Multivariate assessment of the phenetic affinities of

Australasian (Aves: )

by

Allan J. Baker

Department of Ornithology, Royal Ontario Museum, and Department of Zoology,

University of Toronto, Toronto, Ontario, Canada

Abstract tion in morphological characters (Baker, 1975).

The Australian taxa have suffered similar treat- Phenetic affinities of Australasian oystercatchers were eluci-

dated multivariate statistical of seven The first by analysis morpho- ment, as illustrated below. metric characters taken from skins. Nineteen museum opera- to receive formal recognition was a pied , tional taxonomie units (OTUs), representing all the currently described Vieillot as and most of the subordinate by (1817) Haematopus longi- recognized species taxa, were used to calibrate morphological variation in Australasian rostris. Three further pied from northwestern the by the Haematopus against range displayed Haematopo- South , New Wales and Queensland were didae. The rather of variation homogeneous nature among H. described respectively as picatus King, 1826, OTUs was not well suited for analysis by hierarchical but ordination method H. australianus H. clustering methods, a non-hierarchical Gould, 1838 and longirostris

utilizing both principal components and nonmetric scaling Mathews mattingleyi Mathews, 1912. (1912) syn- produced excellent summaries of the similarity matrices. onymized H. australianus with the nominate H. The South Island Pied Oystercatcher (H. o. finschi) of

New Zealand clustered tightly with the Eurasian H. ostra- l. longirostris, and restricted picatus as a sub- and thus be of Palaearctic All the legus seems to origin. of H. The he species longirostris. following year remaining Australasian taxa grouped in another cluster with these taxa under H. the New World and it that the submerged ostralegus longi- forms, appears likely Australian Pied Oystercatcher will have to be split from the rostris Vieillot, ostensibly because of the paucity Comments the ostralegus group as a separate species. on of material then available by which good sub- and of zoogeography the family are postponed species could be All until a current study of a much larger data set is completed. recognized (Mathews, 1913).

Australian pied birds were subsequently lumped “It cannot be but to to divers mischievous any study have under H. ostralegus Linnaeus, 1758, by Mathews systems of nomenclature simultaneously co-existent in the & Iredale (1921), and no subordinate taxa were same dominion.” Milligan (1911). designated. Mathews (1927) assigned longirostris

INTRODUCTION of and picatus as subspecies H. ostralegus, but

Australasian stercatchers distributed Peters considered H. oy are widely (1934) only o. longirostris

around the of sea shores New Zealand, Australia, as valid. Oliver (1955) concluded that specimens

most islands in Torres Straits, Misima, Kai and from New South Wales, northern New Zealand

the Islands Aru Islands, and the southern coast of New Guinea and Chatham were all referable to one

Condon (Condon, 1975). The affinities and hence the species, H. longirostris. (1975) relegated

of H. number of taxa that should be recognized from longirostris to a subspecies ostralegus, but

this region have been subject to lively debate in noted that "the taxonomy is very confused and

the past, and remain largely unresolved today. some authors are inclined to follow Oliver".

I have outlined previously the chaotic nomen- Similar taxonomie confusion is apparent for the

melanistic The clatural history of the New Zealand taxa, and have oystercatchers of Australia. original

the suggested that this instability was due primarily description of Sooty Oystercatcher was given

of varia- to an essentially typological assessment by Gould (1845) as Haematopus fuliginosus. BIJDRAGEN TOT DE DIERKUNDE, 47 (2) - 1977 157

erected the Castelnau & Ramsay ( 1877) a new species, characters and/or the taxa is dependent on

H. for melanistic of variation If the of ophthalmicus (lapsus calumni), a range spanned. placement

with around the taxa in such is to be reliable and specimen a large fleshy patch eye a space a repeat- from Bountiful Island in the Gulf of Carpentaria. able representation of their phenetic affinities,

In Mathews referred both and then it follows that 1912 fuliginosus as many taxa as possible

ophthalmicus to subspecies of the New Zealand should be treated in the analysis. I have used all

H. unicolor 1844 and erected the data available to and H. liatus oystercatcher Forster, me, only pal

bernieri for a new subspecies H. u. a specimen galapagensis, H. p. pitanay and H. p. durnfordi

from Bernier Island in Australia. the in Western On are not represented. As some of the taxa used

basis of he unicolor this supposed priority synonymized study are actually represented by individuals,

with niger in 1913, but retained the above sub- I have followed Sneath & Sokal (1973) hereafter

species. Mathews & Iredale (1921) reversed this in employing the terminology operational taxo-

decision and submerged all taxa under H. uni- nomie units (OTUs).

