Patterns of homoplasy in North American Astragalus L. (Fabaceae).

Item Type text; Dissertation-Reproduction (electronic)

Authors Sanderson, Michael John.

Publisher The University of Arizona.

Rights Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.

Download date 10/10/2021 18:39:52

Link to Item http://hdl.handle.net/10150/184764 INFORMATION TO USERS The most advanced technology has been used to photo­ graph and reproduce this manuscript from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer. The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction.

In the unlikely event that the author did not send UIVn a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion.

Oversize materials (e.g., maps, drawings, charts) are re­ produced by sectioning the original, beginning at the upper left-hand corner and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the back of the book. These are also available as one exposure on a standard 35mm slide or as a 17" x 23" black and white photographic print for an additional charge.

Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6" x 9" black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order.

U·M·I University Microfilms International A 8ell & Howell Information Company 300 North Zeeb Road. Ann Arbor. M148106-1346 USA 313/761-4700 800/521-0600

Order Number 9000144

Patterns of homoplasy in North American Astragalus L. (Fabaceae)

SandersOl~, Michael John, Ph.D.

The University of Arizona, 1989

U·M·I 300 N. Zeeb Rd. Ann Arbor, MI 48106

PATTERNS OF HOMOPLASY

IN NORTH AMERICAN ASTRAGALUS L. (FABACEAE)

by

MICHAEL JOHN SANDERSON

A Dissertation Submitted to the Faculty of the

DEPARTMENT OF ECOLOGY AND EVOLUTIONARY BIOLOGY

In Partial Fulfillment of the Requirements For the Degree of

DOCTOR OF PHILOSOPHY

In the Graduate College

THE UNIVERSITY OF ARIZONA

1 9 8 9 THE UNIVERSITY OF ARIZONA GRADUATE COLLEGE

As members of the Final Examination Committee, we certify that we have read the dissertation prepared by __M_i_c_h_a __ e_l __ J_. __ s__ a_n_d_e_r_s_o __ n______

entitled Patterns of homoplasy in North American Astragalus

and recommend that it be accepted as fulfilling the dissertation requirement

for the Degree of Doctor of Philosophy

M. Donoghue Date ' ~~~~

R. Hoshaw ~~f{-.~~~ R. Robichaux

Date

Final approval and acceptance of this dissertation is contingent upon the candidate's submission of the final copy of the dissertation to the Graduate College.

I hereby certify that I have read this dissertation prepared under my direction and recommend that it be accepted as fulfilling the dissertation requirement. lJ\\~t~:r, .b~, \\-JL- Dissertation Director Date ) 3

STATEMENT BY AUTHOR

This dissertation has been submitted in partial fulfillment of requirements for an advanced degree at The University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the library.

Brief quotations from this dissertation are allowable without special permission, provided that accurate acknowledgment of source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the head of the maj or department or the Dean of the Graduate College when in his or her judgement the proposed use of the material is in the interests of scholarship. In all other instances, however, permission must be obtained from the author.

SIGNED: 4

ACKNOWLEDGMENTS

Michael Donoghue initially proposed the idea of working on the "hopeless" genus Astragalus, although he should not be held accountable for my desire to pursue a proj ect of the present scope. His enthusiasm for studying evolutionary problems via cladistic methods was infectious and greatly influenced the course of my work. As always, his tenacious reading of early drafts of the dissertation improved it immeasurably. Thanks also to the other members of my committee, Dr. William Heed, Dr. Robert Robichaux, Dr. Robert Hoshaw, .and Dr. Richard Strauss, all of whom provided much needed criticism of the present work. Rupert Barneby gave badly needed encouragement to my early thoughts on working in the genus. I thank him for taking the time to respond at such length to my letters and for hosting my visit to the New York Botanical Garden in 1986. The present work has benefited greatly from the intellectual climate provided in Dr. Donoghue's lab by Geeta Bharathan, Mark Porter, and Jim Malusa. Larry Hufford advised me on many developmental issues and the ideas about character resolution in Chapter Three grew from conversations with him. Ned Friedman also helped me on my first attempts at serial sectioning. Renee Rusler deserves much of the credit for all my field work, since I probably would not have gone without her, and she continually prodded me into looking over one more hill for those damn plants. Geeta, George Ferguson, Marty W., and most especially George Davenport also accompanied me on field trips and helped to make all of them memorable. I thank the curators of several herbaria in Ylhich I worked at various times, including NY, RSA, NM, and the especially the staff at UC (Barbara Errter and Elizabeth Neese). Dr. C.T. Mason and Becky van Devender are especially deserving of my thanks for providing a "home" in the U of A herbarium. I would like to acknowledge a debt to four present and former faculty members in the Department of Ecology and Evolutionary Biology, each of whom at early stages in my graduate studies helped awaken in me an interest in "real" biology. They are Dr. James H. Brown, Dr. Charles H. Lowe, Dr. Robert Hoshaw, and Dr. Charles T. Mason. A special word of thanks is also due Rick Michod, my first major professor, who in very many ways is responsible for my involvement in evolutionary biology. Finally, I thank my parents for providing more than occasional financial support throughout my lengthy stay in graduate school. This research was supported by an Herbarium Travel Grant from the American Society of Plant Taxonomists, and by a McGinnies Scholarship from the Office of Arid Lands Studies at the University of Arizona.

------.--~--- 5

TABLE OF CONTENTS

LIST OF TABLES 7 LIST OF FIGURES ...... 8 ABSTRACT 11

1. INTRODUCTION: HOMOPLASY AND EVOLUTION . 13

Homoplasy: Definitions and History . . • . . . 13 Homoplasy and Homology ...... 16 Classical View (Similarity Alone) ..• 17 Evolutionary View (Similarity plus Common Ancestry) ...... •... 20 Transformational View ("Transformation" regardless of Similarity) . • . . . . • . 25 Taxic View (Homology equals Synapomorphy) . 26 Homology as Information continuity . . . . 30 Homoplasy and Homology Defined . • . 31 Homoplasy in Relation to Evolution and Phylogeny •.•.... 33 Homoplasy and orthogenesis · · · · · · 33 Homoplasy and Tendencies · · · · · · 36 General Patterns of Homoplasy · · · · · · 41 2. PHYLOGENETIC ANALYSIS OF NORTH AMERICAN ASTRAGALUS 46

Introduction . . . . 46 Overview of the Genus ...... 46 Monophyly of the Genus and its Relationships . . • ...... 50 Methods .•••.••... • • • 55 Scope of Present Study • • • • 55 Characters and Taxa Used • • • 65 Polymorphism • • . • . . . • • • • • • • 76 Correlation of Characters · • • 77 Quantitative Characters .•.. • • • 80 Rooting ••.•.. . . • . • • • 84 Algorithms Employed .... 87 Results .••.•...... • • • • 88 Overview of Trees Found • • • 88 Notes on Selected Major Clades · . . . 91 Discussion . . . • . • . . . . . • • • • 108 Description of the Phylogeny • •• . 108 Relation of Results to Classical taxonomies . . . . . • • ...... 109

-_.- -.------6

The Nature of Astragalus species: Cladistic Implications . . . • . · 121 Exclusion of South American Species · 123 Evolution of Selected Characters · 125

3. STATISTICAL TESTS FOR HOMOPLASTIC TENDENCIES 148

Introduction ...... • . . . . . 148 Methods . .. • ...... 152 Reconstructing the Distribution of Homoplasy. 152 Measures vf spatial Pattern ...... 156 Null Models for the Expected Distribution . . 162 Tests of Significance . . . • . . . . • . . . 169 The Computer Program ...... • . 170 Studies Employed and Characters Analyzed . . 172 Results ...•..•...... 183 Behavior of the Null Models . . . . •. . 183 Clustering in Three Focal Characters . . . . 186 Clustering in 14 Additional Characters • 191 Clustering in the Iguanid Lizard Study . 192 Discussion • . . . • • • ...... 192

APPENDIX 1: PATTERNS OF VARIATION IN LEVELS OF HOMOPLASY ...... 202

APPENDIX 2: DESCRIPTIONS OF CHARACTERS USED IN ANALYSIS OF CHAPTER TWO . 256

APPENDIX 3: CLADISTIC ANALYSIS OF ASTRAGALUS SECTION ARGOPHYLLI GRAY . . . . . 280

APPENDIX 4: LISTING OF PROGRAM FOR ANALYSIS OF HOMOPLASTIC TENDENCIES . . . . 287

APPENDIX 5: MORPHOMETRIC VARIATION IN FRUITS OF ASTRAGALUS ...... 305

APPENDIX 6: NEW TAXA OF ASTRAGALUS DESCRIBED SINCE BARNEBY (1964) ...... 309

LITERATURE CITED . . . . . • ...... 312 7

LIST OF TABLES

Table Page

1. List of taxa used in cladistic analysis of North American Astragalus • . . 57

2. List of sectional abbreviations 62

3. List of characters used 63 4. Data matrix for cladistic analysis 66

5. comparison of clustering using mean nearest­ neighbor distances and empirical distribution functions • .. ..••• ...... • . . 161

6. Null models used and their assumptions 166

7. Factor loadings of first two principal components used in morphometric analysis of red-flower evolution. . • . . . 175

8. Levels of spatial clustering under null model I for North American Astragalus . · · · · 187 9. Levels of spatial clustering under null models III and IV for N. Am. Astragalus · · · · 188 10. Levels of spatial clustering under null models III and IV for Igaunid lizards . · · · · 193 11. Effects of scope, scale, and resolution on aspects of homoplasy . . . . • • . . . 195

A1-1. Summary of cladistic analyses used. . 242

Al-2. Correlation matrix for factors in homoplasy .. 246

Al-3. Multiple regression analysis of CI . . . .. 247

A3-1. Characters used in cladistic analysis of section Argophylli . . ... • . .. 284

A3-2. Data matrix for analysis of Argophylli . 285 8

LIST OF FIGURES

Figure Page

Illustration of homoplastic tendencies • . . . • 37

2 • Effect of number of taxa on amount of homoplasy. 42

3. Phylogenetic relationships of Astragalus . . . . 51

4. Morphometric variation for fruit size and shape for 162 species of North American Astragalus 81

5. Majority rule consensus tree of 250 equally parsimonious trees of length 595 steps . . . 89

6. Majority rule consensus tree of 250 equally parsimonious trees of length 596 steps . . . 90

7a. One of the 596 step trees (shown in four parts) 93 b. Part of the 596 step tree . • . . . . 94 c. Part of the 596 step tree . . . . . 95 d. Part of the 596 step tree • 96

8. Marcus Jones' phylogeny of North American Astragalus ...... · · · · 114 9. Charac't.er evolution in fruit septum · · 128 10. Character evolution in fruit-cross section · 129 11. Character evolution in fruit incurvature · · · · 130 12. Character· evolution in fruit inflation 131 13. Character evolution in fruit texture · · · 132 14. Character evolution in fruit stipe . . . · · · · 133 15. Character evolution in fruit persistence · 134 16. Character evolution in flower and calyx conformation . • . . . • . • • . • . • . . . . . 135

17. Character evolution in point of renewal .... 136

18. Character evolution in life-cycle ...... 137

19. Character evolution in leaflet shape · . . 138

...... - .. - ...... ----.-...._-_ ... --_._------9

20. Character evolution in stipules . . 139

21. character evolution in chromosome number 140 22. Patterns of clustering and localization of homoplasy . . . . . ~ . . • . . . • . . 151

23. Alternative reconstructions of character state changes . • . . • . • . • . • ...... 153

24. Illustration of hidden homoplasy that can never be inferred using parsimony ...... 155

25. Illustration of various methods of measuring spatial pattern on a tree • • . . • .. .. 159

26. Illustration of completely pectinate and completely dichotomous trees . . . 163

27. Behavior of Model III 165

28. Parallel trend in floral evolution in three clades of North American Astragalus. .. .. 173

29. Topological distribution of homoplasy in red-flowered morph in North American Astragalus...... • . . . 174

30. PCA analysis of parallelism in red flowers ... 176

31. Illustration of variation in the compound leaf ...... •...... 178

32. Topological distribution of homoplasy in terminal leaflet trait in North American Astragalus . . . . • • ...... • ...... 179

33. Topological distribution of homoplasy in septum character in North American Astragalus . . . . • ...... 181

34. Expected mean nearest neighbor distances under null models I, II, and III...... 184

35. Hypothesized effect of increasing character resolution on the occurrence of clustering • 199

AI-I. Effect of number of taxa on homoplasy 250

.. _. __ ._._-- ._------10

Al-2. Effect of number of characters of homoplasy 251

Al-3. Effect of taxonomic rank on homoplasy 252 Al-4. Comparison of plant/animal homoplasy .. 253

Al-5. Comparison of morphological/molecular homoplasy. 254 Al-6. Confidence levels and homoplasy ...... 255

A3-1. Cladogram of Astragalus section Argophylli . 286 11

ABSTRACT

Patterns in the distribution of homoplasy are investigated from theoretical and empirical perspectives. The history of the term "homoplasy" as used by morphologists, evolutionary systematists, cladists, and others is reviewed, especially in relation to its complement, "homology."

Homoplasy is defined relative to homology, which is viewed as any similarity shared through' an unbroken line of common ancestry. An investigation of levels of homoplasy based on a statistical analysis of 60 published phylogenies reveals a strong dependence of homoplasy on the number of taxa included.

This relation is independent of number of characters, type of data, taxonomic rank, or organism, and suggests that large taxa should be the focus of empirical studies of homoplasy.

Hence, a phylogenetic analysis of the large genus

Astragalus was undertaken using 113 representative species

(and varieties) found in North America. Fifty-seven binary and mUltistate characters were scored and the resulting matrix was subj ected to numerical cladistic analysis. Two large sets of equally parsimonious trees were found at 595 and 596 steps.

The sets were analyzed using consensus methods, robust clades were discussed in detail, and the phylogenies were compared to previous classifications. Character evolution of a large set of taxonomically important and morphologically varied traits was investigated. 12 statistical tests were developed to detect patterns of topological clustering of homoplastic character changes in cladograms. The tests use Monte-Carlo computer simulations of four null models of character evolution in an attempt to rej ect the hypothesis of random homoplastic distributions. For the Astragalus data set only two of 17 characters were significantly clustered, and this is close to random expectation. Another data set from the literature was also tested, and in it no characters were clustered at the 5 percent level. The explanation for these negative findings regarding homoplastic "tendencies" is explored with respect to "scope", "scale", and character "resolution," factors believed to play an important role in the analysis of character evolution. 13

CHAPTER ONE INTRODUCTION: HOMOPLASY AND EVOLUTION

" •.. the members of the same class, although only distantly allied, have inherited so much in common in their constitution, that they are apt to vary under similar exciting causes in a similar manner; and this would obviously aid in the acquirement through natural selection of parts or organs, strikingly like each other, independently of their direct inheritance from a common progenitor." (Darwin, 1872:328)

HOMOPLASY: DEFINITIONS AND HISTORY Why study homoplasy, the recurrent origin of similarity? Evolutionary biology seeks to uncover the mechanisms responsible for the origin of differences among organisms, but its complementary task is to explain similarities among those organisms. Unlike differences, however, similarities can arise either through default, by the absence of evolutionary forces leading to changes in characters, or by the positive action of evolutionary forces applied to formerly less similar organisms. The first kind of similarity is important in systematics as an indication of phylogenetic h'istory but is less interesting from an evolutionary perspective because it requires no explanation beyond "descent without modification." The latter kind of

------_._------14 similarity, often viewed as merely a hindrance by systematists (e.g. Patterson, 1982), has long fascinated evolutionary biologists.

since Darwin, mechanistic explanations for homoplasy have been of two kinds, selectionist and orthogenetic, corresponding roughly with the two major schools of evolutionary theory that were at odds before the Modern

Synthesis. Even today the orthogenetic explanation is favored among some systematists (e.g. cronquist, 1968;

Grehan and Ainsworth, 1985), and many others subscribe to a thinly veiled version of it in discussing "tendencies" of characters to evolve in parallel within related groups. The persistence of orthogenetic thinking may stem from the neglect by contemporary microevolutionists of interesting empirical patterns of homoplasy known to many workers engaged in the reconstruction of phylogeny. Orthogenesis was rejected by the Modern synthesis mainly because of the perception that the mechanism of natural selection was sufficient to explain homoplasy at all levels. Since a plausible mechanism for orthogenesis was never demonstrated by its proponents, the concept was abandoned and derided

(Bowler, 1984).

Unfortunately, explanations of macroevolutionary patterns based on natural selection have not greatly enhanced our understanding of the mechanisms responsible for 15 homoplasy. Discussions of homoplasy rarely do more than invoke a sense of wonder at the power of convergent selection. Just as a good case has been made recently for deeper studies of adaptation (Ridley, 1983; Greene, 1986;

Coddington, 1988, Donoghue, 1989), homoplasy deserves closer scrutiny. One reason for the dearth of insightful studies of homoplasy may be the overemphasis on individual case studies focusing on one character in one group of taxa

(e.g., Dahlgren, 1970; Alberch, 1981). General conclusions from such work are difficult to find.

This thesis is an analysis of patterns of homoplasy at several levels. Its goal is to uncover patterns sufficiently general and replicable that they may suggest the evolutionary mechanisms responsible for homoplasy. The first chapter introduces the subject of homoplasy with an analysis of the various meanings that have been associated with the term, and proposes a definition appropriate to the present study. Several very general empirical features of patterns of homoplasy are discussed by reference to a paper by Sanderson and Donoghue (Appendix 1). The second chapter presents a cladistic analysis of North American species of the large flowering plant genus Astragalus which, because of its large size, is an appropriate model system for the study of homoplasy. Finally, the third chapter proposes statistical tests that can detect topological patterns in 16 homoplasy--specifically, the clustering of homoplasy, which is the pattern expected under certain "tendency" models of evolution. These tests are applied to the phylogenetic analysis of Chapter Two, and also to a representative study from the zoological phylogenetic literature (Etheridge and de Queiroz, 1988).

Homoplasy and Homology -- The term "homoplasy" has a long history within evolutionary biology but is less widely known than its counterpart "homology." Since homoplasy is often defined as the complement of homology, it is nearly impossible to adequately define it or relate its historical development without discussing homology. The following is not meant to be an exhaustive review of the complex history of homology, but it is necessary to summarize that history in more than a superficial manner (see reviews in Haas and

Simpson, 1946; Simpson, 1959: Patterson, 1982; Roth, 1988).

Rather than provide a chronological review, the following is a conceptual classification of various ways of thinking about homology (highly modified from Patterson, 1982). The historical transformation of the concept of homoplasy can then best be discussed within that framework. Finally, the views of homology/homoplasy adopted here will be presented and defended. Various ancillary terms such as "analogy," 17

"convergence," "parallelism," and "homoiology" will also be discussed.

Classical View (similarity Alone) -- Owen (1843, 1847, 1848)

is generally credited with first using the term homology for

" ... the same organ in different animals under every variety

of form and function" (1848). His view of analogy was "a

part or organ in one animal which has the same function as

another part or organ in a different animal." Kaplan (1984)

provides recent exposition of this view. He suggests that

it is senseless to define homology in terms of phylogenetic hypotheses that, in practice, are far removed from direct

observation. Instead homology should be assessed through

comparisons made prior to phylogenetic study, based on

certain special criteria of similarity (he cites Remane's

positional and developmental criteria, 1952).

Interestingly, Kaplan advocates Lankester's definition of

terms (1870), in which homology arising via common ancestry

is "homogeny," and homology arising independently is

"homoplasy." Non-homologous (sensu Kaplan) characters

relating to function are then "analogies." It is not clear

what non-homologous, non-functional independent origins of traits would be termed.

Kaplan (1984) initially motivates this non-phylogenetic

definition of homology on "operational" grounds. On the one 18 hand, Kaplan clearly intends that homology be assessed by

the careful application of certain special criteria of

similarity. Anyone, anywhere, ought to be able to study the

same organisms and check whether Remane's criteria are

satisfied or not, in much the same way as anyone could

determine whether a ball is blue. Once it is agreed what

the definitions of similarity are (e.g., Remane's criteria),

most of the work should then be done. Yet, Kaplan

painstakingly discusses two case studies in which he

suggests that only by the most thorough and detailed

analysis can homology be accurately assessed. This suggests

that different scientists might arrive at different

conclusions in assessing homology via these same criteria.

The history of the comparative morphology of actual

characters argues that this is indeed the case (e.g., the

interpretation of homologies of the seed cone in conifers) .

What does this potential disagreement imply about the

process of assessing homology by similarity criteria? It

might mean that such homology assessments are just variant

hypotheses that deserve to be tested further as part of

normal scientific inquiry. However, it is more likely that

the controversy in this kind of homology assessment sterns

from differences over the similarity criteria themselves.

It is not clear what it takes to demonstrate "positional

similarity, for instance. Evidence satisfactory to Kaplan 19 or Remane may not be satisfactory to someone else. The homologies of homoeotic mutants, such as compound leaves that arise in the position of stipules, are particularly troublesome (Sattler, 1988). Hypotheses about homology erected in this way are hypotheses primarily about goodness of fit of observation to some codified similarity criteria.

Unless one adopts an essentialist view, one cannot say that two structures are "truly" positionally the same; one can only say that they are in the same position according to the definition of some person X. Arguments that result from application of XIS definition are merely disputes over what

X "means" by those criteria (Brown, 1977).

This view of homology does not appear especially relevant to evolutionary biology. Possibly at its core it actually is an attempt to characterize something about evolution. For instance, an advocate of the classical view might believe that rigorous application of Remane's criteria will reveal cornmon ancestry and point out independently derived features, prior to any phylogenetic study. One objection to this is the lack of empirical justification for the equivalence of homology and cornmon ancestry. There has never been a phylogenetic test of the validity of Remane's criteria. Moreover, this view of homology suggests that certain characters -- those satisfying Remane's criteria 20 cannot undergo homoplasy. Again, there is little empirical evidence to support this idea. Since Kaplan agrees with Lankester (1870) that certain homologies ~ arise independently as homoplasy, it is clear that his homology does not automatically guarantee common ancestry. Perhaps Remane's criteria are merely meant.to be a better than average clue about common ancestry. However, there is as little hard evidence for this notion as there is for the stricter view that homology always equals common ancestry. Furthermore, why adopt a term for a special kind of similarity that might or might not have something to do with common ancestry? The only justifiable reason is that it might reflect some other interesting feature of the biology or evolution of the organisms involved. But if homology sensu Kaplan has such a connotation, it is unclear what it is.

Evolutionary View (Similarity Plus Common Ancestry) Darwin (1872) used homology in much the same way as Owen (1848) did, as similarity that satisfies certain positional or other criteria. citing the work of Geoffroy st. Hilaire, he emphasized the "high importance of relative position or connexion in homologous parts" (1872:334). Darwin's contribution was to provide an explanation for this "conformity to type" by demonstrating that such homologies 21 often reflect descent from a common ancestor. He was less interested in defining homology as an evolutionary concept than he was in interpreting the then prevalent view of homology in the context of his theory. He reserved the term "analogical resemblances" for those broad similarities that arise in distantly related groups through the action of adaptation. By the publication of the sixth edition of the origin of species (1872), Darwin had been greatly influenced by a paper by E. Ray Lankester (1870), which for the first time proposed an explicitly phylogenetic definition for homology and introduced the term "homoplasy." Lankester also began with the existing concept of homology as similarity of a certain kind. But he was very critical of Owen's view of homology as "essential sameness" on account of its arbitrariness, and sugg'ested +:hat phylogenetic history could provide a suitable criterion to judge similarity. He decomposed homology into two exclusive classes: "homogeny," or homology due to common ancestry~ and "homoplasy," or homology due not to common ancestry but to independent origination. Since functional analogy rarely would satisfy any rigid homology criteria, Lankester viewed it as a concept operating outside the sphere of homogeny/homoplasy. Darwin (1872) regarded this as a major advance, and Lankester's work has had a powerful, though not sufficiently

...... -_ ... _._------_.. _------22 appreciated, effect on the concept of homology ever since.

Among the major contributors to the modern synthesis, the systematists and paleontologists, Simpson and Rensch,

and to a lesser extent, Mayr and Stebbins, were most

concerned with issues of homology and homoplasy. Their general objectives seem to have been twofold: (1) to

establish, once and for all, an evolutionary concept of homology, and (2) to repudiate orthogenetic interpretations

of non-homologous trends and series (parallelisms) that had

been widely discussed in the period following Darwin.

Because of his prolific writings, Simpson's views have

had a wide impact. His first major discussion of homology

can be found in a paper by Haas and Simpson (1946), in which

homology is equated with similarity arising from common

ancestry, and homoplasy is "non-homology." This paper,

though coathored by simpson, is primarily the work of Haas

(see Simpson, 1980), who should be credited with clarifying

a hopelessly confused historical and semantic muddle

concerning concepts of homology and homoplasy. Simpson

(1980) distinguished his view of analogy, however, from

Haas's and from Lankester's. Both those authors regarded

analogy as independently evolved functions. Simpson, on the

other hand, believed that analogy is similarity in function

and the structures associated with that function (see also

Haas and Simpson, 1946:325).

------23

Simpson's view of the difference between convergence and parallelism is interesting because it refers to a process (at least implicitly). He regards convergence as acquired similarity in spite of divergent ancestry, while parallelism is acquired similarity among taxa because of their common ancestry (Simpson, 1961). Some causal property of common ancestry, such as similarity of genetic and developmental systems is implicated.

Although Bernhard Rensch is rarely credited with playing a major role in the Modern Synthesis, his book

Evolution above the Species Level (1959, based on material written in the 1940's) contains a most insightful synthesis of material about homoplasy. His familiarity with a vast body of information about paleontological and systematic patterns, and his intellectual upbringing as part of a long

German tradition of orthogenetic thinking made him uniquely qualified to analyze and repudiate orthogenesis. His

interests were in mechanisms responsible for homoplasy rather than in philosophy and he spent little time in defining terms. Evidently he regarded homology in much the

same way as Simpson did. Homology criteria like those of

Remane (1952) he saw as empirical guideposts (p. 200) only.

His "parallelism" subsumed three aspects of homoplasy: (1)

parallel mutation, (2) parallel selection on homologous

structures, and (3) parallel selection on analogous

------_.. _---- 24 structures (which he used interchangeably with convergence) .

In the first, Rensch included identical mutations in

identical genes, such as eye color mutants in Drosophila or albinism mutants in vertebrates. Unless selected to

increase in frequency this kind of parallelism is not likely to be an important factor in evolution. His point in discussing the other two kinds of parallelism is simply to demonstrate that selection can explain patterns of phylogenetic parallelism without the need to invoke

orthogenesis. While his arguments are frequently

compelling, at times they retreat into elaborate exercises

in adaptive story-telling. His most detailed example

concerns parallel trends for increases in body size in mammals ("Cope's Rule", Cope, 1877). After concluding that the pattern itself is quite robust, that almost all lineages

of mammals have undergone a general trend towards larger body size, he suggests seven reasons why being large ought to be favored by selection (Rensch, 1959:211ff.). Rensch made an important advance by demonstrating that natural

selection was a sufficient explanation for hitherto

orthogenetically interpreted patterns, but it remains to be

shown whether it is a necessary explanation as well. 25

Transformational View ("Transformation" regardless of

Similarity) -- Hennig (1966) did not use the term

"homoplasy," preferring instead to contrast homology with convergence, analogy, "homoiology," and parallelism. He defines homologies as transformations of an original character during the course of descent from an ancestor.

Unfortunately, "transformation" is an elusive concept to define without reference to similarity, and once similarity is introduced the definition of homology becomes somewhat arbitrary. Implicit in the word, however, is the concept of

common ancestry and a temporal sequence of change. These are both inferential elements of the definition, and thus are hypotheses about the world, rather than mere semantic points-- unlike the pure similarity view of homology.

Following Remane (1952) Hennig equated convergence \.,ith analogy. "In convergence, two forms with similarities in directly adaptive structures •.. have come from radically different ancestors with basically different patterns of organization •.• " (1966:117). Homoiology refers to the acquisition of similar forms independently in very closely

related taxa, although what is meant by his "narrow kinship

group" (Hennig l 1966:117) is vague. Finally, parallelism

refers to divergence from "a common primary condition"

(Hennig, 1966:119) followed by a similar sequence of change

over time in the different taxa. It is unclear whether 26 common primary condition means common ancestor, or merely common character state possessed by two taxa that are possibly fairly closely related. Some of the examples cited by Hennig are familiar ones involving long term unidirectional paleontological trends. Others are indistinguishable from his examples of homoiology (e.g. as parallelism: decrease in body size of chironomid Diptera; as homoiology: increase in size of horns of titanotheres) .

Taxic View (Homology=Synapomorphyl -- Eldredge and Cracraft

(1980) do not explicitly refer to transformation in defining homology; instead they define it as "inferred inherited similarities" (p. 36) and equate it with synapomorphy -­ which does not follow, as demonstrated by Patterson (1982)

(i.e., inferred inherited similarity can be plesiomorphic).

Homoplasy is not defined, but is essentially subsumed by convergence, defined as non-homologous similarity. They define parallelism in a peculiar and restrictive way only to then reject the entire concept as unscientific. According to these authors parallelism refers to the origin of similarity independently just in sister taxa (taxa that are each other's closest relatives). They point out that parsimony algorithms would' never permit characters to be optimized on a tree such that sister taxa would exhibit such parallelism, which is true. This is a defect in parsimony 27 algorithms, however, not a defect in the concept of parallelism. certainly it is conceivable that sister-taxon parallelism has occurred in nature. Algorithms other than parsimony, such as maximum likelihood can, at least in principle, uncover such parallelism. Patterson (1982) unequivocally equates homology with synapomorphy. Several critical issues are raised in this paper. The first concerns the relationship between similarity and homology. Patterson's view on this is unclear although he seemingly wants to reject any relationship of similarity and homology. He cites Bock's (1963) argument that homologues need not be at all similar, as in the case of ear ossicles in mammals versus the putatively homologous hyomandibular of sharks. But then he claims that no author has convincingly rejected a relationship between similarity and homology. Patterson's main point is to suggest that homology in the sense of similarity due to common ancestry is simply not interesting in cases where the similarity characterizes paraphyletic groups. He cites the absence of wings in apterygote insects (presumably a paraphyletic group) :

~he absence of wings in all these organisms is homologous by the conventional evolutionary definition (similarity due to common ancestry), but there is no more interest in that "homology" than in the fact that invertebrates, plants, and protists also lack molar teeth, feathers, pentadactyl limbs and an infinity of other homologous structures. [1982:30J 28

To the systematist interested only in reconstructing the nested hierarchy of monophyletic groups, only information relevant to that task may seem interesting -- such as

synapomorphies. There is no necessity that homology, however, must have the same properties as synapomorphy,

simply because synapomorphy is an interesting and very

useful concept. Patterson seems to admit this in an

important section on "Taxa and Transformations." Here he

points out the distinction between the synapomorphic view of

homology, which is "taxic" and aimed at characterizing monophyletic groups of taxa, and the "transformational" view, which is more concerned with the evolutionary process

of character change. He cites an example from tetrapod

phylogeny in which some have argued that the incus and

malleus of mammals (ear ossicles) are transformations of the

quadrate and articular, while others have argued they are

transformations of the stylohyal and ceratohyal. According

to Patterson's taxic view it makes little difference which

evolutionary hypothesis is correct because the mammalian ear

ossicle bones are synapomorphies of mammals in either case,

and the other sets of bones are both synapomorphies of bony

fish plus tetrapods, characterizing the same group. "That

example implies that the transformational approach may be

more informative, and a lot more interesting, a point I will

not deny." He then continues, II •••• concentrating on

------29 transformation at the expense of taxa is not fruitful (p.

36)." Surely, however, we are not forced to choose one or the other. It is appealing to reject any term that is blatantly associated with paraphyly. Patterson is correct in pointing out that homology sensu simpson can in many cases mean shared plesiomorphic similarity. However, this does not impinge on methods of phylogeny reconstruction, since it is now well established that shared derived features are the only ones that delimit monophyletic groups. The important question is whether retaining Simpson's view of homology is at all useful in understanding the processes of evolution and transformation of characters. Patterson views homoplasy as "non-homology" or things that might be mistaken for homology. He points out the historical distinction between convergence and parallelism and defines them in terms of two of his tests of homology (similarity, and congruence -- i.e., agreement with phylogenetic hypotheses suggested by other characters): convergence fails both tests, parallelism fails the congruence test but passes the similarity test. This distinction is a bit hyperbolic, as convergence would not be a long-standing issue in evolutionary biology were it not for the striking similarities that it evinces. Patterson

------30 presumably has stringent criteria in mind such as Remane's for his similarity test.

Homology as continuity of Information -- Roth (1988) has reviewed a relatively recent development in the concept of homology, in which homology is defined as correspondences that arise because of continuity of information (Van Valen,

1982). Note that correspondence is just another word for

similarity. In principle, the continuity can be provided by any number of processes, including but not restricted to, genealogy or ontogeny. According to Roth, this definition

is " .•. the most succinct, comprehensive, and ideologically neutral definition of homology yet proposed" (1988:2). In

fact it is no more "succinct" than any of the previous views of homology. It is more comprehensive than some definitions since it can subsume, for example, the evolutionary and classical views, but this is not necessarily a good property. Nor is "ideological neutrality" a good thing

except insofar as uncritical ideology is rarely

constructive. It is curious that Roth believes that this

kind of broad "comprehensive" definition is

admirable:" ... its beauty lies in its flexibility: the

definition can be used by adherents to any school of

thought ..• " (1988:2). If, as Roth implies (p.2), this view

of homology is equally useful to a Creationist or 31 evolutionist, it may not represent a useful tool for understanding the processes of evolution.

Homology and Homoplasy Defined -- The definitions of

homology/homoplasy adopted in the present work are based

primarily on those of Haas and Simpson (1946): they depart

from those authors with regard to the distinction between

parallelism and convergence.

(1) HOMOLOGY -- Two things that are similar to each other by

some set of criteria are "homologous" relative to those

criteria if and only if they are each descended from a

series of ancestors to which they are also similar by those

criteria, culminating in a single common ancestor to which

they are also similar. Note that this definition could be

applied to separate structures within one individual (i.e.,

serial homology) as well as to separate individuals or taxa.

Some workers have defined hom9logy as transformation

(e.g., Hennig, 1966). In contrast, the definition adopted

here implies homology is the complement of transformation.

Homology is the remnant of similarity conserved during the

course of transformation of characters. Mostly the

distinction between these views of homology is a semantic

one that emphasizes different aspects of the same thing. "A

petal is a transformation of a leaf" emphasizes the

differences, and "a petal is homologous to a leaf" 32 emphasizes the similarities. In both cases common ancestry

is implicit. (2) HOMOPLASY -- Two things that are similar are said to be

"homoplastic" relative to the similarity criteria if any of

the ancestors of either object up to and including the

common ancestor was not similar by the stated criteria.

The term "analogy" refers to a type of similarity

criterion, a criterion of function or adaptive value,

depending on usage. Hence, "analogy" is homoplasy in the

context of function, and is a subset of all hCill0~lasy.

"Convergence" and "parallelism" have been distinguished

on the basis of many factors (Haas and Simpson, 1946),

including amount of morphological similarity of the

homoplasy, degree of relationship of the taxa (or rank), and

process underlying the origin of the homoplasy. It is not

useful to distinguish between convergence and parallelism on

the basis of degree of relationship because there is no

objective measure of that factor. Definitions based on rank

are subject to taxonomic biases. Definitions based on

degree of morphological divergence are subject to biases

introduced by uneven rates of evolution, in addition to

those due to the arbitrary measurement of degree of

divergence. Definitions based on the qualitative

differences in the similarity of the homoplastic traits

depend wholly on the arbitrary similarity criteria that are 33 used. Definitions based on the action of natural selection versus some other process (such as orthogenesis) are interesting but difficult to apply in practice, since rigorous demonstration of the historical action of processes such as selection has proven difficult (Ridley, 1983).

Hence convergence and parallelism are better left under the more general term homoplasy, without regard to any differences between the two.

"Reversal" is also a kind of homoplasy, defined as the homoplastic reacquisition of the plesiomorphic state.

Superficially it seems different from parallelism/ convergence as discussed above. However, once it is realized that "reacquisition" implies that the plesiomorphic state was originated twice, once somewhere among the group's ancestors, and once in the "reversal", the distinction is no longer so important.

HOMOPLASY IN RELATION TO EVOLUTION AND PHYLOGENY

Homoplasy and Orthogenesis -- In the late nineteenth and early twentieth centuries there was a major interest in the phenomenon of homoplasy in relation to mechanisms of evolution -- orthogenesis in particular -(see review in

Grehan and Ainsworth, 1985, for a favorable recent exposition of orthogenesis, or Haas and Simpson, 1946, which is much more critical). orthogenesis is a concept fraught

------34 with difficulties of definition. It may be defined as directional patterns of evolutionary change that arise from

something other than environmentally induced selective

pressures. In particular, the non-random, directional

production of variation was believed to play a role in the

genesis of directional evolution (Eimer, 1898: Bowler,

1979). Today the term is most often associated with

recurrent patterns of non-adaptive overspecialization like huge antlers in Irish Elk (Simpson, 1984:170), or with

Cope's Rule, the paleontological trend towards increasing

body size (Rensch, 1959:206). Originally, however, at least

part of the interest in orthogenesis derived from

observations of parallelism in the fossil record and among

extant taxa (e.g., Haacke, 1893, discussed in Grehan and

Ainsworth, 1985; Scott, 1891: Buckman, 1901). Osborne (a

renowned orthogeneticist) proposed a law of "latent or

potential homoplasy" (1902) to explain parallelisms in

closely related taxa. Although not always couched in terms

of homoplasy per se, investigations of parallel "trends" and

"tendencies" were indeed investigations of homoplasy. The

ubiquity of such patterns in the context of a diverse array

of environmental conditions may have been what led some to

reject the Darwinian explanation in favor of orthogenetic

mechanisms. 35

Orthogenesis as an explanation of recurring patterns of evolution ultimately failed because no plausible mechanism was postulated and because the view of natural selection developing from the Modern synthesists was deemed a sufficient explanation for observed patterns (Wright,

1978:496ff.). In particular, the concept of directional mutation as a force in evolution (one possible orthogenetic mechanism) was rejected outright by the proponents of the

Modern Synthesis. The dogma ever since has been that mutation itself is random in its effects, and therefore only selection (or drift or migration) can ever directionally alter gene frequencies. Recently, however, the renewal of interest in developmental constraints has suggested that the effects of random mutations may be constrained along certain axes of variation (see review in Maynard smith, et. aI,

1985). The phylogenetic history of a group must in some

fashion constrain its future evolutionary trajectory by

imposing certain restrictions on what types of mutants are possible. It is reasonable to expect the probability of change in any character (including the chance of its

origination) to change over time (Riedl, 1978; Donoghue,

1989). This may result either from (1) changes in the

developmental system that alter the probability that random

nucleotide substitutions yield a given effect, or (2)

changes in the fitness effects of mutants such that the

--_._------36

probability that a mutant will ultimately be fixed is

altered. Regardless of whether this phenomenon is termed

"orthogenesis" or not, it deserves investigation by

evolutionary biologists, who have in the main ignored the

patterns that promulgated the idea of orthogenesis, as well

as orthogenesis itself.

Homoplasy and Tendencies The impact of orthogenetic

thinking on systematics was powerful and long-lasting.

Recurring patterns of parallelism seen in particular groups

convinced many systematists that "tendencies" represent

characteristics of the groups themselves, regardless of the

sporadic nature of such tendencies with respect to included

taxa. Although orthogenetic mechanisms such as directional

mutation were often cited as underlying such tendencies

(e.g., Cronquist, 1968), even systematists who rejected

orthogenesis were willing to use parallelisms as evidence of

relationship. Indeed, Simpson's definition of parallelism

implies that its occurrence stems from the close

relationship of the taxa involved; i.e., parallelism arises

because of something in common in the progenitor taxa.

,Evolutionary systematists have almost uniformly embraced the

idea that parallelism is, in some cases, an indication of

relationship (Simpson, 1961; Mayr, 1969; Cronquist, 1968;

Throckmorton, 1965; see reviews in Patterson, 1982, Cantino,

1985; Saether, 1986; Gosliner and Ghiselin, 1984; Stevens, 37

FIG. l. Illustration of homoplastic tendencies. Slack lines indicate the origin of a trait on that internode. A. Parallel origin within small clade, hence "apomorphic" tendency. S. Parallel origin at base of large divergen~ clade, hence "plesiomorphicll tendency. As defined in t!1is paper, tendencies might also include reversals on same ~ree. 38 1986). There is little dispute that certain characters arise repeatedly in some groups and not in others (e.g., Rensch, 1959). Saether (1983), who refers to such tendencies as "underlying synapomorphies" (after Tuomikoski, 1967), or "unique inside parallelism" (after Brundin, 1976), cites three morphological features of the female genitalia of Chironomid midges. Each character arises at least once' in each subfamily of the semifamily Chironomoinae, but is absent in the rest of the semifamily and in other dipterans. Cantino (1982) notes the recurrence of a gynobasic style in the angiosperm order Lamiales and in the family Boraginaceae (along with its absence in other taxa), and uses it as evidence of relationship of the two groups (although it is ultimately outweighed by more numerous tendencies that suggest alternative relationships). Presumably similar genetic and ontogenetic makeups of closely related taxa lead to similar sequences of evolutionary change under the influence of selection. This idea is originally due to Darwin (1872) and has been modified little over the years (see Haas and Simpson, 1946). Many cladists have rejected the utility of parallelism in reconstructing phylogeny (Hennig, 1966; Wiley, 1981; Eldredge and Cracraft, 1980; Patterson, 1982), although a few have favored it (Brundin, 1972, 1976; Saether 1979, 1983, 1986; Cantino 1982, 1985). Two serious problems are 39 apparent (see Fig. 1). Given that a derived character state has arisen independently in each of two clades, X and Y, can it be assumed that X and Yare more closely related to each other than to some other clade, Z, lacking the derived state (as the proponents would have us believe)? The first criticism is simple: there is no way of knowing whether or not taxon Z has the "tendency" but has merely failed to exhibit it as yet, and hence there is no firm evidence that

X and Y share some underlying derived state that is absent from Z (Rasmussen, 1983). The second criticism was noted by cantino himself and then refuted (Cantino, 1985). The putative genetic or ontogenetic similarity that might give rise to repeated parallelisms will often be plesiomorphic similarity, and the groups that exhibit these tendencies will then be paraphyletic. Hence it is impossible to reconstruct the nested hierarchy of monophyletic taxa of a group from plesiomorphic tendencies. cantino's response to this criticism is that rates of evolution may be roughly uniform and hence similar taxa will be monophyletic more often than not. This is a strange attitude from a cladist, and is not supported by the current body of cladistic analyses of real groups which suggests little uniformity in rates within phylogenies.

Despite a profusion of theoretical papers on the validity of applying apomorphic tendencies to phylogeny

------40 reconstruction, neither camp has convincingly refuted the other. Patterns of localized homoplasy appear to exist for some characters, and many systematists will continue to want to use them in their work. The arguments against the utility of tendencies to construct relationships is irrefutable on logical grounds, but even if the evidence of relationship is not good in theory, it might serve the useful purpose of provoking further study of critical taxa. A better way to test the validity of tendencies might be to examine patterns of homoplasy in an actual phylogenetic analysis, in which there is a sufficient degree of repeatability among characters that some robust inferences may emerge. The remainder of this thesis focuses on an analysis of topological patterns of homoplasy in a large phylogenetic study of North American Astragalus (Fabaceae). The aim is to determine whether or not homoplasy tends to be clustered within subgroups in a statistically significant manner, and hence to show evidence of "tendencies". The goal of this work is not to provide a method of using tendencies to reconstruct phylogenies. It aims instead at providing rigorous statistical evidence for or against the.notion that there are significant homoplastic tendencies in character evolution. 41

GENERAL PATTERNS OF HOMOPLASY

Some relevant general features of homoplasy have already been elucidated by comparison of patterns in sets of

cladistic reconstructions (Archie, 1985; see Appendix 1). A basic standard for comparison in these works is the

"consistency index" (CI; Kluge and Farris, 1969). CI is a

real number bounded by 0 and 1, which indicates the overall

amount of homoplasy in a phylogeny. In a data set with no

homoplasy (perfectly "consistent") the CI is 1.0, and it

declines towards 0 with increasing homoplasy. Archie

(1985), in a limited analysis of 7 morphometric studies,

observed that CI was lower in studies with more terminal

taxa. He also suggested that CI was positively correlated

with the number of characters included in the study.

Sanderson and Donoghue (see Appendix 1) compared CI's

in 60 recently published cladistic analyses, taken from

among several major groups of o~ganisms (plants, animals,

fungi), types of data (morphological, allozyme, nucleotide

and protein sequence, chemosystematic), and taxonomic rank

(species level to class/phylum). A very strong effect of

number of taxa was observed, with homoplasy higher in larger

data sets (i.e., the average level per character, see Fig.

2). No other variate exerted a significant influence (see Appendix 1). 42

• .-• - A a • •• 1.° 0 • • ..• - • • • 6 • •••• ". - •• • • • • • • :..l • I • • • • .4 • • • " • • • • • • • • .2

0 0 10 20 30 4.0 ~O 60 -~- 0 • • 1 • 8 -.2 I.'

-.4

-.S -:..l

=: -.9

-1

• -1.4+-____ o --~------~------10 20 40 :0

Number of Taxo FIG. 2. Effect.of number of taxa on amount of homoplasy. Data from 60 published cladistic analyses of various sizes (see Appendix 1). Homoplasy is measured by the consistency index (see text), which is one in the absence of homoplasy and declines with increases in homoplasy. CI values untransformed in A, log-transformed in B. 43

Th~ relationship between number of taxa and level of homoplasy is so robust in the face of tremendous variability in details of the cladistic studies examined that it suggests the presence of either a) an egregious artifact, or b) a real evolutionary pattern. We have considered a number of possible artifacts of taxonomic practice and entailed by algorithms used in phylogeny reconstruction and have rejected them all (see Appendix 1). On the other hand, the pattern is consistent with a simple evolutionary model in which the probability of character state change is proportional to the number of branches (internodes) on a tree, subject to the constraint that only a finite number of states are possible in a given character. Clearly, if such a constraint were not imposed, every character state change might be to a new state, and no homoplasy would result. Of course, this "constraint" may reflect the limits of resolution of the systematist more than it does some feature of evolution. Yet evolution clearly imposes constraints in the number of states of a character; nucleotide sequences have exactly four states, for example.

This model of evolution, if accurate, could be seen as contrary to the hypothesis that homopl·asy is localized in subtrees of a phylogeny. However, the relationship between homoplasy and number of taxa could actually be explained by a model of random homoplasy such as that above, modified

------44 slightly by the constraint that homoplasy is topologically restricted. The ol~ly assumptions necessary in that case are that different characters are restricted to subtrees of different size and/or phylogenetic position (more or less

randomly). The observed dependency of CIon number of taxa would then follow.

It is important to note that homoplasy studied by

comparison of cladistic analyses is a particular subset of

homoplasy. It is homoplasy among character states that the

systematist believed initially to be homologous because they

satisfied a particular set of similarity criteria. One

would expect that if the scope of a taxonomic analysis were

expanded to include more distantly related taxa, then

eventually fewer homoplastic changes would be detected

because they would not pass those initial similarity

criteria and would be scored as separate states. Hence a

lower limit on homoplasy may eventually be found that

depends on the resolution of the similarity criteria

employed. Our study did not address the effects of

expansion of study size within the same taxonomic group

because very few such analyses have been published. Even

so, at every taxonomic level, systematists are apparently

choosing characters that pass initial homology criteria in

disparate taxa, and yet are ultimately discovered to be homoplastic. 45

The relationship between extensive homoplasy and large taxa suggests that large taxa could be an important focus for the study of homoplasy. Moreover, in large taxa replicated occurrences of particular kinds of homoplasy may be discovered, that might be overlooked in studies of smaller taxa. The danger in studies of large taxa is that similarity criteria may be relaxed or less rigorously applied than in smaller studies by the same systematist.

The homoplasy is no more or less "real" in such cases, but it may be important to employ roughly the same standards of similarity across a range of numbers of taxa, because changes in the character "resolution" will affect the detected levels of homoplasy and patterns of tendencies observed (see discussion in Chapter Three) .

------46

CHAPTER TWO

PHYLOGENETIC ANALYSIS OF NORTH AMERICAN ASTRAGALUS

INTRODUCTION

Overview of the Genus -- The number of species in Astragalus may exceed 2000, making it the largest genus of flowering

plants according to some authors (Airy Shaw, 1973; Polhill,

1981). It is distributed primarily in arid regions of the

Northern Hemisphere, but is also found in the Andes of South

America and very sparsely in East Africa (Agnew, 1974).

Three principal centers of diversity are the Irano-Turkish

region, the Himalayas, and the western United States,

particularly the Great Basin. Edaphic specialization,

coupled with limited dispersal ability, have produced

population structures conducive to rampant local

differentiation and geographic speciation, particularly in

the physiographically varied regions where it attains its

highest diversity (Karron, 1987a,b, 1989; Karron, et. al.,

1988). Compared to related genera such as Oxytropis

(Barneby, 1952), geographic ranges in Astragalus are

strongly skewed towards narrow endemics (note the large

number of species on threatened and endangered lists in the 47 western u.s.: Welsh, 1979; Mozingo & Williams, 1980; Meinke, 1982). Relatively few, mostly weedy species, are widespread.

The handful of "cosmopolitan" species found in both Asia and

North America are Arcto-boreal plants of northern North

America and high elevations in the Rocky Mountains; they are completely absent from middle or low elevations in the western u.s. and Mexico, and none are found in South America

(Barneby, 1964).

Astragalus is morphologically diverse, especially in vegetative morphology and the structure of the fruit. Most species are perennial herbs, although major groups in both the Old and New World are annuals. Secondary growth, when present, is usually confined to a woody caudex at the base of the stems. Some Old World groups are "sub-shrubs" and exhibit a more extensive woody region at the base of the stems. None are trees. The vegl'':ative morphology has been highly modified in many groups. Plants vary from caulescent, tall, and robust, to essentially acaulescent, dwarf, and tufted. The imparipinnate leaf has been modified variously, including reduction to phyllodes, to large unifoliate leaflets, and in some taxa to a spine produced through induration of the petiole or rachis (mostly in the

Old World) .

The legume is quite variable and has provided many taxonomically useful characters. It is often tardily

- --"-~------"------48 dehiscent along one or both sutures, either while persisting on the raceme or after abscission. It ranges in size from

just under 3 mm. in some species with only one or two pair~ of ovules (e.g., b. didyrnocarpus) to well over 50 mm. in species with "bladdery" pods in which the mesocarp is membranous and the entire fruit greatly inflated (e.g., b. megacarpus, see data in Appendix 5). Other variations include pods with leathery or fleshy valves, as well as more "typical" legumes which are laterally compressed with papery valves. The pod may be sessile on the receptacle or stipitate; in some the pod is elevated on an elongation of the receptacle (the gynophore). Flowers are less variable than the fruit but still provide useful phylogenetic information. They are papilionoid in structure with interlocking keel and wing auricles, and 9 + 1 arrangement of filaments. There is great variation in size of the flowers, ranging from less than 5 rom to over 40 rom, and some useful variation is found in the degree of curvature of the petals when viewed from the side. The shape of the calyx varies from broadly campanulate to deeply cylindric. The inflorescence is less variable in the New World than in the Old World where it has heavily influenced taxonomic treatments. It is uniformly an axillary raceme, but the length of the raceme axis, and other finer details 49 are variable. other morphological features are discussed in more detail in Appendix 2.

Unfortunately, the size of the genus has evidently dissuaded many botanists from undertaking broad surveys of

cryptic characters, such as anatomy, development, or biochemistry (with the exception of some agricultural

research on selenium to1eranc~ and nitrotoxin accumulation).

