Differentiation at 38 presumptive loci was examined among 30 species of palaeotropic finches (Estrildidae) by protein electrophoresis. Three species of Ploceidae, two of Fringillidae, and one of Emberizidae were included for comparison and the establishment of out-groups. Phenetic and cladistic analyses were employed, and both produced concordant patterns of relationships among the species. I conclude that all four families are closely related, with Estrildidae and Ploceidae grouped on one major sublineage and Fringillidae and Emberizidae on the other. Within the Estrildidae, three distinct radiations are identified, corresponding to the waxbill (Estrildae), mannikin (Lonchurae), and grassfinch (Poephilae) groupings in current classifications. Contrary to prevailing views, Aidemosyne is shown to be a member of Poephilae, not Lonchurae, and to be allied with Neochmia and Aegintha. Similarly, the relationships of Aegintha as currently presumed with the Estrildae are not consistent with the electrophoretic data. Overall, the data suggest a monophyletic origin for the Australasian Poephilae. Within this assemblage, however, Emblema and Poephila are clearly polyphyletic by current classifications. A major point of ambiguity in the data centers on the interconnections between the three major estrildid assemblages; at present, this can be treated only as an unresolved trichotomy. Received 24 October 1985, accepted 15 August 1986.

Department of Population Biology, Research School of Biological Sciences, Australian National University, P.O. Box 475, Canberra City 2601, Australia, and Division of Wildlife and Rangelands Research, C.S.I.R.O., Australia BEFORE 1980, most ornithological studies us- ing multilocus protein electrophoresis were confined to intraspecific variation (Baker 1974, Corbin et al. 1974, Manwell and Baker 1975), and they revealed levels of variation compa- rable to those of other vertebrates (Avise 1977). These studies also indicated a marked lack of differentiation between closely related species (Smith and Zimmerman 1976, Barrowclough and Corbin 1978), a result that has now been found to be a consistent feature in the Aves. In a series of papers on comparative protein electropho- resis within passerine families, Avise et al. (1980a-c, 1982) reported that genetic distances in birds were much lower than those observed in other vertebrates. Similar findings also were reported in nonpasserine families (Barrow- clough et al. 1981, Gutierrez et al. 1983, Adams et al. 1984). Whether the observed low isozymic genetic distances are a consequence of a slow rate of protein evolution (Avise et al. 1980c) or an artifact of overestimating the age of avian taxa (Baker and Hanson 1966) remains unre- solved (Avise and Aquadro 1982). Such genetic properties place obvious constraints on the use of protein electrophoresis in evolutionary stud- ies. Thus, the low genetic distances between populations (Barrett and Vyse 1982) and sub- species (Barrowclough 1980) of birds limit its use in determining levels of gene flow through hybrid zones or between populations. This weakness becomes its strength in determining relationships among species at generic and fam- ily levels. It is only because of the low isozymic divergence encountered among birds that elec- trophoresis can be used productively up to in- terfamilial comparisons (Avise et al. 1980c, Bar- rowclough et al. 1981). Accordingly, I have applied it here in an attempt to resolve species- group relationships in the estrildine finches, Estrildidae. Delacour (1943) divided the estrildine finches into three tribes on the basis of courtship, mouth markings of nestlings, and life habits. They are the waxbills (Estrildae), which are restricted largely to Africa; the grass finches (Poephilae), an Australasian-centered group; and the man- nikins (Lonchurae), which are pan-palaeotropic from Africa through southern Asia to Austra- lasia. While this grouping has been accepted generally, the exact composition of each tribe has been in constant dispute. In particular, the relationships of the genera Aegintha, Erythrura, Chloebia, Arnadina, and Aidernosyne have never been resolved satisfactorily. Mayr (1968) could not group the genera into three discrete tribes TABLE 1. Enzymes examined, buffers used, and tissue distribution of each enzyme. Run- Run- No. ning ning of buf- time b Enzyme (E.C. no.) Abbreviation loci Tissue fer a (h) Isocitrate dehydrogenase (1.1.1.42) IDH Glutamate-oxaloacetate transaminase (2.6.1.1) GOT Glucose-phosphate isomerase (5.3.1.9) GPI Mannose-phosphate isomerase (5.3.1.8) MPI Glycerophosphate dehydrogenase (1.1.1.8) GPDH 6 Phosphogluconate dehydrogenase (1.1.1.44) PGDH Malate dehydrogenase (1.1.1.37) MDH Glyceraldehyde-3-phosphate dehydrogenase (1.2.1.12) GAPDH Malic enzyme (1.1.1.40) ME Peptidase C (3.4.11) PEP L-A Peptidase B (3.4.11) PEP L-G-G Esterase c (3.1.1.1) EST Aldolase (4.1.2.13) ALD Triose-phosphate isomerase (5.3.1.1) TPI Guanine deaminase (3.5.4.3) GDA Glutamate dehydrogenase (1.4.1.3) GDH General