THE RELATIONSHIPS OF THE HIGHER TAXA OF : PROBLEMS IN PHYLOGENETIC REASONING

JOEL CRACRAFT

Department of Anatomy University of Illinois at the Medical Center Chicago, Illinois 60680

It will be the thesis of this paper that the rela- third if they share a common ancestor that is tionships of the higher taxa of birds (here, not at the same time an ancestor of the third subfamilies and higher) are very poorly under- taxon. The term “relationship” defined in this stood and consequently that our classifications way will be devoid of the notion of degrees of reflect this lack of knowledge. Indeed, the morphological resemblance or divergence. A disparity between what many ornithologists more detailed treatment of this approach to might interpret from the classifications that relationships will be presented below. have been published and the reality of avian phylogeny is so great that our classifications RELATIONSHIPS OF THE HIGHER can scarcely be called “phylogenetic.” I want TAXA: STATEMENT OF THE PROBLEM to discuss the reasons for this situation and will It probably is a general impression among attempt to show that avoidance of certain many ornithologists, including some system- kinds of phylogenetic reasoning should help atists, that the phylogenetic relationships of to improve our understanding of avian rela- avian families are fairly well understood. This tionships. A major purpose of this paper, then, conclusion might follow from the various ma- is to review the theories and methods used jor classifications (but less so from their ac- within avian systematics. To my knowledge companying discussions) such as Mayr and this has not been attempted previously, and Amadon (1951), Wetmore (1960), or Storer such an examination will hopefully prove use- ( 1960b, 1971)) none of which is drastically ful since there now seems to be a renewed in- different from the other. Moreover, the most terest in morphological and systematic work widely accepted classification, that of Wet- in ornithology. more (1930, 1951, 1960), has not changed At the outset I want to emphasize that this significantly in three decades. However, as paper is concerned with the study of phylog- Stresemann (1959:274) notes, the considerable eny and not with the theory or methodologies number of detailed systematic studies “have used to construct classifications. I consider made it apparent that the relationship of cer- the problems of classification to be secondary tain species or groups of species is far less to those of phylogeny, because a proper under- unequivocably established than one would con- standing of the latter is a prerequisite for the clude from a study of currently adopted sys- study of numerous evolutionary phenomena tems, the authors of which attempt to present (see below). A given theory of phylogeny a simplified phylogenetic tree of birds.” need not have a close correspondence with A paradox seemingly exists within the cur- any specific theory of classification. Of rent state of systematic knowledge. On the course, classification cannot be divorced en- one hand, our classifications suggest a stable tirely from phylogeny since the former de- system, one in which our ideas about relation- pends (at least in most taxonomic philosophies ships seem to be either rather firmly estab- today) on knowledge of the latter, but I be- lished, or alternatively, one in which we have lieve the central issues confronting avian sys- little confidence in the systematic work on the tematics lie within the domain of phyloge- higher taxa; in either case, we have the ap- netic inquiry rather than that of classification pearance of a system which embodies a seem- per se. Hence, I will restrict this discussion ingly unlimited amount of inertia. On the to the theories and methodologies used to de- other hand, the large number of systematic termine relationships within the higher taxa papers appears to tell us that the classifications of birds. need a major revision, that they may reflect In the following pages I will use the term conservatism of opinion rather than reality. “relationship” in the sense of recency of com- Part of the problem, I think, is that system- mon ancestry (genealogy). Thus, two taxa are atists have emphasized “classification” to the more closely related to each other than to a detriment of determining phylogeny; that is,

[3791 The Condor 74:379392, 1972 380 JOEL CRACRAFT we have not been interested in the construc- sary to present the viewpoint advocated here. tion of specific scientific hypotheses such as Such a procedure will then provide a basis on which family might share a most recent com- which the discussion of previous work can be mon origin with another family but rather in compared and evaluated. In principle, I am more general problems such as to what family supporting many of the ideas formalized by or to what order this or that taxon might con- Hennig (1966) and commonly termed cladis- veniently be placed. The desire to have a tics or phylogenetic systematics. I am not “stable” or “orderly” classification is under- claiming unqualified acceptance of all of his standable but, in our quest for this stability, theories or methods, but I do believe that his classification has become the proverbial cart system contains those methods offering the to the phylogenetic horse. best chance of determining phylogenies within As was implied above, many of the mor- the higher categories of birds, and that his sys- phological studies attempting to resolve the tem is more objective than those in current relationships of “problem taxa” have not been use. For these reasons his work deserves close well received or adopted by the majority of attention by avian systematists. systematists. Dissatisfaction with these stud- A discussion of the methodology involved in ies is leading either to a state of near apathy in determining phylogenetic relationships (as which it is assumed we can only “subjectively” (defined in the introduction) starts with sev- determine relationships (Stresemann 1959) or, eral basic assumptions. Through the phylog- to the use of techniques by some workers that eny of a higher taxon (e.g., the class Aves), appear to them to hold almost unlimited prom- every monophyletic assemblage of species ise in resolving relationships (e.g., Sibley 1960, (i.e., orders, families, genera) evolves certain 1970) but which, in fact, seem to promulgate character-states that are unique to that line- the same theoretical and methodological prob- age. Thus, these character-states have been lems as the work they criticize. derived within these lineages after branching Still, the question must be asked why many from their ancestor, and if we can recognize morphological investigations, whether of mole- such character-states, then we will have a cules, muscle, or bone, fail to arouse wide- basis on which to define monophyletic as- spread support. Is it because the workers have semblages. (In a very real sense, knowledge not produced detailed comparative studies? of the mere existence of these monophyletic I do not think this is the case. We have stead- lineages is dependent by definition on the rec- ily accumulated comparative data but they ognition of derived characters. ) However, the seem to be leading us nowhere in determining recognition of derived character-states is fre- avian relationships. Rather, I suspect that we quently difficult, and systematists have con- have failed to consider in detail the theoret- structed a number of methodologies for their ical and methodological background with recognition (see below). which to analyze the data that have been col- It is basic to cladistic theory that phyloge- lected. Obviously, the theoretical framework netic branching involves dichotomies (Hen- used by a given investigator influences strongly nig 1966). Theoretically, this should be ac- the manner in which he gathers and inter- cepted as a generality by most workers, prets his data. Unfortunately, there has been especially when their concern is the higher little direct discussion as to the working meth- categories. The supposed exceptions at the ods employed by different avian systematists. species level are few in number and lack de- In my opinion, an interchange of theoretical tailed substantiation, thus the arguments raised ideas may enhance the value of future mor- by various critics fail to undermine the hy- phological work, and I hope that this paper pothesis of dichotomous branching (Darlington can contribute to such a discussion, In a 1970). Even at the species level it is very im- sense, the paper is presented with the idea of probable that populations which are postu- challenging the status quo but also, and more lated to have become isolated more or less at importantly, with the desire of finding some the same time will all attain species-level dif- methods that will give us the answers to our ferentiation simultaneously. Hennig ( 1966: questions about avian relationships. 211) correctly notes that “the impossibility of determining with certainty the sequence of DETERMINATION OF dichotomous cleavages in a group never means PHYLOGENETIC RELATIONSHIPS that all the species arose simultaneously (by Before discussing the approaches that have radiation) from one stem species.” The use been used by various ornithologists to deter- of cladistic theory does not guarantee that mine phylogenetic relationships, it is neces- workers will be able to determine dichotomous RELATIONSHIPS OF HIGHER TAXA OF BIRDS 381

