EARLY DEVONIAN FISHES FROM UTAH PART II. HETEROSTRACI ROBERT H. DENISON FIELDIANA: GEOLOGY VOLUME 11, NUMBER 7 Published by CHICAGO NATURAL HISTORY MUSEUM OCTOBER 16, 1953 EARLY DEVONIAN FISHES FROM UTAH PART II. HETEROSTRACI ROBERT H. DENISON Curator of Fossil Fishes FIELDIANA: GEOLOGY VOLUME 11, NUMBER 7 Published by CHICAGO NATURAL HISTORY MUSEUM OCTOBER 16, 1953 PRINTED IN THE UNITED STATES OF AMERICA BV CHICAGO NATURAL HISTORY MUSEUM PRESS INTRODUCTION In the first of this series of papers, which described the Osteo- straci from the Early Devonian of Utah (Denison, 1952), the geo- logical occurrence and the localities were discussed in detail. For convenience a list of the localities is repeated below, but details must be obtained from the paper referred to. All of the specimens come from the lower part of the Water Canyon Formation in Cache County, Utah. 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. The drawings in this paper, with the exception of figures 71 and 74, B, have been made, mostly from my sketches, by Miss Maidi Wiebe, staff artist in Chicago Natural History Museum. Family CYATHASPIDAE Kiaer 1932 Subfamily PORASPINAE, new rank (=PALAEASPINAE Zych 1931) In this paper, the subfamily Poraspinae is used as an approximate equivalent of the "tribe" Poraspidei of Kiaer (1932). The determi- nation of the systematic rank of this group is necessarily a matter of personal opinion, particularly since there is little information regarding its phylogenetic history. It may well be argued that Kiaer, with his extensive knowledge of these primitive Heterostraci, 291 292 FIELDIANA: GEOLOGY, VOLUME 11 was best qualified to devise a sound classification. Yet it is often true that great familiarity with a group may emphasize small dif- ferences and result in a tendency to exaggerate rank without need. This has clearly happened in the case of Kiaer's classification of 1932. Thus, within his "tribe" Poraspidei, he included five families, three of which are represented by only a single genus each, and the other two by only two genera each. The genera are with few ex- ceptions well characterized, but this is not the case with his families. The characters used to distinguish the latter are either those of the genus or genera included, are shared by other families, or are of doubtful familial importance; for example, one of the most charac- teristic features of the Dinaspidae (Kiaer, 1932, p. 17) is given as the peculiar development of the lateral line system, yet it appears that this feature is very different in the two genera; in Dinaspidella it resembles the lateral line system of Poraspis, while in Irregulare- aspis it forms a complicated network. If these two genera require grouping together to distinguish them from other genera, there are available in common usage other ranks, such as the subfamily and tribe (not as used by Kiaer). Kiaer grouped his families into two "tribes," where common usage would have employed the rank of superfamily. These super- familial groups, the Poraspidei and the Cyathaspidei, were distin- guished primarily by the degree of subdivision of the dorsal and ventral shields. One fact that has not been universall}^ understood is that the subdivision in the "Cyathaspidei" is limited to the super- ficial layer, and that the dorsal and ventral shields are undivided in both the "Cyathaspidei" and "Poraspidei." The superficial sub- division is sharpest in Cyathaspis and Archegonaspis, but it is not entirely absent in all of the "Poraspidei." Thus, even Poraspis may show superficially in its dorsal shield a rostral area and paired lateral areas, distinguished by the pattern of the dentine ridges and bounded by the pineal and lateral dorsal sensory canals (Kiaer, 1932, pi. 2). In Anglaspis insignis (op. cit., pi. 7, fig. 2), the sub- division is even more marked. Tolypelepis undulata (op. cit., pi. 10) shows little more superficial subdivision and may be regarded as intermediate in this respect, even though its rostral pattern is es- sentially transverse. Kiaer's two superfamilies were also distin- guished by the development of the lateral line system, which is described as being always strong in the "Poraspidei" and mostly im- perfectly developed in the "Cyathaspidei." It is doubtful whether much weight can be given to this distinction, for a completely developed lateral line system does occur on the dorsal shield of the DENISON: EARLY DEVONIAN FISHES 293 "Cyathaspidei," as is clearly shown in Kiaer's photograph of Toly- pelepis (op. cit., pi. 10). The lateral line system is also present in Cyathaspis, although, as Kiaer says (op. cit., p. 21), the pores are difficult to see. Probably it was also developed in Archegonaspis. In a recent paper. Flower and Wayland-Smith (1952) concluded that Kiaer's two tribes were better used as families, the Poraspidae and Cyathaspidae, and that the boundary between them is "too tenuous to be recognized." In my opinion, there is between the two groups no known morphological distinction of sufficient importance to warrant familial distinction. There is apparently a difference in their geological occurrence, and mainly for this reason they will be retained provisionally as separate subfamilies and classified as below. Family Cyathaspidae Kiaer 1932 Subfamily Poraspinae (new rank) Americaspis White and Moy-Thomas 1941 {=Palaeaspis Claypole) Anglaspis Jaekel 1927 Allocnjptaspis Whitley 1940 (=Cryptaspis Bryant) Dinaspidella Strand 1934 {=Dinaspis Kiaer) Homalaspidella Strand 1934 {=Homalaspis Kiaer) Irregulareaspis Zych 1931 (=Dic(yaspis Kiaer) Poraspis Kiaer 1932 Vernonaspis Flower and Wayland-Smith 1952 ?Corvaspis Woodward 1934 Subfamily Ctenaspinae (new rank) Ctenaspis Kiaer 1930 Subfamily Cyathaspinae (new rank) Cyathaspis Lankester 1865 Archegonaspis Jaekel 1927 Tolypelepis Pander 1856 ( = Tolypaspis Schmidt) In the above classification, it should be noted that Ctenaspis has been separated from the Poraspinae because of its specialized lateral line system and its development of dentine tubercles rather than ridges. The confused status of Eoarchegonaspis Kiaer 1932 has been discussed by Flower and Wayland-Smith (1952, pp. 366- 370) ; to eliminate the insoluble nomenclatorial problems involved, they have designated as lectoholotype of the genotype, Anatifopsis wardelli, an unrecognizable fragment, making further use of Eoarche- gonaspis impossible. Traquairaspis has been shown by White (1946, p. 236; 1950, p. 55) to belong to a family of its own, more closely related to the Pteraspidae. The genus Diplaspis, which Kiaer placed in the "Cyathaspidei," must still be regarded as of uncertain affinities. 294 FIELDIANA: GEOLOGY, VOLUME 11 Allocryptaspis Whitley, May, 1940, p. 243. Cryptaspis Bryant 1934, p. 154, non Pascoe 1872. Bnjantaspis White and Moy-Thomas, June, 1940, p. 507. Cryptaspidisca Strand 1942, p. 384. Amended diagnosis. — Large Poraspinae with the dorsal shield relatively long and narrow, rather flat anteriorly, becoming wider and more arched posteriorly; posterior termination a blunt point; lateral edges turned under anteriorly, forming lateral laminae sepa- rated from the dorsal face by rounded to sharp lateral angles. Ven- tral shield narrow, flat anteriorly and moderately convex posteriorly, with a weak post-branchial deepening and a distinct postero-median lobe. Orbits very small, relatively far forward, and notching the ventro-medially directed laminae of the dorsal shield. Branchial openings posterior to branchial plates, and bounded by notches in the dorsal and ventral shields. Maxillar brim, pineal macula, im- pressions of inner organs, and lateral line system essentially as in Poraspis. Dentine ridges coarse, 3 to 5 per mm., minutely crimped, usually flat-topped, round-topped on median scales, crested near branchial openings. Pattern of dentine ridges basically simple and longitudinal, but with a slightly fan-like arrangement in the pre- pineal triangle of the dorsal shield and in the anterior triangle of the ventral shield, and with whorls modifying the longitudinal arrange- ment locally; no rim of transverse ridges at the rostrum. Scales of the tail as in Poraspis, except that the dorso-lateral scales have strongly curved dentine ridges, and more extended dorsal and ventral overlap areas. Minute structure of the exoskeleton agreeing as far as it is known with that of Poraspis. Allocryptaspis utahensis, sp. nov. Type.— C.N. U.M.FF 737, associated dorsal shield, ventral shield, and numerous scales (figs. 61, 62). Horizon. — Early Devonian, near base of Water Canyon For- mation. Locality. — Cottonwood Canyon (locality G), NW. 3<4> sec. 18, T. 13 N., R. 3 E., Cache County, Utah. Referred specimens. — A nearly complete ventral shield, PF 738 (fig. 63); and an incomplete dorsal shield, PF 739, both from lo- cality G (Cottonwood Canyon). In addition, there are approxi- mately 25 fragmentary specimens, including parts of shields, scales, and two branchial plates; these were collected at localities A (Black- DENISON: EARLY DEVONIAN FISHES 295 Fig. 61. Allocryptaspis lUahensis, sp. nov., type; PF 737 (X 1). shield on left, ventral shield on right, numerous scales in center. Dorsal smith Fork), E (Green Canyon), G (Cottonwood Canyon), and I (Dry Lake). Description. — Allocryptaspis includes the largest known members of the Cyathaspidae. In the type of A. utahensis (PF 737), the total length of the dorsal shield is estimated to be 110 mm., while in A. ellipticus from Beartooth Butte, Wyoming, the corresponding measurement is 78-91 mm. The total length of the ventral shield in A. utahensis (PF 738) is 91 mm., in A. ellipticus it is 79-83 mm., I 296 FIELDIANA: GEOLOGY, VOLUME 11 and in A. flabelliformis, the second species from Beartooth Butte, it is 100 mm. Kiaer and Heintz (1935, pp. 125-126) have shown that there is an evolutionary trend towards increase in size in the species of Poraspis. Such a trend is not too obvious when all the Poras- pinae are considered. Thus, the smallest known Poraspinae, be- longing to Anglaspis, Dinaspidella and Homalaspidella, occur in the Dittonian and its equivalents. The North American genus, Ameri- caspis, may have dorsal shields 55 or more mm. in length; if the beds in which it occurs are correctly correlated with the Upper Ludlow, a moderately large size was attained quite early in the group. Considering all the described Poraspinae, the maximum length of the dorsal shield was about 55 mm. in the Upper Ludlow, 60 mm. in the Downtonian, and 80 mm. in the Dittonian. The still larger size of the species of Allocryptaspis suggests that they were post-Dittonian in age. The dorsal shield of A. utahensis (figs. 61, 64, A, C) is relatively narrow and flat in front, and increases gradually in width posteriorly. The rapid increase in width in the type (PF 737) is due in part to crushing; in life the posterior part of the shield must have been more convex, and this has been accounted for to some extent in the res- toration (fig. 64, A, C). An estimate of the ratio of maximum width to total length is 0.40, indicating that this was one of the most slender-bodied of all Poraspinae. A single species of Poraspis, the Podolian P. sturi, has similar proportions, its width/length ratio being 0.39; other species of this genus range from quite narrow (W/L =0.43) to extremely broad (W/L =0.77). Allocryptaspis ellipticus has a broader shield than A. utahensis (W/L =0.48). In none of the specimens of A. utahensis can a maxillar brim be seen at the anterior end of the dorsal shield; however, it may have been present, since apparently it was well developed in A. ellipticus (Bryant, 1935, pi. 3). Immediately behind the rostrum, the lateral edges of the dorsal shield are folded sharply downward and inward to form distinct lateral laminae that continue posteriorly almost to the branchial notches (figs. 62, 64, C, II). Between the dorsal surface and these lateral laminae, the margins are sharp anteriorly but they become more and more rounded posteriorly. The orbits are indi- cated by very small, semicircular notches, placed entirely in the ventro-lateral laminae of the dorsal shield, so that the eyes must have been directed laterally and downward (fig. 64, C, or). The diameter of the notches is 1.8 mm. in PF 737, less than 2 per cent of the total length of the shield, thus considerably smaller relatively DENISON: EARLY DEVONIAN FISHES 297 than in other Poraspinae, excepting A. ellipticus. Below the orbital notch, the ventral lamina is cut by two flaring edges, presumably for the suborbital plate (fig. 64, C, so) that is known to bound the orbit below in Anglaspis and Irregulareas'pis. The orbits are rela- tively farther forward in AllocrTjptaspis than in other Poraspinae. Fig. 62. Allocryptaspis utahensis, sp. nov., type, PF 737 (XI); ventral view of anterior part of dorsal shield, hrn, branchial notch; II, lateral lamina; or, orbital notch. Fig. 63. Allocryptaspis utahen- sis, sp. nov., PF 738; ventral shield (XI). Their position has been measured by Kiaer and Heintz (1935, pp. 45-46) by the so-called rostral index (the ratio of the "rostral length" to the total length). Since their "rostral length" was measured from the anterior end of the shield posterior to a line connecting the orbits, this ratio is better called the orbital index. In A. utah- ensis, it is 0.11; in A. ellipticus, it is 0.12. Other Poraspinae show a corresponding ratio of 0.14 to 0.23. The position of the branchial opening in Allocryptaspis is shown by a rounded notch at the edge of the dorsal shield, posterior to the lateral lamina (fig. 64, A, brn). Behind it there is no marked descending post-branchial lobe, but the margin of the shield extends nearly straight to the rounded 298 FIELDIANA: GEOLOGY, VOLUME 11 postero-lateral corner. In the median line, the dorsal shield termi- nates posteriorly in a bluntly rounded point; in none of the speci- mens, either from Utah or Wyoming, is there any suggestion of as large a postero-median projecting lobe as is indicated by Bryant in his restoration (1935, fig. 1, A). The pineal macula in A. utahensis (PF 737) is an elongate-oval elevated area (fig. 64, A, pi) similar to that of A. ellipticus (PF 439). Internal impressions of the gills, semicircular canals, and cranial cavity are not preserved in any of the material from the Water Canyon Formation, but they are dis- tinct in A. ellipticus, where they have been described by Bryant (1935, pp. 115-116). The collection from the Water Canyon Formation includes one nearly complete ventral shield (PF 738; fig. 63) and three partial ventral shields (PF 737, PF 741, PF 342), as well as some more fragmentary remains. These specimens indicate that this shield is narrow and only slightly vaulted anteriorly, and slightly broader and more deeply vaulted posteriorly (fig. 64, C). Because of crush- ing, the proportion of width to length is not exactly determinable by the usual method, employing a maximum width. The width has been measured anteriorly just behind the antero-lateral corners, where the arching is slight and the possibility of distortion due to crushing correspondingly small. The ratio of this anterior width to the total length of the ventral shield is 0.27 in PF 738, compared to a corresponding ratio of 0.33 in A. ellipticus (PF 145), and 0.38 in A. flabelliformis (Bryant, 1935, pi. 7). The anterior edge is narrow, straight or slightly concave. From the antero-lateral corner, the edge for the branchial plate extends nearlv straight posteriorly and slightly laterally until mid-length, where it curves up