From-1\ the Mi41ddle -D.Evoni)*;An O)F No(Rthnamerica

Home , Calymene, Ferron Point Formation, Greenops, Phacopidae, Phacops, Phacops rana

S!YSTiEM\ATICS ANID EVO(LUJT(ION* O)F PH-ACOPS RAN (GEEN, 1832) AN1*D PHIACOPS IOWVEN)SIS DELO, 1935 (TR.ILoBITA) FROM-1\ THE MI41DDLE -D.EVONI)*;AN O)F NO(RTHNAMERICA

NILES ELDREDGE

BULLETIN OF THE AMERICAN MUSEUM OF NATURAL HISTORY VOLUME 147:ARTICLE 2 NEW YORK:1,972 SYSTEMATICS AND EVOLUTION OF PHACOPS RANA (GREEN, 1832) AND PHA COPS IOWENSIS DELO, 1935 (TRILOBITA) FROM THE MIDDLE DEVONIAN OF NORTH AMERICA

NILES ELDREDGE Assistant Curator, Department ofInvertebrate Paleontology The American Museum ofNatural History

BULLETIN OF THE AMERICAN MUSEUM OF NATURAL HISTORY VOLUME 147: ARTICLE 2 NEW YORK : 1972 BULLETIN OF THE AMERICAN MUSEUM OF NATURAL HISTORY Volume 147, article 2, pages 45-114, figures 1-28, tables 1-9 Issued February 14, 1972 Price: $2.70 a copy

Printed in Great Britain by Lund Humphries CONTENTS

ABSTRACT...... 49 INTRODUCTION...... 51 Abbreviations...... 52 MIDDLE DEVONIAN STRATIGRAPHY...... 5 MORPHOLOGY AND RELATIONSHIPS OF THE BiOSPECIES Plhacops rana (Green, 1832) and Phacops iowensisDelo, 1935...... 56 Systematic Paleontology...... 58 PhacopsiowensisDelo, 1935...... 58 Phacopsrana (Green, 1832)...... 59 The Affinities ofPhacops rana and Phacops iowensis...... 59 Previous Work on Middle Devonian Phacopid Taxa in North America...... 60 ANALYTICAL TECHNIQUES ...... 61 ONTOGENY OF THE CEPHALON OF Phacops rana...... 65 INTERPOPULATIONAL VARIATION IN Phacops rana...... 69 Interpopulation Variation in Number of Dorsoventral Files...... 70 Factor Analysis and Interpopulational Variability ...... 71 Systematic Paleontology...... 78 Phacops rana crassituberculata Stumm, 1953...... 78 Phacops rana milleri Stewart, 1927...... 79 Phacops rana norwoodensis Stumnm, 1953...... 79 Phacops rana paucituberculata, New Subspecies...... 80 Phacops rana rana (Green, 1832)...... 81 A Theory ofRelationships among Subspecies ofPhacops rana ...... 81 Trends (Biostratigraphic Character Gradients) in Phacops rana...... 86 Mode of'Origin of the Subspecies...... 87 Biostratigraphic Significance ofPhacops rana...... 88 N7ARIATION IN Phacops iowensis ...... 89 Qualitative Analysis ofMorphology...... 89 Subtaxa ofPhacops iowensis ...... 91I Systematic Paleontology...... 91 Phacops iowensi-s alpenensis (Stumm, 1953)...... 91 Phacops iowensis iowensis Delo, 1935...... 92 Phacops iowensis southworthi Stumm, 1953...... 92 INTERACTIONS BETWEEN Pha-cops rana AND Pliacops iowensis...... 93 Summary of the Distribution of the Two Species...... 93 Phacops rana...... 93 ni Plhacops.iw...... 93 Mutual Occurrence ofPhacops rana and Phacops iowensis...... 95 Interactions between the Species...... 96

SUMMARY.~~~~~~~~~~~~~~~~~~~~103..... APPENDIX 1.- Glossary of Morphological Terms .106... APPENDix 2. Locality List...... 108 REFERENCES CITED...... 111

ABSTRACT Twvo TRILOBITE SPECIES, Phacops rana (Green, 1832) the reduction is an allopatric phenomenon, involving and Phacops iowensis Delo, 1935, from the Middle transitional populations acquiring a reduced number Devonian of North America, are analyzed in detail of dorsoventral files on the eastern margin of the from the point ofview ofgeographic and stratigraphic craton (exogeosyncline), which subsequently invade v-ariation. A purely morphological, as opposed to bio- the cratonal interior. Other evolutionary changes stratigraphic, approach is used in analyzing relation- in the P. rana lineage appear to be phyletic trends. ships among subtaxa of the two species. Subsequent The two species are nearly mutually exclusive, comparison of the relative sequence of inferred though coeval and their geographic ranges overlap evolutionary events with documented biostratigraphic considerably. Although P. iowensis occurs from Iowa distributions allows an analysis of trends (biostrati- to New York, it was confined to the Michigan Basin graphic character gradients), character displacement for the larger part of its history. Phacops rana is found between the two species, and mode of origin of the from New York south to Virginia and west as far as subtaxa. Iowa. Phacops iowensis was by far the more The stable schizochroal eye of these species is the most through time; it is invariably rare, and generally important anatomical complex in terms of both confined to purer limestones. The one case ofsympatry intrapopulational and interpopulational variation between rana and iowensis occurs in the Hungry and species discrimination. Lens number per eye Hollow Formation of Ontario, resulting in may be broken down into number mor- of dorsoventral phological changes in the two species which are best files (vertical columns of lenses) and number of explained as character displacement. lenses per dorsoventral file. The adult The P. rana population lineage as a whole converged on iowensis in number number of dorsoventral files per eye in Phacops is ofdorsoventral files and in many ornamental features; reached early in ontogeny and is stabilized; the the convergence was number of dorsoventral files is closest in the Taghanic. the most consistently The distributions of, and interactions between, reliable criterion for discrimination of the two species. rana and Throughout its P. Phacops Phacops iowensis are best explained history, iowensis had 13 dorso- if the two taxa are considered true "bio-species." ventral files in normal adults. This species belongs to a the native North American Although little change that occurred within the phacopid lineage that can iowensis lineage seems to have been phyletic in be traced back at least as far as the Gedinnian the nature, ("Phacops" logani Hall). allopatric model is necessary to account for the Phacops rana, more important evolutionary changes in P. rana. morphologically closest to P. schloth- Five subspecies of Phacops cimi (Bronn) from Europe and elsewhere, has from rana, including P. rana 15 to 18 dorsoventral files. paucituberculata, new subspecies, and three subspecies Most of the history of ofP. are species involved the reduction from 18 to 15 files; iowensis, recognized.

INTRODUCTION INVESTIGATIONS of evolutionary phenomena on Museum of Natural History and Columbia the species level are again coming into vogue in University devoted many hours to discussing paleontology. This renewed interest stems pre- various evolutionary implications, as did Dr. dominantly from the emphasis on populations N. D. Newell of the same institutions. I am also in the "New S,ystematics," the realization that grateful for conversations with Mr. J. C. Boylan, large samples of fossils, particularly of marine of Columbia, Dr. L. Marcus of Queens College, invertebrates, spanning broad geographic areas Dr. R. M. Finks of Queens College, and Dr. G. and thick stratigraphic sequences are available, J. Nelson of the American Museum of Natural and the relatively recent appearance ofsophisti- History. cated multivariate statistical routines per- To Dr. D. V. Manson of the American Mus- formed on computers. The present study gives eum I owe a special debt of gratitude. Dr. the results of a detailed consideration of the Manson wrote thefactoranalysisprogram MUST variation and evolution within two coeval species used extensively herein, and arranged for the of trilobites whose geographic ranges overlap. analysis of the data. He also reinitiated me into Taken together with other similar studies that the intricacies of factor analysis, thereby greatly have appeared recently, it is hoped that the increasing my knowledge of this multivariate feasibility of studying micro-evolutionary phe- procedure. nomena in the fossil record may be demon- Dr. E. N. K. Clarkson of the University of strated. Edinburgh has given me instructive advice on Quantitative techniques, especially various the nature ofphacopid eyes and has also critically forms of factor analysis, are used extensively in reviewed portions of an earlier version of the parts of the present report. By now these tech- manuscript. niques are conventional and need no further I wish to thank the following people for discussion, although the basic elements of the arranging loans of specimens from their re- multivariate models actually used are explained spective institutions: Dr. S. J. Gould, Museum of in the text. The study does stray a bit from con- Comparative Zoology, Harvard University; vention in that a biostratigraphic approach for Mr. C. Kilfoyle, New York State Museum; phylogenetic reconstruction is abandoned in Dr. Steven W. Mitchell, Wayne State Univer- favor of an analysis of relationships based sity; Dr. E. C. Stumm, University of Michigan strictly on morphological features determined Museum of Paleontology; Mr. F. J. Collier, to be "primitive" or "advanced." A purely United States National Museum, Smithsonian "cladistic" approach to the systematics of fossil Institution; Dr. R. Linsley, Colgate University; taxa need not debar application of evolutionary and Mr. H. Strimple, State University of Iowa. models to the explanation of phylogenetic Mr. Strimple also helped to clarify the strati- patterns in the fossil record. This point ofview is graphic positions ofmany ofthe specimens in the elaborated on in the text. A detailed application Iowa collections. ofthe allopatric model to certain elements ofthe I thank Mr. L. Magrum and Mr. M. E. data presented here is contained in Eldredge Widener of Toledo, Ohio, for the donation of (1971). many well-preserved specimens from the Silica shale of the Toledo area. Mr. Widener addition- ACKNOWLEDGMENTS ally compiled an invaluable report on the bio- I am deeply indebted to Dr. H. B. Rollins of stratigraphy of Phacops in the Silica shale based the University of Pittsburgh who originally on his extensive experience in the quarries of the introduced me to the Hamilton fauna of central Medusa Portland Cement Co., Silica, Ohio. New York. Our discussions on Phacops and other Mr. David R. Eldredge gave very helpful faunal elements, as well as the physical stratig- advice and assistance in computation. Mr. raphy of the Middle Devonian of New York, Richard C. Eldredge served as field assistant for were of great help throughout the course of portions of the summers of 1966 and 1967; his this study. Dr. R. L. Batten of the American enthusiasm, diligence, and sharp eyes provided 51 52 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147 many of the better specimens of Phacops used supported by a National Science Foundation in this study. Graduate Fellowship and funds for field work Finally, my wife, Michelle, who has been granted by the Department of Geology, Co- forced to live in a world colored by Phacops for lumbia University. several years, deserves much of the credit for the final appearance of this study. As field assistant ABBREVIATIONS her eyes were unerring and her spirit unflagging. AMNH, the American Museum of Natural History, Her encouragement throughout the subsequent Department of Invertebrate Paleontology months of analysis and final preparation of MCZ, Museum of Comparative Zoology, Harvard the present report made the way far easier for University NYSM, New York State Museum, Albany me. This is attributable to her study largely SUI, State University ofIowa and is to her. patience dedicated UMMP, University of Michigan Museum of Paleon- The contents of the present paper are drawn tology largely from my Ph.D. dissertation at Columbia USNM, National Museum of Natural History, University completed in 1969. Research was Smithsonian Institution MIDDLE DEVONIAN STRATIGRAPHY THE MIDDLE of The DEVONIAN eastern and central main change is the downward extension of North America provides an excellent source of the Givetian to include the entire material for studies Cazenovian of geographic variation and and the reallocation of the Taghanic to the evolution. The rocks are well known and corre- younger Frasnian series. Thus under lations are reasonably currently secure. There is a wide accepted terminology, there is no convenient range in types of sediments deposited synchro- term to embrace nously in different regions, and a Cazenovia, Tioughnioga, and great deal of Taghanic. The Taghanic is Frasnian on the time (some eight to 10 million years) was avail- basis of able for European ammonoid zones, and al- evolutionary change. An added advan- though containing new, European-derived faun- tage to the present study is that Phacops has al elements rarely been utilized (e.g., Scutellum), clearly is faunistically for biostratigraphic purposes, closer to the underlying Hamilton rocks removing any possible from to than circularity the the rocks subsequently deposited. On this framework ofrelative time. basis, the classification of Both the Cooper et al. (1942) is time-stratigraphic classification of perhaps the more convenient as Middle Devonian rocks of the United it provides a States term ("Erian") which exactly corresponds to the and correlation with European series have Phacops-bearingrocks herein discussed. undergone revision in recent years and are not accurate correlation However, entirely stabilized. with European rocks is The "Hamilton" (com- clearly desirable to facilitate close prising the Marcellus, Skaneateles, Ludlowville, of related comparisons and Moscow formations as taxa, and as conodonts and ammo- revised by Cooper, noids indicate a correlation 1930) and the overlying Tully and "Chemung" more along the lines formations and their lateral propounded by Rickard (1964) than by Cooper equivalents, are the et al. (1942), Rickard's classification is Phacops rana and P. iowensis-bearing rocks. here. For accepted Cooper et al. convenience, the term "Middle (1942) classified these rocks as the Devonian" as used Erian series, subdivided into the throughout this study stands Cazenovia, for Cazenovia, and Tioughnioga, and Taghanic stages (see fig. 1). Tioughnioga, Taghanic The lower portion of the stages. Cazenovian, corre- a of sponding to the Marcellus and its equivalents, Although study geographic and temporal was variation must necessarily have an considered Eifelian, and the rest of the established time independently Erian was allocated to the Givetian. framework on a regional basis, A slightly the Middle Devonian of the eastern and central rearranged classification is given United States is far from by Rickard (1964) and adopted here (fig. 1). totally understood. Disagreements on the correct correlation of COOPER, et al., 1942 N.Y.S. SECTFION RICKARD, 1964 TAGHANIC TULLY TAGHANIC SENECAN FRASNIAN MOSCOW GIVETIAN TIOUGHNIOGA TIOUGHNIOGA ERIAN LUDLOWVIL-LE ERIAN GIVETIAN ':KAKJF:T971ORVMINArItLtCceZ CAZENOVIA CAZENOVIA EIFELIAN MARCELLUS

Fig. 1. Comparison of the time-stratigraphic classifications of Middle Devonian rocks by Cooper et al. (1942) and Rickard (1964). 53 Ca) a

u)0 a.) C.)6 c.)

a. CU 1972 ELDREDGE: PHACOPS 55 rocks of certain areas with the standard of the interior to the west. The rock sequences in the Chenango Valley in New York are common, east are also much thicker and evidently repre- and a certain amount of choice exists between sent a more complete record oftime. two or more slightly variant correlations pro- The well-known clastic wedge of the eastern posed by different workers. The works of G. A. exogeosyncline of the Middle Devonian thins Cooper over the past 40 years serve as the and pinches out to the west, with sandstones starting point for all discussions of Middle and siltstones giving way to calcareous shales and Devonian strata of the eastern and central purer limestones. Occasional limestones do states and southwestern Ontario. Figure 2 shows occur in the east and form the basis ofmost ofthe the major stratigraphic sections investigated in understanding of the physical stratigraphy of the present study. Although Cooper et al. (1942) the Hamilton Group, particularly in New York. is the main source, the most recent work in But in general, within any one recognizable each area has been incorporated into the chart. biostratigraphic unit ofshort duration, there is a A noticeable feature of the Middle Devonian great range of sedimentary environments pre- of eastern and central North America is the served. The direction of greatest sediment virtual absence of Tioughnioga sediments west variation is along a line normal to the roughly of the Buffalo arch. The only exceptions are the north-south strike of the ancient shoreline in the widespread basal Tioughnioga "Centerfield" east running west to Michigan. It is along this faunas and the Widder, Norway Point, and axis that the greatest amount ofvariation within Lower Petoskey formations of southwestern both species and subspecies of Phacops is evident. Ontario and eastern and western Michigan. There is much less sediment and biologic No proved Moscow equivalent exists west of variation parallel to the shoreline within the Buffalo, New York where the Windom member Appalachian system and within other local thins to 12 feet on the eastern shore ofLake Erie. tectonic areas such as the Michigan Basin. The stratigraphic section, then, is far more A complete locality list for specimens used complete in the exogeosyncline on the eastern is given in Appendix 2. margin ofthe continent than it is on the cratonal MORPHOLOGY AND RELATIONSHIPS OF THE BIOSPECIES PHACOPS RANA (GREEN, 1832) AND PHACOPS IOWENSIS DELO, 1935 Phacops AND RELATED GENERA are morphologi- of distinguishing P. rana from P. iowensis. In cally quite uniform. Campbell (1967, p. 33-35) addition, as discussed in detail below, the same has recently presented a full diagnosis of the criterion is of utmost importance in the de- genus Phacops which will not be repeated here. lineation of subtaxa within the P. rana group. There is as yet little agreement on reliable, Since Clarkson's (1 966a, 1 966b) work appeared, species-specific characters that may be used the importance of dorsoventral file number in systematically to clarify the definitions of, and phacopid (sensu lato) systematics has been inde- interrelationships among, the many species of pendently discovered by Burton (working on Phacops described in the literature. Although it is some European species of Phacops) and Young unwise to extrapolate the findings of a single (working on American pterygometopids). There study on but two species and apply a priori the can be no question of the great potential that species-specific criteria found to differentiate P. this easily observed character may have in rana from P. iowensis to the anaylsis of other clarifying the interrelationships among phacopid species of Phacops, certain of these differentia do taxa. seem to have importance within the genus as a The lenses are always arranged in 13 dorso- whole and promise to help untangle the difficult ventral files (vertical columns) on the visual problem of species relationships within Phacops surface of the eye in all but the smallest post- and closely allied genera in the Silurian and larval instars of P. iowensis. Phacops rana, on the Devonian. other hand, has from 15 to 18 dorsoventral Foremost among these characters is the eye files, and two intermediate populations are complex, particularly the distribution of lenses known where variation between two or more on the visual surface. Steininger (1831, 1. 351, ff.) dorsoventral file numbers is continuous. prophetically saw the importance of lens Phacops iowensis also tends to have fewer arrangement as an important taxonomic feature lenses per dorsoventral file than does P. rana, and of the phacopids, but it was not until Clarkson the combination of fewer dorsoventral files and (1966a, 1966b) presented lens number counts fewer lenses per dorsoventral file amounts to in terms of dorsoventral files (see Appendix 1 for differences in average number of lenses per eye a glossary of morphological terms) that the between the two species. The average number systematic value of lens arrangement within the of lenses per population ranges from 40 to 48 Phacopidae became clear. Very simply, a in different populations of P. iowensis, and the unique number of dorsoventral files is reached maximum number of lenses observed in an early in holaspid ontogeny within any single individual specimen is 54. Interpopulational population sample and is stabilized throughout variation in average number of lenses is great the remainder of ontogeny. With but few in P. rana, and ranges from a low of 51 to over exceptions known to me, all variation in dorso- 100. ventral file number is interpopulational or In addition to dorsoventral file numbers, other interspecific. Variation in total number of characters serve to differentiate P. rana from P. lenses in the eye, on the other hand, is a function iowensis, and, equally as important, to unite both of dorsoventral file number and the as a whole the various populations of P. rana number of lenses per dorsoventral file. This with different numbers of dorsoventral files latter factor is complex and is discussed at length (see figs. 3, 4, 13-16, 22, 23). The librigenal and elsewhere (Eldredge, MS); variability in number fixigenal moieties of the ocular platform are of lenses per dorsoventral file ranges from far more sharply defined in P. iowensis than in asymmetry within an individual through the P. rana because of the more deeply emplaced interspecific level. facial suture in the former. Differences in dorsoventral file number form The exoskeleton of both species tends to be the primary, but by no means the only, means covered by tubercles over wide areas. Tubercles 56 1972 ELDREDGE: PHACOPS 57 in P. iowensis tend to be of uniform size and are margin is thus slightly recurved posterodorsally generally rounded at the base, rising in a conical anterior to the genal angle in P. iowensis. fashion and often terminating in a point. Phacops There is no consistent difference in the num- rana shows greater variation in tubercle size ber of axial rings or pleura on the pygidia of the and shape, but the tubercles are usually bluntly two species. The pleura are highly arched and rounded on top. the pleural furrows deeply incised in P. iowensis, Tubercle shape is of critical importance on in contrast to the rather flat pleura and shallow certain areas of the exoskeleton. In P. rana, the pleural furrows in P. rana. tubercles of the occipital lobe and axial rings The first anterior interpleural furrow com- of the thorax and pygidium become transversely monly is weakly developed in P. rana, particu- elongated and flattened. Tubercle elongation is larly in specimens from older formations. The also present on the cephalic margin on the genae and particularly on the anterior portion of the composite glabellar lobe, where the elongation is frequently so highly developed as to produce long, wavy transverse ridges similar to those on the external surface of the doublure and hypostoma. No transverse elongation of tubercles is found in P. iowensis. Tubercles in both species tend to be larger on the glabella (including glabellar lobe Ip) than elsewhere on the cephalon. The ocular platform is usually more heavily tuberculated in P. iowensis than in P. rana. Tuberculation is generally heavier over the thorax and pygidium in P. iowensis than in most populations of P. rana. The lateral and posterior margins of the pygidium are heavily tuberculated in P. iowensis, and tuberculation is generally obsolescent on the pygidial margins ofP. rana. Many specimens of P. iowensis, including the holotype (SUI 9 266, see fig. 3), show obsolescence of the lp glabellar furrow mesially and a consequent tendency toward incorporation of glabellar lobe 1 p with the composite glabellar lobe. Whereas other specimens of P. iowensis show moderately deep emplacement of glabellar furrow lp, P. rana never shows any obsolescence of this furrow. Glabellar furrows 2p and 3p are variably developed in both species, but perhaps more commonly encountered in P. iowensis. The genal angle terminates in a moderately sharp point in P. iowensis, and is somewhat more bluntly rounded in all but the earliest populations of P. rana. When the cephalon is oriented with the dorsal margin of the visual surface in the horizontal plane, the genal angle forms the C ventralmost portion of the cephalic margin in rana. In iowensis, with the of some of FIG. 3. Phacops iowensis iowensis Delo, 1935. A, B. exception Holotype, Cedar Valley Formation, SUI 9-266. A. the earlier specimens known, the ventralmost Dorsal view, x2. B. Left lateral view of cephalon, portion of the cephalon is anterior to the genal x4. C. Dorsal view of pygidium, paratype, Cedar angle and is situated below the eye. The ventral Valley Formation, SUI 9-117, x 3.5. 58 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147 anterior two or even three interpleural furrows developed or absent. Cephalon covered by are frequently developed in P. iowensis. round, conical tubercles of uniform size, not Additional criteria for distinguishing the becoming elongated transversely. two species emerge in factor analysis of linear Thorax generally covered with tubercles measurements and are discussed below. similar to those on cephalon. By far the most important differences be- Pygidium with eight to 10 axial rings and tween the two species are the number of dorso- six to eight pleura. Pleura highly arched, pleural ventral files on the visual surface, shape and furrows deeply incised. Anterior two to five distribution of tubercles, particularly on the interpleural furrows frequently developed, with cephalon, and relative degree of differentiation area of fused pleuron anterior to interpleural of the ocular platform into librigenal and furrow more highly arched than posterior fixigenal moieties. region. Tuberculation heavily developed over Although the term "biospecies" should perhaps be applied as a conclusion following an exhaustive study, it is perhaps justifiable at this point to claim such status for both P. rana and P. iowensis as a working hypothesis. The characters enumerated immediately above at the very least justify recognizing both rana and iowensis as "Operational Taxonomic Units." Full status as "biospecies" further depends upon demonstration of historical cohesion or integrity, which can only be deduced following a full analysis of interrelationships among any subtaxa found to exist, and consideration of geographic and stratigraphic distributional data. But the fact remains that dorsoventral file number and all other characters mentioned above point to the existence of two discrete and internally consistent taxa, and for the purposes of subsequent analysis and discussion, the assumption is made that P. rana and P. zowensis in fact are true biospecies. B The two species may be diagnosed as follows:

