The 111(4):873-880, 1994

ONTOGENETIC EVIDENCE FOR RELATIONSHIPS WITHIN THE

W. PARKER CANE Departmentof Ornithology,American Museum of NaturalHistory, Central Park West at 79th Street, New York, New York 10024, USA

Al•sTP,ACT.--Comparative growth patternsof selectedcranial and postcranialskeletal char- actersare used to test hypothesesof phylogeneticrelationships between the major groups within the family Laridae.The developmentof the bill in skimmersand is characterized by highly nonlinearsize-dependent allometries in contrastto the relativelyconstant, linear allometricrelationship found in . Thesecomparative ontogenetic trajectories thus sup- port the hypothesisthat skimmersand terns are more closelyrelated to one anotherthan either is to the gulls. The linear size-dependentallometry found throughoutdevelopment in gulls is similarto that found early in the ontogenyof all larids.In contrast,the highly nonlinearallometry which characterizesthe later stagesof developmentin ternsand skim- merssuggests additional complexity in the geneticcontrol of bill development.The more parsimoniousphylogenetic hypothesis is that the shapeof the bill in gulls and skuasis the more primitive characterstate. The bill shapein skimmersrelative to terns may reflectthe heterochronicextension of the ontogenetictrajectory characteristic of the later stagesof developmentin terns.Ontogenetic transformations found amongpostcranial characters are consistentwith the hypothesisof relationshipsbased on bill ontogenies.However, because of similar nonlinear characteristicsin both and growth trajectories,phylogenetic inferencesdrawn from the relativeontogenies of postcranialdimensions are consideredless compelling.Received 22 January1993, accepted 27 March 1993.

A 1,4UMI•ERof hypothesesof phylogenetic re- the bill. Here I describethe ontogenetic trans- lationshipsamong the major groupsof formationsunderlying these bill-shape differ- comprisingthe Laridae(, gulls, terns,and encesand examinetheir phylogeneticimpli- skimmers)have been suggested(for historical cations. Gulls and terns also can be differentiated summary,see Sibley and Ahlquist 1990). These morphometricallyby several postcranialchar- can be summarized in three phylogenetic trees acters;consequently, the phylogeneticinfer- (Fig. 1). The first hypothesis(Fig. 1A), which ences that can be derived from differences in suggeststhat the ternsand gullsare moreclose- the underlying postcranialdevelopmental tra- ly related than either is to the skuas(Stercora- jectoriesare explored. riinae) and skimmers (),reflects the traditional classificationused by Peters (1934) and Wetmore (1960), and is supported by evi- MATERIALS AND METHODS dence from protein electrophoresis(Hackett Specimensin development series.--During the 1988, 1989)and DNA-DNA hybridization (Sibley and 1989, and 1990 breeding seasons,31 Ahlquist 1990).Behavioral analyses (Moynihan (Sternahirundo) embryos ranging in agefrom 10 days 1959;see also Burger and Gochfeld 1990) sup- of incubationto 22 days(pipping), and 58 youngfrom port a similar relationship. In contrast,a closer 1 to 57 daysposthatching, were collected or salvaged relationshipbetween the skuasand gulls and on Great Gull Island (GGI). The island, which lies at between the skimmersand terns (Figs. lB and the eastern end of Long Island sound (72ø07'W, 1C) has been suggestedby comparative para- 41ø12'N),Suffolk Co., New York, is the siteof a long- sitology(Timmermann 1957) and an analysisof term study of reproductivesuccess in the Common Tern (Haysand Riseborough1972). Each year a team skeletal morphology (Strauch 1978). checksthe island daily, numbering new nestsand Basedon a morphometricanalysis of skeletal eggs.When chickshatch they are bandedand the characters,Schnell (1970b) felt that, pheneti- nest associationrecorded. Consequently, the age of cally, skimmers showed similarities to terns, chicks later recovered is known to within 24 h. with skimmersand terns being differentiated Forother taxa, developmental specimens consisting from gullsand skuasprimarily by charactersof of 10 Herring Gull (Larusargentatus) and two Great 873 874 w. PARKERCANE [Auk, Vol. 111