A relationships at all taxonomic levels. Indeed, , I I suspect that it will be difficult, if not im- , B C possible, to determine dichotomous inter- D E F G generic and interspecific relationships of many avian families, especially passerines. When a species divides into two separate lineages, we can term these lineages “sister- groups” (Hennig 1966). Likewise, when we trace the evolutionary trends of homologous features in these sister-groups, we see that one lineage will possess a more or less primitive character-state of that feature whereas the other lineage will have a relatively derived character-state. If we repeat this procedure for a number of characters, we will discover that each lineage possessesboth primitive and derived characters. It becomes obvious that if derived character-states can be identified, then monophyletic lineages can be constructed. It is equally obvious that taxa which share primi- tive character-states-i.e., those inherited from a distant ancestor-cannot be united on the basis of those characters. Figure 1A summa- B ti rizes these conclusions. Each of the lineages, D E F G A-G, is defined as monophyletic because they each possess a suite of derived characters which have their primitive counterpart in the coordinate sister-group. If figure 1A is as- sumed to represent the true phylogeny, then a character analysis which yields the phylogeny of figure 1B would be incorrect (in this case the characters of suite 2, assumed to unite F and D/E, are in fact not derived). Thus, this theoretical system provides a straightforward method for invalidating a proposed phylogeny, and the invalidation is not based on ambig- uous arguments about which groups share a greater number of “more important charac- ters.” Phylogenies can be refuted by (a) show- ing that a given suite of characters used to define the lineages are either primitive or of FIGURE 1. A, Dendrogram of a phylogeny assumed multiple origins, and (b) arguing for a dif- to be true. Dark rectangles represent derived char- ferent set of derived characters which yields acter-states; open rectangles represent corresponding a different phylogenetic hypothesis. If one primitive character-states of same characters. Taxon B is defined monophyletic by character-state suite 4, is working with higher categories of birds, it taxon C by suite 3, and so forth. B. Phylogeny of should be possible to detect multiple origins taxa A-G arrived at by incorrect analysis of character- of character-states if a sufficient number of state suites 1-4. See text. characters are analyzed. If character 1 argues for a relationship between taxa A and B, and ships in a relatively unambiguous fashion, other characters (also seemingly derived) sup- with a reasonable degree of confidence, and port a closer affinity between B and C, then with a minimum of subjective decision mak- one might suspect character convergence in ing; the other approaches are either reducible character 1, depending upon the number of to this in principle or they have a greater contradictory characters and the clarity of chance of producing erroneous phylogenies. their evolutionary trends. Accepting the above theoretical framework As I will try to support in a subseyuent sec- as valid for determining phylogenies, we must tion, this theoretical model seems to be the now consider the critical problem of develop- best one available for determining relation- ing methods capable of recognizing primitive 382 JOEL CRACRAFT and derived character-states. This problem trends, then time occurrence data should be has been discussed in detail by numerous avoided or interpreted with considerable cau- workers (e.g., Maslin 1952; Simpson 1961; tion. The congruence of evolutionary trends Hennig 1966; Mayr 1969; Crowson 1970; and stratigraphic sequences is not always Kluge 1971) and need not be belabored here. reliable, and comparative analysis is our only Possibly without exception, these and other independent means of checking this reliability. systematists agree that comparative analysis A final method of recognizing a primitive of the distribution of character-states is the character-state might be the use of develop- best method (and parsimony the criterion) mental studies. For example, if we are at- for recognizing the primitive condition. Thus, tempting to determine the primitive-derived “a condition which occurs in related forms sequence of two bones which have apparently outside the group [under investigation] is fused in some taxa, and we find in those taxa probably primitive within it” (Crowson 1970: with a single adult bone that two were present 106; see also the extended discussion in Hen- in an early stage of development, we may be nig 1966:88ff). It is essential that a worker on firm ground in postulating the latter as the specify the hierarchical level for which he primitive condition. Although this method considers the character to be primitive or often can provide important data concerning derived, because these latter characteristics character sequences, the amount of work will are relative to taxa of different levels. For be too time-consuming for most systematic example, whereas the possession of feathers studies. is a derived character for the class Aves, it is The recognition of the primitive condition a primitive character within the group. is sometimes very easy, sometimes impossible. Obviously, it will not be possible to deter- Usually it involves detailed comparative study. mine relationships within a given taxon unless Occasionally, characters will be found that a character analysis of closely related taxa is show an apparent derived condition in two also made. On this basis alone many system- taxa thought to be unrelated. Usually such atic papers in the avian literature are ques- multiple origins of derived characters may be tionable for they provide little or no com- detected by the study of numerous other char- parative basis on which to arrive at reliable acters which clearly define the taxa in ques- judgments of primitiveness. The problem of tion. Furthermore, derived characters can be deciding which “closely related taxa” to com- lost or modified to a secondarily primitive pare is frequently real and sometimes subjec- condition. It is also possible that two sister- tive. When the argument for considering par- groups will each diverge so much from their ticular taxa as close relatives is weak, then it common ancestor that relative degrees of prim- is clear that one should make as broad a com- itiveness cannot be recognized for some char- parative study as possible before deciding acters. These problems merely indicate that what character-states are primitive. this aspect of phylogenetic analysis requires A comparative analysis such as advocated detailed work and the realization that defini- above permits the recognition of evolutionary tive answers are not always possible (see the trends and also the direction of those trends. workers cited above for extended discussions To my knowledge there exist no objective of these problems). In general, if it is not pos- criteria with which to recognize the direction sible to tell primitive-derived sequences using of trends outside of a comparative analysis. distribution patterns of the character-states, Another method used to discern primitive it will usually be advisable to abandon the character-states is the occurrence of these use of that character and thus not weaken character-states at different levels in time. phylogenetic hypotheses. It is my belief that Primitive characters are generally expected to if one is considering the relationships of the occur earlier in time. Whereas this may be higher taxa of birds (especially nonpasserines and is frequently true, numerous exceptions where divergence is greater), an adequate can be found mainly because of the incom- number of characters (skeletal, myological, pleteness of the fossil record. Differential etc.) will almost always be available, and of species possessing derived char- therefore rejection of some characters will not acter-states and/or uneven paleontological be a serious problem. sampling can result in primitive character- The theory and practical application of the states appearing later in time than derived above phylogenetic procedure are discussed character-states. The direction of these trends in great detail by Hennig ( 1966), Brundin can be established only by a comparative ( 1966), and Crowson ( 1970). The central is- analysis, and if the latter does not reveal sue which they raise is that affinities of two RELATIOKSHIPS OF HIGHER TAXA OF BIRDS 3s3

