Variation in Numbers of Scleral Ossicles and Their Phylogenetic Transformations Within the Pelecaniformes

VARIATION IN NUMBERS OF SCLERAL OSSICLES AND THEIR PHYLOGENETIC TRANSFORMATIONS WITHIN THE PELECANIFORMES

KENNETH I. WARHEIT, • DAVID A, GOOD,2 AND KEVIN DE QUEIROZ2 •Departmentof Paleontology,University of California,Berkeley, California 94720 USA, and 2Departmentof Zoologyand Museum of VertebrateZoology, University of California,Berkeley, California 94720 USA

ABSTRACT.--Weexamined scleral rings from 44 speciesof Pelecaniformesand found nonrandom variation in numbersof scleralossicles among genera, but little or no variability within genera.Phaethon, Fregata, and Pelecanusretain the primitive 15 ossiclesper ring, while the most recent common ancestorof the Sulae (Phalacrocoracidae,Anhinga, and Sulidae) is inferred to have had a derived reduction to 12 or 13 ossicles.Within the Sulidae, Sula (sensu stricto)exhibits further reduction to 10 ossicles.These patterns of ossiclereduction are congruentwith both Cracraft'shypothesis of pelecaniformrelationships (1985) and that of Sibley et al. (1988). The presenceof scleral rings in museumspecimens is significantly greater for Phaethonand Fregata,and lessfor Pelecanus,than would be expectedfrom a random distribution. We concludethat the scleralring is of potential systematicimportance, and we make recommendationsfor its preservationin museumcollections. Received 11 August1988, accepted 25 January1989.

THE scleralring (annulusossicularis sclerae) is to be a retained primitive character.Therefore, a ring of smalloverlapping platelike bones,the any explanationof the ring's function within scleral ossicles (ossiculasclerae), found within birds must also consider its widespread occur- the sclerain the cornealhemisphere of the eye rence in Vertebrata. Whatever function the betweenthe retinal margin and the conjuncti- scleralring may serve,the number of ossicles val ring (Edinger1929, Martin 1985).The func- per ring varies in a nonrandom pattern among tion of the scleralring as a whole, and the in- some avian taxa (Lemmrich 1931, Curtis and dividualossicles in particular,is poorlyknown. Miller 1938, de Queiroz and Good 1988). Two functions are most often proposedfor the The use of the scleral ring as a systematic scleralring: first, as supportand protectionof character in birds has been limited. Studies have the eye in the region where it is found; alter- been descriptivesurveys, conducted primarily natively, as an attachmentfor the ciliary mus- at the ordinal or familial levels, on ring mor- cles,specifically, the anteriorcorneal muscle (m. phology, function, and variation in ossicle cornealisanterior), which suggestsa role in cor- number and position (e.g. Lemmrich 1931,Cur- nealaccommodation (Lemmrich 1931, King and tis and Miller 1938).While conductingtwo in- McLelland 1984, Martin 1985). Curtis and Mil- dependentresearch projects, one on the scleral ler (1938)also suggested that eccentricityof the ossiclesand phylogeneticrelationships of the ring aperture may facilitate binocular vision. Hoatzin (de Queiroz and Good 1988), and the The arrangementof the ossiclesthemselves is otheron the phylogeneticrelationships within even more problematic, although Lemmrich the Sulidae (Olson and Warheit 1988, Warheit (1931)proposed that having a scleralring com- in prep.), we discoveredwhat appearsto be posedof overlappingplates permits growth of phylogeneticallyinformative patterns of vari- the ring during ontogeny.Lemmrich also pro- ation in numbersof scleralossicles among the posed that the differences in ossicle number pelecaniformgenera. among taxa may be attributed to "growth or- We summarized the numbers of scleral ossi- ganization" (Organisationmerkmal)particular to cles for extant speciesof Pelecaniformes,and taxa. Becausethe scleral ring existsin a wide examinedthese patterns in the contextof cur- variety of vertebrates(Edinger 1929) including rent hypothesesabout pelecaniform relation- nonavian dinosaurs (references in de Queiroz ships.As a result of our survey,we alsofound and Good 1988), its presencein birds appears that scleral rings are not satisfactorilymain-

