--Phylogenetic relationships and evidence for pectoral reduction in the extinct Auckland Islands Merganser (Mergus australis) were investigated using a plumage-based phylogenetic analysis of the six Recent species of merganser and morphometric comparisons of study skins and skeletal specimens. The hypothesized phylogeny indicates that M. australis diverged from the other Mergus immediately after the Hooded Merganser (Lophodytes cucullatus) and is a member of a basal grade of comparatively small, southern hemisphere mergansers; the Brazilian Merganser (M. octosetaceus) branched next and is the sister-group to the larger, more derived, northern hemisphere species of Mergus. M. australis was unique in its natal and adult plumage characters, including the sexually monochromatic plumage of adults. Morphometric analyses revealed that M. australis was the smallest member of its genus but possessed the longest bills and relatively short wings and tails. Based on regression estimates of body mass and wing area for M. australis, the species had estimated wing-loadings which exceed those for other Mergini and approach the threshold of flightlessness hypothesized by Meunier (1951). Skeletal comparisons confirmed that M. australis had exceptionally long bills, and this also revealed that the species possessed relatively short wing elements and scapulae, as well as sterna characterized by shallow carinae and small caudal widths. These morphological characteristics, hypothesized phylogenetic relationships, and ecological information for M. australis are compared to the typical correlates of insularity in waterfowl listed by Weller (1980), and the evolutionary implications of these characteristics are considered. Received 26 May 1988, accepted 1 Nov. 1988.

The mergansers (Lophodytes cucullatus and five species of Mergus) are highly derived members of the sea ducks (Mergini), most closely related to the Smew (Mergellus albellus) and goldeneyes (Bucephala spp.) (Livezey 1986). Mergansers are capable divers and are largely piscivorous, but unlike several other genera of the Mergini (e.g., Somateria, Melanitta, and Clangula), are strictly foot-propelled divers, i.e., strokes of the feet provide the sole propulsion at submergence and for underwater loco- motion (Townsend 1909; Kelso 1922; Brooks 1945; Humphrey 1957, 1958). Most mergansers, like other Mergini, are limited in distribution to the northern hemisphere, show striking sexual dichromatism, and are migratory (Delacour 1959; Johnsgard 1965, 1978). The Auckland Islands Merganser (Mergus australis) was one of two Recent species of merganser with southern-hemisphere distributions (the Museum of Natural History, Univ. of Kansas, Lawrence, Kansas 66045. 410 Livezey ß AUCKLAND ISLANDS MERGANSER 41 1 other being the Brazilian Merganser [M. octosetaceus]), being limited, at least in historical times, to the Auckland Islands (50øS, 165øE), a small group of islands located 330 km south of the South Island of New Zealand (Phillips 1925, Delacour 1959, Williams 1964, Greenway 1967, Kear and Scarlett 1970, Weller 1980). Subfossil specimens of Mergus, mensurally comparable to M. australis, have been found as well on the main islands of New Zealand (Kear and Scarlett 1970). Extirpated during the first decade of the twentieth century, this insular species is represented in museum collections by some 26 skins (including downy young), three virtually complete skeletons, some additional skeletal parts, and one car- cass preserved in alcohol (Kear and Scarlett 1970). In addition to its southern-hemisphere insular distribution, the Auckland Islands Mergan- ser has been noted for its small size, lack of sexual dichromatism, relatively long bill, and unusual plumage pattern of its downy young (Mathews and Iredale 1913, Humphrey 1955, Kear and Scarlett 1970, Weller 1980). The Auckland Islands Merganser was not, as claimed by Luther (1967), flightless; individuals were observed in flight by Hutton and Ranfurly (fide Kear and Scarlett 1970). The notion that M. australis was flightless may have originated with the mistaken inclusion ofM. australis as a synonym of the Auckland Islands Flightless Teal (Arias aucklandica) by Gray (1844). Based on a study of skeletal measurements, however, Humphrey (1955) concluded that the humerus ofM. australis showed evidence of reduction in length. Humphrey further inferred from proportions of wing elements that the species also showed reduction of the ulna and manus. The unusual distribution and morphology of M. australis, particularly the evidence of wing reduction, prompted its inclusion in an ongoing study of flightlessness in diving birds. In this paper I propose a phylo- genetic hypothesis for the six Recent species ofmergansers using published descriptions of natal and adult plumages. This analysis is followed by morphometric comparisons of external and skeletal measurements to de- termine interspecific patterns of size, shape, and sexual dimorphism with- in the group and to examine in greater detail the evidence for reduction of the pectoral limb in M. australis and the possibility that the species was evolving toward flightlessness. MATERIAL AND METHODS Phylogenetic characters.- Characters of natal and adult plumages, as well as colors of soft parts, were taken from illustrations and text in Partridge (1956), Delacour (1959), Kear and Scarlett (1970), and Bartmann (1988). Kear and Scarlett (1970) pointed out several errors in the illustration of a downy M. australis in Delacour (1959). Characters used in the phylogenetic analysis were those that varied within the in-group (Lophodytes and Mergus) and for which polarities (primitive states) could be determined by comparison with its sister 412 THE WILSON BULLETIN ß Vol. 101, No. 3, September 1989 genera, Bucephala and Mergellus (Livezey 1986). I endeavored to reduce all character com- plexes to binary characters (i.e., having but a primitive and a single derived state); those for which this was not possible were analyzed as unordered. Autapomorphies (derived character states unique to a single terminal taxon) were included in the analysis because such differences have been considered of taxonomic value in traditional, phenetic classifications (e.g., Hum- phrey 1955, Delacour 1959, Johnsgard 1961 a, Kear and Scafiett 1970). The tree was con- structed using the criterion of parsimony (Wiley 1981). It was hoped that phylogenetic relationships inferred from plumage characters would be effectively independent of the morphometric patterns that were investigated. In addition to the plumage characters, I attempted to include courtship behaviors (Johnsgard 1960, 196 l a, b, 1965) and syringeal morphology (Humphrey 1955) in the analyses, but was unsuccessful because of uncertain homologies, polarities, continuous variation (in syringeal bullae), and missing data for crucial species (e.g., M. australis and M. octosetaceus). Specimens and morphometric data.