color. In 1927 Mathews reinstated juliginosus, The OTUs used in this study, the number of

and bernieri as of H. in each and their in the ophthalmicus subspecies species geographic range

unicolor, but Peters (1934) rejected unicolor as a breeding season are given in table I. The nomen-

H. clature checklists species epithet and recognized only f. fuligi- used is that of current and con-

nosus and H. from Australia. notes taxonomie decisions For f. ophthalmicus no on my part.

These latter have endured and used heuristic and taxa since are purposes, to provide a yardstick by

Condon by (1975). which distances among OTUs can be judged, I

The cluttering of the literature with numerous have included hybridizing forms as separate OTUs.

is For and nomenclatural changes perhaps understandable example, the melanistic H. u. unicolor the

in and H. the context of the nature relative recency pied u. reischeki hybridize so extensively in

of ornithological discovery in Australasia. How- northern New Zealand that they might well be

if is be made the in if the ever, progress to beyond point lumped together a single taxon extent

where taxonomie decisions reflect little than of is the final arbiter. more gene exchange Conversely,

the personal preferences of their proponents, it is although H. bachmani hybridizes with H. palliatus

clear that a substantive study of the systematic in lower California, the hybrid population (here

will have made. H. relationships of the taxa to be referred to as p. frazeri) is relatively restricted

The taxonomie outlined above stems in distribution and thus the of the instability integrity gene

principally from an inadequate assessment of the pools of the parental species is essentially main-

tained. affinities of allopatric taxa, especially those from

the Old and New Worlds. This paper presents Selection of characters a multivariate analysis of the phenetic affinities of

Australasian oystercatchers, based on external mor- Although numerical taxonomists such as Sneath &

skin used advocate the of phometric characters traditionally by Sokal (1973) use large suites of

avian characters 60 that simi- systematists. (commonly or more) so

larity coefficients will have reasonable confidence MATERIALS AND METHODS in this levels, study I have used only the seven Selection of taxa for comparative study external morphological characters traditionally

Because the phenetic affinities of Australasian used by avian systematists in assessing the affini-

need be in the broader ties of related I Haematopus to cast per- closely taxa. While acknowledge

of the of variation that the of small number of characters spective range morphological use a may

in the the selection of taxa for decrease the of the statistical Haematopodidae, reliability results, my is mul- comparative study most important. Where intention is not to construct a numerical taxonomy

methods this but rather tivariate statistical are used, sampling as such, to obtain a more complete eval-

factor becomes critical because the configuration uation of the morphometric information contained of the multidimensional the in character suite. The characters used space defined by a traditional 158 A. J. BAKER - PHENETIC AFFINITIES OF OYSTERCATCHERS

TABLE I

OTUs used in this their sizes and study, sample geographic ranges.

x OTU » Geographie range in the breeding season ) no. OTU designation Abbreviation 9~9 ï~è

1 Haematopus ostralegus malacophaga H.o.mala. 18 17 Iceland and Faeroe Islands

occidentalis H.o.occi. British Isles 2 H. o. 22 21

H.o.ostr. 40 east to and south to 3 H. o. ostralegus 37 Europe Arkhangelsk Spain

and Baltic Sea; coasts of Asia Minor, Mace-

donia, the Black and Caspian Seas

H.o 4 H. o. longipes long. 10 1 Inland southern and eastern Russia and western

Siberia

H. H.o.fins. 18 20 South Island of New Zealand 5 o. finschi

6 H. o. osculans H.o.oscu. 5 11 Coasts of northeastern Russia from eastern

Siberia to northern China

7 H.o.rost. 41 New Guinea south Australia to Tas- H. o. longirostris 45 through

mania

8 H. chathamensis H.chat. 14 14 Chatham Islands

H. H.f.fuli. Coasts of Australia 9 fuliginosus fuliginosus 21 28 except ophthalmicus range

10 H. f. ophthalmicus H.f.opht. 3 4 Northern Australia coast from Cape York Peninsula west to Kimberleys

11 H. unicolor unicolor H u.unie. 20 21 New Zealand

reischeki of Zealand 12 H. u. Hu.reis. 17 17 North Island New

H 13 H. moquini moquini m.moqu. 8 8 Coasts of South Africa from Natal to South-

west Africa

14 meadewaldoi H.m. mead. 1 1 Eastern Islands H. m. Canary (probably extinct)

