Bound By BURGMEIER nnnK Bindery. Inc. EARLY DEVONIAN FISHES FROM UTAH PART III. ARTHRODIRA ROBERT H. DENISON FIELDIANA: GEOLOGY VOLUME 11, NUMBER 9 Published by CHICAGO NATURAL HISTORY MUSEUM JUNE 18, 1958 EARLY DEVONIAN FISHES FROM UTAH PART III. ARTHRODIRA ROBERT H. DENISON Curator of Fossil Fishes FIELDIANA: GEOLOGY VOLUME 11, NUMBER 9 Published by CHICAGO NATURAL HISTORY MUSEUM JUNE 18, 1958 Library of Congress Catalog Card Number: 52-2799 PRINTED IN THE UNITED STATES OF AMERICA BY CHICAGO NATURAL HISTORY MUSEUM PRESS INTRODUCTION The Osteostraci and Heterostraci of the Water Canyon Formation of northern Utah have been described in previous papers of this series (Denison, 1952, 1953). In the first of these a general account of the occurrence of the vertebrate remains was presented, together with detailed information on the collecting localities. For easy reference a list of the localities is repeated here. They are all in the Water Can- yon Formation of Cache County, Utah, and all except Locality I are in the Bear River Range. A. Blacksmith Fork, east side of Left Fork. B. Blacksmith Fork, immediately south of forks. C. Blacksmith Fork, immediately northwest of forks. D. Blacksmith Fork, Left Fork, in the two side canyons down- stream from Charlie's Hollow. E. Green Canyon. F. Water Canyon. G. Cottonwood Canyon. H. Crest of ridge, south-southeast of Naomi Peak, I. Northwest of Dry Lake, on the eastern edge of the Wellsville Range. Arthrodires have been found at all of these localities, but those from Locality H near Naomi Peak will not be included in this paper. The small and mostly rather poorly preserved collection from these beds represents a distinct fauna and almost certainly comes from a higher stratigraphic level, though probably below the beds referred to the Jefferson Formation. The only previous account of arthrodires of the Water Canyon For- mation was published by Branson and Mehl (1931). The form that they described as an arthrodire and named Campiaspis utahensis is really Cephalaspis, as was shown previously (Denison, 1952, p. 272). The fragments that they described and figured (pp. 516-517, pi. 1, figs. 4-6) as Aspidichthys sp. indet. probably belong to Aethaspis major, a new species to be described below. The median dorsal plate of Dinichthys {?)jeffersonensis and the jaws of Piyctodus described by Branson and Mehl probably came from the lower part of the over- 461 462 FIELDIANA: GEOLOGY, VOLUME 11 lying Jefferson Formation, as later defined by Williams (1948, pp. 1139-40); they are Late or perhaps Middle Devonian in age. As in my studies of the Osteostraci and Heterostraci from Utah, close comparisons have been made with the fauna from Beartooth Butte, Wyoming. In those two groups the same genera, and in some cases the same species were present in the two regions, but the com- mon arthrodires from Utah are quite distinct genera. No specimens of the Beartooth Butte Anarthraspis, and only a single fragment of Bryantolepis have been found in the Water Canyon Formation. The names employed for the plates of the dermal armor are those in common usage. I have avoided as far as possible the use of plate names that imply homologies with elements in Osteichthyes. New plates not previously named or described include the following: (1) paired antero-ventrals, bounded by the anterior ventro-laterals, in- tero-laterals, and anterior medio-ventrals, and confused with the latter by Stensio (1944, fig. 17, B) and others; (2) posterior dorsals, median plates lying behind the median dorsal in certain early arthro- dires, including those from the Beartooth Butte and Water Canyon Formations; similar plates were described as body scales in Phlyct- aenaspis acadica by Heintz (1933, p. 142, pi. 3, fig. 4). The specimen numbers given in the text refer to Chicago Natural History Museum specimens, unless otherwise noted. The drawings in this paper have been made, many from my original sketches, by Miss Maidi Wiebe, Staff Artist, Chicago Natural History Museum. CLASSIFICATION OF PLACODERMS In view of the current instability of placoderm classification, the following remarks are necessary to define and justify the usage of the higher categories employed here. Placodermi M'Coy 1848 is retained as a name for the class. Aphetohyoidea Watson 1937 is rejected, since the presence of a full hyoidean gill slit has not been clearly demon- strated in either the Acanthodii or Placodermi. Moreover, even if both groups should possess this feature, it would be improper to unite them in Aphetohyoidea since they are not closely related, as shown by many other features. Placodermi was originally proposed (M'Coy, 1848, p. 6) to exclude the ostracoderms now called Osteostraci and to include those forms now known as Coccosteus, Bothriolepis, Pterich- thyodes, Asterolepis, Chelyophorus, and Dipterus. The Dipterus was a cranial roof described as Polyphractus and was not then recognized as a dipnoan. With this exception, the original use of Placodermi was to include the groups now called Euarthrodira, Antiarcha, and Ptycto- DENISON: EARLY DEVONIAN FISHES 463 dontida. It may well be extended to include the more recently dis- covered related groups, Phyllolepida, Petalichthyida, Radotinida, Acanthothoraci, Rhenanida, and very doubtfully the Stegoselachii. There is little justification for the procedure of Stensio (1944, p. 75; 1948, p. 222) and White (1952, p. 298) of expanding the Arthrodira to include all the Placodermi, since the latter name has priority and long usage in this sense. Arthrodira was proposed by Woodward (1891, p. xxi) as an order to include Coccosteus, Brachydirus, Phlyctaenaspis, Chelyophorus, Din- ichthys, Titanichthys, Macropetalichthys, Homosteus, Heterosteus, As- terosteus, Phyllolepis, Holonema, and Mylostoma. The Antiarcha were specifically excluded. The genus Asterosteus, which could not be properly interpreted in 1891, is now known to be a representative of the highly specialized, ray-like Rhenanida. This group is probably best excluded from the Arthrodira, although Westoll (1945, p. 394) believes that they had a common origin with the Petalichthyida, and White (1952, p. 297) considers them to be related to the Brachy- thoraci on the slender evidence of a slight similarity in the shape of the nuchal plate. With this exception, we may follow Woodward by including in the Arthrodira the Euarthrodira, Petalichthyida, Phyllolepida, and Ptyctodontida, and may add as well the recently described Acanthothoraci and Radotinida. There can be no question of the close relationship of all of these groups. The order Euarthrodira was introduced by Gross (1932, pp. 9, 54) to distinguish certain arthrodires from the Petalichthyida and Ptycto- dontida. This name has found general acceptance. Gross subdivided the Euarthrodira into two groups: the suborder Acanthaspida of Heintz (1929, p. 24) with long trunk shields, and the Brachythoraci, a new suborder, with reduced trunk shields. Heintz (1937, p. 14) re- placed Acanthaspida by Arctolepida, since Acanthaspis, the type genus, was found to be a petalichthyid. Stensio (1944, p. 2) later in- troduced a new name, Dolichothoraci, for this group, but Heintz's earlier name is to be preferred. I consider the use of the groups Arc- tolepida and Brachythoraci to be unsatisfactory in that they repre- sent a more or less arbitrary, horizontal classification without regard to phyletic considerations. In spite of the assumptions of Stensio (1944) and White (1952), there is little doubt that the ancestors of the Brachythoraci are to be found among the Arctolepida. But until such time as euarthrodiran evolution is more clearly understood, these two groups are retained provisionally as representing a primi- tive (Arctolepida) and a more specialized (Brachythoraci) complex. 464 FIELDIANA: GEOLOGY, VOLUME 11 The classification adopted in this paper is essentially that of Gross (1937, p. 50): Class Placodermi M'Coy 1848 Superorder Antiarcha Cope 1885 Superorder Arthrodira Woodward 1891 Order Euarthrodira Gross 1932 Suborder Arctolepida Heintz 1937 Suborder Brachythoraci Gross 1932 Order Phyllolepida Stensio 1934 Order Petalichthyida Gross 1932 Order Ptyctodontida Gross 1932 Order Radotinida Gross 1950 Order Acanthothoraci Stensio 1944 Superorder Rhenanida Broili 1930 ?Superorder Stegoselachii Woodward 1924 The relationships and classification of the various genera of Arc- tolepida are poorly understood today. Perhaps this is not surprising when one considers that only three genera were known previous to 1929. Today, including two new genera to be described in this paper, there are 21 valid or probably valid genera, although some of them are inadequately known. For example. Murmur is known only from a single plate of the trunk shield that reveals little about its affinities. One genus (Heterogaspis) is based upon trunk shields, and another iSvalbardaspis) upon cranial roofs, many of which belong to other genera. Lataspis has been referred to Actinolepis by F0yn and Heintz (1943, p. 19) but is probably better retained as a distinct genus at present. Taunaspis is of uncertain affinities and may not belong to the Arctolepida. For these arctolepid genera there are available four family names, not including those based upon synonyms. Two of these families, Actinolepidae Gross (1940) and Williamsaspidae White (1952), have been used to include only the type genus. The family Arctolepidae Heintz 1937 ( = Monaspidae Heintz 1929; Jaekelaspidae Heintz 1932; Heterogaspidae Strand 1933) was proposed to include the Spitsbergen Arctolepida characterized by the fusion of the plates of the shield, a feature of doubtful significance. The fourth family, Phlyctaenas- pidae ( = Mediaspidae Heintz 1929), was apparently first used by Jaekel (1907, p. 176, "Phlyctaenaspiden") to sepsirate Phlyctaenaspis from the Coccosteidae. Heintz (1932a, p. 210) employed it to include genera without completely fused plates (Phlyctaenaspis, Mediaspis, and Huginaspis), and in the 1932 edition of Zittel it included all of the Arctolepida. DENISON: EARLY DEVONIAN FISHES 465 In my opinion, the known Arctolepida, with one or two exceptions, are best retained for the present in a single family, for which the name Phlyctaenaspidae is selected. The exceptions are the Middle Devonian Williamsaspis, for which White's familial name Williams- aspidae is available, and possibly the Late Devonian Gronlandaspis, which, when better known, may well reveal sufficient specializations to warrant the erection of a distinct family to include it. Taunaspis and Murmur must be considered at present of uncertain affinities, but the remaining 17 genera form a relatively compact group in which subdivision into different families, particularly monotypic ones, would serve no useful purpose in our present state of knowledge. Of course, the 17 genera represent various phyletic lines, but the distinction of these lines as families (e.g. Actinolepidae) would tend to overemphasize the differences and to mask the interrelationships of the Arctolepida as a whole. However, a subdivision of the Phlyct- aenaspidae into two subfamilies is made here. Definitions are given below, but a discussion of the significance of this classification is deferred until later. Suborder ARCTOLEPIDA Heintz 1937 These Euarthrodira have unreduced lateral trunk shields, long contacts between the lateral and ventral trunk shields, and small to medium-sized pectoral fenestrae between the anterior laterals and anterior ventro-laterals, indicating the presence of stenobasal pectoral fins. Spinals are always present and are usually large. The median dorsal lacks an internal keel. On the cranial roof, the preorbitals are generally relatively small, the rostral and pineal are fused in adults, and the nuchal characteristically is pointed anteriorly and has the sides subparallel or converging posteriorly. The orbits are generally far anterior, small, and notch the cranial roof only slightly. The sub- orbital has a short suborbital process. The endocranium is platybasic, with a large suborbital shelf and a short occipital region. The ventral shield is relatively short and broad. Most Arctolepida lived in the Early Devonian, though a few survived into the Middle and Late Devonian. Family PHLYCTAENASPIDAE Jaekel 1907 The ventral shield is nearly flat, and the spinal plates are at the lower edges of the lateral faces of the trunk shield. The spinal plates are moderately short to very long and always project posterior to their sutures with the anterior laterals and anterior ventro-laterals. 466 FIELDIANA: GEOLOGY, VOLUME 11 The pectoral fenestrae are small or possibly medium-sized in some genera. Subfamily ACTINOLEPINAE, new rank The median dorsal is short and broad. Paired antero-ventrals are present. The spinal plate is medium-sized to rather short. There is no differentiated exoskeletal articulation between the anterior dorso- laterals and the paranuchals. The following genera are included : Actin- olepis Agassiz 1844; Aethaspis gen. nov.; Anarthraspis Bryant 1934; Bryantolepis Camp, Welles, and Green 1949 ( = Euryaspis Bryant 1932); Kujdanoiviaspis Stensio 1942; Lataspis Strand 1932 { = Plat- aspis Heintz 1929); ?Mediaspis Heintz 1929; Simblaspis gen. nov.; Svalbardaspis Heintz 1929 (in part). Subfamily PHLYCTAENASPINAE Hay 1929 The median dorsal is long and narrow. The paired antero-ventrals are lost or indistinguishably fused with adjacent plates. The spinal plates are long. The anterior dorso-laterals have well-developed trochleae for articulation with the glenoid fossae on the paranuchals. The following genera are included: Arctaspis Heintz 1929; Arctolepis Eastman 1908 ( =^ J aekelaspis Heintz 1929); Diadsomaspis Gross 1937 ; Elegantas pis Heintz 1929; ?Gronlandaspis Heintz 1932b; Hetero- gaspis Strand 1932 (^Monaspis Heintz 1929); Huginaspis Heintz 1929; Phlyctaenaspis Traquair 1890; Prosphymaspis Gross 1937; Svalbardaspis Heintz 1929 (in part). "Phlyctaenaspis" heintzi Gross (1933b) resembles the Phlyct- aenaspinae in most respects, but the median dorsal referred here on the evidence of its ornamentation is short and broad, as in the Acti- nolepinae; if the median dorsal really does belong to "P." heintzi, this species must be intermediate between the two subfamilies. "Coccos- teus" angustus Traquair 1903 was referred to Phlyctaenaspis by Gross (1933a, p. 28) but appears to represent a new genus inter- mediate between the Phlyctaenaspinae and Coccosteidae. The pos- terior ventro-laterals and cranial roof resemble Coccosteus, the median dorsal is close in shape to that of Phlyctaenaspis, and the anterior ventro-laterals, spinals, and suborbitals are comparable both with those of Coccosteus minor and certain Arctolepida. Family WILLIAMSASPIDAE White 1952 The ventral shield is convex and the spinal plates are placed rather high on the sides of the trunk shield. The spinal plates are DENISON: EARLY DEVONIAN FISHES 467 reduced so that they do not project posteriorly to their sutures with the anterior laterals and anterior ventro-laterals. The pectoral fenes- trae are rather large. One genus is included: Williamsaspis White 1952. Incertae sedis Murmur Whitley 1951 (=Ptychaspis Bryant 1935; Euptychaspis White and Moy Thomas 1941); Taunaspis Schmidt 1933. ARCTOLEPIDA FROM UTAH Arthrodires of the Water Canyon Formation occur as isolated plates, except when the plates are fused together as is often the case in cranial roofs, occasionally with the median dorsal and anterior and posterior dorso-laterals, and usually with the anterior ventro-laterals, antero-ventrals, intero-laterals, and spinals. For this reason the deter- mination of which plates belong to the same genus and species is beset with many difficulties. Most helpful has been the type of ornamen- tation, which appears to be characteristic of a species. But this must be used with caution, since the ornament changes with the growth of an individual, owing to the formation of larger tubercles on the younger parts of a plate and the overgrowth by secondary tubercles on the older parts. An isolated plate of a young individual may be neither generically nor specifically identifiable, since its shape, pro- portions, and ornamentation may differ from that of the adult. For this reason there are many plates in the collection from Utah that must be considered as indeterminable. Siniblaspis,^ new genus Type species. — S. cachensis,'- new species. Diagnosis. — Simhlaspis includes Actinolepinae with rather short, broad cranial roofs. The rostral and pineal plates are not fused to the rest of the cranial roof. The preorbital plates are relatively long, with their centers of ossification in front of the centers of the plates and with their anterior borders moderately concave for the reception of the pineal plate. The nuchal plate is very short and relatively broad. The orbital notches are shallow to moderately distinct. The endo- cranium is not ossified. The median dorsal plate is short and broad, with a prominent, rounded, postero-median lobe. The anterior lateral ' From aifiiiXoi, a beehive, for Utah, the beehive state, and aaTTLs, shield. 2 cachensis, from Cache County, Utah. 468 FIELDIANA: GEOLOGY, VOLUME 11 plate is low, has a rather flat external face lacking ridges, a well- developed inner wing, a moderately long edge for the spinal plate, and an expanded postero-dorsal portion. Simblaspis cachensis, new species Type. -CNUM-FF 504, a small but well-preserved and little crushed cranial roof, lacking only the rostral and pineal plates (fig. 86, A). Referred s^Jeamews,— An incomplete cranial roof, PF499 (fig. 86, B) ; fragments of cranial roofs, PF 306, 308, 316, 500, 923 (part); left marginal, PF 521; median dorsal (fig. 88, A) and incomplete anterior or posterior medio-ventral, PF 301; incomplete median dorsal, ante- rior and posterior dorso-laterals, fused together, PF 965; left anterior lateral, PF 562 (fig. 88, B). Horizon. — Early Devonian, near base of Water Canyon Forma- tion. Locality. — The type is from Locality D, Blacksmith Fork, NE. 3^, sec. 26, T. 11 N., R. 2 E. The referred specimens are from Localities A, B, and D in Blacksmith Fork, and Locality G in Cottonwood Canyon. All localities are in Cache County, Utah. Diagnosis. — A species attaining a rather large size for an arc- tolepid; the length of the cranial roof, excluding the rostral and pineal, is 91 mm, in the largest individual. The ornament consists of small, crowded tubercles, lacking any distinctive pattern and over- grown in certain areas by large, crowded, secondary tubercles. Description and discussion. — The considerable difference in size and the lesser differences in the shape of the cranial roofs of the small- est and largest individuals referred here suggest at first that two species are represented. The type, PF 504 (fig. 86, A), has a cranial roof, excluding the rostral and pineal, that is 47 mm. long; in PF 499 (fig. 86, B) the corresponding measurement is 91 mm. The small individual has a prominent postorbital process that forms a distinct orbital notch, and also has a deep notch between the postorbital process and the prominent postero-lateral corner formed by the marginal and postmarginal. In large individuals, the postorbital proc- ess and the orbital notch are less distinct, and the lateral margin of the cranial roof is less irregular. In both large and small individuals, with one exception, the plates of the cranial roof are completely fused together and the sutures indistinct. The fusion of the plates is indica- tive of cessation of growth, but it is not possible to assume at present that this was necessarily permanent. It is possible that at certain B Fig. 86. Simblaspis cachensis, cranial roofs (X 1). A, type, PF 504: B, PF499 , incomplete on left side. 469 470 FIELDIANA: GEOLOGY, VOLUME 11 periods there was resorption at the sutures, followed by another period of growth. It is also possible that there was a high degree of variation in the time at which the sutures closed and growth ceased. A fragment of a cranial roof on PF 923 has the sutures open and the plates free in spite of the fact that this is a moderately large individual. PF 306, 308, and 316 are fragments of cranial roofs of intermediate size, and PF 308, at least, is also intermediate in the configuration of the lateral margin. These intermediate specimens, as well as the similar type of ornamentation found in all specimens, indicate that they represent a single species. The ornament of Simhlaspis is distinctive. The primary tubercles are small and crowded and lack any noticeable arrangement. These are overgrown by large, crowded tubercles, also lacking any pattern. The distribution of these large tubercles, however, appears to be characteristic. On the cranial roof (figs. 86, 87) they develop in a postero-median triangle covering the nuchal and the medial parts of the paranuchals and centrals. They appear on the anterior tips of the preorbitals and in large individuals develop on the rims of the orbital notches, as well as at the center of the marginals. The remainder of the cranial roof apparently retains