A HerringGull A Skuas Skimmers Gulls Terns Common Tern

B Skuas Gulls

Terns Skimmers

C Skimmers Terns Gulls B -- Skuas HerringGull•=••=•=••

Fig. 1. Three cladogramssummarizing alternative ' ' hypothesesof phylogeneticrelationships within the Laridae. Common Tern

Black-backedGull (Larusmarinus) embryos and young were collected near or on Great Gull Island. The Black Skimmer(Rynchops niger) developmental material was obtained from six embryos and juvenile specimens housed in the anatomical collection of the American Museum of Natural History (AMNH). For freshly collected material, embryos were re- moved from the egg and, after removal of the ex- traembryonicmembrane, placed immediately into 10% BlackSkimmer'•'l buffered formalin. Larger specimenswere injected Fig. 2. Bill measurementsused. Bill shapesat about intraperitoneallywith the fixative.After 24 to 48 h time of hatching shown on left, and adults on right. (dependingon size),embryos were transferredto 70% Drawings are not to scale.(A) Upper bills with bars isopropanol.Posthatching specimens were skinned representingthe PRENAR (seeAppendix 2). (B) Man- before being placed in formalin for severaldays to dibles with SYMLEN measure indicated. one week. The few specimensthat showedabnormal osteologicaldevelopment (e.g. cranial deformitiesor abnormalnumbers of digits) have been excludedfrom Tern, one or two specimenswere measured(Appen- the analysis. dix 1). The adult Common Tern data representsthe Specimenswere clearedand stainedusing a mod- averagesof 28 male and 38 female specimens. ification of the Alcian Blue/Trypsin/Alizarin Red S Measurements.--Thebill and postcranialmeasures technique (Wasserzug1976, Dingerkus and Uhler usedare similar to thosedescribed by Schnell (1970a; 1977)in which cartilageis renderedblue and calcified see my Appendix 2 and Fig. 2). The measurements bone red. After staining, specimenswere stored in for the development series encompassboth the os- glycerinwith thymol crystalsadded to inhibit con- sifted and cartilaginousportions of all elements.All tamination. measurementswere madeusing digital calipersunder Adultspecimens.--Adult skeletal measurements were a binocular0.7 to 4.2x dissectingmicroscope equipped obtained from material housed in the anatomical col- with 10 x W.F. eyepieces.Smaller embryos were mea- lections of the AMNH, U.S. National Museum, and sured using a Bauschand Lomb 7.5 x ocular microm- CarnegieMuseum. For taxaother than the Common eter calibrated with a steel rule. October1994] ComparativeLarid Ontogenies 875

T^BLE1. Threeprincipalcomponents extracted from covariancematrix of postcranialskeletal characters ß Gull (seeAppendix 2) for adult larids(taxa in Appendix 4 o 1). [] Tern ß Skimmer Variable P1 P2 P3

COR 0.26 0.15 -0.24 SCP 0.24 0.16 -0.30 ILM 0.26 0.21 0.01 ISH 0.28 0.10 0.04 SAC 0.16 0.44 0.05 FEM 0.29 0.09 -0.07 TBT 0.32 -0.17 0.15 TMT 0.36 -0.52 0.58 TOE 0.33 -0.51 -0.67 10 ' 12' ' 1•4 16 HUM 0.29 0.07 0.09 General Size RAD 0.28 0.19 0.17 CMC 0.26 0.09 0.04 Fig. 3. Logarithmof length of mandibular sym- PRX 0.21 0.29 0.05 physisrelative to generalsize (see text) for gulls,sku- Eigenvalue 1.815 0.046 0.018 as, terns, and skimmers. Solid lines are least-squares Percent trace 95.9 2.5 1.0 fits of data for terns and gulls.