Cytological surveys, on the other hand, have provided much

useful information. Species in the Old World are uniformly

2n=16 (or multiples), as are all other genera in the tribe

Ga1egeae (except Gue1denstadtia). In the New World, most

species fall into an aneuploid series running from 2n = 22

to 30. New World species with 2n=16 are arcto-borea1

species also found in Asia, or relatives (Barneby, 1964;

reviewed in Spe11enberg, 1976). This split between old and

New World taxa is one of the most fascinating puzzles in the

genus, as it does not correspond to any clear morphological

change. In general, polyploidy is very rare in the New

World (with perhaps a half-dozen cases, see Spe11enberg,

1976).' No cytological features other than chromosome number

have been investigated. Little molecular work has been

undertaken to date, although species examined show a loss in

the chloroplast DNA inverted repeat -- a mutation that

characterizes a large group of legume tribes including

Ga1egeae (Lavin, et. a1., in press). 50

Monophyly of Genus and Relationships -- Astragalus is

imbedded in the papilionoid tribe Galegeae which, in its

modern restricted sense, comprises 20 genera (Polhill,

1981). with the exception of Astragalus and oxytropis all

genera are endemic to the Old World (principally Sino­

Himalayan Asia). The tribe itself is probably paraphyletic

since it lacks any clear apomorphies that are not also found

in related tribes. In fact it may be at the phylogenetic

base of a major radiation of herbaceous perennial legumes

into Northern Temperate regions (see Fig. 3). Galegeae and

several related tribes (Trifolieae, Vicieae, Hedysareae,

among others) all share several apomorphies including

epulvinate leaf bases, closed vascular system (Dormer, 1945,

1946), stipules adnate to the petiole (Polhill, 1981), and

the loss of a chloroplast DNA inversion (Palmer, et.al.,

1987; Lavin, et.al, in press). While many of the other

tribes in this clade are characterized by clear apomorphies

(e.g., toothed leaflets in Trifolieae), no such character is

apparent for Galegeae.

Unfortunately, just as the monophyly of Galegeae is

suspect, so too is the monophyly of Astragalus -- and for

much the same reason. No clear apomorphies are unique to

it. One putative apomorphy of the genus, the presence of a

dorsally-derived longitudinal septum in the pod, is far from 51

Astragalus

ASIA. EUROPE NORTH AMERICA S. AM. ( 1600-2000 spp.) (400 sPP.) (100 spp.)

2n=32.30 •....

Septum tendency?

2n= 16 p1es1omorphy

?

loss of cpDNA Inversion herbaceous vegetatfve syndrome

FIG. 3. Hypothesized phylogenetic relationships of Astragalus. Arrows indicate migrations of small groups into the geographical regions shown. Number of branches is schematic and does not indicate actual number of major clades. Question marks indicate unknown synapomorphies needed to delimit clades. 52

universal in the group. Since no cladistic analysis of the

tribe, or genus, has ever been attempted, it is not certain

. that this apomorphic condition evolved at the base of the

genus or merely within some subclade nested within the

group. Moreover, even if it was derived at the base of the

clade, it has been lost independently in a number of

lineages, and it is possible that related genera such as

oxytropis are derived from such groups. Many of the other

genera are more specialized than some primitive members of

Astragalus, and they may be monophyletic segregates of a

paraphyletic Astragalus. This has a strong bearing on

discussions of why Astragalus is so speciose, since its

"depauperate" relatives may in fact be part of the "genus".

A number of large groups within Astragalus are

undoubtedly monophyletic. In fact, cytological evidence

suggests that the entire New World group is monophyletic,

since it possesses chromosome numbers unique in the tribe

(2n = 22-30). This hypothesis is important to the cladistic

work presented below, and therefore warrants further

discussion. The hypothesis is borne out only if it can be

assumed that the aneuploid 2n=22 series is a derived

feature, and not plesiomorphic. It has generally been

assumed that Astragalus originated in the old World

(Barneby, 1964), based on biogeographic considerations such

as the geographic distributions of related genera, and 53 centers of diversity (1500 - 2000 spp. in the Old World vs. perhaps 600 in the New World). The reasons are rarely discussed explicitly (e.g., Barneby, 1964), and it is now known that traditional evidence for centers of origin of taxa, such as centers of diversity, is not compelling (Brown and Gibson, 1983). What is needed is information about outgroups of Astragalus in order to determine the primitive area of origin. Assuming the outgroups of Astragalus are found among the other genera of Galegeae, which all have

2n=16 (except Gueldenstadtia), then the basal condition in

Astragalus is likely 2n=16. It is remotely possible, however, that the rest of Galegeae is derived from at or near the base of Astragalus. In that case outgroups from related tribes would have to be considered (Maddison, et. al., 1984). Chromosome numbers of the tribe Milletieae, which is a putative outgroup of Galegeae, include 2n=20 and

22. Some authors (Turner and Fearing, 1958) have used this similarity to propose a polyphyletic origin of Astragalus

(Spellenberg, 1976). However, the lack of 2n=20 in any known Astragalus species may suggest that the origin of the

22-30 series was independent even if Milletieae is an outgroup. Although single-character systematics is never too reliable, it does raise the specter that Astragalus originated in the New World and subsequently radiated dramatically in the Old World. However, until firm evidence 54 is uncovered suggesting that other genera of Galegeae are not the appropriate outgroups, the body of evidence still favors the traditional view that Astragalus originated in the Old World.

Even granting that the 2n=22 series is derived, it is important to determine how many times it was derived.

Barneby (1964) suggested that multiple invasions of the New

World have occurred. Hence, in his view certain New World groups are most closely related to various Asiatic groups.

The arcto-boreal, 2n=16, species all can be assigned fairly unambiguously to Old World sections of the genus based on morphological similarities. The cytological data provides additional support. On the other hand, the idea that subgroups of the aneuploid New World species (which have never been assigned to Old World groups based on morphology) are most closely related to various Old World sections would require a hypothesis of repeated parallel. acquisition of the

same apomorphic chromosome number, which seems dubious in the absence of support from other characters. Hennig's

auxiliary principle (Hennig, 1966:121), to assume

synapomorphy unless parallelism is demonstrated, must be

invoked at this point. 55

METHODS

Scope of Present Study -- The present cladistic analysis is restricted to the aneuploid Astragalus endemic to North

America. The relationships of North American taxa to those in South America and the implications of the exclusion of the latter from this study are considered in the Discussion below. Eight sections comprising 13 species with chromosome numbers in multiples of 2n=16 are excluded (see Table 1). section Minerales ought to be included; although neither species has been examined cytologically, both are similar to section Polares with 2n=24. Unfortunately, I have been unable to examine specimens of this group. The monotypic section Sesamei (s. wrightii) supposedly belongs to an Old World group on account of its morphology (Barneby, 1964;

Spellenberg, 1976), but has 2n=22. It too has been unavailable for study. A. williamsii of section Hemiphaca is excluded on biogeographic and morphological grounds although its chromosome number is unknown (Barneby, 1964).

Finally, sections orophaca and Sericoleuci, with seven species, are not considered. These species probably have

2n=22 (only one species examined to date), but are morphologically very isolated among the North American taxa, and have often been removed to the segregate genus Orophaca

(e.g. recently by Isely, 1983b; see also Roberts, 1977).

They likely form a very derived monophyletic group (based on 56 synapomorphies of trifoliate leaves, medifixed hairs, enlarged, connate, scarious stipules) of uncertain origin, and it seems better to await a larger scale study with broad outgroup coverage before these taxa are investigated in depth. with these exclusions the scope of the present study includes some 390 species. This is too great a number given the constraints of currently available computer algorithms for phylogeny reconstruction, which practically are limited to less than 150 taxa. In practice few cladistic studies are published with more than 25 taxa (see Appendix 1). An alternative to working at the specific level is to use sections as terminal taxa. In two attempts prior to this one, I undertook such an approach, using Barneby's 92 sections. Heterogeneity within sections necessitated that many be split into sub-sections or individual species, ultimately yielding some 150 terminal taxa. Actual coding of these taxa was not satisfying, however, because of extensive polymorphism in the resulting groups, recurring questions about their monophyly, and the practical uncertainties encountered when trying to score not one but several species simultaneously. Unfortunately,- Astragalus species are separated from one another by small morphological gaps and it is very difficult to recognize

larger monophyletic groups that might be used as terminal 57

TABLE 1. North American species of Astragalus used, based on classification of Barneby, 1964. Underlined taxa are section names with abbreviations in parentheses. Species used in cladistic analysis of North America are indicated by (*); taxa used in analysis of section Argophylli Gray (Appendix 3) are denoted by (+). This analysis includes 113 taxa, including a total of seven varieties of three species (A. trichopodus, magdalanae, lentiginosus). Only varieties used in this analysis are listed. Taxa described since Barneby (1964) are indicated by (%) and are listed in Appendix 6. Authorities for all taxa can be found in Barneby (1964), and Appendix 6. New combinations are not included below, but are found in Appendix 6. penellianus GENISTOIDEI(GN) PHACA tioides *miser umbellatus altus *convallarius americanus hidalgensis diversifolius purpusii ASTRAGALUS pueblae LONCHOCARPI(LNC) alpinus *recurvus *titanophilus leptaleus *rusbyi *xiphoides longissimus cronquistii MINERALES egglestonii pinonis molybdenus *micranthus aequalis shultziorum esperanzae *episcopis *cobrensis *lancearius OROBOIDEI microcymbus duschesnensis eucosmus %cenorrhynchus nidularius robbinsii harrisonii SCALARES col toni HEMIPHRAGMIUM scalaris ripleyi aboriginurn schrnollae cottoni TIOPSIDEI *lonchocarpus scopulorurn hamiltoni HEMIPHACA williamsii SCYTOCARPI(SCYl DRABELLAE (DRB) *flexuosus *spatulatus STRIGULOSI(STR} pictiformis *chloodes guatarnalensis proxirnus drabelliformis legionensis fucatus simplicifolius hintoni subcinereus *detritalis tolucanus *wingatanus strigulosus *gracilis lyonetii *coriaceous SOLITARII(SO} radicans *hallii *alvordensis potosinus *puniceus solitarius zacatecanus castetteri applegatii jaliscensis %shevockii regiornontanus 58

TABLE 1 (CON' D)

CAMPTOPODES (CM) tegateroides NEVINIANICNV) *camptopus %chuskanus *traskiae %tiehrnii nevinnii. COLLINI (CO) *collinus OCREATI(OC) NEONIX(NX) *curvicarpus *flavus rnulfordae gibbsii sophoroides johannis-howellii *rnoencoppensis *peckii TWEEDYANI %yoder-williarnsii tweedyi ALBULI (AL) *albulus ATRATI(ATR) PODOSCLEROCARPI(PO) *atratus *sclerocarpus *salrnonis *sinuatus BISULCATI (BI) *speirocarpus *bisulcatus *racernosus QUINQUEFLORI(Q) CUSICKIANI(CU) *brandegei *filipes OOCALYCES(O) quinqueflorus inversus *oocalycis *californicus INYOENSES(IN) *cusickii PECTINATI(PCT) *inyoensis *whitneyi *pectinatus sterilis nelsonianus JAEGERIANI cerarnicus grayi jaegerianus %knightii *toanus linifolius PACHYPODES ERVOIDEI (EV) rafaelensis pachypus *tenellus saurinus vexilliflexus osterhouti DRUMMONDIANI bourgovii drurnmondii *rnicrocystis WOODRUFFIANI(W) kentrophyta *woodruffi MALACI (MAL) *cibarius POLARES MISELLI(MIS) ensiforrnis bodini congdoni rnalacoides polaris agnicidus *minthorniae urnbraticus vallaris GYNOPHORARIA paysonii cirnae nutzotinensis *ervoides *rnalacus sinaloae charnaerneniscus JEJUNI carrninis jejunis straturensis PRUNIFORMES(PU) lirnnocharis *howelli *accidens *arthuri HUMISTRATICHM) *rnisellus *humistratus oniciforrnis sesquiflorus toquirnanus rnicrornerius 59

TABLE 1 (CON I D)

REVENTI-ARRECTI(RA) mokiacensis +inflexus *reventus *beathii parryi sheldoni *praelongus +castaneiformis *adanus pattersoni +chamaeluce terminal is sabulosus +amphioxys riparius musimonurn arrectus NEGLECTI +cymboides atropubescens neglectus *+missouriensis *remotus +accurnbens *eremeticus ULIGINOSI ani sus scaphoides canadensis %+holrngreniorum orcuttianus oreganus %piutensis obscurus %anserinus %ackermanii ONOBRYCHOIDEI %stocksyi adsurgens %+piscator MICHAUXIANI %phoenix Michauxii HYPOGLOTTOIDEI agrestis NUDI DESPERATI(DS) serenoi naturitensis monumental is CONJUNCTI(CJ) ARGOPHYLLI(ARG) deterior *conjunctus *+argophyllus *desperatus hoodianus +.donis %cottamii reventiformis marianus %equisolensis leibergi desereticus callithrix LAYNEAE HESPERONIX(HX) *+tephrodes layneae *bolanderi +iodopetalus +shortianus MOLLISSIMI CML) BICRISTATI(BC) +cyaneus *mollissimus webberi columbianus helleri *bicristatus tidestromii %+nutriosensis %errterae +waterfallii %hartrnanii +feensis PORRECTI neornexicanus GIGANTEI(G) porrectus +uncialis *giganteus +musiniensis AMPULLARII (AM) +loanus SARCOCARPICSR) *arnpullarius +newberryi *crassicarpus +eurekensis gyps odes PACHYPHYLLUS +coccineus plattensis asclepiadoides +purshii sanguineus +leucolohus %bibullatus PREUSSIANICPRL) sUbvestitus preussii +funereus TENNESSEENSES *eastwoodae +utahensis tennesseensis *crotalariae nudisiliquus 60

TABLE 1 (CON'D) MEGACARPI thurberi VILLOSI (V) mega carpus wardi *distortus oophorus aquilonius soxmaniorum beckwithii cerussatus obcordatus endopterus *villosus LUTOSI serpens lutosus pubentissimus LOTI FLORI pardalinus lotiflorus PTEROCARPICPT) *sabulonum *casei nutans PANAMINTENSES pterocarpus gilmani panamintensis tetrapterus *insularis geyeri HUMILLIMICHU) ANEMOPHILI aridus troglodytus anemophilus wetherlli *gilensis miguelensis sparsiflorus sileceus harbisonii diaphanus cremnophylax *hornii humillimus DENSIFOLIICDNS) %moranii %wittmannii *nutallii %sanctorum *pomonensis %beatleyi LEPTOCARPI CLEP) curtipes *nothoxys pycnostachus pringlei oxyphysus bryantii DIPHYSICDI) *gentry.:i. TRICHOPODICTRI) lentiginosus *arizonicl.i5 asymmetricus *v. diphysus albens trichopodus *v. palans mohavensis *v. trichop. *v. wilsonii parvus *v. phoxus iodanthus hypoxylus %oxyphysopsis pseudiodanthus tricarinatus bernardinus INFLATI CINF) MONOENSES(MN) emoryanus *douglasii *monoensis nyensis macrodon ravenii acutirostris oocarpus *pulsiferae francisquitensis deanei perianus tener gruinus pauperculus palmeri CYSTIELLA rattani prorifer striatiflorus clarianus *idrietorum breweri piscinus CIRCUMDATI coahuilae fastidius circumdatus *nuttallianus magdalanae leptocarpus *v. magd. PLATYTROPIDESCPLT) lindheimeri *v. peirson. *platytropis allochrous amnis-amissi SUCCUMBENTES(SC) wootoni *amblytropis *succumbens 61

TABLE 1 (CON'D)

SCAPOSI SERICOLEUCI calycosus sericoleucus aretoides GREGGIANI tridactylicus greggii barrii

MICRANTHI(MC) OROPHACA *hartwegi gilviflorus vaccarum proimanthus goldmani hyalinus *clevelandi oxyrrhynchus NEW SPP. NOT PLACED %daleae IN SECTIONS iselyi HYPOLEUCI(HY) danicus *hypoleucus ast::::agal inus atwoodii CHAETODONTES(CHT) debequaeus spaldingii bryogenes tyghensis *lyalli lemmoni *caricinus lentiformis andersonii *sepultipes *austinae BRAUNTONIANI(BR) *brauntonii

DIPHACI diphacus

REFLEXI reflexus SCUTANEICSU) scutaneus *brazoensis MICROLOBIUMCMCL) *gambellianus *didymocarpus

SESAMEI wrightii

------62

TABLE 2. List of section abbreviations. sections are from Barneby (1964) as listed in Table 1.

AB Albuli SO Solitarii AL Alvordenses SR Sarcocarpi AM Ampullarii STR strigulosi ARG Argophylli SC Succumbentes ATR Atrati SU scutanei BI Bisulcati TRI Trichopodi BR Brauntoniani V Villosi CHT Chaetodontes W Woodruffiani CJ Conjuncti CM Camptodes CO Collini CU cusickiani DI Diphysi DNS Densifolii DS Desperati DRB Drabellae EV Ervoidei G Gigantei GN Genistoidei HM Humistrati HU Humillimi HY Hypoleuci HX Hesperonix IN Inyoenses INF Inflati LEP Leptocarpi LNC Lonchocarpi MAL Malaci MC Micranthi MCL Microlobium MIS Miselli ML Mollissimi MN Moenenses NV Neviniani NX Neonix o Ooco:lyces OC Ocreat.i PCT Pectinati PLT Platytropides PO Podosclerocarpi PRL Preussiani PT Pterocarpi PU Pruniformes Q Ouingueflori RA Reventi-Arrecti SCY Scytocarpi 63

TABLE 3. Characters used in cladistic analysis of North American Astragalus. See Appendix 2 for further discussion of each character and description of character states.

VEGETATIVE MORPHOLOGY FRUIT

1. Point of renewal 32. Fruit orientation 2. Duration of root 33. Valve texture 3. Type of root 34. Valve surface 4. Growth pattern of stem" 35. Valve pigmentation 5. Robustness of stem 36. Presence of spongy­ 6. Presence of developed pithy mesocarp petiole 37. Presence of pulpy 7. Leaf orientation filaments in locule 8. Presence of oblanceolate 38. septum phyllodes 39. Persistence 9. Presence of pectinate leaf 40. Number of ovules 10. Leaflet number 41. Dehiscence 11. Leaflet size 42. Stipe 12. Leaflet folding 43. Gynophore 13. Leaflet venation 44. Fruit attachment 14. Leaflet shape 45. Incurvature 15. Terminal leaflet attachment 46. Beak 16. Leaflet apex 47. Inflation 17. Spinescence 48. Dorsal surface 18. Stipules 49. Fruit cross-section 19. Stipule size 50. Bisulcate fruits 20. Stipule shape 51. Fruit length 21. Presence of black stipules 52. Fruit breadth 22. Presence of scarious 53. Fruit inclusion in stipules calyx 23. Pubescence type 54. Fruit drying green 24. Pubescence quantity

BIOCHEMISTRY & CYTOLOGY INFLORESCENCE & FLOWER 55. Nitrotoxin 25. Inflorescence type accumulation 26. Length of pedicel 56. Selenium accumulation 27. Calyx and corolla 57. Chromosome number conformation 28. Length of calyx teeth 29. Petal coloration 30. Length of wing petal 31. Presence of irregularly graduated petals

------64 taxa, except in a few cases in which clear synapomorphies are present (e.g., sections Pectinati, Drabellae).

The approach used in the present work is to score a subset of individual species chosen selectively from as broad a taxonomic range as possible (Table 1). This

approach is not often advocated in cladistic studies because

the omission of taxa can obscure phylogenetic relationships

of included taxa, although the precise circumstances are not

understood (Doyle and Donoghue, 1987; Gauthier, Kluge, and

Rowe, 1988; Donoghue, et. al., ms.). In the absence of

homoplasy, shared-derived characters will indicate the true

relationships of those taxa scored regardless of missing (or

extinct) taxa, but homoplasy can introduce uncertainty. In

addition, missing outgroups can alter polarity assessments

needed to properly root a phylogeny. However, all

systematic and phylogenetic work is faced with a sampling

problem at some point. It is usually impossible to

completely sample the individuals in even one species, for

example. Moreover, extinction of taxa can always bias

results but its effects are entirely unknowable unless

fossils are uncovered. Also, since in practice few

consistent standards exist to guide taxonomists who name

taxa, the mere act of sampling all "taxa" recognized by some

taxonomist may not guarantee that the appropriate

evolutionary units have been found. Finally, systematists 65 who work at higher taxonomic levels, such as families, or sections of a genus, often only have information about isolated. members of that group and assume that the information applies to the group as a whole. This is particularly true for characters that are difficult, time­ consuming, or expensive to study, such as developmental anatomy, or DNA sequences. This does not excuse the present work from an obligation to be as thorough as is practically possible, but it does suggest its limitations are wholly shared by most, if not all, phylogenetic work.

Characters and Taxa Used -- In all, 113 taxa (including 109

species and 4 additional varieties; see Table 1) were scored

for a set of 57 unordered binary and mUltistate characters

(112 binary character equivalents), including 24 from the vegetative morphology, 7 from the inflorescence and flower,

23 from the fruit, 2 from biochemistry, and 1 from cytology.

with the exception of the three biochemical and cytological

characters, and two morphological characters, which were

scored from data in the literature, all characters were

scored from examination of herbarium specimens -- generally.

at least 10 specimens per species, usually many more. Most

of the work was based on the large collections of Astragalus

maintained at ARIZ, NY, RSA, and UC, supplemented by loaned material from NY and TEX. Appendix 2 should be consulted 66

TABLE 4. Data matrix for 113 species of North America Astragalus, and 57 characters. Polymorphism indicated by multiple entries in the same column. For characters 30, 51, and 52, the two numbers entered for each taxon refer to the upper and lower bounds of the set of polymorphic states.

1 1 1 1 1 1 1 111 2 2 2 2 2 2 2 2 223 1 2 3 4 567 8 9 0 1 2 3 4 5 678 901 234 5 6 7 8 9 0 accidens 000 2 0 1 0 0 0 0 1 1 o 0 0 1 0 0 1 0·0 0 1 0 6 0 0 0 1 6 1 8 adanus o 0 0 2 0 0 0 000 1 1 o 0 0 0 0 0 1 0 0 0 1 0 401 018 1 1 1 5 9 albulus 000 2 0 1 0 0 0 1 1 ? o 1 0 ? 0 2 1 0 0 0 2 1 4 0 1 015 2 8 alvordensis 1 0 0 2 110 001 o 0 0 0 010 0 0 000 1 1 6 0 2 0 0 ? 1 3 ambylotropis 000 2 ? o 0 0 0 ? 0 0 0 0 020 ? 0 0 0 001 3 0 2 o 0 ? 1 6 1 ampullarius 100 1 000 0 0 1 1 1 o 0 0 0 0 1 1 1 0 110 600 004 112 1 9 argophyllus 000 0 ? o 0 0 0 ? 0 0 0 0 0 2 001 0 0 ? 016010 o 9 1 1 arizonicus 000 1 0 ? 0 0 0 1 1 001 0 0 0 0 0 0 0 021 3 0 200 4 121 1 2 4 5 arthuri 000 0 101 000 1 o 0 0 0 0 0 0 0 0 0 0 1 0 5 0 201 7 1 1 8 atratus 000 0 1 0 1 0 0 1 o 0 0 1 0 2 0 0 0 0 0 o 1 0 3 1 2 014 1 1 1 135 austinae 000 0 ? o 0 0 0 1 0 0 0 0 0 2 01100 1 0 1 3 0 2 104 1 6 beathii 000 2 0 0 0 0 0 1 1 1 0 0 0 0 0 0 100 0 1 0 6 0 1 0 0 7 2 1 9 bicristatus 000 2 0 1 0 001 1 0 0 0 0 0 0 1 100 0 1 0 6 0 1 017 1 1 8 bisulcatus 0002010 0 0 0 1 0 0 0 0 0 0 1 1 0 0 0 1 0 1 000 1 4 1 2 7 bolanderi 000 2 010 0 0 0 1 0 0 0 0 0 0 1 1 0 0 0 1 0 601 015 1 1 2 8 brandegei 010 1 100 001 1 0 0 0 0 0 0 0 0 000 1 0 5 1 3 1 1 ? 1 brauntonii 000 2 0 1 0 0 0 0 1 1 0 0 0 0 0 0 10000 1 1 0 2 1 0 4 2 2 6 brazoensis ? 2 0 1 1 1 0 001 1 1 000 1 0010001 030 2 1 0 1 2 1 2 . californicus 000 201 0 0 0 1 1 0 0 0 0 0 0 1 0 0 0 0 1 1 401 0 1 7 1 1 1 camptodes 100 1 0 1 0 0 0 1 1 0 0 0 0 1 0 0 0 000 1 1 4 0 2 0 1 5 1 6 9 caricinus 000 1 100 0 011 00002 0 1 1 0 0 1 0 1 2 0 3 1 1 1 2 4 2 case! 1002100 0 0 110 0 1 0 0 0 0 0 0 0 0 1 0 4 0 1 006 1 1 6 8 67

TABLE 4 (con I d) 1111111 1 1 1 2 222 2 2 2 222 3 ------,---123 4 567 8 9 0 1 2------3 4 5 678 9 0 1 2 3 4 5 678 9 0 chloodes o 0 0 0 ? ? 1 1 0 2 ? ? 0 ? 120 2 1 0 0 1 2 140 310 1 5 cibarius o 001 0 0 0 0 011 1 0 0 0 0 0 0 1 1 0 010 3 0 100 7 1 2 1 1 6 1 8 clevelandi o 002 0 1 0 0 0 0 1 1 000 0 0 010 001 0 1 0 2 0 1 0 1 5 3 4 cobrensis 1 0 011 0 0 0 0 1 o 1 0 0 010 2 0 0 0 000 4 0 3 0 0 2 1 1 1 6 3 collinus o 0 0 2 0 1 0 0 011 0 0 0 010 0 1 0 0 010 1 o ? 015 1 ' 1 8 conjunctus o 000 0 0 1 0 0 0 0 0 0 1 0 0 0 1 1 0 0 1 1 0 4 0 101 7 11 1 1 1 1 8 convallarius 1 0 0 210 1 0 0 2 1 ? 0 1 1 ? 0 100 0 0 1 0 4 1 301 3 1 7 5 coriacious 1 0 0 2 0 0 0 0 011 0 0 1 0 0 0 1 0 0 0 0 1 040 200 6 1 1 8 crassicarpus o 0 0 1 0 1 o 0 0 0 1 1 0 0 0 0 0 0 1 o 0 0 1 0 6 0 100 6 2 1 9 crotalarieae o 1 0 1 0 0 0 0 0 1 110 0 0 0 0 0 1 0 0 010 4 0 1 0 0 9 2 2 1 6 curvicarpus 000 2 0 1 0 0 0 1 1 1 0 0 0 1 0 o 1 0 0 0 1 0 5 0 o 017 1 1 9 cusickii 1 0 0 2 000 0 0 111 0 1 0 0 0 1 0 0 0 010 4 1 0 016 1 1 5 7 6 desperatus o 0 0 010 0 0 0 1 1 0 0 0 020 010 0 011 4 0 200 4 1 6 6 detritalis 000 0 ? 0 1 1 021 001 020 2 1 0 012 1 3 0 2 1 0 5 1 8 didymocarpus ? 2 011 1 0 0 0 1 o 0 0 0 0 1 0 0 0 000 1 130 200 o 2 1 1 1 distortus 001 2 1 0 0 0 0 1 0 1 0 o 0 0 0 0 0 o 0 010 4 0 100 3 1 1 1 6 2 5 douglassii o 0 0 1 0 0 0 0 0 0 1 1 0 0 0 0 0 000 00104 0 301 3 2 1 1 6 5 episcopus 10020 ? 0 0 0 2 ? ? ? 1 1 ? 0 0 0 0 0 o 1 0 4 020 1 3 1 8 eremeticus o 0 0 2 0 0 0 0 0 1 1 1 0 1 0 0 0 0 1 0 0 010 4 0 100 6 2 1 1 2 9 ervoides 000 210 0 0 001 10000 000 0 0 0 1 050 3 0 1 2 1 7 filipes o 0 0 2 0 0 000 1 1 ? 010 0 0 1 0 0 0 0 1 0 4 1 201 5 1 1 6 flavus o 000 0 0 1 0 011 ? 0 1 0 0 021 0 0 0 2 1 2 0 201 5 1 1 4 7 6 flexuosus 1 0 0 2 0 1 0 0 0 1 1 0 0 0 0 0 0 1 1 0 0 010 4 020 0 4 112 5 1 6 gambellianus ? 2 0 111 0 0 0 1 0 0 0 0 0 1 000 0 001 0 3 0 2 0 0 0 2 1 1 2 68

TABLE 4 (con I d) 1 1 1 1 111 111 2 2 2 222 2 2 2 2 3 1234567------,------8 901 2 3 4 5 678 9 0 1 2 345 678 9 0 gentryi o 1 0 1 1 0 0 0 0 1 010 0 0 0 0 0 0 0 0 0 1 a 3 a 2 0 0 4 2 2 1 5 giganteus o 0 010 0 0 0 0 0 2 ? 0 0 0 2 0 020 0 0 011 0 1 0 0 ? 1 gilensis 000 0 ? 0 0 0 0 ? 0 0 0 0 0 2 0 0 0 0 0 0 2 1 302 1 0 2 114 gracilis 1 0 021 000 0 1 1 0 0 1 0 0 0 1 0 0 0 0 104 030 0 0 5 hallii 1 0 0 2 0'1 0 0 0 0 000 0 0 1 0 1 0 0 0 0 1 0 4 120 1 5 1 1 8 hartwegi o 0 020 1 0 0 0 0 1 0 0 0 000 000 0 0 1 0 502 0 1 o 1 1 2 3 3 hornii o 0 020 1 0 0 a 0 1 1 0 0 0 0 0 000 0 0 1 0 3 0 2 0 1 4 1 5 howelli o 0 011 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 0 4 0 2 0 1 4 211 7 humistratus 00021 1 0 0 0 1 1 0 0 0 0 2 0 210 0 0 2 1 6 0 2 1 0 ? 1 hypoleucus 1 0 0 1 1 1 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 205 o 2 0 0 1 2 1 1 6 3 1 3 idrietorum o 0 010 1 0 0 0 1 0 0 0 0 0 1 0 0 ? 0 0 1 004 o 3 0 0 3 1 2 1 1 6 6 insularis ? 2 0 1 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 106 03000 1 5 inyoensis o 0 021 1 000 110 0 0 0 0 0 000 0 0 114 o 2 0 0 3 4 lancearius 1 0 0 2 0 ? 0 0 0 210 ? 1 1 ? 0 0 0 0 0 0 1 0 4 020 1 4 1 6 lentig-diphysus 0 0 0 1 0 0 0 0 0 1 1 1 0 0 00001 0 0 0 1 0 621 0 0 8 2 1 lentig-palans o 0 010 0 0 0 0 1 1 1 0 0 0 ? 0 0 1 0 0 0 1 0 4 0 1 008 2 1 lentig-w11son1i 0 0 0 1 0 0 0 0 011 1 0 0 0 o 0 0 1 0 0 0 1 0 6 0 1 0 1 8 2 1 1 lonchocarpus 1 0 020 0 0 0 021 0 o 1 10000 0 0 0 1 1 4 1 201 4 8 lyall1 o 0 020 0 0 0 0 1 1 0 0 0 0 2 0 0 1 0 0 0 0 1 203111 1 1 4 3 magdalanae-magd o 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 0 1 3 0 2 0 0 4 1 2 1 1 4 7 6 magdalanae-peirs 0 001 0 0 0 0 0 1 0 0 0 0 1 0'0 0 0 0 0 1 0 1 6 0 200 4 121 7 malacus o 0 0 1 0 0 0 0 0 1 1 0 0 0 0 0 0 0200101 4 0 1 0 0 7 2 6 8 micranthus o 0 021 1 0 0 0 0 1 0 0 0 0 o 0 0 1 0 001 03030 1 2 1 1 1 4 microcystis 000201000 1 1 0 0 0 0 o 0 1 1 1 1 0 1 1 6 0 200 ? 1 1 1 2 1 minthorniae 0001000 0 0 1 1 0 0 0 0 o 0 020 0 1 0 1 4 010 0 7 8

.------69

TABLE 4 (con I d) 111 1 1 1 1 1 1 1 222 2 2 222 2 2 3 123 4 567 8 9 0 1 2 3 4 5 6 7 8 901 2 3 4 567 8 9 0 ------misellus 0 0 0 1 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 1 0 6 0 3 0 1 2 2 1 1 6 miser 0 0 0 1 1 0 1 0 0 1 1 ? 0 0 0 0 0 1 1 0 0 0 1 0 4 1 3 0 1 2 1 2 6 4 missouriensis 0 0 0 0 ? 0 0 0 0 ? 1 0 0 0 0 2 0 0 1 0 0 0 2 1 6 0 1 0 0 8 9 moencoppensis 0 0 0 1 1 ? 0 0 0 1 1 ? 0 1 0 ? 0 1 0 0 0 0 1 0 4 0 2 1 0 3 2 1 1 6 mollissimus 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 1 2 0 1 0 0 8 1 1 2 3 9 monoensis 1 0 0 2 1 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 6 0 2 0 1 3 1 21 1 3 6 nothoxys 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 3 0 2 0 0 3 1 1 1 4 6 nutallianus-aust ? 2 0 1 1 0 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 1 0 3 0 2 0 0 1 2 2 1 3 5 nutallii 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 1 0 1 1 0 0 0 1 0 1 0 0 0 1 5 2 1 1 1 7 oocalycis 0 0 0 1 0 1 0 0 0 0 1 0 0 1 0 0 0 1 1 0 0 0 0 0 1 0 ? 0 ? ? 1 peckii 0 0 0 1 1 0 0 0 0 1 0 0 0 1 1 2 1 2 0 0 0 0 0 1 6 0 3 1 1 0 1 4 pectinatus 0 0 0 2 0 1 0 0 1 1 ? 1 0 1 1 2 0 1 1 0 0 0 1 0 4 1 1 0 1 7 6 8 platytropis 0 0 0 0 ? 0 0 0 0 ? 0 0 0 0 0 2 0 ? 0 0 0 0 0 1 3 0 2 0 0 2 1 6 1 7 pomonensis 0 0 0 1 0 1 0 0 0 0 1 1 1 0 0 0 0 0 1 0 0 0 1 0 1 0 0 0 1 6 2 8 praelongus 0 0 0 2 0 0 0 0 0 0 1 1 0 0 0 0 0 0 1 a 0 0 1 0 5 0 1 0 1 8 1 2 1 6 9 pulsiferae 1 0 0 2 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 6 0 2 1 1 ? 213 puniceus 0 0 0 1 0 1 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 1 1 6 0 1 0 0 7 2 1 1 2 9 racemosus 0 0 0 2 0 1 0 0 0 0 1 0 0 0 0 0 0 1 1 0 0 0 1 0 1 0 0 1 1 6 111 8 recurvus 1 0 0 1 1 0 0 0 0 1 0 1 0 1 0 1 0 1 0 0 0 0 1 0 4 0 3 0 1 2 213 remotus 0 0 0 2 1 0 0 0 0 1 1 0 0 1 0 0 0 0 0 0 0 0 1 0 4 0 2 0 1 5 7 reventus 0 0 0 1 0 0 1 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 1 0 4 0 1 0 1 6 21 1 5 2 9 rusbyi 0 0 0 2 1 0 0 0 0 0 1 1 0 0 0 0 0 1 0 0 0 0 1 0 5 0 3 0 1 2 1 1 6 sabulonum ? 1 0 1 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 6 0 3 0 0 ? 2 1 salmonis 0 0 0 0 1 0 ? 0 0 1 0 0 0 0 0 2 0 0 0 0 0 0 1 0 4 0 2 0 1 0 1 1 6 7 sclerocarpus 0 0 0 2 0 1 0 0 0 1 1 ? 0 1 0 0 0 0 0 0 0 0 1 1 4 0 1 0 0 6 6 2 7

------_._------70

TABLE 4 (con' d) 111 1 1 1 1 1 1 1 2 2 2 2 222 2 223 1234567 8 9 012 3 4 5 678 901 2 3 4 5 6 7 8 9 0 sepulUpes 1 0 0 2 0 0 0 0 0 110 0 0 0 0 0 1 1 0 0 0 0 1 ? 0 210 3 1 7 sinuatus o 0 0 2 0 1 0 0 0 101 0 0 0 1 0 a 0 0 0 011 ? 0 101 6 2 8 spathulatus o 0 0 0 ? ? 1 1 0 2 ? ? 0 ? 120 2 1 0 0 1 2 1 3 0 310 1 5 speirocarpus o 0 0 201 0 0 0 1 1 1 0 0 0 1 0 0 0 0 0 0 1 1 6 0 101 7 8 succUIr\l:)ens a a 010 1 0 a 0 111 a a 0 0 o ? 1 0 0 a 0 1 6 0 2 016 2 2 9 tenellus o 0 0 2 0 1 0 o 0 1 1 0 0 0 0 0 0 1 1 1 1 01060 201 2 1 112 5 tephrodes o 0 0 0 ? o 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 00040 100 9 112 116 titanophilus 1 0 021 0 0 0 0 110 010 0 0 o 1 0 0 010 5 1 301 4 5 toanus o 0 0 2 0 ? o 012 ? ? ? 1 1 ? 0 1 1 0 0 0 1 0 4 0 0 0 1 2 6 8 traskiae o 0 010 0 0 0 0 011 0 0 0 0 0 0 0 000 o 1 3 0 201 4 4 7 trichopodus-tri- 0 0 a 2 0 1 0 0 0 a 1 1 0 a 0 0 0 0 0 0 0 0 1 0 1 0 001 6 1 1 1 8 trichopodus-phox 0 0 0 2 0 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 a 1 0 1 0 0 a 1 6 1 1 1 8 villosus o 1 101 0 0 0 0 1 1 1 000 0 0 0 1 0 0 001 3 0 211 5 1 6 whitneyi 1 0 0 1 1 0 0 001 1 0 000 o 0 1 0 0 0 010 6 o 0 0 0 6 1 1 7 wingatanus 10011 ? ? 0 0 010 o 0 0 0 0 1 1 0 0 0 1 0 4 0 300 1 2 1 4 woodruffi o 0 020 ? 0 002 ? ? 011 0 0 0 2 0 0 011 3 0 210 6 148 xiphoides 1 0 020 0 0 0 0 2 ? ? 0 ? 1 ? 0 0 1 0 0 010 5 1 301 3 1 5

3 3 333 3 3 334 4 4 444 4 4 4 4 555 5 5 555 Node------,------1 234 5 6 7 8 9 0 123 4 5 6 7 8 9 0 1 2 3 4 567 accidens o 1 2 ? ? ? ? 1 2 1 0 2 0 ? ? 0 ? ? 0 0 ? ? 0 0 0 0 ? adanus o 0 200 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 520 0 0 0 ? 1 1 4 albulus o 010 0 0 0 011 0 0 000 0 0 0 0 0 5 2 0 0 ? 1 ? 2 4 alvordensis 1 1 1 0 0 0 0 0 1 ? 1 2 0 ? 2 0 0 0 1 0 6' 1 0 0 1 0 ? 7 3 ambylotropis 1 ? 1 0 1 0 0 ? ? ? 0 0 0 0 0 010 ? 0 7 7 0 0 0 0 ? ampullarius o 0 1 0 2 0 0 011 0 2 0 0 0 010 ? 0 5 600 ? 0 ? 6

"'.' - ,---,------71

TABLE 4 (con' d)

3 3 3 3 3 3 3 3 3 4 4 444 444 4 455 5 5 555 5 Node------1 2 3 4 567 8 901 234 567 8 9 0 1 2 3 4 567 argophyllus o 0 2 0 000 000 0 0 ? 0 110 0 0 0 6 4 0 0 000 7 5 ar1zonicus o 0 1 0 0 0 0 1 0 1 0 0 0 0 100 1 1 0 6 2 0 0 001 7 3 arthuri o 1 1 0 0 0 0 1 1 1 0 2 0 1 0 0 0 1 1 0 8 2 0 0 001 1 9 3 atratus o 1 2 0 0 0 0 0 110 0 0 0 020 1 0 0 5 2 0 0 0 0 ? 1 1 1 7 4 aust1nae o 0 1 0 0 0 0 1 0 2 0 000 ? 0 0 0 101 1 1 0 0 0 ? 1 1 beathii o 0 2 0 001 0 1 0 0 0 0 0 0 0 1 0 0 o 7 500 ? 1 1 1 8 6 bicr1status o 0 2 1 000 0 2 1 1 201 100 2 0 0 7 4 0 0 001 8 5 bisulcatus o 1 1 0 0 0 0 011 0 100 000 0 0 1 4 000 111 152 bolander1 o 0 1 0 0 0 0 1 2 1 0 1 0 0 1 111 1 0 3 500 000 6 brandegei o 1 1 0 000 111 0 000 000 0 1 0 1 ? 0 0 ? 0 1 1 brauntonii o 1 1 0 000 0 1 2 0 0 0 0 1 0 0 1 o 0 2 2 0 0 001 1 3 brazoensis o 1 1 0 000 2 0 2 1 0 1 ? 100 1 0 ? 120 0 ? 0 0 californicus o 1 1 0 0 0 0 0 1 1 1 200 000 010 8300001 9 4 camptodes o 1 1 0 0 0 0 0 1 1 1 2 0 1 200 2 1 0 6 300 101 9 4 caricinus o 1 1 0 0 0 0 1 1 2 0 0 0 0 000 0 1 0 1 0 1 0 0 0 ? 132 casei o 1 2 1 1 0 0 0 0 0 0 0 0 0 010 0 008 300 000 9 4 chloodes o 0 1 0 000 0 1 2 0 000 000 0 1 021 0 0 001 4 cibarius o 0 2 0 1 0 ? 0 2 0 0 0 0 ? 100 1 007 4 0 010 0 1 1 9 7 clevelandi o 1 1 0 000 1 0 2 0 0 0 0 100 1 1 0 1 0 0 0 002 2 1 cobrensis o 1 1 0 0 0 ? 1 1 1 0 0 0 0 000 1 0 0 4 300 1 0 ? collinus 1 ? 1 0 0 0 0 0 1 1 1 2 0 0 0 000 1 0 6 2 0 0 101 7 conjunctus o 020 000 011 0 0 0 0 0 0 0 0 0 0 6 3 0 0 001 7 4 convallarius 1 0 1 0 0 0 1.0 1 1 2 0 0 1 0 0 0 0 1 0 7 2 0 0 0 0 ? 193 1 coriacious o 0 2 0 0 1 0 0 1 1 0 0 0 0 000 0 0 0 5 4 0 0 100 2 7 5 crassicarpus o 0 2 0 0 1 0 2 0 o 0 0 0 0 0 0 1 0 006 6 0 0 0 0 0 1 1 8 8 crotalarieae o 0 2 0 0 0 1 010 0 0 0 0 001 o 007 600 111 1 8 7 72

TABLE 4 (con' d)

3 3 3 3 3 3 3 3 3 4 444 4 4 4 4 4 4 555 5 5 555 Node 1 2 3 4 5 6 7 8 9 0 1 2 3 4 567 8 9 0 1 2 3 4 567 ------curv1carpus o 1 1 0 0 0 0 0 1 1 120 120 0 2 1 0 ? ? 0 0 1 0 ? cus1cki1 o 1 0 0 0 0 0 0 1 111 0 1 0 0 1 0 ? 0 7 7 0 0 0 0 0 9 8 desperatus o 1 0 0 1 0 0 0 0 1 000 01000003201 000 1 5 4 detr1talis o 0 1 0 0 0 0 0 1 1 0 0 0 010 o 1 1 0 8 1 0 0 0 0 ? 3 didymocarpus o 0 1 0 0 0 0 2 0 2 ? 0 011 0 0 1 0 0 ? ? 0 1 ? 0 1 2 distortus ? 010 0 0 0 0 0 1 0 0 0 ? 1 0 0 1 1 0 5 1 0 0 ? 0 2 6 4 douglassii 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 ? 0 7 7 0 1 0 0 0 1 1 9 9 episcopus 1 1 1 0 0 0 ? 0 1 1 0 0 0 0 0 0 001 o 7 3 0 1 000 8 5 eremeticus o 010 0 0 0 1 1 1 020 1 0 2 0 1 1 0 6 4 0 0 ? 0 1 7 5 ervo1des o 110 0 0 0 1 1 1 0 0 0 ? 1 0 0 1 1 0 4 1 o 0 004 filipes o 1 1 0 0 0 0 0 1 1 120 000 0 0 1 0 7 4 o 1 000 8 5 1 flavus o 0 1 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 1 3 1 1 0 ? 1 1 4 2 flexuosus o 010 0 0 ? 011 000 010 0 0 0 0 4 2 0 0 1 0 0 1 2 6 3 gambellianus o 1 1 0 0 0 020 2 ? 0 0 11001 000 101 ? 0 ? gentry1 o 0 1 0 0 0 0 1 0 1 000 000 0 1 1 0 4 1 0 0 0 0 ? 1 6 2 giganteus o 0 1 0 0 0 0 110 000 010 0 1 0 0 5 4 0 0 000 2 1 1 6 6 gilensis o 0 1 0 0 0 0 0 1 2 0 0 0 0 0 000 1 0 1 2 0 0 ? 0 0 2 3 gracilis o 1 2 1 0 0 1 0 1 2 0 0 0 012 000 0 3 2 0 0 1 0 ? hallii o 1 1 0 0 0 1 0 1 1 0 1 0 0 0 0 1 0 0 0 6 4 0 010 0 6 hartwegi o 1 1 0 0 0 0 1 1 200 0 010 0 1 1 0 3 0 0 0 0 0 ? 4 1 hornii o 010 0 0 0 0 1 1 0 0 0 0 0 ? 1 0 ? 0 4 5 0 1 ? 0 ? 5 6 howelli o 1 1 0 0 0 0 1 1 1 020 100 0 1 1 0 620 000 ? 1 7 4 htnnistratus o 010 0 0 ? 0 0 1 000 ? 0 0 0 0 003 3 ? 0 ? 0 1 1 4 5 hypoleucus o 1 1 0 0 0 0 1 0 1 o 0 0·0 0 0 0 1 1 0 3 1 0 0 0 D 3 2 1 4 3 idr1etorum o 0 0 0 0 0 0 0 0 1 000 100 1 0 ? 0 ? ? o 1 ? 0 0 insular1s o 0 0 0 0 0 0 0 0 1 0 0 0 100 1 0 ? 0 6 6 o 1 ? 0 ? 1 7 inyoensis 1 1 1 0 0 0 0 1 1 1 0 2 0 011010032 o 000 ? 5 4

-_. - ----_.-._------_. 73

TABLE 4 (con' d) 3 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4 4 4 5 5 5 5 5 555 Node ------,------1 2 3 4 5 6 7 8 9 0 123 4 5 6 7 8 901 2 3 4 567 lancearius 1 1 1 0 0 0 ? 0 1 1 0 0 000 0 0 0 1 0 6 4 0 100 ? 8 5 lentig-diphysus 0 0 1 0 2 0 0 1 1 0 0 0 0 0 1 1 1 1 0 0 5 5 0 0 ? 0 ? 7 7 lentig-palans o 0 1 0 0 0 011 0 000 0 1 0 0 0 0 0 5 3 0 0 ? 0 0 7 5 1 lentig-wilsonii o 0 1 000 0 1 1 0 000 0 1 o 0 0 0 0 ? ? 0 0 ? 0 ? lonchocarpus o 1 1 0 0 ? 0 ? 1 1 1 2 0 1 0 o 0 0 0 0 8 3 0 0 ? 0 0 2 1 9 4 lyalli o 110 0 0 0 1 1 2 000 0 0 0 o 0 1 021 1 0 0 0 ? 132 magdalanae-magd 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 ? 0 6 6 0 1 ? 0 ? 1 7 7 magdalanae-peirs 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 ? 0 6 6 0 1 ? 0 ? 1 8 7 malacus o 1 2 0 0 0 ? 1 2 ? 0 0 0 ? 1 0 0 1 1 0 9 5 0 0 0 0 ? micranthus o 1 1 0 0 0 ? 1 1 1 0 0 1 1 0 0 0 1 0 0 3 2 0 000 ? 153 microcystis o 110 ? 0 0 0 0 2 000 0 0 010 ? 0 ? ? 0 ? ? 0 ? minthorniae o 1 2 0 0 0 ? 1 2 0 000 ? 0 o 0 010 5 4 0 000 ? 1 8 5 misellus o 1 1 0 0 0 0 111 010 1 0 001 1 o 4 2 0 000 ? 6 miser o 1 1 0 0 0 001 1 200 ? 0 0 0 0 1 051 0 000 0 7 3 missouriensis o 0 2 0 0 0 001 0 0 0 ? 0 0 0 0 2 1 0 6 4 0 0 000 7 6 moencoppensis o 0 0 0 0 0 0 0 1 2 0 0 000 o o ? 0 0 ? ? 1 0 ? 1 ? mollissimus o 0 1 0 0 0 0 1 0 0 0 0 0 0 1 o o 1 0 0 520 0 0 0 0 2 1 1 6 monoensis o ? 1 0 0 0 0 1 0 1 0 0 0 1 1 100 o 0 6 3 0 000 ? nothoxys o 0 1 0 0 0 0 1 0 1 0 0 0 0 1 001 1 062 0 000 3 7 3 nutallianus-aust 0 0 1 0 0 0 0 1 1 1 0 0 0 0 1 0 0 1 1 0 ? ? 01001 nutallii 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 ? 0 ? ? o 0 ? 0 0 oocalycis 0 ? 1 ? ? 0 ? 0 0 2 ? 0 1 7 0 0 0 1 0 1 ? ? 1 ? ? 1 1 1 peckii o 110 0 0 0 1 1 2 000 0 0 001 10010 000 ? 1 2 pectinatus o ? 2 0 0 0 0 0 1 1 0 0 0 0 0 0 1 200 420 001 0 6 5 platytropis 1 ? 1 0 1 0 0 ? ? ? 0 0 0 0 0 010 006 6 0 000 ? 7 8 pomonensis 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 010 ? 0 7 8 0 0 ? 0 0 8 praelongus o 020 0 0 1 0 1 0 0 0 0 0 0 010 0 0 7 600 ? 1 1 8 8 pulsiferae o ? 0 0 0 0 0 0 0 2 0 0 0 1 0 010 ? 0 4 5 0 000 1 1 1 6 6 . 74

TABLE 4 (con I d) 3 3 333 3 3 3 3 4 4 4 4 4 444 4 4 555 5 5 555 Node 1 2 345 678 9 0 1 2 3 4 567 8 9 0 1 2 3 4 567 puniceus o 1 100 0 0 0 1 100 0 0 1 0 1 1 ? 0 5 5 0 000 ? 6 racemosus o 1 100 0 0 0 1 101 0 0 000 1 0 0 7 200 ? 1 1 1 1 3 recurvus o 1 100 0 ? 1 1 1 0 0 0 0 000 110 520 0 ? 0 ? 6 3 remotus o 0 100 0 011 102 0 1 000 1 1 06200 ? 0 ? 7 3 r'eventus o 0 200 0 0 0 100 0 o 0 0 0 0 1 0 o 520 0 0 0 ? 1 7 4 rusbyi o 1 100 0 ? 1 1 1 0 2 o 1 000 1 1 05201 ? 0 ? 2 7 3 sabulonum o 0 100 010 010 o 0 1 100 0 ? 05401 ? 0 1 1 1 salmonis o 1 200 0 0 0 110 0 0 0 020 1 0 o 5 2 o 0 0 0 ? 1 1 1 6 4 sclerocarpus o 0 210 0 ? 0 1 012 0 1 111 2 0 075 o 0 000 8 6 sepultipes o 0 100 0 0 1 0 1 0 0 0 0 0 0 0 0 1 0 4 2 1 0 0 0 ? 1 1 5 3 sinuatus o 1 100 0 0 0 1 112 0 ? 200 010 6 2 o 0 0 0 0 7 3 spathulatus o 0 100 0 0 0 120 0 0 0 000 0 1 0 1 0 o 0 001 2 2 speirocarpus o ? 1 1 0 0 0 0 1 1 ? 2 0 ? 200 2 1 072 o 0 000 9 4 succumbens o 0 100 001 ? ? ? 0 0 0 100 110 ? ? 0 0 001 tenellus o 1 101 0 0 0 1 202 0 0 000 2 1 0 3 1 0 1 ? 0 1 5 3 tephrodes o 0 200 0 0 0 0 000 0 0 110 0 0 0 6 3 0 0 000 7 5 titanophilus 1 1 100 010 110 0 0 0 000 000 ? ? 0 0 ? 0 ? 2 toanu5 o 0 200 0 0 0 1 1 0 0 0 0 000 2 0 0 5 4 0 0 ? 1 0 1 6 traskiae o 0 100 0 0 1 1 1 0 1 0 1 1 0 0 1 o 0 5 3 0 0 0 0 ? 2 6 4 trichopodus-tri- 0 1 0 0 0 0 0 0 1 1 0 2 0 1 0 0 1 0 1 0 7 6 0 1 000 9 8 trichopodus-phoK 0 1 0 0 0 0 0 0 1 1 0 2 0 0 0 0 0 0 1 0 5 3 0 1 0 0 ? 9 6 villosus 0 0 1 1 0 0 0 0 0 1 0 0 1 ? 1 0 0 0 1 0 ? ? 0 ? ? 0 ? 1 whitneyi 0 1 0 0 1 0 0 0 1 1 1 1 0 0 0 0 1 0 ? 0 7 7 0 0 0 0 0 1 8 8 wingatanus 0 1 1 0 1 0 ? 0 1 2 0 0 0 ? 0 0 0 0 1 0 4 3 0 1 ? 0 0 1 5 woodruffi 0 0 1 0 0 0 0 0 1 1 1 0 0 0 0 0 0 2 1 0 ? ? 0 0 0 1 ? xiphoides 1 1 1 0 0 0 ? 0 1 1 0 0 0 1 0 0 0 0 0 0 ? ? 0 1 0 0 ? 75 for detailed discussion of the characters used and the coding schemes employed. Taxa were selected to be representative of the range of character variation found in North America. Of the 92 sections recognized by Barneby (1964; see Table 1), 12 were initially excluded for reasons discussed above (under "Scope of Study"). Species were sampled from among the remaining

80 sections, such that 52 sections in all were included

(Table 2). Of the 26 sections not included, 23 were monotypic, one contained two species, and one contained three species. In every case taxa were excluded because close relatives were apparent in related sections. sampling was aimed at adequately representing morphological variation

in the characters of Table 3, rather than complete sampling of Barneby's classification. Some sections were sampled more heavily than others if higher levels of variation were evident (e.g., Scytocarpi). On the other hand, some large

sections were not sampled as completely as one might expect

(e.g. Argophylli, Inflati) because in those groups

speciation has proceeded faster than divergence and related

species are often very similar to one another. The

resulting data matrix (Table 4) is large and unwieldy and

little can be gleaned from it without the aid of a 'computer. 76

Polymorphism -- The presence of more than one state of a character within a single taxon is polymorphism (Arnold, 1981). It is often argued that polymorphism indicates that the terminal taxon itself should be broken into smaller units. However, this assumes that ultimately a terminal taxon can be found that is not polymorphic for any character (Gauthier, Estes, and de Queiroz, 1988). At least in the data set compiled here, almost every terminal taxon is polymorphic for some character (Table 4). Indeed, this appears to be a general feature of characters, which is related to the way homoplasy is sporadically distributed among taxa and among taxonomic levels (Appendix 1). Even a character that is useful in delimiting higher taxa because of its invariance within most terminal taxa is often found to vary within some terminal taxon somewhere (Stebbins, 1974). Nonetheless, polymorphism certainly can be reduced by including smaller terminal units in the analysis, and that strategy has been employed in this study. In binary characters, polymorphism provides no phylogenetic information because the character-state optimization algorithms assign the terminal taxon whichever of the states is most parsimonious given the states in the rest of the tree. This is functionally equivalent to scoring taxa as "unknown," indicated by a question marks. In multi­ state characters, however, polymorphism that consists of 77 only a subset of the possible states can be informative.