the minute primary tubercles. In an old individual such as PF 499 (fig. 86, B) the secondary tubercles attain a much larger size than in a young individual such as PF 504 (fig. 86, A) ; they may well represent a third or even fourth generation of tubercles. The median dorsal (fig. 88, A) represents an old individ- ual and is completely covered with large tubercles. On this specimen it is possible to see large tubercles, possibly representing a third gen- eration, overgrowing medium-sized tubercles, probably of the second generation; no small, primary tubercles are visible. On the anterior lateral (fig. 88, B) the large tubercles are limited to a dorsal triangular area, while the small, primary tubercles remain on the anterior and ventral parts of the plate. Simblaspis has a relatively unspecialized cranial roof (figs. 86, 87, 105, D). From Kujdanowiaspis (fig. 105, F) it is distinguished mainly by two features: one is the elongation of the preorbitals, which has taken place mostly posterior to the centers of ossification; the other is the shortness of the nuchal, which is a result of the reduction of the posterior part of the cranial roof that has affected the paranuchals as well. The posterior part of the cranial roof is similar to that of Bryant- olepis (fig. 105, C), but Simblaspis does not have the firmly attached rostral and pineal, the greatly enlarged postnasals, and the short, very wide preorbitals that characterize Bryantolepis. DENISON: EARLY DEVONIAN FISHES 471 Fig. 87. Simblaspis cachensis, cranial roof, slightly restored from type, PF 504 (X 3/2). The lateral line canals of the cranial roof are arranged in general as in other Arctolepida, but certain canals or pit lines rarely seen in this group are indicated in the type of S. cachensis (fig. 86, A). The profundus canal ( = premarginal canal of Bryant, 1934, p. 137), which is prominent in Bryantolepis (fig. 105, C, pfc) and sometimes encoun- tered in Kujdanowiaspis (Stensio, 1945, p. 50), is indicated in S. cachensis (fig. 105, D, pfc) by a short groove and pits on the post- orbital immediately anterior to the infraorbital canal, as well as by scattered pits postero-lateral to the end of the supraorbital canal on the preorbital. The middle and posterior pit lines (fig. 105, D, mp, pp) are indicated by short grooves posterior to the termination of the central canal in the middle of the central; these lines are seen also in Aethaspis, Kujdanowiaspis, and Actinolepis. On the paranuchal a short, transverse groove anterior to the opening of the endolympha- tic duct indicates the posterior end of the posterior pit line (fig. 105, D, pp), while a similar groove posterior to the endolymphatic duct represents part of the supratemporal commissure (fig. 105, D, stc); 472 FIELDIANA: GEOLOGY, VOLUME 11 the only other arctolepid showing these canals is Actinolepis (fig. 105, H). The median dorsal of Simblaspis (figs. 88, A; 107, G) is relatively short and broad, as is typical of the Actinolepinae. It is very similar to that of Kujdanowiaspis (fig. 107, I) and Bryantolepis (fig. 107, H), though having a more prominent postero-median lobe than the former, and possessing an antero-median convexity not found in the latter. On either side of this antero-median convexity is a smooth, non-tuber- culated area (fig. 107, G, x), a feature found also in Aethaspis and Actinolepis, as well as in a median dorsal referred to "Phlyctaenaspis" heintzi by Gross (1933b, p. 60, fig. 9:1). This may signify that the median dorsal was actually overridden slightly by the cranial roof when the head was elevated. The anterior lateral (figs. 88, B; 110, C) resembles that of Bryant- olepis (fig. 110, E) and differs from Aethaspis (fig. 110, D) and An- arthraspis (fig. 110, F) in being relatively low. Anteriorly it has a large and sharply marked inner wing (fig. 110, C, iw), covered, even in the large PF 562, with very small tubercles. The dorsal part of the plate is relatively long and has coarse tuberculation. The ventral part of the plate is somewhat shortened, suggesting that the spinal may have been reduced to some extent. The shape of the posterior, concave edge indicates that the pectoral sinus was relatively narrow, especially when compared to Aethaspis. The only anterior and posterior dorso-laterals referred here with certainty (PF 965) are incomplete and so closely fused to each other and to the median dorsal that little can be said of their form. Another specimen (fig. 88, C), representing a smaller individual with firmly fused median dorsal and anterior and posterior dorso-laterals (PF 968, from Locality A), cannot be referred to S. cachensis because of its diff'erent type of ornament, consisting of moderately small, widely spaced tubercles. One small indication of relationship to Simblaspis is the peculiar dorsal twist to the lateral line at its anterior end on the anterior dorso-lateral, a feature also present in PF 965. The bone radiation can be made out in part and indicates a relatively long and shallow anterior dorso-lateral, similar to those of Kujdanowiaspis (fig. 108, I) and Bryantolepis (fig. 108, H) in shape, but with the overlap area for the anterior lateral not as deep, especially anteriorly. The posterior dorso-lateral also shows resemblances to those of Kuj- danowiaspis (fig. 109, I) and Bryantolepis (fig. 109, H), having an exposed area that is of moderate depth anteriorly, constricted at the middle, and deepened greatly posteriorly. I DENISON: EARLY DEVONIAN FISHES 473 Fig. 88. A, B, Simblaspis cachensis (X 3/4): A, median dorsal, PF 301; B, anterior lateral, PF 562. C, ?Simblaspis sp., anterior and posterior dorsolaterals and half of median dorsal, PF 963 (X 3/2). In conclusion, Simblaspis may be said to be a moderately primi- tive arctolepid, most closely comparable to Kujdanoiviaspis and Bryantolepis. It differs from the former especially in the elongated preorbitals and shortened nuchal and paranuchals, as well as in the shape of the anterior laterals. From Bryantolepis it is distinguished particularly by the fact that the rostral and pineal are not fused to 474 FIELDIANA: GEOLOGY, VOLUME 11 the rest of the cranial roof, as well as by the elongate preorbitals and presumably small postnasals. Aethaspis,' new genus Type species. — A. major, new species. Diagnosis. — Aethaspis includes Actinolepinae with the cranial roof broad posteriorly and tapering to a narrow anterior end. The rostral and pineal plates are fused to the preorbitals in large individuals. The postnasal plates are small and sometimes fused to the preorbitals. The preorbital plates are lengthened, especially posterior to the centers of ossification. The nuchal plate is greatly elongate and ex- tends anteriorly between the centrals to meet the preorbitals; it is notched laterally in its posterior half by the paranuchals. The orbital notches are small. The endocranium is ossified perichondrally. The median dorsal plate is short and broad and is crested externally in its posterior half. The anterior and posterior dorso-laterals are relatively high. The anterior lateral is relatively short and high, with a well- developed inner wing, and an edge for the spinal plate of moderate length. The spinal is rather short and stout. The posterior ventro- laterals are relatively short and broad. Aethaspis major, new species Type. - CNHM-PF 503, a nearly complete cranial roof (fig. 89, A). Referred specmews. — Nearly complete cranial roofs, PF 311, 1404 (fig. 89, B); fragments of cranial roofs, PF 300, 302, 313, 506, 507, 514-518, 915-919 (fig. 89, C), 1635, 1636; median dorsals, PF 563 (fig. 92, B), 905-907 (fig. 92, A), 910 (fig. 92, C), 911; posterior dor- sals, type 1, PF 558 (fig. 93, A), and type 2, PF 314, 559 (fig. 93, B), 967; anterior dorso-laterals, PF 305, 532 (fig. 95, A), 922 (fig. 95, B), 971; posterior dorso-lateral, PF 303 (fig. 95, C); anterior laterals, PF 560 (fig. 97, A), 561 (fig. 97, B), 925; posterior lateral, PF 924 (fig. 98, A) ; intero-lateral, PF 994; anterior ventro-laterals, some with attached intero-laterals, spinals, and antero-ventrals, PF 307, 939 (fig. 98, B), 940 (fig. 98, D), 983, 988, 989 (fig. 98, C), 990, 991; ?pos- terior medio-ventral, PF 315 (fig. 98, F); posterior ventro-laterals, PF 913 (fig. 98, E), 968. Horizon. — Early Devonian, near base of Water Canyon Forma- tion, ' From ariOris, curious, unusual; and aairis, shield. DENISON: EARLY DEVONIAN FISHES 475 Locality. — The type is from Locality B, Blacksmith Fork, S. 3^, sec. 3, T. 10 N., R. 2 E. The referred specimens are from Localities A, B, C, and D in Blacksmith Fork, Locality F in Water Canyon, and Locality G in Cottonwood Canyon. All are from Cache County, Utah. Diagnosis. — A large species attaining the following dimensions: total length of cranial roof approximately 136 mm.; length of median dorsal approximately 95 mm. The adult ornamentation is character- istic, though differing in various parts of an individual. Much of the surface is covered with large tubercles that locally in older individuals tend to coalesce to form a ridge-like ornament. Parts of certain plates have a sharply contrasting ornament of fine, crowded tubercles; this occurs on the anterior and lateral margins of the cranial roof and on the anterior face of the anterior laterals. Aethaspis utahensis, new species Typc-CNRM-FF 1405, a nearly complete cranial roof (fig. 90). Referred specmens. — Suborbital, PF 1403 (fig. 94, C); incomplete median dorsal, PF 964 (fig. 94, A) ; posterior dorsals, type 2, PF 975, 976 (fig. 94, B), 978, 1639; anterior ventro-lateral with attached intero-lateral, spinal, and antero-ventral, PF 321 (fig. 94, E, F) ; and posterior ventro-lateral, PF 322 (fig. 94, D). Horizon. — Early Devonian, near base of Water Canyon Formation. Locality. — All of the specimens are from Locality A in a thin, limy sandstone exposed in a gully about 200 yards southwest of the north- east corner of sec. 3, T. 10 N., R. 2 E., Cache County, Utah. Diagnosis. — A moderate-sized species; the total length of the cranial roof is approximately 68 mm. as estimated from the type, and the length of the median dorsal is 50 mm. in PF 964 (probably neither of these specimens is fully adult). The ornamentation consists of very small, moderately spaced, primary tubercles, commonly with a dis- tinctly linear arrangement. These are overgrown by medium-sized, secondary tubercles in the centers of the plates, and primary tubercles of medium size often occur in the younger parts of plates near their periphery. Description and discussion of Aethaspis. — The two species of Aethaspis are distinguished by their size and ornamentation. Since both of these features change during growth, there must be addi- tional evidence that A. utahensis does not represent merely young individuals of A. major. The evidence from the present collection is not as satisfactory as one could wish. Although A. ynajor is the most Fig. 89. Aethaspis major, cranial roofs (X 2/3). A, type, PF 503; B, PF 1404; C, left postero-lateral part of cranial roof, PF 919, showing main lateral line and ,| opercular pit line on paranuchal. ! 476 DENISON: EARLY DEVONIAN FISHES 477 tf^j^gy. -■ Fig. 90. Aethaspis utahensis, cranial roof, type, PF 1405 (X 3/2) common arthrodire in the Water Canyon Formation and shows a certain amount of variation in size, there is only a single specimen that can be determined with certainty as belonging to a juvenile individual. This is an incomplete median dorsal (PF 563, fig. 92, B), whose total length is estimated to be 36 mm. Its median region is missing, but its outer parts are ornamented with moderate-sized tubercles showing a tendency, anteriorly at least, to unite at their bases into ridges. This ornament is distinctly of the A. major type. Among the known specimens of A. utahensis there is not one that can be proved to belong to a fully adult individual, and it is probable that the cranial roof of the type specimen is not fully grown, since its sutures are quite distinct. However, the median dorsal referred here (PF 964, fig. 94, A) is considerably larger than that of the juvenile 478 FIELDIANA: GEOLOGY, VOLUME 11 A. major, yet it retains the characteristic type of ornament of A. utahensis. Additional evidence of the distinctness of these species is provided by the posterior dorsals, type 2. PF 314, 559 (fig. 93, B), and 967 have in their posterior halves relatively large tubercles with an indication of concentric arrangement and a tendency to unite at their bases into ridges; their anterior halves are ornamented with tubercles of a sharply contrasting small size. This ornament agrees with that of A. major. On the other hand, PF 975 and 976 (fig. 94, B), though of similar size and shape, have the very different ornament of A. utahensis. These posterior dorsals also differ in their convexity. In A. major they are strongly vaulted and sharply crested in the poste- rior halves, while in A. utahensis the crest is lacking and the posterior half is only slightly more convex than the anterior. With the exception of size and ornament, the two species of Aethaspis are very similar. There are a few other differences whose systematic importance is not certain. In A. major (fig. 105, B) the nuchal is relatively narrower and longer; posterior to its presumed center of ossification, it is deeply notched on both sides. In A. utah- ensis (fig. 105, A) the nuchal is relatively shorter and broader and has smaller notches in its posterior half. Another difference involves the relations of the central and marginal; in A. major these plates have broad contacts, while in A. utahensis they are separated by slender processes of the postorbital and paranuchal. In A. utahensis the middle and posterior pit lines are indicated on the central by short, shallow grooves (fig. 105, A, mp, pp), and the external opening of the endolymphatic duct is apparent on the paranuchal. None of these are seen in A. major, but this species sometimes shows a groove or series of grooves laterad to the main lateral line canal on the para- nuchal (figs. 89, C; 105, B, ope). If, as seems probable, these represent a pit line of the sensory canal system, they have no known homologue in other arthrodires. However, they may be homologous to the lines in Acanthodes that cross the posterior part of the gill covers and which were identified as opercular lines by Watson ( 1937, p. 121, fig. 20) and as scapular lines by Stensio (1947, pp. 48-49, fig. 11, D). The most distinctive feature of Aethaspis is its greatly elongate nuchal (fig. 105, A, B) that meets the preorbitals and completely separates the centrals from each other. Nothing approaching this condition is found in any other euarthrodiran. The elongate pre- orbitals with the centers of ossification in the anterior halves are also characteristic. Long preorbitals also occur in Actinolepis (fig. 105, H), but there the centers of ossification are far posterior and near or at the midline. The anterior end of the cranial roof is narrow, with a DENISON: EARLY DEVONIAN FISHES 479 small rostral and pineal, probably derived from separate ossification centers but fused in all known specimens. The postnasals v/ere appar- ently extremely small. In PF 507 and 1404 (fig. 89, B) the anterior end of the preorbital is cut off in a concave fashion just laterad to the su- praorbital sensory line; presumably the postnasals are absent here. In PF 311 and 313 the preorbitals are continuous with an anterior lobe that continues the convex rostral contour; probably the anterior parts of the lobes represent the postnasals, completely fused to the preorbitals. A number of cheek plates and jaws occur in the Water Canyon Formation collection, but with one exception they are undetermin- FlG. 91. Aethaspis utahensis, restoration of cranial roof based largely on PF 1405 (X 3/2); lightly shaded anterior and lateral parts restored from A. major. 480 FIELDIANA: GEOLOGY, VOLUME 11 Fig. 92. Aethaspis major, median dorsals (X 3/4). A, inner side, PF 907; B, outer side of juvenile, PF 563; C, outer side, PF 910. able specifically and generically. They will be described separately below, although presumably some of them belong to Aethaspis. A suborbital plate (PF 1403, figs. 94, C; 106, F) resembles A. utahensis in ornament and is referred to that species. It is of particular interest since this plate has been described adequately only in Phlyctaenaspis among the Arctolepida (Heintz, 1933, pp. 128-129, figs. 1, 2). The suborbital of Aethaspis is nearly flat and lacks the medially curved lower edge of Phlyctaenaspis. The anterior, suborbital process is rela- tively deeper, and the posterior "blade" is not as deep as in Phlyct- aenaspis and has nearly straight, sub-parallel, dorsal and ventral edges. The dorsal margin of the "blade" has a narrow, unornamented external surface for articulation with the inner side of the postorbital. The posterior edge is thick and must have had a harmonic suture DENISON: EARLY DEVONIAN FISHES 481 V ■liiiwliiiSiiiil ifcs»i?»-. Fig. 93. Aethaspis, posterior dorsals. A, A. major, type 1, PF 558 (X 4/3); B, A. major, type 2, PF 559 (X 4/3); C, A. sp., type 2, juvenile, PF 557 (X 2). with the postsuborbital, rather than overlapping it as in Coccosteus and Dinichthys. The median dorsal of Aethaspis is a short, broad plate (figs. 92; 94, A; 107, E), having the characteristic proportions of the Actinolepinae. It is considerably arched, has a median crest in its posterior half, and a nearly central center of ossification. Internally it has a small, median ridge (fig. 92, A) in the central part of the plate, but lacks, as do other Arctolepida, a prominent keel. In shape and proportions it agrees very closely with the median dorsal of Actinolepis (fig. 107, D) and Kujdanowiasjns (fig. 107, I). Anteriorly it has a median convex- ity, and on either side of this there is a small unornamented area re- sembling a sutural overlap (fig. 107, E, x). Similar areas are found in Actinolepis and Simhlaspis (fig. 107, G, x) and perhaps indicate over- lap by the cranial roof. There is a distinct postero-median lobe, but it is less prominent than in Simhlaspis and Bryantolepis (fig. 107, H), Several median dorsals, ranging in estimated length from 36 to 95 mm., are referable to A. 7najor on the evidence of their ornament. The tendency to a ridged ornament is quite pronounced in the anterior and lateral parts of some of the plates. Only one median dorsal is referable 482 FIELDIANA: GEOLOGY, VOLUME 11 to A. utahensis (fig. 94, A); it is incomplete at the sides, but its con- tours may be restored with some confidence by reference to the con- centric arrangement of the tubercles. There is evidence in Aethaspis that there were additional median plates lying behind the median dorsal. One of these, here called the posterior dorsal, type 1, is represented by PF 558 (fig. 93, A). This is a highly arched, oval plate, with a median crest extending for two- thirds of its length. Since it is symmetrical, it is certainly median, and its great convexity is a strong indication that it is dorsal. Its orna- ment shows that it belongs to A. major, but it is smaller and of a very different shape than the median dorsal of this species. It resembles the median dorsal of no other arthrodire, except for the problematical Taunaspis (Gross, 1937, fig. 13, D). It is assumed that this plate lay behind the median dorsal in the midline. A second type (type 2) of posterior dorsal of Aethaspis is represented by PF 314, 559 (fig. 93, B), 967, 975, and 976 (fig. 94, B). The first three, as already mentioned above, have the ornament of A. major, and the last two of A. utahensis. These plates are also symmetrical, slightly broader than long, rounded in what is believed to be the anterior end and indented in the opposite end. In A. utahensis (fig. 94, B) they are gently arched anteriorly and moderately arched posteriorly. In A. major (fig. 93, B) the anterior end is gently arched, while posteriorly they are strongly vaulted and provided with a median crest. It is assumed that these plates occupied a position behind the larger posterior dorsal, type 1. PF 557 (fig. 93, C) is a small plate that may represent a juvenile example of the pos- terior dorsal, type 2, of Aethaspis. It is remarkable particularly for the slot in the midline in its posterior half, a feature that is suggested by the tubercle pattern of the larger posterior dorsal, PF 976 (fig. 94, B). Apparently other Arctolepida possessed posterior dorsal plates, although they have never been found in position. Heintz (1933, pp. 141-143, pi. 3, figs. 3-6) has described a number of small plates and scales in Phlyctaenaspis acadica; some of these are relatively large, sharply arched, and incut behind, and resemble the second type of Aethaspis. Bryantolepis has some small, slightly arched, oval plates that may represent the same elements (Bryant, 1935, pi. 6, fig. 2). These cannot represent the median dorsal of a juvenile individual. Fig. 94. Aethaspis utahensis. A, median dorsal, PF 964 (X 3/2); B, posterior dorsal, type 2, PF 976 (X 3/2); C, left suborbital, PF 1403 (X 9/4); D, right posterior ventro-lateral, PF 322 (X 3/2); E, right anterior ventro-lateral, antero- ventral, intero-lateral, and spinal, ventral view, PF 321 (X 3/2); F, right intero- lateral and spinal, dorsal view, PF 321 (X 3/2). 