Geographic Variation in the Milkfish Chanos Chanos 11. Multivariate Morphological Evidence

Copeia, 1985(4), pp. 890-898

Geographic Variation in the Milkfish Chanos chanos 11. Multivariate Morphological Evidence

Morphological variation in six meristic and 19 morphometric characters was examined in the milkfish, Chanos chanos. Samples were collected at 15 locations in the Pacific Ocean, from the Philippine Archipelago in the west, along the equatorial Pacific Ocean to Tahiti and Hawaii in the east. Variation within each sample, estimated as the multivariate generalization of the coefficient of variation, did not vary substantially among the localities. Principal component analyses indicated that multivariate body shape differences existed, but that meristic differences were lacking among the samples. Along a morphometric shape axis, the Hawaiian samples and the equatorial Pacific Island samples generally had larger head features-smaller tails in comparison to the Philippine samples. The data for a repeat sample from the island of Hawaii indicated that the multivar- iable shape criterion increased with the size of fish and thus provided an insight into the principal component results. The patterns of morphological differences and electrophoretic differentiation (Winans, 1980) were compared graphically. The genetic population structure of milkfish in the Pacific Ocean consists of three distinct groups: the Philippine Archipelago; the equatorial Pacific Ocean, including Tahiti; and the Hawaiian Archipelago.

ILKFISH, Chanos chanos (Forskal), are biology of a species is to assess the levels and M tropical marine fish of the Indian and patterns of variability within and among natural Pacific oceans that inhabit shallow, inshore re- populations of the organism (Could and John- gions such as estuaries and sandy reef flats. On ston, 1972). Contemporary descriptions of geo- the basis of their availability, general hardiness graphic variation are commonly based on elec- and rapid growth in pond culture, milkfish are trophoretic and morphological characteristics considered one of the best suited species of fish of populations, as these two sets of traits consist for aquaculture (Bardach et al., 1972). Al- of discrete genetic characters and multigenic though the pond culture of milkfish already morphological characteristics and are differ- contributes significantly to human nutrition, es- ently affected by environmental variation. In pecially in Southeast Asia, there are few data this approach then, different portions of an or- concerning their basic biology (Liaoet al., 1979: ganism's genome are represented and evaluat- Kuo et a]., 1979; and Wainwright, 1982). ing and contrasting these two data sets may be An important initial step in a study of the useful from a systematic, evolutionary and/or

O 1985 by the American Society of ichthyologists and Herpetologists

WINANS-MILKFISH GEOGRAPHIC VARIATION 89 1

Fig. 1. Location of milkfish sample sites. P8, San Thomas, Luzon; P7, Mercedes, Luzon; P6, Hamitique, Panay; P5, Ormoc, Leyte; P4, Argao, Cebu; P3, Cagayan de Oro, Mindanao; P2, Zamboanga, Mindanao; and P1, Sulop, Mindanao. a managerial standpoint (Wright, 1978; Utter, ing electrophoretic studies, subsamples of each 1981; and Ihssen et al., 1981). fish collection were used for morphological To assess geographic variation in milkfish, analyses. Average sample size was 29 fi~h-~er samples were collected where possible in the locality (Table 1). The sample sizein all analyses Pacific Ocean, from the Philippine Archipelago was 47 1 fish. eastward to the Hawaiian Archipelago and Ta- Data were recorded for 19 morphometric hiti. Specimens were subjected to multivariate characters (plus the standard length) listed in electrophoretic and morphological analysis. Table 2. On larger fish, the first four measure- Electrophoretic variability at 38 loci was re- ments in Table 2 were measured with a meter ported in Winans (I 980). In this paper, patterns stick accurate to 0.5 mm. Otherwise, the mor- of multivariate morphological variation in 25 phometric characters were measured with dial characters are viewed for the same set of milk- cali~ersaccurate to 0.1 mm. Six meristic counts fish samples used for the electrophoretic anal- were also made: dorsal-fin rays, pectoral-fin rays, ysis. Data from a Tahiti sample and a repeat pelvic-fin rays, anal-fin rays, pored lateral line sample from the island of Hawaii are also in- scales and scales along the base of the dorsal cluded. Levels of variability in each locale are fin. All measurements and counts were made estimated and tested for equality among sam- on the left side of freshly thawed specimens. ples and patterns of differentiation are assessed Morphological variability was estimated in by principal component analysis. These results each sample by the multivariate generalization are compared with the electrophoretic results. of the coefficient of variation (Van Valen, 1978: 41)