to the post- branchial lobe. The latter is relatively weakly developed, and has at its anterior end a shallow rounded notch that must represent the ventral boundary of the branchial opening (fig. 64, C, br). The postero-lateral corner of the post-branchial lobe is rather sharply rounded, and from there the shield tapers rapidly backwards to a projecting, rounded postero-median lobe (fig. 64, B). This agrees in all essential respects with the ventral shield of A. ellipticus. In the early descriptions of this genus under the name of Crypt- aspis, Bryant (1934, p. 156; 1935, pp. 116, 119) described as branchial plates some elongate rectangular plates that he found in association. Kiaer and Heintz (1935, p. 133) were of the opinion that these were dorso-lateral scales of the tail, and they were correct, as is shown by a number of such scales, both from Utah and Wyoming, that exhibit the characteristic arrangement of the superficial ridges. The DENISON: EARLY DEVONIAN FISHES 299 branchial plate is as yet undiscovered or undescribed in A. ellipticus and A. fiabelliformis. In the collections from Utah, there are two specimens (PF 743 and PF 744) that are believed to be incomplete branchial plates of A. utahensis for the following reasons: In both the dorsal and ventral shields, the dentine ridges are flat-topped except around the branchial notches, where they assume a very characteristic crested form; PF 743 and PF 744 have exactly this type of ridges. Neither of these plates possesses any entire margins, but apparently they were long and narrow in shape and rounded in cross section. This rounding is thought to represent the angulation between the ventral shield and the lateral lamina of the dorsal shield. There is no indication of any branchial notch, and such may not have been present on the branchial plate in this genus. The pattern of the dentine ridges is essentially longitudinal in what is assumed to be the dorsal lamina, and diagonal in the ventral lamina, about as in the branchial plate of Poraspis (Kiaer and Heintz, 1935, fig. 51). There is little doubt that Bryant misinterpreted the specimen on which he based his description of the branchial openings in A. ellipticus (1935, p. 116, pis. 1, 4). This specimen is a natural cast of the inner side of the dorsal shield, and the two slits in the cast that he interpreted as double branchial openings must in reality have been occupied by bone forming a partial inner rim around a single opening. It is certain that there was but a single pair of branchial openings in A. utahensis. They are indicated by nearly semicircular notches in the dorsal shield in PF 737 and PF 739, and by shallower notches in the most anterior part of the post-branchial lobe of the ventral shield in PF 342 and PF 738. It would appear, then, that the branchial openings lay between the dorsal and ven- tral shields (fig. 64, C, br), rather than between the dorsal shield and the branchial plates, as in other Cyathaspidae. The branchial plate must have been shortened (since the relative position of the branchial opening is about as in other members of the family); possibly it still bounded the branchial opening anteriorly, although this is not now demonstrable. The dentine ridges are relatively coarse in Allocryptaspis. In A. utahensis there are 3.2 to 3.6 per mm., in A. flabelliformis about 4 per mm., and in A. ellipticus 4 to 5 per mm. Thus they are much coarser than in Poraspis, Dinaspidella, and Irregulareaspis, but somewhat finer than in Anglaspis and Coruaspis. In general, the ridges are flat- topped with narrow grooves between. At first glance. 300 FIELDIANA: GEOLOGY, VOLUME 11 they appear to be smooth-sided, as is usually the case in this family, but under magnification, well-preserved specimens show a very fine lateral crimping. This crimping is hard to observe in the specimens from Beartooth Butte, but is suggested in an impression of A. ellipticus (PF 439). Where the ridges curve around the branchial notches, as well as on the fragments interpreted as branchial plates, the ridges are not flat, but come to V-shaped, slightly denticulate crests. On the median body scales, they are crested or round- topped. The pattern of the dentine ridges in Allocryptaspis is relatively simple, with, of course, none of the subdivision into areas that is well developed in the Cyathaspinae. The rostral region shows a slightly fan-like arrangement, comparable to that of Poraspis elong- ata (Kiaer and Heintz, 1935, pi. 19). There is no anterior margin of irregular or transverse ridges, such as occurs in the younger spe- cies of Poraspis (op. cit., p. 126). Some of the ridges curve around the pineal macula, while others abut against it. The posterior part of the shield has an essentially longitudinal arrangement of the ridges; this is modified considerably by whorls in the anterior part and laterad to the lateral longitudinal sensory line. On the lateral laminae of the dorsal shield, the ridges are mainly longitudinal ex- cept where they curve up over the orbits and the branchial openings. The ventral shield has an essentially longitudinal ridge pattern modified by a slightly radiating pattern anteriorly, minor whorls medially and along the lateral edge, and curves below the branchial openings. The course of the canals of the lateral line system has been determined with difficulty in the Utah material (fig. 64, A), although it is marked by the same type of external pores that occur in Poraspis. It is much more obvious in occasional specimens of A. ellipticus (fig. 65, A). The pineaP branches do not appear to coalesce in the 1 Different interpretations of the homologies of the lateral lines of the Hetero- straci have led to confusion in terminology (Holmgren, 1942; Save-Soderbergh, 1941; Stensio, 1926). Since there is no ossified endoskeleton, it is not possible to demonstrate innervations, and the identifications that have been made have been based only on topographical relations of the lateral lines to various organs. For this reason, the more noncommittal names of Kiaer and Heintz (1935, fig. 3) will be used, with the following emendations: The term "pineal canal" is restricted to that part of the pineal canal of Kiaer and Heintz lying postero-mediad to the junction of the lateral dorsal line; the more anterior part is termed "the supra- orbital canal"; these two are poorly distinguished in the Cyathaspidae, but sharply so in the Pteraspidae. The lateral branch from the lateral dorsal line posterior to the orbit is here called "the infraorbital canal." The small anterior branch lying above the orbits and between the supraorbital and infraorbital canals (Kiaer and Heintz, 1935, p. 66, fig. 12, "x") is called "the profundus Hne," following Holmgren (1942, p. 12). shie,d;°B'te„t'J:fSrc1att7*vfe'7„?d'''''*r^'''A^'=- ™- « »■ A. dorsal chial opening; 6r», branchi'a „„fi 7 "'dorsal and y^^^^^f ,|^.^y • lateral p„„io'„ „f ;e„Sl;:L";eS'cl''„it';:''« 'iitill' ""f^' ^"'^'^ '"• «, lateral lamina of dorsal shipid- h,J lofl !' , ' ^*®^^^ ^^''^al sensory line- tion of ventral transver e commt ure , S '^^^^^^^ ^■"^' '«^' "^^dial por^ notch; pi, pineal macula; XZe^iZtZh^?/ Plate; /..-., dorsal sensory t'ra^s^vers:^;^^^?^; :::^^^^ -^-^ital 301 302 FIELDIANA: GEOLOGY, VOLUME 11 median line posterior to the pineal macula in A. utahensis, although this junction clearly occurs in A. ellipticus (fig. 65, A), as in the later representatives of Poraspis from Spitsbergen (Kiaer and Heintz, 1935, p. 126). From the pineal region, they extend antero-laterally and then curve towards the lateral border, at least in A. utahensis; the more anterior region is termed the supraorbital canal. The an- terior ends of the median dorsal lines curve medially to join the pineal branches, as in the later species of Poraspis from Spitsbergen. The lateral dorsal lines agree with those of Poraspis except that the anterior portions that curve inwards to join the pineal-supraorbital branches have not been determined. The pineal branch is counted as the first of the dorsal transverse commissures of the sensory line system. The second and third transverse commissures were shown as having anterior branches in A. ellipticus by Bryant (1934, pi. 26, fig. 1; 1935, fig. la); these "paired diagonal canals" are apparently cracks since they are not symmetrical on Bryant's specimen and do not appear on another well-preserved specimen of A. ellipticus (fig. 65, A) or in A. utahensis. The second commissural canals do not quite meet in the midline in these species and extend somewhat anteriorly of laterally nearly to the lateral dorsal lines. The third commissures meet in the midline and extend somewhat posteriorly of laterally, reaching the lateral dorsal lines in A. ellipticus. The last two commissures are shown correctly for A. ellipticus in Bryant's figures, and for A. utahensis in figure 64, A. The pattern of these canals is closely similar to that of Poraspis. On the ventral shield, the course of the lateral line canals has been incompletely determined in A. utahensis (PF 737, PF 738; fig. 64, B). The recognized fragments of the lateral ventral lines and of the ventral transverse commissures agree in pattern with the canals of Poraspis. The scales of the cyathaspid body and tail are best known in nearly complete articulated specimens of Anglaspis heintzi (Kiaer, 1932, fig. 11) and Irregular easpis hoeli (op. cit., pi. 5). As a general rule, only isolated scales have been discovered, as in Allocryptaspis ellipticus, where such scales from the dorso-lateral region were de- scribed as branchial plates by Bryant. A few isolated scales of A. utahensis have been discovered in the Water Canyon Formation, but the type (PF 737) preserves a large number of scales from vari- ous regions of this one individual, a few of them still partially ar- ticulated. The following description is based for the most part on this specimen (fig. 61). DENISON: EARLY DEVONIAN FISHES 303 The median scales (dorsal and ventral are not distinguishable) agree essentially with those of Poraspis as figured by Kiaer and Heintz (1935, figs. 41-44). Among them may be recognized more anterior scales with a relatively broad anterior edge, and more pos- terior scales with a narrow anterior edge. The dentine ridges of these scales are slightly rounded or convex; they are sometimes nearly parallel and sometimes radiate slightly posteriorly, as in Fig. 65. A, Allocnjpfaspis elUpficus, PF 439 (X IM); impression of anterior part of dorsal shield, showing sensory canals. B, A. utahensis, PF 749; right dorso- lateral scale ( X 2). Poraspis. At the anterior edge, there are two transverse ridges or sometimes only one. The anterior and antero-lateral overlap areas are relatively broad. The dorso-lateral scales (fig. 65, B) differ from those of Poraspis in certain details. There is probably a greater number of dentine ridges than in a comparable scale of Poraspis; this is no doubt correlated with the larger size of Allocryptaspis, since the ridges are coarser in the latter. The dentine ridges have a pronounced down- ward curvature towards the anterior edge, where they are bounded by two dorso-ventrally directed ridges, broken into short lengths. The anterior overlap area is relatively narrow, while the antero- dorsal and antero-ventral overlap areas are very long dorsally and ventrally, though narrowed antero-posteriorly. 304 FIELDIANA: GEOLOGY, VOLUME 11 A few ventro-lateral scales are present in PF 737 (fig. 61), recog- nizable by their asymmetrical form, flat-topped dentine ridges, and sensory canal pores. Unfortunately, none is complete, but they appear to differ from those of Poraspis in having wider overlap areas. PF 737 also preserves a number of scales from the caudal region. Some of these, presumably median scales, are slender, symmetrical, and fulcra-like, with as few as three dentine ridges. The preservation of the dermal skeleton is such that it is im- possible in a thin section to see the details that have helped to differentiate the various genera from Spitsbergen and England. The main features are typical of the Cyathaspidae, namely, the thin, laminated basal layer, the thick middle layer formed of rather regu- lar large cancellae, and the superficial layer of dentine ridges. How- ever, the sections do not show the development of the reticular layer or the size of the canals leading to the pulp cavity. Comparisons. — That Allocryptaspis is a member of the Poras- pinae is indicated by the lack of any superficial subdivision of the dorsal and ventral shields and possibly by the well-developed lateral line system. It retains many characters that may perhaps be primi- tive in the subfamily, such as the simple pattern of the dentine ridges and the general resemblance to Poraspis of the scales, lateral line system, structure of the dermal shield, etc. On the other hand, it shows some features that are surely specializations within the family. Among these, the most striking are the large size and slender proportions, the distinctly developed lateral laminae of the dorsal shield, and the position of the branchial openings between the dorsal and ventral shields, posterior to the branchial plate. Allocryptaspis utahensis may be distinguished from A. ellipticus by its larger size, coarser dentine ridges, and more slender propor- tions. A. flahelliformis is known from only a single ventral shield and is not well characterized. From this species, A. utahensis dif- fers possibly in being slightly smaller and in having somewhat coarser dentine ridges, and surely in its more slender proportions. Family CARDIPELTIDAE Bryant 1933 The peculiar heterostracian, Cardipeltis, characterized by its three-lobed dorsal shield, has been poorly understood in both its structure and relationships. In the original description, Branson and Mehl (1931, p. 522) believed that it was related to the Poras- pinae {"Palaeaspis"). Bryant at first (1932, p. 241) considered it DENISON: EARLY DEVONIAN FISHES 305 to be closer to the Pteraspidae, then (1933, p. 307) referred it to a family of its own, the Cardipeltidae, which in 1935 (p. 120) he con- sidered to belong to the "Poraspidei" (Poraspinae of this paper). Obrutchev (1941, p. 21) compared it to the Drepanaspidae (Psam- mosteidae), to which he believed it was related, while Brotzen (1936, p. 10, footnote 2) thought it was more probably an arthrodire. My own conclusions are that Cardipeltis is undoubtedly a heterostracian and probably represents a specialized branch derived from the Po- raspinae. A number of hitherto unknown details of its structure may be determined from the new collections from the Water Can- yon Formation, but a complete reconstruction of this form is not yet possible. Cardipeltis Branson and Mehl 1931 Amended diagnosis. — Large, rather flat-bodied Heterostraci. The dorsal shield probably undivided, having a rounded anterior lobe and paired postero-lateral lobes, separated by deep branchial notches laterally at mid-length and by a deep postero -median emargination. External branchial openings directed postero-dorsad. Dorsal sen- sory canal system similar to that of the Poraspinae and consisting of two pairs of longitudinal canals and probably five transverse com- missures. At least two pairs of plates form the lateral edges. Ven- tral shield not known, and possibly subdivided into small plates. Ornament of dorsal shield with a radiating pattern, and consisting of flat-topped, elongate ridges centrally and of roundish tubercles peripherally; that of lateral plates consisting of irregular, flat- topped