483 484 FIELDIANA: GEOLOGY, VOLUME 11 since such are known possessing the same shape as in the adult. Simi- lar oval plates are known in Anarthraspis (PF 1536-7; 1539-40) and in Kujdanowiaspis. As far as is known, posterior dorsal plates do not occur in the Brachythoraci. The anterior dorso-lateral plate of Aethaspis major (figs. 95, A, B; 108, F) is relatively high both in the exposed face and in the overlap areas for the median dorsal and anterior lateral. In general shape and proportions it is most similar to Anarthraspis (fig. 108, G), and differs considerably from Bryantolepis (fig. 108, H) and Kujdanowiaspis (fig. 108, I), in which the exposed face is very shallow. As in all Actinolep- inae, the anterior edge lacks a trochlea for articulation with the cranial roof, but in the dorsal half there is a wide, unornamented area that was overlapped presumably by the paranuchal. Between this unornamented area and the overlap area for the median dorsal is a narrow, coarsely ornamented process of the exposed surface that extended toward the midline in front of the median dorsal. On the ventral part of the anterior margin there is a smaller, finely tuber- culated process of the exposed face that extended in front of the anterior lateral. The posterior part of the plate has internally a long overlap area for the posterior dorso-lateral (fig. 108, F, s. PDL). Dorsal to the lateral line canal, there is a posteriorly projecting lobe of the anterior dorso-lateral, as is commonly the case in Actinolep- inae; in Aethaspis major this is distinctive in shape in that it usually involves only the lower half of the exposed face above the lateral line. The anterior dorso-lateral is not known with certainty in A. utahensis, but one specimen doubtfully referred here (PF 1638, fig. 96) shows fused left anterior and posterior dorso-laterals and half of the median dorsal. As far as can be determined, this anterior dorso-lateral agrees in all respects except size and ornamentation with that of A. major. The posterior dorso-lateral of A. major (figs. 95, C; 109, F) is also a relatively high plate, with wide external overlap areas for all the adjacent plates. The exposed face is of nearly uniform depth throughout, although its ventral edge is slightly concave to accommo- date the posterior lateral. In this respect it is perhaps most closely comparable to Anarthraspis (fig. 109, E). The posterior dorso-laterals of Kujdanowiaspis (fig. 109, I) and Bryantolepis (fig. 109, H) differ in having the exposed faces shallow anteriorly, high posteriorly, and incut between deeply by the posterior lateral and shallowly by the median dorsal. In Aethaspis major the overlap area for the anterior dorso-lateral (fig. 109, F, s. ADL) is widened dorsally to accommodate the posterior lobe of that plate. The lateral line canal can be traced only a short distance on the posterior dorso-lateral, as is commonly I I Fig. 95. Aethaspis major (X 1). A, right anterior dorso-Iateral, outer side, PF 532; B, right anterior dorso-Iateral, inner side, PF 922; C, right posterior dorso-lateral, outer side, PF 303. Fig. 96. Aethaspis cf. utahensis, PF 1638, fused left anterior and posterior dorso-laterals and incomplete median dorsal (X 2). 485 486 FIELDIANA: GEOLOGY, VOLUME 11 Fig. 97. Aethaspis major, anterior laterals (X 2/3). PF 560; B, right side, inner face, PF 561. A, left side, outer face. the case in Actinolepinae. In PF 1638 (fig. 96), doubtfully referred to A. utahensis, the posterior dorso-lateral agrees in general shape and proportions with that of A. major. The anterior lateral plate is known only in A. major (figs. 97; 110, D), where it has quite distinctive features. The center of ossification is placed anteriorly and low, and in front of it the plate is bent sharply mediad to form a clearly demarked and finely tuberculated inner wing (fig. 110, D, iw). The dorsal part of the plate is unreduced and on the inner side shows wide areas for overlap of the anterior and posterior dorso-laterals (fig. 110, D, s. ADL, s. PDL). Behind the posterior dorso-lateral overlap there is probably a very small overlap area for the posterior lateral (fig. 110, D, s. PL), although this is not clear on the one specimen showing the inner side (PF 561). Lateral and ventral to the posterior lateral overlap area, the posterior edge of the anterior lateral is concave; in its dorso-median part it is thick- ened, indicating the probable position of the pectoral fenestra and fin; in its ventro-lateral part it is thinner where presumably it came in contact with the anterior ventro-lateral. The ventral edge of the plate is somewhat shortened, a feature that is indicative of a reduced attachment for the spinal. The plate is relatively higher than in Simblaspis (fig. 110, C) and Bryantolepis (fig. 110, E). It differs from Anarthraspis (fig. 110, F), in which the dorsal part of the plate is greatly shortened antero-posteriorly, resulting in a relatively short, high anterior lateral. I DENISON: EARLY DEVONIAN FISHES 487 The posterior lateral plate has been recognized in very few Arc- tolepida, probably because it is small and with no very striking shape. In the Water Canyon Formation collection, PF 924 (figs. 98, A; 111, C) is identified as a posterior lateral of a large individual of Aethaspis major. It differs from the posterior lateral of Phlyctaenaspis (fig. Ill, A, B) in two important respects. First, it is relatively short and deep. Second, the area overlapped by the anterior lateral (fig. Ill, C, s. AL) is very small and is widely separated from the posterior ventro-lateral overlap area (fig. Ill, C, s. PVL). In Phlyctaenaspis and also in Anar- thraspis (fig. Ill, D) these two overlap areas meet anteriorly, indicat- ing that the posterior ventro-lateral came into contact with the anterior lateral. In Aethaspis and also in Bryantolepis (fig. Ill, E) the overlap areas indicate that the posterior ventro-lateral and anterior lateral were separated and that the posterior lateral must have bounded the pectoral fenestra posteriorly. It is interesting that in a restoration of Phlyctaenaspis sherwoodi (Denison, 1950, pi. 3, fig. 2) this situation was inferred, although the posterior lateral itself was not preserved. The spinal plate of Aethaspis major (PF 939, figs. 98, B; 112, D) is relatively short and blunt-tipped and is characterized by the pres- ence of large, irregularly arranged, round tubercles on its outer edge, and smaller tubercles on the medial edge. In A. utahensis (PF 321, figs. 94, E, F; 112, F) the spinal is relatively slightly longer; laterally it has a distinctive double row of large tubercles arranged in pairs, as well as a cluster of large tubercles near the tip on the medial side. In addition to these two specimens, there are a number of other spinals in the Utah collection, some of which are doubtfully referable to Aethaspis and others are at present undeterminable. These specimens are of medium or small size, and reveal the interesting fact that the spinal is relatively longer in small individuals and shorter in large individuals. Length, spinal Aethaspis major, PF 939 Aethaspis utahensis, PF 321. Aethaspis sp., PF 536 Aethaspis sp., PF 547 indet., PF 923 indet., PF 553 Length Length Length, ant. ant. vent.-lat. spinal vent.-lat. 69 mm. 55 mm. 0.80 36 32 0,89 35 35 1.00 35 39 1.11 20 27 1.35 11 16 1.45 This correlation raises the interesting question as to whether the small individuals may not represent young of A. ynajor and A. utahen- 488 FIELDIANA: GEOLOGY, VOLUME 11 sis, and suggests that during growth the spinal did not keep pace, at least in length, with the anterior ventro-lateral. However, both of the very small forms, PF 923 (fig. 112, E) and 553, differ from Aethaspis in possessing several prominent, forwardly curved barbs on the inner side of the spinal, and unless this is a juvenile character they are probably distinct from Aethaspis. The intero-lateral is a long, slender, transversely oriented plate that extends from the midline to the anterior end of the spinal. It is sharply folded at its anterior edge so as to form a ventral lamina and a dorsal or inner lamina. In Aethaspis (figs. 94, E, F; 98, C, D; 112, D, F) the inner lamina is not present in the median part of the plate but begins in front of the antero-ventral and widens laterally to its widest point, where it meets the inner wing of the anterior lateral. Together with the latter it forms a constriction in the neck region extending from the upper half of the anterior lateral nearly to the mid- line. The suture of the intero-lateral with the anterior lateral is not overlapping. The ornamentation of the inner lamina is distinctive and would hardly be recognized as belonging to Aethaspis if found separately. In A. utahensis it consists of linear groups of three, four, or five, pointed tubercles, with the central one largest, and all leaning anteriorly. Probably a similar ornament occurs in A. major, and in PF 560 it is continued for a short distance onto the inner wing of the anterior lateral. The ventral lamina of the intero-lateral bounds ante- riorly the anterior medio-ventral, antero-ventral, anterior ventro- lateral, and spinal. It overlaps the anterior medio-ventral externally but appears to be overlapped by the antero-ventral and apparently abuts against the anterior ventro-lateral and spinal in non-overlap- ping sutures. In both species of Aethaspis the antero-ventral edge of the intero-lateral forms a distinct ridge which is undercut in such a way that it actually overhangs ventrally the posterior part of the ventral lamina. Internally, between the dorsal and ventral laminae, lay the bracing cartilage of the shoulder region; this is indicated by perichondral ossification extending from the anterior medio-ventral suture in a lateral direction. The ossification was called the pre-spinal Fig. 98. Aethaspis major (X 1). A, left posterior lateral, PF 924; B, left spinal with incomplete anterior ventro-lateral and intero-lateral, PF 939; C, right anterior ventro-lateral, antero-ventral, intero-lateral, and spinal, PF 989; D, left antero-ventral with parts of intero-lateral and anterior ventro-lateral, inner side, PF 940; E, left posterior ventro-lateral, PF 913; F, presumed posterior medio- ventral, PF 315. .Si/!::- -; 489 490 FIELDIANA: GEOLOGY, VOLUME 11 lamella by Heintz (1929, p. 41), and this part of the cartilage was homologized with the coracoid process by Stensio (1944, p. 56). The anterior ventro-lateral plate shows no particularly remark- able features in Aethaspis (figs. 94, E; 98, B-D; 112, D, F). It overlaps the adjacent antero-ventral, anterior and posterior medio- ventrals, and posterior ventro-lateral, but its sutures with the intero- lateral and spinal are harmonic or only slightly overlapping. There is no suggestion of any contact with or overlap of the posterior lateral, such as occurs in Bryantolepis (p. 528). The edge for the spinal is of moderate length, not as long relatively as in Kujdanowiaspis (fig. 112, H) and many Phlyctaenaspinae (fig. 112, I) and not shortened as much as in Anarthraspis (fig. 112, C) and Actinolepis (fig. 112, B); this feature is apparently correlated with the moderate length of the spinal. The posterior edge of the anterior ventro-lateral is rather deeply concave between the posterior ventro-lateral and spinal con- tacts. In the anterior and antero-median parts of this concavity some specimens show a thick lamina of unornamented bone lying internal to the ornamented external face of the anterior ventro-lateral (fig. 112, D, F, pf). This lamina represents an ossification of endoskeletal cartilage that forms the base of attachment of the pectoral fin. It is the most posterior part of the "scapulo-coracoid" of Stensio (1944) and White (1952). It occupies the pectoral fenestra and presumably bridges the gap of this fenestra between the anterior ventro-lateral and the anterior lateral. It indicates the presence of a stenobasal pectoral fin that has an attachment relatively no wider than in Kuj- danowiaspis. Lateral to the pectoral fenestra the anterior ventro- lateral probably meets the anterior lateral. The medial edge of the fenestra is presumably bounded by the posterior lateral in Aethaspis. Stensio (1944, fig, 17) has indicated the presence of paired "ante- rior median ventral" plates in Kujdanowiaspis. The evidence in Aeth- aspis and other Actinolepinae clearly shows that a median anterior medio-ventral was also present in this subfamily, so the paired plates are here called antero-ventrals to distinguish them. In Aethaspis (figs. 94, E; 98, C, D; 99; 112, D, F) they are small, triangular plates with their centers of ossification lateral to the centers of the plates. They are overlapped by the anterior ventro-laterals posteriorly and in A. utahensis they appear to overlap the intero-laterals, at least medially. PF 940 and 988, fragments of A. major, demonstrate the presence of a wide internal overlap area on the medial side (figs. 98, D; 99, B, s. AMV). This could only represent a suture for the anterior medio- ventral, and could not possibly have contacted the opposite antero- ventral because of its position and the absence of an external overlap DENISON: EARLY DEVONIAN FISHES 491 area on the medial side of any an tero- ventral. This is one of the many indications that the antero-ventrals are distinct from the anterior medio-ventral. Of the median ventral plates, there is only one incomplete speci- men (PF 315) that can be referred to Aethaspis (fig. 98, F). Its size indicates that it belongs to A. major, although there is no very strong Fig. 99. Aethaspis major. A, restoration of ventral shield (X 1/2); B, inner side of antero-ventral and medial parts of intero-lateral and anterior ventro- lateral, showing sutural overlap areas for anterior medio-ventral (s. AMV) and posterior medio-ventral (s. PMV), PF 940 (X 3/4). AMV, anterior medio-ventral; AV, antero-ventral; AVL, anterior ventro- lateral; IL, intero-lateral; PMV, posterior medio-ventral; PVL, posterior ventro- lateral; SP, spinal. tendency toward union of tubercles into ridges. It is a relatively short, broad, five-sided plate, agreeing closely in shape with the posterior medio-ventral of Anarthraspis (fig. 113, D). It is considered to be the posterior medio-ventral of A. major. The posterior ventro-lateral plates are poorly known in Aethaspis (figs. 94, D; 98, E; 114, C, E). From incompletely preserved specimens it is evident that the sutural overlaps are conventional, with the left posterior ventro-lateral overlying the right in the median line. The ventral lamina is relatively short and broad, even more so than in 492 FIELDIANA: GEOLOGY, VOLUME 11 Kujdanowiaspis (fig. 114, F), and is strongly contrasted with the long, narrow ventral lamina of Anarthraspis (fig. 114, B), Bryantolepis (fig. 114, D), and Actinolepis (fig. 114, H); posteriorly it is probably bluntly rounded. The lateral lamina is short and relatively high, rising to overlap the posterior lateral ; it is distinctly set off by a sharp curvature from the ventral lamina. In conclusion it may be said that Aethaspis, though in every re- spect a typical member of the Arctolepida, difi"ers in so many features from previously described genera of this suborder that it cannot be considered as closely related to any of them. The outstanding special- izations of the cranial roof are not even suggested in any other arctolepid; these include the lengthening of the preorbitals posterior to the centers of ossification, and the elongation of the nuchal to separate the centrals and to meet the preorbitals. Some of the lateral plates of the trunk shield resemble those of Anarthraspis, but this may be only an indication of a relatively high trunk shield in both genera. The shortening of the spinal and of the spinal edge of the anterior ventro-lateral reflects only a common evolutionary trend within the Arctolepida. Thus Aethaspis must be considered as repre- senting a distinct phyletic line within the suborder. Bryantolepis Camp, Welles, and Green PF 329, from Green Canyon (Locality E), is the only specimen from the Water Canyon Formation that is referable to one of the genera of Beartooth Butte Arctolepida. It is a small fragment of a cranial roof consisting of the rostral, pineal, and postnasal, and parts of the preorbital and postorbital (fig. 100). The large, five-sided pineal and the very large, wide postnasal leave no doubt that this is Bryant- olepis. As in the latter, these anterior plates are fused to the pre- orbital, and there is an indication that the endocranium was peri- chondrally ossified. The fragment cannot be identified with any of the Beartooth Butte species. B. hrachycephalus (Bryant), the common species at Beartooth Butte, is smaller. B. obscurus (Bryant) is doubt- fully distinct from B. hrachycephalus. B. cristatus (Bryant) was based on the median dorsal, and its cranial roof is not known with certainty. B. major (Bryant, 1935, p. 126) was based on a specimen that was considered to be a cranial roof but is probably an imperfectly pre- served median dorsal of Anarthraspis. Undetermined Arctolepida Because of the lack of association of plates in the Water Canyon Formation, there are a number of specimens that cannot be referred DENISON: EARLY DEVONIAN FISHES 493 with certainty either to Aethaspis or to Simblaspis. Among the most interesting of these are dermal bones of the cheek, lower jaw, and perhaps upper jaw. Hitherto the only jaw elements known in Arc- tolepida were described in Phlyctaenaspis acadica (Heintz, 1933, pp. 132-133, pi. 2, figs. 2-4), but their preservation is not good. There are soc Fig. 100. Bryantolepis sp. indet., restoration of anterior part of cranial roof based on PF 329 (X 3/2). PI, pineal; PN, postnasal; PRO, preorbital; PTO, postorbital; RO, rostral; cc, central canal; ifc, infraorbital canal; soc, supraorbital canal. five rather well-preserved infragnathals in the collection from the Water Canyon Formation (PF 509-512, 984). The infragnathal is well known in many Brachythoraci. In Coccos- teus, which may be used as a typical form for comparison (Heintz, 1931, fig. 4; Watson, 1934, figs. 1, D, E; 2, D), it consists of a posterior blade and an anterior tooth-bearing part. In the Utah Arctolepida no blade is present, and the smooth, thin, posterior edge of the tooth- bearing part in PF 510 and 511 (fig. 101, B-D) shows that this is not caused by breakage. The probable implication of this is that the infragnathal of Brachythoraci is a compound bone with a separate center of ossification posteriorly. In Arctolepida the posterior dermal ossification may not have formed or may not have been discovered or recognized as a distinct element. The anterior, tooth-bearing part of the infragnathal of the Utah Arctolepida is similar in its general structure to that of Coccosteus. PF 509-511 (fig. 101, A-D), all of which may belong to the same species, illustrate the character of this bone well, especially PF 511, which has been prepared free of the rock. The bone is solid only along the dorsal, biting edge. Below this it consists of two nearly parallel, thin laminae, a deep lateral one and a shallow medial one. Between 494 FIELDIANA: GEOLOGY, VOLUME 11 these laminae is a deep groove that must have been occupied in life by Meckel's cartilage. The groove extends to the anterior end of the jaw, where it is closed by the union of the two laminae in the symphy- sial region. As one progresses forward, at midlength the infragnathal curves slightly mediad, and near the anterior end it is bent sharply toward the midline; the bends are stronger ventrally than dorsally, resulting in a twisting of the ventral edge inward. The dorsal edge of PF 509 (fig. 101, A) is set with teeth in much the same way as in Coccosteus. The teeth occur in the anterior part of the posterior half of the bone and include seven rather large, distinct elements posteriorly ; in front of these are four smaller and increasingly indistinct elements, grading anteriorly into a worn ridge in which separate tooth elements are not distinguishable. In PF 511 (fig. 101, C) the tooth-bearing edge is partly broken away. In PF 510 (fig. 101, B), the largest infragnathal, the separate teeth are relatively small, probably because of wear; twelve are distinguishable along the dorsal edge and four occur just below them on the lateral face, where they are definitely attached and not displaced. The anterior part of the dorsal edge shows no distinct teeth in any of these specimens, but is a relatively smooth, worn surface. In distinction to Coccosteus and Dinichthys there is little indication of shearing wear on the outer side, but only a blunting wear on the dorsal edge itself. The only sign of the prominent prongs that occur anteriorly in Coccosteus and Dinichthys are slight convexities of the dorsal edge. In PF 509 (fig. 101, A) these