shown in Table 1 have been standardized to unit Various models for the use of size- or age-depen- length. Allometric coefficients(b) for the equation dent allometrictransformations in phylogeneticanal- ln(Y) = a + b In(general size) were obtained from ysishave been proposed(Alberch et al. 1979, Kluge bootstrapestimates of the reducedmajor axesfits of and Strauss1985, Kluge 1988).Here the analysisfo- the data using a FORTRAN program. cuseson size-dependentstatic (adult) and ontogenetic allometriesof individual elementsrelative to general RESULTS growth or size. I have defined the "general size" of an individual specimento be equal to the sum of the Schnell (1970a) found the primary variables logarithmicallytransformed values of 13 postcranial discriminating between terns and gulls to be measuresfor that specimenmultiplied by 0.277 (the squareroot of the inverseof the numberof variables). the length of the os premaxillaerostral to the The resultantvalue is then comparableto that which anterior edge of the nasal opening (PRENAR) would be derived from principal-componentsanal- and the length of the dentarysymphysis of the ysis in the caseof purely isometricgrowth. In such mandible (SYMLEN; see Appendix 2). The ex- analyses,the generalsize of a specimenis often con- tent to which these measuresvary acrossthe sidered equivalent to the score of the specimen on Laridaecan be seenin Figure 2. Relative to total the first principal component (PC1). In the caseof bill size, these measures in terns and skimmers isometric growth, PC1, when standardized to unit are quite large, while in gulls and skuasthey length, would have all variable coefficientsequal to are relatively small. Figure 3 shows the inter- p(-0.5)where p is the numberof variables(see Pimentel 1979:58+). Standarization of the componentsto unit specificallometric relationshipsbetween SYM- lengthhas the convenientmathematical property that LEN and general size among the Laridae. The the anglebetween vectors can be calculatedby simple adult larid taxa included in this analysis are vector multiplication. listed in Appendix 1. Interspecificallometry ap- Ontogenetictrajectories of individual elementsrel- pearslinear for both gulls and terns. Although ative to sucha general-sizemeasure would resemble the skimmeris quite distantfrom both of these thoserelative to other size surrogatessuch as the cube groups,this could representeither a primitive root of body mass.For the adult larids included in or derived condition, such an array of inter- my study, the correlation between this general-size specificallometries in adults thus offering little vector and body mass (data from Dunning 1992) is insight into evolutionaryrelationships. 0.994. However, becauseof the high degree of indi- vidual variation in body masses,especially in youn- The ontogenetic trajectoriesunderlying the differences in PRENAR and SYMLEN across the ger specimens,the useof body massas a size measure would serveonly to obscurerelative growth patterns. Laridaeare shown in Figure 4. Throughout em- Multivariate analyseswere carriedout using NTSYS bryonic and early posthatchingdevelopment, version 1.5 (Rohlf 1988). Note that the components gulls, terns, and skimmersshow a similar size- 876 w. PARKERCANE [Auk,Vol. 111

A o CommonTern S [] AdultTerns R ß Skimmer S S 0.7 S S S KK L LL ,< S L L C C ""0.6 S L L S S L L L L L L 16 General Size • 315 Pc1

B Fiõ. 5. Scoresof adult larid taxa on first two com- ponentsextracted from ½ovariancematrix of loõarith- micall¾transformed postcranial characters. PC! ap- proximatesõeneral size (seetext). Lettersindicate: õull, (C) skua,(K) , ($) tern, and (R) skimmer.

