sternum length From dorsal spine of sternal rostrum to caudal border of sternal body at midline Sternum width Transversedistance between the first costal processes of each side Synsacrum length From corpus (body) of cranial-mostsynsacral vertebrato that of the caudal-most synsacral vertebra Pelvis width Distance between left and right antitrochantersthat form the dorsal rims of the acetabula Femurlength From tip of femoral trochanterto lateral condyle at its distal end Tibiotarsus length From patellar articularsurface of the cranial cnemial crest to medial condyle at its distal end Tarsometatarsuslength From the intercondylareminence at its proximal end to the most distal metatarsaltrochlea in the region. All urile reference specimens were USNM. After using these specimens to learn to re- from Alaska. It was important that the pelagicus ref- peat our measurements, Brian Schmidt measured the erence series included large birds taken outside the kenyoni type for us. Aleutians because kenyoni is small and has been re- Resolvingproblems with the morphometricdata.-Al- ported only from the Aleutians. For pelagicus males, though most of the Valdez birds were in good con- these specimens came from Prince William Sound, dition, some were too rotten to be sexed by gonads, Alaska (n = 2), Russia (n = 3), and Washington (n = and two were initially misidentified to species by 1); for pelagicus females, these specimens came from their preparators. Additionally, a few specimens Russia (n = 2), Washington (n = 3), and British Co- from the Valdez collection and from other museums lumbia (n = 1). The birds from Russia and Prince had broken elements that could not be measured. We William Sound were large. Each character was scored have dealt with these problems as follows. on each specimen by Bostwick and Peterson to eval- Specimens with missing characters were included uate the hypothesis that these qualitative characters in the morphometric analyses only if two or fewer diagnose the species. characters were missing. Twenty-seven of the 225 Quantitative characters.-For our morphometric specimens we measured had missing characters; 1 analyses, we used the 14 skeletal measurements de- with many missing characters was excluded, 23 had fined in Table 2. All were large enough to be taken a single missing character, and 3 had two missing accurately with dial calipers. Some of these mea- characters. We estimated values for missing charac- surements are difficult to replicate unless other ters in univariate regressions using the character workers first learn to match the measurements we most strongly correlated with the missing character made for individual specimens (available from Roh- as an estimator. All species and sexes were combined wer or Filardi). The problem lies not in the repeat- for this regression analysis because sex and species ability of our measurements, but in interpreting our determinations had not yet been evaluated. We later definitions for them (Table 2). Unless the same mea- reassessed these estimates using correlations be- surements are taken, errors between observers seri- tween characters for within-sex and within-species ously affect results. With one exception (see below), analyses; however, changes in estimates were so Filardi made all measurements. small (typically <1 mm) that reestimates were not The holotype specimen for kenyoni is housed at the used. Less than 1% of our 3,136 morphometric values United States National Museum (USNM 431164), were estimated, making the quality of our data ma- and the two paratypes are at the University of Wash- trix high. ington Burke Museum (UWBM 18613, 18614). Be- With the full matrix of morphometric measure- cause the USNM does not lend type specimens, we ments, our next step was to resolve problems in sex sent three pelagicus measured by Filardi to the determination. Several specimens were not sexed by

