INFORMATION TO USERS
While the most advanced technology has been used to photograph and reproduce this manuscript, the quality of the reproduction is heavily dependent upon the quality of the material submitted. For example:
• Manuscript pages may have indistinct print. In such cases, the best available copy has been filmed.
• Manuscripts may not always be complete. In such cases, a note will indicate that it is not possible to obtain missing pages.
• Copyrighted material may have been removed from the manuscript. In such cases, a not«~·· will indicate the deletion.
Oversize materials (e.g., maps, drawings, and charts) are photographed by sectioning the original, beginning at the upper left-hand corner and continuing from left to right in equal sections with small overlaps. Each oversize page is also filmed as one exposure and is available, for an additional charge, as a standard 35mm slide or as a 17"x 23" black and white photographic print.
Most photographs reproduce acceptably on positive microfilm or n1icrofiche but lack the clarity on xerographic copies made from the microfilm. For an additional charge, 35mm slides of 6"x 9" black and white photographic prints are available for any photographs or illustrations that cannot be reproduced satisfactorily by xerography.
----····------· ------·-·------./
------8713662
Hendrickson, Dean Arthur
GEOGRAPHIC VARIATION IN MORPHOLOGY OF AGOSIA CHRYSOGASTER, A SONORAN DESERT CYPRINID FISH
Arizona State University PH.D. 1987
University Microfilms International 300 N. Zeeb Road, Ann Arbor, Ml48106
. ··-·------·-·-· ~------·------·~----- GEOGRAPHIC VARIATION IN MORPHOLOGY OF AGOSIA CHRYSOGASTER, A SONORAN DESERT CYPRINID FISH by
Dean A. Hendrickson
A Dissertation Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy
ARIZONA STATE UNIVERSITY
May 1987
------·------GEOGRAPHIC VARIATION IN MORPHOLOGY OF AGOSIA CHRYSOGASTER, A SONORAN DESERT CYPRINID FISH by
Dean A. Hendrickson
has been approved
t1ay 1987
APPROVED:
.Chairoerson
Supervisory Committee
ACCEPTED:
-~::?d~--Dean, Graduate College
------···---·- ···------· ------. --··. ABSTRACT
Morphometric analyses of Agosia chrysogaster
northern morph native to Bill Will lams, Gila, Sonoyta and de la
Concepcion basins of Arizona, New Mexico and Sonora, and a southern form from Willcox Playa of Arizona and Rios Sonora, Yaqui, Mayo,
Fuerte an~ Sinaloa of Sonora and Sinaloa, Mexico. The latter is
smaller, and less sexually dimorphic, but has longer pre- and
postdorsal body lengths. Populations in the geographically
intermediate Rios Sonoyta and Sonora are morphologically intermediate.
Males differ more between morphs than do females. Meristic characters overlap between morphs, but the northern form has higher mean lateral line scale counts. Highly tuberculate nuptial males, characteristic
of the northern morph, were not found in the south, nor were •spawning• pits associated with spawning of the former. Morphs differ on a multivariate axis on which temporal variation at single localities is also reflected. Distances among some intra-locality samples on this axis were greater than least inter4norph distances. Measures of morphological disimilarity were
weakly correlated with inter-sample differences in elevation; latitude, and longitude, but more highly correlated with an index of
hydrologic isolation among localities. Differentiation among basins thus appears to reflect hydrographic isolation, rather than ecological
conditions.
Electrophoretic data on&· chrysogaster produced relationships
patterns largely incongruent with results of the morphological
analyses, and with unexpected geographic area relationships.
iii
--- -··- ----· ·------·-· ---·-- ··--···· ACKNOWLEDGEMENTS
Many people provided assistance that enabled me to complete this study. Primarily I wish to thank my wife, Sherry, and son, Garrett, who, along with my parents, supported me throughout this project and tolerated many hours in which my focus was on this instead of them. This work, and many others I have simultaneously undertaken, would have been impossible without their help and understanding. Dr. W. L. Minckley has been an endless source of support and intellectual stimulation throughout, and has made it possible for me to participate in several other professionally profitable, and enjoyable, endeavors.
My work with him has greatly broadened my scientific background and for that I am forever grateful.
Much field work for this study was supported by various grants from the Department of Zoology at Arizona State University, which also provided me with part-time teaching and museum assistantships for four years. I wish to thank all collectors mentioned in the 1 ist of samples for their field help, and all museum curators who lent specimens. Ms. Susan Zneimer did the electrophoretic analyses in the lab of Dr. Don Buth at University of California, Los Angeles. I also wish to thank Terry Johnson for his flexibility in allowing me time to complete this study while employed under his supervision at Arizona Game and Fish Department. Michael E. Douglas and members of my committee provided many suggestions which greatly improved the dissertation.
iv
------TABLE OF CONT8NTS
Page LIST OF TABLES vii
LIST OF FIGURES viii INTRODUCTION Objectives
Distribution, Biology and Ecology of Agosia chrysogaster 4
Taxonomic History 10
METHODS AND MATERIALS 12
Morphometric and Meristic analyses 12
Electrophoretic Analyses 15
RESULTS 17
Univariate Analyses 17
Multivariate Analyses of Morphometric Data 20
Allometric Differences Between Morphs 31
Shape Reconstructions 33
Temporal Variation in Morphology of Agosia chrysogaster 38
Classification of Unknowns to Morph and Basin 47
Results of Analyses of Meristic Data 50
Correlations with Ecology and Geography 51
Other Characters which Corroborate Morphometric Analyses 56
Results of Analyses of Electrophoretic Data 58
DISCUSSICJI.I 75
Physiography and Distribution of Morphotypes 85
Other Fishes 88
v
---· ----.------Page
CONCLUSIONS 93
REFERENCES 97
APPENDIX 112 Appendix A. Acronyms and descriptions of morphometric variables. 112 Appendix B. Descriptions of samples used in analyses 114
Appendix C. Means and 9~/. confidence intervals <±2 standard errors) for all variables by sex and basin 120 Appendix D. Nonlinear regression statistics by sex and morph for all ·variables regressed on SL. 127 Appendix E. Composition of native fish faunas of river basins where Agosia chrysogaster is native. 136
vi
--- -·-·· ·------· ----·------·-- LIST OF TABLES
Page
Table 1. Mean Coefficient of Variation for all variables measured 5 times on each of 5 specimens. 18
Table 2. Summary of meristic characters of Agosia chrysogaster. 49
Table 3. Numbers of specimens of each genotype found in a survey of 15 1oc i i n 25 pop u 1 at i on s from 22 ge ogr ap h i c localities. 51
Table 4. Numbers of specimens of each genotype found in a survey of 15 loci in 25 populations from 22 geographic 1 oc a 1 i t i e s • 59
Table 5. Genetic similarities among samples- Nei 1 s <1978) unbiased genetic identity above diagonal, and Rogers 1 <1972) genetic similarity below diagonal. 62
vii
------·-·· LI ST OF FI GURES
Page Figure 1. Map of distribution of Agosia chrrsogaster with localities of samples used 5
Figure 2. Lateral and ventral views of female Agosia chrysogaster and dorsa 1 v i ew of rna 1e 13
Figure 3_. Plot of sample mean scores on first canonical variates from 3 separate DFA~s maximizing distances among samples
Figur~ 4. Scatter plot of scores on first two canonical variates of pooled, within-sample covariance matrix of females. 25
Figure 5. Plot of scores for all basins on sheared 2nd and 3rd components
and southern (stars> morphs of Agosia chrysogaster at grand mean SL. 34 Figure 7. Overlaid average shapes of females of northern (solid line) and southern (dotted line) morphs of Agosia chrysogaster at grand mean SL. Average measures at mean SL were predicted for each morph from non-linear regressions of each variable on SL. 36
