Proceedings ofthe United StatesNational MuseumSMITHSONIAN INSTITUTION ? WASHINGTON, D.C.
Volume 124 1968 Number 3647The Suborders of Perciform Fishes
By William A. Gosline 1Senior Post-Doctoral Fellow, Division of Fishes
IntroductionThe basic concept and limits of the order Perciformes (Percomorphi)as defined by Regan (in various papers but especially 1929) seem to meto be the best yet proposed. Patterson (1964) has presented the viewthat the Perciformes are polyphyletic. In the same broad sense thatmammals are polyphyletic (cf. Simpson, 1959) this may well be, butthe particular lines of polyphyletic perciform derivation drawn byPatterson (1964) seem highly unconvincing (Gosline, 1966b). Stillmore recently, Greenwood, et al. (1966), have removed some of theforms here included in the perciform fishes to the separate superordersAtherinomorpha and Paracanthopterygii. This action, which seemsto me to involve a confusion between convergence and inheritance, is inmy opinion untenable (see below). Various people, including Regan(1936) and the present author (1962), have tinkered with the boundarylines established by Regan (1929) for the Perciformes. Of such authors,Berg (1940) made the most drastic changes. The question of whetherto include certain groups in or exclude them from the Perciformes iscertainly moot. Here, aside from the exclusion of the callionymoidfishes, I follow the old perciform boundaries of Regan (1929).
1 Department of Zoology, University of Hawaii, Honolulu 96822.
2 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124The present paper is addressed to the problem of how best to arrangeand classify the fishes that make up the order Perciformes. Attentionhas been focused on the subordinal and superfamilial levels. Familieshave been considered only insofar as they have been misplaced orindicate what fishes are included in a suborder or superfamily. Suchformal family classifications as have been included are not original,and the sources from which they have been adopted are stated.It has, of course, been possible to examine only a small proportionof the thousands of species included in the Perciformes. Selection ofmaterial for investigation has been made on two bases. The greatestamount of time has been spent on the most controversial groups,notably the Blennioidei. Within a group the morphologically general-ized members have been investigated.Names used throughout this paper are conventional. In no instancehas an effort been made to solve nomenclatorial problems with regardeither to bone or fish names.Acknowledgments.?Almost all of the work on which this paper isbased has been done during tenure of a Smithsonian Research Associate-ship. For space and facilities in the Fish Division of the U.S. NationalMuseum during the year 1965-1966, I am greatly obligated to thestaff of that Division, especially to its Curator, Dr. E. A. Lachner.The majority of the material investigated is in the U.S. NationalMuseum. I would like also to thank Drs. D. M. Cohen and D. W.Strasburg of the U.S. Fish and Wildlife Service and Drs. J. Bohlkeand J. C. Tyler of the Academy of Natural Sciences of Philadelphia(ANSP) for the loan of specimens. Though I have benefited greatlyfrom discussion with all of the ichthyologists in the U.S. NationalMuseum, I would like specifically to acknowledge the help of Dr. D.M. Cohen with the ophidioids, of Drs. B. B. Collette and R. H. Gibbswith the scombroids, and of Dr. V. M. Springer with the blennioids,all of whom have been kind enough to read one draft or another of thesection on the groups mentioned.The original manuscript of this paper, submitted in December 1966,was revised and brought up to date in August 1967. Both drafts havebeen typed by my wife, whose assistance gratefully is acknowledged.Material ExaminedUnless otherwise noted, all material investigated forms part of theU.S. National Museum fish collections. Specimens that were examinedmerely for superficial characters will not be listed. Other material fallsinto four categories: a very few of the specimens were cleared andstained by the trypsin method developed at the USNM by Dr. W. R.Taylor; a number of forms were X-rayed through the courtesy of theUSNM Fish Division; some of the skeletons in the skeleton collection
no. 3647 PERCIFORM FISHES?GOSLINE 3
of the Fish Division were utilized; and the majority of the materiallisted consists of single preserved specimens, one side of which hasbeen dissected more or less completely with or without alizarin staining.Aside from a few specimens that disintegrated during staining, thespecimens, along with their dissected parts, are now back in thebottles from which they came.Names of species are those on the USNM bottles, except in one ortwo instances wherein the generic name obviously was incorrect.Anabantoidei.?Specimens of Ophicephalus species (148517),Anabas testudineus (102556), and Osphronemus goramy (12876).Specimens of Nandus marmoratus (44785) and Pristolepis fasciatus(107835) were stained and dissected.One specimen of Luciocephalus pulcher (35737) was X-rayed.Among comparative material, one stained and cleared (17428) andone stained and dissected (8568) specimen of Centrogenys marmoratusand one partially dissected Toxotes jaculatrix (174913) were examined.Acanthuroidei.?Partially dissected specimens of Teuthis oramin(195521), Zanclus canescens (82945), and Prionurus sculprum (3882).Scombroidei.?The stained and partially dissected specimen,about four inches long, upon which the account of Scombrolabrax isbased was loaned to me by Dr. D. W. Strasburg. The original de-scription was checked subsequently against a series of S. heterolepis(USNM 187651), one of which was stained.A whole series of tuna and mackerel skeletons in the collections ofthe USNM and the University of Hawaii was examined for the pinealorgan.Ophidioidei.?One X-rayed specimen of Gadopsis marmoratus(ANSP 81566) kindly loaned by the Philadelphia Academy of NaturalSciences. One stained and partly dissected specimen of the samespecies (48813).Two stained and partially dissected specimens of Neobythites gilli(200553) and one oiDicrolene intronigra (200554).One partly dissected specimen of Brotula barbata (131279). Onestained and partly dissected Lepophidium negropinna (197144).Among comparative material, one stained and partly dissectedPhycis regius (190434) and one "Macruridae" (158664) were examined.Blennioidei.?Parapercidae : one partly dissected Prolatilusjugularis (176470) and an X-ray of the same species (77365) ; X-raysof Mugiloides chilensis (114930), Pinguipes brasiliensis (83241), andParapercis allporti (179797); partly dissected P. cephalopunctata(1430785).Trichonotidae (sensu lato) : one partly dissected Hemerocoetes species?(177085); one skeleton (26335) and one stained and partly dissectedspecimen of Bembrops gobioides (158132).
4 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124Cheimarrichthyidae: one stained and partly dissected Cheimar-richthys josteri (198510).Bovictidae: one partly dissected Cottoperca gobio (114925).Nototheniidae: one stained and partly dissected Trematomuspennellii (179676) and one partly dissected Eleginops maclovina(77319).Harpagiferidae: one stained and partly dissected Harpagifer bispinis(77282).Trachinidae: partly dissected Trachinus draco (31064), T. vipera(39473), and T. radiatus (2213).Uranoscopidae : one partly dissected Uranoscopus japonicus (122508)Dactyloscopidae: one slightly dissected Dactyloscopus crossotus(114411).Leptoscopidae: one slightly dissected Leptoscopus angusticeps(39684).Congrogadidae: one stained and partly dissected Congrogadussubducens (173805).Notograptidae: one stained and partly dissected Notograptusguttatus (173798).Tripterygiidae: one stained and partly dissected Enneapterygiusetheostoma (71528).Clinidae: one stained and partly dissected Labrisomus nuchipinnis(uncataloged) ; one partly dissected specimen of Clinus superciliosus(93637).Blenniidae: one partly dissected Blennius cristaius (185376); onestained and partly dissected specimen of Runula tapeinosoma (195704).Bathymasteridae: one skeleton (26230) and one partly dissectedBathymaster signatus (111994); one partly dissected Ronguilus jordani(103689).Anarhichadidae: a partial skeleton of Anarhichas lupus (110814).Cryptacanthodidae: one skeleton of Cryptacanthodes maculatus(26512).Zoarcidae: one stained and partly dissected Lycodes species?(177654); one partly dissected Zoarces viviparus (10065); a partialskeleton of Z. anguillarus (26498).A good deal of additional material, not included in the Blennioidei,was used in delimiting it.The Basis of Perciform ClassificationThe Perciformes are the largest order of modern fishes. The classi-fication, like that of fishes in general, has evolved piecemeal over theyears; nevertheless, out of the efforts of such ichthyologists as Jordanand Regan, the classification of the Perciformes (and of the higherteleostean fish orders) has developed a largely unstated but nonethe-
no. 3647 PERCIFORM FISHES?GOSLINE 5less real structural coherence. This basic structure is accepted here,and such changes in subordinal status as are suggested have beenmade with the idea of strengthening rather than altering it. A briefaccount of the basis of perciform classification may help to explain this.The basal percoid fishes represent the greatest focal point of fishevolution that exists today. Some 50 families of these with thousandsof species generally are recognized, and they dominate all of thericher marine fish faunas. The families are differentiated on relativelyslight bases but to require any other would result in one tremendous,taxanomically meaningless, and unmanageable family. As it is, theSerranidae (sensu lato) has been tending in that direction (Gosline,1966a).It is assumed that from the basal percoids an adaptive radiationhas taken place. Some of the lines of development have differentiatedvery little, in which case they are still included with the basic stock;others, considered separate superfamilies, somewhat more; separatesuborders, more still; and derivative orders, most of all. Thequestion which fish belongs in which taxon and why constitutes thesubject of perciform classification. Some of the theoretical and prac-tical problems will be discussed briefly here.The basic difficulty is the old one of vertical vs. horizontal classi-fications. Stated briefly: if, in figure la, the lineages, represented by
Figure 1.?Diagrammatic representation of perciform radiation: a, hypothetical (see p. 6for lettering); b, with actual suborders included. (At right of broken line in b are thoseforms with dorsal and anal soft rays showing exact 1:1 correspondence with vertebrae;to left of dotted line forms have about 2+ dorsal and anal rays per vertebra; betweendotted and broken lines normal ratio of 1 + ray per vertebra is maintained.)
6 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124the radiating lines, are traced back to their bases, in this case intothe basal mass of percoid families, then how does one distinguishthem? Contrariwise, if a line XY, representing some theoretical stageof structural development, is drawn across the radiating lines andeverything below XY is called a suborder Percoidei, then how doesone classify the parts of the radiating lines above XY?Omitting from present consideration the mugiloids and anabantoids,Regan (1913), followed herein, places all of the Perciformes below atheoretical line XYin the suborder Percoidei. Matsubara (1955, 1963),following the lead of Jordan and others, adopts what is probably amore consistent approach and divides the areas both above and belowXY into separate divisions; e.g., the Percina, Chaetodontina, Caran-gina. As far as I can determine, there are no concrete morphologicalcriteria for the separation of the more basal groups, and a decision asto which of the basal percoid families should be assigned to whichsection has to be made on a largely intuitive or authoritarian basis.Furthermore, I feel no intuitive assurance that such a group as theChaetodontina is not an assemblage of similar-looking but unrelatedfishes. It may well be that when other and sharper tools are devisedfor investigating the relationships of percoid families (see, e.g., Frei-hofer, 1963) elimination of the line XY and the basal suborder Per-coidei will prove feasible. For the moment, however, recognition of acentral group Percoidei seems preferable.Such a recognition, as already noted, causes difficulties in thetreatment of the percoid-derivative taxa. If all of the radiating linesbelow XY (for example c, d, and e) are considered to belong to thesingle suborder Percoidei, then should all the individual lines aboveXY, however close (for example a and b), be considered separatesuborders? Regan (1929; seems to have adopted essentially thiscourse in recognizing the Siganoidea (Teuthidoidea) as distinct fromthe related Acanthuroidea, the Scombroidea distinct from the Tri-chiuroidea, etc. In this, I do not follow him. In the first place, 1 cansee no compelling logic in the procedure. In the second, it has thepractical result of creating a tremendous basal suborder Percoideiwith numerous splinter offshoot suborders. Here, the concept of aderivative percoid suborder is that it should contain fishes moreclosely related to one another than to any other fishes outside theboundary of the suborder Percoidei. This concept admits the possi-bility that a derivative suborder may have been polyphyletic at thetime it crossed the fine XY. In practice (fig. 16) it has the effect ofcombining certain of Regan's (1929) suborders.Another problem of perciform classification is that of determiningwhich lineages should be recognized as derivative suborders ratherthan as full orders. Many factors have a bearing on this question.
no. 3647 PERCIFORM FISHES?GOSLINE 7One is logical consistency. The callionymoids are a case in point.I believe that the callionymoids, like the gobiesocids, are notothe-nioid derivatives. Thus, unlike the other suborders recognized herein,the callionymoids would seem to be derivatives of derivatives of thepercoids. To be consistent, therefore, they should not be placed inparallel with the other suborders recognized here but either shouldbe included in the notothenioids or be removed from the Perciformesentirely. Of these alternatives, I prefer the latter. From the pointof view of classification, the callionymoids then would have a positionanalagous to that of the Tetraodontiformes (which seem to havearisen from the percoid suborder Acanthuroidei)
.
In general, recognition of a group as a separate superfamily, sub-order, or order is based on degree of morphological differentiation,precedent, and the size of the group under consideration. As to thelast factor, the generally accepted dictum "that the size of the gap[between units] be in inverse relation to the size of the unit" (Mayr,1943, p. 139) has been adopted. Thus, the large group Scorpaeni-formes is considered herein a separate order from the Perciformes,though the known differences between the two units are not great(cf. Berg, 1940; Matsubara, 1953). Conversely, though the abovedictum militates against small units, the complete elimination ofcertain small perciform suborders does not appear feasible at thepresent time. Thus, to combine the Kurtoidei, containing but a singlegenus, with any other perciform suborder would seem to abrogatephylogenetic principles. The same is true of the Schindlerioidei.Again, I have come to the somewhat reluctant conclusion that theIstiophoridae, Xiphiidae, and Luvaridae bear no real relationshipto the scombrid fishes and must, at least provisionally, be placedin a separate suborder by themselves (see p. 28).Finally, there arises the question of how to draw the line XYin figure la. One could draw such a line with a view to creating adefinable basal suborder Percoidei. This would leave bits and piecesof radiating lineages outside the line XY to be tucked away in onesuborder or another as decorously as possible. In practice, the lineXY has been drawn with an eye to creating coherent derivativesuborders. In figure la, therefore, XY should have been drawn asa zigzag line, dipping more or less deeply into the basal Percoideiat different points. In practice, then, the Percoidei contains all thoseperciform fishes that do not belong to some other suborder. ThePercoidei presumably contain related fishes, but defining it morpho-logically in positive terms is difficult.With regard to the derivative suborders, as knowledge increases,more and more structurally transitional forms between these and thebasal Percoidei become known. Thus, to a greater or lesser extent,
8 PROCEEDINGS OF THE NATIONAL MUSEUM vol. vuScombrolabrax (see p. 33) closes the structural gap between the per-coids and the scombroids, Gadopsis (see p. 26) that between thepercoids and the ophidioids, and a new family for which only a pro-visional notice has so far been given (Haedrich, 1967b) is stated to beintermediate between the percoids and the stromateoids. With suchgaps being filled in, the separation of perciform suborders into neat,precisely definable pigeon holes becomes increasingly impossible.The classification of the Perciformes to suborder adopted here isas follows: Order PerciformesSuborder Mugiloidei" Anabantoidci" Percoidei" Kurtoidei" Acanthuroidei" Ophidioidei" Stromateoidei" Xiphioidei" Scombroidei" Gobioidei" Blennioidei" Schindlerioidei
"Protopercoid" SubordersThough the great majority of modern perciform fishes belong toto the basal suborder Percoidei and its derivatives, there are twogroups that at least may have developed from a "protopercoid"stock, namely the Mugiloidei and Anabantoidei.The main, and only significant reason for considering this possi-bility is that the Mugiloidei always and the Anabantoidei often lacka direct articulation between the pelvic bones and the cleithra. Thiscondition suggests the subabdominal pelvic position of prepercoidorders. Various interpretations are possible, however, and I am notsure which one is correct. First, as already suggested, the Mugiloideiand/or Anabantoidei may have evolved from a protopercoid stock inwhich a direct connection between the pelvics and cleithra had notyet developed. A variant of this hypothesis, again postulating aprotopercoid ancestry for the Mugiloidei and/or Anabantoidei, wouldbe that in the protopercoids, as in the berycoids, the pelvic-cleithralrelationship remained variable, a more or less fixed articulation be-tween the two elements only becoming established at the percoid stageof development. Under this thesis, the Mugiloidei would represent thenonarticulated aspect of protopercoid inheritance, whereas in theAnabantoidei the whole gamut of protopercoid pelvic variation stillwould be represented. Conversely, it may be, as Dollo (1909) hassuggested, that the lack of a pelvic-cleithral articulation in the
no. 3647 PERCIFORM FISHES?GOSLINE 9Mugiloidei and in some of the Anabantoidei represents a secondary-loss; certainly such a loss has occurred in such other percoid deriva-tives as the Stromateidae, Tetragonuridae, Gempylidae, and Trichiu-ridae (Regan, 1909a).Because of the possibility that the Mugiloidei and Anabantoideidiverged from a protopercoid stock somewhat ahead of the otherexisting Perciformes, they will be dealt with first. Whether these twosuborders, however, are considered as "protopercoid" (fig. lb) orpercoid derivatives is of no great moment for overall Perciformesclassification. Suborder MugiloideiThe suborder Mugiloidei, as understood herein, contains thePolynemidae, Mugilidae, Sphyraenidae, Atherinidae, and phallo-stethoid families. Rosen (1964; and in Greenwood, et al., 1966) re-cently has removed the Atherinidae and phallostethoid families to aseparate order Atheriniformes of the superorder Atherinomorpha.This order and superorder I believe to comprise three unrelatedgroups?the exocoetoids, the cyprinodontoids, and the atherinoids
?
all of which are adapted basically to living at or very close to thewater surface and, consequently, have developed numerous featuresin common. The question of an atherinid-cyprinodontoid relationshiphas been discussed widely in recent years (e.g., Hubbs, 1944; Rosen,1964; Greenwood, et al., 1966; and Foster, 1967). I have nothing toadd to or subtract from what I have said already on the subject(1961b, 1962, 1963). Alexander (1967) recently has discussed the jawstructure of the two groups.In an earlier paper (Gosline, 1962), I advocated the exclusion ofthe Mugiloidei from the Perciformes as a separate order, largelybecause of the consistent lack of a direct articulation between thepelvic girdle and the cleithra. At that time, I was unaware of the wholerange of variation in this characteristic that occurs in the Anaban-toidei. Because of the doubt thrown on the character of the pelvic-pectoral articulation by the anabantoids, as well as on other grounds(Freihofer, 1963), it seems advisable to return the mugiloid fishes tothe Order Perciformes.Suborder AnabantoideiThe suborder Anabantoidei, as recognized herein contains theOphicephaliformes and Anabantoidei of Berg (1940) and Liem (1963),and the Luciocephalidae (Liem, 1967). The morphological divergenceamong these three groups is not contested. It seems to me, howeveras it did to Regan (1909b), that they are related more closely to oneanother than to any other fishes. They hold in common three morpho-logical features that are highly peculiar among acanthopteran fishes:
IQ PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124
a suprabranchial air-breathing organ, a gas bladder that extendsposteriorly well behind the body cavity, and teeth usually presenton the parasphenoid. With regard to the last feature, Liem (1967, p.108) describes the parasphenoid of Luciocephalus as toothless, butaccording to Regan (1909b, p. 768) there are "two or three minuteteeth on the parasphenoid." It may be that in the Luciocephalidaethe presence of parasphenoid teeth is a variable feature, as indeed itis among the ophiocephalids and anabantids (sensu lato). Additionalsuggestions of a relationship among the three groups are their fresh-water, Old World distribution, centering in southeast Asia, and theirnest-building and/or oral-incubating proclivities. It seems most un-likely that all these features are the result of convergent evolutionfrom independent origins.As already noted, the pelvic girdle of some of the Anabantoidei isremote from the cleithra (Ophicephalus=Channa, Anabas); in others,it articulates directly with the cleithra in typical percoid fashion(Betta, Colisa, Trichogaster) . Furthermore, in Ophicephalus the pelvicfin consists of six segmented rays. If the outermost pelvic rays ofOphicephalus represent the usual percoid pelvic spines transformedback into soft rays, such a secondary regression is only representedelsewhere, to my knowledge, among the Fleuronectiformes (Hubbs, 1945).Among the anabantoids are found two seemingly atavistic charac-teristics. One, discussed at length by Liem (1967), is the presence ofa mental ossification that closely resembles the gular plate of elopoidand earlier fishes. My own belief is that the mental ossification ofLuciocephalus is not a true gular plate. The other characteristic is theparasphenoid teeth already mentioned. Aside from two other percoidfamilies (see below), teeth on the parasphenoid are not found in theTeleostei above the elopoids. Why they should reappear in the anaban-toids and two other percoid families I do not know, but again it seemsto me that a postulate of reappearance is preferable to one ofinheritance.In searching for possible anabantoid relatives, one is led naturallyto the two percoid families that also have parasphenoid teeth: theNandidae and Pristolepidae. The "bite" provided by the parasphenoiddentition of Pristolepis is quite different from that of Nandus (whichresembles that of Ophicephalus), just as that of Ophicephalus differsfrom the parasphenoid apparatus of the anabantids (sensu lato).Aside from the parasphenoid dentition, Nandus and Pristolepisappear to be rather normal percoids, lacking such specialized anaban-toid features as the accessory air-breathing organ and the backwardlyextended gas bladder. They do bear certain features, however, suggest-ing an anabantoid relationship. First, all of these fishes have an ex-panded auditory bulla on the cranium. Second, Ophicephalus (fig. 2a),
PERCIFORM FISHES?GOSLINE 11
Figure 2.?Caudal skeletons: a, Ophicephalus species; b. Pristolepis fascialus.(Ep= epural, Gb= gas bladder, Ha=hemal arch, Hy=hypural, Na= neural arch,Un= uroneural, Ur=urostyle.)
