PHYLOGENETIC PATTERNS AND HYBRIDIZATION1 V. A. FUNK2 ABSTRACT Hybridization is an important part of the evolutionary history of flowering plants. If hybridization has occurred among the species of a taxon under cladistic analysis the results are varied but always present additional difficulties. Hybridization results in incongruent intersecting data that obscure the underlying hierarchy. Guidelines and methods are examined for their usefulness in identifying possible hybrids in a cladistic study. Seven genera are analyzed cladistically and the resulting cladograms examined for possible hybrids. These hypotheses of hybridization are then compared to other data, such as distribution and cytology, to see if the hypotheses of hybridization are supported or rejected. The more hybrids an analysis contains and the more complex the interactions, the more difficult it becomes to identify possible hybrids and their parents. It is difficult to overemphasize the importance of hybridization and polyploidy in evolution be- cause they are outstanding features of many plant groups. According to some authorities 30-80% of the species of angiosperms are polyploids (Goldblatt, 1979; Lewis, 1979; Stebbins, 1974), which allows for the possibility of a tremendous amount of hybridization. Of course, these figures do not include diploid hybrids. Despite its im- portance, hybridization has been virtually ig- nored by those who have dominated the discus- sion of evolutionary theory. This is a result, no doubt, of evolutionary theory being largely in the hands of scientists who work on groups in which such phenomena as polyploidy and hybridiza- tion have a strong relationship with unisexuality and are not considered important in evolution. There are different types of hybridization. Fig- ure 1 summarizes some of the possibilities (a more detailed explanation is found in Funk, 1981) but does not include introgression. For the pur- pose of this paper it is important to note that many hybrids are sexually reproducing individ- uals and are morphologically distinct and in some manner reproductively isolated from both par- ents. Thus, they behave as species no matter whose definition you chose to adopt. THE STUDY OF HYBRIDIZATION The basic concept of phylogenetic systematics (sensu Hennig, 1966) is an ever branching pat- tern or hierarchy. The method of cladistics (phy- logenetic systematics) seeks to discover these patterns by grouping together taxa that share apomorphies (evolutionarily novel, unique, or derived characters). Hybridization, or reticulate evolution, is inconsistent with a method de- signed to depict hierarchies. Hybridization is, therefore, a cause of incongruent, intersecting data that obscure phylogenetic information. Cladists have been concerned with this problem for sev- eral years. Most realize that any method that seeks to identify patterns of relationship must be able to accommodate hybridization because of its frequency. Workers in the problems of hy- bridization and phylogenetic systematics include Bremer (1983), Bremer and Wanntorp (1977), Clark (1982), Funk (1981), Humphries (1983), Humphries and Funk (1984), Nelson (1983), Nelson and Platnick (1980), Rosen (1979), Wag- ner (1969, 1983), Wanntorp (1983), and Wiley and Brooks (1982). One favorite method of botanists in estimating the closeness of relationships among taxa is the percentage of hybridization in crossing studies. An important point about such hybridization studies was made by Rosen (1979: 277): "repro- ductive compatibility is a primitive attribute for the members of a lineage and has, therefore, no power to specify relationships within a genea- logical framework." We cannot use the ability of two or more species to hybridize as an indication of close relationship because the ability is rela- 1 A number of people, who do not necessarily agree with everything that I have said, have kindly provided me with data and comments on various drafts of the manuscript and this paper would not have been possible without their assistance. They include: R. Jansen, J. Semple, C. Clark, R. Sanders, C. Humphries, H.-E. Wanntorp, K. Bremer, G. Nelson, N. Platnick, D. Rosen, and P. Weston. I appreciate the assistance of B. Kahn in helping prepare the illustrations. 2 Department of Botany, National Museum of Natural History, Smithsonian Institution, Washington, DC. 20560. ANN. MISSOURI BOT. GARD. 72: 681-715. 1985. 682 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 72 At]|onnlunloirt| . 7n a: re" Hybridization -GO ASEXUAL I SEXUAL FIGURE 1. Possible hybrid and polyploid relation- ships of two species (A and B) and their reproductive capabilities (Funk, 1981). lively ancestral, possessed at one time by all members of the group. In fact, it is the loss of the ability to hybridize that is apomorphic. Be- cause it is never certain that any two taxa are unable to reproduce, whether or not species hy- bridize is uninformative in determining the pat- tern of relationship. Among cladists, three different approaches have been suggested for dealing with the problem of hybridization (Humphries & Funk, 1984). Some suggest using the most parsimonious cladogram(s) and leaving the homoplasies (char- acters appearing more than once on the clado- gram) resulting from the presence of hybrids as the true reflection of the character pattern. Others have advocated removing the hybrids that have been identified by their 'intermediacy' at the be- ginning of the analysis. The third group advises leaving all of the taxa in the analysis and then closely examining the cladograms for polytomies (nodes with more than two branches) and char- acter conflicts that may indicate possible hybrids. There are problems with all three approaches. The first does not accurately reflect the character pattern; as we shall see, hybrids may appear on the cladogram in a polytomy or with character conflicts or even as ancestors when they are none of these. The patterns do not reflect the accurate sister-group relationship (nor are they true rep- resentations of phylogeny). The second approach assumes one can identify the hybrids prior to the analysis and this is not possible in many cases. The third position relies to a great extent on hy- brids causing polytomies. Further analysis has shown this usually does not happen (Humphries, 1983). Wagner (1983) has suggested a method for dealing with hybrids that he calls reticulistics. This method works with groups in which hybrids are intermediate in character and progressively less well with those that are not intermediate. Often the hybrids to which Wagner refers are F,'s that are being formed continually and are sex- ually in viable. Certainly for well-defined plant groups in which the hybrids are intermediate and obvious and are characterized by being rare and either sterile or polyploids (definition, Wagner, 1983: 71), Wagner's method should be consid- ered. These individuals are not units of evolution and thus are not species. In this paper I am con- cerned with hybrids that are regarded as evolu- tionary units and that are usually designated as species, subspecies, or varieties. Theoretically, in cladograms, hybrids show up by causing character conflicts; so also do homo- plasies (Fig. 2, character 4; Appendix A). One must be diligent in trying to distinguish between character conflicts caused by hybridization and those that are the result of parallel or convergent evolution. It is advisable to use the cladogram (developed with the concept of parsimony using all taxa) to examine the apomorphies that appear more than once on the cladogram (homoplasies). Perhaps a closer examination will reveal that some characters originally thought to be apo- morphies are actually different structures (false homologies) or are combinations of characters. For instance, not all black anthers in the Com- positae genus Montanoa Cerv. are homologous. Although originally treated as one apomorphy (Funk, 1982) a close examination showed that some black anthers were black only around the edges of the thecae, some were black only on the top of the connective, while others were com- pletely black. This additional information indi- cated that "black anthers" was not a single apo- morphy but three apomorphies. Character conflicts can also be the result of a designated apomorphy actually containing several charac- ters. Characters such as habit, chromosome number, and flower regularity, all can be divided into several characters. New apomorphies can be added to the cladistic study to reflect the addi- tional information because the "groups of char- acters" should not be viewed as character con- flicts but rather as separate apomorphies. Such changes should be made only when available evi- dence indicates that they are not homologous. Remaining character conflicts are the result of either undetected homoplasy or hybridization. 