Figu~e 8. Overlaid average shapes of males Figure 9. Overlaid average shapes of males Figure 11. Percentages of total numbers of specimens used in morphometric analyses by month of collection and morph viii ------· Page Figure 12. Relationship of Mahalanobis' multivariate distances among samples based on 49 morphometric variables, and •Pair level.• 53 Figure 13. Phenogram of genetic relationships of samples of Agosia chrysogaster 66 Figure 14. Unrooted Wagner network of relationships of samples of Agosia chrysogaster derived by clustering of Roger's (1972) genetic distance coefficients <1 -similarity) with the distance Wagner procedure of BIOSYS-1 Figure 16. Plot of S. i X INTRODUCTI~ Objectives A number of major problems confront biogeographers analyzing evolutionary histories of western North American fishes. Studies of geology and paleoecology are complete enough that hypotheses of area relationships can be formulated for testing against biological data distributions required for such tests Cracraft, 197S; Craw and Weston, 1984; Hendrickson, 1986; Nelson and Platnick, 1981; Nelson and Rosen, 1981; Rosen, 197S, 1978, 1979>, are inadequate or entirely lacking for the region's fishes. This study partially ameliorates the paucity of systematic data on Western North American fishes. The genus Agosia was chosen primarily because it occurs in one of the ichthyologically least-studied regions of North America. Taxonomy within the genus has not been examined since the 18B9s (Jordan, 1886, 1891>, although Miller <19S9> and McNatt (1974> suggested that specimens from the Rio Yaqui system were taxonomically distinct from those of the Gila River basin. Additionally, a second taxon 198S>, shares much of its distribution with Agosia. If both fishes attained their distributions via the same geologic events, they would display congruent phylogenies and area relationships. Failure in this ---- . ------···· 2 regard would indicate either independent derivations of distributions, differential dispersals of species, or errors in reconstruction of phylogenies. Development of phylogenetic hypotheses for Agosia is thus required before the biogeographic hypothesis that it and f. occidentalis share the same history can be tested in the sense of vicariance biogeographers After preliminary analyses, a detailed phenetic approach to relationships analysis was decided upon for several reasons. First, samples from throughout the range of Agosia indicated little inter-sample differentiation, and a lack of discernible osteological or other anatomical characters appropriate for use in a strict cladistic analysis. Furthermore, metric characters appropriate for recoding to discontinuous character states of measurements 1985; Winans, 1984) and Jshearing' of Principal Components et al ., 1981; Strauss, 1984; Bookstein et al ., 1985), provide powerful methods of analyzing morphometric data and comparing shapes and shape ------·-··- -- ·--·· ·------··--·----· 3 changes among groups. Such analyses can lead to discovery of characters useful in cladistic analyses. Several hypotheses, stated as follows, were addressed with the morphometric data set: 1. Populations of each major river basin will be morphologically differentiated from one another. Genetic isolation by hydrographic divides and genetic drift or differential selection since time of isolation are the means by which such differentiation may have evolved. If morphological divergence is constant in rate, and convergence is absent, patterns of phenetic relationships should approach the true phylogeny. 2. Morphological distances among populations will be positively and linearly related to hydrographic distance among them. While panmixia should prevent significant morphological differentiation among local, hydrographically proximal populations, deviation from panmictic conditions will increase as a function of inter-locality distance. Deviations from 1 inearity in this relationship reflect either variations in gene flow, heterogeneity of differentiation rates, or polyphyletic origins of single basin populations. Since drainage distance .is believed the parameter most correlated with extent of genetic exchange, various other measures of inter-locality geographic distances 3. Morphology will be uncorrelated with measures of ecological conditions. Endler <1982a, b, 1983) and Chernoff <1982) have pointed out some difficulties and importance of discriminating ------·-----~- ~~-- -~- ·-·-- -~~ 4 ecological from phylogenetic causes of variation. If morphology is not affected by ecology, usefulness of morphological differentiation as an indicator of phylogeny is increased. 4. Patterns of morphological variation will match patterns of independently analyzed meristic, electrophoretic, and other data sets .on relationships among populations of Agosia chrysogaster. Congruent patterns should be evident as well in phylogenetic relationships of other groups similarly affected by the same events in earth history. Failure to find congruence among independent data sets indicates failure of data to depict true phylogenies or differences in history among the groups being tested. Distribution, Biology and Ecology of Agosia chrysogaster The natural geographic range of Agosia chrysogaster extends from the Bill Williams River drainage <350 N> of Arizona, U.S.A. 1989; Hendrickson, 1984>. It occurs in all major, intervening tributaries of the Sea of Cortez apparent exception of the Rio Matape of Sonora the taxon is associated with a diverse m~alian, avian and herpetofauna of La Brean age Non-native populations of Agosia, apparently the result of use of ------··- --·------····----·--- 5 Figure 1. Map of distribution of Agosia chrysogaster with localities of samples used. Outlined area is approximate natural distribution of the species. Open circles mark samples used in morphometric analysis only. Solid circles are for those used in both electrophoretic and morphometric analyses. Half-solid circles indicate those used only in electrophoresis. ------6 RIO GRANDE HUALAPAI LAKE BILL WILLIAMS _ __.-.~ MIMBRES GILA ..,1.~""'--,.,.. COLORADO~ SONOYTA -..,_..;1-4 CONCEPCION WILLCOX PLAYA SONORA MATAPE -~_; YAQUI COCORAQUI MAYO-- 0 200 400 km FUERTE SINALOA ------7 the species as bait Grande No specimens have been taken in recent studies in the Virgin RiverJ Arizona and Nevada (e.g., Cross, 1985; unpublished data, J. E. Deacon, pers. comm.>. That population Desert region to which Agosia chrysogaster is native were recently reviewed by Turner and Brown <1982>, and aquatic habitats of the region described by Hendrickson et al. <1981>, Hinckley and Brown (1982>, and Hendrickson and Hinckley <1985>. Paleoclimatology was reviewed by Hinckley et al. <1986>, who further described the region's geologic history. Aoosia chrrsogaster is typically the most abundant fish in Soncran Desert stre~s. While mostly at low to mid-elevation (less than 1500 m>, it penetrates to 2873 m elevation Hinckley, 1981; Schreiber, 1978; Schreiber and Hinckley, 1982; Fisher et al ., 1981> cyprinid is highly successful and apparently well adapted to low-order tributaries, persisting through a broad, highly fluctuant and unpredictable spectrum of conditions ranging from dramatic flash floods ~·~'nckley and Meffe, in press>, and individuals survived under algal mats in stream reaches lacking daytime surface flqws.{Minc~ley, 1973; Deacon and Hinckley, 1974), Decreased nocturnal evapo-transpiration renewed surface discharge and allowed the fish to leave algal mats to forage. High fecundity and prolonged reproductive period Barber, 1979; Rinne, 1975; Lewis, 1978a; Kepner, 1982> provide rapid recovery from events which reduce populations. An apparent high vagil ity adaptability were provided by Lewis <1978b>, who provided data on tolerance of Agosia chrysogaster to toxic mine effluents, Lowe et al. <1967>, who studied its survival in low concentrations of dissolved oxygen, and by Marsh and Hinckley's <1982> record of the species from Phoenix metropolitan area irrigation canals. Breeding behavior and spawning habitats of a. chrysogaste~ in the Gila River basin have been described by Hinckley and Barber <1979>, ---·-···------9 Hinckley <1973>, Lewis (197Bb>, and Kepner <1982). Saucer-shaped nests are excavated to a depth of 1 - 4 em in sand or gravel of shallow areas (3- 18 em) with little current, generally at mouths of backwaters. Males do not defend territories, but move with apparent randomness over nesting areas. Single females enter the area and spawn in a brief flurry of sand, with one or rarely 2 males, in the depressions. Females quickly depart from spawning areas and males resume patrolling. Eggs are non-adhesive. Any one female generally carries ova in 3 stages of development. Spawning periodicity is multimodal, and at least some eggs appear ready to be spawned at almost any time of the year. Individuals reach maturity asynchronously, but spawning activity of populations peaks in spring and autumn. Fecundity ranged from 8 to 379 ova per female in Kepner~s (1982) study. Low numbers of eggs per nest indicate fractional spawning and distribution of egg complements across numerous nests. Postlarval anatomy and development have been described by Winn and Miller <1954). Agosia chrysogaster is parasitized by a diversity of organisms significant biological impacts ------· Taxonomic