For example, given a three-state character, a taxon exhibiting states 1 and 2 will be more parsimoniously placed with taxa exhibiting either a 1 or a 2 than it will with taxa exhibiting a o. scoring such a taxon as a question mark is incorrect because optimization may reconstruct a state (e.g., 0) that actually was never present in the terminal taxon. Polymorphism has therefore been explicitly included in the data set (Table 4) used in the present analysis, and given the large number of mUltistate characters it is likely to have an important effect on the results. until very recently, polymorphism has been ignored in cladistic studies, despite the availability of relevant algorithms in computer packages such as MacClade (version

2.87e; Maddison and Maddison, 1988), and the latest versions of PAUP (version 3.0; Swofford, 1989).

When coding polymorphism, however, one must be aware that consistency indices may be calculated either with or without the extra steps that polymorphisms entail in the terminal taxa. Thus, CI's may seem atrociously low when polymorphisms are explicitly coded. In the data set of

Table 4, question marks refer to truly unknown information.

Correlation of Characters -- Suites of logically or functionally correlated characters may produce misleading 78 reconstructions because parsimony algorithms treat them as separate traits, allowing them to outweigh truly independent characters. Confidence in a tree or a clade arises only as a result of an assumption of independence of the evidence brought to bear on it (Felsenstein, 1985; Sanderson, 1989).

Detection of correlation is extremely difficult, of course, particularly in the absence of detailed functional or developmental information on the characters. Ironically, characters that share the same distribution among taxa, and therefore strengthen support for particular hypotheses of relationships, are also the likeliest candidates to be correlated. Apomorphies that support a clade but later undergo reversals in different taxa, may provide somewhat better evidence of independence, though even then the characters may be independent only within the clade.

In the process of selecting and coding characters for analysis in this study, the problem of correlation arose frequently. The general solution employed was to combine correlated characters into a single larger mUltistate character. For example, epidermal hairs in As~ragalus may be attached to the surface of the plant at the base of the hair (basifixed) or towards the middle (medifixed).

Originally, this trait was coded separately from another character dealing with the orientation of hairs: appressed, spreading, or erect. However, it is clear both on logical 79 grounds and from an examination of the distribution of these characters among groups of Astragalus, that medifixed hairs are always appressed hairs by virtue of their structure. Hence, one mUltistate character was employed having the states, "medifixed-appressed", "basifixed-appressed," and "basifixed-spreading." Such recoding has important implications for the evidence the characters reveal. The character was left unordered, which meant that "hasifixed," which previously was a single state that might have united all taxa having basifixed hairs, now did not necessarily unite such taxa. In general, increasing the unordered states of a character redu:es the evidence of relationships contained in that character, but such conservatism may be warranted in many cases in which correlation is likely. In some instances evidence of correlation did not become apparent until after initial reconstructions were examined. An important example concerned inflorescence characters. Repeatedly a group of species united by a diverse array of characters (morphological, cytological, and biochemical) appeared to be polyphyletic because a suite of inflorescence characters were uniting some members with other groups. In particular, a syndrome of long-dense inflorescences with pendulous flowers (originally two traits) appeared to be responsible. On examination of the entire data matrix, it was found that all taxa with long- 80 dense inflorescences had declined flowers. This suggested the operation of some concerted (perhaps pollination) syndrome, and the character was combined.

Quantitative Characters -- Many of the characters used in the analysis are most naturally measured on a meristic scale (number of ovules or leaflets) or a continuous scale (lengths and widths of fruits). The development of cladistic methods of phylogenetic inference has fallen short where morphometric data are concerned (Felsenstein, 1988). Some cladists have even argued that continuous data are not valid cladistic characters (Pimentel and Riggins, 1987) and should not be used to reconstruct phylogeny. Apparently it is easier to reject an entire class of morphological data than it is to derive a usable method for incorporating it. This viewpoint is convenient but indefensible. Surely the fact that one systematist happens to view a character on a discrete scale while another systematist sees it on a continuous scale can ultimately have no bearing on the phylogenetic information contained in different types of characters. One cladistic approach to continuous data has been to seek statistically significant gaps that might indicate some underlying discrete process (Archie, 1985; Goldman, 1988). This is a stopgap approach which concedes that continuous 81

3.5

3 • ••• a ••• • -E 2.5 . .. E ".. . -....c 2 . .. • ~ . .. , .. '. • ...... '11 •• ~ 1.5 • • ~. I •• ... I • ",., ••• - • '='1:1 · .: i I· •• .~ ...... c •• .1 • ...... • ...... 5 • ..

0 1 1.5 2 2.5 3 3.5 4 log length (mm)

FIG. 4. Morphometric variation in fruit size and shape for 162 species of North. American Astragalus. Each point represents the mean of 5 specimens. Length is length from apex of fruit (not including style, if present) to base (not including stipe, if present); width is measured from adaxial face to abaxial face at the center. All data were log­ transformed. Note that all fruits are confined to a triangular region in· the lower half of the diagram. also note the lack of clear gaps' in morphospace. Raw data can be found in Appendix 5. 82 data must ultimately be put on a discrete scale. As this

approach is currently the only one at all favored among

cladists (Felsenstein, 1988), it is worthwhile justifying why it has not been used here. Practical reasons include

limitations on available time for sampling hundreds of

species. Much more important is the critical sensitivity of

gap-coding procedures to thoroughness of sampling. As

Archie (1985) pointed out, it is increasingly difficult to

find gaps between taxa as more intermediate taxa are

sampled. In Astragalus, the species used in this study have

thoroughly filled the character space available to them, and

gaps, if present, are not apparent (see Fig. 4). No doubt

statistical procedures could find gaps somewhere, but they

would likely be dependent on the vagaries of the species

sampled and the error rates of the within-species data.

In a preliminary morphometric analysis of some 28

species in Astragalus section Argophylli Gray, in which over

600 plants were scored for 22 meristic and continuous

traits, only 8 characters had significant gaps at the

resolution of 1/2 to 3/4 of one standard deviation. More

characters exhibited gaps when the resolution was increased

(i.e., when the cutoff level for between group variation was

reduced), but the significance of such gaps is questionable,

and this approach, when applied consistently to all 83 characters, resulted in characters with large numbers of states and hence little information about relationships.

Hence gap coding (and related methods) has not been used in this study. In some cases I have fallen back on arbitrarily defining cutoff points. Cutoff points were

selected so as to minimize as best as possible the number of taxa that would have ambiguous (polymorphic) scores because

of overlap in their distributions in two or more states.

This is an informal way of looking for dips or multimodality

in the frequency distribution, a procedure that \Vas

impractical in a study of this scope. Characters 10, 11,

19, 26,and 40 were treated in this way (see Appendix 2).

In the case of other characters which seemed to require more careful analysis, I devised a new technique,

"polymorphism overlap coding." Traits are scored as mUltistate characters from a through 9, where the integers

are scaled by the range of the character. The integer

corresponding to the sample mean for the taxon was

determined, and then all integer states that occurred in the

range of plus or minus one standard deviation were

determined (the choice of one s.d. is arbitrary). The taxon was coded polymorphically with the set of all integers found above. 84

Rooting -- It is now well-established that outgroup rooting is the best method for determining ancestral character states (Maddison, et. al., 1984). Unfortunately, all outgroup methods require knowledge of related taxa that is unavailable in the current study, because there has been little comprehensive systematic work on the 2000 Old World species of Astragalus. The last monograph on the entire group was published in 1868 (Bunge, 1868). More recently,

Gontsharov (1965) reviewed the Astragalus of the Soviet

Union, comprising some 900+ species, and Chamberlain and

Matthews (1970) surveyed those of Turkey, with over 800

species. Podlech and coworkers (e.g., Podlech, 1988) have published a series of monographs on Old World sections which

should eventually provide useful information, but these are

non-phylogenetic and would merely provide raw data for

cladistic analysis.

One approach to rooting the New World taxa would be to

assume that the outgroups to the New World taxa in the 2n=22

chromosome series can be found among the handful of arcto­ boreal species found in both hemispheres, such as Astragalus

alpinus or b. canadensis. Monographers of the genus in

North America have noted the morphological similarity

between cosmopolitan species and certain members of the

North American section strigulosi (Barneby, 1964; Rydberg,

1929: see his genus Astragalus and especially Atelophragma)

------~ .. -- ._._._-- 85 and have argued for the section's primitiveness. Unfortunately, these "Old World" taxa are themselves quite diverse morphologically, and since they represent exceedingly small samples from distinct, diverse Old World radiations, including them together in a simultaneous outgroup-ingroup cladogram (Maddison, et. al., 1984) is not likely to accurately reflect their relationships to each other or to the New World groups. Moreover, since many of these Old World species are found in the same mesic high­ elevation or high-latitude habitats, there has been extensive morphological convergence that tends to obscure the relationships among them. The same considerations preclude use of the outgroup sUbstitution method, which is most helpful in cases where the ingroup cladogram is fairly robust to changes in the outgroup (Donoghue and cantino, 1984). The cladograms presented in this analysis have been rooted according to established hypotheses about the origin of the North American taxa, which suggest an origin in section strigulosi and related groups. These hypotheses are based on three separate lines of evidence: (1) similarity of members of section Strigulosi to co~mopolitan taxa, (2) presence of high chromosome numbers (2n=26-30) in that section, which is in line with a hypothesis that a hypotetraploid on 2n=16 was primitive in the New World 86

(Spellenberg, 1976, and references therein), and (3) an unusual (for the genus) geographic distribution in Northern and Central Mexico, an area which has long been viewed as relictual for Astragalus species, serving as a Pleistocene refugium (Jones, 1923). I have arbitrarily selected Astragalus ervoides, as the basal taxon in the analysis. It is currently placed in section Miselli (though it may be better-placed among strigulosi), a close relative of strigulosi, and it has 2n=30 and a distribution in Northern Mexico. The cladograms presented below should be examined with the idea that many other nearby taxa might represent closer approximations to the actual root. An alternative hypothesis for the origin of the New World taxa deserves future examination. Vilkomerson (1943) suggested that the primitive chromosome number in the New World is 2n=24, perhaps derived from a hybridization event between 2n=8 and 2n=16, both common numbers in the Old World (Senn, 1938). Over 80 percent of species in North America have 2n=24 or 2n=22, and very few have higher numbers. There is a small group of species comprising sections Pol ares and Gynophoraria with 2n=24, coupled with a neo­ boreal distribution -- an unusual biogeographic pattern for New World Astragalus. To date these have not been given serious attention as possible ancestral groups in the New World. [In the cladograms presented below, section Polares 87 would appear closely related to section Humistrati in the villosi-Inflati group. Rerooting by these taxa would not disturb the monophyly of the large Homaloboid or Megaflorae radiations, but would make the Piptoloboid group paraphyletic, and the formerly basal paraphyletic group, including sections Miselli, Strigulosi, etc., monophyletic.]

Algorithms Employed -- The data were analyzed using Wagner parsimony as implemented on PAUP version 2.991 (Swofford,

1988). Alternative trees were explored using MacClade version 2.87e (Maddison and Maddison, 1988). The bulk of the search for minimal trees was carried out in PAUP using heuristic algorithms that sequentially join taxa together and then search among a set of rearrangements (llbranch­ swapsll) for more parsimonious solutions. Generally, PAUP options used were MULPARS (MAXTREE=250), HOLD=5,

SWAP=GLOBAL, ADDSEQ=CLOSEST. Several rearrangements of the order of data were input since this can affect the success of the search routines. Due to the size of the data matrix, each run required many hours of computer time. In total, about 150 hours of 8 Mhz. 68000 processor time were used in the search, and at least 75 million rearrangements were examined by the program. ervoides hartwegi h~oleucus clevelandi braZoensis gambelli anus ~ didYmocarpus brandegei - micranthus - mise-llus ~'d t'lj rusbYi Pl Pl H arthuri C"liG') howelli t-'1Il (1) tJ· remotus S Ul el"E'metious r-.,) 0 • salmonis ::s atratus tJ· peckii o ;:3:; s:: Pl oobrensis III l-I. ~ recurvus o anZonious rt Ii Ii tJ· nothoxys (1) rt distorl:us (1)'< gentryi III ~ nut alli anus-al Ii o s:: monoensis Hlt-' pulsifer-ae (1) platytropis .~ allbYlotropis (1)­ sabulonum ::s \0 CJQO insularis rtdP ict'ietorum ::r'-" magdalanae-l'Ia magdalanae-pe UlO \00 sepultipes Ul::S villosus III huni strat us III (1) rt ::s desperatus (1) III 9ilensis 'd s:: austinae III III ly,aUi carioinus C"rt Pl Ii detritalis III (1) spathulatus (1) (1) chloodes P- traskiae O o Hl suocumbens ::s iI'\Yoensis r-.,) brauntonii P-Ul mollfssillus p> 0 rt giganteus Pl (1) malacus ..c minthorniae o s:: lentfg-diphYsI.. HlP> I. ~ Ptrlfceus j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j III co malaous ..0 minthorniae o C lentfg-diphYsI" H1 Pl. I-' p~1oeus I-' bolanderf '-<: lent f g-wilsoni lentig-palans cfbarius raoemosus bisulcatus moenooppensi s oooallJois albulus flavus miser wingatanus tenellus mi croclJst is flexuosus hallii corfacious gracilis casei whitneyi cusfckii filipes lonchocarpus woodruffi episoopus lanoearius convallarlus tihnophilus xiphofdes. toanus pectinatus conjunotus re-ventus adanus speirocarpus sinuatus curvicarpus collfnus oalffornicus camptodK 1------1C alvordensis bicristatus sclerooarpus ampullarius hornfi accidens trlchopodus-pl trichopodus-tl douglassf; pomonensis nut allf i crassfoarpus praelongus: beathii orot alari eae missourfensfs argophYllus tephrodes

88

RESULTS

Overview of Trees Found -- Two sets of equally parsimonious trees were uncovered following several complete runs of the data matrix with different orders of input. One set had a length of 595 steps (CI=.17, not including steps due to polymorphism), and the other had a length of 596 steps

(CI=.17). In each case the limit of MAXTREES was reached and 250 equally parsimonious trees were produced.

Undoubtedly many more trees in each set remain to be uncovered. A 90% majority rule consensus tree of each set revealed that the trees within a set were all quite similar to one another: most clades appeared in more than 90% of the trees (Figs. 5,6). Only a few polychotomies were present on the 596 step tree, but a large polychotomy was present in one apical region of the 595 step tree. However, such a polychotomy should not be taken to mean that all possible resolutions of the polychotomy appear among the set of equally parsimonious trees.

In the discussion below I focus on the 596 step tree, instead of the minimal tree, because it embodies fewer unreasonable character-state changes. In particular, within the set of 595-step trees, many trees exhibit a large terminal clade that is diagnosed by a set of reversals to the condition at the base of the tree. The reversals

include the reacquisition of persistent fruits from ervoides ha-twe9i h\l)oleucus clevelandi braZoensis 9ambellianus didYmocarpus bl"andegei micranthus arthlri ~ 't1 .., how.1l1 n> n> H c"t-(C) remotus I-'tn er9meticus CD r" trasldae S 0'1 lanoearlus ~o episcopus: ::I r" convallarlus o ::.:: tihnophilus C III xiphoides tn L.", filipes o rt t-( califomicus t-( r" sinuatus CD rt alvordensis CD'-< colli nus tn t-( oamptodes o C c,,"Yicarpus I-tJI-' ~irocarpus CD sclel"ocarpus I-' CD bicririatus ::l \0 lonohocarpus (JQ 0 woodruff; rtdi' miser :T'-' wingatanus UtO tenellus \00 miorOOl:/stis cr.::! flexuosus (J) tn CD hall;; rt ::! coriacious CD tn graoilis 'd C casei tn tn racemosus c"rt bisuloatus: IlJ t-( moencoppensis (J) (1) oooalYo;5 CD CD albulus Po o flavus o I-tJ toanus ::l pectinatus ~ coojunctus: PoUt IlJ 0 reveMus rt adanus III CD salmonis .0 atratus o C I-tJIlJ peckii I-' cobrensis I-' recurvus: '-< sucol.nbens mollissimus praelon9us beathii crot alar; eae ------..crassicarpus " o.Ut J,,;f----::;: revent us (' III 0 -L. adanus . rt III (1) salmonis .a atratus o ~ peckii t-tJPJ cobrensis t--' t--' recurvus '< suocunbens molli ssi mus: praelongus beathii cl"otalal"ieae r- I. CC crassi oarpus dou91assii pomonensis nutallii hornH ampullarius whitneyi cusickii acoidens t ri chopodus-t t t ri chopodus-pl mi ssouri ensis argophyUus !------tcc tephrodes cibarius lentig-wilsoni lent i g-palans lenti g-di pnyst. puniceus :------Icc bolander; malacus :------1.[: minthorniM giganteus: brauntonii inyoensis monoens:is pu1s1ferae platytropis ambYlot ropi s sabulonum insularis magdalanae-lla idrietorum magdalanae-pe sepulHpes villosus humist rat us desperatus: gilensis austinae Iyalli caricinus detritalis spathulatus chloodes ariZon1cus: nothoxYs distortus: gentryi nutallianus-at. misellus rusbyi 1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1 91 deciduous ones, moderate numbers of leaflets from numerous, papery valves from fleshy, declined pods from ascending ones, moderate to small flowers from large ones, and some other characters. While it is possible that all these reversals did occur, it does not seem likely, and the alternative 596 step tree does not exhibit these.

Notes on Selected Major clades -- A number of clades appeared consistently in almost all equally parsimonious trees on the 596 step trees. As these may represent major radiations in the group, their diagnostic apomorphies and component taxa are summarized below. Numbers in parentheses refer to characters of Appendix 2, with the state transformations indicated following the colon. Character evolution was reconstructed using the default ACCTRANS option of PAUP. Changes indicated are not all unambiguous.

Homoplastic ~haracters can sometimes be optimized in more than one pattern (Swofford and Maddison, 1987). The resolved cladogram shown in Fig. 7 was selected arbitrarily from the set of equally parsimonious trees. Sections recognized by Barneby are indicated above the cladogram of

Fig. 7 (see Table 2 for abbreviations). Two thirds of the clades discussed below appear on all the 595 trees as well.

Without exception, the ones that do not are found in the 92 large polychotomy on the 595 tree, and actually appear in many of the resolved individual 595 trees.

Homaloboid Clade

NODE "k". sections Lonchocarpi (in part) and section Genistoidei (in part) [renewal(l):O->l; lflt#(10) :1->2; terminal lflt (15) :0->1; cavity filaments (37):0->1; petal grad (31) :0->1]. Species in this group are restricted to the Great Basin. The presence of unusually long pedicels unites lh convallarius (section Genistoidei) with members of section Lonchocarpi. The consequent polyphyly of section Genistoidei is somewhat surprising, given the unusual coiling dehiscence of that section, which must now evolve twice. However, the overall size and shape of the fruit, and the vegetative morphology is more consistent with a relationship to Lonchocarpi. section Lonchocarpi itself is distinctively specialized, with a strong tendency towards reduction of the compound leaf, and subterranean point of renewal.

NODE "i". sections Collini, solitarii, Podosclerocarpi, and Bicristati [fr curvature (45) :0->2; inflor (25) :4->6; fr breadth (52):8->7]. STR NX ATR RA NV HIS STR Q HCL SU HC HY Me MIS

11'1 11'1 - --- - :J - - :J - 11'1 11'1 C 11'1 11'1 :J :J .... f: :0 :J 11'1 U J:. (II .. u .& 11'1 II c 11'1 11'1 .. 011 :J .... 11'1 11'1 IV ...... 0- ~ U C IV :J .... :J ." C. "Q c ~ :J L c Go .... 0 QI > ...... II ..... II ...... L 0 ...... :st ::I :J ..... '5o IV "tJ e 0 II ..... 3 :J ~ 0 e 11'1 ell J:. III .0 L C .0 :::J\ N 0 '0 e .. ., IV 11'1 11'1 U > a. u L 3 e "tJ ell 1: u ..... e .. IV > ., ., .... ., L L 0 1: .... :J ...... L .. L ..... :::J\ L L a. ..IA .. L ., .... J:. .. e L e .0 0- :0 .0 U J:. J:. 011

FIG. 7a. One of the 596 step trees (shown in four parts). This tree is discussed in detail in the text, and is used for the analyses of chapter three. certain nodes are labelled so that their character support can be indicated in the text. Barneby's sections are indicated above each cladogram (for abbreviations, see Table 2). Terminal nodes \D A, B, and C are shown on the following pages. W lincurius episcopus r-I r- convallarius ~ "ti'tanophilus xiphoides filipes H-Ig lonchocarpus '"%j woodruffi H (") C) L /h californicus c:: sinuatus (") -.J ilvordensis Q t:7 /' ~'~/~ '"-c colli nus Q camptodes h .- curvicirpus ::t ~ U, "'i -.~ speirocirpus Q rt sclerocarpus 'J>. r- 0 bicrish'tus '"'" miser C') rt z :r win9ihnus (!) tt'nellus U1 microcYstis II !< ~ 01 flexuosus hillii Ul rt coriacious LJQ (!) "0 9ricilis casei I"C rt -I t1 racemosus (!) 15! (!) bisulcitus ~ ..nc.pp.nsis l?l oocilycis l> ilbulus ..f:lJ flavus Q toanus pectinatus I~ con;unctus I~ rt'ventus adanus I~

v6 inYoensis 12: I-Oj brauntonii ;0 succumbtns I ~ molli ssi mus 9i9antf'Us I~ mclacus

'"!j ..., .... minthorniat ~~ G') ///// ~ Itnti9-diphysus ..S2 -...J puni~!w: ::: C'l X 'fl'l Ie V, b~!~ldf'ri n Itnti9-palans -< "':) OJ Illnt i ;-I;;U.sonii.. 1"'1 ("'t cibarius 0 I-!) m; sS'Ouri tnsi s ("'t ar90phYllus :r CD ttphrodu I~ U'1 1.0 pra.lon9us Cl' buthii til ("'t crohlarit.. I~ CD '0 crassicarpus I ~ ("'t 1"'1 dOU91assii CD CD pomontnsis nutillii u~~ hornii ampullarius I ~ wMtntYi cus; cle;; I g I "'0 accidens c: t r; chopodus-t ri t r; chopodus-phc I i1

£6 96

s.poot~o cc sn~~tn4~~dS 0:: Q sH~~~~·P .... snu~o~.J~o :::t: m!tit u itu~~sn!

~ :::t: I S~suin6 en Sn~~..J.dS.p Q I C) 1: sn~~..I~~wn4 C) :I: I l-l +J > I SnSOn~A C. Sidq,tndiS C) +J UJ Wn.JO~i~p~ \0 '6!w-nu!t!p6tW (j'\ L!'l L-. z ..I ~id-nu't!P6'W C)... +J S~..IetnSU~ - \0.; wnu°tnqes 0

.j.,) ~do..l ~ottiqwe S ~ ~ I s ~do.qli ~!td ~ if.J.J~nd ":l ~ I S~SU.OUoW r-- sno~uoz~e .....<.:J Co .sn!-SnU'~nt~nu :::.. ~ ....I ~t;..I~u.6

> sn~o~~p

stixo4~ou 97

This is a diverse group centered in the Pacific Northwest and California, extending to the margins of the Great Basin. The presence of more or less coiled fruits unites this group of sections, which except for Bicristati, have been associated together in the past. The position of section Bicristati, which lacks the coiled fruit, nested up within this clade is surprising, but it is the sister group to a member of Podosclerocarpi without particularly strongly coiled fruits, sharing with it a stiffly fleshy mesocarp. NODE "j". Node i, plus sections cusickiani (in part), Lonchocarpi (in part), and Woodruffiani [stipe (42) :0->2; dehisc (41) :0->1; stems (5):1->0]. The discovery of the polyphyly of section cusickiani is important in the taxonomy of the genus. The section is extremely heterogeneous, encompassing a very wide variety of fruit types. In the current analysis, the species with inflated fruits have been split away from those with laterally compressed, uninflated fruits. Furthermore, the latter members of the section appear to be basally paraphyletic to the radiation of sections Collini, etc., discussed above. Their geographic distribution in the Northwest and California is in line with this conclusion. Astragalus lonchocarpus seems anomalous among these taxa and is united with them only on the basis of some rather 98 questionable character states. Its position remains somewhat elusive. Barneby suspected that A. woodruffi was related to the

Pectinati on account of its selenium tolerance, but noted its isolation in the genus. Here it is united with A. lonchocarpus because of a similar reduction of the leaves.

In this position, however, 7 changes distinguish it from their common ancestor. Its distribution in utah further points to its isolation and hints that the present conclusion may incorrect.

NODE "gil. sections scytocarpi (most), and section

Pterocarpi [renewal (1) :0->1; nitrotoxin (55) :0->1; fr breadth (52):5->6].

Section scytocarpi encompasses a large amount of variation in fruit morphology but was recognized by Barneby on the basis of its subterranean caudex, a conclusion supported by this analysis. section Pterocarpi, on the other hand, was thought to be very distantly related owing to its deciduous fruits (not persistent, as in the rest of scytocarpi). Given the heterogeneity of fruits in

Scytocarpi, the hypothesis that Pterocarpi are derived from it is not unreasonable. The presence of rather rare nitrotoxins in both groups is also suggestive. Jones (1923) agreed with this hypothesis. 99

The taxa in this group are distributed from the Great Basin southward, and eastward onto the Great Plains. Interestingly, at the base of this clade is A. flexuosus, a Plains species, and at the base of this clade's sister group

(see next clade) is a pair of species, A. racemosus and ~ bisulcatus with much the same distribution. It is apparent that each has given rise to parallel radiations into the Great Basin region. NODE "f". sections ocreati, Bisulcati, Albuli, Oocalyces, Pectinati, Conjuncti, Reventi-Arrecti (in part) [selenium (56):0->1; chromosomes (57:0->1; stems (5) :1->0]. This predominantly seleniferous group of the Great Basin and Plains is also united by an aneuploid chromosome number increase from 2n=22 to 2n=24. Part of the heterogeneous section Reventi-Arrecti plus section conjuncti is nested within this group, mainly because they share fleshy valves with the Pectinati.

NODE "hit. sections Ervoidei and scytocarpi (in part) . [valve pigm (35): 0->1; valve green ()54:0->1; ovules (40) :1->2; fr length (51):2->3] Astragalus wingatanus, placed with section Scytocarpi by Barneby, is at the base of a small clade including his section Ervoidei. This is a classic case of a taxon, phenetically similar to nearby species of section 100 scytocarpi, which nevertheless is genealogically most closely related to another group.

NODE "e". Groups f, g, and h, plus section Genistoidei

(A. miser). [stipule connation (18) :0->1; stip size (19):0-

>1].

Astragalus miser is a widespread, polymorphic, species found mostly in higher elevation mesic areas throughout the

Great Basin. Its position at the base of this large clade suggests it is an old group. Also its lack of clear synapomorphies, aside from the peculiar dehiscence, may force its position through default.

NODE "d". All the previous, nodes i,j,k,e,f,g,h.

[septum (38) :1->0; dors groove (48):1->0]

The discovery of a very large clade with uniformly unilocular fruits is rather surprising, given the frequency that the character changes state on the tree as a whole

(note that this clade is also present on the 595 tree). If the clade is supported by future analysis, it may ultimately warrant nomenclatural status. 101

Megafloroid Clade

NODE lit". sections cusickiani (in part), Trichopodi, Densifolii, Inflati (in part), Pruniformes, Ampullarii [fl type (27) :1->0; valve txt (33):2->1]. The position of this group at the apex of a large pectinate clade is the most troubling conclusion of this analysis, particularly because of its diagnosis by several reversals which make its members resemble more primitive taxa elsewhere in the tree, including a reversal from fleshy-ligneous valves to membranous. The group is geographically heterogeneous: the elements from cusickiani and Pruniformes distributed in the Northwest, those from Trichopodi and Inflati in Coastal California and Baja, and Ampullarii from Utah. Barneby suggested that Trichopodi was derived from elements of cusickiani like b. filipes, found in a distant part of the topology (node j). There is little reason to question the hypothesis that the species of Inflati placed near Trichopodi actually belong there. They share the same range, are very similar in almost all respects except the slightly smaller flowers and deciduous fruits. Ampullarii seems more closely related to Preussiani on account of the fruit, but those two taxa are not too distant on this tree. 102

NODE "u". section Preussiani [selenium (56):0->1; chromosomes (57):0->1; cavity filaments (37) :0->1]. The hypothesis that this section is monophyletic is extremely robust, having been found on all trees examined to date. The selenium tolerance and chromosome number is convergent on species in section Ocreati and its relatives, but the strikingly different fruit and floral morphology clearly argues against a relationship. The pod filaments have not been studied thoroughly enough to provide definitive support for this group. The group is scored as having inflated fruits (though very fleshy and inflated), which tends to unite it with species that have bladdery­ inflated fruits, even though it seems as though these kinds of inflation are probably different. Developmental work is needed to clarify this point.

NODE "v". section Argophylli [internodes (4):2->0; vesture (24):0->1; lflt folding (12) :1->0; lflt apex (16):0- >2; fl color (29) :1->0]. The monophyly of this large and heterogeneous section is well-supported by this analysis (see also Appendix 3). Interestingly, b. missouriensis.appears to be basal in the group, because of its plesiomorphic fruit (straight and often laterally-compressed) -- this despite previous belief that it represents specialization in the section. 103

NODE "s". Groups t,u, and v, plus sections Malaci (in part) and Sarcocarpi [valve txt (33):1->2; septum (38) :1-

>0] . The existence of a clade of taxa with fleshy,

unilocular fruits has been borne out by several cladistic

analyses of hig~er level relationships in the North American

group performed by the author. The present analysis is

unusual in that this clade also includes a set of taxa with

a reversal to membranous fruits (see node t above).

NODE "r". Previous four groups plus sections Diphysi, Malaci (in part), scytocarpi (A. puniceus), Hesperonix, Mollissimi, Gigantei, Succumbentes [inflor (25):4->6; fr

orient (32) :1->0; wing length (51):4->8; ovules (40) :1->0;].

This is a large group of taxa, with ascending many­

ovulate fruits, shortened inflorescences, and large flowers.

section Diphysi includes the widespread and polymorphic species A. lentiginosus, three varieties of which were included as terminal taxa. No clear synapomorphies unite

this group of varieties, and in most topologies the species

is paraphyletic with two varieties most closely related to

other members of this clade, and one variety the sister group to section Hesperonix and A. puniceus. Astragalus

lentiginosus, as currently circumscribed, contains over

thirty varieties, encompassing tremendous variation in form

of the pod: from narrow, elongate, and incurved, with 104 papery-fleshy valves, to greatly inflated, globose, and membranous. The group hangs together (loosely at best), only because of a more or less continuous intergradation of intermediate forms. It is a weedy, abundant species, widespread, at low and middle elevations in the arid western U.S, west of the Rocky Mountains and north of Mexico. Though it is hard to believe that "mere" varieties of a species could represent the sister groups to a large and diverse clade of major groups, it is possible that the lentiginosus group is quite old. It may be necessary to more adequately sample the variation within this taxon before firm conclusions about its monophyly can be reached. Sections Mollissimi and Gigantei appear not to be sister groups. Despite a strikingly similar gestalt, their shared characters arise in parallel according to the current topology. This may be a case in which the overall morphological similarities of the two groups have not been adequately captured in the data matrix. Several potential synapomorphies of the group were omitted from the analysis (persistent style, ventral groove of the pod, vesture drying brownish, prominent midrib of leaves). The inclusion in this large clade of members of section scytocarpi CA. puniceus), and section Succumbentes is surprising, because of their traditional placement in very different parts of the genus. 105

Piptoloboid Clade

NODE "gil. sections Inflati (in part), Monoenses (in part), Platytropides [inflation (47):0->1: septum (38):1->0; fr curvature (45) :1->0: fr length (51) :3->6].

This group of species with inflated fruits is supported by several characters. The evolution of inflation can be traced in this group from the early-diverging taxa, which have inflated but primitively papery fruits to more derived taxa with inflated and membranous-transparent fruits. The latter group includes the speciose Inflati (sensu stricto) , abundant in the Southwestern u.s. and the Great Basin. Here the analysis has identified probable outgroups to a taxonomically difficult section of the genus, section

Inflati, which could greatly benefit from further phylogenetic work.

NODE "p". sections Chaetodontes (in part) and

Drabellae. [connation (18) :0->1; scarious stip (22):0->1; fr length (51) :3->1]

section Drabellae, a distinctive monophyletic group united by peculiar oblanceolate phyllodes, is distributed in the Northeastern Great Basin and Northwestern Plains.

Section Chaetodontes, restricted to the Northwest, is much 106 more heterogeneous but shares similarities in the connate stipules and reduction in fruits. It is surprising that they are identified as sister taxa in this analysis, as previous taxonomies have split them on account of the presence of a septum in the fruit of Chaetodontes. certainly convergent evolution for reduced fruits may be responsible for this hypothesized relationship, but on the other hand, A. detrital is of section Drabellae, though lacking the septum, does possess the dorsal groove of the pod so often associated with the presence of a septum in other species. It is basal in the Drabellae clade and may represent retention of the vestiges of the septum.

Node "m". Previous group (node p) plus sections Humistrati, Desperati, Humillimi, Villosi, Chaetodontes (A. sepultipes) [calyx teeth (28):0->1; stip size (19) :0->1; fr breadth (52):6->4].

This clade is one of the most novel hypotheses suggested by this study. Most of the groups included are small sections traditionally viewed as isolated groups with uncertain affinities (Rydberg, 1929; Barneby, 1964). The unifying apomorphy here is elongate calyx teeth, a character never relied upon in prior taxonomic work, but'evidently one with some phylogenetic relevance. Both sections Desperati and Humillimi could benefit from further sampling (only one species sampled in each case). They appear to be placed 107 here because of similarity of fruit and flower reduction. Also section Humistrati has similar reproductive structures.

NODE "I". Previous three groups (nodes q,m,p) [inflor (25) :4->6; dors groove (48):1->0]. This clade is characterized by loss of the dorsal groove of the pod, which occurs just prior to loss of the septum in parallel in the two subsequently diverging clades (nodes q,m). Coincident with this modification of the fruit is a contraction of the length of the raceme.

NODE "b". sections Strigulosi (in part), Neonix, Atrati. [internodes (4):2->0; lflt size (11):1->0; lflt apex (16):0->2]. This clade includes two species from section Strigulosi (A.cobrensis, and A. recurvus), the remainder of which is scattered in the basal region of the tree. They appear somewhat anomalous here since they lack (by reversal) the acaulescent habit of the group as a whole. The entire clade is more in line with taxa having declined pods, which appear at the base of the phylogeny. Their present position higher in this tree stems from similarity in their copious vesture. Note that in the 595 trees they are found more basally. 108

NODE "c". All of the groups above. [stipe (42) :2->0; fr attachment (44):1->0; fr breadth (52):6->5; wing length (30) :7->4] Along with a reduction in size of flowers and fruits, the base of this clade is characterized by astipitate fruits, which is reversed in the Homalobi (see above), though common within the Megaflori. NODE "a". sections Hypoleuci, Micranthi, scutanei, and Microlobium [ovules (40) :1->2; chromosomes (57):1->2; fl. type (27) :3->2; wing length (30) :2->1; petiole (6) :0->1]. This is a group of very specialized Astragalus, centered in Northern Mexico and extending into southern Arizona, and Texas. The group is united by reduction in the flowers and fruits, and by high chromosome numbers. An orderly progression can be observed in the evolution of this clade towards annual species with extraordinarily small bi­ ovulate fruits (e.g., A. didymocarpus).

DISCUSSION Description of the Phylogeny -- In the preceding description of individual clades on the phylogeny of Fig. 7, three major clades were identified and given informal names, the Homalobi, Piptolobi, and Megaflori. These clades are derived from a paraphyletic group of taxa consisting primarily of Barneby's sections Strigulosi, Leptocarpi, 109

Micranthi, and Miselli. The Homalobi clade is characterized by a unilocular fruit, the Piptolobi by a deciduous fruit, and the Megaflori by a large flower, but these characters are not monothetic. Relationships in the Homalobi clade are much more fully resolved and better supported than relationships in the Megaflori clade. There the lack of resolution in the consensus trees of Figs. 5 and 6 corresponds to serious uncertainties about character homologies and further systematic work in this group is warranted. The Piptolobi clade contains some novel hypotheses of relationship, such as the inclusion of Barneby's section Drabellae, but is generally well resolved and supported.

Relation of Results to classical Treatments Groups recognized in classical taxonomies were not in most cases intended to represent monophyletic groups, at least under the cladistic definition of monophyly. It is perhaps unfair, to evaluate such classifications by that standard in the light of a phylogenetic analysis. It is important to do so, however, for several reasons. Even if the traditional work might have recognized paraphyletic taxa, it is unlikely to have deliberately recognized grossly polyphyletic taxa, and the discovery of such taxa by cladistic analysis is therefore revealing. Second, non-systematists who use 110 traditional classifications for evolutionary or ecological studies may often be operating under the presumption that recognized taxa are monophyletic and hence in some sense comparable (e.g., in the study of adaptation). The evaluation of such groups is therefore paramount prior to such work. Barneby (1964) has comprehensively reviewed the taxonomic history of the genus' in North America. Aside from numerous descriptions of new species, and new combinations (see Appendix 6), the only major work on North American Astragalus since Barneby's monograph is a comprehensive species list for the united states by Isely (1983a,b, 1984, 1985, 1986), which does not deal with the higher-level taxonomy of the group. Four eminent American botanists revised the North

American species of Astragalus and attempted to arrange ~hem into higher categories. Asa Gray (1864) was the first, and his emphasis on the morphology of the legume has had an important influence on all subsequent work. At the time only 111 species were known to Gray, which he organized into 27 sections, many of which remain at least in name. Rydberg's revision (1929), recognized 564 species in North America, dispersed among nG less than 28 genera and 82 sections. Those genera, with names like Xylophacos and Hesperastragalus. still litter many floras written prior to 111

Barneby's monograph (1964). His work has been criticized on a number of accounts: its narrow species concept (not allowing sub-specific taxa), its apparent reliance on too few characters (see the keys, for example), and the artificiality (polyphyly) of the taxa that resulted

(Barneby, 1964). Marcus Jones' opinion was clear enough

(1923: p 15): The proposed genera of Rydberg are mostly the product of his idea that no genus should contain more than six species whatever Nature may have said or done about it, which is rather hard on the Almighty, but where genera and species are governed by botanical inspiration and not study or morphological knowledge this state of affairs makes strange bed fellows.

Certainly the numerical criticism is unfair since, for example, Rydberg's Phaca has 99 species and Homalobus has

63; indeed the number of supraspecific taxa recognized by

Rydberg, 28, is near that of Jones, 30 (see below) -- Jones merely called them "sections" instead of genera. More to the point, the present analysis can be used to evaluate the phylogenetic status of Rydberg's genera. Many of his genera are hopelessly polyphyletic. Phaca, for example, includes almost all of the inflated-pod groups together despite the trait's independent origin at least 7 times (see Fig. 12).

Consequently, section Preussiani, which is a clear monophyletic group in the present analysis, has members in different genera according to Rydberg (Phaca and

Jonesiella). Yet many other of his genera are monophyletic, 112 or could be made so by the realignment of a few taxa.

Xylophacos corresponds closely with section Argophylli (NODE

"v") if A. iodanthus, A. webberi, and A. cibarius are removed. Pisophaca is section scytocarpi (NODE "g")i

Hesperastragalus is the group at NODE "a". Interestingly, some clades on the present phylogeny agree better with

Rydberg's classification than with Barneby's. His Diholcos and Cnemidophacos together correspond to the clade at NODE

"f" (see above), and this group includes A. reventus and A. adanus, which Barneby placed in an unrelated section

Reventi-Arrecti. Batidophaca includes isolated elements from several of Barneby's sections scattered throughout his monograph (humistratus, gilensis, villosi, desperatus), and yet these taxa all are found within a clade at NODE "0" though in this case other groups not in Batidophaca are also derived from this node. Finally, many of Rydberg's genera are paraphyletic, containing one or more species that are more closely related to other genera than they are to other species in their genus. Important among these are

Astragalus, Hamosa, Atelophragma, and Homalobus -- all fairly large groups that appear convex on the cladogram but are basal to further radiations that were segregated as separate genera by Rydberg ••

Jones' monograph (1923) could not be more different than Rydberg's. He recognized barely half the number of 113 species (although many more varieties), and disposed them among 30 sections of the one genus, Astragalus. It is very evident from the descriptions of species, and from his introduction, that Jones' experience with species in the field was extensive. His emphasis on ecology and distribution indicates the major role that it played in the genesis of his concepts of higher taxa. The work is appealing to those interested in phylogeny for several reasons. He actually draws a phylogenetic tree for the North American sections -- the only one ever published up to the present (Fig. 8). His higher taxa were based on consideration of many characters simultaneously and hence are "natural" or not grossly polyphyletic (the necessity of distinguishing shared-derived character states had not as yet been recognized). Moreover, by recognizing a fairly limited number of higher groups, and not racognizing a large number of small or monotypic segregate groups (like Barneby does), Jones emphasized relationships instead of differences -- which is where the emphasis should be in a large group like Astragalus. Jones' classification is very similar in spirit to

Ba~neby's except that his sections are generally fewer and larger (never monotypic). There are few grossly polyphyletic higher taxa, but many are paraphyletic. In some ways Jones' work is more interesting phylogenetically 114

Atrati Strigulosi

TriphyJ1i Homalobi I Lattflor; IntlaU Debilas I Alpini I Sparsifton POdo-sclerocarpt 1 Preussi ReyenU-ArrecU I I Ul1g.,Hypogl..Ch aeto

Ocreat f, B1 sui caU Gal eg1 formes I Lonchocarpi Hamost l Laptocarpf Flexuosf I Mfcranth1 I Didymocarpi Malacl I Sarcocarpi -- Mollissimi I Argophyl1i

FIG. 8. Phylogeny of North American Astragalus according ,to Jones (1923). 115 because larger groups imply bolder hypotheses about relationships at a higher level. For example, Jones' section

Homalobi unites much of Barneby's Lonchocarpi and cusickiani with a group consisting of Drabellae, Humistrati, and

Humillimi, the latter three forming a distantly related clade on Fig. 7 (NODE "n"). section Collini corresponds closely to most of the clade at NODE "i", while the remainder of that clade resides in Jones' Podosclerocarpi, which also includes Barneby's sections Pectinati (better placed among the seleniferous groups at NODE "f") , and

Pterocarpi (more closely related to section scytocarpi at NODE "g"). Jones' Bisulcati includes A. oocalycis (as did the corresponding genus of Rydberg's), although according to the present results A. oocalysis is more closely related to this section's sister group, but in any case Barneby's assignment of this species to a monotypic section was not terribly helpful. Paradoxically, after criticizing

Rydberg's genus Phaca as representing a catch-all for species with inflated fruits, Jones circumscribed a section

Inflati even more inclusive than Rydberg's group. Rydberg, at least, considered that the A. lentiginosus complex, with aberrantly bilocular inflated fruits, might belong elsewhere; Jones lumped the complex into Inflati as well.

Jones' work is interesting for students of phylogeny because he actually published a phylogenetic tree of the 116

North American groups -- the only one to date (Fig. 8). comparison of his tree to that of Fig. 7 is difficult because his tree contains relatively few taxa; however, it is interesting to note the areas of congruence. One major agreement is in the progression of sections Hamosi ->

Leptocarpi -> Micranthi ~> Didyrnocarpi, which can be seen at the base of the phylogeny of Fig. 7. On the other hand,

Jones has a major group consisting of sections Argophylli,

Sarcocarpi, Malaci, and Mollissimi arising as the sister

group to the Hamosi radiation just discussed, all stemming

from his Flexuosi, a group which appears monophyletic and

quite derived on the present phylogeny in Fig. 7 (see NODE

"g"). Indeed, its subterranean renewal (char. # 1) and generally inflated fruits (# 47), characters rarely reversed

it seems are specialized features, and it is unlikely that

they are at the base of this pair of major groups.

Rupert Barneby's incomparable Atlas of North American

Astragalus (1964) is an outstanding example of classical

evolutionary taxonomy in its attempt to synthesize all

relevant information into a coherent classification. The

descriptions of individual species are exhaustive in detail

and the entire work is executed in a consistent manner.

Barneby recognized 368 species in 92 sections -- not many

more species than Jones but vastly more sections. Barneby

suggests that this inflation stems from his own uncertainty 117 about relationships of segregate taxa (p.32), but the practice of removing different-looking species to their own higher category is rife among classical taxonomists, and reflects an emphasis on differences instead of similarities. An example of this practice is his monotypic section Pachyphyllus, which contains a species with bizarre unifoliate leaflets. The classification implies nothing about the relatives of this species despite its readily apparent relationship, based on flower, fruit, and non-leaf vegetative morphology, to section Preussiani. His thoughts on phylogeny, reflected in a 12 page discussion, betray the pessimism about angiosperm phylogeny that characterizes much of traditional plant systematics (Prance and White, 1988): A system of classification, except in its purely utilitarian aspect, is no sounder than the assessment of relationships on which it is based, and relationship in the plant kingdom, whether we like it or not, is perceived more intuitively than by conscious reasoning [Barneby, 1964, p. 33]. Nonetheless Barneby's contribution to understanding of higher level relationships has been profound, particularly because he utilized new characters in a consistent, synthetic manner. Unlike any of his predecessor's, Barneby also attempted to group sections into even higher (informal) categories called "phalanxes." In doing so he relied heavily on characters related to the fall (abscission) and dehiscence 118 of the fruit, and the size and orientation of the flower. It is doubtful that Barneby himself put much stock in these groupings, so that their cladistic status is not terribly important. Based on his comments scattered throughout the treatise, they appear to have been mainly a convenient device for organizing the large number of sections into manageable groups, rather than a reflection of phylogeny. Three taxonomic levels could be evaluated by the present analysis: species, sections, and higher taxa inf.Jrmal groups termed "phalanxes" by Barneby. The monophyly of species cannot realistically be addressed because of the insufficient number of specimens of each species examined. Only in the case of A. trichopodus, A. magdalanae, and A. lentiginosus, was more than one sub­ specific group scored, and paraphyly might be detected in those groups. In fact, A lentiginosus is a probable candidate for a paraphyletic species, but it deserves much qrea~er phylogenetic study; the other two species were monophyletic. The evaluation of sections is more appropriate to this analysis, but one caveat must be remembered. Few sections have been sampled completely, and it is therefore impossible to strictly "confirm" the monophyly of any groups, although they can be falsified. In most cases, however, a sampling scheme was used such that the range of character variation within a section was 119 embodied in the taxa included. It is not likely, therefore that if excluded taxa were included they would disturb the results discussed below.

Barneby's sections were evidently constructed using two basic principles of classical evolutionary taxonomy: overall similarity among a set of preferred characters, and chainin~ of taxa (stevens, 1986). The latter refers to the agglomerative practice of combining taxa together one after the other until the perception of too large a gap stops the process. In Barneby's work overall similarity corresponded to synapomorphy in some instances; in other instances it did not. Among the sections confirmed as monophyletic by this study are several that have long been viewed as distinctive groups: sections Argophylli, Drabellae, Pectinati,

Scytocarpi and Lonchocarpi (both if adjusted slightly),

Preussiani, Inflati (in large part), and Microlobium. The monophyly of most other sections is questionable, however.

Among the large sections, most certainly paraphyletic are cusickiani, Collini, strigulosi, Miselli, Micranthi,

Reventi-Arrecti, and Leptocarpi. Some of these paraphyletic sections "give rise" to other paraphyletic sections in a hierarchical array of paraphyly (e.g; paraphyletic cusickiani gives rise to several paraphyletic sections, including COllini) •

. -.-..- ..... ------120

On the other hand, few sections appear to be grossly polyphyletic. section Reventi-Arrecti may be one, and even there further phylogenetic work is needed before much confidence can be placed on this suspicion. Most species in

Barneby's sections appear near one another in terms of path length they simply are not grouped according to the nested set of synapomorphies that accurately reflects their history.

Barneby proposed seven higher-level groups, termed phalanxes, for the North American sections. Four of these contain species from previously recognized Old World subgenera (Bunge, 1868) and contain species in the 2n=16 chromosome series. The bulk of the North American endemics reside in the "Homa1oboid," "Piptoloboid," and "Orophaca" phalanxes -- the last of which is not considered in this paper (see above). Each of the first two contain more than

150 species, dividing the North American group roughly in two equal halves. The character distinguishing these two taxa according to Barneby is the disposition of the pod: persistent in Homalobi versus deciduous in Piptolobi.