1985] FUNK-HYBRIDIZATION 683 FIGURE 2. Cladogram illustrating character conflict that can be the result of either hybridization or parallel evolution. A lack of apomorphies can also be caused by hybridity. The hybrid does not necessarily in- herit all of the apomorphies of both parents. This observation is important: there is no reason to assume that all apomorphies are dominant over the more plesiomorphic characters of a trans- formation series. The data presented later in this paper indicate that in the heterozygous condition of the hybrid there might well be a higher per- centage of plesiomorphic characters showing in the phenotype. Therefore, it is possible for the hybrid to show only plesiomorphies (primitive, ancestral, or more general characters) and appear on the cladogram in an ancestral position. In Figure 2, taxon D could be a hybrid, between taxon C and any other taxon, that inherited the plesiomorphies of both parents. Indeed, it is in- teresting to speculate on whether or not one could use such cladistic studies to identify interesting genetic problems. Often the hybrids in a cladistic analysis will be grouped with one or the other of the parents depending on with which parent they share the most apomorphies (Humphries & Funk, 1984). When the putative parents are sister species (two species that are more closely related to one another than they are to any other species), hy- FIGURES 3-6. Cladograms illustrating the pattern of species A and B and their hybrid, H. 684 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 72 FIGURE 7. Cladogram illustrating the pattern of species A and B and their hybrid, H. brids are quite often apparent regardless of whether or not they form polytomies. Hybrids that form dichotomous branching patterns can be identified as such so long as they have at least one apomorphy of both parents or if they lack an autapomorphy of the parent with which they are grouped. For instance in Figures 3-5 (for characters see Appendix B), species A and B hy- bridize to give species H. If there were an equal number of apomorphies in A and B (characters 4 and 5) and if both were found in the hybrid, the result could be expressed as two equally par- simonious cladograms (Figs. 3, 4) or as a tricho- tomy (Fig. 5), although the latter involves one more character change and is therefore one step longer. If the incongruent character set is inferred to be the result of hybridization, the hybrid could be placed above the diagram connecting with both parents (Fig. 6). However, if one parent taxon had one more autapomorphy than the oth- er, or if the hybrid showed unequal character inheritance, then a single cladogram results. For instance, if taxon A (Fig. 7; Appendix C) had two autapomorphies (5 and 6), and the hybrid in- herited all apomorphies of both parents, the most parsimonious cladogram would give the result shown in Figure 7. Nonetheless, because they are sister taxa, the possibility of hybridization is ap- parent, so long as the hybrid has at least one apomorphy from each parent. However, as men- tioned earlier it is possible for the hybrid not to display all the apomorphies. If the hybrid in Fig- ure 7 did not have character 4 there would be no indication that it was a hybrid (except for the more indirect evidence of lack of apomorphies in taxa A and H). If one parent has very few autapomorphies there is less chance that the hy- brid will have any indication of its history. Sometimes there is more than one hybrid from the same two parents (Figs. 8-12; Appendix D). The two hybrids are most parsimoniously grouped with either parent (Figs. 8, 9) but the two equally parsimonious cladograms indicate the hybridity of HI and H2. The most parsi- monious cladogram has two reticulations (Fig. 12). It is, of course, possible that taxa HI and H2 are the result of a single hybridization event followed by segregation. One way to evaluate this possibility is to examine the distribution of the taxa in question. The possibility of hybridization followed by segregation lessens as the distance between the hybrid taxa increases. The only time it is "most parsimonious" to have a polytomy is when the hybrid does not have any of the autapomorphies of either parent (Fig. 13; Appendix E). However, parent taxa are not always sister taxa. For instance, Wagner (1954) has shown that at least three diploids are involved in producing the hybrids in Asplenium, and Grant (1953, 1964) has shown that species from different species groups are hybridizing in Gilia (Funk, 1981). In cases such as this the task of identifying hybrids becomes more difficult. For instance, given the cladogram in Figure 14 (Figs. 14-17, 19; Ap- pendix F) the most parsimonious cladogram that includes the hybrid, places the hybrid (H, Fig. 15) as the sister taxon of the parent that involved the least number of homoplasious events (the number of character changes or length of this cladogram is L = 11). The length would increase if the hybrid were grouped with the other parent (Fig. 16, L = 12) because there is one more ho- moplasy. It is much longer to form a polytomy (Fig. 17, L = 14). The only time a polytomy would be formed in the most parsimonious cladogram is if the hybrid had none of the apo- morphies of either parent, at least those above the first node shared by both parents (Fig. 18; Appendix G). If the hybrid is identified as such it can be removed from the cladogram and placed above it giving an even shorter cladogram (Fig. 19, L = 9). The identification of possible hybrids is only the beginning of an analysis. Those cladograms indicating hybrids (e.g., Fig. 19) are merely hy- potheses of hybridization and should be tested by using other data, such as distribution, chro- 1985] FUNK- HYBRIDIZATION 685 FIGURES 8-12. Cladograms illustrating some of the results when the same two parent species (A and B) produce more than one hybrid, HI and H2. mosome number, karyotyping, and pollen fer- gram without a reticulation and continues by tility before they can be referred to as hybrids. adding reticulations one at a time so as to min- Nelson (1983) has suggested a method for ana- imize character conflict. It is based on the idea lyzing cladograms for possible hybrids. His pro- that when there are two equally parsimonious cedure begins with the most parsimonious dado- ways of representing a homology on a cladogram 686 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 72 1 3 FIGURE 13. Cladogram illustrating that it is more parsimonious to have a polytomy when the hybrid does not show any of the autapomorphies of either parent. one should investigate the possibility of inserting a reticulation. If the reticulation results in a de- crease of apparent homoplasy and if the taxon exhibits character conflict of the "intermediate" type, the reticulation can be maintained. For a certain set of data (Appendix H) there are two equally parsimonious cladograms, with the same branching pattern. One cladogram (Fig. 20) has a parallel acquisition of character 4 (Figs. 20-22, Appendix