tubercles on one lamina and of elongate, crimped, rounded or crested ridges on the other. Histology of the exoskeleton as in other Hetero- straci, except that the reticular and cancellar layers are not clearly differentiated and the basal laminated layer is relatively thick. Cardipeltis wallacii Branson and Mehl 1931 The cotypes (University of Missouri, 601 VP and 602 VP) are dorsal shields from Blacksmith Fork, Utah. Additional material from the Water Canyon Formation in Chicago Natural History Museum includes a nearly complete dorsal shield (PF 804; fig. 66) from Blacksmith Fork (locality D), the anterior half of a dorsal shield (PF 895; fig. 67) and the rim of a dorsal shield (PF 805) from Cottonwood Canyon (locality G), as well as eight smaller fragments of dorsal shields. Not certainly referable to this species but presumably belonging to Cardipeltis are twelve incomplete mar- 306 FIELDIANA: GEOLOGY, VOLUME 11 ginal elements (fig. 69), six scale-like plates possibly belonging to the ventral armor, and a single scale from the caudal region; these were found in Blacksmith Fork (localities A and D), in Water Can- yon (locality F), and in Cottonwood Canyon (locality G). Three species of Cardipeltis have been described, of which C. wallacii is the genotype. The known dorsal shields are 132 141 mm. in median length; the branchial notches are only moderately deep; the post-branchial lobes are not very prominent; the posterior edge is deeply emarginate with only a small postero-median process; the pre-branchial lateral edge is nearly straight; the anterior edge is emarginate; the ornamentation varies in different parts of the dor- sal shield, but near the middle consists of 8-10 ridges per cm., some as much as 20 mm. in length, although most of them are shorter. Comparisons. — C. sinclairi Bryant from Beartooth Butte, Wyo- ming, does not differ significantly in size, although larger individuals have been recorded. Bryant (1933, p. 310) believed that it differed from C. wallacii in general proportions, but such differences may be due to the flattening of the Beartooth Butte material. It does differ in having prominent, strongly convex post-branchial lobes, sharply marked off anteriorly by the deeper branchial notches and posteri- orly by the deep posterior emargination; the latter contains a promi- nent postero-median process. The lateral margins anterior to the branchial notches are more convex in C. sinclairi and pass into a convex anterior margin. The ornamentation, according to Bryant (loc. cit.), consists of ridges averaging about 1.0 mm. wide and 5-10 mm. long, but they do not appear to be well preserved on the type; in Chicago Natural History Museum specimens PF 151 and UC 2215, they differ from those of C wallacii in being broken up into much shorter lengths in the middle of the shield. C. ohlongus Bryant, the second species from Beartooth Butte, is inadequately known. It does not differ significantly in size from C. sinclairi. Bryant (loc. cit.) described it as being much nar- rower than the latter, but the type is perhaps less flattened than that of C. sinclairi and is probably either incompletely preserved or exposed on its margins, especially laterally. If this is so, the narrow proportions, shallow branchial notches, and shallow posterior emar- ginations cannot be relied on as valid characters. Bryant says the ornamentation is similar to that of C. sinclairi. At the present time, this species cannot be considered to be well established. Discussion. — To date the only remains of Cardipeltis that have been described are the large three-lobed shields that usually have DENISON: EARLY DEVONIAN FISHES 307 Mm Fig. 66. Cardipeltis wallacii, PF 804; dorsal shield (X H)- been considered to represent the dorsal shield of a heterostracian. However, since Brotzen (1936, p. 10, footnote 2) expressed doubt that this belonged to an ostracoderm and compared it with the median-dorsal plate of an arthrodire, it is necessary first to establish the class and order to which the shield belongs. This is done most conclusively by its histological structure, which is preserved moder- ately well in specimens from the Water Canyon Formation, but generally very poorly in those from Beartooth Butte. Histology: In section (fig. 71, A) the shield may be divided into three layers. Superficially, there is a layer formed by the ridges. 308 FIELDIANA: GEOLOGY, VOLUME 11 which are composed of an outer thin enamel-like tissue and an inner thicker dentine-like tissue; they agree closely in structure with the ridges of Poraspis and Pteraspis, but find no counterpart among the placoderms. According to Dr. Tor Orvig, who has seen the sections, the superficial layer is characteristic, being of much the same type as in Oniscolepis and also in specimens from Podolia possibly belong- ing to Corvaspis. The middle layer occupies about half of the total thickness and contains a large number of irregular cavities, smaller in the outer part and larger in the inner part; there is no sharp distinction between reticular and cancellar layers as in the Cyathas- pidae and Pteraspidae. The middle layer resembles that of the Drepanaspidae but is relatively much thinner. It may resemble Corvaspis in structure, although the microstructure in this genus is inadequately figured (Woodward, 1934, pi. 19, fig. 3). It is similar to the middle layer of the shield rim in Traquairaspis (Wills, 1935, pi. 6, fig. 3), although the median part of the shield is more typically pteraspid in the latter genus. The basal layer is laminated and rela- tively thicker than in the Cyathaspidae, Pteraspidae, and Drep- anaspidae; it appears to lack any cell spaces, and in this it resembles typical aspedin. The microstructure is thus clearly heterostracian in nature, and although it is somewhat modified from the typical condition, it surely cannot belong to any placoderm. While considering the histology of Cardipeltis, it should be men- tioned that one of the characters by which Bryant (1933, p. 308) distinguished the Cardipeltidae was the thinness of the exoskeleton. Actually it is 1.3 1.5 mm. thick in C wallacii — not much, relative to the size of the shield, yet considerably thicker than in most Cyathaspidae and Pteraspidae. A thicker exoskeleton among the Heterostraci is found in Corvaspis, where it is about 2 mm. (Wood- ward, 1934, p. 567), and in the Drepanaspidae, where the dorsal and ventral discs may be 6 mm. or more thick. The thinness of the exoskeleton of many of the Beartooth Butte fishes is apparently the result of post-depositional modifications. Branchial Notches: If one accepts the heterostracian nature of the large, notched, symmetrical shields of Cardipeltis, they could only represent the dorsal shield or a part of it — the dorsal disc. This is clearly demonstrated by the pattern of the lateral line canals, which will be described below, and by the presence of the deep, paired notches on each side. The latter were originally interpreted as orbital notches (Branson and Mehl, 1931, p. 520; Bryant, 1933, p. 308) but later recognized as branchial notches (Bryant, 1935, p. 120; Obrutchev, 1941, p. 21). Their position, structure, and rela- DENISON: EARLY DEVONIAN FISHES 309 tionship to lateral line canals all indicate that these are, in fact, the external branchial openings. Each is a deep, postero-medially di- rected notch, lying at about mid-length of the shield on its upper surface and continued internally by smooth, unornamented bone that forms a half tube directed antero-ventrad (fig. 68, hrn). Bryant (1933, pp. 309-310; 1935, p. 120) described these furrows as being directed obliquely backwards and inwards, but this is clearly not the case in the Chicago Natural History Museum specimens of C. wallacii (PF 804 and PF 805) and C. sinclairi (PF 150), nor does it appear to be so in the specimen of C. sinclairi figured by Bryant (1933, pi. 18, fig. 1). In these specimens (fig. 66), the posterior ends of the branchial tubes were directed postero-dorsad. Presuma- bly they were bounded laterally by branchial plates as in most other Heterostraci; a number of such plates from the Water Canyon For- mation will be described below. When the branchial notches of Cardipeltis are compared with those of other Heterostraci, the greatest similarity is seen to be with the Cyathaspidae. In the Pteraspidae, the branchial opening typi- cally lies between the cornual and branchial plates, and while the dorsal disc may be notched for the cornual plate usually it does not bound the branchial opening. In the Traquairaspidae, the branchial opening pierces the branchio-cornual plate, and the dorsal disc does not enter into it. In the Drepanaspidae the branchial opening is far removed from the dorsal disc, and lies between the branchial and cornual plates at the postero-lateral corner of the shield. In the Cyathaspidae, on the other hand, the branchial opening lies between the dorsal shield and the branchial plate; generally it notches the branchial plate more deeply than the dorsal shield, but in some Poraspinae (Allocryptaspis and Dinaspidella) the notch in the dorsal shield may be considerable. Derivation of the Cardipeltis condition from that of the Poraspinae requires merely a migration of the opening in a dorso-median direction so that it invades the dorsal shield deeply. Canals of the Lateral Line System : For the first time, it has been possible to determine the course of the lateral line canals in Cardi- peltis. Some obscure parts of it are doubtfully recognizable in the Beartooth Butte material, but two specimens of C. wallacii from the Water Canyon Formation (PF 804, fig. 66; PF 895, fig. 67) reveal what is possibly the complete pattern of the dorsal shield. The canals are indicated by their external openings, which appear as pores lying between, and often notching, the superficial dentine ridges. There are two pairs of longitudinal canals and probably 310 FIELDIANA: GEOLOGY, VOLUME 11 five transverse commissures. The median pair of longitudinal canals (fig. 68, mdl) may be identified with the median dorsal lines of Poras- pis, while the lateral pair (fig. 68, Idl) closely resembles the lateral dorsal lines of that genus, except for the strong medial curvature around the branchial notches. The transverse commissure that joins the lateral dorsal lines opposite the branchial notches (fig. 68, tcJ/.) is presumably the fourth commissure of Poraspis (see footnote 1, p. 300, on terminology of canals). The fifth or posterior transverse commissure (fig. 68, tc5) is obscure, but the more anterior ones, the second and third (fig. 68, tc2, tc3), are well marked. The continua- tions of the median dorsal lines anterior to tc2, as well as the median commissure joining them (fig. 68, psl), may represent the pineal- supraorbital canals of Poraspis. An alternative explanation is that the most anterior median commissure represents the median part of tc2, and that the paired anterior median canals represent the anterior part of the median dorsal lines. If the latter were true, it would mean that the shield is only a dorsal disc as in the Pteraspidae, and that the pineal, rostral, and orbital regions were anterior to it. The first explanation is favored for reasons to be discussed later but cannot be considered as definitely demonstrated by the available material. Ornamentation of Dorsal Shield : The surface of the shield is orna- mented with flat-topped ridges and tubercles composed of enamel- like and dentine-like tissues. Superficially, they appear to be smooth- edged, yet everywhere below the surface, and locally on the surface, they are laterally crimped, as is well shown in an impression figured by Bryant (1935, pi. 9, fig. 2). The ridges are approximately 1 mm. in width, and in the central part of the shield of C. waUacii they are frequently as long as 15 or 20 mm. Towards the edge they are subdivided into shorter lengths, and in the peripheral part of the shield they are broken into oval or rounded tubercles. This type of ridged ornamentation is most closely comparable to that of the Poraspinae, although the ridges are coarser than in any known mem- ber of this subfamily. Corvaspis (Woodward, 1934, pi. 19, fig. 1) resembles Cardipeltis in having the coarse longitudinal ridges broken into short lengths in the central part of the shield, and into tubercles near the periphery. The Pteraspidae are characterized by much finer ridges, while the Traquairaspidae and Drepanaspidae typically have a tubercular type of ornamentation. The pattern of the ridges is essentially a radiating one, with the center of radiation in the midline near the posterior end of the shield. Although most of the ridges in the posterior part of the DENISON: EARLY DEVONIAN FISHES 311 shield converge towards the "center of radiation," in the more an- terior part the ridges diverge antero-laterally from the midline. In the region anterior to the most anterior transverse sensory commis- sure (?pineal canal), the ridge pattern agrees closely with that of some Poraspinae (e.g. Poraspis intermedia, Kiaer and Heintz, 1935, -CA»«^ Fig. 67. CardipeUis wallacii, PF 895; anterior part of dorsal shield (X 1). pi. 17). No pineal macula has been recognized in any specimen of CardipeUis, however. The radiating pattern is interrupted by paired whorls on either side of the midline between sensory canals tc2 and tcJf. This type of ridge pattern is easily derivable from the essen- tially longitudinal one of the Poraspinae, which in certain forms (Anglaspis and some Poraspis) shows an incipient radiation of ridges from the midline in the anterior part of the shield. It is quite 312 FIELDIANA: GEOLOGY, VOLUME 11 clearly not related to or derived from the concentric pattern of the dorsal disc of the Traquairaspidae or Pteraspidae. Orientation and Extent of the Dorsal Shield : Both the presence of branchial notches and the pattern of the lateral line canals indicate that the described shields of Cardipeltis belong to the dorsal side. Obrutchev (1941, p. 21) was of the opinion that these shields had been incorrectly oriented by Bryant and by Branson and Mehl, and that they should be placed with their broad notched ends forward. This theory was based on the belief that Cardipeltis was a drepanas- pid (psammosteid), related to the Late Devonian Aspidosteus and that the Cardipeltis shield represented only the dorsal disc rather than the complete dorsal shield. Obrutchev's orientation is impos- sible, however, since the external branchial openings would thus be directed antero-dorsally. Moreover, the position of the branchial openings, as well as other features, makes any relationship with the drepanaspids unlikely. We may thus conclude that Bryant and Branson and Mehl oriented the shield correctly, but how much of the dorsal armor is represented by this shield is worth further con- sideration. The available material is not conclusive in this matter. Favoring the view that the Cardipeltis shield corresponds to the entire dorsal shield of the Cyathaspidae, rather than to the dorsal disc of the Traquairaspidae, Pteraspidae, or Drepanaspidae, are the following points: 1. The rounded anterior end of the shield in C. sinclairi would appear to be terminal. It might be argued that the notch in the anterior end of C. wallacii was occupied by a more anterior plate or plates, but since the dorsal shields are otherwise strictly comparable in these two species this is considered to be unlikely. 