Judging by the cladograms of Figs. 5 and 6, these two taxa are unquestionably polyphyletic. The character changes between deciduous and persistent 11 times (see also Fig.

15). Although fruit persistence is a useful character for delimiting higher taxa, it undergoes frequent reversal and 12l parallelism. Although the Homaloboid group hangs together fairly well (except e.g., Preussiani), the deciduous fruit has evolved in it on several independent occasions -- making any group defined on the basis of deciduous pods polyphyletic. The most glaring consequence of this grouping is the placement of certain deciduous groups (sections Leptocarpi, Didymocarpi, etc.) towards the end of the Piptoloboid series, fully 800 pages away in Barneby's monograph from their closest relatives, which are in the Homalobi.

The Nature of Astragalus Species: Cladistic Implications species of Astragalus have generally been recognized by plant taxonomists on the basis of correlation of consistent morphological discontinuities with geographic range (Jones, 1923; Barneby, 1964). Depending on the extent of its range, a typical species is found in few to numerous isolated local populations, often with restricted opportunities for gene flow between populations because of ecological barriers like elevational or edaphic conditions. Ranges of very closely related species or varieties are typically allopatric, and geographic speciation is most likely the universal mode for the group. Moreover, in most groups of related Astragalus species, a pattern consisting of one widespread species with several narrowly endemic peripheral isolates is common. 122 This may imply a general mode of speciation by founder effect. Since the operational taxonomic units of this study are indeed selected species, it is necessary to consider their status as units in a cladistic study. If the prevalent mode of speciation is as described above, then generally the widespread species that gives rise to peripheral isolates will be paraphyletic. However, examination of cladograms that result from this study will suggest that such species occur at the tips of branches, coordinate with other species or clades at the tips (e.g., a species will always have either a sister species or a sister group). This can lead to erroneous interpretations of the nature of the ancestral species that gave rise to the tips. This problem is illustrated by the Newberryani subsection of section Argophylli, which is analyzed in detail in Appendix 3. This section is composed of one species, A. newberryi Gray, widely distributed in Nevada, Arizona, and southern Utah, and several more narrowly distributed species arranged around the periphery of its range. One of these is A. coccineus Brandg., a striking species with very large red flowers. The unusual flowers were sufficient for its relegation to a monotypic sUbsection by Barneby, despite convincing evidence of its relationship to A. newberryi. Cladistic analysis of the section suggests 123 that A. coccineus is actually basal in the clade composed of subsection Newberryani sensu stricto plus A. coccineus, which seems counterintuitive if it is believed that the latter species is actually a peripheral isolate of A. newberryi. Astragalus coccineus is basal because, despite its derived floral morphology, it retains a plesiomorphic leaf morphology consisting of more numerous and smaller leaflets than the species of Newberryani sensu stricto. Hence the ancestor of this group may have combined the primitive vegetative morphology with the primitive floral morphology. The earliest peripheral isolate was A. coccineus which became fixed for the derived flower. The then paraphyletic A. newberryi evidently underwent modification of its vegetative morphology prior to its giving rise to later isolates that share this morphology (e.g., A. loanus, A. musiniensis). Knowledge of geographic variation coupled with inferences about the history of patterns of interbreeding is essential to a proper interpretation of the phylogeny depicted in Fig. A3-1 (Appendix 3). This is a reflection of the problems inherent in species definitions in the context of cladistic analyses at low taxonomic levels (see de Queiroz and Donoghue, 1989).

Exclusion of South American Species -- Most authors have suggested that the 100 or so species of Astragalus in South 124 America are derived from groups in North America (Barneby, 1964; Johnston, 1947; see Fig. 3). This hypothesis is presumably based largely on biogeography. Astragalus is essentially absent from the entire southern hemisphere outside of South America. It presumably migrated into the New World via Beringian land bridges. According to this hypothesis, migration was in a southward direction in the New World, and the North American "group" is paraphyletic. Species in South America are isolated by a large disjunction consisting of almost all of central America and Columbia. All have chromosome numbers in the 2n=22-30 aneuploid series and are morphologically similar to species in North America (Johnston, 1947; Ledingham and Pepper, 1973; Gomez-Sosa, 1979). Taxonomic work on the South American group has not been undertaken at the level above species and hence it is very difficult to draw phylogenetic inferences from the literature, and practical limitations have precluded scoring of South American taxa. Only a very few South American species seem to belong to North American sections. section Inflati and perhaps section Strigulosi have a handful of representatives in South America (Barneby, 1964). The remainder exhibit distinctive combinations of the same characters found in North America. In addition, the adaptation to often high­ elevation habitats along the Andes has resulted in many 125 species of tufted habit with very reduced few-ovulate fruits. other uniquely South American character states include the presence of glandular hairs in the inflorescences of some species (Gomez-Sosa, 1979), and greatly reduced inflorescences in many (sessile, few­ flowered clusters; Johnston, 1947). A character state not prevalent in North America, connate stipules (see Appendix

2) is very common in South America. Also chromosome numbers in the range of 2n=26 or 28 are much more common in South

America than in North America, where 2n=22 or 24 is found in

83% of species examined, and an even higher proportion of sections (Spellenberg, 1976). It is possible that the South

American groups are derived from groups in North America having this combination of characters. certain members of a group of related sections that includes Strigulosi, Miselli,

Leptocarpi. Hypoleuc~ and Micranthi exhibit these character states in various combinations and also have a distribution in temperate Mexico -- the southernmost extent of Astragalus in North America. The vast majority of South American species may represent a single monophyletic radiation from this North American group. As such, its impact on the reconstruction of the phylogeny of Astragalus in North

America is probably minimal, although this hypothesis deserves further scrutiny as more information is gleaned. 126

Evolution of Selected Characters -- Any discussion of the

general pattern of character evolution in Astragalus should

be prefaced with some remarks about the levels of homoplasy

evident. On average every character undergoes about ten

changes on the topologies listed here. The consistency

index of .17 (~ot counting polymorphisms) is not unexpected

for a study with 113 taxa (see Appendix 1), but the level of

homoplasy is still startling at first glance. There is some

variance in the amount of homoplasy in different characters

but much depends on the distribution of apomorphic states

among terminal taxa. Those characters with very few

apomorphic taxa, which are constrained thereby to have only

that number of changes and no more tend to have less

homoplasy than characters with more even distributions of

apomorphic and plesiomorphic states among terminal taxa

(e.g., of the 5 characters that undergo only one change and

hence are not homoplastic, none unites more than three

taxa). The most homoplastic character, petal coloration (#

29), has a CI of .056. Some characters change states deep

in the phylogeny as well as along terminal branches, and

they therefore exert an important influence in diagnosing

major clades. Other characters change state so sporadically

that they are not very informative about relationships.

This issue will be addressed in more detail in Chapter

Three. What follows is a discussion of the reconstructed

------.. -'- _._ .. _.. --. 127 patterns of evolution for a number of characters historically believed to be taxonomically important in the group. Numbers in parenthesis following character names refer to the character number in Appendix 2 and the data matrix of Table 2. See discussion at beginning of "Results" section on methods used to reconstruct character history.

Fruit morphology (Figs. 9-15): Characters associated with the fruit have traditionally been of supreme importance since Asa Gray's revision of species in North America (Gray,

1864). The septum (#38, see Fig. 9) appears to be primitively present and to have been subsequently lost in three major clades, the homaloboid group, eventually in the

Megafloroid group, and in the Piptolobi group. Hence it's loss is important in diagnosing several major clades. The size (#51,52) and shape (#48, 49) of the pod is less consistently useful for higher-level systematics. At the base of the phylogeny are narrow, moderately small fruits either trigonously or laterally compressed, usually with a dorsal groove. The homaloboid radiation begins with an elaboration of fruits on the basic theme of strong lateral compression (#49, Fig. 10). One group explores various degrees of strong coiling (#45, Fig. 11), while the rest remain essentially linear in profile. Another group undergoes a secondary expansion of the lateral walls so as to become terete or slightly dorsiventral in cross-section I, .,filipu ourvloarpus sptlrooUllW soltrocArPUS , __ blorlshtus '" noalftPtodH sl IM/at us IlvordtnSfs lIIIIIIIII[] ooUfrnd I' --,oaltfomlous CD 'C 'C ell " r::::J lonchooArPUS o'l'>j J:I c:::I ., cr , ...... woodru(fl III H C - CD CI:I oonnllar!us -cc CI:I CD In G') <3CD:I tihnophllus (Il • c:::I Q :I ,.. xipholdts 0. g -d ,.. -::r " i!"~!II~'~~~ :::~:~ o '" t_uus ::l n .1 oroollStf. () - "O::r" wi ngat a/'iUII ::r"1U ....---r-...-u toanus '< Ii peotinatus I-'Ill rtvtntus o 0 .cIanus ()Q rt oonJOOOtWl (Il (Il .Ofilooppensl • ::l Ii '< oocall/OI. (Il albulus In <: flaws ::r"o " ~raotmosus o I-' I.!=:I:Jbisuloatus :e: s:: ::s rt .... ~r.I=[@====,:::J~CJ!=- Graolli • ...·0 castl ::s ::s oorlaolous I'>j .... " • ,.1 •.,. .... ::l QQ ~~I H) ampullar! us Ii aooldtnS ...... s:: trlohopedu.-trl­ .... trlchopodu.-Jtlox rt I' E,IIomU In (Il "0 rt s:: :3 " ~oroblar!tM prwlO/l9U' o I' ~mISSourl.nsi'a!"90phyllus ::r" t~rod.. III Ii n." 'LIolbarlus III ~ltntlO-Wl1sonu o .1tntlg-palans rt , ~ltntf"-d'Ph\lS" (Il punlo.us Ii bolanckrl malacus v.> , c:mlnthornl" co !1f\JMIt.us I · .moUfssillUS • suocumb9ns: ... _ ..._& __ '1

s:: bf..thtt !3 I' I n..tI!morot ,lArlu. i t..=::::prMlon!IUg (') ml ..ourl.nsls ::r 'r"90ph1)llus III ttPhrod.s Ii n I' ,olb"..us III ,. .1.nti9-UlbonH n _ .1.ntf9-p.. l~ rt I .-.ltntl\l-dlph1)SU (l) _ ~punlo.us Ii bolandfl"f w m&1acus co I ___mlnthornlDf ____ \l19ant~

rU.:tfffrt""'*+ttff_ :2 . If! ;::E:.tnvotnsts -;:==~~m.lstr'tus II L!:::xJd ..ptntus ______'ustlnH .11),111 owlofnus spathulatus ohloodH dttrthlts 1/1 I' W!!~= ~ ; SfP(JU fPfll m8\ld.lanM-11'1i9d tdrt.toruo m8\ldal_J>.trs L..-..I ...... sabul_ Insularls pUlsl'.,." plat vt ropls aMb1JlotrOl'ls . I' IIOII09nstS ...._- ..... _ .wlZoniCUll dfstortUS I1 [!5_.nothoxlls -----1~~==~---austoobr9nsis r9O\ll"'VUS , ~~tt l-=nlMnls ,tr,tus • • r_t us • ~.,._.ttous .....traSI

__ :.Ioranthus rdl~~!JI~'~'~~~~~~~~~~~~~~~~~:.~usIi brMldt\l9t \leMllIaous df~u • • br.. z~l$ ' ..-- .ol.Y91andf lI~':::::'::::::::::::::::::::::::::::::::::::"'~~"'::::::::::~~US: • .,.vold..

, IIftlfpu II curv1 Q,rpUs sptlrooarpus .0)"00Il1'u. bforflt,tul " __IIf°uPtodt. ___.__lIIlfnuatus .Ilvordtnsi. • -::~!~':nleus , pnlonOhOOarpus -'TI CD 'CI :I cr ___ woodrum () H .QC~" :YG') C:_~CI ....oonvallarlus - CC ~ tthnophllus III <:I CL Ii c C ~ III I-' ~.:!~~: () 0 !.~ , .'_ .)anot.. lus rt ::r .t_llus CD n .Ioroel/stl. Ii (") - .wllI9atanus :Y t01llU1 ~1Il ..-otlMtus 1.0 Ii rtvtnlus '-"Ill () adlllU' rt oonJunotus CD .Otnoepptnsl. Ii 000111101. albulus CD flevu. < :Ulraotmosus o =::J[Jblluloatus I-' ~ ~~~~u;lf1tXUOSUS rt ~I :;hallll .... 9r.allls o caul !j oorlaolous • __.IAr .... whltlW4Jf !j ousIckli HI upullrfus Ii IIOOldtns ~ trl chopodus-t ri­ .... trlchopodus-phox rt homU pOIIOOtIlIf. () nutaUIl Ii dou91u.U o I' • ,crulfcwpu. rJl rJl ~athU crotliarttM rJl prHIOC\9US CD ml ••olrl.nsl. () ar90phllllu5 rt t~odes .... Jn I' ' u-L~clbarlus o 1' ;lflltf9-w11S0nll !j 1_ ~l.ntf9-P.ll!nS ).nt 19-dl phi/sus punlc.us boland.rt .... Il.eus mlnthornl .. oloant.us I' --ijmoUlssIDUS

(') Ii o !Il !Il I !Il (!l n ,....,rt o ::l

I ~r'1tolrYUS ----.u------~1 ...... "s~lmonls""old 1 .... ~tr~tus ,_I'"ctmotus r..__.nlMttous

~:Hlj _ -:E3r: ..1 ol'"ant h= .bl'"~ndfg.1 gamMllIanus dldl/moolrPuS brnotnsfs II [f ---.:~;~!~S! : :.rVoldts

i ,ftltpts , CW"Vloarpus .,.trooarpu. so"rooarpu. • ___btcrtstatus oamptod..

sinuatus ~l~ alvord.nsf. ~ oolltnu. " ~,caltforntous ~>'%j CD 'C I' r;:::cJonohooarpu. V1H ,Q en ... · C* - L!::DlIOOdrufft ...... G) ,. c_ cc - -c :In oonVallarlWl C :I - C ~ ttt_hllus ca xiphoid •• ..... g c & ~ ::r- ..... CIt -d CD ,. I' : I" : .,.tseopus Co L.!! ~ -::r -==::::I:CJ,t.Io.artus n ttlltUUS (') - r;;==== g alorocllstt. ::r I, wtnllatanus III -.-r:::ctoanus Ii I r----. L..,,-,~tlnatus III revtntus (") u,.,.....,,~ rt CD oonJunotus Ii 1OMI'I000000nsl. oooal\lOts CD alhulus ~ fiavus o t-' s:: rt 1-" o lIrAOIltS ::l _I oorIaolous 1-" I' -,.tnr ::l wbHrwvt ous101cft H1 Ii apulhrlus s:: .coldtn. 1-" trlohopoclJs-t rl­ rt trlohopodus-phox I' -,hornft n pocaontMf. s:: nutaUll - Ii dou9lusfl ~ !' ~,crusioVPII' 1\1 rt I' g~athl1 C orotalarl'M Ii prHlon9US CD ,-... " OCC:::'~~ts n ::r 1\1 Ii 1\1 n ~!:~~,!!::::::!:!!!!!:!!!:!:::!:!::!:iS§g,~: rt . bolandtrt CD .. alacus Ii .tnthornt.. _~ OIl/lnt.tIS ~O!::;;::;:;;:;::;;;;;;:::;::;::;::;:::;;:..;;;;::;;:;;;:;;:::::;::;:::;:::::;::;:;;:::;:~=:;:

1"'\ <: III rt C crotalvi_. t-1 pra.lonOUS I alssourl .,,11 s C1> I ,..... rl II IIIIII II II III {C;::::~~US o ::r' It~'::;·:;.::::::;:;::::::::';:.:::·:::::::::::::·i:::::~!::~~~~Ilsonll III .'h' "·;·1 ... "m.!,. .... """ '; .. '.'., "1 .jQI.ntl9-palans Ii "l.ntlo-dlphl/sus III o "punlctIJJ rt bohncMrl (1) .• lIIaIMllls Ii .Inthornt .. ___ uOlllant.uJ ~o!::m:::;:::::::::::::::!::::::::::::;:::!::::::::::::::::::::::::::;:;:;:;:::::2r~i ;.!! ••. !iI.!Ii.Mim.Iiii!.f!ili.,.,h1m!i., ' .. ";Ilie ".I.I.p .•. I.!.!!'''''''''. ii.'utnl/otnStJ liU'ill.jll.,."j" .•.• !e!:"';~:~s austin.. II/.m ortotnus sp.thuhtus ohloodts ;Jchtrlhlls ~III!·u.;:.:~::::::::::d:;!~~.s ~ I OIltnSis .a~IanM-lM9d I drl.tonJll lIIa~hnw.".lrs ~ I __ nmmlsabulonlllll Insularts pUIstf.,... phtl/troPls I/IIblJIotropts

R .•. ''£-'''''''$ ••.• 1. jfi$'·!.""'~11l "ii!! "i.i1i!~::::::::::;:::::::':::::::::::=,~I.:s

ILi!! jiI'i'i!!i!!lI!i!! !!i"'iiIi=:~l~:us-aurt~WWl!lllllllllllllllJnothO>ftl

-OamMIlIanus dt dVaiooarpus brazo~ns1s oIn.hndt ]t~~:::::::::::::::~:::.::::::':::",:::,:::::::::::,:.::::::: .. ::i:::::::::j::::::::::~:::::::::::::::.:::,:::::::::·:::::.:::~g:;~;~i~:;~:~::us

I , ofillpts curvfov-pus ~lrooV'Pus sol.rooarpus blerl ..,tus U::=::::cnOMPtodH sfnuatus .!vord.nsfs 1I1II1(tlll oo1l!nus I' ---;;::jo.!lfornlous CD 'a - C lonohooarpus ..,....., .co=-=- wooli-ufff -...JH c---­- cc - :I oonnllNfus - G'l < 3 ~ ~ tf hnophtlus °OCIICl xiphoid.. t-' ".,1:1.""2. 'a CD .., 1'77o",fsoopus :7 ca. . ===::iolMlC.rlus "" at_llus - ••loroclIstls (") " wlngatanus ::r' toarAll Pl .Pflltt natus Ii Pl r.nntus n aadaoos rt ooonJunotus CD II09nOGppfllsfs Ii oool1l1Ols albulus CD ofl.wl <: r aotlllOSUl o ahisuJo.tus t-' C J iOfl.)(IJOSUS rt _I i ahaUII 1-" !lrllCtlis o caul ::1 oorfaolous I' --';:;.1- 1-" .whltn'lIi ::1 .ouslokfl HJ ••mpullirlus Ii aoold4tns C • t rl ohopodus-t rl­ 1-" otrlohopodus-phox rt • whomft pOlllOn.hsfs 1-" nuhl111 ::1 dOU!llauff HJ • --.or.sslorpus t-' III • b.. t hll rt _orotalNf ... 1-" pr•• lonllUS o mlssourl.nsts ::1 arQophllUus ,..... t.phrod.. J nil' ~~Ciolbarlus n ~ :--1.nti 0-1.11 1101111 ::r' . 1.ntiQ-p.lans III 1 Ii IL r::_l.nttQ-dll>hllsus III . _ • IIIPunlo.us o _bobnd.rf rt :alnalaous CD amlnthorntH Ii tilQIQ.nttus 1" :iUlnolltnllllUS 1. :C1suocumbfl>s:

,-- ::l ==iiinuhl111 t1) .dol/91usll t-' • ..or... lewpu. III rt I E~r~Athtt.eroblarl ... 1-'­ .prulongus o 1~IDI.sourl.nst. ::I ===; Lr,:::lo,,.qophl/llus I" t.=Joiepl\rodu n Jn oolbarlus ::r 1 ;oltntfll-wllsontt III oltntill-pal.ns 11 I' =.lfntfll-dlph4JSUS III .punlo.us n I.behndfrl rt , :OmdACUS

I ...",,11 I •~= --.allh01lls atratus

nU!II'''''IF''+I'.!!-W··III.!I""mflllpu C\r¥io.-pus sp.frooll1'us f01w~ bfor'f.t atu. I"IIIFJI~ UbrzmnmlllODlPtodtS '1~s1nuatus CD-a­ -a :I lJalvordHIsfs .CI 0 CD ell CD llIoolllnus c: - ell i: :I II"I!!! 1!!!·;,aw·;,AAmoaUfornlous '-""1 -CC""C :I::J' 0 CD II".!, 1""%" o H o c2 d=- ""i!i!E:~=::'riS ::TG') g ., c2 ::::J "oonYalhrlus I!> o tttanq>hflus 11 -",:r .... C xfphofdH I!> t-' Cj) .p/soopus W --" (jICD o lI'ii "i1i.lliliI'.iiIi i!III!llano• .-fus rt ::J' en cc: h ....llus 11 (') .1 oroo\lSt I s ::r' wfn9atanus wI!> toanus WI1 pect lnat us -I!> nYflItus 0 ...... adanus rt conJunotus en _ncow-nsfs 11 oooal\lOls en albulus < (laVIIS 0 rao•• osus t-' bfsuloatus C (I.XUOSIIS rt h.1lf1 1-'- :.9I"aoflls 0 cu.f ::s CortaelCUI 1-'- Illwqlllilil"!J!whillllillill ..:;;g"alstr ::s ....JtiDUl wtli t .... 1/1 ...... Ili'1f:=Jaousfcklf H! mallPullarfus 11 .accfdtns ~ t rI chopodus-t rI­ 1-'- trfchopodus-phox rt Ilhllw"w"'liuDmhomlf < pomOjlfllsl s I!> nut.lUl t-' dou91uslf

(1) .b. .. thff • c::crObllrl .... rt prulOl$ls (1) , 1:i:E1D1J'S"""I.nS'S X • arlloph\lllus rt .hphrodu ~ I-( "Ii (0 11!"'I"H' .. '" .. ,"ii. "~:~~~:~!~~IJSOnll ..... I@'. I .. , ,'.'.'.' ..... I· ,'" ""'" ,." · ... mltn! III-Pllan. t.!.! ,... ",.., ...... ,~:!:~~~~'Ph\lSUS !lImbolandtrl •• ma:;aous ~ """""""'''' ""'''',.,' a:::::::!=!a. IIl!:::;:::::.::;;:::::::::;::;:==::=~!~:;:~:~: .11" ' ... " '. '''''N,' ",11,.1.,11'1,1,' ,,.,.,,.di,!II,liiiliia!l!l!lmllllmbrauntonll 1n !II.!.I."'.!.>.,.!.! j .., ii' ii' ",.1,.' '.' •. " '., II '"!.I''' ,,:mmmmmm \lotnsb ~mhumlstratus ~1,=!Udtsptr ..tus " austln.. l11aUI meariotnus mspathuJatus mohloodu mdttrltaU. mllll• nsi• ___ _m villosu• ,. , .• .!I.1i '!111"Wstpultlpu maQdahn ...-maQd Idrl.toru", ••gdalan ..-pelrs IL-.IIlllllJmsabulcnum 1U.Ul.\J-.-.lnsularf • pulslf.ra. i III ~platlitroPls "I il' ", •• if! , iii: f!C"a::~:"IS I. . . Irlzonlcus dlstortu. I I • 9tntr\ll I . nut.allianus-aurl . notho~ cobrM\Sls noUl"tUS ptO\(lt salmonls atr.. tus r .....tus .. H'... UC\lS d! I ~ ...... - d:~~!!ij!ii!g:t!!~~!!!i~~i!!:~!§§!!§§§~£' -....,,~Jllanus dld\IItOOarpus br.azoensb 11 !~:::.:: .. ~::.::::::::::::::::::.::::::::,:.:::::::::::::::::::::::::::::::::::::':.:::.:::::::::::::::,:::::'::'::::::::::::::!~;:;:;:.:;::!:!;~:;::~§~

E "ftll~u ourviOMPUS ~lrooM"PUs lOuroorpus bferlriftu. --,oUlPtoMs .....lII sfnu• tus ~ -- .lvordonsfs oolllnus CD'CI_ en ::3 • -8OIltfornlous .CI e e :re "lj c -::3 e ::3 ~-:!::!:ris H -. CC = • G'l c a ;leD oonv.Uarlus e e tt t anophllus ~ ~ xipholdes ~ -:: ..I, !' : i-solsoOPUS .p- ... 1 ,1ano ..,.lus teMUus IIlorooystts () r;===~ ::r­ II ~wln9.tanus III ~toanus Ii ~ L-w-.~ttnatus III reventus o IUn ...... adaoos rt oonjunotus CD IIQtnOQPp.nsts Ii oool1l1Ols Ilbulus (l) < n.vus o r.o_us ~ blsuloatus ~ =--ne)(\J()sus rt !, r:pm;;;;:rr;hallit ..... grlOl1Is o o... t :;j oorI 101 ous ..... I' ---,alnr ----'1lII'.llIIwhltntV' :;j _F'!:b..r.. oust ckll HJ lIIIPullart us Ii ___lOolden. ~ trlohopodus-trl­ ..... trlohopOdus-phox rt II ,homti pomOMnsfs !II nutalllt .....rt ~doU91u.1t "d I' ,orusfoarpus (l) I ~~.thti I orotalarl ... pr•• longus o II ~' atSlou.:tensts ::r' ~ . ....OOp!IYllus Il> _ , . t.phrod.s Ii J nII 'OlDarlu. III 1' : :1.ntf\l-VIlSOnff o . - ; :l.ntt\l-palan. rt 1 (l) II r;:::=::Dl.ntt\l-dlphl/sus Ii . i Lr.:=opunlot'Us .. bolanderl .p- m.l.ous 'N II : ~ ... Inthornt .. \II\lanhus

pomontnn. til nut-dill ....rt ___dou9lassll "0 l.!.!::::====:x:JicrunOM1>Us (1) ....thlt " orotal...t ... '""' Pl"HIOl\9US o ,~.I ourltrulls ::Y . . ar\JophVllus.. II> ~ I' ' hphrod•• Ii J .- Cl0l'r1u:J II> : :1.ntlQ-ullS001t o . n ; :l.ntlg-palans rt 1 (1) I' ~lentlQ-dtphllsus Ii . i I.I;:::x:lpunlot

II ~~sal.onls .tratus , l:C~tus

~m 2 i~r 10 =:~::~ ~~rot..llianus JII.I 1 rrlP5~~~~

__ filip,s - otrVtoarpu. SPftrooarpus 11II1I£m-1!1111- -I I.sol,rooarpus ~ ~~!:;,\s::::s 'CI ~ Q. co '" Gf'O CD '1~·slnuatu. ,Q " .Jvord#nsf. 00111II1II -u:t: - IIJ., --­" < EI -~c. • -poaUfornlous m eI C lonohoorpus W"1 "g~ " ;: C4 g woodrufft \0 H -::r' ::I C C4 oonva11arfus c;) - :I,.. -----tttanophtlus -n - xiphoid •• I-' I.J1 • II' -:::~~ tfl'l.l1us .-.-J-t'"U:::J:]mtorOOl/stII C) • ~w1ngahnus ::T toanus III PfOttn.tus Ii III r.Vfntus o adanus rt -- oonjunotus (1) IIOfnooPPfnst. Ii oooall1011 albulus (1) flavus < r~ ...osus o bbuloatus I-' s:: rt /-', ~ II!-E-;~7.':-graoill. o ou.t ::l cortaotous • minI' /-', wilHIM\Il ::l oustoldl H! ... MPUllarfus Ii aoofdMs s:: trt chopocIus-t rt­ /-'- ____III.homlt trtchopodus-phox rt POIIIOIlffIIl. '1:l nutaUIt (1) douQ1uslt Ii 1/1 " •• crusloarpu. ~athtl /-', _ _ crotalarf._ III I rt l5pr_lon9us CD ::l , ~~u,:ts o '=iEiit.phrodf. (1) fl'" I I '::~!!::~~nsontt o _ ;:;.lHlttg-pal.... ::T .1HlUQ-dlphl,jsus III puntclUS Ii bollnMrt III mmalaou. o , nmmmtnthorntH rt ...... gtgllltfU. CD Ii I . o::=:: _ .br~tontt II! .lnlJoftlSts _____r=ohtnnfstratus

'0 IL-rre;;;i-.utaUit !l) ~dou91alStl 11 en " ncrusfevpus 1-" I ~b.athtt en crotalarl.M rt -prMlongus !l) IIIlssocrf_ls ::l &r9Of>/IVUus n t.phr~s !l) I, i "I,i1, ,11 .. ,,,,,, i.! I" 1"""':~!::~\1SOnlt .,..... (I. n .1~nt i 9-p.alan~ :T I r;_ltntl9-dtphy5US III _puntc.us 11 mbolan

d!' I §;=_Invo-'s --r.=;xlhulalstratus -l..!::xJd.S!'ft'atus ~austl"" lllaUi oarlollllls spathulatus OhlooMS .~trltalls .11 11 " E!~~= 1 .l!PUltipu malJdalanM-tN1!d I drl.tonJll ....\Jdalan_frs OSibuloolllll clnsularts ----.I r---'opulslf.rM platlltropls amblliotropis I : !' ;umOnMnsls III oarlzonlous -dfstorlus lI,nt,...,I _ nut am anus-aust ,-=onothoX\ls .r.ourvus.ocbrensls I .,~.p,okll __salmonls .atratu. B ~r.motus trtmttlous trukl.. dt W :~.!: E:~~usmtoranthus _brandtll.I dfdl/lftooarpus I .te~-"'~"'brazo.nsfs I' IYD=:::S·hartWtlli ••rvofd ..

~ arge. broad cal yx large. narrow calyx smull. straight calyx strongly Incurvad .polymorphic equivocal

FIG. 16. Character evolution in corolla and calyx conformation (character 30).

I"

~ ::l Po ~orot~Mi'M o pntlon9\lS ~ .iuoc.rltnSts ~ ~us '-< t.phrodH ~ l.~~:.:; ... :.:::.::::: .. ::' :.. ': .. :" :~~!~:~!~t1sonlt m1.ntlg-palans ~ml.ntfg-dlphlJsus punlct!Us boboo.rl ",abcus IItlll!l,".i!! 11,.,.11, .i!!l!i " .~::~!=M Ir" j~I!~~P!~~~~~~~~'~'~ ~~m:?~ ;;;bramtonll fn\lOfllSI s humfstntus III dup.,..t US _--=-111 aust fna. IIIIyallf 'lIIoarioinus 1IISJ>lthubtus lIIohloodH

~1~='==-:-::::::::'n!.~fS... I·vUlosus 1 lIIs.pUltf~s lIIa\ldallllM1a9d f drf .torull ma9dal."M-p4lIrs lllsabulonum lIIinsularls PU1SI,trU pbtvtropls ambYlotropis , II ~I" L-lE ::mOna.nSIS .. "" lOanZonloul ~dfstOl"tus o.ntr\ll nutallll1lus-~ust , -UnothoxYs oobr.nsls rtourvus PtOklf Usalmonls atratus I lEur.motus - , .r.mttlous traskfu

, omlullus I'USbYi ",Ionnthus ,d11~~1~'~'~~~~~~~~~~~~~~~~~:~~~~_.brandtO.I -FlllIoamb.llhnusE ~dld\llllOo.rpus mbrazo.nsls l' iii "" iii! i OO!J:f~f;;;::E~~j~~~~ _ lII·rvoldu

I ,ffllp•• , ourvioarpus sp.froo.,-peJ. .,..,If.t.~Jarooarpus II IIHII'i odH sfnuatUJ alvord.ns1. oollfnus 1-''-::1 I' ~oalfforntcus ...... H CD " CoO CI) Q .CI C C C I' ~lonoboowpu. C - 'CI CI' . L!::xJ woodruffl -c.c:: CD '" oC3lCD oonValllrfUS I-' g C _ ~ tit onophllus -...J a .a n a lCiphofd.. ~::r"'::J_ 1:1 CD I' I l!5· ~flloopus .. .1ano.. rtus () n - II ::r' :I t.nouus III i ~;:=:==~ IIloroovstfs Ii II ~ wfn\Jatanus III ~toaoos 0 I i--P ~peotln.tus rt Avtntus (t> Un '->adMlUS Ii ooruunotus lIoenc~nst. (t> < oooalUof. 0 albulus I-' l1avus ~ I ~r.loaOSlU rt ~bfsulo.tus t-" 11.)(\10_ 0 ::l IlfShlllll9'10111. OHot t-" ::l I' .1_oortaolous 'd whftM\lf 0 oustclcfl t-" upullirfus ::l aoofdHIs rt 'I J ~ trtohopodus-trf- trlohopodus-pl\ox 0 JI --homlt H1 POll_f•• Ii nut.Utl (t> ~dougl.. stl ::l II crlSstcarpu. (t> I ~"'athtt ~ I' : ~orota1Irl.M III , -prMlon9US t-' ",Iuour-lfilsls .-qop/lllUus 0 t~od" ::r' Hi II' --olburlus III Ii 1II : :~:::::~~=" III lfiltf 9-dIJ>hllsus 0 PIIflto",s rt bolllllCHrt (t> .. alleus Ii .. fnthornl .. qlQliflteus jll' ::~~;;::

PQlIOOtnSt.. Ii oot.UII CD douglu$U ::l " ,crusfolrP\ls CD ( I ~b.. thtl I' orot~rft.. Pl I pr.. l~ t-' I' ~ralssourltnSlS ",...... Mllophl/Uus n hphrod.s :T Jnil' :nfb.rfus Pl "· : :l.ntfo-wtlsonff H 1. : Iftltfg-p~W\S Pl n I' r.=:=c. Iftltf9-dll'hllsu. ("t . i Lr;:=::cPllllfows I bol.oo.rf (l) H lIalllOUS I' 'n,!:::Jolllnthomlu t.=:=cllf9ant.us ,0rrfO~ ..::~~=:f === r===r;:::ohtmlstntus I,r L!=Ddtsp.ntus .-r;=:=owsttn.. II/aut o.noloos spathulatus ohloodts dttrft.Us:

J 1" I' 1.• -----::~=npultfpn ra.g~MIM-lIia9d 1drf.torUill m.-]dalanw-p.frs nbulOllUll fnsulrfs pulsff.r.. pllJtl/troPls MbI/lotroPls I' 'I a!!lOOOtnsfs 'IIII .. , ..fzonfou$ dfstortUS II. 9tntrl/f ~bllfanus-MJst L.:-===::l[]notho)(IJs~ aoobrftlsls artourVUs PtOkft Alaonls .tr.tus rfPOtu$ tr_tfous traskf .. ,0°inll~ ~~=:;-!~ ~1lbt11fanus dl dI/inOOlrpUS br~o.nsfs illl ~rl-a§E'"!! perennial shortly perennIal annual polymorphIc equ1YOCDl

FIG. 18. Character evolution in life-cycle (character 2). 1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1 ,-.., n ::r III Ii III n missotrf-is rt r90PhYUus (1) t~od.. Ii j n I" --"':;ofbar.us N 11 :wlflltfo-vllsonff ...... 1_ :w1tnt.9-P.t_ 1.nUlhflph\ISUS puniCfUS bolandtrf .al~ .fnthomf.. 9f9MltfUS rrITll" --.0= 10 _ ~~::f humtstl".tus dts!>H' .tus IUStfnM 11/.tU OM'iotnus sp.thul.tus chloodts dttrftalf. 9fl.nsfs nIII ill! iii.iipi.IIIIB!IijI!liilllvtUosus , ' ICI StPUlt1P'S !I\I9d.lana_a9d fdrl.torum ••9

.. ,,"""" ,,""" .. '." ,".'" ,""., ,... '" ~Ijl ,~.t,,,,..,., "".,,"'''''''"'',.,,''''',]§;.• oamb.lll .. nus. .d.d\llllOorpus .bl"'ZOtnsts J IIII rn fl-'"""""'-:::"'~oltnland' i jl ~~~~:=us 1, - ~'I"VOfdts