I. SYSTEMATIC PALEONTOLOGY FAMILY PHACOPIDAE HAWLE AND CORDA, 1847 GENUS PH,4COPS EMMRICH, 1839 D Phacops iowensis Delo, 1935 Phacops iowensis DELO, 1935, p. 422-423, p1. 48. Phacops iowensis: STUMM, 1953, p. 140-142, pl. 12. Figure 3; see also figures 22, 23 FIG. 4. Phacops rana rana (Green, 1832). A, B. EMENDED DIAGNOSIS: Eyes moderately large, Hamilton Group, Eighteen Mile Creek, Erie County, bearing 13 dorsoventral files of lenses in normal New York, AMNH 5499/1. Figured by Hall and adults. Trace of facial suture over ocular Clarke, 1888, pl. 8, fig. 7. A. Dorsal view of cephalon, platform deeply incised. Genal angles termi- x 3. B. Frontal view of cephalon, x 3. C. Left lateral in view of cephalon, Hungry Hollow Formation, nating blunt point, recurved dorsally from UMMP locality B, UMMP 24313, x2. See also ventral cephalic margin. Glabellar furrow lp figure 24. D. Dorsal view of pygidium, Brandon variably incised, occasionally obsolescent mes- Substage, Cedar Valley Formation, Belanski collec- ially. Glabellar furrows 2p and 3p weakly tion, SUI 6267, x 3. 1972 -A.dJL.141%.A-iJ--lN-7L ELDREDT.hRm* PRArf DPE 59 entire surface of pygidium, including pygidial Gedinnian-Siegenian species of Phacops cur- margins. rently recognized from North America include HOLOTYPE: State University ofIowa 9-266. P. logani Hall, 1861, P. hudsonicus P. Hall, 1861, Phacops rana (Green, raymondi Delo, 1935, and P. 1832) 1969. Emsian-Eifelian claviger Haas, var. species include P. cristata Calymene bufo rana GREEN, 1832, p. 42, casts 11, 12. Hall, P. cristata var. Calymenebufo: and pipa Hall and Clarke, HALL, 1843, p. 201, fig. 80. 1888, as well as variants of P. cristata Phacops bufo: EMMONS, 1860, p. 138, fig. 124. described recently Phacops rana: HALL, 1861, P. 55. by Stumm (1954). Each of these rana: species is characterized essen- Phacops HALL AND CLARKE, 1888, p. 19-26, pls. by tuberculation 7-8. tially identical in shape and distribution to the condition in P. Phacops rana: DELO, 1940, p. 22-23, pl. 1. iowensis described above. In rana: P. has Phacops STUMM, 1953, p. 135-140, pis. 9-12 addition, logani 17 dorsoventral files, (in part). whereas P. hudsonicus has probably 16 files, and Figure 4; see also figures 13-16 P. raymondi has 15. Phacops cristata and P. cristata var. EMENDED DIAGNOsIs: Eyes pipa each have 14 dorsoventral large, bearing from files. It seems 15 to 18 dorsoventral files of lenses in normal quite probable that these species are closely related and that there is a definite adults. Trace of facial suture over ocular character platform shallow. Genal angles gently rounded gradient involving reduction in and number of dorsoventral files the near ventral cephalic margin. Glabella Lower and Lower throughout furrow lp deeply incised, glabellar furrows 2p Middle Devonian. Phacops and iowensis with 13 dorsoventral files also to 3p weakly developed or absent. Cephalon this belongs covered by low, rounded tubercles "Native North American" species complex. becoming In most other aspects of P. iowensis transversely elongate at the anterior margin is morphology, of the glabella, on the very similar to P. cristata var. pipa. In addition, genae, and on the occip- a ital lobe. Tubercles largest on single specimen in the collections of the central region National Museum of Natural ofcomposite glabellar lobe and glabellar Jobe lp. History (USNM Axis of thorax covered with 78923) from the Gedinnian Linden Formation transversely of Tennessee with 13 elongate tubercles. Tuberculation on pleura dorsoventral files is variably surprisingly similar in all respects to developed. of P. specimens Pygidium with from seven to iowensis from the Hungry Hollow 10 axial rings tion of Ontario. Forma- and six or seven pleura. Tubercles moderately Phacops iowensis was almost elongate on certainly derived from a native North American transversely axis; tubercles cover Devonian pleura, becoming obsolescent on pygidial mar- species. gin. Interpleural furrows generally Among well-known species, the closest relative obsolescent, to P. rana anteriormost interpleural furrow occasionally seems to have been P. schlotheimi present as shallow groove (Bronn) from the Eifelian of and else- set off by parallel where. Germany rows oftubercles. Pleural furrows rather shallow, Details of genal angle and ocular plat- pleura only moderately arched. form morphology, and mode of tuberculation HOLOTYPE: NYSM 13887/1. are closely comparable in the two species. Furthermore, P. schlotheimi has 18 dorsoventral THE AFFINITIES OF files and a 15 file variant, and the 18 dorso- PHACOPS RANA AND ventral file PHACOPS 10 WENSIS specimens can be further subdivided into two eye variants Hall and Clarke (1888, p. (Burton, MS.a, MS.b). 24) observed that P. This situation resembles rana seemed more closely related to closely the pattern of species of variation seen in the earliest Phacops from Germany than to any other species (Cazenovian) populations of rana as found in the Devonian of North America. Phacops discussed below. In P. rana has Examination of specimens of Lower and addition, recently been found to Lower occur in certain areas of the Middle Devonian species of Phacops in the Spanish Sahara of collections of northwest Africa (Burton and the Amnerican Museum of Natural If Eldredge, MS). History and the Phacops rana or its immediate ancestor did Museum of Comparative from the Zoology, Harvard University, strongly migrate Devonian European or North con- African areas to firms the opinion ofHall and Clarke. Devonian North America, it was not the only trilobite to do so. Greenops 60 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147 (Greenops) boothi (Green, 1837) is a native nupera Hall from the Chemung sandstone and dalmanitacean of the predominantly European P. cacapona Hall from the Hamilton of Virginia subfamily Asteropyginae. This subfamily ranges as probably valid. from Lower Devonian into the Upper Devonian Stewart (1927, p. 58) recognized a new in Europe, but is represented in North America "variety" ofP. rana, P. rana var. milleri from the by the single subgenus Greenops (Greenops) found Silica shale of Ohio. The new variety was based in Cazenovian, Tioughniogan, and Taghanic largely on details of lens number and mor- rocks (Stumm, 1954, p. 202). Its earliest un- phology ofthe visual surface. questionable occurrence is in the Solsville Delo (1935, p. 422-423) recognized a new member of the Marcellus Formation (AMNH species, P. iowensis, from the Cedar Vallev of loc. 3013) slightly higher stratigraphically than Iowa. In a later paper (1940, p. 22-23), Delo the earliest known occurrence of P. rana. As it is indicated that not all Phacops from the Traverse evident that there were marine connections Group of Michigan, the Milwaukee Formation, between Europe-Africa and North America in and the Devonian of Virginia were conspecific upper Eifelian or lower Givetian times suitable with P. rana sensu stricto, though he recognized no for the migration of Greenops to North America, new taxa. In addition, Delo (1940, p. 21) it is conceivable that species of Phacops may also correctly considered P. nupera as unrecognizable have come in from the European-northwest and probably referable to P. rana. Delo (1940, Africa faunal province. In view of the striking p. 16) also correctly regarded the cotypes of P. similarity between P. rana and European- cacapona as a mixture of true P. rana and P. African species of Phacops, and the great dis- cristata var. pipa Hall and Clarke. crepancies between P. rana and species ofPhacops Stumm (1953) reviewed the systematics of found in the Lower and Lower Middle Devonian Phacops from the Traverse Group of Michigan ofNorth America, a European-African ancestry and the Devonian of northwestern and north- for Phacops rana seems well founded. central Ohio, and southwestern Ontario. He recognized seven subspecies of P. rana and two PREVIOUS WORK ON MIDDLE subspecies ofP. iowensis from these beds. Criteria DEVONIAN PHACOPID TAXA IN NORTH of differentiation of these taxa primarily in- AMERICA volved differences in total lens number in rela- Although Phacops has attracted much attention to cephalic size, number of lenses per tion in paleontological and biostratigraphic "vertical row," mode of ornamental tubercu- research in the Middle Devonian of North lation, and gross proportional differences eval- America, there exist only a relatively few com- uated qualitatively. prehensive studies of the systematics of these Finally, Stumm (1954) described Phacops trilobites. ohioensis on the basis of two pygidia from the Hall and Clarke (1888, p. 19-26) gave a Dundee limestone of Michigan and Ohio. These detailed description and discussion of P. rana pygidia are probably not phacopid, and seem that remains the most complete and generally closely comparable to the pygidium of a species excellent treatment of the anatomy of this of the proetid genus Dechenella in the Jefferson- trilobite to be found in the literature. However, ville limestone of southern Indiana (Stumm, they accorded little recognition to variation in 1964b). specimens from outside New York State, and It may be concluded, then, that P. rana and knowledge of regional Middle Devonian stratig- P. iowensis are the only recognizable taxa raphy was probably insufficient to allow correctly referred to Phacops in the span of recognition of variation through time. In addi- Middle Devonian time in North America under tion to P. rana, Hall and Clarke recognized P. consideration here. ANALYTICAL TECHNIQUES REPEATABILITY AND BIOLOGICAL interpretability interval; ideally this interval would be a single of measurements on any organism depend to a bedding plane, but few collections on the available large degree manner of orientation of the came from such a short interval. The term specimen. Clarkson (1966c), in his discussion "population" is thus extended of "PhacoPs" musheni Salter, here to include argued that the all specimens collected within a short interval most likely orientation ofthe cephalon in life was of rock with the dorsal representing an interval of time far margin ofthe visual surface in the shorter than the most abbreviated of time horizontal plane. This conclusion is based on or Clarkson's stratigraphic biostratigraphic units regionally studies on the physiological implica- recognizable in the Middle Devonian of North tions oflens arrangement in P. musheni, and seems America. to be applicable to phacopids in general; this The general of orientation was family multivariate clustering consequently adopted for all techniques known as factor has been cephalic measurements on P. rana and analysis P. shown to have a wide range of applicability to iowenszs. paleontological problems (Eldredge, Each specimen was mounted with Plasteline Various forms of factor 1968). on a square analysis have proved wooden block with the dorsal useful in this study in the clarification of charac- margins of both visual surfaces in the same ter horizontal gradients, discrimination between P. rana plane. Thus the dorsal surface of the and P. iowensis, and the of palpebral lobe was normal to the line of study interrelation- vision. ships among variables during the ontogeny Rotation of the block enabled the specimen to of a be viewed population sample ofP. rana. anteriorly, laterally, or posteriorly Factor is the in a systematic manner. Pygidia were mounted analysis essentially resolution of an Nxn data matrix Z into an Nxm factor with the surface of the first axial ring in the coefficient horizontal plane. matrix A and an mxn factor components matrix where N is the All measurements were made with an F, number of ocular samples n is the micrometer mounted in the eyepiece of (specimens here), number of a low- variables, and m is the rank of the data matrix, powered, binocular stereoscopic microscope, to the number where 1 millimeter was equal equal of mutually perpendicular to 7.4 micrometer theoretical vectors needed to units. Angles were measured with a goniometer explain a chosen attached to percentage of the variance of the data matrix. the microscope. The basic is thus Shaw (1957, p. equation Z=AF. The aim of 193) has discussed the problem factor analysis is to reduce the of comparing measurements made on trilobites ofthe data dimensionality with the exoskeleton space from n, the number ofvariables intact with those on ex- or original reference to foliated specimens. His conclusion, vectors, m, which is that the hopefully much smaller than n. The first measurements are usually quite comparable if made theoretical vector represents the axis of major to the deepest part of bounding furrows, variation of the data seems sound and has been original swarm, and the followed here, following vectors represent minor although preference was given to nonexfoliated directions successively specimens. ofvariation. A further step, where the calculated factor axes The measurements are may be rotated to achieve described in table 1 a better fit to the data be and illustrated (cephalon only) in figure 5. swarm, may carried As most of out by a best-fit procedure such as the Varimax the measurements involve simple criterion. each linear dimensions of clearly defined anatomical Thus, specimen (Ni) will have parts, they need not be discussed a score or coefficient on each factor, and each further. variable will have a certain The terms "sample" and "population" are (nj) score, or used relative importance, in the composition of each nearly interchangeably herein. Most of the factor. This collections used, particularly for measuring general series of procedures has purposes, been referred to as "Q-mode" factor analysis. come from a single locality and in "R-mode" factor most cases from a very short stratigraphic analysis simply factors a transposed data matrix Z', so that the factor 61 62 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147 TABLE 1 MEASUREMENTS OF THE CEPHALON AND PYGIDIUM CEPHALON TCL Total cephalic length, sagittally from front of glabella to transverse line joining posterior margirns ofgenal angles CL Cephalic length, sagittally from front ofglabella to rear ofoccipital lobe GL lp Length of composite glabellar lobe, sagittally from front ofglabella to glabellar furrow lp IpL Sagittal length ofglabellar lobe 1 p OCL Sagittal length ofoccipital lobe PEG Distance from anterior margin ofvisual surface to anterior (sag.) margin ofcephalon LVS Maximum length ofvisual surface LLVS Anterior length ofvisual surface, from anterior margin to palpebral line LPA Length ofpalpebral area, from occipital furrow to intersection ofaxial and palpebral furrows L Distance from palpebral line to posterior margin (sag.) ofoccipital lobe LOP Length ofocular platform from posterior margin ofvisual surface to occipital furrow MI lpA Distance between proximal margins of 1 p apodemal invaginations MlpA Distance between distal margins of lp apodemal invaginations OMIA Distance between proximal margins ofoccipital furrow apodemal invaginations OMA Distance between distal margins ofoccipital furrow apodemal invaginations MIGW Glabellar width immediately anterior to glabellar furrow lp CW Maximum cephalic widthjust anterior to posterior margins ofgenal angles WBVS Width between visual surfaces along palpebral line GWCM Glabellar width where axial furrows intersect anterior cephalic margins GWE Glabellar width where dorsal portion ofvisual surface intersects axial furrow WVS Width ofvisual surface along palpebral line WPA Width ofpalpebral area along palpebral line WPL Width ofpalpebral lobe along palpebral line WPAO Width ofpalpebral area along occipital furrow 1pW Width ofglabellar lobe 1p CWlp Width of lp glabellar lobe, exclusive ofdistal nodes WOC Width ofoccipital lobe WOP Width ofocular platform from axial furrow to distal margin HGL Anterior height ofglabella (sag.) FARCH Height offrontal arch, from anterior ventral margin to line connecting genal angles (sag.) ARCH A Angle of line connecting anterior margin of cephalon to lowest margin of the genae, from the horizontal HVS Height ofvisual surface along palpebral line HEYE Total eye height, from top of visual surface to cheek, including ocular platform if distinguishable from cheek COUNTS MADE ON CEPHALON No. TPL Number oftubercles on palpebral lobe No. TPA Number oftubercles on palpebral area No. TOR Number oftubercles on occipital ring No. Tlp Number oftubercles on glabellar lobe Ip No. TFC Number oftubercles on fixigena, including ocular platform No. TSL Tubercles on sagittal line ofcomposite glabellar lobe No. TTL Tubercles on composite glabellar lobe along transverse line at anterior margins ofthe visual surface No. COL Number ofdorsoventral files oflenses/eye Eye Count Number of lenses/dorsoventral file written in groups of three starting with most anterior dorsoventral file number 1 Total number oflenses - Summation ofeye count formula MEASUREMENTS MADE ON PYGIDIUM TLPYG Total sagittal length, from ring furrow ofanterior segment to posterior margin ofpygidium LPAX Length (sag.) from ring furrow ofanterior segment to posterior margin ofaxis MLPYG Length (sag.) from line connecting anterolateral pygidial margins to posterior margin WAPYG Width between anterolateral margins 1972 ELDREDGE: PHACOPS 63 TABLE 1-(Continued) WS5P Total width at level ofaxial segment 5 LUR Length (sag.) ofarticulating half-ring N4R Width ofarticulating half-ring LSl, LS5 Length (sag.) ofaxial segments 1 and 5 WVSI, WVS5 W idth ofaxial segments 1 and 5 WVPLF WNidth ofpleural furrow ofanterior pleuron COUNTS MADE ON PYGIDIUM -No. PYG Number ofaxial segments No. PL Number of pleura MEASUREMENTS MADE ON THORACIC SEGMENTS LTS 1, 6, 11 Length (sag.) ofsegments 1, 6, and 1 1 WVTS 1, 6, 11 Width ofsegments 1, 6, and 11 WVPLF 1, 6, 11 WVidth ofpleural furrow ofsegments 1, 6, and 1 1 WVTPL 1, 6, 11 Total width ofsegments 1, 6, and 11

F1G. 5. Cephalic measurements. See table 1 for abbreviations. 64 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY V'OL. 147 coefficients matrix (A) consists of the scores of divided by the square root of the sum of the each variable on the m factors and the factor squared observations. The effects of these two components matrix (F) reflects the relative transformations on the results are discussed in contribution of each specimen to each of the conjunction with the actual analyses presented m factors. R-mode factor analysis is useful in the below. study of the interrelationships of variables and In R-type analysis, each observation is gives results comparable to those of the "mor- subtracted from the mean and the data is then phological integration" procedure given by normalized by variables, so that Z'Z becomes Olson and Miller (1958). the familiar correlation coefficient matrix. Transformation of the original data matrix The factor analyses were performed on a before factoring is frequently useful to clarify CDC 6600 computer operated by the CEIR relationships between samples or between corporation. Programs included MUST, a variables. For example, conversion to loga- multifaceted factor analysis package developed rithms is useful in removing allometric effects. by V. Manson; MUST includes COVAP, an Three such data transformations are used here. earlier program developed by Manson and In analysis of sample interrelationships ("Q- Imbrie (1964) used here for R-mode analysis. mode"), it is often convenient to make all the A multiple regression program developed by sample vectors equal in length. The sum of the the IBM corporation for the IBM 360 computer squared elements of a vector is equal to the as part of their Scientific Subroutine package variance or total information of the vector, was run on an IBM 360-50/75 system at and it is further convenient to give each speci- Columbia University. Multiple regression is men a variance equal to unity. This "normal- similar in all its essentials to standard bivariate ization by cases" is accomplished by dividing regression analysis, except that a hyperplane in each original observation on a specimen by the multivariate space is fitted to the data swarm square root of the sum of the squared obser- instead of the conventional straight line. How- vations on that specimen. ever, the object is still to predict one "depen- An alternative transformation is normaliza- dent" variable by one or more predictor, tion by variables, which gives each of the n "independent" variables, and the application variables a variance equal to one. Normaliza- of this technique to the study of ontogeny is tion by variables is of course accomplished in basically the same as it is in standard bivariate the same way as is normalization by cases, regression analysis. except that each element in a column vector is ONTOGENY OF THE CEPHALON OF PHACOPS RANA INTRODUCTION The general mode of growth in Phacops is THE ONTOGENY of trilobites may be sufficiently simple so that ignorance of the conven- ontogeny of individual iently divided into three periods. The protaspis specimens is not a is characterized by a serious problem. However, to distin- larval cephalon directly the inability articulated with a protopygidium. guish "trajectory" of the of an In the individual within ontogeny meraspis period, thoracic segments are added the data swarm creates serious progressively until the adult number is reached; difficulties in studies of the development of the anv individual with the lenses on the visual surface of the adult number of of the visual eye. Ontogeny thoracic segments belongs to the holaspis period surface is discussed in the present (see Whittington, 1957, 1959, for recent reviews paper only in conjunction with systematic oftrilobite ontogeny). problems. Analysis of ontogeny and statistical comparisons of populations of Phacops are limited to SOME GENERAL CONSIDERATIONS the holaspis period in the present study. Minute Ontogenetic allometry, in its broadest sense, specimens probably referable to the protaspis is any change in proportion during development period of P. rana are occasionally found, but (Gould, 1966). Thus are far too rare to enter into statistical allometry may occur even Lack analysis. though growth may be described by the of complete specimens where the number for a simple equation straight line, y=ax+b, where a ofthoracic segments is known precludes accurate is the and b the separation oflate slope intercept of the line. If b is meraspis "degrees" from early significantly different from the holaspid instars. between zero, proportion As in any two variables being compared will other arthropods, growth continues in as the postlarval (holaspis) stage. change growth occurs. As Shaw (1956) has Since onto- shown, the use genetic size increase proceeds in widespread of ratios in the discrete molt systematic literature of trilobites is undesirable stages, postlarval ontogeny of fossil arthropods as these must be ratios may change in Ratios studied on the population level are of ontogeny. collection of by course useful if comparable molt suites of specimens of different can be stages sizes. Thus the ontogenetic identified, although in practice the record of an indi- identification of discrete molts is vidual is lost, or at least unrecognizable, and difficult. Com- parison of ratios between two or more generalizations on ontogenetic processes are tions also popula- based on over-all population tendencies. may be appropriate if specimens of Fossil comparable size are used, the arthropods therefore occupy a somewhat inter- of although possibility mediate position between some groups phylogenetic or even geographically based size molluscs) (e.g., increase or decrease may invalidate the in which individuals retain a con- It is results. tinuous record of the ontogeny of hard by now apparent that, in terms of linear and other parts, dimensions of the groups (e.g., vertebrates) in which in exoskeleton, holaspid growth only the latest stage trilobites may usually be described a of ontogeny is generally linear by observable in any one specimen (Gould, 1966). simple equation. The power function, The ontogeny of y=bxk, commonly to both fossil vertebrates must be and applied ontogenetic studied on the basis of a suite of specimens phylogenetic cases ofallometry, is generally arranged according to size, but as smaller unnecessary to describe growth within the specimens, particularly of terrestrial species, holaspid period. Hunt's (1967) work on the commonly represent unsuccessful individuals agnostid Trinodus elspethi (Raymond) from the that died before reaching full Ordovician Edinburgh Saul's adult size, popu- and the Formation, (1967) lational ontogenetic analyses of fossil verte- author's unpublished data on homalo- brates are probably less truly representative of notids, Bright's (1959) study of the ontogeny of the actual ontogenetic process than are analyses the ptychopariid Elrathia kingii (Meek), Pabian based on molt series offossil arthropods. and Fagerstrom's (1968) work on Ameura sangamonensis (Meek and Worthen) and Whit- 65 66 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147 tington's (1957) analysis of Isotelus gigas DeKay PLOTTING AND BIVARIATE ANALYSIS are but a few examples where linear equations Bivariate plotting and regression analysis are adequate to describe holaspid growth. In show that cephalic growth in P. rana is linear addition, Waterman (1954) has shown that in and closely conforms to patterns seen in other terms of the major dimensions of both the trilobites. WVith the exception of certain linear prosoma and opisthosoma of Limulus polyphemus dimensions of the eye, bivariate scattergrans (L.) and Tachypleus tridentatus Leach, post- show a very tight unimodal distribution. A larval growth in the Xiphosura is not only single example, based on data on the crassi- linear, but essentially isometric. tuberculata eye variant from the Silica shale The degree of allometry present in holaspid ("SILC": UMMP loc. E), where cephalic ontogeny must be understood before comparisons length (CL) is plotted against the major between populations are made, as samples from reliable width measurement (WBVS) is shown different areas and formations commonly repre- in figure 6. The reduced major axis (Imbrie, sent different segments of ontogeny. Such 1956, p. 234) has an intercept of 1.45 units sampling biases could easily lead to spurious (=0.2 mm.), indicating that the proportion of results if allometry is a significant factor in the CL to WBVS undergoes little change. Schwim- postlarval growth of a tagma. Although Giles mer (MS) has come to a similar conclusion as a (1960, p. 373) claimed that allometric effects result of his study of P. rana from the Wanakah would be negligible in interpopulational com- shale ofwestern New York. parison of coyote samples, and that "testing for allometry would be a study in itself," cer- MULTIPLE REGRESSION tainly some of is degree testing necessary before Multiple regression analyses of the same results of statistical discrimination between population (SILC) of cephalic length (CL) populations are accepted. versus one set of 13 gross cephalic measurements (predictors) and another set of measure- QUALITATIVE HOLASPID ONTOGENY OF P. rana ments of the eye complex are summarized in No consistent changes in the ontogeny of tables 2 and 3. Absolute values of the standard ornamentation and other qualitative features partial regression coefficients (beta weights) for were noted in either P. rana or P. iowensis. As the predictors indicate the relative contribu- Hall and Clarke (1888, p. 23) pointed out, the tions of the corresponding variables to the pre- smallest P. rana specimens are in most respects diction of the criterion (CL). R2 is an estimate miniature adults. of the total variance of the criterion explained