2. The radiating arrangement of the ornament can only be compared to that of the Poraspinae. In the Traquairaspidae and Pteraspidae, which have separate rostral, pineal, and orbital plates, the ornament of the dorsal disc is concentric and parallel to the edges, even anteriorly. The ornament appears to have little pattern in the Drepanaspidae but there are suggestions of a concentric ar- rangement in some dorsal discs (e.g., Psammolepis, Gross, 1933a, fig. 1, A). Even some of the Cyathaspidae (Cyathaspis, Vernon- aspis), where the subdivision of the shield is only incipient, show a tendency for the ornament to be concentric in the part of the dorsal shield that corresponds to the dorsal disc. 3. The pattern of the lateral lines is consistent with the view that there are no separate anterior dorsal plates; it does not prove this, however. DENISON: EARLY DEVONIAN FISHES 313 tc5 Fig. 68. Restoration of dorsal shield of Cardipeltis ivallacii (X H)- brn, branchial notch; Idl, lateral dorsal sensory line; mdl, median dorsal sensory- canal; psl, ?pineal-supraorbital sensory canal; ic2-5, dorsal sensory transverse commissures. 4. No specimens have been collected that might represent ros- tral, pineal, or orbital plates. This is, of course, purely negative evidence, but perhaps sufficient collecting has been done so that it has some weight. The following points may be interpreted as supporting the view that the Cardipeltis shield is only a dorsal disc, not the complete dorsal shield: 1. The absence of a pineal macula. This is not conclusive since the macula is indistinct in some Cyathaspidae, as Dinaspidella and Irregulareaspis. 314 FIELDIANA: GEOLOGY, VOLUME 11 2. The absence of any orbital notches. This represents an ob- vious modification of the cyathaspid condition, and while it indicates that the eyes were lateral or antero-lateral to the median shield, it does not prove the presence of orbital plates such as occur in the Pteraspidae and Traquairaspidae. 8. As pointed out above, the lateral line system of canals may be interpreted in such a way that the pineal-supraorbital line is anterior to the large median shield, presumably on separate rostral and pineal plates. Plates of the Lateral Rim : The collections from the Water Can- yon Formation include at least two types of plates that clearly formed part of the lateral rim (fig. 69). In cross section they are V-shaped, with one arm of the V shorter than the other and with the point forming a rather sharply rounded edge. On the narrower (?ventral) laminae, they are ornamented with long, narrow, crimped, and sharply crested ridges (fig. 69, B, D, F). This type of ridge carries over onto the edges of the wider (?dorsal) laminae, where it gives way to large, flat-topped, and very irregularly shaped tubercles, which may in turn grade into small, flat, circular tubercles near the medial margins (fig. 69, A, C, E, G, H). The histological structure is very similar to that of the dorsal shield of Cardipeltis, except that the middle layer is thicker, particularly at the rim; there is every reason to believe that these are lateral plates of this genus. Unfortunately, it is not possible to demonstrate the position and orientation of these plates with respect to the dorsal shield. It is probable that the wider laminae are dorsal, because the flat-topped tubercles agree more closely with the lateral ornament of the dorsal shield. There is a correspondence between one end of each of the two types, which suggests that they may have been joined along these edges. If these assumptions are correct, one plate might be comparable to the branchial plate of the Pteraspidae, while the other would be an approximate equivalent of the cornual plate of the same family; together they might be considered as homologous to the branchial plate of the Cyathaspidae. It is unlikely that the rim plates were separated from the dorsal shield by small anamestic plates, as in the Drepanaspidae ; if this were so the external branchial opening would have an improbable position far from the lateral edge. Ventral Shield: In all known Heterostraci, the ventral side of the anterior part of the body is covered by a ventral shield, and one would be expected in Cardipeltis. Strangely enough, none has yet been collected and described. This might be attributed to the E B G '^^' D H Fig. 69. Cardipeltis sp.; lateral rim plates, oriented ^'^hthe rim down (X IM) A B PF 819; C, D, PF 943; E, F, PF 821; G, PF 325; H, PF 824. A, C, E, G, and'n probably dorsal; B, D, and F probably ventral. 315 316 FIELDIANA: GEOLOGY, VOLUME 11 chances of collecting, yet to my knowledge ten dorsal shields of varying completeness have been collected at Beartooth Butte and eight from the Water Canyon Formation of Utah. These have not been found in "pockets," but have been scattered, particularly in Utah, so that the occurrence cannot be attributed to water sorting. Assuming that the chances of finding dorsal and ventral shields are equal, the probability that one would find 18 dorsals and no ventrals is only (^y^, or 0.0000038, or considering only the Utah occurrences, the probability of finding eight dorsals and no ventrals is (hY, or 0.0039. Thus, there is a strong possibility that the ventral shield as such did not exist in Cardipeltis. There are in the collections from Utah, however, a few small plates of varying size and shape, covered with large, flat-topped tubercles, except perhaps near the edge where the tubercles may be small (fig. 70, A). That they belong to Cardipeltis is indicated by their histological structure, which agrees exactly with that of the dorsal shield except in the larger size of the tubercles. If these are not anamestic plates joining the dorsal shield to the lateral plates, they may be part of a ventral armor that is not formed by a single shield but has been subdivided into small plates. Scales of the Caudal Region: There is a single scale (PF 828; fig. 70, B), from the Water Canyon Formation of Blacksmith Fork, Utah (locality A), that is presumed to belong to Cardipeltis. Rela- tive to the size of the members of this genus, the scale is small, 12.5 mm. in length and an estimated 11.5 mm. in width; this may mean either that Cardipeltis had small scales or that this scale is from the posterior part of the tail. Since it appears to be symmetrical, it is probably a median scale. Its exposed area is ornamented with shiny tubercles that are small and subcircular peripherally, and narrow, elongate, laterally crimped, and crested more medially. The median tubercle, though not well preserved, was large. There are relatively wide anterior and antero-lateral areas without ornamentation for overlap by the scales anterior to it, and on its under surface there are roughly ridged areas by which it overlapped the scales posterior to it. The ornamentation closely resembles that of the ventral lami- nae of the plates of the lateral rim, and for this reason the scale is presumed to belong to Cardipeltis. Conclusions. — Cardipeltis is a heterostracian that was probably derived from the Poraspinae. Its specialization is considerable, dif- ferent from that of other Heterostraci, and clearly warrants its retention in a family of its own. No adequately described genus Fig. 70. Cardipeltis sp. (X 5). A, small plate possibly belonging to the ventral armor, PF 814; B, scale, PF 828. ^--z?/ Fig. 71. Sections of the exoskeleton (X 30). A, Cardipeltis sp., PF 812; B, Protaspis sp., PF 944. bl, basal laminated layer; cl, cancellar layer; rl, reticular layer; si, superficial layer of dentine and enamel tubercles. 317 318 FIELDIANA: GEOLOGY, VOLUME 11 can be considered as a close relative. It is possible that Corvaspis had evolved in a similar direction, but it is too incompletely known to make close comparisons. Weigeltaspis has an ornamentation that is similar to that of parts of Cardipeltis, but only fragments have been described, so its relationships are still obscure. The Late Devonian Aspidosteus has a similarly shaped shield, but it is more probably related to the drepanaspids. Family PTERASPIDAE Claypole 1885 Next to the arthrodires, pteraspids are the most common of the Beartooth Butte and Water Canyon Formation vertebrates. A number of species have been discovered, the majority of which are referable to Protaspis. Protaspis Bryant 1933 In the first description of the Water Canyon Formation fishes, Branson and Mehl (1931, pp. 523-529, pi. 3, figs. 1-5) described a large pteraspid dorsal disc under the name of Glossoidaspis giganteus. There was nothing in their description, or in their figures of the type, that would serve to distinguish this form from Pteraspis. Un- fortunately, the type (Univ. of Missouri, 599 VP) has been mislaid, so it is not possible, at least at present, to characterize either the species or the genus. When the Beartooth Butte pteraspids were first described by Bryant (1932), he referred them to Pteraspis and a new genus, Cyrtaspis. In 1933, he erected a second new genus, Protaspis, for the species earlier referred to Pteraspis, as well as a number of new ones. There is every probability that Protaspis is identical with Glossoidaspis of Branson and Mehl, but, since this is not demon- strable, Bryant's name will be used. White (1935, p. 442) pointed out that the characters used to dis- tinguish Protaspis and Cyrtaspis were hardly of generic rank, and he considered both as subgenera of the genus Cyrtaspis; but since Cyrtaspis is preoccupied, the name Protaspis Bryant 1933 is here retained for the genus. It is used to include the subgenus Protaspis and the subgenus Cyrtaspidichthys Whitley 1940 (here new rank), replacing Cyrtaspis Bryant 1932. Amended diagnosis of the genus Protaspis. — Pteraspidae with a broad shield, low or moderately vaulted. Branchial plates extremely long, extending anteriorly to meet the rostral plates and posteriorly DENISON: EARLY DEVONIAN FISHES 319 B Fig. 72. Protaspis cf. bucheri; postero-lateral part of dorsal shield, U.S.N.M. 20524 (X 4). A, dorsal; B, ventral, bds, suture between branchial plate and dorsal disc; br, branchial opening; brp, branchial plate; cpa, anterior lamina of cornual plate; cpp, posterior scale-like process of cornual plate; dd, dorsal disc; pbr, posterior process of branchial plate. to or beyond the postero-lateral corners of the dorsal disc. Branchial openings posteriorly placed, opening upward between notches near the postero-lateral corners of the dorsal disc and the posterior ends of the branchial plates. Posterior ends of the branchial tubes bounded internally by laminae of bone that may represent the an- terior parts of the cornual plates. Small, scale-like plates attached to the posterior ends of the branchial plates possibly represent the posterior parts of the cornual plates. Median dorsal spine small and scale-like. Rostrum very short to moderately long. Pattern of the lateral line canals and histological structure of the shield as in Pteraspis. Posterior part of the body covered with fewer and larger scales than in Pteraspis; the tail nearly symmetrical. Discussion. — Protaspis resembles Pteraspis in most respects and is distinguished from it mainly by the length of the branchial plates, the position of the branchial openings, and the development of the 320 FIELDIAXA: GEOLOGY, VOLUME 11 plates bounding these openings. Neither the microstructure of the exoskeleton nor the pattern of the sensory canals shows any signifi- cant differences such as were described by Bryant. Presumably Protaspis was derived from some primitive Pteraspis, resembling P. primaeva, that had not developed a highly arched shield, strongly projecting dorsal spine and cornual plates, an elongate rostrum, or a heterocercal tail. Its derivation would involve the posterior mi- gration of the external branchial openings, together wiith the length- ening of the branchial plates and reduction and modification of the cornuals. Branchial Openings and Cornual Plates : As the most characteris- tic features of the genus, the structure of the external gill openings, the branchial tubes, and the related plates merits description in some detail. This is best shown in a number of specimens from the Water Canyon Formation of Utah (figs. 72, 73). The branchial opening lies on the dorsal surface of the shield and is bounded ex- ternally by the branchial plate and the dorsal disc, both of which are notched to form its margins. A peculiarity of this genus is that these two plates meet not only anterior to the branchial opening, as is usual in pteraspids, but also behind it ; the position of the posterior suture between the dorsal disc and the branchial plate is determin- able by the pattern of the dentine ridges (figs. 72, A; 73, A, hds). Preparation of the branchial opening shows that it is floored internally by a lamina of bone (fig. 72, B, cpa), while sections and internal preparations reveal that this lamina extends from the inner surface of the dorsal disc to the ventral lamina of the branchial plate in the immediate vicinity of the gill opening (fig. 73, B-D, cpa). Although sutures have not been observed, a schematic cross section through this region (^g. 73, D, cpa) shows that the lamina has the same fundamental relationships to the dorsal disc and the branchial plate as does the cornual plate of other pteraspids (Brotzen, 1936, figs. 1, 3). It is believed to represent the modified anterior part of the cornual plate. The common gill ducts of Protaspis extend anteriorly from their external openings over the anterior cornual Fig. 73. Protaspis sp.; PF 869 (X 2). A, dorsal view of right postero- lateral part of dorsal shield; B, lateral view of part medial to line x-y in A; C, medial view of part lateral to line x-y in A; D, transverse section through bran- chial opening; E, transverse section about 15 mm. anterior to branchial opening. bds, suture between branchial plate and dorsal disc; bpd, dorsal lamina of bran- chial plate; hpr, ventral lamina of branchial plate; br, branchial opening; brp, branchial plate; brt, common branchial duct; cpa, anterior lamina of cornual plate; dd, dorsal disc. 321 322 FIELDIANA: GEOLOGY, VOLUME 11 laminae into distinct grooves between the dorsal and ventral lami- nae of the branchial plates (fig. 73, C-E, brt) ; the branchial grooves may be traced nearly to the anterior end of the branchial plates, decreasing gradually in size forward. Faint internal impressions on a specimen of Protaspis dorfi from Beartooth Butte (fig. 74, A) suggest that there were at least seven and perhaps eight pairs of gills well within the anterior half of the shield. Posterior to the branchial openings the branchial plates are vari- ously developed in detail. In Protaspis sp. (PF 869), they project as blunt, posteriorly directed processes, triangular in section (fig. 73, A-C, brp). In others P. cf. bucheri (U.S.N.M. no. 20524; fig. 72, pbr) and P. dorfi (PF 868; figs. 75, 76)— the medial edge of the branchial plate is extended posteriorly as a thin, vertical process that embraces the medial side of a thick, triangular, scale-like plate. The position of this scale-like plate (fig. 72, cpp) is much the same as that of the posterior part of the cornuals of other pteraspids, and in the Beartooth Butte material it has been identified as the cornual (Bryant, 1983, p. 291; White, 1935, p. 441). Bryant, however, in- fluenced by a restoration of Pteraspis longirostra, figured by Zych (1931, fig. 49), believed that the cornual plate represented a modified scale in series with a row of lateral scales on the posterior part of the body. White (1935, p. 418) doubted the presence of this series of enlarged scales and