This content downloaded from 129.237.46.100 on Thu, 18 Sep 2014 14:09:14 PM All use subject to JSTOR Terms and Conditions April 2000] Evaluation of Kenyon's Shag 311 their preparators,and others apparently were sexed posterior probabilities refer to the probabilities of erroneously. In preparing the Valdez series and a group affiliation assigned by the analysis after the group of 23 freshly collected pelagicusfrom Wash- best discrimination had been achieved using the a ington, we learned that testes of first-yearmales are priori groupings. flattened and so little developed that, without care- We also used PCA to evaluate variation in the full fully checking for paired gonads, testes are easily matrix of 224 specimens (kenyoni, urile, pelagicus) and mistaken for ovaries (S. Rohwer and C. Filardi un- in certain subsets of this matrix. Principal compo- publ. data). Forthis reason, we felt that correctingthe nents analysis is useful because it economicallysum- sex on specimens lacking measurements of gonad marizes variationin morphologicaldata without us- size was appropriate. ing information about group membership. We ex- To assign sexes we pooled all sexed specimens (n tracted principal components from correlation ma- = 214) in a discriminant analysis. We pooled species trices to avoid weighting larger characters more for this analysis because this analysis needed to pre- heavily than small characters,as occurs with covari- cede the evaluation of species identity and because ance matrices. All of our analyses were done using our samples of urileand kenyoniwere too small to be JMP3.1 and StatView5.0 from SAS Institute. treated separately. Two of the three specimens designated as kenyoniby Siegel-Causey (1991) are unsexed, and we measured only 25 specimens of urile. RESULTS We accepted the sexes assigned to specimens by our discriminant analysis except for three University of QUALITATIVE ANALYSES Kansas Natural History Museum (KUNHM) specimens collected by Siegel-Causey in the Aleutian Is- Here we summarize the results of our in- lands, each of which had gonad measurements. Fi- spection of the seven derived characters for ken- nally, we corrected two errors of species identifica- yoni (Siegel-Causey 1991: appendix 2). We first tion in which urilespecimens had been identified as attempt to establish the synonymy of Siegel- pelagicus(UWBM 50583 and 52123). Both were Val- Causey's anatomical names with those of Bau- dez casualties, presumably encased in tar-likecrude mel and Witmer (1993), pointing out ambigui- oil, which made identification by their preparators ties. We then describe the features emphasized difficult. in our examination of variation in these quali- Morphometricanalyses.-Terminology surrounding tative characters;this represents our best at- is because distinc- discriminant analysis confusing tempt to reconstructwhat Siegel-Causey(1991) tions are often drawn between the two-group case and the case of more than two groups. In two-group examined. Finally, we report the results of our comparisons, the new linear combinationof the orig- comparison of the paratype with the reference inal variables that best separates them is called the series. Character descriptions (anatomical discriminant function. When more than two groups names in italics) are followed by a list of pos- are to be discriminated, most references shift names sible characterstates (see Siegel-Causey 1991: and apply the name canonical analysis. Thus, the appendix 1). See Seigel-Causey(1991: appendix first discriminatoris called the first canonicalaxis, or 2) for a description of how the characterstates canonical variate 1, and so on for the n - 1 variates apply to the three taxa. needed to plot the n centroids for the known groups Character1 [mandible].-"Fossaaditus: (a) ex- (see Overall and Klett 1972).Both analyses follow the tends usually 1/3 (but no more than 1/2) the same principle of maximizing the ratio of the between-group variance to the pooled within-group length of the insertion of M. pseudotemporalis; variance. As in principal components analysis (b) extends at least 3/4 the length of the inser- (PCA), the canonical axes obtained from a multiple tion" (Siegel-Causey1991). The fossa aditus ca- discriminant analysis are orthogonal to each other, nalis mandibulae (Baumel and Witmer 1993) with the first axis explaining the greatest difference presumably is the same feature as Siegel-Cau- between the groups, the second the next greatest, sey's fossa aditus. Its general location is illus- and so on. trated, but the illustrationis insufficient to per- In all of our multivariateanalyses, we used the cor- mit judging its extent with confidence.Baumel rected data matrix (missing measurementsreplaced, et al. (1979) recognized two muscles that create sexes and species assigned or corrected). We evalu- the fossa: M. ated the validity of kenyoniusing discriminant and pseudotemporalis superficialis canonical analyses. Our first approachwas a canon- and M. pseudotemporalis profundus; they not- ical analysis using urilemales and females and pela- ed that M. pseudotemporalis profundus is gicus males and females as known groups. The three sometimes synonymous with M. pseudotem- kenyonispecimens were then projected onto these poralis, but they did not illustrate either.In our axes. In our discussions of discriminant analysis, reference series, a muscle scar was found near