mm21 CL 2C

14 13 12 11 I ic

WBVS s 6 7 8 9 10 11 12 13 14 15 16 17 IS 19 20 21 22 23 24 25 26 27 28 mni FIG. 6. Bivariate scattergram of cephalic length (CL, ordinate) vs. width between visual surfaces (WBVS, abscissa). Phacops rana crassituberculata, Silica shale, UMMP locality E. Reduced major axis shown by dashed line: CL=0.75WBVS+ 1.45, in original micrometerunits, where 1.45 units=0.2 mm. N-36. 1972 ELDREDGE: PHACOPS 67 TABLE 2 by the predictors,and its statistical significance is MULTIPLE REGRESSION OF 13 GROSS CEPHALIC also included in the tables. DIMENSIONS (PREDICTORS) AGAINST CEPHALIC LENGTH As is shown in table 2, 99.91 cent of the (CRITERION) a per variance in CL is explained by the 13 gross cephalic measurements. It is Variable Beta Weight that hardly surprising glabellar lengths, rather than widths, are GLIP 0.72 better predictors of cephalic length. What is of 1PL 0.15 interest here is that the hyperplane has an OCL 0.18 intercept of +0.45 units, equal to 0.06 mm. In GWCM -0.05 terms of glabellar lengths, at least, there is GWE -0.02 apparently no significant allometry in relation MIGW 0.08 to total Axial IPW cephalic length. widths are rela- -0.03 tively poor predictors of CL; it be GWIP 0.00 that although may MIIPA -0.01 suspected this is owing simply to the MIPA masking effects ofglabellar of 0.02 axial lengths, allometry OMIA 0.01 widths in relation to cephalic length cannot OMA 0.02 be ruled out entirely on the basis of this analysis. WOC -0.04 The various linear measurements of the eye explain 99.74 per cent of the variance of R2 0.99908 CL, and the intercept is 0.93 units, or 0.13 mm. Intercept 0.45013 units=0.06 mm. There is a p <0c01 (table 3). more even contribution of the variables to the prediction of CL, probably ap. rana crassituberculata (SILC) only (UMMP loc. E). reflecting the more homogeneous size range of N=30. the variables. There seems to have been little Svmbols: as in table 1. significant allometry between the entire complex and cephalic length. eye

FACTOR TABLE 3 ANALYSIS factor of the same MULTIPLE REGRESSION OF R-type analysis population 13 VARIABLES OF THE EYE clusters variables in COMPLEX (PREDICTORS) AGAINST CEPHALIC LENGTH (SILC) terms of specimens. The first factor extracted shows a (CRITERION) a of tight grouping most of the variables and is basically a simple Variable Beta Weight growth factor. Two other factors extracted yield more interesting results in rotated solution. WVS -0.02 When factors 2 and 3 are WPL plotted against factor 0.18 1, (figs. 7, 8) a very tight cluster ofmain LPA 0.02 width and cephalic WPAO -0.16 length measurements is apparent. WPA 0.16 However, measurements of the eye can be WBVS -0.23 discriminated from the main cluster by all WOP 0.25 three factors. The close association of gross L 0.37 measurements, including axial widths, along HVS 0.12 all three factors, is a strong indication that HEYE -0.03 allometry is insignificant in the ontogeny of the LVS 0.06 glabella and over-all LLVS 0.20 cephalic dimensions. PEG 0.13 Although multiple regression indicated little allometry between the eye complex as a whole R2 0.99736 and cephalic length, it does appear that the Intercept 0.93036 units=0.13 mm. eye complex maintains a measure of independ- p <0.01 ence from the axial portion of the cephalon during ontogeny. of variables ap. rana crassituberculata (SILC) only. N= 30. Interrelationships Symbols: as in table 1. within the eye complex are discussed in a forth- coming paper (Eldredge, Ms). 68 BUlLLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147

I I I^ el LLVS ;. A'

,.a c i *d w.t

l CW1p WO * LOP 6

*HGL

NFAC WVS. El 1 2 3 4 5 6 7 8 9 FIG. 7. R-mode factor analysis (rotated solution) plot offactor I (ordinate) vs. factor II (abscissa) for P. rana crassituberculata cephala, Silica shale, UMMP locality E. N=28. Two groups, A and B, are outlined by dashed lines: Group A consists ofmajor linear dimensions ofthe cephalic axial structures; Group B: a - HVS; b - HEYE; c - WPA; d - LPA; e - WPAO; f- WPL. Other abbreviations as in table 1.

II Several variables show rather high inde- -NFAC pendence. For example, factor 2 seems to measure diagenetic deformation. The width of -8 the visual surface (WVS) and width of the ocular platform (WOC) could easily have been distorted by deformation of the genae, struc- turally the weakest portion of the cephalon. -7. Measurements made proximal to the distal margin of the palpebral lobe seem far less prone to variation induced by deformation. A -6 few variables (e.g., MI1 PA, Ml PA) are omitted because of incorrect results caused by errors in -HGL the original data matrix.

-5 -LOP CONCLUSIONS it'% t I X. !,I Postlarval growth in Phacops, as in other -4 ipL trilobites and living xiphosurans, is linear and r, : % can be adequately described by the equation IL for a straight line. Furthermore, allometry, . I I while present between certain -3 .Cw1p -' variables, is *WOP .WPL generally sufficiently negligible that populations .L may be compared even when biased in terms of .WVS :L portions ofthe ontogenetic sequence represented. -2- i 3 4 5 6 7 8 The cephalon is basically well "integrated," reflecting its nature as a solid tagma. The FIG. 8. R-mode factor analysis (rotated solutior ocular complex, however, maintains a degree plot of Factor III vs. a) (ordinate) Factor I (abscissg of for P. rana crassituberculata cephala, Silica shall independence from the major cluster of UMMP locality E. N=28. Two groups, I and I gross and axial dimensions. are outlined by dashed lines. Group I consists ofHV' LPA, HEYE, and WPA; Group II consists of maj(or cephalic dimensions. Other abbreviations as iin table 1. INTERPOPULATIONAL VARIATION IN PHACOPS RANA

THE VERY CHARACTERS that may be used to slope, specimens from the Taghanic of the west differentiate P. rana from P. iowensis are the (Milwaukee Dolomite, Cedar Valley Forma- ones that show the greatest amount of inter- tion) show almost complete loss of flattening populational variation within the P. rana com- and elongation, retaining the elongate tubercles plex as a whole. This fact suggests a significant only on the anterior margin of the glabella amount of interaction between the two species, and on the posteriormost row of tubercles on and that consequently their histories cannot the occipital lobe. properly be understood separately. The details of the apparent interactions between these two GENAL ANGLE species are discussed following ageneral discussion The genal angle is situated on the ventral of interpopulational variation within each of the lateral margin of the cephalon in lateral view species. with the dorsal margin of the visual surface in The interspecific differentia discussed pre- the horizontal plane. In Cazenovian samples of viously included aspects of the eye complex P. rana the genal angle is generally quite acute, (particularly number of dorsoventral files), whereas the genal angles of Tioughniogan and shape and distribution of ornamental tubercles, Taghanic samples are generally more rounded. genal angle morphology, and pygidial characters. In addition size and shape differences of OCULAR PLATFORM various anatomical areas of the exoskeleton All Cazenovian have shallow facial which emerged in factor samples analysis may be added. suture furrows and a relatively flat librigenal The facial suture ORNAMENTATION moiety. furrow is deeper and there is a tendency for the librigenal moiety to On the cephalon of P. rana the low, rounded become vertical and merge imperceptibly with tubercles become elongated transversely, often the area under the visual surface in specimens into ridges, on the occipital lobe, genae, and from the This condition is particularly on the Tioughniogan. anterior margin of the com- developed to an extreme in specimens of P. rana posite glabellar lobe. This elongation into from the Taghanic. ridges is progressive, so that at the cephalic margins, the ridges are similar to those on the THE PYGIDIUM doublure and hypostoma. The is a solid that Within the P. rana pygidium tagma undergoes complex, there is a spec- little postlarval ontogenetic change in qualitative trum ranging from a strong development of characteristics and there is interpopulational flattened and elongated tubercles, to a virtual variability in only a few characters. absence of this feature. In Cazenovian samples, Large, gently rounded tubercles generally flattening and elongation of the tubercles begins cover the axis and on the proximal regions of the horizontal posterior surface of the pleura in Cazenovian P. rana and become obso- composite glabellar lobe, and increases in lescent near the intensity margin. Larger specimens of the anteriorly down the frontal slope to the milleri eye variant often show incipient obso- anterior cephalic margin. Virtually all but the lescence over the most entire thorax and pygidium. anterior tubercles on the occipital lobe of P. rana are similarly Specimens from the Taghanic, par- transversely elongate. This state is ticularly those from the Cedar Valley and Mil- most extremely developed in the populations waukee of the formations, frequently show smaller, Cazenovian. In Tioughniogan samples, more conical tubercles flattening begins distributed over the farther down the anterior entire surface of the the slope of the glabella and is less extreme at the pygidium, including anterior cephalic posterior and lateral margins. Strength and margin. Although specimens distribution of tuberculation is variable of P. rana from the Taghanic (Tully) of New quite between populations of P. rana throughout the York retain this elongation on the anterior Tioughniogan. 69 70 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147 All but the first interpleural furrows are limestone), Indiana (Deputy and Silver Creek generally obsolescent in P. rana. Cazenovian limestones), Ohio (Silica shale), and south- samples, including P. rana paucituberculata, in western Ontario (Arkona shale). In addition, many cases have the anterior interpleural they are known from float material in Iowa and furrow developed as a faint transverse groove, from lower Hamilton rocks in the southern commonly set off by two parallel rows of tuber- Appalachians. Two eye variants primarily cles. Tioughniogan specimens display the inter- involving differences in number of lenses per pleural furrow less frequently, and this feature dorsoventral file, have led to recognition of two is generally absent in Taghanic samples. Degree subtaxa within the 18 dorsoventral file group: of incision of the pleural furrows and degree P. rana milleri Stewart (fig. 14), and P. rana of arching of the pleura seem to converge with crassituberculata Stumm (fig. 13). The number of the interpleural furrow. Shallow furrows and lenses per dorsoventral file is a component of relatively flat pleura characterize most Caze- interpopulational variation not encountered novian populations. Tioughniogan P. rana, again in the subsequent history of P. rana, and again, are variable, but Taghanic samples the general nature of this geographic variability show deeply incised pleural furrows and more and the validity ofP. r. crassituberculata and P. r. highly arched pleura. milleri as distinct subspecies are discussed below. The ontogeny of the eye of populations with INTERPOPULATIONAL VARIATION IN 18 dorsoventral files suggests that the first NUMBER OF DORSOVENTRAL FILES (anterior) file is added last in ontogeny. This As in intrapopulational and ontogenetic file is reduced to a single lens in the one speci- variability, much of the variation seen on the men known from the lower shale of the Solsville interpopulational level within P. rana involves member of the Marcellus (AMNH loc. 3013) of eye morphology. The number of dorsoventral New York (fig. 13D). A small sample from the files per eye is particularly significant. As noted overlying sandstone unit ofthe Solsville (AMNH previously, the normal adult complement of loc. 3013) shows great variation in the develop- dorsoventral files is virtually invariant within ment of the first file. One specimen has the a population. Smallest holaspid cephala ob- file fully developed, whereas in others it is served have but two or three fewer dorsoventral partially developed or totally absent. The files than the adult number, and the normal oldest Phacops-bearing unit above the Solsville adult complement is quickly reached and in New York is the Stafford limestone, which is stabilized. Variations in dorsoventral file num- either uppermost Marcellus or lowermost ber within some populations are not uncommon Skaneateles in age. All Phacops from this for- but are rarely symmetrical. At least one of the mation have 17 dorsoventral files, and all 17 eyes on any specimen is likely to be normal for file populations of P. rana were limited to the the population. Only two populations have been eastern exogeosyncline throughout the duration encountered that exhibit significant variation ofthe Cazenovian. As the holotype ofP. rana has in this feature. 17 dorsoventral files, specimens with the dorso- The oldest specimens of P. rana known occur ventral file complement may be referred to P. as rare specimens in the Dundee and Delaware rana rana. limestones of Ohio and in the Solsville member The widespread Centerfield (basal Tiough- of the Marcellus Formation in central New niogan) fauna marks the disappearance ofall 18 York (fig. 1 3D). They are uniformly character- dorsoventral file P. rana and the first incursion of ized by 18 dorsoventral files, a characteristic 17 dorsoventral P. rana west ofthe exogeosyncline. they share with their European relative, P. Phacops rana has 17 dorsoventral files throughout schlotheimi. Phacops rana populations with 18 the entire Tioughniogan (fig. 4). dorsoventral files persist into the upper Caze- Another dorsoventral file number variant novian Skaneateles equivalents generally west appears in the Taghanic. Again, a variable of the exogeosyncline of New York and the population occurs in the exogeosyncline of the Appalachians. They occur in Skaneateles equiv- east. Specimens from the Apulia member of alents in Michigan (Bell shale, Rockport the Tully limestone of central New York show a Quarry limestone, Ferron Point Formation, range of 15 to 17 files. A collection from AMNH Genshaw Formation), Illinois (St. Laurent locality 3039 (see fig. 15D), for example, has pre- 1972 ELDREDGE: PHACOPS 71 dominantly 15 or 17. Collections from other contribution of each variable to the discrimina- localities (USNM 18, 19; AMNH 3040, AMNH tion. 4) in the Tully show the same variation. Two different data transformations were Loss of dorsoventral files is evidently pedo- used to study the interrelationships of samples; morphic in nature, as the larger specimens in the the mathematics of these transformations were transitional populations in both the Marcellus explained previously. The effects of these trans- and Tully of New York tend to reach the more formations remain to be considered. Normaliza- primitive complement of dorsoventral files. tion by cases converts each specimen (sample) Addition of the final dorsoventral file(s) is vector to unit length and thus minimizes the seemingly retarded into progressively later effects of ontogeny and size differences in stages ofontogeny, until they are ultimately lost. samples resulting from collecting bias. The Phacops rana with 15 dorsoventral files from factor components, however, reflect the unequal the Taghanic are all referable to P. rana norwood- magnitude of the variables; the larger variables ensis Stumm, 1953. This subspecies occurs in have larger absolute variances and thus domi- most Taghanic formations known (Upper nate over smaller variables. The net effect is Cedar Valley of Iowa; Milwaukee Dolomite, that the contributions of smaller variables to Wisconsin; Upper Petoskey, western Michigan; population discrimination are ignored in the Tully limestone, New York). No P. rana are analysis. known from the Taghanic of eastern Michigan Normalization by variable gives all variables or northern Illinois. a variance equal to unity. Sample vectors are The one specimen of P. rana of possible then of unequal length and reflect the absolute Chemung age available ("P. nupera" Hall, size of the specimens. The result is to magnify AMNH 496911) is poorly preserved, but has the effects of ontogeny and for factor coefficients 17 dorsoventral files. Apparently, then, the main to reflect the absolute size of the specimen. stock ofP. rana rana survived for a while in post- Differences in mean sample length vectors Tully times. (reflecting sampling biases) could therefore Five specimens of P. rana from the Caze- lead to incorrect discrimination if there were novian of Ohio and Michigan have 15 dorso- significant allometry in the ontogeny of Phacops. ventral files. Four are from the Plum Brook Normalization by cases following normalization shale of Ohio (UMMP loc. D) and a single by variables removes the problem, although specimen is known from the upper Ferron Point allometry is insignificant in Phacops, and popula- shale (USNM loc. 2). Although these five tions can be successfully compared no matter specimens have but 15 dorsoventral files, in what size range and mean sample vector differ- terms of ornamentation and other morphology ences exist between samples. Two different they are more similar to the contemporaneous analyses, normalizing the data by variable, P. rana with 18 files than they are to P. rana one with and one without subsequent normaliza- norwoodensis (15 files) from the Taghanic. tion by cases, were performed. The two differed little; only the former is presented here. FACTOR ANALYSIS AND NORMALIZATION BY CASES ONLY INTERPOPULATIONAL VARIABILITY A matrix normalized by cases consisting of The form of Factor Analysis used to analyze 329 specimens and 30 variables yielded two intersample relationships of both cephala and significant factors in rotated solution. Com- pygidia factors the transformed data matrix ponents of the two significant rotated factors into a matrix of factor coefficients, which are in are given in table 4. Factor 1 essentially consists effect scores the of each individual specimen on of high positive loadings on all gross cephalic each reference vector. Each reference vector, in measurements and a high negative loading on turn, consists ofthe score, or relative weight, ofthe total lens number. This means that a specimen contribution of each of the individual variables with a high positive coefficient for factor 1 has to the definition of the vector. Samples thus have relatively fewer lenses per given cephalon size a particular distribution when factor coefficients than does a specimen with a lower coefficient. are plotted against a reference vector, and the Factor 2 is merely the mirror image of factor 1, components of the vector reveal the relative i.e., specimens are discriminated on lens number 72 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147 TABLE 4 factor plotted against the stratigraphic Positions FACTOR COMPONENTS, CEPHALIC DATA NORMALIZED of the samples. Such a diagram is not intended BY CASES to be read as a simple procession ofphylogenetic (Variables are defined in Table 1) events, but merely as a means of assessing variation among quasicontemporaneous sam- Variable Factor 1 Factor 2 ples. Sample abbreviations and size are given GLIP 0.238 0.109 in table 5. The nine Cazenovian samples lpL 0.032 0.012 plotted show a considerable range in variation. OCL 0.044 0.026 CARD represents the early, variable popula- GWCM 0.343 0.149 tion, with both 18 and 17 dorsoventral files. All CL 0.304 0.157 other samples have 18 dorsoventral files GWE 0.313 0.113 except PLUM (15) and STAF (17). Since MIGW 0.177 0.062 high loadings on factor 1 indicate relatively few lpW 0.169 0.057 lenses per cephalon size, it is apparent that the CW1P 0.093 0.035 MIIPA 0.106 0.027 main variation within the Cazenovian in total MIPA 0.148 0.058 lens number is based on number of lenses per OMIA 0.116 0.047 dorsoventral file rather than number of OMA 0.162 0.063 dorsoventral files. Discrimination is therefore WOC 0.191 0.154 based on the milleri and crassituberculata eye WVS 0.017 0.013 variants (both with 18 dorsoventral files), WPL 0.028 0.033 discussed below. LPA 0.087 0.060 In terms of general eye morphology, the 17 WPAO 0.078 0.016 dorsoventral file population (STAF) is more WPA 0.047 0.037 similar to the WBVS 0.450 0.156 crassituberculata variant (e.g., LOP 0.027 -0.012 SILC, SCRK), and figure 9 shows that P. rana WOP 0.134 0.075 samples above the Cazenovian apparently are L 0.116 0.028 closer morphologically to P. rana with the HGL 0.107 0.046 crassituberculata eye variant. Typical crassitubercu- HVS 0.055 0.048 lata show very high scores, i.e., have relatively HEYE 0.100 0.047 fewer lenses per cephalon size than typical LVS 0.093 0.127 milleri-variant populations (e.g., SILM, ARKA). LLVS 0.060 0.081 Variation is considerably compressed and PEG 0.137 0.043 dispersion relatively uniform throughout the 4 FAC -0.383 0.915 Tioughniogan history of P. rana. In the upper Tioughniogan there is an increase in lens number per cephalon size which amounts to a per cephalon size, but here a high loading indi- convergence on the earlier milleri eye variants cates a relatively large number oflenses. In factor of the Cazenovian. Clarkson's specimen (1966b, 1, the relative importance of each variable seems p. 468) of a "P. rana milleri" from western New directly proportional to its absolute magnitude, York probably came from a unit such as the but some smaller variables are also important Windom member of the Moscow Formation in factor 2. Specifically, the length of the visual (WINC). Figure 9 shows that the eye mor- surface (LVS) is positively related to lens phology of the 17 dorsoventral file specimens number (#FAC). The length of the ocular from upper Tioughniogan formations are sec- platform from the rear margin of the eye to the ondarily convergent on the 18 dorsoventral posterior border furrow (LOP) is negatively file milleri eye variants. related to lens number. In general, normaliza- The variation shown by quasicontempora- tion by cases allows discrimination on the basis neous populations within the Tioughniogan seems of over-all eye size, particularly length and total to be correlated with gross sedimentological lens number, all at a given over-all cephalic features of the various formations. Most Caze- size. novian formations that produce milleri-eye Figure 9 shows the distribution of mean variant specimens in abundance are highly sample factor coefficients for the first rotated calcareous shales. Throughout the Tioughnio- 1972 ELDREDGE: PHACOPS 73