taxa can bc demonstrated only by the use of The concept of overall resemblance, while shared derived character-states and that prim- having provided ornithology with the genera1 itive similarities contain no phylogenetic in- outlines of avian relationships, has been one formation. of the major causes for the ambiguity and conflict that I discussed in a previous section. RELATIONSHIPS OF THE HIGHER This follows from the fact that such an ap- TAXA: EXAMINATION OF proach has a large component of subjective PREVIOUS METHODOLOGY decision-making and is easily susceptible to It will be the purpose of this section to re- errors, stemming not so much from observa- examine and analyze some past and present tion but from interpretation. Overall resem- approaches advocated or used to determine blance also provides no objective basis on phylogenetic relationships in birds. I will at- which to judge the validity of the conclusions tempt to show that many of these approaches produced by this method. Cladistic method- are more likely to lead to erroneous conclusions ology, while certainly not devoid of all sub- or produce relatively useless and ambiguous jectivity, uses a more critical and less subjec- statements about relationships than is the tive character analysis than does that of method discussed above. It should be under- overall resemblance. stood that this criticism is formulated from The primary reason for rejecting the con- the preceding theoretical remarks. cept of overall resemblance as a working method in avian systematics is that we never GENERAL OVERALL RESEMBLANCE know whether the similarities being used to argue relationships are primitive or derived. That most ornithologists, past and present, Too often they are primitive and thus lead to have based or do base the relationships of incorrect conclusions. A classic example is the birds on some notion of general overall re- interfamilial relationships of the Galliformes. semblance can hardly be questioned. Nearly On the basis of overall resemblance several all of the 19th and 20th century morphologists opinions have gained wide acceptance: have followed this approach. A characteristic (1) the and Megapodiidae are more statement, for example, would be that taxa A closely related to each other than to the other and B resemble each other more than either galliforms (e.g., Holman 1964; Wetmore 1960); does C, therefore A and B are more closely and (2) the family Gallinuloididae is related. Such an approach often leads to the more closely related to the Cracidae than to vague and ambiguous conclusion that taxon the Phasianidae (Tordoff and MacDonald A is intermediate in structure between B and 1957; Wetmore 1960; Brodkorb 1964). Re- C and therefore is related to both. Imprecise cently, Lester L. Short and I undertook a study statements such as this abound in the system- of the interfamilial relationships of the fossil atic Iiterature. With few exceptions, avian and recent galliforms, and we based our phy- paleontologists use the overall resemblance logeny on a detailed analysis of primitive and argument for relationships. General overall derived character sequences (see fig. 2). Al- resemblance has provided the foundation for though this study will be published elsewhere, a number of well-known classifications and is, certain results can be mentioned here for they I believe, the basis for the differences one provide an example of how the use of overall sometimes sees in them (e.g., Mayr and Ama- resemblance has led to erroneous conclusions. don 1951; Verheyen 1960; Wetmore 1960). In figure 2 we see that the suites of characters Most, if not all, of the statements concerning used to unite the megapodiids and cracids relationships proposed by the molecular sys- and the gallinuloidids and cracids (suites 10 tematists (e.g., Sibley 1960, 1970; Sibley et al. and 11, respectively) consist of primitive char- 1969; Hendrickson 1969; to name only a few) acters which are devoid of phylogenetic in- have genera1 overa resemblance as their formation since they were inherited from their principal working method. [The concept of common ancestor (character suite 1). Some general overall resemblance applies to those primitive characters included in both suites studies that “weight” their taxonomic char- 10 and 11 are the absence of an intermeta- acters as well as to those that consider their carpal process on the carpometacarpus and a characters of equal weight (e.g., Schnell shallow internal notch of the sternum. HoI- 1970a, 1970b). In terms of their basic ap- man (1964) lists numerous other similarities proach, “weighting” seems to be of little im- between cracids and megapodiids, and in portance to most ornithologists, especially nearly every case he himself claims they are since it is rarely discussed and procedures for primitive within the galliforms. Study of the weighting are not mentioned.] skeletal anatomy of galliforms does indeed 384 JOEL CRACRAFT

MEGAPODIIDAE CRACIDAE GALLINULOIDIDAE N”M,D,NAE PHASIANINAE Because of the above argument, I suggest that the concept of overall resemblance be dropped from avian systematics. I would rec- ommend using it only when a detailed char-

IO acter analysis has failed to yield primitive- derived sequences of the character-states. If overall resemblance is used in such a case, it should be made explicitly clear that the sim- ilarities used to define groups may be primi- tive and not derived. I suspect that reliance on overall resemblance should be unnecessary in any sophisticated systematic study involv- ing a reasonable number of characters. The use of overall resemblance may be inevitable in some paleontological studies where frag- mentary material can often make a character analysis very difficult.

USE OF CONSERVATIVE FIGURE 2. A phylogeny of the Galliformes based CHARACTERS on an analysis of primitive-derived character-states. Note that suites of character-states 10 and 11 con- A number of systematists have argued that sideredby some authors to unite Megapodiidae with conservative characters, i.e., those which have Cracidae and Gallinuloididae with Cracidae are primi- evolved slowly and changed relatively little, tive and thus devoid of phylogenetic information. Dark rectangles represent derived character-states; are to be preferred in trying to discern phy- open rectangles represent corresponding primitive logeny (e.g., Farris 1966). Compared to the character-statesof same characters.See text. concept of general overall resemblance, ap- plication of the theory of conservative char- acters has received scant attention from orni- reveal that the Megapodiidae and Cracidae thologists. A notable recent example is Jehls’ share a large number of striking features, but (1968) study of downy plumage patterns in in the absence of a character analysis based on the Charadriiformes. Conservatism of taxo- valid systematic theory one cannot draw phy- nomic characters is also a strong influence on logenetic conclusions from these similarities. the thinking of molecular systematists (Sib- A previous study of the relationships of ley 1967: 12,197O: 19). some fossil gruiform families (Cracraft 1968) Those people who support the use of con- used general overall resemblance as a work- servative characters have, in my opinion, over- ing method, and the affinities of some of these looked certain aspects of character analysis. families were misinterpreted. A more recent Many of these conservative, slowly evolving investigation using primitive-derived se- characters are present in the group under quences has shown, for example, that the study and do indeed help these workers to Bathornithidae are not more closely related recognize subordinate taxa (e.g., families to the Cariamidae (Wetmore 1960; Cracraft within an order ), but these characters do not 1968) but rather to other fossil gruiform fam- necessarily indicate phylogenetic affinity be- ilies such as the Idiornithidae and (some- tween these taxa because they may be primi- what more distantly) the Geranoididae (Cra- tive for the group as a whole. Indeed, primi- craft 1972). tive characters are by definition relatively The influence of the concept of overall conservative. Constancy (i.e., conservatism) resemblance cannot be overestimated even to of a character throughout a group is of no this day, since most systematic papers of the significance unless one can show that this is last decade or so have employed this theo- a derived constancy and not primitive. Far- retical framework (e.g., Cottam 1957; Johns- ris ’ suggestion (1966) that the relative taxo- gard 1960, 1964; Ligon 1967; Maclean 1967; nomic value of a character is proportional (or Cracraft 1967; Hudson et al. 1969; Stegmann 1969; to name a few). Any of these papers somehow related) to the degree of conserva- may be correct in their conclusions, but until tism shows a lack of understanding of how the results are verified by a character analy- characters are used and evaluated to define sis, we do not know whether the relationships monophyletic groups. they advocated are based on primitive or In general, the use of conservative charac- derived similarities, ters is unnecessary in phylogenetic studies RELATIONSHIPS OF HIGHER TAXA OF BIRDS 385