12 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124Anabas, and Pristolepis (fig. 26) hold in common certain peculiarities ofthe caudal skeleton. In all three, there is the full percoid complementof five hypurals (using Nybelin's [1963] system of counting) ; these areall subequal in width and splayed out like the spokes of a fan. There isonly one epural, and the last hemal arch is not in contact with theurostyle. (Judging from X-ray photographs [e.g., Liem, 1967, fig. 9],Luciocephalus seems to have a specialized version of the same basictype of caudal structure.) Finally, there is the fact that the Nandidaeand Pristolepidae, like the Anabantoidei, are freshwater fishes with adistribution center in southeast Asia.The Percoidei and Derivative SubordersThe suborder Percoidei comprises the central mass of the perciformfishes; its members dominate the richer marine fish faunas today,notably those of coral reefs.An ecological peculiarity that is at least worth noting is that manyof the percoid families that, on morphological grounds, seem tostand at the base of the suborder contain or comprise euryhalineand/or freshwater forms; e.g., the Centropomidae, Percichthy-idae, Kuhliidae, Centrarchidae, Percidae, Nandidae. The same istrue of the "prepercoid" families Mugilidae, Atherinidae, Phallos-tethidae, Ophicephalidae, Anabantidae, and Luciocephalidae.As compared with the presumably ancestral Beryciformes, thepercoids seem to differ in no one important character (Gosline, 1966b)
;
rather, judging by living forms, they appear to have integrated anumber of minor features in what amounts to an advance over theBeryciformes in general adaptiveness. Again judging from the observa-tion of living forms, the most satisfactory answer to the question ofwherein this advance lies seems to be in an increase in swimmingabilities in the percoids.On the other hand, if the suborders and orders derived from thepercoids are compared with the Percoidei, it becomes clear that each ofthese derived taxa has adopted some specialized mode of life; thus, ofderivative percoid suborders, the xiphioids, scombroids, and schindl-erioids have taken up an existence in the open sea, the gobioids andblennioids have adopted a life in direct contact with the bottom, theacanthuroids and stromateoids have developed specialized food habits,etc. But, again, most of these specializations have involved furtherchanges in methods of swimming and maneuvering. Indeed, thisaspect of existence runs so continuously through the evolution of thepercoids and their derivatives that it seems well to take it up by wayof an introduction to these groups.The adult percoids are mostly maneuverers living close enough tothe bottom to use it for protection but not maintaining direct physical
no. 3647 PERCIFORM FISHES?GOSLINE 13
contact with the bottom (at least during the day). Though the defen-sive armature of percoids is less extensive than that of most livingberycoids, the percoids seem to have provided the pelvic spines with afirmer base in the development of a direct pelvic-cleithral attachment.In bringing the pelvics forward under the pectorals, the percoids alsoseem to have increased their ability to maneuver. Harris (1938)showed that acanthopteran pectorals are so constructed as to give anupward thrust to the front of the fish when erected for the purpose ofturning or stopping and that erection of the pelvics at the same timeoffsets this. In this respect, the pelvics seem to counteract the pectoralsmore efficiently if they are directly below the pectorals rather thanbehind them, as they are in lower fishes and still, to some extent, inmost Beryciformes.For the paired fins to be effective in stopping or turning, a forwardspeed ("headway") must have been generated previously. This is usu-ally developed by the vertical fins and the body. Among the lowerpercoids, the forked caudal fin, a basal teleostean feature, plays alarge role. Gero (1952) has shown that, for a swimming fish, a forkedtail shape is the most efficient. From this basal type, found in such alower percoid as Roccus (=Morone), two divergent lines of develop-ment have occurred. One is carried to its extreme in the Scombridae.Here, the widely forked fin has a short, high, relatively stiff bladefirmly attached to the caudal skeleton at the end of a slender caudalpeduncle. This type provides great power and speed, but it has itslimitations. Harris (1953, pp. 26, 27) stated: "Tails of this type arefound in fishes which are fast continuous swimmers (scombroids) ; ifa sudden burst of speed from a standing start is required, the angle ofattack of this type of tail would be too high and the tail would 'stall'."At the opposite extreme is the rounded caudal that has been developedagain and again in percoids and their derivatives. Such a caudal shapenot only provides a better "getaway" mechanism but seems to be amore efficient (or perhaps accurate) propulsive force at slow speedsand in enclosed areas.Aside from caudal shape, there are other factors that affect theforward locomotion of the percoids and their derivatives (fig. 3).Thus, when a fish becomes either very deep-bodied or very elongate,the potentiality for rapid locomotion seems to be lost. At both ex-tremes, the importance of the caudal fin as a source of forward thrustdiminishes. Such a deep-bodied form as Chaetodon has a relativelylong posterior border to the body, covered by the soft dorsal and analfins, and a short, brushlike tail. In moving forward, it flaps the wholerear portion of the body, of which the tail is only an insignificant part.The end point in such a line of development is of course the tetraodon-tiform Mola, which has no caudal fin at all.
14 PROCEEDINGS OF THE NATIONAL MUSEUMElongate perciform fishes usually move forward by undulation, butthis may be by two very distinct methods. In one, the fish holds itsbody more or less rigid and undulates the dorsal and anal fins only.In these forms, the dorsal and anal soft fins tend to be long and thefin rays to be closely spaced (i.e., two or more per vertebra), insertedbasally on a sort of ball and socket axis, and with well-developedmusculature. Locomotion by means of fin undulation seems to provideprecision of movement rather than speed and enables the fish to movebackward or forward with almost equal ease. Such a method of loco-motion has been developed frequently among the lower teleosts; e.g.,gymnarchids, gymnotids, probably halosaurids and macrourids, andthe Syngnathiformes. It occurs, however, only in the ophidioids amongthe Perciformes (fig. 16), and in the Tetraodontiformes.
Flop thedorsal and analMOLIDAE
Deep-bodied Fishes
^ Undulate finsseparately fronthe body
__-? ~BR? VUWElongate Fishes
CEPOLIDAEUndulate body andvertical fins togetherFigure 3.?Diagram of certain types of forward motion in the perciform fishes and theirderivatives.
The other, more usual method of locomotion among elongate perci-form fishes and their derivatives is for the fish to undulate its body andfins together. Here, the vertical fins tend to coordinate their structureas well as movement with that of the body, the relationship betweensoft dorsal and anal fin rays and vertebrae becoming 1:1. Generally,also, the number of vertebrae in such fishes is increased over the basalpercoid number of 24 or 25. This development of a 1:1 ratio betweensoft fin rays and vertebrae in elongate perciform fishes occurs again andagain (fig. 16). Sometimes it occurs in free-swimming forms like theCepolidae or Schindleriidae, but more frequently it develops in bottom-resting forms.
no. 3847 PERCIFORM FISHES?GOSLINE 15Whereas the great majority of the lower percoids and, for that mat-ter, of lower teleosts, live constantly in midwater, i.e., off the bottom, agreat many adult percoid derivatives have taken up a life in directcontact with the bottom, making only short dashes to obtain food or toavoid enemies. Some of the various percoid derivatives that haveadopted this habit are the Blennioidei, Gobioidei, Scorpaeniformes,Pleuronectiformes, Gobiesociformes, and many Lophiiformes. The finrequirements of such forms are in many respects almost opposite tothose of a swimming fish. An account of them can be deferred best tothe section on the suborder Blennioidei (see p. 48).Of the suborders among the Percoidei and their presumed deriva-tives, there are some for which I can add little or nothing to existingknowledge. It seems well to deal with these first, leaving until lastthose suborders to which the major portion of the present investigationhas been devoted. Suborder PercoideiFor purposes of the present paper, the superfamily (division) classi-fication of Regan (1913, p. 112) will be accepted, except that hisGadopsiformes, Nototheniiformes, Callionymiformes, and most of hisTrachiniformes have been removed and, following Norman (1929), theChiasmodontoidae have been added. Here, Regan's Gadopsiformesare included in the Ophidioidei; the Nototheniiformes and most of theTrachiniformes have been placed in the Blennioidei; and the Calliony-miformes have been taken out of the order Perciformes. The onlyfamily of Trachiniformes retained in the suborder Percoidei is theOpistognathidae, and this seems to belong in the superfamily Percoi-dae, close to the Acanthoclinidae. The Trachiniformes of Regan, thus,is abolished. Suborder KurtoideiThis suborder consists of a single genus. The anatomy of this peculiarfish has been described by de Beaufort (1914). I can add only that, insix specimens of Kurtus indicus examined, five had 15 and one had 14branched caudal rays; de Beaufort and Chapman's (1951, p. 82) state-ment that the suborder has the "Caudal with 17 divided rays" seemsto be in error. Suborder SchindlerioideiThis is another perciform suborder based upon a single isolatedgenus. The fish is neotenic, but its peculiar caudal supporting structureseems to be unique among fishes of any stage of ontogenetic develop-ment (Gosline, 1959). The most recent of the varied suggestions con-cerning the relationships of Schindleria is that it might have evolvedfrom something near the ammodytoid Hypoptychus (Gosline, 1963).
16 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124Suborder StromateoideiNo examination has been made of any stromateoid by the presentauthor. A recent review of the group, however, has been providedby Haedrich (1967a). Suborder GobioideiCertain of the families formerly placed in the blennioids have beenmoved to the Gobioidei by me (Gosline, 1955), but I have nothingto add to that paper.Suborder AcanthuroideiThe zanclids, acanthurids, and siganids (teuthidids) herein areconsidered members of a single suborder. The relationships among thethree groups, to my knowledge, have not been disputed. The questionmerely is whether the siganids represent a sufficiently aberrant off-shoot of the acanthurid stock to warrant a separate suborder. Starks(1907) was in doubt about the matter. From the overall view ofperciform fishes taken in this paper it seems preferable to considerthe siganids as one of the two superfamilies in the suborderAcanthuroidei.The primary specialization of these fishes seems to be the develop-ment of a nipping type of jaw structure. Gregory's (1933, pp. 279-283) analysis of this structural complex and the relationships of thesefishes appears to me to be entirely correct. He raises what seemsto be the only important taxonomic question regarding the group;namely, whether or not it should be removed entirely to the Tetra-odontiformes, which it foreshadows.In this connection, the "prepalatine" bone (Starks, 1907, 1926)of the Siganidae (Teuthididae) warrants brief mention. In the sig-anids, as in the Tetraodontiformes, the upper jaw, instead of beingprotrusile as in most percoids, rocks in and out on the tip of the pala-tine as a fixed point. In the Tetraodontiformes, the whole palatinemay become attached rigidly to the cranium and remain free fromthe rest of the suspensorium. In the siganids, a somewhat differentsystem has been developed to accomplish the same end. The palatinebone has become divided into two parts, with the rear portionattached to the rest of the suspensorium as usual. The front portion,i.e., the "prepalatine" bone, however, has developed as a separateelement from the rest of the palatine and has developed a firm attach-ment to the inner surface of the expanded nasal bone above andof the lacrimal below. The nasal in turn has a rigid, sutured attach-ment on the front of the cranium.
no. 3G47 PERCIFORM FISHES?GOSLINE 17Suborder OphidioideiThe suborder Ophidioidei (treated as an order by Mead, Bertel-sen, and Cohen [1964, p. 580] without comment), as generallyunderstood, contains the fishes included in the families Brotulidae,Aphyonidae, Ophidiidae, Pyramodontidae, and Carapidae. To theseI add the family Gadopsidae for reasons dealt with below.The suborder may be defined as follows: pelvies, when present,of one or two filamentous rays on each side, originating ahead of thepectoral fins; dorsal and anal long, without spines except in Gadopsis,the rays more numerous than the vertebrae between them; one ormore of the first few ribs usually expanded.To the end of the last century, the ophidioids, along with thegadoids, blennioids, and other fishes with anterior pelvics, generallywere placed in an assemblage known as "Jugulares." In 1903b,Regan concluded (p. 460) "that the Blennioid fishes [in which Reganat that time included the ophidioids] are modified Acanthopterygii,but that the Gadoids have originated from some less specializedstock, and that the absence of non-articulated fin-rays, the largenumber of rays in the ventrals, and the lack of direct attachment ofthe pelvic bones to the clavicles, taken together must be regarded asprimitive characters." Between 1903 and 1966 (Greenwood, et al.,1966) this separation of the gadoids from the blennioids and ophi-diods generally has been accepted.In 1903b, as noted, and again in 1912d, Regan included the ophi-dioids in his perciform suborder Blennioidea. In 1929, however, hesegregated them as a separate perciform suborder "Ophidioidea."The later allocation appears to me to be correct.The clarity of the distinction between the percoids and the ophi-diods, however, is obscured considerably by the Australian genusGadopsis, a morphological intermediate usually placed among thepercoids but herein assigned to the ophidioids.In my opinion, the basic specializations of the ophidioid fishes liealong two probably interrelated lines. One involves locomotion andthe other sensory systems. The presumed nature of these will bediscussed before dealing with general characters.In the basal percoids (see p. 5), there are somewhat more fin raysthan vertebrae, but the relationship is indeterminate (Francois,1959). Gadopsis shows a fairly typical condition, with 28 soft dorsalrays whose pterygiophores extend downward over 25 neural spinesand with 18 soft anal rays under 14 hemal spines (in the X-rayedANSP specimen). The other ophidioids, instead of going the usualway of elongate percoid derivatives in developing an exact 1 : 1 re-lationship between soft dorsal and anal rays and vertebrae, have280-S35?6S 2
18 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124developed an approximately 2 : 1 ratio between rays and vertebrae(fig. 16).Specimens of Brotula multibarbata in the Honolulu aquarium,though they remained with the body curved and in contact with thesubstrate during the period I was able to observe them, continuallypassed undulations along the free portions of the dorsal and anal fins.Suggestions of similar fin undulations are found in the observationsof living brotulids by Whitley (1935) and Dawson (1966). This isnot to say that all brotulid locomotion is carried on by fin undulationalone, for all brotulids can doubtless undulate the body in coordinationwith the fins and probably do when greater speed is needed. Certainlysuch coordination occurs in ophidiids (Herald, 1953; and Briggs andand Caldwell, 1955) and carapids (Arnold, 1956).Phylogenetically, the argument regarding ophidioids herein ad-vanced is not that they all swim in a manner very different from,say, the zoarcids (which have a 1 : 1 fin ray to vertebra relationship),but that their capability for independent fin undulation has ledtoward a morphological endpoint contrary in direction to that atwhich the basal percoids (with about 1.1 or 1.2 fin rays per vertebra)almost have arrived, and in a direction that has been followed byrelatively few other percoid derivatives. Consequently, this develop-ment (of an approximately 2 : 1 fin-ray-to-vertebra ratio) in ophi-dioids appears to be systematically significant.With regard to the sensory peculiarities of the ophidioid fishes,it seems to me that these are basic and that most, if not all, of theother ophidioid specializations are secondary to and related to them.Because of this, certain structural complexes that are not in themselvesstrictly sensory will be included in the discussion here.Morphologically, one of the peculiarities common to all brotulids,ophidiids, Gadopsis, and certain gadids, e.g., Urophycis, is the develop-ment of the pelvic fin into one or two well-developed filamentsoriginating more or less far forward. Functionally, the pelvies ofbrotulids and ophidiids have not been studied beyond the few pre-liminary observations of Herald (1953) and Briggs and Caldwell(1955). The function of the Urophycis pelvies, however, has been thesubject of an excellent recent investigation by Bardach and Case(1965).With regard to behavior, Bardach and Case (1965, p. 198) wrote inpart:Fishes swimming along the bottom ordinarily direct their [pelvic] fins forward,with the branches spread apart to an angle of up to 45?, the entire fin sweepingfrom slightly forward of the snout back toward the flank (Fig. 5) [their figure].Each fin encompasses an arc of approximately 120? ahead and to the side of thefish. Upon touching a morsel of food with a fin tip, the fish often has to back upto veer down and ingest what it found.
no. 3647 PERCIFORM FISHES?GOSLINE 19Although, as just mentioned, observations on living brotulids andophidiids are only preliminary, there are two pieces of circumstantialevidence beyond gross pelvic morphology that suggest these fishesuse their pelvies as Urophycis does. One piece of evidence is that theophidiids, at the expense of considerable elongation of the cleithra,have brought their pelvics forward under the chin and, hence, nearerthe mouth. The other is that the brotulids and ophidiids, like Urophycisand gadoids in general, have developed a direct route of innervationfor taste perception in the pelvic fins. As Freihofer (1963, p. 141)has noted, in these fishes, the pelvic branch of the ramus lateralisaccessorius "passes anterior to the base of the pectoral fin and lateralto the cleithrum and the pectoral actinosts." This does not occur inGadopsis, which retains the inherited and, for fishes with anteriorpelvics, circuitous nerve "route of passing down the postcleithra andthen turning and coursing anteriorly en route to the distant pelvicfin" (loc. cit.).There is, I believe, a close relationship between the method oflocating food by means of pelvic filaments, as noted by Bardach andCase, and the jaw structure of brotulids, ophidiids, Urophycis, and,for that matter, polynemids (which presumably locate food by meansof pectoral filaments) . In all of these fishes, the food items are detectedunder the fish rather than ahead of it, and, in all, the mouth is inferior.In all also, such premaxillary protrusion as occurs extends the upperjaw vertically downward or even downward and slightly backward(rather than forward as in most percoids) ; the premaxillary pedicelis short and vertical, or it even extends up and somewhat forward.Finally, there is a peculiar development of a muscle to the maxillarythat Rosen (1964; and in Greenwood, et al., 1966) called a levatormaxillae superioris.In Merluccius, which differs from most gadoids in having a prog-nathous lower jaw, I can find no "levator maxillae superioris." Thatsome fishes with prognathous lower jaws, however, do have a muscleof this sort is clear from the batrachoid fishes (see Rosen, in Green-wood, et al., 1966). For a further account of this muscle in the cod,see Holmqvist (1910) and van Dobben (1935).With regard to senses other than that of taste in the ophidioids,morphological data suggest that the acustico-lateralis system is de-veloped highly, olfaction is normal, and the eyes are degenerate.In Gadopsis, as in other ophidioids, the lateralis system of the headlies in enlarged canals that, in the pterotic (fig. 4) and circumorbitalbones, are partially or completely open, bony troughs. There is alsoa large median opening (mucous or sensory pit) without a bony roofon the middorsal line between the two halves of the interorbitalcommissure.
20 PROCEEDINGS OF THE NATIONAL MUSEUMIn one respect, the lateralis system of Gadopsis is specialized con-siderably less than that of the ophidioids. In Gadopsis, as in mostpercoids, the ep axial body musculature extends forward over thedorsal surface of the skull and attaches in part to a low supraoccipitalcrest. The supratemporal commissure, as in most percoids and inthe gadoids, is incomplete; it extends upward on each side of the headthrough the lateral extrascapular and then ends blindly over theepaxial musculature noted above. In the other ophidioids, the epaxialbody musculature does not extend in over the skull; there is no supra-occipital crest rising above the cranial surface; and the supratemporalcommissure is complete. There appears to be, as in the northern andmany tropical blennies, a medial (as well as a lateral) extrascapularthat has become fused completely with the parietal bones.
me lo Po moFigure 4.?Cranium of Gadopsis marmoratus (ab= attachment surface for Baudelot'sligament, af= anterior facet for hyomandibular articulation, Ba= basioccipital, ca=cartilage, Ep= epiotic, Ex=exoccipital, Fr= frontal, ho= hyomandibular opening oftrigemino-facialis chamber, In= intercalar, me= membrane, Me=mesethmoid, mo=main opening of trigemino-facialis chamber, op= opening of supraorbital sensory canal,Pa= parasphenoid, pf= posterior facet for hyomandibular articulation, Pi= parietal,Pl= pleurosphenoid, Po=prootic, Pr=lateral ethmoid, Pt=pterotic, Sp= sphenotic,Su= supraoccipital, Vo= vomer).In Gadopsis, as in other ophidioids, the eyes are relatively small orcompletely absent. In all, the eyeball seems to be capable of slightrotation or none. The eye muscles are weak and usually flabby in thepreserved specimens, and there are no eye muscle canals (myodomes)
.
