I examined hypotheses of Aphelocoma jay phylogeny derived from allozyme data. Results from various algorithms differ in details, but the overall patterns are consistent: Scrub Jays (A. coerulescens) and Unicolored Jays (A. unicolor) were derived independently from different populations of Gray-breasted Jays (A. ultramarina). Within Scrub Jays, the californica subspecies group was derived from the populations of interior North America (woodhouseii group). One Unicolored Jay population and two Scrub Jay populations, all strongly differentiated, are placed consistently at the base of the phylogeny, but phenotypic, biogeographic, and theoretical evidence suggests that these populations represent rapidly evolving populations derived from within populations of their respective species. Because analyses of rates of molecular evolution demonstrate significant rate heterogeneity, I suggest that the application of a molecular clock to date-splitting events in the Aphelocoma jays is not a valid approach. Received 18 February 1991, accepted 3 July 1991.

Committee on Evolutionary Biology, University of Chicago, Chicago, Illinois 60637, USA THE THREE species of Aphelocoma jays range throughout western and southern North Amer- ica and northern Central America (Fig. 1; Pi- telka 1951). Scrub Jays (A. coerulescens) range from Oregon and Wyoming south to the Isth- mus of Tehuantepec, with disjunct populations on Santa Cruz Island off the coast of southern California and in peninsular Florida. The 15 subspecies form five groups, each characterized by unique combinations of plumage, morpho- logical, and behavioral characters (Fig. 1; Pi- telka 1951, Peterson 1991a): (1) woodhouseii group (Wyoming and southeastern Oregon south through Great Basin and along both sides of Rocky Mountains, and then along interior slopes of Sierra Madre Oriental and Sierra Madre Oc- cidental of northern Mexico to southern Chi- huahuan Desert and vicinity of Mexico City); (2) californica group (western Oregon, Califor- nia, and Baja California); (3) sumichrasti group (southern Mexico); (4) coerulescens group (pen- insular Florida); and (5) insularis group (Santa Cruz Island). The Gray-breasted Jay (A. ultra- marina) ranges throughout the mountains of northern and central Mexico and the south- western United States. The seven subspecies fall into three groups characterized by unique com- binations of morphological and behavioral characters (Fig. 1; Pitelka 1951 ): (1) pot osina group (Sierra Madre Oriental); (2) wollweberi group (Si- Present address: Department of Zoology, Field Museum of Natural History, Roosevelt Road at Lake Shore Drive, Chicago, Illinois 60605, USA. erra Madre Occidental); and (3) ultramarina group (Transvolcanic Belt). Finally, the Unicolored Jay (A. unicolor) consists of five allopatric popula- tions, each a separate subspecies, in southern Mexico and northern Central America (Pitelka 1951). The Aphelocoma jays have been the subject of numerous comparative studies. Evaluations of social systems in the genus have led to advances in understanding ecological factors important in the evolution of sociality (Woolfenden and Fitzpatrick 1984, Fitzpatrick and Woolfenden 1986, Brown 1987, Peterson and Burt, in press). Differential habitat use in Scrub Jays (Peterson and Vargas 1992) is correlated with geographic variation in beak shape, suggesting that beak shapes have responded to natural selection (Pe- terson, in prep.). Integration of phylogenetic information into such investigations will allow an important new dimension of understanding (Brooks and McLennan 1990). Hence, I have attempted to estimate the phylogeny of the dif- ferentiated forms in the genus. Rates of molecular evolution.--Since the pub- lication of the influential paper of Zuckerkandl and Pauling (1962), the idea of a "molecular clock" has been controversial in molecular bi- ology and systematics. The clock concept is based on the assumption of a uniform rate of molec- ular evolution in a group. If the uniform rate assumption were correct, the accumulation of genetic differentiation between sister taxa would be time-invariant, and divergence times could be estimated from genetic distances. The clock concept is an important feature of many sectors of molecular systematics. At- tempts have been made to calibrate clocks for a variety of taxa and biochemical data sets (e.g. Sarich 1977, Wilson et al. 1977, Sibley and Ahlquist 1981). Several commonly used tree- building algorithms depend on the assumption of a clock (Felsenstein 1982, 1989). The molec- ular clock is now a common method of dating divergence times between taxa (e.g. Zink 1982, Nevo et al. 1987, Johnson and Marten 1988). Hence, a uniform rate of molecular evolution and the consequent existence of a clock can be important assumptions in molecular systemat- ics. However, the assumption that evolutionary rates are homogeneous often goes untested. If molecular evolutionary rates differ among lin- eages, the clock becomes unreliable. Equality of rates in sister lineages can be tested with the relative-rate test: in the three-taxon statement ((A, B), C), A and B should show similar degrees of differentiation from C (Wilson et al. 1977). Significant departure from equality indicates that A and B have evolved at different rates. Rate uniformity has been tested in a number of taxa, with some studies documenting homo- geneity (e.g. Sibley and Ahlquist 1983, Bledsoe 1987), and others finding heterogeneity (e.g. Britten 1986, Sibley et al. 1987, Sheldon 1987, Springer and Kirsch 1989). A uniform rate and molecular clock are commonly and uncritically assumed in systematic studies, so a test of the assumption of rate uniformity was desirable in my study. METHODS Sampling design.--During 1986-1989, I studied and collected Aphelocoma jays at 35 sites in the United States and Mexico. Tissue from the endangered Flor- ida population of Scrub Jays was salvaged from re- cently dead individuals at nests and roadsides. Sam- ples of heart, liver, and pectoral muscle from each individual were stored in cryotubes in liquid nitro- gen. All tissue was deposited in the Frozen Tissue Collection of the Field Museum of Natural History. In total, I analyzed 615 Aphelocoma jays (458 Scrub Jays, 138 Gray-breasted Jays, and 19 Unicolored Jays). Specimens of Scrub, Gray-breasted, and Unicolored jays were collected at 24, 8, and 3 sites, respectively, in the United States and Mexico (Table 1, Fig. 1). I attempted to obtain samples of 20 individuals at each site to maximize tree stability (Archie et al. 1989): Scrub Jay samples averaged 19.2 (range 13-27) indi- viduals; Gray-breasted Jay samples averaged 17.3 (range 7-22) individuals; and Unicolored Jay samples Fig. 1. Summary of collecting localities and sub- species groups of three species of Aphelocoma jays; Florida range indicated in upper right-hand corner. (A) Scrub Jays: open circle, Santa Cruz Island (ACINS); closed circles; californica group (CO CAL); closed tri- angles, northern woodhouseii group (CO WOO); open triangles, southern woodhouseii group (CO MEX); squares, sumichrasti group; dotted circle, Florida (ACCO2). (B) Gray-breasted Jays: triangles, wollweberi group (UL WOL); circles, potosina group (UL POT); squares, ultramarina group (UL ULT). (C) Unicolored Jays: triangles, east slope populations (UN E); circle, west slope population (ANGUE). TABLE 1. Sample localities and sample sizes (complete locality data in Peterson 1990). Sample Group Sample Locality size Scrub Jays CO CAL CO MEX CO WOO CO SUM Gray-breasted Jays UL WOL UL POT UL ULT Unicolored Jays UN E ACCAL CA, Monterey Co., Bradley 24 ACCAU CA, Trinity Co., Hyampom 16 ACHY3 BCS, Sierra de la Laguna 16 ACIMM OR, Adair Village 15 ACOBc CA, Orange Co., Silverado Canyon 16 ACOBk BCN, Rosa de Castilla (Ojos Negros) 20 ACOOC CA, San Ramon 16 ACSP1 CA, Alturas 22 ACSP3 CA, Bodfish 19 ACCY1 COA, E1 Diamante Pass 21 ACCY2 SLP, Sierra de Bledos 21 ACGR! DUR, Villa Ocampo 21 ACGR3 JAL, Lagos de Moreno 20 ACNE! NV, Toiyabe Mountains 14 ACNE4 NV, Mt. Charleston 27 ACNE6 AZ, Drake 25 ACTEX TX, Edwards Plateau, Carta Valley 22 ACWO! CO, Front Range, Gardner ! 1 ACWO3 NM, Manzano Mountains 26 ACWO4 TX, Davis Mountains 21 ACREM GRO, Xocomanatl/n 13 ACSM2 OAX, Sierra de los Mijes, Zempoaltpetl 20 ACINS CA, Santa Cruz Island 14 ACCO2 FL, Archbold Biol. Station 18 AUAZ1 AZ, Pefia Blanca Lake 16 AUGRA JAL, Sierra de Bolafios 7 AUWO1 ZAC, Valparaiso 20 AUCO2 COA, Sierra E1 Carmen 22 AUPO2 SLP, Sierra de Bledos 22 AUPO3 HID, Jacala 20 AUULT MOR, Huitzilac 14 AUCOL JAL, Sayula 17 ANGUE GRO, Sierra de Atoyac 7 ANOAX OAX, Sierra de los Mijes, Zempoaltpetl 10 ANCON HID, Sierra de la Huasteca, Tlanchinol 2 averaged 6.3 (range 2-10) individuals. To avoid prob- lems with nonindependence of individual genotypes when related individuals (e.g. parents and offspring, multiple siblings) were collected, I included only one member of each potentially related pair of individ- uals. For convenience in referring to population sam- ples, I assigned each a five-letter code, in which the first letter indicates the genus, the second indicates the species, and the remaining three indicate sub- species and (if necessary) the sample from that sub- species (e.g. ACNE2 is the second sample of A__phelo- coma coerulescens nevadae; Table I). For reference to larger infraspecific divisions, I divided the species in a manner equivalent to the subspecies groups of Pi- telka (1951), except that the woodhouseii group was divided into northern (United States populations) and southern (Mexican populations) parts (Fig. I, Table 1). To provide outgroups for phylogenetic analyses, I included 70 individuals of 17 other corvid species. To maximize detection of plesiomorphic alleles for po- tentially closely related genera (Hardy 1969, Hope 1989), I included all material available (up to 20 in- dividuals) of each of the following species: Cyanocitta cristata, C. stelleri, Cyanolyca mirabilis, C. cucullata, C. viridicyana, and Gymnorhinus cyanocephalus. For less closely related corvid genera, I included one or two individuals of each species available: Calocitta formosa, C. colliei, Cyanocorax violaceus, C. cyanomelas, C. yncas, Corvus brachyrhynchos, C. cryptoleucus, Pica pica, P. nut- tallii, Nucifraga columbiana, and Perisoreus canadensis. Electrophoresis.--I homogenized approximately 0.4 g each of heart, muscle, and liver from each sample in a 1 ram disodium EDTA/100 ram Trizma base/0.2 ram NAD, NADP, and ATP buffer. Homogenates were centrifuged at 12,000 rpm for 30-45 min. Superna- tants were drawn into capillary tubes, and stored at -76øC for later use. Gels for electrophoresis were made of 10-12% starch in the appropriate buffer. After electrophoresis, gels were sliced, and stained differ- entially following Shaw and Prasad (1970) and Harris and Hopkinson (1978). Loci studied included ACON (enzyme number 4.2.1.3; Harris and Hopkinson 1978), ACP (3.1.3.2), ADA (3.5.4.4), ADH (1.1.1.1), AK (2 loci, 2.7.4.3), CK (2 loci, 2.7.3.2), ES (2 loci, 3.1.1.1), FUM (2 loci, 4.2.1.2), GDA (3.5.4.3), GOT (2.6.1.1), GPD (1.1.1.8), GPI (5.3.1.9), IDH (1.1.1.42), LDH (2 loci, 1.1.1.27), MDH (2 loci, 1.1.1.37), MPI (5.3.1.8), PEP (leucyl-alanine, 2 loci; leucyl-aminopeptide, 1 locus; 3.4.11), PGD (1.1.1.44), PGM (2.7.5.1), SOD (1.15.1.1), and SORDH (1.1.1.14). Details of electrophoretic con- ditions are given in Peterson (1990). Loci showing difficult-to-interpret patterns of variation were ex- cluded. Stained gels were digitized with an optical imaging system, and visual images were stored in computer files; gels were later discarded. To assure correct assignment of homologies, I in- cluded reference individuals on each gel. Alleles from different populations were not considered equivalent until found indistinguishable in side-by-side com- parisons on the same gel. I rechecked hornology as- signments for each locus under different buffer con- ditions and/or at maximum migration distances (an average of two recheck runs per variable locus; Pe- terson 1990). Tree-building analyses.--I employed a variety of methods to estimate phylogenetic relationships of the Aphelocoma jays. I began with simple methods with restrictive assumptions; then, with successively more complex approaches, I attempted to relax these as- sumptions. I initially analyzed relationships of all 35 population samples, but later combined individual samples and investigated relationships of the groups of samples described above. Although an explicit test of the monophyly of the Aphelocoma jays has yet to be conducted, several lines of evidence indicate that the group is monophyletic. Several characters of plumage, vocalizations, and morphology are characteristic of Aphelocoma jays (Pi- telka 1951). Hope (1989) found that Aphelocoma, Cy- anocitta, and Gymnorhinus form a well-supported group based on phylogenetic analysis of qualitative and mensural osteological characters. Aphelocoma and Cy- anocitta share a synapomorphy (a new bony element in the sclerotic ring; Curtis and Miller 1938) not pre- sent in Gymnorhinus (Peterson, unpubl. data). While Cyanocitta has a number of unique features, Aphelo- coma is not so well defined. Because it seems unlikely that Cyanocitta was derived from within Aphelocoma, I assume for the purposes of this paper that Aphelo- coma is monophyletic. I performed phenetic analyses on matrices of mod- ified Rogers' genetic distances (as modified by Wright 1978), which were calculated using programs devel- oped by Scott M. Lanyon. Allele-frequency data and genetic-distance matrices are available in Peterson (1990) or on request from the author. Exploratory analyses based on other genetic distance measures either produced similar results (e.g. Cavalli-Sforza chord distance) or were difficult to interpret (e.g. Nei's genetic distance), so I report only results from Rogers' distances. Two different clustering methods, the un- weighted pair-group method (UPGMA) and single- linkage cluster analysis, were applied to the genetic distance matrix using NTSYS-pc (version 1.40; Rohlf 1988). One potential problem with strictly phenetic meth- ods is the assumption of equal evolutionary rates in each lineage (Felsenstein 1982). A first step in relaxing this assumption is the Fitch-Margoliash algorithm, which allows for some variation in evolutionary rates (Fitch and Margoliash 1967, Felsenstein 1989). I used Steller's Jays (Cyanocitta stelleri, CYSTE) as the out- group, because it is sufficiently close phylogenetically to be informative, and the sample size (20) was suf- ficiently large to permit accurate estimation of gene frequencies. Because of limited computer memory, the entire data set could not be analyzed at one time, so I analyzed subsets of the data in two ways. I either assumed species to be monophyletic and analyzed each separately, or I used representative populations from each of the subspecies groups. All trees were checked for stability using a iackknife manipulation (Lanyon 1985), in which the data were reanalyzed sequentially omitting each ingroup taxon and the re- sulting trees combined by strict consensus using pro- grams developed by Scott M. Lanyon to complement those of PHYLIP (Felsenstein 1989). Cladistic analysis of electrophoretic data is difficult both logically and methodologically, chiefly because of problems with the treatment of polymorphic char- acters (Buth 1984, Swofford and Berlocher 1987, Swof- ford and Olsen 1990). Computer packages such as PAUP deal with the problem by selecting the most- parsimonious character state from among the possi- bilities and ignoring potentially conflicting infor- mation in other alleles present in the population (D. Swofford, pets. comm.). Because loci in my study were highly polymorphic (Peterson 1990), coding loci as characters would not be very informative; instead, I chose to code alleles as characters. Using the heuristics/subtree-pruning and regraft- ing branch-swapping option of PAUP (version 3.0; TAIIE 2. Matrix of character states representing presence and absence of 38 phylogenetically informative alleles at 29 loci in 35 Aphelocoma jay populations and a composite outgroup. Sample 1 11 21 31 OUTGR 7110010110 0100011000 0000001000 00100000 ACCAL 0000010000 0001110000 0100111000 00000110 ACCAU 0000000000 1000110100 0000010000 00000100 ACCO2 0100000000 0000010000 0010000000 01000000 ACCY1 0100000000 0000010001 0000011001 00011111 ACCY2 0100000000 0010010001 0100011000 10000100 ACGR1 0000110100 1010010001 0100011000 10000110 ACGR3 0000000000 0000010000 0101011000 10000110 ACHY3 0000010000 0000110000 0110011000 00110010 ACIMM 0000000000 0000110000 0000011000 00000010 ACINS 0000000000 0000010000 0000000000 00000000 ACNE1 1000010110 0000010000 0010011110 00000110 ACNE4 0000010100 0000010101 0110011111 00000110 ACNE6 0000000010 0000010000 0010011000 00000110 ACOBc 1000000001 0000110000 1010011000 00000111 ACOBk 0010000000 0000110000 0000010000 00001010 ACOOC 0000100000 1000010001 0000111000 00000110 ACREM 0000000010 0000010000 0000011000 01000000 ACSM2 0000100010 0000010010 0000011000 01101000 ACSP1 1000000000 1000110100 0000011010 00000111 ACSP3 0000000000 0000110000 0000011000 00001110 ACTEX 0000000000 0000010000 0000011000 00000010 ACWO1 1000000000 0000010001 0001001010 00000110 ACWO3 0000000001 0010010001 0010011010 10000110 ACWO4 1000000000 0010010101 0000001110 00000110 AUPO3 0000001000 0000010010 0000011100 00000000 AUGRA 0000001000 0000011000 0000001000 00000000 AUCOL 0000000000 0100001000 0000001000 00000000 AUULT 0000000000 0000001000 0000011000 00100000 AUPO2 0011000000 0000010010 1000011100 00000000 AUWO1 0000001000 0000011010 0000001000 00000000 AUAZ1 0000000000 0000001000 0000000000 00000000 AUCO2 1001000000 0000011010 0000011100 00000000 ANCON 0000000000 0001001000 0000000000 00000000 ANOAX 0000000000 0100001000 0000011000 00000000 ANGUE 0000000000 0000001000 0000000000 00000000 Swofford 1989), I analyzed a binary coding of 102 alleles from 29 loci (Table 2). Of these alleles, 38 were potentially informative about relationships among ingroup taxa. This approach can have the undesirable consequence of hypothetical ancestors lacking alleles at particular loci (Buth 1984, Swofford and Berlocher 1987). However, this problem can be ignored if pop- ulation samples are large enough that the presence or absence of one allele in the sample can be assumed to be independent of the presence of other alleles. In other words, the alleles under study were indeed lacking in those problematic hypothetical ancestors, and some other allele is assumed to have been present at that particular locus. If the assumption of independence of alleles at a locus were unreasonable, weighting alleles so that each locus contributes equally to the analysis would be desirable. In this particular data set, however, pre- liminary runs indicated that weighting serves only to overemphasize loci with few alleles and deem- phasize apparently informative alleles at loci with more alleles. Hence, succeeding analyses were based on unweighted presence-and-absence data for alleles. I used a composite outgroup representing allele pres- ences and absences in all outgroup taxa to root re- sulting trees. All data sets were tested for significant phylogenetic information with the randomization tests of Archie (1989). To assess the stability of nodes on the resulting trees, a jackknife manipulation (Lanyon 1985) was conducted, in which the analyses were re- peated sequentially omitting each of the ingroup taxa. In each repetition, the heuristic search was allowed to run to completion, with up to 2,500 trees stored for branch-swapping, and the resulting 35 sets of 13 to 2,500 equally-parsimonious trees were combined using a majority-rule consensus. One problem plaguing cladistic analyses of elec- trophoretic data is failure to detect alleles actually present in populations because of inadequate sam- pling (Swofford and Berlocher 1987). This nondetec- T^SLE 3. Matrix of character states representing presence and absence of 33 phylogenetically informative alleles at 29 loci in eight subspecies groups of Aphelocoma jays and a composite outgroup. Sample 1 11 21 31 OUTGR 7110010110 1001100000 0000000100 000 CO CAL 1011100011 0011010111 1010100111 111 CO WOO 1000101110 0101010101 1111111000 110 CO MEX 0101101001 0101000101 0110011011 CO SUM 0001000100 0001001000 0010000101 000 UL POT 1010010000 0001101010 0011000000 000 ULWOL 0000010000 0001101000 0000000000 000 UL ULT 0000000000 1000100000 0010000100 000 UNICE 0000000000 1010100000 0010000000 000 tion causes extra homoplasy in the resulting trees in the form of "reversals," which are actually undetected synapomorphic alleles. In my study, several popu- lation samples were available from most of the sub- species groups, which reduced considerably the prob- ability of nondetection of alleles. I conducted a cladistic analysis of relationships among regions within the three species (i.e. subspecies groups; Fig. 1, Table 1) by combining all alleles found in populations of each region into a subspecies group (Table 3). I found 33 alleles to be phylogenetically informative about re- lationships of ingroup taxa (Table 3). Results based on other phylogeny-estimation algorithms, such as the polymorphism parsimony of Felsenstein (1979), yielded similar results, and are not presented below. Rate tests.--I tested for evolutionary rate uniformity by focusing on three populations previously identi- fied as distinctive in studies of genetic differentiation (Peterson 1990): Scrub Jays from Florida; Scrub Jays from Santa Cruz Island; and Unicolored Jays from Guerrero. The hypothesis tested was that historical independence of lineages could account for zones of extreme genetic differentiation among populations of Aphelocoma jays. Using three-taxon statements from the phylogenetic results, I conducted relative-rate tests that compared genetic distances in all representatives of the two taxa connected by the interior node of these trees to all representatives of the third taxon. Because at least two of the groups are represented by numer- ous population samples, replicate comparisons were possible. To visualize patterns of nonuniformity of rates, I also estimated branch lengths and their variability for two subspecies-group phylogenies with a resam- pling technique. The resampling manipulation was an attempt to estimate how robust branch-length es- timates are to choice of populations for sampling. I used the user-tree option of the FITCH program in PHYLIP to estimate branch lengths in 30 replicate analyses. Each replicate included ACINS, ANGUE, ACCO2, and randomly selected population samples representing each of the remaining eight subspecies groups (Table 1). The wollweberi group of Gray-breast- ed Jays was excluded from some analyses because it is part of an unresolved trichotomy in some trees, and the user-tree option of PHYLIP requires dichotomous branching at internal nodes. Branch lengths from the replicate analyses were compared using Mann-Whit- ney U-tests. RESULTS Phenetic methods.--The two phenetic analyses differed in the placement of a few groups and in the degree of resolution (Fig. 2). However, both established one cluster that consisted of populations of the californica group of Scrub Jays, and another of populations of the woodhouseii and sumichrasti groups of Scrub Jays (Pitelka 1951). The only exceptions are the inclusion of the populations of the southern tip of Baja Cal- ifornia and northwestern California (ACHY3 and ACCAU) as lineages basal to both of the clusters in the single-linkage method, and the inclusion of four populations of Gray-breasted Jays at various points within these clusters. The remaining populations of Gray-breasted Jays and the two east slope Unicolored Jay populations (ANCON and ANOAX) form a sister cluster to the two just described. Scrub Jay populations from Florida and Santa Cruz Island, and the Guerrero Unicolored Jay population fall outside all of these clusters. Fitch-Margoliash analyses.--Analyses of the re- lationships of the 24 population samples of Scrub Jays (Fig. 3A) showed basal branches leading to the Santa Cruz Island and Florida populations (ACINS and ACCO2), and then branches to populations from the interior western United States, northern Mexico, and southern Mexico (woodhouseii and sumichrasti groups). Derived from within the latter group are the populations of the californica group. A jackknife manipula- tion of this data set preserves the same basic  ACCAL ACOOC A COBC A CIMM ACOBk A CSP3 ACSP 1  ACCY1  ACGR 1 ß ACCY2 A CGR3 ACNE6 ACW03 ACW01 ' ACNE4  ACNE 1 ß ACTEX A CREM -F--- ACSM2  AUP03 AUP02  ACW04 A CHY3 A CCAU AUGR AUC02 z AUCOL AUAZl ANCON AUUL T AUWO 1 ANOAX A CC02 ANGUE A ClNS Single-linkage UPGMA 0.3 0.2 0.1 0.0 0.3 0.2 0.1 A CCAL ACOOC ACOBc A CIMM A COBk A CSP3 A CSP 1 A CCAU ACHY3 ACCY1 A CGR I  ACCY2 A CGR3 A CNE6 ACW03 ACW01 -- ACNE1 ACNE4 ACTEX  ACREM A CSM2 AUP03 AUP02 ACW04 AUGRA AUC02 AUCOL AUAZ 1 ANCON -- AUUL T  AUWO 1 ANOAX ACCO2 AClNS ANGUE 0.0 Modified Rogers' Genetic Distance Fig. 2. Phenetic relationships among 35 populations of Aphelocorna jays based on (A) single-linkage cluster analysis and (B) UPGMA. See Table 1 for abbreviations. structure, with 8 of the 23 nodes lost due to instability of the trees (Fig. 3A). Removal of populations with very long branches (Santa Cruz Island, ACINS, and Florida, ACCO2), as sug- gested by Swofford and Olsen (1990), preserves the same branching pattern in the remaining populations. The eight populations of Gray-breasted Jays divided into two main lineages, one consisting of the two southern populations and AUAZ1, and the other of the eastern populations plus AUGRA and AUWO1 (Fig. 3B). None of the nodes was unstable under a jackknife manip- ulation. Analyses of the three Unicolored Jay populations indicated that the east-slope pop- ulations (ANOAX and ANCON) are sister taxa, well differentiated from the west-slope popu- lation (ANGUE, Fig. 3C). The analysis of representatives from each subspecies group in the genus yielded unex- pected results. Inclusion or exclusion of the bas- al-split populations ANGUE, ACINS, and ACCO2 did not change the placement of other taxa, so these populations were excluded from analysis. In nine replications (all jackknifed) of randomly chosen representatives of each sub- species group (for three examples, see Fig. 4), the southern Gray-breasted Jay population (AUCOL or AUULT) clustered with the Uni- colored Jay population (ANCON or ANOAX), and the eastern Gray-breasted Jay population (AUPO2, AUPO3, or AUCO2) clustered with the various Scrub Jay populations. The western Gray-breasted Jay population (AUAZ1, AU- GRA, or AUWO1) joined with the southern Gray-breasted Jay-Unicolored Jay group, joined with the Scrub Jay-eastern Gray-breasted Jay group, or separated in a basal trichotomy. A1- Fig. 3. Fitch-Margoliash analysis of relationships among populations within each of three species of Aphelocoma jays: (A) Scrub Jays; (B) Gray-breasted Jays; and (C) Unicolored Jays. Asterisks indicate nodes unstable to a jackknife manipulation. Numbers indicate fitted branch lengths; unmarked branches sized proportionally. CYSTE is Cyanocitta stelleri, the outgroup; see Table ! for other abbreviations. though the outgroup (CYSTE) is linked to the network in a fairly consistent position, it could be argued that inaccurate rooting could produce the paraphyly in Gray-breasted Jays. In this case, however, the two-branched form of the phy- logeny makes any possible rooting lead to para- phyly in that species. From these results, I sug- gest that Gray-breased Jays represent a paraphyletic taxon that gave rise to both Scrub Jays and Unicolored Jays. Cladistic analyses.--In cladistic analyses of re- lationships of the 35 individual populations, PAUP found at least 2,500 trees of 108 steps. A majority-rule concensus of these trees is shown in Figure 5. Of the 12 nodes in the concensus- tree, eight were retained in all jackknife pseu- doreplicate analyses. The randomization test of Archie (1989) indicated the existence of signif- icant phylogenetic information in the data (P -< 0.03). These trees have several interesting features. First, they identify several pairs and trios of populations as sister taxa: central California and southern Baja California (ACCAL and ACHY3); northern and central Great Basin populations (ACNE1 and ACNE4); northern and southern Rocky Mountain populations (ACWO1 and ACWO4); southern Mexican populations of Scrub Jays (ACREM and ACSM2); and central- western Mexican (AUGRA and AUWO1) and eastern Mexican (AUPO2 and AUCO2 and then AUPO3) populations of Gray-breasted Jays. Many of these groupings correspond to those described based on morphology and geography (Pitelka 1951). Within the woodhouseii group, northern (ACNE1) and southern (ACNE4) neva- dae populations are sister taxa, and northern (ACWO1) and southern (ACWO4) woodhouseii are also sister taxa. This finding confirms geo- graphic patterns described by Pitelka (1951), and ANCON i AUCOL AUGRA AUPO2 ACCY2  ACREM ACWO3 ACOBk ANCON i AUAZ1 AUULT AUCO2 ACCY1 ACSM2  ACNE1 ACSP3 ANOAX i AUULT AUWO1 AUPO3 ACGR3  ACSM2 ACNE6 ACCAL Fig. 4. Three replicate Fitch-Margoliash analyses with randomly chosen representative populations of eight subspecies groups of Aphelocoma jays. Bars in- dicate species membership: white, Scrub Jay; gray, Gray-breasted Jay; and black, Unicolored Jay. See Ta- ble 1 for abbreviations. contradicts patterns described by Phillips (1986), who suggested that southern Great Basin (southern nevadae) and southern Rocky Moun- tain (southern woodhouseii) populations are more closely related to each other than to populations to the north. In the original consensus tree and in 33 of 35 of the jackknife pseudoreplicates, the eastern ACCAL ACHY3 ACCAU ACSP1 ACCY1 ACNE1 ACNE4  75-99% ACWO1 ' 50-74% ACWO4 ACOOC length 108 steps  ACCY2 c.I. = 0.352 ACGR1 ACGR3 AClMM .............  ^co.  ACSP3 ACTEX ACREM ACSM2  AUPO3 AUPO2 AUCO2 ACCO2 ACINS ANCON  ANGUE AUAZ1 AUGRA AUWO1  AUCOL AUULT ANOAX Fig. 5. Cladistic analysis of 35 Aphelocoma jay pop- ulations, based on binary coding of 38 alleles. A ma- jority-rule consensus of 2,500 108-step trees shown by line type. Nodes marked with asterisk found to be unstable to a jackknife manipulation. See Table 1 for abbreviations. Consistency index (c.i.) indicated. populations of Gray-breasted Jays are united with the Scrub Jay clade, making Gray-breasted Jays as a taxon paraphyletic (Fig. 5). In addition, in all trees examined, a northwestern Gray- breasted Jay population sample (AUAZ1) groups with two Unicolored Jay populations (but see Discussion). Again, the exact position of the root does not change the paraphyly in Gray-breasted Jays. Therefore, the currently recognized spe- cies Gray-breasted Jay appears to be a paraphy- letic taxon, with the other two recognized spe- cies derived from different populations within it. Inspection of patterns of character evolution on this tree indicates that the Scrub Jay clade (minus ACTEX, ACREM, and ACSM2) is sup- ported by two synapomorphies, and that sev- eral other internal nodes are supported by syn- apomorphic alleles. Because the GDA locus shows an odd pattern of variation in Gray- breasted Jay populations (Peterson 1990), I be- came concerned that it might be responsible for the patterns of paraphyly in Gray-breasted Jays. This locus might be suspect because the two alleles observed differed in mobility by only about 2 mm. (Several rechecks of allele homol- ogies at maximum mobility failed to distinguish more than two alleles at this locus.) ! reran the analyses omitting the information at the GDA locus. Although the strict consensus of the re- suiting 2,500 most-parsimonious trees is less well resolved than the overall analysis, AUAZ1 was still more closely related to two Unicolored Jay populations than to other Gray-breasted Jay populations (but see Discussion). A node unit- ing eastern Gray-breasted Jay populations with Scrub Jays was found in 69% of the trees. There- fore, support exists for the hypothesis of para- phyly of Gray-breasted Jays at loci other than GDA. In analyses of relationships of the subspecies groups, PAUP found one 58-step tree (Fig. 6), nine 59-step trees, and many longer trees. The shortest tree is almost identical to those pro- duced by the Fitch-Margoliash analyses (Fig. 4), but differs in that eastern and western Gray- breasted Jays group together; it provides better resolution among populations of Scrub Jays. The 59-step trees differ in placement of a few branches, but the basic structure is similar. Sev- eral alleles provide support for most nodes, in- cluding 10 alleles that represent synapomor- phies for all Scrub Jay populations except the sumichrasti group, and one to five synapomor- phies that support other nodes (Fig. 6). A jack- knife manipulation of the data set preserved the general structure of the tree, with two un- stable nodes. The randomization test of Archie (1989) indicated the presence of significant phy- logenetic information in the data set (P < 0.01). Omitting the two GDA alleles from analysis did not change the tree structure, except in that two nodes are unresolved. Although the tree of the eight subspecies groups in the three species is well resolved, placement of the remaining three distinct lin- eages is equivocal. The Scrub Jays of Santa Cruz Island (ACINS) and Florida (ACCOE), and the Unicolored Jays of Guerrero (ANGUE) are in- OUTGR !- Apomorphy []- Parallelism - Revereal length 58 steps c.I. = 0.569 3 ! ! 2 ! 2 1 10 I 21 _ CO CAL co woo * L CO MEX 1 I CO SUM 21 I I la  UL POT  UL WOE UL ULT I *2 UN E Fig. 6. Cladistic analysis of eight subspecies groups of Aphelocoma jays and outgroup showing character- state changes and support for nodes. Nodes marked with asterisk found to be unstable to a jackknife ma- nipulation. See Table 1 for abbreviations; consistency index (c.i.) indicated. variably placed as basal offshoots of the trees (Fig. 7B). For reasons outlined below, this ar- rangement seems unlikely, so a placement of these three populations that agrees more closely with geography (Fig. 7A) is also considered in the analyses of rate uniformity. Rates of molecular evolution.--Because the phy- logenetic studies described above are equivocal as to the placement of three populations (ACINS, ANGUE, and ACCO2), ! used both trees of Fig- ure 7 in the rate tests. ! conducted a relative- rate test based on the following three-taxon statements (from Figs. 7A and 7B): ((ACINS, cal- ifornica group), woodhouseii group); ((ANGUE, east-slope unicolor), southern Gray-breasted Jays); ((ACCO2, rest of Scrub Jays), eastern ul- tramarina); ((ACINS, rest of ingroup), CYSTE); ((ANGUE, rest of ingroup), CYSTE); and ((ACCO2, rest of ingroup), CYSTE). In the six tests, apparent departures from rate equality ex- ist; ACINS, ANGUE, and ACCO2 in every com- parison had greater genetic distances and pa- 0 x  0 x _    oo z z O O O O G.I. = 1.0 PLUMAGE TYPE  ß Unicolored Jay [] Gray-br. Jay [] Scrub Jay Fig. 7. Two possible phylogenies combining the results of Figures 2-6, showing most-parsimonious re- constructions of patterns of plumage evolution. Abbreviations follow Table 1; consistency index (c.i.) indicated. tristic distances to the third taxon than did their sister lineages (sign test; all P < 0.001). Presum- ably, higher rates of molecular evolution have characterized these three lineages relative to the rest of the group. Analyses involving resampling of popula- tions revealed a high degree of heterogeneity of rates among lineages (Fig. 8). Statistical com- parisons indicate that evolutionary rates are sig- nificantly different in every pair of sister lin- eages (Mann-Whitney U-test; all P < 0.05). The validity of these statistical tests depends on the effectiveness of the resampling manipulation in estimating true levels of branch-length vari- ability. The manipulation employed here eval- uates error introduced by choice of sample lo- cality, but not sampling error in estimation of genetic distances. Nevertheless, it is clear that molecular evolutionary rates have been uneven in the history of the Aphelocoma jays. DISCUSSION The consistency with which geographically related populations clustered together indicates that these data contain phylogenetic informa- tion. For example, in most analyses, the 10 wood- houseii-group populations cluster together, as do the nine californica-group populations. Below I discuss the unexpected features of the trees in an attempt to understand what factors, meth- odological or biological, led to those patterns. Paraphyly of the Gray-breasted Jay.--In all anal- yses, the Gray-breasted Jay is a paraphyletic tax- on. Unicolored Jays invariably group with the southern Gray-breasted Jay lineage, and Scrub Jays with the eastern Gray-breasted Jay lineage. The position of the western Gray-breasted Jay lineage is less clear. On the whole, it appears that the Gray-breasted Jay is a relatively old lineage that has been in the mountains of north- ern Mexico for a long period of time, and that Scrub Jays and Unicolored Jays were derived independently from different parts of the Gray- breasted Jay complex. Although unexpected (Pitelka 1951), this idea receives some support from other evidence. First, Gray-breasted Jays are intermediate in habitat use between the other two species. Where the three species' ranges overlap in central Vera- cruz, Scrub Jays live on dry interior slopes and Unicolored Jays on humid coastal slopes. Gray- breasted Jays inhabit the pine-oak forests at higher elevations intermediate between the two, and so direct invasion of either habitat would have been possible. Second, Gray-breasted Jays in the southern and western portions of the range show delayed maturation of beak color, a character present in rudimentary form in Uni- colored Jays and otherwise rare in birds (Pe- terson 199 lb). Eastern Gray-breasted Jays do not delay beak color maturation, which is similar to the ontogenetic patterns of Scrub Jays (Pe- terson, in press). Similarly, western and south- ern Gray-breasted Jays do not give the female- specific "rattle" vocalization, which is found in eastern Gray-breasted Jay and all Scrub Jay pop- ulations (Brown and Horvath 1989). Therefore, the hypothesis of independent derivations of Scrub and Unicolored jays from Gray-breasted Jays is supported by other characters. Taxonom- ic implications of these findings will be treated elsewhere (Peterson, in prep.). Origin of the californica group of Scrub Jays.- In a number of Fitch-Margoliash trees (e.g. Fig. 3), the californica group of Scrub Jays is derived from within the woodhouseii group of Scrub Jays. Some support for this hypothesis is also found in cladistic results although these nodes are not preserved in the consensus analyses (Fig. 5). This pattern is to be expected based on a prob- able Central American/Mexican origin of the species with subsequent invasion into the Great Basin and then into California (Peterson 1990). Rapid evolution versus old splits.--Three dis- junct populations--the Scrub Jays of Santa Cruz Island (ACINS), the Scrub Jays of Florida (ACCO2), and the Unicolored Jays of Guerrero (ANGUE)--consistently are placed as basal lin- eages in the phylogenetic analyses (Fig. 7; in a few alternative trees, the Florida population ACCO2 grouped with southern Mexican Scrub Jay populations based on one allele unique to the two groups). This result is surprising, given that each shares obvious similarities of plumage coloration, shape and size, vocalizations, and behavior with other members of its species. For example, ACINS shares several unique plumage characters with populations of California and, on that basis as well as on grounds of geograph- ic proximity, seems to have been derived from adjacent mainland populations (Pitelka 1951). Two alternative hypotheses are apparent: (A) that these lineages are exceptionally divergent populations derived from within Scrub Jays and Unicolored Jays, but for nonhistorical reasons appear as basal splits on the trees (Fig. 7A); or (B) that these lineages are exceptionally old, and the characters that each shares with its conspe- cific populations are actually retained similar- ities (Fig. 7B). Alternative B is better supported by the allozyme data than alternative A (tree- length of 62 steps in B; 73 steps in A). Still, two biological points argue against al- co wco Fig. 8. Trees showing results of resampling anal- yses of rates of molecular evolution in different lin- eages, based on two trees of Figure 7. See Table 1 for abbreviations. ternative B. First, although placement of these three populations as basal splits (Fig. 7B) re- duces hypothesized levels of heterogeneity of rates of molecular evolution, it leads to prob- lems in understanding the evolution of plum- age coloration. For the tree in Figure 7B, plum- age coloration must have undergone a number of reversals and episodes of rapid evolution (plumage coloration mapped onto tree; consis- tency index = 0.5). Under the view in Figure 7A, plumage coloration shows only two changes: one at the derivation of Scrub Jays; and another at the derivation of Unicolored Jays (consisten- cy index = 1.0). Hence, hypothesis B is less consistent with patterns of variation in plum- age-color characters than hypothesis A. Fur- thermore, these populations share not only as- pects of plumage coloration, but a wide range of characters (e.g. body size and shape, ecolog- ical requirements, vocalizations, behaviors) with conspecifics (Pitelka 1951), so whole suites of characters not included in my analyses would have had to have changed unpredictably. Second, even if these populations represent basal splits (alternative B), they still have ex- ceptional branch lengths. Unequal branch lengths can result from incorrect choice of out- group (i.e. an outgroup taxon actually belongs to the ingroup). Outgroups in this study include representatives of the entire family Corvidae, so derivation of outgroup taxa from the ingroup is unlikely. Nevertheless, if to avoid unequal branch lengths, the trees are rerooted at the midpoint of the longest branch (in this case, that leading to ACINS or ANGUE), not only do at least two other ingroup taxa remain with ex- ceptional branch lengths, but the outgroup ap- pears as an exceptionally divergent branch de- rived from within ApheIocoma. Removal of the three long branches from the tree (after Swofford and Olsen 1990) produces results that agree with geography and mor- phology. The populations on the three long branches all represent small, isolated popula- tions isolated from the remainder of the species' ranges. Hence, ! considered features of small populations that could possibly affect accurate reconstruction of phylogenies. A possible resolution of these difficulties may lie in the extremely reduced levels of within- population genetic variation in these same three populations (Peterson 1990). Heterozygosities for the three populations averaged 1.9, as com- pared with 3.9 for the remainder of the popu- lations sampled (Mann-Whitney U-test, P < 0.05; Peterson 1990). Other measures of within-pop- ulation variation (e.g. proportion of loci poly- morphic, number of alleles) are similarly re- duced in the three populations. Many of the characters on which the phylogenies are based are alleles at low or moderate frequencies (e.g. PGD-e, which is found in all but one of caIifor- nica-group populations, never at a frequency >0.136). If a population with low-frequency al- leles that identify its sister-taxon relationships loses variation, perhaps due to population bot- tlenecks, those same low-frequency alleles would most likely be lost. Consequently, many of the characters shared with the ingroup (syn- apomorphies) are secondarily absent. These changes will move such lineages to the base of cladograms. Phenetic techniques that assume rate equality obviously will be sensitive to in- creases in evolutionary rates in particular lin- eages and, potentially, also to secondary losses of alleles. Thus, the tempo and mode of evo- lution in a group can affect the outcome of phy- logenetic analyses; in this case, it leads to the basal placement of ACINS, ANGUE, and ACCO2. This problem may affect other parts of the phylogenetic reconstruction of the ApheIocoma jays as well. For example, it could be responsible for the basal placement of ACTEX within Scrub Jays and the grouping of ANCON, ANGUE, and AUAZ1, all of which show reduced genetic variation (Fig. 5). Grouping of Florida and Santa Cruz Island populations (ACCO2 and ACINS) in some analyses (a most unlikely outcome geo- graphically) is almost certainly a result of this phenomenon. Unexpected features of the phy- logenies discussed above, such as the derivation of the caIifornica group from the woodhouseii group and paraphyly of Gray-breasted Jays, most likely do not result from this problem because they do not involve groups that differ markedly in levels of genetic variability. The best hypothesis for the position of ACINS, ACCO2, and ANGUE in the Aphelocoma jay phy- logeny appears to be that they are exceptionally derived lineages from within Scrub Jays and Unicolored Jays. This conclusion is supported by several logical problems with the alternative hypothesis. The methodological reason for their basal placement in the analyses is clearly iden- tifiable, and has nothing to do with time since divergence from the ancestral stock. My results serve to illustrate some of the dif- ficulties of using electrophoretic characters in phylogenetic analyses, as have been pointed out by other workers (e.g. Buth 1984, Swofford and Berlocher 1987, Swofford and Olsen 1990). The highly polymorphic nature of these characters makes coding and analysis difficult; effects of population history can reduce the probability of correctly reconstructing the history of the group. Although ! believe that useful and ac- curate information was retrieved from the data set, the difficulty of exact placement of several important populations demands further study with other types of biochemical characters. Rates of molecular evoIution.--The analyses of rates of molecular evolution indicated that at least three ApheIocoma jay populations have un- dergone marked accelerations in rate of differ- entiation relative to their sister populations. Re- gardless of whether they are placed as basal splits or with apparently conspecific popula- tions (Figs. 7B or 7A), these populations have evolved faster than other populations of Aphelo- coma jays. Inspection of branch lengths in the tree (Fig. 8) suggests that heterogeneity of evo- lutionary rates may not be restricted to just those three lineages, but that rate heterogeneity on a finer scale may be common in the genus. Because all tree-building methods are to some degree sensitive to rate differences (Swofford and Olsen 1990), trees estimated are not inde- pendent of the question of whether rates of evolution are uniform. When presented with a case of extreme rate inequality, algorithms for phylogeny estimation err on the side of reduc- ing heterogeneity of branch lengths. A search for rate differences based on trees potentially affected by such differences is a conservative test; errors introduced will be of not finding a difference that exists, rather than of finding spurious heterogeneity. Studies of variation in other molecules that will permit critical inde- pendence of phylogeny estimation and branch- length estimation are in progress. More interesting is the degree to which evo- lutionary rates differ. If ACINS and ANGUE represent basal splits from the genus, they have been evolving 20-25% more rapidly than the rest of the genus. If, as seems more likely, they were derived from within Scrub Jays and Uni- colored Jays, their evolutionary rates have been three to four times greater than those of their sister lineages. Correlations between rate of dif- ferentiation and features of population biology are explored in Peterson (1990, in prep.). Electrophoretic data for populations of Aphe- locoma jays clearly violate the assumption of rate equality. Although studies of several other avi- an groups have come to the opposite conclusion (e.g. Bledsoe 1987), my results and the results of others (e.g. Britten 1986, Springer and Kirsch 1989) indicate that the assumption of rate uni- formity is not always reasonable. Rate-depen- dent algorithms and interpretations should be tested first (e.g. Page 1990) and used cautiously, if at all, in studies purporting to derive phy- logenetic or biogeographic information from molecular data alone. ACKNOWLEDGMENTS Many people have lent invaluable assistance to me during this project. I thank especially Scott M. Lan- yon, Frank A. Pitelka, and John W. Fitzpatrick for their generous interest throughout the project. Thanks to John Bates, Shannon Hackett, Scott Lanyon, and two anonymous reviewers for helpful comments on the manuscript. I thank Pamela Austin, John Hall, Victor Pacheco, Mary Anne Rogers, Douglas Stotz, Thomas Schulenberg, David Willard, Adolfo Navarro S., and Patricia Escalante P. for their interest in my work. Shannon Hackett provided invaluable assis- tance with testing phylogenetic information in the cladistic data sets. I thank the Direcci6n de Flora y Fauna Silvestre, Secretaria de Desarollo Urbano y Ecologla, of the Mexican government, and the state governments of Oregon, California, Nevada, Arizona, Utah, New Mexico, Colorado, and Texas for providing permits for scientific collecting. The Field Museum generously provided access to its facilities for bio- chemical analysis. Financial support was provided by the National Science Foundation Dissertation Im- provement Grant Program (BSR-8700850), National Geographic Society, Field Museum of Natural His- tory, Chapman Fund (American Museum of Natural History), Sigma Xi, and the University of Chicago. LITERATURE CITED ARCHIE, J. W. 1989. A randomization test for phy- logenetic information in systematic data. Syst. Zool. 38:239-252. ARCHIE, J. W., C. SIMON, AND A. MARTIN. 1989. Small sample size does decrease the stability of den- drograms calculated from allozyme-frequency data. Evolution 43:678-683. BLEDSOE, A.H. 1987. DNA evolutionary rates in nine- primaried passerine birds. Mol. Biol. Evol. 4:559- 571. BRITTEN, R.J. 1986. Rates of DNA sequence evolu- tion differ between taxonomic groups. Science 231:1393-1398. BROOKS, D. R., AND D. A. MCLENNAN. 1990. Phylog- eny, ecology, and behavior. Univ. Chicago Press, Chicago. BROWN, J.L. 1987. Helping and communal breeding in birds. Princeton Univ. Press, Princeton, New Jersey. BROWN, J. L., AND E.G. HORVATH. 1989. Geographic variation of group size, ontogeny, rattle calls, and body size in Aphelocoma ultramarina. Auk 106:124- 128. BUTH, D.G. 1984. The application of electrophoretic data in systematic studies. Annu. Rev. Ecol. Syst. 15:501-522. CURTIS, E. L., AND R. C. MILLER. 1938. The sclerotic ring in North American birds. Auk 55:225-243. FELSENSTEIN, J. 1979. Alternative methods of phy- logenetic inference and their interrelationship. Syst. Zool. 28:49-62. FELSENSTEIN, J. 1982. Numerical methods for infer- ring evolutionary trees. Q. Rev. Biol. 57:379-404. FELSENSTEIN, J. 1989. PHYLIP (phylogeny inference package), version 3.2 manual. Univ. Washington, Seattle. FITCH, W., AND E. MARGOLIASH. 1967. Construction of phylogenetic trees. Science 155:279-284. FITZPATRICK, J. W., AND G. E. WOOLFENDEN. 1986. De- mographic routes to cooperative breeding in some New World jays. Pages 137-160 in Evolution of animal behavior (M. H. Nitecki and J. A. Kitcheil, Eds.). Oxford Univ. Press, Oxford. HARDY, J.W. 1969. A taxonomic revision of the New World jays. Condor 71:360-375. HARRIS, H., AND D. g. HOPKINSON. 1978. Handbook of enzyme electrophoresis in human genetics. North-Holland Publishing Co., Amsterdam. HOPE, $. 1989. Phylogeny of the avian family Corvi- dae. Ph.D. dissertation, City Univ. New York, New York. JOHNSON, N. K., AND J. A. MARTEN. 1988. Evolution- ary genetics of flycatchers. II. Differentiation in the Empidonax difficilis complex. Auk 105:177-191. LANYON, S.M. 1985. Detecting internal inconsisten- cies in distance data. Syst. Zool. 34:397-403. NEVO, E., R. BEN=$HLOMO, g. BEILES, J. U. g. JARVIS, AND G. C. HICKMAN. 1987. Allozyme differen- tiation and systematics of the endemic subter- ranean mole rats of South Africa. Blochem. Syst. Ecol. 15:489-502. PAGE, R. D.M. 1990. Temporal congruence and cla- distic analysis of biogeography and cospeciation. Syst. Zool. 39:205-226. PETERSON, A.T. 1990. Evolutionary relationships of the Aphelocoma jays. Ph.D. dissertation, Univ. Chicago, Chicago. PETERSON, h. T. 1991a. New distributional infor- mation on the Aphelocoma jays. Bull. Br. Ornithol. Club 111:28-33. PETERSON, A. T. 1991b. Sociality and ontogeny of coloration in the blue-and-black jays. Wilson Bull. 103:59-67. PETERSON, h. T. 1992. Geographic variation in the ontogeny of beak coloration of Gray-breasted Jays (Aphelocoma ultramarina). Condor 93:448-452. PETERSON, A. T., AND N. VARGAS B. 1992. Ecological diversity in Scrub Jays (Aphelocoma coerulescens). In Biological diversity of Mexico: Origins and distribution (T. P. Ramamoorthy, R. Bye, A. Lot, and J. Fa, Eds.). Oxford Univ. Press, Oxford. PHILLIPS, A.R. 1986. The known birds of North and Middle America, vol. 1. A. R. Phillips, Denver. PITELKA, F.A. 1951. Speciation and ecologic distri- bution in American jays of the genus Aphelocoma. Univ. Calif. Publ. Zool. 50:195-464. ROHLF, F. J. 1988. NTSYS-pc: Numerical taxonomy and multivariate analysis system, version 1.40. Exeter Software, Setauket, New York. $ARICH, V. M. 1977. Rates, sample sizes, and the neutrality hypothesis for electrophoresis in evo- lutionary studies. Nature 265:24-28. SHAW, C. R., AND R. PRASAD. 1970. Starch gel elec- trophoresis--A compilation of recipes. Blochem. Genet. 4:297-330. SHELDON, F.H. 1987. Rates of single-copy DNA evo- lution in herons. Mol. Biol. Evol. 4:56-69. $IBLEY, C. G., AND J. E. AHLQUIST. 1981. The phylog- eny and relationships of the ratite birds as in- dicated by DNA-DNA hybridization. Pages 301- 335 in Evolution today: Proceedings of the Sec- ond International Congress of Systematic and Evolutionary Biology (G. G. E. Scudder and J. L. Reveal, Eds.). Carnegie-Mellon Univ., Pitts- burgh. $IBLEY, C. G., AND J. E. ALQUIST. 1983. Phylogeny and classification of birds based on the data of DNA-DNA hybridization. Curr. Ornithol. 1:245- 292. $IBLEY, C. G., J. E. ALQUIST, AND F. H. SHELDON. 1987. DNA hybridization and avian phylogenetics: Re- ply to Cracraft. Evol. Biol. 21:97-125. SPRINGER, M. S., AND J. g. W. KIRSCH. 1989. Rates of single-copy DNA evolution in phalangeriform marsupials. Mol. Biol. Evol. 6:331-341. SWOEEORD, D.L. 1989. PAUP: Phylogenetic analysis using parsimony, version 3.0. Illinois Natural History Survey, Champaign. $WOFFORD, D. L., AND $. H. BERLOCHER. 1987. Infer- ring evolutionary trees from gene frequency data under the principle of maximum parsimony. Syst. Zool. 36:293-325. $WOFFORD, D. W., AND G. J. OLSEN. 1990. Phylogeny reconstruction. Pages 411-501 in Molecular sys- tematics (D. M. Hillis and C. Moritz, Eds.). Si- nauer Associates, Sunderland, Massachusetts. WOOLFENDEN, G. E., AND J. W. FITZPATRICK. 1984. The Florida Scrub Jay: Demography of a cooperative breeding bird. Princeton Univ. Press, Princeton, New Jersey. WILSON, A. C., S. CARLSON, AND T. J. WHITE. 1977. Biochemical evolution. Annu. Rev. Blochem. 46: 573-639. WRIGHT, S. 1978. Evolution and the genetics of pop- ulations. Vol. 4, Variability within and among populations. Univ. Chicago Press, Chicago. ZINK, R.M. 1982. Patterns of genic and morphologic variation among sparrows in the genera Zono- trichia, Melospiza, Junco, and Passerella. Auk 99: 632-649. ZUCKERKANDL, E., AND L. PAULING. 1962. Molecular disease, evolution, and genic heterogeneity. Pages 189-225 in Horizons in biochemistry (M. Kasha and B. Pullman, Eds.). Academic Press, New York.