History The genus Agosia has a relatively simple taxonomic history. Girard <1856) described 2 species, a. chrrsogaster and a. metallica, respectively from the Santa Cruz and San Pedro rivers of the Gila River drainage, and shortly afterwards published figures These taxa were subsequently synonymized with B· chrysogaster page priority) by Evermann and Rutter <1895). In the interim, Cope and Yarrow <1875) examined other material from Fort Lowell, Arizona, on the Gila River, which they described as Hyborynchus siderius. Jordan and Gilbert (1883) referred these specimens to the genus Zophendum. Jordan (1891) noted the Camp Lowell Zophendum did not differ from specimens of a. chrysogaster from the Rio Sonora. Over the last 59 years, taxonomic status has been stable, and the genus has been accepted as monotypic aside from comments by Miller <1959> and McNatt <1974> who noted that Rio Yaqui populations may represent a distinct, undescribed form. Relationships of North American cyprinids have long been unclear, and phyletic affinities of the genus Agosia remain so. Many taxa now believed to be distant relatives of AQosia have at one time been placed into the genus and Miller (1948a) considered Moapa to be its nearest relative. Coad 1 inked Moapa with Eremichthys, at some distance from the group containing Agosia. Recently, Hinckley et al. (1986) 1 based on recent geological data, speculated that relationships of various western North ------·-~-- ·----···-· ---- . ---·-·------···-· 11 American taxa may I ie to the south, and in particular that nearest relatives of Agosia may be found within Algansea, a central Mexican endemic. At least phenetically, some taxa currently in Algansea Algansea aphanea) are remarkably similar to Agosia. However, Barbour and Miller (1978> hypothesized the sister group of Algansea to be subgenus Temeculina of genus Gila. This relationship, which makes Gila paraphyletic, is based principally on a single synapomorphy taxonomy of certain taxa assigned provisionally to Notropis Although I do not further address relationships of the genus Agosia, a comment relevant to future studies is appropriate. While taxa previously hypothesized as its sister groups are western North American forms, such need not be the case. For example, Howes (1984) considers that Pogonichthys and Ptychocheilus may be North American representatives of the otherwise eastern Asian aspinine cyprinids. His work lends support to earlier speculations of such trans-Pacific relationships Morphometric and Meristic Analyses Specimens were collected by seines and electrofishing, preserved in 19/. formalin and transferred to 70/. ethanol after rinsing in water. Host preserved material is in the Co11 ect ion of Fishes, Arizona State University specimens from the same samples used for morphometric analyses. Meristics data were compiled on at least one sample of 39 specimens from each major basin. A truss system dimensions, was used for morphometric measurements. Data were taken on 49 dimensions specimens representing 69 spatially and/or temporally separated samples the IBM 3981 and 3999 mainframe computers at ASU, employing programs in versions 5.98 and 5.16 of the Statistical Analysis System Figure 2. Lateral and ventral views of female Agosia chrysogaster and dorsal view of mate.· Measurements used in moprhometric analyses are indicated. ------·-- ·-··-·- 14 15 characters were ambiguous. Each major sub-basin of the Gila River drainage was represented by similar sample sizes. Morphometric data en non-native populations of Hualapai Lake, Mimbres River, and Rio Grande drainages also were analyzed. Data were subjected to descriptive, univariate statistical analyses specimen deletion. Unless otherwise specified, all analyses were conducted separately on each sex and significance levels of PS.05 applied in all tests. Since sexual dimorphism confounded many analyses, results presented here are of analyses on females only unless stated otherwise. Multivariate analyses of morphometric data were done using SAS routines the PROC MATRIX code presented by Bookstein et al. (1985); after corrections of minor errors, including those pointed out by Rohlf (in press), Electrophoretic Analyses Electrophoretic analyses were done by Zneimer <1986>. Specimens were frozen fresh on dry ice in the field and maintained at -800 C for a maximum of 5 years prior to analysis. Populations used in the ------·· ------·-·--····------· ------. ·-··-·· 16 electrophoretic survey are indicated in Appendix B. Buffer systems and stains employed to survey 15 presumptive loci resolved from mixed tissue extracts are described in Zneimer <1986>. All analyses of electrophoretic data were done using BIOSYS-1 Univariate Analyses Measurement precision was analyzed by compiling 5 repeated measurements of all variables on each of 5 individual specimens of varied size and sex from 5 different collections. Means, standard errors and coefficients cf variation Measurement precision varied over nearly 3 orders of magnitude among characters, but most <42 of 49} were measured relatively precisely, with cv~s ranging from 9.118 to 9.747. Greatest lengths Sex-specific means and 9~/. confidence intervals for all raw variables are summarized in Appendix C. 18 VARIABLE MEAN CV VARIABLE MEAN CV VARIABLE MEAN CV NARORB 3.227 SNTBARB 0.458 HEADW 0.274 ISTHMUS 1.934 OPERNAR 0.425 DOOPERCO 0.264 NARL 1.684 MOUT!id 0.429 DORSFL 0.250 BARB ORB 1.264 BODYW 0.404 PCFL 0.237 PECOPER 9.747 PELVGW 0.490 BPECTO 0.216 NARSNT 0.735 IORB 0.392 PECDO 0.209 OPERCORB 9.704 PELFL 0.372 ANFL 0.205 I NARES 0.677 ORBDIA 0.372 PECOPELO 0.203 IPREMAX 9.629 PELOANO 0.370 PELODI 0.180 SNTBMEM 0.618 ALIPB 8.367 NPECTO 0.172 I BARB 9.616 AN BASE 8.360 ANODO 0.167 PLIPB 0.506 CAUDPD 0.356 PELODO 0.137 BOPERCO 9.505 ANOHP 0.349 HEADD 0.124 DORSBASE 0.501 IOPERCO 0.315 HEADL 9.110 LOPERCO 9.485 ANODI 9.396 SL 9.068 BARBNAR 9.466 HPDORSI 9.284 FL 9.953 TL 0.035 Table 1. Mean Coefficient of Variation for all variables measured 5 times on each of 5 specimens. CV was computed for each specimen, then averaged across specimens. See Appendix A for acronyms and descriptions of variables. Since size differences were present ~ong samples, basins and sexes, it was desirable to isolate these effects. Bookstein et al. <1985) recently reviewed methods of isolating size effects from those of shape in morphometric studies. Basically, three methods have been employed: Principal Components Analysis 1978; Albrecht, 1978; Hills, 1978; Pimentel, 1979; Humphries et al ., 1981; Mosimann and James, 1979; Blackith and Reyment, 1984; Bookstein e t a 1 • , 1985) . Regression techniques were applied to partition variance into size-related and size-free components. For each sex, plots of raw variables ·against SL revealed non-1 inear relationships, which remained after log transformations. Log-log 1 inear regression consistent deviations from linearity. The same was true if individual variables were regressed onto scores of the first principal component indicating better fit to the data. Therefore "size-free• data for each sex were produced by non-linear regression of measures onto the first principal component of pooled, within-sample, covariance matrices. Resultant residuals were used in subsequent analyses. The possibility that the pooled population displayl\'d heterogeneity of allometric relationships was also investigated by regression techniques. Tests revealed differences in regression parameters ~ong basins, which are discussed later. ---· ,_, ______213 Multivariate Analyses of Morphometric Data PCA was used not only to summarize the data set and analyze relationships ~ong variables, but as a means of evaluating •natural• groupings of observations. PCA of the pooled covariance matrix of log-transformed raw measures of all variables produced a single cluster of data points on all plots of component scores. Upon closer ex~ination, however, mean scores for some drainage basins differ~d from others, despite broad overlap in scatters for individual specimens. All variables loaded heavily and positively on the first component Component 2 loaded heavily <-.92) on the imprecisely measured ISTHMUS, with the next largest being ANFL (0.13). Component 3 may be interpreted as a contrast of several vertical and horizontal measures in the anterior head region (loading on BARBORB is 8.38, NARSNT 8.33, NARORB 9.29) with predorsal body lengths NARORB Upon failure of PCA to resolve •natural• clusters in the morphometric data set, DFA was applied. While PCA extr-acts 1 inear combinations of variables (components> accounting for maximal ~ounts ------·------·------21 of variation, it does this with