since conservatism becomes irrelevant once a is difficult for me to see just how functional primitive-derived character analysis has been analysis can do this in the absence of a com- completed. Whereas the notion of conserva- parative study, which is the only context in tism is of no help in inferring phylogenies, it which this method would have any theoretical may be of some value in interpreting various validity; moreover, a comparative study will evolutionary phenomena once phylogenetic yield information about unique occurrence. relationships are known. Multiple origins of characters can only be known after a phylogeny has been determined THE PROBLEM OF ADAPTIVE AND on other characters. Functional analysis does FUNCTIONAL SIGNIFICANCE OF not provide any objective criteria for deter- TAXONOMIC CHARACTERS mining the probability of unique origin, and It has long been thought by many systematists I do not know of a single example in which that the more significant a character is adap- multiple origins cannot be recognized from tively or functionally, the less important it is conventional comparative study. Likewise, I taxonomically. Bock (1967) quite rightly do not know of a single example in which criticized the reasoning behind this conclu- unique origin ( or multiple origins ) has been sion. Still some ornithologists continue to im- demonstrated by functional analysis in the ab- ply that “adaptive” characters are questionable sence of comparative data. for taxonomic studies. What is probably meant The above does not imply, of course, that in most of these cases is that characters which functional analysis has no role in the study of undergo rather rapid evolutionary changes phylogeny nor does it imply that those work- are of little taxonomic value; this is close to ers using functional analysis disregard com- the idea that if a character or character-state parative data. If one includes the study of has a high probability of multiple origins, it evolutionary mechanisms and the explanation becomes relatively useless as an indicator of of adaptive trends in the realm of phyloge- phylogenetic relationships. In essence, this is netic study, then functional analysis is of attacking the problem in the wrong manner. paramount importance. Functional analysis It would be best to undertake a primitive- can be helpful in defining characters them- derived character analysis, determine the rela- selves. In general, the more functionally in- tionships, and then discuss such factors as terrelated features are, the less valid it is to multiple origins of characters (see below). treat them as separate characters in order not The concept of multiple origins or unique to bias ones’ argument. Nevertheless, we occurrence leads us to a discussion of the role should eliminate the notion that functional of functional analysis in determining the reli- analysis is essential for the determination of ability of taxonomic characters (see Bock phylogenetic relationships. This hinders non- 1960). Unfortunately, there has been too functional comparative studies, leads to un- great an emphasis on functional analysis as due criticism of many systematists who do not the philosophers’ stone of phylogenetic studies feel compelled to include functional analysis, (e.g., Cain 1959), and it has led to a mis- and perhaps causes some anxiety in those in- understanding as to what role functional anal- vestigators who have not included such analy- ysis should play in the study of phylogeny ses. (see Bock 1969a and Zusi 1971, for a discus- THE THEORY OF HOMOLOGY sion of systematics and functional analysis). This misunderstanding has also led some I do not intend to dwell upon the theory of workers to criticize the conclusions of others homology for it has been discussed by numer- who have neglected to provide a functional ous authors. While many workers accept the analysis (see, for example, Bocks’ 1960, 1963, definition of homology as pertaining to struc- critiques of Tordoff 1954, and McDowell tures that can be traced back to the same 1948, respectively); a critique centered upon structure in the common ancestor, a number methods of comparison rather than the ab- of workers question the usefulness of the sence of functional analysis would have been theoretical construct in explaining historical more appropriate. phenomena (e.g., Nelson 1970:378-379). The One of the major uses of functional analy- differences between the ideas expressed here sis is said to be in determining the relative and those contained in the concept of homol- probability of unique occurrence of taxonomic ogy as advocated by some evolutionary sys- characters (Bock 1969a:443-444); other work- tematists (Simpson 1961; Bock 1969a, 1969b) ers who have discussed this problem (Wilson may not be significant. One of the reasons 1965) do not consider functional analysis. It why the use of derived characters is less am- 386 JOEL CRACRAFT biguous than the usual notion of homology is ingly in contradiction, Bock (1969a:441) also that the former is not tied so closely with the notes that paradaptive aspects are those dif- problem of levels of monophyly (Simpson ferences that arise between phyletic lines and 1961). Bocks’ statement (1969b:72) “A hi- “are either adaptive or nonadaptive, accord- erarchical series of homologies can be es- ing to whether they are selected or rejected by tablished by repetitive determinations using selection.” ever-increasingly narrower and more precise Bock argues that because paradaptive as- conditional phrases” appears to be the neces- pects involve comparisons between lineages sary approach if one is to argue relationships (i.e., horizontal comparisons ), the distinction using homologies. Yet I am unaware of any between monophyletic assemblages is that they study in which the branching sequences of a possess different paradaptive aspects of taxo- higher taxon of birds (or within any other class nomic characters (Bock 1969a:441). Thus, all of vertebrates) have been determined using a we need to do is develop some measure of the hierarchy of well-argued homologies. probability of multiple origins of these par- Followers of the phylogenetics of Hennig adaptive aspects and WC will have a criterion do not use the concept of homology in the to judge the value of taxonomic characters. same sense as do evolutionary systematists, Those paradaptations with a higher probabil- because the former workers do not believe it ity of unique origin would have a greater is possible to trace structures back to the cor- taxonomic value. responding structure in an ancestor (they be- I believe the concept of paradaptation is lieve actual identification or observation of unnecessary for the determination of taxo- ancestors is impossible; see below). Rather, nomic value of characters and can lead to phylogeneticists are interested in primitive- erroneous conclusions. In support of this derived sequences of the structures they are statement we can examine Bocks’ example of studying. Phylogeneticists recognize the im- the arrangement of the toes in perching birds portance of comparing the “same” (i.e., ho- (1969a:444) : “The arrangements of the toes mologous ) structures in making primitive- for perching feet in birds appear to be re- derived analyses, but the ability to identify stricted to the anisodactyl, syndactyl, hetero- these characters as “homologous” is almost dactyl, and zygodactyl types (Bock and Miller never a problem. The criteria used to discern 1959). If the ancestral condition of three primitive-derived character sequences are less anterior toes and a short, elevated posterior ambiguous and laborious than those criteria hallux is considered, then the probability that normally used to establish the hierarchy of any of these four paradaptive toe arrange- homologies. The latter are founded on some ments will arise only once is about 0.25.” criterion of overall resemblance, and state- From this he concludes: “the arrangement of ments almost always must be subjective and/ the toes is a feature with little taxonomic or ambiguous, e.g., “these features are so value.” Using this reasoning, it should be similar in these taxa that they could have come possible to eliminate nearly all taxonomic about only by origin from a common ances- characters that have been proposed in avian tor . . . therefore they are homologous . . . systematics. For example, how many basic therefore they indicate affinity. . . .” The con- patterns (i.e., “paradaptive aspects”) are there cept of primitive-derived sequences avoids for jaw muscles, leg muscles, number of pri- this form of argument by using methods that maries, bill structure, egg-white proteins, and are not strictly dependent upon a measure of so forth? Can it be claimed that for any char- overall similarity. acter in which there are four basic arrange- ments (not to speak of more than four), the THE CONCEPT OF PARADAPTATION probability of unique occurrence is about 0.25 According to Bock ( 1967:67), “Those aspects and therefore that these arrangements are of of a feature that are dependent upon, result- little taxonomic value? One could postulate, ing from, or under the control of chance-based on the other hand, that the heterodactyl foot evolutionary mechanisms may be termed of trogons is of great taxonomic value. It par&apt&e (from para‘ ’ and adaptive‘ ),’ is clearly a derived character that helps dcm- meaning besides‘ adaptive ’ in the sense that onstrate that the trogons are a monophyletic these aspects are not dependent upon selection group. Zygodactylism cannot bc used to claim and hence cannot be judged in the range of relationship between genera of the Piciformes, adaptive to nonadaptive. Paradaptive aspects Psittaciformes, or Cuculidae, but it can be of a feature are dependent only upon the acci- used to define the monophyletic nature of dental evolutionary mechanisms. . . .” Seem- these higher taxa if it can be shown that this RELATIONSHIPS OF HIGHER TAXA OF BIRDS 387 toe arrangement is derived for each group. relationship. The latter can only be proposed It is difficult to imagine any meaningful way once a relationship has been demonstrated in which the probability of unique origin can between one of these taxa and a third taxon. be estimated prior to the determination of a I therefore suggest that arguments of con- phylogeny. vergcncc not be used to refute or support phy- I therefore suggest that the concept of par- logenetic inferences. Rather, this approach adaptation not be used to evaluate taxonomic should be replaced by an argument based on characters. derived character sequences.