they are clearly absent on a specimen of Pro- taspis dorfi (PF 226; fig. 74) in which the posterior part of the body is well preserved. If a connection existed between this scale-like posterior part of the cornual and the anterior lamina of the cornual that underlies the branchial opening, it is obscure in available speci- mens and must have been nearly completely surrounded by the posterior end of the branchial plate. Its absence in many specimens may be due to non-preservation, but in those that have a blunt, projecting, posterior tip on the branchial plate, the posterior part of the cornual may have no longer existed. The description above is based entirely on specimens from Utah, but apparently the Beartooth Butte Protaspis had a similar struc- ture. This region is not well shown in any of the material from Wyoming in Chicago Natural History Museum, but in a few speci- mens figured by Bryant the branchial opening can be seen lying posteriorly between the dorsal disc and the branchial plate. It is best displayed in a dorsal shield of Protaspis ampins (Bryant, 1935, pi. 14) and is also clear in another specimen of the same species figured by Bryant (1933, pi. 10). The branchial notch in the dorsal disc is seen in a shield of P. (Cyrtaspidichthys) sculptus (op. cit., pi. DENISON: EARLY DEVONIAN FISHES 323 16, fig. 1). Paired impressions on the type of P. (C.) falcatus (op. cit., pi. 17, fig. 1) mark the external branchial openings and indicate that Bryant has misinterpreted the inner margins of the branchial plates in this form. Lateral Line System: The sensory canals of Protaspis, as illus- trated by Bryant (1933, fig. 1, A), showed important differences from those of Pteraspis. Brotzen (1936, p. 29) doubted Bryant's reconstruction, and indeed there seems to be little evidence for the presence of the questionable intermediate pair of dorsal longitudinal canals. Two juvenile dorsal discs of Protaspis figured by Bryant (1935, pi. 10, fig. 1; pi. 11, fig. 2), as well as a dorsal shield of P. (Cyrtaspidichthys) sp. (PF 217) in Chicago Natural History Museum show a typical Pteraspis arrangement. In a few specimens of Pro- taspis from Utah (figs. 75-81), the lateral line system of the dorsal shield may be seen in considerable detail and there is complete agreement with the Pteraspis pattern. It should be noted that the most anterior transverse commissure of the dorsal disc is here identi- fied with the third of the Cyathaspidae. If this is correct, the second commissure has been lost in the Pteraspidae, except for possible remnants in Pteraspis leathensis (White, 1952, figs. 3, 7). Histology: Another character by which Bryant (1933, p. 293) distinguished the "Protaspidae" was the thinness of the exoskele- ton, which he described as being only about one-fifth of a millimeter in thickness. However, Beartooth Butte material of all groups is often characterized by thin bone, which may be due in some cases to solution and replacement, and in others to compression; the sec- tions that Bryant figures (op. cit., pi. 4) have a dense middle layer that is doubtless the result of compression. Sections of fragments of the shield of Protaspis sp. (fig. 71, B) from the Water Canyon Formation reveal a typical PteraspisAike structure with well-devel- oped cancellar and reticular layers. The total thickness is 0.80-0.95 mm. The exoskeletal histology, therefore, cannot be used to dis- tinguish Protaspis from Pteraspis. Posterior Part of Body: Bryant (1934, pis. 22, 23) has figured two specimens of Protaspis from Beartooth Butte in which the scaled part of the body was preserved; however, they do not give a clear picture of the shape and arrangement of the scales. In 1949, I was fortunate enough to obtain at this locality a specimen of Protaspis dorfi (PF 226; fig. 74, A), which, due to unusual circumstances of preservation, shows the entire posterior part of the body and tail in lateral view. The animal apparently lay on its side in the mud so that the anterior shield became much crushed obliquely, while the 324 FIELDIANA: GEOLOGY, VOLUME 11 scaled posterior part was flattened but otherwise little distorted. It exhibits important differences from Pteraspis, as described by White (1935, pp. 412 418). The length of the scaled part of this specimen (123 mm.) is only slightly less than that of the armored anterior part (135 mm.). Its depth anteriorly (26 mm.) is presumably about the same as that of the posterior part of the shield. It deepens gradually to 30 mm. at a distance of 90 mm. behind the shield and then abruptly narrows to 22 mm. on the tail proper, which begins behind a notch in the ventral edge. Bryant's specimens indicate that it was very slender laterally. The tail of PF 226 has a rounded posterior termination, modified by a small posterior lobe. The most obvious difference from Pteraspis is the shortness of the tail itself and the absence of any suggestion of an enlarged ventral lobe on it. Protaspis dorfi has a nearly symmetrical tail that could be described as a modified protocercal type. The scales, which were poorly preserved in themselves, have been removed to leave a reasonably sharp impression, except anteriorly where their arrangement is obscure. There is a series of dorsal ridge scales, probably 11 in number (fig. 74, B, drs). The third is the largest one; from here they decrease in size anteriorly and also posteriorly until they are quite small where they extend onto the tail proper. These scales are relatively much larger than those of Pteraspis rostrata, and differ also in that they overlap. They are much more closely comparable to the dorsal ridge scales of the cya- thaspid Anglaspis (Kiaer, 1932, fig. 11), although they are fewer in number. The ventral ridge scales (fig. 74, B, vrs) are probably five in number and are even longer than those on the dorsal side; the second is the longest, the first and fourth the shortest. The last of these reaches to the ventral notch about 90 mm. behind the armor. The ventral ridge scales are more closely comparable to those of Anglaspis than Pteraspis. Protaspis differs from both, however, in the absence of distinct fulcral scales dorsally or ventrally on the tail. The flank scales of Protaspis are relatively smaller and more numerous than in the Cyathaspidae, yet larger and fewer than in Pteraspis. The scales are arranged in oblique rows, each of which probably contains from 6 to 9 scales in the anterior half of the scaled region. According to White (1935, p. 415 416), there are from 9 to 15 per row in Pteraspis toombsi, while the Cyathaspidae have only two. Protaspis also has fewer rows of scales than Pteraspis; counting them to the base of the tail, there are probably 12 antero-dorsally pq X O " bo qT 73 -4^ -^ F-H 3 ,_. m rt l-M ti 4^ 43 C -t-i 0) o o > •2 X o • a c3 325 326 FIELDIANA: GEOLOGY, VOLUME 11 directed rows, while Pteraspis has from 15 to 25. The flank scales have been somewhat distorted in preservation but appear to be nearly square anteriorly, rhombic posteriorly, and elongate rhombic or irregular in shape on the tail itself. For a pteraspid that is somewhat specialized in other respects, there are many unexpected features in the scaled part of the body of Protaspis. Certainly primitive for the Pteraspidae are the large ridge scales and relatively large flank scales. The nearly symmetrical form of the tail may also be primitive, although a tendency towards an enlarged lower lobe was already present in Downtonian Cyathas- pidae, judging by Kiaer's restoration of Anglaspis. From a func- tional point of view, the Protaspis tail is more poorly adapted for active swimming than that of Pteraspis. A nearly symmetrical tail such as this would be fitted only for forward propulsion, and would not raise or lower the posterior end as do heterocercal and hypocercal tails respectively. The association of this type of tail with a heavily armored anterior body, typically showing a tendency towards a broad, flat shape, also with relatively small eyes usually directed more dorsally than in Pteraspis and most cyathaspids, and with the branchial openings removed as far as possible from the mouth, sug- gests that Protaspis was becoming adapted to a more benthonic life. The symmetrical tail would be sufficient for wriggling around on the bottom but is poorly adapted for getting the animal off the bottom or for free swimming. European Occurrences of Protaspis: Protaspis arnelli Brotzen, from the "Old Red" of Podolia, has a rather broad, flat shield with elongate branchial plates as is typical of this genus, and the branchial openings are placed at the postero-lateral corners of the dorsal disc. But Brotzen's illustrations (1936, figs. 3, 6, and pi. 1) show that the branchial openings are not enclosed posteriorly by a junction be- tween the dorsal disc and branchial plates. The very small cornual plates exclude the dorsal disc from the external branchial opening, and lack the scale-like posterior process. These differences suggest that the Podolian form may have acquired its Protaspis-\ike charac- teristics independently, yet they are hardly adequate to characterize a distinct genus. The only other European Protaspis is P. wiheriesiensis Brotzen from the Gres de Wih^ries of Belgium (originally figured by Leriche, 1926, pi. 2, fig. 1, as Pteraspis dunensis). In this species, the bran- chial opening is enclosed behind, and there is a postero-lateral proc- ess that resembles the scale-like, posterior part of the cornuals of DENISON: EARLY DEVONIAN FISHES 327 the North American species. The shape and proportions of the dor- sal shield agree with those of some members of the subgenus Cyrt- aspidichthys. The differences from the North American Protaspis, as far as they can be determined from Leriche's illustration, are very slight. Specific Characters in the Pteraspidae The systematic determination of the specimens of Protaspis from Utah and Wyoming is difficult and it cannot be satisfactory without a much larger collection of well-preserved material. Nearly every determinable specimen shows differences that may represent only individual variations, or, on the other hand, may be valid specific characters. In modern vertebrates, there is usually some basis for assessing the significance of such differences, at least by comparisons with the situation in other related groups. But among the ostracoderms, there is not a single study of variation sufficiently detailed to offer a sound starting point for a systematic evaluation. I know of only one described pteraspid occurrence that is suitable for a study of intraspecific variation; this is at the Wayne Herbert quarry, near Newton, in southwestern Herefordshire, England. From a thin, lenticular band of siltstone, representing the deposits of a single pool, White (1935) reports the occurrence of a score of individuals of Pteraspis. Most of them belong to P. rostrata and presumably represent members of a single population and perhaps a school of one age group. Before considering the pteraspids of Utah and Wyoming, it will be instructive to consider the kind and degree of individual variation encountered in the Wayne Herbert quarry specimens. An attempt will be made to evaluate the charac- ters that have been used commonly to distinguish species in this family. Size. — There is clear evidence of growth of the pteraspid shield. In the first place, it is not unusual to find remains of juvenile indi- viduals with their small plates still unfused. In the second place, growth lines occur on the different plates, represented, for example, by concentric ridges on the dorsal and ventral discs, or by ridges parallel to the posterior edge on the dorsal side of the rostral plate. It may be questioned, however, whether the shield continued to grow after the fusion of the plates. Larger individuals almost al- ways have the rostral, pineal, orbital, and branchial plates and the dorsal disc ossified together with sutures indistinct, especially inter- nally. Further growth would then be possible only after resorption 328 FIELDIANA: GEOLOGY, VOLUME 11 along the sutures. In modern vertebrates, such a process would be most improbable, but there is at least a possibility that it occurred in the Pteraspidae. It is much more probable, however, that the size attained before fusion of the plates was subject to individual as well as specific variation. If this were so, the size of any individual would not be a particularly reliable indication of its specific identity. According to White, the individuals of Pteraspis rostrata from the siltstone lenticle of the Wayne Herbert quarry agree closely in size. My measurements from his figures of five of these indicate that the length of the dorsal disc anterior to the growth center has a range from 35.0 to 40.5 mm. fM=38.5, (7=2.5, V=6.6). If these fishes are of the same age and belong to a single school, a higher variability might be expected in any population. In six specimens of Protaspis (Cyrtaspidichthys) ovatus the total length of the dorsal disc ranges between 65 and 92 mm. The type, whose dorsal disc is about 75 mm. in length, is one of the few known specimens of a large individual with the plates of the dorsal shield unfused; it is at least consistent with the view that fusion of the plates took place at different sizes within a species. The specimens that have been referred to Protaspis amplus and P. perlatus have dorsal shields ranging between 115 and 180 mm. in length. Nearly all the characters by which they are distinguished are clearly cor- related with size, so their specific distinction must be considered to be uncertain. General proportions. — These may be given by the proportions of the entire dorsal shield or by the maximum width length of the dorsal disc or ventral shield. It is a character which must be used with caution for two reasons: (1) The proportions are commonly greatly altered by distortion during preservation. Generally this results in a dorso-ventral flattening, which may increase the length only slightly but usually increases the width considerably, especially posteriorly, where the original arching of the shield is greatest. The degree to which this distortion may affect the proportions is shown by a single specimen of Protaspis amplus (PF 243). Measuring the width of the dorsal disc on the badly flattened right side, its W L is 1.41; if the width is measured on the relatively uncrushed left side, W L =1.02 (lengths measured as described below). (2) Pro- portions change with growth. This is readily demonstrable by the concentric, ridge-like thickenings on the dorsal and ventral discs, which almost certainly represent growth lines and allow measure- ment of the size and proportions of a single disc at various stages. DENISON: EARLY DEVONIAN FISHES 329 On casual inspection, it is obvious that the growth increments along the midline of the dorsal disc anterior to the center of growth are in general greater than those in a lateral direction on either side of the midline. The increments posterior to the center of growth are relatively small and unfortunately rarely completely preserved. For this reason, in measuring the dorsal discs, the length has been taken in the midline anterior to the center of growth and thus is not the total length; anteriorly, it has been measured to the line connecting the most anterior points of the disc at any stage. The widths measured have been the maxima, generally well anterior to the center of growth. Data for comparison have been obtained from dorsal discs of Protaspis from Utah and Wyoming