protein (--) GP Fumerase (4.2.1.2) FUM Aconitase (4.2.1.3) ACON Lactate dehydrogenase (1.1.1.27) LDH Phosphoglucomutase (2.7.5.1) PGM Hexokinase (2.7.1.1) HK Superoxide dismutase (1.15.1.1) SOD NADP specific dehydrogenase a NADP nDH NAD specific dehydrogenase d NAD nDH 2 Muscle E 2 2 Muscle E 2.5 1 Muscle B 3 1 Muscle C 1.5 1 Muscle A 3 1 Muscle C 2.5 2 Muscle A 1.5 1 Muscle D 3 2 Muscle A 1.5 1 Muscle A 2 1 Muscle B 1.5 2 Muscle A 1 1 Muscle D 3 1 Liver D 3 1 Liver C 1 1 Muscle D 3 5 Muscle A 3 1 Liver E 2 2 Liver, muscle F 2 2 Muscle, heart D 3 2 Liver E 3 2 Muscle C 2 1 Muscle A 2 1 Liver C 2 1 Muscle, liver D 3 a A 50 mM TEM, B - 15 mM TEB, C 50 mM TEM + NADP, D = 50 mM TEM + NAD, E = 0.1 M Tris-citrate, F = 0.01 M citrate-phosphate. See Baverstock et al. (1980) for buffer recipes. b At 5 mA per 12 cm gel. ß Used method A in Harris and Hopkinson (1976) with 4-methyl-umbelliferyl-acetate. d "Nothing" dehydrogenases, i.e. bands were observed without the addition of any substrate to the staining mixture. and arbitrarily accepted Delacour's (1943) re- vision with minor modifications. This reflects the fact that the Estrildidae are unusual among birds in that plumage patterns are a relatively poor clue to relationships owing to extensive convergences and parallelisms (Harrison 1963, Mayr 1968). Moreover, the value given to the various morphological and behavioral charac- ters (Goodwin 1982) used is often subjective, both in interpretation and in application. For example, there is a gradation in the patterns of nestling mouth markings between the grass- finches and mannikins (Delacour 1943) such that the separation between these two tribes is ar- bitrary. Thus, although Mayr's (1968) arrange- ment was intended to be temporary, it still re- mains the only practical classification. I analyzed protein products of 38 presump- tive loci in 30 species of estrildid finches by both phenetic and cladistic methods. Six species of Ploceidae, Fringillidae, and Emberizidae were included for comparison and as out-groups in cladistic analyses. The resulting data are used to reassess relationships within the Estrildidae and the affinities of this family to other seed- eating birds. MATERIALS AND METHODS Specimens used were obtained from the wild and from aviaries. Wild-caught specimens of several species came from the Fitzroy River region of north- western Australia, and, whenever possible, prefer- ence was given to material from such sources. For other species only aviary-bred birds were readily available. In such cases, each sample was obtained from diverse sources in Victoria and New South Wales to minimize the chance of obtaining related individ- uals. A list of the specimens examined and their col- lection localities is presented in the Appendix. Liver, muscle, and heart tissue were excised and stored in liquid nitrogen. Tissues were homogenized using the recipe of Baverstock et al. (1980). Individ- uals were screened for 25 enzyme systems repre- senting 38 presumptive loci (Table 1). Enzymes were stained according to the method of Harris and Hop- kinson (1976) except GOT, where the procedure of Shaw and Prasad (1970) was used. All systems were run in a cellulose acetate matrix on a paper support (Cellogel, Chemetron, Italy). The measures of Nei (1978) and Rogers (1972) were used to estimate genetic distances between taxa. A phenetic analysis by the UPGMA method (Sneath and Sokal 1973) was performed summarizing the matrix of Nei's (1978)/, while the matrix of Rogers' (1972) / was subjected to a distance-Wagner procedure (Far- ris 1972). In the latter computation, three species of Ploceidae were used as an out-group. Both these anal- yses were performed on the BIOSYS package (Swof- ford and Selander 1981). In addition, a qualitative analysis using the method of FIennig (1966) was per- formed. The rationale and procedure of this analysis has been outlined in Baverstock et al. (1979) and Pat- ton and Avise (1983). After an initial dichotomy with- in the Estrildidae was determined through the plo- ceid out-group, each of the sister estrildine lineages were then treated as out-groups to each other from dichotomy to dichotomy. RESULTS ALLELIC FREQUENCIES AND HETEROZYGOSITIES AND GENETIC DISTANCE DATA Allelic frequencies for the 30 variable loci are presented in an appendix that is available from the author. The following loci were monomor- phic in all species: GOT-2, HK-2, ME-1, ME-2, MDH-1, MDH-2, NADnDH, GP-4, GP-5, TPI, and GDH. The range of heterozygosity mea- sures (Appendix) is large (0.00-0.06, mean = 0.03), and the average proportion of polymor- phic loci is 12% (see Appendix). These values are comparable to the levels of genetic variation reported for birds in general (Selander 1976, Avise and Aquadro 1982). Matrices of genetic distance values for the Estrildidae are presented in Table 2. Except for Poephila bichenovii, in which two subspecies were examined, all comparisons are between species. The Nei (1978) distance between P. b. bichenovii and P. b. annulosa of 0.01 (Table 