The eyeball is covered by a heavy membrane. In the ophidioids thisis taut over the eyeball, but in Gadopsis it appears to be infoldedaround the eyeball, perhaps permitting greater eye rotation.The relatively small eye and weak eyeball musculature are containedin a small eye socket. This I think is associated with certain featuresof the skull in the interorbital region and of the brain and olfactory
no. 3647 PERCIFORM FISHES?GOSLINE 21
nerve location. In this connection, I propose the working hypothesisthat degeneration of the eye and its musculature is followed in timeby the loss of the myodome and the basisphenoid and that a longitu-dinal trough bounded by membrane or bone and containing the an-terior portion of the brain eventually will extend forward between theorbits. Extreme examples of this sort of development are found par-ticularly in such small-eyed, broad-headed fishes as the salmonoidGalaxias, the gadoid Lota (Svetovidov, 1948), the zoarceoid Crypta-canthodes (Makushok, 1961a), and the ophidioid "Dinematichthys"(Gosline, 1953).In Gadopsis, as in ophidioids and numerous other fishes, thebasisphenoid is absent. The interorbital space has been encroachedupon from both the posterior and elsewhere. In Gadopsis the anteriorportion of the interorbital space is filled medianly in large part by acrest rising from the parasphenoid (fig. 4). Above and behind thiscrest is a V-shaped trough comprising a pair of membranes leadingupward and outward from the parasphenoid crest to attachmentson the lower surfaces of the frontals. At the posterior end of theorbital cavities in Gadopsis, the internal orbital bony walls are ex-tended anteromedially well beyond the trigemino-facialis opening(fig. 4).In ophidioids, as in the gadoids and other fishes, the anteromedialextension of the bony orbital rims is developed further. In Brotula,for example, lateral flanges from the parasphenoid meet the frontalsahead of the pleurosphenoid ("alisphenoid" of Regan, 1903b, p. 461,fig. 1a). The latter bone, now completely surrounded by otherossifications, seems to disappear completely in some brotulids.The olfactory organ of Gadopsis and ophidioids seems to be de-veloped normally. In the forms examined, the two well-separatednostrils on each side lead in over an elongate-oval rosette. In Gadopsis,the olfactory nerve to each rosette passes back through the lateralethmoid and, for a short distance, through the anterior end of theorbital cavity and alongside the parasphenoid crest. About one-thirdof the way back in the orbits, the olfactory nerves of each side passinto the membranous trough described above. They extend posteriorlyinto this trough to the olfactory lobes of the brain, which projectforward into the trough. (Unlike many gadoids, the olfactory bulbsof Gadopsis and ophidioids are at the front of the olfactory lobes ofthe brain; see Svetovidov, 1948, pp. 13-17.)In the otic system of Gadopsis and ophidioids, there is always amore or less enlarged auditory bulla. In the juvenile Gadopsis dissected(106 mm SL), the wall of the central portion of this enlarged bulla ismembranous (fig. 4), and the intercalar (opisthotic) has only aminute extension on it. In Brotula, the expansion of the bulla is
22 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124
relatively slight and almost entirely comprises the exoccipital andprootic. In Microbrotula, the expansion is greater but comprising thesame two bones. In another brotulid, "Dinematichihys" (see Gosline,1953), in Benthocometes robustus (see Bougis and Kuivo, 1954, fig. 17),and apparently in the carapid "Fierasjer acus" (see Emery, 1880),the intercalar forms a part of the bulla wall.One seems to be on fairly firm ground in associating auditorybulla expansion with some specialization in hearing though, to myknowledge, the exact nature of the association remains unknown.It is probably more controversial to attempt to relate the gas bladderpeculiarities of ophidioids with hearing; however, I agree with Marshall(1965, p.. 314) that there is such a relationship. In the ophidioidsexcept Gadopsis, there always appears to be ligamentous tissueextending between the anterior end of the gas bladder and the anteriorribs, one or more pairs of which are modified considerably (Regan,1903b; Arnold, 1956). In the ophidiids (Rose, 1961) and oviparousbrotulids (Marshall, 1965, p. 314 quoting Courtenay, in litt.), it hasbeen suggested that the ligaments to the forward end of the gasbladder are used in sound production.Something should be said at this point about the Carapidae andPyramodontidae. These families, most if not all the members of whichlive as inquilines in the cavities of invertebrates, generally are agreedto be related to the brotulids and ophidiids. Among the numerousfeatures probably associated with their mode of life, however, are theloss of the pelvic fins and the development of a more or less terminalmouth, often with enlarged teeth. The other systems dealt with aboveseem to be essentially the same as those in the brotulids andophidiids.To summarize briefly the sensory systems and related structures inthe ophidioids, these fishes seem to have become modified extensivelyin association with the development of filamentous pelvics that areused presumably as probes for finding food. Though various fisheshave developed similar probes from other structures, the gadoids,ophidioids, some anabantoids, and pegasids are, to my knowledge, theonly fishes that have developed filamentous pelvics of this type. Bycontrast, the hypertrophy of the acustico-lateralis system and thedegeneration of the eyes have occurred repeatedly, especially amongdeep-sea forms. (Whether these features are brotulid preadaptations toor have been developed in association with a deep-sea existence hasno bearing on the present argument.) The unique feature, presumablyassociated with the acustico-lateralis system, that the ophidioids seemto have developed is the gas bladder-rib relationship.In the following paragraphs no attempt will be made to give anycomplete structural account of Gadopsis or other ophidioids. Regard-
no. 3647 PERCIFORM FISHES?GOSLESTE 23ing Gadopsis, only those features not previously considered, in whichit differs from the ophidioids, will be mentioned. In addition, in viewof the recent reassignment of the ophidioids and zoarcids to theGadiformes by Greenwood, et al. (1966), it seems necessary to discussonce again some of those features that provide the basis for believingthat the similarities among these three groups are due to convergenceand not to genetic inheritance.Jaw structure.?In addition to characters already discussed, twoother aspects of ophidioid jaw structure will be noted herein. First,most, if not all, of the brotulids and ophidiids retain a supramaxillary.In this minor feature, Gadopsis has advanced farther from the basalpercoid condition, for it has no supramaxillary. Second, Gadopsis andthe ophidioids, like most percoids, have the premaxillary subequal tothe maxillary in length. In this they differ from such groups as thezoarcids, uranoscopids, and batrachoids, which often have very shortpremaxillaries and the much longer maxillaries to some extent includedin the gape.Suspensorium and associated structures.?The major peculiar-ity of the suspensorium of Gadopsis and the ophidioids is a trend to-ward the fusion of the mesopterygoid and ectopterygoid. This fusion,which seems to be a constant feature of ophidiids, pyramodontids,and carapids (see Regan, 1912d, and Gosline, 1960) occurs in Gadopsis.Here again, Gadopsis is somewhat more advanced than brotulids, inwhich, so far as known, the ectopterygoid and mesopterygoid areseparate.The suspensorium of the gadiform fishes and its innervation is verydifferent from anything found in Gadopsis, the ophidiids, or, for thatmatter, in the percoid fishes. Regan (1903b, p. 464) has commented onsome of the gadiform peculiarities as follows
:
Certain features of the suspensory apparatus seem to be constant throughout thesuborder, and may prove to be of some importance. The head of the hyomandibu-lar articulates with a single socket, to the formation of which the squamosaland postfrontal contribute. The entopterygoid is well developed, attached to theectopterygoid below and in front by a vertical suture to the palatine. The palatineis attached anteriorly only to the praefrontal, and has a long maxillary processBy contrast, Gadopsis and other ophidioids have two more or lessseparate articular heads on the hyomandibular, and the mesopterygoid(entopterygoid) is attached to and forms a continuous surface withthe metapterygoid and sometimes posteriorly with the hyomandibular.The most peculiar feature of the Gadiformes is the course of thehyomandibular branch of the facial nerve. In most teleosts that havebeen investigated (Patterson, 1964, p. 435), as in Gadopsis and ophidi-oids, the hyomandibular branch and the main trunk of the facialisnerve exit from the cranium by separate openings, that of the hyoman-
24 PROCEEDINGS OF THE NATIONAL MUSEUM vol. hmdibular branch being posterior and more or less internal to thehyomandibular bone (fig. 4: ho). After exiting from the skull, thehyomandibular branch enters the medial face of the hyomandibularbone and passes downward within it. In the gadoids (Stannius, 1849,p. 33), the hyomandibular branch has the same cranial exit as themain facialis trunk, after which it swings backward and penetratesthe front of the hyomandibular bone.Branchiostegal rays.?The ophidioids are said to have six toeight branchiostegal rays (Kegan, 1912d, p. 277); in Gadopsis thereare seven. This is a rather high number for percoid derivatives. Inthe stichaeoid blennies (Makushok, 1958, p. 21), these are rarelyseven, generally fewer.Pelvic fins and pelvic girdle.?The filamentous fins and theirpresumed function in Gadopsis, the ophidiids, brotulids, and certaingadids already have been discussed. (Zoarcids never have filamentouspelvics.) Despite the general similarity between the pelvic fins of theOphidioidei and certain of the Gadiformes, there are minor differences,some of which suggest different ancestries for the two groups. Thus,even when, as in the gadoid Laemonema, the pelvics become reducedto two main filamentous rays, there are rudimentary rays medial tothese; in the ophidioids, when there is a rudimentary structure inaddition to the filaments, it is a small ossicle lateral to the main raysand presumably represents a reduced spine (as in the Blenniidae andZoarcidae) . At the other extreme, however, the maximum number ofsoft pelvic rays in gadoids is twelve, but the ophidioids never havemore than two. The pelvic fins of the Gadiformes, when present, arewide set and articulate with pelvic bones that are never attacheddirectly to the cleithra; the pelvic fins of ophidioids, when present, areclose set and articulate with pelvic bones that are usually, though notalways (D. M. Cohen, pers. comm.), attached directly to the cleithra.Freihofer (1963, p. 141) recently has noted the similarity of theramus lateralis accessorius pattern in the gadoids, ophidioids, zoarcids,and (in litt.) nototheniids. In all of these, the pelvic branch of theramus lateralis accessorius extends downward across the base of thepectorals instead of downward along the postcleithrum behind thepectorals and thence forward to the pelvics. But all four groups offishes mentioned have the pelvics far forward, where the normalpercoid nerve course would be highly circuitous. Furthermore, all fourare groups living near the bottom, which may or do (Phycis, see above)use their pelvic fins to locate food. That the shorter and presumablymore efficient course of the ramus lateralis to the pelvics developedindependently in these groups is suggested by the fact that Gadopsis,herein considered to be at the base of the ophidioids, and the Bathy-masteridae, at the base of the zoarcids, have a perfectly normal percoid
no. 3647 PERCIFORM FISHES?GOSLINE 25ramus lateralis pattern (Freihofer, 1963, p. 136). In this instance, then,I would view the similarities in nerve course as an adaptive trait thathas been elicited more than once by similar circumstances.Pectoral.?In Gadopsis and ophidioids, there are four actinosts.In the Gadiformes, the number varies from three to 13. The scapularforamen of Gadiformes is usually between the scapula and coracoid;in Gadopsis and ophidioids, it is contained in the scapula.Dorsal and anal fins.?It is in the structure of the vertical finsthat the percoid affinities of Gadopsis are most plainly manifest. Inthat fish, there is a single dorsal fin with 10 pungent spines anteriorly,followed by 27 or 28 soft rays. Anterior to the dorsal fin, there are twowell-developed predorsal bones, the anterior interdigitating betweenthe second and third neural spines and the posterior between thethird and fourth. The anal fin has three sharp, graduated spines atthe front of the fin and 18 or 19 soft rays. The pterygiophores of thesespines are separate, but the second is considerably enlarged and ex-tends up in front of the first hemal arch. (One peculiarity of the dorsaland anal fins of Gadopsis is that its last dorsal and anal rays are notdivided to the base.)Ribs.?In Gadopsis, Baudelot's ligament originates on the basioc-cipital. There are epipleurals from the first vertebra and pleural ribsfrom the third. The anterior pleural ribs are enlarged only slightly,if at all. The gas bladder is large, firm walled, simple, and withoutspecial ligaments to either the ribs or skull. In all these respects,Gadopsis is typically percoid.The ribs of ophidioids are modified in various ways as alreadynoted. In one of the less-marked modifications, Brotula has epipleuralribs from the first vertebra and pleural ribs from the third (Regan,1912d, p. 278). Baudelot's ligament is attached to the basioccipital;however, in Brotula, the first two pleural ribs are expanded, and thereis a sheath of ligamentous tissue extending up and forward from thegas bladder over the anterior ribs. In no known ophidioid is the firstvertebra fused to the skull.The gadoids differ in the above features in several respects. Thereare never any epipleural or pleural ribs on the first two vertebrae.In most macrourids, the first vertebra is free from the skull andBaudelot's ligament, so far as known, is attached to the first vertebra.In the gadids, by contrast, the neural arch of the first vertebra isattached firmly to and its centrum completely fused into the cranium;here, Baudelot's ligament originates on the rear of the skull. Unlikeophidioids, there may be a direct connection between the gas bladderand inner ear in gadoids (in Moridae; Svetovidov, 1948), and whenthe gadoids have "drumming muscles," these usually are not attachedto the ribs or skull (Marshall, 1965, pp. 312-313).
26 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124Caudal fin and skeleton.?Aside from the rounded shape, thecaudal fin of Gadopsis and its supporting structure (fig. 5a) seem tobe of a fairly normal percoid type. There are five hypurals (countingas in Nybelin's 1963 system), one uroneural, and two epurals?allautogenous?and 15 branched caudal fin rays.Among the brotulids, at least one member (Gosline, 1953) has 15branched caudal rays, but there are more or less fusion and/or re-duction in the caudal skeletons of all. In carapids, the caudal skeletonand fin are absent.The caudal fin of the gadids has been the subject of much discussion.The caudal skeleton at least seems to represent a modification froma perfectly normal teleostean type (see, e.g., Barrington, 1936, andGosline, 1964) but so reduced as to be morphologically similar tothat of some brotulids.Summary.?To summarize Gadopsis, this fish seems in manyrespects to present a mosaic of characters, some percoid and othersophidioid. In the sense organs and associated structures, Gadopsisseems to have developed most of the basic peculiarities of theophidioids: it has the anteriorly located, filamentous pelvies, thesubterminal mouth and jaw structure, the at least partially reducedeyes, the expanded auditory bullae, and the troughlike sensory canalsof the head. In the following features, however, Gadopsis retains thepercoid condition rather than the more advanced ophidioid type:the ramus lateralis innervation of the pelvics, the incomplete supra-temporal commissure, and the simple gas bladder without specialrelationships to the anterior ribs.In fin structure, aside from the pelvics, Gadopsis shows a generalizedpercoid rather than the ophidioid condition. There are pungentspines at the front of the dorsal and anal, three in the anal, with thepterygiophore of the second extending in front of the first interhemal.There are two predorsal bones. The caudal fin has 15 branched raysand five autogenous hypurals. Finally, the dorsal and anal soft rayrelationship to vertebrae is percoid and does not show the crowdingof the rays found in ophidioids.In a few minor characters, Gadopsis is more specialized than atleast the more generalized ophidioids. It has no supramaxillary, andthe entopterygoid and ectopterygoid are fused. Perhaps into thiscategory should be added the fact that Gadopsis is a freshwater fish.The question arises as to whether or not Gadopsis should be retainedamong the percoids or placed among the ophidioids. Zoologically, Icannot see any clearcut basis for decision. From the viewpoint ofindicating the type of fish from which the ophidioids arose, Gadopsisand the Gadopsidae perhaps can be allocated best to the ophidioids,where the spiny-rayed Gadopsis would hold a position somewhat
PERCIFORM FISHES?GOSLINE 27
Figure 5.?Caudal skeletons: a, Gadopsis marmoratus; b, Trachinus draco; c, Bathymastersignatus; d, Scombrolabrax heterolepis; e, Scomber japonicus; f, Tkunnus albacares. (a, b,d drawn from preserved material; c, e, f from dried skeletons; Ep=epural, HA= hemalarch, Hy=hypural, and NA= nural arch; broken line in fig. J=basal limits of caudarays.)
28 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124
analagous to that of Psettodes among the Pleuronectiforrnes (Norman,1934).As noted above, Greenwood, et al. (1966, p. 397), have added theophidioids and zoarcids to the order Gadiformes. In the present paperthe more generally held view that the ophidioids and zoarcids (seebelow) have no close relationship to the gadoid fishes or to one anotheris supported. In agreement with Makushok (1958 and elsewhere), thezoarcids are assigned herein to the Blennioidei, close to the stichaeidfamilies. The ophidioids differ from these and all blennioids in having,among other things, the dorsal and anal fin rays more numerous thanthe vertebrae between them and, except in the Carapidae, which lackpelvics, in the one or two rayed filamentous pelvic fins. That thesimilarity between the zoarcids and ophidioids in ramus lateralisaccessorius nerve pattern (Freihofer, 1963) may be the result ofconvergent evolution has been suggested above.Regarding the fin-ray-to-vertebra relationship and the filamentouspelvic fins, some of the gadoids are similar to the ophidioids. Further-more, there seems to be no one well investigated character by whichall of the gadoids can be separated from all ophidioids; for example,no pelvic differences can be used to differentiate the two groupsbecause the carapids among the ophidioids and the gadiform genusMacruroides completely lack pelvics. Again, Svetovidov (1948)placed considerable emphasis on the penetration of the intercalar bythe glossopharyngeal nerve in gadoids, but this did not occur in themacrurids that Pftiller (1914, p. 76) investigated.Despite the lack of criteria that will separate all gadoids from allophidioids, I follow Regan (1903b), Svetovidov (194S), and othersin separating these two groups widely. If, as I have tried to show, theophidioids can be traced back through a fish very much like Gadopsis,then the percoid derivation of the ophidioids seems assured; by con-trast, no one in recent years has suggested a percoid derivation forthe gadoids (see, e.g., Rosen, 1964; Gosline, 1964). Leaving asidepresumed ancestries, however, many of the central tendencies in thetwo groups are very different. Regan (1903b) noted a number ofthese tendencies long ago, and more have been added by subsequentauthors. Suborder XiphioideiThe suborder Xiphioidei, as herein understood, comprises thefamilies Istiophoridae, Xiphiidae, and, provisionally, the Luvaridae.The Istiophoridae and Xiphiidae usually have been considered "ahighly specialized end-stage of the scombriform series" (Gregory andConrad, 1937, p. 2 3). The Luvaridae, containing only Luvarus
no. 3647 PERCIFORM FISHES?GOSLINE 29imperialis, has been allocated variously; Regan (1903a, p. 372)considered it "a most abnormal and specialized Scombroid."A principal reason why Regan (1903a; 1909a) placed Luvarusamong the scombroids seems to have been that in it, as in the Scom-bridae and Xiphiidae, "the deeply forked bases of the rays of thecaudal fin are inserted nearly vertically and extend over the hypuralso as to almost entirely conceal that bone, those of the upper andlower series nearly meeting in the middle line on each side" (1903a,p. 372). Additionally, in Luvarus, "the ossified sclerotic and broadopercular bones are typically Scombroid features" (1903a, p. 374).In the Xiphiidae and Istiophoridae, along with the peculiar caudalray bases noted above, the rostral structure has been considered amorphological extrapolation of the type found in the scombrids ingeneral, most notably in Acanthocybium (cf. fig. in Regan, 1909a).To the present author, it seems that all of the morphological featuresmentioned above may well be merely adjustments of large, power-fully swimming fishes to the requirements of hydrodynamic efficiency.(Hertel, 1966, e.g., p. 255, stresses the difference in what constituteshydrodynamic efficiency in large, powerfully swimming animals andin small, weak swimmers.) With increase of body size and swimmingspeed, the role of hydrodynamic forces in the existence of the animalbecomes, of course, increasingly important. It is probably significantthat among the members of the percoid family Carangidae, whichalso contains large, powerful swimmers, almost all of the morphologicalcharacters discussed above have been duplicated. Another, at leastcurious, parallel in the Carangidae is that, in those forms with a high,blunt head, the premaxillary remains protrusile, as in Luvarus;however, in the pointed-headed Chorineminae (Suzuki, 1962, p. 147),the premaxillaries are rigid and form a beaklike structure similar tothat of Scomber.If, however, one excludes from consideration those features that maybe related to hydrodynamic efficiency, there seems to be slight re-semblance between the Istiophoridae, Xiphiidae, and Luvaridae onthe one hand, and the Scombridae, on the other. In the former group,the vertebrae number from 23 to 26 (a typically percoid condition);in the Scombridae, the vertebrae are 30 or more. In Xiphias andTetrapterus (Gregory and Conrad, 1937, fig. 5), the caudal skeletonis only about as specialized as that of Scomber, certainly far lessmodified than the caudal skeleton of the tunas. In Luvarus, with thefusion of the last two vertebrae, the caudal skeleton (Gregory andConrad, 1943, fig. 7) has become modified in a different fashion thanthat of the Scombridae.Probably of greater importance, the istiophorids Xiphias andLuvarus seem to guide then forward trajectory in a somewhat different
30 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124way than do the Scombridae. In the Scombridae, the route of forwardtrajectory seems to be controlled, at least in part, in usual percoidfashion by a combination of well-developed 6-rayed pelvics directlybelow the highly placed pectoral fins (Harris, 1938). In the istiophorids,xiphiids, and Luvarus, the pelvic fins have a reduced number of raysor none. The pectorals are low on the body and have become fixedin extended position in the adults of Xiphias and of the istiophoridIstiompax indicus (thus secondarily resembling the shark condition).In this regard, it should be noted that, in the trichiurids and in manygempylids, the pectorals are low and the pelvics reduced or absent,but such forms are relatively small, weakly swimming fishes.Finally, the dorsal fin of the Scombridae commences well behindthe head. That of the Istiophoridae and Xiphiidae originates overthe back of the head. The first interneurals of Xiphias are shown byGregory and Conrad (1937, fig. 3) to extend downward into theregion of the skull-vertebrae articulation. In the juvenile Luvarus(Gregory and Conrad, 1943, fig. 38), the dorsal fin again originatesfar forward, but, with growth, moves back, leaving, however, a pairof large interneurals that interdigitate between the cranium and thefirst vertebra (Gregory and Conrad, 1943, fig. 8).In certain respects, e.g., the 23-26 vertebrae, the Istiophoridae,Xiphiidae, and Luvaridae are more generalized than the Scombridae.That they are specialized scombrid offshoots seems an impossibleconclusion, and that they are even related to the Scombridae, animprobable one.A more difficult problem is to determine what the Xiphioidei isrelated to and/or derived from. Before this matter can be profitablydiscussed, the question arises as to whether or not the Istiophoridae,Xiphiidae, and Luvaridae are interrelated. Regan (1909a), Gregoryand Conrad (1937), and others have postulated that the Xiphiidaeand Istiophoridae extend back separately into Eocene times. Thatthe two families are related more closely to one another than to anyother modern family has not, to my knowledge, been questioned.Whether or not the Luvaridae are related to the Istiophoridae andXiphiidae is more doubtful. Certainly Luvarus has many featuresthat separate it widely from all other living fishes. In mouth andsnout structure, Luvarus differs widely from the istiophorids andxiphiids. It may be that these features provide good indications ofphylogenetic relationships, but the alternative possibility at least issuggested here that the anterior profiles of Luvurus, on the one hand,and of the istipohorids and xiphiids, on the other, represent alterna-tive attainments of hydrodynamic efficiency in large, strongly swim-ming fishes and, hence, are not necessarily of great phylogeneticsignificance. In any event, the Luvaridae herein are included pro-
no. 3647 PERCIFORM FISHES?GOSLINE 31
visionally in the Xiphioidei. What appear to me to be the moreimportant unifying elements of the Xiphioidei, as understood herein,are the following:Vertebrae 23-26. Pelvic fins, if present, with not more thanthree rays. Pectorals inserted low on sides. Dorsal and anal fin raysat least somewhat more numerous than the vertebrae. Anteriorinterneurals interdigitating between the skull and the vertebralcolumn. Frontal bones without a median crest (though the supra-occipital extends forward over the frontals in Luvarus). Nasal bonesforming a rigid portion of the head skeleton (or possibly absent inLuvarus: see Gregory and Conrad, 1943, p. 254).The Xiphioidei seem to have originated among the basal percoidstock, though no modern percoid group suggests any obvious rela-tionship with the xiphioids. That the group is an old one, extendingback at least to the Eocene, is well attested to by fossil evidence(though the usual attribution of the Palaeorhynchidae, with 50-60vertebrae, to the xiphioids seems dubious).Suborder ScombroideiThe fishes herein included in the suborder Scombroidei are theScombridae as defined by Kegan (1909a), Fraser-Brunner (1950),Collette and Gibbs (1963) and the trichiuroid fishes, i.e., the familiesGempylidae (cf. Matsubara and Iwai, 1958), the Trichiuridae (cf.Tucker, 1956), and the Scombrolabracidae (Roule, 1922). The Istio-phoridae, Xiphiidae, and Luvaridae, usually included in the Scombro-idei (e.g., Regan, 1909a; Gregory and Conrad, 1937, 1943), hereinhave been removed to a separate suborder, Xiphioidei, for reasonsgiven in the previous section.Among the trichiuroid families, the relationship between the Gem-pylidae and the Trichiuridae has never, to my knowledge, been ques-tioned. Scombrolabrax, discussed below, has been placed near theGempylidae since its discovery in 1922.Again, a postulate of relationship between the Scombridae and thetrichiuroid families, particularly the Gempylidae, generally has beenaccepted. The only question has been whether or not the two groupsshould be placed together in a single suborder (e.g., Regan, 1909a)or allocated to separate suborders (e.g., Regan, 1929). It is true thatthe principal evolutionary trends in the two groups have been verydifferent. That of the trichiuroids has been toward large-fanged,ribbon-shaped forms, whereas the scombrids have developed into thebulky, powerfully swimming tunas. Nevertheless, in many of whatwould appear to be basic structures, the trichiuroids and scombridsoverlap. Indeed, the presumed gempylid Lepidocybiwm shows so many
32 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124
scombrid characters (Matsubara and Iwai, 1958) that its transfer tothe family Scombridae has been advocated. Conversely, the scombridGrammatorcynus has a number of gempylid characters (Matsubaraand Iwai, 1958). Finally, it seems that, except in a few characters,the genus Scombrolabrax (fig. 6), could serve morphologically asan ancestral form for the trichiuroids and, in most respects, for theScombridae as well.