,fiU~s - o!rVfO.trpus spdrlo.tl'PUl .oJ".~ blorfshtus t-l>%:l I' --'OllDPtOdH +"-H sinuatus '-" G) fIIIIIR-O ~lvord.ns1s oolUnus CD'O .... O I' ,0~lfomtous t-l .Q 0 - C lonch~rpus \0 C-':II" -cc CD,. woodruff' c 3 ct CD oonv~artus OC).,~ -Ir---tttanophtlus o n., CD :J~ ct '0 ,. xiphoid" III .... ::r n · I L - .. ~tsoopus Ii -n . _ ~lano.. rius III hnlUus o ,.torool/sUs rt wlnllatanus (1) toanus Ii PlOttMtUS (1) rlvlntus damll o< ..,. -oonjlllOtus t-' lIotl\OOPPff\Sl. s:: oooal\1Ol. rt ~lbulus ..,­ flaVIII o I' , ,.....no.!I>O.us ::3 . l!::XJblsuloatus ..,- ::3 Jf.~~~:~g~~t'f'UI --or~oms t-l OUttt (1) __.oort~ous III II 0,.1..,- I"t) whttne\lt t-l ousIddt CD rt ....w~us ~dIns I U r-----trichopodus-tri- III ::r' trichopodus-phox III "tl IIIJ!I==~=lCIa~~lms CD ,..... ~dou!Ilusil o I' ::JOorusiO.t/"'PUS ::r' bla\hli III crot~larf._ Ii pr.. longus III I ~~"ISSotrfftlsiS o I. . arIIOPh',/Uus rt t~hrodH CD I :aolbarfUS Ii Jn I' .. ~u lentf!HIllson" 1. ulentlHIl_ I I ~!:~~Phllsus bolMldlrl ...lacus I' I n.!:::JwllltnthomlH '-==:IwIII lIant.us Ifll I' :a:!!!:!:: 1- ;~br.untonll j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j t..=:IO_:OII r-.. I (') I1.:_-====:=lOClorustoarpus ::T b.. \hH III crotabrfu. 1'1 prHlon<}Us III IIIlssourfl1lsis ('l &r9ophYllus rt Cl t tPhrodu (1) J II' :Clofbarfus 1'1 -.-:n :ol.ntfg-viJsonfi 1_ Clltntl\H>alans a Itntf!HIfl'hl/sus Opmfceus bol.oo.rt Cllllilacus ClmfnthornfH ...J II' --- . Cl919antttlS InO :Cl= ,0 :~~~f r====r=:IClMlafstratus 10 • !..nndtsp.,..tusClMlStfnH oil/1m owf 01 nus spathulltus ohloodts • dttrftalfs J II' I' ~i09fltnsfS -- !uvlllosus 1 ClStPultfPts omagdllanat-rll!ld mfdrf.lorum Clma9dlllnat;>tlrs onbulonum Clfnsuhrfs ClPutslftrH oplltl/lropls Clamb\llot ropf. I I' iomonotllsls I_. .arfzonfous dfstortus Cl9t11trl/I 'Cl nut am lnuS-aust i =Clnotho)(\IS oobnnsis

~~~Ij ...... w.w...... ""-a ... _ ....- ...... §iz:. dlc!Vrnooarpus obrazo.nsfs r.=: fllocltVthndl 1Jill I [1 ~--~~ otrvcldts j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j Ii' UI. .1.. • I "1 " "lIDflll~s IUcurvio.rpus Ic,p,lrooarpUl ulKOO¥PUJ bfc,.t.tatus '"Ij t..:=:=JoolllPtod.. H 111111(1-1111 I ,ostnuatus G') oalvorcHnsis CD "CI I» (II .... .t:I 0:: 0-' 0ool11nus C - :I CD II I ," lllIoallfornlcus -CCC'llI'D CII lonohooarpus o'" < :I 0 ° o C'lI owoocruffl C'lI 0 = 0 lID oonvallirl us (") a -d I» = tlhnophllus ::T - =: ,...CD =a lox!phold.. III C'lI ... I'D -' t.=:=Jo·plsoopus Ii 1~.===::J·ulinotlrius III n ~lIDt",.llUS rt 1/iI'II fj.. i. IE:~~:=s (l) Ii IIDtO'nus l!II~tlnatus (l) nVHltus < oadanus o l!ll00njWlOtus t-' l!II_noop~sls ~ 1!II0000al\lOis ,....rt .a1bulus o .flavus ::l I!IIrllO.IlOSUS I!IIblsuloatus ,.... ::l ~r;i!:!:!;;;;;:=.=!:!:Il3;~~I~sUS gr.OI1lS III . oastl ,....rt ~oorlllOlous 1ft iiiillllstr '"d --whltrwVI ~ t-' cuslolcll (l) - ampuu.nus III aOOlcHM t n ohopodus-t n­ ,...... t rl ohopodus-phox n I' ~.hornU ::r' ~PO!loMllsls III I Ii ~~nu\ault cfouQla.. 1I III n rt (!l I' ~cr .. t~art.M Ii p...... lo!l9us I' ~"'SS0Urfonsts t-' .r9ophllllus CO tltPh ... Cldts J I" uol~nus 1n BlftltlQ-wllsonlI l"'tlg-pallns r::=x::Jltnti o-ctI phI/sus '"lID punl MUS hollOO.,.1 lIIalleus II ~lIlnthornl" II ' 9Ig~nt.us J r:!!!:.n: on ~brauntonll InuoHlsts

III 5;~t;ilii-·- Ii II rr dou9las.11 III o I' -- - - nJDrusfoarpus rt I @btathll (l) I : croblart'M Ii prulo!l9us Rltssourt"nsf. t-' argophllllus co t~hrodu J nIl'ualblrlus 1! :u1flltlg-wll.ontt 1. - ;0 I.ntl9-palans r=J91.ntlg-dlphVsus !llIIIIPunla.us bolandwt IUlacus II I n.!::xJmfnthornfH ~glgant_ .rrID1'1 ~u=!!::= 0 . brauntonft 1. _ f~fllsfs -.hunlistratus dHp.,.atus· IUStfnM In!LJ~ll1illt oarlotnus .spathulatus .chlooc!H -,....Mtrttalts J\ \1 I' !;~~=! 1m; " :-m~UlttP" aa\ldalanw-tnaOd fdr1.toru. aagdalan..-p.frs I ~ , : :'abulOllUl!ltnsularl. pUlsif.rH plat\Jtropt. amblllot rOPf. I I' '~III. II a-·larlZontouS•

/I L!~==::xJnothoxlJle=~:-aust .-.cobr_fs :"1I:!:3r:5r.ourvUJ

I' ~~okttsalmonl. atratus I'. ~r.motusenMttcus mlllll~\I ~~~t r;: n. . = .~.fS'UUS 10 "..... '. .. '@~~~~~¥s 9allbtUfanus dldl/lDoOarpus I braZOfllsls J I'I UY*0§~~£~~~~ l. ~~fl"YoldtS

1,6 .. •...... fllfpu ourvlcarpus sp.frocarpus scl.rocu-pus bfcrlshtus IlIlIIf;J.~~1IIU.. *!!,i OamptocMs I L...... I ,...-....tnultu. 1D'a~f\)f\)f\)~ alVOI"dIMt. '! !. II if if if " oollfm. c '§ (.ol PI) ~ PI) PI) calf fornl ous Co;JOCDC7I~PI) I' r;::::xJ~~onOhooil'PuS UI"6j n CI I L-, woodruffl --..JH !.i:l convlUlI'fus ,-"G") :. tthnophl1us n xlpholdts J1111 ,~Iscopus N _- lanonrlus .-. ttntUus .1 oroc..,st! s C") wln<,latlnus ::r­ ----.r:::::x:Jtomus II> ~ ...... ;>totinatus Ii nvtntus II> n IdIOUS rt 'mconJunotus (l) .otl'lOOpptnrls Ii 00011\101. Ilbulus (l) mllivus ~ mrlo.lDOsus o mblsulOitus .-. r;;::==:JCl==:::ICllltXUOSus s:: ~I r.=;t;hlllU rt \11"10111 • ..... ' o OIs.I ::s oorlaolous I' ',.1..,. ..." .....r:=JClwhHneVl ::s -----1 M L-,Cl ourl okf I n ampullarlus ::r­ lOofdMs Ii ---r u-:::JCltrlohopodus-trl­ o trlcllopodus-phox 8 horn" o Cl pOllontnsl. III L----' M !:::JCl nut 11111 o ::ICldou91lSsll 3 II Clcru.lcarpus (l) I l5,jmbnthll ::s I__ , : morotl1U"1nt s:: , mprltlon9us 3 I' ~CllIIlssourltnJI' c:r' - , _ Cllr9cph..,lIus (l) ohph.. odts t1 J I t= :::Jootbarfus r' n ltntio-wilsonll 1_ ::Clltntill-P alans n -- Itntio-dll>h\lsU. ::r­ punlcNS II> Ii obolancMrI II> mabou. n mlnthomlM rt ~Cl9f9anttu. (l) JrIIll 'omo11lssl,.u. t1 ll:::::-::;;:,:,::',:::::::::: .:.::::::.:::::::.:::.::' ::':.::: :'::::: ::::.:.:::;:.::m:~~~:::~t

;3 I' ----;~~~;;;Io;;;,u, CD It ,i," ~:~~~!~~n. ::s , mprulon9U1 s:: ;3 IOllllssouritnSl1 0' O·r90ph\lllUi CD ottphrod•• t-t J nI'LL..= ===::JO01 barius 1 : 1.nWrullsonli ol.ntfo-palans n - 1.ntlQ-dlph\jsus ::r' punloeus III Obolan

~ll! ~i!l!!!l!!!!~~![!!ljl!jjj~jiijj!§~llj§!!i!§!§~~~g" ... b~l1hr.'t$ ~dfdl/moo.rpus jill '::;:;;;;at:::~:~~usIII !i :;.'!:::.::: _..-voldts

141

(Fig. 10). Generally, however, the pods in these groups are laterally compressed at least early in development. The

Megafloroid group undergoes modification of the trigonous cross-section in a basically dorsi-ventral mode, often with rudimentary inflation of the walls -- or in some cases extreme inflation (Fig. 12). These fruits tend to be gently incurved or strongly beaked and incurved (#45,46), rarely straight as in the Homalobi. The Inflati-Villosi group is quite heterogeneous with respect to fruits: either strongly inflated (#47) or greatly reduced in size and then variable in shape and cross-section.

As Barneby predicted, the valves are primitively papery in texture, later modified to either fleshy in the

Megafloroid clade or membranous in elements of that clade and the Inflati-Villosi group (#33; Fig. 13). Likewise the primitive ovule complement appears to be about 8-28 (#40), with more ovules among Megaflorae and several recurrences of few ovules throughout the basal paraphyletic assemblage,

Homalobi, and Piptolobi. The stipe (#42; Fig. 14) is polymorphic in the basal taxa and hence it is difficult to determine whether it is primitively present or absent; it is prevalent in the Homalobi -- though absent in certain subcladesi it is essentially absent in the Megaflorae and piptolobi. 142

The persistence of the fruit (#39; Fig. 15), a character that played a major role in Barneby's conception of higher taxa, is primitively persistent. Deciduous fruits arise in two major clades, once in the Micranthi radiation and once in the Piptolobi group, as well as in a few small groups within the Megaflorae. Persistent fruits characterize the rest of the Megaflori and Homalobi and recur once within the Piptolobi. The deciduous fruit appears correlated with the annual habit and adaptation to low desert conditions.

Flower and inflorescence morphology: The flower in

North America is exceedingly variable in size (#30), though less variable in shape (#27, Fig. 16). At the base of the group flowers are moderately small to very small in some derived groups. Generally, such flowers are only moderately incurved, but strongly incurved petals have evolved in each of the three large apical clades (although only in the small-flowered base of the Megafloroid clade). Small flowers are common in the Piptolobi group. Along with characterizing much of the Megaflorae group, large flowers have evolved independently several times, most frequently in the homaloboid clade. They are then frequent in species that are robust and leafy with long inflorescences. Barneby suggested that the large flowers of certain Megafloroid groups (e.g., Sarcocarpi, Mollissimi, Argophylli) are 143 qualitatively different than other large flowers, but there is no phylogenetic support for the single origin of any such flower. The large flowers of many Coastal and Plains groups

(e.g. Trichopodi, Bisulcati) tend to be declined in a dense inflorescence (#25), with an oblique and rather abbreviated calyx tube. This syndrome has arisen in both the

Megafloroid and Homaloboid clades.

The pattern of inflorescence evolution is less clear

(#25). The basal groups generally have long, loose

inflorescences, giving rise on several occasions to long,

subcapitate inflorescences. The long and loose raceme

becomes contracted with a reduction in flower number in the

Megafloroid group, and more sporadically in the Homaloboid,

and Piptolobi groups. In the first two of these, several

origins of long, very dense racemes also are indicated. In

a few groups there is polymorphic reduction of the peduncle

such that a sub-sessile inflorescence.is occasionally found (A nutriosensis Sanderson of Mollissimi, A. tenellus, some

members of the Kentrophyta group and Drabellae).

Vegetative morphology: This analysis uses a large

number of characters based on vegetative morphology (24 of

57), which heretofore has been a neglected source of

taxonomic characters. Only a few of these (stipules, hair

attachment, nature of the root) were used by Barneby to any

great degree in his assessment of higher level 144 relationships. Although the homoplasy in these characters is not systematically greater than other characters, they do not appear to consistently define as many higher-level taxa as the reproductive characters do although there are exceptions, such as the subterranean caudex uniting much of section Lonchocarpi. The subterranean caudex (#1, Fig. 17) arises 12 times on the phylogeny, but is never reversed once (this despite the use of the ACCTRANS option of PAUP which tends to prefer reversals over parallelisms in character state reconstructions). Similarly, the annual or shortly perennial (#2; Fig. 18) life-cycle arises 6 times but never reverses. Evidently these apomorphic states represent specializations which tend to inhibit the reacquisition of the plesiomorphic state. Both conditions are mostly absent from the Megafloroid clade. The annual habit is confined to the basal group and to the Piptolobi group. The number of leaflets per leaf (#10) is a surprisingly useful indicator of larger-scale relationships. Though the character was unordered, its reconstruction on the final tree never entailed a jump between few and many leaflets

(i.e., 0 <-> 2, omitting the intermediate state). Such behavior is reassuring for continuous characters. Few leaflets is often coincident with general trends in leaf reduction such as in the Lonchocarpi group. Numerous 145 leaflets characterize the "Pacific Clade" of inflated-pod species (NODE "t") , as well as a group including the

Pectinati and Reventi-Arrecti (in part).

The shape and degree of folding of the leaflet along the midrib were also useful characters. Linear leaflets

(#14; Fig. 19) arise 10 times and are lost twice. The support for a large clade within the Homalobi (NODE "W") includes the origin of linear leaflets. Likewise much of the spiral-fruit group is diagnosed by the loss of linear leaflets. Presence of flat leaflets (#12) comes and goes more sporadically, but flat leaflets do seem characteristic of some smaller groups of species (e.g., Sarcocarpi, preussiani) .

Connation of the stipules into a sheath on the side of the stem opposite the petiole (#18, Fig. 20) is a character that heavily influenced Barneby's classification. Dimorphic stipules (basally connate-apically free) arise 12 times from free ones and fully connate stipules arise 4 times. Free stipules are derived by reversal on three occasions. The reversals never unite taxa except once, in which case they unite a pair of species. Connate stipules define one very large clade in the Homalobi (see NODE "e" above), and originate frequently in the rest of that group. They are almost entirely absent in the Megaflorae and sporadic in the remainder of the tree. 146

Medifixed hairs (#23) arise a number of times, usually in one or a few closely related species (e.g., A. flavus and A. albulus). Their most significant impact is at NODE "0", where they unite sections Humistrati, Drabellae, Desperati, and Chaetodontes -- the latter two groups of which lack medifixed hairs. Although it is by no means impossible that medifixed hairs could have reversed to basifixed, this clade remains suspicious for this reason.

cytology: Chromosome number (#57, Fig. 21) is a character that has been heavily used at one taxonomic level and completely ignored at another. It is the only reliable evidence that suggests the cohesion and monophyly of the New

World taxa on account of a difference in basic number, and yet the variation within the New World has rarely influenced

taxonomic work there. Spellenberg (1976) suggested a close

relationship of several groups that are at the base of the

present topology based on the preponderance of higher than

average chromosome numbers in those groups. The present

analysis confirms this observation. Counts of 2n= 26,28,

and 30 are more or less restricted to the basal groups. [Note that part of the reason the tree was at A. ervoides was because of its count of 2n=30; see discussion above

under "Rooting".] Most members of the three large terminal

clades all have 2n=22 or 24. Aneuploid reduction has been

the most frequently postulated mechanism of chromosomal 147 evolution in New World Astragalus (Barneby, 1964; Ledingham and Pepper, 1973; Spellenberg, 1976), but of 16 state changes on the final topology, 9 involve increases in number. Spellenberg's prediction (Spellenberg, 1976:475) that aneuploid increase has been a factor in Astragalus evolution is supported. In fact, aneuploid increase from

2n=22 to 24 defines at least one major group (see NODE "f" above) . 148

CHAPTER THREE

STATISTICAL TESTS FOR HOMOPLASTIC TENDENCIES

INTRODUCTION

Interest in rigorous studies of patterns of character evolution has increased with the increasing availability of cladograms for many groups of organisms. Studies of adaptation (Ridley, 1983; Greene, 1986; Coddington, 1988), correlated evolution of functionally dependent characters

(Sessions and Larson, 1987), and coevolution (Sillen­

Tullberg, 1988), have benefited significantly from the incorporation of phylogenetic information (Donoghue, 1989).

At the same time, techniques for uncovering pattern in nature have been refined by the introduction of null models, which allow patterns to be distinguished from randomness

(e.g., Simberloff, 1987). Null models of character evolution are in their infancy but promise to play an important role in detecting patterns requiring explanation from evolutionary biologists (Ridley, 1983; Felsenstein,

1985bi W. Maddison, ms.).

An important pattern of character evolution that has been discussed by systematists from Darwin to the present is

"tendencies," or recurring instances of homoplasy confined

------149 to some groups and absent from related groups (Darwin, 1909;

Haacke, 1893; Scott, 1891; Eimer, 1898; Osborne, 1902:

Wernham, 1912: Arber, 1925; Rensch, 1959: Simpson, 1961;

Throckmorton, 1965: Cronquist, 1968: Mayr, 1969: Saether,

1983; Gosliner and Ghiselin, 1984: cantino, 1982, 1985: see discussion in Chapter One). Often the repeated origin of a peculiar morphology in more than one group is viewed as evidence of relationship of those groups. Thus, in addition to the evolutionary implications that such tendency patterns may have, they can affect the way phylogenies are reconstructed (e.g., cantino, 1982). However, despite many anecdotal references to tendencies, there has never been a systematic analysis of tendencies among a large suite of characters within one group, or across groups. Nor has any sort of statistical test aimed at detecting tendencies with confidence been proposed. Proponents of the use of statistical models in ecology and biogeography have long noted the ability of the human mind to extract pattern even from random data. It is therefore imperative that putative patterns like "tendencies" be subjected to statistical tests to determine if there actually is a pattern. In this chapter such a test is proposed and applied to a suite of characters used to reconstruct the phylogeny of North

American Astragalus (Chapter Two). In addition, a cladistic 150 analysis of Iguanid lizard genera (Etheridge and de Queiroz,

1988) is examined using the same methods.

The relationship between homoplasy in general and tendencies as a specific topological pattern of homoplasy has never been explored. In principle, tendencies might appear in at least two logically distinct forms. In one type, all homoplasy in a given character might be confined to a particular subregion of a tree. In the other, homoplasy might be more broadly distributed, but wherever it

is found, additional instances of it are nearby. The first

form of tendencies, "localization," can be measured as the size of the smallest group containing all of the homoplasy versus the size of the study group as a whole. The second

form of tendencies, "clustering," can be measured by the average distance of each homoplastic event to its nearest neighbor. These two forms of tendencies could often be

associated with each other, but need not be; a character might be relatively clustered but unlocalized, or vice versa

(see Fig. 22). In this study, I focus on clustering rather

than localization per see Clustering implies that

regardless of the position of other homoplastic changes on a

tree, homoplastic changes are likely to occur near some

other change. This pattern represents non-independence of

character evolution in the sense that a change in one

position on the tree makes it more likely that a similar 151

A

FIG. 22. Patterns of clustering and localization of homoplasy. A) Homoplasy is all confined to a subclade 0: the tree, but is not very clustered within that subclade. 3: Homoplasy is found in two regio~s distant from each other (unlocalized), but is clustered within each region. 152

change in the same character occurs in closely related taxa. Such a general pattern seems in keeping with the original meaning of the term "tendency." As discussed in Chapter One, both manifestations of "tendencies" may be apomorphic or plesiomorphic (Fig. 1), because they supposedly arise from underlying genetic similarity of taxa, and similarity need not be apomorphic. Any measure aimed at the detection of tendencies should do so without regard for the polarity of the characters involved.

METHODS Reconstructing the Distribution of Homoplasy -- Before topological patterns of homoplasy can be investigated, they must first be reconstructed using the topology and the distribution of character states among the terminal taxa. This is the general problem of character "optimization" (Swofford and Maddison, 1987). Parsimony is used to determine the ancestral states of characters at each node and the location of all changes in characters. Parsimony algorithms minimize the number of changes in the character that is being optimized. In some cases when there is .homoplasy in a character, there is more than one topological distribution that entails the same number of steps; hence the optimization is ambiguous or equivocal. Usually this occurs when instances of homoplasy are near each other (see 153

o o o o

FIG. 23. Alternative reconstructions of character state changes (hence the distribution of homoplasy). For the tree given and the data as shown, two equally parsimonious reconstructions are possible. The first (labelled a and a') represents parallel origin of the apomorphy~ the second (b and b / ) represents origin and then loss. The former would be reconstructed using the DELTRANS option of PAUP and MacClade, the latter using the ACCTRANS option, which is the option programmed in the present work.

------_._------154

Fig. 23). Available algorithms offer various options for selecting among alternative equally parsimonious reconstructions. In PAUP and MacClade the ACCTRANS option prefers origination followed by reversal, while the DELTRANS option prefers parallel origination. In the present analysis, the ACCTRANS option is used throughout and is written into the program listed in Appendix 4. Maddison

(ms.) has investigated the effect of using one or the other option on the generation of null models of character evolution and has found that it has very little effect. It is likely to have little effect in the present case as well, because ~lternative optimizations do not generate topological distributions of homoplasy that vary widely in their levels of clustering (Fig. 23).

Another problem is that parsimony optimization will fail to detect certain instances of clustered homoplasy. In particular, parallel origins in two adjacent branches will never be detected, because it will always be more parsimonious to infer one change rather than two (Fig. 24).

For purposes of detecting clustering, parsimony will lead to a conservative test.

It was noted in Chapter Two that extensive polymorphism among terminal taxa was scored in the North American

Astragalus study. Such polymorphism clearly entails additional instances of homoplasy. In general, such 155 o o 1 1

FIG. 24. Illustration of hidden homoplasy that can never be inferred using parsimony. Shown in open boxes are the parallel origins of a trait, while the single closed box represents the non-homoplastic state that parsimony would actually infer to have arisen. 156 homoplasy has been ignored for the present and has not been incorporated into the computer program developed for this study. The main reason is uncertainty about how to score the occurrence of such homoplasy topologically. This problem is investigated below in some detail for the

Astragalus leaf reduction character, and some possible solutions are discussed.

Measures of spatial pattern -- An objective statistical test

is needed to detect spatial patterns in the distribution of homoplasy on trees, and to distinguish significant. clustering of homoplastic changes from chance associations.

Such an analysis requires two elements: first, a measure of the deviation of the observed from the expected pattern, the

expected pattern having been derived from some appropriately chosen model, and second, the probability distribution of the chosen statistic, which is needed for tests of

significance of observed deviations.

The problem of detecting spatial pattern has received

serious attention from statisticians (Bartlett, 1975;

Diggle, 1979), motivated primarily by interests of

ecologists and geographers (Pielou, 1969; Boots and Getis,

1988). Such work is typically aimed at distributions in a

one- or two-dimensional space. However, there is no reason 157 why the methods cannot be carried directly over to the analysis of spatial patterns constrained to lie on a tree. One major type of spatial pattern is the distribution of points on a line or plane, which is entirely analogous to the distribution of homoplastic "events" on a tree. Deviations from a "random" (see below) distribution of such points can be in the direction of clustering or regularity. Two classes of analyses used to detect such deviations rely on quadrat sampling or distance methods. In the former, the number of points that fall within selected subareas is analyzed; in the latter, the distances between selected points is the focus. Distance methods seem more appropriate in cases with few data points, such as the present one, in which only two to ten "points" (homoplastic changes) are likely in a given character. There are statistical arguments for and agains'c its general superiority to quadrat methods (Diggle, 1979; Ludwig, 1979). The decision about how to. measure distances between changes on a phylogeny is not trivial. In the present study two measures are used: (1) "branch number," the number of internodes between changes, and (2) "branch length," the total number of character state changes (among all characters) per internode, summed over the intervening internodes (see Fig. 25). The former is sensitive to the rate of cladogenesis, the latter to the rate of anagenesis. 158 A third measure would be absolute time of divergence, but this information is rarely available in phylogenetic studies. Many distance methods have been discussed in the extensive literature on the subject (see reviews in Boots and Getis, 1988; Diggle, 1979). Most rely on variations of "nearest neighbor" methods, in which the distance measured for a given point is the distance to that point's nearest neighboring point. Higher order methods which simultaneously consider the nearest neighbor and more distant neighbors have received much less attention (Diggle, 1979). When studying spatial patterns in the real world a sampling procedure in which only a subset of points is scored must generally be employed. For "mapped" data, in which the locations of all points are known, more refined statistical techniques are available, and these are the appropriate analogs to the case of interest in this paper, because the location of all instances of homoplasy are known for a given tree and character. Clark and Evans (1954) used a test based on the mean nearest neighbor distance, which is the average for all points of the distance to that point's nearest neighbor. Although it is perhaps the simplest possible test, Diggle (1979) claims that it is generally not a particularly powerful test of goodness of fit of observed to expected .

. --.. -. --'-"-.'.- -.------~ 159

c

distances

a-b 3~3 a-c 4~13 b-c 2~10

FIG. 25. Illustration of various methods of measuring spatial pattern on a tree. In this tree three changes occur. Next to each branch (internode) is the path length distance. In box are distances to between changes: the first is the distance in terms of simple number of intervening branches; the second is the branch-length distance. 160

He proposes an alternative ("refined nearest-neighbor analysis") based on the distribution function of observed nearest neighbor distances for all points. For a set of nearest neighbor distances, the empirical distribution function (EDF) is a function which, for each distance X, gives the probability that observed points have distances less than or equal to X. In other words, it is a probability distribution function in the strict probabilistic sense (see Rohatgi, 1976). A mean EDF is generated by Monte-Carlo simulation. For N replicates of a simulation using some null model to generate points, an EDF for each replicate is generated, and its fit to the mean EDF is measured using some criterion of goodness of fit. The distribution of that fit statistic is then used to test the significance of the observed fit of the data's EDF to the mean simulated EDF. I investigated the relative merits of both traditional and refined nearest neighbor analysis using the program described below (see especially the functions

"do- EDF", and "translate-- NN EDF" in listing), and concluded that few differences resulted from the choice of approach.

Levels of significance were comparable for all characters; characters with higher confidence levels for clustering using refined NN analysis were about equally matched by other characters with somewhat lower levels (Table 5). For 161 TABLE 5. Comparison of topological clustering using mean nearest neighbor statistic and empirical distribution function (EDF) for four selected characters from Table 6. Model III used for mean NN distance (branch number metric). The expected value for the EDF goodness of fit statistic is always zero, hence it is not listed in its own column Mean NN distance EDF goodness fit Character Obs. Exp. Probe Obs. Probe

(3) 7.0 14.2 .15 12.5 .02 (34) 8.0 9.5 .33 1.2 .40 (55) 6.3 7.7 .32 1.4 .27 (56) 8.0 9.5 .33 1.4 .35 162

simplicity, the simpler mean nearest-neighbor statistic was used in the following analyses.

Null models for the expected distribution -- In the case of points distributed on a plane an appropriate null model is "complete spatial randomness," or equiprobability of a point regardless of position. In ecological and biogeographic studies much debate has centered on the choice of null model. For example, Simberloff (1987) recently studied the issue of how improbable it might be for two area cladograms to match by chance. The null model in his case was that all distinct classes of tree topologies are equally likely to be selected at random, but many other null models are possible, such as equiprobability of all distinct trees (Farris, 1981) . Null models for the evolution of characters on trees have often taken the form of a Brownian motion model (Cavender, 1978, 1981; Sessions and Larson, 1987; Felsenstein, 1988), in which the probability of change along some branch depends only on some rate parameter and the length (age/time) of the branch. Generally practical constraints require that the rate parameters be homogeneous either among branches and/or among characters, which lessens the model's interest and utility (see reviews in Pagel and Harvey, 1988; Donoghue, 1989). In an analysis of correlated

------~------~ 163

A

FIG. 26. Illustration of A) completely pectinate tree, S) completely dichotomous tree. 164 character evolution, W. Maddison (ms.) suggested a null model in which character state changes are equally likely to occur on any branch, regardless of the length of the branch.

One shortcoming of such a model, which Maddison recognized, is that on a pectinate tree (see Fig. 26), some branches represent much longer periods of time than others, and it may be unrealistic to assume equiprobability of character state change along branches with very different lengths. In such a model, regions of a topology subject to rapid cladogenesis (e.g., speciation) will have higher proportions of character changes (DeSalle and Templeton, 1988; Mindell, et. al., 1989). This may implicitly presuppose a rather narrow model of macroevolution for the null model, in which anagenesis is highly correlated with speciation rate as opposed to time of divergence (which may be more appropriate in some cases). Another feature of this model is that the distribution of apomorphies among terminal taxa fluctuates wildly with the position of the randomly selected character changes. For example, two changes located at the base of the tree in one random replication of the null model might mean that 95% of the terminal taxa have the apomorphy, whereas if those two changes occur in parallel along terminal branches, only two terminal taxa will nave it (see

Fig. 27). Maddison's null model is termed "model III" here. 165

• A

FIG. 27. Behavior of Model III. For a fixed number of changes on a tree (here two), the number of apomorphic taxa fluctuqtes greatly depending on the placement of those changes. A. Changes appearing apically and close together result in few apomorphic taxa. B. Changes that are separa~ed more yield more taxa that are apomorphic.

------.. ------166

TABLE 6. Null models used and their assumptions.

Model I random permutation of N terminal taxa, where N is the number of taxa with apomorphy distances measured by either branch number or branch length metric Model II random permutation of N terminal taxa constrained so that exactly M changes in the character occur on tree, where M is the number of changes seen in the tree reconstructed from the data distances measured as in model I Model III random permutation of M character state changes where every branch is equally likely to have a change, where M is determined from data distances measured by branch number metric (although branch length could be used) Model IV random permutation of M character state changes where probability of branch receiving change is proportional to its length (number of changes along branch for all characters); M is determined from data distances measured by branch length metric (although branch number metric could be used) 167

An alternative null model ("model I") is suggested by analogy with those in use in compatibility methods of phylogeny reconstruction and character weighting (Meacham, 1981; Penny and Hendy, 1985). Two characters are "compatible" if one character can be mapped onto the phylogeny suggested by the other with no instances of homoplasy. Meacham studied the likelihood of character compatibility arising at random under a null model in which the observed number of apomorphic states of each character were distributed randomly among taxa. Clearly this is very different than a null model that distributes character changes among branches. Whereas the previous model constrained the number of changes and allowed the number of apomorphic states to fluctuate, this model constrains the number of states to that actually observed among the terminal taxa in the data set, and allows the reconstructed number of changes to fluctuate. Models I and III are most similar when the number of apomorphic states is very small and each arises independently in terminal branches on the tree. In such cases the number of changes equal the number of apomorphic states, each arising in parallel. The difference between the two models is that the model III will allow non-terminal changes in the character, which drastically alters the distribution of the character among taxa. 168

In some replications of the null model, the distribution of apomorphic states among taxa will be such that neighboring taxa have the apomorphy and in fact are sister taxa united by that apomorphy. In such cases the number of changes is reduced. This suggests a third null model ("model II") , jointly constrained as in both the previous ones, so that exactly M terminal taxa have the apomorphic state and exactly N changes occur on the phylogeny. Some characters are found in a large number of terminal taxa, yet change state only a few times (N«M). such characters change deep within a topology so that each time they do change they diagnose rather large groups and are therefore quite important. This model is computationally nearly intractable for all cases except when M and N are nearly equal and both are rather small, and therefore has not been investigated in much detail (see below) . Finally, a null model based on the total amount of observed character change on branches is suggested. Model IV entails weighting the random distribution of character change by the length of branches. Thus a change is twice as likely to occur along a branch of length 10 than along one of length 5, and no changes occur along zero length branches. This model is meant to incorporate the observation that observed rates of anagenesis fluctuate 169

widely on trees. Some branches have evidently undergone

very rapid evolution simultaneously among many unrelated

characters. Alternatively, such branches may imply

unusually long periods of absolute time. It is therefore

possible that instances of homoplasy are more likely to

occur along such a branch than along branches that are

relatively static.

In each model, both the branch number and the branch

length metric of distance could in principle be used to

assess the spatial pattern of homoplasy. However, it seems

more consistent to use the metric that is naturally

associated with each of the null models; in model III the

branch number metric is the natural choice because

characters are randomly assigned to branches regardless of

their length, while in model IV the branch length metric

seems more appropriate since the probability of a "hit" is

proportional to length. In models I and II, either met~ic

is appropriate, since in these it is the terminal taxa that

are randomized, not the branches.

The assumptions of all the null models are summarized in Table 6.

Tests of significance -- Given any of the null models

discussed above, it is possible to generate the expected

distribution of some measure of goodness of fit of observed

------170 to expected. A computer program has been written (Appendix

4) that generates these distributions by Monte-Carlo simulation. Both measures of spatial pattern discussed above have been investigated. In each case, for the given observed topology, a large number of replicates (500) of the null model is effected, the distribution of the spatial statistic is recorded, and the significance of the statistic observed for the data is compared to the distribution from the Monte-Carlo runs. For example, if the observed mean nearest-neighbor distance for a character in the data is smaller than the mean observed in 95% of the Monte-Carlo replicates, then the null hypothesis can be rejected at the

95% confidence level.

The computer program -- The program listed in Appendix 4 is written in C, and is entirely portable to implementations of

C that have standard IO functions for reading and writing to disc files and the terminal. Its memory requirements are modest, but it does rely extensively on recursive functions for moving through tree topologies, and the stack may have to be increased to ca. 4-8 kbytes when trees have over 100 taxa in them. The program runs quickly, even on the 8 Mhz.

Leading Edge Model 0 used for the analyses described below.

A simulation with 500 replicates, working on a tree with 113 taxa in it, ran in under 15 minutes. considering that for 171

Model I each replicate included a complete re-optimization of all ancestral character states, this is quite fast. Each of the four null models discussed above are implemented. Model III is the simplest and quickest, since it merely requires the program to randomly assign changes to branches and then calculate the nearest neighbor distances. Model I is much more involved, since after the apomorphic states are randomly assigned to the terminal taxa, the location of character state changes must be inferred by assessing ancestral character states. Farris optimization procedures were used following the algorithms given in Swofford and Maddison (1987). Note that the ACCTRANS option was selected (see above). Model II was implemented with a "brute force" approach in which the number of apomorphic states is fixed at the outset of a loop, and then the loop is simply repeated until the precise number of changes on the tree is achieved. In practice if these two numbers are very close to one another, it does not take many iterations of the loop, but when they differ significantly the time required to randomly generate exactly the specified number is prohibitive, and other algorithms would be needed in such a case. In Model IV, a list of nodes on the tree is generated in which each node appears exactly k times, where k is the number of changes along the internode underneath it. Random 172 samples from this list produce a selection of changes weighted by the length of the internode.

Input to the program consists of a file containing a description of the phylogeny, including branch lengths for model IV runs (see Appendix 4 for format), a data matrix, and a specification of the model requested. Output consists of a table such as is found in Tables 8-10, which gives observed and expected values of the mean nearest-neighbor distances (with significance levels) for each character under various null models.

studies employed and characters analyzed -- The methods were applied to the cladistic analysis of North American

Astragalus discussed in Chapter Two and to a recently published cladistic analysis of Iguanid lizards (Etheridge and de Queiroz, 1988).

These methods were applied to the cladistic analysis of

Chapter Two because an understanding of the details of the characters involved should prove helpful. In particular, three focal characters were selected prior to the present analysis because of their phylogenetic and evolutionary interest. The first character is a striking red pigmented flower, with an associated pseudo-tubular morphology probably representing a syndrome for hummingbird pollination

(see Fig. 28). The flower has evolved three times in North FIG. 28. Parallel trend in floral evolution in three clades of North American Astragalus. Camera lucida drawings of flowers from three species pairs are shown (plus A. argophyllus see Appendix 3) in the last pair. Arrows are meant to suggest the transformation from the inferred ancestral state. Marking such ancestral states by extant species is not strictly correct, but these taxa have floral morphologies characteristic of other outgroup taxa that would be used to correctly infer ancestral conditions. A. ~ crassicarpus var. crassicarpus --> A. sanguineus. B. A. mollissimus var. bigelovii --> ~ helleri. C. A. argophyllus --> A. newberryi --> A. coccineus. Specimens drawn respectively from Warnock 20024 (TEX), Riskind and Patterson 1965 (TEX), Sanderson 501, Balls and Gourlay 4462 (NY), Sanderson 562, Sanderson 517, Sanderson 312. 173 IflfiIJI,IIlililltll,lftllfl£IIJifllllllllliilll,ltltlfllllltlllllilililillllflilliillllilltllilliflill,llllilill

FIG. 29. Topological distribution of homoplasy in red­ flowered morph in North American Astragalus. Shaded boxes represent originations, shaded diamonds represent reversals.

I-' ...... *'" 175

TABLE 7. Factor loadings of first two principal components used in morphometric analysis of red-flower evolution. Analysis included six species: ~. mollissimus, ~ helleri, ~ crassicarpus, ~. sanguineus, ~. newberryi, and ~. coccineus. Characters used (all log-transformed): (1) calyx length, (2) length of calyx teeth, (3) calyx width at teeth; (4) wing length, (5) length of blade of wing, (6) width of blade; (7) keel length, (8) keel width; (9) banner length, (10) distance from top of banner to bottom of keel. Total N = 65. Character Component 1 Component 2 ------1 -.385 -.081 2 -.316 .043 3 -.251 .398 4 -.388 -.073 5 -.363 .072 6 -.080 .819 7 -.381 -.074 8 -.178 .045 9 -.385 -.059 10 -.274 -.376 ------%variance .616 .121 FIG. 30. Parallel trend in floral evolution in three clades of North American Astragalus. Axes represent first two factors of peA on ten linear measurements of flowers from three species pairs. Arrows connect the centroids of the species distributions, and are meant to suggest the transformation from an inferred ancestral state. Marking such ancestral states by extant species is not strictly correct, but these taxa have floral morphologies characteristic of other outgroup taxa that would be used to correctly infer ancestral conditions. In the case of section Argophylli, the centroid of A. tephrodes is also included, since it is more representative of the prevalent outgroup condition (see Appendix 3). See Table 7 for characters used in analysis.

______.. ____ ....______.. ----t___ _ 176

Accccinaus Onawbarryi [Jsanguinaus Acrassicarpus ehellari • mollissimus 4r------+------~~----~------~

c

3

2 c c c c

c • • he .. N ..0 • • a ~ a C E-o .. ... Co) • .. .. • a :: 0 0 erne 0 0 Ite .. ne .. • ..---e .. trcr .. co ()o . , .. 0 0 a 0 ..II .. ..

.. 0 .. 0 ·2 .. ..

FACTOR 1

".-- -._._------177

America (Barneby, 1964; confirmed by the cladistic analysis of Chapter Two, see Fig. 29), and each origin has entailed lengthening, relative narrowing, and straightening of the petals, as well as modification of the purple color to bright red. A morphometric analysis of the flowers of the three red-flowered species, as well as their three respective outgroup taxa is shown in Fig. 30. The PCA diagram shows a parallel trend towards larger flowers in each group, despite some evident divergence along the second

PCA axis, which represents a relative width factor (Table

7). Barneby (1964:753) explicitly suggested that this parallelism represented a tendency uniting the three sections of the genus containing the red-flowered species.

The second character is a feature of the compound leaf usually associated with a trend towards reduction of the leaf (Fig. 31). In some taxa, the terminal leaflet (and often the lateral ones as well) loses the joint, or petiolule, with the rachis of the leaf, becoming confluent.

Often the terminal region of this "rachis" then becomes lengthened or flattened-expanded or both. In extreme cases all lateral leaflets are lost and the leaf forms a unifoliate "phyllode," although the development in such cases has not been studied and it is probably inappropriate to apply that term. Reduction of the compound leaf occurs

in at least three other ways in North American species. In FIG. 31. Illustration of types of leaf morphology. A) Typical imparipinnate leaf with jointed terminal leaflet (~ mollissimus var. bigelovii, Sanderson 507). B) Leaf reduced to solitary jointed terminal leaflet (A. musiniensis, Sanderson 566). C) Terminal leaflet confluent with rachis, usually in the context of linear leaflets (A. ceramicus, Sanderson 554). D) Lateral leaflets lost entirely, leaving "phyllode." The latter is interpreted as terminal confluence, based on related taxa (A. lonchocarpus, Sanderson 558). E. Confluence of terminal leaflet perhaps by paedomorphosis in A. magdalanae var. peirsoni (Alexander and Kellogg 1937, ARIZ).

fll!~ i II-~! ! ~I ~ ~I- , I ! I-I - -~ !Ii! II' · 21--:' 9 - IICI I -,f-. ,. I J 11-]1 ~'Iiililli 11!I!lfIIJlill ti!IJJ~!lilllilssilil!i~=!llljjjllllli illiljlili!~il§ liiilllli~lliii!illilillli611 It

FIG. 32. Topological distribution of homoplasy in terminal leaflet character (character 15) in North American Astragalus. Shaded boxes represent originations, shaded diamonds represent reversals.

I-' ~ \D 180

a number of groups typified by relatively few leaflets, such as A. calycosus, A. newberryi and relatives, and species in section Orophaca, the terminal leaflets remain elliptic­ orbicular and jointed, but often the lateral leaflets are lost. In A. asclepiadoides, a truly bizarre unifoliate leaf is produced, which is large, thick, leathery, orbicular, palmately veined and essentially clasping the stern. In A. magdalanae var. peirsonii, a potentially very instructive variety, the terminal and lateral leaflets are confluent, but they appear not fully developed, as though the marginal meristems aborted very early leaving rather short linear­ truncate lateral projections on the rachis. It would be possible to examine the localization of homoplasy in the character "leaf reduction," but given the diversity of probable developmental types, it is more instructive to examine only the specific leaf reduction involving confluence of the terminal leaflet. This seems more likely to stem from the kind of minor underlying synapomorphy that proponents of tendencies argue will produce clustering of homoplasy. Initial phylogenetic analyses suggested a pattern of clustering of this character in the general clade shown in Fig. 32, a group of taxa with unilocular, laterally compressed fruits. The third focal character studied, the presence of a longitudinal septum in the legume, never appeared clustered 1 , 11 fJlg- i II-i ! ~I' II- I' I ')-1· -I !ji 11' • 11-- '. Jiell 'ii-. ; - I J li-11 ~~lij;l!flili!illfIIJlill fJj~jJj!)i.lltl!!IJiJ!ii:!lliljIIJIIII iljii~lllil~i i lii~IJll1111lillliijltJlli611 It I

FIG. 33. Topological distribution of homoplasy in septum character (character 38) in North American Astragalus. Shaded boxes represent originations, shaded diamonds represent reversals.

I-' 00 I-' 182 in any previous analysis; nor does it on the phylogeny shown in Chapter Two (see Fig. 33). Details of the anatomical structure of the legume are discussed in Appendix 2 in the character discussion. I examined the development of the septum in three species using serial sectioning. Variation in the degree of intrusion of the septum seems to stem from variation in the time of initiation of growth of the subepidermal meristem near the dorsal suture. In A. gypsodes, a species with a fully bilocular pod, the septum is initiated early and is in contact with the ventral (adaxial) suture prior to anthesis. In A. mollissimus var. bigelovii it is initiated later and is only about 1/2 of the way across the locule at anthesis. At maturity, this septum is not fully fused with the adaxial suture. In A. tephrodes the septum is not at all developed prior to anthesis, and often does not develop at all; when it does, it merely projects slightly into the locule. A continuous character of this type might be expected to be labile. In addition to the three characters discussed above, 14 other binary characters were selected from the data set of Chapter Two so that the presence of a general trend towards a particular spatial pattern over all characters could be detected. These characters do not represent all of the binary characters from that data set, but they were selected to cover all aspects of the morphology of the organism, 183 including vegetative, floral, and fruit characters, as well as its known biochemistry. Not all characters were included because not all characters were scored with equal reliability. The chosen subset represents characters with well-characterized discrete states, whose apomorphic conditions were scored unambiguously across all taxa. A few mUltistate characters were included by converting them to binary characters by considering only transitions to and from a single apomorphic state. Thus for character 45, the acquisition of spirally incurved fruits, which is one state of a three state character was studied by noting character changes that only involved that one state. In Etheridge and de Queiroz's (1988) study, all 21 binary characters in their data set were analyzed (see Table 10 for identification of characters used and their states).

RESULTS Behavior of the null models -- The behavior of the first three null models was examined for two test topologies, each with 75 taxa. One topology was fully "pectinate," or asymmetrically branching, whereas the other was "dichotomous," or symmetrically branching (see Fig. 26). Model IV was not studied theoretically because there are an infinite number of ways that different length branches can FIG. 34. Expected mean nearest neighbor differences under null models I, II, and III, for completely pectinate and completely dichotomous trees with 75 taxa (see Fig. 11). Abscissa is the number of homoplastic changes for model III, or the number of terminal taxa with apomorphic states for Models I and II. Model II is run under the assumption that the number of apomorphic taxa equals the number of changes (M=N in text). 184

fit MOdel I (PICI) _Madel II (PICt) AMOCIal III (Pee:) o MOdIi I (OichOI) CModel II (CichOI) ,6MOCIal III (Oiehct) 28

26

24

22

20

18

:. " ...~ 16 ~ z z , 4 ;; :. 2:

12

, 0

8

6

2~ __.- __~ __~ __~ ______~ ______~ ______~ ______--+ o 2 4 6 8 , 0 • 2

Xumber of C.,an~es 185 be distributed on a tree, and no simple cases other than all branches being equal were evident. Note that for models I through III path lengths are measured in terms of the "branch number" metric -- the number of internodes between changes. The mean nearest neighbor distance for various values of the number of apomorphies/changes is plotted for combinations of the three models and topologies in Fig. 35. It is always true that the mean nearest-neighbor distance declines as more changes/apomorphies are included in the null model, which is intuitively reasonable. The striking effect of topology is somewhat surprising, however, but can also be understood readily once it is realized that path lengths on a pectinate tree tend to be longer than on a dichotomous tree. In fact, the maximum path length on a pectinate tree increases linearly with the number of taxa included, but the maximum path length on a dichotomous tree increases only as the base-two logarithm of number of taxa. Hence in large trees the difference in expected path length under almost any sort of null model will be much greater, which is the case illustrated in Fig. 34. With 128 taxa, for example, the maximum path length on a pectinate tree is 128 branches, while on a dichotomous tree it is only 18 branches. On a pectinate tree the differences between the mean nearest-neighbor differences for the three null models are 186 not very significant. This is counterintuitive given the fundamental differences in the null models. It arises because any assignment of an apomorphic state to a terminal taxon yields the equivalent path length to a given point as the assignment of a change to some internal branch.

Therefore the expectations are very similar. This is not true for a non-pectinate tree. In trees with more dichotomies, there are regions where assignments to terminal taxa will always yield path lengths longer than any assignments of changes to internal internodes. Hence in non-pectinate trees, it is expected that the mean path lengths for homoplastic changes will be shorter under model

III than models I or II. Models I and II themselves will always be very similar in large trees -- less so in smaller

ones and in cases with large numbers of changes. That is because in both instances it is more likely that apomorphic

states will be assigned by chance to neighboring taxa,

leading to synapomorphy instead of parallelism, and therefore reducing the number of changes scored in model I while model II looks for another replicate. In general,

both the topology and the particular null model used can

affect the outcome of a test of significance of some

apparent spatial pattern. In the following studies, Model

II is not considered since it gives results so similar to

Model I for the M=N case.

-_._--_. __ ... _--_ .. _--_ ..... --_.. _------187

TABLE 8. Levels of topological clustering in homoplasy for 17 characters of North American Astragalus. Observed and expected mean nearest neighbor measures for null model I (see text), with both branch number and branch length metrics used, are tabulated. Character numbers refer to characters in Appendix 2 and Table 2 of Chapter 2. Some characters could not be analyzed due to limitations on array size for Model I routines. Character numbers refer to characters in Appendix 2. Character (*) refers to the red-flowered character discussed in text. Column "#tax" indicates the number of terminal taxa possessing apomorphic state for that character, which is the key replicated in model I.

(Branch number) (Branch length)

Char #tax Obs. Exp. Probe Obs. Exp. Probe ------(1) (character out of array bounds) (3) 2 7.0000 14.9460 0.0960 14.0000 39.5520 0.0540 (14) (character out of array bounds) (15) 12 6.2857 6.0154 0.6680 19.8571 16.0042 0.9320 (23) 10 7.4286 6.5647 0.7840 18.5714 17.2696 0.6900 (26 ) 12 6.6000 6.0154 0.7800 17.9000 16.0042 0.8040 (28) (character out of array bounds) (31) (character out of array bounds) (34) 6 8.0000 8.5397 0.4420 23.0000 22.7348 0.5820 (38) (character out of array bounds) ( 40) (character out of array bounds) (45 ) 5 3.0000 9.2240 0.0020 9.6667 24.3677 0.0040 (47) (character out of array bounds) ( 50) 3 3.6667 11. 8367 0.0160 10.3333 31.4460 0.0180 (55 ) 9 6.5000 7.0155 0.4160 19.0000 18.5484 0.5780 (56) 12 8.0000 6.0154 0.9800 22.2500 16.0042 0.9860 (*) 3 6.3333 11.8367 0.0760 16.3333 31.4460 0.0840 188

TABLE 9. Levels of topological clustering in homoplasy for 17 characters of North American Astragalus. Observed and expected mean nearest neighbor measures for null models III and IV (see text), with corresponding metrics of branch number and branch lengths respectively, are tabulated. Character numbers refer to characters in Appendix 2 and Table 2 of Chapter 2. Character (*) refers to the red-flowered character discussed in text.

Model III Model IV (Branch number) (Branch length)

Char #ch Obs. EXp. Probe Obs. EXp. Probe

(1) 12 5.5833 4.3097 0.9220 14.5833 11.7928 0.8720 (3) 2 7.0000 13.5560 0.2080 14.0000 35.7160 0.1000 (14) 12 4.8333 4.3097 0.7560 14.1667 11.7928 0.8280 (15) 7 6.2857 6.1629 0.5580 19.8571 16.3614 0.792 0 (23 ) 7 7.4286 6.1629 0.7780 18.5714 16.3614 0.7140 (26) 9 7.0000 5.1913 0.9240 19.2222 14.2700 0.9120 (28) 10 5.5000 4.8774 0.7520 16.9000 12.6970 0.8960 (31) 6 7.5000 6.7777 0.6560 20.0000 17.7307 0.6660 (34 ) 4 8.0000 8.4715 0.4800 23.0000 22.4185 0.5840 (38) 9 4.4444 5.1913 0.3280 10.7778 14.2700 0.1860 (40) 9 5.8889 5.1913 0.7220 17.7778 14.2700 0.8420 ( 45) 3 3.0000 10.1167 0.0160 9.6667 27.6567 0.0500 (47) 8 7.5000 5.6213 0.9040 20.7500 14.9730 0.9300 (50) 3 3.6667 10.1167 0.0380 10.3333 27.6567 0.0500 (55) 6 6.5000 6.7777 0.5180 19.0000 17.7307 0.6200 (56) 4 8.0000 8.4715 0.4800 22.2500 22.4185 0.5400 (*) 3 6.3333 10.1167 0.1900 16.3333 27.6567 0.1680 189

Clustering in three focal characters (Table 8,9 and Figs.

29, 32, and 33) -- The pseudo-tubular red flowered morph arises three times on the cladogram. Although the three red-flowered species, b. sanguineus, b. coccineus, and b. helleri, were not included in the data matrix in Chapter

Two, their very close relatives, b. crassicarpus, b. argophyllus, and b. mollissimus, respectively, were (with regard to the characters of Table 3, these red-flowered species are scored identically with the taxa actually used in the cladistic analysis of Chapter 2). Indeed, as far as the characters used to reconstruct the phylogeny are concerned, the red-flowered species could have been sUbstituted identically for the species that were used. The observed mean nearest-neighbor distance is 6.3 branches and