convexi- ties form the tops of two oblique ridges on the lateral face, extending postero-ventrally to the ventral edge. They may mark the position of two relatively large teeth that are completely surrounded by bone. The convexity at the top of the anterior ridge forms the most ante- rior part of the dorsal edge and is the region from which the bone radiation spreads. Below this convexity is the antero-median edge of the infragnathal, which is set with three teeth comparable to the 3-6 "symphysial teeth" of Coccosteus. In PF 511 (fig. 101, C) the upper- most of these teeth is worn on top, its wear surface being continuous with that of the dorsal edge of the bone behind. In PF 509 (fig. 101, A) the dorsal wear surface is not worn down to the upper symphysial tooth, but there is wear on the uppermost part of the antero-lateral face, and this has involved slight wear of the side of the upper tooth. This approaches the condition found in a specimen of Coccosteus deci- piens (PF 1335) where the antero-lateral faces of the upper four of the six symphysial teeth are worn, as well as the antero-lateral face of the infragnathal itself. This anterior wear is, no doubt, the result of occlusion with the "symphysial teeth" of the anterior supragnathal. DENISON: EARLY DEVONIAN FISHES 495 PF 512 (fig. 101, E) is a very small infragnathal, incomplete pos- teriorly, and may represent a young individual of the same type as is represented by the lower jaws described above. On its dorsal edge fourteen small teeth are distinguishable; they decrease in size anteriorly and blend into the smooth, worn, anterior biting edge. The lateral face shows the two ridges described above, the posterior one extend- c D ^ Fig. 101. Arctolepida indet., infragnathals (X 3/2, except E). A, PF 509, lateral view; B, PF 510, lateral view; C, D, PF 511, medial and ventral views; E, PF 512, lateral view (X 3); F, PF 984, lateral view. 496 FIELDIANA: GEOLOGY, VOLUME 11 ing postero-ventrally and lacking any convexity on the biting edge, the anterior one nearly vertical and terminating dorsally in a high, tooth-like projection on the biting edge. There are two prominent symphysial teeth. PF 984 (fig. 101, F) is quite a different type of infragnathal and may represent an acanthodian rather than an arctolepid. It is incom- plete anteriorly and is missing a section in its posterior half, though there is a contact on the underlying rock that gives its length. The antero-posterior convexity indicates that it is the lateral face that is exposed. This infragnathal is relatively long and shallow compared to those described above. Ventrally it is grooved for Meckel's cartilage. The medial lamina bordering the groove is deep posteriorly and shallow anteriorly, while in the lateral lamina the reverse is true; as a result the groove is open laterally in its posterior part. The middle section of the preserved part is set with small teeth on its biting edge, twenty of them being distinguishable. Posteriorly they are larger, closely spaced, and inclined somewhat backward; anteriorly they decrease in size, perhaps because of wear, until they blend anteriorly into the smooth anterior part of the biting edge. The symphysial region is not preserved. The collection of the Water Canyon Formation contains several specimens that may be dermal bones of the upper jaw, but I am un- able to identify them definitely with any of the anterior or posterior supragnathals described in the Brachythoraci or in Ph.lyctaenaspis. One type is represented by PF 981 and 1645. The face illustrated in figure 102, A, lacks a smooth bone surface and consists of honeycomb- like reticular bone; this may have attached to the palatoquadrate. Adjacent to it is a narrow face (fig. 102, A, B) that is quite flat and smooth except for an isolated tubercle or tooth at one end. Opposite the latter is a narrow, convex face (fig. 102, C) that is nearly smooth except for several tubercles at one end. The face opposite that for the palatoquadrate (fig. 102, C) is broad, lacks any ornament, and may have occluded with the infragnathal. Another type of bone (PF 526, 979) is built on a similar plan, but is smaller, of different shape, and without any tubercles or teeth. The broadest face (fig. 102, D) has exposed reticular bone, similar to that of PF 1645, and may have attached to the palatoquadrate. The opposite face (fig. 102, E) is relatively narrow and deeply grooved and may have been occlusal. Finally, there are three rather flat, externally tuberculated, dermal bones that probably belong to the cheek. PF 540 (fig. 103, A, B) is a DENISON: EARLY DEVONIAN FISHES 497 five-sided plate with two long, subparallel sides ia-b, d-e), with one end pointed {h-c, c-d) and the other slightly convex ia-e). There are no external overlap areas, and while internal ones are possible on sides a-h, h-c, and d-e, they are not certain. At all of the margins except a-e, the bone tapers to a thin edge. Margin a-e is thick and bluntly cut off and probably abutted against another plate. It is possible that PF 540 is a postsuborbital and that it joined the suborbital at this edge. In Dinichthys and Coccosteus the suborbital externally overlaps the postsuborbital, and the latter is a relatively small, triangular plate, but the suborbital of Aethaspis utahensis (PF 1403) has a bluntly I Fig. 102. Arctolepida indet., possible supragnathals (X 3). A-C, PF 1645; D, E, PF 526. 498 FIELDIANA: GEOLOGY, VOLUME 11 truncated posterior end and must have had a harmonic rather than an overlapping suture here. The internal face of PF 540 (fig. 103, B) has a sinuous ridge extending from the pointed end (c) toward the middle of the opposite edge (a-e). This ridge is of solid bone and so is almost certainly not the quadrate; since its surface is relatively smooth it is not at all certain that the quadrate attached here. For this reason a definite identification of this bone as a postsuborbital is not possible. PF 530 (fig. 103, C, D) is an elongate bone with one long edge (a-b) somewhat concave and the opposite one (c-d) strongly convex, and with short, blunt ends (h-c, a-d). At the margins it tapers to a thin edge all around, so that it could not form a harmonic suture with the posterior edge of the suborbital. The external face (fig. 103, C) is tuberculated except for an area extending from near corner c to the middle of edge a-b; this smooth area presumably was overlapped by an adjacent bone. The inner surface (fig. 103, D) is rather featureless except for an area underlying the external overlap area ; here there is an elongate depression marked off by a ridge extending from corner c toward the center of the bone, and then curving toward the middle of edge a-b. The appearance of this depression indicates that cartilage may have attached here, and this, of course, suggests that it may be a postsuborbital, but other characters of the bone do not support this theory. It may be the internal bone, lying between the suborbital and postsuborbital and the cranial roof, but it is not closely comparable to the internal bone where it is known in Brachythoraci and Phlyctaenaspis. PF 1641 (fig. 103, E, F) may represent the same element as PF 530, although its shape is more complex. The edge a-g-f-e may corre- spond to edge c-d of PF 530 but differs in having the notch e-f-g. The edge a-b-c-d may correspond to edge a-b of PF 530, but it differs in having a sharply marked projection b-c-d. However, the external face has a smooth, external overlap area exactly corresponding to that of PF 530, and beneath it on the inner surface (fig. 103, F) is a groove corresponding again to that on the inner surface of PF 530. It is possible that this also is an internal bone, and if this is so, the external, smooth area was probably overlapped by the postsuborbital. GEOLOGIC RANGE AND HABITAT OF ARCTOLEPIDA The earliest known arctolepid, and the first of the Order Euarthro- dira to appear (fig. 104), is Kujdayioioiaspis, one of the Actinolepinae. It is found in the Middle Dittonian of Great Britain and in the Old -™JiL-«. mlmmmm»mn:m,n Fig. 103. Arctolepida indet., cheek plates, outer and inner views (X 3/2) A, B, PF 540; C, D, PF 530; E, F, PF 1641. 499 500 FIELDIANA: GEOLOGY, VOLUME 11 Red, Stage I, of Podolia, both probably Upper Gedinnian in age (Denison, 1956, pp. 404, 406; White, 1956, p. 2). Other orders of Arthrodira appear slightly earlier: the Acanthothoraci (Palaeacan- thaspis) occur in the Czortkow Stage of Podolia, probably of Middle Gedinnian age, and the Radotinida (Radotina) are found in the Lochkov Limestone of Bohemia, generally considered to be Late Si- lurian but possibly Downtonian in age (Denison, 1956, p. 398). Kuj- danoiviaspis is also found in the Upper Dittonian of Great Britain, and in Stages II and III of the Podolian Old Red, thus probably persisting into the Lower Siegenian. This genus is not reported else- where, but there are forms from the lower or Kapp Kjeldsen Division of the Wood Bay Series in Spitsbergen that resemble it, at least in the pattern of the dermal bones of the cranial roof. One of these is "Sval- bardaspis" angulatus (Heintz, 1929, p. 58), and others are yet to be described. These are also Lower Siegenian in age. No later Actinolepinae are reported from Great Britain, and none at all are known with certainty from France, Belgium, or Germany. In Spitsbergen Lataspis is found in the middle or Lykta Division of the Wood Bay Series, whose age is perhaps Emsian. It has been re- ferred by F0yn and Heintz (1943, p. 19) to Actinolepis but is best retained as a distinct genus at present. Relatives of Actinolepis from the Lykta Division include "Svalbardaspis" rotundus, "S." polaris, and certain undescribed forms. A large member of the Actinolepinae from Stage III of the Podolian Old Red has been referred to Lataspis by Brotzen (1934, p. 119), but it is not generically determinable at present. The only other Spitsbergen genus that may belong to the Actinolepinae is the inadequately known Mediaspis from the Grey Hoek Series at Grahuken. Its age is uncertain, but it may be Emsian and thus older than the Grey Hoek Series elsewhere. The typical Actinolepis occurs in the Baltic Middle Devonian, specifically in the Pernau, Narowa, and Luga Stages of Givetian age. It is the latest surviving representative of the Actinolepinae. In North America, Actinolepinae occur in the Beartooth Butte Formation at Beartooth Butte, Wyoming (Anarthraspis and Bryant- olepis), and in the Water Canyon Formation of Utah (Aethaspis, Bryantolepis, and Simhlaspis) . A provisional correlation, based largely on pteraspids, assigns these formations to the Lower or Middle Sie- genian, approximately equivalent to the Upper Dittonian or Lower Breconian of Great Britain. The Phlyctaenaspinae do not appear as early as the Actinolepinae, the first recorded being Lower Siegenian in age. They occur in the DENISON: EARLY DEVONIAN FISHES 501 lower or Kapp Kjeldsen Division of the Wood Bay Series of Spits- bergen. Here the typical form is Arctaspis, a genus that approaches the Actinolepinae in some respects. The same beds contain Elegant- aspis and some "Svalbardaspis" that may belong to this subfamily {"S." stensioi and perhaps "S." typicus). Upper Siegenian Phlyctaen- aspinae occur in the Wahnbachschichten of Germany; these are Prosphymaspis and forms referred, though perhaps not correctly, to Phlyctaenaspis. In the Upper Siegenian Hunsriickschiefer is another " Phlyctaenaspis," as well as "Coccosteus" angustus, a form that is intermediate between the Phlyctaenaspinae and Coccosteidae. In the upper part of the Early Devonian, probably Emsian, the typical Phlyctaenaspis occurs in New Brunswick and possibly in Quebec also. This genus is reported from the German Emsian, but the identification is not certain. In the latter there is also Prosphym- aspis and Diodsomaspis. In probably equivalent beds in Spitsbergen, the Lykta Division of the Wood Bay Series, Arctolepis is the charac- teristic form. "Svalbardaspis" stensioi, which appears to be related, possibly persists into the Lykta Division. In the upper or Stj0rdalen Division of the Wood Bay Series there occurs Heterogaspis, a poorly defined genus, some or all of which may belong to the Phlyctaen- aspinae. This division may belong, at least in part, to the Middle Devonian, and this is certainly true of the Grey Hoek Series, which also contains Heterogaspis and another of the Phlyctaenaspinae, Huginaspis. One species of Phlyctaeyiaspis, P. shenvoodi, is reported from the base of the Late Devonian in New York state (Denison, 1950). This is the last of undoubted Phlyctaenaspinae, although Gronlayidaspis from the Arthrodire Sandstone of eastern Greenland may also belong here; its age is very late Devonian or possibly even early Carboniferous. Williamsaspis, from the Middle Devonian of Australia, is a spec- ialized arctolepid that is placed in a family of its own. Arctolepida occur in fresh-water, marginal-marine, and marine sediments. However, the earliest Actinolepinae (Kujdanowiaspis) and Phlyctaenaspinae (Arctaspis, Elegantaspis, and "Svalbardaspis" stensioi) are restricted to fresh-water deposits. This and other consid- erations suggest that it was in rivers that Euarthrodira had their origin (Denison, 1956, pp. 426-427). Certain later genera in both sub- families remained in this environment; these include Lataspis and Actinolepis among the Actinolepinae, and the typical Phlyctaenaspis, Arctolepis, some Heterogaspis, and Gronlandaspis among the Phlyc- SPITSBERGEN GERMANY 5 z v: ■< :z; GIVETIAN >■ < ^ W 9 iJ UJ Q O tn Huginaspis 1— H 5 Heterogaspis Z < ^ EIFELIAN i ^■ ^v Huginaspis •J w Ci. s z a < Q o >/Arcto/epis 3 X Diadsomaspis CO \A/ediaspis ■5 N m Heterogaspis\ 5 ^ ? Phlyctaenaspis EMSIAN 5 w Lataspis \. § X K ? ^ J -x Prosphymaspis >■ < C3 ;< > Arctolepis ii ? Phlyctaenaspis c "Svalbardaspis" 1 ^ :z; 1 ^ < = s Prosphymaspis h— I " R ;z; z to J2 o ? Phlyctaenaspis > J S g SIEGENIAN Ei3 3 § 1 "Coccosteus" angustus a. ^ Elegantaspis M £- 5 s Arctaspis Z X 5 g Taunaspis H "Svalbardaspis" H O" z 5 a X < >■ < z z GEDINNIAN 5 —^ 5 .^-^— - Fig. 104. Correlation chart showing the occurrence of Arctolepida in the Early and Middle Devonian. 502 AUSTRALIA BALTIC REGION NEW YORK f f PODSNETOGOR Oredesch LUGA Narowa Pernau Phlyctaenaspis Actinolepis Actinolepis Actinolepis Williamsaspis QUEBEC NEW BRUNSWICK GREAT BRITAIN ? Phlyctaenaspis Phlyctaenaspis PODOLIA UTAH WYOMING Kujdanowiaspis Kujdanowiaspis 5 ^ cj. Lataspis Kujdanowiaspis Anarthraspis Aethaspis Bryantolepis Simblaspis Kujdanowiaspis Kujdanowiaspis < I— I > o :z; < P— H fa iz; t-H :zi •—I o C/2 :z; ;z; l-H P Fig. 104 {continued). Correlation chart showing the occurrence of Arctolepida in the Early and Middle Devonian. 503 504 FIELDIANA: GEOLOGY, VOLUME 11 taenaspinae. Others moved into habitats along the sea margin; these include Aethaspis, Simblaspis, possibly Anarthraspis and Bryantole- pis, as well as some of the German forms referred to Phlyctaenaspis. Still others were more definitely inhabitants of the sea. In Spitsbergen these include Mediaspis, Huginaspis, and some Heterogaspis. In Ger- many the marine genera are Prosphymaspis, Diadsomaspis, some ''Phlyctaenaspis" and "Coccosteus" angustus. The Australian Wil- liamsaspis of the family Williamsaspidae is a marine form. MORPHOLOGY OF THE ARCTOLEPIDA Before considering the phylogenetic interrelationships of the va- rious genera of Arctolepida, it is necessary to have some conception of what is primitive and what is specialized within the suborder. This problem is confused by a difference of opinion regarding the evolution of the pectoral fins. Stensio (1944) and White (1952) have assumed that a eurybasal fin, such as occurs in some Brachythoraci, is primi- tive, although they may concede that Arctolepida are primitive in other respects. There is no evidence to support this theory, and the geological occurrence is strongly opposed to it. Heintz (1932a, 1933, 1938) and Westoll (1945), on the other hand, have considered Arc- tolepis to represent a primitive member of the Euarthrodira and have seen the general evolutionary trends within the order to be exempli- fied by the series: Arctolepis-Phhjctaenaspis-Coccosteus-Dinichthys. Evidence of the evolution of the group may be derived from their geological occurrence, and according to this, Kujdanowiaspis should be a primitive euarthrodire, as it is the earliest known representative. However, since the occurrence may in some cases be misleading be- cause of the persistence of primitive forms or the precocious develop- ment of advanced forms, the comparative morphology of the Arctole- pida will be discussed in some detail to see what light it throws on the problem. Cranial Roof The nuchal plate (fig. 105, NU) is commonly relatively long and narrow, pointed in front, tapering slightly posteriorly, and with its center of ossification near the center of the plate {Kujdanowiaspis, Arctaspis, some "Svalbardaspis," and Actinolepis) . The anterior pointed part projects between the posterior parts of the centrals, and in one specimen of Kujdanowiaspis this has been interpreted as a separate ossification and named the median central plate (Stensio, 1945, pp. 42, 44, fig. 10, A). Stensio considers this median central DENISON: EARLY DEVONIAN FISHES 505 as a relic, usually fused to the nuchal, of a once distinct series of median plates. To interpret this as a distinct plate is open to question, since the evidence is only bone radiation; but even if it exists, the plate could more logically be interpreted as a new variant in view of the fact that it has not been observed in other euarthrodires. In Arctolepis (fig. 105, 1) and Anarthraspis (fig. 105, E) the nuchal differs only in having the center of ossification posteriorly placed, while in Phlydaenaspis (fig. 105, G), Bryantolepis (fig. 105, C), and Simblaspis (fig. 105, D) the nuchal is relatively short and broad with the center of growth far back, particularly in the first two genera. The posterior reduction of the nuchal approaches the condition of the Brachythor- aci, but the posterior widening of the nuchal that characterizes the latter group is not known among the Arctolepida. At the other ex- treme, the most striking condition of the nuchal is found in Aethaspis (fig. 105, A, B), where it is so greatly elongated anteriorly that it meets the preorbitals and completely separates the centrals; this condition is not found in other Euarthrodira. This greatly elongate nuchal, as well as the short, broad nuchal of other genera, could best be derived from the common intermediate condition that occurs in Kujdanowiaspis (fig. 105, F), Arctaspis (fig. 105, L), and others. For this reason a relatively long, narrow nuchal with central growth center is considered to be primitive within the Arctolepida. The paired paranuchal plates (fig. 105, PAN) are typically large plates at the postero-lateral corners of the cranial roof. Their con- tacts with the adjacent plates show considerable variation in restor- ations, but their precise position may not be correct where the sutures are not visible and the boundaries have been determined from bone radiation. One peculiarity occurs in Aethaspis (fig. 105, A, B) where the paranuchals project into notches in the lateral edges of the nuchal. Very commonly each paranuchal has an anterior projection extending between the central and marginal to meet the postorbital; this is known in Kujdanowiaspis, Arctaspis, some "Sval- bardaspis," Anarthraspis, Bryantolepis, Simblaspis, one species of Aethaspis, and a relative of Actinolepis from Spitsbergen. There may be variation in this character, even in a single individual, so it is not necessarily of great systematic importance. In the Phlyctaen- aspinae there are glenoids developed on the posterior edges of the paranuchals for articulation with the anterior dorso-laterals; their significance will be considered below. Each paranuchal is traversed by the main lateral line canal (fig. 105, Ic), which makes a distinct bend on the plate; at this bend are given off the posterior pit line and the supratemporal commissure (fig. 105, pp, stc) where these 506 FIELDIANA: GEOLOGY, VOLUME 11 canals have been observed in the Arctolepida. Where the bone radiation is visible, the center of ossification occurs at the bend in the lateral line canal, and its position here has been inferred in other cases. The center of ossification is placed centrally in the plate in Kujdanoiviaspis, Arctaspis, some '^Svalbardaspis," Anarthraspis, and a Spitsbergen form related to Actinolepis. It is slightly anterior in Adinolepis and somewhat posterior in Aethaspis, Siviblaspis, Bryant- olepis, and Arctolepis. In Phlyctaenaspis (fig. 105, G) it is near the posterior edge of the plate, a condition resembling that in the Brachythoraci and involving a shortening of the posterior part of the cranial roof. The various modifications of the paranuchals in the Arctolepida are best derived from the condition in Kujdanoiviaspis, Arctaspis, and Anarthraspis, where the plates are large with central growth centers, and this