Fish were sampled in 1977, 1978 and 1979 from 15 locations in the Pacific Ocean (Fig. 1). The three major island groups and their acro- where @ and x. are the sums of the vari- nyms were the Philippines (PHIL, eight locali- ance and mean squared, respectively, of the six ties), the equatorial Pacific Islands (EPI; Palau, meristic traits in a sample over j populations Tarawa, Fanning and Christmas islands) and the (see below). Hawaiian Islands (HI, two localities). The sam- To examine among-sample differences multi- ple from Tahiti was considered separately. Gill, variately, principal component analyses were hand, or surround nets were used to capture performed on character variance-covariance the fish in coastal ponds or waterways. Follow- matrices using a combination of subroutines 892 COPEIA, 1985, NO. 4

TABLE1. SUMMARYSTATISTICS FOR FORKLENGTH TABLF.2. VARIABLECOEFF~CIENTS ON PRINCIPAL IN CM. Locality acronyms given in Fig. 1. COMPONENTSI AND I1 FROM AN ANALYSISOF 19 MORPHOMETRICCHARACTERS. PC I1 was adjusted or Fork length sheared by algorithm in Humphries et al. (1981). Stan- dard devi- Mini- Max,- Sheared Sample N Mean ation mum mum CV,** Morphometric character PC I PC 11 P 1 30 22.05 0.69 20.8 23.8 3.88 Fork length P2 31 21.89 0.97 20.1 26.2 3.12 Length snout-anal fin origin P3 33 24.69 1.02 22.0 26.1 3.49 Length snout-pelvic fin origin P4 37 20.90 1.69 18.6 26.1 3.47 Length snout-pectoral fin origin P5 36 21.39 0.92 20.0 24.1 3.62 Length snout-dorsal fin origin P6 29 22.15 1.66 18.1 25.9 3.98 Head length P7 21 21.75 0.49 20.9 22.7 3.50 Snout length P8 37 20.40 0.73 19.2 21.9 3.10 Postorbital length PAL 20 9.43 1.00 7.3 11.8 3.90 Orbital length TAR 30 14.76 1.64 9.9 17.0 5.65 Caudal depth FAN 15 16.54 0.61 15.1 17.3 5.57 Body depth at anus CHR 32 21.19 0.70 20.1 22.5 3.77 Length dorsal fin base HAW 29 22.30 4.89 15.3 30.5 3.73 Length anal fin base HAWS@ 30 36.77 2.88 28.8 43.0 3.80 Length pectoral fin base OAH 30 11.7 0.66 10.3 13.0 4.73 Pectoral fin length TAH 31 25.5 1.80 22.4 30.2 3.70 Ilead width Nares width *' Multivariate generalization of the roeffi~ientof variation for SIX meristic characters. Bony interorbital width @ HAW2 (sample taken one year after HAW). Premaxilla length % of total variance from the International Mathematical and Sta- tistical Library for numerical software (IMSL, 1977). Meristic data were transformed into cant for the remaining 19 morphometric char- square roots and morphometric data into com- acters. mon logarithms to reduce the correlation of the measurement means and variances (Sokal and Rohlf, 1969). Bivariate confidence ellipses were calculated by program A3.13 in Sokal and Rohlf Sample statistics for the fork length are pre- (1969). sented in Table 1. Since most of the morpho- To estimate measurement error for the 20 metric characters were correlated with the size morphometric characters, 10 fish (20-30 cm, of the fish, or fork length (see below), an ex- fork length) were randomly chosen from three amination of these data provides a good idea of samples and each morphometric character was the range of variation in the morphometric measured independently on each fish five times. characters. The standard deviation of each distance character over the five replications was estimated Variation within samples.-To estimate morpho- for each fish; the average standard deviation of logical variability within each sample, only the a character over the 30 fish was used as an es- meristic traits were considered. Within the 16 timate of measurement error. The average samples, correlation coefficients of meristic measurement error for the 20 characters was characters with the length of the fish (fork