H) and the other has one acquisition of character 4 and a subsequent loss (Fig. 21). However, one can introduce one reticulation and eliminate the need for homoplasy and/or loss (Fig. 22). All this diagram indicates is that if hybridization is involved it is most likely that taxon B is of hybrid origin. A more complicated example involves nine taxa and 12 characters. There are two equally parsimonious cladograms with the same branch- ing pattern with different amounts of homoplasy and loss (Figs. 23, 24; for characters for Figs. 23- 25 see Appendix I). By progressively adding re- ticulations, all need of reversal and/or homoplasy can be eliminated. Taxa H and I may be hybrids (Fig. 25). Nelson's method of examination of character conflict is only a beginning. Microloma illustrates the use of the Nelson method to change the cladogram in Figure 26 (for characters for Figs. 26, 27 see Appendix J) to the one in Figure 27 by adding one reticulation and thereby shortening the cladogram. Some groups have characteristics that cause difficulties when using Nelson's method. For in- stance, some have their origin in hybridization but have since speciated (developed autapomor- phies), some have numerous hybrids and even hybridization among hybrids, and some hybrids do not have all of the apomorphies of the parents (as in Fig. 2). The workability of Nelson's meth- od is dependent on the hybrid inheriting the apo- morphic characters from the parent taxa without too many character losses; otherwise the clado- gram with the reticulations (Fig. 28; for char- acters for Figs. 28, 29 see Appendix K) will be longer than the most parsimonious cladogram without reticulations (Fig. 29). Also, this method is only feasible when the percentage of hybrids in the data set is low because the possibilities become endless, especially when the hybrids are hybridizing. There are additional guidelines and methods that can be used when examining cladograms for possible hybrids. Use of these on seven data sets indicates that insights into the identification of possible hybrids and their parents can be gained by studying the character patterns of the clado- grams as well as the distributions and ploidy levels of the taxa involved. Some of these guidelines and methods are elaborations and evaluations of previously published ideas and others are new. GUIDELINES AND METHODS FOR IDENTIFYING POSSIBLE HYBRIDS AND THEIR PARENTS 1. When there are two cladograms of similar length and one taxon position changes, the taxon that is moving may be a hybrid and the two taxa between which it is moving may be the parents. In Figures 3 and 4, taxon H (the hybrid) shifts between taxa A and B in the two most parsi- monious cladograms. Taxon H may be a hybrid and A and B may be its parents. In Figures 15 and 16 taxon H shifts between taxa C and D and may be a hybrid. 2. As an extension of number 1, it is possible to follow a path of character conflicts. Figure 15 has characters 6 and 7 appearing twice and this identifies taxon D, the parent with which H (the hybrid) is not grouped. Figure 16 has 1, 3, and 5 appearing twice and this identifies taxon C, the parent with which H is not grouped. You do not have to have all of the characters. For instance, in Figure 16, H might not have character 3 (Fig. 30; for characters for Figs. 30, 31 see Appendix L) but as long as it had 1 and 5 taxon C would emerge as the other parent (Fig. 31). 3. Taxa that are defined solely by character conflicts may be hybrids or parents. In Figure 1985] FUNK-HYBRIDIZATION 687 14 15 16 18 1 9 FIGURES 14-19.?14-17, 19. Cladograms illustrating the pattern of species C and D and their hybrid, H. 18. Cladogram illustrating when a polytomy is formed in the most parsimonious cladogram. 15, taxa H (the hybrid) and D are defined only by characters that appear elsewhere on the clado- gram and in Figure 16, taxa C and H also have only homoplasies. The same is true for taxa C and B in Figure 20, taxa A and I in Figure 23, and taxa B and C in Figure 26. 4. Taxa with reversals may be hybrids. In Fig- ure 32 (for characters for Figs. 32, 33 see Ap- pendix M) taxon H (the hybrid) has not inherited all of the apomorphies of both parents and there- fore is defined not only by characters that appear more than once on the cladogram but also by 688 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 72 FIGURES 20-22. Cladograms illustrating a simple example of the Nelson (1983) method for analyzing clado- grams for possible hybrids. ABCDIEFHG 23 4 / 25 FIGURES 23-25. Cladograms illustrating an example of the Nelson (1983) method for analyzing cladograms for possible hybrids. 1985] FUNK-HYBRIDIZATION 689 26 27 FIGURES 26, 27. Cladograms illustrating the use of Nelson's method for analyzing cladograms for possible hybrids in the genus Microloma. character loss. Taxon H does not have characters 3 and 8 from taxon C and is also missing char- acter 10 from taxon D. In fact, there has been enough inheritance of plesiomorphies to make the cladogram with reticulations (Fig. 33) the same length as the most parsimonious cladogram (Fig. 32). Another possibility is when there is one parent species rich in apomorphies and another lacking them altogether?hybrids might be in- termediate or they might evidence multiple loss. 5. Taxa without autapomorphies may be par- ents. If the hybrid inherits all of the apomorphies of the parent with which it is grouped, then the parent will have no autapomorphies (Fig. 2, tax- on A; Fig. 7, taxon A; Fig. 15, taxon C; Fig. 16, taxon D). If the hybrid fails to inherit any of the apomorphies of the parent with which it is not grouped, the hybrid will have no autapomor- phies (Fig. 13, taxon H)?normally in evidence as homoplasies. 6. Consensus Trees?Consensus analysis is developing rapidly as an aid in evaluating a num- ber of equally or nearly equally parsimonious cladograms. Consensus trees represent the in- formation shared by two or more cladograms. The consensus tree is a compromise classifica- A B /^C 0 E a\ T~l / / 28 29 FIGURES 28, 29. Cladograms illustrating that the cladograms with reticulations are not necessarily the most parsimonious. 690 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 72 30 31 FIGURES 30, 31. Cladograms illustrating that possible parents can be identified even if all apomorphies of both parents are not present in the hybrid. tion. Consensus trees were first developed by Adams (1972) and have been used in the context of comparing cladistic versus phenetic methods (Mickevich, 1978; Schuh & Polhemus, 1980; Sokal & Rohlf, 1981) or to compare a clado- gram constructed from a chemical data set and an intuitive tree (Seaman & Funk, 1983). There are many different types of consensus analysis, Adams consensus (Adams, 1972), strict consen- sus (Sokal & Rohlf, 1981), majority consensus (Margush & McMorris, 1981), and durchschnitt consensus (Neuman, 1983). Only Adams con- sensus trees and Nelson component analysis will be discussed in this paper because of ease of use and explanation. For further investigation, other references include Mickevich (1978), McMorris et al. (1983), and a series of papers in Felsenstein (1983). An Adams consensus tree (ACT) may have a topology different from any of the cladograms used to construct them. Figures 15 and 16 have a different topology, but if we concentrate on agreements we get Figure 14. Then we can add the taxon left out to the first node common to both of its positions (Fig. 17). A detailed dis- cussion of how to construct an ACT is found in Adams (1972). 7. Component Analysis (NCA) was devel- oped by Nelson (1979) as a consensus method. A component is any monophyletic group on a cladogram or phylogenetic tree. Every cladogram can be divided into a certain number of com- ponents that have more than one terminal taxon. In order to construct a Nelson Consensus Tree it is possible to add together components that are common to two or more cladograms. The 32 FIGURES 32, 33. Cladograms illustrating that character loss may be an indication of hybridization. 