TAGHANIC "MILD . CRVR *"TULL

U. TIOUGHNIOGA *WINC oWINE

M. TIOUGHNIOGA * WGRB * WKHJ *LHCH .WIDD

L. TIOUGHNIOGA * CBUF *CEBY *HHWR * CFAY

U. CAZENOVIA .*PLUM *ARKA *SILM *SILC OGNSW * fPTM *SCRK |*STAF

*CARD L. CAZENOVIA

.55 .60 .65 .70 .75 .80 .85 FIG. 9. Factor analysis of P. rana cephala. Data normalized by samples. Mean score for each sample for factor 1 (rotated) plotted against approximate stratigraphic position. Sample abbreviations as in table 5. Factor components are given in table 4. gan lineage, samples with the highest mean limestone in western New York, which scores scores on factor 1 occur in more clastic rocks. rather low, and WIDD, representing a sample WINE and LHCH are samples from the from the Widder Formation of Ontario (UMMP Moscow and Ludlowville formations respec- 4993 and 27066). The Widder is a variable tively from the Chenango Valley in New York. sequence of calcareous shales and limestones; Both formations are variable shale-siltstone the specimens are very similar to those collected sequences with a low calcareous content. Speci- by the author from the underlying Hungry mens from purer limestones (CEBY, HHWR) Hollow Formation, which is a Centerfield also tend to score relatively high. Samples from equivalent. The Hungry Hollow was named by fine-grained argillaceous sediments with a high Cooper and Warthin (1941), and includes 6 calcareous content (e.g., Tioughniogan shales feet of shale and limestone formerly assigned from western New York-WINC, WGRB) to the Widder Formation. The UMMP collec- similar to milleri-bearing rocks of the Cazeno- tion was made prior to recognition of the Hun- vian, show relatively lower loadings, i.e., a gry Hollow, and in view of the identical scores greater proportional number of lenses per ceph- of WIDD and HHWR, there is perhaps some alon size. justification in considering the UMMP Widder The two major apparent exceptions include sample as having come from the Hungry Hollow. CBUF, a population from the Centerfield There is an abrupt break between specimens 74 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147 TABLE 5 FACTOR ANALYSIS PLOTS Sample Formation Locality N1 N2 No. Dorsoventral Files Taxon ALPI Alpena limestone AMNH 3059 10 10 13 Pia APPM Hamilton s. Appalachians 3 3 18 Prm APPR Hamilton s. Appalachians 2 2 17 Prr ARKA Arkona shale UMMP A, B 17 15 18 Prm BELI Bell shale UMMP 31 1 1 13 Pia CARD Cardiff AMNH 3013 5 4 17-18 Prc-Prr CBUF Centerfield AMNH 3047 9 9 17 Prr CEBY Centerfield AMNH 3044 13 13 17 Prr CFAY Centerfield AMNH 3033 6 6 17 Prr CRVI Cedar Valley SUI 4 3 3 13 Pii CRVM Cedar Valley float 1 1 18 Prm CRVN Cedar Valley SUI, various 6 5 15 PM CRVR Cedar Valley SUI, various 3 3 17 Prr DSCI Dock Street UMMP 53 1 1 13 Pii FDMI Four Mile Dam AMNH 3061 3 3 13 Pii FPTM Ferron Point UMMP IA 5 5 18 Prm GNSWI-3 Genshaw USNM 3 3 3 13 Pia GNSW4-5 Genshaw USNM 3 2 2 18 Prm GPTI Gravel Point AMNH 3060 2 2 13 Pia HHWI Hungry Hollow AMNH 3054 1 1 13 Pis HHWR Hungry Hollow AMNH 3054 5 5 17 Prr LHCH Ludlowville AMNH 3031 11 11 17 Prr MILD Milwaukee UMMP G 11 11 15 PM NRPT Norway Point UMMP III 1 1 13 Pii PLUM Plum Brook UMMP D 4 4 15 Prp SCRK Silver Creek AMNH 6, 6A 13 13 18 Prc SILC Silica Shale UMMP E 28 28 18 Prc SILM Silica Shale UMMP E 35 34 18 Prm STAF Stafford NYSM 2 13 13 17 Prr THBY Thunder Bay UMMP 35 1 1 13 Pii TULL Tully AMNH 3039 16 16 15-16-17 Prr-Prn UPET Upper Petoskey UMMP 7b 2 1 15 PM WGRB Wanakah AMNH 3049 28 27 17 Prr WIDD Widder UMMP A, B 7 7 17 Prr WINC Windom AMNH 3052 20 18 17 Prr WINE Cooperstown AMNH 3034 15 15 17 PIT WKHJ Wanakah AMNH 3041 23 23 17 Prr Totals 329 320 Symbols: N1, number ofspecimens per sample, normalizing by cases. N2, number of specimens per sample, normalizing by variable, then by cases. Pia, P. iowensis alpenensis; Pii, P. iowensis iowensis; Pis, P. iowensis southworthi; Prc, P. rana crassituberculata; Prm, P. rana milleri; Pm, P. rana norwoodensis; Prp, P. rana paucituberculata; Prr, P. rana rana. with a relatively large number of lenses per NORMALIZATION BY VARIABLES cephalon size in the Upper Tioughniogan, and AND BY CASES those with relatively fewer in the Taghanic. A matrix normalized first by columns (var- There is little variation along factor 1 among the iables), then by rows (cases) consisting of 329 15 dorsoventral file Taghanic samples. The specimens and 28 variables yielded four signifi- sudden displacement toward a high loading is cant factors in rotated solution. In this analysis, partially explained by the fact that most the width of the visual surface and the total specimens in the Taghanic have but 15 dorso- number of lenses were omitted. The third and ventral files. fourth factors, although representing a very 1972 _SELDRRDTR-SFTF ,-P-[ArnP1 1-auvX 1) 75 small percentage of the total variance of the graphically as the mean scores of all samples of matrix, are considered "significant" by virtue P. rana fall within a narrow range. The only of the seemingly important biological informa- samples that are effectively discriminated by tion contained in the vector equations and the this factor are: APPM, APPR, LHCH, SCRK, apparent "sense" made by the relative scores STAF, WINE, HHWR, and WIDD. Of these, of the various samples. In a previous study the first six consist of exfoliated (internal mold) (Eldredge, 1968) a similar case was presented specimens where LOP is increased in which a to artificially single factor, explaining a very small owing the absence of exoskeleton at the rear amount of the variance of the matrix, had margin of the eye. WIDD and HHWR have important consequences in the analysis of much lower scores (i.e., greater value of convergence between two gastropod species. than any other nonexfoliated LOP) The sample of P. rana. vector equations (factor components) for The reasons for this apparent is the analysis are given anomaly in table 6. Factor 1 is discussed when the interactions between P. rana based largely on the character LOP-i.e., the and P. iowensis are considered below. distance from the rear of the eye to the posterior Factor 2 seems at first to be a border furrow glance "mirror measured exsagitally. High image" of factor 1, where a high score indicates scores of samples on this vector indicate a a value for LOP smaller relatively high in relation to value for LOP in relation to total total cephalon size. Mean scores of cephalon size. This samples of factor is not illustrated P. rana are plotted according to approximate stratigraphic position in figure 10. The scores TABLE 6 for APPM, APPR, LHCH, SCRK, STAF, and FACTOR WINE have been omitted for the reason COMPONENTS, CEPHALIC DATA NORMALIZED above. But given BY VARIABLES, THEN BY CASES in factor 2 there does seem to be (Variables are defined in Table 1) important interpopulational variability in the relative size of LOP. The range of variation is Variable Factor 2 Factor 3 Factor 4 greatest, again, in the Upper Cazenovian. Tioughniogan and Taghanic GLIP 0.117 0.044 -0.178 in much samples show, IPL 0.125 -0.126 general, less variability. Tioughniogan 0.004 are closest OCL 0.087 0.067 -0.152 samples to the upper end of the GWCM 0.114 0.112 -0.023 Cazenovian spectrum, whereas four of the five CL 0.110 0.032 -0.095 Taghanic samples fall to the lower end. The GWE 0.136 0.086 0.111 similarity in the pattern of variation among MIGW 0.124 0.073 0.175 samples shown by this factor with the pattern IPW 0.130 0.059 0.154 seen in factor 1 CW1P 0.174 0.103 (normalizing by cases; fig. 9) 0.423 indicates that the variation in LOP size to a MI1P 0.145 0.086 0.367 is, MIPA degree, correlated positively with lens number 0.115 0.058 0.240 per size. The OMIA 0.115 0.119 0.162 cephalon only discrepancy is in the OMA Taghanic samples where the absolute number 0.108 0.064 0.149 of lenses WOC 0.129 0.051 0.121 decreases but LOP remains relatively WPL -0.004 0.002 -0.208 small. This suggests that the loss of the two LPA 0.101 0.062 -0.105 dorsoventral files in the Taghanic samples WPAO 0.171 0.046 -0.128 probably was compensated by greater thickness WPA 0.081 0.105 -0.206 of interlensar scleral rather WBVS 0.106 tissue than by 0.055 -0.045 reduction in absolute eye size. the LOP 0.800 -0.041 -0.175 Finally, high WOP 0.099 (and close) scores of HHWR and WIDD repre- 0.123 -0.217 sent a real L 0.193 -0.036 0.013 variation best explained with refer- HGL 0.095 -0.571 0.339 ence to the interactions between P. rana and P. HVS 0.036 -0.505 -0.140 iowenszs. HEYE 0.072 -0.510 -0.180 Factor 3 (fig. 11) shows a pattern of inter- LVS 0.009 0.118 -0.161 populational variation similar to that LLVS exhibited 0.007 0.112 -0.201 by factor 2, although the amount of variation PEG 0.149 0.017 -0.094 among quasicontemporaneous samples is apparently more constant at different periods of 76 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147

TAGHANIC UPET.MILD-TL*CRVN *TULL . CRVR

U. TIOUGHNIOGA *WINC

M. TIOUGHNIOGA *WGRB*WKHJ

* WIDD

L. TIOUGHNIOGA sCBUF *CFAY"CEBY *HHWR

*PLUM

*ARKA *SILM *SILC * FPTM U. CAZENOVIA

*CARD L. CAZENOVIA

.55 .60 .65 .70 .75 FIG. 10. Factor analysis ofP. rana cephala. Data normalized by variables, then by cases. Mean score for each sample for factor 2 (rotated) plotted against approximate stratigraphic position. Sample abbreviations as in table 5. Factor components are given in table 6. time in factor 3. Here again, the characters the cephalon are varying together; the explana- underlying this variation are largely related to tion for the covariance of total eye height and eye morphology. In factor 3 a high score indi- visual surface height with glabellar height is cates a small (short) visual surface height and not clear. total eye height, as well as a small glabellar The fact that the patterns of interpopula- height, in relation to total cephalon size. That tional variation are similar for both factors 2 and is to say, all vertical (height) measurements of 3 again suggests they are related, and that 1972 ELDREDGE: PHACOPS 77

TAGHANIC * UPET .MILD . TULL *CRVN CRVR

U. TIOUGHNIOGA WINC WINE

M..TIOUGHNIOGA *WKHJ .WGRB eLHCH

* WIDD

L. TIOUGHNIOGA * CBUF * CFAY *CEBY* HHWR

*PLUM U. CAZENOVIA i *SILCFPT

STAF * SCRK

*CARD

L. CAZENOVIA

-.06 -.04 -.02 0.0 .02 .04 .06 .08 FIG. 1 1. Factor analysis ofP. rana cephala. Data normalized by variables, then by cases. Mean score for each sample for factor 3 (rotated) plotted against approximate stratigraphic position. Sample abbreviations as in table 5. Factor components are given in table 6. specimens with small LOP values per cephalon tissue, and not by reduction in height ofthe eye. size also tend to have a more inflated glabella The equation for factor 4 (table 6) is per- and, in particular, taller eyes in relation to haps the most interesting of the entire analysis, total cephalon size. In other words, factors 2 as it effectively segregates into three distinct and 3 show different aspects of interpopulational groups all length and width measurements and variation in total eye size in P. rana. Most measurements related to the eye. Very simply, a Cazenovian populations, especially those of the high mean population sample score on factor 4 milleri eye variant, tend to have a relatively (fig. 12) indicates a broader, shorter cephalon larger number of lenses per dorsoventral file (in terms of axial measurements) with a smaller than other samples, and it is not surprising that eye complex. Here glabellar height varies posi- their eyes are relatively taller than those of most tively with axial widths. Again, the spectrum of Tioughniogan samples. However, the milleri variation seems widest in the Upper Cazeno- sample from the Arkona shale of Ontario vian. Variability is also extensive among "Cen- (ARKA) has an anomalously high score in terfield" (basal Tioughniogan) populations, factor 3. Again, the low scores of the 15 dorso- but somewhat less so in the Middle Tioughnio- ventral file samples in the Taghanic imply that gan (but see WKHJ). Here again, the scores loss of two dorsoventral files was compensated of the Tioughniogan samples fall at one end by an increase in thickness of interlensar scleral of the spectrum of variation among Cazenovian 78 BULLETIN AMERICAN ]MUSEUM OF NATURAL HISTORY V'OL. 147

TAGHANIC *MILD *UPET *CRVN . TULL .CRVR

U TIOUGHNIOGA *WINE WINC

NI. TIOUGHNIOGA *LHCH *WGRB *WKHJ

*WIDD

L. TIOUGHNIOGA *CBUF *CFAY * HHWR *CEBY

PLUM U. CAZENOVIA *SILM *SILC *ARKA M SCRK *FPTM *STAF

*CARD

L. CAZENOVIA

-04 -02 0.0 02 04 06 08 FIG. 12. Factor analysis of P. rana cephala. Data normalized by variables, then by cases. Mean score for each sample for Factor 4 (rotated) plotted against approximate stratigraphic position. Sample abbreviations as in table 5. Factor components are given in table 6. samples. Cazenovian samples tend to have SYSTEMATIC PALEONTOLOGY larger eyes, and narrower, longer cephala, Four subspecific nomina (P. rana crassitubercui- than Tioughniogan samples; two samples from lata, P. rana milleri, P. rana rana, and P. rana the Taghanic (MILD, UPET), fall within the norwoodensis) were somewhat casually introduced range of some of the Upper Cazenovian sam- during the discussion of interpopulational ples, but the remainder of the Taghanic sam- variability in dorsoventral file number. These ples are similar to those of the Tioughniogan. subtaxa, plus a fifth yet to be introduced, occupy In summary, factor analysis has added a few definite subsets of the spectrum of variation in linear measurements to the list of important the other characters discussed. These differ- interpopulational variables in P. rana. These entiating characters are set forth formally in are: (1) number of lenses per cephalon size, the following taxonomic diagnoses. (2) LOP, (3) HVS, HEYE, HGL, and (4) clusters of axial lengths, axial widths, and eye Phacops rana crassituberculata Stumm, 1953 taken size together. It is clear, then, as in the Phacops rana crassituberculata STUMM, 1953, p. 136-137, qualitative analysis presented above, that it is PlS. 9, 10. the various aspects of eye morphology that show Phacops rana crassituberculata: CLARKSON, 1966b, p. 470, the most significant amount of interpopula- pl. 73 tional variability in P. rana. Figure 13A-D 1972 ELDREDGE: PHACOPS 79 EMENDED DIAGNOSIS: Eye with 18 dorso- anteriorly, tubercles becoming wavy ridges at ventral files in normal adults. Maximum num- anterior cephalic margin. Tubercles of occipital ber of lenses per file six, rarely seven. Interlensar lobe similarly transversely elongate. Tubercu- sclera well developed, hexagonal; lenses gener- lation only moderately developed over genae. ally flush with sclera except on the ventral Tubercles usually small, occasionally sharply portion of the visual surface in immature pointed, on palpebrum. specimens. Trace of facial suture over ocular platform Tubercles large and bluntly rounded on quite shallow. Genal angles bluntly pointed. composite glabellar lobe, decreasing in size Pleura of pygidium rather flat; pleural furrows and becoming transversely elongate posterior rather shallow. Anterior interpleural furrow to anterior glabellar slope. Transverse elonga- occasionally developed. tion of tubercles becoming increasingly extreme HOLOTYPE: UMMP 25537. STRATIGRAPHIC DISTRIBUTION: Cazenovian. The distribution of this subspecies is given in table 8. Phacops rana milleri Stewart, 1927 Phacops rana var. milleri STEWART, 1927, p. 58, pl. 5. Phacops rana var. milleri: DELO, 1940, p. 23-24, pl. 1. Phacops rana milleri: STUMM, 1953, p. 137-138, pls. 9, 10. Phacops rana arkonensis STUMM, 1953, p. 138-139, pl. 10. A non Phacops rana milleri CLARKSON, 1966b, p. 468-470, pl. 73. Figure 14A-D EMENDED DIAGNOSIS: Eye with 18 dorsoventral files in normal adults. Maximum number of lenses per file nine. Horizontal interlensar sclera variably developed, usually in upper one-third of each dorsoventral file only. Lenses generally protruding from scleral surface in all but dorsalmost rows. Remainder of cephalic features as in P. rana crassituberculata, except tuberculation more weakly developed, and more rounded on palpebrum. Thorax and pygidium as in P. rana crassituberculata except tubercles less strongly developed, approaching obsolescence. SYNTYPE: Department of Geology, Ohio State University 16266 (not examined). STRATIGRAPHIC DISTRIBUTION: Cazenovian. The distribution of this subspecies is given in table 8.

Phacops rana norwoodensis Stumm, 1953 Phacops rana norwoodensis STUMM, 1953, p. 140, pl. 12. FIG. 13. Phacops rana crassituberculata Stumm, 1953. A-C. Silica shale, AMNH locality 3063, AMNH Figure 15A-D 28898. A. Dorsal view of cephalon, x 2. B. Left DIAGNOSIS: with rarely 16 3. C. Frontal view of EMENDED Eye 15, lateral view of cephalon, X files. Trace of facial suture cephalon, x2. D. Right lateral view of cephalon, or 17, dorsoventral Solsville Member, AMNH locality 3013, AMNH across ocular platform moderately deeply 29119, x3. incised. 80 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 1 47

C D C FIG. 15. Phacops rana norwoodensis Stumm, 1953. A. Dorsal view of cephalon, paratype, Milwaukee dolomite, UMMP locality G, UMMP 14727, x 3. B. Oblique lateral view of cephalon, holotype, Upper Petoskey Formation, UMMP locality 7b, UMMP 25524, x 3. C. Dorsal view of pygidium, Brandon Substage, Cedar Valley Formation, Belanski collec- D tion, SUI locality 183-5, SUI 6427, x 2. D. Right lateral view of cephalon, Tully Formation, AMNH FIG. 14. Phacops rana mitteri Stewart, 1927. Silica shale, AMNH locality 3063. A-C. AMNH 28896. A. locality 3039, AMNH 28899, x 3. Dorsal view of cephalon, x 2. B. Left lateral view of cephalon, x 3. C. Frontal view of cephalon, x 2. D. STRATIGRAPHIC DISTRIBUTION: Taghanic. The Dorsal view ofpygidium, AMNH 28897, x 3. distribution of this subspecies is given in table 8. Phacops rana paucituberculata, Elongation of new subspecies cephalic tubercles usually Figure 16A-C restricted to anterior glabellar margin, genal margins, and posterior row of tubercles on DIAGNOSIS: Eye with 15 dorsoventral files in occipital lobe. Tubercles variably densely or normal adults. Tubercles extremely sparse sparsely distributed over composite glabellar over composite glabellar lobe, except in single lobe. Genal angles as in P. rana rana. known specimen from outside type area. Pleural furrows deeply incised and pleura Remainder of features as in P. rana crassituber- highly arched for species. culata. HOLOTYPE: UMMP 25524. HYPODIGM: Four specimens from the Plum 1972 ELDREDGE: PHACOPS 81 Brook shale of Ohio, (UMMP loc. D). UMMP developed partially up front slope of glabella, 25538, 28849, 28850, and 57135 (holotype). becoming progressively more elongate approach- STRATIGRAPHIC DISTRIBUTION: Cazenovian. ing anterior cephalic margin. Tuberculation of The only other specimen referred to this sub- genae and palpebral areas variably developed. species occurs in the Upper Ferron Point Genal angles bluntly rounded. Formation. Axial rings of thorax and pygidium covered with tubercles moderately elongate transversely. Phacops rana rana (Green, 1832) Interpleural furrows obsolescent on pygidium. Figure 4A-D Pleural furrows moderately incised; pleura gently rounded. Tuberculation obsolescent on See Diagnosis ofspecies for synonymy. pygidial margins. DIAGNOSIS: Eye with 17 dorsoventral files in HOLOTYPE: NYSM 13887/1. normal adult. Trace of facial suture across STRATIGRAPHIC DISTRIBUTION: Cazenovian, ocular platform moderately incised. Tioughniogan, Taghanic, ?Finger Lakes. The Tuberculation generally densely distributed distribution of this subspecies is given in table 8. over glabella. Transverse elongation of tubercles A THEORY OF RELATIONSHIPS AMONG SUBSPECIES OF PHACOPS RANA Two general problems must be discussed before a general theory of relationships among these five taxa can be elaborated. The first of these involves the new subspecies P. rana paucituberculata. Although resembling P. rana norwoodensis in having but 15 dorsoventral files, the five specimens known of P. r. paucituberculata are identical in all other important interpopulational variation characteristics to P. rana crassituberculata and P. rana milleri. As mentioned previously, the closest relative to P. rana known is the Eifelian European species P. schlotheimi, sensu lato, recently revised by Burton (MS.b). Burton has shown that the P. schlotheimi in the Trilobitenfelder at Gees, Germany, includes a 15 dorsoventral file variant known as P. schlotheimi latreillii. In view of the close similarity between P. r. paucituberculata and the P. r. milleri-crassituberculata complex, and the existence of an apparently equivalent variation in B the closest known relative of the P. rana group, it is here concluded that the five known Caze- novian specimens of P. rana with 15 dorsoventral files are not closely related to the 15 dorsoventral file subspecies P. r. norwoodensis, but in fact represent a rare subspecies of P. rana whose affinities lie with the 18 dorsoventral file milleri- crassituberculata complex. C The second problem to be faced involves the relationship between P. rana crassituberculata and FIG. 16. Phacops rana paucituberculata, new subspecies. Plum Brook Shale, UMMP locality D; holotype, P. rana milleri. These two taxa are fundamentally UMMP 57135. A. Dorsal view, x2. B. Frontal view alike in terms of the criteria used to distinguish of cephalon, x3. C. Left lateral view of cephalon, the other subspecies, but differ from each other x 4. in other aspects of eye morphology. Phacops rana 82 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147 milleri and P. rana crassituberculata differ primarily may be a polymorphism reflecting a broad onto- in relative number of lenses per dorsoventral genetic norm of reaction reflecting local ecologi- file. Both have 18 dorsoventral files. There are cal conditions. A third possibility is that the up to eight or nine lenses per dorsoventral file in variants represent true geographic races. milleri, and crowding of so many lenses onto the The distribution of the two variants is given visual surface left little room for interlensar in table 7. Both Stewart's types of milleri and scleral tissue. Generally, the lenses within a Stumm's types of crassituberculata come from the file actually touch each other three-quarters Silica shale of Ohio and represent extremes. of the way up the height of the visual surface, Specimens from other stratigraphic units are and the lenses bulge out beyond the scleral occasionally more difficult to classify. Although surface. specimens from the Silver Creek limestone of Thus, the appearance of the lower three- Indiana are definitely crassituberculata, collec- quarters of the visual surface in milleri is similar tions from the St. Laurent limestone of Illinois to the lower one-third or so of the visual surface and units elsewhere are of a more intermediate of other subspecies. Apparently the rate of character. Generally, however, any single for- lens addition is very high and persists later mation will produce one or the other, but not into ontogeny than is usual in other subtaxa of both variants. The important exception is the P. rana. The crassituberculata variant generally Silica shale. The biostratigraphy of Phacops has a maximum of six lenses per dorsoventral within the Silica shale reveals that most of the file and is distinguished from P. rana rana in units recognized by Ehlers, Stumm, and terms of eye morphology solely by possession of Kesling (1951) contain only one of the eye 18 dorsoventral files. Factor analysis (fig. 9) variants, or at least an overwhelming predomi- shows that in terms of lens number per eye, nation ofone variant over the other. For example, typical populations of crassituberculata (e.g., unit 9 is a relatively thin calcareous shale unit SILC, SCRK) are close to later populations of containing abundant specimens of the milleri P. rana rana, whereas typical populations of the variant. The crassituberculata variant is extremely milleri variant (ARKA, SILM) seem less rare in unit 9. Conversely, crassituberculata far closely related to P. rana rana. Furthermore, outnumbers milleri in the underlying unit 8, this type of eye variation is not encountered which is a limestone (M. Widener, personal above the Cazenovian, or even within Caze- commun., 1968). Although specimens have been novian populations with 17 dorsoventral files, found with cephala of both of the two variants although upper Tioughniogan P. rana rana associated on the same bedding plane, in general secondarily converges on the milleri condition. crassituberculata is to be found in the limestones At least three plausible explanations may be and the denser, more calcareous shales of the considered for the existence of the two eye Silica Formation, whereas milleri occurs in the variants. Clarkson (1966b, p. 467) has tentatively more argillaceous units. The distribution of the suggested that the eye variants may reflect variants given in table 7 indicates that this sedi- sexual dimorphism. Alternatively, the variation ment preference holds generally true throughout the distribution of the two eye variants. It TABLE 7 will be recalled that, for the species as a whole, DISTRIBUTIONS OF Phacops rana crassituberculata AND factor analysis (fig. 9) indicated that samples Phacops rana milleri from highly calcareous shales in general tended P. r. crassituberculata P. r. milleri to have a greater number of lenses per cephalon size than those from more argillaceous or Silica shale Silica shale purely limestone units. Dundee limestone Bell shale The eyes of all small post-larval instars of Saint Laurent limestone Upper Ferron Point shale P. rana collected from the Silica shale appear to Lower Arkona shale Upper Arkona shale be of the milleri variety. Absence of any small Silver Creek limestone Genshaw Formation specimens clearly referable to the crassituberculata Lower Ferron Point shale Lower Romney variant indicates that the early ontogenetic Deputy limestone Lower Rockport Quarry variants is limestone development of the two identical. Cardiffshale and siltstone Rare but definite occurrences of specimens referable to both variants on the same bedding 1972 T.TT,TRF.?P.T .* PDLr ntD -xJ_ 83 plane also indicates that the hypotheses of geographic distribution and the relative ease in ecological or geographic races is incorrect. recognition of the two variants, it is Sexual perhaps dimorphism must also be rejected in best to continue to recognize the variants as view of the stratigraphic and geographic distri- distinct subspecies: P. rana milleri and P. rana butions of the two variants. The precise nature crassituberculata. It must be remembered, of the how- relationship between milleri and crassi- ever, that the relationship between those two tuberculata seems an insoluble problem with the subspecies is of a different nature than the given data as the correct explanation appears to relationships between both of them and the lie in the realm of population genetics. Perhaps other three subspecies recognized here. the ontogeny of the eye was flexible and capable In terms of dorsoventral file number of reaction to the the only, peculiar environmental five subtaxa of P. rana can be arranged in a conditions obtaining at a given place at a given gradational manner as in 1 7A. time. figure Figure 1 7B Alternatively, a more strictly genetically shows a similar gradational arrangement based based adaptation to local environments may on ornamental and other characters underlie the distributions discussed of the two variants. above. These character gradients are No model seems be available highly to that satisfac- intercorrelated; there is no "mosaic" torily accounts for all the pattern morphological and of different combinations of character states distributional data of these closely related different variants. among samples. The two diagrams differ only in that P. rana paucituberculata is Nevertheless, population samples can gen- included with the crassituberculata-milleri erally be referred to one or the other group eye variant. in figure 1 7B. It has alreadybeen concluded that The variants were also generally distinguished this second arrangement is probably a more in factor analysis (see figs. 9-12). In view of the accurate reflection of the true affinities of P. r.