0.48 mm (k0.36 mm), with values ranging from length) were generally small and not statistically . ~ 0.16 mm for the caudal peduncle depth to 1.64 significant (i.e., in the 96 comparisons, only 6 mm for the standard length. Fork length mea- were significant). On the other hand, all of the surements had a measurement error of 0.93 morphometric variables were correlated with mm and were therefore used in lieu of standard the length of the fish (mean correlation coeffi- length in all analyses. As 70% of the characters cient = 0.95). Hence, a good estimate of size- had a measurement error of less than 0.5 mm, independent morphological variability in a sam- measurement error was considered insignifi- ple was a measure of meristic variability. WINANS-MILKFISH GEOGRAPHIC VARIATION 893

Fig. 3. Illustration of important morphometric measurements for sheared PC 11. Coefficients are x 100.

CHR Chclrrmarlrland FI\N Fannmg ,stad MI\,, Hewall. ,978 HAW 2 Haw811.1019 (the variable loadings and the multivariate TAH Tihlli TAR Tan*= scores) was therefore adjusted or sheared by the I -0 70 0 0.70 algorithm of Humphries et al. (1981) to pro- Prncoll ."n"""anr r duce a size-free shape component, shared PC Fig. 2. 95% confidence eIlipses for bivariate means I1 (SPC 11). SPC I1 contrasted several head char- of sheared PC I1 and PC I from an analysis of 19 acteristics (orbital, snout and snout-premaxil- morphometric characteristics of Chanos. Locale ac- lary length) with several tail width characters ronyms are given in Fig. 1. (caudal depth, body depth and anal-fin base) and was considered a shape indicator (Fig. 3). Gen- erally, the Hawaiian samples and the Equatorial Values of CV,, given in Table 1, differed little Pacific Island (EPI) samples had large SPC I1 among the samples. Values ranged from 3.10 values (larger head features-smaller tails), while (P8) to 5.65 (TAR). A test of equality of the 16 the Philippine samples had smaller SPC I1 val- CV, values was not significant (P = 0.20 in ues (smaller head features-larger tails) (Fig. 3). Smith's test; from Van Valen 1978:35). When The Philippine sample P6 and the EPI sample the sample values were compared by island from Christmas Island (CHR) were exceptions group (i.e., PHIL, EPI and HI), a statistically to this pattern. The mean SPC I1 value for the significant among-group difference was re- Tahiti sample was close to the typical Philippine vealed (0.0 1 < P < 0.025; Kruskal-Wallis value. H-test). This result may be explained by larger A principal component analysis of the six me- CV, values in the EPI group (Z = 4.72) which ristic characters indicated little among-sample were significantly greater than the PHTL values variability. Of the total variance, 84% was ex- (2 = 3.52; P = 0.01), Wilcoxon two-sample test). plained along the first two axes (Table 3), with The PHIL and HI values were not significantly subsequent functions explaining less than 6% different (U = 4, P > 0.05). The inclusion of of the total variance. The scale count on the the Tahiti sample into the EPI group (justified dorsal-fin base predominated on PC I and the on electrophoretic criteria below) did not alter lateral line scale count predominated on PC IT. the above results. Finally, the CV, values did In the PC I-PC I1 scattergram, 11 diagonal not vary significantly with latitude (r = -0.27) columns appeared along the PC I axis corre-

or longitude (r = 0.47). s~onding.n to the 11 observed values of scale counts on the dorsal-fin base, as illustrated in Among sample differences. -In the principal com- Fig. 4. A scattergram of the bivariate means ponent analysis of the morphometric data, 98% along the first two axes is presented in Fig. 5. of the total variance was explained along the Despite the relative distinctiveness of the sam- first two axes (Table 2). Character loadings were ples from Oahu, Fanning Island and Tarawa approximately equal on principal component I long PC 11, the most remarkable feature of the (PC I) and mean fork length per sample in- results depicted in Fig. 5 is the extensive overlap creased from left to right along the PC I axis among the samples. (Fig. 2). Thus, PC I was considered a measure of general size. A preliminary analysis indicated that PC I1 was influenced to a small degree by within-group size variability, i.e., bivariate el- Morphological variability within each sample, lipses tilted with respect to the PC I axis. PC 11 when estimated as CV,, varied little over the 15 894 COPEIA, 1985, NO. 4

TABLE3. VARIABLECOEFFICIENTS ON THE FIRST THREEPRINCIPAL COMPONENTS FROM AN ANAI.YSISOF SIXMERISTIC CHARACTERS OF Chanos chanos.