This content downloaded from 129.237.46.100 on Thu, 18 Sep 2014 14:09:14 PM All use subject to JSTOR Terms and Conditions 312 ROHWER ET AL. [Auk, Vol. 117 the fossa aditus, but the nature of that impres- and Witmer (1993) recognized a sulcus trans- sion varied greatly among specimens, such that versus. Siegel-Causey (1988) followed the no- comparing its extent among specimens re- menclatureof Owre (1967) and Howard (1929). quired some subjectivity. Howard illustrates the furrow clearly,but from The fossa aditus and the insertion of the M. an orientation different from that required to pseudotemporalis were evaluated as follows. see the character.We looked at the dorsolateral The fossa aditus was considered to be the sub- profile of the head of the humerus for a notch tile depression at the proximal end of the man- and found it on the paratype, although its dis- dible on the inside of the ramus. The insertion tinctness depended on the orientation from of M. pseudotemporalis was assumed to be the which the bone was viewed. We then examined long, narrow groove running underneath, and the reference series to see if the same notched often beyond, the fossa aditus. The extent of the profile existed. In all, 25%of pelagicusand 18% insertion varied among individuals, such that of urile showed the notch, although not as no- the fossa aditus extended relatively farther in ticeably as the paratype. the paratype than in most of the reference se- Character13 [femur].-"Attachmentof M. ob- ries. However, we did not find this character turatorexternus + internus:(a) elliptical, shal- state to be unique to the paratype; at least one low, indistinct; (b) deeply excavated, broad, urile and one pelagicus in the reference series subcircular; (c) deeply excavated, narrow were as extreme as the paratype. (Character 5 of Siegel-Causey and Lefevre, Character5 [humerus].-'Attachmentof M. dor- 1989)" (Siegel-Causey 1991). This illustration salis scapulae: (a) proximal most scar lateral to of this characterwas unclear and lacked labels distal scar; (b) both scars in line on bicipital in Siegel-Causey and Lefevre (1989). Baumel crest" (Siegel-Causey 1991). M. dorsalis scap- and Witmer (1993) did not recognize M. obtu- ulae is synonymous with M. scapulohumeralis ratorexternus or M. obturatorinternus, placing posterior (George and Berger 1966) and M. sca- them in the synonymy of M. obturatorius la- pulohumeralis caudalis (Baumel and Witmer teralis and M. obturatorius medialis, respec- 1993). George and Berger (1966) describe this tively. We assumed that the impressiones ob- muscle in detail. M. scapulohumeraliscaudalis turatoriae (Baumelet al. 1979) were the attach- is a single muscle that inserts onto the humerus ments to which Siegel-Causey (1991) referred. by a single tendon, but Siegel-Causey's (1991) However, illustrations and descriptions in Bau- characterdescription refers to the existence of mel et al. (1979) were insufficient to determine two scars. Furthermore,"lateral" on the bicip- to which feature of the femur Siegel-Causey ital crest is difficult to determine, given the dif- (1991) referred. Scrutiny of the impressiones ference of orientationbetween the bone and the obturatoriaein our reference series did not re- axis of the bird's body, as well as the curved veal the patterns described. The attachmentin surface of the bicipital crest. We assumed that urile was smaller than in the other species, Siegel-Causey's "scars" were equivalent to the which does not agree with Siegel-Causey's two most obvious protuberances on the bicip- (1991) general description ("deeply excavated, ital crest, and that he used "lateral" in refer- broad, subcircular").No discrete variationwas ence to the bird's body, rather than to the axis observed in the reference series; the paratype of the bone. We found that 33%of urile and 42% may be narrower in a quantitative (but not of pelagicus in the reference series exhibited the qualitative) sense. state that we understood to be derived in ken- Character14 [femur].-"Attachmentof M. flexor yoni, and that the paratype failed to show the perforatusdigiti II: (a) indistinct; (b) deeply ex- derived state. cavated without noticeable lateral bony mar- Character7 [humerus].-"Ligamentalfurrow: gins; (c) deeply excavated with robust lateral (a) does not reach head; (b) distinctly notches bony crest" (Siegel-Causey 1991). This muscle head (Character 64 of Siegel-Causey, 1988)" is synonymous with M. flexor perforanset per- (Siegel-Causey 1991). This character is not il- foratus digiti II (Baumeland Witmer1993). Nei- lustrated and is defined exactly the same way ther Baumel and Witmer (1993) nor Howard in both references. Baumel et al. (1979) did not (1929) described the bony feature attributedto recognize a "ligamental furrow," but referred this attachmenton the femur.The paratype was to a sulcus ligamentus transversus, and Baumel indeed deeply excavated with a robust lateral

This content downloaded from 129.237.46.100 on Thu, 18 Sep 2014 14:09:14 PM All use subject to JSTOR Terms and Conditions April 2000] Evaluation of Kenyon's Shag 313

crest; however, we also found this state on 50% 100 of pelagicusand 18%of urilein our referencese- ries. 801 Character16 [femur].-"Attachmentof M. flexor hallucis longis: (a) medial margin marked by 60 distinct line just adjacentto medial prominence Z

of external condyle; (b) medial margin coinci- 4-, 40 dent with medial prominence, causing the superior aspect to appear sharply produced" 20 (Siegel-Causey 1991). Attachmentof this muscle was described in detail by George and Ber- ger (1966);its proximity to the external condyle 0.0 0.2 0.4 0.6 0.8 1.0 made localization of the features described Probabilityof BeingFemale straightforward. Nevertheless, we found little variation among individuals, and no discrete FIG.2. Posteriorprobability of sex assignment, as computed from the discriminant variation among taxa. analysis using preparator'ssexing. Character 19 [tibiotarsus].-"Supratendinal bridge:(a) lateral (inferior) width greater than medial (superior); (b) widths equal" (Siegel- Causey 1991). This feature is the pons supra- sex; all had posterior probabilities of 0.806 or tendineus illustrated by Baumel and Witmer greater of being either male or female (Table 3). (1993) and was easily identifiable in our refer- Two of these 10 specimens were the unsexed ence series. The variation described by Siegel- paratypes of kenyoni (UWBM 18613, 18614), Causey (1991) was unclear: instead of having a whose probabilities of being female exceeded definable width at each extreme, the supraten- 0.999. Five specimens in the Valdez dinal bridge is double-concave in shape, mak- sample were sexed as females their ing relative widths difficult to characterize.The by preparators but were suggested to be males our discriminant paratype and almost all of the reference series by analysis. Three of these five were are best described as having the medial width specimens young birds with bursae, and all were listed as greater than the lateral, opposite to the scoring having "smooth ovaries" (Table 3). We suspect provided by Siegel-Causey (1991). The only in- these birds were young males in which the dividual that was somewhat different, a pela- right testis was not found, either because the gicus, could best be described as having widths carcass was too rotten or because the left testis equal. Thus, the limited variation we observed was mistaken for an ovary, causing the prepar- in this characterdid not correspond to Siegel- ator not to look carefully enough for a right go- Causey's (1991) characterstates, and we found nad. All had posterior probabilities of 0.941 or no discrete differences among taxa. greater of being males, so we changed their sex assignment accordingly. QUANTITATIVE ANALYSES Our discriminant analysis challenges the sex assigned to six additional specimens from Determiningsex.-The discriminant analysis KUNHM and the Museum of Vertebrate Zool- for correcting or assigning sex cleanly assigned ogy at Berkeley (MVZ). Three had neither go- sex to most of the specimens (see Fig. 2). Only nad measurements nor age data, so we changed 15 of the 224 specimens had sexes assigned their sex assignments to correspond to those with a probability less than 0.9 of being either from the discriminant analysis; posterior prob- male or female. No change in assignment of sex abilities for these corrected sexes ranged from was suggested for any of the 25 urilespecimens. 0.691 (a very marginal bird) to more than 0.99 Sexing problems occurred for 22 specimens, (Table 3). The other three specimens were col- all pelagicusand kenyoni(Table 3). Ten of these lected by Siegel-Causey. All had gonad de- specimens were not sexed by their preparators. scriptions, so we did not change their sex as- For these 10 specimens, we used the sex they signments, despite high posterior probabilities were assigned by our discriminant analysis favoring the change (0.872 to 0.998). All of when our analyses required that we know the these birds were collected in the Aleutians