THE ARGUMENT OF CONVERGENCE DISCUSSION There arc several notable examples of ornithol- THE STUDY OF PHYLOGENY ogists arguing against a particular relation- ship of taxa by claiming that the organisms All workers agree that phylogeny involves the are convergent and not really related; this I branching of lineages and the subsequent call the argument of convergence. The “con- divergence of these lineages. To most the vergence” of the is a case in point (Mc- study of phylogeny also includes a discussion Dowell 1948). A more striking example (at of functional differences between and within least it is cited more often) is the argument lineages, differences in evolutionary rates, that the loons, grebes, Hesperornis, and other aspects of adaptive radiation, etc. Despite divers are convergent and therefore not re- the considerable agreement among the stu- lated. The argument has its basis in Stolpes’ dents of phylogeny, honest differences of paper (1935) on the hindlimb and has been opinion are present; it is important to attempt accepted by numerous workers (e.g., Mayr and a resolution of this discord. Because many Amadon 1951; Storer 1956, 195Y, 1960a, 1971; ornithologists may not be aware of the con- Stresemann 1959; Wetmore 1960). A re-eval- flicts between the different workers, I want uation of this problem raises doubts about the to present a brief discussion of some of these validity of this type of phylogenetic argu- views. Hopefully, each reader will examine mentation. I want to discuss briefly these an d evaluate these views impartially, espe- methods, but a detailed systematic analysis of cially with regard to what they might do for the taxa involved will be presented at a later the study of avian phylogeny. date ( Cracraft, unpubl. ) . For our purposes, we will discuss the dif- A strong argument can be made that most ferences between the followers of Hennig of the shared morphological (principally ( 1966)) sometimes termed “cladists” ( genealo- osteological) features of loons, grebes, and gists or phylogeneticists) because they em- their fossil allies are derived within the class phasize the importance of phyletic branching Aves and therefore one must conclude that or cladogenesis, and those of Simpson ( 1961), these taxa evolved from a common ancestor. Mayr (1965, 1969), and others, who place a The features that many call convergent are great emphasis on divergence and similarity actually derived and indicate monophyly. (“evolutionary taxonomists”). I do not plan to Stolpe (1935) did not present a valid argu- review all of the literature. Rather, it seems ment for concluding convergence in the diving appropriate to compare the views of Mayr birds. In his paper he merely showed that ( 1965), whose stimulating work has greatly there are differences among these taxa, and influenced avian systematics and who has from this he concluded nonrelationship and given probably the most detailed and articu- convergence. late criticism of the cladists, with those of Hen- We may ask whether it is acceptable to nig. [Darlington ( 1970) has also presented a critique of Hennig and his followers, but his refute a given relationship by the argument of discussion raises no serious theoretical ob- convergence. In none of the papers cited jections to cladism that were not treated ear- above is there ( 1) a primitive-derived char- lier by Mayr.] It will be necessary to examine acter analysis, or (2) any evidence presented Mayrs’ views in terms of my own experience that convincingly argues a relationship of one in studying the higher taxa of birds, thus I of these groups to a separate lineage of birds. will not claim to support or defend all of the The last point is critical because, by definition, cladists ’ thinking. I am more interested in a judgment of convergence must be based on developing a system that will help us under- an a priori assumption of relationship. The stand the history of avian taxa rather than de- fact that two taxa (e.g., loons and grebes) bating every detail and exception of cladistics. might show a number of morphological dif- The conflicts between the cladists and evo- ferences cannot serve as an argument for non- lutionary taxonomists begin with the meaning 388 JOEL CRACRAFT