383 The Auk 106:383-388. July 1989 384 WARHEIT,GOOD, AND DE QUEIROZ [Auk,Vol. 106 tained in osteologicalcollections, and are ab- acrocorax(sensu lato; see below), however, we sent even in some specimensthat are consid- found 14 specieswith a modal number of 13 ered "complete." Accordingly, we make ossicles,4 specieswith 12 ossicles,and 3 species recommendationsfor the preparation of scleral with 12 or 13 ossicles (the modal number of rings in osteologicalspecimens. ossicles could not be resolved for the latter 3 speciesbecause of small samplesizes). Similar- MATERIALS AND METHODS ly, modal numbers of ossiclesvary among the specieswithin the single sulid genus(Sula) rec- We sampled 695 skeletonsfrom 51 speciesof pele- ognizedmost recently by Nelson (1978)and the caniforms. The number of specimenssampled for a American Ornithologists'Union (A.O.U. 1983), particular specieswas determined by its availability in the collections.For specieswith a limited number but doesnot vary within eachof the three sulid of skeletons,all specimenswere examined. If the sam- generarecognized by Olson and Warheit (1988). pie size for a specieswas large (n > 20), skeletons Availabilityof scleralrings.--Availability of were chosenrandomly, but with an effort to maintain scleral rings was not distributed randomly equal numbers of males and females. among the genera (X2 = 42.8, P < 0.001; Table We recordedthe number of completerings present 1); however, an analysisof the residualsof the (0, 1, or 2) for each specimen.When ossicleoverlap Chi-square(see Haberman 1973)shows that not is extreme, some plates may be visible on only one all genera deviate significantlyfrom indepen- side of the ring. Therefore, the number of ossicles dence.The number of scleralrings presentfor that constitute a ring was determined by counting both Phaethonand Fregata are significantly plates on both the inner and outer sides.For speci- greater (X2 = 14.83,SR (standarizedresidual ad- mens with two rings, both rings were examined. In most cases (87%) the number of ossicles in the left justedby estimateof variance) = 1.76,P < 0.05; and right rings was the same, but occasionallythe X2 = 27.54, SR = 2.13, P < 0.05; respectively), number differed (see Curtis and Miller 1938 for uni- while the number of rings presentfor Pelecanus variate statisticsand discussionof constancyof scleral is significantly less (X2 = 37.28, SR = -1.78, P ossiclenumbers between right and left eyes for var- < 0.05) than expected from a random distri- ious taxa of birds). bution. The number of scleralrings present in Mean counts were recorded for those specimens the other genera was not significantly different with both left and right rings preserved.For example, from a random distribution (P > 0.05). The if the number of ossiclesin the left ring was 14 and probabilitiesassociated with the numberof rings the right ring was 15, a count of 14.5 was used for that specimen.This was donebecause using the total absent were not significant (P > 0.05) for all number of scleralrings to calculatethe mode, rather genera,but Phaethon,Fregata, and Pelecanusap- than the preferred total number of specimens,would proachedsignificance (P = 0.10), with Phaethon not only violate the assumptionof independent sam- and Fregatahaving fewer rings absent,and Pele- pies,but would alsobias the mode towardsspecimens canushaving more rings absent than expected. with two rings over those with only one ring. We summarizedthe data in two ways. First, modal DISCUSSION numbersof ossidesare reported for 44 of the 51 species surveyed(no scleralrings were availablefor 7 species). Second,the availability of ossiclesfor each species Basedon comparisonswith ratites, tinamous, was determinedby comparingthe number of speci- and nonavian dinosaurs (de Queiroz and Good mens containingossicles (either one or two) to the 1988), the primitive number of scleral ossicles total numberof specimensexamined. Specimens with for neognathbirds appears to be 14 or 15. Num- broken or disarticulatedrings were excluded. bersin this range are taxonomicallywidespread within Neognathae (Lemmrich 1931,Curtis and RESULTS Miller 1938, de Queiroz and Good 1988). Cra- craft (1985) argued that the pelecaniformsare Variationin thenumber of scleralossicles.--The monophyletic and suggestedthat the Procel- modal numbersof ossiclesper scleralring for 44 lariiformes, Sphenisciformes,and Gaviiformes speciesof pelecaniformsare listed in Table 1. are the most significant outgroups. These out- Ossiclenumbers are distributed in a phyloge- group taxapossess either 14 or 15 scleralossicles neticallyinformative pattern within and among (Lemmrich 1931,Curtis and Miller 1938).Sibley pelecaniformgenera. Within generathere is lit- et al. (1988) suggestedthat the Falconiformes tle variability in ossiclenumber. Within Phal- and Charadriiformesare the closestoutgroups July1989] PelecaniformScleral Ossicles 385