c m r n P A

c m P r n

B x FIG. 17. Diagram of character gradients. A. Taxa arranged strictly according to number of dorsoventral files. B. Taxa arranged all other according to qualitative characters as discussed in text. Along each axis the primitive condition is an indicated by X and the direction of the arrow indicates the derived c condition(s). -P. rana crassituberculata; m -P. rana

- milleri; n P. rana norwoodensis; p - rana P. paucituberculata; r - P. rana rana. 84 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VIOL. 147 paucituberculata. Consequently, the 15 dorso- dorsoventral files must be assumed to have ventral file condition is thought to have evolved been derived directly and independently from twice within the P. rana complex as a whole. In the 18 file condition. The variation in dorso- view of the parallel, intercorrelated nature of all ventral file number in various samples of P. rana these character states (including ornament, from the Tully limestone of New York has been genal angle and ocular platform morphology discussed above. Amount of variation differs of the cephalon, and morphology of the pygid- from place to place. In the more argillaceous ium), there are but three possibilities as far as West Brook Member in the more eastern part primitive versus derived conditions of these of New York, all specimens retain 17 dorso- characters are concerned. These possibilities ventral files. In the Apulia Member (a fine- are indicated by arrows under figure 1 7B, where grained limestone) farther west, populations the primitive condition is indicated by an X, showing differing amounts of variation in and the derived condition by the arrow point. dorsoventral file number have been collected. Characters shown by factor analysis to be im- In a collection from South Lansing, New York portant interpopulational discriminants in gen- (AMNH loc. 3039), most cephala have but eral do not show the same type of gradation 15 dorsoventral files, but a few of the larger between subtaxa depicted in figure 1 7B and are cephala possess 16 or 17. Although the data are not discussed further herein. not quite so definitive as in the case of the 18-17 Although arrangements 1 or 2 may seem more dorsoventral file relationship, there are grounds economical hypotheses than arrangement 3, for preferring the hypothesis that 17 dorso- there is really little basis for preference for any ventral files is primitive with respect to the 15 one of the three possible primitive-derived file condition of Taghanic P. rana. A theorv of relationships outlined in figure 17B. The only relationships based on dorsoventral file number character that affords some definite data by is presented in figure 18. which primitive and derived conditions can be If this view of the relationships of the subtaxa assessed is dorsoventra] file number. Here of P. rana is substantially correct, we have some ontogeny and interpopulational variation do basis for preferring arrow no. 1 in figure 1 7B as indeed suggest a definite pattern of primitive most likely representing primitive-derived con- and derived conditions. ditions in all the other characters of the charac- Ontogeny of the eye in both P. r. milleri and ter gradient. However, there is no real biological P. r. crassituberculata indicate that the first proof that this is so. When this gradient, and the dorsoventral file is the last of the 18 files to apparent sequence of primitive-derived dorso- appear in ontogeny. The Solsville Formation ventral file number conditions are reviewed near Morrisville, New York (AMNH loc. 3013; below in conjunction with P. iowensis, stronger see Appendix 2) provides the only known sample reasons emerge for viewing arrow no. 1 in in which there is intrapopulational variability figure 1 7B as the true axis of primitive-derived between the 18 and 17 dorsoventral file condi- conditions. tions. As generally only the larger cephala in this When this view ofthe relationships among the population sample possess an eighteenth file five subtaxa ofP. rana is compared to the known (dorsoventral file no. 1), and this file is incom- stratigraphic record, and particularly first pletely developed in some specimens, the data appearance, of these subspecies, a rough coinci- suggest that the 17 dorsoventral file condition dence between stratigraphic position and ad- was derived pedomorphically from the 18 dorso- judged primitive-derived conditions emerges. ventral file condition. Furthermore, the main Not surprisingly, there is a basic agreement variants of the P. schlotheimi complex, here between the order of inferred phylogenetic considered the closest known relative of P. rana, events and the stratigraphic occurrence of the have 18 dorsoventral files. The best hypothesis, various subspecies. But to what extent may it be then, seems to be that the 18 file condition is said that the stratigraphic record provides an primitive for P. rana. ancestral-descendant sequence for these taxa? No ontogenetic or interpopulational variabil- The theory of relationships is, in a sense, ity data exist for P. r. paucituberculata, but in testable, but a rigid ordering of an ancestral- view of its close relationship by all other criteria descendant sequence, with or without con- to the milleri-crassituberculata complex, its 15 current stratigraphic data, is not. Consequently, 1972 ELDREDGE: PHACOPS 85

FIG. 18. Theory of relationships among the five subspecies of Phacops rana. Abbreviations as in figure 17. no claim that P. rana crassituberculata is ancestral appropriate attention paid to all assumptions, to P. rana rana, or that P. rana rana is ancestral to a best fit "model" for a reconstructed pattern of P. rana norwoodensis, is admissible in any strict phylogenetic events may be elaborated, with the sense. It must be recognized that the samples potential for rejection should new, contradictory on which the analysis of the interrelationships evidence appear. was based should not be construed as actual Various aspects of the evolutionary process ancestors to any such (stratigraphically) subse- may be investigated under these conditions. quent samples, but rather constitute samples of Among these, "trends," modes of speciation, taxa related in an unknown manner by common and interspecific interactions are discussed in ancestry to the subsequent samples. conjunction with P. rana and P. iowensis. The With this proviso in mind, once a theory of term "trend" is used herein perhaps in a slightly interrelationships ofthe taxa concerned has been different sense than is usual. Two difficulties, in worked out on strictly biological criteria, and if particular, lie in the usage of this term. The the stratigraphic record is at least in some preponderant model used in paleontology for accord with the inferred sequence of phylo- the origin of new infrataxa (species, etc.) has genetic events, the paleontologist is free to involved gradual, linear change instead of the speculate on the nature of these phylogenetic allopatric model accepted by most neontologists events on the basis of his samples. His state- (Eldredge, 1971). As a consequence the tend- ments, however, are largely untestable. But the ency is to look to a stratigraphic sequence to concept of the fossil record as a laboratory for reveal gradual, progressive evolutionary changes analyzing evolutionary experiments is very -i.e., trends. Thus, demonstration of bio- appealing, affording as it does the only oppor- stratigraphic trends may form the basis for the tunity available to the evolutionist to study analysis of the interrelationships among the phylogenetic patterns over a formidable period taxa involved, which is tantamount to assuming of time. It would my be fruitless, in view, to what is to be proved, and which may easily abandon this potential for comparing apparent result in an incorrect interpretation of phylo- patterns with known processes simply because geny. This constitutes one of the problems that the concept that preserves a the fossil record may be encountered when establishing ancestral- biostratigraphic pattern of phylogeny which descendant relationships on the basis of bio- may be read directly is biologically naive. With stratigraphical data. 86 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VJOL. 147 Lack of adequate attention to allopatric raphic ranges of these subtaxa, all of these phenomena raises a second, more substantive important variables turn out to possess charac- problem regarding trends. The problem is one teristics of trends. A simple list of these trends, of reconciling the process of adaptation to local then, is a trivial reiteration of these characters edaphic conditions by a peripheral isolate-an discussed extensively above. However, differ- important point in the general theory of allo- ences in rate and direction among these charac- patric processes-with long term, net change in ter gradients are instructive, as they demon- morphological structures which we refer to as strate (once again) that, to a degree, different "trends." As Wright (1967) has remarked, in characters may change in different ways the context of the evolution of higher taxa, (directions, rates) all the while being an speciation is a stochastic process. It seems that important part of a highly "integrated" pheno- adaptation to local edaphic conditions by type. It is well to remember, too, that the great peripheral isolates should also be stochastic in majority of anatomical features in P. rana the context oflong-term trends. In this light, it is remained essentially stable throughout eight possible that the pervasiveness and importance to 10 million years. of anagenesis (Huxley, 1958), with its attendant Most of these "biostratigraphic character connotation of "general improvement" may gradients" involve aspects of eye morphology; have been overemphasized in recent discussions six major trends involving ornamental and other (e.g., see Gould, 1970). features are thought to be progressive, i.e., In the present study, then, the term trend is gradual, involving a linear shift at a more or less applied to a demonstrated character gradient constant rate from the initial character state (between interrelated taxa), which has been in all Cazenovian populations through to the shown to appear in the stratigraphic record in final character state exhibited by populations the inferred order of its evolution. The samples in the Taghanic. It is essential to emphasize that used to document the trend are related in "gradual" and "progressive" imply changes in terms of recency of common ancestry, and no "mean" character state; no claim is made that ancestral-descendant relationships between sam- these changes can be documented stratigraphi- ples are claimed. The discussion is strictly cally, e.g., within successive samples of P. rana within the context of the history of the entire rana within the Tioughniogan. The significance group being considered (in this case the entire ofmost ofthese trends lies in the gradual approx- biospecies P. rana) and involves taxa definitely imation of P. iowensis characteristics by the known not to have been ancestral to subsequent entire P. rana complex through time. These samples as well as those which may or may not six trends may be summarized briefly as: (a) have been in fact ancestral to subsequent rounding of the genal angle; (b) deeper incision samples. It is simply a "statistical" characteriza- of the facial suture furrow on the ocular plat- tion ofthe status ofthe entire species Phacops rana form, and shift of the librigenal moiety of the at successive stratigraphic intervals. Amount ocular platform to a more nearly vertical orien- of variability among quasicontemporaneous tation, so that it becomes confluent with the samples, as well as any apparent linearity in area under the visual surface; (c) decrease in the direction of change can be related to adaptation amount of flattening and the degree of trans- to ecological parameters (including other species) verse elongation of ornamental tubercles on just as easily as if the assumption were made that the anterior region of the composite glabellar the successive samples were actual documents lobe, occipital lobe, and genal margins; (d) of progressive, phyletic change strung out in an increase in strength of tuberculation on the ancestral-descendant sequence. thorax and pygidium; (e) deeper incision of pleural furrows on the pygidium; and (f) TRENDS (BIOSTRATIGRAPHIC eventual loss of interpleural furrows on the CHARACTER GRADIENTS) IN pygidium. PHACOPS RANA Changes in mean character state values can All characters, including linear measurements, perhaps be documented more closely in the shown to be important interpopulational var- trends involving eye morphology, particularly iables, have particular "values" in each of the those seen in factor analysis. Within the context subspecies. In view of the different biostratig- of the preceding discussion on the nature and 1972 ELDREDGE: PHACOPS 87 significance of trends, the four graphs (figs. event taking a relatively short period of time. 9-12) showing mean population sample scores Throughout the greater period of the history of for a particular factor plotted against approxi- P. rana, dorsoventral file number was rigidly mate stratigraphic position reveal interesting stable. But, if the theory of relationships among information on modes of morphological change the various subtaxa as presented above is within the P. rana group. Thus one may speak largely correct, it is nevertheless true that all of a trend within Tioughniogan P. rana rana for phylogenetic changes in dorsoventral file num- increase in total number of lenses per cephalon ber involved reduction and never addition of size (fig. 9). Starting from a mean character files. A similar over-all trend in dorsoventral file state in the general range of the Cazenovian reduction is probably true of the P. logani-P. crassituberculata eye variant (e.g., SILC), this iowensis lineage in the Lower through Middle increase in total lens number through time Devonian of North America. Reduction and progressed until Upper Tioughniogan samples ultimate loss of eyes in other phacopid lineages, (WINC, WINE) fall into the general range of especially in the Upper Devonian of Europe P. rana milleri. Reduction in dorsoventral file (Richter and Richter, 1955) are of course number from 18 to 17, then, is unrelated to this well known, but it remains to be seen whether trend. The abrupt shift in this trend as seen in these trends were accomplished in the same the Taghanic samples of P. rana norwoodensis manner as presented here for P. rana. I will is due simply to deletion of two additional not pursue here the adaptive significance of dorsoventral files. such loss (but see Clarkson, 1967); in view of the In factors 2 and 3 (figs. 10, 11) of the second convergence in many characters of P. rana on P. factor analysis, aspects of eye size (height and iowensis, however, phylogenetic reduction in dor- posterior extent of the ocular platform) result soventral file number within the P. rana group in similar dog-legged plots. Here, the eye in may reflect convergence toward the 13 dorso- most samples of P. rana is reduced in size from ventral file condition of P. iowensis, although the the Upper Cazenovian into the Middle Tiough- functional anatomical reasons for this remain niogan; the trend is then reversed. Again, obscure. reduction in number of dorsoventral files is unrelated to trends seen in eye size, since earliest P. rana samples with 17 dorsoventral MODE OF ORIGIN OF THE SUBSPECIES files fall within a portion of the general Caze- Elsewhere (Eldredge, 1971) I have discussed novian variation, and reduction from 17 to 15 the probable mode of origin of the subspecies files necessitates no change in eye size whatso- of P. rana in terms of the allopatric model. It ever. Increase in eye size, rather, seems to be should be noted that the pattern presented related to increasing the total number of lenses in that paper represents merely a hypothesis of the eye within the Upper Tioughniogan based on the assumption that the theory of samples of P. rana - a phenomenon solely of relationships given in detail above is substan- increased number of lenses per dorsoventral tially correct. Again, the statement "P. rana rana file. is the ancestor of P. rana norwoodensis" should be Factor 4 (fig. 12), involving axial length and rewritten: "The common ancestor ofP. rana rana width proportions, and over-all size of the ocular and P. rana norwoodensis possessed (among other complex, shows no pattern of gradual change; important characteristics) 17 dorsoventral files." Tioughniogan rana are all rather similar (though The further hypothesis that this "common variable), and score closest to a portion of the ancestor," if somehow known, would be classi- variation seen in the Cazenovian. Some simi- fied as P. rana rana is essentially untestable. larity to the other trends (factors 2 and 3; However, for reasons set forth in Eldredge figs. 10, 11) does appear with the anomalously (1971), this hypothesis is to be preferred over low scores ofMILD and UPET in the Taghanic. others, and if correct, leads to interesting Finally, the grandest trend of all involved insights into modes of speciation among Paleo- reduction from 18 to 15 dorsoventral files. It will zoic invertebrates in the epicontinental seas. be argued immediately below that, far from For a further discussion, see Eldredge (1971). being gradual, the reductions took place in The hypothesis of phylogenetic descent in P. isolated, peripheral populations in steps, each rana is shown in figure 19. 88 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147

FIG. 19. Hypothesized phylogeny of the P. rana stock. Numbers at the base of the diagram refer to the population number of dorsoventral files. Dotted lines: origin of new (reduced) number of dorsoventral files in a peripheral isolate; horizontal dashed lines: migration; vertical solid lines: presence of taxon in indicated area; dashed vertical lines: persistence of ancestral stock in a portion of the marginal sea ("exogeosyncline") other than that in which the derived taxon occurs. Crosses denote final disappearance. The derivation of the 15 dorsoventral file subspecies P. rana paucituberculata is omitted from the diagram.

BIOSTRATIGRAPHIC SIGNIFICANCE OF Middle Cazenovian the PHACOPS through Taghanic RANA (and perhaps Finger Lakes) of the exogeosyn- The stratigraphic and geographic distribu- cline, but is restricted to the Tioughniogan tions of the five recognized subspecies of Phacops west of New York and the Appalachians to the rana are summarized in table 8. So far as is south; two specimens of apparent Taghanic age known, P. rana milleri and P. rana crassituberculata are known from the Cedar Valley Formation of are confined to the Lower Cazenovian of the Iowa. Phacops rana norwoodensis is known only eastern exogeosynclinal sediments, and range from the Taghanic. Thus a simple count of throughout the Cazenovian in epeiric deposits dorsoventral files could materially aid the field to the west. Phacops rana paucituberculata is known biostratigrapher working in the Middle Dev- only from the Upper Cazenovian of Ohio and onian ofeastern and central North America. Michigan. Phacops rana rana ranges from the VARIATION IN PHACOPS IOWENSIS

Phacops iowensis is not found abundantly in occurrence of P. iowensis east of the Michigan any formation and was presumably a rare Basin proper is in the Hungry Hollow Forma- species throughout its history. Originally de- tion of nearby southwestern Ontario in lower- scribed from the Cedar Valley of Iowa, its most Tioughniogan times, and in the Frame distribution is primarily within the Michigan Member of the Mahantango Formation at Basin where it occurs sporadically throughout Huntington, Pennsylvania, in the Upper Tiou- the Traverse Group. ghniogan. In sharp contrast to P. rana, stability was the Figure 21 shows the distribution of P. iowensis main feature of the history ofP. iowensis through- along factor 1 (normalizing by cases) plotted out the Middle Devonian. Such variation as can against stratigraphy. Factor components are be documented is of an order of magnitude less given in table 4 and sample size and abbrevia- than comparable variation in P. rana. tions in table 5. Although sampling is poor, Phacops iowensis specimens never show any particularly in post-Cazenovian sediments, it is deviation from the 13 dorsoventral files present apparent that there was no change in total lens in the earliest known specimen (AMNH 29105) number per cephalon size throughout the history from the Solsville Formation of central New of P. iowensis. As no change in number of dorso- York (fig. 20). The only other record of Mar- ventral files occurred, the number of lenses cellus (i.e., Lower Cazenovian) Phacops from per dorsoventral file itselfremained constant. the native North American lineage occurs in the Results of the factor analysis of the 26 cephala Ragland sandstone of Alabama. Specimens of P. iowensis where the data were normalized from the Ragland (USNM 71697) have 14 by variables, then by cases, are rather inconclu- dorsoventral files and on this basis as well as sive and show no easily interpretable, consistent other criteria are referable to the P. cristata-pipa patterns of variation. Discussion of these results complex of the Eifelian rather than to P. iowensis will be delayed until the interactions of the two sensu stricto. The origin of the 13 file P. iowensis species are considered below. seems to have occurred in much the same manner and at the same time as the reduction QUALITATIVE ANALYSIS OF from 18 to 17 dorsoventral files occurred in the MORPHOLOGY P. rana lineage. Phacops iowensis seems to have The facial suture furrow across the ocular arisen in the marginal exogeosyncline in the platform is invariably strongly developed in early Cazenovian and to have migrated into P. iowensis. However, it is more strongly devel- the Michigan Basin area with the reinvasion oped in Tioughniogan and Taghanic specimens of the Hamilton fauna in Lower Middle Caze- than in the Cazenovian specimens from the novian times. Indeed, the only subsequent Alpena and Gravel Point limestones and the Bell shale. Specimens from the basal Tiough- niogan Hungry Hollow limestone, however, show only a slight development ofthe furrow. The genal angle is invariably more sharply pointed (less rounded) in P. iowensis than in P. rana. In Cazenovian samples of P. iowensis, the genal angle is situated on the ventral margin of the cephalon when seen in lateral view and the dorsal margin ofthe visual surface is oriented in the horizontal plane. The genal angle is raised above the ventral margin in Tiough- FIG. 20. Dorsal view of cephalon ofP. iowensis from niogan and Taghanic specimens. the Solsville Member near Morrisville, New York, Glabellar furrow 1 p is variably developed AMNH locality 3013, AMNH 29105, x 2. and is less deeply incised mesially in younger 89 90 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147