A 6 systematist must ask himself which definition has the greater consistency, greater objectivity, and greater usefulness in representing the rela- tionships of birds and in explaining the phe- nomenon of phylogeny. In my viewpoint the answer is the definition of the cladists. My reasons are as follows: 1. Because we assumed the phylogeny of figure 3 as given, the relationships recognized by cladists will always remain the same as long as the branching sequence remains the same. This is not true for the evolutionary taxonomist whose decisions about relation- ships will change with differences in diver- gence. How much divergence does there have to be between Tr and Tz before B is considered more closely related to C/D than to A? Is it not possible for two or more evo- FIGURE 3. Hypothetical dendrogram. Little mor- lutionary taxonomists to disagree on the inter- phological divergence has taken place between times T1 and T, compared to that between T, and T::. See pretation of the amount of divergence and text. therefore have contrary opinions as to the closest relative of B? What kinds of objective criteria are to be used? The viewpoint of the of relationship. (I am not interested here in evolutionary taxonomist is ambiguous and the definition per se but in the applicability subjective. On the other hand, any number of this definition to the study of phylogeny. of cladists would always agree on how to ex- We can argue ad nuuseum about definitions press relationship given the same branching without realizing that it is the implications of sequence regardless of the degrees of diver- the definition that are of paramount impor- gence. Science, it seems to me, should choose tance.) The differences in viewpoint can be the objective over the subjective, the unam- illustrated using figure 3, which we can con- biguous over the ambiguous; students of phy- sider to represent a true phylogeny of an logeny should do no less. order of birds with four families (this example 2. In terms of practicality, a cladistic defini- is slightly modified from Mayr 1965). We tion of relationship is preferable. Suppose we will assume that there has been little mor- assume the taxa in figure 3 to be the following phological divergence between times Tr and (which I believe to be most probably the ac- T, (i.e., little divergence between taxa A and tual phylogeny): A, Megapodiidae; B, Crac- B ) but a greater amount of divergence be- idae; C, Gallinuoididae; and D, Phasianidae tween T, and T,. Cladistics would define phy- (sensu lato). Any statements about relation- logenetic relationship in terms of the branch- ships by a cladist (i.e., B is more closely re- ing sequence, i.e., B is more closely related to lated to C/D than to A) tell us information C/D than to A because B and C/D share a about the branching sequences of the taxa. more recent common ancestry that is not at A statement by the evolutionary taxonomist to the same time an ancestor of A. According to the effect that B is closer to A would tell us Mayr (1965:79), “If there was very little evo- A and B are most similar (at least in his lutionary change between T, and Tz, and a opinion), which in turn may lead others to great deal of evolutionary change between T, infer that A and B also had a more recent and Ts, then obviously [emphasis mine] B is common ancestry (which is not true in this more closely related to A than it is to C/D. example). Thus, using the definition of the The argument of the cladist fails to recognize evolutionary taxonomists may lead to incor- that the term relationship has two different rect assumptions about ancestry; that of the meanings, genetic relationship and genealog- cladist cannot imply incorrect information ical relationship.” Mayr continues (p. 79) about common ancestry nor does it imply any- that “it is no longer legitimate to express rela- thing about similarity. tionship in terms of genealogy.” This example Mayr ( 1965), Bock ( 1968), and Darlington indicates that there is a difference in defini- (1970) make a point of emphasizing that clad- tion between the two schools, and no amount ists ignore differential rates of evolution or of rhetoric can obscure this point. Thus, each assume equal rates. Thus, according to Mayr RELATIONSHIPS OF HIGHER TAXA OF BIRDS 389

(1965:80), “The most serious weakness of both questionable (except on a purely theoretical the straight cladistic and straight phenetic level) and simply a hoped for, but still approach is that they completely disregard mythical, picture of reality. the frequent occurrence of exceedingly dif- The central problem of phylogeny is not the ferent evolutionary rates.” This viewpoint of interpretation of dendrograms (or cladograms) the evolutionary taxonomists is clearly in- but their construction. Evolutionary taxon- correct as it pertains to cladistic analysis. omists would give us a dendrogram in which Although he does not discuss evolutionary origin from a common ancestor is subservient rates in detail, Hennig (1966:88) rejects the to degrees of similarity. Their philosophy and possibility of equal rates and recognizes “the working methods are usually inadequate to proven fact that there are differences in the recognize cladistic relationships. Certainly rate of evolution.” In my opinion, cladists within birds, the sister-group relationships of probably pay more attention to differential the families are very poorly known despite rates than does the evolutionary taxonomist, decades of evolutionary . Avian because the entire system of the former rests systematists should be concerned with the his- on the use of primitive and derived character- tory of birds. This should begin with the de- states, which are by definition reflections of termination of cladistic relationships and then different rates of evolution. This is less true proceed to a study of divergence and asso- of evolutionary taxonomists, whose system is ciated phenomena. founded principally on overall similarity. An important point often ignored (or not under- THE IMPORTANCE OF PHYLOGENY stood) by evolutionary taxonomists in their The importance of knowing the phylogenetic work or in their criticism of cladistics is that relationships of the avian taxa is obvious. evolutionary rates are only meaningful after Once phylogenetic relationships have been the branching sequences of taxa are known. determined, the interpretation of numerous The study of evolutionary rates depends on a historical events will become much more sig- cladistic approach to phylogeny rather than nificant. The examples of convergence, mul- on one of overall resemblance. tiple origins of characters, and evolutionary Mayr (1965:79) believes “genetic similar- rates have been discussed already. Several ity” should be the dominant theme of phylo- other aspects can be mentioned. geny, and in his opinion “When a biologist The use of cladistic relationships has more speaks of phylogenetic relationship, he means predictive value in the study of biogeography relationship in gene content rather than clad- than does evolutionary taxonomy (Nelson istic genealogy.” This concept of “genetic re- 1969). Returning to figure 3, if we consider latedness” has been adopted by Sibley (1970: the cracids to be more closely related to the 21) in his discussion of passerine relation- megapodes than to the gallinuoidids-phasianids ships. Sibley has carried the term much fur- as evolutionary taxonomists have proposed, ther than what Mayr would probably want then our statements as to the place of origin and even speaks of the genetic relatedness of and probable paths of dispersal will be dif- families rather than of gene pools of species. ferent from those postulated by a cladist. This use of “genetic relatedness” is even more Perhaps the major problem of the endless curious when one considers that Sibley studied discussions concerning the biogeographical only five or six genes (assuming one gene, affinities of the African and Australasian avi- one protein) at most, using techniques which faunas is that the phylogenetic affinities he himself admits tell us nothing about the (sister-group relationships) of the organisms actual structure of the genes (see Cracraft are so poorly understood. One of the strong- 1971). But even on a theoretical level “ge- est arguments for cladistics and against the netic similarity” has its problems since any emphasis of similarity by the evolutionary statement comparing genetic relatedness of taxonomists is that the historical biogeography organisms must of necessity be inferred from of -their patterns of origin and dis- some measure of overall resemblance. Mayr persal-is not a reflection of morphological (1965:85-86) rejects a close correspondence resemblance. On the other hand, these fac- between genotype and phenotype (at least in tors do reflect or have a close relationship to any exact, predictable manner), yet he still the cladistic affinities of the taxa. Meaningful uses this principle as a basis for his whole explanations of biogeographical events are im- system. As any ornithologist will realize, the possible given incorrect cladistic relationships. practical use of “genetic relatedness” within Darlington ( 1970: 10-17) has criticized the the higher (or lower ) taxa of birds is highly biogeographic reasoning of several cladists, 390 JOEL CRACRAFT