TABLE1. Total number of specimensexamined, number (and percentage)of specimenswith ossicles,and modal number of ossiclesfor speciesof Pelecaniformes.Data associatedwith each genusheading equals the totalsfor all the specieswithin that genus.

Specimens Specimens Total No. with Modal Total No. with Modal Taxon no. ossicles(%) no. Taxon no. ossicles(%) no. PHAETHONTIDAE P. harrisi 8 2 (25) 13 Phaethon 35 24(69) 15 P.magellanicus 9 5 (56) 12 P. aethereus 4 2 (50) 14-15 P.melanoleucos 13 2 (15) 12-13 P.lepturus 8 5 (63) 15 P.neglectus 8 8 (100) 13 P. rubricauda 23 17(74) 15 P. olivaceus 24 15(63) 13 P. pelagicus 20 2 (10) 12-13 FREGATIDAE P.penicillatus 21 4 (19) 13 Fregata 65 43 (66) 15 P. punctatus 7 0 (0) -- F. aquila 11 5 (45) 15 P. pygmeus 2 1 (50) 13 F. ariel 24 22 (92) 14-15 P. sulcirostris 4 2 (50) 13 F. magnificens 8 1 (13) 14 P. urile 19 7 (37) 12 F. minor 22 15 (68) 15 P. varius 9 1 (11) 12-13 PELECANIDAE ANHINGIDAE Pelecanus 88 22 (25) 15 Anhinga 36 10 (28) 12 P. conspicillatus 4 0 (0) -- A. anhinga 24 6 (25) 12 P. crispus 2 0 (0) -- A. melanogaster 1 0 (0) -- P. erythrorhynchos 19 7 (37) 15 A. novaehollandiae 3 1 (33) 12 P. occidentalis 46 12 (26) 15 A. tufa 8 3 (38) 12 P. onocrotalus 8 2 (25) 16 SULIDAE P. philippensis 2 1 (50) 14 P.rufescens 7 0 (0) -- Morus 48 20(42) 12 M. bassanus 37 14 (38) 12 PHALACROCORACIDAE M. capensis 9 5 (56) 12 Phalacrocorax 274 123 (45) 13 M. serrator 2 1 (50) 12 P. africanus 7 2 (29) 13 Sula 148 52(35) 10 P.aristotelis 1 0 (0) -- S.dactylatra 37 12(32) 10 P. atriceps 24 12(50) 13 S.nebouxii 15 7 (47) 10 P.auritus 34 20(59) 13 S.variegata 16 5 (31) 10 P.bougainvillii 14 8 (57) 12 S.leucogaster 39 14(36) 10 P. capensis 14 14(100) 13 S.sula 41 14(34) 10 P. carbo 23 8 (35) 13 P. carunculatus 2 0 (0) -- Papasula 1 1 (100) 12 P. abbotti 1 1 (100) 12 P. coronatus 9 9 (100) 13 P. gaimardi 2 1 (50) 12 Totals 695 295 (42)