1985] FUNK-HYBRIDIZATION 691 34 35 FIGURES 34, 35. Cladograms illustrating that by comparing the components (Table 1) one can identify the source of incompatibility. movement of one taxon from one side of the cladogram to the other can have the effect of changing all of the components (Figs. 34, 35, Table 1) and no consensus tree can be construct- ed. However, instead of searching for complete agreement among the components, we can ex- amine in what way they are different and identify which taxa are responsible for the lack of con- gruence. Examining the list of components for the two figures (Figs. 34,35, Table 1) it is obvious that taxon H is causing the incompatibility. These various suggestions are not to be used individually but collectively to generate hypoth- eses of hybridization. We can then turn to other forms of data to corroborate or falsify our hy- potheses. Two of the most readily available in many groups are distribution and polyploidy, but others such as karyotyping and pollen fertility can be employed. We will see how such infor- mation can be used in the examples below. All cladograms were constructed manually. EXAMPLE 1. MICROLOMA (ASCLEPIADACEAE) BREMMER AND WANNTORP (1979) Microloma is an African genus of 19 species, nine of which are represented in the most par- simonious cladogram (Fig. 36, Appendix J) and an alternative cladogram that is three steps long- er (Fig. 37). Figure 37 shows taxon C sharing an apomorphy with taxon B and Figure 36 shows taxon C sharing four apomorphies with taxon D indicating that B and D may be the parents if C is a hybrid. Also, there is a path of parallel char- acters to follow to the parents especially in Figure 37. Furthermore, taxon D (Fig. 36) and taxon B (Fig. 37) have no autapomorphies, indicating they might be the parents with which the hybrid is grouped. In addition, an ACT (Fig. 38) can be constructed that identifies the hybrid by placing it at the first node shared by the two parents. A NCA (Table 2) shows that C is responsible for the incompatibility between the two sets of com- ponents. EXAMPLE 2. ANACYCLUS (ASTERACEAE) HUMPHRIES (1979, 1981) Anacyclus (Appendix N) is a genus of Medi- terranean distribution with 14 species (Figs. 39- 42). The monograph and subsequent paper by Humphries (1979, 1981) included a cladistic analysis and speculations on the hybrid origin for three of the species. Figure 39 is the most parsimonious cladogram (L = 45) for the data furnished by Humphries (1983); the number in parentheses next to the lower case letter indicates the number of apomorphies that have that pat- tern of distribution. If we concentrate on the character types that conflict with the most par- simonious cladogram (p, b, f, q, 1) we can draw alternative cladograms that are slightly longer (Fig. 40, L = 50; Fig. 41, L = 52). Figure 40 has four character types in conflict (k, p, q, o) and a TABLE 1. Components for Figures 34, 35. Figure 34 Figure 35 AH AHB AHBC EF DEF AB ABC HF EHF DEHF 692 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 72 A B 37 FIGURES 36-38. Two cladograms of Microloma and their Adams consensus tree. gain and subsequent loss of two character types (c, d). Figure 41 also has four types of conflict (o, b, c, q) and two types of loss (c, d). The con- sensus tree for these three cladograms is illus- trated in Figure 42 and shows that four taxa are potential hybrids (N, C, I, J) because they have dropped in resolution and because they are de- fined either completely or primarily by charac- ters that are found elsewhere on the cladogram. In addition by listing the components of the three figures (Table 3) it is evident that all of the com- TABLE 2. Components for Figures 36, 37. Figure 36 Figure 37 CD BC BCD ED FGECD FGED HIFGECD HIFGED BHIFGECD BCHIFGED ABHIFGECD ABCHIFGED ponent incompatibility is caused by taxa I, J, C, and N. These four taxa are potential hybrids (Fig. 43) and their possible parents are as follows: possible possible hybrids parents N AB x EF-LM I G x H J G x H C AB x D-LM Figure 43 is 42 steps long. Based on his analyses Humphries (1979, 1981) suggested three puta- tive hybrids (N, I, J) and indicated corroboration for the hypothesis of hybridization for these three species. According to Humphries in addition to sharing apomorphies with AB and EF-LM, A. officinarum (N) is known only as an extinct cul- tivar grown in the nineteenth century for phar- maceutical reasons. Anacyclus valentinus (I) and A. inconstans (J) have florets of intermediate length between A. clavatus (G) and A. homoga- mos (H). Also, A. valentinus (I) is a weedy taxon 1985] FUNK- HYBRIDIZATION 693 m(3) n(5) -a(5) FIGURES 39-42. Cladograms of Anacyclus. occurring only on disturbed land in the southwest Mediterranean region, and A. inconstans (J) oc- curs sympatrically with the Algerian population of A. clavatus (G). Cytogenetic studies carried out by Humphries (1981) corroborated the hy- potheses of hybridity for A. valentinus (I) and A. inconstans (J). There was no material for cyto- genetic studies available for either W. officinarum (N) or A. monanthos (C). There is, then, good reason to list both A. valentinus (I) and A. in- constans (J) as possible hybrids because the hy- potheses of hybridity have been supported by independent data. Anacyclus officinarum (N) has been supported as a hybrid by its cultivated na- ture but is more difficult to analyze because it is not known to be extant. Anacyclus monanthos (C) was not indicated by Humphries to be of hybrid origin. In his monograph, Humphries TABLE 3. Components for Figures 39-41. Figure 39 Figure 40 Figure 41 Conflicts IJ IJ IJ IJH GJJ GU IJ GIJH GIJH GIJH IJ GIJHK GIJHK GIJHK GIJHKLM GIJHKLM GIJHKLM EFGIJHKLM EFGIJHKLM EFGIJHKLM DEFGIJHKLM NEFGIJHKLM NEFGIJHKLM N CDEFGIJHKLM DNEFGIJHKLM DNEFGIJHKLM C, N ABN CDNEFGIJHKLM ABC C, N ABNCDEFGIJHKLM ABCDNEFGIJHKLM ABCDNEFGIJHKLM 694 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 72 H K LM -a(5) FIGURE 43. Cladogram of Anacyclus with reticulations. (1979) listed it as a pioneer of sandy soil and in some areas a dominant weed; these are charac- teristics of some hybrids. He also mentioned that it resembles one of the two subspecies of A ho- mogamos (H). However, Humphries thinks that A. monanthos (C) is a taxon exhibiting inter- mediate characters and not a hybrid because it has many autapomorphies (Humphries, pers. comm.). Because of the lack of additional infor- mation on this taxon I have left it as a dotted line on the cladogram (Fig. 43). The cladogram in Figure 43 differs from that of Humphries because I emphasized reducing the number of parallel characters and was not as concerned if this created additional character losses. Therefore, taxon N is connected as an internode further along the diagram. This is im- portant because it is, in my opinion, unlikely that a hybrid will inherit all of the apomorphies of both parents. Thus, some plesiomorphies will be inherited resulting in the loss of some characters. The apomorphies that are inherited are the guide to the identification of possible parents. Anacyclus is a good example of the use of cla- distics to identify possible hybrids and then em- ploying other techniques including karyotyping and chemical analysis, distribution, and habitats to further corroborate or falsify the hypotheses of hybridization (Humphries, 1979, 1981). EXAMPLE 3. AGAST ACHE (LAMIACEAE) SANDERS (1982, AND IN PREP.) The 14 species of Agastache Cl&yX. sect. Brit- tonastrum (Appendix O) are all diploid and are confined to the Cordilleras of the southwest United 1985] micrantha comp. mexicana comp FUNK-HYBRIDIZATION pallidiflora comp. 695 cana comp. 28(1') 5(2) * 25, 26, 27 % 24 $ 11(2) FIGURES 44, 45. Cladograms of Agastache. States and Northwest Mexico. Looking at the parsimony cladogram (Figs. 44,45) there are taxa with no autapomorphies (A. pallidijlora