is considered to be primitive within the suborder. The central plates (fig. 105, CE) show little variation within the Arctolepida. Commonly they are six-sided, each side corresponding to a contact with an adjacent plate; one of the lateral sides may not be present in those forms where the paranuchal meets the postorbital. The postero-medial borders accommodate the pointed anterior end of the nuchal except in Aethaspis, where the nuchal and preorbitals completely separate the centrals. The anterior borders form more or less transverse sutures with the preorbitals except in Aethaspis, where the preorbitals extend backward between the centrals, in Actinolepis and related forms, where the preorbitals form a pointed projection between the centrals, and in Arctolepis, where the pineal projects between the anterior ends of the centrals. The center of ossification is near the center of each plate except in Anarthraspis, where it is posterior in position, and in Actinolepis, where it ap- proaches close to the midline. The central sensory canal (fig. 105, cc) and, where they are indicated, the middle and posterior pit lines (fig. 105, mp, pp) radiate laterally from the center of ossification. In only one arctolepid, Arctolepis, does the supraorbital line (fig. 105, I, soc) extend posteriorly onto the central, a condition that is typical in the Brachythoraci. All the various features of the genera mentioned above could be derived from the usual condition found in the remaining genera, that is, a five- or six-sided central with a central ossification center, separated posteriorly by the pointed anterior end of the nuchal, and with a transverse suture with the preorbital. This condition is considered primitive within the Arcto- lepida. DENISON: EARLY DEVONIAN FISHES 507 The preorbital plates (fig. 105, PRO) show the most varied development of any of the dermal bones of the arctolepid cranial roof, with the possible exception of the rostral, pineal, and post- nasals. This may be taken as an indication of the relative plasticity of the anterior part of the skull. In Kujdanowiaspis, Arctaspis, "Spalbardaspis" angulatus, and Anarthraspis they are somewhat wider than long and are five-sided, with the sides bounding respec- tively the opposite preorbital, the pineal and postnasal, the orbit, the postorbital, and the central. Anteriorly the preorbitals are notched shallowly for the pineal. The center of ossification is near the center of each plate, and from this point the supraorbital line (fig. 105, soc) extends antero-laterally. A very shght modification of this pattern is found in Phlyctaenaspis (fig. 105, G) where the pos- terior sutures with the centrals are directed obliquely postero- medially, instead of being transverse. In Simblaspis (fig. 105, D) the preorbitals are relatively much larger, especially their posterior parts, with the result that the centers of ossification are somewhat in front of the centers of the plates. In Bryantolepis (fig. 105, C) these plates are also large but are widened laterally in such a fashion that the centers of ossification lie mediad to the centers of the plates. The preorbitals of this genus are distinctly six-sided as a result of the deep notching anteriorly by the pineal and the enlargement of the postnasals. More striking differences in the form of the preorbitals are found in other genera. In Arctolepis (fig. 105, I), and to a lesser extent in "Svalbardaspis" stensioi (fig. 105, K), the pineal and rostral have pushed their way backward between the preorbitals and the orbits have notched the cranial roof deeply, even more than in Bryantolepis, resulting in relatively narrow preorbitals that are completely separated from each other in Arctolepis. This condi- tion approaches that of the Brachythoraci, where the preorbitals are characteristically separated by the pineal and part of the rostral. In Aethaspis (fig. 105, A, B) the preorbitals are greatly elongate posterior to the centers of ossification, and their posterior edges haveS-shaped sutures with the nuchal and centrals; anteriorly they are notched quite deeply by the pineal and rostral. The most re- markable modification of the preorbitals occurs in Actinolepis (fig. 105, H) and related forms from Spitsbergen (fig. 105, J). In these the centers of ossification are far posterior and near or at the mid- line, resulting in the fusion of the two plates. The growth is almost entirely anterior to the centers of ossification, and the anterior end of the combined preorbitals is deeply notched by the rostral and pineal. A close approach to the Actinolepis condition is seen in -, ,PN Fig. 105. Restorations of cranial roofs of Arctolepida, all reduced to the same area; centers of ossification stippled. A, Aethaspis utahensis, from PF 1405 (X 0.6 ); B, A. major, from PF 503, 917-919 (X 0.3); C, Bryantolepis brachycephalus, from PF 158, 1542, 1544 (X 1.3); D, Simblaspis cachensis, from PF 504 (X 0.7); E, Anarfhraspis sp., from PF 1533 and Princeton 13659 (X 0.3) ; F, Kujdanowiaspis sp., based largely on specimen from Old Red, II, Podolia (X 1.2). 508 />'PN PMG Fig. 105 (continued). G, Phlyctaenaspis acadica, from British Museum (Nat- ural History), P 5972 (X 0.5); H, Actinolepis tubercnlata, from specimen figured by Gross, 1940 (X 0.5); I, Ardolepis solnordali, modified from Heintz, 1929 (X 0.8); J, Spitsbergen genus related to Actinolepis, from specimen in Stockholm (X 0.4); K, " Svalbardaspis" stensioi, from specimen figured by Heintz, 1929 (X 0.4); L, ?Arctaspis sp., based on specimen from lower part of Wood Bay Series, Spits- bergen (X 0.5). CE, central; MG, marginal; NU, nuchal; PAN, paranuchal; PI, pineal; PMG, postmarginal; PN, postnasal; PRO, preorbital; PTO, postorbital; RO, rostral; cc, central canal; ifc, infraorbital canal; Ic, main lateral line; mp, middle pit line; ope, "opercular" canal; pfc, profundus canal; poc, preopercular canal; pp, posterior pit line; soc, supraorbital canal; stc, supratemporal commissure. 509 510 FIELDIANA: GEOLOGY, VOLUME 11 ''Svalbardaspis" rotundus and "S." polaris (Heintz, 1929, figs. 23-24). All of the modifications of the preorbitals that occur in the Arcto- lepida can be derived most easily from the condition of Kujdanowi- aspis, Arctaspis, and Anarthraspis simply by changes in the position of the center of ossification and of the relative amount of growth in different directions, and by the posterior movement of the rostral and pineal in the midline. For this reason the type of preorbital exhibited by these genera is considered to be primitive within the suborder. The postorhital plates (fig. 105, PTO) show only minor variations within the Arctolepida. Each plate is traversed by the central sensory canal (fig. 105, cc), the infraorbital canal (fig. 105, ijc), and, in Bryontolepis, by the profundus line (fig. 105, C, pfc). The center of ossification is at the point where these lines meet. It is nearly central in the plate in Arctolepis, Actinolepis, Aethaspis, Simhlaspis, and some "Svalbardaspis," slightly lateral to the center in Arctaspis and Bryantolepis, and antero-lateral in position in Kujdanowiaspis, Anarthraspis, Phlyctaenaspis, and "Svalbardaspis" stensioi. The only important modification of the preorbital is a deep notching of the lateral edge by the orbit; this is pronounced only in Arctolepis (fig. 105, I), although it is foreshadowed in "Svalbardaspis" stensioi (fig. 105, K). It is an approach to the condition of the Brachythoraci, where the orbits are characteristically large and form deep notches in the cranial roof. In view of their rare occurrence, the deep orbital notches must be considered as a specialization within the Arcto- lepida. The marginal plates (fig. 105, MG) are relatively small elements traversed by the main lateral line canal (fig. 105, Ic), the otic branch of the infraorbital line (fig. 105, ifc), and the preopercular line (fig. 105, poc). The center of ossification is at the point where these canals meet, and generally it lies in the median or postero-median part of the plate. In other words, there is very little growth posterior to the preopercular canal and median to the main lateral line and infraorbital canal. In Phlyctaenaspis (fig. 105, G) and Actinolepis (fig. 105, H) the ossification center is more centrally placed, and this is related to the fact that the paranuchal and postorbital do not grow medial to the marginal, at least in the individuals upon which the restorations are based. There is little form variety upon which to base any opinion on what is primitive within the Arcto- lepida. The postmarginal plates (fig. 105, PMG) are small elements forming the lateral corners of the cranial roof. Since they are some- DENISON: EARLY DEVONIAN FISHES 511 times detached, or incompletely known for other reasons, little can be said about their development within the Arctolepida. Only two genera show any outstanding differences: in Phlyctaenaspis (fig. 105, G) they are relatively large, and in Aethnspis (fig. 105, A, B) they are very small and crescentic in shape. Both of these develop- ments may be considered as specializations within the suborder. The rostral (fig. 105, RO), pineal (fig. 105, PI), and paired postnasal (fig. 105, PN) plates are almost invariably fused together in known specimens and are distinguishable only by their radiation from distinct centers of ossification. One specimen figured by Bryant (1934, pi. 24, fig. 3) represents a separate rostral, presumably of Anorthraspis. In Aethaspis the postnasals may or may not be fused to adjacent bones. The rostral, pineal, and postnasals form the roof over separate endocranial elements, which are often ossified perichondrally as paired circumcapsular bones (Stensio, 1945, p. 8). This may account for the fact that these dermal roofing bones are frequently not fused to the rest of the cranial roof and are absent in many specimens. They are known to be fused in Aethaspis, Bnjant- olepis, Arctolepis, and "Svalhardaspis" stensioi, they are sometimes fused in Kujdanowiaspis, Actinolepis and related forms from Spits- bergen and possibly in Arctaspis and Phlyctaenaspis, while they appear to be detached always in Anarthraspis, Simblaspis, and the other species of "Svalbardaspis." In Kujdanowiaspis (fig. 105, F) the rostral and pineal are rather short, wide plates; the pineal usually has a slightly convex posterior edge that fits into the concave anterior edges of the paired preorbitals; and the postnasals are moderately large plates with strong lateral projections that form the anterior edges of the orbits. The situation is essentially the same in Phlyct- aenaspis (fig. 105, G), while Anarthraspis (fig. 105, E) differs only in having relatively longer, narrower rostrals, and pineals and post- nasals that do not project so far laterally in front of the orbits. Other genera show more considerable differences. In Bryantolepis (fig. 105, C) the pineal is enlarged and deeply notches the anterior borders of the preorbitals; the postnasals are strikingly large, although they do not appear to enter into the anterior borders of the orbits. In Aethaspis (fig. 105, A, B) all of these anterior plates are greatly reduced ; the rostral and pineal form a small plate occupy- ing a notch in the anterior ends of the preorbitals, while the post- nasals are mere vestiges that are sometimes missing and at other times form a small projection at the anterior ends of the preorbitals. In Actinolepis (fig. 105, H) the postnasals are probably much re- duced, while the rostral and pineal are enlarged antero-posteriorly 512 FIELDIANA: GEOLOGY, VOLUME 11 and lie almost entirely between the expanded anterior ends of the preorbitals. This situation is foreshadowed in related forms from Spitsbergen (fig. 105, J). In Arctolepis (fig. 105, I) the pineal and the greatly enlarged rostral completely separate the preorbitals, the pineal even notching the anterior ends of the centrals; it approaches in this respect typical Brachythoraci. " Svalhardaspis" stensioi (fig. 105, K) has a condition intermediate between Arctolepis and Kujdan- owiaspis. These anterior cranial plates are surely specialized in Arctolepis, Actinolepis, and Aethaspis, and probably in Bryantolepis. Since they could all be derived from the condition in Kujdanowiaspis and Phlyctaenaspis, and since intermediates are known in some cases, the latter genera are considered to be primitive in this respect. Cheek Plates The suborbital plate (fig. 106) is known in a few Arctolepida, while other plates of the cheek region are known with certainty only in Phlyctaenaspis acadica, where they have been described by Heintz (1933). The type of Anarthraspis montanus, described by Bryant (1932, p. 249, pi. 10, fig. 1) as a rostral and pineal, is actually a suborbital, as is shown by its asymmetrical radiation, and there are other suborbitals known in this genus (fig. 106, E). Bryant has figured (1932, pi. 7, fig. 1) a poorly preserved suborbital of Bryant- olepis, and another suborbital of this genus (PF 1597) is restored in figure 106, C. Finally, there is a well-preserved suborbital of Aethaspis utahensis (figs. 94, C ; 106, F) described above. The most striking feature of the suborbital of Aethaspis and Anarthraspis is the short and relatively deep anterior or suborbital process. This process bounds the orbit below and meets the postnasal in front, so that its shortness in these genera is a clear indication of the small- ness of the orbits. This character is also suggested by the small orbital notches on the cranial roof, and is rather characteristic of the Arctolepida. Somewhat larger orbital notches occur in Phlyctaen- aspis, Actinolepis, Simblaspis, and Bryantolepis, and correlated with this the suborbitals of Phlyctaenaspis and Bryantolepis are seen to have relatively longer suborbital processes. Because of their common occurrence, small eyes and accompanying features are assumed to be primitive within the Arctolepida. In Brachythoraci the eyes are usually large and may be huge in some of the Wildungen genera, and in correlation with this character the suborbitals may be greatly modified. The orbits of Arctolepis, although they notch the cranial roof deeply because of their more dorsal position, were not much enlarged; the suborbital is not known in this genus. DENISON: EARLY DEVONIAN FISHES 513 The blade or posterior part of the suborbital is relatively large in Aethaspis, Phlyctaenaspis, and Anarthraspis. In Anarthraspis it has a posterior projection that is not known in other euarthrodires. In Aethaspis utahensis the blade has an unornamented dorsal border Fig. 106. Right suborbital plate of Arctolepida and Coccosteidae, all reduced to the same area. A, Coccosteus decipiens, after Heintz, 1931 (X 0.8); B, C. minor, PF 1101 (X 4.4 ) ; C, Bryantolepis brachycephalus, PF 1597 (X 2.8) ; D, Phlyctaenaspis acadica, British Museum (Natural History), P 6555 (X 1.0); E, Anarthraspis sp., restored from PF 1538, 1535 (X 0.9); F, Aethaspis utahensis, PF 1403 (X 1.2). which must have formed a movable joint with the postorbital; the blade is very long and squarely truncate behind. Phlyctaenaspis has an angulate posterior border divided into a postero-dorsal edge presumably for the internal plate and a postero-ventral border for the postsuborbital. In Bryantolepis the blade is relatively small. The infraorbital sensory canal always passes ventrally down the blade, then anteriorly along the suborbital process. From this the supramaxillary canal extends postero-ventrally, but this canal has not been seen in Anarthraspis and Bryantolepis. In Coccosteus deci- piens (fig. 106, A) the center of ossification is at the junction of these canals, but in Aethaspis it is on the infraorbital canal some distance postero-dorsal to this junction. Two other euarthrodires have suborbitals that resemble those of the Arctolepida. One is "Coccosteus" angustus (Traquair, 1903, pi. 6, fig, 1), in which the suborbital appears to be closely com- 514 FIELDIANA: GEOLOGY, VOLUME 11 parable to that of Phlyctaenaspis. The other is Coccosteus minor (fig. 106, B), whose suborbital resembles that of Bryantolepis very closely. Of the other plates of the cheek region very little is known in the Arctolepida. Heintz (1933, pp. 131-132) has described what may be an internal plate in Phlyctaenaspis acadica. Some plates that have been described above (pp. 496-498) may represent the post- suborbital and internal of Aethaspis. A full understanding of this region must await discovery of better material. Endocranium Since the endocranium of Aethaspis has not been fully studied and is not described in the present paper, this structure will be considered only briefly here. This discussion is based largely on the work of Stensio in several publications (1934, 1944, 1945, 1948), and also on his manuscripts and figures, which have very generously been made available to me. Additional information on the arthro- dire endocranium has been published by Gross (1937, ?Phlyctaen- aspis; 1940, Actinolepis) and White (1952, Buchanosteus) . The endocranium was largely cartilaginous but sometimes was provided with perichondral ossification. Among the Water Canyon Formation Arctolepida, perichondral bone is invariably present in Aethaspis but no sign of it has been seen in Siynblaspis. In the Bear- tooth Butte forms, it is present in Bryantolepis but appears to be absent in Anarthraspis. It is present in Kujdonowiaspis and in some Brachythoraci, but in many others there was apparently no ossi- fication. When present in the Arctolepida, the ossification may con- sist of a single element or the nasal capsules (circumcapsular bones) may be separate. Separate circumcapsular bones occur in some Kujdanoiviaspis and in an undetermined form from the lower Wood Bay Series of Spitsbergen. They are fused to the postethmoid ossi- fication in other Kujdanowiaspis and in Bryantolepis. In the Brachy- thoraci they are fused where known, with the exception of Buchan- osteus. It is probable that separate circumcapsular bones are asso- ciated with rostral, pineal, and postnasals that are not fused to the rest of the cranial roof. If this is so it would indicate separately ossified nasal capsules in Anarthraspis, Si^nblaspis, Phlyctaenaspis, and some Svalhardaspis, besides those mentioned above. In certain Brachythoraci (Pholidosteus and Leiosteus) the endocranium is divided into several ossifications and may develop large dorsal fontanelles. This would appear to be the result of a secondary DENISON: EARLY DEVONIAN FISHES 515 reduction in ossification, but the separately ossified nasal capsules may be primitive. The endocranium of Arctolepida is low and broad (platybasic), with thick lateral walls. The orbits are small and anteriorly placed, with wide suborbital shelves and well-developed supraorbital proc- esses. There are prominent paired lateral processes (posterior post- orbital processes of Stensio) behind the postorbital processes, and the occipital region is short. In the Brachythoraci the endocranium is usually relatively higher and narrower, and in the several genera with large eyes it may be tropibasic, with the orbital cavities very large and with the suborbital shelves and supraorbital processes reduced. The postorbital processes are placed farther back, the posterior postorbital processes are absent, and the occipital region is relatively long. The platybasic endocranium with a short occipital region and small orbits is probably primitive among placoderms, since it occurs not only in Arctolepida but also in certain Brachythoraci, probably in the Acanthothoraci, and in a modified form in the Antiarcha. The endocranium of many Brachythoraci is surely specialized in its narrowness and height, in the large orbits, and in the lengthened occipital region. Buchanosteus and perhaps Coccosteus are inter- mediate between the Arctolepida and Brachythoraci in possessing the primitive arctolepid type of endocranium in conjunction with the typical dermal cranial roof of the Brachythoraci. Trunk Shield The median dorsal plate takes two very different forms in the Arctolepida. In the Actinolepinae (fig. 107, D-I) it is short and broad and has a nearly central ossification center. In the Phlyctaenaspinae (fig. 107, A-C) it is long and narrow and has the center of ossification placed well behind the center. Strictly intermediate forms are not certainly known in Arctolepida; however, "Coccosteus" angustus has a relatively short median dorsal compared to that of the Phlyctaenaspinae, and the typical Coccosteus median dorsal is inter- mediate in shape if one forgets the posterior spine. Determination of which condition is primitive within the Arctolepida is not possible on purely morphological grounds, though the central ossification center of most Actinolepinae has the appearance of a simple and unmodified condition, while the posterior ossification center of the Phlyctaenaspinae may be a modification involving an anterior elon- gation. Stensio (1944, pp. 25-26; 1948, p. 209) is of the opinion that 516 FIELDIANA: GEOLOGY, VOLUME 11 the single median dorsal of the Euarthrodira is a complex