This content downloaded from 129.237.46.100 on Thu, 18 Sep 2014 14:09:14 PM All use subject to JSTOR Terms and Conditions TABLE 3. Summary of changes in assignment of sex suggested by discriminant analysis in which the sex assigned by preparators identified known groups (all problems were with pelagicus or kenyoni). Posterior probabilities are those for predicted sex from the discriminant function analysis. Comments in quotes are from specimen labels.

Assigned Predicted Posterior Museum number sex sex probability Gonad/ age comments Miscellaneous sexing problems KUNHM 85947 M F 0.872 Testes (20 x 6 and 26 x 7 mm); no age data KUNHM 85951 M F 0.998 Testes (9 x 2 and 8 x 2 mm); no age data KUNHM 45963 M F 0.691 No data KUNHM 85912 F M 0.979 "Ovary damaged"; no bursa MVZ 19089 M F 0.999 No data MVZ 15184 M F 0.996 No data UWBM 17531 F M 0.968 No data o Valdez sexing problems UWBM 52003 F M 0.941 Smooth ovary (16 x 7 mm); "fleshy bursa" x UWBM 52082 F M 0.980 Smooth ovary (25 x 7 mm); "no bursa" UWBM 52088 F M 0.999 "Smooth ovary"; "deteriorated?"; "pit bursa" UWBM 52103 F M 0.982 "Smooth ovary" (12 x 7 mm); "thin-walled bursa" UWBM 52124 F M 0.990 "Smooth ovary" (25 x 12 mm); "thin-walled bursa" Specimens not sexed by preparator UWBM 18613 (kenyoni) ? F 0.999 No data UWBM 18614 (kenyoni) ? F 0.999 No data UWBM 21146 ? M 0.994 No data UWBM 48054 ? M 0.997 Salvaged specimen; "decomposed" UWBM 50574 ? M 0.999 No gonad data; "fleshy bursa" UWBM 51966 ? F 0.806 "Sexed by size, no gonads seen"; "no bursa" KUNHM 88895 ? F 0.999 No data MVZ 19088 ? M 0.989 No data MVZ 124050 ? M 0.999 No data MVZ 175977 ? F 0.995 No data D

This content downloaded from 129.237.46.100 on Thu, 18 Sep 2014 14:09:14 PM All use subject to JSTOR Terms and Conditions April 2000] Evaluation of Kenyon's Shag 315

-1.05 ; urile * E females seu -1.1 5* *

0 ? urile males -1.35 ,

0 c -1.45-1 .265 o, eaiu 5

uCpelagicus 0 Ai females -1.55 2.0~~~~~~2. . .6 28 30 . I ~~~~~~00 0 -1.657 ~O pelagicus -1.75~~~~~~~~~0 males