no constraints on orientation of components other than that they be orthogonal and uncorrelated. The first component is oriented on the axis of greatest overall variation, and subsequent components describe sequentially lesser proportions of total variance. DFA likewise extracts 1 inear combinations of variables describing sequentially decreasing amounts of total variance, but does so under the constraints that they maximize inter-class variance while simultaneously minimizing intra-class variation. The technique thus requires some a priori basis of classification of observations. It has a further drawback when compared to PCA. Whereas loadings of individual variables onto components can be interpreted in PCA as shedding 1 ight on structure, intercorrelations among variables greatly confound such interpretations of 1 inear Discriminant Functions. Thus, although these functions may provide discrimination among groups, contributions of indi~Jidual variables remain largely unknown. DFA performed on log-transformed raw measurements for all variables and all specimens produced a discriminant function maximizing multivariate distances among samples, yet a plot of scores on the first 2 canonical variates revealed 1 ittle discrimination among them. Figure 3 illustrates that although scatters of scores for each sample are broadly overlapping, sample means on the first axis The same DFA 1 s were also run on data segregated by sex. Each sex 22 Figure 3. Plot of sample mean scores on first canonical var·iates from 3 separate DFA~s maximizing distances among samples ------23 w ...J :>:"- a: (F) ...... "'I z tn ex IXl IJ) "'I I 24 had comparable patterns of relationships among samples and basins. Based on analyses of raw, log-transformed data on all variables for each sex, and using major group as the class variable in DFA, morphological separation of the 2 major groups was greatest among males Use of size-free data in DFA improved segregation of the same 2 groups. Separation of the forms on the first discriminant function of size-free residuals input to DFA classifying samples was roughly comparable to that of the second discriminant function of raw data on the pooled, within-basin covariance matrix. The 2 major groups were, as found with raw data, a northern one consisting of samples from the Bill Williams, Gila, Sonoyta and Concepcion basins and a southern one comprised of samples from all remaining drainages The function developed from the southern group successfully classified 97/. of females and 99/. of males from the northern form, the northern group 1 s function correctly classified all females, but only 55/. of ·- -··· -- ---·-··------.-----. 25 Figure 4. Scatt~r plot of scores on first two canonical variates of pool~d, within-sample covariance matrix of females. Matrix was of residuals of 49 morphometric variables from nonlinear regressions on PCl from a pooled, within sample covariance matrix of females. Specimens of northern morph indicated by"+", those of southern morph by •x•. CANONICAL DISCRNINANT ANALYSIS ON RESIDUALS OF NONLINEAR REGRESSION ON PC1. CLASS =SAMPLE. CAN2 6 6 X X 4 X X + + ++ + X -1-t +* 3 X lE,c" ~ X +if.-11-,~t+ ++ + ~ ~ )C Xx X .+ +of~-++ + :1=1:+ +-fl'"+ .... +..&...... 2 XX X X~ + i" • X X lC ilf. *X ~ X 't.jJ/ X~ X X ~ x,j6c ~X Xx ~ X :x 0 X lC\l )( XX X X X X )( X x· X X c -1 X ~ X,CX ~ X xX X~ ~ >§c X ~ ~ ~ )( X X X X X )0( XX X -~ A x x *xxx XX x -2 X X ~X >)c X X X X~ XX X N X X X X X X xlll: X ~ X 2 -3 XX X Xx XX -4 -6 -6 -7 -8 -9 I -8 -7 -6 -6 -4 -3 -2 -1 0 2 3 4 5 6 CAN1 X • WUCOX, VACU. SONORA, + • BlJ.. WLLJAUS. COCOlAQU, UAVO, FlERTE SONOYTA&~ t-v &stW.OA 0.. males of the other group. Generalized multivariate (all variables included) distance between sexes was 11.7 in the southern, and 15.8 in the northern group. As was the case with PCA's on pooled matrices, initial sheared components from the matrix of all variables failed to discriminate samples, but on closer analysis provided slightly less overlap among basins than did PCA. Shearing components on the major groups recognized from DFA slightly improved discrimination between them, but still left extensive overlap between scatters of the 2 apparent morphs. Although these sheared components did not provide the high levels of inter-group discrimination seen with DFA, interpretabl il ity in terms of morphology makes PCA of interest at this point since at least some discrimination exists and loadings on Discriminant Functions are not easily interpretted. Although variance within each major group was high, differences between them were present on components which have interpretable loading structures. Sheared PC-3 for females, for example, on which differences between morphs were detected, contrasted measures of the anterior head region with pre-dorsal and pre-pelvic body lengths and depths. Since certain imprecisely-measured characters consistently loaded highly on principal components extracted from all variables, and interpretation of loadings from 49 variables was complex, sub-sets of variables were input to sheared PCA analyses to simplify analyses. First, variables were ranked by overall estimates of measurement precision, and the 36 most precise characters were used with major grcup (mor~~) as the grouping variable. This reduced data set provided 28 slightly more graphical separation of major groups on PC plots than those based on the entire data matrix. Differences between major groups were greatest for males on H3 (sheared PC3>, which principally described variation H4 which is a combination of PELOANO <-.51), CAUDPD <.36) 1 and HPDORSI (-.42). By focusing on discrete areas of body shape, the character set was further reduced, yet still retained discriminatory power. Using only four characters from the anterior body region, and producing sheared PC's on the pooled within-group covariance matrix, H2 of the analysis on males becomes a contrast between PELODO (-.8) and DOOPERCO (,5). The major groups differ noticeably on mean scores on this axis major groups on these axes is more complete, and interpretation of loadings is clearer than for axes based on the full data set. After some experimentation with effects of various subsets of variables on outcome of sheared PCA, a 19-variable subset was selected that maximized inter-basin and inter-morph discrimination. While not as distant as in plots of discriminant functions, the same groups recognized with DFA are evident in Figure 5, a plot of scores on sheared components extracted from this subset of variables for males. ------·- -- ·- --· ·---·-·· ------. ···- -·· 29 Figure 5. Plot of scores for all basins on sheared 2nd and 3rd components H2 H3 SNTBARB 9.732 -9.337 BODYW 9.199 9.391 DOOPERCO -9.288 -0.379 PECDO -0.216 -9.059 PECOPELO 0.162 -9.038 PELODI -9.933 -9.299 ANODO -9.235 0.251 PELODO 9.105 IL269 HPDORSI -9.038 0.596 PELOANO -9.444 -9.263 30 (J) (V")' . N I ~~------~------~~ ~ ~ (Y') L() I • . I 0 . ·-·· ------31 Plots of the same components from females only display an equivalent geometry of relationships, but with slightly less distance between basins. Allometric Differences Between Morphs After recognition of morphometric heterogeneity, a variation on the regression method of size correction applied earlier to pooled data was performed on partitioned data sets for variable-by-variable analysis of among-group differences. Appendix D presents r.onl inear regression statistics for each variable regressed on SL by sex in each major group. These relationships were later used in shape reconstructions (see below>, but also demonstrate differences in allometries between groups and sexes. For example, a large proportion of variables have nonlinear log-log relationships with SL in at least one sex in one or both major groups. In the northern group, 35 <7~/.) of the 48 variables regressed on SL had significantly nonlinear relationships with that variable in at least one sex. Twenty-three (66/.) were positively nonlinear in both sexes while 4 <11/.) had negative quadratic terms. Twelve were nonlinear in males only <5 negative, 7 positive). In contrast, the southern group had only one variable nonlinearly related to SL in both sexes. A total of 14 additional variables were non-linear on SL in one sex only <14 positive, 1 negative), Females had a positive quadratic term on 19 variables, in which the same term was positive for both sexes in the other major group. Non-linearity with SL was demonstrated for only 3 variables (2 positive, 1 negative) in males of the southern group, besides the one for which both sexes are non-1 inear. Non-linearity was found in 3 variables in the southern group, for which no significant deviations from I inearity were indicated in the northern group, 2 of these were in males. Only one character, pectoral fin length As indicated by lack of overlap of asymptotically-estimated, 9~/. confidence intervals, a number of variables differed in one or more regression parameters between sexes and/or major groups. Within the northern group, sexes were significantly different in each intercept, slope and quadratic term of the pelvic-fin-length to SL. Differences between sexes were not demonstrated in the southern group for the relationship of any variable to SL. Although not different between sexes in the northern group, barbel to nares other variables differed in regression relationships on SL between opposite sexes of different major groups. 