2) is similar to that observed between subspecies of other birds (Corbin 1977). In comparison, genetic distances between congeneric species range from 0.10 to 0.25, while those between genera are higher, ranging from 0.30 to 0.77 (Table 3). These values are generally higher than those reported in sev- eral other passerine families where mean intra- and intergeneric values of/ are 0.10 and 0.26, respectively (summarized in Avise and Aquad- ro 1982). Distance values for species representing the three related families Ploceidae, Fringillidae, and Emberizidae are given in Table 4. The in- terfamilial genetic distances are low, ranging from 0.765 to 0.455, in contrast to the recorded distance across avian families of 1.00 (Avise and Aquadro 1982). CLUSTER ANALYSIS Phenetic analysis.--The UPGMA-generated phenogram (Fig. 1) weakly separates the Estril- didae from the Ploceidae and groups the es- trildids in three clusters corroborating the tribes of Delacour (1943) and Mayr (1968). They are the Poephilae (cluster 1), Estrildae (cluster 2), and Lonchurae (cluster 3) in the nomenclature of Mayr (1968). Apart from Aegintha temporalis and Aidemo- syne modesta, the species in cluster 1--all Aus- tralasian-always have been considered to be closely related (Morris 1958, Mayr 1968). Both A. temporalis and A. modesta are included here in the Poephilae, aligned with Neochmia ruff- cauda (Fig. 1). The most decisive split is between Poephila guttata, P. bichenovii, and the rest of the tribe. Although the genetic distance between P. guttata and P. bichenovii is relatively high (/ = 0.21), they stand together apart from other Poe- philae. The remaining species of Poephila cluster closely with the Neochmia-Aegintha-Aidemosyne assemblage, while Emblema guttata is grouped with Neochmia phaeton and not its congener, E. picta. The limited number of species and genera examined precludes comprehensive analysis of the relationships within cluster 2, the African Estrildae. Two points, however, merit com- ment. First, Pytilia melba and P. phoenicoptera cluster together and are close to the rest of the Estrildae, despite their extensive karyotypic dif- ferences (Christidis 1983). If Nei's (1978) / is related to time since divergence, then the branching patterns (Fig. 1) suggest that the two species of Pytilia have diverged rather recently both from each other and from the rest of the Estrildae. This in turn indicates that the marked chromosomal differences within Pytilia (Chris- tidis 1983) also have arisen recently. Secondly, the species of Estrildae are phenetically more distant from one another than are those within the Poephilae. Such a result is not unexpected because it is generally agreed that the Estril- didae arose in Africa (Goodwin 1982), so Afri- can estrildine taxa could be expected to be older and more distantly radiated than their Asian and Australian counterparts. The third cluster, Lonchurae, is subdivided into five groups (Fig. 1). The first group, which includes the genera Erythrura and Chloebia, and the second, which is the genus Amadina, are distinct from the rest of the Lonchurae. With the exception of Padda oryzivora, the remaining species belong to Lonchura itself and separate into three groups. The first of these is mono- typic with L. pectoralis. The second is also mono- typic with L. bicolor, though it may reflect the fact that neither L. cucullata nor L. fringilloides-- both supposed close allies of L. bicolor--was in- cluded in the analysis. Padda oryzivora and the remaining five species of Lonchura comprise the final group. Here, extremely short branch points separate the species even though some of them, particularly P. oryzivora, are morphologically distinct. Wagner analysis.--Unlike phenetic analyses, the Wagner tree does not assume a constant rate of protein evolution (Fig. 2). The dendrogram generated in Fig. 2 uses three species of Plo- ceidae as an out-group. It is clear from the branch lengths that the rate of protein evolution varied along the different lineages. Despite this, there is considerable concordance between the UPGMA and Wagner networks. The branching patterns for the Poephilae in both networks (Figs. 1 and 2) are essentially the same, the only difference being a minor switch in position between Aegintha temporalis and Neochmia ruficauda. Moreover, the five clusters within the Lonchurae are the same. In the branching sequences of the Lonchurae, how- ever, there are differences. These are due to unequal amounts of protein change along a lin- eage, as shown in the differences in the lengths of branches leading to Amadina and Lonchura pectoralis in both networks. As a result the UPGMA analysis places L. pectoralis closer to the genus Lonchura than to the genus Amadina, while the Wagner network clusters Amadina closer. The arrangement in the Wagner tree reflects the greater number of ancestral character states (symplesiomorphies) and derived character states (synapomorphies) that L. pectoralis and Amadina share with the rest of Lonchura, re- spectively. This