plate formed by the fusion of homologues of the anterior and posterior median dorsal plates of the Antiarcha. If this were true, the long median dorsal of the Phlyctaenaspinae might be primitive, and the laterally notched median dorsal of Prosphymaspis (fig. 107, A) might be taken to indicate an incomplete fusion of the two elements. Unfortunately for this theory there is no indication of two centers of ossification in any euarthrodiran median dorsal and, as Parrington has pointed out (1956, p. 410), assumptions of bone fusions without direct evidence is unjustified. A single median dorsal growing from a single ossification center is certainly primitive within the Arctolepida, although the possibility cannot be ruled out that the euarthrodiran ancestor possessed two median dorsals, one of which has been lost. All Arctolepida lack the median keel on the inner side — a feature that is characteristic of the Brachythoraci. Some Arctolepida do possess a low, median, inner ridge from which the keel could have been derived. The median dorsals of the Actinolepinae (fig. 107, D-I) show little variation in shape. All have a well-developed postero-median projec- tion, and usually there is a slight convexity of the anterior margin in the median line, although this is absent in Bryantolepis and Anarthraspis. Simhlaspis, Aethaspis, Actinolepis (fig. 107, D, E, G, x), and "Phlyctaenaspis" heintzi have a smooth, untuberculated area on either side of this anterior convexity. Since the trunk shield and cranial roof are very close in Arctolepida, these areas may actually have been overridden by the cranial roof when the head was elevated. Typically the median dorsal of the Actinolepinae has a dorsal crest in the posterior half of the plate. Among the Phlyctaenaspinae, the shape of the median dorsal is not adequately known in most Spitsbergen forms because of the complete fusion of the bones of the trunk shield, but it is certainly long and relatively narrow in all. The most elongate median dorsal occurs in the Late Devonian Phhjctaenaspis sherivoodi (Denison, 1950, fig. 1, B), while P. acadica (fig. 107, B) has a more moderately proportioned median dorsal. The latter also has a very distinct postero-median process, which foreshadows the posterior spine of Coccosteus. " Phlyctaenaspis" heintzi resembles other members of this subfamily in most respects but differs in having a short, wide median dorsal like that of the Actinolepinae; if this median dorsal has been correctly assigned to this species, this form may represent an inter- mediate between the two subfamilies. The most remarkable median dorsals occur in Huginaspis (Heintz, 1929, pi. 20, fig. 2), where the S ADL ■^-s PDL" ,s ADL- ■~s. PDL ^, ADL Fig. 107. Median dorsal plates of Arctolepida, all reduced to the same area; centers of ossification stippled. A, Prosphymaspis consirida, after Gross, 1933a (X 2.0); B, Phlydaenaspis acadica, modified from Heintz, 1933 (X 0.5); C, Arcto- lepis decipiens, restored from Heintz, 1929 (X 0.9); D, Adinolepis tuberculata, after Gross, 1940 (X 3.8); E, Aethaspis major, restored from PF 907, 910 (X 0.4); F, Anarthraspis sp., restored from PF 1531 (X 0.4); G, Simblaspis cachensis, restored from PF 301 (X 0.4); H, Bryantolepis brachycephalus, restored from PF 162-164 (X 1.2); I,Kujdanowiaspiss-p., based on cast of Podolian specimen, PF 1178 (X 0.8). s. ADL, s. PDL, extent of inner overlap areas for anterior and posterior dorso-laterals; x, external unornamented area. 517 518 FIELDIANA: GEOLOGY, VOLUME 11 plate has a very high median dorsal crest, in Prosphymaspis (fig, 107, A), where it is deeply notched laterally by the posterior dorso- laterals, and in Diadsomaspis (Gross, 1937, fig, 12, N), where it is narrow anteriorly between the anterior dorso-laterals and widens posteriorly between the posterior dorso-laterals. All of these condi- tions are considered to be specializations because of their unique occurrences. Posterior dorsal plates have been described above (p. 482) in Aethaspis, where they have been compared with similar plates of Bryantolepis, Anarthraspis, and Kujdanowiaspis. Their bilateral symmetry and their highly arched form in Aethaspis (figs. 93, 94, B) leave little doubt that they lay in the dorsal midline behind the median dorsal. It is possible that they were generally present in Arctolepida but have not yet been recognized in many genera. Thus the plate described by Gross (1937, fig. 12, B) as the median dorsal of Phlydaenaspis pusilla could be a posterior dorsal, and the same is true of the so-called median dorsal of Taunaspis eurystethes (op. cit,, fig, 13, D) and of Murmur arctatus (Bryant, 1935, pi. 10, fig. 2). All of these differ from typical median dorsals in their shape. The known posterior dorsals of Actinolepinae are relatively large plates. Those of Phlydaenaspis acadica, on the other hand, are quite small. These plates are not known in any Brachythoraci, and it is probable that they were absent in this suborder. But it is possible that com- parable elements were present in the Petalichthyida; Lunaspis has the posterior part of the body covered with scales, and at least two of these in the median line behind the median dorsal are enlarged to resemble small posterior dorsals. The posterior median dorsal of the Antiarcha may be a homologue of the arctolepid posterior dorsal which has become incorporated into the lengthened trunk shield. The occurrence of these plates in two or three distinct orders of Arthrodira suggests that they were present in the common ancestor; Fig. 108. Right anterior dorso-lateral plates of Arctolepida, all reduced to the same area. A, Prosphymaspis subtilis, modified from Gross, 1933b (X 1.8); B, Phlydaenaspis acadica, after Heintz, 1933 (X 0.8); C, "Phlydaenaspis" heintzi, modified from Gross, 1933b (X 1.0); D, Diadsomaspis elongala, after Gross, 1933a (X 1.1); E, relative of Adinolepis from Spitsbergen, from cast, PF 1183 (X 0.9); F, Aethaspis major, restored from PF 532, 922, 971 (X 0.7); G, Anarthraspis sp., PF 263 (X 0.9); H, Bryantolepis brachycephalus, restored from PF 438, 1600 fX 2.7); L Kujdanowiaspis sp., based on cast of Podolian specimen, PF 1180 (X 1.5). Ic, main lateral line; s. AL, s. MD, external overlap areas for anterior lateral and median dorsal; s. PDL, extent of internal overlap area for posterior dorso- lateral; tr, trochlea for articulation with paranuchal. ! I s MD 519 520 FIELDIANA: GEOLOGY, VOLUME 11 in this case their loss in the Brachythoraci and other groups would be secondary. The paired anterior dorso-lateral plates differ in the Actinolepinae (fig. 108, E-I) and Phlyctaenaspinae (fig. 108, A-D) in one important respect: in the latter there is a well-developed trochlea which was received in a glenoid fossa on the paranuchal to form an exoskeletal articulation between the head and trunk; in the Actinolepinae there is no trochlea or glenoid, but on the external face there is a smooth, antero-dorsal, untuberculated area which was overlapped by the posterior edge of the paranuchal. The Phlyctaenaspinae resemble the Brachythoraci in which the trochlea-glenoid articulation is almost always well developed. The absence of the differentiated articular surfaces in the Actinolepinae does not mean that no movement was possible between the head and trunk but does suggest that this movement was very limited. An endoskeletal articulation was, of course, also present (Stensio, 1945, p. 7). Because of its highly differentiated nature, the articulation of the Phlyctaenaspinae and Brachythoraci must be considered as a specialization, while the simple overlap of the anterior dorso-laterals by the paranuchals is presumed to be primitive in euarthrodires. No clearly intermediate forms showing how the articulation developed have been described. It is interesting that in the Antiarcha the articulation is reversed, with the trochlea on the cranial roof and the glenoid fossa on the anterior dorso-lateral; this situation may signify that the differenti- ated articulation was independently acquired in the Antiarcha and Euarthrodira. Among the latter, the Actinolepinae surely represent an early stage in the development of the exoskeletal articulation. Limited movement between the head and trunk is indicated not only by the simple overlapping of the anterior dorso-laterals by the paranuchals but also by the small space dorsally between the head and trunk shields. In many Phlyctaenaspinae there is also very little space between the cranial roof and the trunk shield; in these forms (Arctaspis, Arctolepis, Elegantaspis, Gronlandaspis) the articular trochleae are placed very close to each other in the most antero- dorso-median part of the anterior dorso-laterals. Movement of the head on the trunk must have been very small here also and probably was more definitely limited to a vertical direction than in the Actinolepinae. In Phlyctaenaspis acadica (fig. 108, B) the trochleae are better developed and more widely spaced, so probably a greater degree of movement was possible (Heintz, 1933, fig. 5, A). In the Brachythoraci the trochleae are typically well developed and placed far laterally; usually there is such a wide gap between the posterior DENISON: EARLY DEVONIAN FISHES 521 edge of the cranial roof and the median dorsal that considerable movement may have been possible, although in at least one form, Coccosteus minor, the gap was partially filled by paired plates. The shape and proportions of the anterior dorso-laterals show considerable differences among the Arctolepida. In Kujdanowiaspis, Bryantolepis, and Diadsomaspis the exposed face is long and low, suggesting a relatively low trunk shield; in Aethaspis, Anarthraspis, Phlyctaenaspis and a relative of Actinolepis from Spitsbergen it is relatively high, indicating that the trunk shield was high. The extremely short, high anterior dorso-lateral of Prosphymaspis (fig. 108, A), with its angulate cross section, is surely a specialized form. The Actinolepinae are characterized by a projection of the posterior margin dorsal to the lateral line; this feature is absent or little developed in the Phlyctaenaspinae. The Actinolepinae also have ornamented processes of the exposed face, one extending dorsally in front of the median dorsal and the other extending ventrally in front of the anterior lateral; the dorsal process and sometimes the ventral process are reduced in the Phlyctaenaspinae. The posterior dorso-lateral plates of various Actinolepinae differ in their proportions. In Kujdanowiaspis (fig. 109, I) and Bryantolepis (fig. 109, H) the exposed face of the posterior dorso-lateral is rather shallow anteriorly, deep posteriorly, and constricted in between; the anterior shallowness agrees with the assumed low trunk shields of these genera. In Aethaspis (fig. 109, F) and Anarthraspis (fig. 109, E) the posterior dorso-lateral is deep throughout, indicating a higher trunk shield. The same is true of Phlyctaenaspis (fig. 109, A), although the ventral edge is deeply notched for the posterior lateral plate in "P." heintzi (fig. 109, D). The most striking modifications of the posterior dorso-lateral occur in Prosphymaspis (fig. 109, B) and Gronlandaspis (fig. 109, C), both of which have a very short, high posterior dorso-lateral with a deep ventral notch for the posterior lateral plate; Prosphymaspis is also distinctive in having the dorsal edge strongly convex where it projects into the notched median dorsal. In the Brachythoraci the relatively high trunk shield has resulted in a high posterior dorso-lateral; in addition this plate is reduced posteriorly, especially postero-ventrally, in connection with the reduction of the lateral wall of the trunk shield. The anterior laterals are among the most interesting plates of the trunk shield, although the modifications in Arctolepida (fig. 110) are I relatively slight compared to those that appear in Brachythoraci. t Certain genera have relatively flat plates, only slightly convex in S-MD s. PL Fig. 109. Right posterior dorso-lateral plates of Arctolepida, all reduced to the same area. A, Phlycfaenaspis acadica, after Heintz, 1933 (X 0.8); B, Pro- sphymaspis subtilis, after Gross, 1933b (X 1.2); C, Gronlandaspis mirabilis, after Stensio, 1939 (X 0.5); D, " Phhjctaenaspis" heintzi, after Gross, 1937 (X 1.0); E, Anarthraspis sp., restored from PF 1527, 261 (X 0.8); F, Aethaspis major, PF 303 (X 0.6); G, undetermined arctolepid from Utah, PF 534 (X 2.8); H, Bryantolepis brachycephalus, PF 180 (X 2.6); I, Kujdanowiaspis sp., from cast of Podolian specimen, PF 1173 (X 1.6). Ic, main lateral line; s. ADL, s. AL, s. MD, s. PL, external overlap areas for anterior dorso-lateral, anterior lateral, median dorsal, posterior lateral. 522 DENISON: EARLY DEVONIAN FISHES 523 the center, with a central ossification center and with ridges of varying strength radiating to the four corners, dividing the plate into quadrants. This is true in Kujdanowiaspis (fig. 110, H), Arctaspis, Elegantaspis, and "Phlyctaenaspis" heintzi. The central ossification center is of course a feature of an early ontogenetic stage and may be correlated to some extent with size. It could be expected in juvenile individuals, and is, in fact, found in an extremely small individual assigned by Gross (1937, p. 20) to " Phlyctaenaspis" pusilla (fig. 110, G). It could be argued that this is a primitive character when it occurs in adults in that there has been no major modification due to differential growth. In other Arctolepida there are various departures from this simple scheme. For one thing, the ossification center comes to occupy a more anterior and ventral position as a result especially of a reduction in size of the anterior quadrant. At the same time the anterior quadrant becomes turned inward (fig. 110, iiv) and is then called the "inner wing" by Heintz (1932a, p. 168), the "medial lamina" by Stensio (1944, p. 26), and the "apron" by White (1952, p. 292). Arctolepis (fig. 110, A) shows this condition to a slight extent; it is marked in Aethaspis (fig. 110, D), Anarthraspis (fig. 110, F), Bryantolepis (fig. 110, E), and Simhlaspis (fig. 110, C), while a more extreme development is found in Williamsaspis and most Brachythoraci (and of course the ptyctodonts) . Though projecting inward, the inner wing is ornamented and was formed in the dermis, thus indicating an infolding of the skin in the post-cranial, pre-shoulder region. The infolding is con- tinued ventrally by the intero-lateral, which is joined to the ventral part of the inner wing of the anterior lateral. The significance of this infolding has been discussed recently by White (loc. cit.), who has come tentatively to the conclusion that it does not represent a post-branchial wall but is the result of a neck constriction. Its development is perhaps correlated with an increased mobility of the articulation between the anterior dorso-laterals and paranuchals; where the inner wing is absent, motion of the head on the trunk must have been very slight; where it reaches its maximum develop- ment, in the Brachythoraci and Ptyctodontida, a greater amount of movement was possible. The lower quadrant of the anterior lateral forms at its ventro- lateral edge a suture for the spinal (fig. 110, s. SP). This is a non- overlapping, harmonic suture in Aethaspis, Bryantolepis, Simhlaspis, and perhaps in other euarthrodires. The spinal edge is relatively long in Arctaspis, Arctolepis, Elegantaspis, Huginaspis, Kujdanowiaspis, "Phlyctaenaspis" heintzi and P. sherwoodi, in all of which a long spinal 524 FIELDIANA: GEOLOGY, VOLUME 11 occurs; it is relatively short in Aethaspis and Anarthraspis, in which the spinal is short, and is intermediate in Bryantolepis, Phlyctaenaspis acadica, and Simhlaspis. The reduction of the spinal edge and of the whole ventral triangle of the anterior lateral is extreme in the Brachythoraci, in which the spinal may be reduced to a vestige or be absent. There can be little question that the well-developed ventral triangle with a long spinal edge is primitive in the Euarthrodira and that this part of the anterior lateral has suffered great reduction in certain Arctolepida and almost all Brachythoraci. Williamsaspis is unique among the Arctolepida in possessing a rather long spinal edge in connection with a short, non-projecting spinal; it is reasonable to conclude that this is a specialized condition. The upper triangle of the anterior lateral, which overlaps the anterior dorso-lateral and part of the posterior dorso-lateral, shows little modification within the Arctolepida. It is relatively high in Aethaspis, Anarthraspis, and Phlyctaenaspis acadica and helps to form the rather high trunk shield in these forms. In certain high- shielded Brachythoraci in which the lateral walls of the trunk shield are reduced, the anterior lateral is greatly lengthened dorso-ventrally and shortened antero-posteriorly, a condition not approached by any arctolepid. The posterior edge of the anterior lateral typically overlaps the anterior or antero-dorsal edge of the posterior lateral. More ventro- laterally it does not form a suture with the underlying anterior ventro-lateral, and its exact relationships have been somewhat obscure. Heintz (1933, p. 141) considered that the anterior ventro- lateral and anterior lateral were in contact along this edge in Phlyctaenaspis acadica. However, a specimen of Kujdanowiaspis figured by Stensio (1944, fig. 16) clearly demonstrates that the anterior lateral and anterior ventro-lateral are separated to form a fenestra filled internally by perichondral bone that formed the attachment of the pectoral fin. The same situation occurs in a specimen of Williamsaspis described by White (1952). A reconstruction of Phlyctaenaspis sherwoodi (Denison, 1950, pi. 2) indicated the presence of a pectoral fenestra in this species, and one is definitely present in Arctolepis (fig. 115), though Westoll (1945, p. 384) and others have believed it to be absent in this and related Spitsbergen genera. These are a few of the many reasons for believing that the pectoral fenestra was present in all Arctolepida. This fenestra may completely separate the anterior lateral and anterior ventro-lateral, as in Williamsaspis, or, on the other hand, the anterior lateral and anterior ventro-lateral s. SP Fig. 110. Right anterior lateral plates of Arctolepida, all reduced to the same area; centers of ossification stippled. A, Arctolepis decipiens, after Heintz, 1938 (X 1.3); B, Phlyctaenaspis acadica, after Heintz, 1933 (X 0.8); C, Simblaspis cachensis, PF 562 (X 0.6); D, Aethaspis major, restored from PF 560, 561 (X 0.5); E, Bnjantolepis brachycephalus, restored from PF 1549, 1553, UC 2210 (X 2.1); F, Anarthraspis sp., restored from PF 255, 1524 (X 0.9); G, "Phlyctaenaspis" pusilla, after Gross, 1937 (X 10.9); H, Kujdanowiaspis sp., restored from casts of Podolian specimens, PF 1163, 1164, 1166, 1176 (X 1.5). iw, inner wing; s. ADL, s. PDL, s. PL, extent of inner overlap areas for anterior and posterior dorso-laterals and posterior lateral; s. SP, edge for spinal. 