var. neomexicana ne-n, A. pallidijlora var. greenei pf-r, A. mexicana mex), taxa with only character conflicts (A. breviflora brv, A. mearnsii mm, A. coccinea coc, A. pringlei prn), and taxa with character losses {A. pallida var. coriacea pd-c, A. pallidijlora var. gilensis pf-i). Three of these taxa (prn, brv, pd-c) are immediately identifiable as possible hybrids, A. pringlei and A. breviflora because they have four character conflicts each and A. pallida var. coriacea because it has two character losses. Less strongly indicated are A. pallidijlora var. gilensis (because of the absence of character 24 and some connection with A. coccinea) and A. mearnsii [because of character 21(1')]. Examining each one individually, A. breviflora has characters 25, 26, and 27 (and sometimes 11). If A. breviflora is a hybrid then one parent (A. mearnsii) could be its sister species. In order to identify a candidate for the second parent, emphasize the characters^, breviflora has and not the characters it does not have. Then locate where else on the cladogram these char- acters are found (25, 26, and 27; Fig. 44). The presence of these three characters indicates A. wrightii (wrt) as the other possible parent. Com- paring the geographical distributions (Sanders, pers. comm.), A. breviflora and A. wrightii are sympatric in southern Arizona, southwest New Mexico, and northern Sonora and Chihuahua. Agastache mearnsii is found a short distance from A. wrightii in southern Sonora and Chihuahua. The distribution pattern would suggest a hybrid- ization event is possible. A second possible hybrid, A. pringlei (prn) is sister to A. micrantha (one subspecies of which has no autapomorphies). Agastache micrantha has four character conflicts (13, 16, 21, and sometimes 11, see Appendix O). Locating the apomorphies on the cladogram identifies the "Pallidiflora complex," which contains A. brev- iflora, A. mearnsii, and the five subspecific taxa of A. pallidijlora. Agastache pringlei is some- times sympatric with A. micrantha (southern 696 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 72 micrantha comp. mexicana comp. brv pallidif lor a comp. cana comp. A ' -^13 / wrtv'mex epl pirn / pdp pdcxmrn \ pf-r / : //\\ / / / / /\ \\ i \ jJ pf-i 4-13(1') 28(1') pfp ne-h ne-n aur can rup 11(2) 5(2") 5(2') 2 FIGURE 46. Cladogram of Agastache with reticulations. Chihuahua and southern New Mexico) and para- pa trie with A. mearnsii (southern Chihuahua and the Chihuahua/Sonora border). The A. pallidi- flora subspecific taxa and A. breviflora are dis- tributed in the southwest United States. If A. pringlei is a hybrid the most likely putative par- ents are A. micrantha and A. mearnsii. A third potential hybrid, A. pallida var. co- riacea (pd-c), is sister to A. pallida var. pallida and has two character losses. If A. pallida var. coriacea is a hybrid and one parent is A. pallida var. pallida then the other parent is probably something that lacks characters 10 and 11; pos- sibly something in the "Pallidiflora complex" be- cause the putative hybrid has characters 16 and 21. Based on geographical distribution, the most likely parent in the "Pallidiflora complex" is A. mearnsii that has a parapatric distribution with A. pallida var. coriacea. The last two possible hybrids, A. coccinea (coc) and A. pallidiflora var. gilensis (pf-i), are more difficult to place because they have less infor- mation (fewer characters). We can estimate that A. coccinea is a hybrid and that one parent may be found in the group defined by character 32; because of the presence of character 21(1') the other parent would probably be A. mearnsii. The distribution of the species does not help in this case because several of the taxa are sympatric. The final taxon considered as a hybrid is A. pallidiflora var. gilensis. This taxon might be the result of a cross between A. pallidiflora var. greenii and some taxon without character 24. All taxa in the "Pallidiflora complex" that have a distri- bution that would allow for the formation of this hybrid have character 24 except A. mearnsii. However, A. pallidiflora var. gilensis has an aut- apomorphy, and this makes it less likely that it is of hybrid origin (but does not exclude it). Based on this analysis the following hybrids are pos- 1985] FUNK- HYBRIDIZATION 697 x=3 x=4 x-9 gdv lan lat gs-gd sea BR pit mar gs-gg gs-c gs-h sub gs-g1 ill li-d FIGURES 47-49. Cladograms of Chrysopsis and Bradburia. sible: brv = wrt x mm, prn = mic x "Pallidiflora complex," pd-c = pd-d x mm, coc = mrn x mex-plm, and pf-i = pf-r x some taxon without 24 (possibly mrn). Figure 46 shows the hypothesized hybrids and their possible parents. According to Sanders (1981), the hybrids and their parents are as follows: brv = wrt x mrn, prn = mic x mrn, pd-c = pd-p x mrn, coc = pfr x mrn (mex-epl + mrn), and pf-i = pf-r x mrn (pf-f ? pf-r + mrn). Although Sanders' estimates are in some cases more specific, there are no conflicts. EXAMPLE 4. CHRYSOPSIS AND BRADBURIA (ASTERACEAE) SEMPLE (1981, AND PERS. COMM.), SEMPLE AND CHINNAPPA (1984) Chrysopsis (Nutt.) Elliot (Appendix P; 10 species) and Bradburia Semple & Chinnappa (Appendix P; 1 species) are yellow-rayed golden- asters distributed in the southeast United States (especially Florida) except for one species of Chrysopsis that occurs in the eastern United States. Using the characters furnished by Semple three cladograms were constructed (Figs. 47-49). There is a high level of homoplasy in the clado- gram because only nine apomorphies lacked character conflicts and only four of these 11 are synapomorphies (10, 20, 11, 2A). Such a high level of character conflict and character loss is an indication of possible hybridization. The chromosome numbers were not used as char- acters in the analysis and are indicated on the cladogram to facilitate the discussion. Also, there is one report of x = 4 for Bradburia that is not indicated on the cladogram. Taxa C. lanuginosa (lan), C. gossypina subsp. gossypina f. gossypina (gs-gg), C. gossypina subsp. hyssopifolia (gs-h), and C. linearifolia subsp. li- nearifolia (li-1) are defined only by character con- flicts and/or character losses and may be hybrids themselves or the parent with which another hy- brid is not grouped. Two taxa C. godfreyi f. viri- dis (gd-v) and C. gossypina subsp. cruiseana (gs-c) are possible parents with which the hybrids are grouped because they have no autapomorphies 698 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 72 FIGURE 50. Cladogram of Chrysopsis and Bradburia with reticulations. and have a single taxon as a sister taxa. If we construct hypotheses based on these data we ob- tain the following results: Ian = gd-g x li-1, gs-c & h = sub x pil or mar, and gs-gg = sub x some taxon without 8. Because so many of the taxa are sympatric or parapatric, the distributions are not of much assistance in refining the hypotheses. The exceptions are C. gossypina subspp. cru- iseana and hyssopifolia (gs-c & h) that can be attributed to a cross between C. subulata and C. mariana (sub x mar). There are four taxa with no autapomorphies [gd-g (C. godfreyi f. godfreyi), gs-gd (C gossypina subsp. gossypina f. decumbens), gs-gt, li-d)] that are potential hybrids or parents. Their chro- mosome numbers show that all of the gs taxa are x = 9, and based on outgroup comparison the base number for the genera is probably x = 5 (this agrees with Semple, 1981). Therefore, the subspecific taxa of C. gossypina are most likely hybrids. This supports the hypotheses of hybrid- ization for three of the subspecific taxa of C. gossypina (gs-gg, gs-c & h), however C. gossypina f. trichophylla and f. decumbens (gs-gt, gs-gd) were not identified, except for noting that they lacked autapomorphies. The cladogram clearly indicates why C. gossypina f. decumbens was overlooked, it has no synapomorphies or autapo- morphies. If it is a hybrid it is an excellent ex- ample of a hybrid inheriting all plesiomorphies of both parents and appearing in an ancestral position on the cladogram. Likewise, C. gossy- pina f. trichophylla was overlooked because it inherited most, but not all, of the plesiomor- phies. The chromosome number of x ? 