33 Shape Reconstructions The truss system of measurement is a geometrically logical improvement over traditional means of morphometric data collection, which often included measurements between non-homologous points and typically were heavily biased toward disjointed and compounded length dimensions as well as disjunct depth measures 1982; Bookstein et al ., 1985; Strauss and Fuiman, 1985). Perhaps greatest advantages of the truss system result from an ability to derive average shapes for groups of observations, permitting graphic comparison of size-free shape differences among groups relationship, provide graphic confirmation of results from DFA and PCA. Again, it is obvious that greatest differences between major groups in each sex are in pre-dorsal the head, especially the mouth variation from the overall Size/SL relationship. Overlaid reconstructions of mean shape for each sex within the ------·------34 Figure 6. Overlaid average shapes of males of northern (dashes) and southern ------35 ~ :1 ·-1 :; ==- 12:=· I ...... '.,~\ i2 \ ...... , ...... I ..••••••••••- ...... - . 36 Figure 7. Overlaid average shapes of females of northern ·------· ·----·· ----- ·-··. ..,.., •J{ mm..... '''''''' 38 major groups illustrate the trend toward greater body depths and shorter fins of females in each major group. Overall, the extent of sexual dimorphism is slightly greater in the northern than in the southern major group. This fact is evident from overlaid shape reconstructions Temporal Variation in Morphology of Agosia chrrsogaster A series of 8 samples taken from a single locality over a period of 39 months (7 over a 12-month period) was used to assess extent of temporal variation. Since growth and changes in condition factor were expected among samples in this series, it was not surprising that DFA of raw data discriminated among samples. Inputting size-free data for females into DFA, which maximized distances among samples, revealed that samples taken less than 2 months apart had mean scores on that axis that are as different as means on the same axis of the nearest inter-major group pairs Figure 8. Overlaid average shapes of males (dashes) and females . ------43 --- I I I I I I I I I I I I I I I ------41 Figure 9. Overlaid average ~hapes of males -·-·------· ·---·--·------··------· ------. --·-·-. 42 • I ._, .... J:: ...... IJ ...... li I '\ . \ I- ~·-·-...... ------~ 43 Figu~e 10. Mean scores for samples on first two canonical variates of pooled, within-morph covariance matrix of females. Matrix was of residuals of 19 morphometric variables ------·- --- ·--· -----····------·--·· 44 m (..) "' m 01 "' "" "'m Ill ., m "' "" m < 0 0 :z ct u :z 0 ct UJ :z:: > > > "' "' > 30 "' C\J o .... ' "'"' "' "' 0 ... en ~ N -.--.-.. -.·:--· . .--~~~~-r~~--~~~~-r~-T~~~-r~----~-C\J_L~ "'~--- 0 u ' :z ct w :0: 45 both inter-morph and intra-locality differences are great along the s&me axis indicates that temporal variaton parallels shape differences between the major morphological groups. On discriminant functions constructed to maximize between-morph and minimize within-morph variance, distances &mong Bonita Creek samples became small in comparison to inter-morph distances, but still encompassed the full range of variation within the northern morph on this improved between-morph discriminator. The same geometry of inter-morph and inter-sample 19-variable reduced data set. Although multiple, temporally-isolated samples from single localities in the range of the southern morph were not available for detailed analyses, temporally discrete pairs of samples were available for 2 localities. These also display large multivariate distances on the same inter-morph discriminators. Thus the morphs apparently do not differ in the way their shapes vary over time. High levels of temporal variability indicated that morphs detected by DFA could be an artifact of seasonal variation, if seasonal distributions of samples differed significantly between the 2 geographic areas. To evaluate this possibility, frequency distributions of numbers of specimens by month of collection for each group were compared Despite distributional differences, the average time of collection of --- ~ ------~ ~ ------~--~ ~~. 46 35 + ss I I ss I I ss I I ss ss I I ss ss 38 + I ss ss I ss ss I I tfl ss ss I I ss tfl ss ss I I ss tfl ss ss 25 + tfl ss ss I ss I ttl I ss ss ss p I ss tfl ss I ss E I tfl I ss ss ss R I ss tfl ss ss c + tfl 28 I ss ss ss E I ss tfl tfl ss ss I N I tfl tfl I ss ss ss T I ss tfl tfl ss ss I A I ss tfl tfl ss ss G + tfl tfl 15 I ss ss ss E I tfl tfl I ss ss ss I ss ttl tfl ss I ss I ss tfl tfl ss ss I I ss tfl tfl tfl ss ss 18 + tfl tfl tfl I ss ss ss I ss tfl tfl tfl ss ss I I tfl tfl tfl tfl ss ss ttl I ss I ss tfl tfl tfl tfl ss ss tfl I I ss tfl tfl tfl ttl ss tfl ss tfl tfl 5 + ss tfl tfl tfl tfl ss tfl ss tfl ss tfl tfl tfl tfl I I ss tfl tfl tfl ttl ss tfl ss tfl ss tfl tfl tfl ttl : ttl ss ttl ttl ttl ttl ss tf-1 ss tf-1 ss tf-1 ttl ttl ttl : tfi ss tfl tfi tfl tfl ss tfi ss tfi ss tfi tfl tfl ttl : tfi ss tfl tfi ss tfi tfi ss tfi ss tfi ss tfi tfi tfi tfi ------J F 11 A 11 J J A s 0 N D Figure 11. Percentages of total numbers of specimens used in morphometric analyses by month of collection and morph ------47 southern morph specimens is only about one month earlier than that for the northern form. Additionally, temporal variation at Bonita Creek on canonical variates scores from refined DFA~s maximizing discrimination between major groups (by utilization of subsets of variables and a priori classification to major groups) is proportionally far less than seen on canonical variates maximizing inter-sample distances. Therefore, differences between major groups appear to be fixed exaggerations of seasonal morphological variation. Classification of Unknowns to Morph and Basin Using a subset of 19 morphometric variables, mostly from the body-trunk region, a series of classification functions was derived from a subset of 769 specimens of mixed sex. The first of these, designed to provide classification to morphotype, correctly classified 418 (more than 9~/.) of the 443 mixed-sex specimens not used in its development. Interestingly, 11 specimens), were from a single sample Sonoyta, and 19 of those were females. Only one female, but 16 of 17 males, from that sample were correctly classified by this function. Despite this, the function was relatively successful and data on all specimens were used to derive a similar discriminant function usable for classification of future unKnowns. Scores on this discrimator, which maximizes between basin distances, but also provides accurate classification to morphotype, may be computed using the following formula: ------~- -~------~------· 48 X= 12.94966592 - 19.61163418 + 4.57912251 + 15.11293161 where values entered for each variable are logarithms (base 19) of the measurement in millimeters. Specimens of the northern morph score higher on this function than those of the southern group. Using a score of -6.95 as the point of division, this function misclassified to morph only 4.95/. <69 of 1212 total specimens- 21 of 471 males and 39 of 741 females) of all specimens used in this study. Although classification of single specimens to basin of origin is less accurate, the classification of unknown, relatively small samples to basin of origin should be possible. Basin mean scores on this function 9~1. confidence limits>, as well as maxima and minima, are presented in Table 2. ------···- 49 BASIN MEAN L(l.JER95"/.C I UPPER95"/.C I Mimbres -2.9966 -3.634159 -2.179959 Hualapai -3.3942 -3.848842 -2.759558 Concepcion -3.6403 -3.868214 -3.412386 Gi 1a -3.8422 -3.935474 -3.748926 Bi 11 Wi 11 i ams -4.1677 -4.493588 -3.931812 Sonoyta -5.3749 -5.689682 -5.967318 Sonora -6.4728 -6.681924 -6.264576 Yaqui -7.2993 -7.479387 -7.128213 Cocoraqui -7.7926 -8.333794 -7.251496 Fuerte -8.1631 -8.369244 -7.965956 Wi 11 cox Playa -8.2491 -8.462812 -8.917388 Mayo -8.4819 -8.669519 -8.393281 Sinaloa -8.7326 -9.346988 -8.119112 Table 2 - Mean and 95"/. confidence intervals <±2 Standard Errors) of the mean for scores on the 19-variable discriminant function defined in text. Statistics based on analysis of all specimens of Agosia chrysogaster 1 isted in Appendix B. Efficiency and accuracy of this discriminating function clearly demonstrate morphologic distinctiveness of the 2 morphotypes, and, in some instances, differ·entiation among basins. Intermediacy of scores on th1s discriminator for specimens fr-om Rios Sonoyta and Sonora is notable, as is the nature of misclassified specimens. Rio Sonoyta specimens, though previously indicated as part of the southern morphotype, are misclassed with a frequency of 32.