yond that in terns, which, if a derived condition (see Discussion),would suggest hypermor- 4 • 8 10 12 14 16 phosis(sensu Alberch et al. 1979;Fig. 4). For General Size gulls,no suchshift occurs.Rather, throughout posthatchingdevelopment the growth of the Fig. 4. (A) Developmentaltrajectories for rostral rostral end of the premaxilla (Fig. 4A) and of end of premaxilla (PRENAR) relative to general size. the dentarysymphysis (Fig. 4B) relative to over- Solid line is approximatepolynomial fit of ontoge- all growth continuesalong the sametrajectory netic trajectorydefined by 89 Common Tern speci- mens.H representsadult Common Terns.Dashed line seen early in development;change in rostral is developmentaltrajectory in Herring Gullsand Great bill form is primarily a changein size that par- Black-backedGulls with adultsof thesespecies rep- allelsgeneral growth with little or no change resentedat end of gull trajectory.Other adult gulls in shape. arrayed along this trajectory.Gulls having a general In skimmers the rostral end of the mandible size lessthan 12 reflectvalues for embryosand early projectsbeyond the rostraltip of the premaxilla. posthatchingspecimens of Herring Gulls and Great The limited datasuggest that this, in part, may Black-backedGulls. (B) Ontogeny of the mandibular be due to either relatively rapid growth very symphysis(SYMLEN) relative to general size. On- early in embryonicdevelopment or larger ini- togenetictrajectory for CommonTern representedby tial mesenchymalcondensations in the man- solid curve, which reflectspolynomial least-squares fit (r 2 = 0.85). dibular anlage.Either would resultin the pos- itive displacement in the developmental trajectoryof the mandibular symphysissug- dependentallometric growth pattern in the ros- gestedby the currentdata set (Fig. 4B). If con- tral end of the bill. However, in the Common firmed by further data, such a displacement Tern at about three weeks posthatching, al- would seem unique to skimmersand, thus, of thoughgeneral body growth is essentiallycom- little use in the elucidationof phylogeneticaf- plete, bill growth as representedby PRENAR finities. and SYMLEN, continues. Consequently, a Comparativepostcranial ontogenies.--Discrimi- marked shift occursin the allometric growth of nation amongterns, skimmers, gulls, and skuas PRENAR (Fig. 4A) and SYMLEN (Fig. 4B) rel- also can be achieved using only postcranial ative to body size. A similar shift in develop- skeletal characters. Table 1 shows the results of mental allometries is found in the skimmer, al- a principal-componentsanalysis of 13 postcra- though the final development phase in nial skeletalcharacters for 28 larid taxa (Ap- skimmers, especiallyin SYMLEN, extends be- pendix 2). The first component(PC1) is similar October1994] ComparativeLaridOntogenies 877

TABLI/2. Ratio to body massa of three postcranial 3.5 measurescomprising major sources of interspecific Black-backedGulls ß variation among larids as indicatedby principal- componentsanalysis (see Table 1). o Common Tern 2.5 ß BlackSkimmer ß

No. Oo• speciesmeas- Ratioto body mass of Taxon ured SAC TMT TOE

Terns 10 2.6 4.4 1.7 Skimmers ! 2.6 5.2 !.3 0.5 Gulls 15 2.0 6.5 2.3 Skuas 2 2.4 7.3 2.4