16.3 character changes (Table 8). Under Model I (randomized terminal taxa), the expected values are 11.8 (P<.08) and

31.4 (P<.08) respectively (Table 8). However, under Model

III and IV (changes assigned randomly to branches), the expected means are lower and the corresponding levels of significance less (Table 9). The pattern is not significantly clustered, although it is close for Model I.

Confluent terminal leaflets (character # 15 in Table 3) changes state 7 times on the cladogram and is found in 12 terminal taxa (non-polymorphically). Its observed mean NN 190 distance is 6.2 branches and 19.6 character changes which exceeds the expected mean for all models. Hence there is no reason to reject the null hypothesis in this case. There are at least two ad hoc explanations for this somewhat surprising result. First, there are two topological outliers on the phylogeny. Astragalus magdalanae var. peirsonii has confluent terminal and lateral leaflets and is remote from the main activity in this character. Interestingly its leaves appear quite different from those in other taxa (see discussion above), and hence this instance of homoplasy may represent the action of a different developmental mechanism. Also A. cusickii is polymorphic for confluent leaflets and in the current topology this species is far removed from its relatives according to previous taxonomic work (see Chapter TWo). It is quite possible that the position of this species is actually much closer to taxa with confluent leaflets. Second, in general the inclusion of polymorphism tends to change the pattern significantly. To investigate this effect, a single change was placed along the terminal branch leading to each polymorphic taxon, but the nearest neighbor distance was corrected by one unit upward, so as not to bias the results. To be fair, the null models should have been re-run with one additional taxon for each instance of polymorphism, but this was not done, favoring the null 191

model. An additional 5 changes were now entered onto the cladogram and the observed mean NN distance declined to 5.3

(but expected for 12 changes is 5.2, hence P = .58). However, when the two outliers were removed, the observed

mean NN distance declined to 3.2 (P = .02), a significant pattern of clustering, but it was achieved at some cost to credibility. This example highlights the sensitivity that the results have to initial decisions about what comprises a character state of a character. The issue will be discussed again below in terms of character resolution. The longitudinal septum of the legume (character # 38)

changes state 9 times, often deep in the tree and characterizing large groups of taxa. surprisingly, its observed mean NN distance is somewhat below the expected Model III and IV values but still not significantly below expectation.

Clustering in 14 other binary characters -- A set of 14

,.r9ther characters, spanning the range of phenotypic variation in the group, were also examined for spatial clustering of homoplasy (Tables 8,9). Only two characters showed evidence of clustering significant at the 5% level in all models. These were the spirally coiled pod and bisulcate ventral surface of the pod, both of which change state three times). It is presumably not coincidental that significant results

.------. ------_._------._- -...... -- ._- .. __ ..------192 are found among the characters that change state very rarely, as the test is most powerful in such cases. Two other characters are localized at about the 10% confidence level in model I (though only one of these is also significant under models III and IV), but it is not surprising that among 17 characters a few would show this level of significance by chance under the null model. Clustering in the Iguanid Lizard study -- Out of 21 binary characters that were studied from Etheridge and de Queiroz's (1988) analysis, none showed significant levels of clustering at the 5% level (Table 10) for either model III or IV. Two characters were clustered at the 10% level, precisely as would be expected by chance.

DISCUSSION In the Astragalus study only 2 of 17 characters showed significantly clustered patterns of homoplasy on the topology. This is true regardless of the null model used. with respect to a subset (17 of 57) of characters used in the cladistic analysis of North American Astragalus, there is no evidence of a general trend towards character "tendencies." A typical character is no more likely to undergo homoplasy near another instance of homoplasy than would be expected at random. Particular characters do deviate from the null model significantly, but the number of 193

TABLE 10. Levels of topological clustering in homoplasy for 21 binary characters of Iguanid lizards (Etheridge and de Queiroz, 1988). Observed and expected mean nearest neighbor measures for null models III and IV (see text), with corresponding metrics of branch number and branch lengths respectively, are tabulated. Character column is the character number in the published analysis. Characters 22 and 26 are multistate, each with a lone autapomorphic state that is ignored, so that these are treated as binary for the purposes of this table.

Model III Model IV (Branch number) (Branch length)

Char #ch Obs. Exp. Prob. Obs. Exp. Prob. ------(1) 3 12.6667 8.8473 0.8520 38.0000 30.8840 0.6960 (4) 2 3.0000 13.8840 0.0980 15.0000 46.9320 0.1680 (6) 8 3.7500 3.7188 0.5620 19.7500 12.7587 0.9480 (11) 2 11.0000 13.8840 0.4500 29.0000 46.9320 0.3560 (13) (non-homoplastic) (17) 4 4.5000 6.7345 0.2000 21. 5000 23.7740 0.4160 ( 20) (non-homoplastic) ( 21) 6 5.3333 4.8187 0.6500 29.8333 16.1257 0.9920 (22) 3 9.6667 8.8473 0.6020 31. 3333 30.8840 0.5460 (24) 7 5.1429 4.1174 0.8240 21. 8571 14.3528 0.9560 (25) 6 5.6667 4.8187 0.7400 19.1667 16.1257 0.7080 (26) 7 4.4286 4.1174 0.6480 15.7143 14.3528 0.6140 (34) 11 3.4545 2.9438 0.7860 1.2.1818 9.3869 0.8540 (35 ) 2 24.0000 13.8840 0.8540 95.0000 46.9320 0.9160 (38) (non-homoplastic) (39) 3 4.3333 8.8473 0.1380 21. 0000 30.8840 0.3020 ( 40) (non-homoplastic) ( 41) 5 4.4000 5.4872 0.3560 19.2000 19.1048 0.5380 (42) (non-homoplastic) (48) 2 2.0000 13.8840 0.0620 11. 0000 46.9320 0.0960 (49 ) 6 4.3333 4;8187 0.3800 12.0000 16.1257 0.2500 194

these is small and consistent with the expectation that they

simply represent outliers of the random distribution. In a

sample of 17 null characters, about one character would be

expected to be significant at the 5% level. Here there are

two. In the Iguanid study (Etheridge and de Queiroz, 1988),

no characters were significantly clustered. The conclusion

that must be drawn from these two studies is that the

homoplasy observed in this set of characters cannot be

distinguished from what would be expected if all homoplasy

were distributed at random topologically. The characters

that appear by eye to be clustered are simply chance

extremes.

Systematists impressed by apparent tendencies among

characters may have simply focused on those characters that

happen to be extreme deviates from an underlying random

distribution of homoplasy. By ignoring the lack of

tendencies in the majority of characters, systematists have

overlooked "chance -- the everpresent rival conjecture"

(Polya, 1968).

Given the negative conclusion of the present analysis,

it is fair to inquire whether the methods used were

appropriate. This kind of work is fraught with potential

biases and artifacts, some of which may be serious under

certain patterns of evolutionary change (Maddison, ms.;

Sanderson and Donoghue, in press, see Appendix 1). Three

----_. ----_. __ .- 195 TABLE 11. Hypothesized effects of various factors on observed levels of homoplasy, clustering, and localization. Plus and minus signs refer to increases and decreases respectively. Symbols in parentheses refer to effects likely to be rare (usually because of a violation of the auxiliary principle see discussion in text).

Amount of Amount of Amount of As ..• homoplasy clustering localization

Scope + + + +

Scale + + - (+) - (+) Resolution + + (-) + (-) 196 factors, which I term scope, scale, and character resolution, could potentially introduce artifacts into the present analysis, and a discussion of their influence is warranted. The "scope" of an analysis is the breadth of taxonomic coverage, in terms of the inclusion of more and more distantly related groups. The "scale" is related to the degree of splitting of the terminal taxa in the study

the finer the scale, the more terminal taxa from within previous broader taxa. The number of taxa in any analysis increases both with increased scope and finer scale. The

"character resolution" is related to the information content of character and character state definitions. Finer resolution, based on detailed study of a given character, illuminates more information about the character and is apt to result in more and more finely differentiated character states. Given that homoplasy, as it is defined in Chapter

One, is tied to the similarity criteria adopted by each investigator, the resolution of a character will directly affect the likelihood that it is judged homoplastic: the more vague the criteria, the more instances of homoplasy that will be inferred.

The question is therefore how these factors combine to influence the overall level of homoplasy, and particularly the amount of clustering (and localization) that may be detectable. In Table 11, I attempt to summarize the 197

expected effect of these factors. It is useful to invoke an

"auxiliary principle" at times in this discussion, according

to which homoplastic changes are likely to be most similar

in most closely related taxa, and less similar in distantly

related groups. This principle has long been held among

workers interested in parallelism (from Darwin on) but it

has not been tested phylogenetically.

Since increases in scope or scale result in more taxa,

it is clear that levels of homoplasy will increase in such

cases, based on the results of Sanderson and Donoghue (in

press, Appendix 1, and see discussion in Chapter 1). It may

be, however, that levels of homoplasy will achieve a maximum

value with increased scope, because eventually, new

instances of homoplasy will be sufficiently distinct as to

be scored as a different state (assuming the auxiliary

principle holds). On the other hand, increases in

resolution must reduce the overall level of homoplasy by

eliminating character changes that formerly arose because

taxa were considered similar enough to be coded as the same

state but now are seen to be different.

The effects on clustering and localization are less

obvious. Increases in scope are likely to increase

(eventually) the observed levels of clustering and

localization if the auxiliary principle holds, because

ultimately distant homoplasy will be viewed as distinct by

------198 the investigator, and subsequent inclusion of additional taxa will only serve to make the original homoplasy seem more and more localized (and clustered). As the scale increases (becomes finer), the observed levels of clustering and localization may increase or decline depending on the way that the scale is increased. If terminal taxa are replaced with large numbers of their component infra-taxa and these do not undergo any additional homoplasy then clustering will become more evident; although the observed NN distances remain the same, the expected values under any null models increase because there are more branches on the tree. But if additional taxa are added that are interposed between taxa in the original study, the observed NN distances increase, and levels of clustering may in fact decline. In general, the additional taxa, wherever they are placed, will usually undergo some additional homoplasy unless only very particular terminal taxa are chosen for increases in scale (more subdivision), so levels of clustering should rarely be expected to increase. The effect of character resolution depends critically on the validity of the auxiliary principle. Increased resolution should generally lead to increased levels of clustering or localization if distant homoplasy is less similar than homoplasy in closely related groups (see Fig.

------199 A*

FIG. 35. Hypothesized effect of increasing character resolution on the discovery of clustering. As resolution is increased, it is expected that homoplastic changes will disappear roughly in the sequence indicated (1 first, etc.). Taxa more distantly related from those indicated by "--" will likely have homoplasy that is slightly different. 200

35). Thus as resolution increases, some homoplastic events disappear. If they disappear first in distantly related groups, then clustering is enhanced. This was the case for the terminal leaflet character discussed above for Astragalus. Increasing resolution by "undefining" the type of leaf in A. magdalanae var. peirsonii, meant an increase in clustering. Clustering will also be easier to detect simply because fewer instances of homoplasy on a tree lead to more powerful tests (see above under results). However, a caveat is that overall homoplasy may decline faster than clustering can increase so that in practice it is all but impossible to detect it. If the auxiliary principle does not hold, the events that disappear might be the nearby ones, which would cause a decline in observed clustering (and perhaps localization as well). Given the behavior of these factors, it is now possible to more fully interpret the results of the analyses of spatial clustering described above. Ordinarily, systematists have no objective way to measure the resolution of their characters. The lack of observed clustering found in the analyses above suggests that the resolution used is too low to detect clustering in the data, but alternatively there may simply be no clustering in these characters. Another possibility is that these organisms are all so similar genetically that any trait is likely to evolve 201

anywhere (Cantino, 1985). This is difficult to believe for Astragalus, a genus notable for its morphological diversity. In any case, it would be inappropriate to attempt to utilize

supposed "tendenciesll among such characters, because they are likely to be spurious. It may often be possible, by more detailed study of characters, to increase resolution to the point that clustering relative to the scope and scale of the study can be detected (as occurred in the terminal leaflet character discussed above). As noted earlier, however, there is no guarantee that increased resolution automatically implies increased clustering. until more detailed character analysis is undertaken in cladistic studies, it will not be known whether IItendencies" exist, and if so whether they should play any significant role in taxonomic practice .

.. - ..... -..------_._------202

APPENDIX 1

PATTERNS OF VARIATION IN LEVELS OF HOMOPLASY

by

Michael J. Sanderson and Michael J. Donoghue

Department of Ecology and Evolutionary Biology,

University of Arizona, Tucson, AZ 85721

-----...------203

Abstract. -- Patterns of variation in levels of homoplasy were explored through statistical analyses of standardized consistency indices. Studies were obtained from 60 recent cladistic analyses of a wide variety of organisms based on several different kinds of characters. Consistency index is highly correlated with the number of taxa included in an analysis, with homoplasy increasing as the number of taxa increases. This observation is compatible with a simple model of character evolution in which (1) the probability of character state change increases with the total number of branches in a tree, and (2) the number of possible states of a character is limited. Consistency index does not show a significant relationship to the number of characters utilized in an analysis or to the taxonomic rank of the terminal taxa. When the relationship between consistency index and number of taxa is taken into account, there is no significant difference in the amount of homoplasy in plant versus animal data sets. Likewise, the level of homoplasy

in morphological and molecular data sets does not appear to differ significantly, although there are still too few molecular studies to be confident of this result. Future comparisons of consistency indices, including studies along the lines established here, must take into account the

influence of the number of taxa on homoplasy. 204

INTRODUCTION

Phylogenetic studies of character evolution have focused on selected traits within individual groups of organisms (e.g., Ridley, 1983; Wake and Larson, 1987; Huey and Bennett, 1987; Sessions and Larson, 1987;

Sillen-Tullberg, 1988; Hufford, 1988). Lauder (1981, 1982) suggested searching for general patterns of character evolution by comparing c1adograms of different groups, but this has seldom been done and such comparisons been restricted to only a few groups (e.g., Emerson, 1988). No attempt has been made to document general patterns of character evolution by comparing cladograms of many groups at once. Here we undertake such an analysis of one general feature of character evolution, homoplasy (Lankester, 1870), which is the independent evolution of the same character state.

Although the term homoplasy is sometimes used to refer to any kind of convergence, parallelism, or reversal

(Futuyma, 1986), its connotation is narrower in cladistic analyses that provide the basic data upon which our analysis is based. Homoplasy in cladistic analyses results when features hypothesized at the outset of an analysis to be homologous are found to arise more than once on a cladogram, or to originate and then be lost (Farris, 1983; Patterson,

1982). The criteria used to assess homology £ priori are a 205 special subset of all possible criteria, since they are intended to permit recognition of character states similar by virtue of common ancestry. In particular, these include the positional and developmental criteria of Remane (1952; see Patterson, 1982).

Systematists often view homoplasy resulting from rejection of these initial hypotheses of homology as no more than a collection of mistakes. Indeed, the most frequently used measure of homoplasy in cladistic studies, the

"consistency index" or CI (Kluge and Farris, 1969, see below), is generally interpreted as a measure of goodness of fit of data to a tree topology, rather than a description of some evolutionary pattern. In some cases the quality of a particular study has been judged largely by this measure

(e.g. Riggins and Farris, 1983 p. 99).

Although we agree that homoplasy always represents a

"mistake" in the sense of a mistaken hypothesis of homology, it is clear that not all such mistakes are of equal interest to evolutionary biologists. Mistakes due to measurement errors (e.g., misreading a caliper) do not necessitate any explanations of an evolutionary kind, whereas "mistakes" such as the discovery of an independent origin of some character do. Moreover, if the latter mistake occurs among characters that have already passed some a priori test of homology, they are especially interesting, because the 206 evolutionary explanation for their recurrence may differ from an explanation adequate to explain broad convergences that are not similar positionally or developmentally

(Rensch, 1959 p.191).

Assertions about homoplasy are widespread in the literature (see Discussion). statements regarding levels of homoplasy in plants versus animals, and in morphological versus molecular data have seldom been tested. Hypothesized relationships between homoplasy and the number of characters, taxonomic rank, or number of taxa included in cladistic studies have received even less attention. Here we explore general patterns of variation in levels of homoplasy utilizing data obtained from recent cladistic analyses of a wide variety of organisms. We focus on the relationship between consistency index and the variables mentioned above, and consider the evolutionary implications of the patterns observed.

METHODS

Cladistic studies were selected according to the following criteria. First, in order to exclude analyses utilizing earlier versions of parsimony algorithms, we considered only studies conducted since 1980. Second, only those analyses based on discrete, "qualitative" characters were selected; studies involving continuous (morphometric) 207

characters and distance data (e.g., DNA hybridization data) were not considered, because there is no clear analog of consistency index for such data. Third, only data sets analyzed under Wagner parsimony (allowing both forward state changes and reversals; Felsenstein, 1982) were selected for comparison; analyses using restricted parsimony approaches (e.g., "00110 parsimony"), character compatibility, or phenetic algorithms were excluded. Finally, analyses were included only if sufficient information was provided to permit us to standardize consistency indices in the manner described in Appendix I. Sixty data sets were selected that satisfied these requirements (Table AI-I), encompassing a broad range of variation in the type of organism and the type of characters

involved. Of these, 26 are analyses of plant taxa, 30 are of animals (both interpreted broadly), and 4 are of "other" groups (two of fungi, one of all major groups of eukaryotes, and one involving invented "organisms," the Dendrogramaceae). Forty-two of the studies are analyses of "morphological" data (interpreted broadly to include one study of secondary chemical characters), while the remaining 18 are analyses of "molecular" data, (also interpreted broadly to include 4 protein electrophoretic studies, and 14 restriction fragment, amino acid, and nucleotide sequence studies). The smaller number of molecular studies reflects

------208 the fact that fewer of these are currently available that meet the requirements specified above. We selected studies covering a wide range of number of taxa (4 to 68), number of characters (15 to 868), and taxonomic rank of terminal taxa

(species to classes/phyla).

The consistency index (Kluge and Farris, 1969) was .used as the standard for comparison of levels of homoplasy among studies. Although CI measures only the overall amount of homoplasy, its simplicity and widespread use make it an obvious choice for comparing data sets. A character is perfectly consistent, or shows no homoplasy on a cladogram, if all state changes occur only once. For example, a perfectly consistent binary (two-state) character would entail only a single state change (step), say from state 0 to state 1. If more than one state change is required to fit parsimoniously, or "optimize," the character on the tree

for example, two changes from 0 to 1 or a change from 0 to 1 and then a reversal to 0 then the character is not perfectly consistent -- it is inconsistent, or homoplastic.

The overall level of homoplasy entailed by a cladogram is routinely reported as its consistency index. This measure is defined as the minimum number ~f character state changes required by a particular data set (summed over all characters) divided by the total number of state changes required to most parsimoniously fit all of the characters on 209 the tree under consideration. If homoplasy is absent the consistency index is 1.0, and CI decreases toward a as homoplasy increases.

Possible Sources of Bias

Despite its frequent use, it is not widely appreciated that CI is affected by the inclusion of characters that are necessarily perfectly consistent on all trees (Carpenter,

1988). Inclusion of characters in which a single state is possessed by all groups under consideration (invariant characters) and/or characters in which only one of the

included taxa possesses a particular derived state

(autapomorphies), will inflate CI by adding one unit to both the numerator and denominator in the calculation.

Unfortunately, cladists have not been entirely consistent in

omitting these phylogenetically uninformative characters

from their data sets. Recognizing the possibility that this

inconsistency might seriously bias our comparison of CI's

obtained from different studies, we recalculated CI's after

eliminating necessarily consistent characters according to a

set of rules detailed in Appendix I.

Several other potential biases are more difficult to

eliminate. We are unaware of any systematic bias associated

with these factors in relation to type of organism, type of

data, or any other variable discussed below, and therefore 210 doubt that they exert any significant effect on our results.

Nevertheless, these factors are worthy of mention as possible sources of residual variation in Cl's, and their effects should be tested when a larger number of appropriate cladistic studies have become available.

Character distribution. -- The effects of autapomorphies and invariant characters on Cl are best seen as extreme cases of the influence of the distribution of character states among the taxa. Characters in which an almost equal number of taxa possess the alternate states of a binary character have a greater probability of showing lower consistencies than characters in which the distribution of states is markedly unequal, simply because there are more opportunities for homoplasy in the first case

(Meacham, 1981, 1984). Thus, data sets that include a large proportion of characters with highly unequal distributions of states will tend to have higher Cl's than those in which most characters have nearly equal distributions among the taxa.

Multistate characters. -- Studies involving ordered mUltistate characters will tend to have lower consistencies

than those in which the multistate characters are unordered, because cladograms may necessitate state transitions in

ordered characters that add more than one step (e.g., from

state "1" to "3"). This effect is apparent in Baum's (1984) 211 compatibility analysis of Avena, which involved ordered characters with as many as 16 states. Not surprisingly, he found that the largest sets of mutually compatible characters (cliques) were exceedingly small, indicating a very high level of inconsistency. Most molecular studies utilize either binary coding for presence-absence of restriction fragments or unordered mUltistate coding for nucleotide data. Such data are therefore less susceptible to this kind of bias than ordered morphological data. Missing data. -- Parsimony algorithms assign characters scored as "unknown" or "missing" whichever state is most parsimonious given the position of the taxon in the tree (based on "known" characters). Since such coding can never increase inconsistency, whereas homoplasy might be introduced if "unknowns" were replaced by definite scores, data sets with more unknowns will tend to have higher consistencies than those with fewer. This is likely to be a significant factor only in data sets in which many taxa are scored as unknown due to high levels of polymorphism or in studies that include fossil groups for which information on many characters is lacking (e.g., Gauthier, 1976) and/or about which there are uncertainties regarding initial homology assessments (e.g., Doyle and Donoghue, 1986). 212

statistical Analyses

We utilized multiple linear regression to test for the dependence of CIon several variates of interest, including the number of taxa included in an analysis, the number of characters used, and the taxonomic rank of the terminal taxa

(using a linear scale corresponding to the categories of the Linnaean hierarchy: from species = 1, to class or above = 5). Because preliminary analyses suggested a curvilinear relationship of CI to one of the independent variates, number of taxa (see Fig. A1-1a), log-transformed CI values were used in subsequent analyses (Fig. A1-1b). Log transformation was effective in linearizing the data, as indicated by a low autocorrelation of residuals in a Durbon-Watson test.

Differences in regression parameters between subsets of the data were examined using analysis of covariance. Class variables considered were type of organism ("plant" versus

"animal") and type of character data ("morphological" versus

"molecular"). All statistical analyses were performed using

SAS version 6.02 on an IBM PC-AT microcomputer.

RESULTS

Relationships among the variables are shown in Table

Al-2 and Figures 1-5. consistency index is highly correlated with number of taxa (p < 0.001), but not with 213 number of characters or taxonomic rank. Number of characters is strongly correlated with number of taxa (p <

0.01) but not with CI, despite the strong correlation between CI and number of taxa. Multiple regression analysis applied to the entire data set confirmed the pattern evident from the correlation matrix (Table A1-3). The intercept and slope of number of taxa are both highly significant . components of the regression (p < 0.001), while other variates are not significant. The linear regression model explains 50% of the variance in CIa Because of potential sensitivity of the regression model to extreme outliers for character number (those few studies with between 200 and 800 characters), an additional analysis was conducted omitting these studies. A negative, but non-significant (R-squared = 9%), relationship was found between CI and character number in a simple regression analysis; however, when other variables were included in a multiple regression model, number of characters was still not a significant component.

Conclusions based on the entire data set are the most robust results of our study. When the data set is subdivided, sampling error becomes a more serious problem, and we hesitate to draw strong conclusions. For this reason, we did not further subdivide the "molecular" subset of studies. We suspected that number of taxa would be the most important factor affecting CI in subsets of the data 214

(as it was in the entire data set), and multiple regression

applied separately to plants and animals, and to

morphological and molecular data, provided no reason to

question this expectation. Regressions for plants and

animals are strikingly similar, as are estimates for

morphological and molecular data; in each case only number

of taxa is a significant variate. Although the regression

parameters for the molecular data set are not significantly

different from zero, they nevertheless suggest a negative

relationship between CI and number of taxa.

Plots of CI versus number of taxa for subsets of the

data are shown in Figures 4 and 5. Simple regressions of a

dependent variate on a single independent variate may be

misleading in the context of multiple regression if other

variates exert important effects (Montgomery and Peck,

1982). However, the similarity between plant and animal

data sets in the multiple regression analyses (Table A1-3)

is mimicked by a similarity in the simple regressions (Fig.

Al-4). Analysis of covariance for number of taxa alone

showed no significant difference in the slopes of these two

regressions. The comparison of morphological and molecular

characters was less clear. Multiple regression parameters

suggest some differences in the patterns for the two kinds

of data (Table Al-3), but the simple regressions of CIon

number of taxa differ from one another at a significance

-----. -_._-_.. _------215 level of p = 0.057 (Fig. Al-5). More molecular studies must be considered before any importance can be attached to this "almost significant" difference.

DISCUSSION Number of Taxa Our results indicate that as the number of taxa included in an analysis increases, the consistency index decreases (Fig. Al-1). Therefore, the average number of state changes per character increases with the addition of taxa. Although this relationship has been suggested previously on theoretical (e.g., Riggins and Farris, 1983), and empirical grounds (Archie's 1985 analysis of 7 morphometric data sets), the present analysis, based on 60 studies, provides definitive support for this relationship. Several taxonomic artifacts might yield the observed results, but, in our opinion none provides a compelling explanation. For example, systematists working on large

groups might be more prone to errors in £ priori homology assessments and in scoring taxa (perhaps because of the additional work required), leading to an increase in homoplasy. There is, however, sociological pressure opposing this bias, since low CI values are widely considered to be symptomatic of poor character analysis and dubious results. Furthermore, systematists (at least in 216 morphological studies) sometimes remove excessively homoplastic characters after consulting a preliminary parsimony analysis, either as part of a formal successive weighting scheme (Farris, 1969), or because initial indications of homoplasy lead to a reconsideration of certain characters and the discovery of mistakes. Even so, systematists have been unable to maintain high Clls in large studies. Bias would also be introduced if systematists studying large groups included more homoplastic characters than they would in studying smaller groups. A study of a large number of taxa necessarily requires the identification of a larger number of individual relationships, and this requires more characters to achieve the same degree of resolution. The strong positive correlation seen between number of taxa and number of characters (Table Al-2) may reflect a conscious or unconscious desire to include more characters, even if this means settling for characters that are more likely to show homoplasy. However, there is no reason why additional characters must necessarily be more homoplastic; in fact, larger numbers of characters do not appear to be correlated with lower Clls (see below). A third possible bias may result from the algorithms used to generate most-parsimonious trees. For small data sets (fewer than 15 to 20 taxa) "branch-and-bound"

------217 algorithms are available that guarantee the recovery of the most parsimonious tree(s). For larger data sets, however, there are no efficient algorithms that always identify the shortest trees, and therefore larger studies may report tree-lengths somewhat longer than the actual minimal tree(s), therefore entailing more homoplasy. However, we consider it unlikely that this accounts for the large differences in CI observed in Fig. Al-l. One way to investigate this bias would be to use branch-and-bound algorithms to study variation in CI among small studies. We predict that the observed relationship to number of taxa would still hold. Another bias stemming from algorithms used to reconstruct trees may be important in studies with very few taxa. Parsimony algorithms place an absolute upper limit on the amount of homoplasy that can be detected. For example, in the case of three taxa, a consistency index below 0.60 is not possible, because of the way parsimony algorithms minimize character state changes on a tree. However, this minimum CI approaches zero rapidly as number of taxa increases and is unlikely to be an important bias over the large range of number of taxa studied here. An obvious alternative to all of the explanations given above is that systematists studying large groups are unable to find characters with low levels of homoplasy because such 218 characters are rare. In other words, the correlation of CI with number of taxa may be primarily a reflection of the evolutionary process itself, not of some taxonomic artifact.

The increase in homoplasy with number of taxa is consistent with a model of character evolution in which the probability that a character will change somewhere on a tree is monotonically related to the total number of internodes

(branch segments) in the tree, and hence to the number of taxa. Given an average non-zero probability of character state change along internodes as branches are added, the number of state changes will tend to increase, and this will result in increased homoplasy if a limited number of states is possible. This last point is critical. If there were an unlimited number of states, then an increase in the amount of change in a character might not result in more homoplasy--instead all changes might be to completely new states. It is the existence of a limited number of alternative states (due to some biological constraint; e.g., the existence of only four nucleotides), or the imposition of a limited number of discrete states by the systematist, that may be responsible for the basic result.

The data compiled here do not allow a direct test of this model. What is needed are studies of nested subsets of taxa taken from within the same larger group that differ in the number of terminal taxa they contain. In any case, the 219 view that systematic characters should undergo more homoplasy in broader analyses accords well with empirical generalizations suggested by systematists. It is widely

recognized, for example, that characters found to be useful

in delimiting groups in one taxon will often be found less

useful when additional taxa are examined (e.g., Mayr, 1969;

Davis and Heywood, 1973 p.116). In addition, characters

thought to be conservative because they delimit "higher"

taxa are often found to vary among closely related species

when additional taxa are considered (e.g., Stebbins, 1974).

Number of Characters

There has been little explicit discussion of the

relationship between homoplasy and number of characters.

Although it is clear that the number of character

incompatibilities in a data set can only increase as

characters are added, the consistency. index can increase or

decrease depending on the distribution of the new character

states among taxa (in contrast, when taxa are added to an

analysis, CI can declines or stays the same). Archie (1985)

found that CI declined slightly with the addition of

characters in seven morphometric data sets.

Despite a strong correlation between number of taxa and

CI, and between number of characters and number of taxa, we

did not find a significant correlation between number of 220 characters and CI (Fig. Al-2). Nor does number of characters enter significantly into the regression estimates. When the effect of number of taxa is taken into account, studies with hundreds of characters exhibit levels of homoplasy similar to studies employing only a handful of characters (e.g. compare Fink, 1985, with Wighton and wilson, 1987). This result might be interpreted as a refutation of the widespread view that the quality or reliability of phylogenetic analyses increases with the addition of characters. Indeed, if CI were considered a measure of quality, our results would cast doubt on the intuition that

it is better to include more characters. However, the discussion above on the decline of CI with an increasing number of taxa suggests that this is an inappropriate

interpretation, unless we suppose that studies involving more taxa are necessarily inferior. We suggest that CI is not a measure of quality, but simply a measure of the overall level of homoplasy. These attributes are logically distinct.

The standard view on the desirability of adding characters can be defended on other grounds. The reason to gather more data is not to increase consistency index (which

it does not), but rather to take advantage of "statistical consistency" (Felsenstein, 1978), that is, convergence on 221

truth as a more complete sample of characters is considered.

There is also a relationship between the number of

characters and the confidence that can be associated with

particular clades, at least when confidence is measured by

the bootstrap technique (Felsenstein, 1985; Sanderson,

1989). It is important to recognize that CI and confidence

are not directly related. Although confidence may in some

cases be associated with the CI, it need not be. In Fig.

Al-3, we present a data set possessing a CI much lower than

the expectation for that number of taxa, which nevertheless

has component clades supported with high levels of

confidence as measured by bootstrap resampling -- higher

levels in fact than are often found in real data sets

(Sanderson, 1989).

Taxonomic Rank

There is little consensus in the literature regarding

the expected relationship between homoplasy and taxonomic

rank. A number of authors have suggested that unravelling

phylogenetic relationships among higher taxa, such as

angiosperm families or mammalian orders, will be

particularly difficult owing to rampant parallelism and

convergence at such levels (e.g., Cronquist, 1987 p. 46).

Other authors suggest that parallelism ought to be more

prevalent at lower taxonomic levels, because very closely

... -. __ .. _.. _------222 related organisms are genetically and developmentally more similar, and therefore more prone to parallel responses to similar selective or mutational pressures (e.g., Vavilov,

1922; Rensch, 1959; Davis and Heywood, 1973; stevens, 1986).

Our analysis suggests that levels of homoplasy are largely independent of taxonomic rank (Fig. Al-4).

Homoplasy in characters used to assess relationships among higher taxa is similar to that in characters used at lower taxonomic levels. This result might simply reflect the fact that ranks are not equivalent, especially among divergent groups (e.g., species or families of angiosperms versus mammals or insects; Mishler and Donoghue, 1982; Gauthier,

Estes, and de Queiroz, 1988). Perhaps if rank were adjusted to reflect age, as suggested by Hennig (1966), a correlation between homoplasy and rank would be discovered, with older groups showing more homoplasy. However, this expectation may be misguided. Higher taxa, if monophyletic, are conceptually equivalent to single ancestral species, and a cladistic analysis of a set of higher taxa can be viewed as an analysis of their ancestral species. If there is no reason to expect homoplasy to be any more or less prevalent among such ancestral species than it is among a set of recent species, then levels of homoplasy should be about equal at different ranks. 223

We suspect that intuition on the subject of rank has often been mistaken for two reasons. First, higher taxa are not viewed as ancestral species. Instead the myriad polyrnorphisms, parallelisms, and reversals, which seem characteristic of all large taxa, are taken into consideration, even if only subconsciously. Consideration of such variation is operationally equivalent to the inclusion of all the taxa within the higher group, which leads to an increase in homoplasy owing to the increase in number of taxa (see above). Also, when higher taxa reflect old lineages, it might seem more likely that extensive homoplasy would have time to appear. However, if there are few terminal taxa, then that homoplasy (present or not) is difficult to detect given the upper limit on homoplasy detectable by parsimony. These conclusions do not contradict the observation that any particular character might be expected to show progressively higher levels of homoplasy as the scope of a study were expanded upward in rank. In practice systematists undoubtedly choose their character sets to reflect an appropriate amount of variation for the rank of the terminal taxa used. Although workers using molecular data are more constrained than morphologists in this regard, they too can choose the appropriate molecule, or the appropriate set of restriction enzymes (some restriction 224 mutations appear more or less frequently in a given genome). Hence, in all cases the relative invariance of levels of homoplasy to taxonomic rank may reflect compensation on the part of taxonomists.

Plant versus Animal Data In contrast to the factors discussed above, there has been much commentary on the level of homoplasy in plants versus animals. In particular, it is widely believed that plants are more homoplastic than most animals. Thus, according to Wagner (1984 p.11S), "plants are evidently unusually inclined towards parallelisms." Likewise, cronquist (1987 p.24) suggested that " .•. the relative lack of morphological integration in plants, and poor correlation of evolutionary advances with adaptive zones and ecologic niches, combine to permit rampant parallelism, in contrast to the more rigid (though imperfect) evolutionary channelling in animals." Accounting for increased homoplasy by reference to the relative simplicity and indeterminate growth of plants has been an especially popular argument.

Th~s Stebbins (1974 p.143), observing that " ... patterns of development of individual organs are, in general, much simpler in plants than in animals," concluded that" similar but independent evolutionary modifications of structures ••. are much more common in plants than in

..------.------225 animals." Similarly, Gottleib (1984) postulated that relatively few genes of major effect may underlie character differences between plant species, and that plants may therefore be more prone to parallelism than animals. Such sentiments have led to the widespread belief that phylogeny reconstruction is more difficult in plants than in animals (e.g., Funk, 1981 p.73; Davis and Heywood, 1973 p.112). Stebbins (1950 p.506) presumed, for instance, that " ... morphological similarity is much less indicative of phylogenetic relationships in plants than it is in animals." And cronquist (1987 p.24) pointed to the "difference in amount of parallelism in plants and animals" in commenting on the reluctance among botanists to embrace cladistic methods. The perception of rampant homoplasy in plant morphological characters has even led to the suggestion that only molecular data will provide characters with acceptable levels of homoplasy (Sytsma and Gottlieb, 1986b). Despite considerable interest in this issue, levels of homoplasy in plants and animals have never been compared quantitatively. Surprisingly, our comparison indicates that levels of homoplasy are remarkably similar; the regressions of consistency index on number of taxa match closely (Fig. Al-5). We conclude from this that there is little difference in the average level of homoplasy between plants and animals, at least among those characters actually 226 selected by systematists in analyzing cladistic relationships. We can think of two explanations for this result. First, it is possible that CI fails to capture what most biologists have in mind when they use the term "homoplasy." Perhaps if all characters (or a truly random sample of characters) were considered, plants would indeed be more homoplastic than animals. Systematists often eliminate characters prior to phylogenetic analysis, and it is possible that plant systematists eliminate more characters because a higher proportion of characters are excessively variable in plants than in animals. This is a difficult proposition to test because the number of "rejected" characters is not usually reported in published studies. Unfortunately, one cannot simply compare the average number of characters included in plant studies versus animal studies. Plant studies are generally based on fewer characters, but it is well known that animals (at least the "higher" metazoa) have many more basic tissue types than plants (e.g., Gottlieb, 1984) and consequently are likely to have many more taxonomic characters. Alternatively, preconceived ideas about homoplasy may have been incorrect. Animals and plants may in fact be approximately equally homoplastic. It is not possible at this time to distinguish between these two alternative 227

hypotheses, which is unfortunate from the perspective of

attempting to understand the interesting evolutionary

implications of homoplasy. On the other hand, botanists

should find these results encouraging, because regardless of which of the above explanations is correct, it is clearly possible to gather charact~r data on plants that exhibit as

little homoplasy as data on animals. Indeed, plant

systematists have been doing it for years.

Morphological versus Molecular Data

The value of morphological data in elucidating

phylogeny has frequently been questioned because of the

susceptibility of morphology to convergent evolution by

natural selection. Some have suggested that molecular data

can and will resolve phylogenetic questions that morphology

cannot (Hillis, 1987; Patterson, 1987; Sibley and Ahlquist,

1987 p.118; Bobrova, et. al., 1987). For example, Sytsma

and Gottlieb (1986b) state that a " ... primary reliance on

morphological data to model phylogenetic relationships may

be misleading, no matter how many characters are examined."

Furthermore, molecular systematists have touted high Clls in

support of the contention that such data are less

homoplastic (e.g., Jansen and Palmer, 1988 p.764), although

these Clls are substantially lower when numerous 228 autapomorphies are removed (Table A1-1). On the other hand, some systematists have pointed out difficulties in establishing homology in molecular characters (e.g.,

Patterson, 1987 p.18), and it has even been suggested

(Mishler, et. al., 1988) that molecular data may show ~ homoplasy due to the limited number of character states for a character (locus) and the difficulty in eliminating homoplasy by careful ~ priori character analysis (in

contrast to morphological characters which undergo development) .

A few studies have compared levels of homoplasy in molecular versus morphological data in particular taxa.

Wyss, et. ale (1987) examined homoplasy in amino acid

sequences of mammal orders and found it comparable to that

seen in morphological data. In a similar analysis of amino

acid sequences in angiosperms, Bremer (1988) concluded that

levels of homoplasy were too high to yield robust phylogenies. Our study is more comprehensive in that it

includes 18 molecular studies and 42 morphological studies,

and it considers the effects of other factors such as the

number of taxa and characters. We find little support for

the belief that molecular data show either lower or higher

levels of homoplasy (Fig. Al-6). H6wever, because our

sample of molecular studies is still small and because of

the high variability in consistency indices, the regression 229 estimates for molecular studies are not statistically significant (Table Al-3). Had they been significant they would have implied that molecular data are more homoplastic than morphological data in smaller studies and less homoplastic in larger ones. Unfortunately, small sample sizes prevented us from examining further subsets of the

"molecular" data, to see, for example, if sequence and restriction-site data are more or less homoplastic than protein electrophoretic data. We do find it suggestive, however, that the large study of Jansen and Palmer (with 55 taxa) has a much lower CI than other restriction-fragment studies with fewer taxa.

In general, our result is not a criticism of the use of molecular data. Molecular data may prove to be better than morphological data for reasons other than lower levels of homoplasy. Perhaps the ability to generate large numbers of characters will improve the statistical consistency of the tree estimation process (Felsenstein, 1978), and less ambiguous character state assignments may reduce artifacts introduced through human error. However, there is no evidence to date that molecular data are less homoplastic than morphological data. 230

CONCLUSIONS Our comparison of consistency indices, which has become possible with the availability of large numbers of cladistic studies, allows a test of hypotheses about character evolution and exposes unexpected patterns that require evolutionary explanation. We have found that CI is strongly correlated with number of taxa, but not with number of characters or taxonomic rank. Moreover, when other factors are taken into account, there does not appear to be any significant difference in homoplasy between plants and animals, or between morphological and molecular data. Many other comparisons are suggested by this study. First, more refined analyses based on a larger sample of studies will soon be possible. The contrast of "morphological" versus "molecular" data sets is crude, and can be refined as studies become available. It would be useful to compare morphological versus behavioral, reproductive versus non-reproductive, sequence versus electrophoretic, and nuclear versus organellar genome data. Second, CI could be decomposed into its components, parallelism and reversal, as it has been in some individual studies (Eckenwalder and Barrett, 1986; Gauthier, Kluge, and Rowe, 1988). Finally, it should be possible to compare patterns in the topological distribution of homoplasy: for 231 instance, the localization of homoplasy to particular regions of a tree, and the topological correlation of homoplastic changes in different characters.

Analyses along these lines, and any other quantitative comparisons of homoplasy using the consistency index, should take into account the strong relationship of consistency index to the number of taxa studied. Our results provide an initial benchmark for comparative purposes. Inferences about the implications of levels of homoplasy in particular groups must be drawn with the knowledge that the size of the group is a critical factor. This has rarely been recognized. Thorpe and Dickenson (1988) studied the use of regulatory sequences in reconstructing phylogenies in

Drosophila and concluded that " •.. the amount of homoplasy within the groups of closely related species is sufficiently small that regulatory proteins may be useful for inferring relationships at this level ..• (p. 97)" Their conclusions were based on a comparison of CI's in small subsets of taxa relative to the CI of the overall data set, and in large part reflect differences in the numbers of taxa involved.

Systematists studying large groups frequently encounter what seem to be inordinately high levels of homoplasy (e.g.,

Cronquist, 1987 p.32). For example, Funk (1981 p.82) suggested that the level of parallelism in the genus

Montanoa had reached "staggering proportions." In fact, the 232 consistency index that we calculate from her cladogram, 0.46, is only slightly below the value of 0.52 expected with the 25 taxa included in her study. The observed deviation is well within the standard error of the predicted value based on the regression parameters in Table Al-3 (Snedecor and Cochran, 1980 p. 166). Observations of extreme levels of homoplasy should be made not in terms of absolute values of consistency indices, but by reference to extreme deviations from the regression line in Fig 1.

ACKNOWLEDGMENTS J. D. Palmer, R. K. Jansen, P. F. stevens, E. Zimmer and W. B. Heed kindly made available information on their analyses in press. We are grateful to R. E. strauss and B. Walsh for statistical advice and helpful comments on the manuscript. W.H. Wagner, P.F. Stevens, and J.M. Porter provided stimulating comments on the ideas proposed here, and two anonymous reviewers provided useful criticisms and suggestions. This research was facilitated by an NSF grant awarded to MJD (BSR-8414450).

LITERATURE CITED Anderberg, A. 1986. The genus Pegolettia (Compositae: Inuleae). Cladistics 2:158-186. Archie, J. 1985. Methods for coding variable morphological features for numerical taxonomic analyses. Syst. Zool. 34:326-345. Arnold, E. N. 1981. Estimating phylogenies at low taxonomic levels. Z. Zool. Syst. Evol-Forsch. 19:1-35. Baum, B. R. 1983. A phylogenetic analysis of the tribe Triticeae (Poaceae) based on morphological characters of the genera. Canad. J. Bot. 61:518-535 • • 1984. Application of compatibility, and parsimony 233 methods at the infraspecific, specific, and generic levels in Poaceae, pp. 192-220. In T. Duncan and T. stuessy (eds.), Cladistics: Perspectives on the Reconstruction of Evolutionary History. Columbia Univ. Press, New York. Baum, B. R. and D. B. o. Saville. 1985. Rusts (Uredinales) of Triticeae: evolution and extent of coevolution, a cladistic analysis. Bot. J. Linn. Soc. 91:367-394. Bobrova, V., A. Troitsky, A. Ponomareo, and A. Antonov. 1987. Low molecular weight rRNA sequences and plant phylogeny reconstruction: nucleotide sequences of chloroplast 4.5s rRNAs from Acoris calamus (Araceae) and Ligularia calthifolia (Asteraceae). Pl. syst. Evol. 156:13-27. Bremer, K. 1988. The limits of amino acid sequence data in anglosperm phylogenetic reconstruction. Evolution 42:795-803. Bremer, K., C. J. Humphries, B. D. Mishler, and s. P. Churchill. 1987. On cladistic relationships in green plants. Taxon 36:339-349. Brooks, D., R. Daniel, R. T. O'Grady, and D. R. Glen. 1985. The phylogeny of the Cercomeria Brooks, 1982 (Platyhelminthes). Proc. Helminthol. Soc. Wash. 52:1-20. Campbell, C. s. 1986. Phylogenetic reconstructions and two new varieties in the Andropogon virginicus complex (Poaceae: Andropogoneae). Syst. Bot. 11:280-292. Cane, J. H. 1983. preliminary chemosystematics of the Andrenidae and exocrine lipid evolution of the short-tongued bees (Hymenoptera:Apoideae). Syst. Zool. 32:417-430. Carpenter, J. M. 1987. Phylogenetic relationships and classification of the Vespinae (Hymenoptera:Vespidae). Syst. Entomol. 12:413-431. ____ . 1988. Choosing among multiple equally parsimonious trees. Cladistics 4:291-296. Churchill, S. P., E. o. Wiley, and L. A. Hauser. 1984. A critique of Wagner Groundplan-Divergence studies and a comparison with other methods of phylogenetic analysis. Taxon 33:212-232.

Cracraft, J. 1985. Monophyly and phylogenetic relationships of the Pelicaniforrnes: A numerical cladistic analysis. Auk 102:835-853.