Figure 6.
?
Scombrolabrax heterolepis: sketch to show external appearance, based on speci-men 5}i inches SL (USNM 197651) taken off Mississippi delta by the "Oregon" (drawnby Barbara Downs).
Regarding scombrid phylogeny, Kishinouye (1923) consideredcertain of the tunas to be so specialized as to warrant a separate order,Plecostei. This classification, though adopted by Berg (1940), wasshown long ago to be based on inadequate grounds (Takahasi, 1926).At the base of the scombrid series, Fraser-Brunner (1950) placedGasterochisma. It appears to me, however, that Gasterochisma, whichI have examined only superficially, bears at least as much resemblanceto the Bramidae as to the Scombridae; if Gasterochisma is a scombridat all, it is at best a highly aberrant one.Starks (1910) seems to have been correct in considering Scomberas the least specialized living scombrid. Among the percoid-likecharacters retained by Scomber but lost by most or all of the rest ofthe Scombridae are the following:Mesethmoid with a low median crest anterodorsally (see Allis,1903, pi. 4: fig. 5). Intercalar not expanded on the posterodorsal faceof the skull, not separating the exoccipital form the pterotic; lowerlimb of the posttemporal articulating with an intercalar projectionthat extends downward and backward from the ventral cranial sur-face. Premaxillaries with separate articular and ascending processes,the latter not greatly expanded (ibid., pi. 5: fig. 16). Circumorbitalseries of bones complete (ibid., pi. 3: fig. 4). Operculum without asmoothly rounded free border but rather with a moderately deepindention above (ibid., pi. 3: fig. 4). An anal spine present (Matsui,
no. 3617 PERCIFORM FISHES?GOSLINE 331967). In the caudal skeleton of Scomber (fig. be), the upper and thelower hypural plates remain separate with a notch between them,and the preurostylar vertebra has no attached neural arch; by con-trast, in such an advanced scombrid as Thunnus (fig. bf), the upperand lower hypurals have fused into a single plate without a mediannotch and the preurostylar vertebra seems to have a well developedneural arch (fig. 5/: NA), though this may represent a fusion betweenthe anterior epural of Scomber (fig. be) and the preurostylar centrum(cf. Gregory and Conrad, 1943, fig. 5d).Among the trichiuroid fishes, increasing degrees of morphologicalspecialization are shown by the series Scombrolabracidae-Gempylidae-Trichiuridae. Since no account of the osteology of the basal memberof the series, namely Scombrolabrax (fig. 6), has ever been given, oneis presented below. The Osteology of ScombrolabraxFigures bd, 6Teeth.?The jaw teeth are all well separated from one another,and all point more or less backward. They are in single rows exceptfor one to three inner teeth near the midline of each jaw; these innerteeth of the upper jaw are needle-like fangs and are by far the largestin the mouth, but the inner teeth of the lower jaw are small. The outerrow in each jaw is made up of well separated, sharp, distally-proximallyflattened teeth; those along the sides of the lower jaw are much thelarger. There is a single row of small teeth on each palatine and aV-shaped row on the vomer. Mesopterygoid toothless.There are three patches of needle-like teeth on the upper pharyngealsof each side; the separate lower pharyngeals have similar teeth.On the first arch are five lathlike gill rakers that, however, havespines projecting from their posterior border. The other gill rakers arein the form of low, spinulose platelets. On the rear face of the anteriorarch and on succeeding arches are numerous rakers consisting of single,upright, needle-like spines (cf. Matsubara and Iwai, 1952).There are no teeth on the hypobranchials, basibranchials, or tongue.Sensory canals of head and associated bones.?The infraorbitalcanal is complete and joins the supraorbital canal between the frontaland pterotic as usual. The lacrimal is a long bone that does not overlapthe maxillary except far forward. It has no serrations but has theusual three canal exits along the lower surface. The first circumorbitalis essentially a continuation of the lacrimal. The second circumorbitalbears a very large subocular shelf that extends somewhat forward asweU as somewhat back of its canal-bearing portion. Above the secondcircumorbital are 11 bony half rings (the medial halves) that carry280-835?68 3
34 PROCEEDINGS OF THE NATIONAL MUSEUM vol.. rathe infraorbital canal up to its junction with the supraorbital canal(three of these are slightly larger than the others and may representthe usual percoid circumorbital bones).The supraorbital canal starts in a tubular nasal bone that is attachedmovably to the frontal behind it. The canal then passes back throughthe frontals, giving off two major lateral exits and one median. Themedian exit apparently represents the interorbital commissure; oneach side, it passes in through a low frontal rise and opens out ontothe surface of the skull on the interior slope of this rise; the open-ing is covered with a membrane, and there is no sign of any connectionbetween canals of the two sides of the head.The temporal canal extends the full length of the pterotic in atrough, open externally. The preopercular canal joins the temporalcanal via a membranous tube.Jaws.?The upper jaw is distinctly protrusile. The usual ethmoid-maxillary and palatine-premaxillary ligaments are present.The maxillary has a long, subtriangular supramaxillary.The premaxillary is uot beaklike. Its ascending process is nearlyvertical, with the usual deep groove between it and its well-developedif low articular process over which the maxillary head rides.Suspensorium.?The top of the interopercle and the lower portionsof the subopercle and preopercle have weak serrations. There are twoweak points on the opercle separated by a deep indentation; abovethe upper of these, the opercular edge is more or less ragged edged.There is no metapterygoid lamina (cf. Katayama, 1959).Hyoid apparatus.?There are seven branchiostegals on each side,not six as reported by Roule (1922).There is a groove along the epihyal continued forward into theceratohyal, also one anteriorly on the ceratohyal; the grooves at thetwo ends of the ceratohyal are connected by a completely enclosedtunnel.The usual gill arch bones are present.There is a well-developed pseudobranch.Cranium.?The inner face of the maxillary head rides on the side ofthe vomerine portion of the ethmovomerine keel. The ethmoid con-tributes to the keel but also has a broad, flat upper portion under andbetween the frontals.Posteriorly, the frontals become slightly raised medially. Appressedagainst a portion of the lower surface of this rise is the "pineal organ"(Rivas, 1953). Laterally, there are two low ridges over the supra-orbital canal. The whole top of the skull looks like that shown byMatsubara and Iwai (1958, fig. 5) for Ruvettus.The parasphenoid is slightly arched. There is no posterior opening tothe myodome.
no. 3647 PERCIFORM FISHES?GOSLENTE 35The pleurosphenoids do not meet on the midline.A basisphenoid is present.The auditory bulla is swollen somewhat, with a peculiar, lateral,puffed-out area in the exoccipital. There are no soft areas on thebullae walls.The round facets for vertebral articulation on the exoccipitals seemto be separate from each other and from the round area on thebasioccipital.Paired fins and girdles.?There are four actinosts. In the wetspecimen, the bottom one articulates with the cartilage over andbetween the scapula and the coracoid. The very long pectoral fin has18 rays, the uppermost of which inserts below the level of the main(lowermost) opercular projecting point.The upper, laminar postcleithrum is attached entirely to the clei-thrum above. To its anterior edge is attached the lower, long sword-like postcleithrum, Avhich runs down in back of, and has a ligamentextending to, the pelvis.The pelvis extends between and is attached tightly to the cleithrain normal percoid fashion. The pelvic bones are long and somewhatseparate on the midline. They have relatively long posterior processes.The pelvic fin has a well-developed spine and five soft rays.Axial skeleton.?Vertebrae 13+ 17. The first vertebra with a well-developed hemal spine is the fourteenth. Vertebrae five through 13have parapophyses, the anterior more or less laterally directed, chang-ing to vertically posteriorly. Pleural ribs articulate with notches inand behind the tips of the parapophyses.In the caudal skeleton (fig. 5d), the urostyle extends back betweenthe fourth and fifth hypurals (counting as in Nybelin's 1963 system),leaving the uppermost hypural alongside the two autogenous uro-neurals. (The possibility that Scornbrolabrax has only four hypuralsand three uroneurals was investigated and dismissed because thelowermost of the three bones immediately above the urostyle [fig. 5:?Hy5] ends posteriorly in a cartilaginous plate that forms a continuousedge with that of the hypurals below, whereas the upper two bones ofthe series, i.e., the uroneurals, do not.) There are three separate epuralsand three autogenous hemal arches. The caudal rays only slightlyoverlap the hypurals (about as shown by Matsubara and Iwai, 1958,fig. 9).The first dorsal pterygiophore interdigitates between neural archestwo and three. There are no predorsal bones.In the anal fin there are three close-set, graduated anal spines. Thefirst two anal spines articulate with one pterygiophore, the third witha separate one.
36 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124Internal organs.?The peritoneum is black. The stomach isstraight, elongate, and thick walled. There are six finger-like pyloriccaeca. The gas bladder extends nearly the full length of the abdominalcavity and is rounded at both ends.DiscussionRoule (1922, 1929) and Grey (1960) agreed that Scombrolabrax isrelated to the gempylid fishes. Both authors have noted the similarityin general appearance between Scombrolabrax and the gempylid genusEpinnula. Grey demonstrated in some detail the similarities betweenthe peculiar lateral-line scales of gempylids and those of Scombrolabrax.The upper jaw structure with its long supramaxillary and its fangduplicates that of the Gempylidae as illustrated by Matsubara andIwai (1958, fig. 3). The skull roof, as previously noted, seems to bethat of the gempylid Ruvettus. The spinulose gill rakers again are likethose of gempylids. Indeed, there seems nothing about Scombrolabraxthat would militate against a Scombrolabrax-gempylid relationship.In most instances wherein Scombrolabrax differs from the gemylids,it differs in the direction of the percoids. Thus, in Scombrolabrax, theupper jaw is protrusile, some of the opercular bones are spinous orserrate, the pelvic girdle is relatively strong and firmly attached tothe cleithra, the parts of the caudal skeleton are not fused, the lateralline is simple, the lateral-line scales bear a groove rather than a com-pletely bone-enclosed tunnel (Grey, 1960), the number of vertebraeis relatively low, etc.If Scombrolabrax is included in the trichiuroid fishes and if thetrichiuroids and Scombridae are combined in a single suborder, thedifficulties of defining the suborder become considerably greater. Thebest that I can do in this regard is as follows.The suborder Scombroidei are perciform fishes with nonprotrusileupper jaws (except Scombrolabrax), the postorbital members of thecircumorbital ring of bones represented either by numerous smallpieces or absent, the interorbital commissure of the supraorbitalcanals widely incomplete or lacking, the predorsal bones (Smith andBailey, 1961) lacking, and the vertebrae numbering 30 or more.Regarding the origins of the Scombroidei and more especially theScombridae, these frequently have been postulated to lie in the areaof the percoid family Carangidae (e.g., Starks, 1911). My own workhas led to the conviction that Regan (1909a) was correct in separatingthe Scombridae widely from the Carangidae and that the rathernumerous morphological features held in common by members of thetwo families (Starks, 1911) are the result of convergence. The reasonsfor this conclusion are as follows
:
(1) In the Carangidae (see Suzuki, 1962), the supraoccipital crestalways is carried forward on the frontals to the ethmoid region and
no. 3C47 PERCIFORM FISHES?GOSLINE 37provides a source of attachment for the body musculature, whichextends anteriorly along either side of it. The interorbital commissureof the lateralis canals is always complete and has a median openingbetween the frontals on the top of the crest.In the scombroid fishes (including the trichiuroids) , thesupraoccipital crest and the body musculature do not extend forwardover the head medially beyond the supraoccipital, except, to myknowledge, in Gasterochisma, Scomberomorus, and Acanthocybium.Other than in these genera, there is either a median open spacebetween the frontals posteriorly or a transparent area in the frontalsdirectly under which is an expanded "pineal organ" (Rivas, 1953).The interorbital commissure of the lateralis system is never complete(it was not located in the large skull of Gasterochisma examined).Except in Scomberomorus and presumably Acanthocybium, the twolateral portions of the commissure are widely incomplete on themidline; in Scomberomorus and presumably Acanthocybium, thetwo halves of the commissure extend up the outside surfaces of thehalves of the frontal crest and open by separate exits on either sideof its rim. If Rivas (1953) is correct in postulating the pineal body asa light receptor in scombroids, then the scombroids, except Scom-beromorus and Acanthocybium, have a rather different system ofsensory perception on the top of the head than the carangids, and thetwo exceptional genera would represent an incomplete return towardthe carangid system.(2) In the Carangidae, the usual five suborbital bones are present(see Suzuki, 1962), forming a typical complete circumorbital ring.In the scombroids, the suborbital bones behind the eye arevariously modified or absent. In Scomber and Rastrelliger (Allis, 1903,pi. 3; fig. 4; Starks, 1910), they form a series of flat, somewhatexpanded plates that appear to be variable in number. In Scombrola-brax, they occur as rather numerous small ringlike ossicles (see above)
.
In most of the other scombroids, the posterior suborbitals, alongwith the postorbital section of the infraorbital canal, are absent orrepresented by scalelike ossifications.(3) In the Carangidae, the vertebrae are almost always 24 andnever exceed 26 (Suzuki, 1962).In the Scombridae, the vertebrae are 30 or more.To me, a more promising area of scombroid origin among the percoidfishes is that represented today by the Pomatomidae, especiallyScombrojps. It is not so much that the pomatomids positively fore-shadow the scombroids as that they appear to be more generalizedpercoids, lacking the rather numerous nonscombroid specializationsfound in the Carangidae; e.g., the median frontal crest bearing theinterorbital lateral-line commissure. The Pomatomidae have the
38 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124
anterior portion of the cranial roof flat or with a low arch, the inter-orbital commissure of the lateralis system is broadly incomplete,and the vertebrae number 26.Suborder BlennioideiThe fishes united here under the Blennioidei form one of the mostunsatisfactory suborders of the Perciformes. The blennioids are percoidderivatives that basically have taken up a mode of life in contact withthe bottom. This mode of life, however, has been adopted repeatedlyby percoid derivatives; indeed, it is the most successful of postpercoiddevelopments among fishes. All of the various fishes that live in contactwith the bottom have developed certain specializations in common.For one thing, all of the sense organs in which perception depends onambient water tend to move toward the upper surface of the head andbody. More important are the changes associated with locomotion.Insofar as the basal percoid must maintain at least equilibrium in afluid environment, it is always "swimming" or at least "treadingwater." By contrast, a fish maintaining contact with the bottom isbasically sedentary (unless it is a continuous "grazer") and swims onlyin short dashes from a standing start. These differences in swimmingrequirements are reflected in fin structure.The problem with the bottom-living percoid derivatives is to dis-tinguish the convergent characters associated with a life in contactwith the substrate from the indicators of similar genetic inheritance.Beyond that lies the difficulty of defining groups and of separatingthem from the basal Percoidei.From Linnaeus (1758) to the present, the position of the pelvic finshas formed a major basis for fish classification. The majority of thepercoids and their derivatives have the pelvics more or less under thepectorals. Most or all of the derivative forms with pelvics ahead of thepectorals usually have been allocated to the Jugulares. Such a divisionassumes that the pelvics, once they have moved forward of the pec-torals, do not return. To my knowledge, this assumption is correct.The question of how many different times the pelvics have movedforward is more difficult. The refinements in the Jugulares proposed byBoulenger (1901, 1904) and Jordan (1923) have consisted primarily inexcluding from the Jugulares polyphyletic elements in which anteriorpelvics had been developed independently. Jordan's (1923) concept ofthe Jugulares is closest to the suborder Blennioidei, as accepted here,of any classification previously proposed (see table 1).Since the Jugulares of Jordan and Boulenger are percoid derivatives,one difficulty is to determine where the percoids end and the Jugularesstart. In many percoids, e.g., the Serranidae, Cepolidae, Chiasmodon-tidae, and the whole series of families around the Pseudochromidae-
no. 3647 PERCIFORM FISHES?GOSLINE 39
Plesiopidae, the pelvics are sometimes behind and sometimes in frontof the pectorals. Under the circumstances, it seems impossible toadopt pelvic position alone as a basis for distinction. As an additionalcharacter, Regan (1912d) used reduction in the pelvic to four or fewerrays to separate out a group (table 1), which he called the SuborderBlennioidea. Various aspects of the artificiality of Regan's Blennioidea,however, have been pointed out by Starks (1923), Regan himself(1929), Hubbs (1952), Smith (1952), Gosline (1955), and Makushok(1958). In this paper, a different supplementary character to definethe Jugulares will be adopted, namely, the presence of an exact 1 :
1
ratio between the vertebrae and the dorsal and posterior anal soft rays.One result of adopting this additional criterion is to exclude from theJugulares a number of fishes with anterior pelvics such as serranids andserranid-like families and the Opistognathidae. It also excludes fromthe Jugulares some almost certainly extraneous elements such as theMastacembeliformes and Gadopsidae and three "Series" included byJordan (1923), namely, the Brotuliformes, Ophidiiformes, and Carapi-formes. If this supplementary criterion clarifies the limits of theJugulares, it adds certain phylogenetic complications that will benoted below.Even if the Jugulares are defined as acanthopteran fishes with thepelvics ahead of the pectorals and an exact correspondence betweenthe dorsal and anal rays and the vertebrae, certain groups would beincluded that do not seem to belong there. These are the champso-dontoids, the ammodytoids, the schindleroids, certain gobioids, thePleuronectiformes, and possibly the Symbranchiformes. Of these, theSymbranchiformes can be at least technically excluded because theyhave no dorsal and anal rays at all. The flatfishes are set aside easilyon the basis of asymmetry. The schindlerioids have no pelvics, butneither do a number of specialized Jugulares. Under the circumstances,it is easiest to exclude Schindleria on the basis of its fused caudalvertebrae. Among the gobioids, certain burrowing forms, e.g., Trypau-chen, Microdesmus, Kraemeria, have anterior pelvics; these may beremoved on the basis of their lack of parietals.The champsodontoids and ammodytoids provide more seriousproblems. In the first place, it is not absolutely certain, in my opinion,that they should be excluded from the Jugulares. On the assumptionfollowed here that they should be, the best means of doing so wouldseem to be their forked caudal fin preceded by a long, constrictedcaudal peduncle supported by bladelike neural and hemal arches.The only remaining problem in defining the Jugulares is that ofcertain specialized groups that may well hve been derived from them.Such groups are the batrachoids and lophioids, the Callionymidae,Draconettidae, and Gobiesocidae. What the batrachoids and lophioids
40 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124
evolved from is not clear to me. They are, in any event, much morehighly specialized than the Jugulares in a number of respects (Regan,1912b), and perhaps they are excluded most easily because of theirrigid attachment of the post-temporal to the cranium. The Calliony-midae, Draconettidae, and Gobiesocidae appear to have been de-rived from one of the Jugulares groups (see fig. 12). Once again, how-ever, they are specialized sufficiently to warrant separation. Theymay be removed most easily by the absence of a metaperygoid.The net effect of the restrictions outlined above is to eliminate anumber of groups from Jordan's (1923) Jugulares. Such excludedgroups are: the suborder Haplodoci, the series Callionymiformes,Ammodytiform.es, Brotuliform.es, Ophidiiformes, and Carapiformes,and the families Chiasmodontidae, Opistognathidae, Owstoniidae,Champsodontidae, and Cerdalidae. (In the families Chiasmodontidae[Norman, 1929] and Owstoniidae [Kamohara, 1935], the position ofthe pelvics, judging from illustrations, is somewhat variable buthardly warrants their inclusion in Jordan's Jugulares. These twofamilies will not be mentioned further here.) The fishes in the re-maining families of Jordan's (1923) Jugulares are those comprisingthe group to be dealt with here. These fishes may be defined as follows:Symmetrical acanthopteran fishes with the pelvic fins, whenpresent, inserted ahead of the pectorals. Dorsal and posterior soft analrays exactly equal in number to the vertebrae between them. Caudalfin usually rounded; when forked, it is not preceded by a constrictedpeduncle supported by several fused vertebrae or by blade like neuraland hemal spines. Metapterygoid and parietal bones present. Post-temporal movably attached to cranium.Though the group herein dealt with is closest to the Jugulares ofJordan (1923), as noted above, it will be called, henceforth, the sub-order Blennioidei, to bring the subordinal nomenclature into linewith that usually used in fishes.Morphological CharactersGeneral features.?As compared with the percoids, the Blen-nioidei (for the families included in this suborder as herein understoodsee table 3) have less deep, compressed bodies. The abdominal regionof the Blennioidei frequently is rather short, with the anus relativelyfar forward and with the caudal portion of the body always more orless attenuated. Dorsal and anal fins are low and long, usually endingposteriorly close to the outer caudal rays, and frequently extendingfarther forward than is usual in percoids. The caudal and pectoralfins usually are rounded. The gas bladder is generally absent in theadult.