2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 Canonical Axis I

FIG. 3. Multiple discriminant analysis of urile and pelagicus separated by sex (four groups); each of the three kenyoni types that are projected onto the plot are denoted by an X. where pelagicusis small (see below). Small size canonical analysis we used corrected sexes (20 increases the likelihood of our discriminant assigned or changed; Table 3) and corrected analysis classifying a specimen as female. species identifications (two changed). We ex- As a check on the validity of using the dis- cluded kenyoni as a known group but projected criminant analysis to change assignment of it onto the canonical axes. Excluding kenyoni as sex, we summarized the morphological varia- a known group was essential to the validity of tion in all of our kenyoniand pelagicusspeci- this analysis because canonical analyses may mens using PCA. When we used the assign- have little generality when sample sizes are not ment changes suggested by the discriminant considerably larger than the number of char- analysis, the sexes were almost nonoverlap- acters measured (Stevens 1996), especially ping in a plot of PC I versus PC II. Furthermore, when the groups being compared are as similar only two of the birds that fell out as interme- as the sex and species classes of these cormo- diates in this PCA (and, thus, whose sex might rants. In this four-group analysis (urile and pe- be questioned) were birds whose sex was as- lagicus, males and females), the first two canon- signed using the discriminant analysis. The ical axes explained 89% of the variance in the three Siegel-Causey specimens from the Aleu- data. Although the measurements for male pe- tians were problematic:KUNHM 85912, sexed lagicus and female urile specimens overlapped as female, fell firmly in the male cluster, as the considerably, the four sex-species clusters were discriminant analysis suggested it should cleanly discriminated (Fig. 3). When projected (though its sex was not changed); KUNHM onto these axes, the three kenyonispecimens fell 85947, sexed as male, fell marginally in the with pelagicus females, suggesting that kenyoni male cluster, corroborating Siegel-Causey's may simply be pelagicus females (Fig. 3). As was sexing; and KUNHM 85951, sexed as male, fell true for the discriminant analysis used to as- at the edge of the female cluster, as the discrim- sign sexes, two Aleutian pelagicus specimens inant analysis suggested it should (although its collected by Seigel-Causey fell in the "wrong" sex was not changed). To save space, the figure sex class (Fig. 3). for this analysis is not presented. Discriminating kenyoni and pelagicus.-Here Discriminatingpelagicus and urile.-For this we present a discriminant analysis attempting

This content downloaded from 129.237.46.100 on Thu, 18 Sep 2014 14:09:14 PM All use subject to JSTOR Terms and Conditions 316 ROHWER ET AL. [Auk, Vol. 117 to separate the three kenyoni specimens from the 94 pelagicus females. This analysis was de- signed to give the validity of kenyonithe benefit 1.0 0 of the doubt. All three kenyoni specimens ap- 0 pear to be females: one was sexed as such by the preparator, and all three were assigned a sex of o 0 female with posterior probabilities greater than 0.8 8 0.999 in our discriminant analysis of sex as- 0 signment. 0 0 Because we used 14 skeletal characters, and there are only three kenyoni specimens, kenyoni a 0.6 is almost certain to fall out as correctly identi- 0 0 fied in a posterior assignment of species identity. Consequently, this result will not reliably 0 assess the validity of kenyoni. However, a good , 0.4 0 ratio of specimens to characters is available for 0 0 pelagicus females. Thus, if this analysis success- 0 0 fully divided the pelagicus females clearly into either kenyoni or pelagicus, without intermedi- 0.2 8 ates, then the specific distinctness of kenyoni would be supported. 8 0? We present the results of this discriminant analysis of female kenyoni and pelagicus by plot- 0.0 0 ting the posterior probabilities of these speci- n=85 n=9 n-3 mens being kenyoni (Fig. 4). Figure 4 divides birds into kenyoni (all from the Aleutians), pelagicus females from the Aleutians, and pelagicus females from all other localities. The pelagicus females from the Aleutians were not identified as a group to be discriminated; instead, FIG. 4. Posterior probabilitiesof being kenyoni, as they were plotted separately after the two- computed from the "two-group" discriminant anal- group discriminant analysis had separated pe- ysis of all pelagicusfemales and kenyoni.Aleutian pelagicus and kenyoni (Fig. 4). Because all three lagicus are identified on the plot but were not an a kenyoni are from Amchitka Island in the Aleu- priorigroup in the analysis. tians, we plotted pelagicus from the Aleutians separately to assess their similarity to kenyoni. Although the three kenyoni specimens are iden- urile (n = 25) and pelagicus (n = 188) is invalid. tified as kenyoni, only one was identified as such For this reason, we repeated the analysis using with an extremely high probability (0.999); the all pelagicus (n = 196) and the three kenyoni other two had posterior probabilities of 0.854 specimens, but without assigning sex. The re- and 0.710 of being kenyoni. The looseness of this sults were similar but, surprisingly, the groups cluster raises doubt about the validity of ken- were even less clearly separated because small yoni. More important, the combined sample of males from Washington, California, and the pelagicus and kenyoni females does not fall into Aleutians fell out between the centroids for the two clearly separate clusters. This result seri- two groups. ously challenges the validity of kenyonibecause Size variation in pelagicus.-An interesting our sample of female pelagicus was large (n = result of the preceding analysis is that almost 94). half of the pelagicus from the Aleutians fell clos- This analysis could be challenged by arguing er to the three specimens Siegel-Causey desig- that determining the sex of kenyoni specimens nated as kenyoni than to other pelagicus, most of using a generalized discriminant function which came from Prince William Sound, Alas- based on sexes assigned to birds by their pre- ka (Fig. 4). This suggests that pelagicus from the parators for a mixed sample of kenyoni (n = 1), Aleutians are smaller than those from sur-