525 526 FIELDIANA: GEOLOGY, VOLUME 11 may meet both mediad and laterad to a relatively small fenestra, as in Arctolepis. In other forms the pectoral fenestra extends medially to reach the posterior ventro-lateral and posterior lateral, while the anterior ventro-lateral and anterior lateral meet laterally near the spinal. This is probably the case in Aethaspis, where an internal thickening of the posterior edge of the anterior lateral indicates the presumed extent of the attachment of the pectoral fin. As restored, Phlyctaenaspis shenvoodi shows the same situation. It may be true in Kujdanowiaspis, since Stensio's specimen is incomplete laterally. The significance of the development of the pectoral fenestra will be deferred to the discussion of the pectoral fin. The posterior lateral plates have hitherto been recognized among the Arctolepida only in Phlyctaenaspis acadica (Heintz, 1933, fig. 3), "P." heintzi (Gross, 1933a, fig. 7:5), and Williamsaspis (White, 1952). In the course of the present study the posterior lateral has been found also in Aethaspis, Anarthraspis, and Bryantolepis. In all except Williamsaspis it is a small plate. In Phlyctaenaspis (fig. Ill, A, B) it is relatively long and low, overlapping the posterior dorso- lateral dorsally, with a long external overlap area for the posterior ventro-lateral along its whole ventral edge, and with a small external overlap area anteriorly or antero-dorsally for the posterior tip of the anterior lateral. The posterior lateral of Anarthraspis (fig. Ill, D) is similar except that it is relatively higher and has a longer overlap area for the anterior lateral. A plate figured by Bryant (1934, pi. 13, fig. 4) as a posterior ventro-lateral of Bryantolepis is probably a posterior lateral of Anarthraspis. In all of these the anterior lateral and posterior ventro-lateral overlap areas meet, signifying that the pectoral fenestra did not reach the posterior lateral. In Bryantolepis (fig. Ill, E) the posterior lateral is long and low with its exposed face high posteriorly but very low anteriorly; the anterior lateral overlap area is long and does not meet the ventral overlap area, being separated by a forward extension of the exposed face that must have bounded the pectoral fenestra; as will be shown below, both the anterior and posterior ventro-laterals appear to overlap the posterior lateral in this genus. Aethaspis (fig. Ill, C) has a short, high posterior lateral with a very small anterior lateral overlap area that is widely separated from the posterior ventro-lateral area; the posterior lateral must have had a wide contact with the pectoral fenestra. Kujdanowiaspis may have been incorrectly restored by Stensio (1944, fig. 17, A) as far as the relationships of the anterior lateral, posterior lateral, and posterior ventro-lateral are concerned; it seems probable that the posterior lateral bounded the pectoral I DENISON: EARLY DEVONIAN FISHES 527 fenestra, at least in the form upon which this part of the restoration was based. Williamsaspis has a very large posterior lateral, relatively short and very high, peculiarly shaped, laterally ridged, and with a rather large edge for the pectoral fenestra; presumably this is a specialized condition, since nothing approaching it is found in other euarthrodires. In the Brachythoraci the posterior lateral is almost -\^-^W s.PVL s.PVL Fig. 111. Right posterior lateral plates of Arctolepida, all reduced to the same area. A, Phlyctaenaspis acadica, after Heintz, 1933 (X l.D; B, "P." heintzi, after Gross, 1933b (X 1.2); C, Aefhaspis major, restored from PF 924 (X 0.5); D, Anarthraspis sp., PF 1541 (X 1.4); E, Bryantolepis brachycephalus, PF 183 (X 2.2). s. AL, s. PVL, external overlap areas for anterior lateral and posterior ventro- lateral. invariably reduced in connection with the posterior reduction of the trunk shield. A small but well-developed posterior lateral, relatively long and low in proportions, is believed to be primitive in the Arctolepida. The anterior ventro-lateral plates (fig. 112, AVL) are constructed on a plan very similar to that of the anterior laterals. The lateral border attaches to the spinal in a simple harmonic suture and shows modifications similar to those of the spinal edge of the anterior lateral; it is long in forms with a long spinal and tends to be short in those with a short spinal. The anterior border unites with the intero-lateral in a harmonic suture, while the antero-medial, medial. 528 FIELDIANA: GEOLOGY, VOLUME 11 and postero-medial borders overlap the antero-ventral (in Actinolepinae), anterior and posterior medio-ventrals, and posterior ventro-lateral. The postero-lateral part of the plate is of particular interest. It is arched inward to bound the pectoral fenestra ventrally, while its posterior edge is concave to form the boundary of the pectoral sinus, which lies between the body wall and the spinal. The extent of the pectoral fenestra has been considered in the discus- sion of the anterior lateral, but additional evidence may be obtained from the anterior ventro-lateral. The fenestra may extend all of the way from the spinal to the point where the anterior ventro-lateral overlies the posterior ventro-lateral, as in Williamsaspis. In Arctolepis, at the other extreme, it is probable that the anterior ventro-lateral and anterior lateral meet both medially and laterally to the small pectoral fenestra (fig. 115). An anterior ventro-lateral of Bryantolepis (PF 194, fig. 112, G, s. PL) shows very distinctly an internal overlap area in its postero-lateral corner, which could only have overlapped the anterior lateral or the posterior lateral ; an overlap of the anterior lateral by the anterior ventro-lateral is most improbable, and while the anterior ventro-lateral is not known to suture with the posterior lateral in other Arctolepida, possibly excepting Williamsaspis, this appears to be the more probable explanation of this overlap. Above the ornamented edge of the pectoral sinus of Aethaspis and Anarlhraspis there rises a smooth, unornamented flange (fig. 112, C, D, F, pf) attached to the anterior ventro-lateral and probably representing the endoskeletal attachment of the pectoral fin ; this occupies the antero- medial part of the pectoral sinus and does not extend to the spinal, suggesting that the pectoral fin base was considerably more restricted than in Williamsaspis, The postero-lateral part of the anterior ventro-lateral also varies in the shape and position of the pectoral sinus. In many forms (Arctaspis, Arctolepis, Kujdanowiaspis, " Phlyctaenaspis" heintzi, P. sherwoodi, Prosphymaspis, Heterogaspis, and Huginaspis) the sinus is rather narrow laterally and deep antero-posteriorly, with its lateral edge bounded by the spinal and its anterior and medial edges bounded by the anterior ventro-lateral and anterior lateral. Bryantolepis (fig. 112, G) differs in having an extremely narrow sinus. In other genera the pectoral sinus is more open laterally because of the shortness of the spinal and of the spinal edge of the anterior lateral and anterior ventro-lateral, and also takes a more anterior position as a result of the shortening of the antero-lateral part of the anterior ventro-lateral. This situation is found in Aethaspis, Anarlhraspis, Actinolepis, Lataspis, Phlyctaenaspis acadica, and Mediaspis; its DENISON: EARLY DEVONIAN FISHES 529 greatest development is seen in a relative of Actinolepis from Spits- bergen (fig. 112, A), where the spinal projects only slightly, and in Williamsaspis (fig. 112, J), where the sinus is not bounded laterally by the spinal. Among the Brachythoraci, only the Coccosteidae and Pkolidosteus resemble the Arctolepida in the development of the pectoral sinus. The deep pectoral sinus is considered to be primitive within the Arctolepida, while the relatively open sinus with a more laterally directed pectoral fin is considered to be specialized. The spmaZpk^es (fig. 112, 5P) are invariably present in Arctolepida, and for that matter in all Arthrodira, except for certain Brachythoraci where their absence is due to secondary loss. This alone is sufficient demonstration that the spinal represents part of the heritage of the ancestral arthrodire. Within the Euarthrodira there are all variations, from the extremely long, slender spinal of Arctolepis and Elegantaspis (Heintz, 1929, pis. 2, 16) to a mere vestige in Dinichthys and to complete loss in a number of Brachythoraci. Opinions diff'er as to what represents the primitive condition. Stensio (1944) maintains that the large spinal is specialized and was formed from the exoskeleton of the anterior part of a once more extensive pectoral fin, while White (1952, pp. 296-297) considers the small spinal of Williamsaspis to represent a primitive condition in Arctolepida. On the other hand, Heintz (1932a, 1938), Westoll (1945), and Gross (1954) consider the long spinal to be primitive. This matter will be considered in more detail in the discussion of the pectoral fin. In my opinion the primitive euarthrodires had spinals of moderate length, such as occur in Arctaspis (fig. 112, I) and Kujdanowiaspis (fig. 112, H). The Phlyctaenaspinae tended to retain long spinals and even to lengthen them in Elegantaspis and Arctolepis, while the short spinals of certain Actinolepinae such as Aethaspis, Anarthraspis, Actinolepis, and Williaynsaspis are believed to indicate a tendency toward reduc- tion, leading to the conditions found in the Brachythoraci. This belief fits in well with the stratigraphic occurrence and also receives some support from a possible ontogenetic development. As was shown above (p. 487), there is a suggestion in certain forms from the Water Canyon Formation that the spinals were relatively large in small individuals and relatively small in large individuals. If this is actually a growth phenomenon, the long-spined juveniles may be recapitu- lating a long-spined ancestral stage. The intero-lateral plates (fig. 112, IL) are adequately known in very few Arctolepida. They have been described above (p. 488) in Aethaspis, where they were shown to consist of a ventral or external 530 Fig. 112. Right anterior ventro-lateral, antero- ventral, intero-lateral, and spinal plates of Arctolepida, the AVL's all reduced to the same area. A, genus from Spitsbergen related to Adinolepis, from cast, PF 1171 (X 0.6); B, Actinolepis tuberculata, modified from Gross, 1940 (X 1.3); C, Anarihraspis sp., restored from PF 262, 1525, 1528 (X 0.4); D, Aethaspis major, restored from PF 939, 940, 983, 988-990 (X 0.6) ; E, undetermined arctolepid from Utah, PF 923 ( X 1.9 ) ; F, Aethaspis utahe7isis, PF 321 (X 1.0); G, Bryantolepis brachycephalus, restored from PF 189, 194, 1543 (X 1.7); H, Kujdanowiaspis sp., modified from Stensio, 1944 (X 1.1); I, Arctaspis hoegi, interpreted from Heintz, 1929 (X 0.6); J, Williamsaspis bedfordi, modified from White, 1952 (X 1.1); K, Phlyctaenaspis acadica, after Heintz, 1933 (X 0.6). AMV, anterior medio- ventral; AV, antero-ventral; AVL, anterior ventro- lateral; IL, intero-lateral; SP, spinal; pf, endoskeletal attachment for pectoral fin; s. AMV, s. PMV, s. PVL, extent of internal overlap areas for anterior and posterior medio- ventrals and posterior ventro-lateral; s. PL, internal overlap area, probably for posterior lateral; vp, transverse ventral pit line. 531 532 FIELDIANA: GEOLOGY, VOLUME 11 lamina and a dorsal or internal lamina; the latter meets the infolded anterior quadrant of the anterior lateral, and together they form the inner wing of the trunk shield. This intero-lateral agrees in all essen- tial respects with that of Phlyctaenaspis acadica as described by Heintz (1933, figs. 3, 5). The intero-lateral of Bryantolepis is probably similar. Williamsaspis has a very well-developed inner wing and I believe, contrary to the opinion of White (1952, p. 257), that the intero-lateral had a strong dorsal lamina which contributed impor- tantly to its formation. In Arctolepis the inner wing is rather small, and was probably formed as usual of both the intero-lateral and anterior lateral; figures of Heintz (1929, figs. 12, 13) show a distinct dorsal lamina on the intero-lateral. A specimen of Kujdanoiviaspis that has been figured by Stensio (1944, fig. 16) is interesting in that it shows the inner wing in an incipient stage of development; the anterior quadrant of the anterior lateral is turned in only slightly and meets the dorsal lamina of the intero-lateral, which is small and restricted to the lateral half of the plate. Presumably, the inner wing was small or absent in the primitive euarthrodiran. The paired antero-ventral plates (figs. 94, E; 99; 112, AV), as stated above, have not been recognized previously, or have been con- fused with the anterior medio- ventral. Characteristically they are present in the Actinolepinae and absent in the Phlyctaenaspinae, Williamsaspidae, and Brachythoraci. It is possible that they repre- sent a new ossification in the Actinolepinae that was never developed in the other groups, but I know of no evidence to support this. On the other hand, they may have been a primitive character that was lost in most later Euarthrodira. This loss may have been a result simply of failure to ossify, or it could have resulted from fusion with the neighboring intero-lateral, anterior medio-ventral, or anterior ventro- lateral. A fusion with the medial part of the intero-lateral is suggested in certain cases. The center of ossification of the antero-ventral is near its lateral point and approaches that of the intero-lateral in Bryantolepis (fig. 112, G) and some undetermined arthrodires from Utah, PF 536 and 923 (fig. 112, E). In Anarthraspis (fig. 112, C) the antero-ventral is not clearly distinguishable from the intero-lateral in any specimen that I have seen and complete fusion may have taken place, although the presence of the antero-ventral is still indicated by the expanded medial end of the combined intero-lateral and antero- ventral. I know of no case where there has been a fusion of the antero- ventral with the anterior ventro-lateral, resulting in a single center of ossification. Evidence for fusion with the anterior medio-ventral is also absent, although in Phlyctaenaspis acadica the expanded anterior DENISON: EARLY DEVONIAN FISHES 533 end of the anterior medio-ventral may include antero-ventral equiva- lents. Another possibility is that the paired antero-ventrals fused to a single median plate and that the original anterior medio-ventral was reduced and lost. If the bone radiation is correctly indicated in Heintz's photograph of Arctaspis hoegi (1929, pi. 12, fig. 1), the antero-ventrals Fig. 113. Anterior medio-ventrals (AMV) and posterior medio-ventrals (PMV) of Arctolepida, all reduced to the same area. A, "Phlydaenaspis" heintzi, after Gross, 1933b (X 0.8); B, P. acadica, after Heintz, 1933 (X 0.6); C, Bryant- olepis sp., AMV from Bryant, 1934, PMV from PF 1551 (X 2.2); D, Anarthraspis sp., AMV from Bryant, 1932, PMV from PF 258 (X 0.6). s. AMV, s. AV, s. AVL, s. IL, s. PMV, s. PVL, external overlap areas for anterior medio-ventral, antero-ventral, anterior ventro-lateral, intero-lateral, posterior medio-ventral, and posterior ventro-lateral. may have fused so as to have a common center of radiation at the midline anteriorly (fig. 112, I). If this is so, the median plate just behind may be the posterior medio-ventral as Heintz identified it, or the anterior medio-ventral which has been crowded posteriorly. All of these points favor the view that the antero-ventrals were present primitively in the Arctolepida and that their loss in various ways is secondary in the Phlyctaenaspinae, Williamsaspidae, and Brachythoraci. The anterior medio-ve^itral and posterior medio-ventral plates (fig. 113, AMV, PMV) lie in the midline largely between the anterior ventro- laterals; the posterior medio-ventral is also bounded posteriorly by the posterior ventro-laterals, while the anterior medio-ventral is bounded anteriorly by the intero-laterals and in the Actinolepinae by the antero-ventrals. The anterior and posterior medio-ventrals are without known exception overlapped externally by all of these paired 534 FIELDIANA: GEOLOGY, VOLUME 11 plates. There appears to be less consistency about the overlap be- tween the two median ventral plates. In Coccosteus, Dinichthys, Pholidosteus, and probably in other Brachythoraci the anterior medio- ventral externally overlaps the posterior medio-ventral. The same appears to be the case in Anarthraspis (fig. 113, D) and in Bryanto- lepis (fig. 113, C). However, in Phlyctaenaspis acadica (fig. 113, B), "P" heintzi (fig. 113, A), and Prosphymaspis the reverse is true, and the posterior medio-ventral overlaps the anterior medio-ventral. In Phlyctaenaspis the two anterior sutural areas of the anterior medio-ventral are overlapped by the intero-lateral. Two similar over- lap areas occur in Bryantolepis and Anarthraspis, but here they are probably overlain by the antero-ventrals ; if this is so, the anterior medio-ventral would not meet the intero-laterals except perhaps in the midline. In Aethaspis (fig. 99) the anterior medio-ventral is not known, but its approximate shape is indicated by the overlap areas and margins of the anterior ventro-lateral, antero- ventral, and intero-lateral in PF 940. The anterior medio-ventral must have had a quite distinctive form with two pairs of lateral points bounding the antero-ventral overlaps, and with sharply distinct intero-lateral over- laps anteriorly. The posterior ventro-lateral plates (fig. 114, PVL) are interesting for the manner in which they overlap each other near the midline. Usually the left posterior ventro-lateral externally overlaps the right, but in Williainsaspis (White, 1952, p. 257) the reverse is true, and in Prosphymaspis (fig. 114, G) the right plate overlaps the left anteriorly, though posteriorly the normal overlap occurs. Otherwise the posterior ventro-laterals are overlapped by the anterior ventro-laterals and overlap the posterior medio-ventral and posterior laterals. The upward bend of the posterior ventro-lateral at the ventro- lateral edge serves to separate a lateral and a ventral face. These plates are often somewhat flattened in preservation and are so drawn in figure 114. In Brachythoraci, where the lateral trunk shield is reduced posteriorly, the lateral face of the posterior ventro-lateral is small or absent, but in the Arctolepida it is usually strongly developed ; it is particularly long in several Phlyctaenaspinae {Arctolepis, Phlyct- aenaspis, Prosphymaspis, Huginaspis) and in Williamsaspis, while in the Actinolepinae it is of moderate length in some {Kujdanowiaspis, Aethaspis, and Bryantolepis) and quite short in a few (Actinolepis and Anarthraspis). The short lateral face is believed to be specialized within the Arctolepida. The ventral face of the posterior ventro-lateral is short and broad in Kujdanowiaspis, Arctaspis, Aethaspis, Prosphymaspis, Phlyctaen- Fig. 114. Posterior ventro-lateral plates of Arctolepida, flattened and all reduced to the same area. A, Phlyclaenaspis acadica, modified from Heintz, 1933 (X 0.6); B, Anarthraspis sp., right PVL, from Bryant, 1934 (X 0.7); C, Aeihaspis major, left PVL, restored from PF 913 (X 0.4); D, Bryantolepis sp., restored from PF 166, 1546, 1547 (X 1.4); E, Aeihaspis utahensis, right PVL, restored from PF 322 (X 0.8); F, Kujdanowiaspis sp., based on casts of Podolian specimens, PF 1177, 1179 (X 1.6); G, Prosphymaspis subtilis, from Gross, 1937 (X 1.6); H, Actinolepis tuberculata, left PVL, after Gross, 1940 (X 4.2). s. AVL, s. PVL, external overlap areas for anterior ventro-lateral and oppo- site posterior ventro-lateral; s. PL, s. PMV, s. PVL, extent of internal overlap areas for posterior lateral, posterior medio-ventral, and opposite posterior ventro- lateral. 535 536 FIELDIANA: GEOLOGY, VOLUME 11 aspis acadica, and some Heterogaspis; it is long and narrow in Anarthraspis, Actinolepis, " Phlyctaenaspis" heintzi, P. sherwoodi, and Williamsaspis, while in Ardolepis, Huginaspis, and Bryantolepis it is intermediate in proportions. The Brachythoraci typically have long, slender posterior ventro-laterals, indicating a slender trunk region. The rather short, broad trunk may be primitive in Euarthrodira. The two posterior ventro-laterals usually form a bluntly rounded posterior end to the trunk shield; however, in Ardolepis, Prosphymaspis, and especially in Huginaspis, the posterior ventro-laterals are incut pos- teriorly to form a concave posterior end of the shield. Pectoral Fins A difference of opinion regarding the origin and evolution of their pectoral fins has led to widely divergent views on the phylogeny and classification of the placoderms. Heintz (1932a, 1938) believed that the Arctolepida were primitive and possessed a simple pectoral fin consisting of an immovable skin fold from the spinal to the posterior corners of the anterior lateral and anterior ventro-lateral. From them he derived the Brachythoraci with a reduced spinal and enlarged or open pectoral fenestra, suggesting the presence of a large pectoral fin. Gregory and Raven (1941) agreed with Heintz that the long-spined Arctolepida were primitive, as did Westoll (1945), Romer (1946), and Gross (1954), although Westoll thought that a fin membrane was absent in early Arctolepida. On the other hand, Stensio (1944) has argued that a long-based pectoral fin such as is found among the Brachythoraci is primitive in arthrodires. He interpreted the steno- basal fin, whose presence he demonstrated in Arctolepida, as the result of a reduction of a long-based fin. The endoskeleton lying between the anterior lateral and anterior vento-lateral represents, according to Stensio, the anterior part of the originally eurybasal fin, while the spinal is believed to represent the skeleton of the anterior part of the fin. White (1952) has accepted this theory of Stensio. The evidence of the nature of the pectoral fin in arthrodires is of three types: (1) The radials of the pectoral fin skeleton are rarely preserved. They were first noted by Heintz (1932a, p. 198, fig. 86) in an undeter- mined arthrodire which displayed four radials lying parallel to each other between the anterior ventro-lateral and anterior lateral. Later (1938, p. 21, fig. 5) he found in Coccosteus decipiens six or seven radials arranged parallel to each other along the posterior border of the anterior lateral. Gross (1938, p. 199, fig. 5, A) described a specimen of DENISON: EARLY DEVONIAN FISHES 537 the nearly related Rachiosteus which shows five pectoral fin radials, probably somewhat displaced yet still adjoining the anterior lateral. The specimen of Coccosteus indicates the presence of a pectoral fin with a moderately broad, unconstricted base, which must have attached at a pectoral fenestra between the anterior lateral and anterior ventro- lateral. The pectoral fin was probably similar in the other two speci- mens. Fin radials were presumably cartilaginous in other arthrodires and thus not preserved. They are known, however, in Gemundina, which had extremely eurybasal, ray-like pectoral fins, and in Pseudo- petalichthys, in which these fins