particularly Brundin ( 1966). In my opinion, by the time dimension. Of course, the nature of much of this criticism misses the point be- the paleontological material makes analysis cause “a biogeographic analysis implies, log- more difficult. ically follows from, and at best can be no Second, it is impossible to make definitive more reliable than, a prior phyletic analysis” statements about the systematic position of (Nelson 1969:246). Darlington ( 1970: 8-9) fossils until one has an understanding of the believes that most taxonomists use the same relationships of the recent groups. Once methods as the cladists in determining rela- branching sequences are determined for recent tionships. Although I cannot speak for other groups, then fossil taxa can be incorporated. specialties, this is unequivocally not true in Third, it is pointless either to look for an- ornithology. Avian phylogeny and classifica- cestors or to consider fossils to be ancestors. tion have their basis in the concept of overall Ancestral-descendant relationships are un- resemblance, which some of us reject as an knowable and unprovable; no criteria exist for unreliable approach to phyletic relationships. recognizing a particular fossil as an ancestor. I would therefore predict that much of the Statements can be made about cladistic rela- biogeographic analysis of passerine distribu- tionships, and this provides a basis for in- tion will be fruitless until cladistic relation- telligible evolutionary interpretations. For ships have been resolved in more detail. It is example, I believe (Cracraft 1972) that the frustrating and somewhat futile to compare Eocene Eogruidae and the Ergil- the avifaunas of the southern continents unless ornithidae share a most recent common an- one knows the relationships of the taxa in- cestry within the order . The volved. eogruids are clearly more primitive in a large Another, minor predictive aspect of cladism number of characters and are probably near would be the possible explanation of relation- the ancestry of the ergilornithids. But how ships and distribution patterns of parasites, can one know whether the eogruids are really especially Mallophaga. Many of the peculiar the ancestors? It is possible to discuss all the distribution patterns of Mallophaga, which relevant evolutionary events without resorting have led to needless speculation about avian to proposing the eogruids as the ancestral relationships, might disappear if the cladistic group. relationships of the hosts were known. Naturally the loss of the idea that fossils are ancestors (in the sense that they can be rec- THE ROLE OF PALEONTOLOGY ognized as ancestors) will be difficult for The present state of avian paleontology is some. But in doing so we will increase the very poorly understood by ornithologists as precision of the working methods in paleontol- well as by other vertebrate systematists. Un- ogy and eliminate a considerable number of fortunately, a re-evaluation of much of the irresponsible statements, which have not nec- earlier work has only begun and many un- essarily come from the paleontologists them- described forms remain unstudied. A sig- selves, but from those who interpret and apply nificant number of the described species, their results. especially those from the early Tertiary, are incorrectly placed as to family, sometimes FROM PHYLOGENY TO CLASSIFICATION even to order. However, the avian fossil record The emphasis of this paper is phylogeny, not is not as poor as some would believe, and in classification. As do most systematists, I be- certain cases phyletic lines can be drawn lieve phylogeny should precede classification (e.g., Cracraft 1972). On the basis of my work -that is, if the classification is to be called in paleontology I have come to several gen- “phylogenetic.” The strongest differences be- eral conclusions regarding the fossil record tween the cladists and evolutionary taxono- of birds and the role it should play in avian mists are perhaps those involving the transla- phylogeny. These conclusions are not re- tion of phylogeny into classification. I do not wish to enter the controversy in this paper. stricted to birds and in fact may be derived Rather than discuss theory at this time, I theoretically using any group of organisms would prefer to see possible methods of classi- (see Schaeffer et al. 1972 for an excellent dis- fication debated after the phylogenetic rela- cussion of the theory and working methods of tionships of various orders of birds are known. paleontology). Preliminary findings within some orders sug- First, the working methods of paleontology gest that a purely cladistic classification in- are exactly the same as those of neontology. volving dichotomies is feasible and very The methods of comparison and interpretation satisfactory and should not disturb the sensi- are identical and should not be influenced bilities of most ornithologists. RELATIONSHIPS OF HIGHER TAXA OF BIRDS 391

CONCLUSIONS BOCK, W. J. 1968. Phylogenetic systematics, clad- istics and evolution. Review of phylogenetic It is hoped that the contents of this paper will systematics. Evolution 22:646-648. help to improve and promote the study of BOCK, W. J. 1969a. Comparative morphology in phylogeny in the class Aves, and I have tried systematics, p. 411-448. In Systematic biology. to raise and discuss some issues which focus Nat. Acad. Sci. Publ. No. 1692. on this improvement. The crux of the prob- BOCK, W. J. 1969b. Discussion: the concept of homology. Ann. New York Acad. Sci. 167:71- lem, it seems to mc, is that the notion of gen- 73. eral overall resemblance is unreliable and BOCK, W. J., AND W. DE MILLER. 1959. The scan- should be replaced by the method of charac- sorial foot of the woodpeckers, with comments ter analysis of primitive-derived sequences. on the evolution of perching and climbing feet I have tried to give my reasons for this con- in birds. Amer. Mus. Novitates No. 1931:1-45. BRO~I~ORB, P. 1964. Catalogue of fossil birds: clusion, the reasons being based on actual ap- Part 2 ( Anseriformes through Galliformes ). Bull. plication to several orders of birds. Florida State Mus. 8: 195-335. Every avian systematist, including myself, BRUNDIS, L. 1966. Transantarctic relationships and would freely admit that the method of overall their significance, as evidenced by chironomid resemblance has provided us with the skeletal midges. Kgl. Svenska Vcten. Hand]. 11( 1) :l- 472. framework of avian relationships. Thus, most CAIS, A. J. 1959. Function and taxonomic impor- of the families are probably correctly placed tance. Syst. Assoc. Publ. 3:5-19. as to order. But relationships of families COTTAhI, P. A. 1957. The pelecaniform characters within orders are very poorly known, and it of the skeleton of the Shoe-bill Stork, Balaeniceps is at this level that cladistic analysis will be rex. Bull. Brit. Mus. (Natur. Hist.), Zool. 5 (3):1-71. superior to overall resemblance. I have great CRACRAFT, J. 1967. On the systematic position of optimism that with careful analysis of char- the Boat-billed Heron. Auk 84:529-533. acters we will be able to resolve the phylo- CRACHAFT, J. 1968. A review of the Bathornith- genetic relationships of avian families, par- idae ( Aves, Gruiformes ), with remarks on the ticularly nonpasseriforms where divergence is relationships of the suborder Cariamae. Amer. Mus. Novitates No. 2326: l-46. greatest, in the near future. I believe that the CRACRAFT, J. 1971. Review of “A comparative incessant speculation that has characterized study of the egg-white proteins of passerine the study of avian phylogeny and classification birds” by C. G. Sibley. -Banding 42:157- will be eliminated only with the adoption of 161. the cladistic approach. Can we afford to con- CRACRAFT, J. 1972. The systematics and evolution of the Gruiformes (Class Aves) 3. Phylogeny tinue as we have? of the suborder Grues. Bull. Amer. Mus. Natur. Hist. In press. ACKNOWLEDGMENTS CROWSOn-, R. A. 1970. Classification and biology. Nearly all of the ideas expressed in this paper were Heinemann Educational Books Ltd., London. germinated while I held a Frank M. Chaoman Fel- DARLINGTON, P. J. 1970. A practical criticism of ibwship, 1969-70, at the American Museum of Nat- Hennig-Brundin “phylogenetic systematics” and ural History. Hours of intense debate with several Antarctic biogeography. Syst. Zool. 19: l-18. staff members, particularly Gareth J. Nelson, did FARRIS, J. S. 1966. Estimation of conservatism of much to dampen my enthusiasm for many of the characters by constancy within biological pop- concepts that have come to bc expected of American ulations. Evolution 20:587-591. evolutionary systematists. I also want to thank the HESDRICKSON, H. T. 1969. A comparative study of following individuals who either read and/or dis- the egg white proteins of some species of the cussed, sometimes with vigor and rigor, the contents avian order Gruiformes. Ibis 111:80-91. of this paper: Walter J. Bock, John R. Bolt, George HESSIG, W. 1966. Phylogenetic systematics. Univ. A. Clark, Jr., Niles Eldridge, Ernst Mayr, Gareth J. Illinois Press, Urbana. Nelson. Kenneth C. Parkes. Donn E. Rosen. Lester L. HOL~~AN, J. A. 1964. Osteology of gallinaceous Short, Leigh‘ Van Valen, and Richard L. Zusi. It is birds. Quart. J. Florida Acad. Sci. 27(3) :230- necessary to state, as probably will be obvious, that 252. there was sometimes a lack of unanimity with my HUDSON, G. E., K. M. HOFF, J. VAIWER ‘ BERGE, AK~ ideas; nevertheless I am most grateful for the com- E. C. TRIVETTE. 1969. A numerical study of ments of the above persons. the wing and leg muscles of Lari and Alcae. Ibis 111:459-524. LITERATURE CITED JEHL, J. R., JR. 1968. Relationships in the Charad- BOCK, W. J. 1960. The palatine process of the pre- rii (shorebirds): a taxonomic study based on maxilla in the passeres. Bull. MLIS. Camp. Zool. color patterns of the downy young. San Diego 122:361-488. Sot. Natur. Hist. Mem. 3:1-54. BOCK, W. J. 1963. The cranial evidence for JOHSSGARD, P. A. 1960. Classification and evolu- affinities. Proc. XIII Int. Ornithol. Congr., tionary relationships of the sea ducks. Condor Ithaca ( 1962 ) : 39-54. 62:426-433. BOCK, W. J. 1967. The use of adaptive characters JOHNSGARD, P. A. 1964. Comparative behavior and in avian classification. Proc. XIV Int. Ornithol. relationships of the eiders. Condor 66:113-129. Congr., Oxford ( 1966) :61-74. KLUGE, -4. G. 1971. Concepts and principles of 392 JOEL CRACRAFT