MassesfrOm Dunning (1992). Ratiosto cube roots of body mass. General Size

Fig. 6. Ontogenetictrajectory for developmentof to a general-sizevector that would have all co- TOE. Upward- and downward-facingarrows indicate efficientsequal to 0.277 (the angle between such time of hatchingin CommonTerns and Herring Gulls, a theoretical vector and PC1 is 10.6ø). The sec- respectively.All trajectoriesterminate with adulttaxa. ond component(PC2) contrastssynsacral width The four gulls in upper-right corner are two adult (SAC) with distal hindlimb elements (TMT and HerringGulls and two adultGreat Black-backed Gulls. TOE). Figure5 showsthe scoreof the individual taxaon thesemajor axes of postcranialvariation. I interpret PC2 to reflect the major axis of in- skimmers (data not shown). The most notice- terspecificpostcranial "shape" variation. able difference in postcranialontogenetic tra- In comparisonto gulls and skuas(for a given jectoriesis found for TOE (Fig. 6). In terns and general size), skimmersand terns have a rela- skimmers,there is an abrupt slowing of growth tively larger ratio of synsacralwidths (SAC) to after hatching.In gulls,a similarbut lessmarked distal hindlimb lengths (see Table 2). This is a slowing occursduring posthatching develop- consequenceof quite different allometric growth ment. patterns in several skeletal elements relative to overall growth. During prehatching develop- DISCUSSION ment in terns,and apparentlyalso in skimmers, the growth allometryof SAC relative to overall In the rostral development of the bill, the growth appearsto be significantly more posi- differencesin the ontogenetictrajectories seen tive than that in gulls. The prehatching allo- for terns and skimmers versus gulls (and ap- metric coefficient,b, for the allometric equation parently skuas;Fig. 4) are not due to a simple size- or age-dependent transposition; rather, in(SAC) = a + b in(general size) there are two fundamentally different growth for the Common Tern is patterns. One is common throughout devel- opment in gulls and skuas,and in early devel- b = 0.272 _+ 0.012, opmental stagesin skimmersand terns, while n = 27, while in the Herring Gull the other is seen only in the late stagesof de- velopmentof skimmersand terns.Similarly, as b = 0.216 _+ 0.016, demonstratedby SAC and TOE (Fig. 6), the on- n = 7. During posthatchingdevelopment, gulls, togenetic trajectoriesfor postcranialgrowth in terns and skimmersall appear to exhibit similar terns more closely resemble those found in allometric growth (Common Tern, b = 0.217 _+ skimmersrather than the trajectoriesfound in 0.006, n = 58; Herring Gull, b = 0.198 _+ 0.050, gulls. However, in these postcranial characters n = 5). Tarsometatarsal(TMT) growth in terns there are qualitative similarities in all larid and skimmers slows during the posthatching growth trajectories: although SAC develop- period, although somereacceleration is seen in mental allometry is more positive in terns and the final stagesof growth. A similar TMT de- skimmersduring the prehatchingperiod, gulls velopmental trajectory is seen in gulls, al- as well as terns and skimmers exhibit a shift to though posthatching growth deceleration ap- lesspositive SAC allometry during posthatch- pears to be somewhat less than in terns and ing development;the ontogenetictrajectory for 878 W. PARKERCANE [Auk,Vol. 111