15 H. palliatus palliatus H.p.pall. 4 11 Atlantic and Pacific coasts of North and

Central South America

California south 16 H. p. frazeri H.p.fraz. 11 18 Lower to Jalisco

from about 17 H. ater H.ater 14 29 Coasts of South America 10°S on

about Falkland the west to 43°S on the east;

Islands

18 H. bachmani H.bach. 17 15 Aleutian Islands through Alaska west to Lower

California

19 H. leucopodus H leuc. 1 3 Southern South Africa from Chiloé Island

-round to Chubut; Falkland Islands

Totals 285 321

Baker !) These ranges are mapped more exactly in (1973).

outlined the and their methods of measurement are the claw. (6) Tail length — measured from below: the of of central rectrices Bill — measured the (1) length along point emergence two to

dorsal surface of the culmen from the tip to the the tip of the longest rectrix. (7) Wing length —

of Bill — junction the feathers. (2) depth mea- measured as the chord of the flattened wing, from sured from the junction of the feathers with the the carpal flexure to the tip of the outermost pri-

feather. The first five culmen on the upper mandible to the exposed mary measurements were

of mandibular rami the lower taken with dial the and portions the on calipers to nearest 0.1 mm,

— the the measured with steel mandible. (3) Bill width measured at remaining two were a widest the the dorsal point of bill beneath feather rule accurate to 1 mm.

the mandible. Tarsus margin on upper (4) length Statistical methods — measured diagonally from the posterior aspect of the joint between the tibiotarsus and the tarso- As oystercatchers are markedly sexually dimorphic metatarsus, to the anterior aspect of the joint in size (Baker, 1975), character variation in males

and middle and between the tarsometatarsus the toe. females was analyzed separately beyond. The

( 5 ) Middle toe length — measured dorsally from raw data matrix for each sex was too large to

the joint with the tarsometatarsus to the base of handle practically at the level of individuals, so 47 - BIJDRAGEN TOT DE DIERKUNDE, (2) 1977 159

means were calculated for each OTU for but MDSCALE (except similar, more faithfully preserves

those where n = 1) and the dimensions of the the distances between the OTUs in the reduced

data matrix reduced The of fit of the were correspondingly to space (Rohlf, 1970). goodness

of char- MDSCALE solution distance 19 X 7 (number OTUs X number of to the matrix was

acters). assessed by computing stress values (see Kruskal,

* Prior to the calculationof and the matrix correlation r similarity coefficients, 1964) M (Rohlf,

the characters were standardized across OTUs so 1972).

that had of and In this they a mean zero a standard study the OTUs were projected into a

deviation of and thus each character theoreti- three-dimensional and minimum one, space, a spanning

cally can contribute equal information in the en- tree (MST) computed from the original distance

taxonomie distances matrix suing analyses. Average (d) was superimposed to provide additional

and correlation of product-moment coefficients (r) insights on the phenetic relationships the OTUs.

were then and Distinctive clusters of OTUs in the three-dimen- computed among OTUs, separate

cluster analyses were performed using the un- sional MDSCALE ordinations were isolated using

arithmetic finds weighted pair-group method on aver- subsets analysis. This technique all subsets

— Sneath & of OTUs the condition that the ages (UPGMA Sokal, 1973). satisfying greatest

used it between OTUs in the subset is less UPGMA clustering was initially because dissimilarity

the least distortion when than the least between OTU in usually gives represented dissimilarity any

two-dimensional & the subset and OTU in the subset. by a phenogram (Sokal Rohlf, any not Only

the OTUs and OTUs 1962; Farris, 1969; Rohlf, 1970). Cophenetic complete set of single are correlations between the phenograms and the orig- not permitted as subsets. inal distance and correlation matrices were cal- RESULTS culated to determine how accurately the pheno- Cluster analysis grams summarized the similarity coefficients.