were rather short-based. (2) The endoskeletal shoulder girdle bearing an articular crest for the pectoral fin has been described by Stensio (1944) in a few arthrodires. In Enseosteus (op. cit., fig. 14) both the shoulder girdle and the articular crest are very long, demonstrating the presence of a eurybasal pectoral fin. In Palaeacanthaspis (op. cit., fig. 9) and in Kujdanoiviaspis (op. cit., fig. 16) the articular area for the pectoral fin was short and the fin must have been stenobasal. White (1952, figs. 16-19) has described the endoskeletal shoulder girdle of Williamsaspis, which has an articular ridge of intermediate length for the pectoral fin. In Aethaspis an unornamented flange of the anterior ventro-lateral has been interpreted above (p. 490) as an ossification of the endo- skeletal attachment of the pectoral fin, and indicates that the latter was narrow-based. Westoll (1945, p. 384) believed that Arctolepis lacked a pectoral fenestra and fin, but both were present, as is proved by a section of a specimen in Naturhistoriska Riksmuseet in Stockholm (fig. 115, B) which shows the perichondrally ossified articulation of the pectoral fins; the latter must have been very narrow-based in this genus. (3) In the many arthrodires where neither the pectoral fins nor their endoskeletal girdles are known, evidence of the fin development may be obtained from the structure of the lateral wall of the trunk shield. In Kujdanoiviaspis (Stensio, 1944, fig. 17) the anterior lateral, anterior ventro-lateral and probably posterior lateral (see p. 526) bound a small pectoral fenestra through which the pectoral fin attached to its endoskeletal girdle ; the size of the fenestra necessarily indicates a narrow-based pectoral fin. In Phlyctaenaspis sherwoodi a reconstruc- tion (Denison, 1950, pi. 3) indicates that a similar pectoral fenestra was present here. The section of Arctolepis mentioned above shows that this genus also had a small pectoral fenestra. Presumably this fenestra was present in all Arctolepida, and there is evidence of its presence in Aethaspis, Anarthraspis, and Bryantolepis. It is well shown 538 FIELDIANA: GEOLOGY, VOLUME 11 Fig. 115. Arcfolepis sp. A, right half of trunk shield in dorsal view (X 3/2) showing position of section, x-y, and extent of pectoral fenestra, pf; B, section through lateral part of trunk shield (X 3), demonstrating the presence of a pectoral fenestra, pf; bones stippled, presumed cartilage cross-hatched; endoskeletal peri- chondral bone (end) black. ADL, anterior dorso-lateral; AL, anterior lateral; AVL, anterior ventro- lateral; IL, intero-lateral; MD, median dorsal; PDL, posterior dorso-lateral; PVL, posterior ventro-lateral; SP, spinal. in Williamsaspis (White, 1952, figs. 3, 4), where it is relatively larger than in other Arctolepida, its size indicating the longer base of the pectoral fin. Among the Brachythoraci, the situation must have been similar in the Coccosteidae, Pholidosteidae, and perhaps the Holonem- idae. The size of the pectoral fenestra is not known exactly in these forms, but it was certainly present and perhaps somewhat larger than in typical Arctolepida. In most Brachythoraci the pectoral fenestra is widely open behind, forming a pectoral incisure. This results from absence of contact between the ventral and lateral parts of the trunk DENISON: EARLY DEVONIAN FISHES 539 shield behind the anterior lateral and anterior ventro-lateral. Al- though the open pectoral incisure cannot be taken as proof of the presence of a long-based pectoral fin, such a fin could occur only in forms that have the lateral wall of the trunk shield incompletely covered by dermal bones. If we return to the theory of Stensio that the eurybasal fin is primitive in the Arthrodira, we may inquire with what sort of trunk shield it occurred originally. As far as I know, Stensio has not pub- lished his opinion on this matter. White (1952, p. 286) believed that the primitive form had not only a long-based pectoral fin, but a full body armor as in the Arctolepida. It is difficult to imagine how this could be. The primitive fin of Stensio's theory extended forward to the anterior end of the trunk shield and included the ancestral spinal at its anterior edge. A fin of this sort would completely separate the lateral and ventral trunk shields and leave them with only a carti- laginous endoskeletal connection. Such a shield would have been so inefficient either as a protective armor or as an external shoulder girdle that it is hardly to be considered as a possible ancestral condition. Another possibility, if we follow Stensio's theory, is that the long- based, primitive pectoral fin occurred in forms with the lateral wall of the trunk shield little developed. Such a condition is exemplified by many Brachythoraci, and it is necessary to consider whether these forms, though clearly specialized in other respects, could have had primitive trunk shields. One difficulty of this possibility has to do with the spinal. This plate is always very small or completely absent in those Brachythoraci that have the deep pectoral incisure. It is surely a vestigial plate rather than a rudiment that could give rise to the large and functional spinal of the Arctolepida, Coccosteidae, and Pholidosteidae. The spinal was certainly a part of the ancestral inheri- tance of arthrodires, since it is present in all orders, and possibly was present in the placoderm ancestor, if itshomologue is represented in the Antiarcha and Rhenanida. A primitive euarthrodiran should possess a well-developed spinal, which is not present in those Brachythoraci with eurybasal pectoral fins. In known euarthrodires the size of the spinal is inversely correlated with the size of the pectoral fin, so it seems improbable that there ever was a long-finned member of the group with a large spinal. The posterior dorso-lateral and posterior lateral may be very small in the Brachythoraci, and the anterior lateral may be very narrow and produced far under the head. These features appear to be the result of reduction and specialization and are difficult to picture as a primitive condition. Finally, the closure of 540 FIELDIANA: GEOLOGY, VOLUME 11 a pectoral fenestra in Arctolepida and Coccosteidae requires the junc- tion of the posterior lateral and posterior ventro-lateral. But how can one explain the presence of the posterior ventro-laterals in the long- finned Brachythoraci where they have no apparent function and are sometimes lost, unless it is as an inheritance from an ancestor with a fully developed trunk shield? It is also possible according to the Stensio theory that the ancestral arthrodire was unarmored and that the exoskeleton was indepen- dently acquired in the various orders, as well as in the suborders Arctolepida and Brachythoraci. This seems hardly worthy of serious consideration since the armor is universally present in placoderms and is so similar in general plan that it was undoubtedly derived by modi- fications of a common armored ancestor. The discussion above indicates some of the difficulties in Stensio's theory, and these are not overbalanced by evidence to support it. None of these problems arise if the Arctolepida are considered as primitive Euarthrodira, ancestral as a group to the Brachythoraci. All of the elements of the trunk shield of the Brachythoraci are present and well developed in the Arctolepida. A small pectoral fenes- tra is the primitive place of attachment of the narrow-based pectoral fin. As the fin lengthens, the pectoral fenestra enlarges and finally becomes open behind, a change accomplished by the posterior reduc- tion of the lateral trunk shield and the loss of its original posterior connection to the ventral shield. The opening of the pectoral fenestra allows the lengthening of the pectoral fin, and this change is accom- panied by a reduction and sometimes the loss of the spinal. A number of genera are known that are intermediate not only in their trunk shield and pectoral fins but also in cranial characters, and these are strong support for the derivation of the Brachythoraci from the Arctolepida. Thus Williamsaspis is an arctolepid with a reduced spinal and slightly elongate pectoral fin base. Coccosteus and Pholidosteus are Brachythoraci as regards their cranial structure yet they still retain the trunk shield of an arctolepid with a closed pectoral fenestra ; they approach the Brachythoraci, however, in the probable enlarge- ment of the fenestra and in the posterior reduction of the lateral trunk shield. Finally, the stratigraphic occurrence of arthrodires presents major difficulties to Stensio's theory. A small pectoral fenestra indicating a stenobasal fin, and a relatively large spinal are characteristic of Early Devonian Arctolepida. A narrow-based fin also occurs in the Early Devonian Acanthothoraci and in Pseudopetalichthys but is exceptional DENISON: EARLY DEVONIAN FISHES 541 in later Arthrodira, occurring in the Middle Devonian Actinolepis (where it may be somewhat enlarged), and in the Late Devonian Phlyctaenaspis shenvoodi. A somewhat longer fin base associated with a reduced spinal may occur in ''Coccosteus" angustus from the Early Devonian Hunsriickschiefer; it is typically developed in the Middle Devonian Williamsaspis, Coccosteidae, and probably in the Holonem- idae; it persists into the Late Devonian in the Coccosteidae, Holonem- idae, and Pholidosteus. The Brachythoraci with an open pectoral incisure and very small or no spinal are the characteristic forms of the Late Devonian, although a few of them appear in the Middle Devonian (Heterosteidae and Homosteidae). The only placoderm known to have a eurybasal fin in the Early Devonian is the rhenanid, Gemun- dina, which is a ray-like form of undoubtedly extreme specialization. The paleontological evidence thus strongly supports the primitive nature of the stenobasal pectoral fin in arthrodires and indicates that the eurybasal fin was specialized in this group. Sensory Canal System The lateral line canals are developed very uniformly among the Euarthrodira. The deeper canals, which form marked grooves in the dermal bones, are rather consistently present, although they may be reduced in a few of the Wildungen Brachythoraci. The pit lines may form shallow grooves on the dermal bones or they may be entirely superficial, in which case there is no record of their presence in fossils. The supraorbital canal (fig. 105, soc) in Arctolepida almost always begins at the center of ossification of the preorbital and extends anteriorly onto the postnasal. In the Brachythoraci (as well as the Petalichthyida and Phyllolepida) it also extends posteriorly toward the center of ossification of the central. In one specimen of Ar dole pis (fig. 105, I, soc) from Wijde Bay, Spitsbergen, the supraorbital canal extends onto the central, although this is not usually the case in this genus. In the coccosteid, Rachiosteus (Gross, 1938, p. 186), the supra- orbital canal extends forward from the center of the preorbital as a wide canal, while its posterior part, which extends onto the central, is narrow and probably shallow. IVIany Arctolepida may have pos- sessed the posterior part of this canal as a superficial pit line, but whether this condition was primitive or a secondary reduction is not clear. In either case the Wijde Bay Arctolepis and Rachiosteiis repre- sent intermediate conditions. The canal called premarginal by Bryant (1934, p. 137) was considered by Stensio (1945, p. 50) to have been supplied by the 542 FIELDIANA: GEOLOGY, VOLUME 11 profundus nerve and was named by him the profundus canal. It is well developed only in Bryantolepis (fig. 105, C, pfc) although traces of it are seen as a pit line on the preorbital and postorbital of Kujdanowiaspis and Simblaspis (fig. 105, D, pfc). This canal may have been present as a superficial pit line in other Arctolepida. It has not been detected in Brachythoraci. The central, infraorbital, and preopercular canals, as well as the cephalic division of the main lateral line canal, are with few excep- tions well developed in Euarthrodira, but the middle and posterior pit lines and the supratemporal cross commissure are less commonly indicated. Among the Arctolepida, Actinolepis shows a full development of the posterior pit line and supratemporal commissure (fig. 105, H, pp, stc) ; Simblaspis (fig. 105, D) sometimes shows the antero-medial portions of the middle and posterior pit lines and the postero-lateral portions of the posterior pit line and supratemporal commissure; Kujdanowiaspis (Stensio, 1945, pp. 50-51) may display the antero- medial part of the middle pit line and both ends of the posterior pit line. In the Brachythoraci the ends of the middle and posterior pit lines are not uncommonly indicated, as in Coccosteidae, Rhinosteus, Brachyosteus, Leiosteus, and Pholidosteus; the supratemporal cross commissure is known only in Coccosteus minor, Brachyosteus, and Rhinosteus, where medially it passes onto the gap between the nuchal and median dorsal. The groove or grooves on the lateral part of the paranuchal of Aethaspis (fig. 105, B, ope) probably indicate a pit line and may, as suggested above (p. 478), be homologous to the canal that crosses the posterior part of the gill cover in Acanthodes. It is not known in any other arthrodire. In Acanthodii, Watson (1937, p. 121) identified it as the opercular canal, while Stensio (1947, pp. 40-41) compared it to the scapular canal of Batoidea. Little is known of the sensory canals of the cheek in Arctolepida. The infraorbital canal (fig. 106) passes behind and under the eye on the suborbital and is usually well developed, though on the dorsal part of the suborbital it may be only a shallow pit line. The supra- maxillary (or jugal) line may be a deep canal or may be a shallow pit line that leaves no mark on the bone, as in Anarthraspis and Bryantolepis (fig. 106, E, C). On the trunk shield the main lateral line canal passes posteriorly across the anterior and posterior dorso-laterals (figs. 108, 109), al- though commonly the posterior part is superficial and leaves no mark on the bones. Usually this is the only post-cranial canal to be seen. DENISGN: EARLY DEVONIAN FISHES 543 but in a few cases others are indicated. On the lateral trunk shield a canal or groove may branch dorsally from the main lateral line canal on the anterior dorso-lateral, cross the antero-dorsal corner of the posterior dorso-lateral, and pass onto the median dorsal probably nearly to its center of ossification. This canal is not known in Arc- tolepida but is seen in some Coccosteidae where it may represent a vestige of the dorsal lateral line. A probable homologue occurs in some Antiarcha in the anterior oblique abdominal pit line (Stensio, 1948, pp. 185-186). On the ventral shield a sensory line may originate on the anterior medio-ventral, pass laterally onto the anterior ventro- lateral, and curve anteriorly to cross the lateral part of the intero- lateral. A part of this canal is seen on the anterior ventro-lateral of a relative of Actinolepis from Spitsbergen (fig. 112, A, vp), and it is also known in some Coccosteidae and in Holonema. It is probably a vestige of a transverse ventral pit line and has a homologue among Antiarcha on the anterior ventro-lateral of Bothriolepis (Stensio, 1948, pp. 187-188). It is probable that the sensory canal system was more fully developed in the ancestral placoderms and that it has been reduced to some extent in known arthrodires. The reduction may have in- volved merely a transformation into superficial pit lines or in many cases it may have led to complete loss of a canal. The Arctolepida and Coccosteidae retain a number of canals otherwise not indicated or rare in placoderms. These include the profundus canal, the middle and posterior pit lines, the supratemporal cross commissure, the oper- cular canal, a part of the dorsal lateral line, and a ventral pit line. On the other hand, in certain of the Late Devonian Wildungen arthro- dires, especially Oxyosteus, Synauchenia, and Brachydirus, the sensory canals are greatly reduced. According to Geuenich (1939) these were free-swimming forms, living under favorable light conditions, and no longer had as great a need for the lateral line sensory system. EVOLUTION OF THE ARCTOLEPIDA From the preceding discussion of arctolepid morphology, the characters of a primitive member of the group may be summarized as follows: In the cranial roof, the nuchal was moderately long and narrow. The paranuchals were rather large and lacked differentiated glenoids for the anterior dorso-laterals. The centrals were six-sided. The pre- orbitals were wider than long and five-sided. The rostral, pineal, and 544 FIELDIANA: GEOLOGY, VOLUME 11 postnasals were rather small, possibly not fused to the cranial roof behind, and the pineal notched the preorbitals only slightly. The bones had for the most part nearly central ossification centers. The orbits were small, and the suborbitals had short suborbital processes and large blades. The sensory canal system was quite fully developed, though many canals may have been present only as pit lines. The Fig. 116. Cranial roof and cheek plates of hypothetical primitive arctolepid. CE, central; IN, internal; MG, marginal; NU, nuchal; PAN, paranuchal; PI, pineal; PMG, postmarginal; PN, postnasal; PRO, preorbital; PSO, post- suborbital; PTO, postorbital; RO, rostral; SO, suborbital; cc, central canal; ifc, infraorbital canal; Ic, main lateral line; mp, middle pit line; ope, "opercular" canal; pfc, profundus canal; poc, preopercular canal; pp, posterior pit line; smc, supramaxillary canal; soc, supraorbital canal; stc, supratemporal commissure. endocranium was low and broad, with large suborbital shelves and supraorbital processes and a short occipital region; probably it was perichondrally ossified, with the nasal capsules separately ossified. The possible arrangement of the cranial plates of a primitive euar- throdire is shown in figure 116. The trunk shield was low and broad. The median dorsal was short and broad with a central growth center, and the posterior dorsals were large. The anterior dorso-laterals lacked trochleae for the cranial roof, and both the anterior and posterior dorso-laterals were probably relatively long and low. The anterior laterals had nearly central ossi- fication centers, their anterior triangles were not modified into an inner wing, and their ventral triangles were large. The posterior laterals were well developed and probably rather long and low. The spinals were of moderate length. The anterior ventro-laterals had long spinal edges and bounded deep pectoral sinuses. The dorsal laminae of the intero-laterals were small and formed little or no inner wings. Paired antero-ventrals were present, and the posterior ventro-laterals DENISON: EARLY DEVONIAN FISHES 545 had well-developed lateral faces and rather short ventral faces. The pectoral fenestrae were small and the pectoral fins were narrow-based. Kujdanoiviaspis approaches most closely to this ideally primitive condition and both in its early appearance and in its structure can be considered to be near the ancestry of other Arctolepida. This suborder is not well enough known at present to permit any detailed determi- nation of its phyletic development. However, certain general evolu- tionary trends are clear, and some specific phyletic lines may be distinguished. The Arctolepinae as a group retain a number of primi- tive features, including the short, broad median dorsal, the paired antero-ventrals, and the lack of any differentiated articulations be- tween the anterior dorso-laterals and paranuchals. Kujdanowiaspis itself is considered to be the most primitive, while " Svalhardaspis" angulatus and some undescribed forms from the lower part of the Wood Bay Series of Spitsbergen are very similar in known cranial characters. Simblaspis and Bryantolepis are possibly related primitive forms that show only a few distinctive specializations. In both, the nuchal and paranuchals are shortened posteriorly, the preorbitals are relatively large, and the anterior laterals have well-developed inner wings; Bryantolepis is also specialized in its large postnasals and pineal and very narrow pectoral sinuses. Anarthraspis, while it retains a rather primitive cranial roof, is advanced in having a high trunk shield with well-developed inner wings on the anterior laterals, shortened spinals and pectoral sinuses, and probably a long ventral shield. Aethaspis is specialized both in its cranial roof and its trunk shield. The former has a greatly elongate nuchal completely sep- arating the centrals, preorbitals much elongated posteriorly, and reduced postmarginals, rostral, pineal and postnasals. Its trunk shield is high, the anterior laterals have inner wings, the spinals are reduced, and the pectoral fenestrae slightly enlarged. One dis- tinct phyletic line of Actinolepinae is represented by the Middle Devonian Actinolepis and related forms from the Lykta and Stj0rdalen Divisions of the Wood Bay Series of Spitsbergen, including "Sval- bardaspis" rotundus and "