TOE (Fig. 6) seen in terns and skimmersis dif- characterstate) and, although possible,is less ferent from that seen in gulls, but the gulls probableon the basisof parsimony(see below). examined here (as well as terns and skimmers) I argue that evidence for characterpolarity show a similar slowing of growth in the post- alsocan be derivedfrom the relativecomplexity hatchingperiod. The differenceis one of degree of the ontogenetictransformations leading to and perhaps in the timing of this transforma- the diversity of larid bill shapes.Consider the tion during posthatching development (the relative growth patterns found in two taxa, A limited data here suggestthat the transforma- and B. Taxon A has highly nonlinear size-de- tion to slower TOE growth in gulls takesplace pendentallometric growth, while no suchshift somewhatlater in the posthatchingperiod). in relative growth patternsis found in taxonB. Although the qualitative similarities in the The ontogenetictrajectory in A, involving not trajectoriesfor all larid taxaexamined here limit only simple size changesduring development the phylogenetic inferencesthat can be drawn but alsoage-dependent shape changes, is more from the postcranialevidence, the relative on- complexthan the developmentpattern seenin togeniesof hindlimb and especiallybill char- taxon B (most likely reflecting greater com- acters suggesta phylogenetic hypothesis that plexity in geneticcontrol of development).This reflectsa closerrelationship between terns and greaterdevelopmental complexity would sug- skimmersand between gulls and skuas(Figs. gest a more derived state. Such an interpreta- lB and 1C). Which hypothesisis to be favored tion would be further supportedif, along with depends on decisions regarding character po- having less complexity,the allometric growth larity (i.e. whether the conditionfound in gulls pattern present throughout development in and skuas is more primitive [Fig. lB], or more taxonB is the sameas that occurringat an earlier derived [Fig. 1C] relative to the condition in the stageof developmentof A. This commonon- terns and skimmers). togeneticallometry would be assumedto be the Theoretical arguments for the use of com- more primitive or ancestralgrowth trajectory. parative ontogenies in determining character The marked shifts in relative growth of the polarity(e.g. see critique in Wheeler1990) would bill seen late in developmentin the skimmers seem to support the hypothesis that the con- and terns would suggestthe existencein these dition of the bill in gulls and skuasis the phy- taxaof more localizedcontrols (here resulting logeneticallymore primitive characterstate (Fig. in age-dependent shape changes) added to a lB). Workers (e.g. Hennig 1966, Fink 1982,Pat- basiclarid developmentpattern. In contrast,the terson 1983) have argued that phylogenetically relative simplicity of size-dependentallometry more primitive characterstates often precede in gulls,as well as its occurrencein the early more derived ones during ontogeny. Others stagesof developmentin all larids,suggests that (Patterson 1983, de Queiroz 1984) have noted the bill shapeand underlying ontogenetictra- that ontogenetictransformations themselves are jectory found in gulls (as well as in skuas)is the shared-derived characters sought in phy- the more primitive condition,thus favoring the logenetic analyses. Empirical evidence (e.g. hypothesisreflected in Figure lB. To argue the Kraus 1988)generally supportsontogenetic-po- reverse (i.e. that the common ancestor of larids larity arguments(but seealso Mabee 1989). Such had a bill ontogenysimilar to that of skimmers criteria depend on the modification of onto- and terns;Fig. 1C)would imply neotenyin gulls, geniesthrough terminal additions;nontermin- or the gain and subsequentloss of complexity al additions (e.g. see Kluge and Strauss1985, in relative growth patterns, which would be Mabee 1989) and paedomorphosis can con- lessparsimonious. Such a hypothesiswould re- found the useof ontogeny in the determination quire the introduction of an extra step in the of characterpolarity. With this caveat,it can be evolution of the group: first the evolution of argued(Crowson 1970, Nelson 1978)that, when the tern/skimmer bill ontogenetictrajectory immature stagesof one taxon (terns and skim- from an ancestorthat had only the prehatching mers here) show resemblance to adults of an- developmentalpattern; and then the lossof the other(gulls, skuas), this is evidencethat the first tern/skimmer posthatchinggrowth pattern in taxonhas descended from ancestorsresembling gulls and skuas. the second.The reverse would imply neoteny For postcranial characters,while there are dif- (or the gain and subsequentloss of a terminal ferencesin size-dependentgrowth trajectories October1994] ComparativeLarid Ontogenies $79 that suggestaffinities between terns and skim- to phylogenetic systematics.Syst. Zool. 34:280- mersrelative to gulls (and skuas),qualitatively 299. the shapesof the trajectoriesunderlying post- DINGERKUS,G., ANDL. D. UHLER. 1977. Enzyme clear- cranial differentiation are similar across the lar- ing of alcianblue stainedwhole smallvertebrates for demonstrationof cartilage.Stain Technol. 52: ids examined.Consequently, while the greater 229-232. degree of nonlinearity in the ontogenetic tra- DUNNING, J. B., JR. 1992. CRC handbook of avian jectories in terns and skimmers for characters body masses.CRC Press,Boca Raton, Florida. such as SAC or TOE may indicate a more de- FINIC,W. L. 1982. The conceptualrelationship be- rived condition, polarity arguments based on tween ontogeny and phylogeny. Paleobiology the relative complexity of postcranialdevel- 8:254-264. opmental trajectoriesmust be considered more H^ClCETT,S.J. 1989. Effectsof varied electrophoretic equivocal. conditions on detection of evolutionary patterns AlthoughZusi (1962)felt his osteologicaland in the Laridae. Condor 92:73-90. myological data were inconclusive, he tenta- HAYS, H., AND R. W. RISEBOROUGH.1972. Pollutant concentrationsin abnormal young terns from tively suggestedthat the skimmersare a highly Long Island Sound. Auk 89:19-35. specializedoffshoot of the terns. The ontoge- HENNIG,W. 1966. Phylogenetic systematics.Univ. netic evidence presented here would support Illinois Press, Urbana. sucha hypothesis.The relatively lesscomplex KLUGE,A.G. 1988. The characterizationof ontogeny. ontogenetictrajectories for the bill suggestthat Pages57-81 in Ontogeny and systematics(C. J. skuasand gulls lie closer to the ancestralroot Humphries, Ed.). Columbia Univ. Press, New of the Laridae. These characteristics of bill on- York. togeny, as well as the differing patterns of de- KLU•E, A. G., ^ND R. E. STRAUSS.1985. Ontogeny velopment in distal hindlimb elements, indi- and systematics.Annu. Rev. Ecol. Syst. 16:247- catethat terns and skimmersare more closely 268. related to each other than either is to the gulls KRAUS,F. 1988. An empirical evaluation of the use of the ontogeny polarization criterion in phy- or skuas,with skimmersbeing the morehighly logenetic inference. Syst. Zool. 37:106-141. derived taxon. M^I•EE,P. 1989. An empirical rejection of the on- togenetic polarization criterion in phylogenetic inference. Cladistics 5:409-416. ACKNOWLEDGMENTS MOYNII-•N, M. 1959. A revision of the family Lar- I thank Helen Hays, Director of the Great Gull Is- idae (Aves). Am. Mus. Novit. 1928. land (GGI) Project for her permission to carry out NELSON,G. 1978. Ontogeny, phylogeny, paleontol- much of this work on the island, and the many vol- ogy and the biogeneticlaw. Syst.Zool. 27:324- unteersand studentson GGI who helped in obtaining 345. specimens.The manuscript benefitted from the com- PAI'I•RSON,C. 1983. How doesphylogeny differ from ments of G. Barrowclough,M. LeCroy, and several ontogeny?Pages 1-31 in Development and evo- reviewers of an earlier version of the manuscript.This lution (B.C. Goodwin, N. Holder, and C. C. Wy- is contribution number 71 of the Great Gull Island lie, Eds.). CambridgeUniv. Press,Cambridge. Project. PETERS,J.L. 1934. Check- of the world, vol. 2. Harvard Univ. Press, Cambridge, Massa- chusetts. LITERATURE CITED PIMEN•,R.A. 1979. Morphometrics.Kendall/Hunt, ALBERCH,P.S., S. J. GOULD, G. F. OS•-•, ^ND D. B. Dubuque, Iowa. W.•CE. 1979. Size and shape in ontogeny and ROI-ILE,F.J. 1988. NTSYS-PC, Numerical phylogeny.Paleobiology 5:296-317. and multivariate analysissystem, version 1.5. Ex- BAUMEL,J. J., A. S. KING, A.M. LUCAS,J. E. BREAZILE, eter Publishing, Setauket,New York. AND H. E. EVANS(Eds.). 1979. Nomina Anatom- SCrtNELL,G.D. 1970a. A phenetic study of the sub- ica Avium. Academic Press, London. (Aves).I. Methodsand resultsof prin- BURGER,J., AND M. GOCHFELD. 1990. The Black Skim- cipal componentsanalyses. Syst. Zool. 19:35-57. mer. Social dynamics of a colonial .Co- SCrtNELL,G.D. 1970b. A pheneticstudy of the sub- lumbia Univ. Press, New York. orderLari (Aves).II. Phenograms,discussion, and CROWSON,R. A. 1970. Classificationand biology. conclusions.Syst. Zool. 19:264-302. Atherton Press, New York. $IBLEY,C. G., AND J. E. AHLQUIST.1990. Phylogeny DE QUEIROZ,K. 1984. The ontogenetic method for and classification of birds. Yale Univ. Press, New determiningcharacter polarity and its relevance Haven, Connecticut. 880 W. PARKERCA/qE [Auk,Vol. 111