TAGHANIC .CRVI *THBY

U. TIOUGHNIOGA

M. TIOUGHNIOGA *NRPT

L. TIOUGHNIOGA *FDMI * DSCI HHWI

. ALPI * GPTI

U. CAZENOVIA * GNSW

. BELI .75 .80 .85 .90 FIG. 21. Factor analysis of P. iowensis cephala. Data normalized by samples. Mean score for each sample for factor 1 (rotated) plotted against approximate stratigraphic position. Sample abbreviations as in table 5. Factor components are given in table 4. samples. Thus glabellar lobe ip tends to be presumably from the Genshaw Formation better developed in Cazenovian P. iowensis than near Alpena, Michigan. In this there is a very those ofthe Tioughniogan and Taghanic. slight development of transverse elongation of The only conspicuous deviation from the the tubercles on the anterior region of the typical iowensis pattern ofsharp, round tubercles composite glabellar lobe near the anterior covering the entire surface of the cephalon cephalic margin. These specimens are normal occurs in a single small sample (USNM 78934), P. iowensis in all other respects and no other 1972 ELDREDGE: PHACOPS 91 samples of P. iowensis show any signs whatever tened on the axial rings. Two specimens from the of flattening and elongation of cephalic tuber- Hungry Hollow Formation show strong tuberculation. culation along the pleura posterior to the Tubercles covering the entire exoskeleton pleural furrow. Tuberculation is usually obso- are not as sharply conical in Cazenovian sam- lescent in this area in most specimens of both ples as they are in Tioughniogan and Taghanic P. iowensis and P. rana. specimens ofP. iowensis. Pleural furrows are deeply incised in P. SUBTAXA OF Phacops iowensis iowensis and the pleura are rather highly inflated. The Cazenovian of P. iowensis At least populations the first two anterior interpleural form a discrete subunit for which the name P. furrows are developed in most specimens, and zowensis alpenensis (Stumm) (fig. 22) is available. occasionally three or four are apparent. The All post-Cazenovian populations are referable two portions of the fused pleuron separated by to P. iowensis iowensis Delo (fig. 3) except a few the interpleural furrows are of subequal height rather aberrant specimens from the Lower in Cazenovian P. iowensis, but are unequal in Tioughniogan (Centerfield) Hungry Hollow younger specimens. The larger section Formation, for which the name P. iowensis anterior to the interpleural furrow is more southworthi Stumm is available. These highly arched and (fig. 23) thus higher than the posterior subspecies are formally diagnosed below. section in Tioughniogan and Taghanic specimens. SYSTEMATIC PALEONTOLOGY Tubercles cover the entire pygidium and reach the lateral and posterior margins. Tuber- Phacops iowensis alpenensis (Stumm, 1953) cles are slightly higher and more conical in Phacops rana alpenensis STUMM, 1953, p. 139-140, pl. 11. Tioughniogan and Taghanic specimens than Phacops rana bellensis STUMM, 1953, p. 139, pl. 10. they are in the Cazenovian specimens. Tuberculation of the thorax in P. iowensis is Figure 22A-D conservative and similar to that of P. rana, with EMENDED DIAGNOSIS: Glabellar furrow ip the exception that in iowensis, the tubercles deeply incised. Tubercles moderately conical never become transversely elongated and flat- over entire exoskeleton. Genal angle near

A C FIG. 22. Phacops iowensis alpenensis (Stumm, 1953). A. Dorsal view ofholotype, Alpena Formation, UMMP locality 40, UMMP 25516, x 1. B. Left lateral view of cephalon, Alpena Formation, UMMP locality 53, UMMP 29628, cast of paratype, Buffalo Museum of Science E 15211, x 2. C. Dorsal view of paratype pygidium, same horizon and locality as B, UMMP 29559, x 2. D. Dorsal view of cephalon of enrolled specimen, Alpena Formation, AMNH locality 3059, AMNH 28895, x2. 92 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY V;OL. 1 47

_ ~~~~B FIG. 23. Phacops iowensis southworthi Stumm, 1953. Hungry Hollow Formation, UMMP locality B. A. Left lateral xiew of cephalon, UMIMP 24306, x 2. B. Pygidium and posterior portion of thorax, UMMP 25450, x 1. C. Dorsal view of cephalon of holotype, UMMP 24313, x 2. See also figure 24. ventral margin of cephalon, less steeply re- HOLOTYPE: SUI 9-266. curved than in other subspecies. Trace of facial STRATIGRAPHIc RANGE: Tioughniogan, Tag- suture over ocular platform moderately deeply hanic. Distribution of this subspecies is sum- incised. marized in table 8. Pleural furrows only moderately deeply incised and pleura moderately arched for Phacops iowensis southworthi Stumm, 1953 species. HOLOTYPE: UMMP 25516. Phacops iowensis southworthi STUMM, 1953, p. 14 1-142, STRATIGRAPHIc RANGE: Cazenovian. The pl. 12. precise distribution of this subspecies is given in Figure 23A-C table 8. EMENDED DIAGNOSIS: Exoskeleton very large for species. Glabellar furrow 1 p moderately deeply impressed. Genal angle slightly re- Phacops lowensis iowensis Delo, 1935 curved dorsally, raised slightly above ventral cephalic margin. Trace of facial suture across Figure 3A-C See Diagnosis of species for synonymy. ocular platform shallow for the species. Tu- bercles on exoskeleton large, sharply conical. EMENDED DIAGNOSIS: Glabella furrow lp Numerous small tubercles interspersed with variably incised, occasionally obsolescent me- large tubercles on glabella. sially. Tubercles sharply conical over entire Pleural furrows deeply incised and pleura exoskeleton. Genal angles recurved dorsally, highly arched. raised off ventral margin of cephalon. Trace HOLOTYPE: UMMP 24313. of facial suture over ocular platform deeply STRATIGRAPHIc DISTRIBUTION: Lower Tiough- incised. niogan (Centerfield). This species is based Pleural furrows deeply incised; pleura highly solely on specimens from the Hungry Hollow arched. Formation ofsouthwestern Ontario. INTERACTIONS BETWEEN PHACOPS RANA AND PHACOPS IO WENSIS

SUMMARY OF THE DISTRIBUTION OF P. iowensis and the disappearance of P. rana rana THE TWO SPECIES except in eastern New York and Maryland. A ALTHOUGH THE DISTRIBUTIONS of P. rana and P. few specimens have been recovered from iowensis have been discussed in the previous two Frasnian "Chemung" sandstones of New York chapters in conjunction with interpopulational and Maryland. Although poorly preserved, variation, their distributions are briefly clarified these specimens seem referable to P. rana rana, and summarized here and tabulated in table 8. not to P. rana norwoodensis, and simply represent a greatly diminished population near extinction. Phacops rana Phacops rana crassituberculata first appears as Phacops iowensis isolated occurrences in the Lower Cazenovian Phacops iowensis has been regarded as strictly Delaware and Dundee limestones of Ohio and ofTaghanic age, with the exception ofP. iowensis the Solsville siltstone of east-central New York. southworthi of the Hungry Hollow Formation of This subspecies and P. rana milleri are nearly southwestern Ontario. It is a rather rare species ubiquitous higher in the Cazenovian and show a in comparison with P. rana. In areas of geo- marked preference for calcareous shales. Phacops graphic overlap, P. iowensis tends to occur in the rana paucituberculata is known only from the purer limestones, and the subspecies of P. rana Cazenovian Plum Brook shale of north-central in the shalier units, although there are numer- Ohio and as a single specimen from the Caze- ous and important exceptions to this general- novian Upper Ferron Point shale of north- ization. The actual temporal range of P. eastern Michigan. iowensis is approximately as great as that of P. Phacops rana rana first appears in the Middle rana. The bulk of the history of iowensis was Cazenovian of western New York in the argil- limited to the Michigan Basin. The earliest laceous Stafford limestone, replaces P. rana occurrence is a single specimen from the Sols- milleri and P. rana crassituberculata in the east, and ville Formation of central New York. The gradually spreads westward. Although P. rana second oldest specimen comes from the Caze- rana occurs as far west as Iowa and adjacent novian Bell shale of eastern Michigan. Phacops states, it is perhaps most abundant in the iowensis is most abundant in the Alpena and Tioughniogan of the Appalachians and New Gravel Point limestones of Michigan, both of York. This subspecies shows a marked prefer- which are upper Cazenovian. Phacops iowensis ence for soft calcareous gray shales; any fossili- iowensis is found rarely in the Dock Street Clay ferous formation of this general facies in the and Four Mile Dam Formation, and is present Hamilton of New York will contain specimens as a single known specimen from the Tioughnio- of P. rana rana. Its distribution in the "encrinal" gan Norway Point Formation of eastern Michi- limestones and silt and sandstones is more gan. It also occurs rarely in the Lower Petoskey erratic. Phacops rana rana usually preferred fine- Formation of western Michigan. Its subsequent grained, slightly argillaceous calcareous sub- disappearance in the central United States is strates, but rich occurrences in more clastic attributable to lack ofdeposition or preservation formations are not uncommon. of sediments in the Upper Tioughnioga of the Phacops rana norwoodensis is the most widely Michigan Basin region, since it reappears in distributed of the subspecies of P. rana. It is some units of the Cedar Valley Formation as found in the calcareous shales and limestones well as in the Potter Farm and T'hunderbay of the Tully Formation and its equivalents limestones of eastern Michigan. A single westward through Michigan, Wisconsin, and pygidium from the Frame shale (AMNH loc. Iowa. Its lack of geographic restrictions reflects 3072) in Pennsylvania is the only known Upper the disappearance and presumed extinction of Tioughniogan occurrence ofP. iowvensis. 93 94 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147 TABLE 8 DISTRIBUTION OF THE SUBSPECIES OF Phacops rana AND Phacops iowensis Formation Prc Prm Prp Prr Prn Pia Pis Pii Central New York "Chemung" X Tully ~~xa Cooperstown -Ax Portland Point Panther Mountain Stone Mill Butternut Pompey Delphi Station Mottville XU Cardiff (Solsville) xa x Western New York Windom _ x Kashong x Menteth X Deep Run X Tichenor X Wanakah X Ledyard X Centerfield X Levanna X Stafford X Oatka Creek Southwestern Ontario Widder - X Hungry Hollow (Coral M.) x Hungry Hollow (Ls. M.) x Arkona (U.) - Arkona (L.) - Northeastern Michigan Thunderbay x Potter Farm x Norway Point x Four Mile Dam x Dock Street x Alpena x Newton Creek ?~~ Upper Genshaw x Lower Genshaw Upper Ferron Point X X x Lower Ferron Point Rockport Quarry (basal 18") X Bell x Rogers City Dundee Northwestern Michigan- Upper Petoskey x Lower Petoskey x Upper Charlevoix Gravel Point (emmetcnsis zone) _. x x Lower Gravel Point _ x 1972 ELDREDGE: PHACOPS 95 TABLE 8-(Continued) Formation Prc Prm Prp Prr Prn Pia Pis Pii Northwestern Michigan

Ten Mile Creek x - Upper Silica x x Lower Silica (with Paraspirifer) x x Dundee North-central Ohio Prout ? Plum Brook Delaware x - - Eastern Wisconsin Milwaukee (Zones B, C) Northeastern Iowa and Northwestern Illinois

Coralville - - - - x

Rapid - - - - x Solon x Wapsipinicon Southwestern Illinois Saint Laurent (basal 10') x Southern Indiana Beechwood Swanville Silver Creek Deputy Speeds Maryland, Virginia, West Virginia "Chemung" Harrell

- Upper Romney -_ x Lower Romney _ X _. _ South-central Pennsylvania Tully Member of the Harrell Formation Upper Shale (Frame) x Montebello Lower Shale Eastern Pennsylvania

"Tully" - - -? Upper Mahantango - X "Centerfield" -_ - Lower Mahantango Alabama Ragland Key: Prc Phacops rana crassiluberculata; Prm, Phacops rana milleri; Prp, Phacops rana paucituberculata; Prr, Phacops rana rana; Prn, Phacops rana norwvoodensis; Pia, Phacops iowensis alpenensis; Pis, Phacops iowensis southworthi; Pii, Phacops iowensis iowensis. a variable population (see text). X, present in formation. ?, probably present in formation.

MUTUAL OCCURRENCE OF PHACOPS and only a few isolated tagmata of P. iowensis, RANA AND PHACOPS 10 W1VENSIS' but one specimen from the University of Michi- Although there is a broad overlap in both the gan Museum of Paleontology shows a cephalon geographic and time distributions of P. rana and of P. iowensis southworthi actually touching a P. iowensis, the two almost never coexisted. The cephalon of P. rana rana on the same bedding Hungry Hollow Formation of southwestern plane (fig. 24). This specimen is the only direct Ontario has yielded many specimens of P. rana evidence for coexistence of the two species. 96 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY N7OL. 147 Table 9 lists all additional formations in which both species are found. In most formations, one species heavily outnumbers the other. Formations such as the Alpena limestone and Gravel Point limestone are rather thick sequences with several subunits readily recognized on lithologic and faunal criteria; no distinction of subunits was made in the majority of museum collections studied, and there is no direct evidence for coexistence in these few formations which have produced both species. There is no doubt that marine connections for faunal migration existed between the present- day areas of outcrop of Middle Devonian rocks from Iowa to New York. Such connections were probably more widespread than paleogeo- graphic maps have traditionally shown. An example is the occurrence of P. rana crassituberculata in the Cazenovian of southwestern Illinois, and the occurrence of a single large specimen of P. rana milleri found in float material at Shells- burg, Iowa, earlier this century. A complete locality list is included as Appen- dix 2. INTERACTIONS BETWEEN THE SPECIES FIG. 24. Slab from Coral Member, Hungry Hollow The virtually complete nonoverlap in occur- Formation, UMMP locality B, showing association rence of the two species is strongly indicative of of two cephala of Phacops rana rana with a cephalon of mutual exclusion. Although it is to some degree Phacops iowensis southworthi on the same bedding plane. true that, within the Michigan Basin, P. iowensis The upper two specimens are rana; the lower cephalon occurs in the purer limestones, whereas P. rana is is the holotype of southworthi. UMMP 24313, x 1.5. most commonly found in more argillaceous See also figure 4C and 23C. units. P. rana is found in pure, dense limestones throughout the Middle Devonian outside the TABLE 9 geographic distribution range of P. iowensis. It may be hypothesized that the two species FORMATIONS IN WHICH BOTH Phacops rana AND Phacops shared too of their in lowensis OCCUR many aspects ecology common to allow them to live sympatrically. Formation Prc Prm Prr Pia Pis Pii WN1hen closely related species with essentially similar ecological requirements and modes of Dock St. Clay R R life become sympatric, one species will either Gravel Point resorb the other, or competitively exclude the emmetensis Z. R A other, or a mutual subdivision of the range of Genshaw A R - Bell A R variation of one or more ecological parameters will allowing coexistence to continue. Hungry Hollow A - R result, Solsville R R In the one seemingly valid instance of coexist- Frame A - - R ence of rana and iowensis (in the Hungry Hollow (Upper Mahantango) Formation, fig. 24), rana is a typical P. rana rana with 17 dorsoventral files. All the other quali- KEY: Prc, Phacops iana crassituberculata; Prm, Phacops i-ania tative features of the morphology of these rana millei-i; Prr, Phacops rana rana; Pia, Phacops iowensis alpenensis; of Pis, Phacops iowensis southwow thi; Pui, Phacops iowensis iowensis; seem well within the range of variation other A, subspecies relati-ely abundant; R, stubspecies relatixely Centerfield rana. Phacops iowensis is known from rare. three cephala, two pygidia, and one thorax from 1972 ELDREDGE: PHACOPS 97 the Hungry Hollow; all of these specimens are in parameters (e.g., Lack, 1947, on bill morphology the collections of the Museum of Paleontology in relation to seed size in birds). Unfortunately, of the University of Michigan (UMMP 24306, the adaptive significance of most of the mor- 25450, and 24313). They differ from other P. phological features in Phacops are unknown. iowensis in a variety of ways and are very large But though the rana of Hungry Hollow appear compared to the normal maximum size seen "normal," P. iowensis southworthi is far more in other populations. This suite of specimens dissimilar to typical rana morphology than are forms the hypodigm of the subspecies P. iowensis any other iowensis specimens known. Phacops southworthi. iowensis in the Four Mile Dam and Dock Street Hybridization upon sympatric occurrence Formations (correlatives of the Hungry Hollow of the two species would have been very unlikely in eastern Michigan) are much more typical in view of the probable separate ancestries of the of the species in over-all appearance and more two species. Subdivision of ecological realms similar to rana than are the P. iowensis southworthi in sympatric situations frequently results in specimens. "character displacement," essentially an empha- Factor analysis clarifies the relationships sis of some of the morphological differences between the two species considerably. In between the species. Most discussions of charac- rotated factor 1, in which the data were normal- ter displacement deal with features directly ized by cases, figure 25 shows the distributions related to subdivision of specific ecological of both species simultaneously plotted strati-

TAGHANIC *MUPET *CRVN CRVI *CRVR * TULL °THBY

U. TIOUGHNIOGA *WINC . WINE

M. TIOUGHNIOGA * WGRB * WKHJ * LHCH . WIDD a NRPT

L. TIOUGHNIOGA *CBUF * HHWRI DSCI *CFAY OCBFDMI HHWI oALPI oGPTI U. CAZENOVIA PLUM

*ARKA *SILM .SILC*GNSW R GNSW SCRK * FPTM * *STAF

oBELI

*CARD L. CAZENOVIA

.55 .60 .65 .70 .75 .80 .85 .90 FIG. 25. Factor analysis of P. iowensis (O) and P. rana (.) cephala. Data normalized by samples. Mean score for each sample for factor 1 (rotated) plotted against approximate stratigraphic position. Sample abbreviations as in table 5. Factor components are given in table 4. 98 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147 graphically. Cazenovian P. iowensis are rela- It is to be remembered that WIDD is here tively close to the main line ofP. rana in terms of thought to be misidentified and actually comes lens number per cephalon size, although iowensis from the Hungry Hollow. In terms of lens generally has even fewer lenses per cephalon number per cephalon size, rana from Hungry size than the extreme condition in P. rana. Hollow seems to be actually convergent on Whereas P. rana shows its variation toward zowensis. According to factor 1, then, the picture increase in number of lenses per cephalon size, does not seem to be one of classic, mutual diver- iowensis remains relatively constant. The one gence of sympatric individuals of two species, cephalon of P. iowensis southworthi sufficiently but convergence of one (rana) toward the other well preserved to be measured is shown on the (iowensis) and consequent responsive divergence plot (fig. 25). As expected, southworthi shows the of iowensis away from general rana morphology. highest loading of all iowensis on factor 1, Although the general trend of increase in lens corroborating the hypothesis of character dis- number per cephalon size in P. rana rana is placement to a certain extent. Perhaps ofgreater away (i.e., divergent) from iowensis, P. rana interest is the fact that HHWR and WIDD are norwoodensis from the Taghanic reconverge on closer to the average iowensis than any other iowensis in this respect. Reduction of number of post-Lower-Middle Cazenovian rana population. dorsoventral files from 18 to 15 within the rana

*UPET .CRVN *TULL oTHBY o CRVI .*CRVR

U. TIOUGHNIOGA _WINC

M. TIOUGHNIOGA *WGRB oNRPT oWKHJ * WIDD .CFAY L. TIOUGHNIOGA *CBUF oDSCI * HHWR a FDMI H CEBY HHWI

o ALPI o GPTI PLUM

oGNSW *ARKA *SILM SILC . FPTM

U. CAZENOVIA O BELI

*CARD L. CAZENOVIA

.55 .60 .65 .70 .75 .80 .85 .90 FIG. 26. Factor analysis ofP. iowensis (D) and P. rana (.) cephala. Data normalized by variables, then by cases. Mean score for each sample for factor 2 (rotated) plotted against approximate stratigraphic position. Sample abbreviations as in table 5. Factor components are given in table 6. 1972 ELDREDGE: PHACOPS 99 lineage, of course, amounts to a partial con- LHCH, SCRK, STAF, and WINE are omitted vergence on the iowensis condition of 13 dorso- because they consist of internal molds. As in the ventral files. case offactor 1, in which the data were normal- Communalities in a factor analysis where the ized by cases only, WIDD and HHWR have data were normalized by variables are propor- the highest scores of any P. rana sample and tional to the over-all size of each specimen. actually fall close to the grand mean score of P. Although not illustrated, these communalities iowensis. HHWI (P. iowensis southworthi) is again show that throughout the stratigraphic range, displaced farther away from rana than any other all specimens of P. iowensis fall within the varia- P. iowensis sample, although specimens from the tion in size of P. rana, with the sole exception of Four Mile Dam limestone also score high. The one very large specimen of P. iowensis south- pattern is thus repeated: P. Zowensis southworthi worthi included in the analysis. seems to be "rebounding" away from P. rana, and Simultaneous plots of P. rana and P. iowensis samples of P. rana assumed to be sympatric are scores for the factor analysis where the data converging on P. iowensis. were normalized by variables, then by cases, Factor 3 (fig. 27) shows a somewhat different are presented in figures 26, 27, and 28. Factor 2, pattern. Based on eye height, HHWI again based primarily on variation in LOP, is shown falls at one end of the extreme of variation in in figure 26; here again, APPM, APPR, P. iowensis, farthest away from P. rana. Among all

TAGHANIC oCRVI *UPET -MILD *TULL *CRVNCV oTHBY .CRVR

U. TIOUGHNIOGA .WINC *WINE

M. TIOUGHNIOGA *WKHJ *WGRB *LHCH o NRPT . WIDD

oHHWI *CBUF oFDMI o DSCI *CFAY .CEBY.HHWR L. TIOUGHNIOGA

o ALPI ° GPTI

*PLUM o GNSW . .SILC U. CAZENOVIA SILM *FPTM *ARKA *STAF *SCRK a BELI

*CARD

L. CAZENOVIA

-.06 -.04 -.02 0.0 .02 .04 .06 .08 FIG. 27. Factor analysis ofP. iowensis (D) and P. rana (.) cephala. Data normalized by variables, then by cases. Mean score for each sample for factor 3 (rotated) plotted against approximate stratigraphic position. Sample abbreviations as in table 5. Factor components are given in table 6. 100 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147 samples of P. rana from Centerfield limestone 1. Normal iowensis tuberculation is developed to an the equivalents, however, Hungry Hollow P. rana extreme, where very large conical tubercles cover score farthest from P. iowensis. Here the scores of entire surface ofthe carapace. The few known specimens of southworthi are very HHWR and WIDD are not identical. There is 2. much larger than any other known specimen of thus some evidence for a more classic picture of iowensis. This accounts in large measure for the character displacement involving eye height divergence in lens number per cephalon size of P. between these two assumed sympatric popula- iowensis southworthi from the normal rana condition. tions. 3. P. iowensis southworthi shows a relatively shorter Factor 4 (fig. 28), involving proportions of (lengthwise) and taller visual surface (factors 2 and 3, normalized by variables, then by cases) than is usual glabellar lengths, widths, and eye complex size, for zowenszs. shows no peculiarities in the scores of either P. iowensis southworthi shows no apparent con- HHWI or HHWR and seems not to be involved vergence toward P. rana in any respect. in the interactions of the two species in the ofP. rana from the Hungry Hollow Hungry Hollow. Specimens Formation show convergence on iowensis in the The interrelationships between the only following characteristics: of the two species known to have populations 1. Smaller number of lenses per cephalon size than coexisted are therefore complex. The small is usual for the species in Centerfield times. southworthi sample shows particularly strong 2. Larger posterior length of the ocular platform divergence from rana in the following character- and concomitant shorter eye length than is usual for istics: P. rana ofCenterfield age.