Meristic character PC I PC I1 PC I11 Dorsal-fin rays 0.05 0.03 -0.11 Pectoral-fin rays 0.04 -0.09 -0.91 Pelvic-fin rays 0.01 -0.08 -0.36 Anal-fin rays 0.00 -0.01 -0.09 Lateral line scales 0.12 -0.98 0.11 Scales dorsal-fin base 0.99 0.12 0.03 % of total variance 64.1 19.8 5.9

locations in the Pacific Ocean (Table 1). When the 15 CV, values were compared, no significant Fig. 4. Scatterplot of scores for three samples on differences were detected. When the data were the first two axes from a principal component analysis pooled and examined by island groups, the four of six meristic characteristics of Chanos. samples from the equatorial Pacific Ocean (the EPI group) had significantly greater values than the eight Philippine samples; but no other pair- that includes the Philippine Archipelago (Ra- wise group comparisons were significantly dif- banal and Ronquillo, 1975). Therefore, if the ferent. Two samples in the EPI group account- mutation-drift hypothesis is applicable to these ed for the significant island group differences: data, it is not surprising to find the highest levels Tarawa (CV, = 5.65) and Fanning Island of electrophoretic variability in the Phillippine (CV, = 5.57). Thus, there was no compelling samples. And, by this reasoning, the paucity of evidence for a consistent trans-eauatorial in- electrophoretic variation observed in the most crease in morphological variation for all four geograhhically peripheral samples, the HI and EPI samples. Finally, no latitudinal or longitu- TAH samples (Winans, 1980) may also reflect dinal trends in CV, were detected; and it is con- a historical (founder effect) or a contemporary cluded that morphological variability within the decrease in the effective population size of milk- samples was not diffeient or only ;lightly dif- fish in these areas. The lack of a similar pattern ferent among the locales sampled here. Similar in within-sample morphological variability sug- results were obtained when variability was es- gests that relevant environmental or population timated in each sample as the mean cbefficient parameters affecting variability did not vary sys- of variation over the six meristic characters. tematically among the samples. These results are somewhat dissimilar from A principal component analysis of six meristic the patterns of electrophoretic variability mea- characters indicated little among-group differ- sured at 38 loci in the same Chanos samples ences, whereas a like analysis of morphometric (Winans, 1980). Although the estimates of the characters separated milkfish samples by size on average heterozygosity per locus and the per- PC I and by shape on SPC I1 (Fig. 3). Since size centage of polymorphic loci (Po.,, level did not differences may reflect only sampling variation, change significantly over the same region, the and separation of fish by size can be achieved average number of alleles per locus and the per- by eye without statistical analyses, the discrim- centage of polymorphic loci (Po,,, level) de- ination among samples by the body shape cri- creased significantly from the west to the east terion is essentially the more useful result. In (e.g., PHIL > EPI > HI). According to the mu- general, the Philippine samples were distinctive tation-drift hypothesis, the highest levels of ge- along the shape axis. netic variability are found in regions of greatest The repeat sample from the island of Hawaii effective population size (Nei, 1975). Whereas (HAW2) provides an insight into shape differ- Chanos are distributed throughout the tropical, ences along SPC I1 since its analysis represents inshore marine habitats of the Pacific and In- the only opportunity in the present study to dian Oceans, the most abundant populations view, in one location, the degree of between- have been noted in a region of Southeast Asia year morphological variation in Chanos. The WINANS-MILKFISH GEOGRAPHIC VARIATION 893

P1 Philippino ,.lend. PAL Pala" CHR Chr8srmol !%land FAN Fanning Island HAW Harsii. 1978 O,# ?h ,,,O., ".,1."d. pa, CHS ch",,ms,,.,mdPa,." ,AM 8,U.d "An *..., i, l97l "A" 2 "".i,. 79,s "PA T4R Tarara T*" 0.8" T.h,,, ,An T...-* r --