STRAUCH,J. G., JR. 1978. The phylogeny of the Char- Cranial measures adriiformes (Aves): A new estimate using the methodof charactercompatibility analysis. Trans. PRENAR. Rostrocaudallength from anterior end Zool. Soc. Lond. 34:263-345. of nasal opening to rostral tip of os premaxillare TIMMERMANN,G. 1957. Studien zu einer verglei- (Schnell's SK2). SYMLEN. Rostrocaudallength of chendenParasitologie der Charadriiformesoder symphysismandibularis (Schnell's SK13). Regenpfeiferv6gel.Tiel I: Mallophaga. Parasito- logische Schriftenreihe. Heft. 8. Gustav Fischer, Postcranial measures Jena. W^SSERZUG,R.J. 1976. A procedurefor differential COR. Lengthof coracoideumfrom mostdistal edge stainingof cartilageand bone in whole formalin- of Processusacrocoracoideus to ventral edge of pro- fixed vertebrates. Stain Technol. 51:131-134. cessuslateralis. SCP. Maximum length of scapulafrom WETMORE, A. 1960. A classification for the birds of cranial edge of acromion to extremitascaudalis. ILM. the world. Smithson. Misc. Coil. 139(11):1-37. Length of ilium from cranial end to midpoint of mar- WHEEI.m•, Q. D. 1990. Ontogeny and characterphy- go caudalis.ISH. Length of ishium from caudaledge logeny. Cladistics6:225-268. of foramen acetabulito caudal end of processuster- ZusL R.L. 1962. Structural adaptationsof the head minalisischii. SAC. Transversely,maximum width of and neck in the Black Skimmer. Publ. Nuttall synsacrumbetween roedial marginsof alae ilii. FEM. Ornithol. Club no. 9. Maximum length of femur from proximal edge of cristatrochanteris to distal edge of condyluslateralis.