PERCIFORM FISHES?GOSLINE 41Members of the Blennioidei that I have seen in life, primarilytropical blennies, move forward by undulation of the body and fins;even when at rest on the bottom, they maintain a sinuous bodyconfiguration.Nasal organs.?The Blennioidei are somewhat unusual in thatthe two nostrils have become reduced to one in two different groups.All of the cold-water blennies (Zoarceoidae) have only a single nostrilon each side. The same is true of the Bovictidae, Nototheniidae,Harpagiferidae, Bathydraconidae, and Channichthyidae, thoughother members of the notothenioid stock, e.g., the parapercids,trichonotids and cheimarrichthyids, have two on each side. Attemptsto relate nostril number to gross olfactory rosette structure have beenunsuccessful. There does, however, seem to be a correlation betweennostril number and geography?perhaps 90 percent of all frigid-waterfishes, including Blennioidei, have one nostril on each side, whereassome 90 percent of all tropical fishes, including Blennioidei, have two.Circumorbital bones (fig. 7).?The circumorbital bones havebeen used extensively in the classification of certain groups of Blen-nioidei (e.g., Regan, 1912d; Stephens, 1963; and Springer, 1964).Nevertheless, for distinguishing major groups, they must be utilizedwith considerable circumspection. The basal percoid pattern com-prises a lacrimal and five circumorbitals, the uppermost (dermophen-otic) movably attached to the cranium. The second circumorbitalnormally bears a subocular shelf in marine forms (Smith and Bailey,1962). The sensory canal of the lacrimal contains several neuromastorgans; that of the second circumorbital, two; the other circumorbit-als have a single neuromast. Among the Blennioidei, the percoidpattern just described breaks down in many ways, though the basictrends are only two.
Figure 7.?Right circumorbital bones: a, Cheimarrichthys fosteri; b, Harpagifer btspinis;c, Trachinus draco. Lateral views, except that in f a top view of the anterior end ofthe series is shown below. (ds= Dermosphenotic, la? lacrimal, so= subocular shelf.)The first trend, occurring in most of the notothenioid and zoarceoidseries, is toward a disintegration of the circumorbital system. Thefirst stage in such a trend is shown by the notothenioid Para-percis (Gosline, 1963, fig. 2a). There, the subocular shelf is missing
42 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124and six circumorbital bones are present; undoubtedly, this increasehas occurred by the breaking up of the percoid second suborbital intotwo components, each with a single neuromast. Further change inthe system may take place in three fashions. First, disintegration maycome to involve the lacrimal, as apparently occurs in the zoarceoidLycodes, in which the lacrimal is divided into two almost separateportions. Second, in fat-cheeked forms, the lower circumorbitals mayleave the orbital border, as occurs in the notothenioid CJieimarrichthys(fig. 7a) and again in the zoarceoid Lycodes. Finally, the central por-tion of the circumorbital system may drop out entirely, as occurs inthe notothenioid Bembrops or the zoarceoids Lumpenus and Ptilichthys(Makushok 1961b, p. 235, fig. 4).In addition to the various stages and types of circumorbital dis-integration occurring in the notothenioids and zoarceoids, there arefrequent instances of a complete reversal of the trend itself. Thus,among zoarceoids, the anarhichadids have a strongly constructed,nearly rigid circumorbital chain of bones (Barsukov, 1959, pis. 7-16).Among the notothenioids, the circumorbital series forms a more or lessrigid ring of bones in Hemerocoetes and Harpagifer (fig. 76), and inCrystallodytes , this ring is made up of only three bones (Gosline, 1963).As contrasted with the notothenioids and zoarceoids, the generaltrend of circumorbital bones in the tropical blennies, trachinoids, andcongrogadoids is toward a strengthening of the ring and a consolidationof its elements. Again, various processes are involved. Some of theseare well indicated within the single genus Trachinus. In T. draco (fig.7c), which approaches the percoid condition more closely than anyother member of the Blennioidei, there are a lacrimal and five cir-cumorbitals, with a well developed subocular shelf on the second.In T. radiatus, the whole chain forms a rigidly interlocked series ofbones; the lacrimal and what was the first circumorbital of T. dracoare united rigidly; there is a subocular ledge running all the way aroundthe bottom of the orbit; and the first circumorbitals have also expandeddownward over the cheek, foreshadowing the condition in the "urano-scopoid families."At the posterodorsal end of the circumorbital series, two differentthings may happen. One appears to be a simple loss of elements.Thus, in T. vipera, I can find only two circumorbital bones abovethat which bears the subocular shelf, instead of the three of T. draco.Again, among the congrogadoids, there are two circumorbital bonesabove that bearing the subocular shelf of Congrogadus, but in therelated Notograptus, there is only one.A different development of the uppermost circumorbital bone occursin the topical blennies. In Enneapterygius and to some extent in Clinus,
no. 3647 PERCIFORM FISHES?GOSLINE 43the uppermost circumorbital retains its usual superficial positionbehind the orbit. But in Labrisomus and Blennius, this uppermostelement becomes largely buried in the flesh and forms what appears tobe a cranial bone rather than a member of the circumorbital series(Springer, 1966).Once again, however, it must be noted that consolidation of thecircumorbital series is not a universal feature in the tropical blennies,congrogadoids, and trachinoids. Indeed, in the clinid blenny Exerpesasper, the circumorbital chain is widely incomplete, being representedanteriorly only by an isolated lacrimal (Springer, 1955).Jaw apparatus.?There is no supramaxillary in the Blennioidei.In forms with relatively long premaxillary pedicels, there seemsto be two kinds of jaw protrusion. In one, represented by Congrogadus,the pedicels are stout and affixed firmly to the toothed portions. Insuch fishes, protrusion of the upper jaw may be great, but there islittle possibility of expanding the gape laterally. A different systemoccurs in most trichonotids and in certain of the tropical blennies.Here, the premaxillary pedicel is hinged at its base with the resultthat the distal ends of the premaxillaries can expand outward at thesame time the whole bone is protruded forward.In the zoarceoids especially but also in the unrelated gobiesocidand batrachiform fishes, the toothed portion of the premaxillary isrelatively short, with the maxillary extending well out behind it.Indeed, in such a zoarceoid as Anarhichas, it cannot be said that themaxillary is excluded from the gape.Opercular and hyoid apparatus.?Opercular armature is un-common in the Blennioidei.As in other bottom fishes, water tends to be expelled from the upperportion of the gill cavity. Among many of the Blennioidei there is aspecial valve for this purpose (Makushok, 1958, pp. 20, 21, fig. 8).By contrast, the gill openings usually are restricted more or less be-low, with the gill membranes attached to one another across theisthmus or broadly attached to the isthmus. The trachinoids and thenotothenioid families Trichonotidae (sensu lato) and Bovictidaeare exceptional in having the gill openings extending far forward.The trachinoid fishes (Gill, 1907) and at least some trichonotidsbury themselves up to the eyes in sand or mud. Baglioni (1908)has shown that Trachinus and Uranoscopus, at least, pump waterover the gills by sliding the branchiostegal membranes up and downover the cleithral region. Inasmuch as the branchiostegals of thesefishes and of such trichonotids as Crystallodytes (Gosline, 1963) arelargely covered by the operculi, this method of breathing must causea minimum of disturbance in the surrounding sand or mud.
44 PROCEEDINGS OF THE NATIONAL MUSEUMLycodapus, generally placed among the zoarceoids, is another fishwith the gill openings extending far forward, but the relationshipsof this fish seem open to question.Another feature that may be associated with wide gill openings isthe branchiostegal ray number. Thus, in most Blennioidei, there aresix branchiostegal rays, but in the Bovictidae and frequently in theTrichonotidae, there are seven. Among the zoarceoids, however, theAnarhichadidae, with the gill membranes broadly joined to theisthmus, also have seven. Makushok (1958, p. 21) considers the con-dition in anarchichadids to represent a secondary increase.Suspensorium (fig. 8).?The suspensorium develops various modi-fications among the Blennioidei, but it is difficult to evaluate thesephylogenetically.
Figure 8.?Right suspensoria and opercular bones: a, Prolatilus jugularis; b, Bathymastersignatus; c, Trachinus draco; d, Notograptus guttatus. (hc= Hyomandibular crest, hs=hyomandibular spine, in= interspace between upper and lower portions of suspensorium,mc= metapterygoid crest, mo=mesopterygoid, ms= metapterygoid strut, sy=symplectic.)The parapercid genus Prolatilus (fig. 8a) seems to be the onlymember of the Blennioidei to retain the rather typical percoid metap-terygoid strut (Katayama, 1958).Across the surface of the back of the suspensorium, various crestsdevelop for muscular attachment. Among the parapercids, such aridge runs anteroventrally across the hyomandibidar. In zoarcids(fig. 86), it is usually on the metapterygoid. In Trachinus and urano-scopids, the hyomandibular sends forward a hooklike process (fig. 8c).Various members of the Blennioidei lose a firm attachment betweenthe anterior and posterior portions of the suspensorium. Among such
PERCIFORM FISHES?GOSLINE 45fishes are the trichonotid Crystallodytes, notograptids (fig. 8d), con-grogadids, and possibly the zoarceoid Ptilichthys (Makushok, 1958,p. 66, fig. 38b).The mesopterygoid is developed variously. In Trachinus, it isbroad and in T. draco, it bears teeth. Another family in which, so faras known, it is consistently broad is the Trichonotidae. On the otherhand, the mesopterygoid appears to be narrow throughout thezoarceoids.Gill arch system.?The gill arch system of the Blennioidei isbasically percoid, with the lower pharyngeals always separate. Onlytwo modifications in the Blennioidei will be noted. The first, occurringin the congrogadoid Notograptus, is that the posterior basibranchialshave dropped out. The second, which recurs repeatedly, is that thethree upper pharyngeal tooth patches become reduced to two or,in blenniids, to one.Dorsal portion of the head (fig. 9).?The frontals usually arepaired in the Blennioidei; however, in at least the tropical blennyRunula, the frontals of the two sides seem to have fused.
ns pb po le me a I
Figure 9.?Diagram of certain structures in Prolatilus jugularis (al= anterior level reachedby the body musculature extending over the cranium, le= lateral extrascapular, ns= upperend of neural spine, pb= predorsal bone, po= posterior rim of supraoccipital, pt= pteryg-iophore of first two dorsal spines).On the dorsal surface of the cranium posteriorly, the Blennioidei ingeneral differ from the typical percoid in that the body musculaturedoes not extend forward over the skull, and the supraoccipital andfrontal-parietal crests, which, in part, form surfaces of attachment
46 PROCEEDINGS OF THE NATIONAL MUSEUMfor such musculature, usually are missing; however, in Prolatilus(fig. 9), a very generalized genus of the notothenioid family Paraper-cidae, the usual percoid condition is retained; furthermore, as in thepercoids, the supratemporal commissure is incomplete, ending blindlyover the musculature.Generally, in the Blennioidei, the supratemporal commissure iscomplete. In such fishes as most notothenioids, all congrogadoids,trachinids, and certain tropical blennies of the families Triptery-giidae (Rosenblatt, 1957, unpubl. Ph.D. dissertation) and Clinidae,the supratemporal canal runs up on each side through the lateraland medial extrascapulars and then crosses the midline in a mem-branous tube; however, in the "uranoscopoid families," in mosttropical blennies, and in all the zoarceoids, the medial extrascapularappears to have fused with the parietal.Certain tropical blennies and zoarceoids have secondarily developedcrests on the skull; e.g., a median crest along the frontals. Suchcrests, however, are for the attachment of jaw musculature, not bodymusculature (Makushok, 1958, p. 51).Even though a supraoccipital crest rarely occurs on the dorsalsurface of the skull in the Blennioidei, a small crest may be retainedon the posterior surface. In the zoarceoids, with the single exception(known to me) of Cryptacanthodes (Makushok, 1961a, fig. 3), eventhis section of the crest is lost.Sphenoid region of the cranium (fig. 10).?So far as thedifferentiation of lineages among the Blennioidei is concerned, thesphenoid region of the cranium seems to be one of the most diag-nostic parts of the whole fish. The features of importance here arethe basisphenoid and the postorbital bar.
>&>,
'? '?'?'?' ;^rTrJ:y:J:^\
no. 3647 PERCIFORM FISHES?GOSLINE 47thenioids, the basisphenoid, though generally present, is lacking inthe flat-headed Bembrops and is said to be absent (Regan, 1913,p. 141) in the Hemerocoetidae. Among the tropical blennies andtheir relatives, it is apparently always present. In the zoarceoids,there is no basisphenoid.In the zoarceoid Bathymaster, the brain cavity is separated fromthe posterior myodome only by membrane anteriorly (fig. 106),though posteriorly there appears to be the usual horizontal prooticledge separating the two cavities. In such a fish as the parapercidProlatilus (fig. 10a), by contrast, the myodome is separated almostcompletely from the cranial cavity by the wings of the basisphenoidanteriorly and by a well-developed prootic ledge posteriorly. Thisis the usual percoid condition. (For an account of variations of theposterior myodome in scorpaeniform fishes, see Quast, 1965, pp. 574,584.)In the basal percoids, the ascending wing of the parasphenoid islow (as in fig. 4) and does not extend up to a junction with thepleurosphenoid in front of the prootic. But again and again in thepercoid derivatives?and, for that matter, in lower teleosts (see,e.g., figs, in Svetovidov, 1948)?the ascending wing of the para-sphenoid becomes prolonged upward in front of the prootic to thepleurosphenoid and, in extreme instances, meets a descending wingof the frontal ahead of both the prootic and pleurosphenoid. Starks(1923, pp. 261-263), Makushok (1958, pp. 41, 42), and Quast (1965,pp. 572-574) discuss variations in this character.Among the Blennioidei, the parasphenoid always extends up to thepleurosphenoid or frontal ahead of the prootic in the Zoarceoidae andTrachinoidae; it does not do this in the Notohenioidae. In the tropicalblennies, it is variable (Starks, 1923, p. 263, and Springer, 1966).Among congrogadoids, a long sliver of prootic extends forward to theorbital border between the pleurosphenoid and parasphenoid inCongrogadus, but in Notograptus the parasphenoid and pleuro-sphenoid meet.Fin structure.-?-With a few exceptions, the differentiation betweenspines and soft rays is not as clear in the Blennioidei as it is in mostpercoids. On the one hand, pungent spines and their large pterygio-phores tend to be reduced or lost. The tropical blennies are the onlygroup that consistently has dorsal fin spines. On the other hand, thebranching of the soft rays usually is reduced; where it does occur inthe vertical fins, the posterior half of each branch rebranches soonerthan the anterior half. In many blenniid genera, e.g., Medusablennius(Springer, 1966), there are no branched fin rays at all.Paired fins and their girdles.?As noted above, the functionand structure of the paired fins in the Blennioidei are different from
48 PROCEEDINGS OF THE NATIONAL MUSEUMwhat they are in the percoids. In the percoids, the pectorals may beused to govern the vertical plane of forward movement, for stopping,turning, "treading water," and even in some?-e.g., the labrids?forforward locomotion. One of the structural features that permits all ofthese activities is the ability to rotate the pectoral base aroundthe upper ray as an axis. In the percoids, the uppermost pectoral rayarticulates with the scapula (as in fig. 11a), but the lower rays articu-late with progressively longer and independently movable actinosts.(If the outer ends of these actonosts are swung outward and downward,the pectoral fin base is brought into a plane vertical to the water; ifthey are swung up and back, the fin base moves toward a horizontalplane.) Among all but the most generalized of the Blennioidei (fig.11a), both the function and structure of the pectoral change consider-
Figure 11.?Primary pectoral girdle, right side: a, Prolatilus jugularis; b, Hemerocoetesspecies; c, Labrisomus nuchtpinnis. (In each figure, position of base of uppermost pectoralray is shown.)
ably. These fins (except in tropical blennies), instead of being used inmaneuvering, may act as props against the bottom and, by beingbrought back sharply against the body from a somewhat erect position,may provide a fast standing start from the normal stationary position.Structurally, the pectorals of the Blennioidei, except where secondar-ily reduced as in the Congrogadidae, almost always are rounded andbroad based. The pectoral girdle tends to have broad actinosts rigidlyattached to the scapula and coracoid and to one another in order toform a rather rigid, platelike surface of attachment for the pectoralrays. In one group of the Blennioidei, the Notothenioidae, the plate-like nature of the primary girdle frequently has been increasedfurther by the fusion of the uppermost actinost with the scapula, re-ducing the autogenous actinosts to three. This has occurred in theBovictidae, Nototheniidae, Harpagiferidae, Bathydraconidae, Chan-nichthyidae, and the trichonotid (sensu lato) Hemerocoetes (fig. 116).
no. 3647 PERCIFORM FISHES?GOSLINE 49Regan (1913, p. 141) states that the trichonotid (sensu lato) Bembropsalso has three actinosts, but I find four in two specimens identified asB. gobioides.Inasmuch as the possibility for pectoral rotation has been largelylost in most of the Blennioidei, the differentiation between the upper-most ray articulation and that of the lower rays diminishes. Indeed,several of the upper pectoral rays usually move up to an articulationon the scapula along with the uppermost.The tropical blennies, with the batrachoids and lophioids (Starks,1930), are unique among teleosts in that they have developed second-arily an ability to rotate the fins?but not on the uppermost rayarticulating with the scapula as an axis. Except in the Tripterygiidae,the pectoral rays all articulate with separately movable actinosts(fig. lie). (The axis for maximum rotation for such a fin theoreticallywould lie between the two middle actinosts.)The pelvic fins in the Blennioidei, when present, are always in ad-vance of the pectoral bases, though in a few of the generalized forms,like the parapercid Prolatilus, not much so.Among the Blennioidei, three things happen to the pelvic fins.One, which seems to have no phylogenetic significance, is that, inelongate fishes, the pelvics tend to dwindle in size and disappearcompletely. A sequence of this sort can be followed in the nototheni-oid family Trichonotidae (Apocreedia) , in the congrogadoids, and inthe zoarceoids (Makushok, 1958).Those Blennioidei in which the pelvics are not minute or absentseem to have put them to two rather different uses. In one, representedby the Dactyloscopidae, almost all the Blennioidae, and to someextent the Trachinidae, the two or three outer soft rays are simple,somewhat strengthened, and recurved at their tips, which extend wellbeyond the membrane between them. Usually such fins are held moreor less erect under the body.In the other type of development, the pelvic fins are held back flatagainst the abdomen, but all five rays are retained, none are strength-ened, and the inner are at least somewhat the longest. In this type ofdevelopment, which occurs in almost all of the the Notothenioidaeand in the Leptoscopidae, the pelvics frequently become separatedwidely from one another. Such fishes must rest with their thoracicareas between the pelvics in direct contact with the substrate.The pelvic girdles of the Blennioidei are very varied. The onlytaxonomically meaningful structural peculiarity that I could find isthat mentioned under the Trachinoidae (see p. 59).Vertical fins.?The basic dorsal fin arrangement that runs throughmany of the Blennioidei is a short, anterior spinous dorsal followedby a long, low fin of soft rays. Especially in the eel-shaped forms,280-835?6S i
50 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124the separate anterior spinous dorsal is lost, and there is a single longdorsal fin that may be made up entirely of soft rays {Congrogadus),almost entirely of spines (Notograptus) , or partly of each (Blenniidae)
.