Fig. 6. Electrophoretic relationships among Chan- os samples based on UPMGA clustering of Nei's genetic distance values (Nei, 1978). Fig. 5 95% confidence ellipses for bivariate means on PC I1 and PC I from an analysis of six meristic characteristics of Chanos. Locale acronyms given in Fig. 1. meristic traits may vary as a function of a population's physical milieu (Hubbs, 1922; Barlow, 196 1; Fowler, 1970). Since there were no me- principal component analysis of the morpho- ristic differences between HAW and HAW2- metric data showed that there were consider- their confidence ellipses overlapped in Fig. 5- able multivariate differences between the two there is no reason to suspect, at least on this samples along the first two PC axes (Fig. 3). basis, that the two yearly samples experienced Since the HAW2 fish were considerably larger varying environmental conditions. than the HAW fish (mean fork length: 37 cm The third possibility, that the SPC I1 differ- and 22 cm, respectively), the separation of ences are predictable shape changes that accom- HAW2 and HAW along the size-related axis pany different size distributions, may be eval- was expected. The degree of separation along uated by considering the third Flawaiian sample SPC 11, a multivariate shape axis, was also large, OAH, the sample most genetically similar (by indicating a considerable shape difference. Since electrophoretic criteria) to HAW and HAWS. primary emphasis is placed on shape discrimi- OAH, which had the smallest sized fish in the nation among samples, further consideration of HI samples, had the lowest mean value along this observation is warranted. SPC I1 with respect to HAW and HAW2. In There are several possible explanations for fact, the three Hawaiian centroids were linearly the shape difference between HAW and HAW2. arranged along SPC I1 in a pattern correspond- Differences may be due to: 1) genetic differ- ing to their mean fish size along PC I: the larger ences between the samples, 2) morphological the fish, the greater the SPC I1 value (Fig. 3). changes associated with yearly environmental With the available data therefore, the best ex- changes and/or 3) shape changes associated with planation for the relationship of HAW and size differences. The electrophoretic analysis of HAW2 along SPC I1 is that there is a size-re- these samples indicated that the fish did not lated or ontogenetic change in a body shape come from dissimilar genetic stocks. Allele fre- among the milkfish in Hawaii. Importantly, this quencies at the polymorphic loci were highly finding points to the need for more complete similar, as reflected in a small Nei's genetic dis- sampling of fish (with respect to size) from each tance value between the samples (Fig. 6). Thus, locale to completely understand how ontoge- there is no reason. based on electro~horeticin- netic shape changes within each locale affect formation, to expect morphological difference~ our evaluation of among-locale shape differ- between HAW and EIAMr2. It may be argued ences. that the fish in the two samples experienced A comparison of the similarity in patterns of different environmental variability during de- morphological and electrophoretic differentia- velopment and in response, have slightly altered tion is depicted in Fig. 7. The pattern of mor- morphological characteristics. Unfortunately, phological differentiation is a summary of the no precise data are available with which to assess variation along the SPC I1 axis in Fig. 3. The this possibility directly. It is well established that pattern of electrophoretic differentiation is a COPEIA, 1985, NO. 4

AUSTRALIA

Electrophoretic ,..