The The results of hierarchical clustering were tested morphometric relationships of the 19 OTUs, against a non-hierarchical ordination technique as elucidated by UPGMA cluster analyses of the

and shown for adult females combining both principal component analysis r d matrices, are and

males in (PCA) and nonmetric multidimensional scaling fig. 1. Results for both sexes are quite analysis (MDSCALE). This dual approach utilizes disparate for the two coefficients of similarity. the results of PCA initial For three from as an input configuration females, clusters are produced for the MDSCALE routine. correlation coefficients and two from average

The value of principal component transforma- taxonomie distances. Moreover, the composition

data derives the somewhat the tion of morphometric from assess- of clusters changes for two coeffi-

of simultaneous variation in all characters cients. While cluster of the matrix ment analysis r (fig.

all and the reduction of the known close over OTUs, consequent la) faithfully preserves phenetic

data matrix much smaller between the forms of a large to a set of relationships hybridizing

thus northern New Zealand unicolor and H. orthogonal (and uncorrelated) principal com- (H. u. u.

other ponents. The MDSCALE technique, on the reischeki) and lower California (H. p. palliatus

OTUs and H. it of hand, represents / geometrically by / points p. frazeri), has the disruptive effect in reduced of k dimensions k < the OTUs the western a space (where t\ separating representing

k = typically 2 or 3) such that the interpoint European subspecies of H. ostralegus into different

have with More distances a monotone relationship the clusters. specifically, it breaks off H. o. distances the OTUs in the dis- occidentalis and H. Ice- among original o. malacophaga (British, similarity matrix. MDSCALE differs from PCA landic and Faeroe birds) from the cluster including in that is nominate requiring only the monotone relationship the H. o. ostralegus (the remaining

in the reduced rather than The central preserved space, ex- western European birds). European

maximum of the variation. and and H. plaining a percentage Eurasian subspecies (H. o. longipes o.

Results of the in osculans with H. two techniques are practice quite respectively) group o. ostralegus, 160 A. J. BAKER - PHENETIC AFFINITIES OF OYSTERCATCHERS

Fig. 1. Phenograms from UPGMA cluster analysis of similarity matrices: (a) correlation phenogram for females, (b) distance

for for distance for identified phenogram females, (c) correlation phenogram males, (d) phenogram males. OTUs are as in

table and cophenetic correlations are denoted r I by c .

and with H. o. finschi of New Zealand. The only because they are largely consistent with taxon-

Australian Pied Oystercatcher (H. o. longirostris) omies adopted by recent checklists (e.g. Kinsky,

clusters with western European H. ostralegus and 1970; Condon, 1975), but also because they are

the melanistic H. m. moquini of South Africa. explicable in terms of Zoogeographie and evo-

Both Australian of the All H. subspecies Sooty Oyster- lutionary processes. the subspecies of ostra-

catcher The Chat- H. lon fall within (H. fuliginosus) grouptogether. legus except o. girostri s a major

Islands New H. ham Oystercatcher (H. chathamensis) cluster, and the Zealand o. finschi shows

form related the closest with the Eurasian H. appears as an aberrant distantly to phenetic affinities o.

Canary Islands H. m. meadewaldoi. longipes and H. o. osculans. H. chathamensis of

Results of the cluster analysis of the matrix of the Chatham Islands clusters with western Euro-

taxonomie distances intu- of H. and the New average (fig. lb) are pean subspecies ostralegus

itively more appealing (though note the comments World H. leucopodus and H. bachmani. Both

Sokal to the H. of Rohlf & (1965) contrary regarding Australian subspecies of fuliginosus group

size-ordered classification from d matrices), not together and link with the South American Black BIJDRAGEN TOT DE DIERKUNDE, 47 (2) - 1977 161

and the Australian Pied of variation found in the Oystercatcher (H. ater) geneous range phenetic

Oystercatcher (H. o. longirostris). The hybridizing Haematopodidae.

forms of both H. unicolor and H. palliatus form Principal components analysis and form broader compact groups together a Because the PCA of the OTUs with the South H. projections are cluster, African m. moquini

similar to those of MDSCALE in this and the H. quite study, Canary Islands m. meadewaldoi as they will not be here. However, it is more distant outliers. presented

useful to the of of the Neither the inspect pattern loadings of phenograms are particularly good characters on the first three principal in summarizing the information in the similarity components, as they aid considerably in interpreting the axes matrices. Cophenetic correlations are only 0.735 of the MDSCALE ordination. For females, 91.7% for the correlation coefficient phenogram and of the character variance is accounted for the 0.743 for the distance phenogram. Cluster analysis on first three principal components; 59-8% on com- of OTUs of males, as with females, produces three