APPENDIX I TBT. Maximum length of tibotarsusfrom proximal edge of crista cnemialis lateralis to distal end of ep- Adult taxa(and number measured)used in analysis icondylus lateralis. TMT. Maximum length of tarso- of interspecific bill allometries and principal-com- metatarsus from proximal surface to distal end of ponentsanalysis of postcranialcharacters. When both trochleametatarsi tertii. TOE. length of phalanxprox- sexesmeasured, averages of pooled male and female imails of digitus tertius of pedis. HUM. Maximum measurementsused in analyses. length of humerus from proximal surfaceof caput Skuas:Catharacta skua (1), Stercorariuslongicaudus humeri to distal edge of condylus ventralis. RAD. (1). Gulls: heermanni(1), L. delawarensis(2), L. Maximum length of radius from proximal surface of canus(2), L. argentatus(2), L. fuscus(1), L. californicuscaput radii to distal surfaceof faciesarticularis radi- (2), L. occidentalis(1), L. marinus(2), L. glaucescens(2), ocarpalis.CMC. Length of carpometacarpusbetween L. hyperboreus(1), L. atricilla(2), L. pipixcan(1), L. phi- proximal edge of Trochlea carpalls and Facies arti- ladelphia(1), Rissatridactyla (2), R. brevirostris(2). Terns: cularisdigitalis major. PRX. Length of phalanxprox- Chlidoniasniger (2), nilotica (1), caspia imails digiti majoris from proximal surfaceof facies (1), S. hirundo(66), S. dougallii(2), S. albifrons(2), S. articularis metacarpalisto Faciesarticularis phalan- maxima(2), Larosternainca (1), Anousminutus (1), Gygis gealis. alba(1). Skimmers:Rynchops niger (3).

APPENDIX 2

Descriptionof measurementsused. Terminology follows Baumel et al. (1979); see also Schnell (1970a).