Fig. 7. Patterns of geographic variation in the milkfish Chanos chanos. Circles refer to the samples described in Fig. 1. Patterns represent: A) morphometric variation along the sheared principal component I1 axis in Fig. 3 (i.e., either positive or negative SPC I1 mean sample values) and B) electrophoretic similarity for 38 loci (Fig. 6). summary of the single locus and multilocus anal- Philippine samples and the two Hawaiian Sam- yses over 38 loci presented in Winans (1980) ples) or across adjacent island groups (for the and here (Fig. 6). 'The dichotomous summary four equatorial Pacific Island samples plus Ta- of the morphometric study basically divides the hiti). In sum, both data sets indicate that the samples into Philippine and non-Philippine Philippine samples are different from the other samples. The values for one Philippine sample samples; but, only in the electrophoretic data (P6), Tahiti and Christmas Island, however, set are all eight Philippine samples one homo- produce considerable heterogeneity in this sim- geneous group and the Hawaiian Island samples ple model of morphological-variation. On the a distinctive group. Since environmental influ- other hand, the pattern of electrophoretic sim- ences on morphological variation may be sub- ilarity reflects the geographic proximity of the stantial but not assessable here, further evalu- samples, i.e., electrophoretic similarity coin- ation of the levels and patterns of morphological cides with archipelago boundaries (for the eight characters requires repeat sampling, tabulation SVINANS-MILKFISH GEOGRAPHIC VARIATION 897 and analysis of environmental data ana/or LITERATURECITED breeding 'studies (Riddell et al., 198 1; Tocid et BAXDACH,~.E.,J. W.RYTHF.R AKD W. 0.MCLARNEY. al., 1981). Since the expression of electropho- 1972. Aquaculture. Wiley Interscience, New York. retic variation is virtually ilnaffected by envi- BARLOW,G. W. 1961. Causes anri significance of ronmental variation and the patterns of eiel- morphological variation in fishes. Syst. Zool. 10: rrophoretic similarity follow patterns of 105-116. geographic proximity, the electrophoretic re- CHERRY,L. M.; S. M. CASEAND A. @. WILSON.1978. sults are presumed to reflect best the subpop- Frog perspective on the morphological difference olation st;ucturing among the samples of &an- hetween humans and chimpanzees. Science 200: a:. I conclude, therefore, that the genetic 209-2 1 1. ~opulationstructure of milkfish in the Pacific FOWLER,^. A. 1370. Control of vertebral number in teleosts-an embryological problem. Q. Rev. Bioi. Ocean consists of three distinct groups as illus- 45:148-167. :?ated in Fig. 7B. Whereas this genetic structure GOLJLD,S.J., AND R. F.JOHNSTON.1972. Geographic may result from numerous evolutionary factors variation. Annu. Rev. Ecol. Syst. 3:457-498. (Selander and Whittam, 1983), the simplest ex- HUBBS,C. k. 1922. Variations in the number of ver- planation is that it results from oceanographic tebrae and other meristic characters of fishes cor- conditions affecting gene flow within and among related with the temperature of water during de- island groups (Winans, 1980). velopment. Am. Nat. 56:360-372. That there are both similarities and dissimi- HUMPHRIES,J.M., F. L. BOOKSTEIN,F. CHERNOPF, G. !arities between the patterns of morphological R. SMITH,R. E. ELDERAND S. G. POSS.1981. Multi- aild electrophoretic variation in these Chanos variate discrimination by shape in relation to size. Syst. Zool. 30:291-308. samples is not an unusual finding. Other work IHSSEN,P. E.; H. E. BOOKE,J. M. CASSELMAN,J. with a variety of vertebrate species has also MCGLADE,N. R. PAYNEAND F. M. UTTER. 1981. shown an independence of morphological and Stock identification: materiais and methods. Can. molecular evolution (Cherry et al., 1978; Lar- j. Fish. Aquat. Sci. 38:1838-1855. son, 1980; Shaklee and Tamaru, 1981; Wake, PMSL LIBRARYREFERENCE MAUUAL, EDITION 7. 1977. 1981). What this independence of morpholog- Internationai Mathematics and Statistics Library, ical and electrophoretic variation means there- Houston, Texas. fore is that both character sets will be useful in Kuo, C. M., C.E. NASHAND W. 0.WATANABE. 1979. future attempts to understand the evolution of Induced breeding experiments with milkfish, Chan- os chanos Forskal, in Hawaii. Aquaculture 18:95- this species and to manage milkfish stocks for 106. aquaculture. Further studies of the relationship LARSOK,A. 1980. Paedomorphosis in relation to rates of these two character sets in milkfish will in- of morphological and molecular evolution in the volve examining the association of electropho- salamander Aneides jZnvipunctuatus (Arnphibia, retic phenotypes and bilateral symmetry of me- Plethodontidae). Evolution 34:1-17. ristic characters at both the individual and LEARY,R. F., F. W. ALLENDORFAND K. L. KNUDSEN. population level (Vrijenhoek and Eerman, 1982; 1983. Developmental stability and enzyme hetero- Leary et al., 1983). zygosity in rainbow trout. Nature 301:71-72. ~AO,1. C.,J. V.JUARIO, S. KUMAGAI,H. NAKAJIMA, M. NATIVIDADAND P. BURI.1979. On the induced spawning and larval rearing of milkfish, Chanos chanos (Forskal). Aquaculture 18:75-94. For their assistance in the sampling program, NEI, M. 1975. Molecular Population Genetics and 1 would like to thank R. Schleser, Oceanic In- Evolution. North-Holland Publishing Co., Amster- stitute, Hawaii; P. Siu, Papeete, Tahiti (the TAH dam, The Netherlands. sample); and C. Tamaru and Mr.Watanabe, NLI,M. 1978. Estimation of average heterozygosity Universityof Hawaii (the HAW2 sample). My and genetic distance from a small number of indi- special thanks to J. B. Shaklee for his invaluable viduals. Genetics 89:583-590. input into this project. G. Carey and H. Carey RABAUAL,H. R., AND I. A. RONQUILLO.1975. Dis- provided computer assistance. Parts of this pa- tribution and occurrence of milkfish Chanos chnnos per comprise a doctoral dissertation completed (Forskal). Proceedings National Bangos Sympo- sium, Philippines. 1975:20-33. at the University of Hawaii. This research was RIDDELL,B. E., W. C.LEGCETT AND R. L. SAUNDERS. supported in part by the Jessie Smith Noyes 198 I. Evidence of adaptive polygenic variation be- Foundation, New York and a postdoctoral fel- tween two populations of Atlantic salmon (Sal7r~o lowship from the National Research Council, salar) native to tributaries of the S. W. Miramichi Washington D.C. River, N. B. Can. J. Fish. Aquat. Sci. 38:321-333. 898 COPEIA, 1985, NO. 4