to the most recent common ancestor of the taxa thon,Fregata, and Pelecanusretain the primitive included in the traditional Pelecaniformes, number of 15 ossiclesper ring, but the clade which they considerto be polyphyletic.These stemmingfrom the most recentcommon ances- outgroup taxa also have either 14 or 15 scleral tor of the sulids, cormorants, and anhingas ossicles (Curtis and Miller 1938). Therefore 14 (=Sulae Sharpe 1891) is derived in that the or 15 scleral ossiclesappears to be the primitive number of ossicles is reduced to 12 or 13. A conditionnot only for the Neognathaebut also further decrease in ossicle numbers occurs for the Pelecaniformes. within the Sulidae;Sula has only 10 ossiclesper At least two reductions in ossicle number must ring (somespecimens show as few as 8), com- have occurred within the Pelecaniformes. Fur- pared with 12 in both Morusand Papasula.The thermore, these reductionsare congruent with 10 scleralossicles per ring in Sulais the smallest the two currenthypotheses of pelecaniformre- number known for birds (see Lemmrich 1931, lationships (Figs. 1, 2). The traditional and Curtis and Miller 1938). widely acceptedhypothesis (Fig. 1) has been These proposedtransformations in numbers supportedcladistically by Cracraft(1985). Phae- of ossiclesis alsocongruent with an alternative 386 WARHEIT,GOOD, AND DE QUEIROZ [Auk,Vol. 106

Fig. 2. Phylogenetic hypothesisbased on Sibley Fig. 1. Phylogenetichypothesis based on Cracraft et al. (1988). Threskiornithidae (15 ossicles[Lemmrich (1985), exceptrelationships within the Sulidae (Pa- 1931]), Phoenicopteridae(14 ossicles[this study]), pasula,Morus, and Sula)which are based on Olsonand Scopidae(15 ossicles[this study]), and Ardeidae (14- Warheit (1988).The number associatedwith eachtax- 15 ossicles[Lemmrich 1931, Curtis and Miller 1938]) on representsthe modal number of ossicles/scleral are successiveoutgroups to the Ciconiidae-Procellar- ring for that taxon. iiformesclade and have been omittedto simplify the figure. Relationshipswithin the Sulidaeare basedon Olson and Warheir (1988), and the number associated hypothesisfor the relationshipsof the "pele- with eachtaxon representsthe modal number of os- caniforms"(Fig. 2). Sibleyet al. (1988)proposed sicles/scleralring for that taxon. that the Pelecaniformesis polyphyletic,with Fregatidaeand Pelecanidaemore closelyrelat- ed to ciconiids,sphenisciforms, gaviiforms, and (Fig. 2). Theserelationships also require at least procellariiformsthan to the Sulae(=Sulida in three ossicle transformations: either reduction Sibleyet al. [1988]).Nevertheless, because the to 12 ossicles for the Sulae, increase to 13 os- Sulaeis consideredto be monophyleticand be- sicles for Phalacrocorax,and reduction to 10 os- causeits hypothesizedoutgroups all have 14 or siclesfor Sula (Fig. 3c); or, reduction to 13 os- 15 ossiclesper ring, this hypothesisalso implies sicles for the Sulae, reduction to 12 ossicles for reductions in ossicle numbers in the most re- Anhingaplus the Sulidae, and reduction to 10 cent common ancestor of Sulae and of Sula. ossiclesfor Sula(Fig. 3d). From the perspectiveof ossicletransforma- Of these four alternative hypothesesof os- tions,an importantdistinction between the tra- sicle evolution within Sulae (Fig. 3), the hy- ditionalhypothesis and that of Sibleyet al. (1988) pothesisrepresented by Figure 3b requiresone concernsthe relationshipswithin the Sulae.Both more transformationthan thoserepresented by hypothesesrequire the samenumber of ossicle Figures3a, 3c, and 3d. It is also the only alter- transformations,yet the transformationsthem- native involving homoplasy,specifically con- selvesdiffer (Fig. 3). In the traditionalhypoth- vergent reduction from 13 to 12 ossiclesin esis(Fig. 1), Phalacrocoraxand Anhingaare sister Anhingaand the Sulidae.Although the hypoth- taxa.This hypothesisrequires at leastthree os- esized transformationsconcerning Phalacrocor- side transformations: reduction to 12 ossicles ax in Figures 3a and 3c involve increasesin for the Sulae, increase to 13 ossicles for Phala- ossicle numbers (from 12 to 13 ossicles),these crocorax,and reduction to 10 ossicles for Sula neednot be interpretedas homoplasies. Because (Fig. 3a). An alternativehypothesis of ossicle there is no evidence that the common ancestor evolutioninvolving reductionto 13 ossiclesfor of the Sulae passedthrough a 13 ossiclestage the Sulae,convergent reduction to 12 ossicles (transformation could have been from 14 or 15 in Anhingaand Sulidae,and reductionto 10 ossiclesdirectly to 12 ossicles),an increasefrom ossiclesfor Sularequires at leastfour transfor- 12 to 13 ossicles does not necessitate a character mations(Fig. 3b). In the hypothesisof Sibleyet reversal. al. (1988),Anhinga and the Sulidaeare sistertaxa Although the number of ossiclesper ring var- July1989] PelecaniformScleral Ossicles 387