and from White's figures of Pteraspis from England. They suggest that (a) at an early stage, such as when the dorsal disc is only 10 or 15 mm. in length, the interspecific differences in proportions are smaller and in some cases of doubtful significance, (b) if one plots the lengths against the widths of a single dorsal disc at different stages of growth, the resulting curve may be of value in determining or distinguishing the species. In any case, the use of the proportions of a dorsal disc without consideration of its relation to the size of the disc has little significance, as is obvious by reference to the table below, derived from measurements of different growth stages of the dorsal disc of a specimen of Protaspis dorfi (PF 868; figs. 75, 76): Length W/L Length W/L 20.6 mm. 1.56 49.0 mm. 1.16 29.4 mm. 1.33 56.2 mm. 1.11 34.4 mm. 1.25 66.7 mm. 1.08 40.5 mm. 1.17 75.7 mm. 1.04 To test the variation in proportions within a single species, five examples of Pteraspis rostrata from the Wayne Herbert quarry have been measured as described above^from figures of White (1935, fig. 3; pi. 25, fig. 99; pi. 26, figs. 104, fo5, 106). For each growth stage of the dorsal disc, the length has been plotted against the width, and curves have been fitted approximately to the plotted points. From this graph, the proportions were computed for a dorsal disc of a length of 35 mm. and found to range between 1.13 and 1.24 mm. Since these specimens are more or less crushed and distorted, this gives a fair estimate of the amount of variation in proportions to be expected in a small sample of fossilized individuals of the same size and same species. Shape of the dorsal disc. — White's (1935) study of Pteraspis ros- trata indicates that there may be considerable individual variation 330 FIELDIANA: GEOLOGY, VOLUME 11 in such features as the width of the anterior emargination and the roundness of the antero-lateral corners. Many characters, such as the curvature of the lateral margin, may be altered considerably by crushing and during growth. There is some reason to suspect that the thin, projecting posterior margin may be commonly broken off, and if this is so, its shape cannot be used without first establishing its completeness. White (1935, p. 420) states that growth "stops at an early stage along the hinder margin of the dorsal disc, except in P. leathensis and P. jackana," yet, in a specimen of P. rostrata that he illustrates in his figure 9 this is not the case. Among the Beartooth Butte Protaspis and Cyrtaspidichthys, Bryant has figured a number of specimens in which growth is clearly continuous around the pos- terior edge, and yet in other specimens the margins are commonly incomplete. There are differences in the arching or convexity of the pteraspid shield, but because of flattening in preservation it can be used only rarely as a systematic character. The arching of the dorsal disc is moderate in most Pteraspis but high in such forms as P. stensioi and P. jackana, while Protaspis usually has a rather fiat shield. One could not expect to determine the extent of intraspecific variation from material available at present. Proportions of the rostrum. — Within the Pteraspidae, this charac- ter has a systematic importance that is obvious when one compares the extremes, such as Rhinopteraspis dunensis with its long, slender rostrum, and Pteraspis vogti with its short, stubby rostrum. The amount of variation within a species can best be studied in Pteraspis rostrata. The relative length of the rostrum is given by the ratio of the rostral length (RL) to the total length of the shield (TL). In five specimens from the Wayne Herbert quarry, measured from White, this ratio is 0.33-0.37; three specimens from other localities have corresponding ratios of 0.32-0.33. Apparently the relative length of the rostrum is not particularly variable, at least within this species. However, the possibility is suggested in some Protaspis from Beartooth Butte that the relative length of the rostrum in- creases with growth. Thus, the largest individuals in the P. amplus- perlatus-perryi group have relatively the longest rostra, and the same applies, though perhaps less clearly, in P. bucheri and P. brevirostris. The relative width of the rostrum is given by the ratio of the rostral width (RW) to RL. In five specimens from the Wayne Herbert quarry, this ranges between 0.60 and 0.77, while the other three specimens have the ratio 0.74-0.89 (White, 1935, p. 400, gives DENISON: EARLY DEVONIAN FISHES 331 the ratio as 0.50-0.87 in this species). Apparently there is con- siderable intraspecific variation. It may be due in some degree to crushing, which, because of the greater lateral convexity, is apt to increase the width more than the length. It may also be due to changes of proportion with growth. In the rostrum, the growth lines and dentine ridges of the dorsal side run parallel to the pos- terior edge of the plate. It seems probable that the anterior part was formed first, and that growth was by the addition of increments to this posterior edge. If this theory is correct, the size and propor- tions of the plate can be determined at different stages of growth as in the dorsal disc, and this has been attempted in a few specimens of Protaspis and Cyrtaspidichthys from Utah and Wyoming. Changes in proportions with growth are shown in all cases, of which two examples are given below: Protaspis (Cyrtaspidichthys) sculptus (from Bryant, 1933, pi. 16, fig. 2) Protaspis (Protaspis) bucheri (from Bryant, 1933, pi. 2, fig. 1) RL RW/RL RL RW/RL 12.8 mm. 2.81 11.5 mm. 2.74 16.0 mm. 2.56 16.5 mm. 2.21 18.6 mm. 2.40 23.5 mm. 1.70 23.1 mm. 2.14 Both Protaspis and Cyrtaspidichthys regularly show such a pro- portional change with increase in size, resulting in a relatively nar- rower rostrum at a larger size. It has not been possible to make similar measurements from published illustrations of Pteraspis ros- trata. But when one plots the RL against the RW of different-sized individuals of this species, there is a suggestion of narrower propor- tions in larger individuals, comparable to the situation in Protaspis. The change in proportions during growth appears to be different in Rhinopteraspis dunensis, where the smallest specimen figured by Gross (1933b, fig. 2, D) is proportionately narrower than the large individuals (op. cit., fig. 2, A, C, E) ; this suggests that the rostrum became relatively broader with growth in this species. The data are sufficient to indicate clearly that size must be considered when comparing rostral proportions. Shape of the rostrum. — The amount of intraspecific variation that can occur is indicated by White's (1935, figs. 86-90) restorations of the varieties of Pteraspis rostrata. The tip of the rostrum may be sharply pointed or broadly rounded. The angle between the lateral edges may be as little as 37° or as much as 58°. The posterior edge may be broadly or sharply curved. The interspecific variations are usually very much greater, as between Pteraspis rostrata, P. crouchi. 332 FIELDIANA: GEOLOGY, VOLUME 11 and Rhinopteraspis dunensis. On the other hand, the rostra are very similar in P. jackana and P. stensioi, also in the four species referred to the subgenus Simopteraspis. Pineal and orbital plates. — The shape of the pineal plate and the presence or absence of a contact between the pineal and orbital plates are clear-cut characteristics that have been used to define a species. A small, rounded or triangular pineal, widely separated from the orbitals, is a primitive feature in the Pteraspidae, and is characteristic of the early subgenus Simopteraspis. Most later Pter- aspidae have developed on the orbital plates long medial processes that acquire blunt contacts with the pineal; the pineal plate then assumes a more or less rectangular shape. That this feature is in- constant is shown by the examples of Pteraspis crouchi, P. rostrata, and Rhinopteraspis dunensis; in these species, some individuals lack the long medial process of the orbitals and the orbital-pineal contact, and may also have a triangular pineal. Dorsal spine. — Since this element is easily detached and rarely preserved, it is not always usable systematically. It does show some extreme modifications in the Pteraspidae. In Rhinopteraspis, it is very elongate, slender and rounded; in Pteraspis jackana and P. stensioi, it forms a dorsally directed, blade-like cut- water; in the primitive Simopteraspis, it is small and scale-like. There is not suf- ficient information to assess the extent of intraspecific variation, although White (1935, p. 401) indicates that the relative lengths of the spine and of its inserted portion are subject to individual varia- tion in Pteraspis rostrata and P. crouchi. Branchial plates. — The width of the branchial plate has been used by Bryant (1933) as a specific character in Protaspis and Cyrt- aspidichthys. Its length compared to that of the cornual plate has been shown by White (1935, p. 402) to be subject to variation in Pteraspis rostrata. The latter feature is related to the position of the branchial opening, and with the extreme lengthening of the branchial plates in Protaspis becomes a characteristic of generic weight. The shape of the posterior termination of the branchials may be of systematic importance in Protaspis. Cornual plates. — The preservation of these plates is so uncertain that their use in systematics is limited and a study of their variation is not now possible. Primitively, they are small, triangular plates in Simopteraspis. Extreme modifications include the following: great lateral projections in Pteraspis podolica and P. lerichei; elongate, slender form in Rhinopteraspis, projecting neither laterally nor pos- teriorly; extreme reduction in Protaspis. DENISON: EARLY DEVONIAN FISHES 333 Ventral disc. — The remarks that were made above about the dorsal disc apply for the most part equally well here. The propor- tions and shape can be altered considerably by flattening as well as during growth (White, 1935, pp. 403 404). Ornamentation. — The surface of the pteraspid shield is covered with rather fine, crenate ridges that in some cases are crossed by transverse grooves, or subdivided into papillae. The ridges may be flat-topped (as in Pteraspis crouchi and Rhinopteraspis leachi), round- topped (as in P. rostrata, P. jackana, and P. dixoni), or sharp-crested (as in P. leathensis). The pattern of the ridges is in general similar in known pteraspids, except for P. dixoni, which shows a peculiar and distinctive development (White, 1938, p. 105). The shape of the ridges may not always be a reliable systematic character, as is indicated by a specimen of Protaspis dorfi (PF 868) , where it varies from flat to crested within a small area. The coarseness of the dentine ridges has been used to distinguish species, particularly by Bryant (1932, 1933). At first, White (1935, p. 420) found little evidence of specific differences in this character among the British species, which he reported as having 50 80 ridges per cm. Later (1938, p. 106), he recognized Pteraspis dixoni as having particularly coarse ridges, numbering 38-50 per cm. It is clear that among the North American pteraspids the coarseness of the ridges may differ significantly between species, but at present there is not much information about intraspecific variation. Speci- mens referred to Protaspis bucheri show the following variation in the midline of the dorsal disc: Princeton no. 13519 (measured from Bryant, 1933, pi. 2, fig. 1) has 48 ridges per cm.; C.N.H.M.-UC 2209 has 53 per cm.; C.N.H.M.-PF 240 has 57 64 per cm. Along the midline of the dorsal disc, there are minor but noticeable varia- tions in the coarseness of the ridges within a small area. As White (1935, p. 420) has pointed out, the ridges in the mid- line of the dorsal disc are coarser than the ones at the sides of the disc. The difference may be considerable: thus, in the type of Pro- taspis priscillae (PF 867), there is an average of 84 ridges per cm. in the midline and 113 per cm. laterally. This is clearly related to differences in the relative rate of growth of diff"erent parts of the disc, but the available data do not suggest any immediately obvious correlation with the proportions of the disc. The ridges are generally somewhat finer on the midline of the rostrum than on the midline of the dorsal disc; PF 867 has 97 per cm. on the rostrum. This indicates that while the coarseness of the ridges may be an important 334 FIELDIANA: GEOLOGY, VOLUME 11 and valid specific character, counts given without specifying where they were taken have little significance. Species of Protaspis from Wyoming In papers on vertebrates from the Beartooth Butte Formation, Bryant (1932, 1933) described nine species of the subgenus Protaspis and four of the subgenus Cyrtaspidichthys. This is a large number of species of a genus for a single locality, and since it will be neces- sary to compare the Utah forms with them, it is considered necessary to review Bryant's species in light of the remarks above on variation in the family. The type material from Beartooth Butte has not been studied, so the following revisionary comments must be con- sidered as provisional. Protaspis dorfi is clearly distinguished from all other species by its fine ornament; Bryant says there are 80 ridges per cm., while PF 241 shows 93 per cm. in the midline of the dorsal disc and about 120 per cm. on the midline of the rostrum. In general proportions, the short rostrum (RL/TL= 0.14 0.18), and the very narrow bran- chial plate, this species is close to P. bucheri and P. brevirostris. But the rostrum terminates more sharply, and one uncrushed specimen (Princeton no. 13649) indicates that the dorsal disc may be nar- rower, more highly arched, and with rounder lateral contours. The size attained is large; one ventral disc measures 150 mm. in length. Protaspis bucheri and P. brevirostris resemble each other in the moderate proportions of the dorsal disc, the narrow branchial plates, the short rostrum, and the coarse ornamentation (about 45-65 ridges per cm. in the midline of the dorsal disc). The relative length of the rostrum (RL TL) is 0.11-0.19, and it tends to be proportionately longer in larger individuals, suggesting that it is related to individual growth; there is no clear distinction between the species in this re- spect, although the type of P. brevirostris, a small individual, has a very short rostrum. The proportions of the rostrum (RW RL) are variable between 1.4 and 2.6. In connection with the suggestion made above that this feature is correlated with growth, it is in- teresting to note that the widest rostrum (RW RL=2.6) occurs in the smallest individual, and the narrowest rostrum (RW/RL=1.4) occurs in the largest. The two species cannot be differentiated by this character. In shape, the rostrum may be very blunt to broadly rounded, but the differences are no greater than those in Pteraspis rostrata. Since there appears to be no valid reason for distinguishing these two species, P. brevirostris is here referred to P. bucheri. DENISON: EARLY DEVONIAN FISHES 335 Protaspis amplus, P. perlatus, and P. perryi are all species with a broad shield, broad branchial plates, moderate to long rostrum, and coarse ornament (about 50 ridges per cm.). P. perryi was dis- tinguished by having the dentine ridges broken into short lengths. This may be a valid character, but on the other hand it may well be only an individual variation, and in the absence of other distin- guishing features this