5 does not support, and in fact falsifies, the hypothesis of hybridization for C. lanuginosa because one of the hypothesized parents has x = 5 and the other x = 9. Figures 47-49 show that the x = 9 taxa do not 1985] FUNK- HYBRIDIZATION 699 californica group frutescens group FIGURE 51. Cladogram of Encelia. form a monophyletic group. Semple (1981) pro- posed that the x = 9 group is from one hybrid- ization event between an individual of C. su- bulata and one of C. mariana with subsequent selections to give different combinations of pa- rental genes. The cladogram does not support that statement because of the different combi- nations of characters in the five different sub- specific taxa. However, it also does not reject Semple's suggestion. It simply suggests that one should also consider the possibility of several different hybridization events involving the same species as parents with different characters being inherited each time (Fig. 50). Using cladistics alone has not given us a clear answer to the question of hybridization in Chry- sopsis and Bradburia, however when the hy- potheses of hybridization are tested with addi- tional information, such as distribution and ploidy level, we have been able to make five putative hybrids and gain some insight into pos- sible parents. EXAMPLE 5. ENCELIA (ASTERACEAE) (CLARK, PERS. COMM.) Encelia Adanson (Appendix Q) comprises 18 taxa distributed in the western United States. All taxa are diploids. The data matrix and the taxon distributions were furnished by Curtis Clark. There are at least two cladograms that are equally parsimonious (Figs. 51, 52) and one that is one step longer (Fig. 53); all have different branching patterns. There are several additional cladograms that have slightly different character distributions but do not have different branching patterns, and these are not illustrated. On the three cladograms there are several taxa that change positions. Encelia farinosa can be placed as the sister taxa of E. farinosa var. phenicodonta (Figs. 52, 53) or near the base of the cladogram (Fig. 51). In either case it is only defined by char- acter loss and/or homoplasy. Likewise E. asperi- folia is either the sister taxon of E. ventorum/E. laciniata (Fig. 51) or it is near the basal node (Figs. 52, 53). In all three positions E. asperifolia 700 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 72 californica group frutescens group FIGURE 52. Cladogram of Encelia. is identified by character loss and/or homoplasy. Encelia canescens is either the sister taxon of E. palmeri (Fig. 51) or it shares the node defined by character 7(1) with several other taxa (Figs. 52, 53). Two other taxa have some indication that they may be hybrids, E. laciniata with in- termediate character 10(1) and the unidentified taxon from Santa Carla (SC) that is always de- fined by character homoplasy. There are three taxa that can be indicated as possibilities but cannot be placed as hybrids, ancestors, or parents because they have few apomorphies. Two taxa (E. virginensis and E. actoni) have only one apo- morphy each and they appear on the cladogram at the basal node, and E. californica has a slight change in position depending on whether char- acter 3 is treated as a character loss or not (Figs. 51-53). Encelia farinosa and E. asperifolia stand out because of character losses and homoplasy. The best estimate for E. farinosa is that it is a hybrid between E. farinosa var. phenicodonta (its sister taxon in Figs. 52, 53) and something without characters 1, 2, and 3. Encelia asperifolia may be a hybrid between something without char- acters 3 and 12 (perhaps something in the "Fru- tescens group") and something in the "Califor- nica group" with character 8 (perhaps E. californica with which it is closely grouped and whose lack of apomorphies may account for E. asperifolia's similar situation). Encelia canes- cens might be a hybrid, between E. palmeri its sister taxon in Figure 51 and E. farinosa var. phenicodonta, because of the intermediate na- ture of character 14. Encelia laciniata may be a hybrid between E. ventorum and some other tax- on that has not left a trace. If so, E. laciniata is an example of a hybrid inheriting all of the apo- morphies of one parent. If the taxon from Santa Clara (SC) is a hybrid, one parent might be E. ventorum because of characters 13 and 16 and the other parent something from the "Californica group" that has character 12. A summation of possible hybrids is as follows: far = phe x some- 1985] FUNK - HYBRIDIZATION 701 califomica group frutescens group FIGURE 53. Cladogram of Encelia. thing without 1,2, and 3, asp = cal x something in the "Frutescens group" with 8, can = pal x phe, lac = ven x ?, and SC = ven x "Frutescens group" with 12. Listed above are five hypotheses of hybridiza- tion and some possible parents. Examination of the distribution patterns and other data do not support two of the hypotheses (E. farinosa and SC) and an additional one is added. An F, that has been identified as E. virginensis has been found growing with E. actoni so that E. virgi- nensis should be investigated as a possible hybrid with E. actoni as one of the parents. Using the distributions we can narrow down the choice of possible parents to the following: asp = cal x SF, can = phe x pal, lac = ven x ?, and vir = act x ?. The reticulate cladogram is illustrated in Fig- ure 54. The hybrids indicated with solid lines are those that were supported as hybrids by both the cladistic analysis and the distributional data; those with dotted lines were supported by only one of the two. EXAMPLE 6. MONT ANOA ( ASTER ACEAE) (FUNK, 1982) Montanoa Cerv. (Appendix R) has 20 species in Mexico and Central America and five in northern South America. Examining one of the equally parsimonious cladograms (Fig. 55), only one taxon (M. hexagona) shows any strong in- dication of hybridization. This taxon has the only two character losses on the cladogram. There are however, three known high level polyploids in the genus (Fig. 55). Two of these (M. revealii and M. guatemalensis) show no evidence of being hybrids. The third is M. hexagona and, it could be a hybrid between its sister taxon, M. hibis- cifolia, and something outside of the group de- fined by character 34. None of the polyploids are sympatric with any other species and they all have at least 95% pollen viability, and during meiosis there is at least one stage where a com- plete bivalent can be observed. So, we are left with these three species being either very old polyploids, two of which have developed autapo- 702 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 72 FIGURE 54. Cladogram of Encelia with reticulations. morphies, or the parents are extinct so the rela- tionships do not show up on the diagram. Or, less likely, they are autopolyploids with the dip- loids no longer extant. The cladogram cannot help us in resolving this matter. EXAMPLE 7. ACMELLA (ASTERACEAE) (JANSEN, IN PRESS) This last example presents the most difficult case: one where there are more hybrids than non- hybrids in a genus, where the hybrids are hy- bridizing, where there are few characters in the analysis and some of the hybrids have inherited mostly plesiomorphies, and where the hybrids are weeds that disperse readily and tend to hy- bridize wherever they are. Although not the rule, such situations are not that unusual in the As- teraceae family. One such genus is Acmella (Ap- pendix S). The species of Acmella used to be part of Spilanthes Jacq. but were removed by Jansen (1981). There are 39 taxa (30 species), 16 of which are diploids (23 polyploids; ploidy level is esti- mated in some species, see Appendix S). The genus is pantropical with one species in the southeast