This content downloaded from 129.237.46.100 on Thu, 18 Sep 2014 14:09:14 PM All use subject to JSTOR Terms and Conditions April 2000] Evaluationof Kenyon'sShag 317

TABLE 4. Summary of principal components anal- 8 ysis for pelagicusand kenyoni(n = 199). 7 Aleutianbirds kenyoni 6 0~~~~~~ PC I loadings 5 Character Males Females 4 Cranium depth 0.629 0.598 3 Cranium length 0.898 0.904 2 Maxilla length 0.813 0.888 Coracoid length 0.958 0.964 1 Humerus length 0.951 0.965 0 Ulna length 0.949 0.971 Carpometacarpus length 0.929 0.944 35 Sternum length 0.819 0.854 >, 30 - _ N. Pacificbirds Sternum width 0.844 0.858 Synsacrum length 0.835 0.846 25 Pelvis width 0.800 0.837 20 Femur length 0.921 0.942 Tibiotarsus length 0.934 0.965 15 Tarsometatarsus length 0.919 0.946 5

0 F,3 rounding populations. To further explore size variation in pelagicus, we summarized the var- 8 iation in our morphometric measurements us- 7 Southernbirds ing separate PCAs for males and females. By 6 separating the sexes into different analyses, we 5 avoided confusing within-sex size variation 4 with sex differences in the shape of the crani- 3 um. In both analyses, loadings for all charac- 2 ters on PC I were strong and positive (Table 4); furthermore, the 14 character coefficients were 0 very strongly correlated between the sexes (r = -4.0-3.0-2.0-1.0 0.0 1.0 2.0 3.0 0.974), indicating that patterns of size variation were comparable in the two sexes. Because Principal Component I Scores these coefficients were so similar (Table 4), we combined the sexes in our plots of results. FIG. 5. Principal components scores summariz- Thus, the standardized PC scores, computed ing size differences among Aleutian (including ken- separately for males and females, were treated yoni), North Pacific,and southern (BritishColumbia, as though they were the same measure of size. Washington,and California)pelagicus. By combining these independent assessments of size variation, our histograms summarizing information about geographic variation in size gicus we measured from British Columbia, are as general as we could make them (Fig. 5). Washington, and California were small and This analysis shows that most Aleutian spec- were similar in size to birds from the Aleutians imens are small, including the three designated (Fig. 5). Although we could have borrowed by Seigel-Causey (1991) as kenyoni. The distinc- specimens from more museums, there are tion between the Aleutian birds and the other probably too few breeding-season specimens Alaskan specimens, including three pelagicus to determine if the size transitions in pelagicus from the Pribilof Islands, is strong and surpris- are gradual or abrupt on the Alaska Peninsula ing. Specimens of pelagicus from the Aleutian and around the west coast of North America Islands do indeed seem to be small birds. In (Campbell et al. 1990, Johnsgard 1993). contrast, pelagicus from Prince William Sound and from the Pribilofs are large (Fig. 5). Apart DISCUSSION from Prince William Sound, we have poor samples from other regions of the west coast of Qualitative evaluations.-The characters used North America. However, most of the 14 pela- to describe new taxa should be readily acces-