also accounts for the discrep- ancies in the position of Lagonosticta senegala of the Estrildae (Figs. 1 and 2). An alternative to the distance-Wagner pro- cedure for analyzing electrophoretic data is the use of percent fixed differences (Baverstock et al. 1982, Adams et al. 1984). For most species this method produces a Wagner network sim- ilar to that presented in Fig. 2, and so it is not repeated here. Some discrepancies occurred. On percent fixed differences, N. ruficauda is 8 units from N. phaeton, Aegintha temporalis, and Aide- mosyne modesta, while A. temporalis and A. mo- desta are 6 units from one another. The problem lies in the position of N. phaeton, which is 19 units from A. modesta, 14 units from A. tem- poralis, but only 8 units from N. ruficauda, a result that produces negative branch lengths. Such inconsistencies are quite common in electro- phoretic (Avise et al. 1980c) and immunological (Farris 1981) data and are probably the result of back-mutations and convergences. Here, fixed allelic differences indicate that it is more ap- propriate to group N. phaeton within the Neoch- mia-Aidemosyne-Aegintha cluster than to align it with Emblema guttata (Figs. 1 and 2). This align- ment is corroborated by morphological, behav- ioral (Goodwin 1982), and chromosomal (Chris- tidis 1986) data. Cladistic analysis.--A Hennigian cladogram drawn from qualitative analysis of electro- morphs (Fig. 3) was based on three species of Ploceidae as an out-group. This method of anal- ysis takes account of convergence and back-mu- tation in the construction of phylogenetic trees (Baverstock et al. 1979, Patton and Avise 1983). The specific characters defining each of the branches in the cladogram, the most robust of which are those defined by three or more char- acter states, are listed in Table 5. In agreement with UPGMA and distance- Wagner analyses, the cladogram resolves the Estrildidae into the three same base clades. The relationships of the Lonchurae to the two other clades, Estrildae and Poephilae, however, are ambivalent. Three derived character states of the electromorphs defining these clades--LDH- l(c), PGDH(g), and PGM-2(g)--link the Lon- churae with the Estrildae; an alternative group-- GPDH(c), SOD(j), and NADP nDH(g)--links the Lonchurae to the Poephilae instead, an arrange- ment supported by the distance-Wagner den- drogram. The UPGMA phenogram groups the 384 L. CHRISTIDIS [Auk, Vol. 104 Biochemical Systematics Estrildid Finches . . ...... TABLE 3. Mean genetic distances at different taxonomic levels within the Comparison a,b b (Nei 1978) SD SE Congeneric in Poephilae (10) Congeneric in Lonchurae (17) Congeneric in Estrildae (2) Intergeneric in Poephilae (45) Intergeneric in Lonchurae (49) Intergeneric in Estrildae (19) Interfamilial; Estrildidae vs. Ploceidae (93) 0.124 0.073 0.007 0.048 0.049 0.003 0.208 0.084 0.042 0.281 0.097 0.002 0.395 0.085 0.002 0.352 0.078 0.004 0.630 0.046 0.002 Number of pair-wise comparisons is in parentheses. Species compositions of the genera are based on the revision by Christidis (1987). Estrildae with the Poephilae on the GOT-i(c) synapomorphy. There is no single dominant factor that favors any of these three alternative branching pat- terns, although a closer connection between Poephilae and Lonchurae is supported by the Wagner dendrogram and one of the alternative Hennigian cladograms. Without other indepen- dent evidence corroborating any of the alter- native alignments, the relationship among the tribes remains an unresolved trichotomy. Al- though the degree of resolution of the Hen- nigian cladogram (Fig. 3) is less than in the UPGMA (Fig. 1) and Wagner (Fig. 2) networks, the composition and relationships of species within each of the estrildid tribes is generally consistent across all three analyses. Thus, in the Hennigian cladogram both Poephila guttata and P. bichenovii form a distinct clade within the Poephilae; Emblema picta is separated from all other grassfinches on one electromorph; and the two Neochmia species form a distinct clade, consistent with data from fixed allelic differ- ences. Clades within the Lonchurae are all de- fined by two or more derived character states (Fig. 3), and the clusters of species are similar to those obtained by the Wagner and UPGMA methods. There is parallel correspondence in the Estrildae. DISCUSSION The electrophoretic data were analyzed by three different methods, after the approach of Patton and Avise (1983). I produced a best-fit phylogeny by comparing the results of all three methods for agreement or discrepancy. In most instances there was good agreement among the three analyses, and consequently phylogenetic conclusions can be drawn with some confi- dence. The comparison also pinpointed areas of conflict, identifying discrepancies such as in the relationships between Neochmia ruficauda and N. phaeton. I tested the phylogenies obtained in the present study by an analysis of estrildid karyotypes (Christidis 1986) and found broad agreement with the electrophoretic data in the