I, on II and on III. For clusters with correlation coefficients and two with ponent 20.2% 11.7%

of the character variance is taxonomie distances. The of males, 86.9% ac- average membership counted for the first three clusters differs between the but the overall by principal com- sexes, ponents; 54.4% on component I, 21.2% on II and patterns of groupings of OTUs are broadly similar. III. Cluster the matrix 11.3% on analysis of r again partitions The the loading are similar for sexes the subspecies of H. ostralegus into different patterns I has clusters, but more disturbingly H. chathamensis (table II). Component high positive loadings

for all characters bill and therefore is inserted between the hybridizing forms of H. except length, the OTUs the serves to differentiate largely on unicolor (fig. lc). The integrity of the H . palliatus basis of size. Component II is effective in dis- complex is maintained, however, and the two sub-

OTUs on an inverse relation- of H. Western tinguishing chiefly species fuliginosus group together. between bill tarsus of H. be ship length to a European subspecies ostralegus appear to (and length lesser and tail Middle closest the South H. degree) length. toe length phenetically to African m. and bill width load and and tail moquini and the Australian H . o. longirostris. positively, wing load fe- results of cluster matrix lengths negatively on III of The analysis on the d component males, and the reverse is apparent for males. Hence of males (fig. Id) are again intuitively much more III from the All the component seems to involve similar contrasts satisfactory than those r matrix. in these characters OTUs. H. across Eurasian subspecies of H. ostralegus except o.

osculans in the same with the appear cluster, along TABLE II

ostralegus- like H. leucopodus of South America Loading of characters on the first three principal components

H. of the Chatham Islands. H . based on a matrix of correlations and chathamensis o. among morphological

characters. osculans is outlier in cluster a distant a including

the hybridizing forms of H. palliatus from lower Character Sex Com- Com- Com-

ponent I ponent II ponent III California and the Canary Island Oystercatcher

(H. m. meadewaldoi). The OTUs representing Bill length Male 0.163 —0.880 0.405

Female 0.236 —0.919 —0.261 H. unicolor and H . fuliginosus are clustered to- Bill depth Male 0.849 —0.248 —0.031 gether with H. ater of South America. Female 0.894 —0.142 —0.103 Both phenograms of males provide relatively Bill width Male 0.837 —0.220 —0.381 Female fits with the values in the and d 0.901 —0.088 0.320 poor similarity r Tarsus length Male 0.804 0.304 0.132 matrices. The cophenetic correlations are only Female 0.808 —0.411 0.056 0.688 for the correlation and coefficients 0.695 Middle toe length Male 0.797 0.140 —0.433

Female 0.184 for the taxonomie distances. This dis- 0.763 0.575 average Wing length Male 0.821 0.350 0.296 appointing result reflects the weakly hierarchical Female 0.863 0.277 —0.376 the which in nature of similarity matrices, turn Tail length Male 0.634 0.605 0.433

Female 0.733 0.512 0.402 derives from the relatively constrained and homo- 162 A. J. BAKER - PHENETIC AFFINITIES OF OYSTERCATCHERS

Nonmetric multidimensional scaling and subsets catchers, and have accordingly classified them in

analysis a single monogeneric family. Consensus has ended

at the generic level, however, because of the The three-dimensional MDSCALE ordinations of seemingly intractable nature of morphometric females and males differ somewhat in the place- variation of allopatric forms, the lack of a global ment of particular OTUs, but the general pattern view in assessing affinities, and the confounding of phenetic relationships expressed therein is influence of melanism which has probably arisen similar (fig. 2). In both ordinations the Eurasian independently in several geographically wide- H. and subspecies of ostralegus group together spread taxa (e.g. H. unicolor in New Zealand, with H. have clear phenetic affinities o. finschi H. fuliginosus in Australia, H. moquini in Africa, of New Zealand, H. leucopodus of South America H. ater in South America and H. bachmani in

and H. bachmani of western North America. All North America). the remaining Australasian taxa together group Taxonomie problems within the Haematopodi- within another major grouping containing all the dae are thus similar to those encountered within New World OTUs. The African H. moquini homogeneous assemblages such as species-com- is phenetically intermediate, appearing nearer the plexes, and their successful resolution often de- New World of females and the group (fig. 2a) of relevant data pends on a careful synthesis from Old World of males (fig. 2b). group morphology, ecology, ethology, physiology etc. Distinctive clusters of OTUs within the two (Mayr, 1969). This approach has been fruitful indicated subsets major are by the proce- groups for the New Zealand taxa (Baker, 1974), but dure. For the western sub- females, European data reliable from such diverse disciplines as these species H. o. occidentalis and H. o. malacophaga are as yet unavailable even for the Australasian