This content downloaded from 129.237.46.100 on Thu, 18 Sep 2014 14:09:14 PM All use subject to JSTOR Terms and Conditions 318 ROHWER ET AL. [Auk, Vol. 117 sible to independent investigators. We encoun- into two clusters, as would be expected if they tered difficulties in identifying the exact char- were different species (Fig. 4). When we re- acters on which Siegel-Causey (1991) based his peated this analysis using all pelagicusand ken- description of kenyoni. Because anatomical yoni specimens, thus making no assumption names were used without an authoritativeref- about sex of the specimens, the separation was erence (e.g. Baumelet al. 1979) or an illustration even lower. These results suggest that kenyoniis of the features in question, most of the quali- a subset of pelagicusthat cannot be discrimi- tative charactersused to identify kenyoniwere nated morphometrically. ambiguous. Why do our morphometricanalyses generate After making the best judgments possible re- these messy results when Siegel-Causey's garding the interpretation of Siegel-Causey's (1991) canonical analysis so cleanly separated (1991) characters,we encountered furthercom- kenyoni,pelagicus, and urile?The contrastlies in plications. The discrete taxonomic differences the adequacy of samples. Siegel-Causeyinclud- described by Siegel-Causey (1991: appendix 2) ed a total of 40 specimens in his analysis (17 could not be confirmed for any of his seven di- urile, 20 pelagicus,and 3 kenyoni).However, he agnostic characters.Thus, no derived character, included the measurements of 37 morphomet- or autapomorphy,was verifiablein kenyoni.Sie- ric characters in his canonical analysis. As is gel-Causey (1991) mentioned that several char- true of multiple regression, discriminant and acters showed "modal" variation, although canonical analyses solve a set of p simultaneous most were not specified and one was referred equations (where p is the number of variables to in error (Siegel-Causey 1991:11). measured) such that the ratio of between-group Siegel-Causey (1991) stated that kenyoniis di- variance to pooled within-group variance is agnosable based on "its small size and by six maximized (Overall and Klett 1972). Because [really seven] autapomorphic skeletal charac- the number of equations in these multivariate ters." In our reanalysis, none of these seven analyses equals the number of characters,the characterswas represented by a derived state number of specimens included in a discrimi- in kenyoni.If kenyoniis a valid biological entity, nant analysis must considerably exceed the then differences in size and shape would be the number of characters.When this is not the case, only basis for its diagnosis, a basis that Siegel- the solution to these equations risks being Causey (1991) portrayed as unreliable. unique to the particular set of specimens mea- Morphometricanalyses.-Our analyses of skel- sured. Only when considerably more speci- etal morphometrics also failed to distinguish mens than charactersare measured can the re- kenyoni.When the three "kenyoni"specimens sults of such multivariateanalyses be trusted to were plotted on the two canonical axes that represent general differences between the cleanly separate urile males, urile females, pe- groups being discriminated (Stevens 1996:265). lagicus males, and pelagicusfemales, all three Given the similarity of the three cormorants fell into the cluster of pelagicusfemales. This re- Siegel-Causey (1991) was comparing, he sim- sult is consistent with our generalized sexing ply did not measure enough specimens for his function that identified the kenyonispecimens canonical axes to be generally applicable to as females. other samples representing the same groups. Assuming that the three kenyonispecimens Although our qualitative and morphometric were females, we included them in a discrimi- analyses failed to support the validity of ken- nant analysis contrasting them with all other yoni, our morphometricanalyses revealed that pelagicus females. Although these specimens pelagicusfrom the Aleutians are quite small, were successfully discriminated (an artifact of even though they are surrounded by popula- small sample size), many pelagicusfemales fell tions that have consistently been characterized into the kenyonicluster. If this analysis had as large (Palmer 1962, Hobson 1997). Indeed, yielded two discrete groups of birds, separated we found that the size differences between by a gap in morphological space, the validity of Aleutian and nearby Alaskan populations (Fig. kenyoniwould have been upheld. Instead, spec- 5) were as great as those between Alaskan pop- imens from our large sample of pelagicusfe- ulations and populations to the south, which males (n = 94) were drawn into the loose clus- are treated as P p. resplendensAudubon because ter of kenyoni.No gap divided the specimens of their small size (Palmer1962). The pattern of