species cannot be considered to be firmly es- tablished. The relative length of the rostrum shows considerable variability: in P. perlatus, RL/TL= 0.23-0.24; in P. amplus it is 0.27 0.32; in P. perryi it is estimated to be about 0.25-0.27. But when the rostral length is plotted against the total length, the points fall nearly on a straight line, suggesting that the proportions are a function of the individual size. For example, the longest rostrum (RL/TL=0.32) occurs in an extremely large individual of P. amplus (TL=170 mm.), while a smaller individual is not sharply distinct from a large P. perlatus. In its proportions, the rostrum is broader in P. perlatus (RW/RL=2.0), and relatively narrower in P. amplus (RWRL= 1.4-1.5); this character is clearly correlated with indi- vidual size and is not a valid specific distinction. P. amplus appears to have relatively broader branchial plates; there is no information on the variability of this feature, but it may also be dependent on growth. Additional study of these species is necessary. For the present, they will be referred to as the P. amplus-perlatus-perryi group, leaving open the question of their distinction. Protaspis nanus was characterized particularly by its slender form. Bryant's figure of the type (1933, pi. 9, fig. 1) shows beyond any question that the dorsal disc is incompletely preserved, for the dentine ridges in its antero-lateral region are directed laterally. If the pattern of these ridges is considered in restoring the type, it appears that this must be a form with a relatively broad shield. This, of course, excludes the possibility that the slender ventral shield referred here by Bryant (op. cit., pi. 9, fig. 2) belongs to the same species. The oval pineal plate not in contact with the orbitals can- not be considered as more than an individual variation. The so- called transverse sensory canal on the middle of the dorsal disc has no counterpart in any other pteraspid and is either a growth line or a crack. It would thus appear that the type of P. nanus belongs to a broad-shielded species with a moderately long, broad rostrum, and coarse ornamentation. Presumably, it should be referred to the P. amplus-perlatus-perryi group. Protaspis cingulus and P. constrictus are based on single ventral shields and are not well characterized. The sharply depressed flange 336 FIELDIANA: GEOLOGY, VOLUME 11 on the posterior border of P. cingulus (Bryant, 1933, pi. 13, fig. 1) is not found in other pteraspids and is possibly the result of a break or crushing. The constriction near the posterior end that Bryant (1933, p. 303, pi. 13, fig. 2) used to characterize P. constrictus occurs in other well-preserved though somewhat flattened ventral shields and is probably not by itself a valid specific character. P. cingulus has about 50 ridges per cm. and may belong to P. bucheri. The ornament of P. cingulus is unknown; this may be the ventral shield of P. bucheri or P. dorfi. P. (Cyrtaspidichthys) ovatus and P. (C.) papillatus are probably not distinct. There are differences in the shape of the rostrum, but less than occur between individual varieties of Pteraspis rostrata. According to Bryant (1933, pp. 305-306), there is a difference in the relative length of the rostra. According to him, RL/TL=0.33 in P. (C) ovatus (Princeton no. 13533), but as I restore this speci- men it is more nearly 0.28; in another specimen (Bryant, 1935, pi. 16) it is 0.25. In P. (C.) papillatus RL/TL=0.25, so the relative length of the rostrum is about the same in the two species. The posterior termination of the dorsal disc was described as being more acutely pointed in P. (C.) papillatus, but because of probable in- complete preservation this is not a reliable distinction. The de- pressed area on the rostrum described by Bryant (1933, p. 306) in the type of P. (C.) papillatus coincides approximately with the thickened ventral rostral rim (as indicated in specimens of Protaspis from Utah); it is probably a result of crushing around this rim. P. (C.) papillatus is referred to P. (C) ovatus. Protaspis (Cyrtaspidichthys) falcatus was based on an incomplete dorsal shield that was apparently misinterpreted by Bryant (1933, p. 307, pi. 17, fig. 1). Points of reference on this specimen are the pineal foramen anteriorly, and paired markings postero-laterally that clearly indicate the branchial openings. Since the branchial openings in this genus lie between the branchial plates and the dorsal disc, the branchial plates cannot have been enormously wide, as Bryant believed. This leaves nothing to distinguish P. (C.) falcatus from P. (C.) ovatus. Protaspis (Cyrtaspidichthys) sculptus has more slender propor- tions than P. (C.) ovatus, at least in adult specimens. The rostrum is somewhat shorter and more bluntly rounded; RL/TL=0.20 in PF 217, compared to 0.25-0.28 in P. (C.) ovatus. This species is probably valid. DENISON: EARLY DEVONIAN FISHES 337 In summary, the following species of Protaspis from Wyoming can be considered as well established: P. {Protaspis) dorfi, P. (P.) hiicheri, P. (P.) amplus-perlatus-perryi group, P. (Cyrtaspidichthys) ovatus, P. (C.) sculptus. Species of Protaspis from Utah The only pteraspid previously described from Utah is Glossoid- aspis giganteus Branson and Mehl (1931, p. 523), which, as was mentioned above, is probably identical with the Beartooth Butte Pro- taspis. The type is a dorsal disc of moderately slender proportions, with rounded lateral contours and fine ornamentation (100 ridges per cm. according to Branson and Mehl, or 85 per cm. in the antero- lateral region as measured from their illustration, pi. 8, fig. 3). The only Beartooth Butte species with which it is comparable is P. dorfi, but identity cannot be established from the figures of this specimen. A number of species belonging to the subgenus Protaspis are repre- sented in the Chicago Natural History Museum collections, and one is in the United States National Museum. Unfortunately, only a few of them are complete enough to allow a satisfactory determination. Protaspis (Protaspis) dorfi (Bryant) PF 868, a badly crushed dorsal shield (fig. 75) from locality A, Utah, is indistinguishable from the Wyoming specimens of P. dorfi. The dimensions and proportions of the dorsal disc, as meas- ured from the less crushed right side, are given at different growth stages on page 329. The total length of the shield, excluding the dorsal spine, is 121 mm.- rather small for this species. The rostrum is short (RL TL=0.17) and has a rather sharply rounded tip, agree- ing closely with that of P. dorfi. The branchial plates are narrow, and the ornamentation is very fine; there are 89 ridges per cm. in the midline and 114 per cm. laterally on the dorsal disc, while on the midline of the rostrum there are 107 per cm. The fine dentine ridges are deeply crimped and may be either flat-topped or sharply crested within a small area. The dorsal spine (fig. 76, dsp) is scale- like, but considerably longer than shown in Bryant's restoration (1933, fig. 1, B). On the right side, the branchial opening is fairly well shown; posterior to it the scale-like posterior process of the cornual plate is closely joined medially to the postero-lateral edge of the dorsal disc. On the dorsal surface the branchial plate does not extend far posterior to the branchial opening, but on the ventral side it sends a thin, posterior process along the inner side of the scale-like process of the cornual plate. This region agrees in all 338 FIELDIANA: GEOLOGY, VOLUME 11 Fig. 75. Protaspis dorfi; crushed dorsal shield, PF 868 (X 1). essentials with that of P. cf. bucheri, described on page 322. A new restoration of P. (P.) dorfi, based on PF 868, is given in figure 76. The pattern of the dorsal sensory canals agrees with that of Pteraspis. Protaspis (Protaspis) cf. bucheri (Bryant) A dorsal shield (figs. 77, 78), little crushed, and complete except postero-medially, is doubtfully referred to this species. It was col- lected by a student in Utah State Agricultural College from Water DENISON: EARLY DEVONIAN FISHES 339 Fig. 76. Restoration of dorsal shield of Proia^pis dorfi (X 1). br, branchial opening; brp, branchial plate; cpp, posterior scale-like process of cornual plate; dd, dorsal disk; dsp, dorsal spine; gl, growth lines; Ml, lateral dorsal sensory line; mdl, median dorsal sensory line; or, orbital plate; pi, pineal plate; ro, rostral plate; sol, supraorbital sensory line; (c3-5, dorsal transverse sensory commissures. Canyon Formation in Cottonwood Canyon, near locality G, and is now in the United States National Museum collections (no. 20524). It has been beautifully prepared on both sides and shows many interesting anatomical details. Dr. David H. Dunkle, through whose kindness I have been able to study this specimen, originally compared 340 FIELDIANA: GEOLOGY, VOLUME 11 it (Williams, 1948, p. 1139) to P. brevirostris, a species that is con- sidered in this paper to belong to P. hucheri. It resembles P. bucheri in having moderately broad shield and branchial plates, short, broadly rounded rostrum, and coarse ornamentation. However, it is very small and has the plates of the dorsal shield completely Fig. 77. (X IH). Protaspis cf. bucheri, dorsal view of dorsal shield; U.S.N.M. 20524 fused so that the sutures cannot be determined on the inner side. The total length of the shield, excluding the dorsal spine, is esti- mated to be 70 mm., while the smallest specimen of this species from Beartooth Butte (except for the obviously juvenile individuals with unfused plates) is about 90 mm. long. The small size may or may not be a valid specific criterion, but in the absence of other clearly distinctive features, it is considered to be unwise to refer the speci- men to a new species. DENISON: EARLY DEVONIAN FISHES 341 The ornamentation consists of flat-topped ridges that number, on the dorsal disc, about 57 per cm. in the midhne anteriorly and about 79 per cm. laterally. In the dorsal disc, the maximum width (W) is 47.4 mm. and the length anterior to the growth center is estimated to be 48 mm., giving a W L ratio of 0.99. This is somewhat Fig. 78. Protaspis cf. bucheri, ventral view of dorsal shield; U.S.N.M. 20524 (X VA). narrower than in the Beartooth Butte P. bucheri of comparable size (estimated from growth rings), but the difference may be accounted for by the crushing of the latter. The rostrum is short (RL=12.5 mm.), but RL estimated TL==0.18, which is relatively long for a small P. bucheri; in the smallest specimens from Beartooth Butte, this ratio is 0.11 0.12. The rostral width is 31.4 mm., giving RW/RL=2.5; this is similar to the rostral proportions of P. bucheri 342 FIELDIANA: GEOLOGY, VOLUME 11 of comparable size. The structure of the posterior part of the bran- chial plate, the branchial opening, and the cornual plate (figs. 72, 77-79) has been described above (p. 322) and is similar to that of Fig. 79. Restoration of dorsal shield of Protaspis cf. bucheri, based on U.S. N.M. 20524 (X 1'2)- br, branchial opening; brp, branchial plate; cpp, posterior scale-like process of cornual plate; dd, dorsal disc; Idl, lateral dorsal sensory line; mdl, median dorsal sensory line; or, orbital plate; pbr, posterior process of branchial plate; pi, pineal plate; ro, rostral plate. P. dorfi. This region is not clearly displayed in any of the Bear- tooth Butte P. hucheri, but a specimen figured by Bryant (1933, pi. 1) appears to have a similar arrangement. Because of the dif- ferences in size and rostral proportions, the reference of this speci- men to P. hucheri must be considered as doubtful. DENISON: EARLY DEVONIAN FISHES 343 The inner side shows some interesting details. The anterior lamina of the cornual plate that forms the floor of the branchial tube below the branchial opening is attached to the ventral lamina of the branchial plate but is displaced by minor crushing from its usual contact with the dorsal disc (figs. 72, B, cpa, 78). The struc- ture of the suborbital region (fig. 78) differs in detail from that of Pteraspis vogti as figured by Kiaer (1928, fig. 2). The branchial plate extends forward to meet the rostral plate, thus excluding the orbital plate from contact with the lateral plate; this feature is ap- parently characteristic of Protaspis. The lateral plate is preserved in its natural position on the medial side of the rostral and branchial plates, but the small oral plates are gone. There is a well-developed ventral rostral beak but no indication of any maxillary tooth plates such as Kiaer described. At the anterior part of the inner side of the rostrum, there are paired depressions that presumably housed the nasal capsules. Three specimens in Chicago Natural History Museum from lo- calities A and G (PF 345, PF 877, PF 902) are also comparable to P. hucheri. These are larger than U.S.N.M. no. 20524, the smallest (PF 345) having an estimated total length of 94 mm. The orna- mentation is moderately coarse (60-65 ridges per cm. in the midline of the dorsal disc), while the proportions of the dorsal disc and ros- trum are similar to those of U.S.N.M. no. 20524, Protaspis (Protaspis) tenuistriatus,' sp. nov. Type. — C.N.H.M.-PF 333, a relatively uncrushed, but incom- plete dorsal shield (fig. 80). Horizon. — Early Devonian, near base of Water Canyon Formation. Locality. — Cottonwood Canyon (locality G), NE. 3^, sec. 19, T. 13 N., R. 3 E., Cache County, Utah. Diagnosis. — A Protaspis with very fine ornament, about 104 ridges per cm. in the midline of the dorsal disc. Dorsal disc and branchial plates broader than in P. dorfi, narrower than in the P. amplus-perlatus-perryi group. Rostrum moderately long and nar- row, with a broadly rounded anterior end; RL/TL estimated to be 0.21, and RW/RL=1.6 in the type. Discussion. — As this incomplete shield has been restored (fig. 81), its total length (excluding the dorsal spine) is about 140 mm. The proportions cannot be determined exactly but the shield appears to 1 tenuis, fine-^siriatits, furrows; in reference to the extremely fine ornament. 344 FIELDIANA: GEOLOGY, VOLUME 11 Fig. 80. Protaspis ienuistnatus, sp. nov., type; PF 333, incomplete dorsal shield (XI). be slightly broader than in P. dorfi and narrower than in the P. amplus-perlatus-perryi group. The ornamentation is even finer than that of P. dorfi, with about 104 ridges per cm. in the midline, 116 per cm. laterally on the dorsal disc, and 106 per cm. on the midline of the rostrum. The form of the ridges is variable. The rostrum is longer and more broadly rounded anteriorly than in P. dorfi, and shorter and narrower than the rostrum of a member of the P. amplus- perlatus-perryi group of comparable size. The branchial plates are very incompletely preserved, but are rather narrow, and meet the rostral plates anteriorly. The branchial opening is not preserved -Idl Fig. 81. Restoration of Protaspis tenuistriatus; dorsal shield (X 1). brp, branchial plate; dd, dorsal disc; gl, growth line; Idl, lateral dorsal sensory line; mdl, median dorsal sensory line; or, orbital plate; pi, pineal plate; ro, rostral plate; sol, supraorbital sensory line; tc3, third dorsal transverse sensory commissure. 