are similar as are the Eurasian very phenetically, taxa. H. and H. osculans. As o. longipes o. subspecies Yet, there is still much valuable taxonomie in- might be for both sexes the New World expected, formation to be gleaned from morphological data and New Zealand hybridizing forms fall into sets. The use of non-hierarchical ordination

subsets, as do the of the Austra- specific subspecies for methods seems particularly appropriate analyz- lian H. A different of fuliginosus. slightly group ing phenetic relationships of homogenous taxa occi- subsets emerges for Old World males; H. o. as the methods do not force them into a rigid with H. dentalis has closest phenetic affinity o. and possibly artificial hierarchy. In this study and H. of New Zealand ostralegus, o. finschi of multivariate analysis of an admittedly small set H. of in the same subset as leucopodus appears characters long used by avian systematists has given South America. new insights into the nature of morphological The low stress values (s) and the high matrix variation in oystercatchers. Of particular impor- correlations (r*j*) indicate that the ordinations tance in this respect is the finding that two principal excellent fit to the relation- provide an phenetic Old groups of OTUs can be defined, one an in the matrix of taxo- ships represented average with solid nucleus of Eurasian World group a sizes nomie distances. Given the small sample on H. ostralegus and the other a New World group,

which some OTUs are based, and the small number which also includes all the Australasian OTUs of characters used, the of concordance of degree H. of Zealand. This except o. finschi New latter the ordination patterns of males and females at- taxon seems better considered as a Palaearctic

tests to the robustness and of the PCA/ utility joint element the New in Zealand fauna (following a MDSCALE approach. secondary invasion) rather than a Holarctic ele-

ment as suggested by Fleming (1962). Conversely,

the Australian H. o. to be DISCUSSION longirostris appears phenetically similar to the New World forms, and

Taxonomists have long been impressed by the should probably be given full species rank. How-

distinctive but uniform morphology of oyster- ever, Zoogeographie and taxonomie speculation 47 - BIJDRAGEN TOT DE DIERKUNDE, (2) 1977 163

Fig. 2. Three-dimensional ordinations using the PCA/MDSCALE method: (a) females, (b) males. Dimension III is represented by the height of the projections. OTUs are identified as in table I, stress is denoted by s and the matrix correlation by rdd*. The MST is indicated by the dashed lines, and joins phenetically similar OTUs together. Distinctive clusters of OTUs defined by the subsets procedure are enclosed by solid lines.

11 164 A. J. BAKER - PHENETIC AFFINITIES OF OYSTERCATCHERS

of birds Gulf data is a collection from the Norman River, of based on the present set obviously prema- Carpentaria, with descriptions of some new species. ture, and must await the results of a study of a Proc. Linn. Soc. N.S.W., 1 (4): 379-386.

much set of skeletal and biochemical char- larger CONDON, H. T., 1975. Checklist of the birds of Australia,

in 1. Australian Ornitholo- acters, currently progress. Non-passerines: 1-311 (Royal gists' Union, Melbourne).