composition of Poephilae, Lonchurae, and Es- trildae, and the alignment of species internally. Although the electrophoretic data subdivide the Estrildidae into three groups corresponding largely to the "tribes" of Delacour (1943) and Mayr (1968), they do not establish definitive relationships among the tribes. Other evidence sheds little light on the question. From appen- dicular musculature, Bentz (1979) suggested that the Estrildae and Lonchurae form a cluster apart from the Poephilae within the Estrildidae; but this is not supported consistently by the elec- TABLE 4. Genetic distance measures in the families examined. Above diagonal: Rogers (1972) genetic distance; below diagonal: Nei (1978) unbiased distance. Species Family 1 2 3 4 5 6 7 1. Poephila guttata Estrildidae -- 0.433 0.450 0.419 0.533 0.504 0.538 2. Passer domesticus Ploceidae 0.547 -- 0.081 0.373 0.406 0.363 0.466 3. P. montanus Ploceidae 0.579 0.050 -- 0.403 0.422 0.361 0.473 4. Foudia madagascariensis Ploceidae 0.530 0.456 0.504 -- 0.445 0.478 0.502 5. Carduelis carduelis Fringillidae 0.751 0.514 0.530 0.589 -- 0.254 0.375 6. C. chloris Fringillidae 0.697 0.442 0.423 0.641 0.282 -- 0.375 7. Tiaris canora Emberizidae 0.765 0.621 0.632 0.693 0.460 0.455 -- Fig. 1. UPGMA phenogram based on Nei's (1978)/. trophoretic evidence (compare Figs. 1-3). Opin- ion has fluctuated, moreover, as to whether the Australian grassfinch fauna has been derived from one or several successive invasions (Mor- ris 1958; Harrison 1963, 1967). Immelmann (1962) and Harrison (1967) maintained that the Australian genera Emblema, Neochmia, and Ae- gintha were all direct descendants of a waxbill (Estrildae) invasion from Africa. This hypoth- esis was based on the assumption that these genera were more closely related to African Es- trilda than to other Australian elements. A sec- ond invasion by primitive mannikins (Lon- churae) was said to have given rise to Poephila and Aidemosyne. In contradicting this hypoth- esis, my results and the myological data of Bentz (1979) suggest instead a single, monophyletic origin for the Australian grassfinches (Poephi- lae) that includes Emblema, Neochmia, and Ae- gintha. The behavioral and morphological sim- ilarities between members of these genera and African Lagonosticta (Mitchell 1962) and Estrilda (Delacour 1943, Mitchell 1962) are evidently the result of convergences and parallelisms, not common ancestry. Neochmia, Aegintha, and Aidemosyne often are placed in three separate tribes (Wolters 1957, 1981; Mitchell 1962). Goodwin (1982) argued for a close relationship among them on the basis of nestling mouth markings and plumage pat- terns, but the evidence was inconclusive be- cause Aidemosyne shared several behavioral characters with the Lonchurae. The electropho- retic data nonetheless support a close relation- ship among these three genera. The phylogeny of the Poephilae (Figs. 1-3) also highlights the shortcomings in plumage patterns as determi- AFA LAT LMA / AER "?' / EAS PPH ' EGU ATE UBE N ASU Fig. 2. Distance-Wagner network based on Rogers' (1972) b. The tree is rooted by the "out-group method" (Farris 1972) using the Ploceidae (PDO, PMO, FMA). Species abbreviations are defined in the Appendix. 15 16 1 19 .PGU PBI PAC PCI PPE AMO NRU NPH ATE EGU EPI LPE CGO ETR LCA LFL LAT LMA LPU POR LBI AFA AER ASU AAM EAS PME PPH UBE LSE Fig. 3. Cladogram based on electromorphs. Num- bers at branch refer to synapomorphies (Table 5). nants of relationship. A case in point is the red upper-tail coverts of Emblema picta, E. guttata, and Aegintha temporalis, which have been used to unite these species in a single genus (McKean 1975). This grouping, however, is clearly poly- phyletic (Figs. 1-3). The electrophoretic results also demonstrate that the Lonchurae are a monophyletic assem- blage and include Amadina. Gutfinger (1976) and TABLE 5. Electromorphs that define dades (Fig. 3). Clade Apomorphic characters 1 ACON-2(b), PGM-2(d), Ndh(h) a 2 PGDH(k) 3 PGDH(h) 4 IDH-2(b), PGM~i(g) 5 GP-2(e) 6 GPDH(a), ACON-i(e), a PGM-2(c), LDH-i(f), a LDH-2(b) 7 IDH-i(e), GOT-i(e), SOD(j) 8 PGDH(g), MPI(d), HK-I(b), LDH-2(a) 9 GDA(a) 10 PGM-2(g) 11 GOT-i(d), Aid(d), GP-2(c), CP-3(a), LDH- l(e) 12 IDH-i(a), GOT-i(c), ACON-i(a), LA-2(h) 1 GOT-i(i) 14 IDH-i(f), ACON-i(i), a PGM-i(c), PGM-2(f), Est-l(d) 15 LDH-i(c), PGDH(g), PGM-2(g) 16 GAPDH(a), ACON-I(h), SOD(c) 17 Ndh(i), Est-l(c) 18 GDA"(e) 19 LA-2(f) 20 GP-2(a) Characters in which the plesiomorphic state could not be deter- mined. Goodwin (1982) contended that Amadina was a specialized offshoot of the Estrildae because it shared several morphological and behavioral traits (song and nestling mouth markings) with Pytilia. Such similarities, however, are evidently convergent or parallelisms. The genera Chloebia and Erythrura, moreover, are sister members of the Lonchurae and not related to Poephila (De- lacour 1943). The closeness of Chloebia and Erythrura (Fig. 1) is consistent with the views of Mitchell (1958) and Schodde and McKean (1976) that the former is an arid-adapted