--Mensural data from 15 skins of (fully grown) M. australis were collected by me or provided by colleagues. Where possible, I also made counts of the primary remiges in M. australis. I measured 10 skins of adults of each sex of the Hooded Merganser (Lophodytes cucullatus), Red-breasted Merganser (M. serrator), and Common Merganser (M. merganser); smaller samples of skins of the Brazilian Merganser (N = 5 males, 5 females) and the Chinese Merganser (M. squamatus; N = 8, 6) were available for measurement. Measurements taken on skins were as follows: total body length; length of exposed culmen on midline; nail width; length of the chord of the unstraightened wing; length of the tail, measured medially' length of the tarsus on anterior surface; and length of the middle toe, excluding nail. Total lengths were measured only on properly prepared study skins (i.e., realistically extended, unmounted skins), had coefficients of variation (s/œ) com- parable to those for other external variables, and their means were used, in part, for estimates of body size ofM. australis, M. octosetaceus, and M. squamatus  for which published records of body mass were not available. Specifically, body masses were estimated using external "body lengths" (mean total lengths minus the mean lengths of culmen and tail), which were independent of interspecific variation in relative lengths of bills and appendages. Together with estimates of wing area based on wing length, these estimated body masses were used to infer the approximate wing-loadings of the three species of Mergus lacking these data. Wing areas, where available, were measured as described by Raikow (1973), and wing- loadings for species (g of mean body mass divided by cm 2 of mean total wing area) were calculated as recommended by Clark (1971). Additional data on body masses and wing areas of mergansers and other Mergini were taken from Miillenhoff (1885), Magnan (1912, 1922), Poole (1938), Meunier (1959), Dement'ev and Gladkov (1967), Raikow (1973), Madge and Burn (1988), and P.S. Humphrey (unpubl. data). Three virtually complete skeletons and a few disassociated skeletal elements ofM. australis were available for study. I sought to measure 15 skeletons of each sex of the other mergansers, although only one complete skeleton of M. squamatus and a partial skeleton of M. octo- setaceus were available. Thirty-three skeletal variables were used in the comparisons, most of which were described in Livezey and Humphrey (1984, 1986) and Livezey (1988, 1989). Several nontraditional measurements require definition, however: LWM's of limb elements refer to "least widths at the midpoints" of shafts; LMW of the tarsometatarsus is the "lateromedial width" of the shaft (at midpoint); and length of digit-III (middle toe) is the sum of lengths of the three proximal phalanges of the digit. All skeletal measurements were made with dial calipers to within 0.1 mm. Statistical analyses.--Simple measurements of skins and skeletons were compared using two-way analysis of variance (ANOVA); missing data resulted in slight variations in sample sizes in univariate analyses. Wing-loadings were log-transformed for tests with ANOVA. Livezey ß AUCKLAND ISLANDS MERGANSER 413 Significances of differences (and associated P-values) based on correlated data were judged using the Bonferroni method of simultaneous inference (Milliken and Johnson 1984). Intra- alar proportions of skeletal wing elements were calculated as lengths of the single elements divided by skeletal wing length (sum of the lengths of the humerus, ulna, carpometacarpus, and the two phalanges of the major digit); arcsines of square roots of proportions were used in ANOVA (Sokal and Rohlf 1981). Allometry of wing areas with body mass, in which both variables include error (model II; Sokal and Rohlf 1981 ), was investigated through "geometric mean" regressions or central trend lines (Ricker 1984) of log-transformed species means. The slopes of regressions using log-transformed means are estimates of the "allometric coefficients" (b) and reflect inter- specific scaling between the variables (Gould 1966); in this application the slope reflects scaling of wing area (and wing-loading) with body size. Stepwise multivariate analysis of variance (MANOVA; see appendix 23 in Dixon 1985) was used to test for interspecific, intersexual, and species-sex interactive differences in mean vectors of external and skeletal measurements; summary statistics (Wilks' lambda and F-statistics) were based on backstep-selected subsets (P < 0.05) of the variables submitted (Jennrich and Sampson 1985). Canonical analysis (CA), a technique which defines mutually orthogonal multivariate axes that maximally separate predefined groups (relative to their pooled within-group covariances), was used to explore multivariate differences among species and sexes (Pimentel 1979). CAs also were used to isolate the multivariate differences between M. australis and the other species (referred to as "canonical contrasts"). Importance of interspecific and intersexual differences on the canonical variates (CVs) was tested through two-way ANOVAs of scores on these CVs (referred to as "ANOVAs of scores"). For MAN- OVA and CA only, external and skeletal data sets were subjected to a missing-data program (Frane 1985) in which measurements missing for reasons of deformity or bilateral breakage (for specimens lacking a small minority of data) were estimated using separate stepwise regressions on specimens grouped by species; these estimates permitted the inclusion of 99 skins and 91 skeletons in MANOVAs and CAs and comprised only 1.0% and 2.0% of the sets, respectively. Both data sets were log-transformed for CA and MANOVA (Jolicoeur 1963). Statistical programs employed are part of the 1987 version of the Biomedical Computer Programs (BMDP; Dixon 1985), performed on an IBM computer at the University of Kansas. PHYLOGENETIC ANALYSIS Based on published descriptions and illustrations, I defined 11 characters of natal plumages and 11 of adult plumages of mergansers (Table 1). Four of the latter included two derived states (characters 16, 17, 18, 20), of which the first three involved the unique plumage aspect ofM. australis. The problematic black-and-orange bill ofM. australis was coded as a unique character state (15c); the tree depicted remains most parsimonious if bill color ofM. australis is treated instead as a (partial) reversal from an orange bill (15b). Both M. australis and M. octosetaceus lack significant sexual dichromatism (character 16), but these conditions are not homologous (illustrated in Delacour 1959, Scott 1972). In M. australis, plumages of both sexes closely resemble females and juveniles of Mergus generally, and monochromatism of this species evidently is the result of the loss of the distinctive alternate plumage of males. This plumage condition rendered several characters (17-19) not comparable in this species (Appendix I). In contrast, the plumage pattern shared by the sexes ofM. octosetaceus differs significantly from the rest of the genus and, although not as striking as those for adult males of most other mergansers, is not juvenal-like and is evidently uniquely derived. Although complicated by these few instances of multiple and/or noncomparable states 414 THE WILSON BULLETIN ß Vol. 101, No. 3, September 1989 (Appendix I), the 22 plumage characters formed the basis for a phylogenetic tree for the Recent mergansers (Fig. 1) of high consistency (consistency index = 1.0). The sister-group relationship of Lophodytes with Mergus, supported by osteological comparisons in Livezey (1986), was corroborated by three plumage characters of downy young (Table 1, Fig. 1). M. australis emerged as the second branch in the basal grade ofmergansers and was characterized by four autapomorphies. Based on characters of its natal plumages, M. octosetaceus was the next branch to diverge in Mergus, and this neotropical species possessed four autapomorphies of the adult plumage (Fig. 1). The presence of a pale suborbital stripe in the natal plumage (character 4b) is shown as a synapomorphy uniting M. octosetaceus with the three remaining species (Fig. 1); this stripe was shown as conspicuous in M. octosetaceus by Partridge (1956) and Delacour (1959), but it was not indicated in photographs included by Bartmann (1988). If the latter state is taken as representative, then the change in character 4 supports the monophyly of only M. serratot, M. squamatus, and M. merganser, but the topology shown (Fig. 1) is unchanged. The sister clade of M. octosetaceus comprised the more "typical" mergansers of the northern hemisphere; of these three species, M. serratot and M. squamatus appear to be closest relatives (Fig. 1), although this relationship is supported by only one synapomorphy, and the topology for these three species conservatively might be considered a trichotomy. Based on this preliminary analysis, M. australis represents an early but unique- ly derived branch in the merganser clade and, although comparable in grade to M. octose- taceus, the two southern-hemisphere species are not closest relatives. MORPHOMETRIC COMPARISONS Univariate comparisons of external measurements.--A substantial range of body size and a diversity of shape was indicated by comparisons of skin characters of mergansers. Total length of the body, probably the best directly measurable "size" measure which is independent of wing size and available for all species, differed significantly among species (F = 35.75; df = 5, 84; P < 0.0001) and ranked L. cucullatus as the smallest, followed (in order of increasing size) by M. australis, M. octosetaceus, M. serrator, M. squamatus, and M. merganser (Table 2); males were larger than females within species (F = 10.43; df = 1, 84; P < 0.005). Available mean body masses--617 g (N = 97) for L. cucullatus, 984 g (79) for M. serrator, and 1382 g (197) for M. merganser--confirmed these size rank- ings. "Body lengths" (total lengths minus lengths of culmen and tail) mirrored these interspecific (F = 27.35; df = 5, 83; P < 0.0001) and intersexual differences (F = 5.68; df = 1, 83; P < 0.05). Regressions of mean body masses on mean "body length" (log-transformed data, N = 3, r = 0.988) provided rough estimates of body mass for the remaining three species: M. australis, 898 g; M. octosetaceus, 983 g; and M. squa- matus, 1234 g. Comparison of actual and estimated body masses with the phylogenetic tree (Fig. 1) indicates that the merganser clade is char- acterized by a strong evolutionary trend toward increased body size. Tarsus length differed significantly among species (F = 195.40; df = 5, 87; P < 0.0001) and between the sexes (F = 157.61; df = 1, 87; P < 0.0001), and conformed closely with body length in interspecific rankings Livezey ß AUCKLAND ISLANDS MERGANSER 4 15 TABLE 1 CHARACTERS AND CHARACTER STATES USED IN PHYLOGENETIC TREE DEPICTED IN FIG. 1 Group Character States  Natal plumage (1) Breast band a = present; b = absent (2) Flank spots a = confined by dark; b = open ven- trally (3) Head color a = gray; b = reddish brown (4) Subocular stripe a = absent; b = present (5) Distal forewing a = dark; b = white (6) Cheek patch a = large, above gape; b = small, ven- tral to gape (7) White back spots a = present; b = absent (8) White preorbital spots a = absent; b = present (9) Breast color a = gray or white; b = tawny washed (10) Dorsal spotting a = distinct; b = vestigial (11) Chestnut subocular spot a = none; b = present (12) Chin a = dark; b = whitish (13) Breast (males) a = white; b = dark red (14) Dark nape stripe (males) a = absent; b = present (15) Bill color b a = grayish; b = orange; c = orange with culmen and nail black (16) Sexual dichromatism b a = present; b = absent, males like fe- males; c = absent, unique plumage (17) a = white; b = red; c = gray Adult plumage (definitive alternate) Background color of sides, flanks (males) b (18) Barring of sides (males) b (19) Barring of sides (20) Crest b (21) Crest (22) Ventral barring a = black; b = white; c = none a = linear vermiculations; b = scal- loping a = 1-parted, from crown; b = emerges continuously from crown to occiput; c = 2-parted a = present in both sexes; b = absent in males a = absent; b = present Primitive state = a, derived states b-c. Analyzed as unordered. (Table 2); middle-toe lengths (not tabulated) mirrored tarsus lengths in interspecific and intersexual differences. Culmen length, however, de- viated markedly from this pattern. M. australis, one of the smallest species overall, had the largest culmen lengths, followed by M. serrator, the ap- proximately equal M. merganser and M. squamatus, M. octosetaceus, and the much shorter-billed L. cucullatus (Table 2); both interspecific (F = 131.29; df = 5, 87; P < 0.0001) and intersexual differences (F = 58.99; 416 THE WILSON BULLETIN ß Vol. 101, No. 3, September 1989  1/2. 1/2- 1/2- 1/2. FIO. 1. Phylogenetic tree for Recent mergansers (Lophodytes and Meres). Chacters (numbers) and states (letters) are defined in Table 1 and compiled in the Appendix; black recnes indicate skelel synamohies deschbed in Livezey (1986). df = 1, 87; P < 0.0001) in culmen length were significant. Nail width (not tabulated) also differed significantly among species (F = 15.19; df = 5, 85; P < 0.0001) and sexes (F = 26.28; df = 1, 85; P < 0.0001); this variable tended to follow body length in interspecific rankings, except that M. australis and especially L. cucullatus had large nail widths relative to their body size. Wing lengths showed substantial interspecific (F = 376.44; df = 5, 87; P < 0.0001) and intersexual variation (F = 123.78; df = 1, 87; P < 0.0001), and although means tended to follow total body length in inter- specific rankings, M. australis had disproportionately short wings (Table 2); body lengths ofM. australis averaged 67 mm longer than those of L. cucullatus, but wing lengths of the former averaged 8 mm shorter. All specimens ofM. australis retained the full complement of primary remiges found in most anseriforms and other mergansers (10 functional plus 1 reduced). Tail lengths also varied significantly among species (F = 58.90; df = 5, 86; P < 0.0001), and paralleled wing lengths in interspecific differences; M. australis had disproportionately short tails (Table 2). Wing-loadings of the Mergini. -- Wing areas were available for all species Livezey ß AUCKLAND ISLANDS MERGANSER 417 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 418 THE WILSON BULLETIN ß Vol. 101, No. 3, September 1989 I WA = 1.32 + (0.74_+ 0.08)MA S. mollissimaO ZO  [2= 935 M. merganser M. fusca spectabilis  B. /s/and/ca J iS. œ1$cheri  M. perspcllata     M. squamatus  .4 M. serrator l .2 H. histrion/cs ellri '   C. h, emal,'s . albeola M. albellus  M. octos*taceus .0 /  I -M austral,'s .8  L. cucullatus .0 . .4 . .8 0 7. 