The anteriormost dorsal ray is almost always far forward, and gener-ally there are no predorsal bones (Smith and Bailey, 1961); however,the notothenioid genus Cheimarrichthys does have the basal percoidnumber of three predorsals, and Congrogadus has two.The anal fin of the Blennioidei rarely contains pungent spines(see, however, Makushok, 1958, p. 34), though one or two unseg-mented anterior rays frequently are present. Among the percoidsthere is usually a more or less constant relationship between theanterior anal pterygiophores and the first hemal spine. Amongpercoids with large, pungent anal spines, the two or three first analpterygiophores frequently are fused; however, in forms with smalleranal spines, such as Acanthoclinus or the opistognathid Gnathypops,they remain separate. In Acanthoclinus, the first anal pterygiophoreextends up behind the first hemal spine; in Gnathypops, the firstpterygiophore is short, and the second extends up behind the anterior-most hemal spine. This more or less constant relationship betweenthe anterior anal pterygiophores and the first hemal spine is main-tained in the members of the family Tripterygiidae, Clinidae, andBlenniidae that I have examined; however, it is lost in the othergroups of Blennioidei. Most frequently, e.g., in the Parapercidae,Trachinidae, and Bathymasteridae, the first few anal pterygiophoresare short and well forward of the first hemal spine. The great variationthat may occur even within a group has been demonstrated byMakushok (1958, p. 29) for zoarceoid families.Posteriorly, the dorsal and anal usually approach and sometimesare connected membranously with the caudal fin. Only in some ofthe clinine clinids is there a lengthy, constricted caudal pedunclebehind the dorsal and anal. Where it does occur, it is supported, aselsewhere, by expanded, bladelike neural and hemal arches?e.g.,among the ammod3^toids (Gosline, 1963).Caudal fin and caudal skeleton.?In the Blennioidei, thefin is generally rounded or it is brushlike. Exceptions may be dividedinto two categories. One contains certain of the secondarily pelagicforms that have a somewhat lunate caudal fin, e.g., the tropicalblenny Runula. The other is made up of certain basal notothenioidswith bilobed tails. Certain species of Parapercis (Cantw ell, 1964)and possibly Cheimarrichthys fall into this category.As so often happens among fishes with rounded caudal fins, thenumber of branched rays becomes variable (Makushok, 1958). In theBlennioidei, the notothenioid Parapercidae is the only family that
no. 3647 PERCIFORM FISHES?GOSLINE 51
maintains the usual percoid number of 15 branched rays, all otherfamilies having a reduced number.In the caudal skeleton, the amount of fusion and/or loss variesall the way from an almost basal percoid condition (Gosline, 1961b)to a single bone (Makushok, 1958). The whole gamut is covered inthe notothenioid group and to a lesser extent in the others. At thevery base is Parapercis (Gosline, 1963, p. 95, fig. 6) with five hypurals(counting as in Nybelin's 1963 system), one uroneural, three epurals,and three hemal arches?all of these elements autogenous, i.e.,separate. In Trachinus (fig. 5b), at the base of the trachinoid-blennioid-congrogadoid series, there are only 11 branched rays in the caudalfin, and the two hypurals to the lower portion of the caudal fin havebecome fused, but the other elements are as in Parapercis. In Bathy-master, at the base of the zoarceoids, there are 14 branched rays; thelast hemal arch has fused to the lower hypurals to form a singleelement supporting the bottom half of the caudal fin, but there arestill three separate upper hypurals, a uroneural, and three separateepurals (fig. 5c) . All of the above fishes show less fusion in the caudalskeleton than such percoids as Acanthoclinus and Opistognathus.The pathways of fusion seem to be about the same in the variousgroups of Blennioidei. Thus, a general first stage seems to be a fusionof the lower hypurals (Trachinus, fig. 5b) followed by an ankylosisof these with the last hemal arch (Bathymaster, fig. 5c). This singleelement fused to the lower part of the caudal fin remains separatefrom the last centrum until after all of the upper hypurals and theuroneural have fused to the urostyle.Vertebral column and ribs.?The basal percoids tend to havea rather standardized vertebral column with 24 or 25 vertebrae, 10abdominal and 14 or 15 caudal. This basal number always is exceededamong the Blennioidei. The increase in the vertebral number occursfirst in the caudal section of the column; in the abdominal section,members of the Parapercidae (Cantwell, 1964), Tripterygiidae(Gosline, 1963), and Leptoscopidae (Regan, 1913) all are recordedwith 10 abdominal vertebrae.Ribs may be quite variable among the Blennioidei. Among theflatter forms, pleural ribs may be lacking completely, as in Bembropsand the leptoscopids. Pleural ribs also are lacking in the elongatePholidae (Makushok, 1958, p. 28). In the Uranoscopidae, pleuraland epipleural ribs both are attached to independent bony strutsthat Starks (1923, p. 279) has called basipleurals. More frequently,however, the usual percoid configuration of epipleural ribs from thefirst, pleural ribs from the third vetebra, is present. From structure,it is sometimes difficult (e.g., among congrogadoids) to distinguishpleural from epipleural elements.
52 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124Relationships of the BlennioideiIn the first part of this section, the Blennioidei have been delimited.Within the group, there is a whole series of what might be calledcentral tendencies that will distinguish the group from its ancestralpercoid type. Thus, in the Blennioidei, the body musculature (exceptProlatilus) does not extend forward over the top of the head, andsupraoccipital and frontal-parietal crests for its attachment arelacking. The supratemporal commissure usually is complete, extendingacross the supraoccipital. There are always more than 25 vertebrae.Predorsal bones usually are absent. The anal fin rarely has pungentspines, though there may be one of two unsegmented rays anteriorly;the anterior interhemals are not enlarged and do not abut againstthe first hemal arch. The pelvic fins either have fewer than five softrays or the inner rays are the longer. Pectoral and caudal fins usuallyare rounded. In the caudal (except Parapercidae), there are fewerthan 15 branched rays. The gas bladder usually is absent in the adult.Most or all of the above characters are associated with the basalmode of life of the Blennioidei noted previously; however, somemembers show secondary modifications. These cause some of theprincipal difficulties in distinguishing the lineages within the suborder(fig. 12) and, for that matter, in defining the Blennioidei. Thus, cer-tain members of the Blennioidei of various ancestries have developedsecondarily a more or less pelagic existence, e.g., the petroscirtinesamong tropical blennies and Zaprora among zoarceoids. When thishappens, the tail may be more or less lunate, as in Runula, instead ofhaving the rounded form typical of the Blennioidei. Furthermore, thepelvic rays of the petroscirtines tend to become filiform and weak(completely lacking in Plagiotremus) rather than sturdy, as in othertropical blennies.Some forms of the Notothenioidae and Blennioidae and all of theTrachinoidae apparently bury themselves at least up to the eyes insand. This obviously creates several problems in breathing andprobably is associated with the wide gill openings of the notothenioidfamily Trichonotidae and the superfamily Trachinoidae (see previoussection), as contrasted with the usual ventral restriction of the gillslits in the Blennioidei.Certain members of the notothenioid family Trichonotidae havebecome sand divers. Here, as elsew7here when this habit occurs
?
e.g., in the Ammodytidae and Kraemeriidae?certain morphologicalfeatures seem to develop. Thus, unlike the rest of the notothenioids,the pelvics of sand-diving trichonotids are close together and maybe reduced or disappear completely, as in, e.g., Apocreedia.Among eel-shaped forms, there is the usual tendency for the pelvicsto dwindle away and disappear first, followed by the pectorals. These
PERCIFORM FISHES?GOSLINE 53
5 ?
3 ~?3, -c^3 to
Q
fa
54 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124trends are encountered not only in the more elongate zoarceoids, butalso in the congrogadoids.In regard to what would seem to be primary phylogenetic differ-ences (fig. 12), the Blennioidei appear to be divisible into three maingroups: a notothenioid, a zoarceoid, and a trachinoid-blennioid-congrogadoid series. These three groups (see fig. 12) are contrasted intable 2.In many respects, evolution within the three groups has progressedalong parallel fines; for example, in the progressive fusion of elementsof the caudal skeleton. On the other hand, the same trend of develop-ment may occur in all three, but apparently has progressed at differentrates. Thus, among the notothenioids, the medial extrascapularremains free of the parietal in all of the forms I have examined exceptHarpagifer; in the trachinoid-blennioid-congrogadoid series, it fre-quently fuses with the parietal; and in the zoarceoids there is notrace of a free medial extrascapular. Finally, there are indications ofsecondary convergence of characters in the three series, as among thepelagic and the burrowing forms already noted.But for all these difficulties, the three do show certain primarydifferences in evolutionary development. Thus, the notothenioids havetended to flatten the head and anterior portion of the body. Probablyassociated with this are peculiarities of paired fin structure. With theexception of certain sand-diving trichonotids, the pelvics are wellseparated and maintain a full complement of five soft rays, with theinner usually the longest; these pelvics normally are held back flatagainst the body. In the pectoral girdle, the actinosts are alwaysbroad and platelike, and the uppermost frequently fuses with thescapula. In the above features, the notothenioids have developed alonglines that are not duplicated elsewhere in the Blennioidei. Conversely,the notothenioids retain certain percoid features that most of theother members have lost. Of these, the independent medial extra-scapular already has been mentioned. More important, the para-sphenoid in notothenioids (fig. 10a) has no wing extending in front ofthe prootic in such a way as to exclude the prootic from the internalorbital border.The notothenioids have, morphologically speaking, the longestlineage in the Blennioidei. They extend from the Parapercidae, themost percoid-like family of the Blennioidei, out to the Callionymidaeand Gobiesocidae. The latter groups show quite clearly all the trendsof notothenioid development listed above (except that the Gobiesocidahave only four pelvic rays), and the specializations that seem towarrant their exclusion from the Perciformes altogether lie along otherlines.
no. 3647 PERCIFORM FISHES?GOSLINE 55The other two main series of the Blennioidei rarely are flattenedanteriorly, generally tend (for very different reasons) to reduce thenumber of pelvic rays, and, with the exception of the Leptoscopidae,never have the pelvics widely separated.Of these two series, the Zoarceoidae, so far as known (but seebelow) , are structurally the most homogeneous. There is among thezoarceoids a general trend toward elongation, and concurrently (asnoted) for the pelvic fins, followed by the pectorals, to diminish anddisappear. In skull characters, the zoarceoids are all specialized:there is no separate, medial extrascapular; the prootic always isexcluded from the orbital border by the parasphenoid (fig. 106) ; andthere is no basisphenoid. In this last feature, the zoarceoids differfrom all but a few of the other Blennioidei. The single nostril oneach side of the head will distinguish immediately the zoarceoidsfrom all tropical Blennioidei.The trachinoid-blennioid-congrogadoid series is internally diverse.It is defined more easily in terms of lack of peculiarities that thenotothenioid and zoarceoid lines have developed than in terms of itsown specializations; nevertheless, there are two weak trends ofdevelopment that may be noted for the trachinoid-blennioid-con-grogadoid series. The first trend is toward a consolidation of thebones of the circumorbital ring. Trachinus (fig. 7c) and the con-grogadoids are the only members of the Blennioidei with a welldeveloped percoid-type subocular shelf, and from here there is usuallya further fusion of circumorbital elements, rather than a disintegra-tion of the circumorbital ring that tends to occur in the zoarceoidsand notothenioids. This differentiation in circumorbitals, however,is not constant (see above). The second trend seems to be a tendencyin the Trachinidae, Uranoscopidae, Dactyloscopidae, and tropicalblennies to erect the close-set pelvics and use them as props underthe body. This trend, however, does not extend to the Leptoscopidae,Congrogadidae, and Notograptidae.If the specializations held in common by the trachinoid-blennioid-congrogadoid series are unimpressive, those that differentiate thethree components of the series are well marked. In the first place,the three groups making up the series appear to have very differentmodes of life. The trachinoids, made up of the Trachinidae, Urano-scopidae, Leptoscopidae, and Dactyloscopidae, partially bury them-selves in sand or mud (Gill, 1907) and apparently wait for or positivelyattract passing prey. Of the various morphological characteristicsrelated to this habit, only one associated with respiration need benoted here. The gill covers extend down over the branchiostegalmembranes, which are completely free from each other and from theisthmus (see p. 43). In the tropical blennies and congrogadoids, by
56 PROCEEDINGS OF THE NATIONAL MUSEUM vol. vu
contrast, the gill covers are more or less broadly attached to oneanother or to the isthmus or both, and a different method of respira-tion must be used.The tropical blennies (Blennioidae), though some members second-arily have taken up a different mode of life, are fishes that basicallyprop themselves off a hard bottom by means of one or more strength-ened pelvic rays. Though the number of pelvic rays always is reducedfrom the five usually found in the trachinids and others, the pelvics,except in secondarily pelagic forms, are never rudimentary as theyare in the congrogadoid group. Another feature found in all but themost generalized tropical blennies, i.e., the Tripterygiidae, is thatthe uppermost pectoral ray articulates with an actinost rather thanthe scapula. In this character, unique, to my knowledge, among theBlennioidei, the tropical blennies approach the batrachoid fishes(Starks, 1930). Also, the Blennioidae are the only superfamily in thesuborder in which a large anterior portion of the dorsal fin (or fins)is made up usually of spines.The congrogadids, with their allies the notograptids and possiblythe peronedyids, are enigmatic eel-like forms. In these, the front andback of the suspensorium are associated loosely. They hold with theTrachinidae, alone among the Blennioidei, a subocular shelf, but thisis a trait inherited from the percoids.An attempt to establish the most generalized, i.e., percoid-like,families among the Blennioidei leads down to the Parapercidae (noto-thenioids), on the one hand, and the Trachinidae (trachinoid-blennioid-congrogadoid series), on the other (fig. 12). Yet the percoidcharacteristics that these two families retain are rather different. Inthe parapercid genus Prolatilus, there is a percoid supraoccipitalcrest and incomplete supratemporal commissure, no strut from theparasphenoid extending up in front of the prootic, 10 abdominalvertebrae in Parapercis, five separate hypurals (counting as in Nybe-lin's 1963 system), and 15 branched caudal rays. The generalizedfeatures of Trachinus, on the other hand, are the broad subocularshelf and the toothed mesopterygoid of T. draco. Though the para-percids and trachinids already have evolved in somewhat differentdirections, a basal percoid family such as the Branchiostegidae coidd,so far as morphology is concerned, stand at the base of both. Indeed,the superficial similarities are such that it is sometimes difficult toseparate the members of the Branchiostegidae from the Parapercidae(however, see p. 43). As for the trachinids, it is not necessary to goso deeply into the percoid stock to find a fish that would provide amorphologically ancestral type. Except for certain specializations,e.g., fusion of elements in the caudal skeleton, Opistognathus or Acan-thoclinus seem to serve fairly well. These genera already have the
no. 3647 PERCIFORM FISHES?GOSLINE 57
erectile pelvic fins well ahead of the pectorals and other typical (ifnot universal) trachinoid features; however, as already suggested,there is no morphological reason why the opistognathids and acan-thoclinids, as well as the trachinoids, should not have been derivedfrom some basal percoid near the Branchiostegidae.In the section that follows, the Blennioidei, essentially the Jugularesof Jordan (1923), will be considered a suborder of the Perciformes.The reasons for this are as follows: First, the members of the Blen-nioidei form a recognizable, definable group of fishes. Second, thoughI am as dubious about a strictly monophyletic origin for the Blen-nioidei (within the limits of that suborder as herein defined) as thosewho have investigated the group before me?e.g., Kegan (1913, p.138) and Starks (1923, p. 264, ftn. 1)?it seems possible that the an-cestors of the various groups of Blennioidei lie deep in the basalpercoids not too distant from one another. Finally, those who insiston strictly monophyletic groups would be forced, I think, into thealternative of recognizing at least three and probably five separatesuborders among the Blennioidei. This possibility has been consideredand rejected. Classification of the BlennioideiIn the present section, for the sake of completeness, the classifica-tion of the suborder Blennioidei (= order Jugulares of Jordan, 1923 inpart) is carried down to family. For the contents and a definition ofthis suborder as understood here, see p. 40.Superfamily Notothenioidae (= Superfamily Notothenioidae -f-Trachinoidae, in part, of Berg, Regan, and Norman).?Head andanterior part of body usually more or less flattened. One nostril oneach side in the nototheniid fishes (sensu lato), two on each side inthe rest. Gill openings extending far forward in the Bovictidae andTrichonotidae (sensu lato), the gill membranes attached to one anotheror broadly attached to the isthmus in the rest. Branchiostegal raysseven in the Bovictidae and most Trichonotidae (sensu lato), six inthe rest. Circumorbital series of bones usually movably connected,sometimes incomplete, without a subocular shelf on the second.Front and rear portions of suspensorium firmly attached except insome Trichonotidae (sensu lato). Prootic forming a part of the in-ternal orbital border. Basiphenoid usually present.Pectoral actinosts platelike, three or four in number, the upperpectoral ray or rays articulating with the scapula. Pelvic fins, exceptin some Trichonotidae (sensu lato), with a spine and five branchedsoft rays, the interspace between pectoral bases usually broader thanthe distance across one pelvic base.The Notothenioidae are the only superfamily of the Blennioideirepresented in both tropical and cold waters. Around the Antarctic
58 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124
continent, this is the dominant group of fishes. The Notothenioidaealso are the only superfamily to contain freshwater members (Cheimar-richthys and Pseudaphritis)
.
Aside from the Gobiesocidae, Draconettidae, and Callionymidae,which herein are removed from the Perciformes entirely, the membersof the notothenioid lineage (fig. 12) seem to fall into three or fourgroups
:
At the base of the whole lineage are the two families Paraper-cidae and Cheimarrichthyidae. These retain predorsal bones and anumber of other percoid features that have been lost by the rest ofthe notothenioids and, for that matter, the other members of theBlennioidei. (Cheimarrichthys does not, however, have an orbito-sphenoid as stated by Lane, 1965).A second group is made up of the notothenioids (sensu stricto)
,
namely the Bovictidae, Nototheniidae, Harpagiferidae, Bathydra-conidae, and Channichthyidae (Norman, 1957). This group is char-acterized by the three pectoral actinosts, by a single nostril on eachside of the head, and by its primarily Antarctic distribution; however,the distinction between this and other groups is not as clear-cut asit appears from the literature. The presence of only three actinostsoccurs in the notothenioid (sensu lato) derivative Callionymidae andin the "trichonotid" Hemerocoetes, which, with other "trichonotids,"has two nostrils on each side of the head, although the first may bevery small; but the derivative Callionymidae and also Melanostigma(see under Zoarceoidae) have only one.The third group is made up of the Trichonotidae (sensu lato)(Schultz, 1960, pp. 273-277; except Cheimarrichthys, among the generaI have seen). This group contains a wide spectrum of morphologicalvariation; however, the members I have been able to examine havethe following features in common : The gill openings extend far forwardunder the throat, as in the Bovictidae among notothenioid (sensustricto) families. The branchiostegal rays are seven, except Hemero-coetes, which has six. The ascending process of the premaxillary isattached movably to the toothed portion. At least in Crystallodytes,Bembrops, and Hemerocoetes, the mesopterygoid forms a broad shelf,free posteriorly, but attached to the palatine anteriorly; the palatine,in turn, is attached movably to the pterygoid. Though these charactersare quite distinctive, Hemerocoetes with three actinosts may be inter-mediate between the Trichonotidae and the Bovictidae amongnotothenioid families.A possible fourth group is represented by Melanostigma, which(see p. 63) may prove to be merely a pelagic notothenioid (sensustricto).
no. 3647 PERCIFORM FISHES?GOSLINE 59Superfamily Trachinoidae ( =Trachinidae, Uranoscopidae, Lep-toscopidae and Dactyloscopidae).?Head compressed or rounded.Two external nostrils on each side. Gill openings extending far for-ward. Circumorbital bones firmly connected, more or less expandedonto the cheek, sometimes with a subocular shelf on the second.Medial tabular firmly attached, but not fused to parietal. Front andrear of the suspensorium firmly connected. Prootic not forming a partof the internal border of the orbit. Basisphenoid present.Pectoral actinosts four, broad or columnar, the upper pectoral rayarticulating with the scapula. Pelvic fins with a spine and five softrays (except Dactyloscopidae), the interspace between them less thanthe distance across one pelvic base (except Leptoscopidae).The trachinoids possess two additional characters in which, to myknowledge, they are unique among the suborder Blennioidei. In thepelvic girdle, the ridge on which the pelvic spine rides extends forwardinto a point. This point may lie adjacent to its fellow on the oppositeside of the midline, as in Trachinus, Leptoscopus, and Dactyloscopus,or form a more laterally located projection from the flesh, as in theUranoscopidae. The second peculiarity is that at least Trachinus andUranoscopus have a bony point extending forward from the outersurface of the posterior rim of the hyomandibular (fig. 8c). Furthersimilarities are as follows : In all four families, the scapular foramen isvery large and, except in the Leptoscopidae (Starks, 1930, p. 226),extends to the cleithrum. All four families have a low number ofabdominal vertebrae (10-12) for the Blennioidei. Certain othertendencies among the trachinoids may be associated with their habitof living in the sand or mud. One is the development, in some trachinidsand uranoscopids, of a continuity between adjacent scale edges toform ridges extending down and back across the body. Another is forthe mouth to have a fringed border. Finally, the circumorbital bonesare more or less expanded down over the cheek; armature is usuallydeveloped; and the top of the head is frequently rugose.The Trachinoidae is made up of tropical and temperate marinefishes occurring on soft bottoms in which they bury themselves up tothe eyes (Gill, 1907).On the basis of the reduction in pelvic ray number in the Dactylo-scopidae, Kegan (1912d) placed this family in a different suborderfrom the Uranoscopidae and Leptoscopidae. Starks (1923) pointed outthe artificiality of this procedure. On the other hand, Starks deniedany relationship between the Trachinidae and the "uranoscopoid"families. To me, the evidence to the contrary given above seemswholly convincing.Superfamily Congrogadoidae (=Congrogadidae, Notograptidae,and provisionally the Peronedyidae) .?Head compressed or rounded.