Ph A Pa M S Ph A Pa M S 13 12 12 12 10 13 12 12 12 10

A ,•"-•"12 B '•,,•-•"13 13"•'12*

Ph A Pa M S Ph A Pa M S 13 12 12 12 10 13 12 12 12 10

O ,• -'•"13

Fig. 3. Relationshipswithin the Sulae,according to Cracraft(1985) (a, b) and Sibleyet al. (1988)(c, d). Relationshipswithin the Sulidaeare basedon Olsonand Warheit(1988). Enclosed circles represent proposed scleralossicle transformations (see text). Size of enclosedcircles identifies specific points on the dadograms. The smallestcircle representsthe 12 to 10 ossicletransformation associated with the genusSula; the largest circlerepresents the ancestralSulae ossicle transformation (from the outgroupcondition to either 12 or 13 ossicles).The medium-sizedcircle representsalternative transformations within the Sulae.Asterisk (*) in- dicatesconvergent transformations. S = Sula,M = Morus,Pa = Papasula,A = Anhinga,and Ph = Phalacro- coracidae.The number associatedwith each taxon representsthe modal number of ossicles/scleralring for that taxon.

ies among families and genera of pelecani- the modal number of 13 present in the other forms, it is relatively constantwithin genera. Phalacrocoracidae.Siegel-Causey (1988) placed This conservativepattern of ossiclevariability P. gaimardi,P. magellanicus,P. pelagicus,and P. within generashould not be construedas being urile in the genus Stictocarbo.As a result, the of phylogeneticsignificance because it depends intragenericvariation in ossiclenumber within on subjectivetaxonomic ranking. Although the the cormorantsis decreased.P. bougainvilliiis gannets (Morus) always have been considered hypothesizedto be convergentwith Stictocarbo "taxonomicallydistinct" from the boobies(Sula), in possessing12 ossiclesper ring. Although Sie- the two groupsare often placedin a single ge- gel-Causey(1988) did not usemodal numberof nus. In the six editions of the A.O.U. check-list, ossiclesin his analysis,their distribution within gannetshave alternatelybeen consideredeither the Phalacrocoracidaeis largely congruentwith Morusor Sula,switching three timesin 73 years the phylogeny he proposed. (editions 3-6). Based on the most recent A.O.U. Reductions in scleral ossicle number are de- check-list (1983), in which all sulid speciesare rived and diagnostic for the Sulae, Sula, and includedwithin the genusSula, there would be Stictocarbo within Pelecaniformes. Derived re- intragenericvariation in modal ossiclenumber. ductions in ossicle numbers also occur in sev- If three sulid generaare recognized(Olson and eral distantly related taxa. For example,within Warheit 1988), however, there is no variation the Charadriiformes, the Alcidae are derived in modal ossiclenumber within each genus. in having 13 scleral ossiclescompared with 15 The amount of variation within the Sulidae has ossicles for most other charadriiforms (Curtis not decreased,it is simply partitioned differ- and Miller 1938). Other arian taxa characterized ently. A similar situationexists within the Phal- by 11 or 12 scleral ossiclesare Columbiforrnes, acrocoracidae.Phalacrocorax gaimardi, P. magel- Psittaciformes, and Cuculiformes (Curtis and lanicus,P. pelagicus,P. urile, and P. bougainvillii Miller 1938, de Queiroz and Good 1988). The have 12 (or 12-13) ossiclesper ring, rather than scleralring has been investigatedin detail for 388 W•R•E•T,GOOD, AND DE QUEIROZ [Auk, Vol. 106 very few groups,however, and more work is (National Museumof Natural History) for permitting necessaryto reachan understandingof the evo- us to examine the specimensin their care. We also lution of this character within birds. thank J. J. Becker,G. R. Graves,B.C. Livezey,and