345 Fig. 82. Protaspis priscillae, sp. shield (XI). nov., type; PF 867, incomplete dorsal Fig. 83. Restoration of dorsal shield of Protaspis priscillae (X 1). brp, branchial plate; dd, dorsal disc; gl, growth line; Ml, lateral dorsal sen.sory line; mdl, median donsal sensory line; or, orbital plate; pi, pineal plate; ro, rostral plate; sol, supraorbital sensory line. 346 n brp Idl 347 348 FIELDIANA: GEOLOGY, VOLUME 11 but it must have been far posterior, indicating that this form be- longs to Protaspis. Another fragment of a dorsal shield (PF 344) from Blacksmith Fork (locality A) has extremely fine ornamentation (120 ridges per cm. laterally on the dorsal disc) and is probably referable to the same species. It preserves a branchial plate posteriorly as far as the anterior edge of the branchial opening and indicates that the latter is in the usual position for Protaspis. Protaspis (Protaspis) priscillae/ sp. nov. Type.—C.N.B.M.-FF 867, a dorsal shield lacking the posterior end and the tip of the rostrum (fig. 82) . Horizon. — Early Devonian, Water Canyon Formation, horizon undetermined. Locality. — Northwest of Dry Lake, on the east edge of the Wells- ville Range, SW. H, sec. 29, T. 10 N., R. 1 W., Cache County, Utah (locality I). Diagnosis. — A species attaining rather large size; the dorsal shield of the type has an estimated total length (excluding the dor- sal spine) of 165 mm. Proportions very slender. Rostrum long and tapering gradually, with its dentine ridges forming sharp, anteriorly directed points in front of the antero-lateral processes of the orbital plates. Pineal plate narrow, straight-edged anteriorly, convex pos- teriorly, and strongly concave laterally. Orbital plates with sharply pointed antero-lateral processes and very slender postero-lateral proc- esses. Branchial plates extremely slender, apparently meeting the postero-lateral processes of the rostral plates. Dorsal disc long, slender, and rather strongly arched, with only slightly curved lateral margins. Ornamentation consisting of fine, probably round-topped ridges, 84 per cm. in the midline and 113 per cm. laterally on the dorsal disc. Discussion. — In many features this species is sharply distinct from any previously described Protaspis. Since the posterior part of the shield is not preserved, its reference to this genus cannot be demonstrated easily but the absence of a deep groove for the bran- chial duct in the posterior end of the preserved part of the branchial plates supports it. Where the latter are broken off, they show in section only a small notch for this duct, while other specimens of ' Named after Mrs. William D. Turnbull, who discovered the type and only known specimen. DENISON: EARLY DEVONIAN FISHES 349 Protaspis sp., such as PF 869 and PF 878, have a rather deep groove in the branchial plate extending 20 25 mm. anterior to the branchial opening. This indicates that the branchial openings of P. priscillae must be far posterior and thus its reference to Protaspis is extremely probable. This is also indicated by the fact that the branchial plates extend far forward and appear to contact the postero-lateral processes of the rostral plates. In its general shape and proportions (fig. 88), P. priscillae may be compared with the British Rkinopteraspis leachi and the Podolian Pteraspis longirostra and P. elongata. It is probable that the rostrum was not as long as in the first two species, but all of these forms have a very slender, elongate, rather strongly convex dorsal shield, suggesting perhaps more active swimming habits than are likely in the species of Protaspis with a broad, fiat shield. Protaspis sp. indet. A number of fragmentary remains from the Water Canyon For- mation suggest the presence of additional species, although in those that lack the branchial openings, reference to Protaspis cannot be certain. Forms with fine ornament include PF 335, with moderately broad branchial plates; PF 874, with a short, broad, bluntly rounded rostrum; and PF 904, with a relatively long, slender rostrum with- out the dentine ridge angulation that characterizes P. priscillae. Two specimens show only the postero-lateral part of the dorsal shield and exhibit differences in detail from the structure of this region in P. dorfi and P. cf. bucheri. PF 869 (fig. 73) does not pre- serve any scale-like posterior cornual process; it may not have ex- isted in this form, for the branchial plate projects some distance posterior to the branchial opening and lacks the slender posterior process that in P. cf. hucheri and P. dorfi, lies on the medial side of the posterior cornual process. Otherwise the structure of this re- gion is typical of the genus, as described above (p. 320). PF 878 also has the branchial plate extending well posterior to the branchial opening, but its posterior end is concave and possesses a very small median process, suggesting that a posterior cornual process may have existed. A number of ventral shields in varying degrees of completeness have been collected, largely from localities A and G, and are pre- sumed to belong to Protaspis. Compared to the dorsal shields, they present relatively few characters that can be used for specific deter- mination, the most important being proportions and coarseness of ornamentation. The available specimens all have moderate to coarse 350 FIELDIANA: GEOLOGY, VOLUME 11 ornament (37-62 ridges per cm.) and thus probably cannot belong to P. dorfi, P. tenuistriatus, or P. priscillae. Some of them are relatively broad and may well belong to P. bucheri, but the avail- able material is not adequate to attempt any specific determinations. Protaspis sp., representing an undescribed subgenus: Seven fragmentary specimens, from localities A, D, and G, indi- cate the presence in the Water Canyon Formation of what must represent by current standards a new subgenus of Protaspis. Un- fortunately, this material is too poorly preserved to provide an adequate definition, either specific or subgeneric. The most charac- teristic feature of this form is its ornamentation (fig. 84, A), which differs from that of any previously described heterostracian. The basic ornament pattern of pteraspids is present, consisting of ridges that are concentric on the dorsal and ventral discs and pineal plate, transverse on the rostrum, and diagonal on the branchial plates. These ridges are very coarse (about 20 per cm. in the midline of the dorsal disc) and are subdivided by cross-grooves so that the actual dentine ridges are short lengths at right angles to the usual directions. These short ridges are flat-topped or slightly rounded, and number about 50-55 per cm. in the midline and about 65 per cm. laterally on the dorsal disc. This type of ornament represents a simple modification of the usual pteraspid condition, and one that occurs regularly on the scales of the posterior part of the body of both Pteraspis and Protaspis. It differs from the papillate subdivision of the ridges characterizing the shield in the subgenus Cyrtaspidich- thys, and also differs from the ornament of Doryaspis (White, 1935, pi. 25, fig. 102), in which the secondary cross-ridges distinctly radiate from the center of growth on the dorsal disc. Two fragmentary specimens (PF 953 and PF 954) indicate that the dorsal shield was exceedingly broad and flat. In PF 953, the length of the dorsal disc anterior to the growth center is estimated to be 100 mm., and its maximum width is estimated as 142 mm. This is larger and relatively broader than in any described species of Cyrtaspidichthys. A large part of one branchial plate is preserved in this specimen; it is moderately broad and clearly demonstrates that the branchial opening had a posterior position as in all Protaspis. A fragment of a rostral plate (PF 952) and two incomplete ventral shields (PF 345 and PF 955) offer no distinctive characteristics be- sides the ornament. In thin-section, the exoskeleton structure is typically pteraspid. DENISON: EARLY DEVONIAN FISHES 351 HETEROSTRACI OF UNDETERMINED FAMILY A specimen from locality C in the Water Canyon Formation (PF 945) is clearly heterostracian but is not certainly referable to any of the recognized families. It consists in part of an incomplete ventral shield (fig. 85, B), ovoid in shape and truncate anteriorly. Laterally, it is bounded by upturned laminae, which are not present A B Fig. 84. A, Protaspis sp., representing undescribed subgenus; ornament of shield, PF 951 (X 8). B, heterostracian of undetermined family; ornament of ventral shield, showing pores of sensory canal system, PF 945 (X 8). in the ventral shields of other Heterostraci. The ornament (fig. 84, B) consists of very fine, denticulate ridges, 110 to 120 per cm. laterally. The general pattern of the ridges is concentric, but in detail they are wavy, with those between one set of growth lines often at a considerable angle to those between adjacent growth lines. The lateral sensory line canals are indicated by relatively large, distinct pores (fig, 84, B). Associated with the ventral shield is the postero-lateral part of the dorsal shield (fig. 85, A). At first glance, this appears to be typically pteraspid in structure, with a postero-median projection, concave postero-lateral edges, and laterally a rim resembling a bran- chial plate. But closer examination reveals that the postero-median projection is a process of the dorsal shield rather than a separate dorsal spine; it also shows that the lateral rim is not a distinct branchial plate, and lacks a branchial opening externally and a branchial tube internally. The ornament on this fragment is similar 352 FIELDIANA: GEOLOGY, VOLUME 11 C Fig. 85. Heterostracian of undetermined family. A, B, PF 945 (X 1): A, left postero-lateral part of dorsal shield; B, incomplete ventral shield. C, fragment of branchial plate with branchial opening, PF 946 (X 1,^2)- to that of the ventral shield except that it is less irregular and some- what finer. The position of the branchial opening in this form is indicated by another specimen (PF 946) from locality D (fig. 85, C). This is a fragmentary plate, externally covered with the same type of orna- ment as on PF 945 and characterized particularly by an elongate opening that pierces the plate obliquely. There can be no question that this is the branchial opening, and that this plate represents a branchial plate. It can be compared only to the so-called "branchial- cornual" plates of Traquairaspis (White, 1946, figs. 20-23), the only previously known heterostracian in which the branchial opening lies within the branchial plate rather than on its margin. DENISON: EARLY DEVONIAN FISHES 353 Until better material is discovered, the reference of this form to family must remain uncertain. It resembles the Traquairaspidae in the position of the branchial opening and in having the dorsal spine formed by part of the dorsal shield. However, it differs from Tra- quairaspis in possessing upturned lateral edges on the ventral shield, in lacking any differentiated marginal areas on the dorsal and ven- tral shields, and in the type of ornament (stellate, elongated tubercles in Traquairaspis). REFERENCES Branson, E. B., and Mehl, M. G. 1931. Fishes of the Jefferson Formation of Utah. Jour. Geol., 39, pp. 509-531, figs. 1-2, pis. 1-3. Brotzen, F. 1936. Beitrage zur Vertebratenfauna des westpodolischen Silurs und Devons. I. Protaspis arnelli n. sp. und Brachipteraspis n. gen. latissima Zych. Ark. Zool., 28A, nr. 22, pp. 1-52, figs. 1-17, pis. 1-10. Bryant, W. L. 1926. On the structure of Palaeaspis and on the occurrence in the United States of fossil fishes belonging to the family Pteraspidae. Proc. Amer. Phil. Soc, 65, pp. 256-271, figs. 1-3, pis. 1-4. 1932. Lower Devonian fishes of Bear Tooth Butte, Wyoming. Proc. Amer. Phil. Soc, 71, pp. 225-254, figs. 1-6, pis. 1-10. 1933. The fish fauna of Beartooth Butte, Wyoming. Part I. Proc. Amer. Phil. Soc, 72, pp. 285-314, figs. 1-8, pis. 1-21. 1934. The fish fauna of Beartooth Butte, Wyoming. Parts II and III. Proc. Amer. Phil. Soc, 73, pp. 127-162, figs. 1-8, pis. 1-26. 1935. Cniptaspis and other Lower Devonian fos.sil fishes from Beartooth Butte, Wyoming. Proc. Amer. Phil. Soc, 75, pp. 111-128, figs. 1-2, pis. 1-18. Denison, R. H. 1952. Early Devonian fishes from Utah. Part I. Osteostraci. Fieldiana: Geology, 11, no. 6, pp. 265-287, figs. 50-60. Flower, R. H., and Wayland-Smith, R. 1952. Cyathaspid fishes from the Vernon Shale of New York. Bull. Mus. Comp. Zool., 107, pp. 355-387, figs. 1-2, pis. 1-8. Gross, W. 1933a. Die Fische des baltischen Devons. Palaeontographica, 79, Abt. A, pp. 1-74, figs. 1-34, pis. 1-6, 1 map. 1933b. Die unterdevonischen Fische und Gigantostraken von Overath. Abh. Preuss. Geol. Landesan.st., (N.F.), 145, pp. 41-77, figs. 1-16, pis. 1-7. Holmgren, N. 1942. General morphology of the lateral sensory line system of the head in fish. K. Svenska Vetenskapsakad. Handl., (3), 20, pp. 1-46, figs. 1-27. Kiaer, J. 1928. The structure of the mouth of the oldest known vertebrates, ptera.spids and cephalaspids. Palaeobiologica, 1, pp. 117-134, figs. 1-7, pis. 12-13. 1932. The Downtonian and Devonian vertebrates of Spitsbergen. IV. Sub- order Cyathaspida. Skr. Svalbard Ishavet, 52, pp. 1-26, figs. 1-12, pis. 1-11. 354 FIELDIANA: GEOLOGY, VOLUME 11 KiAER, J., and Heintz, A. 1935. The Downtonian and Devonian vertebrates of Spitsbergen. V. Suborder Cyathaspida. Part L Tribe Poraspidei, Family Poraspidae Kiaer. Skr. Svalbard Ishavet, 40, pp. 1-138, figs. 1-57, pis. 1-40. Leriche, M. 1926. Deuxieme note complementaire sur le "Pteraspis" de Wiheries (P. dunensis F. Roemer). Bull. Soc. Beige Geol. Pal. Hydrol., 35, pp. 19-29, figs. 1-3, pis. 1-4. Obrutchev, D. V. 1941. Remains of Aspidosteus gen. nov. (Heterostraci) from the Upper De- vonian of River Lovat. Trav. Inst. Pal., Acad. Sci. U. R. S. S., 8, no. 4, pp. 7-22, 1 fig., pis. 1-2. (Russian with English summary.) Save-Soderbergh, G. 1941. Notes on the dermal bones of the head in Osteolepis macrolepidotus Ag. and the interpretation of the lateral line system in certain primitive verte- brates. Zool. Bidrag Uppsala, 20, pp. 523-541, figs. 1-9, pis. 1-3. Stensio, E. a. 1926. On the sensory canals of Pteraspis and Palaeaspis. Ark. Zool., 18A, no. 19, pp. 1-14, figs. 1-7. Strand, E. 1942. Nomenklatorische Bemerkungen iiber einige fossile Fische. Folia Zool. Hydrobiol., 11, pp. 383-386. White, E. L 1935. The ostracoderm Pteraspis Kner and the relationships of the agnathous vertebrates. Phil. Trans. Roy. Soc. London, (B), 225, pp. 381-457, figs. 1-97, pis. 25-27. 1938. New pteraspids from South Wales. Quart. Jour. Geol. Soc. London, 94, pp. 85-115, figs. 1-26. 1946. The genus Phialaspis and the "Psammosteus Limestones." Quart. Jour. Geol. Soc. London, 101, pp. 207-242, figs. 1-60, pis. 12-13. 1950. The vertebrate faunas of the Lower Old Red Sandstone of the Welsh borders. Bull. Brit. Mus. (Nat. Hist.), Geol., 1, pp. 51-67, figs. 1-2. 1952. Pteraspis leathensis White, a Dittonian zone-fossil. Bull. Brit. Mus. (Nat. Hist.), Geol., 1, pp. 69-89, figs. 1-25, pi. 5. White, E. L, and Moy-Thomas, J. A. 1940. Notes on the nomenclature of fossil fishes. Part I. Homonyms A-C. Ann. Mag. Nat. Hist., (11), 5, pp. 502-507. Whitley, G. P. 1940. The Nomenclator Zoologicus and some new fish names. Australian Nat., 10, pp. 241-243. Williams, J. S. 1948. Geology of the Paleozoic rocks, Logan quadrangle, Utah. Bull. Geol. Soc. Amer., 59, pp. 1121-1164, figs. 1-2, pis. 1-6. Wills, L. J. 1935. Rare and new ostracoderm fishes from the Downtonian of Shropshire. Trans. Roy. Soc. Edinburgh, 58, pp. 427-447, figs. 1-4, pis. 1-7. Woodward, A. S. 1934. Note on a new cyathaspidian fish from the Upper Downtonian rocks of Corvedale. Quart. Jour. Geol. Soc. London, 90, pp. 566-570, pi. 19. DENISON: EARLY DEVONIAN FISHES 355 Zych, W. 1931. Fauna ryb Dewonu i Downtonu Podola. Pteraspidomorphi : Hetero- straci. Cz^sc, lA, Lwow, pp. 1-91, 52 drawings, 31 photographs. (In Polish.)