This content downloaded from 129.237.46.100 on Thu, 18 Sep 2014 14:09:14 PM All use subject to JSTOR Terms and Conditions April 2000] Evaluation of Kenyon's Shag 319 variation that our results revealed for the Aleu- LITERATURECITED tians is consistent with the pattern of variation in at least one other Aleutian bird, the Rock AGER, T. A. 1983. Vegetational history of western Alaska during the Wisconsin glacial interval and Ptarmigan(Lagopus mutus gabrielsoni; Jacobsen the Holocene. Pages 75-93 in Late-Quaternary et al. 1983). Unlike the situation for southeast environments of the United States (H. E. Wright, Alaska and British Columbia (Warner et al. Jr., Ed.). Volume 1. The late Pleistocene (S. C. 1982, Rogers et al. 1991), the literature is still Porter, Ed.). University of Minnesota Press, Min- controversial concerning ice-free refugia in the neapolis. western Aleutians. Ager (1983) and Hamilton AMERICAN ORNITHOLOGISTS' UNION. 1998. Check- and Thorson (1983) leave open the possibility list of North American birds, 7th ed. American of ice-free areas in the westernmost Aleutians Ornithologists' Union, Washington, D.C. BAUMEL, J. J., A. S. KING, A. M. LUCAS, J. E. BREAZILE, when the Alaska Peninsula at peak glaciation, AND H. E. EVANS. 1979. Handbook of avian anat- and the eastern Aleutians were fully glaciated. omy: Nomina anatomica avium. Academic Our data on the small size of Aleutian pela- Press, London. gicus conflict with the pattern of geographic BAUMEL, J. J., AND L. M. WITMER. 1993. Osteologia. variation summarized in Hobson (1997:appen- Pages 45-132 in Handbook of avian anatomy: dix 2; based on Siegel-Causey's data for seven Nomina anatomica avium, 2nd ed. (J. J. Baumel, major populations of pelagicus).We cannot re- A. S. King, J. E. Breazile, H. E. Evans, and J. C. Vanden Berge, Eds.). Publications of the Nuttall this but we remain convinced solve conflict, Ornithological Club No. 23. that on the basis of skeletal characters,there is CAMPBELL, R. W., N. K. DAWE, I. MCTAGGART-COW- no reason to treat either the kenyonitype spec- AN, J. M. COOPER, G. W KAISER, AND M. C. E. imens, or the Aleutian populations of pelagicus, MCNALL. 1990. The birds of British Columbia, as a distinct species. However,both Siegel-Cau- Vol. 1. Nonpasserines: Introduction and loons sey's (1991) original description and this eval- through waterfowl. Royal British Columbia Mu- uation are based only on skeletal characters. seum, Victoria. Until the behavior, ecology, and external ap- GEORGE, J. C., AND A. J. BERGER. 1966. Avian myol- ogy. Academic Press, New York. pearance of these small Aleutian cormorants HAMILTON, T. D., AND R. M. THORSON. 1983. The are better known, we see no evidence that ken- Cordilleran ice sheet in Alaska. Pages 38-52 in yoni represents a valid taxon, and we suggest Late-Quaternary environments of the United that it be considered a synonym of Phalacrocor- States (H. E. Wright, Jr., Ed.). Volume 1. The late ax pelagicus. Pleistocene (S. C. Porter, Ed.). University of Min- nesota Press, Minneapolis. HOBSON, K. A. 1997. Pelagic Cormorant (Phalacrocor- ACKNOWLEDGMENTS ax pelagicus). In The birds of North America, no. 282. (A. Poole and F Gill, Eds.). Academy of Nat- Brian Schmidt measured the type specimen of ken- ural Sciences, Philadelphia, and American Or- yoni. Brigitte Rohwer typed the first draft of the man- nithologists' Union, Washington, D.C. uscript. Julian Baumel helped us describe our mor- HOWARD, H. 1929. The avifauna of Emeryville Shell- phometric measurements. The following people mound. University of California Publications in from the Rohwer lab commented on the manuscript: Zoology 32:301-394. Sharon Luke Michael Birks, Butler, Donahue, Sergei JACOBSEN, E. E., JR., C. M. WHITE, AND W. B. EMISON. Drovetski, and Catherine Smith. Two NSF grants 1983. Molting adaptations of Rock Ptarmigan on supported sorting, transporting, preparing, and cu- Amchitka Island, Alaska. Condor 85:420-426. rating the specimens salvaged from the Exxon Valdez JOHNSGARD, P. A. 1993. Cormorants, darters and pel- spill, and a grant from the Exxon Valdez Trustees icans of the world. Smithsonian Institution Council enabled us to complete the preparation of Press, Washington, D.C. these challenging specimens. Filardi was supported OVERALL, J. H., AND C. J. KLETT. 1972. Applied mul- as an Eddy Fellow of the Burke Museum during his tivariate analysis. McGraw-Hill, New York. work on this project. This project was based on spec- OWRE, 0. T. 1967. Adaptations for locomotion and imens at the University of Washington Burke Muse- feeding in the Anhinga and Double-crested Cor- um; the Museum of Vertebrate Zoology at the Uni- morant. Ornithological Monographs No. 6. versity of California, Berkeley; the University of Kan- PALMER, R. S. 1962. Handbook of North American sas Natural History Museum; and the United States birds,Vol. 1. Loons through flamingos. Yale Uni- National Museum. Thanks to all of these individuals versity Press, New Haven, Connecticut. and institutions! ROGERS, R. A., L. A. ROGERS, R. S. HOFFMANN, AND L.

This content downloaded from 129.237.46.100 on Thu, 18 Sep 2014 14:09:14 PM All use subject to JSTOR Terms and Conditions 320 ROHWER ET AL. [Auk, Vol. 117

D. MARTIN. 1991. Native Americanbiological di- SIEGEL-CAUSEY,D., C. LEFEVRE,AND A. B. SAVINET- versity and the biogeographicinfluence of ice age SKII. 1991. Historical diversity of cormorants refugia. Journalof Biogeography18:623-630. and shags from Amchitka Island, Alaska. Con- SIEGEL-CAUSEY,D. 1988. Phylogeny of the Phalacro- dor 93:840-852. coracidae.Condor 90:885-905. STEVENS,J. 1996. Applied multivariate statistics for SIEGEL-CAUSEY,D. 1991. Systematics and biogeog- the social sciences, 3rd ed. Lawrence Erlbaum raphy of North Pacificshags, with a description Associates, Mahwah, New Jersey. of a new species. Occasional Papers of the Uni- WARNER, B. G., R. W. MATHEWES, AND J. J. CLAGUE. versity of Kansas Museum of Natural History 1982. Ice-free conditions on the Queen Charlotte 140:1-17. Islands, British Columbia, at the height of late SIEGEL-CAUSEY,D., AND C. LEFEVRE.1989. Holocene Wisconsin glaciation. Science 218:675-677. records of the AntarcticShag (Phalacrocorax[No- tocarbo] bransfieldensis) in Fuegian waters. Con- dor 91:408-415. Associate Editor: K. R Dial

This content downloaded from 129.237.46.100 on Thu, 18 Sep 2014 14:09:14 PM All use subject to JSTOR Terms and Conditions