~/. <21 of 64). Although 56.~/. of the Rio Sonoyta fish are females, 71.4/. of those misclassified were of that sex, indicating a pronounced tendency toward northern morphology in females. Fish from Rio Sonora were similarly misclassified <27.96/.) 1 with a tendency for males to be most liKely erroneously categorized. The female/male ratio in combined samples from the basin was 1.66, while the ratio for misclassified specimens was 1.39. Specimens of all other basins were accurately ------·· -·-· - ·-··· ·-·· ··--·. ·------. -·· .. 50 classified to morphotype by the 10-variable function. Remaining misclassified specimens were from Rios de la Concepcion (1) 1 Fuerte (2) 1 Yaqui <4>, and the Gila basin (9). For all these basins, rates of correct classification to morph are >911.. Using the same 10-variable discriminant function, 24 specimens from 2 populations (1 and 67 1 Appendix B> introduced outside the native range of Agosia were classified to the northern morphotype. Mean scores from these samples are greater than any basin mean Results of Analyses of Meristic Data ANOVA was used to test the null hypotheses of no differences among basin and morphotype means for each meristic variable. Among basins, significant differences were found for number of scales in the lateral 1 ine DF=10, P=.0992) and number of pectoral fin rays P=.9095), However, significant differences between morphotypes indicated by morphometric analyses, occurred only in lateral-1 ine scales Std. Mean Error- Min. Max. Lateral 1 ine scales (northern morph) 75.10379 9.3763599 69 89 (southern morph) 78.12969 9.3674279 64 89 Dorsa 1 fin rays (northern morph) 8.97925 9.9189748 8 11 (southern morph) 8.95035 9.0299361 8 10 Anal fin rays (northern morph) 7 .975Hl 9.0139592 7 9 (southern morph) 7.97872 0.0187459 7 9 Pectoral fin rays (northern morph) 15.99879 9.9753949 6 18 (southern morph) 15.87949 9 .1908959 6 18 Pelvic fin rays (nor-ther-n morph) 8.96224 9.9227996 7 9 Correlations with Ecology and Geography Mahalanobis' distances from DFA of all 49 size-free morphometric variables from females and classifying on sample were correlated with various measures of inter-locality geographic distance. Straight-1 ine distances among all pairs of localities were geometr-ically computed from latitude/longitude coordinates. Drainage distances between localities were extracted from drainage maps using a rotameter and following coastlines between mouths of basins. Elevational differences among paired localities were also used, as was an index of hydrographic relationships of localities. In the latter, drainages were partitioned into major hydrographic subbasins ~- -~------~----~- C''")_,_ intra-subbasin (2) and intra-locality (1), The last level consists of comparisons of temporally isolated samples from single localities. Of all geographic distance measures, highest correlation of morphological distances among localities was with Pair-level (r2 = 0.49), MahalanobisJ distance is correlated with drainage distance (r2 = 0.23>, while the same statistic for correlation with straight 1 ine distance is r2 = 9.20. Elevational difference between samples shows 1 ittle correlation with inter-locality Mahalanobis distances, whether analyzed within or among·basins. The relationship of Mahalanobis distance to Pair-level is presented in Figure 12. Scores on the first Discriminant Function maximizing inter~orph distances with all morphometric variables were correlated with latitude . Adding latitude to a multiple regression model explains no more variation alone Similar correlations of Principal Components scores (from the pooled within-morph covariance matrix) were also analyzed. Size r2=,133) 1 but not with longitude were correlated with latitude elevation 3 variables at P<.0091 (r2=,337, .089, and .979, respectively). A correlation between latitude and elevation ----·---~- -~-·---~-~------~~---~-~------53 Figure 12. Relationship of Mahalanobis' multivariate distances among samples based on 49 morphometric variables, and "Pair level." Pair level = 1 for comparisons of temporally isolated samples from single localities, 2 for comparisons of geographically separate localities within the same hydrographic subbasin of a major drainage, 3 for inter-subbasin comparisons within single drainages, and 4 for inter-drainage comparisons. Score for· each comparison is indicated by •+•, Means and their 9~/. confidence intervals <±2 Standard Errors) are indicated by vertical 1 i nes. ------54 IMAHALANOBIS DISTANCE AND PAIR LEVEL I PALEVEL lj + + IH 111111_1_ II 3 + * + I IIIIIIIIIIIIIRIIBIIIP••IIIHII'.IIII HI Ill !I -llH+I- 2 + + + -tt+ 1&1 t -itH- + + ++++ + #+t-tt+- -tH- + * 0 2 3 lj 5 6 7 8 9 10 11 12 13 MAHALD PAIRLEVEL INTRA-LOCALITY •1 INTRA-SUBBASIN • 2 INTRA-BASIN • 3 INTER-BASIN • 4 ------._1._1"'"' demonstrates a sampling confuses interpretation of correlations with aspects of morphology. High elevations were sampled only in the Gila River basin; all collections available from southern basins wer·e from low-elevation localities. However·, weak relation:.hips of PC1 the restricted latitudes of the Gila River basin. Linear regressions of meristic variables onto latitude, longitude, and elevation were done to evaluate possible effects of ecology on meristic characters. The highest single correlation was that of lateral-1 ine scales with latitude (r2 = 9.37>. Lateral-1 ine scales correlate little with elevation (r2 = 0.17) and longitude (r2 = 9.12). Other meristic characters were not significantly, or poorly correlated (r2 < 9.99) with these same environmental variables. This was true whether analyses were performed separately on sub-samples An additional indication that morphological characters defining morphs is not highly influenced by ecological conditions was provided by experimental starvation of a sample of fish. Two s~ples <19 and 11) were taken on the same date from a single population. One apparent starvation at the time the remainder were preserved. Subsequent inclusion of these samples in multivariate analyses ------· ------·---···· 56 demonstrated them to differ less on axes which discriminated morphs than did any of the temporally-separated pairs of samples from other localities (see below). Other Characters which Corroborate Morphometric Analyses Preliminary evidence compiled on other characters appears to corroborate the outcome of morphometric analyses. Principal among these development of tuberculation in nuptial males and the possible failure of the southern form to construct a spawning pit. Based on series examined in this study, and many others not specifically used, it appears that tuberculation in southern~. chrysogaster does not reach levels commonly seen in specimens of the northern morphotype. Additionally, I have not observed spawning pits associated with southern~· chrysogaster and know of no reports of such pits by others. Both of these observations may be biased, however, by inadequate representation of samples from Mexico during the breeding season. Though~· chrysogaster in the Gila River basin spawn periodically over a protracted period, a major peaK in reproduction occurs in spring (January- April) tuberculate males are rare during all other seasons Seasonality of reproduction in the southern morphotype has not been studied. Frequency distributions of numbers of specimens used in the morphometric analysis by month of collection light on this matter. Some months are not represented among southern samples, in particular those of what might be peaK spawning periods, ------57 but, unless spawning occurs over a much more restricted time period in the south than it does in the north, highly tuberculate males should have been observed. Mean time of collection of southern specimens is about a month earlier than for northern specimens. In the northern morphotype, large, coarse, cream-colored, horny, dermal