morphologic and functional studies, p. 3-41. In STEGMANN, B. 1969. Uber die systematische Stel- A. J. Waterman [ed.] structure and lung der Tauben und Flughiihner. Zool. Jahrb. function. Macmillan Co., New York. Abt. Syst. 96: 1-51. LIGON, J. D. 1967. Relationships of the cathartid STOLPE, M. 1935. Colymbus, Hesperornis, Pocliceps: vultures. Oct. Papers Mus. Zool. Univ. Michigan ein Vergleich ihrer hinteren Extremitat. J. No. 651: l-26. Ornithol. 83: 115-128. MACLEAN, G. L. 1967. Die systematische Stellung STORER, R. W. 1956. The fossil loon, Colymboides der Flughiihner ( Pteroclididae ) . J. Ornithol. minutus. Condor 58:413-426. 108:203-217. STORER, R. W. 1958. Loons and their wings. Evo- MASLI~\T,T. P. 1952. Morphological criteria of phy- lution. 12: 262-263.’ logenetic relationship. Syst. Zool. 1:49-70. STORER, R. W. 1960a. Evohrtion in the diving MAYR, E. 1965. Numerical phenetics and taxo- birds. Proc. XII Int. Omithol. Congr. Helsinki nomic theory. Syst. Zool. 14:73-97. ( 1958) :694-707. MAYR, E. 1969. Principles of systematic zoology. STORER, R. W. 1960b. The classification of birds, McGraw-Hill Book Co., New York. p. 57-93. In A. J. Marshall led.1 Biology and MAYR, E., ANU D. A~UADON. 1951. A classification comparative physiology of hirds. Academic of recent birds. Amer. Mus. Novitates No. 1496: Press, New York. 142. STORER, R. W. 1971. Classification of birds, p. l- MCDOWELL, S. B. 1948. The bony palate of birds. 18. In D. S. Farner and J. R. King [eds.] Avian Part 1. The . Auk 65:520-549. biology. Academic Press, New York. NELSOS, G. J. 1969. The problem of historical STRESEMANN, E. 1959. The status of avian system- biogeography. Syst. Zool. 183243-246. atics and its unsolved problems. Auk 76:269- NELSON, G. J. 1970. Outline of a theory of com- 280. parative biology. Syst. Zool. 19:373-384. TORDOFF, H. B. 1954. A systematic study of the SCHAEFFER, B., M. HECHT, AND N. ELDRIDGE. 1972. avian family Fringillidae based on the structure Phylogeny and paleontology. Evol. Biol. In of the skull. Misc. Publ. Mus. Zool. Univ. Mich- press. igan No. 81:1-41. SCHNELL, G. D. 1970a. A phenetic study of the TORUOFF, H. B., AND J. R. MACDONALD. 1957. A suborder Lari (Aves) 1. Methods and results of new bird (family Cracidae) from the early Olig- principle components analysis. Syst. Zool. 19: ocene of South Dakota. Auk 74:174-184. 35-57. VERHEYEN, R. 1960. A new classification for the SCHPIELL,’ G. D. 197Ob. A phenetic study of the non-passerine birds of the world. Inst. Roy. Sci. suborder Lari (Aves) II. Phenograms, discus- Nat. Belg. 37( 27) : l-36. sion, and conclusions. Syst. Zool. 19:264-302. WETMORE, A. 1930. A systematic classification for SIBLEY, C. G. 1960. The electrophoretic patterns the birds of the world. Proc. U.S. Nat. Mus. of avian egg-white proteins as taxonomic char- 76( 24) : l-8. acters. Ibis 102:215-284. WETMORE, A. 1951. A revised classification for the SIBLEY, C. G. 1967. Proteins: history books of birds of the world. Smithsonian Misc. Coll. 117 evolution. Discovery (Yale Univ.) 3:5-20. (4):1-22. SIBLEY, C. G. 1970. A comparative study of the WETMORE, A. 1960. A classification for the birds egg-white proteins of passerine birds. Peabody of the world. Smithsonian Misc. Coil. 139( 11) : I , Mus. Natur. Hist. Bull. (Yale Univ.) No. 32: l-37. l-131. WILSON, E. 0. 1965. A consistency test of phylog- SIBLEY, C. G., K. W. CORBIN, AND J. H. HAAVIE. enies based on contemporaneous species. Syst. 1969. The relationships of the flamingos as in- Zool. 14:214-220. dicated by the egg-white proteins and hemo- ZUSI, R. L. 1971. Functional anatomy in system- globins. Condor 71: 155179. atics. Taxon 208:75-84. SIMPSON, G. G. 1961. Principles of taxonomy. Columbia Univ. Press, New York. Accepted for publication 31 January 1972.