60 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124Two nostrils on each side. Gill openings somewhat restricted below.Circumorbital series of bones firmly connected, complete, with a sub-ocular shelf from the second. Medial extrascapular not fused to theparietal. Front and rear of the suspensorium loosely connected. Prooticforming a part of the internal orbital border or not. Basisphenoidpresent.Pectoral actinosts columnar, four in number, the upper pectoral rayarticulating with the scapula. Pelvic fins minute or absent; if present,the interspace between them less than the distance across one pelvicbase.The suborder is entirely inshore, tropical Indo-West Pacific indistribution.The families included here in the Congrogadoidae are the Congro-gadidae, Notograptidae, and very provisionally the Peronedyidae. In1952 Smith divided the Congrogadidae of Regan (1912d) into twofamilies, the Congrogadidae and Haliophidae. This seems, however,an unnecessary proliferation of families among obviously related fishes.Besides, the type of Congrogadus heirichthys and, for that matter,juveniles of C. subducens fall between the two families as Smith definesthem.So far as the congrogadids and notograptids are concerned, arelationship between the two families needs demonstration. This isby no means easy, despite the general eel-like form in both; however,both have a subocular shelf on the second suborbital bone, a featureheld in common with Trachinus and many percoids. Second, thoughthe mechanism is different in the two families, both have a suspen-sorium in which the anterior half is connected only weakly with theposterior portion. Third, the soft dorsal and anal rays show a type ofbranching that does not extend to the base but in which the posterior,but not the anterior branch, redivides. (The Peronedyidae are basedon a single Australian species I have not seen, the affinities of whichare doubtful. It will not be discussed here.)Granting a relationship between notograptids and congrogadids,the question then arises as to what the two families are in turn relatedto. Smith (1952, p. 87) suggests that the congrogadids may be aberrantpercoids. This is a distinct possibility, but Smith's further suggestionof "Spariform relations" seems most improbable. The anterior pelvicposition of Notograptus and certain congrogadid genera and the 1 :
1
relationship between dorsal and anal rays and vertebrae suggest theBlennioidei, and there seems to be no reason to deny them suchan allocation.An effort to locate possible relatives of the Congrogadoidae hasled to an investigation of certain other eel-shaped fishes. The results,though negative, may be noted briefly.
no. 3647 PERCIFORM FISHES?GOSLINE 61In Mastacembelus liberiensis (USNM 118751), there are no pelvicfins. The dorsal and anal rays are somewhat more numerous thanthe vertebrae. The structure of the trunklike snout seems to beunique in fishes. The nasal bone (Regan, 1912a, fig.) forms a long lidover the nasal cavity. It is attached tightly by ligament to theethmoid medially and along its outer surface to the lacrimal. In thecavity below the nasal bone, there is a long nasal organ of the samegross shape as that of Anguilla; however, the nasal organ ofMastacembelus is folded over on itself with the fold hinge medial.The nasal epithelium extends down from the top fold and up from thebottom one as a series of transverse leaves, and the water apparentlypasses between the two folds. The posterior nostril is just ahead of theeye, but the anterior is at the end of a tube at the front of the trunk.Just above the anterior nostril on each side is the opening to anotherlong, membranous tube that connects posteriorly with the supra-orbital sensory canal at the front of the nasal bone. The upper jaw issuspended far forward, below the rostral "trunk," from a membranousextension of the mesethmoid. The maxillary has no connection what-soever with the palatines, and neither the premaxillary nor themaxillary have the usual articular surfaces or pedicels.A fish that possibly is related more closely to the Congrogadoidaethan Mastacembelus is Alabes. In Alabes, the premaxillary pedicelsextend up under the nasal bones, as in Congrogadus, and the anteriorand posterior portions of the suspensorium are disconnected. Alabes,however, is so specialized (degenerate) as to have obscured any realevidence of relationship; Alabes has no supratemporal canal, nodorsal or anal fin rays, and no primary pectoral girdle. Under thecircumstances, it seems best to leave Alabes, at least provisionally,in the Symbranchiformes, where it usually is placed (Regan, 1912c).Superfamily Blennioidae (= Tripterygiidae, Clinidae, Chae-nopsidae and Blenniidae).?Head compressed or rounded. Twonostrils on each side. Gill openings more or less restricted below, thegill membranes attached to one another or to the isthmus. Circum-orbital bones usually firmly connected, without a subocular shelffrom the second. Medial tabular usually fused to the parietal. Frontand rear of suspensorium firmly connected. Prootic usually excludedfrom the internal orbital border. Basisphenoid present.Pectoral actinosts columnar, longer than the scapula and coracoidare broad (fig. lie), the upper pectoral ray articulating with an actinost(except Tripterygiidae) . Pelvic fins with two to four soft rays of whichthe outer are strengthened and the membrane between the raysdeeply incised (except such secondarily pelagic forms as Aspidontus,Runula, Xiphasia). Dorsal and anal soft rays usually unbranched.
62 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124An additional feature that seems to separate the Blennioidaefrom all other members of the suborder is that the members I haveexamined, at least, retain a constant relationship between the anterior-most anal pterygiophores and the first hemal spine.Members of this superfamily are abundant inhabitants of alltropical inshore areas, and some extend their ranges well into temperatewaters.The relationship of the tropical blennies to any other fish group isby no means clear.Superfamily Zoarceoidae (=Zoarceoidae+Stichaeoidae-j-Cryp-tacanthodidae of Makushok-fBathymasteridae+Zaproridae+?De-repodichthyidae+?Scytalinidae).?As Norman (1957, p. 477) in-dicates, Zoarcaeus Nilsson, 1832, appears to be the first Latinizedversion of Cuvier's (1829, p. 400) "Les Zoarces." But Zoarcaeus is anobjective synonym of Enchelyopus Gronow, an invalid name that,depending upon interpretation, may have been validated nomencla-torially by Scopoli (1777). Though the proper generic name to beused herein is by no means clear, the family group names Zoarceoidaeand Zoarcidae are available whether or not the generic name on whichthey are based is a synonym ("International Code of ZoologicalNomenclature," 1964, p. 11).Head compressed or rounded. The body is long and more or lesstapering posteriorly, with a short, usually poorly demarcated caudalpeduncle. A single nostril on each side of head. Gill openings rarely(Derepodichthys) extending far forward below the head. Medialextrascapular of the usually well developed lateral line (seismosensoryof Makushok) system fused to the parietals. Front and rear of thesuspensorium usually firmly connected (apparently weakly connectedin Ptilichthys; see Makushok 1958, p. 66, fig. 38b). Prootic excludedfrom the interior orbital rim. Basisphenoid absent.Pectoral actinosts broad, usually four in number (said to be threesometimes in Cebedichthys [Starks, 1930, p. 83] and altogether absentin Azygopsis [Makushok, 1958, p. 106, fig. 72]), the uppermost pectoralray articulating with the scapula. Pelvic fins with fewer than five softrays (except Bathymasteridae), frequently absent; if present, none ofthe soft pelvic rays are strengthened or the interradial membranesbetween them deeply incised. Interspace between pelvic fins less thanthe distance across one pelvic base.In addition to the above features, there are others common to mostor all zoarceoids that will separate them from many of the otherBlennioidei. First, the maxillary is much longer than the premaxillary,sometimes more than twice as long in such extreme instances asAnarrhichas. Second, the dorsal fin is always continuous (except forPtilichthys, which has separate spines anteriorly). Third, the entop-
no. 3647 PERCIFORM FISHES?GOSLINE 63terygoid is never, to my knowledge, more than a narrow strut, andthe metapterygoid frequently has a vertical crest along the posteriorborder of its outer face (fig. 86).The zoarceoids are one of the major marine, cold-water groups.They are found in both hemispheres but are primarily and basicallynorthern. In depth, they range from the intertidal region to the deepsea. They are generally demersal but at least Zaprora and Lycodapushave developed secondarily a pelagic habit. "Zoarces" viviparus isunusual in being a viviparous form that frequently occurs in water ofreduced salinity (Schmidt, 1917).The foregoing account has been based largely on inshore forms thatare more readily available and that have been investigated much morethoroughly (e.g., Makushok, 1958, 1961a, 1961b). These give an im-pression of homogeneity that may be belied when the more peculiarof the deep-water and pelagic "zoarceoids" have been studied moreintensively. Of those that have been reported on, Zaprora (Chapmanand Townsend, 1938) is, as McAllister and Krejsa (1961) pointed out,a not too abnormal stichaeid-like form; however, the so-called zoarcidMelanostigma, judging from Yarberry's (1965) description, givesevery indication of being a modified notothenioid and not a zoarceoidat all. Thus, a basisphenoid, unknown in zoarceoids, is present inMelanostigma. Its parasphenoid wings are low and do not extend up tothe pleurosphenoids in front of the prootic (Yarberry, 1965, p. 445,fig. 2). There are only three pectoral actionosts. Finally, Melanostigmahas seven branchiostegal rays, a number found throughout the Bovicti-dae and in most of the Trichonotidae, but only among the Anarhi-chadidae of the zoarceoids (see p. 44).Even the inshore zoarceoids, however, despite their morphologicalhomogeneity and peculiarity, have caused what would seem to be anunnecessary amount of taxonomic confusion. Regan (1912d, 1913),for example, placed the Bathymasteridae in the suborder Percoideiand then mixed the remaining zoarceoid families in with the tropicalblennies. Hubbs (1952) and Makushok (1958) rectified the lattererror. As already noted, the two groups differ significantly in skull(see fig. 10) and fin structure and even in the number of nostrils.The Bathymasteridae seem to be a perfectly good zoarceoid familyin both skeleton and soft anatomy. In two features it stands on thepercoid side of the Zoarceoidae and, hence, may be considered themost generalized family in the group. First, the pelvic fin contains aspine and five soft rays; in all other zoarceoids, the pelvic fin is reduced.Second, the ramus lateralis accessorius (Freihofer, 1963, p. 136) hasa percoid-type pattern, rather than one which is of the ophidiid-brotulid type (in Zoarcidae), or reduced (in Pholidae or Stichaeidae)
.
Rosen (in Greenwood, et al., 1966, pp. 389, 397), primarily on the
64 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124basis of Freihofer's data, has assigned the Zoarcidae to the orderGadiformes of the superorder Paracanthopterygii. Since, however,in other respects the Zoarcidae are very similar to the Bathymaster-idae, with a typical percoid accessorius nerve pattern, I prefer to viewthe peculiar accessorius configuration in the Zoarcidae as a speciali-zation within that group (as in the Brotulidae and Ophidiidae; seep. 24) rather than as an indication of relationship with the codfishes.Table 3 will serve as a summary of the classification of the suborderBlennioidei adopted here. SummaryThe higher classification of the Order Perciformes adopted here canbe summarized in synoptic form as follows.To attempt a definition of this order that would exclude the Beryci-formes and Zeiformes on the one hand and the various orders pre-sumably derived from the Perciformes on the other is almost impossible(see Norman, 1957, pp. 58, 59) ; in any event, it would require moredetail than seems warranted here.Suborder Percoidei.?(For reasons dealt with at the beginning ofthis paper, the Percoidei can be defined only in terms of central perci-form tendencies, or negatively by lacking the combination of pecu-liarities that characterize the other perciform suborders.) Pelvicbones extending between and attached by a direct articulation to thecleithra; pelvic fins usually inserted about below the pectoral bases,normally with a spine and five, but sometimes fewer, soft rays, rarelyaltogether absent; dorsal and anal soft rays generally somewhat morenumerous than the vertebrae between them. Basal counts in the Per-coidei (and Perciformes) are as follows: vertebrae 24 or 25, fre-quently more, rarely fewer; anal spines three, predorsal bones three,and branched caudal rays 15, all frequently fewer, rarely more; andbranchiostegal rays six, ranging from four to nine. (Compiled.)Superfamilies (mainly following Regan, 1913, and Norman, 1957,but modified from the preceding account) : Percoidae, Cirrhitoidae,Embiotocoidae, Pomacentroidae, Labroidae, Trichodontoidae, Ara-modytoidae, Champsodontoidae, and Chiasmodontoidae.Suborder Mugiloidei.?Pelvic bones without a cleithral articula-tion. (1) The pectoral fins are divided into two separate parts (Poly-nemoidar) ; or (2) the pelvic fins have been modified into a specializedclasping organ in the males (Phallostethoidae) ; or (3) the spinousdorsal is represented by a short fin well separated from the soft portion.(Compiled.)Superfamilies (following Myers, 1935): Polynemoidae, Mugiloidae,and Phallostethoidae.
no. 3647 PERCIFORM FISHES?GOSLINE 65Suborder Anabantoidei.?An epibranchial air-breathing organ;gas bladder extending posteriorly well beyond the body cavity; teethusually present on the parasphenoid. (Compiled.)Superfamilies : Anabantoidae, Ophicephaloidae, and Lucio-cephaloidae.Suborder Kurtoidei.?Ribs much expanded, enclosing theanterior portion of the gas bladder partially, the posterior portioncompletely; males with an occipital hook, formed by the supraoccipital,used for carrying eggs. (From de Beaufort and Chapman, 1951.)This suborder contains the single genus Kurtus.Suborder Acanthuroidei.?High-headed, compressed fishes withmore or less lunate caudal fins, the gill openings restricted below, andsmall mouths; nasal bones elongate, more or less rigidly attached tothe cranium; teeth specialized, setiform in the Zanclidae, bicuspid tomulticuspid in the rest; cleithra expanded below; additional armaturepresent in the form of (1) a spine at the corner of the mouth in juvenilezanclids, (2) one or more spines on the caudal peduncle of acanthurids,or (3) a second pelvic spine in teuthids. (Compiled.)Superfamilies: Acanthuroidae and Teuthidoidae (= Siganoidae)
.
Suborder Ophidioidei.?Pelvics, when present, consisting of oneor two filamentous rays inserted ahead of the pectoral fins; dorsaland anal fins long and low, spineless except in Gadopsis, the raysconsiderably more numerous than the vertebrae between them; oneor more of the first few ribs usually expanded. (Reworded from pre-ceding account.)The Ophidioidei generally have not been divided into separatesuperfamilies.Suborder Stromateoidei.?"Perciform fishes with toothed sac-cular outgrowths in the gullet immediately behind the last gill arch"(Haedrich, 1967a, but see also Haedrich, 1967b).Haedrich (1967a) recognizes only a single superfamily (includingthe Tetragonuridae)
.
Suborder Xiphioidei.?Large oceanic fishes with 23-26 vertebraeand the anteriormost interneurals interdigitating between the craniumand the first vertebra; pelvic fins absent or reduced to three or fewerrays; pectorals inserted low on the sides; mouth inferior except inLwoarus. (Compiled.)Superfamilies : Xiphioidae and Luvaroidae.Suborder Scombroidei.?Vertebrae 30 or more; predorsal boneslacking; postorbital members of the circumorbital series of boneseither fragmented or absent; upper jaw fixed except in Scombrolabrax.(Reworded from preceding account.)Superfamilies: Scombroidae and Trichiuroidae.
280-835?68 5
66 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124Suborder Gobioidei.?"Parietals lacking. Branchiostegals (4)5 or 6, the first one or two well separated from the others. Mesoptery-goid narrow or absent. Preopercle and symplectic widely divergentabove, with an interspace between them. Hypurals with a splint-likebone above and below" (Gosline, 1955, p. 166).The Gobioidei generally have not been divided into separate super-families.Suborder Blennioidei.?Pelvic fins, when present, inserted aheadof the pectorals; dorsal and posterior soft anal rays exactly equal innumber to the vertebrae between them; caudal fin usually rounded.(Reworded from the preceding account.)Superfamilies: Notothenioidae, Trachinoidae, Congrogadoidae,Blennioidae, and Zoarceoidae.Suborder Schindlerioidae.?Minute, transparent, neotenic,oceanic fishes with the last few vertebrae and the hypural fan fusedinto a single plate. (Compiled.)This suborder contains only the genus Schindleria.
ReferencesAlexander, R. McN.1967. Mechanisms of the jaws of some atheriniform fishes. Journ. Zool.,London, vol. 151, pp. 233-255, 10 figs.Allis, E. P., Jr.1903. The skull, and the cranial and first spinal muscles and nerves inScomber scomber. Journ. Morph., vol. 18, pp. 45-328, pis. 3-12.Arnold, D. C.1956. A systematic revision of the fishes of the teleost family Carapidae(Percomorphi, Blennioidea), with descriptions of two new species.Bull. British Mus. (Nat. Hist.), Zool., vol. 4, no. 6, pp. 247-307,20 figs.Baglioni, S.1908. Der Atmungsmechanismus der Fische. Zeitschr. Allgemeine Pbys.,vol. 7, pp. 177-282.Bardach, J. E., and Case, J.1965. Sensory capabilities of the modified fins of squirrel hake (Urophycischuss) and searobins (Prionotus carolinus and P. evolans). Copeia,1965, pp. 194-206, 10 figs.Barrington, E. J. W.1937. The structure and development of the tail in the plaice (Pleuronectesplatessa) and the cod (Gadus morrhua). Quart. Journ. Microsc.Sci., new ser., vol. 79, pp. 447-469, 25 figs.Barsukov, V. V.1959. Family Anarhichadidae. Fauna SSSR, Ryby, vol. 5, pt. 5, 173 pp.,22 pis., 42 figs.de Beaufort, L. F.1914. Die Anatomie und systematische Stellung des Genus Kurtus Bloch.Morph. Jahrb., vol. 48, pp. 391-410, pi. 12, 3 text figs.de Beaufort, L. F., and Chapman, W. M.1951. The fishes of the Indo-Australian Archipelago, vol. IX, xi+ 484 pp.,88 figs. Leiden: E. J. Brill.Berg, L. S.1940. Classification of fishes, both recent and fossil. Trav. Inst. Zool.Acad. Sci. URSS, vol. 5, pt. 2, pp. 87-517, 190 figs.Bougis, P., and Ruivo, M.1954. Recherches sur le poisson de profondeur Benthocomeles robustus(Goode et Bean) (=Pteridium armatum, Doederlein) (Brotulidae)
.