FARRIS, J. S., 1969. On the cophenetic correlation coefficient. ACKNOWLEDGEMENTS Syst. Zool., 13 (3): 279-285. examine collections in their For permission to care I am Fleming, C. A., 1962. New Zealand biogeography; a grateful to the following colleagues: Mr. H. J. de S. Disney, paleontologist's approach. Tuatara, 10 (2): 53-108. Australian Museum, Mr. H. T. Condon, South Australian GOULD, J., 1845. Haematopus fuliginosus, Gould. Sooty Museum, Dr. N. K. Johnson, Museum of Vertebrate Zoolog/, Oyster-catcher. In: J. GOULD, Birds of Australia, 6 University of California, Berkeley, Mr. F. C. Kinsky, [(18)]: [8], pl. VIII (J. Gould, London). National Museum of New Zealand, Mr. A. R. McEvey, KINSKY, F. C., 1970. Annotated checklist of the birds of National Museum of Victoria, Dr. G. F. Mees, Rijksmuseum New Zealand including the birds of the Ross depen- van Natuurlijke Historie, Leiden, Dr. F. Salomonsen, Uni- dency: 1-96 (Reed, Wellington). versitets Zoologiske Museum, Copenhagen, Mr. R. Scarlett, KRUSKAL, J. B., 1964. Multidimensional scaling by optimizing Canterbury Museum, Dr. L. L. Shortt, American Museum of goodness of fit to a nonmetric hypothesis. Psycho- Natural History, Dr. D. W. Snow, British Museum (Natural metrika, 29 (1): 1-27. History), Dr. P. J. Stanbury, Macleay Museum, University MATHEWS, G. M., 1912. A reference list to the birds of of Sydney, Dr. R. W. Storer, Museum of Zoology, University Australia. Novit. Zool., 13 (3): 171-455. of Michigan, Dr. G. F. Storr, Western Australian Museum, 1913. A list of the birds of Australia: 1-453 (Witherby , Mrs. A. M. Sudilovskaja, Zoological Museum, University of & Co., London). Moscow, Mr. E. G. Turbott, Auckland Institute and Museum, 1927. Systema avium Australasianarum. A , systematic Mr. D. P. Vernon, Queensland Museum and Dr. J. Wattel, list of the birds of the Australasian region: 1-1027 Zoological Museum, University of Amsterdam. The figures (British Ornithologists' Union, London). were drafted skilfully by Maria Tran Thi Vinh-Hao, and MATHEWS, G. M. & T. [REDALE, 1921. A manual of the the manuscript was typed by Greta Vooren. I would also birds of Australia, 1: 1-279 (Witherby & Co., London). like to thank Drs. F. Peeters-Pieters and Mr. Paul for J. M AYR, E., 1969. Principles of systematic zoology: 1-428 of earlier literature. assistance in locating some the Compu- (McGraw-Hill, New York). carried out at of tations were the University Toronto Com- OLIVER, W. R. B., 1955. New Zealand birds (2nd ed., puter Centre using an IBM 370/165 computer. All multi- revised): 1-661 (Reed, Wellington). variate analyses were run using the NT-SYS of system pro- PETERS, J. L., 1934. Check-list of birds of the world, 2: and Kirk grams developed by F. J. Rohlf, J. Kishpaugh D. 1-401 (Harvard University Press, Cambridge, Mass.), at the State University of New York, Stony Brook. This ROHLF, F. J., 1970. Adaptive hierarchical clustering schemes financial study was made possible by support from the Syst. Zool., 19 (1): 58-82. Royal Ontario Museum. The present paper was prepared 1972. An of three ordination , empirical comparison Ian R. I during a visit with Dr. Ball, and especially want techniques in numerical taxonomy. Syst. Zool., 21 (3): to thank him and other staff members of the Zoological 271-280. Museum, University of Amsterdam, for providing facilities ROHLF, F. J. & R. R. SOKAL, 1965. Coefficients of correlation and support. and distance in numerical taxonomy. Univ. Kansas Sei.

Bull., 45 (1): 3-27. REFERENCES A. Numerical SNEATH, P. H. & R. R. SOKAL, 1973. taxonomy:

BAKER, A. J., 1973. Distribution and numbers of New 1-573 (Freeman, San Francisco).

Zealand oystercatchers. Notornis, 20 (2): 128-144. SOKAL, R. R. & F. J. ROHLF, 1962. The comparison of

, 1974. Ecological and behavioural evidence for the dendrograms by objective methods. Taxon, 11 (1):

systematic status of New Zealand oystercatchers (Chara- 33-44.

driiformes: Haematopodidae). Life Sei. Contr., R. Ont. VIEILLOT, L. J. P., 1817. L'Huitrier à long bee, Haematopus

Mus., 96: 1-34. longirostris, Vieill. In: Nouveau dictionnaire d'histoire

1975. variation, hybridization and naturelle, aux arts, à à l'éco- , Morphological system- appliquée l'agriculture, of atics New Zealand oystercatchers (Charadriiformes: nomie rurale et domestique, à la médecine, etc. par une

Haematopodidae). J. Zool., Lond., 161: 357-390. société de naturalistes et d'agriculteurs [éd. 2], 15: 410

CASTELNAU, LE COMTE DE & H. P. RAMSAY, 1877. Notes of (Déterville, Paris).

Received: 5 September 1977