TAGHANIC oCRVI OTHBY .MILD *UPET *CRVN -TULL *CRVR

U. TIOUGHNIOGA .WINE * WINC

M. TIOUGHNIOGA . LHCH .WGRB *WKHJ ONRPT *WIDD

o FDMI aDSCI oHHWI *CBUF *CFAY *HHWR *CEBY L. TIOUGHNIOGA

o ALPI o GPTI *PLUM

U. CAZENOVIA oGNSW * SILC ARKA . *FPTM *SCRK *STAF o BELI *CARD

L. CAZENOVIA FiG. 28. Factor analysis ofP. iowensis (O) and P. rana (.) cephala. Data normalized by variables, then by cases. Mean score for each sample for factor 4 (rotated) plotted against approximate stratigraphic position. Sample abbreviations as in table 5. Factor components are given in table 6. 1972 ELDREDGE: PHACOPS 101 Phacops rana from Hungry Hollow show no conjunction with deeper incision of the pleural change from normal P. rana rana of Centerfield furrows. age in having: 6. Phacops rana norwoodensis has fewest (15) number of dorsoventral files and is thus closer to P. iowensis 1. Seventeen dorsoventral files. in that respect than any other subspecies of P. rana 2. Normal ornamentation. (except P. r. paucituberculata). Phacops rana from Hungry Hollow show diver- This list of characters essentially repeats the gence from iowensis in the following: list of differentia that serve to define the various 1. Narrower cephalon and shorter visual surface subspecies of P. rana. In other words, the very height than in other P. rana of Centerfield age (factor characters that serve to distinguish P. rana from 3). P. iowensis are also those that show the greatest Coexistence, then, produced obvious effects amount of variation among various samples of in the morphology of the two species. In general, P. rana. The character states considered ad- rana tended to converge on, perhaps even to vanced in P. rana are closest to the condition mimic, P. iowensis. Phacops iowensis tended to exhibited in the (generally) stable P. iowensis accentuate the differences already established group. It seems likely, then, that the phylo- between the species. But in other respects, a genetic changes in character states noted in the more classical picture of character displacement history of P. rana stem from a general over-all is shown by the divergence ofthe two species. "interaction" with P. iowensis, and that far In terms of over-all phylogenetic history of from being a stochastic series of adjustments, the two species, factor analysis revealed several e.g., to local edaphic conditions, P. rana actually instances of trends in rana which varied in converged on P. iowensis. It is concluded that direction vis a vis iowensis. Divergence simply the various trends exhibited by P. rana represent reflects independent histories of the species neither "gradual improvement" nor a guiding living in different geographic areas at any one "orthoselection" principle controlling adjust- time. Evolutionary changes in one stock for the ment to local edaphic conditions, but an actual most part may have had little to do with changes morphological (and presumably ecological- in the other stock, at least through the Caze- ethological) convergence on P. iowensis for novian and Tioughniogan. However, the con- some unknown set ofadaptive reasons. vergence of P. rana norwoodensis in the Taghanic It may further be concluded that the initial on P. iowensis iowensis as seen in factor 1 (nor- assumption that P. rana and P. iowensis in fact malizing by cases) and factors 3 and 4 (normal- were true "biospecies" was not premature. Their izing by variables) is strengthened by conver- apparently interwoven histories can best be gences in qualitative morphological features. interpreted if this assumption is made, and there This close resemblance between norwoodensis and seems no reason, in retrospect, to challenge this iowensis therefore may not have been wholly characterization ofthe two taxa. fortuitous. Aside from the convergences seen Phacops rana norwoodensis replaces P. iowensis in factor analysis, norwoodensis is similar to iowensis in the Cedar Valley Group and in the iowensis in the following details, most of which Petoskey Formation, so that the only definite are not applicable to the variable norwoodensis Upper Taghanic occurrence of P. iowensis is in population ofthe Tully limestone: the Thunderbay and Potter Farm formations of eastern Michigan, where no rana are known. 1. The facial suture furrow of the ocular platform It is quite possible that the convergence of rana is more deeply incised than in normal P. rana rana. on iowensis in Upper Tioughniogan and Taghan- 2. Flattening and transverse elongation of tubercles times had to with the is reduced to the extreme anterior edge of the com- ic much do disappearance posite glabellar lobe and posterior row on the occipi- (possibly eradication) ofiowensis in the Taghanic tal lobe. from all but a small portion of its range in the 3. Tuberculation is denser and tubercles are gen- uppermost Tioughniogan and possibly Lower erally smaller and taller and subconical. Taghanic. 4. The pygidium is covered more densely with If the broad conclusions reached on the tubercles which often reach the lateral and posterior interactions between P. rana and P. iowensis are margins. substantially correct as presented above, further 5. The pleura become more highly arched in light is thrown on the one specific instance 102 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147 thought to involve character displacement. The displacement depends to a degree on the posi- apparent case of character displacement in- tion of the relevant population within the volving samples of P. rana and P. iowensis from evolutionary history of its own species, as well the Hungry Hollow Formation is actually very as it depends on the interactions with the other complex and is best viewed as a part of the species at the particular moment. Although total history of the interactions between the character displacement is largely a phenomenon two species. The reasons whyP. rana rana from the of sympatry, present results indicate that the Hungry Hollow converges on P. iowensis in full explanation for any particular case must certain respects and diverges or remains include the previous and subsequent histories "neutral" in others, are only interpretable in the of the lineages involved and the general nature context of previous and subsequent evolutionary of the interactions between the lineages through- changes within the P. rana lineage. At least in out their entire histories. this instance, the precise nature of the character SUMMARY Two TRILOBITE SPECIES, Phacops rana (Green, of the ontogenetic sequence represented. The 1832) and P. iowensis Delo, 1935 from the Middle cephalon is basically well "integrated," reflec- Devonian (Cazenovian, Tioughniogan, Tag- ting its nature as a solid tagma. The ocular hanic, and ?Finger Lakes) of eastern and complex maintains a degree of independence central North America, are analyzed in the from the major cluster ofgross and axial cephalic present study. Other species ofPhacops previously dimensions. described from these rocks, including P. cacapona Theories of relationship should be based on Hall, P. nupera Hall, and P. ohioensis, are con- morphologic (including ontogenetic) data, rather sidered invalid. than relative position in a biostratigraphic Several largely invariant, nonoverlapping sequence. Hypotheses of relative degree of morphological characters serve to differentiate common ancestry are more economical state- P. rana from P. iowensis: ments, and more easily capable ofrejection than I. P. iowensis invariably has 13 dorsoventral are hypotheses of ancestral-descendant rela- files, whereas P. rana has from 15 to 18. tionships. However, if biostratigraphic data are 2. The fixigenal and librigenal moieties of the in rough accord with a previously established ocular platform are more sharply defined in theory of relationships, various aspects of the iowensis than in rana. evolutionary process may be investigated. 3. Tubercles in iowensis are generally rounded Among such phenomena discussed here are at the base and subconical, whereas those in "trends" (biostratigraphic character gradients), rana are usually bluntly rounded on top. Phacops modes of speciation, and interspecific inter- rana shows elongation and flattening of cephalic actions. tubercles on the anterior margin of the compos- The primitive number of dorsoventral files is ite glabellar lobe, on the margins of the genae, determined to be 18 in P. rana, based on con- and on the occipital lobe, as well as on the siderations of (1) the condition in P. schlotheimi, entire axis of the thorax and pygidium. No such herein considered the sister species of P. rana, flattening and elongation is developed in P. and (2) ontogenetic information, particularly zowensis. that derived from rare samples where the 4. The l,p glabellar furrow is commonly number of dorsoventral files is variable. The obsolescent mesially in P. iowensis and always 17 dorsoventral file condition is derived (from sharply incised in P. rana. 18), whereas the 15 dorsoventral file condition 5. The genal angle in iowensis terminates in a is thought to have arisen twice, once directly from moderately sharp point, but is generally bluntly the 18 dorsoventral file condition, and once from rounded in rana. the 17. 6. Interpleural furrows are more extensively All characters, including linear measure- developed on the pygidium of iowensis than ments, shown to be important interpopulational they are in rana. variables, have particular "values" in each of Phacops rana is morphologically closest to the subspecies. In view of the different bio- P. schlotheimi of the Eifelian of Europe and stratigraphic ranges of the subtaxa, all of these Africa and probably migrated to Devonian important variables in fact possess characteris- North America along with Greenops (Greenops) tics of "trends" or "biostratigraphic character boothi (Green) in the Lower Cazenovian (Lower gradients." A summary ofthese trends follows: Givetian). Phacops iowensis apparently evolved (1) Rounding of the genal angle; (2) deeper from native North American species of Phacops incision of the facial suture furrow on the ocular ofthe Lower and Lower Middle Devonian. platform, and shift of the librigenal moiety Post-larval growth in Phacops is linear and of the ocular platform to a more nearly vertical usually adequately described by the equation orientation; (3) decrease in the amount of for a straight line. Allometry between most flattening and degree of transverse elongation variables is negligible, and populations may be of ornamental tubercles on the anterior region compared even when biased in terms of portions of the composite glabellar lobe, occipital lobe, 103 104 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147 and genal margins; (4) increase in strength of rare species. Although its range (Lower Caz- tuberculation on the thorax and pygidium; enovian-Taghanic) is virtually the same as (5) deeper incision of pleural furrows on the that ofP. rana, P. iowensis was far more stable and pygidium; (6) loss ofinterpleural furrows on the evolutionarily conservative than was rana. Phacops pygidium; (7) reduction of length and height iowensis was confined to the immediate vicinity of eye through the Middle Tioughniogan, then of the Michigan Basin throughout most of its subsequent reversal in the Upper Tioughniogan history. It apparently arose in the exogeosyn- and Taghanic; (8) increase in number of lenses cline from the P. cristata-pipa complex of the per cephalon size within the 17 dorsoventral file Eifelian. There was no reduction in dorso- lineage of the Cazenovian and Tioughniogan; ventral files within the P. iowensis lineage. (9) reduction in number of dorsoventral files Factor analysis revealed no phylogenetic change from 18 to 15. in lens number per cephalon size. Since the Five subspecies of P. rana are recognized. number of dorsoventral files was also constant Two subtaxa in the Cazenovian are character- (13), the number of lenses per dorsoventral ized primarily in having 18 dorsoventral files. file in homologous files was also constant. Phacops rana milleri Stewart, 1927 is distinguished Phacops iowensis showed no significant change from P. rana crassituberculata Stumm, 1953, in when plotted against any of the other factor having a greater number of lenses per dorso- axes calculated in factor analysis when the ventral file. Variation in this feature is not data were normalized by variables. encountered in the post-Cazenovian history of "Trends" in P. iowensis: P. rana. The two subspecies are similar in other (1) The trace of the facial suture over the respects, and seem actually to represent popula- ocular platform becomes more deeply incised tions adapted to local edaphic conditions. through time; (2) the genal angle becomes Recognition of two distinct subtaxa is therefore reflexed dorsally and raised above the ventral solely a matter ofconvenience, and the differences cephalic margin; (3) glabellar furrow 1 p between these two taxa are different in kind from becomes less deeply incised mesially. the differences between the other three sub- Phacops iowensis southworthi Stumm, 1953, species recognized herein. Phacops rana crassi- of the basal Tioughniogan of southwestern tuberculata and P. rana milleri are restricted to Ontario, exhibits morphological features dis- Cazenovian rocks, and though widely distrib- tinct from the other subspecies that reflect a uted, are more abundant in shales and lime- geographically based interaction with P. rana. stones of the cratonal interior. Phacops rana Phacops iowensis alpenensis (Stumm, 1953) is paucituberculata, new subspecies, known from restricted to the Cazenovian of the Michigan five specimens in the Upper Cazenovian, is Basin. Phacops iowensis iowensis occurs in Tiough- similar to P. rana crassituberculata, but differs in niogan and Taghanic rocks of the Michigan having only 15 dorsoventral files. Phacops rana Basin, Iowa, and (rarely) Pennsylvania. crassituberculata, P. rana milleri, and P. rana Although there is a broad overlap in both the paucituberculata seem to comprise a closely knit geographic and time distributions of P. rana complex that resembles a comparable trivariant and P. iowensis, only one probable instance of subdivision of the Eifelian species, P. schlotheimi, sympatry is known. In areas of geographic in Europe. Phacops rana rana, with 17 dorso- overlap, P. iowensis tends to occur in the purer ventral files, arose in the exogeosyncline in the limestones, whereas P. rana shows a preference Cazenovian. Restricted to the exogeosyncline for more clastic and argillaceous calcareous throughout the Cazenovian, P. rana rana is the sediments. only subspecies of P. rana known from Tiough- The one possible case of sympatry between niogan sediments. Phacops rana rana may have rana and iowensis occurs in the Hungry Hollow persisted through the Taghanic in the exogeo- Formation of southwestern Ontario. Sympatry syncline, and survived into Finger Lakes time. seems to have induced readily observed mor- Phacops rana norwoodensis Stumm, 1953 with 15 phological changes in the two species: dorsoventral files, is widespread in the Taghanic. Phacops iowensis of the Hungry Hollow shows Populations from New York show a transition particularly strong divergence from rana in from 17 to 15 dorsoventral files. the following characteristics: Phacops iowensis was apparently always a (1) Normal iowensis tuberculation is developed 1972 ELDREDGE: PHACOPS 105 to an extreme, where very large conical tubercles in some of the reference vector plots in factor cover the entire surface of the exoskeleton; analysis. The convergence was closest in the (2) the few known specimens of iowensis from Taghanic. Hungry Hollow are very much larger than any The greater amount of morphological change other known specimen of iowensis. This accounts shown by P. rana when compared with P. in large measure for the divergence in lens iowensis indicates that evolutionary change is number per cephalon size of P. iowensis south- more directly related to extent of geographic worthi from the rana condition. (3) Phacops distribution (greater in rana than in iowensis) iowensis southworthi shows a relatively shorter than to persistence through time (about equal (lengthwise) and taller visual surface than is for the two species). The allopatric model, usual for iowensis. where new character states arise on the periphery Phacops iowensis southworthi shows no conver- of the range of a species, is directly applicable gence toward P. rana in any respect except to the history of P. rana. Important changes in absolute size. eye morphology originate in the exogeosyncline Specimens of P. rana from the Hungry Hollow to the east, and subsequently spread through the Formation show convergence on iowensis in the epeiric seas on the cratonal interior. The following characteristics: allopatric mode of speciation or subspeciation (1) Smaller number of lenses per cephalon probably underlies the "sudden" appearance size than is usual for Centerfield P. rana; (2) of many Paleozoic taxa in cratonal sediments. larger posterior length of the ocular platform Evolutionary phenomena (e.g., development of (LOP) and concomitant shorter eye length clines, character displacement) affecting mor- than is usual for P. rana ofCenterfield age. phology on the species level as seen in recent Phacops rana from Hungry Hollow are similar organisms are applicable to the analysis of to normal P. rana rana of Centerfield age in fossil species, particularly those for which having: large samples with good stratigraphic control (1) 17 dorsoventral files; (2) normal orna- are available. The fossil record may be used to mentation. clarify certain processes seen on a single time Hungry Hollow rana diverge from iowensis in plane. For example, although character dis- the following: placement is largely a phenomenon of sym- (1) Narrower cephalon and shorter visual patry, the full explanation for any particular surface than in other rana ofCenterfield age. case must include the previous and subsequent The P. rana lineage as a whole converged on histories of the lineages involved and the general P. iowensis in many characteristics, including nature of the interactions between the lineages number of dorsoventral files and many orna- throughout their entire histories. mental features. This convergence is also seen APPENDIX 1 GLOSSARY OF MORPHOLOGICAL TERMS

CEPHALON librigenal moiety of ocular platform, and portion Apodemal pit, transverse slit on dorsal surface near ofgena anterior or distal to facial suture. distal margins of 1 p glabellar and occipital furrows, Librigenal moiety ofocular platform, anterior portion marking site ofapodeme. of ocular platform, separated from posterior Apodeme, exoskeletal invagination forming rod or (fixigenal) portion by a groove representing trace ridge on visceral surface for attachment of muscles ofobsolescent facial suture. or ligaments. Occipital furrow, transverse furrow separating occi- Area under visual surface, vertical area of eye below pital lobe from glabella. lens-bearing surface, separated from ocular plat- Occipital lobe, posterior axial lobe. form by a shallow furrow. Ocular platform, flat ridge supporting eye, contin- Ascending diagonal row, oblique row of lenses on uous posteroproximally with palpebral area. visual surface inclined anterodorsally. Separated posteriorly from cephalic border by Axial furrow, groove separating glabella from the shallow posterior border furrow, and from distal fixigena and librigena. portions of gena by a shallow furrow. Bounded Axis, median region, including glabella and occipital dorsally by area under visual surface and divided ring. into fixigenal and librigenal moieties by a trans- Composite glabellar lobe, large, inflated, roughly verse groove representing trace of obsolescent pentagonal anterior portion of glabella consisting facial suture. ofa fusion ofall glabellar lobes except Ip. Ip Glabellar furrow, continuous transverse furrow Descending diagonal row, oblique row of lenses on between composite glabellar lobe (anteriorly) and visual surface inclined anteroventrally. lp glabellar lobe (posteriorly). Dorsoventral file, vertical column of lenses on visual lp Glabellar lobe, complete transverse axial segment surface. of glabella anterior to occipital lobe and posterior Doublure, reflexed continuation of dorsal exoskeleton to composite glabellar lobe. as a ventral shelf. Palpebral area, inflated area of fixigena continuous Eye, large visual organ situated on librigena. posteroproximally with ocular platform and Facial suture, obsolescent molting suture running bounded anteroproximally by axial furrow and transversely from area anterior to genal angle distally by palpebral furrow. across the ocular platform then anteriorly behind Palpebral furrow, faint distally convex groove sep- margin of visual surface, connecting at the anterior arating the palpebral area (prox.) from the pal- cephalic margin around front ofglabella. pebral lobe (dist.). Fixigena, portion ofcephalon between axis and facial Palpebral line, imaginary transverse line connecting suture. Includes palpebral lobe, palpebral area, points of maximum convexity of the palpebral genal angle, and fixigenal moiety of ocular plat- lobes. form. Palpebral lobe, reniform inflated area of fixigena Fixigenal moiety ofocular platform, posterior portion bounded distally by facial suture and proximally of ocular platform, separated from anterior (libri- by palpebral furrow. genal) portion by a groove representing trace of Posterior border furrow, transverse groove imme- obsolescent facial suture. diately anterior to posterior cephalic margin, continuous with occipital furrow. Frontal arch, dorsally convex curvilinear profile of Sclera, interlensar exoskeletal tissue of visual surface. ventral margin ofcephalon, when viewed anteriorly. Tagma, regionally distinct area of exoskeleton Frontal arch sag, ventrally convex sagittal depression generally consisting of a number of similar seg- sometimes developed in frontal arch. ments, each bearing similar appendages. Genal angle, posterolateral corner of cephalon. 3p Glabellar furrow, anteriormost noncontinuous Glabella, axial region bounded distally by axial furrow of composite glabellar lobe consisting of furrows, posteriorly by occipital furrow, and two branches, the anterior one of which runs anteriorly by cephalic margin. anterodistally parallel with the axial furrow. The Horizontal row, row of lenses parallel to dorsal and posterior ramus is transverse and situated on the ventral margins ofvisual surface. distal portion ofthe composite glabellar lobe. Librigena, portion of cephalon distal to facial suture. Tubercle, projection of variable size and shape on Includes visual surface, area under visual surface, dorsal surface ofexoskeleton. 106 1972 ELDREDGE: PHACOPS 107

2p Furrow, noncontinuous transverse furrow of THORAX AND PYGIDIUM composite glabellar lobe, posterior to 3p glabellar Articulating half-ring, crescentic anterior extension furrow. of axial ring (anterior ring only on pygidium) Vincular furrow, continuous semicircular groove which passes under axial ring (or occipital lobe) on doublure just proximal to ventral margin. immediately anterior. Groove is smooth anteriorly and notched posteriorly. Axial furrow, longitudinal groove separating pleura Visual Surface, lens-bearing portion ofeye. from axial ring. Axial ring, central raised portion of tergite bounded HYPOSTOMA distally by axial furrows. Anterior wing, dorsal extension of anterolateral Interpleural Furrow, line of separation of pleura in corners. thorax, becoming very faint or lost in fused pygidial (Anterior) wing process, invagination forming a pleura. process on the visceral side ofthe anterior wing. Pleural furrow, groove along pleural surface. Central body, inflated central region ofhypostoma. Pleuron, lateral portion of thoracic or pygidial seg- Hypostomal suture, line ofjunction between anterior ment. edge of hypostoma and posterior margin of frontal Ring furrow, transverse furrow separating axial rings cephalic doublure. ofpygidium. APPENDIX 2 LOCALITY LIST