tubercles develop profusely within a matrix of fine, evenly distributed, sandpaper-1 iKe tubercles on dorsal and lateral aspects of the cranium of adult, breeding males. Tuberculated area of the head extends from nape anteri·orly to about the anterior edge of the nares. Tubercles are largest on the dorsal surface where arranged in an apparently-random pattern; the size and density decrease only slightly over the preopercle and opercle. Concentrations of tubercles are sometimes found along the dorsal and ventro-posterior rim of the orbit. All fins bear tubercles. Those on paired fins develop only on dorsal surfaces of major rays, from fin base distally to approximately 7~/. of fin length. On the pectoral fin, tuber·cles ar·e 1 imited to 2 parallel rows on the first ray. Each pelvic ray may be vested with a single tubercle row. All medial fins are bilaterally tuberculated, with single rows that may appear on all rays. The first dorsal ray, however, has 2 anteriorly-direch>d, par·allel rows of tubercles on its leading edge. Tubercles do not extend completely to the distal margin on medial fins, but dorsal and caudal rays may have tubercles extending from the base along 59-7~1. of their lengths. Anal fin rays bear tubercles only on medial areas. Each scale of the dorsal and lateral body surfaces may also bear fine tubercles, similar to those of the head. ---- . ---··-··----- ··-- 58 The large, coarse, heavy tuberculation of the head in northern breeding males appears absent in the southern group, as is body tuberculation. However, the same fine, evenly-distributed tubercles which cover the head of the northern subspecies, similarly cover dorsal and lateral head surfaces of this form. Distribution of tubercles on fins appears the same as that in the northern group. While aspects of breeding behavior of e. chrrsogaster from the Gila River basin have been reported (Winn and Miller, 1954; Hinckley and Barber, 1979; Lewis, 1978a; Kepner, 1982), comprehensive studies are not available, and observations on the southern morphotype have not been made. However, as previously noted, nests, or spawning pits such as reported by Hinckley and Barber <1979) and Kepner (1982) for the northern form were not observed during the course of extensive field collections for this study and others in the range of the southern morphotype. Results of Analyses of Electrophoretic Data Isoz~es of Agosia chrysogaster were studied electrophoretically by Zneimer <1986) using specimens from collections also analyzed in this study. Relevant parts of her results are presented here also for comparison with results of morphometric analyses. Table 4 presents numbers of specimens of each genotype found in each population for all loci sampled electrophoretically. Genetic heterozygosity of populations 9.035±9.91). Both Rogers' (1972> and Nei's <1978) indices of overall 59 Table 4. Numbe~s of specimens of each genotype found in a su~vey of 15 loci in 25 populations from 22 geographic localities. Samples numbered as in Appendix B. Adapted from Zneime~ ( 1986). 6B I I SAMPLE : ------:2 5 9 14 18 23 32 33 35 37 38 44 46 : LOCUS ~ru : ======:I I 1. Cbp-1 aa 12 12 18 12 8 18 19 19 18 18 12 12 12 : 2. Ck-A ------:aa 12 18 18 12 14 18 19 19 18 18 12 12 12 : ------:ab : 3. Pep-A bb 12 18 18 18 14 18 19 22 18 8 12 7 : be 5 6 4 : ee 5 12 I : ------: 4. Pep-B aa 12 18 18 12 14 12 13 22 18 12 12 13 12 : ------:aa 1 11 13 : 5. Est-1 ab 1 4 4 1 : bb 18 IB 4 1 14 18 18 19 17 18 18 13 11 : ------: 6. Gp-1 aa 12 17 17 12 14 18 18 18 18 18 12 12 6 : ------: aa 2 13 13 1 ab 2 4 4 11 7. Gpi-A bb 9 18 1 1 14 18 18 18 15 18 18 k 2 cr 1 ------:aa 9 2 17 18 13 16 17 17 14 2 : 8. Gpi-8 ab 3 4 1 2 1 I 3 4 : bb 11 18 18 12 6 : be : ------: 9. Ldh-A aa 12 12 9 9 14 12 13 13 18 12 12 12 12 : ab 3 3 : ------: 18. Ldh-B aa 12 13 12 9 14 12 13 13 18 12 12 12 12 : ~ 3 : ------:aa 12 18 18 18 15 18 19 18 18 13 12 : 1t. 11-Hdh-A ab 1 : ce 18 18 : 12. S-Hdh-8 ------:aa 12 18 18 18 14 18 19 19 18 18 18 13 12 : ------~------: ab 1 : 13. 11-He-A bb 12 18 18 18 8 16 19 18 18 17 14 13 12 : be 6 2 5 : 14. S-He-A ------:aa 12 18 18 18 13 18 19 18 18 18 18 12 12 : ~ : I ------:I aa I 15. PIJ1-A ab I bb 12 18 18 16 14 18 19 19 18 18 18 13 12 : be 2 :I I ·-··------61 Tabl~ 31 eontinu~d !W!PLE ------:47 48 58 51 52 53 56 61 62 63 65 68 l I ======LOCUS ALLELE I 1. Cbp-1 aa 18 12 18 18 18 18 18 12 29 18 12 14 l ------: 2. Ck-A aa 16 18 18 18 18 18 18 18 29 18 12 14 l ------: ~ 2 l 3. Pep-A bb 16 19 18 18 18 18 18 18 18 18 9 6 l be l ee l ------: 4. P~p-8 aa 16 13 12 12 18 12 12 12 18 12 12 12 l ------: aa l 5. E~t-1 ab 1 l bb 16 19 18 18 18 18 17 18 29 18 13 15 l ------: 6. Gp-1 aa 16 18 18 18 18 18 18 18 16 18 12 6 l ------·aa 4 4 ab 5 8 7. Gpi-A bb 16 18 18 3 6 5 9 6 19 18 18 13 be 3 3 6 2 ee 12 6 7 aa 16 18 16 17 16 15 14 18 4 12 B. Gpi-8 ab 2 1 2 2 3 2 1 bb 1 18 17 1 l be 1 l 9. Ldh-A ------:aa 16 13 12 12 18 12 12 12 18 12 12 12 l ab l ------: 18. Ldh-B aa 16 13 12 12 18 12 12 12 18 12 12 12 l ab l ------: aa 16 19 18 18 18 18 18 18 18 18 12 19 l l 11. M-Hdh-A ab I ee I ------: 12. S-Hdh-B aa 16 19 18 18 12 18 18 18 18 18 13 11 : ------:ab 1 l 13. M-H~-A bb 12 18 18 17 18 18 18 16 14 18 9 18 l be 4 1 2 4 2 l ------: 14. S-H~-A aa 16 18 18 18 18 18 12 18 16 18 11 18 l ab 2 l ------:I aa 6 1 2 18 18 18 4 I 15. PS1-A ab 9 18 8 1 7 bb 1 1 8 17 18 28 18 1 11 be -----· --·- --· ------·· ----··--··-----·· --···. 62 TableS. Genetic similarities among samples- Nei's (1978) unbiased genetic identity above diagonal 1 and Rogers' <1972> genetic similarity below diagonal. Samples numbered as in Appendix B. Data from Zneimer (1986). 63 Sfl'IPLE 2 5 9 14 18 23 32 33 35 37 38 44 46 2 .97 .98 .92 1.88 1.88 1.88 1.80 1.88 .87 .83 .95 .98 5 .94 .92 .a6 .96 .97 .97 .96 .97 .92 .8a 1.88 1.88 9 .94 .aa .97 .96 .96 .96 .96 .96 .81 .77 .98 .93 14 .a8 .a2 .94 .as .86 .87 .87 .86 .75 .71 .83 .88 18 .97 .94 .91 .98 1.88 1.88 1.88 1.88 .85 .82 .94 .97 23 .98 .95 .92 .98 .99 1.98 1.88 1.88 .86 .82 .94 .97 32 .98 .95 .93 .98 .98 1.88 1.88 1.88 .86 .82 .94 .97 33 .98 .95 .93 .98 .98 .99 1.88 .9a .86 .82 .94 .97 35 .98 .95 .92 .91 .97 .98 .99 1.88 .87 .83 .95 .98 s 37 .a3 .a9 .77 .71 .a3 .a4 .a4 .84 .84 .99 .92 .92 A 38 .a8 .as .73 .67 .81 .81 .81 .81 .88 .96 .88 .98 f1 44 .93 .98 .a6 .88 .92 .93 .94 .93 .94 .98 .87 .95 p 46 .94 .97 .a8 .a2 .92 .93 .93 .93 .94 .89 .86 .99 L 47 .93 .98 .a8 .a2 .95 .95 .94 .94 .93 .79 .77 .88 .88 E 48 .96 .93 .91 .85 .96 .97 .97 .97 .96 .82 .78 .91 .91 58 .96 .93 .98 .as .96 .97 .9a .97 .97 .82 .7a .92 .91 51 .87 .83 .83 .81 .87 .sa .sa .88 .88 .72 .69 .82 .82 52 .a9 .as .a5 .at .as .98 .98 .98 .98 .74 .71 .a4 .a4 53 .a9 .as .a4 .at .a8 .a9 .89 .a9 .98 .74 .71 .a4 .a4 56 .92 .96 .a7 .a3 .89 .91 .91 .91 .92 .sa .87 .99 .99 61 .91 .95 .a6 .83 .98 .91 .91 .98 .91 .a7 .87 .99 .98 62 .97 .95 .91 .as .98 .99 .98 .98 .9a .84 .83 .95 .98 63 .98 .95 .93 .87 .98 .99 1.88 1.88 .98 .a4 .a2 .93 .97 65 .93 .98 .87 .81 .93 .94 .93 .93 .94 .88 .at .93 .95 68 .98 .95 .93 .87 .98 .99 1.88 1.88 .98 .85 .a2 .93 .97 Sfl'IPLE 47 48 58 51 52 53 56 61 62 63 65 68 2 .97 .99 .99 .98 .92 .91 .94 .94 1.88 1.88 .97 1.88 5 .93 .95 .96 .86 .88 .88 .99 .98 .97 .96 .95 .96 9 .93 .96 .96 .87 .8a .Sa .98 .98 .96 .96 .93 .96 14 .a7 .98 .98 .a4 .85 .a4 .a7 .a7 .98 .98 .aa .98 18 .97 .99 .99 .a9 .91 .91 .93 .92 1.88 1.88 .97 1.88 23 .97 .93 .93 .a9 .92 .91 .93 .93 1.88 1.88 .97 1.88 32 .97 .99 .99 .89 .92 .91 .93 .93 1.88 1.88 .97 1.88 33 .97 .99 .99 .a9 .92 .91 .93 .92 1.98 1.88 .97 1.88 35 .97 .99 .99 .98 .92 .92 .94 .94 1.88 1.88 .97 1.88 s 37 .82 .84 .as .75 .77 .77 .91 .91 .a7 .as .84 .a4 A 38 .7a .81 .81 .71 .73 .74 .a4 .84 .82 .88 .78 .88 f1 44 .98 .93 .93 .a3 .86 .86 .97 .97 .93 .93 .as .93 p 46 .93 .96 .96 .86 .89 .89 .94 .93 .93 .93 .98 .93 L 47 .99 .99 .95 .97 .97 .86 .86 .94 .95 .97 .95 E 48 .97 1.88 .93 .95 .95 .89 .88 .96 .98 .95 .98 58 .97 1.98 .93 .95 .95 .89 .a8 .96 .97 .95 .97 51 .92 .98 .98 1.88 1.88 .82 .82 .87 .88 .91 .88 52 .93 .92 .92 .98 1.88 .84 .84 .88 .98 .93 .98 53 .92 .91 .92 .98 .99 .84 .84 .88 .89 .93 .89 56 .a9 .92 .92 .84 .a6 .a6 .99 .98 .91 .a7 .91 61 .89 .91 .92 .84 .86 .86 1.88 .91 .98 .87 .98 62 .97 .99 .99 .89 .91 .91 .94 .94 1.88 .94 .98 63 .97 .99 .99 .98 .92 .91 .92 .92 .98 .93 1.88 65 1.88 .99 1.98 .96 .9a .98 .92 .92 .97 .97 .93 68 .97 .99 .99 .98 .92 .91 .92 .92 1.88 1.88 .97 ---·------· 64 genetic similarity are in Table 5. Scores for genetically identical pairs of samples are 1.9 in both indices, with increasing divergence lowering scores. Minimum similarity between samples, as measured by Rogers' index, was 9.67, but 95.3/. of all 399 possible sample pairs had scores~ 9.89. Genetically identical sample pairs comprised 4.