7,4 7. k (Body Moss) FIG. 2. Logadthmic plot of mean wing areas on me body mass for extant Merni; meansers shown as squares, other scies as circles. Recession equation {geometdc-mmn fit) excludes three estimated points for Meres {hollow squares); slo ven 2 stdard eors. of sea duck except the extinct Labrador Duck (Camptorhynchus labra- dorius) and three species of Mergus, including M. australis. For the latter, wing areas were estimated using a regression of mean wing areas on mean wing lengths for the three species of merganser for which both data were available (N = 3, log-transformed data, r -- 0.999). These estimates were 371 cm 2 for M. australis, 450 cm 2 for M. octosetaceus, and 722 cm 2 for M. squamatus; together with the corresponding estimates for body mass (detailed above), rough estimates of wing-loadings (õ.cm -2) were 2.42, 2.18, and 1.71, respectively. The estimates for M. octosetaceus and es- pecially M. australis were high relative to those for other Mergus (un- weighted mean of species = 1.67, 2 species, N = 4 wing areas), Lophodytes (1.50, I species, N = 6), Mergellus (1.52, I species, N = 3), Bucephala (1.50, 3 species, N = 20), Clangula (1.76, 1 species, N = 5), Melanitta (1.62, 3 species, N = 18), Histrionicus (1.32, I species, N = 8), $omateria (1.93, 3 species, N = 6), and Polysticta (1.68, 1 species, N = 1). A logarithmic plot of mean wing areas and body masses of 16 extant Livezey ß AUCKLAND ISLANDS MERGANSER 419 2 I SKINS  t- CONTRAST. ........... ............. ............. Y   / / / ß CANONICAL VARIATE  zG. . ]ot of Qrst tree canonical adates of see extea] measurements of sets. e scores of species-sex oups are plotted; males e represented b solid drc]es, /eraales b oHo circles. Oblique dashed Js abe]]ed "contrast" depicts te approximate odetatJo of coJc contrast betee . trlis d oter spedes. Mergini revealed a strong allometric relationship between the variables (Fig. 2), one which was not significantly different (P > 0.05) from that preserving geometric similarity (slope relating an area with mass expected to be 0.67; Gould 1966). the estimated points for the three Mergus lacking data were plotted but were not included in the regression. The pronounced negative deviation of M. australis (Euclidean distance from line -- 0.35) and M. octosetaceus (distance = 0.25) from the tribal line reflects their relatively heavy wing-loadings (Fig. 2). The clarity of the allometric re- lationship indicates that wing-loadings were largely a function of body mass (with the possible exception of the two austral Mergus); there was little clustering of congeners, including most mergansers. Canonical analysis of external measurements.--A CA of external mea- surements produced significant dispersion of species-sex groups (Wilks' lambda = 0.0002; df-- 6, 11, 87; P < 0.001). Stepwise MANOVAs revealed that these multivariate differences involved significant interspe- cific (Wilks' lambda = 0.11; df = 3, 1, 87; P < 0.001) and intersexual (Wilks' lambda = 0.38; df = 2, 1, 87; P < 0.001) effects. The first three canonical variates together accounted for over 95% of the total intergroup dispersion and included those dimensions in which M. australis differed from the other taxa. The first canonical variate (CV-I) encapsulated roughly two-thirds of the total intergroup dispersion in external measurements (Table 3, Fig. 420 THE WILSON BULLETIN ß Vol. 101, No. 3, September 1989 TABLE 3 STANDARDIZED COEFFICIENTS AND SUMMARY STATISTICS FOR CANONICAL VARIATES OF EXTERNAL MEASUREMENTS SEPARATING SPECIES AND SEXES OF MERGANSERS All-group analysis Contrast of Character CV-I CV~II CV-III Mergus australis Total length -0.06 -0.09 0.28 0.04 Culmen length 0.16 -0.77 0.06 0.65 Nail width 0.44 -0.01 0.58 0.43 Wing length -0.98 0.55 -0.46 - 1.10 Tail length 0.11 0.23 0.76 -0.03 Tarsus length -0.39 -0.41 0.17 -0.03 Middle-toe length 0.02 -0.10 0.34 0.07 Eigenvalue 34.3 13.0 2.3 13.6 Variance (%) 67.0 25.3 4.5 -- Canonical R 0.99 0.96 0.83 0.97 3) and involved both interspecific (F -- 572.43; df = 5, 87; P < 0.0001) and intersexual differences (F = 90.50; df = 1, 87; P < 0.0001; ANOVA of scores). Canonical coefficients indicate that CV-I primarily contrasted lengths of the wing and tarsus with nail width (Table 3); positions of species and sexes on this axis shows that CV-I is strongly inversely cor- related with overall body size (Fig. 3). Large M. merganser had the lowest scores on CV-I, followed closely by M. squamatus and M. serrator, and ultimately the high-scoring M. octosetaceus, M. australis, and L. cucul- latus; males had lower scores than females in all species (Fig. 3). CV-II described a strikingly different multivariate axis, one involving strong interspecific (F = 215.83; df = 5, 87; P < 0.0001) and less pro- nounced intersexual differences (F = 7.46; df = 1, 87; P < 0.01; ANOVA of scores). This variate contrasted lengths of the culmen and tarsus with the lengths of the wing and tail (Table 3). M. australis had the lowest scores on CV-II, indicating that this species had long bills and tarsi relative to their wing and tail lengths. L. cucullatus had opposite proportions, and the remaining Mergus had intermediate scores; males tended to have slightly lower scores than females within species (Fig. 3). The third canonical variate (CV-III) for external measurements incor- porated both interspecific (F = 27.96; df = 5, 87; P < 0.0001) and in- tersexual differences (F = 39.20; df = 1, 87; P < 0.0001; ANOVA of scores). CV-III contrasted wing length with other measurements, espe- cially nail width, tail length, and middle-toe length (Table 3). M. serrator had comparatively low scores on this axis, indicating that this species had Livezey ß AUCKLAND ISLANDS MERGANSER 421 long wings relative to their tail lengths, nail widths, and (to a lesser degree) other external measurements (Fig. 3). The high scores of M. merganser and M. octosetaceus reflected their opposite proportions, and other species were intermediate on this axis; males exceeded females in all species on CV-III. A canonical contrast between M. australis and the other species using external measurements provided good separation of the insular form from other mergansers (F = 382.30; df = 3, 85; P < 0.0001). Canonical coef- ficients show that this contrast primarily contrasted wing length with culmen length and nail width (Table 3). Positions of species-sex groups along the contrast indicate that the axis was oriented diagonally with respect to canonical variates I and II and that M. australis was distin- guished most clearly by its relatively short wings (Fig. 3). Univariate comparisons of skeletal measurements.--With few excep- tions, means of skeletal measurements followed overall body size (total length or body mass) in interspecific rankings (Table 4); M. squamatus (not tabulated), represented by a single skeleton, was intermediate between M. serrator and M. merganser in skeletal measurements. As in external measurements, males exceeded (P _< 0.01) females in skeletal dimensions (Table 4). Several skeletal measurements, however, deviated from the general interspecific size rankings: bill length, scapula length, sternal cau- dal width, and sternal carina depth (Table 4). Skeletal bill lengths of M. australis were disproportionately large, exceeding those of the generally larger M. serrator, M. squamatus, and M. merganser. Scapula lengths, sternal caudal widths, and sternal carina lengths ofM. australis, however, were the smallest of all the species compared, including L. cucullatus (Table 4). Intra-alar skeletal proportions.--Mean skeletal wing lengths differed among species (F = 748.48; df = 3, 70; P < 0.0001) and between sexes (F = 59.06; df = 1, 70; P < 0.0001) and followed interspecific rankings in body size (œ + s, n): L. cucullatus (191.5 _+ 6.4, 24), M. australis (194.6 _+ 6.3, 3), M. serrator (245.9 _+ 9.9, 25), M. squamatus (250.6, 1), and M. merganser (268.6 + 15.8, 26). However, the proportions of skeletal wing length contributed by constituent elements differed (ANOVAs of transformed proportions) among species, many of these differences being attributed to the unusual proportions ofM. australis and, to a lesser degree, overall size; intersexual differences in proportions were not significant in any element (P > 0.10). Humeral proportions differed among species (F = 17.14; df = 4, 74; P < 0.0001), constituted an average of 35.7% of skeletal wing length in M. australis, but accounted for only 34.1-34.9% of the wing in the other species compared. Ulnar proportions increased with overall size of species (F = 14.65; df = 4, 74; P < 0.0001), ranging 422 THE WILSON BULLETIN ß Vol. 101, No. 3, September 1989 Livezey ß AUCKLAND ISLANDS MERGANSER 423 from 28.4% in L. cucullatus to 29.0% in M. merganser. Intra-alar pro- portions of the carpometacarpus (F = 21.19; df = 4, 74; P < 0.0001) and proximal (first) phalanx of digit-II (F = 14.65; df = 4, 74; P < 0.0001) opposed those of the humerus, being smaller in M. australis (20.2% and 8.1%, respectively) than in the other species (20.4--20.9% and 8.5-8.7%, respectively). The distal (second) phalanx of digit-II had equal (F = 0.07; df-- 4, 74; P > 0.95) proportions in all species (7.3%). Canonical analysis of complete skeletons.