United States. Some species form au- topolyploid series that would allow them to cross with allopolyploids. Also, some reproduce asex- ually so odd level polyploids persist in nature. There are at least ten equally parsimonious cladograms of Acmella and a large number (over 100) that are only a few steps longer. Many of these cladograms have very different structures. I have selected one to discuss as a representative (Fig. 56), but in no way am I indicating that this particular cladogram is to be preferred over any other. In Figure 56 there are only three apo- morphies (excluding autapomorphies) that are not either subsequently lost or found elsewhere on the cladogram [7, 1, 2(2)]. The three major 1985] FUNK- HYBRIDIZATION 703 23.24,25,28 28 FIGURE 55. Cladogram of Montanoa, numbers above taxa indicate ploidy level of known polyploids. groups of Jansen (in press; indicated in Fig. 56 by the large numbers 1, 2, and 3) are obvious on the cladogram and only one, number 2, is non- monophyletic (this group was non-monophyletic on all of the cladograms that I constructed). Some taxa may be hybrids between the three major groups, some of the more obvious ones are as follows: 1. A. decumbens var. decumbens (23a) may be a hybrid between some taxon in group 1 and one in group 2 because it has apomorphies 1 and 2. 2. A. poliolepidica (1) may be a hybrid be- tween a taxon in group 1 that has apomorphy 20 and a taxon in group 3 with apomorphy 16. 3. The ancestor of A. darmnii (7) and A. so- diroi (8) may have been a hybrid between a taxon in group 2 and one in group 3 that has apomor- phies 18(2) and 17(2). 4. A. paniculata (19) may be the hybrid of a taxon in group 3 and one outside of it because of its lack of apomorphy 22 (only two taxa in group 3 lack apomorphy 22). Other taxa show some indication of hybrid- ization within the groups. However, there is no strong indication of hybrids and their parents within the groups because there are so many al- ternative groupings and so few characters. Some taxa are obviously hybrids or of hybrid ancestry because of their ploidy level (Fig. 56, Appendix S) but there is little indication of what their his- tory might be. The different parsimony clado- grams give us different possible hybrids and par- ents. With the exception of two or three small groups of species that appear repeatedly on many, if not all, of the cladograms there are a few ad- ditional questions, such as biogeography, char- acter evolution, or ecology that can be investi- gated using these cladograms. To a large extent we are dealing with straight character patterns. We have apparently reached the limits of cla- distics with genera such as Acmella. I say this because cladistics is merely an organized way of looking at the relevant data that have been gath- ered. If no consistent pattern develops using cla- distics then the data are responsible, not the 704 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 72 .1* % \ \. \. V2./ / / \^ / ^^ / .' / \ ' \ t ' / \ \ \ \ \ . . . ! ^ v.- = hexaploid = tetraploid FIGURE 56. One of the many equally parsimonious cladograms for Acmella, many of which have different topologies. method. So the lack of resolution in genera such as Acmella is simply a reflection of the data. In the future we may be able to gain more infor- mation from genetic level research to increase the data base and obtain further resolution from cladistic analyses. CLASSIFICATION OF HYBRIDS A number of papers have been published that discuss the possibilities of classifying hybrids (Wiley, 1979; Wagner, 1980; Humphries, 1983; Humphries & Funk, 1984; Nelson, 1973) so the alternatives need not be discussed in this paper. As discussed in Humphries and Funk (1984), I prefer the method called phyletic sequencing or the annotated Linnean Hierarchy. This method works on the basic principle that all information from a cladogram is available in a classification and that a cladogram can be reconstituted from a classification. In such a classification only monophyletic groups are recognized. An exam- ple using the genus Anacyclus from Figure 39 is listed below: Classification of Anacyclus without hybrids Anacyclus A. pyrethrum A. monanthos A. maroccanus A. radiatus Clavatus species group A. linearilobus A. clavatus A. homogamos A. latealatus A. nigellifolius The hybrids can be added in several ways; one 1985] FUNK-HYBRIDIZATION 705 is to make them the sister taxon of either one of the parents. Using the first parent in the phyletic sequence we arrive at the following classification: Classification of Anacyclus with hybrids Anacyclus Sect. Pyrethraria A. pyrethrum +A. offkinarum sedis mutabilis (A. pyrethrum x radiatus) Sect. Anacyclus A. monanthos A. maroccanus A. radiatus Clavatus species group A. linearilobus A. homogamos +A. inconstans sedis mutabilis {A. homogamos x clavatus) +A. valentinus sedis mutabilis (A. monogamos x clavatus) A. clavatus A. latealatus A. nigellifolius The special notations include a plus sign (+) for hybrids, the parental species listed are the hy- brids in parentheses, and the latin phrase sedis mutabilis ("changeable position") that is used to mean a polytomy in the cladogram. The clado- gram is recovered in the following manner. Sec- tion Pyrethraria is the sister group of sect. An- acyclus, and A. pyrethrum and A. officinarum are sister taxa within sect. Pyrethraria (Fig. 39). In succession; A. monanthos is the sister taxon of the remaining species; A. marocannus is the sister taxon to the remaining species; A. radiatus is the sister taxon to the remaining species; the "clavatus group" is the sister group to A. lateala- tus and A. nigellifolius (Fig. 39); within the "cla- vatus group," A. linearilobus is the sister taxa of A. clavatus and A. homogamos and their two hybrids; and A. homogamos is the sister taxon of A. clavatus, but forms a polytomy with the two hybrids. I add to this the provision that should the two parents occur in different subgeneric groups (or different genera) then the hybrid should be listed in both groups. CONCLUSION In general, phylogenetic systematics can be used to help identify possible hybrids and their par- ents for further study. However, several condi- tions exist in some groups that make a cladistic analysis more difficult. Such conditions include the following: 1) having an increased percentage of species within a group that are hybrids, 2) having hybrids that have mainly plesiomorphies, 3) having hybrids that hybridize with one another, 4) having species that can reproduce asexually, and/or 5) having introgression occur. LITERATURE CITED ADAMS, E. N., III. 1972. Consensus techniques and the comparison of taxonomic trees. Syst. Zool. 21: 390-397. BREMER, K. 1983. Angiosperms and phylogenetic systematics: some problems and examples. Verh. Naturwiss. Vereins Hamburg 26: 343-354. & H.-E. WANNTORP. 1977. Phylogenetic sys- tematics in botany. Taxon 27: 317-329. & . 1979. Hierarchy and reticulations in systematics. Syst. Zool. 28: 624-627. CLARK, C. 1982. Relationships between experimental and phylogenetic systematics: an overview. Bot. Soc. Amer. Misc. Publ. 162: 88. FELSENSTEIN, J. 1983. Numerical Taxonomy. Spring- er-Verlag, Berlin, Heidelberg, New York, Tokyo. FUNK, V. A. 1981. Special concerns in estimating plant phytogenies. Pp. 73-86 in V. A. Funk & D. R. Brooks (editors), Advances in Cladistics: Pro- ceedings of the First Meeting of the Willi Hennig Society. New York Botanical Garden, New York. . 1982. The systematics of Montanoa (Aster- aceae, Heliantheae). Mem. New York Bot. Card. 36: 1-133. GOLDBLATT, P. 1979. Polyploidy in angiosperms: monocotyledons. In W. H. Lewis (editor), Poly- ploidy: Biological Relevance. Plenum Press, New York. GRANT, V. 1953. The role of hybridization in the evolution of the leafy-stemmed Gilias. Evolution 7: 51-64. . 