AMNH Jaycox Run, 0.25 mile west of N. Y. 39, 6.5 3013. Solsville Member, Marcellus Formation. miles south ofAvon, New York. Borrow pit in pasture on right side of Swamp 3042. Deep Run Member, Ludlowville Formation. Road, 2.6 miles north of Morrisville, New Same as loc. 3041. York. 3043. Menteth Member, Moscow Formation. Same 3028. ?Pompey Member, Skaneateles Formation. as loc. 3041. Borrow pit on No. 4 Road just east ofjunction 3044. Centerfield Member, Ludlowville Formation. with Pompey Center Road, Pompey Center, Cut along tracks of Delaware and Western New York. Railroad, 3 miles west of East Bethany, New 3029. Pompey Member, Skaneateles Formation. York, near Francis road. Road cut on U. S. Route 20, top of hill west of 3045. Ledyard-Wanakah Members, Ludlowville Pompey Center, New York. Formation. Same as loc. 3044. 3030. Butternut Member, Skaneateles Formation. 3046. Stafford Member, Skaneateles Formation. Borrow pit on secondary road approximately Buffalo Creek east of intersection of Indian 0.5 mile east of southeastern shore of Lake Church and Mineral Spring Roads, Buffalo, Moraine, Madison County, New York. New York. 3031. Ludlowville Formation. Borrow pit near 3047. Centerfield Member, Ludlowville Formation. Hatch's Lake, at intersection of Bradley Brook Buffalo Creek, near Blossom, New York. and Soule roads, 3.6 miles south ofintersection 3048. Wanakah Member, Ludlowville Formation. of Bradley Brook Road and N. Y. 26, near Cazenovia Creek near Transit Road, Erie Eaton, New York. County, New York. 3032. Lower Ludlowville (including Stone Mill 3049. Wanakah Member, Ludlowville Formation. limestone). Robert's Road Quarry, north of Ledges along shore of Lake Erie, just south N. Y. 26 on Robert's Road, West Eaton, ofmouth ofEighteen Mile Creek. New York. 3050. Windom Member, Moscow Formation. Ex- 3033. Lower Centerfield Member, Skaneateles or posure in cliff at shore of Lake Erie, about Ludlowville Formation. Fayette Town Quarry, 0.5 mile south ofloc. 3049. 0.25 mile west ofFayette, New York. 3051. Tichenor Member, Ludlowville Formation. 3034. Cooperstown Member, Moscow Formation. Cazenovia Creek at Northrup Road, Erie Borrow pit in New York State-owned land County, New York. on dirt road approximately 1 mile east of 3052. Windom Member, Moscow Formation. Caze- loc. 3035. novia Creek at Springbrook, New York, 3035. Cooperstown Member, Moscow Formation. upstream from falls. Borrow pit off Deep Spring Road, 1.3 miles 3053. Arkona Formation. Same as UMMP loc. B. from intersection of Deep Spring Road with 3054. Hungry Hollow Formation. Same as loc. 3053. Lebanon Road, south ofLebanon, New York. 3055. Hungry Hollow Formation. Same as UMMP 3036. Windom Member, Moscow Formation. 1.3 loc. A. miles west of N. Y. 80 on Kingsley Road 3056. Widder Formation. Same as loc. 3055. (intersection of Kingsley Road with N. Y. 3057. Upper Ferron Point Formation. Same as 80 is 1.0 mile south ofloc. 3037). UMMP loc. 38. 3037. Portland Point Member, Moscow Formation. 3058. ?Genshaw Formation. Roadcut on Michigan Road cut on N. Y. 80, 0.8 mile south of inter- Rt. 23, 10.6 miles south ofjunction of rts. 23 section ofN. Y. 80 and U. S. 20. and 65. 3038. Windom Member, Moscow Formation. Port- 3059. Alpena Formation. Quarry of the Huron land Point Quarry, 1 mile southwest of South Portland Cement Company, Alpena, Michi- Lansing, New York. gan. 3039. Tully Formation. Same as loc. 3038. 3060. Gravel Point Formation. Same as UMMP 3040. Tully Formation. Carpenter Road, 1 mile loc. 14. from "T" intersection with N. Y. 80, south of 3061. Four Mile Dam Formation. Exposures just Sheds, New York. below Four Mile Dam, south side of Thunder 3041. Wanakah Member, Ludlowville Formation. Bay River, west ofAlpena, Michigan. 108 1972 ELDREDGE: PHACOPS 109 3062. Thunderbay Formation. Same as UMMP AMNH JAMES HALL AND G. K. GREENE COLLECTIONS loc. 35. 1. Milwaukee Formation. Same as UMMP 3063. Silica Formation. Same as UMMP loc. E. loc. G. 3064. Mahantango Formation. Vicinity of Ding- 2. Microcyclus zone of St. Laurent limestone. man's Falls, Pennsylvania. Exposure in east bank of Mississippi River at 3065. Lower Mahantango Formation. 0.2 mile Devils Bakeoven, north of Grand Tower, south ofloc. 3066, on Pennsylvania Rt. 191. Illinois. 3066. "Centerfield" Member, Mahantango For- 3. "Hamilton" (Cedar Valley), Rock Island, mation. Exposure on Pennsylvania Rt. 191, Illinois. about 4 miles north of Stroudsburg, Pennsyl- 4. Tully Formation. Penn Yan, New York. vania. 5. ?Widder Formation. Bosanquet, Ontario. 3067. "Centerfield" Member, Mahantango For- 6. Silver Creek Formation. G. K. Greene Collec- mation. County road between Pennsylvania tion. Near Watson, Clark County, Indiana. Rt. 209 and Saylorsburg, Pennsylvania, 2.0 6A. Silver Creek Formation. G. K. Greene Collec- miles south ofRt. 209. tion. Near Charlestown, Clark County, 3068. "Centerfield" Member, Mahantango For- Indiana. mation. Vicinity of Pohopoco Creek, in road 7. ?Windom Member, Moscow Formation. cut on county road 2.5 miles east ofMahoning Moscow, New York. Valley exit of the Northeast extension of the 8. "Upper Helderberg" (almost certainly Dun- Pennsylvania Turnpike. Near Forest Inn, dee or Delaware Formations), possibly from Pennsylvania. "Kelly's Island," Ohio. 3069. ?"Centerfield" Member, Mahantango For- 9. Chemung. Chemung Creek, New York. mation. Exposure on north side of Pennsyl- 10. Mahantango Formation (?upper). Top of vania Rt. 443, on western edge of Lehighton, hill, north side ofCumberland, Maryland. Pennsylvania. 11. Romney Formation. Mouth of the Cacapon 3070. Upper shale Member, Mahantango Forma- River, Virginia. tion. Near Girty's Notch, near nose of Half 12. Hamilton equivalent. Probably from Virginia. Falls Mountain, in stream bed on west side of Susquehanna River, 6 miles north of inter- UNIVERSITY OF MICHIGAN MUSEUM OF PALEONTOLOGY, section of Rt. 15 and U. S. Rt. 22 at Amity ALL LOCALITIES FROM STUMM, 1953, EXCEPT THOSE IN Hall, Pennsylvania. PARENTHESES 3071. Upper shale Member, Mahantango Forma- A. Arkona, Hungry Hollow, and Widder For- tion. Borrow pit along road, south flank of mations. Rock Glen on the Ausable River, Mahanoy Ridge, east from Pennsylvania Rt. 1 mile east and i mile north of Arkona, 34, south ofNew Bloomfield, Pennsylvania. Ontario. 3072. Frame Member, upper Mahantango Forma- B. Arkona, Hungry Hollow, and Widder For- tion. Roadcut along U. S. Rt. 22, opposite mations. Hungry Hollow on the Ausable Huntingdon, Pennsylvania. River, 2 miles north and i mile east of 3073. Frame Member, upper Mahantango Forma- Arkona, Ontario. tion. Roadcut on U. S. 220, 0.8 mile south of C. Widder Formation. Brick and tile yard, i Newry, Pennsylvania. mile north ofThedford, Ontario. 3074. ?Upper Mahantango Formation. Behind (I). Arkona Formation. Fraser Farm, about 3 State Police barracks in La Vale, Maryland, miles northwest ofArkona, Ontario. along U. S. Rt. 40. D. Plum Brook Formation. Ledges along tribu- 3075. Upper Mahantango Formation. Railroad cut, tary of Pipe Creek, i mile east of Blooming- just south of crossing of twenty-first Lane, off ville, Ohio. U. S. Rt. 220 in Maryland, 1 mile north of E. Silica Formation. (North) quarry of the Keyser, West Virginia. Medusa Portland Cement Company at 3076. Romney Formation. 0.5 mile north ofjunction Silica, Ij miles southwest of Sylvania, ofRt. 50 and Rt. 28 in Romney, West Virginia, Ohio. on Rt. 28 atjunction with Central Avenue. F. Ten Mile Creek For-mation. Ten Mile Creek, 3077. Upper Romney Formation. Exposure on Rt. i mile south ofSilica, Ohio. 50 opposite Methodist Children's Camp, on G. Milwaukee Formation. Milwaukee, Wisconsin. western edge ofBurlington, West Virginia. 7B. Upper Petoskey Formation, Schizophoria bed. 3078. Dundee limestone. Same as UMMP loc. E. Limestone ledges along shore of Lake Michi- 3079. Skaneateles Formation. Borrow pit on south- gan, from 1 *-I miles north of Norwood, ern side of road along southern shore of Michigan, near center of North line, sec. 26, Boone Reservoir, Madison County, New York. T. 33 N, R. 9W. 110 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147

14. Gravel Point Formation. Quarry of Petoskey NATIONAL MUSEUM OF NATURAL HISTORY Portland Cement Company (now Penn-Dixie 1. Basal 18 inches, Rockport Quarry Limestone. Portland Cement Company), about 1 _ Same as UMMP loc. 38. miles west of Petoskey, Michigan, SW i sec. 1A. Ferron Point Formation. Same as UMMP 2 and SE J sec. 3, T. 34 N, R. 6W. loc. 38. 14e. Gravel Point Formation. Abandoned "Bell" 2. Upper Ferron Point Formation. Quarry ofthe Quarry and ledges on shore about 2 miles Huron Portland Cement Company, Alpena, east of Bay Shore, Michigan, near NE corner Michigan. sec. 8, T. 34 N, R. 6W. 3. "Hamilton" (?Genshaw Formation). Ledges 1 Sb. Gravel Point Formation. Shore of Little along Long Lake Road, north of Alpena, Traverse Bay from 9-Mile Point to 2.5 miles Michigan. west, Charlevoix County, Michigan, T. 34 N, 4. Upper Alpena Formation. Same as UMMP R. 7W. loc. III. 21. Lower Petoskey Formation. Kegomic Quarry 5. Thunderbay Formation. Same as UMMP on south shore of Mud Lake just east of loc. 35. Harbor Springs Road (Mich. highway 131), 6. Lower Potter Farm Formation, big Cranaena about 4 mile north of its termination on zone. Shale pit at Alpena Cemetery, on west U.S. 31, 1 mile east of Bay View, Michigan, side ofAlpena, Michigan. SEI SW I sec. 27, T. 35 N, R. 5W. 7. Potter Farm Formation. Bank of Thunderbay Formation. of River west ofAlpena, Michigan. 31. Bell Quarry Michigan Lime- miles west of stone and Chemical Company at Calcite, 8. Potter Farm Formation. 2 Michigan, near Rogers City, 10 sections on Alpena, Michigan. SE T. 35 R. 5E and 9. Gravel Point Formation. Same as UMMP part N, adjacent town- loc. 14. ships. 10. Milwaukee Formation. Same as UMMP 35. Thunderbay Formation. Bluffs on northeast loc. G. shore of Partridge Point, 4 miles south of 11. Upper Cedar Valley Formation. Banks of Alpena, Michigan; extends from center into Mississippi River and quarries near Buffalo, SE sec. ll,T.30N,R.8E. Iowa. 38. Ferron Point Formation. Abandoned quarry 12. Cedar Valley Formation, independensis zone. of Kelley's Island Lime and Transport West edge ofSolon, Iowa. Company, Rockport, Michigan, sec. 6, T. 13. Ragland Formation. Ragland, St. Clair 32 N, R. 9E. County, Alabama. (II). Alpena Formation. Quarry of the Huron 14. Stafford Formation. 2 miles north of Canada, Portland Cement Company, Alpena, Michi- New York. gan. 14A. Stafford Formation. Stafford, New York. (III). Norway Point Formation. Shale bank on the 14B. Stafford Formation. Le Roy, New York. south side of Thunder Bay River, about 1 14C. Stafford Formation. Batavia, New York. mile below Four Mile Dam, west of Alpena, 15. Pompey member, Skaneateles Formation. Michigan. Pratt's Falls, Pompey, New York. (IV). Potter Farm Formation. Shale pit on Potter 16. ?Skaneateles Formation. Morrisville, New Farmjust west ofEvergreen Cemetery, western York. limits ofAlpena, Michigan. 17. Skaneateles shale. Cayuga County, New York. 18. West Brook member, Tully Formation. 51. Upper Ferron Point and Genshaw Forma- Arab Hill Ravine, north end of Peruyter tions. Abandoned shale pit of Alpena Port- Reservoir, New York. land Cement Company, about 1 mile east 19. West Brook member, Tully Formation. and W mile north of Genshaw School and Tinkers Falls, Truxton, New York. 8 miles northeast of Alpena, Michigan, SE 20. Chemung. Spencer Lake Ravine, 2.5 miles 4,sec. 18, T. 32 N, R. 9E. north ofSpencer, New York. 53. Alpena and Dock Street Formations. Aban- 21. Mahantango Formation. Potts Grove, Penn- doned quarry of the Thunderbay Limestone sylvania. Company, eastern edge of Alpena, Michi- 22. ?Upper shale Member, Mahantango For- gan, SE j, sec. 14, T. 31 N, R. 8E. mation. Same as AMNH loc. 3071. 114. Genshaw Formation. Ledges along Long 23. Romney Formation. U.S. 52, 10 miles east Lake Road, near junction with Bell Road, ofBerkeley Springs, West Virginia. south of Long Lake, Alpena County, Michi- 24. Romney Formation. West Virginia. gan, sec. 22, T. 32 N, R. 8E. 25. Romney Formation. Fort Valley, Virginia. 1972 ELDREDGE: PHACOPS III 26. "Hamilton." Strasburg, Virginia. 191-13. "Brandon substage," Cedar Valley For- 27. Delaware Formation. Columbus, Ohio. mation. Locality unknown. 191-16. "Brandon substage, Euryocrinus zonule," STATE UNIVERSITY OF IOWA ?Rapid Member, Cedar Valley Formation, Belanski collection, listed according to Bel- near Brandon, Iowa. anski's station numbers, some of which have "Brandon substage," Cedar Valley Forma- been identified by Harrell L. Strimple. tion. Locality unknown. 139-2. Probably Rapid Member, Cedar Valley UNIVERSITY OF IOWA NON-BELANSKI MATERIAL Formation. Sanders Creek, north of Iowa City in Coralville Reservoir area. 1. Cedar Valley Formation. South side of 145- "Pholidostrophia zonule," Coralville Mem- Spillway, Lake McBride, Iowa. ber, Cedar Valley Formation. Same as 2. Cedar Valley Formation. N j NE i, sec. loc. 139-2. 26, T.81N, R.6W.,Johnson County, Iowa. 147-2. "triangulatus zonule." Probably Solon Mem- 3. Milwaukee Formation. Same as UMMP ber, Cedar Valley Formation. Locality loc. G. unknown. 4. ?Solon Member, Cedar Valley Formation. 168-3. "Gypidula comis zonule," Cedar Valley LinnJunction, Iowa. Formation. SW NW sec. 24, T81, Rt. 5. 5. ?Solon Member, Cedar Valley Formation. 170-1. "Brandon substage, Phacops rana zonule," Solon, Iowa. Cedar Valley Formation, NW NE sec. 26, T81, Rt. 5. NEW YORK STATE MUSEUM 183-5. "Brandon substage, P. rana zonule," Cedar 1. Tully Formation. Ovid, New York. Valley Formation. Locality unknown. 2. Stafford Formation. Stafford, New York. 190. "Brandon substage, Streptalasma zonule," 3. Hamilton Group. Seneca, New York. Cedar Valley Formation. Locality unknown. MUSEUM OF COMPARATIVE ZOOLOGY, HARVARD 191-8. "Vinton substage, Ptyctodus ferox zonule," UNIVERSITY Cedar Valley Formation. Locality unknown. 1. Silica Formation. Same as UMMP loc. E. 191-9. "Brandon substage, Phacops rana zonule," Cedar Valley Formation. Locality un- COLGATE UNIVERSITY known. 1. Silica Formation. Same as UMMP loc. E.

REFERENCES CITED ANON. Bull. Geol. Soc. Amer., vol. 53, pp. 1055- 1956. The Devonian strata of the London-Sarnia 1072. area, southwestern Ontario. Michigan CAMPBELL, K. S. W. Basin Geological Society, Annual Field 1967. Henryhouse trilobites. Oklahoma Geol. Trip Guide, 20 pp. Surv. Bull. 1 15, 68 pp. BRIGHT, R. C. CLARKSON, E. N. K. 1959. A paleoecologic and biometric study of the 1966a. Schizochroal eyes and vision of some Middle Cambrian trilobite Elrathia kingii. Silurian acastid trilobites. Palaeontology, Jour. Paleont., vol. 33, no. 1, pp. 83-98. vol. 9, pt. 1, pp. 1-29. BURTON, CHRISTOPHERJ. 1966b. Schizochroal eyes and vision of some [MS.a] Variation in Phacops latifrons and Phacops phacopid trilobites. Ibid., vol. 9, pt. 3, pp. schlotheimi. 464-487. [MS.b] Variation studies of some phacopid trilo- 1966c. The life attitude of the Silurian trilobite bites of Eurasia and northwest Africa. Phacops musheni Salter 1864. Scottish Jour. Ph.D. Thesis, University of Exeter, 222 pp. Geol., vol. 2, pt. 1, pp. 76-83. BURTON, CHRISTOPIERJ., AND NILEs ELDREDGE 1967. Environmental significance ofeye-reduction [MS.] New subspecies of Phacops rana from the in trilobites and recent arthropods. Marine Eifelian ofthe Spanish Sahara. Geol., vol. 5, pp. 367-375. CAMPBELL, G. COOPER, G. A. 1942. Middle Devonian stratigraphy of Indiana. 1930. Stratigraphy of the Hamilton Group of 112 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 147 New York, pts. 1 and 2. Amer. Jour. Sci., GOULD, STEPHENJAY 5th series, vol. 19, pp. 116-134, 214-236. 1966. Allometry and size in ontogeny and phylo- 1933-1934. Stratigraphy of the Hamilton Group geny. Bio. Revs., vol. 41, pp. 587-640. of eastern New York. Ibid. 5th ser., pt. 1, 1970. Evolutionary paleontology and the science vol. 26, pp. 537-551;pt. 2, vol. 27, pp. 1-12. of form. Earth-Sci. Revs., vol. 6, pp. 1957. Paleoecology of Middle Devonian ofeastern 77-119. and central United States. Geol. Soc. Amer., GREEN, J. Memoir 67, ch. 11, pp. 249-278. 1832. A monograph of the trilobites of North 1967. Age and correlation of the Tully and Cedar America. Philadelphia, 93 pp. Valley formations in the United States. HAs, WINFRIED Proc. Int. Symp. on Devonian System, 1969. Lower Devonian trilobites from central vol. 2, pp. 701-709. Nevada and northern Mexico. Jour. Paleont., vol. 43, no. 3, pp. 641-659. COOPER, G. A., AND P. E. CLOUD HALL,JAMES 1938. New Devonian fossils from Calhoun County, 1843. Geology of New York. Part IV. Comprising Illinois. Jour. Paleont., vol. 12, no. 5, pp. the survey of the fourth geological district. 444- 460. Albany, New York, 683 pp. COOPER, G. A., AND A. S. WARTHIN 1861. Descriptions of new species of fossils from 1941. New Devonian stratigraphic names. Jour. the Upper Helderberg, Hamilton, and Washington Acad. Sci., vol. 31, no. 5, pp. Chemung Groups. Advance sheets of the 259-260. Fifteenth Annual Report of the New York COOPER, G. A. ET AL. State Cabinet of Natural History, Albany, 1942. Correlation of the Devonian sedimentary New York, 83 pp. formations of North America. Bull. Geol. HALL, JAMES, ANDJOHN M. CLARKE Soc. Amer., vol. 53, pp. 1729-1794. 1888. Palaeontology: Vol. VII. Containing de- DELO, DAvID M. scriptions of the trilobites and other Crusta- 1935. New Phacopinae from the Devonian of cea of the Oriskany, Upper Helderberg, Oklahoma and Iowa. Jour. Paleont., Hamilton, Portage, Chemung, and Catskill vol. 9, no. 5, pp. 421-423. Groups. New York Geol. Surv. (Albany), 1940. Phacopid trilobites of North America. 236 pp. Geol. Soc. Amer., Special Paper 29, 135 pp. HUNT, A. S. EHLERS, G. M., E. C. STUMM, AND R. V. KESLING 1967. Growth, variation, and instar development 1952. Devonian rocks of southeastern Michigan of an agnostid trilobite. Jour. Paleont., and northwestern Ohio. Michigan Basin vol. 41, no. 1, pp. 203-208. Geol. Soc., Annual Field Trip Guide, 30 pp. HUXLEY,J. S. 1958. Evolutionary processes and taxonomy with ELDREDGE, NILES special reference to grades. Uppsala Univ. 1968. Convergence between two Pennsylvanian Arsskr., 1958, pp. 19-39. gastropod species: a multivariate mathe- IMBRIE, JOHN matical approach. Jour. Paleont., vol. 42, 1956. Biometrical methods in the study ofinverte- no. l,pp. 186-196. brate fossils. Bull. Amer. Mus. Nat. Hist., 1971. The allopatric model and phylogeny in vol. 108, art. 2, pp. 215-252. Paleozoic invertebrates. Evolution, vol. 25, LACK, DAVID no. l,pp. 156-167. 1947. Darwin's finches. Cambridge, Cambridge [MS.] The ontogeny ofthe eye ofPhacops rana. Univ. Press, 204 pp. ELLISON, R. L. MANSON, D. V., ANDJ. IMBRIE 1965. Stratigraphy and paleontology of the 1964. Fortran program for factor and vector Mahantango Formation in south-central analysis of geologic data using an IBM Pennsylvania. Pennsylvania Geol. Surv., 7090 or 7091/1401 computer system. fourth ser., 298 pp. Kansas Geol. Surv. Comp. Contr., no. 13, EMMONS, E. 47 pp. 1860. Manual of Geology. Philadelphia, Sower, OLSON, E. C., AND R. L. MILLER Barnes and Co., 290 pp. 1958. Morphological integration. Chicago, Univ. GILES, E. Chicago Press, 317 pp. 1960. Multivariate analysis of Pleistocene and PABIAN, R. K., ANDJ. A. FAGERSTROM Recent coyotes (Canis latrans) from Cali- 1968. Biometrical study of morphology and fornia. Univ. California Pub. in Geol. Sci., development of the Pennsylvanian trilobite vol. 36, no. 8, pp. 369-390. Ameura sangamonensis (Meek and Worthen) 1972 ELDREDGE: PHACOPS 113 from Nebraska. Bull. Univ. Nebraska STUMM, ERWIN C. State Mus., vol. 8, no. 2, pp. 189-207. 1953. Trilobites of the Devonian Traverse Group RICHTER, R., AN) E. RICHTER of Michigan. Contrib. Mus. Paleont. Univ. 1955. Oberdevornische Trilobiten, Na(chtrige, 2, Michigan, vol. 10, no. 6, pp. 101-157. Phylogenic der oberdevonischen Phacopi- 1954. Lower Middle Devonian phacopid trilo- dac. Senckenberg leth., vol. 36, pp. 56-72. bites from Michigan, southwestern Ontario, RICKARD, L. V. and the Ohio Valley. Ibid., vol. 11, no. 1964. Correlation of the Devonian rocks of Ncw 1 1, pp. 201-221. York State. New York State Museum and 1964a. Silurian and Devonian corals of the falls Science Serv., Geological Survey, Map ancl of the Ohio. Geol. Soc. Amer., Mem. 93, Chart Series no. 4. 184 pp. SAUL,J. M. 1964b. Silicified trilobites from the Devonian 1967. Burmeisteriai (Digonus) azccrazensis, a new Jeffersonville limestone at the falls of the homalonotid trilobite from the Devonian Ohio. Contrib. Mus. Paleont. Univ. Michi- of Ghana. Jour. Paleont., vol. 41, no. 5, gan, vol. 19, no. 1, pp. 1-14. WATERMAN, T. H. pPP- 126-1136. SCHWIMMER, 1). 1954. Relative growth and the compound eye in [MS.1 'T'he growth ancl allometry of the trilobite Xiphosura. Jour. Morph., vol. 95, no. 1, Phlacops raina. Master of Arts 'T'hesis, Statc pp. 125-158. of New York at 44 WHITTINGTON, H. B. University Bluffalo, pp. 1957. The ontogeny of trilobites. Bio. Revs., SHAW,) A. B. vol. 32, pp. 421469. 1956. Quantitative trilobite studies I. rhe statisti- 1959. Ontogeny of Trilobita. In Moore, R. C. cal description of trilobites. .Jour. Palcont., (ed.), Treatise on invertebrate paleon- vol. 30, no. 5, pp. 1209-1224. tology, part 0, Arthropoda 1, pp. 0 127- 1957. Quantitative trilobite Studi('1S II. Measure- 0145, Lawrence, Kansas, Univ. of Kansas ment of the dorsal shell of non-agnostidean Press and Geol. Soc. Amer. trilobites. Ibied., vol. 31, no. 1, pp. 193-207. WRIGHT, SEWALL STEININGER, J. 1967. Comments on the preliminary working 183 1. Observat ions suir les fossiles clu calcaire papers of Eden and Waddington. In interm&liaire die I'Fifel. Mem. Soc. G-ol. Moorehead, P. S., and M. M. Kaplan FIrance, vol. 1, no. 15,pp. 331-371. (eds.), Mathematical challenges to the STEWART, G. A. neo-Darwinian interpretation of evolution. 1927. IFaunria of the Silica shale of Lucas Couxnty. Wistar Institute Sympos., no. 5, Phila- Geol. Surv. Ohio, F'ourth Series, BIull. delphia, 'The Wistar Institute Press, pp. 32, 76 pp. 117-120.