- ~. -.-----~-~.~---. ------234

• 1986a. origin and evolution of continental biotas: --speciation and historical congruence within the Australian avifauna. Evolution 40:977-996.

• 1986b. The origin and early diversification of ~rds. Paleobiology 12:383-399.

Crane, P. R. 1985. Phylogenetic relationships in seed plants. Cladistics 1:329-348.

cronquist, A. 1968. The evolution and classification of flowering plants. Houghton Mifflin Co., Boston.

· 1987. A botanical critique of cladism. Bot. Rev. 53:1-52. Crother, B. I., M. M. Miyamota, and W. F. Presch. 1986. Phylogeny and biogeography of the lizard family Xantusidae. Syst. Zool. 35:37-45.

cutler, E. B., and P. E. Gibbs. 1985. A phylogenetic analysis of higher taxa in the Sipincula. Syst. Zool. 34:162-173.

Dahlgren, R., and K. Bremer. 1985. Major clades of the angiosperms. Cladistics 1:349-368.

Dahlgren, R., and F. Rasmussen. 1983. Monocotyledon evolution: Characters and phylogenetic estimation. Evol. BioI. 16:255-395.

Davis, P., and V. Heywood. 1973. Principles of Angiosperm Taxonomy. Krieger Publishing Co., Huntington, N.Y.

Doyle, J. A., and M. J. Donoghue. 1986. Seed plant phylogeny and the origin of angiosperms: An experimental cladistic approach. Bot. Rev. 52:321-431 •

. 1987. The importance of fossils in elucidating seed plant phylogeny and macroevolution. Rev. Paleobot. palyn. 50:63-95.

Eckenwalder, J. E., and S. C. H. Barrett. 1986. Phylogenetic systematics of the Pontederiaceae. Syst. Bot. 11:373-391.

Emerson, S. B. 1988. Testing for historical patterns of change: A case study with frog pectoral girdles. paleobiology 14:174-186.

Farris, J. S. 1969. A successive approximations approach to

------235

character weighting. Syst. Zool. 18:374-385.

· 1983. The logical basis of phylogenetic analysis, pp. 7-36. In N. Platnick and V. Funk (eds.), Advances in Cladistics, vol. II. Columbia Univ. Press, New York.

Felsenstein, J. 1978. Cases in which parsimony or compatibility will be positively misleading. Syst. Zool. 34: 152-161.

· 1982. Numerical methods for inferring evolutionary trees. Quart. Rev. BioI. 57:127-141.

· 1985. Confidence limits on phylogenies: An approach using the bootstrap. Evolution 39:783-791.

Fink, W. L. 1985. Phylogenetic relationships of the stomiid fishes (Teleostei:Stomiiformes). Misc. Publ. Mus. Zool. Univ. Mich. No. 171.

Fink, W. L., and S. V. Fink. 1986. A phylogenetic analysis of the genus Stomias, including the synonymy of Macrostomias. copeia 1986:494-503.

Funk, V. A. 1981. Special concerns in estimating plant phylogenies, pp. 73-86. In V. Funk and D. Brooks (eds.), Advances in Cladistics. Oxford Univ. Press, Oxford.

Futuyma, D. J. 1986. Evolutionary Biology. Sinauer Assocs., Sunderland, Mass. 2nd. ed. Gabrielson, P. w., and D. J. Garbary. 1987. A cladistic analysis of Rhodophyta: Floridophycean orders. Br. Phycol. J. 22: 125-138.

Gauthier, J. A. 1986. Saurischian monophyly and the origin of birds, pp. 1-55. In K.Padian (ed.), The origin of Birds and the Evolution of Flight. Mem. Calif. Acad. Sci. No.8.

Gauthier, J., R. Estes, and K. de Queiroz. 1988. A phylogenetic analysis of Lepidosauromorpha, pp. 15-98. In R. Estes and G. Pregill (eds.). Phylogenetic Relationships of the Lizard Families: Essays commemorating Charles·L. Camp. Stanford Univ. Press, Stanford, California.

Gauthier, J. A., A. G. Kluge, and T. Rowe. 1988. Amniote phylogeny and the importance of fossils. Cladistics 4:105-209.

George, M., and o. A. Ryder. 1986. Mitochondrial DNA evolution in the genus Equus. Mol. BioI. Evol. 3:535-546 .

...... ,.-- .' . __ ...._._---_. ---- ._.--.--.--- _-_ ------_.__ __ _._- 236

Gottlieb, L. D. 1984. Genetics and morphological evolution in plants. Am. Nat. 123:681-709. Grismer, L. 1983. A reevaluation of the North American Gekkonid genus Anarbylus Murphy and its cladistic relationships to Coleonyx Gray. Herpetologica 39:394-399. Guyer, C., and J. M. Savage. 1986. Cladistic relationships among anoles (sauria:Iguanidae). Syst. Zool. 35:509-531. Hamby, R. K., and E. A. Zimmer. 1988. Ribosomal RNA sequences for inferring phylogeny within the grass family (Poaceae). Plant Syst. and Evol. 34:393-400. Hennig, W. 1966. Phylogenetic Systematics. Univ. Illinois Press, Urbana. Herman, L. H. 1986. Revision of Bledius. Part IV. Classification of species groups, phylogeny, natural history, and catalogue (Coleoptera, Staphylinidae, Oxytelenae). Bull. Amer. Mus. Nat. Hist. 184:1-368. Hillis, D. M. 1987. Molecular versus morphological approaches to systematics. Ann. Rev. Ecol. Syst. 18:23-42. Hillis, D. M., and S. K. Davis. 1986. Evolution of ribosomal DNA: Fifty million years of recorded history in the frog genus Rana. Evolution 40:1275-1288. Huey, R. and A. Bennett. 1987. Phylogenetic studies of coadaptation: Preferred temperatures versus optimal performance temperatures of lizards. Evolution 41:1098-1115. Hufford, L.D. 1988. The evolution of floral morphological diversity in Eucnide (Loasaceae): The implications of modes and timing of ontogenetic changes on phylogenetic diversification, pp. 103-119. In P. Leins, S.C. Tucker, and P.K. Endress (eds.), Aspects of Floral Development. J. Cramer, Berlin. Jamieson, B. G. M., C. Edwards, and M. Ferraguti. 1987. parsimony analysis of the phylogeny of some Oligochaeta (Annelida) using spermatazoal ultrastructure. Cladistics 3:145-155. Jansen, R. K. 1981. Systematics of Spilanthes (Compositae: Heliantheae). Syst. Bot. 6:231-257. Jansen, R. K., and J. Palmer. 1988. Phylogenetic implications of chloroplast DNA restriction site variation in the Mutisieae (Asteraceae). Amer. J. Bot. 75:753-766. 237

Kellogg, E. and C. Campbell. 1987. Phylogenetic analysis of the Gramineae, pp. 310-322. In T. Soderstrom, et.al. (eds.), Grass systematics and Evolution. smithsonian Institution Press, Washington, DC.

Kimura, M. 1983. The Neutral Theory of Molecular Evolution. Cambridge Univ. Press, Cambridge.

Kitching, I. 1987. Spectacles and silver Y's: A synthesis of the systematics, cladistics, and biology of the Plusiinae (Lepidoptera: Noctuidae). Bull. British. Mus. Nat. Hist. (Entomol.) 54:75-261.

Kluge, A. G. and J. S. Farris. 1969. Quantitative phyletics and the evolution of Anurans. Syst. Zool. 18:1-32.

Ladiges, P. Y., and C. J. Humphries. 1983. A cladistic study of Arillastrum, Angophora, and Eucalyptus (Myrtaceae). Bot. J. Linn. Soc. 87:105-134.

Ladiges, P. Y., C. J. Humphries, and M. Brookes. 1987. Cladistic and biogeographic analysis of Western Australian species of Eucalyptus L'Herit, Informal subgenus Monocalyptus Pryor & Johnson. Aust. J. Bot. 35:251-281.

Lankester, E. 1870. On the use of the term homology in modern zoology. Annals Mag. Nat. Hist. series 4, 6:34-43.

Lauder, G. 1981. Form and function: Structural analysis in evolutionary morphology. Paleobiology 7:430-442.

____ . 1982. Historical biology and the problem of design. J. Theor. BioI. 97:57-67.

Lipscomb, D. L. 1985. The eukaryotic kingdoms. Cladistics 1:127-140.

Livezy, B. C. 1986. Phylogeny and historical biogeography of steamer-ducks (Anatidae:Tachyeres). Syst. Zool. 35:458-469.

Maxon, L. R., and A. C.·Wilson. 1974. Convergent morphological evolution by studying proteins of tree frogs in the Hyla eximia species group. science 185:66-68.

Mayr, E. 1969. Principles of Systematic Zoology. McGraw-Hill, New York.

Meacham, C. A. 1981. A probability measure for character compatibility. Math. Biosci. 57:1-18. 238

• 1984. Evaluating characters by character compatibility analysis, pp. 152-165. In T.Duncan and T.Stuessy (eds.), Cladistics: Perspectives on the Reconstruction of Evolutionary History. Columbia univ. Press, New York. Mishler, B., and S. P. Churchill. 1985. Transition to a land flora: Phylogenetic relationships of the green algae and bryophytes. Cladistics 1:305-328. Mishler, B, K. Bremer, C. Humphries, and S. Churchill. 1988. The use of nucleic acid sequence data in phylogenetic reconstruction. Taxon 37:391-395. Mishler, B., and M. J. Donoghue. 1982. Species concepts: A case for pluralism. Syst. Zool. 31:491-503. Miyamoto, M. M. 1983. Biochemical variation in the frog Eleutherodactylus brandfordii: Geographical patterns and cryptic species. Syst. Zool. 32:43-51. Montgomery, D., and E. Peck. 1982. Introduction to Linear Regression Analysis. John Wiley & Sons, New York. Nelson, C. H. 1984. Numerical cladistic analysis of phylogenetic relationships in Plecoptera. Ann. Entomol. Soc. Am. 77:466-473. Patterson, C. 1982. Morphological characters and homology, pp. 21-74. In K. A. Joysey and A. E. Friday (eds.), Problems of Phylogenetic Reconstruction. Academic Press, London. Patterson, C. (ed.). 1987. Molecules and Morphology in Evolution: Conflict or compromise? Cambridge Univ. Press, cambridge. Remane, A. 1952. Die Grundlagen des Naturlichen Systems, der vergleichenden Anatomie und der Phylogenetik. Akademische Verlagsgesellschaft, Leipzig. Rensch, B.. 1959. Evolution Above the Species Level. Columbia University Press, New York. Ridley, M. 1983. The Explanation of organic Diversity: The comparative Method and Adaptations for Mating. Clarendon Press, Oxford. Riggins, R. , and J. S. Farris. 1983. Cladistics and the root of angiosperms. Syst. Bot. 8:96-101. 239

Rodman, J. E., M. Oliver, R. Nakamura, J. U. McClammer Jr., and A. H. Bledsoe. 1984. A taxonomic analysis and revised classification of Centrospermae. Syst. Bot. 9:297-323.

Rohatgi, v. K. 1976. An Introduction to Probability Theory and Mathematical statistics. John Wiley & Sons, New York.

Sanders, R. 1983. Cladistic analysis of Agastache (Lamiaceae), pp. 95-114. In V. Funk and D. Brooks (eds.), Advances in Cladistics. New York Botanical Garden, Bronx, NY.

Sanderson, M.J. 1989. Confidence limits in phylogenies: The bootstrap revisited. Cladistics. In press.

Schuh, R. T., and J. T. Polhemus. 1980. Analysis of taxonomic congruence among morphological, ecological, and biogeographic data sets for the Leptopodomorpha (Hemiptera). Syst. Zool. 29:1-26.

Sessions, S., and A. Larson. 1987. Developmental correlates of genome size in Plethodontid salamanders and their implications for genome evolution. Evolution 41:1239-1251. sibley, C., and J. Ahlquist. 1987. Avian phylogeny reconstructed from comparisons of the genetic material, DNA, pp. 95-121. In C. Patterson (ed.), Molecules and Morphology in Evolution: Conflict or compromise? Cambridge Univ. Press, Cambridge.

Sillen-Tullberg, B. 1988. Evolution of gregariousness in aposematic butterfly larvae: A phylogenetic analysis. Evolution 42:293-305. sites, J. W., J. W. Bickham, B. A. Pytel, I. F. Greenbaum, B. A. Bates. 1984. Biochemical characters and reconstruction of turtle phylogenies: Relationships among Batagurine genera. Syst. Zool. 33:137-158.

Snedecor, G.W., and W.G. Cochran. 1980. Statistical Methods. Iowa State Univ. Press, Ames.

Sokal, R., and K. Shao. 1985. Character stability in 39 data sets. Syst. Zool. 34:83-89.

Stebbins, G. L. 1950. variation and Evolution in Plants. Columbia Univ. Press, New York •

• 1974. Flowering plants. Evolution Above the Species Level. Harvard Univ. Press, Cambridge, Mass. 240

stevens, P. F. 1986. Evolutionary classification in botany, 1960-1985. J. Arnold Arbor. 67:313-339. straney, D. 1981. The stream of heredity: Genetics in the study of phylogeny, pp. 100-138. In M. Smith and J. Joule, (eds.), Mammalian Population Genetics. Univ. of Georgia Press, Athens, Georgia. Sytsma, K. J., and L. D. Gottlieb. 1986a. Chloroplast DNA evolution and phylogenetic relationships in Clarkia sect. Peripetasma (Onagraceae). Evolution 40:1248-1261. • 1986b. Chloroplast DNA evidence for the origin of the genus Heterogaura from a species of Clarkia (Onagraceae). Proc. Nat. Acad. Sci. 83:5554-5557. Thorpe, P.A., and W.J. Dickinson. 1988. The use of regulatory patterns in constructing phylogenies. Syst. Zool. 37:97-105. Vavilov, N. 1922. The law of homologous series in variation. J. Genet. 12:47-89. Vilgalys, R. 1986. Phenetic and cladistic relationships in collybia sect. Levipedes (Fungi:Basidiomycetes). Taxon 35:225-233. Wagner, W. H. 1980. Origin and philosphy of the groundplan­ divergence method of cladistics. Syst. Bot. 5:173-193. ____ . 1984. Applications of the concepts of groundplan­ divergence, pp. 95-118. In T. Duncan and T. stuessy (eds.), .Cladistics: Perspectives on the Reconstruction of Evolutionary History. Columbia Univ. Press, New York. Wake, D. B., and A. Larson. 1987. Multidimensional analysis of an evolving lineage. Science 238:42-48. Wighton, D.C., and M. V. H. Wilson. 1987. The Gomphaeshninae (Odonata: Aeshnidae): New fossil genus, reconstructed phylogeny, and geographical history. Syst. Entomol. 11:505-522. Wyss, A. R., M. J. Novack, and M. C. McKenna. 1987. Amino acid sequence versus morphological data and the interordinal relationships of mammals. Mol. BioI. Evol. 4:99-116. 241

Appendix to "Appendix 1" Correction of Consistency Index for Autapomorphies and Invariant Characters In the case of binary characters, a character was omitted from the calculation if all of the taxa in the study were scored as having the same state (invariant) or if only one taxon was scored as having the presumed derived state (autapomorphic). This was done even when one or more taxa were scored as "unknown" for the character in question; in such cases parsimony algorithms assign the most parsimonious state to the questionable taxon based upon its position in the tree determined from other character data, and thus there is no possibility of inconsistency in that character. If only one taxon possessed the presumed ancestral (plesiomorphic) state, the character was retained on the grounds that it might still be inconsistent, if it were later found to be most parsimonious to assume a transition to the derived state between the outgroup(s) and the ingroup and a reversal to the outgroup condition within the ingroup cladogram. We assumed that outgroups were consulted in assessing character polarities, unless it was explicitly stated that some other method was employed. If a character was explicitly treated as undirected (polarity unknown) then it was eliminated if only one taxon had a particular state, whether this was coded as 0 or 1 in the matrix. The same rules were also applied in standardizing multistate characters that were treated as unordered, that is, entailing only one step between any pair of states. However, in such cases entire characters were not eliminated; instead, a step was subtracted from the numerator and denominator in calculating the CI when only one taxon possessed a particular derived state. In contrast, in the case of ordered mUltistate characters, steps were not subtracted when only one taxon possessed a derived state on the grounds that the derivation of that state could entail more than one step. This would be the case if the state possessed by closely related taxa in the tree under consideration were more than one step removed in the ordered transformation series. 242

TABLE A1-1. Cladistic analyses used in this study. "Morph" refers to morphological data (broadly construed), "molec" to molecular data. See text for details. Number of characters is standardized to number of binary character equivalents (Sokal and Shao, 1985).

DATA NUMBER NUMBER- NUMBER

KINGDOM TYPE RANK TAXA CHARS AUTAPOS CI SOURCE

plant morph spp 6 17 6 0.79 Jansen, 1981 plant morph spp 9 19 o 0.79 Anderberg, 1986 plant morph spp 14 31 3 0.72 Ladiges & Humphries, 1983 plant morph spp 20 28 4 0.44 Campbell, 1986 plant morph spp 25 51 12 0.63 Funk, 1982 plant morph spp 29 65 3 0.43 Ladiges, et.al., 1987 plant morph spp 37 65 3 0.39 Eckenwalder & Barrett, 1986 plant morph gen 29 126 o 0.31 Baum, 1983 plant morph gen 29 81 34 0.67 Bremer, et.al., 1987 plant morph gen 65 39 o 0.37 Kellogg & Campbell, 1987 plant morph gen 68 94 o 0.32 P.stevens, pers.comm. plant morph fam 15 56 12 0.60 Dahlgren & Rasmussen, 1983 plant morph fam 20 71 2 0.35 Rodman, et.al., 1984 plant morph fam 47 61 2 0.26 Dahlgren & Bremer, 1985 plant morph ord 15 37 6 0.55 Gabrielson & Garbary, 1987 plant morph cla 11 41 13 0.82 Mishler & Churchill, 1985 243

TABLE Al-l. (eon'd) DATA NUMBER NUMBER NUMBER

KINGDOM TYPE RANK TAXA CHARS AUTAPOS CI SOURCE ------plant morph ela 20 31 0 0.62 Crane, 1985 plant morph ela 20 62 0 0.50 Doyle & Donoghue, 1986 plant moled spp 9 119 64 0.90 Sytsma & Gottlieb, 1986a plant molecf gen 10 117 0 0.73 Hamby & Zimmer, 1988 plant molee' gen 16 211 156 0.60 Jansen & Palmer, 1988 plant moled gen 12 390 273 0.65 Jansen & Palmer, 1988 plant moled gen 55 868 568 0.52 Jansen & Palmer, pers.eomrn. plant molecf fam 6 111 35 0.60 Bremer, 1988 plant molecf fam 9 88 26 0.50 Bremer, 1988 plant molecf ord 26 84 0 0.53 Bremer, et.al., 1987

animal morph spp 4 23 10 1. 00 Livezy, 1986 animal morph spp 6 34 13 0.84 Grisner, 1983 animal morph spp 9 33 3 0.81 Cracraft, 1986 animal morph spp 11 83 41 0.64 Jamieson, et.al., 1987 animal morph spp 12 32 0 0.70 Fink & Fink, 1986 animal morph spp 20 74 1 0.68 Collette & Russo, 1985 animal morph spp 24 31 1 0.60 Guyer & Savage, 1986 animal morph spp 27 94 5 0.40 Cane, 1983

animal morph spp 35 72 7 0.43 Herman, 1986 animal morph gen 5 36 13 0.72 Crother, et.al. , 1986 animal morph gen 7 26 0 0.76 Carpenter, 1987 244

TABLE Al-l. (con'd)

DATA NUMBER NUMBER NUMBER

KINGDOM TYPE RANK TAXA CHARS AUTAPOS CI SOURCE ------animal morph qen 10 49 12 0.80 Schuh & Polhemus, 1980 animal morph qen 16 15 0 0.50 Wiqhton & Wilson, 1987

animal morph qen 17 18 5 0.54 CUtler & Gibbs, 1985

animal morph qen 27 323 78 0.49 Fink, 1985 animal morph gen 57 307 24 0.45 Kitching, 1987 animal morph fam 13 59 0 0.66 Cracraft, 1985 animal morph cla 8 84 0 0.89 Gauthier, 1986 animal morph fam 22 113 43 0.63 Nelson, 1984 animal morph cla 9 39 11 0.93 Brooks, et.al. , 1985 animal molec' spp 6 118 87 0.56 Georqe & Ryder, 1986 animal molec' spp 32 76 55 0.66 Hillis & Davis, 1986 animal moled ord 13 19 0 0.59 Wyss, et.al. , 1987

animal moled ord 14 18 0 0.49 It

animal moled ord 14 27 0 0.60 It

animal moled ord 19 17 0 0.63 It

animal molee' spp 12 21 2 0.95 Miyamoto, 1983

animal mOlee' spp 25 90 38 0.38 Sites, et.al. , 1984 animal molee' spp 26 68 0 0.37 W.Heed, pers.comm.

animal molee' qen 14 92 46 0.47 Straney, 1981

other morph spp 11 25 5 0.53 vilgalys, 1986

other morph gen 17 24 0 0.53 Baum & Saville, 1985

._...... -.-_._------245

TABLE A1-1. (con'd)

DATA NUMBER NUMBER NUMBER

KINGDOM TYPE RANK TAXA CHARS AUTAPOS CI SOURCE ------other morph fam 18 105 8 0.63 Churchill,et.al., 1984 other morph cla 36 77 38 0.34 Lipscomb, 1985

Restriction-fragment analysis

~ucleotide sequence analysis

~rotein polymorphism electrophoretic analysis 246

TABLE Al-2. Total Correlation Matrix. Significance levels: *** = p < 0.001, ** = P < 0.01.

Number Number Rank of Taxa of Characters

log(CI) -0.68*** -0.12 -0.01

Number of taxa 0.38** -0.04 Number of characters -0.05 247

TABLE Al-3. Regression parameter estimates of log transformed consistency indices for subsets of the data. Significance levels: *** = p < 0.001, **= P < 0.01, * = P < 0.05.

PARAMETER ESTIMATE

SAMPLE Number of Number of SUBSET SIZE INTERCEPT Taxa characters Rank

------All data 60 -0.2528** -0.0158*** 0.0003 -0.0092

Animals 30 -0.2507* -0.0160** 0.0001 0.0151

Plants 26 -0.2740* -0.0143*** 0.0004 -0.0236

Morph. 42 -0.2126* -0.0170*** 0.0002 -0.0029

Molec. 18 -0.3552* -0.0104 0.0003 -0.0191 248

FIGURE LEGENDS

Fig. Al-l. Graphs of consistency index (CI) versus number of taxa included in 60 cladistic analyses (Table Al-l). A.

Untransformed consistency indices versus number of taxa. B.

Log-transformed consistency indices and linear regression of

CIon number of taxa (see Table Al-3) .

Fig. Al-2. Log-transformed consisteucy index versus number of characters in 60 cladistic analyses (Table Al-l) .

Studies with more than 400 characters have been omitted from the graph (see text).

Fig. Al-3. Example data set with high homoplasy and yet high confidence as measured by Felsenstein's bootstrap confidence assessment. Numbers along nodes refer to number of replicates out of 100 in which clade occurred. The first four characters in the data set are represented three times each, so that 24 characters are included in all. Tree constructed using the Branch and Bound algorithm in PAUP, which produced one minimal tree of length 36 steps. The CI 249 is much below empirically observed values for six taxa (see

Fig. AI-IA), indicating sUbstantial homoplasy.

Fig. Al-4. Log-transformed consistency index versus taxonomic rank (Table AI-I), in which ranks are assigned an integer value corresponding to the Linnaean hierarchy: from species = 1 to classes (and above) = 5.

Fig. Al-5. comparison of log-transformed consistency index versus number of taxa in plant versus animal data sets

(Table AI-I). Simple regressions of CIon number of taxa are plotted for each group. Regression with shallower slope is for plants. 250

• o·• • A o 8 · t.· • • • •• ••• • • • .6 •••• • • •••• • • • •• • •• • I • • • • .4 ••• • • • • • • • • .2

o~~ __~ __~ ____~--~--~--~~------~_~ o 10 20 30 40 ~O 60 o • .J • 8 -.2 •••

-.4

~ -.6 ;, .: -.8

-I

-1.2

-1.4+-__-..,..-_-..,..-_-~------• o 10 20 lO 40 !O

Number- of Taxc

Fig. A1-1. Graphs of consistency index (Cl) versus number of taxa included in 60 cladistic analyses (Table Al-l). A. untransformed consistency indices versus number of taxa. B. Log-transformed consistency indices and linear regression of CIon number of taxa (see Table Al-3) . 251

0 • • • • • -.2 . ,.. • • -.4 • • ..,. • ~ •• ••• • c..l -.6 . • - • =" ~.. - • • • = -.S • 0 •• • - 0 -I 0 • • •·0 -1.2 • 0

-1.4 0

-1.6 0 SO 100 ISO 200 2'0 lOO lSO 400 J.S: Number a r Characters

Fig. A2-2. Log-transformed consistency index versus number of characters in 60 cladistic analyses (Table Al-I) . Studies with more than 400 characters have been omitted from the graph (see text). 252

a 1 1 1 1 (3x) a (3x) 1 1 a a a a a 1 I 0 0 (3x) a ( 3x) 0 a 0 a 1 I 1 a 1 a a 0 1 0 a I a 0 1 a a a 1 0 1 a a 0 a I 0 1 1 0 0 0 0 1 0 1 0 0 a 1 a a 1 a 0 1 0 a a 1 0 0 1 a 1 0 0 0 1 0 0 1 a 0 0 1 1 0 0 a 0 I 0 1

Fig. A1-3. Example data set with high homoplasy and yet high confidence as measured by Felsenstein's bootstrap confidence assessment. Numbers along nodes refer to number of replicates out of 100 in which clade occurred. The first four characters in the data set are represented three times each, so that 24 characters are included in all. Tree constructed using the Branch and Bound algorithm in PAUP, which produced one minimal tree of length 36 steps. The CI is much below empirically observed values for six taxa (see Fig. A1-1A), indicating substantial homoplasy. 253

0 • • • -.2 I • • I -.4 I I • • • • • • -.6 • • (.J • I I :') • • • • = -.8 • • -1 I • • • -1.2 I

-1.4 •

-1.6 2 :;

Taxonomf cRank

Fig. A1-4. Log-transformed consistency index versus taxonomic rank (Table Al-l), in which ranks are assigned an integer value corresponding to the Linnaean hierarchy: from species = :I.. to classes (and above) = 5. 254

0 0 0 Animol -.2 • ?Iont

-.4 0 0 o. 0 o • • 0 -.6 0• c...l • 0 C'I • 8 0 • = -.8 • -I • -1.2

-1.4 •

-1.6 0 10 20 30 40 50 60 --. -

Number of Ta)(~

Fig. A~-5. Comparison of log-transformed consistency index versus number of taxa in plant versus animal data sets (Table Al-l). Simple regressions of CIon number of taxa are plotted for each group. Regression with shallower slope is for plants. 255

0 • -.2 • MOrpholog1C~1 Dete o Molecul~r Det~ -.4 • 0

(..l -.6 0 • ~ • -= -.8 • -I • -1.2

-1.4 •

-1.6 0 10 20 30 40 SO ... -~ ~"- -

Number of Taxa

Fig. A1-6. Comparison of log-transformed consistency index versus number of taxa in morphological versus molecular data sets (Table A~-l). Simple regressions of CIon number of taxa are plotted for each group. Regression with shallower slope is for molecular data.

-----_. ---.--- .------.--- 256

APPENDIX 2

CHARACTERS USED IN CLADISTIC ANALYSIS OF CHAPTER TWO

State 0 is generally the primitive state for binary characters, although this was not assumed in the construction of cladograms. Multistate characters are also unpolarized, and are all unordered except character 33. For discussion of rooting see Chapter Two.

Authorities for all species discussed below can be found in Barneby (1964).

Vegetative Characters

1. Point of renewal of perennial growth [0 = at soil surface; 1 = subterranean].

In some groups of perennials with caudices, particularly those with well-developed internodes, the caudices may develop for some length underground. Usually several branches originate from the root crown and extend upward through several leafless internodes before emerging above ground. Branching of the stem often occurs at or near ground level (as is also typical of plants with a superficial caudex), which gives the pressed plant the appearance of branching upwardly. 257

Annuals are scored as "unkown" for this character since they have no point of renewal; this allows them to be equally parsimoniously derived from perennials with either type of caudex.

2. Duration of root [0 = perennial with well-developed caudex; 1 = short-lived perennial with poorly developed caudex; 2 =annual, without caudex].

3. Type of root [0 = taproot; 1 = fibrous root system].

In most species the root is uniformly a simple taproot; in some annual groups the taproot is ramified into fine fibers.

4. Growth pattern of stem. [0 = acaulescent, leaves usually much longer than entire stem; 1 = moderately caulescent, internodes often lengthening upward, leaves usually longer than internodes; 2 = strongly caulescent, internodes often uniformly long leaves usually about the length of internodes].

The distinction between the two caulescent states is subtle and sometimes questionable in practice.

5. Robustness of stems [0 = robust, thick, often hollow, 1

= gracile, thin, not hollow].

6. Presence of developed petiole [0 = lowest leaflets at least 1/3 of length of leaf from leaf base; 1 = lowest leaves sub-sessile, the petiole very short]. 258

Typically, leaves on acaulescent plants are not sessile, perhaps for developmental reasons. On caulescent plants the uppermost leaves are generally the ones with the shortest petioles; hence the lower leaves were examined for this character. 7. Leaf orientation [0 = spreading from the stem; 1 = narrowly appressed to the stem ].

In a few species the leaves are closely appressed to the stems, diverging at an angle less than 15 to 20 degrees.

Other characters of the leaf such as small, deciduous leaflets are often correlated so that the plant assumes a grass-like or rush-like habit. In some instances the close appression may be correlated with connate-sheathing

stipules, which may act as a functional constraint

restricting ability of the leaf to spread away from the

stem. 8. Presence of oblanceolate phyllodes [0 = absent; 1 = present].

The species comprising section Drabellae exhibit

phyllodes that are dorsiventrally flattened and expanded

distally into an oblanceolate, spoonlike structure. Even in

A. detritalis, which has most leaves more normally pinnate,

early leaves show this oblanceolate phyllode. 9. Presence of "pectinate" leaf [0 = absent; 1 = present]. 259

Several species comprising the aptly named section

Pectinati share a peculiar leaf morphology characterized by elongate, flattened, decurrent leaflets that spread from the rachis and th~r curve inward so as to parallel the rachis.

10. Leaflet number [0 = 20 or above: 1 = 8-20; 2 = less than 8].

The majority of species fall into the intermediate class. The character state boundaries are arbitrary and are coded as polymorphic when significant overlap between states is observed; see Discussion of quantitative characters in

Chapter Two.

11. Leaflet size [0 = less than 4 rom long; 1 = between 4 rom and 15 rom long; 2 = more than 15 rom long].

See discussion of character 10.

12. Leaflet folding [0 = folded conduplicately; 1 = flat]. This is a highly variable character but in some taxa the leaflets are consistently flat (i.e., not folded along the midrib) .

13. Leaflet venation [0 = at most the midrib visible; 1 = secondary veinlets also visible; 2 = conspicuously "veiny": secondaries and tertiaries visible].

In most species only the midrib is visible, but in a few the leaflets are visibly pinnately nerved. Usually this occurs when the leaflets are unusually thin and glabrescent, but it also can occur in very thick-textured leaflets (e.g., 260

some varieties of A. mollissimus). Reticulate-veined leaflets are confined primarily to alpine plants of the Old World chromosome series.

14. Leaflet shape [0 = orbicular-oblong-elliptic; 1 = narrowly linear]. Leaflet shape is very plastic within and between species, varying from sub-orbiculoar to narrowly linear, including various grades of oblong, elliptical, obovate, rhombic, etc. Only a rough distinction between linear and "not linear" appears to be consistent enough within terminal taxa to be useful at this level of analysis. "Linear" may be defined as a leaflet four or more times as long as broad with approximately parallel sides and rounded apex. Numerous cases occur in which species mostly exhibiting linear leaflets have narrowly elliptic, acute or even rounded leaflets (e.g., A. wootoni); nonetheless the character appears useful at this stage. 15. Terminal leaflet attachment [0 = jointed; 1 = confluent]. In a number of taxa the terminal leaflet, as well as the lateral leaflets in many cases, has become confluent with the leaf rachis, apparently by loss of the petiolule. Correlated with this loss or preceding it in some cases is a general lengthening of the terminal leaflet. Rarely different leaves on the same plant may have jointed or

_____ 0--_--- ___ ------. ---- - _.. _-_._------261 confluent terminal leaflets, but this presumably reflects developmental instability. In some species all lateral leaflets are lost and the leaf assumes the form of a phy110de. The character has not been studied in detail developmentally, and it is not clear whether it stems from an elongation of the petiole and loss of the terminal leaflet, or merely loss of the lateral leaflets with the rachis and petiole remaining intact. The latter explanation seems more likely.

16. Leaflet apex [0 = rounded; 1 = notched; 2 = acute].

17. Spinescence [0 = leaf rachises neither persistent nor indurated; 1 = rachises persistent and indurated].

Although common in the Old World, spinescence is relatively rare in North America with the exception of the A. kentrophyta complex. In several other groups the rachis becomes indurated as well (e.g., A. humi11imus).

18. Stipu1es [0 = free; 1 = connate at base of plant but free apically; 2 = connate throughout plant].

This mU1tistate character clearly reflects a continuum of one developmental sequence initiated or terminated at different times. The stipu1es in some groups are connate on the side of the stem opposite from the petiole. This connation'a1most always is most pronounced at the lowest, earliest nodes, and the stipu1es become less connate, more free, apically. In very few groups (e.g., Humistrati, some 262 ocreati), the stipules are connate even at the uppermost nodes. Correlated with this character is a tendency for the stipules to become narrower and longer apically. In some

cases a strong dimorphism exists between upper and lower

stipules, but it is difficult to quantify the differences. 19. stipule size [0 = less than 3 rom; 1 = between 3 and 8

rom; 2 = greater than 8 rom].

Within an individual plant there is often much size and

shape variation in the stipules; in some the stipules get

longer apically (~ argophyllus) , while in others they get

smaller. Measurements were taken on the largest stipules.

20. Stipule shape [0 = deltoid, lanceolate, caudate; 1 =

broad and truncate-rounded].

21. Presence of black stipules [0 = absent; 1 = present] ; In a few species of section Ervoidei the stipules are

black, rather than green or dingy brown.

22. Presence of scarious stipules [0 = absent; 1 = present] .

Many species have stipules with hyaline margins but the

apomorphy here refers only to those that are uniformly

transparent and scarious.

23. Trichome and Pubescence type [0 = basi fixed and spreading; 1 = basifixed and appressed; 2 = medifixed and appressed].

The orientation, length, and quantity of the pubescence

is one the most useful suites of characters for marking 263 individual species in Astragalus; its use at higher levels has always been somewhat controversial. All hairs in the North American species of the genus are simple and unicellular (exclusive of the epidermal bases) and are either attached at the base (basifixed) or attached away from the base (medifixed). The latter state varies from attached medially to attached near one end. The strigose, basifixed hair lies flat on the surface of the plant and arches sharply at its base. In some groups, notably section Argophylli, several species with essentially strigose basifixed hairs are minutely spurred at base, which represents either an independent origin of medifixed hairs or is indicative of the underlying mechanism by which such hairs are produced. All three types of hair may be found among the many varieties of A. miser. In addition, near relatives of species with medifixed hairs are uniformly strigose if basifixed, suggesting that basifixed and strigose is intermediate between medifixed hairs and hairs that are more erect. 24. Pubescence density [0 = scanty; 1 = copious]. Relatively few groups have species with generally copious vestiture, here defined as pubescence sufficiently dense to markedly alter the overall reflectance or color of the plant. 264

Inflorescence Characters

25. Inflorescence type [1 = long peduncle, dense and long

raceme (1/3 to 1/2 of infl. is raceme) -- flowers declined;

2 = as state 0, but flowers ascending; 3 = long peduncle,

dense and short raceme, subcapitate (less than 1/4 of axis

is raceme) -- flowers ascending; 4 = long peduncle,

moderately dense but elongate raceme (more than the space of

one flower between flowers) -- flowers ascending; 5 = as

state 4, but flowers declined; 6 = short peduncle (length of

leaf), few-moderate number of flowers (1-4 fls.) -- flowers

ascending; 7 = sub-sessile, few-flowered].

Variation in the length and density of the raceme, as

well as the length of the peduncle, has never been utilized

to the same extent in New World species as it has in Old

World taxonomies. The New World lacks the very dense,

spikelike axillary racemes that occur in some Old World

groups. However, the raceme does vary strikingly between

higher taxonomic groups in the New World. At one extreme

are long, dense racemes held on long peduncles (mostly in

robustly caulescent plants of prairies or weedy species,

like A. bisulcatus or A. praelongus). At the other extreme

are a very few species with essentially obsolete peduncles

and one or a few flowers in the axils of the leaves (section

Humillimi, Orophaca, A. nutriosensis Sanderson). An

important third type is a short dense raceme on a long

. -_ ..._ .•.. _--- 265

peduncle ("subcapitate") which occurs in some annual groups

(also Humillimi). The problem in scoring this character is

that a continuous set of intermediates exists between

character states, and there is much variation within the

same species and even the same plant. This character is

therefore expected to exhibit rather high levels of

homoplasy, but it should prove useful in resolving

relationships of some groups.

This character has been combined with the floral

orientation character. See discussion under "character

correlation" in Chapter Two.

26. Length of pedicel [0 = less than 2.5 mID.; 1 = greater

than 2. 5 rom.].

In only a few groups does the pedicel elongate prior to

anthesis so that the flower is held away from the raceme

axis. In many more groups, the pedicels ultimately elongate

in fruit but the character refers only to the length of the

pedicel at earlier stages.

27. Calyx and corolla conformation [0 = moderately large, straight, short broad calyx: 1 = large straight and narrow with long calyx; 2 = small but straight, camp. calyx; 3 = small and sharply incurved, camp. calyx].

Barneby recognizes two basic types of flowers in North

American Astragalus, large flowers with cylindrical calyces,

and smaller flowers with campanulate calyces. A careful

.. ------..---- .... ------.. - ._-_ .. _._------266 morphometric analysis of flower variation would probably

reveal much hitherto unutilized phylogenetic variation. I have recognized five basic types. Among large flowers with

cylindrical calyces, some have short, broad calyces (often with long teeth) with relatively broad wings and keel and moderate incurvature of petals (e.g., section Megacarpi,

some Preussiani; usually in a syndrome involving yellow,

drooping flowers), while others have much narrower calyces,

with narrower petals that are straighter (section

Argophylli). Among smaller flowers with campanulate

calyces, there are two types that are more or less

distinguishable based on the degree of incurvature of the

petals. One distinctive type (e.g. A. ceramicus) has very

strongly incurved petals when viewed in profile, the banner

recurved through ninety degrees or more.

28. Length of calyx teeth [0 = much shorter than the tube; 1 = at least 2/3 as long to longer than the tube]. 29. Petal Color [0 = pink-purple; 1 = ochroleucous or whitish-yellow] •

Flower color is one of the most variable characters

above the species level. Even within species many varieties

are partly (or entirely) characterized by differences in flower color.

... .-_.. _------267

30. Length of wing petal [0,1, .•.. ,9].

This continuous character is scored by polymorphism

overlap coding (see Chapter TWo, under "Character

Correlation". The length is measured from tip of the wing

petal to base of the calyx and is log-transformed. 31. Presence of irregularly graduated petals [0 = absent; 1

= present].

Typically, the papilionoid flower in North American

Astragalus has the banner longer than the wings which in

turn are longer than the keel (Barneby, 1964). Rarely, the

banner is shorter than the wings, which is termed irregular

graduation. In Old World groups such as ~. alpinus, the

keel is large, blunt and longer than the other petals, which

is a different sort of irregular graduation.

Fruit Characters 32. Fruit orientation [0 = erect-ascending; 1 = declined]. 33. Valve texture [0 = membranous, transparent; 1 = papery,

opaque but not stiff; 2 = stiff, leathery, ligneous].

The thickness of the mesocarp is quite variable at

higher taxonomic levels. Membranous valves generally are

found in inflated fruits. I presume that papery valves is

intermediate between the other two states and hence this character is treated as ordered.

------268

34. Valve surface [0 = smooth or faintly reticulate or nerved: 1 = coarsely reticulate]. In a few species, typically with leathery valves, the surface is coarsely wrinkled-reticulate from the elevated veins.

35. Valve pigmentation [0 = clear, or uniformly colored; 1

= uniformly mottled in small patches; 2 = minutely speckled] . Many species have pods that are stained at maturity with patches of reddish pigment (often drying dark) , probably anthocyanins. Since these are often evident on the surface exposed to most direct sunlight it may be an environmentally triggered trait. However, its distribution among species suggests a genetic basis for the character. 36. Presence of spongy-pithy mesocarp [0 = absent; 1 = present] .

This is a rare specialization of the mesocarp found in sections Sarcocarpi (including the recently described ~ bibullatus) and Tennesseenses. It is also found in one species of section Argophylli, A. chamaeluce, although this is probably homoplastic. The mesocarp becomes unusually fleshy and thick. Species of section sarcocarpi are the

"ground plums" of the Great Plains.

37. Presence of pulpy filaments in locule [0 = absent, 1 = present] • 269

Pulpy filaments are found inside the cavity of the

fruit of many species. Rydberg used the character to help

diagnose his genus Pisophaca, a hodgepodge of species now

found among several of Barneby's sections (Lonchocarpi,

Scytocarpi). Unfortunately, the quantity of pulp appears to

vary continuously, most species having at least a small

amount of it. This character deserves a careful survey that

was not feasible in this study.

38. septum [0 = unilocular or with rudimentary septum; 1

bilocular, but usually unilocular in the beak (the apical

laterally compressed region associated with the style); 2

fully bilocular, dehiscence then into carpel-like halves].

No character has had a greater impact on Astragalus

systematics in the New World than the septum in the legume.

The septum is a longitudinal invagination of the endocarp

arising from the activity of a subepidermal meristem in the

region of the dorsal suture (Baum, 1948; Roth, 1977). Roth

interprets it is as a wing or rib wholly analogous to

similar structures on the outside of carpels. It does not

fuse with the ventral (adaxial) suture except in a few cases

CA. crassicarpus), nor does it usually extend fully into the

apical, beak region of the pod. Unlike the Old World

species, which are predominantly bilocular, the septum is

evolutionarily labile in the New World. It is consistently

------270 present in many groups, but is absent in other large parts of the genus.

39. Fruit persistence [0 = deciduous from receptacle; 1 = persistent on receptacle; 2 = persistent but fruit and pedicel falling as a unit]. In many generic keys to Fabaceae, Astragalus is suggested to have indehiscent or tardily dehiscent fruits. In fact, the dehiscence varies greatly depending on species. Barneby relied heavily on this character for higher level taxonomic groupings. In some taxa, the pod remains firmly attached to the raceme long after it matures, and dehiscence occurs at that point. In other taxa, dehiscence occurs after the fruit has fallen off the raceme, which generally occurs soon after maturity. Abscission occurs at a joint between the receptacle and pod. In rare cases, the fruit is firmly attached to the pedicel, which falls as a unit (often belatedly) • 40. Number of ovules [0 = more than 30; 1 = 8-30; 2 = less than 8]. This character is undoubtedly correlated with the size of the pod (characters 51,52), in that very small pods necessarily have very few ovules. Some very large pods can also have relatively few ovules, however, so these traits are not necessarily coupled. Data are taken from Barneby

(1964) • 271

41. Dehiscence [0 = apical/ventral; 1 = apical/ventral/ through stipe; 2 = ventral and dorsal, with spiral coiling]. Dehiscence is quite uniformly via the apex (beak) of the pod and through the ventral suture (hence belying the classical definition of "legume"!), although eventually the dorsal suture often becomes involved. In stipitate fruits, dehiscence also may occur from the base of the stipe upward in addition to through the standard means. In a few groups the valves became separated from the apex downward and coil elastically as in Lotus and some other legumes. The mode of dehiscence is governed by the relative orientation of various fibrous layers. Inner and outer strata may cross each other at 30, 45, or 90 degree angles, producing various kinds of bending on dehiscence (Roth, 1977: p. 210). 42. Stipe [0 = none; 1 = shorter than calyx; 2 = longer than calyx]. 43. Gynophore [0 = absent; 1 = present]. The stipe is an elongated basal portion of the gynoecium, continuous with the ovary proper but below it; the gynophore is an elongation of the receptacle and a joint exists between it and the gynoecium. These two characters might have been coded as separate states of the same character since one almost never finds both a stipe and gynophore in the same species. Rarely, however, some specimens in Barneby's Pacific Piptolobi group do exhibit 272 both a stipe and gynophore, and hence it seems appropriate to retain them as separate characters, despite the functional analogy that they represent. The morphology of the stipe varies from a long filiform structure that suddenly broadens into the base of the ovary to a short-cuneate sterile base. The gynophore when small is most often a short conical prolongation of the receptacle. In the Pacific Piptolobi it is often long and filiform and mimics the filiform stipe seen in very closely related species. Questions of homology abound here, and one wonders if a joint may have evolved between the stipe and the ovary in some of these species.

44. Upper suture shape [0 = upper suture arched; 1 = upper suture straight]. In some species (e.g. A. trichopodus and relatives), the fruit appears very asymmetric in profile, with the upper suture almost straight from beak to receptacle. Ordinarily the upper suture arches up from the joint with the receptacle before becoming parallel to the axis of the fruit. The presumed apomorphic state is most noticeable in inflated fruits, and is often difficult to score in uninrlated ones. Only the most exceptional cases have been scored as the apomorphic state. Note that the apomorphic state for this character is logically correlated with state 0 of the next character.

----.------273

45. Fruit curvature [0 = straight; 1 = gentle and uniform; 2 = coiled]. Differences in growth rates of tissues around the dorsal and ventral sutures lead to various degrees of curvature or even spirally curved fruits (Roth, 1977) in other leguminous genera like Medicago. Most species of

Astragalus have pods that are either straight or gently curved. The legumes of a few (the Collini group and relatives, for example) are coiled inward through 180 to 360 degrees or more. It is not at all obvious that this state should be derived from state 1, since "gently incurved" does not really resemble less extreme specimens of the coiled species or early developmental stages, so the character is left unordered. A few species (A. recurvus, ~ drummondii) have decurved (curved downward) fruits, which are probably derived from state 0 and are therefore scored as such.

46. Beak [0 = obscure to cuspidate; 1 = strongly differentiated and upturned; 2 = weakly differentiated but decurved] .

47. Inflation [0 = absent; 1 = present]. Inflation of the fruit, involving expansion of the interior of the pod so that the seeds are suspended in a space much greater than they require, is common in the genus and elsewhere in the tribe. The extent of inflation varies over a wide range: the largest fruits in North America are 274 greatly inflated (e.g., those in A. megacarpus can be over 8 cm. across), but in smaller examples inflation can occur in pods less than 1 cm in diameter.

48. Dorsal surface [0 = smooth; 1 = dorsally grooved; 2 = carinate]. The dorsal (abaxial) suture of the fruit varies in prominence and in the degree to which it is sunken into the surrounding tissue. Often in bilocular fruits (though not always), the suture will be buried in a longitudinal dorsal groove. Alternatively, if it is prominent, there may be a dorsal carina or narrow wing. If the suture is not prominent the dorsal suture may simply be smooth.

49. Fruit cross-section [0 = more or less dorsi-ventrally compressed or wider than broad (tends to dorsi-ventral

compression, terete, inflated); 1 = narrow (tends to lateral compression or triangular)]. There is too much within-species variation in cross­ section of the pod to score all the various types described in the taxonomic literature of the genus. "Dorsi-ventral" grades into "terete" and "trigonous" and so on. In this study I have settled for a very crude distinction between fruits that are essentially narrow (because of strictly lateral compression or strongly triangular compression), and ones that are not. A morphometric analysis is warranted here.

"-" -.-.~------275

Because of the uncertain developmental derivation of inflated fruits, they are scored? for this character unless they are known to be laterally compressed early in development. otherwise, the necessary dorsiventral compression that goes hand-in-hand with inflation will tend to bias the results and make all inflated taxa derived from dorsi-ventral uninflated ones.

50. "Bisulcate fruits" [0 = absent; 1 = present].

Rarely, the ventral suture of the pod is prominent but lies in a longitudinal yentral sulcus, hence producing a dual longitudinal sulcus on the adaxial side of the pod.

51. Fruit length [0,1, •.. ,9]. See character 52.

52. Fruit breadth [0,1, •.• ,9].

The coding of the fruit length and width are discussed in Chapter Two under "Quantitative Characters."

Polymorphism overlap coding was used. Length was measured from the base of the non-stipe region of the ovary to the tip of the legume below the style. Breadth is measured from ventral side of the fruit to the dorsal side at its widest point, usually about the middle.

53. Fruit inclusion in calyx [0 = absent; 1 = present]. In a few species the fruit is small, persistent, and enveloped by the marcescent calyx.

54. Fruit drying green [0 = brown or straw-colored at maturity; 1 = green at or just prior to maturity]. 276

To my knowledge this character has never been used in taxonomic treatments of the New World species. In some species, particularly those with membranous valves of the pod (but also in some with thicker valves such as ~ tenellus), the valves are green at or just before maturity. In the majority of species, the pod is dark brown to straw colored very early in development. scoring of this character is based on dried herbarium specimens and hence may reflect something about the mode of drying rather than the state in living plants. The character is one of many that deserves further field study.

Biochemical and Cytological Characters 55. Nitrotoxin accumulation. See below. Many species of Astragalus in the New World synthesize

nitrogenous toxic compounds. Mis~rotoxin, the beta-D­ glucoside of 3-nitro-1-propanol (3-NPOH), was first identified in A. miser (Stermitz, et. aI, 1969). Since then a number of other related compounds have been identified. All can be classified according to their toxic catabolite, either 3-NPOH or 3-nitro-propionic acid (3-NPA). Williams and Barnaby (1977a) conducted a survey of 505 taxa (species and varieties) in North America and found nitrotoxins in at least trace levels in about half. The quantitative assay 277

used was reported on a linear scale of "T" for "trace" through 1, .... ,5 (most concentrated). Williams and Barneby suggest that values of 4 and 5 generally correspond to the presence of miserotoxin (3-NPOH catabolites), while lower values indicate 3-NPA catabolic products. Coding of this character presents difficulties because the published work does not list species assayed yet found lacking nitrotoxinsi hence it is not clear whether a species not listed in their Table I lacks nitrotoxins or was simply omitted from the study. Also, many taxa not included in the list because of readings below "trace" may actually have nitrotoxins at levels below the sensitivity of the assay. This is suggested by the sporadic occurrence of trace values within many sections (e.g., section Inflati). Nitrotoxins are entirely absent from the groups that accumUlate selenium, except for A. toanus of section Pectinati. Nitrotoxins are also absent from some non-seleniferous sections notably Megacarpi and three sections comprising the "Pacific Piptolobi," sensu Barneby. Nitrotoxins are present in most South American species examined but are relatively rare in the Old World (Williams and Barneby, 1977b), suggesting that nitrotoxins may be a synapomorphy of the New World radiation plus a few groups in the Old World. Until information can be obtained about which taxa were

not assayed, this character will be coded conserva~ively as

------278 a binary trait, with 0 corresponding to the putative primitive presence of nitrotoxins of the 3-NPOH type (T through 3 on th assay scale), 1 corresponding to the 3-NPA type (4 or 5 on the assay), and unknown ("?") scored for taxa not listed in the table.

56. Selenium accumulation [0 = none; 1 = selenium present].

Selenium is accumulated and concentrated in the tissues of a small number of species (Rosenfeld and Beath, 1964), giving rise to a powerful odor in fresh and dried specimens.

Distribution of this character is taken from Barneby (1964), but was verified by field and herbarium study for the taxa used here.

57. Chromosome number [0 = (2n=22): 1 = (2n=24): 2 =

(2n=26); 3 = (2n=28): 4 = (2n=30); 5 = (2n=32,16)].

Data were taken from Spellenberg (1976) and Barneby

(1964). Spellenberg (1976) suggests that a large non-staining region on one pair of chromosomes (causing one pair to appear as two) may have lead to miscounts of 24 for

22. This is consistent with the haphazard taxonomic

distribution of 22's and 24's in many groups. Surprisingly,

however, a major clade appears to be characterized by an

aneuploid increase from 2n=22 to 24, based on the analysis'

of Chapter Two. I have coded the two states differently.

Spellenberg's counts for the species in the Ocreati clade

are all 2n=24 as well. The numbers 2n=22, and 24 represent 279

82% of the North American species assayed to date

(Spellenberg, 1976). 280

APPENDIX 3

CLADISTIC ANALYSIS OF ASTRAGALUS SECTION ARGOPHYLLI GRAY

Aims -- section Argophylli Gray is one of the two largest taxonomic assemblages currently recognized in North America, comprising some 40 species (Barneby, 1964, 1989). The aim of this analysis is to uncover major groups within this large assemblage so that representative species can be identified for use in a broader phylogenetic study of all the North American taxa (see Chapter Two) . overview of Section -- The center of diversity of section

Argophylli is in the physiographically diverse Great Basin of the western United states. A handful of species extend into the Great Plains, the arid regions of the Northwest, the Mojave desert, and the mountains of central Arizona and

New Mexico. Species of this large and heterogeneous group are marked morphologically by their generally acaulescent, tufted habit of growth, their large flowers, and fleshy, many-ovulate, dorsiventrally-compressed, unilocular fruits.

Barneby has subdivided the section into nine subsections, based mainly on pubescence characters and variations in the shape of the fruit. No phylogenetic analysis of the section has been attempted before now, although Barneby speculated that ~. argophyllus or its close relatives may be ancestral within the section. Although. one approach to sampling this 281 large section for the full cladistic analysis would be to take representatives of each subsection, there is £ priori evidence suggesting that these subsections are highly artificial and deserve to be tested more carefully.

Methods -- Qualitative and quantitative characters were used to score a large sample (300) of specimens from 27 species within the section plus Astragalus nutriosensis Sanderson, a recently described species in the closely related section

Mollissimi. Eleven qualitative traits (22 binary equivalents) and initially 23 quantitative characters were scored (see Table 1 and 2). Gap-coding was used to convert continuous data to discrete codes for cladistic analysis

(Archie, 1985; see Chapter Two). This procedure lead to the elimination of two thirds the quantitative characters because of either lack of significant gaps between species, or generation of only autapomorphic states, which provide no phylogenetic information. In all, the full morphometric data set consisted of about 12000 data points.

Polymorphism was coded as described in Chapter Two.

All character states were coded as unordered although there is good justification for coding the morphometric data as ordered when mUltistate. Mapping such characters onto the final phylogeny provides an interesting test of the results, because the occurence of orderly progression from e.g.,

"smaller" to "larger" without skipping states is satisfying. 282

The data set was analyzed with both PAUP and MACCLADEi the latter found a tree one step shorter after global branch­ swapping based on my own intuition about relationships.

Results -- The shortest tree found to date has 63 steps (w/o polymorphism counted) and a CI of .56. Five major groups were uncovered, three of which correspond fairly closely with Barneby's sUbsections Newberryani, Eriocarpi, and

Missourienses. The monophyly of the first two subsections is not ~urprising since each is characterized by rather unusual synapomorphies. The relationships among these five groups is not particularly clear because of lack of resolution at the base of the tree.

The cohesion of section Missouriensis, based on the presence of medifixed hairs is contrary to Barneby's intuition, which is based on perhaps a subtler assessment of similarity of undefined characters. At this point there is little reason to suspect that this sUbsection is polyphyletic. In particular, several species form a robust clade nested within this section and marked by a peculiar lateral, bicarinate compression of the fruit. other species at the base of this group are somewhat more variable in the position seen in equally parsimonious trees.

The most surpising result is the dismemberment of

Barneby's subsection Argophylli into two major groups. One consists of ~ argophyllus and three or four close relatives 283