--Suites of 33 skeletal mea- surements provided the basis for powerful separation of the eight species- sex groups available (Wilks' lambda = 0.0002; dr= 10, 7, 83; P < 0.001); the single skeleton ofM. squamatus was plotted on the axes based on the other three species. Stepwise MANOVAs documented significant inter- specific (Wilks' lambda = 0.053; df = 13, 1,83; P < 0.001) and intersexual differences (Wilks' lambda = 0.29; df = 5, 1, 83; P < 0.001) in these 33- dimensional comparisons. The first three canonical variates accounted for over 95% of the total intergroup variation in skeletons (Table 5). The first canonical variate (CV-I) displayed significant interspecific (F = 5079.52; df = 3, 83; P < 0.0001) and intersexual differences (F = 46.70; df = 1, 83; P < 0.0001; ANOVA of scores) but primarily discriminated Lophodytes from Mergus; M. australis was slightly more similar to Lo- phodytes than its congeners on this axis (Fig. 4). Canonical coefficients for CV-I indicate that Lophodytes differed from other merõansers on a complex size-shape axis involving relative proportions of the skull, among elements within the wing and leg, and among dimensions of the pectoral and pelvic girdles (Table 5). Notable proportionalities of Lophodytes in- dicated by the coefficients were: skulls characterized by relatively short bills and long, deep crania; wings with relatively short ulnae and radii; and legs with relatively large femora. Within each species, males and slightly lower scores than females on CV-I (Fig. 4). CV-II for complete skeletons accounted for an additional 7% of the total intergroup dispersion and included both interspecific (F = 346.83; df-- 37.41; P < 0.0001) and intersexual differences (F = 37.41; df = 1, 83; P 0.0001; ANOVA of scores). Mean scores (Fig. 4) indicate that CV-II, together with CV-III (discussed below) primarily discriminated M. aus- tralis from the other groups (including Lophodytes). Because of the small sample available for M. australis, interpretation of multivariate differ- ences between M. australis and other mergansers must be conservative. Coefficients for CV-II (Table 5), together with the low scores of M. aus- tralis, primarily indicate that the species was characterized by relatively long bills, relatively short mid-wing elements, legs with relatively robust proximal elements and long tarsometatarsi, and sterna with shallow ca- rinae. 424 THE WILSON BULLETIN ß Vol. 101, No. 3, September 1989 TABLE 5 STANDARDI7_ED COEFFICIENTS AND SUMMARY STATISTICS FOR CANONICAL VARIATES OF SKELETAL DIMENSIONS SEPARATING SPECIES AND SEXES OF MERGANSER All-group analysis Contrast of Character CV-I CV-II CV-III M. australis Bill length -0.48 -0.65 0.15 -0.68 Cranium length 0.55 0.07 0.38 -0.08 Height 0.64 -0.03 0.03 0.03 Width -0.38 0.12 -0.22 0.18 Humerus length 0.18 - 0.26 - 1.09 0.36 Head width 0.34 0.35 -0.13 0.39 LWM 0.03 -0.26 -0.28 -0.06 Radius length - 0.51 - 0.38 0.13 - 0.38 LWM -0.05 0.17 -0.27 0.29 Ulna length -0.66 0.24 -0.33 0.25 LWM -0.48 0.26 0.16 0.08 Carpometacarpus length 0.60 0.87 0.25 0.64 Digit-II, Ph. 1 length -0.08 0.39 -0.53 0.60 Ph. 2 length -0.23 -0.53 0.23 -0.58 Femur length 0.65 0.05 0.76 -0.29 Head width -0.68 -0.39 0.33 -0.59 LWM 0.25 0.23 0.03 0.23 Tibiotarsus length -0.38 0.07 0.07 0.01 LWM -0.43 -0.34 0.03 -0.36 Tarsometatarsus length -0.41 -0.65 -0.23 -0.47 LMW -0.01 -0.23 0.39 -0.39 Digit-III length 0.30 0.06 0.15 0.02 Scapula length 0.27 0.14 - 0.14 0.21 Coracoid length - 0.31 0.24 0.52 - 0.08 Basal width 0.20 0.28 0.08 0.23 Stemal carina length 0.14 0.10 0.07 0.05 Basin length -0.20 0.19 -0.07 0.18 Least width 0.03 -0.29 0.28 -0.39 Caudal width 0.15 0.14 - 0.13 0.20 Carina depth -0.00 0.41 0.02 0.37 Furcula height - 0.16 - 0.08 0.39 - 0.30 Synsacrum length -0.15 -0.06 0.07 -0.09 Intemcetabular width 0.28 0.13 0.35 -0.05 Eigenvalue 186.8 15.2 11.7 12.7 Variance (%) 84.1 6.8 5.3 -- Canonical R 1.00 0.97 0.96 0.96 Livezey ß AUCKLAND ISLANDS MERGANSER 425 Fio. 4. Plot of first three canonical variates of 3 3 skeletal measurements of mergansers. Mean scores of species-sex groups are plotted; males are represented by solid circles, females by hollow circles. Oblique dashed axis labelled "contrast" depicts the approximate orien- tation of canonical contrast between M. australis and other species. The third axis (CV-III) for complete skeletons also incorporated sig- nificant interspecific (F = 227.47; df = 3, 83; P < 0.0001) and intersexual differences (F = 126.26; df = 1, 83; P < 0.0001; ANOVA of scores) but mostly represented residual differences between M. australis and other species not accounted for by CV-II; CV-III also provided a lesser dis- crimination of M. serrator (Fig. 4). Coefficients for CV-III, together with the high scores of M. australis, reflected the small size of most wing elements relative to the lengths of the femur, coracoid, furcula, and in- teracetabular width (Table 5). The fourth canonical variate (not tabulated) accounted for another 3.2% of the intergroup variance but served largely to separate further M. serrator from its congeners (F = 73.03; df = 3, 83; P < 0.0001) and to discriminate sexes (F = 97.04; df = 1, 83; P < 0.0001) within species (ANOVA of scores). The remaining three variates for complete skeletons together in- corporated less than 1% of the total dispersion, contributed mostly to intersexual discrimination, and are not considered further. The distinctive skeletal proportions of M. australis were summarized in a canonical contrast between this species and the other species, which 426 THE WILSON BULLETIN ß Vol. 101, No. 3, September 1989 TABLE 6 STANDARDIZED COEFFICIENTS AND SUMMARY STATISTICS FOR CANONICAL VARIATES OF STERNAL DIMENSIONS SEPARATING SPECIES AND SEXES OF MERGANSER All-group analysis Contrast of Character CV-I CV-II CV-11I M. australis Carina length -0.45 -0.40 - 1.06 0.18 Basin length -0.23 -0.29 1.04 -0.27 Least width -0.12 -0.39 -0.54 0.40 Caudal width -0.01 0.22 0.90 -0.43 Carina depth -0.49 0.90 0.11 -0.92 Eigenvalue 38.6 2.5 ! .3 3.2 Variance (%) 90.6 5.9 3. ! -- Canonical R 0.99 0.85 0.76 0.87 maximally distinguished M. australis from all other groups independently of other interspecific differences and all intersexual differences (F = 51.00; df = 15, 69; P < 0.001). Mean scores of species and coefficients of variables for the axis showed the contrast to be largely a linear combination of CV- II and CV-III in the all-group analysis, with a lesser contribution from CV-I (Fig. 4). The coefficients for the contrast (Table 5) and the low scores ofM. australis on this axis primarily reflected the following relative pro- portions: long bills; wing bones generally small, specifically the humerus, carpometacarpus, and proximal phalanx of the major digit; proximal and distal leg elements disproportionately long; and sterna with relatively shallow carinae and small caudal widths. Canonical analysis ofsternal measurements. --A separate CA of the five sternal measurements permitted the plotting of the single specimen of M. octosetaceus and a closer examination of this important and relatively complex part of the pectoral girdle. The analysis documented significant differences among groups (Wilks' lambda = 0.003; df = 5, 7, 81; P < 0.001), and included both interspecific (Wilks' lambda = 0.39; df = 5, 1, 81; P < 0.001) and intersexual differences (Wilks' lambda = 0.38; df = 2, 1, 81; P < 0.001). Interspecific differences were significant on each of the first three variates (F = 895.73, 64.28, and 30.71, respectively; df = 3, 81; P < 0.0001), but sexual differences were limited to the first variate (F = 134.95; df= 1, 81; P.< 0.00001; ANOVA of scores). Coefficients of CV-I for sterna were of the same sign, suggesting that the axis essentially reflected "sternal size" (Table 6); differences in mag- nitude among coefficients, however, show that this was not isometric size and that widths contributed little to this axis. CV-I primarily separated Livezey ß AUCKLAND ISLANDS MERGANSER 427  H o STERNA CONTRAST _, / FiG. 5. Plot of first three canonical vahates of five stemal measurements of mersers. Mean scores of species-sex groups e plotted; males are represented by solid circles, females by hollow circles. Oblique shed axis labelled "contrast" depicts the approximate ohen- tation of conical contrast between M. atralis and other species. the three smaller, "primitive" species (L. cucullatus, M. australis, and M. octosetaceus) from the larger, more derived species and distinguished sexes within species (Fig. 5). The two austral species were poorly represented, however, and multivariate inferences must be made conservatively. CV- II clearly separated M. australis and M. octosetaceus from the other mer- gansers (Fig. 5); coefficients indicate that the former were characterized by sterna with relatively shallow carinae and small caudal widths (Table 6). CV-III provided further separation between M. australis and other mergansers, including M. octosetaceus (Fig. 5), and primarily reflected the small differences between lengths of the carina and basin and between the two sternal widths in M. australis (Table 6). A canonical contrast of M. australis (Wilks' lambda = 0.24; df = 3, 1, 81; P < 0.001) summarized these sternal differences, and combined the reduced carina depth with the "squarish" ventral and lateral aspects of the sterna of M. australis indi- cated in CV-III of the all-group CA (Table 6, Fig. 5). DISCUSSION Phylogenetic and biogeographic inferences. --The evolutionary tree pro- posed (Fig. 1) differs from previous hypotheses concerning M. australis; earlier workers either considered the relationships of M. australis to be unresolved (Johnsgard 196 la) or suggested that M. australis was an iso- lated relative of an extant northern-hemisphere congener, most probably M. squarnatus or M. rnerganser (Humphrey 1955, Johnsgard 1965, Kear and Scarlett 1970, Weller 1980). The tree presented in this paper also 428 THE WILSON BULLETIN ß Vol. 101, No. 3, September 1989 suggests several unexpected historical patterns. First, there is a pro- nounced trend toward increasing body size, and the relatively small size of M. australis is apparently the direct result of phylogeny rather than a secondary (derived) decrease in size. Second, M. australis and M. octo- setaceus are representatives of an early grade of mergansers limited to the southern hemisphere. Although Mergus is known from subfossil remains on the mainland of New Zealand, the phylogenetic relationship of these to M. australis is not known. It should be noted that a comparatively long (82-mm) subfossil humerus from South Island, New Zealand, attributed by Kear and Scarlett (1970) to a possibly larger congener ofM. australis, was reidentitled by Olson (1977) as that of a coot (Fulica). Third, mor- phological similarities between M. australis and M. octosetaceus are the result of shared primitive characters (symplesiomorphies; e.g., head color of downy young) or convergence (homoplasies; e.g., sexual monochro- matism), and do not reflect close relationship. Fourth, many of the unique plumage characters ofM. australis are uniquely derived (autapomorphies) and, although not useful for determining relationships, are informative regarding the evolutionary characteristics of this peculiar insular form. Morphometrics ofmergansers.--Mergansers, like all sea ducks, are typ- ical power-fliers, using rapid wing beats in normal flight (Meinertzhagen 1955) and being characterized by moderately high wing-loadings and comparatively pointed wings with reduced slotting between remiges (Rai- kow 1973; Greenewalt 1962; Liveze. y and Humphrey 1986, this study). A direct relationship between wing-loading and body size in the Mergini (Fig. 2), a simple outcome of the more rapid increases in mass (a three- dimensional variable) than wing area (a two-dimensional variable), gen- erally produces the greatest wing-loadings in the largest species. A similar allometry of wing area with body mass pertains to birds generally (Poole 1938, Hartman 1961, Greenewalt 1962). Projected estimates for M. aus- tralis and M. octosetaceus, however, suggest that these species possess(ed) wing-loadings comparatively high for the Mergini (Fig. 2). The estimated wing-loading for M. australis, roughly 2.4 g. cm -2, is especially high (even allowing for possible over-estimation) and approximates the threshold of flightlessness (2.5 g. cm -2) hypothesized by Meunier (1951). The estimate for M. australis is comparable to the wing-loadings of Flying Steamer- Ducks (Tachyeres patachonicus), a species in which flight is heavy in most individuals and impossible for the heaviest males in some marine pop- ulations (Humphrey and Livezey 1982). Consequently, it is noteworthy that the M. australis observed in flight by Lord Ranfurly was, based on its vocalizations, a female (Kear and Scarlett 1970). I suspect that the heaviest males ofM. australis, if not permanently flightless, were able to take flight only when aided by the lift provided by a strong head-wind Livezey ß AUCKLAND ISLANDS MERGANSER 429 (presumably common in the Aucklands), and with substantial room for take-off (on water) or from a decline (on land). Statistical comparisons of external measurements not only confirmed the comparatively small size and relatively short wings of M. australis but revealed that the species possessed relatively short tails and excep- tionally long bills (Tables 2, 3; Fig. 3). A significant portion of the mor- phometric differences among mergansers, however, was related to phy- logeny in that L. cucullatus, M. australis, and M. octosetaceus were more similar (symplesiomorphically) to each other than to other Mergus on the most important axis of variation (CV-I, Fig. 3). M. octosetaceus was most similar to M. australis in its external dimensions (Fig. 3). M. octosetaceus is specialized for life along fast-flowing tropical rivers (Partridge 1956), and M. australis is believed to have inhabited primarily riverine habitats in the Auckland Islands (Delacour 1959, Kear and Scarlett 1970). There- fore, although the riverine habitats of M. octosetaceus and M. australis differed greatly in ecological characteristics, morphometric similarity be- tween the austral species of Mergus may reflect, in part, convergence associated with similar feeding habits. Analyses of skeletal measurements confirmed the long bills and rela- tively short wing elements of M. australis as well as a diversity of other, less obvious differences in proportions (Tables 4, 5; Fig. 4). Absolute length of the wing skeleton in M. australis was very similar to that in L. cucullatus; lengths of the manus (carpometacarpus and major digit) were equal in the two species, indicating that the shorter external wing lengths of M. australis result from a roughly 8-mm shortening of the primary remiges. M. australis also was characterized by disproportionately short mid-wing elements and relatively weakly keeled, caudally flared sterna (Table 6; Figs. 4, 5). Disproportionately short mid-wing elements and (particularly) reduced sternal carinae are typical of most flight-impaired and flightless species ofcarinate birds, including those of grebes (Podicipedidae; Livezey 1989), steamer-ducks (Anatidae, Tachyeres; Livezey and Humphrey 1986), and rails (Rallidae; Olson 1973). Although not represented by a complete skeleton, M. octosetaceus was most similar to M. australis in sternal conformation (Fig. 5); given the more typical sterna ofL. cucullatus and the remaining Mergus, the similarity between the two austral species is evidently convergent. Paedomorphosis in M. australis.--Paedomorphosis, "... retention of ancestral juvenile characters in the descendant adult phase ..." (Mc- Namara 1986), is the result of the alteration of developmental patterns; a frequently cited form of such heterochrony is neoteny, the retardation of the growth of a body part relative to general developmental state (Gould 1977). Neoteny of the pectoral girdle has been inferred as the ontogenetic 430 THE WILSON BULLETIN ß Vol. 101, No. 3, September 1989 basis for the "degenerate" conditions of sterna and wing elements of a number offlightless birds (Lowe 1928, Olson 1973, Diamond 1981, James and Olson 1983) as well as the more subtle changes in skeletal proportions of flightless steamer-ducks (Livezey and Humphrey 1986). The wing skel- eton and sternum, and the short primary remiges of M. australis, could be characterized similarly. Both sexes ofM. australis have adult plumages that closely resemble that of juveniles ofmergansers generally, and juvenal plumages of Mergus can be presumed to be primitive. Therefore, by definition, the adult plumage of male M. australis also is paedomorphic; perhaps the condition might be more precisely termed "paedochromatic." Whether this plumage change is the result of neoteny is not known, but such a developmental mechanism may underlie the reduced sexual di- chromatism characteristic of many insular populations of anatids (Lack 1970, Weller 1980). Morphological and ecological correlates of insularity in M. australis.-- Weller (1980) described a number of unusual phenotypic features, averred to represent adaptive changes from the ancestral condition, which tend to characterize insular waterfowl: (1) reduced mobility, sometimes flight- lessness; (2) reduced body size; (3) reduced sexual dichromatism; (4) de- velopment of white eye-rings; (5) comparatively dark natal plumages; (6) exceptional tameness of adults; (7) niche expansion, often to include ma- rine habitats; (8) relaxed seasonality of nesting; (9) long-term pairbonds; (10) small clutch size; and (11) large egg size. Of these, four clearly char- acterized M. australis--reduced flight capacity, reduced sexual dichro- matism, dark natal plumages, and tameness (the last based on a virtual lack of escape behavior observed by early collectors). The small size of M. australis conforms with the general pattern among insular ducks, but it appears to have been the result ofphylogeny (i.e., the species represents a branch from the early grade of smaller mergansers; Fig. 1) instead of representing a decrease in size from a larger "continental" ancestor. The small size ofM. australis is notable, however, in that increased body size is typical ofavian lineages undergoing the loss of right (Pennycuick 1975). Two other correlates of insularity-- increased utilization of marine hab- itat and long-term pairbonds--find weak support in M. australis. Al- though it is believed to have been primarily reliant on freshwater rivers, M. australis were observed on salt water near river mouths, and specimens have contained marine organisms (Kear and Scarlett 1970); given that most mergansers have largely freshwater habitats (the sole exception being M. serratoe), this evidence suggests a comparatively broad, presumably derived, feeding 'niche. The disproportionately long bills of M. australis also may provide evidence of a diversified, more marine diet; for example, M. serrator, a comparatively marine species of moderate size, has a longer Livezey ß AUCKLAND ISLANDS MERGANSER 431 bill than the larger freshwater M. merganser (Table 2). Existence of long- term pair bonds in M. australis is suggested by the presence of both adults with broods and the observation of paired adults from October to (per- haps) July (Kear and Scarlett 1970). The development of a white eye-ring clearly does not apply to M. australis, and the remaining three correlates-- relaxed nesting schedule, decreased clutch size, and increased egg size-- cannot be evaluated for M. australis because of lack of information. Significance of fiight impairment in M. australis.-- It appears likely that adults and especially the downy young of M. australis were vulnerable, even prior to the arrival of humans, to a variety of avian and subaquatic predators, including Skuas (Catharacta skua), Peregrine Falcons (Falco peregrinus), predaceous fish, eels, and sea lions (Kear and Scarlett 1970). Although the introduction of exotic predators to the Auckland Islands, notably cats and pigs, posed a significant threat to the endemic birds of the Auckland Islands (Kear and Scarlett 1970), the reduced flight capacity ofM. australis cannot be considered an evolutionary "response" to a lack of predators. Reduced predation pressure is invoked frequently in dis- cussions of flightlessness in insular birds (e.g., Snow 1966; Weller 1975, 1980); paradoxically, concealment from predators is proposed by some to explain, at least in part, the reduced sexual dichromatism and dark natal plumages of some of these same species (Sibley 1957, Kear and Scarlett 1970, Lack 1970, Weller 1980). It appears that the evolutionary reduction or loss of flight capacity in M. australis, a species which did not use its wings in diving, reflects instead the developmentally econom- ical reduction of the pectoral apparatus rendered of decreased selective importance in environments permitting year-round habitation. ACKNOWLEDGMENTS This research was funded by National Science Foundation grant BSR-8516623 and by visitation grants awarded by the American Museum of Natural History and U.S. National Museum of Natural History. I am grateful to B. Gillies, A. Isles, G. Mack, and H. Levenson for their generosity, and to P.S. Humphrey for numerous discussions and provision of unpublished data. Curatorial personnel of the following institutions permitted access to specimens: American Museum of Natural History, New York; Museum of Natural History, University of Kansas, Lawrence; National Museum of Natural History, Washington, D.C.; Museum 0fZoology, University of Michigan, Ann Arbor; British Museum (Natural History), Tring, England; Royal Ontario Museum, Toronto, Ontario, Canada; San Diego Museum of Natural History, San Diego; Museum of Vertebrate Zoology, University of California, Berke- ley; Field Museum of Natural History, Chicago; Carnegie Museum of Natural History, Pittsburgh; Australian Museum, Sydney, Australia; New Zealand National Museum (Natural History), Wellington, New Zealand; and Otago Museum, Dunedin, New Zealand. Data from two skins ofM. australis were provided by the staffof the Staatliche Museen for Tierkunde and Volkerkunde, Dresden, East Germany. P.S. Humphrey, C. R. Blem, S. L. Olson, and three anonymous reviewers commented on the manuscript, and K. Corbin and M. Schmalz typed its several drafts. 432 THE WILSON BULLETIN ß Vol. 101, No. 3, September 1989 LITERATURE CITED BARTMANN, W. 1988. New observations on the Brazilian Merganser. Wildfowl 39:7-14. BROOKS, A. 1945. The under-water actions of diving ducks. Auk 62:517-523. CLARK, R.J. 1971. Wing-loading--a plea for consistency in usage. Auk 88:927-928. DELACOUR, J. 1959. The waterfowl of the world. Vol. 3. Country Life Ltd., London, England. DEMENT'EV, G. P. AND N. A. GLADKOV (EDS.). 1967. Birds of the Soviet Union. Vol. IV. Israel Progr. Sci. Trans., Jerusalem, Israel. DIAMOND, J.M. 1981. Flightlessness and fear of flying in island species. Nature 293:507- 508. DIXON, W. J. (CHIEF ED.). 1985. BMDP statistical software. Univ. California Press, Berke- ley, California. FP,NE, J. 1985. PAM: description and estimation of missing data. Pp. 217-234 in BMDP statistical software (W. J. Dixon, chief ed.). Univ. California Press, Berkeley, California. GOULD, S.J. 1966. Allometry and size in ontogeny and phylogeny. Biol. Rev. 41:587- 640. 1977. Ontogeny and phylogeny. Belknap, Cambridge, Massachusetts. GP,3/4, G.R. 1844. The genera of birds. Vol. 3. Longroans, London, England. GREENEWALX, C.H. 1962. Dimensional relationships for flying animals. Smithsonian Misc. Coil. 144:1-46. GREENWAY, J. C., JR. 1967. Extinct and vanishing birds of the world. 2nd ed. Dover Publ., New York, New York. HARXM^N, F.A. 1961. Locomotor mechanisms of birds. Smithsonian Misc. Coll. 143: 1-91. HUMPHREY, P.S. 1955. The relationships of the sea ducks (Tribe Mergini). Ph.D. diss., Univ. Michigan, Ann Arbor, Michigan. ß 1957. Observations on the diving of the SurfScoter (Melanitta perspicillata). Auk 74:392-394. 1958. Diving of a captive Common Eider. Condor 60:408-4 10. -- AND B.C. LIVEZEY. 1982. Flightlessness in Flying Steamer-Ducks. Auk 99:369- 372. JAMES, H. F. AND S. L. OLSON. 1983. Flightless birds. Natur. Hist. 92:30-40. JENNmCH, R. AND P. SAMPSON. 1985. P7M: stepwise discriminant analysis. Pp. 519-537 in BMDP statistical software (W. J. Dixon, chief ed.). Univ. California Press, Berkeley, California. JOHNSGARD, P.A. 1960. Classification and evolutionary relationships of the sea ducks. Condor 62:426-433. 1961 a. The taxonomy of the Anatidae--a behavioral analysis. Ibis 103a:71-85. 196 lb. The behavior and systematic position of the Hooded Merganser. Wilson Bull. 73:227-236. ß 1965. Handbook of waterfowl behavior. Constable & Co., London, England. 1978. Ducks, geese, and swans of the world. Univ. Nebraska Press, Lincoln, Nebraska. JOLICOEUR, P. 1963. The multivariate generalization of the allometry equation. Biometrics 19:497-499. KEAR, J. AND R. J. SCARLETT. 1970. The Auckland Islands Merganser. Wildfowl 21: 78-86. KELSO, J. E.H. 1922. Birds using their wings as a means of propulsion under water. Auk 39:426-428. LACK, D. 1970. The endemic ducks of remote islands. Wildfowl 21:5-10. Livezey ß AUCKLAND ISLANDS MERGANSER 433 LIVEZEY, B.C. 1986. A phylogenetic analysis of Recent anseriform genera using morpho- logical characters. Auk 103:737-754. 1988. Morphometrics of flightlessness in the Alcidae. Auk 105:681-698. 1989. Flightlessness in grebes (Aves, Podicipedidae): its independent evolution in three genera. Evolution 43:29-54. -- AND P.S. HUMPHREY. 1984. Sexual dimorphism in continental steamer-ducks. Condor 86:368-377. -- ArD . 1986. Flightlessness in steamer-ducks (Anatidae: Tachyeres): its mor- phological bases and probable evolution. Evolution 40:540-558. LowE, P.R. 1928. A description ofAtlantisia rogersi, the diminutive and flightless rail of Inaccessible Island (southern Atlantic), with some notes on flightless rails. Ibis 70:99- 131. LUTHER, D. 1967. Augestorbene und aussterbende Vogel: V. Falke 14:166-169. MADGE, S. AND H. BURN. 1988. Waterfowl: an identification guide to the ducks, geese and swans of the world. Houghton Mifflin, Boston, Massachusetts. MAGNA, M.A. 1912. Modifications organiques cons6ctives chez les oiseaux  l'absence de vol. Bull. Mus. Nat. Paris 18:524-530. 1922. Les caracteristiques des oiseaux suivant le mode de vol. Ann. Sci. Nat. (Ser. 10) 5:125-334. MATHEWS, G. M. AD T. IREDALE. 1913. A reference list to the birds of New Zealand. Ibis 55:402-452. MCNAMARA, K.J. 1986. A guide to the nomenclature of heterochrony. J. Paleontol. 60: 4-13. MEINERTZHAGEN, R. 1955. The speed and altitude of bird flight (with notes on other animals). Ibis 97:81-117. MEUNIER, K. 1951. KorrelationundUmkonstruktionindenGr6ssenbeziehungenzwischen Vogelfliigel und Vogelk6rper. Biol. Gener. 19:403-443. 1959ß Die Gr6ssenabhangigkeit der Korperform bei Vogeln. Z. Wiss. Zool. 162: 328-355. MILLIKEN, G. A. AND D. E. JOHNSON. 1984. Analysis of messy data. Van Nostrand, New York, New York. M'ErLENIOFF, K. 1885. Die Grosse der Flugflachen. Pflueger's Arch. Ges. Physiol. 35:407- 453. OLSON, S. L. 1973. Evolution of the rails of the South Atlantic islands. (Aves: Rallidae). Smithsonian Contr. Zool. No. 152. ß 1977. Notes on subfossil Anatidae from New Zealand, including a new species of Pink-eared Duck Malacorhynchus. Emu 77:132-135. PARTRIDGE, W.H. 1956. Notes on the Brazilian Merganser in Argentina. Auk 73:473- 488. PENNVCtnCK, C.J. 1975. Mechanics of flightß Pp. 1-75 in Avian biology. Vol. 5 (D. S. Farner and J. R. King, eds.). Academic Press, New York, New York. PHILLIPS, J. C. 1925. A natural history of the ducks. Vol. 3. Houghton Mifflin, Boston, Massachusetts. PMqT[, R.A. 1979. Morphometrics: the multivariate analysis of biological data. Ken- dall/Hunt Publ. Co., Dubuque, Iowa. POOLE, E.L. 1938. Weights and wing areas in North American birds. Auk 55:511-517. RAIKOW, R.J. 1973. Locomotor mechanisms in North American ducks. Wilson Bull. 85: 295-307. RICKER, W.E. 1984. Computation and uses of central trend lines. Can. J. Zool. 62:1897- 1905. 434 THE WILSON BULLETIN ß Vol. 101, No. 3, September 1989 SCOTT, P. 1972. A coloured key to the wildfowl of the world. Wildfowl Trust, Slimbridge, England. SmLEY, C.G. 1957. The evolutionary and taxonomic significance of sexual dimorphism and hybridization in birds. Condor 59:166-191. SNow, B.K. 1966. Observations on the behaviour and ecology of the flightless cormorant Nannopterum harrisi. Ibis 108:265-280. SOKAL, R. R. AND F. J. ROHLF. 1981. Biometry. 2nd ed. W. H. Freeman, New York, New York. TOWNSEND, C. W. 1909. The use of wings and feet by diving birds. Auk 26:234-248. WELLER, M.W. 1975. Ecological studies of the Auckland Islands Flightless Teal. Auk 92: 280-297. 1980. The island waterfowl. Iowa State Univ. Press, Ames, Iowa. WILEY, E. O. 1981. Phylogenetics: the theory and practice of phylogenetic systematics. John Wiley and Son, New York, New York. WmLIAMS, G.R. 1964. Extinction and the Anatidae of New Zealand. Wildfowl 15:140- 146. ERRATUM In "Song repertoires and the singing behavior of male Northern Cardinals" by Gary Ritchison (Wilson Bull. 100:583-603), the penultimate sentence in paragraph 2 on page 597 should read "Groschupf (1985) reported that male Five-striped Sparrows (Amphiza quin- questriata) used more complex songs in intersexual contexts." Livezey ß AUCKLAND ISLANDS MERGANSER 435