SELANDER,R. K., AND T. S. WHITTAM.1983. Protein WAINWRIGHT,T. 1982. Milkfish fry seasonality on polymorphism and the genetic structure of popu- Tarawa, Kiribati, its relationship to fry seasons else- lations, p. 89-144.171: Evolution of genes and pro- where, and to a sea surface temperature (SST). teins. M. Nei and R. K. Koehn (eds.). Sinauer Assoc. Aquaculture 26:265-271. Inc., Sunderland, Massachusetts. WAKE,D. B. 1981. The application of allozyme evi- SHAKLEE,J. B., AND C. S. TAMARU.1981. Riochem- dence to problems in the study of morphology, p. ical and morphological evolution of Hawaiian bone- 257-270. In: Evolution today; Proc. 2nd Int. Congr. fishes (Albula).Syst. Zool. 30:125-146. Syst. Evol. Biol. G. G. E. Scudder and J. L. Reveal SOKAL,R. R., AND F. J. ROHLF.1969. Biometry. Free- (eds.). Carnegie Mellon University, Pittsburgh, man and Co., San Francisco, Calif. Pennsylvania. TODD,T. N., G. R. SMITHAND L. E. CABLE.1981. WINANS,G. A. 1980. Geographic variation in the Environmental and genetic contributions to mor- milkfish Chanos chanos. I. Biochemical evidence. phological differentiation in ciscoes (Coregoninae) Evolution 34:558-574. of the Great Lakes. Can. J. Fish. Aquat. Sci. 38:59- WRIGHT,S. 1978. Evolution and the Genetics of Pop- 67. ulations. Vol. 4. Variability Within and Among Nat- UTTER,F. M. 1981. Biological criteria for definition ural Populations. University of Chicago Press, Chi- of species and distinct intraspecific populations of cago, Illinois. anadromous salmonids under the U.S. Endangered Species Act of 1973. Can. J. Fish. Aquat. Sci. 38: 1626-1635. VANVALEN, L. 1978. The statistics of variation. Evol. TIONAL OCEAXICAND AT~~OSPHERICADMI- Theory 4:33-43. NISTRATION, 2725 MONTLAKEBOULEVARD VRIJFNHOEK,R. C., AND S. LEXMAN.1982. Hetero- EAST, SEATTLE, WASHINGTON Ac- zygosity and developmental stability under sexual 98 1 12. and asexual breeding systems. Evolution 36:768- cepted 26 Feb. 1985. 776.