~/. of the tota 1 ; Zneimer <1986) found significant deviations from Hardy-weinberg expectations, indicating lack of gene flow, at both inter- and intra-basin levels in e. chrrsogaster. Great distances among localities and hydrographic divides may be responsible for impeding gene flow at these levels, however, 5 significant, intra-sample deviations from Hardy~einberg expectations were also noted 32, 46, 51 and 52>. Reasons for these are not known, however, 3 of the 5 deviations were at the Gpi-A locus. Five presumptive loci were monomorphic across all samples. Three polymorphic loci Yaqui) and Turkey Creek unique Peptidase allele Est-1 in Coon CreeK and Verde River rare; and synapomorphic presence of the b allele of Ldh-A only in Coon CreeK and Verde River samples <14 and 9). An UPGMA phenogr~ based on the genetic identity coefficient of Nei <1978) is presented in Figure 13. It clusters populations by the unweighted pair-group method with arithmetic means Electrophoretic data were input to BIOSYS-1 1972; Lundberg, 1972; SWofford, 1981) based on Rogers~ Genetic Similarity Coefficients the longest branch. Results are in Figure 14. Obvious discrepancies were noted between relation?hips patterns extracted from electrophoretic data and those indicated by morphology. Since lack of congruence between morphological and electrophoretic patterns might be the result of analysis of different sets of samples, the morphometric data set was re-analyzed after deletion of samples not represented in electrophoretic analyses. Re-analysis clearly indicated existence of the same northern and southern morphological groups indicated by analyses of the full morphometric data set. DFA designed 66 Figure 13. Phenogram of genetic relationships of s~ples of Agosia chrysogaster. Derived by clustering in BIOSYS-1 -----~ ----~-~~ ----- ~~---- GENETIC DISTANCE 0.16 0.12 0.08 0.04 0 .------.- 2 HUALAPAI 35 GILA 33 GILA 23 GILA 32 GILA 68 MIMBRES 63 CONCEPCION 62 CONCEPCION 18 GILA 65 SONOYTA 47 MAYO 48 MAYO 50 MAYO 5 BILL WILLIAMS • I 44 YAQUI 46 YAQUI I I ~~ ~g~g~~ L 9 GILA ------{:::::: 14 GILA 51 FUERTE 52 FUERTE 53FUERTE 38YAQUI 37 WILLCOX 0.84 0.88 0.92 0.96 1.00 GENETIC SIMILARITY ~ '-J 68 to maximize distances among various groupings of samples recognized by Zneimer (1986) 1 or apparent in the Wagner network, discriminated morphologically among them, but greatest differences were consistently between the 2 previously recognized, maJor morphological groups, and not those recognized electrophoretically ------69 Figure 14. Unrooted Wagner network of relationships of samples of Agosia chrrsogaster derived by clustering of RogerJs <1972> genetic distance coefficients (1 - similarity) with the distance Wagner procedure of BIOSYS-1 ------70 71 Figure 15. Plot of sample mean scores ------·------· -----· . --··------72 N N N N N N« N N "' "' PI N ....N ....N N N N N N N 'ill N .... N N N NN .. "' N N "' .., N .., z a: (.) .., ,., .., ..,.., -- "' - -.r_ -- .., -- 0 .., "'.:pr> N- --=------.., ...,~ ------=.::~~ -~-;.- - ~ .;: ------.- - _-c:. - .,::- - .., I .., ,.., -_..,; ------..... -- - .., .., _.., ------"' C\J ..,..,-- -- I "' .., .., tn ... I ... .., "' .., .., .., ::I' .., I "' ~ ~~------~------~------r------r------OT------T-----~N~------~------~~------~"? a: I I I u -----· ----·------73 Figure 16. Plot of sample mean scores (sexes combined) on Canonical Variates l and 2 of DFA on 49 log-transformed morphometric variables. Samples are those common to both electrophoretic and morphometric analyses. Canonical Variates maximize multivariate distances among smallest groups of Distance Wagner tree, lettered as in Figure 14. ------·-- . F F E [E F F F F { F F F f t E F EE rtf FE F f FF F E E E E c F F f £rEf fF E E FF ~ft ~ E E E E E E g F f' f FE F t ~ F~ E c E Ff F E E c F ~ ft:JE ~ F E E lj: F Vf F EE ,IE IF r E[E ~ : E c c c c c F EF E t- EjF E c c c c c F F jEEE Ef E E F E Fn: c cccec FEgFi E c c c r' E E E E 8 B c ccc c ~ c ~ c F f F B E E B B c r F r Cc ftQ:cc EB E ~ sm c B E B B c ~ Cl: cc!f:Cc E c c E F F'E E EB B EFiiBii B faa~ B a ccc c c c c c c 0 C C CC C F E E E B8 B BB sB \ B B ~Slit! B D C 8 C CC Q: C C(t C C C C 1119 B BBB CCC f E B ' 9 BB I!.. cc ccccc c 9 B It 88 B lj'fB BB c " Q; C D c E 8A 8 8 8. 8 BBB BB liB BB Bs D A B 8 8 B ~ c c c 9_ ?._% B B A B B -., B 8 B 8 ¥ BBB B B oD " B c B B 8 9 D A A A A A AA AA B B B AA. AA A BA.~ A B t A ---- " " A -."A" __," A A A D A A ""A A " A A A A A " A ~ A A " A A -6 -5 -II -3 -2 -I 0 2 3 li 5 6 1 CAN! ~-J .to DISCUSSION All analyses of the morphometric data set indicate existence of two distinct, allopatric morphs of Agosia chrysogaster. A northern morph is native to the Bill Williams, and Gila Rivers of Arizona and New Mexico, as well as Rios Sonoyta and de la Concepcion of Arizona and Sonora, while the southern morph is from the endorheic Willcox Playa, and rios Yaqui, Sonora, Cocoraqui, Mayo, Fuerte and Sinaloa. The southern morph, furthermore, shows indications of differentiation between populations of the Mayo, Fuerte and Sinaloa basins and those of the Willcox, Yaqui, Cocoraqui and Sonora basins, though these are morphologically more similar to one another than either is to the northern group. Fishes from Rios Sonoyta and Sonora show tendencies toward body shapes intermediate between morphs. Although the relationship between morphs is affected by size and sex, at average sizes the southern form has proportionately longer pre-dorsal and PECDO> lengths, but shorter post-dorsal barbel differ in size as well. Differences in size and shape appear to be the result of different growth relationships. While log-log relationships of individual variables to SL were significantly non-1 inear for most variables in the - ~----~~------~-~----- ~-~~ 76 northern morph, those of the southern were predominately not significantly different from 1 inearity. Northern males had significantly non-linear log-log relationships on SL for 35 of 48 variables; the same figure for females was 23. In the southern form, fewer non-linear relationships were found, and more were in females (10) than males (3). Though sexes are dimorphic in body and fin characters in both groups, differences are most pronounced in the northern. Morphology of &· chrysogaster in temporally separated samples from single localities varied in the same characters that discriminated between morphs. While there were differences between samples of each morph in distribution of times of collection, average month of collection for the southern morph was only about one earlier than that for the northern form. Aspects of morphology important in discrimination between morphs were correlated with latitude; correlations with longitude and elevation were not predictive. Correlation of morphology with latitude may be an artifact of its high correlation with distribution of the morphs. An experiment in which fish from a single sample were separated randomly into a treatment group starved before preservation and measuring, and another group preserved at time of sampling, also demonstrated lack of effects of environment on body shape characteristics. These specimens differed less on multivariate axes discriminating among basins and between morphs than did those from multiple collections from single localities at intervals over a year. It thus appears that aspects of morphology varying between morphs are ·--- 77 little influenced by ecological conditions. Despite temporal variation and that due to sexual dimorphism and size, it was possible to develop a classification function using 19 variables which should provide 9~/. correct classification of unknowns of either sex to morph, and will likely allow assignment of most ~all samples to basin of origin. Morphological patterns of variation were correlated with other observations. Though ranges were broadly overlapping, the southern form had on average about 3 more lateral-1 ine scales than in northern fish and apparently lacks development of large tubercles in nuptial males as in the northern form. Differences may also exist in breeding behaviour. Although either morphological or electrophoretic data set could be used to accurately identify many unknown samples to basin of origin, and morphology is capable of near-perfect discrimination of morphotypes, the two data sets produce different conclusions regarding relationships of the groups and leave morphologically-defined groups without isozymic definition. Similarly, isoyme-based relationships patterns are mostly unpredictive of morphology. For example, the UPGMA phenogram based on electrophoretic data of Zneimer <1986) and loadings on the sheared components are as follows: