Genetic differentiation and evolutionary relationships were surveyed on 12 species of the Old and seven species of the New World Emberizidae by allozyme electrophoresis of 20 loci. Genetic variability of the Emberizidae is similar to those of the other Passeriformes. The degree of genetic differentiation in the family were large among species of the genus Emberiza of the Old World, and also among genera of the New World Emberizidae. Evolutionary relationships of the Emberizidae based on these genetic distances differed considerably from those of previous publications on some points: (1) Emberiza tristrami, E. elegans, E. bruniceps and E. schoeniclus were genetically much diverged from the other Emberiza as a species of the same genus. (2) Some genetic distances between Emberiza were larger than distances between subfamilies of the New World Emberizidae. (3) Species of the Cardinarinae examined genetically, belonged to the Emberizinae of the New World. Genetic data did not support the current classification that the Old World buntings arose from the New World forms by recent colonization. Discussion was made on the evolution of the Old and New World Emberizidae from the genetic view point.
The Emberizidae, with about 550 species, is a large family of Passeriformes which have made a considerable adaptive radiation. This family is usually divided into three main groups: Emberizinae (the buntings and American sparrows), Cardinalinae (the cardinal-grosbeaks), and Thraupinae (the tanagers). Most species of the Emberizidae are distributed in the New World. The cardinal-grosbeaks and tanagers are found only in North and South America, where many species of buntings and their allies naturally occur. However, about 40 species of the Old World buntings, most of them in the genus Emberiza, are mainly found in Asia.
It is customarily regarded that the Emberizidae is one of the most recently developed and diversified large groups of the Oscines (Passeriformes), and evolution of the Emberizinae groups is thought to have been a geologically recent event. The Old World buntings are usually regarded as forms arising from a single or a few colonizations of ancestors from the New World.
The taxonomy of the Old and New World Emberizidae has been a controversial issue. The Emberizinae were frequently grouped with the cardueline finches and Galapagos finches, and they were treated as members of a single expanded family Fringillidae. Recent publications place the buntings and allies in the same family as cardinal-grosbeaks and tanagers . Morony et al.  subdivided the Emberizinae into seven groups, and group I included Emberiza, Calcarius, Zonotrichia, Junco, Ammodramus, Spizella and others. On the other hand, Wolters  divides the buntings and American sparrows into two large subfamilies. One subfamily includes Emberiza and Calcarius, while the other makes a large group of the buntings and American sparrows, including Zonotrichia, Junco, Ammodramus and Spizella.
Most buntings of the Old World are Emberiza, and the remaining three genera are monotypic. They are generally of quiet appearance, patterned in brown, black and white, having stout bills adapted for seed eating. Morphologically they show little difference and are therefore thought to be a uniform group which evolved quite recently from the colonization by New World forms. Two further genera, Calcarius (the longspurs) and the monotypic Plectrophenax are circumpolar, therefore not typical elements of the Old World. The longspurs, however, share with Emberiza similar characters such as sexual dimorphism, with a brightly patterned head and breast in the male. Most closely related, and physically similar to the Old World buntings, are principally Nearctic genera such as Junco, Zonotrichia and Ammodramus .
Genetic analyses of American sparrows and relatives were made by electrophoresis of protein loci . Avise et al. examined genetic divergence of Calcarius, Zonotrichia, Junco, Ammodramus, Spizella, Amphispiza (group I of Morony et al.) and Pipilo (group VI). With some exceptions, their data demonstrate very close genetic relationships among all the species studied. Calcarius appears genetically very distant from the other group I species, while Pipilo is very similar to the North American sparrows.
Genetic data on birds have accumulated since electrophoretic techniques have been applied in ornithology. The data on protein variations have made it possible to analyze genetic relationships among avian species quantitatively, and they are reliable when many protein loci are studied. Comparative studies of protein polymorphisms among avian groups have now become available as shown above. In addition, Avise et al.  reported that vireos showed a larger degree of genetic differentiation than did other bird groups. For the Old and New world Emberizidae, electrophoretic studies should supply useful information on evolutionary genetics.
In this paper, we report on genetic differentiation and evolutionary relationships by allozyme electrophoresis of 20 loci, dealing with 12 species of Emberiza in the Old World and seven species of the Emberizinae and Cardinalinae in the New World. We will compare the genetic distance of the two subfamilies against previously reported distances of the Emberizinae and other avian groups, and also discuss the evolutionary relationship of the Old and New World Emberizinae, and the systematic position within and between the Emberizinae and Cardinalinae, on the basis of genetic data.
MATERIALS AND METHODS
Samples of the 19 species of Emberizidae were obtained mainly from Asada Birds Shop, Tokyo. The species and the number of individuals used in the present study are given in Table 1, and follows the classification used by Yamashina . Species of genus Emberiza are of the Old World origin, and other species studied are of the New World. Emberiza shoeniclus, 15 individuals of E. spodocephala and 29 individuals of E. rustica were collected in Fukushima-gata, Niigata. E. spodocephala and E. rustica obtained from the bird shop were collections in China. An E. shoeniclus accidentally killed by collision was obtainded in Tokyo.
Proportion of polymorphic loci and heterozygosities of Emberizidae species
Samples of blood, liver and muscle were treated as described by Kuroda et al.  and Kakizawa and Watada , and stored at −80°C until electrophoresis. Livers were used for the detection of the following 13 enzymes: acid phosphatase (ACPH), diaphorase (DIA), esterase D (ESD), glutamate oxzaloacetate transisomerase (GOT), glucose phosphate isomerase (GPI), isocitrate dehydrogenase (IDH), lactate dehydrogenase (LDH), malate dehydrogenase (MDH), peptidase (PEP), 6-phosphogluconate dehydrogenase (6PGD), phosphoglucomutase (PGM), sorbitol dehydrogenase (SDH) and superoxide dismutase (SOD). Muscle samples were used for detecting creatine kinase (CK), glutamate dehydrogenase (GDH) and α-glycerophosphate dehydrogenase (α-GPDH). Hemoglobin (Hb) and one phosphoglucomutase (PGM) were detected using erythrocyte samples.
Four buffer systems were used for starch gel electrophoresis. Most of the enzymes were detected using the Amine-citrate buffer system . The other three systems were used to detect some enzymes and hemoglobin: Tris-citrate (PGM-1 and hemoglobin), Poulik (ACPH, DIA and PEP) and Tris-HCl (CK). The methods of staining enzymes, and data analysis were described in previous papers [11, 15, 17].
The sample size and genetic variability of all 19 species are shown in Table 1. Out of 20 loci examined, no genetic variation was found in Emberiza variabilis and Paroaria gularis, whose sample sizes were very small. The average values of proportion of polymorphic loci was 0.14, and ranged from 0 to 0.35; the high value for Emberiza spodocephala seemed to be caused by the inclusion of chinese subspecies in the sample. The mean values of observed and expected heterozygosities were 0.036 and 0.042, respectively. In the present study, Emberiza spodocephela, Volatinia jacarina and Paroaria coronata showed low observed heterozygosities, with values of one third to a half of those expected. The observed heterozygosities in this study were not different from values reported in other avian studies, including research of New World Emberizidae .
Out of 20 loci examined, intra- and inter-generic variations were not found at two loci, GOT-2 and MDH-2. Table 2 shows the allele frequencies of the 18 variable loci in 19 species of the Emberizidae. There was no intra-specific variation at the loci of CK, GDH and Hb. Only one or two species had a species specific allele and most species had a single fixed allele at CK and Hb. In addition to CK and Hb, six loci (GPI, IDH, LDH, MDH-1, SDH and SOD) showed a similar pattern that one common allele was shared by most species, but unique alleles were fixed or polymorphic for some species. At the GDH locus, all seven species of New World Emberizidae were fixed at the same allele, on the other hand, the Old World Emberiza shared seven different alleles in 12 species of the same genus. GOT-1 was also a diagnostic enzyme which distinguished the Old World Emberiza from the New World species with the few exceptions that Emberiza spodocephala and Passerina versicolor were polymorphic and shared the same alleles at the locus. At the 6PGD and PGM-2 loci, half of the species were polymorphic, and the others fixed at species specific alleles or shared the same allele with a few species. ACPH, DIA and ESD were moderately polymorphic loci. Although five or six species were polymorphic at ACPH and ESD, most of the species in the genus Emberiza were fixed at the same alleles with some exceptions. Similarly, almost all of the Old World Emberiza shared the same allele at α-GPDH, PEP and PGM-1.
Alleles and their frequencies from the variable loci in 19 species of the Emberizidae
Out of the 12 species of Emberiza examined, E. tristrami, E. elegans, E. bruniceps and E. schoeniclus had several species specific alleles, and clearly differed from other Emberiza species of the same genus. On the other hand, seven species of Emberiza (E. pusilla, E. chrysophrys, E. rustica, E. rutila, E. sulphurata, E. spodocephala and E. variabilis) shared most of the same alleles at 20 loci, showing the genetic similarities among species of the genus. At DIA, E. spodocephala collected from Japan and China was fixed at different alleles c and b, respectively. This result suggests that E. spodocephala collected from Japanese and Chinese populations may be differentiated to some extent. In the New World Emberizidae, genetic variability within the genus could only be compared between Paroaria coronata and P. gularis. The two species had only one genus specific allele at PGM-1, and were fixed at different alleles at DIA and 6PGD.
As a means of quantifying genetic differentiation, genetic identities and genetic distances between species were calculated according to the formula presented in Nei . Table 3 shows genetic identities and genetic distances based upon the allele frequencies at 20 loci shown in Table 2. Genetic distance between Paroaria coronata and P. gularis was 0.128, but the values of the Emberizia at species level were much higher than that of the Paroaria. Although some species combinations of the Emberizia showed genetic distances less than 0.1, the values at the species level were much higher when the distances were calculated including the following four species: E. tristrami, E. elegans, E. bruniceps and E. schoeniclus. The largest value of genetic distance within the Emberiza was 0.838 between E. tristrami and E. bruniceps. The average genetic distance between species of the Emberiza was 0.334, which was much larger than the avian means between congeneric species .
Matrix of genetic identity (above diagonal) and genetic distance (below diagonal) between species of the Emberizidae, based upon allele frequencies of 20 loci. Species numbers are the same as Table 2
The mean genetic distance between species in the Emberizidae was 0.331 in the present study. At the genus level, the smallest and the largest genetic distances were 0.176 between Paroaria coronata and Passerina versicolor and 1.000 between Emberiza tristrami and Rhodospingus cruentus, respectively. According to Morony et al. , Lophospringus, Sicalis, Volatinia and Rhodospringus are different group of the Emberizinae in the New World, whose mean genetic distance is 0.449 (0.342–0.565), showing relatively large genetic divergence between genera. The mean genetic distance between members of the genus in the Emberizidae was 0.617, which was also higher than the avian mean .
Based upon the genetic distances in Table 3, a dendrogram (Fig. 1) was constructed for 19 species of the Emberizidae using UPGMA method . Figure 1 reveals that the 19 species are divided into two distinct groups, Emberiza of the Old World and the New World Emberizidae, separated by a genetic distance of 0.672. In the genus Emberiza, seven species consisted of a compact cluster in which genetic distances were within the range from 0.050 to 0.128. E. aureola was split from that compact cluster at a distance of 0.261, and E. schoeniclus and E. tristrami were separated from the above cluster at 0.375 and 0.469, respectively. E. elegans and E. bruniceps were segregated from the cluster at a distance of 0.510, which was much larger than genetic distances at the species level.
Seven species of the New world Emberizidae formed two groups including four and two genera, and they were separated at a distance of 0.440. It is important to point out that Paroaria and Passerina, which are classified as Cardinalinae were included in Lophospingus-Rhodospingus cluster in the New World Emberizinae.
Genetic variation and differentiation of the Emberizidae
At protein level, the amount of genetic variation in birds is average when compared with other animals . The proportion of polymorphic loci of the Emberizidae in this study was 0.14, which was nearly same as that found in other research on the Emberizidae . However, these values were lower than the avian mean. Proportion of polymorphic loci depends upon many factors such as combinations of loci, populations and sample size. Sample size in this study effected the proportion of polymorphic loci, because the correlation was significant (r = 0.61, p < 0.01) between sample size and the proportion of polymorphic loci (Table 1).
An other estimate of genetic variability is genetic heterozygosity, which is usually expressed in avian studies by observed and expected heterozygosity. The mean values of observed and expected heterozygosities were 0.036 and 0.042, respectively. Neither values were correlated with sample size in the present study, and were similar to previous data for the Emberizidae .
In avian research, it is difficult to use a single Mendelian population of birds, because the samples are usually made up of wandering birds or those from bird shops. If this type of mixed samples of the same species were used, the observed heterozygosity would be much lower than that of the expected heterozygosity in general . Emberiza rustica and E. spodocephala were mixed samples collected from Japanese and Chinese populations, and the latter species showed a lower value of observed heterozygosity than expected. Similar cases were also found in Volatinia jacarina and Paroaria coronata, which were collected from a bird shop. Observed heterozygosity is more reliable than expected heterozygosity in bird research because of the difficulty in sampling single Mendelian population of birds for genetic study.
At corresponding levels of the taxonomic hierarchy, birds consistently exhibit far smaller genetic divergence than do other orgnisms . An exceptional case was reported for Vireonidae, where the mean genetic distance among the five congeneric species of Vireo was far higher, D = 0.360 (0.027–0.578), although these distances remained small in comparison with those of many nonavian genera. The observed genetic distance in the genus Vireo is about 4–10 times greater than mean values between other avian congeneric species. One possible explanation for this result is that speciation of some Vireo might be relatively ancient in spite of their morphological uniformity .
Genetic divergence between the congeneric species of Emberiza in the present study is a second example of high genetic distances. The mean genetic distance is 0.334 (0.050–0.838) between the congeneric species of Emberiza. Genetic distances were much larger between four species (E. tristrami, E. elegans, E. bruniceps and E. schoeniclus) and the other Emberiza species. These species, however, had been treated as different genus , unlike the Vireo's case.
Phylogenetic relationships in the Emberizidae
As shown in Figure 1, seven species of Emberiza in the Old World formed a compact cluster and were joined by E. aureola, E. schoeniclus, E. tristrami, E. elegans and E. bruniceps in that order. The genetic distances at the branching points (0.261–0.510) were much larger than usual values within the same genus . These results appear inconsistent with the prediction by the current classification that species of Emberiza evolved recently in the Old World. According to Bock and Farrand , mean species number per genus for all birds is 4.11 and they analyzed several hypotheses why the 39 large genera of birds might be larger than average.
Emberiza is one of the largest genera and is considered to be a taxon that is artificially large, being composed of a single or several monophyletic subgroups each of which are equivalent to other genera in the family. In previous papers we showed small genetic distances among species, less than 0.2, of Anas in the Anatidae and Lonchura in the Estrildidae. These taxons are thought to contain many insular species, or those that have undergone adaptive radiations in several different parts of the world, respectively [11, 15]. In Emberiza, on the other hand, genetic distances were much larger, usually over 0.4, among species of the compact cluster and the other five Emberiza.
Species of Emberiza have often been treated as different genera by previous ornithologists. Despite the present tendency to group Emberiza, Wolters treats E. aureola, E. schoeniclus, E. tristrami, E. elegans and E. bruniceps as a different genus . There is, however, large difference between his classification and our genetic data, in that E. tristrami and E. elegans belong to the same genus and E. variabilis is treated as different subfamily from Emberizinae in his classification. The present study proposes from a genetic point that Emberiza is supposed to be subdivided into several genera, and future clasification of Emberiza will be different from the current.
Avise et al.  reported close genetic similarity of Pipilo (group VI of Norony et al.), and large genetic distance of Calcarius (group I), to other group I species of the Emberizinae in the New World. They showed, in addition, that Carpodacus, a species of Fringillidae, did not appear significantly more divergent from the New world Emberizinae than did Calcarius. Our genetic data revealed that species of the New World Emberizinae, including group II–IV, VII and the Cardinalinae, are comprised of a relatively compact cluster which is largely diverged from group I species of the Old World Emberizinae, that is Emberiza.
Distinctive behavior, called bilateral or double scratching, is thought to be very important for classification because this behavior is restricted to Pipilo and group I species in the New World, and is absent in the circumpolar species such as Calcarius. Bilateral scratching is also absent in the Old World buntings (Emberiza) and snow buntings (Plectrophenax) [8, 9]. The Old World buntings should be more closely related to the circumpolar buntings than to the New World Emberizinae, although direct genetic comparison has not been made among them.
We demonstrate that genetic distances among some of the Emberiza are larger than those found between subfamilies or families of New World Oscines. In our view, Emberiza represents an early divergence in the Old World prior to the speciation of the Emberizinae in the New World, and circumpolar genera such as Calcarius and Plectrophenax represent secondary speciation after colonization of North America by the ancestoral Old World Emberiza. Based on genetic data, we can not support the conventinal thought that Emberizinae have originated and diversified in the New World, while the Old World Emberizinae represents descendents of an invasion by the New World forms.
As shown in Figure 1, the Emberizidae studied here were subdivided into two main groups, the Old and New World Emberizidae, at the genetic distance of 0.672. Genetic distance between Paroaria coronata and P. gularis was 0.128, an average value acceptable for members of the same genus. Three species of Cardinalinae in this paper made a relatively compact cluster, but belonged to the Emberizinae in the New World. This result is inconsistent with the current classification of the Emberizidae, and demonstrates the difficulty of phylogenetic study of New World Oscines such as finches, sparrows, cardinals and tanagers, which are supposed to be the most recently developed and diversified large group of birds.
Recently, molecular techniques of DNA sequencing have been applied to evolutionary studies of organisms, and have been providing much information using restricted regions of single or a few genes. In avian studies, however, these data have not accumulated so far and comparative studies are difficult at present. On the other hand, data from protein electrophoresis have abundantly accumulated using many loci and species. It seems necessary to compare the genetic date from DNA sequencing with those from protein electrophoresis. Further genetic studies of Emberizidae and allies at DNA level are needed for understanding their evolutionary relationships, although they include too many species and genera.
We thank Asada Bird Shop, Tokyo and Bird Migration Research Center, Yamashina Institute for Ornithology, for providing bird samples. We also thank Drs. N. Kuroda and H. Kobayashi for their encouragement during this study, and Mr. J. Minton for critical reading of the manuscript.