rep- resentative of rain forest-inhabiting Erythrura. In this assemblage, only Lonchura as presently constituted is paraphyletic. In particular, L. pec- toralis and L. bicolor are separate from each other and from the rest of the genus, a distinction consistent with their behavior (Guttinger 1976). Apart from A mandava, the species of Estrildae examined here have always been grouped to- gether. Harrison (1961) believed that Amandava was a specialized derivative of the Lonchurae, while Wolters (1957, 1981) saw similarities in behavior and plumage patterns between Aman- dava and poephiline Aegintha. Neither view is supported by the protein data, which demon- strate conclusively that Amandava is a member of the Estrildae. The electrophoretic results can also be used to provide relative time scales of divergence. According to the calibration of the avian "mo- lecular clock" by Gutierrez et al. (1983), most estrildid genera diverged about 8 million years ago and the three underlying tribes about 14 million years ago. Based on an incomplete and depauperate fossil record, however, the average age of most passerine genera is 3.75 million years (Romer 1966, Prager and Wilson 1980). Given the large discrepancy between the pro- tein and fossil time scales--a factor of 2--the accuracy of the latter must be questioned seri- ously. Here a combination of protein electro- phoresis with other corroborative data sets such as DNA-DNA hybridization, microcomplement fixation, and mitochondrial DNA sequence analysis has the potential to provide much more accurate estimates of times of divergence among passerine taxa. Relationships among the Estrildidae, Ploceidae, Fringillidae, and Emberizidae.--In attempting to determine an outgroup for the Estrildidae, three related families were examined (Fig. 4). The phenogram supports the current view that the Ploceidae and Estrildidae are more closely re- lated to one another than they are to any other group (Bock 1960, Sibley 1970, Bock and Mo- rony 1978). This is substantiated further by the low overall genetic distance of 0.63 (Table 3) between the ploceids and estrildids compared with the average of 1.0 across avian families in general (Avise and Aquadro 1982: fig. 2). How- ever, the placing of Passer in the Ploceidae was questioned by Pocock (1966) and Sibley (1970), who argued in favor of fringillid affinities. The electrophoretic data do net support such a con- clusion (Fig. 4); instead they show that Passer is allied to other ploceids and the Estrildidae, cor- roborating the results of DNA-DNA hybridiza- tion (Sibley and Ahlquist 1985). The relationships between Old World and New World seed-eating oscines have been con- troversial, centering on whether similarities in the bony palate are due to monophyly (Tordoff 1954, Mayr 1955) or reflect convergence result- ing from similar feeding strategies (Bock 1960, Raikow 1978). Although my electrophoretic data are limited, they are relevant to these questions. First, there is a clear separation between the ploceid-estrildid and the fringillid-emberizid assemblages. The similarity between the Frin- gillidae and Emberizidae is surprisingly high Fig. 4. UPGMA of the relationships between the Estrildidae (Poephila guttata), Ploceidae (Passer domes- ticus, P. montanus, and Foudia madagascariensis), Frin- gillidae (Caraduelis chloris and C. carduelis), and Em- berizidae (Tiaris canora). [Nei's (1978)/3 = 0.46; see Table 4] and consis- tent with the classification of Mayr and Amadon (1951), who treated them as subfamilies within the Fringillidae. Second, the separation be- tween Estrildidae-Ploecidae and Fringillidae- Emberizidae (Table 4) is not as great as would be expected if the two groups were unrelated (Bock 1960). On the other hand, affinity be- tween these two assemblages is corroborated by DNA-DNA hybridization data (Sibley and Ahl- quist 1985). Thus, the similar palatine processes in these four families may be homologous and evidence for monophyly, and are not conver- gent as argued by Bock (1960, 1963). The av- erage Nei (1978) distance among the four fam- ilies is only 0.59, which indicates that they are close phylogenetically. Before their phylogeny can be placed in context among other passer- ines, however, further work is required on such families as the Nectariinidae, Alaudidae, and Motacillidae. According to evidence from DNA- DNA hybridization (Sibley and Ahlquist 1985), these three families form a natural assemblage with the Ploceidae, Estrildidae, Fringillidae, and Emberizidae. The results of my study clearly indicate that multilocus protein electrophoresis could make a significant contribution in ex- amining the affinities within this assemblage. ACKNOWLEDGMENTS I thank Prof. B. John, Dr. B. J. Richardson, Dr. R. Schodde, and Dr. D. D. Shaw for reading the manu- script critically. Financial support for this study was provided through a Commonwealth Postgraduate Re- search Award, an Australian Museum Postgraduate Award, and the Australian National University. LITERATURE CITED ADAMS, M., P. R. BAVERSTOCK, D. A. SAUNDERS, R. SCHODDE, & I-I. Y. SMITH. 