(probable peripheral isolates) with ranges in the Great Basin and Colorado Plateau. The other consists of A. tephrodes and relatives with ranges along the Mogollon rim in Arizona (extending barely northwest into Nevada) and throughout much of New Mexico. The latter group also includes species of Barneby's subsection Pseudoargophylli which have anomalous bilocular pods.

Implications for Higher Level Phylogeny Since this analysis is not rooted, it is not possible to identify basal members of clades per se for use in the higher level analysis. Because two groups identified in this analysis,

Newberryani and Eriocarpi, are marked by characters not found in any putatively related sections, it seems likely that they actually do represent monophyletic groups nested within the section. Representative species should then be chosen from the remaining groups. A. argophyllus, and A. tephrodes are obvious choices. The peculiar fruit of A. missouriensis is not so peculiar in the wider context of related sections, and hence this species should be included to test for the possibility that it is basal. These three taxa from section Argophylli will be included in the North

American study. That analysis will simultaneously allow tests of the monophyly of the section and determination of the basal groups. 284

TABLE 1. Characters used in cladistic analysis of section Argophylli.

1. Pod Cross-section: 1 = dorsiventrally compressed; 2=trigonously compressed with dorsal groove; 3=bicarinate and laterally expanded; 4=turgid-inflated and terete.

2. Valve Texture: O=papery, thin; l=ligneous, leathery, fleshy; 2=fleshy-pithy

3. Fruit Persistence: O=deciduous; l=persistent

4. Valve Coloration: o=uniformly colored; l=mottled. 5. Ovule Number: Not used.

6. Flower Color: o=pink-purple; l=white; 2=ochroleucous; 3=scarlet.

7. Fruit Pubescence: O=straight and capillary; 1= flattened and spirally twisted.

8. Herbage Pubescnence: O=Medially medifixed; l=subapically medifixed; 2=basifixed and appressed; 3=basifixed and spreading; 4=basifixed and hirsute­ erect; 5=basifixed but very thin and curly hairs.

9. Leaflet Shape: O=elliptic-acute; l=elliptic-obtuse­ oval; 2=obovate,oblanceolate, and obtuse; 3=obovate­ apiculate.

10. Leaflet Coloration: O=same on both sides; l=bicolored (lighter greener above) .

11. Petiole Persistence: O=decidous; l=persistent­ thickened.

12. Caudex Type: O=branched and diffuse; l=condensed, knotty.

[The following are quantitative characters -- see above] 13. Internode Length. 17. Calyx Length.

14. Petiole Length. 18. Fruit Length.

15. Leaflet Length. 19. Leaflet Number per Leaf. 16. Peduncle Length. 20. Flower Number per Raceme.

------285

TABLE 2. Data matrix for section Argophylli. Polymorphism indicated by multiple adjacent digits. Missing data denoted by question marks.

1 2 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 ------argo 1 1 0 0 ? 0 01 23 023 0 0 0 0 0 0 0 2 0 1 0 zion 1 1 0 1 ? 0 0 23 0 0 0 0 0 0 0 0 2 0 1 0 teph 1 1 0 0 ? 0 0 23 012 1 0 0 0 1 0 1 2 0 2 2 iodo 1 1 0 0 ? 0 ? 3 2 1 0 0 1 1 0 0 2 0 2 1 shor 1 1 0 0 ? 0 0 2 12 0 0 1 0 1 1 0 2 1 1 0 cyan 1 1 0 0 ? 0 0 2 23 1 0 0 1 1 0 1 2 1 2 2 holm 2 1 0 0 ? 0 ? 4 2 1 0 ? ? ? ? ? ? ? ? ? wate 2 1 0 1 ? 0 0 2 ? 1 0 0 ? ? ? ? ? ? ? ? feen 2 1 0 0 ? 0 0 ? 2 1 0 0 1 0 0 0 1 0 1 1 nutr 2 0 0 0 ? 1 0 34 01 1 0 0 ? ? ? ? ? ? ? ? purs 1 1 0 0 ? 02 1 5 0 0 0 0 0 0 0 0 2 0 1 0 utah 1 1 0 0 ? 0 1 5 2 0 0 0 0 0 0 0 2 0 1 0 leuc 2 2 0 0 ? 0 1 5 ? 0 0 0 0 0 0 0 2 0 1 0 infl 1 1 0 0 ? 0 1 5 01 01 0 0 3 0 0 0 2 0 2 1 fune 1 1 0 0 ? 0 1 5 2 0 0 0 ? ? ? ? ? ? ? ? newb 14 1 0 0 ? 02 1 23 23 0 01 1 0 0 0 0 3 0 0 0 cocc 1 1 0 0 ? 3 1 3 23 0 1 1 0 0 0 0 4 1 1 0 unci 1 I 0 0 ? 0 0 23 0 0 1 1 0 0 0 0 3 0 0 0 loan 4 2 0 0 ? 2 1 1 2 0 1 1 0 0 1 0 2 1 0 0 musi 4 1 0 0 ? 2 01 1 ? 0 1 1 0 0 2 0 3 0 0 0 eure 1 1 0 0 ? 2 1 2 0 0 1 1 0 0 0 0 2 0 0 0 pisc 13 2 ? 0 ? 0 0 0 ? 1 0 1 ? ? ? ? ? ? ? ? cham 1 2 0 1 ? 02 0 0 2 0 0 0 0 0 0 0 2 1 1 0 amph 1 1 0 01 ? 0 0 0 01 0 0 0 ? ? ? ? ? ? ? ? cast ? 1 0 0 ? 1 0 0 012 0 0 0 ? ? ? ? ? ? ? ? miss 3 1 1 0 ? 0 0 0 01 0 0 0 0 0 0 0 2 0 1 0 cymb 3 2 0 0 ? 12 0 0 2 0 0 0 ? ? ? ? ? ? ? ? accu 3 1 1 0 ? 02 0 0 03 0 0 0 0 0 0 0 0 0 1 0 ! ~ j ~ i t ~ ~ i J! I ~ i I ~ i I! ~ ~ j ~ lillJllltillilill! lil!lfJil

FIG. A3-1. Cladogram of Astragalus section Argophylli based on data matrix of Table A3-2. Length of tree is 63 steps.

N CXl 0) 287

APPENDIX 4

********** Monte Carlo simulation of topological patterns of homoplasy. Written by Michael J. Sanderson April, 1989 Department of Ecology and Evolutionary Biology University of Arizona, Tucson, 85721 ************ #include #include #define ROOT -1 /* code for root in ancestor arrays */ #define TIP -2 /* code for descendants of tips in desc arrays #define EMPTY -99 /* used for array elements not yet initialized #define NUMLOOPS 500 /* number of iterations of simulation *1 #define NUMINTERVALS 40 /* histogram intervals *1 #define MAXSQ 25 /* size of pwdist array */ #define MAXTAXA 125 1* maximum allowable number of taxa in tree */ #define MAXNODES 249 /* 2*(MAXTAXA)-1*/ #define MAXCHAR 25 /* maximum number of characters working simult.*/ /*----- Type definitions ------*/

typedef int node_array[MAXNODES+l]; FILE *fopen(), *fpointer; char fn[12]; float Idata[NUMLOOPS+l] , 1* stores mean NN distances for each rep */ Xdata[NUMLOOPS+l], 1* stores mean weighted NN distances . .. *1 EDF_charl[MAXTAXA+l] , 1* Empirical Distribution function for specific character *1 mean_EDF[MAXTAXA+l], 1* Mean EDF over replicates *1 stat_mean [MAXSQ], 1* Mean cluster stat for this # of changes*/ cumstatx[MAXSQ] [NUMINTERVALS] ,I*cumulative distribution of statistic*/ cumstatProb[MAXSQ] [NUMINTERVALS];I*for different valuesof numchanges*/ /* Oth element of both arrays is 2 (!) changes, since this is the minimum meaningful number of changes */ int Ipwdist[MAXSQ] [MAXSQ] , 1* Matrix of distances between changes*1 Xpwdist[MAXSQ) [MAXSQ) , 1* Matrix of weighted distances between .. */ NN matrix[NUMLOOPS] [MAXSQ] ,/* Set of NN distances for each reR *1 NN=charl[MAXSQ) , 1* NN distances for specific character */ dmatrix[MAXCHAR) [MAXTAXA) ,I*character data matrix*/ numtaxa, numnodes, basal_node, 1* note basal_node must be assigned correct value prior to calling 'optimize' routines *1 pwdimension, ichar, 1* loop counter for character currently used*/

------'------288 cumflag[MAXSQ]; 1* Tells if this number of changes has already been simulated --i.e., don't repeat!*1 node_array ancestor; long iix; 1* This is the random number seed */

1*----- Fundamental Phylogeny data structure ------*1 struct treestruct (int anc; 1* ancestor of node *1 inc left; 1* left descendent, ... . *1 int right; int setmin; 1* minimum value of Farris state set*/ int setmax; 1* maximum value ~I int plesio; 1* iff a change blow this node, then plesio has plesiomorphic state of ancestor *1 int apo;. 1* . .. and apo has apomorphic state of this node *1 int length; 1* branch length on branch beneath*/ ) tree[MAXNODES+l];

1*----- Function Prototypes ------*1 float test_stat(float a, int b); float do one char stat(int a, int *b, int *c, float *d); float calc cluster(); float rnd{); float dist_comp(); float myabs(float x); void get_parameters(long *a, int *d); 1**********************************************1 maine ) ( 1*------Initializations ------.. * .. int i,j.,loop,numapos,try,numchanges,model,numchrs,num; float cX,Istatl,Xstatl,plevel; get_datamatrix(&numtaxa,&numchrs); printf("Enter name of output file: "); scanf( "%s", fn) ; fpointer-fopen(fn, "a"); get_parameters(&iix, &model); numtaxa-get_tree(); 1* read tree file *1 numnodes - 2*numtaxa-l; 289 for (i-l;i<-numnodes;i++) ( tree[ij .apo-~~PTY; tree[ij .plesio-EMPTY; ) for (i-l;i~~~SQ;i++) { NN_charl[ij-EMPTY; cumflag[i-lj-EMPTY; for (j-l;jl) && (cumflag[num-2] -- EMPTY» /* for homopl. characters that haven't been simulated and tha~ ha'le fewer than MAXSQ changes or apomorphies (array bound), do the following simulation*1 { for (loop-O;loop

._._ ...... - ._ .... ----- 290 Xdata[loop) - calc_cluster(Xpwdist,pwdimension); printf( "%8 .4f %8 .4f\n", Idata[ loop) ,Xdata[ loop) ; ) 1* end of FOR loop *1 histogram(Idata,NUMLOOPS,NUMINTERVALS,num); histogram(Xdata,NUMLOOPS,NUMINTERVALS,num) ; ) 1* end of loop statement when simul hasn't been done for this num */ if (numchanges>l) I*this is a homoplastic characterl */ { if (num>MAXSQ-l) I*error checl:ing*1 { fprintf(fpointer,"%3i (character out of array bounds)\n",ichar+l); printf("%3i (character out of array bounds)\n",ichar+l); ) else 1* normal homoplastic character *1 ( plevel-test stat(Istatl,num); cumflag[num:2)-1; 1* Note that this simul. has been aone";' fprintf(fpointer,"%3i %3i %8.4f %8.4f %6.4f\n", ichar+l,num,Istatl,stat_mean[num-2) ,plevel); printf("%3i %3i %8.4f %8.4f %6.4f\n", ichar+l,num,Istatl,stat_mean[num-2) ,plevel);

else { fprintf(fpointer, "%3i (non-homoplastic)\n", ichar+l); printf("%3i (non-homoplastic)\n",ichar+l); ) I*do EDF(NN matrix, mean EDF); tran;late NN EDF(NN charI,EDF charl); statl-dist_c;mp(EDF=charl,mea~_EDF) ; printf("Fit of point-%8.4f\n",statl);*1 ) I*end chars loop*1 fclose(fpointer) ; ) I*end main*1 1*------*! 1* for (i-l; i

----- _.. __ .. --. 291

1***********************************************************************/ float test_stat(float xstat, int nchanges) /* Determines significance level of xstat for simulation with nchanges */ ( int i; float plevel-l.O; for (i-O;i

return (plevel); ) 1***********************************************************************/ float do_onechar_stateint ichar,int *nchanges, int *numapos, float *Xsta:\ 1* This routine takes one char from the data matrix and determines its two cluster statistics: one returned by pointer, one by function; returns number of changes and apomorphies */

int i,sum-O; float Istat; for (i-O;i

... _--_._------292 nt-l;nc-l; for (;;) ( val-fgetc(fp_in); if (val JEOF) break; if (val - EOL) ( numtaxa-nt-l; nt-l; ++nc; if (nc>MAXCHAR) break; ) else ( val--48; dmatrix[nc-l] (nt-l]-val; ++nt; ) ; } --nc; printf( "Matrix of %3i taxa and %3i characters successfully read .. \r." . numtaxa,nc); *numtanumtaxa; *numc-nc; return )

1************************************************************************ int get_tree() 1* Prompts for and reads a disc file containing the ancestor function 0: a phylogeny. For NT taxa, the file must contain 2*NT-l integers, one ?e~ line. The first number (an element of (1, ... ,NT]) is the taxon ser:ing as the (possibly arbitrary) outgroup ('OG') for the tree. Call the next entries A(l), .... ,A(k) , .... A(2*NT-2). A(k) is the node ancestral to node k, where the first NT nodes are simply the number labels for the terminal taxa. From PAUP linkage list output the node which the OG is attached (call this node 'tempnode') and its length can be determined. PAUP stores trees undirected and hence has one fewer node than my storage format. An additional node is inserted as the ancestor of tempnode and OG. It becomes 'basal_node' in this program. The ancestor of temp_node can be set to any value, (-99 is a good choice) in the disc file, but some value must be present (same goes for its length. :: ~i:: be set to 'basal_node' by the program. The length between OG and tempnode is assigned to the branch beneath tempnode; length beneath OG is zero, as is length beneath basal node.' Function returns the number of taxa in the phylogeny. *; ( int node-l,anc ,lastnode,og,oglength, tempnode; FILE *fopen(), *fp in; char fn(12]; -

--_._------_. --.--- 293 int numtaxa; printf("Enter name of tree file: It); scanf("%s",fn); fp_in-fopen(fn, "r"); fscanf(fp_in,"%3i\n",&og); I*first line in file must have outgroup taxon*1 while (!feof(fp_in» ( fscanf(fp_in, 1%3i\n" ,&cree[node].anc); fscanf( fp_in, "%3i \n" ,&tree [node] . length) ; ++node; ) basal node-node; tempnode-tree[og].anc; oglength-tree[og].length; tree[og].anc-basal_node; tree[og] .length-O; tree[tempnode].anc-basal_node; tree[tempnodej.length-oglength; tree [basal_node] .anc-ROOT; tree [basal_nodej .length-O; numtaxa - (basal_node+l) 12; printf(IITree successfully read: %3i taxa\n",nurntaxa); return (numtaxa); } 1*********************************************************************,\,,\,,', void get_parameters(long * seed, int *model) ( printf(IIInput Model, Seed\n"); printf( "X XXXX\n"); scanf("%li %4i",model,seed); return ; ) 1************************************************'''***************.",**** .' permute_branches(int numchanges,int Ipwdist[] (MAXSQ] , int Xpwdist(] (MAXSQ]) 1* Selects 'numchanges' random nodes (above the changes), and calcula~es path lengths between all pairs. Does not allow changes below basal node. At this time neighboring branches may be chosen -- might consider eliminating this later since couldn't be reconstruc~ed by parsimony*1 ( int rndvect(MAXSQ],i,j ,Ilengthl,Ilength2,Xlengthl,Xlength2: int_list(numnodes-l,numchanges,rndvecc); for (i-l;i<-numchanges;i++) (Xpwdist(i] (i]-EMPTY; Ipwdist(i] [i]-~~PTY; for (j-i+l;j<-nurnchanges';j++) ( if (i--j) (Ipwdist(ij[jj-EMPTY;Xpwdist[ij ~j:-~~PTY;) else ( Ilengthl-path_length(rndvect[i] ,rndvect(j] ,&Xlengchl); Ilength2-path_length(rndvect(i] ,tree(rndvect[j]] .anc,&Xlength2): 294

Ipwdist[i] [j]-min(Ilengthl,Ilengch2); 1* ... see find_distances routine . . ·*1

Ipwdist[j] [i]-Ipwdist[i][j]; Xpwdist[i] [j]-min(Xlengthl,Xlength2); Xpwdist[j] [i]-Xpwdist[i] [j]; I

return; I 1********************************************************************/ do_EDF(int NN_matrix[] [MAXSQ], float mean_EDF[MAXTAXA+l]) ( int i,loop; float EDF_array[MAXTAXA+l], stat[NUMLOOPS]; 1* First, calculate mean EDF for simulations *1 for Ci-l;i<-numtaxa;i++) mean EDF[i]-O.O; for Cloop-O;loop

1* Next, calculate distribution of EDFs according to some statistic function DIST_COMPARE; dist. of stats stored in STAT~: *1 for (loop-O;loop

int i;

. ------.---- 295 float stat-O.O,max-O.O; for (i-l;i<-numtaxa;i++) stat+- l*myabs*1 ( EDF_array[iJ - mean_EDF[i: ): 1* ( stat-myabs(EDF_array[iJ-mean_EDF(iJ); if (stat>max) max-stat; ) ;*1 1* (now cales maximum deviation) *1 return (stat); ) 1*********************************************************************/ translate_NN_EDF(int NN_array[MAXSQ], float EDF_array[MAXTAXA+1J) 1* takes a list of integer nearest neighbor distances, and calculates the empirical distribution function (EDF) from it *1 int i,numpoints; float prob,probincrement; for (i-l;i

1*******************************************************************/ get_NN_array (int NN_array[MAXSQJ, int pd) 1* Sets up an array of length MAXSQ having a list of pd nearest- neighbor distances for a tree, based on the P~DIST array *1

int i,j,imin,temp,isum-O; for (i-l;i

for (j-l;j<- pd;j++) { temp - Ipwdist(ij(jj; if (temp !- ~~PTY) if (temp < imin) imin- temp;

) ; NN_array(ij-imin; ) ; return; ) 1*******************************************************************/ setup_pectinate() 1* sets up a pectinate tree */ ( int halfway,i; halfway-(numnodes+l)/2; ancestor [0 j-EMPTY; ancestor [numnodes j-ROOT; for (i-l;i<-halfway-l;i++) ancestor[ij-numnodes+l-i; for (i-halfway;i<-numnodes-l;i++) ancestor[ij-i+l; return; ) setup_dichot( ) /* sets up symmetrically branching tree */ { int i,intnode-numtaxa+l; ancestor(Oj-EMPTY;ancestor[numnodes]-ROOT; for (i-l;i<-numnodes-l;i+-2) ( ancestor[i]-intnode; ancestor(i+l]-intnode; ++intnode; ) return; ) /********************************************************************, permute_taxa(scruct treestruct tree[] ,int numapos) 1* This routine assigns the apomorphic state to 'nurnapos' terminal taxa chosen randomly. Farris state sets (min and max) are set to 1 :o~ those taxa; all others set to O. */ int i,node; node_array temp, rnd_vector; int_list(numtaxa,numapos,rnd_vector); for (i-l;i<-numtaxa;i++) (tree[ij .setrnin-O; tree[i] .setmax-O;); 297 for (i-1;i<-numapos;i++) ( node-rnd vector[i); tree[n~de] .setmin-1; tree[node].setmax-l; } return; } /**********************************************************************,<, float ca1c_c1uster(int pwdist[) [MAXSQ] , int pd) /* uses a matrix of pairwise distances to calculate some cluster metric -- here a nearest neighbor metric *1 int i,j,imin,temp,isum-O; float x; if (pd <- 1) return (0.0); 1* only one change on tree! */ for (i-1;i<- pd;i++) ( imin - +1000; for (j-1;j<- pd;j++) ( temp - pwdist[i][j]; if (temp !- EMPTY) if (temp < imin) imin- temp;

} ; isum+-imin; } ; x- (float) isum/(pd); 1* gets mean nearest neighbor distance */ return (x); } /********************************************************************,' find_distances(int Ipwdist[) [MAXSQ) ,int Xpwdist[)lMAX£Q], int *pd) /* uses path length routine to find pairwise distances between all changes; creates square matrix pwdist containing these distances; indices of this matrix do not correspond to nodes -- rather they are simply a list of changes, considered sequentially; rank of matrix is returned as pd. */

int i,j,ii,jj ,count-O,Ilengthl,Ilength2,Xlength1,X1ength2,plength, temp [MAXSQ) ; for (i-1;i<-numnodes-l;i++) if (tree[i] .apo !-EMPTY) temp[count++]-i; 1* sets up array temp which is list of nodes where changes are. Array extends from 0 to count-1 th element.*/

for (ii-0;ii<-count-2;ii++) for (jj-ii+1;jj<-count-l;jj++) ( i-temp [ ii) ;

_._._._ .. -.-- ... ------298 j -temp [j j ) ; Ilengthl-path_length(i,j,&Xlengthl); Ilength2-path_length(i,tree[jj.anc,&Xlength2); plength-min(Ilengthl,Ilength2); /* path length between changes must be adjusted for topology, since changes actually occur along internodes; hence examine both nodes surrounding change for one of two changes *1 I*printf( "i ,j ,length: %3i %3i %3i\n", i ,j ,plength) ; printf(" counti, countj : %3i %3i\n", counti, countj ) ; *1 Ipwdist[ii+l) [jj+l)-plength; Ipwdist[jj+l) [ii+l)-plength; 1* make symmetric *1 Xpwdist[ii+l)[jj+l)-~in(Xlengthl,Xlength2); Xpwdist[jj+l) [ii+l]-min(Xlengthl,Xlength2); } *pd-count; /*saves the dimension of the pwdist array*/ return; } 1*********************************************************************/ find_changes(struct treestruct tree[) 1* uses ancestrally optimized states to find nodes below which changes have occurred: apo and plesio states of a node refer to changes along internode BELOW that node; hence BASALNODE changes are undefined; effect is that node above a change will indicate its apomorphic state and its ancestor's plesiomorphic state. If no changes occurred then these values will be EMPTY. Changes to polymorphic are not detected. *1 int node,ap,pl; for (node-l;node<-numnodes-l;node++) ( tree[node) .apo-EMPTY; tree[nodej.plesio-EMPTY; ap-tree[node) .setmin; pl-tree[ tree[nodeJ .anc ] .setmin; if (Cap !- pl) && (ap!-9) && (pl!-9» /*dont count changes to poly!*/ { tree[nodej .apo-ap; tree[node] .plesio-pl; }; I*store only for nodes with changes! */ return; } 1********************************************************************/ optimize(struct treestruct tree[]) ( passl(basal_node,tree); pass2(basal_node,tree); return; } 299

1*******************************************************************1 inc median(inc i1, inc i2,int 13) ( if (i2<-i1) swap (&i1,&i2); if (i3<-i2) swap (&i2,&i3); if (i2<-i1) swap (&il,&i2); return (i2); ) 1*******************************************************************1 swap (int *addrl, int *addr2) ( int temp; temp-*addrl; *addrl-*addr2; *addr2-temp; return; ) 1*****************************************************************/ passl (int node, struct treestruct tree[]) Iprocedure traverses tree in postorder, and sets up Farris state sets/ 1* WARNING, procedure cannot be called with a TIP argument *1 1* This is pass one of Farris optimization. Recursive procedure only visits ancestor nodes -- not tips *1 int ld,rd, slmin,slmax,srmin,srmax,y,z; ld-tree[node].left; rd-tree[node] .right; if (tree[ld] .left !- TIP) 1* ensures that only ancestors-not tips are visited *1 passl(ld, tree); if (tree[rd] .left !- TIP) passl(rd, tree): 1* use statesets of descendants to create sets at node */ slmin-tree[ld] .setmin; slmax-tree[ldJ .setmax: srmin-tree[rd] .setmin; srmax-tree[rdJ .setmax;

1* Following takes care of polymorphism: if one descnedar.: :'s po::':;. ::-"li=:-. make ancestral scaCe secs like thac of the non-poly descendant: if both descendants poly, then make ancestor poly :00. */ if «slmin~9) I I (srmin--9» ( if «slmin--9) && (srmin!-9» ( tree[node] .setmin-srmin; tree[node] .setmax-srmax; ) 300 if «slmin!-9) && (srmin--9) ( tree[node] .setmin-s1min; tree [node] .setmax-s1max; ) if «slmin--9) && (srmin-9)) ( tree [node] .setmin-9; tree [node] .setmax-9; J else 1* following is eqn.(3) of Swofford & Maddison 1987 *1 ( y-median(slmin,srmin,slmax); z-median(slmax,srmin,srmax); tree[node] .setmin-min(y,z); tree[node] .setmax-max(y,z); J return ; J 1*******************************************************************/ pass2 (int node, struct treestruct tree[]) 1* traverses tree in preorder and converts state sets to singletons *1 1* assigns setmin e1em&nt to be the final optimized state for node *: 1* what if basal node is not singleton???? *1 ( int 1d,rd, smin,smax; smin-tree[node] .setmin; smax-tree[node] .setmax; if (smin !- smax) { if (node -- basal_node) tree[node] . setmax-tree [node} .setmin; 1* assign basal state arbitrarily to its lowest possible value *! 1* should make this conform to out group instead ?!!! *1 if (smin--9) ( tree[node] . setmin-tree [tree [node) .anc] .setmin; tree[node] .setmax-tree[nodeJ .setmin; ) 1* if node is polymorphic, just assign its ancestor's values to it */ else ( tree[node] .setmin median(tree[tree[node] .anc: .setmin,smin,smax\: tree[node) .setmax - tree[nodeJ .setmin; J ); 1* .. . otherwise assign median of its state set and ancestor'~,: 301 ld-tree[node] .left; rd-tree[node].right; if (tree[ld].left !- TIP) 1* ensures that only ancestors-not tips are visited *1 pass2 (ld, tree); if (tree[rd].left !- TIP) pass2 (r~, tree); return; ) 1********************************************************************/ descendant_setup (struct treestruct tre,e []) 1* procedure determines the left and right descendants of all nodes and stores them in left_desc and right_desc *1 int node-l,knode,ancnode; for (node-l;node<-numnodes; node++) (tree[node].left-TIP; tree [node] .right-TIP;); for (node-l;node<-numtaxa; node++) ( knode-node; ancnode-tree[knode] .anc; while (ancnode !- ROOT) ( if (tree [ancnode] .left -- TIP) 1* code misleading here */ tree [ancnode] .left - knode; else (tree[ancnode].right - knode; break;); knode-ancnode; ancnode-tree[knode1 .anc; ) ) return ) /**********************************************************************-:', int path length (int nodeA, int no deB , int *xdis~) 1* procedure calculates two kinds of path length distance between nodes A and B on tree given by tree structure function. The number of branches between nodes is returned by the function. The path leng:h weighted by branch length of internodes is returned by pointer */ ( int xa[MAXNODES+l] ,d, z,zdist[MAXNODES+l:; for (d-l; d<-numnodes; d++) \xa[d]--l; zdist[d]--l; 1 d-O; z-O; while (nodeA !- ROOT) ( xa[nodeA]-d; zdist(nooeA]-z; z+-tree[nodeA] . length; nodeA-tree(nodeA] .anc; ++d; ) 302 d-O; z-O; while (nodeB !- ROOT) ( if (xa[nodeB] > -1) ( d+- xa [no deB] ; z+-zdis t [nodeB] ; break; ) z+-tree[nodeB] .length; nodeB-tree[nodeB] .anc; ++d; ) *xdist-z; return (d);

1********************************************************************** /' float rndO 1* procedure returns a real random value on [0,1]. The variable iix must be declared globally as a long integer by the main program and must be initialized as a seed integer *1 ( long iiy; float yfl; iiy-iix*65539 ; if (iiy < 0) iiy-iiy+2l47483647+l; yfl-(float) iiy; yfl-yfl*-.4656613E-9; iix-iiy; return(yfl) ; ) 1***************************************************** .):***********.:.***,'~ ... int list(int n, int k, int rnd vector[]) 1* procedure finds k random integers out of the set [1 .... n}. and retur~s them in vector as rnd_vector[lj ... [k] */

node_array temp; int i,d,rnum,vcount; for (i-l;i<-n;i++) temp[i]-i; for (vcount-l;vcounc<-k;vcounc++) ( rnum-n*rnd()+1 ; I*gets random number on : 1. .... n: */ rnd_vector[vcount]-temp[rnum];

for (d-rnum+l;d<-n;d++) temp[d-l]-temp(d:; /*delece element and move down*1 - on; ) ;

return; 303

/************************************************************************1 histogram (float datal] ,int num_data,int num_intervals,int nchanges) ( float max,min,mean,variance , cum histo[SO] ,prob,probincrement,xinc,xval; int i,index,histo[SO]; /*maxi;i;um number of intervals for histogram */ for (i-l;i<-num_intervals;i++) histo[i]-O; array_bounds(data,num_data,&max,&min,&mean,&variance); xinc-(max-min)/num_intervals; for (i-O;inum_intervals) --index;/* corrects for missing last interv*/ ++his to [ index] ; I /*for (i-l;i<-num_intervals;i++) printf("%3i %3i\n",i,histo[i]);*1 prob-O.O; probincrement-(float) l/num_data; xval-min; /*fprintf(fpointer,"Distribution of Goodness of Fit Values\n"); fprintf(fpointer," Fit ~ts. Prob<-\n");*1 for (i-l;i<-num_intervals;i++) ( if (histo[i] !- 0) prob+-probincrement*histo[i}; cum_histo[i]-prob; xval+-xinc; cumstatx[nchanges-2] [i-l]-xval; 1* for ichar character store values of*1 cumstatProb[nchanges-2] [i-l]-prob;/* stat and corresponding P level * ... /* fprintf( fpointer, "%6. 4f %4i %8. 4f\n" ,xval, histo [i], clUn_histo ~ i:) ; *.' /* this table reports cumulative prob that fit is <- xval *1 I stat_mean[nchanges-2]-mean; /*fprintf(fpointer, "mean- %6.3f, variance- %8.3!\n",mean,variance);*/

return; I /*********************************************************************:' array_bounds(float datal] ,int nwn_data, float *max, float *mi~. float *mean, float *variance) /* finds min and max elements in a float array and simple stats */ ( int i; float tmin-100DO.O,tmax--10000.O,tmean-O,tvariance-O; for (i-O;itmax) tmax-data(i}; tmean+-data( i} ;

------.------.-.------. 304 tvariance+-data[il*data[iJ; ) *mean-tmean/num data; *Variance-(tvarIance-(tmean*tmean)/num_data)/(num_data-l); *max-tmax; *min-tmin; return; ) float myabs(float x) ( if (x

APPENDIX 5: FRUIT HEASUREHENTS ON ASTRAGALUS SPECIES (in log millimeters, sample size of 5)

Species Length Std. Dev. Width Std. Dev. --.. ---- .. --- .. - .' - .. --- .... - .. --- .... -- - ...... - .. - - .. - -- .. - .. - ...... 1 aborigi 2.49473 0.350175 1.41802 0.210544 2 acutiro 2.86093 0.058908 1. 09861 0.000000 3 adanus 2.56495 0.000000 1. 36478 0.256092 4 aequali 3.09001 0.064320 2.67355 0.048779 5 albulis 2.56258 O. 077087 1.35231 0.284671 6 alpinus 2.42297 0.107184 1. 07898 0.221292 7 alvord 2.98627 0.166099 0.96346 0.234095 8 ambylo 2.99573 0.000000 2.56495 0.000000 9 ampulla 2.63650 0.101177 2.19722 0.000000 10 anderso 2.71970 0.115132 1. 38629 0.000000 11 anemoph 3.03765 0.143966 2.54671 0.235532 12 aridus 2.58751 0.031884 1.70060 0.128921 13 arizoni 2.99981 0.169317 1.14656 0.117446 14 arrectu 2.85851 0.121518 1.24245 0.203421 15 arthuri* 3.46343 0.165544 1.17053 0.143841 16 arthuri* 3.68888 0.000000 1.09861 0.000000 17 asclepi 3.21594 0.216324 2.39513 0.091192 18 asymmet 3.73767 0.000000 3.21888 0.000000 19 atratus* 2.79867 0.182731 1. 23170 0.217915 20 atratus* 3.09104 0.000000 1.60944 0.000000 21 austin 1.60944 0.000000 0.69315 0.000000 22 beathii 3.27392 0.131143 2.04231 0.174246 23 beckwit 3. 096 70 0.178021 1. 87037 0.204859 24 bicrist* 3.25810 0.000000 1. 79176 0.000000 25 bicrist* 3.23472 0.140574 1. 67420 0.179266 26 bisulca 2.45642 0.182165 0.77326 0.235788 27 bolander* 2.72004 0.352843 1.98742 0.25':'951 28 bolander* 2.45264 0.361211 1.79176 O.OOOOC:) 29 braunt 1. 94076 0.117595 1.09861 0.000000 30 brazoen 1.43508 0.000000 0.91629 0.000000 31 bryanti 3.39897 0.094439 1.49787 0.157785 32 calif 3.48174 0.190507 1.37554 0.162394- 33 calycos 2.42965 0.099401 1.15615 0.128655 34 camp top 3.14715 0.381833 1.46067 0.128831 35 canaden 2.42340 0.160269 1.27122 0.157570 36 carie 1.78049 0.168355 0.63561 0.39632S 37 casei 3.44877 0.156307 1. 48766 0.202733 38 ceramic 3. 08972 0.146952 2.42297 0.326283 39 chloo 1.96203 0.310985 0.69315 0.00024':' 40 cibariu 3.27052 0.319597 2.04717 0.361207 41 cimae 3.21888 0.000000 2.56495 0.000000 42 clevel 1.56481 0.099792 0.27726 0.379652 43 cobrens 2.32239 0.000000 1.25276 0.000000 44 collin 2.94104 0.190575 1.09861 0.000000 ",,-

Species Length Std. Dev. Width Std. Dev. ------_.------45 coltoni 3.12977 0.131281 1. 53506 0.128832 46 congdon 2.99573 0.000000 0.40547 0.000000 47 conjunc* 2.87082 0.144340 l. 41802 0.210544 48 conjunc* 2.94444 0.000000 l. 60944 0.000000 49 conva11 3.24153 0.260103 l. 17053 0.143841 50 coriaCE 2.81739 0.252209 1.70060 0.128921 51 crota1a 3.16172 0.141342 2.36089 0.218317 52 cusicki 3.41642 0.210610 2.56258 0.217537 53 despera 2.38867 0.272339 l. 39694 0.307875 54 detrit 3.31743 0.092056 0.96346 0.234095 55 distort 2.68453 0.136519 l. 19687 0.395595 56 douglas 3.37827 0.290327 2.87409 0.305025 57 drummon 3.13549 0.000000 l. 09861 0.000000 58 duchesn 3.29584 0.000000 l. 09861 0.000000 59 eastwoo 2.83148 0.083300 l. 93560 0.203421 60 egg1est 3.04925 0.109762 l. 21368 0.157570 61 emoryan 2.64413 0.348232 0.89588 0.286707 62 ensifor 3.36730 0.000000 l. 94591 0.000000 63 episcop* 3.15739 0.125015 l.65191 0.231715 64 epsicop* 3.25810 0.000000 1.60944 0.000000 65 eremeti 2.97382 0.189265 l. 69198 0.172475 66 ervoid 2.30259 0.000000 0.69315 0.000000 67 esperan 2.39790 0.000000 l.09861 0.000000 68 eucosmu 2.07944 0.000000 l.38629 0.000000 69 fastidi 3.28165 0.190360 2.48046 0.107186 70 filipes 3.11723 0.162408 l.63302 0.150292 71 flavus 2.16354 0.167049 0.85533 0.222083 72 flexuos 2.58144 0.189053 l.15615 0.128655 73 fucatus 3.06384 0.100369 2.48933 0.154416 74 gambell l. 10806 0.069758 0.70554 0.110306 75 gentryi 2.55628 0.261999 0.89588 0.234096 76 geyeri 2.56495 0.000000 l. 78237 0.168434 77 gigant::e 2.65391 0.086494 l. 89518 0.200623 78 gilensi 1.70201 0.185303 l. 06202 0.155920 79 gracilu 1.97929 0.066769 l.09861 0.000000 80 guatama 2.58597 0.113041 l. 19451 0.166092 81 hallii 2.86782 0.057779 l. 91242 0.348101 82 hart',oleg 2.17552 0.232083 0.57762 0.282976 83 hida1ge 2.83321 0.000000 l. 38629 0.000000 84 ho1mgre 3.72212 0.119556 2.07157 0.1;770; 85 hornii 2.42690 0.050234 2.07419 0.125738 86 howell 3.03897 0.130600 l. 29040 0.166092 87 hurnistr 2.08303 0.168045 1.53855 0.283498 88 hypogl0 1.99965 0.247629 1.01931 0.203422 89 hypoleuc~~ 2.13833 0.232286 0.92815 0.332343 90 hypo1euc* 2.39790 0.000000 0.69315 0.000000 91 hypoxy1 2.11626 0.000000 0.99325 0.000000 92 insular 2.75303 0.059632 2.29853 0.096025 3(; ,-

Species Length Std. Dev. width S cd. Dev. -_ ... --- -_ .. -- ---_ .. -- -_ ...... ------_ ... -- ... --_ ... -- ---_ ...... __ ..... -_ ... 93 inyoe 2.28898 0.203076 1.36478 0.256090 94 iodanth 2.99817 0.241751 1.41574 0.285012 95 kaibens 3.15677 0.030100 1.70060 0.128921 96 kentrop 1.63672 0.335456 0.61365 0.132632 97 knighti 2.35024 0.067390 1.70060 0.128921 98 1ancear 3.13845 0.268804 1.73564 0.130027 99 1ayneae 3.46359 0.190003 1. 56481 0.099792 100 1eptoca 3.17927 0.156303 0.89588 0.234095 101 1eucoph 3.34231 0.174562 2.47082 0.237919 102 lindhei 3.16018 0.141203 1. 52731 0.298443 103 lonchoc 3.60692 0.221867 1. 46264 0.225274 104 longiss 2.91447 0.114919 1.49787 0.157785 105 lotiflo 3.02013 0.034471 1.58903 0.693551 106 1yall 1.88425 0.084432 0.77424 0.181330 107 ma1acus 3.55535 0.000000 1. 79176 0.000000 108 micrant 2.33977 0.209087 1.17151 0.189181 109 minthor 2.93762 0.274160 1.73099 0.105263 110 mise11 2.54671 0.235532 1.09861 0.000000 111 miser 2.86267 0.190940 0.98233 0.296109 112 mojaven 2.97532 0.286704 2.01268 0.094424 113 monoen 2.70805 0.000000 1.38629 0.000000 114 nigresc 1.11501 0.023192 0.61482 0.038232 115 nothoxy 2.85791 0.157593 1.10985 0.148820 116 obscuri 2.88416 0.157781 l. 41425 0.188536 117 oophoru 3.29087 0.110166 2.64706 0.123631 118 oreganu 2.48491 0.000000 1.38629 0.000000 119 oxyphys 3.41484 0.286534 2.38527 0.275828 120 pachypu 2.96154 0.029653 1.70799 0.289061 121 palmeri 2.57555 0.350912 l.67021 0.105263 122 peck 1.49787 0.128831 0.79451 0.202733 123 pectin 2.53770 0.321250 1. 50887 0.392991 124 p1atytr* 2.81719 0.056540 2.36818 0.131182 125 p1atytr* 3.08811 0.087172 2.76865 0.102705 126 prae10n 3.14933 0.158183 2.43319 0.255804 127 preussi 2.86505 0.199105 2.18382 0.246':'63 128 pro rife 2.76734 0.125746 l.83375 0.194262 129 pteroca 3.73767 0.000000 1.79176 0.000000 130 pu1sifer* 2.35811 0.227958 1. 87449 0.179794 131 pu1sifer* 2.53028 0.159977 1. 93904 0.143964 132 punic 2.77063 0.088502 l.79176 0.000000 133 pycnost 2.07944 0.000000 1.38629 0.000000 134 racemos 3.16387 0.024570 1.12971 0.191646 135 rattani 3.55535 0.000000 0.83291 0.000000 136 recurvu 2.76977 0.085971 1.11993 0.095657 137 remotus 2.98627 0.166099 1.09526 0.100376 138 revent 2.81177 0.172183 1. 38441 0.320078 139 robinsi 2.77259 0.000000 l.48161 0.000000 140 rusbyi 2.79949 0.154923 1.11385 0.166528 3 (:;,

Species Length Std. Dev. width Std. Dev. ------_ ... -- -_ .. ------_ ..... ------_ .. 141 sabulon 2.57637 0.097637 1. 59031 0.133810 142 salmon 2.74264 0.164770 1.43916 0.294926 143 scalari 2.00148 0.000000 1.16315 0.000000 144 sclero 3.27532 0.178487 2.01876 0.154961 145 seatoni 2.22361 0.379688 1.20948 a . 13t.. 873 146 sepultip* 2.42181 0.132802 1.19451 0.166092 147 sepultip* 2.66723 0.000000 1.66771 0.000000 148 serenoi 3.13549 0.000000 2.39790 0.000000 149 sinuat 2.86118 0.129050 1.15615 0.128655 150 spathulat* l. 91508 0.068939 0.69315 0.000000 151 spathulat* l. 62441 0.139096 0.56246 0.319602 152 speiro 3.30678 0.168767 1.36048 0.255820 153 sterilu 3.09104 0.000000 2.39790 0.000000 154 striati 2.30259 0.000000 1. 94591 0.000000 155 strigu1 2.57839 0.070104 1.39256 0.258212 156 tenellu 2.35264 0.210696 1.03859 0.256121 157 tener 2.63906 0.000000 0.83291 0.000000 158 toanus 2.76406 0.175960 1.54620 0.063928 159 trask 2.67355 0.048779 1.49787 0.157785 160 trog1od 1.60944 0.000000 1. 38629 0.000000 161 umbrati 3.03997 0.135098 1. 28093 0.000000 162 whitney 3.18520 0.053480 2.60169 0.145560 163 wingata 2.46723 0.140136 1. 37604 0.022934 164 woo toni 3.27306 0.097701 2.56663 0.135919 165 yukonis 2.30259 0.000000 1.22378 0.000000

------.. ------...... ------* Replicates of same species

------309

APPENDIX 6

Taxa of Astragalus described since Barneby (1964) (from the Gray Herbarium Index and other sources)

~ aboriginum vars. Boivin Phytologia 15:383 1967 ~ ackermannii Barneby Brittonia 32:26 1980 ~ adsurgens f. chandonnetii (Lunnel) Boivin Phytologia 15:384 ~ agrestis f. virgultulus (Sheld.) Scoggan Nat. Canada 87:28 1960 ------(Sheld.) Boivin Flora Canada 52 1978 ~ allochrous var. playanus (Jones) Isely Syst. Bot. 8:420 1983 ~ alpinus f. alpinus (Rousseau) Boivin Nat. Canada 94:652 1967 ~ anserinus Atwood, Goodrich & Welsh Gr. Bas. Nat. 44:263 1984 ~ astragalinus (Hook.) Love & Love Bot. Notis. 128:515 1976 ~ atwoodii Welsh & Thorne Gr. Bas. Nat. 37:103 1977 ~ australis vars. Isely Syst. Bot. 8:421 1983 ~ austraL~~ var. qlabriscutus (Gray) Isely Iowa st. J. Res. 59:130 1984 ~ barnebyi Welsh & Atwood Great Basin Nat. 35:346 1975 ~ beatleyi Barneby Aliso 7:161 1970 ~ bibul1atus Barneby & Bridges Brittonia 39:358 1987 ~ bigelovii var. marcidus (Rydb.) Isely Syst. Bot. 8:421 1983 ~ bisulcatus f. albiflorus Boivin Phytologia 15:381 1967 ~ bisulcatus ssp. haydenianus (Brandg.) W. Weber Phytologia 51:369 1982 ~ bisulcatus var. major (Jones) Welsh Great Basin Nat. 38:266 1978 ~ bodinii var. yukonis (Jones) Boivin Phytologia 15:377 1967 ~ bryogenes Barneby Brittonia 36:171 1984 ~ cenorrhynchus Barneby Brittonia 34:78 1982 ~chuskanus Barneby & Spellenberg ~ consobrinus (Barneby) Welsh Great Basin Nat. 38:271 1978 ~ convallarius var. rnargaretae Barneby Brittonia 36:167 1984 ~ cottamii Welsh Rhodora 72:189 1970 ~ cremnophylax var. rnyriorrhaphis Barneby Brittonia 31:463 1979 ~ daleae Ere-recognized] Sida 13:251 1988 ~ danicus var. dasyglottis (Fisch.) Boivin Phytologia 15:384 1967 ~ debequaeus Welsh Great Basin Nat. 45:31 1985

------310

~ didymocarpus ssp. Milesianus (Rydb.) Hoover Leafl. West. Bot 10:348 1966 A . ., episcopus var. lancerius (Gray) Isely Syst. Bot. 8:421 1983 ~ eguisolensis Neese & Welsh Rhodora 83:457 1981 ~ ertterae Barneby & Shevock Aliso 11:585 1987 ~ eucosmus vars. Boivin Nat. Canada 94:521 1967 Phytologia 15:382 1967 ------Hulten Ark. Bot. 7:77 1968 ~ eurylobus (Barneby) Barneby Brittonia 36:167 1984 ~ hartmanii Rydberg Ere-recognized] Sida 13:251 1988 ~ henrimontanus Welsh Great Basin Nat. 38:281 1978 [=A. stocksii Welsh see below] ~ holmgreniorum Barneby Brittonia 32:24-29 1980 A. howellii var. pauper (Barneby) Isely Syst. Bot. 8:422 1983 ~ iselyi Welsh Great Basin Nat. 34:305 1974 ~ kentrophyta ssp. coloradoensis W.A.Weber Phytologia 53:187 1983 ~ kentrophyta ssp. danicus W.A.Weber Phytologia 53:187 1983 ~ knightii Barneby Brittonia 35:109 1983 ~ lentiginosus var. albifloru8 (Gray) Schoener Great Basin Nat. 34:180 1974 ~ lentiginosus var. higginsii Welsh & Thorne Brittonia 33:296 1981 ~ lentiginosus var. piscinensis Barneby Brittonia 29:378 1977 ~ lentiginosus var. pohlii Welsh & Barneby Iselya 2:1 1981 ~ lentiginosus var. wahweapensis Welsh Great Basin Nat. 38:286 1978 ~ moranii Barneby Brittonia 28:278 1976 ~ nutriosensis Sanderson Madrono 35:325 1988 ~ oophorus var. lavinii Barneby Brittonia 36:168 1984 oxyphysopsis Barneby Brittonia 28:273 1976 phoenix Barneby Madrono 20:395 1971 piscator piutensis Barneby & Mabersly Taxon 34:453 1985 preussii var. peabodianus (Jones) Welsh Great Basin Nat.38:297 1978 ~ robbinsii subsp. Harringtonii (Rydb.) Hulten Ark. Bot. 7:78 1968 ~ sanctorum Barneby Brittonia 28:275 1976 ~ serenoi var. shockleyi (Jones) Barneby Brittonia 36:168 1984 ~ shevockii Barneby Brittonia 29:376 1977

------311

~ shultziorum Barneby Brittonia 33:156 1981 ~ stocksii Welsh Great Bas. Nat. 34:307 1974

~ subcinereus var. basalticus Welsh Great Bas. Nat. 38:302 1978 ~ tiehmii Barneby Brittonia 35:169 1984 ~ wittmannii Barneby Brittonia 31:459 1979 ~ yoder-williamsii Barneby Brittonia 32:30 1980 312

LITERATURE CITED

Agnew, A.D.Q. 1974. Upland Kenya wildflowers: A Flora of the Ferns and Herbaceous Flowering Plants of Upland Kenya. Oxford University Press, Oxford.

Airy Shaw, H.K. 1973. A Dictionary of the Flowering Plants and Ferns. Cambridge University press, Cambridge. 8th ed.

Alberch, P. 1981. Convergence and parallelism in foot morphology in the neotropical salamander genus Bolitoglossa. I. Function. Evolution 35: 84-100.

Arber, A. 1925. Monocotyledons: A Morphological Study. Cambridge University Press, cambridge.

Archie, J. 1985. Methods for coding variable morphological features for numerical taxonomic analyses. Syst. Zool. 34:326-345.

Arnold, E.N. 1981. Estimating phylogenies at low taxonomic levels. Zeit. Zool. Syst. Evol. Forsch. 19:1-35.

Barneby, R.C. 1952. A revision of the North American species of Oxytropis DC. Proc. Calif. Acad. Sci. IV, 27:177-312.

------. 1964. Atlas of North American Astragalus. Memoirs New York Bot. Gard. 13:1-1188.

1989. Astragalus In "Intermountain Flora." In press.

Bartlett, M.S. 1975. The Statistical Analysis of spatial Pattern. Chapman & Hall, London.

Bock, W.J. 1963. Evolution and phylogeny in morphologically uniform groups. Am. Nat. 97:265-285.

------. 1977. Foundations and methods of evolutionary classification, pp. 851-~95. In M.K. Hecht, P.C. Goody, and B.M. Hecht (eds.), Major Patterns in Vertebrate Evolution. Plenum, New York. History of homology

Boots, B.N., and A. Getis. 1988. Point Pattern Analysis. Sage Publications, Newbury Park, California.

Bowler, P.J. 1979. Theodore Eimer and orthogenesis: Evolution by "definitely directed variation." J. Hist. Med. Allied Sci. 34:40-73. 313

------1984. Evolution: The History of an Idea. U·niversity of California Press, Berkeley. Boyden, A. 1943. Homology and analogy: A century after the definitions of "homologue" and "analogue" of Richard Owen. Quart. Rev. Biol. 18:228-241. Brown, H. 1977. Perception, Theory, and Commitment. Universi ty of Chicago P::cess, chicago. Brown, J;H., and A.C. Gibson. 1983. Biogeography. C.V. Mosby, st. Louis. Brundin, L. 1972. Evolution, causal biology, and classification. Zool. Scr. 1:107-120. ------. 1976. A neocomian chironomid and Podonominae­ Aphroteniinae (Diptera) in the light of phylogenetics and biogeography. Zool. Scr. 5:139-160. Buckman, S.S. 1901. Homoeomorphy among Jurassic Brachiopoda. Cotteswold Natur. Field Club 13:231-290. Bunge, A. von. 1868. Generis Astragali Species Gerontogeae. Pars Prior. Claves Diagnosticae. Mem. Acad. Imp. Sci. st. Petersb., sere VII, 11 (16): 1-140. Burtt, B.L. 1964. Angiosperm taxonomy in practice. Syst. Assoc. Publ. 6: 5-16. cain, A.J. 1982. On homology and convergence, pp. 1-19. In K. Joysey and A. Friday (eds.), Problems of Phylogenetic Reconstruction. Academic Press, London. cantino, P. 1982. Affinities of the Lamiales. Syst. Bot. 7: 237-248. cantino, P. 1985. Phylogenetic inference from non-universal derived character states. Syst. Bot. 11: 119-122. Cavender, J.A. 1978. Taxonomy with confidence. Math. Biosci. 40:271-280. ------. 1981. Tests of phylogenetic hypotheses under generalized models. Math. Biosci. 54:217-229. Chamberlain, D.F., and ·V.A. Matthews. 1970. Astragalus, pp. 49-254. In P.H. Davis (ed.), Flora of Turkey and the East Aegean Islands. Edinburgh University Press, Edinburgh. 314

Charlesworth, B, R. Lande, and M. Slatkin. 1982. A neo­ Darwinian commentary on macroevolution. Evolution 36: 474- 498. Clark, P.J. and F.C. Evans. 1954. Distance to nearest neighbor as a measure of spatial relationships in populations. Ecology 35:445-453. Coddington, J.A. 1988. Cladistic tests of adaptationa1 hypotheses. Cladistics 4:3-22. Cope, E.D. 1877. Report upon the extinct vertebrata obtained in New Mexico by parties of the expedition of 1874. Rep. U.S. Geogr. Surv. west of the 100th meridian 4, Paleontology, Pt. II. cronquist, A. 1963. The taxonomic significance of evolutionary parallelism. Sida 1: 109-116. -----. 1968. The Evolution and Classification of Flowering Plants. Houghton Mifflin Co., Boston. Dahlgren, R. 1970. Current topics -- parallelism, convergence, and analogy in some South African genera of Leguminosae. Bot. Not. 123: 552-568. Darwin, C. 1872. The origin of Species. Modern Library, New York. 6th ed. de Queiroz, K., and M.J. Donoghue. Phylogenetic systematics and the species problem. Cladistics 4:317-338. DeSalle, R., and A.R. Templeton. 1988. Founder effects and the rate of mitochondrial DNA evolution in Hawaiian Drosophila. Evolution 42:1076-1084. Diggle, P.J. 1979. Statistical methods for spatial point patterns in ecology, pp. 95-150. In R.M. Cormack and J.K. Ord (eds.), Spatial and Temporal Analysis in Ecology. International Cooperative Publishing House, Fairland, Maryland. Donoghue, M.J. Phylogenies and the analysis of evolutionary sequences, with examples from seed plants. Evolution, in press.

Donoghue, M.J., and P.D. cantino. 1984. The logic and limitations of the outgroup sUbstitution approach to cladistic analysis. Syst. Bot. 9:192-202. 315

Donoghue, M.J. J. Doyle, J. Gauthier, A. Kluge, and T. Rowe. The importance of fossils phylogeny reconstruction. Ms.

Dormer, K.J. 1945. An investigation of the taxonomic value of shoot structure in angiosperms with especial reference to Leguminosae. Annals of Botany, London, n.s. 9:141-153.

-----. 194~ vegetative morphology as a guide to the classification of Papilonatae. New Phytol. 45:145-161.

Doyle, J.A., and M.J. Donoghue. 1987. The importance of fossils in elucidating seed plant phylogeny and macroevolution. Rev. Paleobot. Palyn. 50:63-95.

Eig, A. 1955. Systematic studies on Astragalus of the Near East. pp. 1-187.

Eimer, G.H.T. 1898. On orthogenesis and the impotence of natural selection in species formation. The Open Court Publishing Co., Chicago.

Eldredge, N., and J. Cracraft. 1980. Phylogenetic Patterns and the Evolutionary Process: Method and Theory in comparative Biology. Columbia University Press, New York.

Etheridge, and K. de Queiroz. 1988. A phylogeny of the Iguanidae, pp. 283-367. In R. Estes and G. Pregill (eds.), Phylogenetic Relationships of the Lizard Families. Stanford University Press, Stanford.

Farris, J.S. 1981. Discussion [in response to Simberloff, et. al.], pp. 73-84. In G. Nelson and D.E. Rosen (eds.), Vicariance Biogeography: A Critique. Columbia University Press, New York.

Felsenstein, J. 1985a. Confidence limits on phylogenies: An approach using the bootstrap. Evolution 39: 783-791.

------. 1985b. Phylogenies and the comparative method. Amer. Nat. 125:1-15.

------. 1988. Phylogenies and quantitative characters. Ann. Rev. Ecol. Syst. 19:445-471.

Foster, A.S., and E.M. Gifford, Jr. 1974. Comparative Morphology of Vascular Plants. W.H. Freeman, San Francisco.

Futuyma, D.J. 1986. Evolutionary Biology. Sinauer Associates, Sunderland, Mass. 2nd. ed .

....------.. - .. _. __ . ------316

Gauthier, J. R. Estes, and K. de Queiroz. 1988. A phylogenetic analysis of Lepidosauromorpha, pp. 15-98. In R. Estes and G. Pregill (eds.), Phylogenetic Relationships of the Lizard Families. Stanford University Press, Stanford, California.

Gauthier, J. A. Kluge, and T. Rowe. 1989. Amniote phylogeny and the importance of fossils. Cladistics 4:105-209.

Goldman, N. 1988. Methods for discrete coding of morphological characters for numerical analysis. Cladistics 4:59-71.

Gomez-Sosa, E. 1979. Las especies sudamericanas del genero Astragalus (Leguminosae) I. Las especies patagonicas argentinas. Darwiniana 22:313-376.

Gontscharov, N. 1965. Astragalus, pp. 1-918. In V.L. Komarov (ed.), Flora of the U.S.S.R., vol. 12, Leguminosae: Astragalus. Transl. ed. published by the smithsonian and NSF, US Dept. of Commerce.

Gosliner, T.M., and M.T. Ghiselin. 1984. Parallel evolution in opisthobranch gastropods and its implications for phylogenetic methodology. Syst. Zool. 33: 255-274

Gray, A. 1864. A revision and arrangement (mainly by the fruit) of the North American species of Astragalus and oxytropis. Proc. Amer. Acad. 6:188-236.

Greene, H.W. 1986. Diet and arboreality in the emerald moditor, Varanus prasinus, with comments on the study of adaptation. Fieldiana, Zool. (NS) 31:1-12.

Green, T.W., and G.E. Bohart. 1975. The pollination ecology of Astragalus cibarius and Astragalus utahensis (Leguminosae). Amer. J. Bot. 62:379-386.

Grehan, J.R., and R. Ainsworth. 1985. orthogenesis and evolution. Syst. Zool. 34: 174-192.

Guise, A., D. Peacock, and T. Gleaves. 1982. A method for identification of parallelism in discrete character sets. zool. J. Linn. Soc. 74: 293-303.

Haacke, W. 1893. Gestaltung und Vererbung. O.W. Nachfolger, Leipzig. 317

Haas, 0., and G.G. Simpson. 1946. Analysis of some phylogenetic terms, with attempts at redefinition. Proc. Amer. Phil. Soc. 90: 319-349. Hennig, W. 1966. Phylogenetic Systematics. University of Illinois Press, Urbana, Ill. Isely, D. 1933a. Astragalus L. (Leguminosae: Papilionoideae) I.: Keys to United States Species. Iowa st. Jour. Res. 58:1- 172. ------" 1983b. New combinations and two new varieties in Astragalus, Orophaca. and oxytropis (Leguminosae). syst. Bot. 8:420-426. ------. 1984. Astragalus L. (Leguminosae: Papilionoideae) II.: Species summary A-E. Iowa st. Jour. Res. 59:97-216. ------. 1985. Leguminosae of the United states. Astragalus L. species summary F-M. Iowa st. Jour. Res. 60:183-320. ------. 1986. Leguminosae of the United states. Astragalus L. species summary N-Z. Iowa st. Jour. Res. 61:157-289. Jardine,N, and R. Sibson. 1971. Mathematical Taxonomy. John Wiley & Sons, London. Johnston, I.M. 1938. Notes on sone Astragalus species of Ecuador and Peru. Jour. Arnold Arbor. 19:88-96. ------. 1947. Astragalus in Argentina, Bolivia and Chile. Jour. Arn. Arb. 28:336-409. Jones, M.C. 1923. Revision of North American Astragalus. Published by the author, Salt Lake City. Kaplan, D.R. 1984. The concept of homology and its central role in the elucidation of plant systematic relationships, pp. 51-69. In T. Duncan and T. stuessy (eds.), Cladistics: Perspectives on the reconstruction of evolutionary history. Columbia University Press, New York. Karron, J.D. 1987a. A comparison of levels of genetic polymorphism and self-compatibility in geographically restricted and widespread plant congeners. Evolutionary Ecology 1:47-58. ------. 1987b. The pollination ecology of co-occurrihg geographically restricted and widespread species of Astragalus (Fabaceae). BioI. Conservation 39:179-193.

--.----.-.------....------_._-_._------318

------. 1939. Breeding systems and levels of inbreeding depression in geographically restricted and widespread species of Astragalus (Fabaceae). Amer. J. Bot. 331-340. Karron, J.D., Y.B. Linhart, C.A. Chalk, and C.A. Robertson. 1988. The genetic structure of populations of geographically restricted and widespread species of Astragalus (Fabaceae). Amer. J. Bot. 75:1114-1119. Kluge, A.G., and J.S. Farris. 1969. Quantitative phyletics and the evolution of Anurans. Syst. Zool. 18: 1-32. Lankester, E.R. 1870. On the use of the term homology in modern zoology. Annals and Magazine of Natural History, series 4, 6: 34-43. Lavin, M., J.J. Doyle, and J.D. Palmer. Evolutionary significance of the loss of the chloroplast DNA inverted repeat in the Leguminosae subfamily Papilionoideae. Ms. Ledingham, G.F., and B.M. Pepper. 1973. Chromosome numbers of some South American species of Astragalus. Kurtziana 7:27-37. Ludwig, J.A. 1979. A test of different quadrat variance methods for the analysis of spatial pattern, pp. 289-304. In R.M. Cormack and J.K. Ord (eds.), spatial and Temporal Analysis in Ecology. International Cooperative Publishing House, Fairland, Maryland. Maddison, W.A. A method for testing the correlated evolution of two characters. Ms. Maddison, W., M.J. Donoghue, and D.R. Maddison. 1984. Outgroup analysis and parsimony. Syst. Zool. 33:83-103. Maddison, W., and D.R. Maddison. 1987. MacClade Users Manual, version #2.1. Maynard smith; J., R. Burian, S. Kauffman, P. Alberch, J. Campbell, B. Goodwin, R. Lande, D. Raup, and L. Wolpert. 1985. Developmental constraints and evolution. Quart. Rev. BioI. 60: 265-287. Mayr, E. 1969. Principles of Systematic Zoology. McGraw­ Hill, New York. Meacham, C.A. 1981. A probability measure for character compatibility. Math. Biosci. 57:1-18. 319

Meinke, R.J. 1982. Threatened and endangered vascular plants of Oregon. u.s. Dept. Interior. Mindell, D.P, J.W. Sites, and D. Graur. 1989. Speciational evolution: A phylogenetic test with allozymes in Sceloporus (Reptilia). Cladistics 5:49-61. Mozingo, H.N., and M. Williams. 1980. Threatened and endangered plants of Nevada. u.s. Dept. of the Interior. Osborne, H.F. 1902. Homoplasy as a law of latent or potential homology. Amer. Nat. 36:36:259-271. Owen, R. 1843. Lectures on the comparative anatomy and physiology of the invertebrate animals. Longman, Brown, Greene, and Longmans, London. ------. 1847. Report on the archetype and homologies of the vertebrate skeleton. Rep. 16th Meeting British Assoc. Adv. sci., 169-340. ------. 1848. On the Archetype and Homologies of the Vertebrate Skeleton. R. and J.E. Taylor, London Pagel, M.D., and P.H. Harvey. 1988. Recent developments in the analysis of comparative data. Q. Rev. BioI. 63:413-440. Palmer, J., B. Osorio, J. Aldrich, and W. Thompson. 1987. Chloroplast DNA evolution among legumes: loss of a large inverted repeat occurred prior to other sequence rearrangements. Curro Genet. 11:275-286. Patterson, C. 1982. Morphological characters and homology, pp. 21-74. In K.A. Joysey and A.E. Friday (eds.), Problems of Phylogenetic Reconstruction. Academic Press, London. Penny, D. and M.D. Hendy. 1985. Testing methods of evolutionary tree construction. Cladistics 1:266-278. Pielou, E.C. 1969. An Introduction to Mathematical Ecology. John Wiley & Sons, New York. Pimentel, R.A., and R. Riggins. 1987. The nature of cladistic data. Cladistics 3:201-209. Pod1ech, D. 1988. Revision von Astragalus L. sect. Caprini. Mitt. Bot. Munchen 25:1-924. 320

Polhill, R.M. 1981. Galegeae, pp. 357-363. In R.M Polhill and P.H. Raven (eds.), Advances in Legume systematics. Royal Botanic Gardens, Kew. Polya, G. 1968. Patterns of Plausible Inference. Princeton University' Press, Princeton. 2nd ed. Prance, G.T., and F. White. 1988. The genera of Chrysobalanaceae: A study in practical and theoretical taxonomy and its relevance to evolutionary biology. Phil. Trans. R. Soc. Lond. B. 320:1-184. Rasmussen, F.N. 1983. On "apomorphic tendencies" and phylogenetic inference. Syst. Bot. 8:334-337. Remane, A. 1952. Die Grundlagen des Naturlichen Systems, der vergleichenden Anatomie und der Phylogenetik. Akademische Verlagsgesellschaft, Leipzig. Rensch, B. 1959. Evolution above the Species Level. Columbia University Press, New York. Ridley, M. 1983. The Explanation of Organic Diversity: The Comparative Method and Adaptations for Mating. Clarendon Press, Oxford. Riedl, R. 1978. Order in Living Systems: A Systems Analysis of Evolution. John Wiley & Sons, New York. Roberts, M.L. 1977. Systematics of the Orophaca astragali. Master's Thesis. University of Wyoming. Rohatgi, V.H. 1976. An Introduction to probability Theory and Mathematical Statistics. John Wiley, New York. Rosenfeld, I., and O.A. Beath. 1964. Selenium: Geobotany, biochemistry, toxicity, and nutrition. Academic Press, New York. Roth, I. 1977. Fruits of Angiosperms. Gebrueder Borntraeger, Berlin. Roth, V.L. 1988. The biological basis of homology, pp. 1-26. In C.J. Humphries (ed.), ontogeny and Systematics. Columbia University Press, New York. Rydberg, P.A. 1929. Astragalanae. North American Flora 24:251-462. 321

Saether, O. 1979. Underlying synapomorphies and anagenetic analysis. Zool. Scr. 8:305-312.

Saether, O. 1983. The canalized evolutionary potential: Inconsistencies in phylogenetic reasoning. Syst. Zool. 32: 343-359.

Saether, O. 1986. The myth of objectivity -- post Hennigian deviations. Cladistics 2: 1-13.

Sanderson, M.J. 1989. Astragalus nutriosensis: A new species from eastern Arizona. Madrono.

------. 1989. Confidence limits on phylogenies: The bootstrap revisited. Cladistics, in press.

Sanderson, M.J., and M.J. Donoghue. 1989. Patterns of variation in levels of homoplasy. Evolution, in press.

Sattler, R. 1988. Homeosis in plants. Amer. J. Bot. 75:1606- 1617.

Scott, W.B. 1891. On the osteology of Mesohippus and Leptome~yx, with observations on the modes and factors of evolution in the Mammalia. Jour. Morph. 5:301-402.

Senn, H.A. 1938. Chromosome number relationships in the Leguminosae. Bibliogr. Genet. 12:175-336.

Sessions, S.R., and A. Larson. 1987. Developmental correlates of genome size in plethodontid salamanders and their implications for genome evolution. Evolution 41: 1239- 1251.

Sillen-Tullberg, B. 1988. Evolution of gregariousness in aposematic butterfly larvae: A phylogenetic analysis. Evolution 42:293-305.

Simberloff, D. 1987. Calculating probabilities that cladograms match: A method of biogeographical inference. Syst. Zool. 36:175-195.

Simberloff, D., K.L. Heck, E.D. McCoy, and E.F. Connor. 1981. There have been no statistical tests of cladistic biogeographic hypotheses, pp. 40-63. In G. Nelson and D.E. Rosen (eds.), Vicariance Biogeography: A critique. Columbia University Press, New York.

Simpson, G.G. 1944. Tempo and Mode in Evolution. Columbia University Press, New York.

--~-~------322

------1959. Anatomy and morphology: Classification and evolution: 1859 and 1959. Proc. Amer. Phil. Soc. 103: 286- 306.

------1961. principles of Animal Taxonomy. Columbia University Press, New York.

------. 1980. Why and How? Some Problems and Methods in Historical Biology. Pergamon Press, Oxford.

Sneath, P.H.A., and R.R. Sokal. 1973. Numerical Taxonomy. W.H. Freeman, San Francisco.

Spellenberg, R. Chromosome numbers and their cytotaxonomic significance for North American Astragalus (Fabaceae). Taxon 25:463-476.

Stebbins, G.L. 1971. Chromosomal Evolution in Higher Plants. Edward Arnold Ltd., London.

------. 1974. Flowering Plants: Evolution Above the species Level. Harvard University Press, Cambridge.

Stermitz, F.R., F.A. Norris, and M.C. Williams. 1969. Miserotoxin, a new naturally occurring nitro compound. J. Amer. Chem. Soc. 91:4599-4600.

Stevens, P.F. 1986. Evolutionary classification in botany, 1960-1985. J. Arn. Arb .. 67: 313-339

Swofford, D. 1985. Phylogenetic Analysis Using Parsimony, verso 2.4. Illinois Natural History Survey.

Swofford, D., and W. Maddison. 1987. Reconstructing ancestral character states under Wagner parsimony. Math. Biosci. 87:199-229.

Throckmorton, L. 1965. Similarity versus relationship in Drosophila. Syst. Zool. 14: 221-236.

Tuomikoski, R. 1967. Notes on some principles of phylogenetic systematics. Suom. Hyonteistiet Aikak. 33:133- 147.

Turner, B.L., and O.S. Fearing. 1959. Chromosome numbers in the Leguminosae II: African species, including phyletic interpretations. Amer. J. Bot. 46:49-57.

------323

Van Valen, L. 1982. Homology and causes. J. Morphol. 173:305-312.

Vavilov, N. 1922. The law of homologous series in variation. J. Genet. 12: 47-89.

Vilkmomerson, H. 1943. Chromosomes of Astragalus. Bull. Torr. Bot. Club 70:430-435.

Welsh, S.L. 1979. Illustrated manual of proposed endangered and threatened plants of Utah. u.s. Dept. Interior.

Wernham, H.F. 1912. Floral evolution: with particular reference to the syrnpetalous dicotylodons. IX. Summary and conclusions. New Phytol. 11:373-397.

Wiley, E.O. 1981. Phylogenetics: The Theory and Practice of Phylogenetic Systematics. John wiley & Sons, New York.

williams, M.C., and R.C. Barneby. 1977a. The occurrence of nitrotoxins in North American Astragalus (Fabaceae). Brittonia 29:310-326.

-----. 1977b. The occurrence of nitrotoxins in Old World and South American Astragalus (Fabaceae). Brittonia 29:327-331.

wright, S. 1978. Evolution and the Genetics of Populations, vol. IV: Variability within and among Natural Populations. University of Chicago Press.

- -.------