S. L. Olson for their constructive comments on the From our experience,a comparativestudy of ossiclevariation at any level within the avian manuscript.C. G. Sibley and D. Siegel-Causeypro- vided uswith prepublicationcopies of their now pub- hierarchy would be difficult to accomplish,be- lished manuscripts.This projectwas funded, in part, causethe scleral rings of different taxa are not by grantsto K. I. Warheit from the Officeof Fellow- always well representedin museumcollections. ships and Grants, Smithsonian Institution, and the We found that there are more specimensof Fre- Alexander Wetmore Fund of the American Orni- gataand Phaethon,and fewer of Pelecanusin mu- thologists'Union. seumcollections that include scleralrings, than would be expectedfrom a random distribution. LITERATURE CITED Becauseof the potential systematicimportance AMERICaaq ORN•TI-IOLOG•STS' UNION. 1983. Check-list of scleral rings, we recommendthat museums increasetheir effortsto preservethese elements. of North American birds, 6th ed. Washington, D.C., Am. Ornithol. Union. If a museum uses maceration to clean skeletons, CRACRAFr,J. 1985. Monophyly and phylogeneticre- the eyesshould be removedbeforehand. Macer- lationshipsof the Pelecaniformes:a numerical ation can still be used to prepare the isolated cladisticanalysis. Auk 102: 834-853. ring, but it must be checkedfrequently to pre- CURTIS,E. L., & R. C. MILLER.1938. The scleroticring vent disarticulation.If the skeletonis prepared in North American birds. Auk 55: 225-243. by mechanicalmeans, we recommendthat the DE QUEIROZ, K., & D. A. GOOD. 1988. The scleral eyes not be removed; beetle larvae are capable ossiclesof Opisthocomusand their phylogenetic of cleaning scleral rings. Nevertheless,when significance.Auk 105: 88-106. removing the preparedskeleton from the beetle EDINGER,T. 1929. '0ber knocherne Scleralringe. Zool. Jahrb., Abt. Anat. 51: 163-226. frass,special attention must be paid to scleral HABERMAN,$. J. 1973. The analysisof residualsin rings and other small elements. The larvae cross-classified tables. Biometrics 29: 205-220. themselvesmay complicatethe processby re- K•NG, A. S., & J. McLELI,AND. 1984. Birds, their struc- movingand transportingrings. This occurs most ture and function, 2nd ed. London, Bailliere Tin- frequently with small rings when the larvae dal. becometrapped within the center of the ring, LEMMRICH,W. 1931. DerSkleralringderV6gel. Jena. wearing it as if it were a life-preserver. Zeit. Nat. 65: 513-586. MARTIN,G. R. 1985. Eye. Pp. 311-373 in Form and

ACKNOWLEDGMENTS function in birds, vol. 3 (A. S. King and J. Mc- Lelland, Eds.). London, Academic Press. We thank G. F. Barrowclough (American Museum NELSON,B. 1978. The Sulidae:gannets and boobies. of Natural History), K. L. Garrett and the late R. W. Oxford, Oxford Univ. Press. Schreiber(Natural History Museum of Los Angeles OLSON,S. L., & K. I. W•RHEIT. 1988. A new genus County), J. V. Reinsen (Louisiana State University, for Sula abbotti. Bull. Brit. Ornithol. Club 108: 9- Museum of Zoology), R. A. Paynter Jr. (Museum of 12. Comparative Zoology), N. K. Johnson(Museum of SIBLEY,C. G., J. E. AI-rLQUIST,& B. L. MONROEJR. 1988. VertebrateZoology), A.M. Rea(San Diego Societyof A classificationof the living birds of the world, Natural History, Natural History Museum), J. H. basedon DNA-DNA hybridization studies.Auk Hutchison (University of California, Museum of Pa- 105: 409-423. leontology),R. B.Payne (University of Michigan,Mu- SIEGEL-CAusEY,D. 1988. Phylogenyof the Phalacro- seum of Zoology), and S. L. Olson and R. L. Zusi coracidae. Condor 90: 885-905.