1984. Biochemical sys- tematics of the Australian cockatoos (Psittaci- formes: Cacatuinae). Australian J. Zool. 32: 363- 377. AvIsE, J. C. 1977. Is evolution gradular or rectan- gular? Evidence from living fishes. Proc. Natl. Acad. Sci. USA 74: 5083-5087. , & C. F. AQUADRO. 1982. A comparative sum- mary of genetic distances in the vertebrates. Pat- terns and correlations. Evol. Biol. 15: 151-185. --, & J. C. PATTON. 1982. Evolutionary genetics of birds. V. Genetic distances within Mimidae (mimic thrushes) and Vireonidae (vir- eos). Blochem. Genet. 20: 95-100. --, J. C. PATTONß & C. F. AQUADRO. 1980a. Evo- lutionary genetics of birds. I. Relationships among North American thrushes and allies. Auk 97:135- 147. ß --, & --.. 1980b. Evolutionary ge- netics of birds. II. Conservative protein evolution in North American sparrows and relatives. Syst. Zool. 29: 323-334. --, & --. 1980c. Evolutionary ge- netics of birds. J. Hered. 71: 303-310. BAKERß C. M. A., & H. C. HANSON. 1966. Molecular genetics of avian proteins. VI. Evolutionary im- plications of blood proteins of eleven species of geese. Comp. Blochem. Physiol. 17: 997-1006. BAKER, M. C. 1974. Genetic structure of two popu- lations of White-crowned Sparrows with differ- ent song dialects. Condor 76: 351-356. BARRETTß V. g., & E. R. VYSE. 1982. Comparative genetics of three Trumpeter Swan populations. Auk 99: 103-108. BARROWCLOUGHß G.F. 1980. Genetic and phenotypic differentiation in a wood warbler (genus Den- droica) hybrid zone. Auk 97: 655-668. ß & K. W. CORIN. 1978. Genetic variation and differentiation in the Parulidae. Auk 95: 691-702. --., & R. M. ZINK. 1981. Genetic differ- entiation in the Procellariiformes. Comp. Blo- chem. Physiol. 69B: 629-632. BAVERSTOCK, P. R., M. ARCHERß M. ADAMSß & B. J. RICHARDSON. 1982. Genetic relationships among 32 species of Australian dasyurid marsupials. Pp. 641-650 in Carnivorous marsupials (M. Archerß Ed.). Sydneyß Australia, Royal Zoological Society of New South Wales. ß S. R. COLEß B. J. RICHARDSON, & C. H. S. WATTS. 1979. Electrophoresis and cladistics. Syst. Zool. 28: 214-219. -, C. H. S. WATTSß M. ADAMSß & M. GELDER. 1980. Chromosomal and electrophoretic studies of Australian Melomys (Rodentia: Muridae). Austra- lian J. Zool. 28: 553-574. BENTZ, G. C. 1979. The appendicular myology and phylogenetic relationships of the Ploceidae and Estrildidae (Aves: Passeriformes). Bull. Carnegie Mus. Nat. Hist. No. 15. BOCK, W. J. 1960. The palatine process of the pre- maxilla in the passeres. Bull. Mus. Comp. Zool. 122: 361-488. 1963. Evolution and phylogeny in morpho- logically similar groups. Amer. Natur. 97: 265- 285. ß & J. J. MORONY, JR. 1978. Relationships of the passerine finches (Passeriformes: Passeridae). Bonner Zool. Beitr. 29: 122-147. CHRISTIDIS, L. 1983. Extensive chromosomal repat- terning in two congeneric species: Pytilia melba, L. and Pytilia phoenicoptera Swainson (Estrildidae: Aves). Cytogenet. Cell Genet. 36: 641-648. 1986. Chromosomal evolution within the family Estrildidae (Aves) I. The Poephilae. Ge- netica 71: 81-97. --. 1987. Phylogeny and systematics of estrildid finches and their relationships to other seed-eat- ing passerines. Emu 87: in press. CORBIN, K.W. 1977. Genetic diversity in arian pop- ulations. Pp. 291-302 in Endangered birds (S. A. Temple, Ed.). Madison, Univ. Wisconsin. --, C. G. SIBLEY, A. FERGUSON, A. C. WILSON, A. H. BRUSHß & J. E. AHLQUIST. 1974. Genetic poly- morphism in New Guinea starlings of the genus Aplonis. Condor 76: 307-318. DELACOUR, J. 1943. A revision of the subfamily Es- trildinae of the family Ploceidae. Zoologica 28: 69-86. FARRIS, J. S. 1972. Estimating phylogenetic trees from distance matrices. Amer. Natur. 106: 645-668. --. 1981. Distance data in phylogenetic analysis. Pp. 3-23 in Advances in cladistics. Proc. 1st meet- ing. Willi Hennig Soc. (V. A. Funk and D. R. Brooksß Eds.). GOODWIN, D. 1982. Estrildid finches of the world. Londonß British Museum (Natural History) and Oxford Univ. Press. GUTIERREZ, R. J., R. M. ZINK, & D. Y. YANG. 1983. Genetic variationß systematicß and biogeographic relationships of some galliform birds. Auk 100: 33-47. GUTWINGER, H. R. 1976. Ethology and taxonomy of the genera Amadina, Lepidopygia and Lonchura (Es- trildidae). Bonn. Zool. Beitr. 27: 218-244. HARRIS, H., & D. g. HOPKINSON. 1976. Handbook of enzyme electrophoresis in human genetics. Am- sterdam, North Holland Publ. Co. HARRISONß C. J.O. 1961. Affinities of the Red Ava- davat, Amandava amandava (L.). Bull. Brit. Orni- thol. Club 82: 126-132. 1963. The taxonomic position of the Crim- son Finch and Red-browed Finch. Emu 63.' 48- 56. 1967. Apparent zoogeographical dispersal patterns in two avian families. Bull. Brit. Orni- thol. Club 87: 63-72. HENNIG, W. 1966. Phylogenetic systematics. Chica- go, Univ. Illinois Press. IMMELMANN, K. 1962. Beitrage zu einer vergleich- enden Biologie australischer Prachtfinken (Sper- mestidae). Zool. Jahrb. Syst. 90: 1-196. MANWELL, C., & C. M. A. BAKER. 