1964. Genetic and taxonomic studies in Gil- ia. XII. Fertility relationships of the polyploid cob- webby Gilias. Aliso 5: 479-507. HENNIG, W. 1966. Phylogenetic Systematics. Univ. of Illinois Press, Urbana. HUMPHRIES, C. I. 1979. A revision of the genus An- acyclus L. (Compositae: Anthemideae). Bull. Brit. Mus. (Nat. Hist.), Bot. 7: 83-142. . 1981. Cytogenetic and cladistic studies in Anacyclus L. (Compositae: Anthemideae). Nordic I. Bot. 1: 83-96. . 1983. Primary data in hybrid analysis. Pp. 89-103 in N. I. Platnick & V. A. Funk (editors), Advances in Cladistics: Proceedings of the Second Meeting of the Willi Hennig Society. Columbia Univ. Press, New York. & V. A. FUNK. 1984. Cladistic methodology. Pp. 323-361 in V. H. Heywood & D. M. Moore (editors), Current Topics in Plant Taxonomy. Ac- ademic Press Inc., London. IANSEN, R. K. 1981. Systematics ofSpilanthes (Com- positae: Heliantheae). Syst. Bot. 6: 231-257. 706 ANNALS OF THE MISSOURI BOTANICAL GARDEN [VOL. 72 . In press. Systematics of Acmella (Asteraceae, Heliantheae). Syst. Bot. Monographs. LEWIS, W. H. 1979. Polyploidy in angiosperms: di- cotyledons. In W. H. Lewis (editor), Polyploidy: Biological Relevance. Plenum Press, New York. MAROUSH, T. & F. R. MCMORRIS. 1981. Consensus n-trees. Bull. Math. Biol. 43: 239-244. MCMORRIS, F. R., D. B. MERONK & D. A. NEUMANN. 1983. A view of some consensus methods for trees. Pp. 122-126 in J. Felsenstein (editor), Nu- merical Taxonomy. Springer-Verlag, Berlin, Hei- delberg, New York, Tokyo. MICKEVICH, M. R. 1978. Taxonomic congruence. Syst. Zool. 27: 143-158. NELSON, G. 1973. Classification as an expression of phylogenetic relationships. Syst. Zool. 22: 344- 359. . 1979. Cladistic analysis and synthesis: prin- ciples and definitions, with a historical note on Adanson's Families des Plantes (1763-1764). Syst. Zool. 28: 1-21. . 1983. Reticulation in cladograms. Pp. 105- 111 in N. I. Platnick & V. A. Funk (editors), Ad- vances in Cladistics: Proceedings of the Second Meeting of the Willi Hennig Society. Columbia Univ. Press, New York. & N. PLATNICK. 1980. Multiple branching in cladograms: two interpretations. Syst. Zool. 29: 86-91. NEUMANN, D. A. 1983. Faithful consensus methods for n-trees. Math. Biosci. 63: 271-287. ROSEN, D. E. 1979. Fishes from the uplands and intermontane basins of Guatemala: re visionary studies and comparative biogeography. Bull. Amer. Mus. Nat. Hist. 162: 267-376. SANDERS, R. W. 1981. Cladistic analysis ofAgastache (Lamiaceae). Pp. 95-114 in V. A. Funk & D. R. Brooks (editors), Advances in Cladistics: Proceed- ings of the First Meeting of the Willi Hennig So- ciety. New York Botanical Garden, New York. SCHUH, R. T. & J. T. POLHEMUS. 1980. Analysis of taxonomic congruence among morphological, ecological, and biogeographical data sets for the Leptopodomorpha (Hemiptera). Syst. Zool. 29: 1- 26. SEAMAN, F. C. & V. A. FUNK. 1983. Cladistic analysis of complex natural products: developing transfor- mation series from sesquiterpene lactone data. Taxon 32: 1-27. SEMPLE, J. C. 1981. A revision of the goldenaster genus Chrysopsis (Nutt.) Ell. nom. cons. (Com- positae?Astereae). Rhodora 83: 323-384. & C. C. CHINNAPPA. 1984. Observations on the cytology, morphology and ecology of Brad- buria hirtella (Compositae?Astereae). Syst. Bot. 9: 95-101. SOKAL, R. R. & F. J. ROHLF. 1981. Taxonomic con- gruence in the Leptopodomorpha re-examined. Syst. Zool. 30: 304-325. STEBBINS, G. L. 1974. Flowering Plants: Evolution Above the Species Level. Belknap-Harvard Press, Cambridge. WAGNER, W. H. 1954. Reticulate evolution in the Appalachian Aspleniums. Evolution 8: 103-108. . 1969. The role and taxonomic treatment of hybrids. BioScience 19: 785-789. . 1980. Origin and philosophy of the ground- plan divergence method of cladistics. Syst. Bot. 5: 173-193. . 1983. Reticulistics: the recognition of hybrids and their role in cladistics and classification. Pp. 63-79 in N. I. Platnick & V. A. Funk (editors), Advances in Cladistics: Proceedings of the Second Meeting of the Willi Hennig Society. Columbia Univ. Press, New York. WANNTORP, H.-E. 1983. Reticulated cladograms and the identification of hybrid taxa. Pp. 81-88 in N. I. Platnick & V. A. Funk (editors), Advances in Cladistics: Proceedings of the Second Meeting of the Willi Hennig Society. Columbia Univ. Press, New York. WILEY, E. O. 1979. An annotated Linnaean hier- archy, with comments on natural taxa and com- peting systems. Syst. Zool. 28: 308-337. & D. R. BROOKS. 1982. Victims of history- a nonequilibrium approach to evolution. Syst. Zool. 31: 1-24. APPENDICES A-S. Only the apomorphies are indi- cated in the data matrices because in phylogenetic sys- tematics only the apomorphies are used to group taxa. APPENDIX A. Data matrix for Figure 2. Apomorphies Taxa 1 2 3 4 5 6 7 A 1 1 1 1 1 B 1 1 1 1 1 1 C 1 1 1 D 1 1 E 1 APPENDIX B. Data matrix for Figures 3-6. Apomorphies Taxa 1 2 3 4 5 A 1 1 1 B 1 1 1 C 1 1 H 1 1 1 APPENDIX C. Data matrix for Figure 7. Apomorphies Taxa 1 2 3 4 5 6 A 1 1 1 1 B 1 1 1 C 1 1 H 1 1 1 1 1 1985] FUNK-HYBRIDIZATION 707 APPENDIX D. Data matrix for Figures 8-12. Apomorphies Taxa A B C HI H2 APPENDIX E. Data matrix for Figure 13. Apomorphies Taxa 1 2 3 4 5 A B C H APPENDIX F. Data matrix for Figures 14-17, 19. Apomorphies Taxa 1 A 1 1 B 1 1 C 1 1 D E H 1 1 APPENDIX G. Data matrix for Figure 18. Apomorphies Taxa 12 3 4 5 6 7 8 9 A B C D E H APPENDIX H. Data matrix for Figures 20-22. APPENDIX I. Data matrix for Figures 23-25. Apomorphies Taxa 12 3 4 5 6 7 9 10 11 12 A B C D E F G H I 1 1 1 1 1 1 1 APPENDIX J. Microloma R. Br. (Asclepidaceae). Abbreviations.?A. M. incanum Decne. ?B. M. lon- gitubum Schlechter.?C. M. burchellii N. E. Brown.? D. M. armatum. ? E. M. campanulatum. ? F. M. dolichanthum.?G. M. spinosumN. E. Brown.?H. M. viridiflorum N. E. Brown.?I. M. lanalum. Charac- ters. ? Data published in Bremer and Wanntorp (1979) and Humphries (1983) but no character list was fur- nished in either publication. Data matrix.?For Figures 26, 27, 36-38. Apomorphies Taxa A B C D E F G H I APPENDIX K. Data matrix for Figures 28, 29. Apomorphies Taxa 12 3 4 5 6 7 9 10 11 12 13 A B C D E 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Apomorphies Taxa 1 2 3 4 5 A 1 1 1 1 B 1 1 1 1 1 F 1 1 1 C 1 1 1 G 1 1 1 D 1 H 1 1 1 708 ANNALS OF THE MISSOURI BOTANICAL GARDEN APPENDIX L. Data matrix for Figures 30, 31. [VOL. 72 APPENDIX M. Data matrix for Figures 32, 33. Apomorphies Taxa Apomorphies Taxa 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 10 11 12 13 A 1 1 A 1 1 1 1 1 1 B 1 1 1 B 1 1 1 1 1 1 C 1 1 1 C 1 1 1 1 1 1 D 1 1 D 1 1 1 E 1 1 E 1 1 1 H 1 1 1 1 H 1 1 1 1 1 1 APPENDIX N. Anacyclus L. (Asteraceae, Anthemideae). Abbreviations.?A. A. pyrethrum (L.) Link var. pyrethrum.?B. A.pyrethrum(h.)lAnkvar.depressus(Ba.lV)M!dTe.?C. A. monanthos (L.)Thell. ? D. A.maroc- canus (Ball) Ball.?E. A. radiatus Loisel.?F. A. coronatus (Murb.) Humphries.?G. A. clavatus (Desf.) Pers.? H. A. homogamos (Maire) Humphries.?I. A. mlentinus L.?J. A. inconstans Pomel.?K. A. linearilobus Boiss. & Renter.?L. A. latealatus Hub.-Mor. ? M. A. nigellifolius Boiss.?N. A. qffkinarum Hayne. Characters.? Published in Humphries (1979). Data matrix.?Lower case letters represent groups of characters that display that pattern so that there are five apomorphies that have the distribution patterns of a, etc. For Figures 39-43. Taxa Apomorphies abode fghi j klmnopq (5) (3) (1) (3) (3) (1) (3) (1) (1) (1) (1) (1) (3) (5) (4) (1) (1) AB C 1 1 D 1 1 1 EF 1111 1 G 1111 1111 H 1111 1111 1 I 1111 1111 1 J 1111 1111 1 K 1111 LM 1111 N 1 1985] FUNK- HYBRIDIZATION 709 ^ 4* a .9 1.3 an * .5 ? "a 55 ? ^ I B I ^ t I <* ?>* ??as W .9 c 5 ft?< I ^ 05 3 Kg 73 5- X a % o ^ d -?! #8: g 2 _ g =3 ^ PC. o ^ 2 d& 2 | I o BO a) .a v. ?a ? W C3 v.4 a ?c ft ?^ O & < I -3 2 | # q a .2 7, <% 13 a PC o ^ O o ^ a a g -a A O = II 13 ^ I. o T3 g C 00 6 a aw '. a oo ?" a ^3 ^ ft ?S J? ?J ^ - I. ?~ a a Q j o '. ea >> p -S R T3 a 3 ^3 -s o ^ o S fe o . o O BC u I >, T3 iJ a a ^ a 2 2 O o ^< ? "2 < ?? j ^ "5 > ? ?? '^# S3 c ?g 'C a ? | 55 8 a ^ o 8 "< 3 c S, >> ^ > a ft 9 ? a 00 ^ s SB PC sr