1975. Molecular genetics of avian proteins XIII. Protein polymor- phism in three species of Australian passerines. Australian J. Biol. Sci. 28: 545-557. MAYR, E. 1955. Comments on some recent studies on song bird phylogeny. Wilson Bull. 67: 33-34. 1968. The sequence of genera in the Estril- didae (Aves). Breviora 287: 1-14. --, & D. AMADON. 1951. A classification of re- cent birds. Amer. Mus. Novitates No. 1496. MCKEAN, J. L. 1975. "Ploceidae." Pp. 22-23 in In- terim list of Australian songbirds (R. Schodde, Ed.). Melbourne, Royal Australasian Ornitholo- gists Union. MITCHELL, J.G. 1958. The taxonomic position of the GouldJan Finch. Emu 58: 395-411. 1962. The taxonomic position of the Crim- son Finch. Emu 62: 115-125. MORRS, D. 1958. The comparative ethology of grass- finches (Erythruae) and mannikins (Amadinae). Proc. Zool. Soc. London 131: 389-439. NEI, M. 1978. Estimation of average heterozygosity and genetic distance from a small number of in- dividuals. Genetics 89: 583-590. PAttON, J. C., & J. C. gvIsE. 1983. An empirical eval- uation of qualitative Hennigian analyses of pro- tein electrophoretic data. J. Mol. Evol. 19: 244- 254. Po1/2ocK, T.N. 1966. Contributions to the osteology of African birds. Proc. 2nd Pan African Ornithol. Congr. 14: 83-94. PRGER, E. M., & A. C. WLSON. 1980. Phylogenetic relationships and rates of evolution in birds. Proc. 17th Intern. Ornithol. Congr.: 1209-1214. RAIcow, R.J. 1978. The appendicular myology and relationships of the New World nine-primaried oscines (Aves: Passeriformes). Bull. Carnegie Mus. Nat. Hist. 7: 1-43. ROGERS, J. S. 1972. Measures of genetic similarity and genetic distance. Studies in Genetics 7. Univ. Texas Publ. 7213: 145-153. ROMER, A. S. 1966. Vertebrate paleontology, 3rd ed. Chicago, Univ. Chicago Press. SCHODDE, R., & J. MCKEAN. 1976. The relationships of some monotypic genera of Australian oscines. Proc. 16th Intern. Ornithol. Congr.: 531-541. SELANDER, R. K. 1976. Genic variation in natural populations. Pp. 21-45 in Molecular evolution (J. Ayala, Ed.). Sunderland, Massachusetts, Sinauer Assoc. SHAW, C. R., & R. PRASAD. 1970. Starch gel electro- phoresis--a compilation of recipes. Blochem. Ge- net. 4: 297-320. SIBLEY, C.G. 1970. A comparative study of the egg- white proteins of passerine birds. Bull. Peabody Mus. Nat. Hist. 32. --, & J. E. AHLQUIST. 1985. The phylogeny and classification of the Australo-Papuan passerine birds. Emu 85: 1-14. SMITH, J. K., & E.G. ZIMMERMAN. 1976. Biochemical genetics and evolution of North American black- birds (family: Icteridae). Comp. Blochem. Phys- iol. 538: 319-324. SNEATH, P. H. A., & R. R. SOAL. 1973. Numerical .taxonomy. San Francisco, W. H. Freeman. SWOFFORD, D. L., & R. K. SELANDER. 1981. Acomputer program for the analysis of allelic variation in genetics. J. Hered. 72: 281-283. TORDOFF, H.B. 1954. A systematic study of the avian family Fringillidae based on the structure of the skull. Misc. Publ. Mus. Zool. Univ. Michigan 81: 1-41. WOETERS, H.E. 1957. On the generaEstrilda (Swains) and Lagonosticta (Cub). Bull. Brit. Ornithol. Club 77: 62-63. 1981. Die systematische Stellung des dor- nastrilds, Aegintha temporalis (Latham) (Aves, Es- trildidae). Bonn. Zool. Beitr. 32: 137-144. APPENDIX. Species examined, sample sizes, localities, and genetic variability measures. Percentage Locality of loci Mean Species Abbreviation (sample size) polymorphic b heterozygosity Estrildidae Poephila guttata PGU A (15) 21.1 0.053 P. bichenovii bichenovii PBI A (4), C (6) 15.8 0.016 P. b. annulosa PAN B (3) 5.3 0.026 P. acuticauda PAC A (8), B (9) 26.3 0.062 P. cincta PCI A (4) 18.4 0.059 P. personata PPE A (2), B (2) 7.9 0.026 Aidemosyne modesta AMO A (5) 18.4 0.058 Aegintha temporalis ATE A (5), C (2) 15.8 0.041 Emblema guttata EGU A (4) 13.2 0.046 E. picta EPI A (3) 10.5 0.440 Neochmia ruficauda NRU A (9), B (16) 26.3 0.011 N. phaeton NPH B (5) 10.5 0.037 Chloebia gouldiae CGO A (6) 10.5 0.013 Erythrura trichroa ETR A (3) 2.6 0.009 Lonchura pectoralis LPE A (1), B (11) 13.2 0.015 L. fiaviprymna LFL A (3), B (1) 2.6 0.013 L. malacca atricapilla LAT A (2) 2.6 0.013 L. maja LMA A (1) 0.0 0.000 L. castaneothorax LCA A (4), B (5) 15.8 0.012 L. bicolor LB! A (2) 2.6 0.013 Amadina fasciata AFA A (4) 2.6 0.000 A. erythrocephala AER A (4) 5.3 0.013 Amandava subfiava ASU A (3) 5.3 0.018 A. amandava AAM A (1) 0.0 0.000 Estrilda astrild EAS A (3) 2.6 0.000 Uraeginthus bengalus UBE A (3) 7.4 0.044 Lagonosticta senegala LSE A (3) 5.3 0.009 Pytilia melba PME A (5) 5.3 0.011 P. phoenicoptera PPH A (6) 2.6 0.000 Ploceidae Passer domesticus PDO C (16) 13.2 0.023 P. montanus PMO D (3) 7.9 0.018 Foudia madagascariensis FMA A (2) 0.0 0.000 Fringillidae Carduelis carduelis CCA D (2) 2.6 0.009 C. chloris CCH D (2) 5.3 0.026 Emberizidae Tiaris canora TCA A (3) 5.3 0.018 = aviary stock; B = Fitzroy River region, northwestern Australia; C = Australian Capital Territory region; D = Melbourne, Australia. locus was considered polymorphic if the frequency of the most common allele did not exceed 0.99.