Vie et Milieu, suppl. no. 3, pp. 155-209, 33 figs.BOULENGER, G. A.1901. On the Tracbinidae and their allies. Ann. Mag. Nat. Hist., ser. 7,vol. 8, pp. 261-271.1904. Fishes (Systematic account of the Teleostei). In The Cambridgenatural history, vol. 7, pp. 541-727, figs. 325-440. 67
68 PROCEEDINGS OF THE NATIONAL MUSEUM vol. mBriggs, J. C, and Caldwell, D. K.1955. The characteristics and distribution of the spotted cusk eel Otophidiumomostigmum (Jordan and Gilbert). Quart. Journ. Florida Acad.Sci., vol. 18, no. 4, pp. 285-291.Cantwell, G. E.1964. A revision of the genus Parapercis, family Mugiloididae. PacificSci., vol. 18, pp. 239-280, 9 figs.Chapman, W. M., and Townsend, L. D.1938. The osteology of Zaprora silenus, with notes on its distribution andearly life history. Ann. Mag. Nat. Hist., ser. 11, vol. 2, pp. 89-117,10 figs.Collette, B. B., and Gibbs, R. H., Jr.1963. A preliminary review of the fishes of the family Scombridae. FAOFish. Rep., no. 6, vol. 1, pp. 23-32.Cuvier, G. L. C. F. D.1829. Le regne animal . . . , ed. 2, 5 vols. Paris.Dawson, C. E.1966. Gunterichthys longipenis, a new genus and species of ophidioid fishfrom the northern Gulf of Mexico. Proc. Biol. Soc. Washington,vol. 79, pp. 205-214, 3 figs.VAN DOBBEN, W. H.1935. Uber den Kiefermechanismus der Knochenfische. Arch. Neerland-aises Zool., vol. 2, pt. 1, pp. 1-72, 50 figs.Dollo, L.1909. Les t61eost66ns a ventrales abdominales secondares. Verh. Zool.-Bot. Ges. Wien, vol. 59, pp. 135-140.Emery, C.1880. Le specie del genere Fierasfer del Golfo di Napoli e regione limitrofe.Fauna und Flora Golfes Neapel, monogr. no. 2, 76 pp., 9 pis.Foster, N. R.1967. The utility of egg and larval characters in the classification of atherini-form fishes. American Soc. Ichth. Herp., Forty-seventh Ann.Meeting, Abstr.: Ichthy., p. 10.Francois, Y.1959. La nageoire dorsale?anatomie comparee et evolution. Ann. Biol,ser. 3, vol. 35, pp. 81-113, 18 text figs.Fraser-Brunner, A.1950. The fishes of the family Scombridae. Ann. Mag. Nat. Hist., ser. 12,vol. 3, pp. 131-163, 35 figs.Freihofer, W. C.1963. Patterns of the ramus lateralis accessorius and their systematic signif-icance in teleostean fishes. Stanford Ichth. Bull., vol. 8, no. 2,pp. 80-189, 29 figs.Gero, D. R.1952. The hydrodynamic aspects of fish propulsion. American Mus. Novit.,no. 1601, 22 pp., 21 figs.Gill, T.1863. On a remarkable new type of fishes allied to Nemophis. Ann. Lye.Nat. Hist., New York, vol. 8, pp. 138-141, 1 pi.1907 Life histories of toadfishes (batrachoidids) compared with those ofweevers (trachinids) and stargazers (uranoscopids). SmithsonianMisc. Coll., vol. 48, pt. 4, pp. 388-427, figs. 106-123.
no. 3647 PERCIFORM FISHES?GOSLINE 69Gosline, W. A.1953. Hawaiian shallow-water fishes of the family Brotulidae, with the de-scription of a new genus and notes on brotulid anatomy. Copeia,1953, no. 4, pp. 215-225, 5 figs.1955. The osteology and relationships of certain gobioid fishes, with partic-ular reference to the genera Kraemeria and Microdesmus. PacificScL, vol.9, pp. 158-170, 7 figs.1959. Four new species, a new genus, and a new suborder of Hawaiian fishes.Pacific Sci., vol. 13, pp. 67-77, 6 figs.1960. Hawaiian lava-flow fishes, pt. IV: Snyderidia canina Gilbert, withnotes on the osteology of ophidioid families. Pacific Sci., vol. 14,no. 4, pp. 373-381, 4 figs.1961a. Some osteological features of modern lower teleostean fishes. Smith-sonian Misc. Coll., vol. 43, no. 3, 42 pp., 8 figs.1961b. The perciform caudal skeleton. Copeia, 1961, pp. 265-270, 3 figs.1962. Systematic position and relationships of the percesocine fishes.Pacific Sci., vol. 16, no. 2, pp. 207-217, 3 figs.1963. Notes on the osteology and systematic position of Hypoplychusdybowskii Steindachner and other elongate perciform fishes.Pacific Sci., vol. 17, pp. 90-101, 8 figs.1964. Considerations regarding the relationships of the percopsiform,cyprinodontiform, and gadiform fishes. Occas. Pap. Mus. Zool.,Univ. Michigan, no. 629, 38 pp.1966a. The limits of the fish family Serranidae, with notes on other lowerpercoids. Proc. California Acad. Sci., ser. 4, vol. 33, no. 6, pp.91-112, 10 figs.1966b. Comments on the classification of the percoid fishes. Pacific Sci.,vol. 20, pp. 409-418, 2 figs.Greenwood, P. H.; Rosen, D. E.; Weitzman, S. H.; and Myers, G. S.1966. Phyletic studies of teleostean fishes, with a provisional classificationof living forms. Bull. American Mus. Nat. Hist., vol. 13, art. 4,pp. 341-455, pis. 21-23, 32 charts, 9 text figs.Gregory, W. K.1933. Fish skulls: A study of the evolution of natural mechanisms. Trans.American Phil. Soc, vol. 23, pt. 2, pp. vii + 75-481, 302 figs.Gregory, W. K., and Conrad, G. M.1937. The comparative osteology of the swordfish (Xiphias) and the sailfish(Isliophorus). American Mus. Novit., no. 952, 25 pp., 12 figs.1943. The osteology of Luvarus imperialis, a scombroid fish: A study inadaptive evolution. Bull. American Mus. Nat. Hist., vol. 81,pp. 225-283, 38 figs.Grey, M.1960. Description of a western Atlantic specimen of Scombrolabrax heterolepisRoule and notes on fishes of the family Gempylidae. Copeia,1960, pp. 210-215, 3 figs.Haedrich, R. L.1967a. The stromateoid fishes: Systematics and a classification. Bull.Mus. Comp. Zool., vol. 135, pp. 31-139, 56 figs.1967b. A new family of aberrant stromateoid fishes from the equatorialPacific. American Soc. Ichth. Herp., Forty-seventh Ann.Meeting, Abstr.: Ichth., p. 12.
70 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124Harris, J. E.1938. The role of the fins in the equilibrium of the swimming fish, II:The role of the pelvic fins. Journ. Experi. Biol., vol. 15, no. 1,pp. 32-47, 8 figs.1953. Fin patterns and mode of life in fishes. In Essays in marine biology,pp. 17-28, 4 figs.Herald E. S.1953. Spotted cusk-eel, the strange fish that stands on its tail. CaliforniaFish and Game, vol. 39, pp. 381-384, 3 figs.Hertel, H.1966. Structure, form and movement, 251 pp., 297 figs. New York:Reinhold Publ. Corp.Holmqvist, O.1910. Der Musculus protractor hyoidei (Geniohyoideus auct.) und derSenkungsmechanismus des Unterkiefers bei den Knochenfischen.Lund Universitetets Arsskrift, new ser., vol. 2, no. 6, 24 pp., 1 pi.Hubbs, C. L.1944. Fin structure and relationships of the phallostethid fishes. Copeia,1944, pp. 69-79.1945. Phylogenetic position of the Citharidae, a family of flatfishes. Misc.Publ. Mus. Zool. Univ. Michigan, no. 63, 38 pp., 1 fig.Hubbs, Clark1952. A contribution to the classification of the blennioid fishes of thefamily Clinidae, with a partial revision of the eastern Pacificforms. Stanford Ichth. Bull., vol. 4, no. 2, pp. 41-165, 64 figs.Jordan, D. S.1923. A classification of fishes, including families and genera as far asknown. Stanford Univ. Publ., Univ. Ser., Biol. Sci., vol. 3, no. 2,pp. 79-243.Jordan, D. S., and Evermann, B. W.1898. Fishes of North and Middle America, vol. 3. Bull. United StatesNat. Mus., no. 47, pp. xxiv+2183-3136.Kamohara, T.1935. On the Owstoniidae of Japan. Annot. Zool. Japonensis, vol. 15,pp. 130-138, 4 figs.Katayama, M.1959. Studies on the serranid fishes of Japan, 1. Bull. Fac. Educ, Yama-guchi Univ., vol. 8, pp. 103-180, 39 figs.Kishinouye, K.1923. Contributions to the comparative study of the so-called scombroidfishes. Journ. Coll. Agric, Univ. Tokyo, vol. 8, pp. 293-475,pis. 13-34, text figs.Lane, E. D.1965. The osteology of Cheimarrichthtjs fosttri Haast (Pisces, Percomorphi).Trans. Roy. Soc. New Zealand, Zool., vol. 6, pp. 207-213, 10 figs.Liem, K. F.1963. The comparative osteology and phyolgeny of the Anabantoidei(Teleostei, Pisces). Illinois Biol. Monogr. No. 30, 149 pp.,104 figs.1967. A morphological study of Luciocephalus pulcher, with notes on gularelements in other recent teleosts. Journ. Morph., vol. 121,pp. 103-133, 23 figs.
no. 3647 PERCIFORM FISHES?GOSLINE 71Linnaeus, C.1758. Systeraa naturae, ed. X, vol. 1, 824 pp., Holmiae.Makushok, V. M.1958. The morphology and classification of the northern blennioid fishes(Stichaeoidae, Blennioidei, Pisces). Trudy Zool. Inst. Akad. NaukSSSR, vol. 25, pp. 3-129, 83 figs. [In Russian, English translationmimeographed, 1959.]1961a. Some additional data on morphology of the wrymouths (Crypta-canthodidae, Blennioidei, Pisces). Trudy Inst. Okean., Akad.Nauk SSSR, vol. 43, pp. 184-197, 4 figs.1961b. Some peculiarities in the structure of the seismosensory system of thenorthern blenniids (Stichaeoidae). Trudy Inst. Okean., Akad.Nauk SSSR, vol. 43, pp. 225-269, 9 figs.Marshall, N. B.1965. Systematic and biological studies of the macrourid fishes (Anacan-thini?Teleostei). Deep-Sea Res., vol. 12, no. 3, pp. 299-322, 9 figs.Matsubara, K.1943. Studies of the scorpaenoid fishes of Japan, I. Trans. SigenkagakuKenkyusyo, no. 1, 170 pp., 66 figs.1955. Fish morphology and hierarchy, pt. I, xi-f 789 pp., 289 figs. Tokyo:Ishizaki-shoten. [In Japanese.]1963. Fishes. In Animal systematic taxonomy, vol. 9, pt. 2, Vertebrata(lb), pp. 197-531, figs. 195-657. Tokyo. [In Japanese.]Matsubara, K., and Iwai, T.1952. Studies on some Japanese fishes of the family Gempylidae. PacificSci., vol. 6, pp. 193-212, 12 figs.1958. Anatomy and relationships of the Japanese fishes of the familyGempylidae. Mem. Coll. Agric, Kyoto Univ., Fish. Ser., Spec.No., pp. 23-54, 14 figs.Matsui, T.1967. Review of the mackerel genera Scomber and Rastrelliger with de-scription of a new species of Rastrelliger. Copeia, 1967, pp. 71-83,7 figs.Mayr, E.1943. Criteria of subspecies, species, and genera in ornithology. Ann. NewYork Acad. Sci., vol. 44, pp. 133-139.McAllister, D. E., and Krejsa, R. J.1961. Placement of the prowfishes, Zaproridae, in the SuperfamilyStichaeoidae. Nat. Hist. Pap., Nat. Mus. Canada, no. 11, 4 pp.Mead, G. W. ; Bertelsen, E. ; and Cohen, D. M.1964. Reproduction among deep-sea fishes. Deep-Sea Res., vol. 11,pp. 569-596.Myers, G. S.1935. A new phallostethid fish from Palawan. Proc. Biol. Soc. Washington,vol. 48, pp. 5, 6.NlLSSON, S.1832. Prodromus ichthyologiae Scandinavicae, iv-f 124 pp. Lund.Norman, J. R.1929. The teleostean fishes of the family Chiasmodontidae. Ann. Mag.Nat. Hist., ser. 10, vol. 3, pp. 529-544, 11 figs.1934. A systematic monograph of the flatfishes (Heterosomata), vol. 1,viii-f-459 pp., 317 figs. London: British Museum.
72 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 1241957. A draft synopsis of the orders, families and genera of recent fishes andfish-like vertebrates, 649 pp. British Museum (Natural History).[Photoduplicatcd.]Nybelin, O.1963. Zur Morphologie und Terminologie des Schwanzskelettes derActinopterygier. Ark. Zool., ser. 2, vol. 15, pp. 485-516, 22 figs.Patterson, C.1964. A review of Mesozoic acanthopterygian fishes, with special referenceto the English Chalk. Phil. Trans. Roy. Soc. London, ser. B,vol. 247, pp. 313-482, pis. 2-5, 103 text figs.PFULLER, A.1914. Beitrage zur Kenntnis der Seitensinnesorgane und Kopfanatomie derMacruriden. Jenaische Zeitschr. Naturw., vol. 52, pp. 1-134, 2 pis.Quast, J. C.1965. Osteological characteristics and affinities of the hexagrammid fishes,with a synopsis. Proc. California Acad. Sci., ser. 4, vol. 31, pp.563-600, 3 figs.Regan, C. T.1903a. On the skeleton and systematic position of Lnvarus imperialis.Ann. Mag. Nat. Hist., ser. 7, vol. 11, pp. 372-374, 1 fig.1903b. On the systematic position and classification of the gadoid or ana-canthine fishes. Ann. Mag. Nat. Hist., ser. 7, vol. 11, pp. 459-466,2 figs.1909a. On the anatomy and classification of the scombroid fishes. Ann.Mag. Nat. Hist., ser. 8, vol. 3, pp. 66-75, 1 fig.1909b. The Asiatic fishes of the family Anabantidae. Proc. Zool. Soc.London, pp. 767-787.1912a. The osteology of the teleostean fishes of the order Opisthomi. Ann.Mag. Nat. Hist., ser. 8, vol. 9, pp. 217-219, 1 fig.1912b. The classification of the teleostean fishes of the order Pediculati.Ann. Mag. Nat. Hist., ser. 8, vol. 9, pp. 278-289, 6 figs.1912c. The anatomy and classification of the symbranchoid eels. Ann. Mag.Nat. Hist., ser. 8, vol. 9, pp. 387-390, pi. 9.1912d. The classification of the blennioid fishes. Ann. Mag. Nat. Hist.,ser. 8, vol. 10, pp. 265-280, 4 figs.1913. The classification of the percoid fishes. Ann. Mag. Nat. Hist., ser.8, vol. 12, pp. 111-145.1929. Fishes. In Encyclopaedia Britannica, ed. 14, vol. 9, pp. 305-329.1936. Natural history, 896 pp., illustr. London: Ward, Lock & Co.Rivas, L. R.1953. The pineal apparatus of tunas and relative scombrid fishes as a possi-ble light receptor controlling phototactic movements. Bull. Mar.Sci. Gulf and Caribbean, vol. 3, pp. 168-180, 5 figs.Rose, J. A.1961. Anatomy and sexual dimorphism of the swim bladder and vertebralcolumn in Ophidicn holbrooki (Pisces: Ophidiidae). Bull. Mar.Sci. Gulf and Caribbean, vol. 11, no. 2, pp. 280-308, 13 figs.Rosen, D. E.1964. The relationships and taxonomic position of the halfbeaks, killifishes,silversides, and their relatives. Bull. American Mus. Nat. Hist.,vol. 127, art. 5, pp. 217-268.
no. 3647 PERCIFORM FISHES?GOSLINE 73Roule, L.1922. Description de Scombrolabrax heterolepis nov. gen. nov. sp., poissonabyssal nouveau de l'lle de Madere. Bull. Inst. Oceanogr. Monaco,no. 40S, 8 pp., 4 figs.1929. Considerations sur la nature teratologique probable de quelquesformes de poissons abyssaux. Proc. 10th Intern. Zool. Congr.,Budapest, 1927, vol. 1, pp. 795-798.Schaeffer, B., and Rosen, D. E.1961. Major adaptive levels in the evolution of the actinopterygian feedingmechanism. American Zool., vol. 1, pp. 187-204, 7 figs.Schmidt, J.1917. Racial investigations, no. 1: Zoarces viviparus and local races of same.Compt. Rend. Lab. Carlsberg, vol. 13, pp. 279-396.Schultz, L. P., et al.1960. Fishes of the Marshall and Marianas Islands, vol. 2. United StatesNat. Mus. Bull. 202, ix 4- 438 pp., pis. 75-123, text figs. 91-132.Scopoli, J. A.1777. Introductio ad historiam naturalem, x 4- 506 pp. Prague.Simpson, G. G.1959. Mesozoic mammals and the polyphyletic origin of mammals. Evolu-tion, vol. 13, pp. 405-414.Smith, C. L., and Bailey, R. M.1961. Evolution of the dorsal-fin supports of percoid fishes. Pap. MichiganAcad. Sci., Arts, Letters, vol. 46, pp. 345-362, 1 pi., 1 text fig.1962. The subocular shelf of fishes. Journ. Morph., vol. 110, pp. 1-18,3 pis.Smith, J. L. B.1952. The fishes of the family Haliophidae. Ann. Mag. Nat. Hist., ser.12, vol. 5, pp. 85-101, pi. 6, 2 text figs.Springer, V. G.1955. The taxonomic status of the fishes of the genus Statmonolus, includinga review of the Atlantic species. Bull. Mar. Sci. Gulf and Carib-bean, vol. 5, no. 1, pp. 67-80, 2 figs.1964. Review of "A revised classification of the blennioid fishes of thefamily Chaenopsidae" by J. S. Stephens, Jr. Copeia, 1964, pp.591-593.1966. Medusablennius chant, a new genus and species of blennioid fish fromthe Tuamotu Archipelago: Its implication on blennioid classifica-tion. Copeia, 1966, no. 1, pp. 56-60, 3 figs.Stannitjs, H.1849. Das peripherische Nervensystem der Fische, 156 pp., 5 pis. Rostock.Starks, E. C.1907. On the relationships of the fishes of the family Siganidae. Biol.Bull., vol. 13, no. 4, pp. 211-218, 1 fig.1910. The osteology and mutual relationships of the fishes belonging to thefamily Scombridae. Journ. Morph., vol. 21, pp. 77-99, 2 pis., 2text figs.1911. Osteology of certain scombroid fishes. Leland Stanford Junior Univ.Publ., Univ. Ser., no. 5, 49 pp., 2 pis., 1 text fig.1923. The osteology and relationships of the uranoscopoid fishes. StanfordUniv. Publ., Univ. Ser., Biol. Sci., vol. 3, no. 3, pp. 259-290, 5 pis.1930. The primary shoulder girdle of bony fishes. Stanford Univ. Publ.,Univ. Ser., Biol. Sci., vol. 6, no. 2, pp. 149-239, 38 figs.
74 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124Stephens, J. S., Jr.1963. A revised classification of the blennioid fishes of the American familyChaenopsidae. Univ. California Publ. Zool., vol. 68, pp. iv+ 133,15 pis., 11 figs.Suzuki, K.1962. Anatomical and taxonomical studies of the carangid fishes of Japan.Rep. Fac. Fish., Pref. Univ. Mie, vol. 4, pp. 43-232, 61 figs.Svetovidov, A. N.1948. Gadiformes. In Fauna of the U.S.S.R., Fishes, vol. IX, no. 4. Zool.Inst. Akad. Nauk, 221 pp., 72 pis., 39 text figs. [In Russian; Englishtranslation, 1962.]Takahasi, N.1926. On the Plecostei, an order of Teleostomi, established by Prof. Kishin-ouye. Journ. Coll. Agric, Univ. Tokyo, vol. 7, no. 4, pp. 83-98.Tucker, D. W.1956. Studies on the trichiuroid fishes, 3: A preliminary revision of thefamily Trichiuridae. Bull. British Mus. (Nat. Hist.), Zool., vol. 3,pp. 73-130, pi. 10, 23 text figs.Whitley, G. P.1935. Studies in ichthyology, no. 9. Rec. Australian Mus., vol. 19, pp.217-250, pi. 18, text figs.Yarberry, E. L.1965. Osteology of a zoarcid fish Melanostigma pammelas. Copeia, 1965,pp. 442-462, 9 figs.
N0 . 3647 PERCIFORM FISHES?GOSLINE 75Table L?Families included under various classificationsRegan (1912d) Fishes that WiU be included0^?S? border B????i in the Bienmoide, hereSeries Trachiniformes xFamily TrachinidaeSeries NototheniiformesFamily Nototheniidae" Bathydraconidae
" Channichthyidae
" Bovichidae
" HarpagiferidaeScries CallionymiformesFamily Draconettidae" CallionymidaeSeries PercophidiformesFamily Percophididae" Mugiloididae
" Parapercidae
" Pteropsaridae
" Hemerocoetidae
" Chimarrichthyidae
" Creediidae
" Limnichthyidae
" Trichonotidae
" OxudercidaeSeries AmmodytiformesFamily Ammodytidae" Bleekeriidae
" HypoptychidaeSeries BathymasteriformesFamily Bathymasteridae" ZaproridaeSeries UranoscopiformesFamily Chiasmodontidae" Opisthognathidae
" Owstoniidae
" Champsodontidae
" Uranoscopidae
" Leptoscopidae
" Dactyloscopidae x
xXXXXXXXXX
76 PROCEEDINGS OF THE NATIONAL MUSEUMTable 1.
?
Families included under various classifications?ContinuedJordan (1923)Order JugulakesSeries BlenniiformesFamily Clinidae" Notograptidae" Peronedyidae
" Ophioclinidae
" Blenniidae" Emblemariidae
" Runulidae
" Atopoclinidae" Chaenopsidae" Cebedichthyidae
" Pholidae
" Xiphisteridae
" Stichaeidae" Lumpenidae" Ptilichthyidae
" Cryptacanthodidae
" Anarhichadidae" Anarrhichthyidae
" Xiphasiidae
" XenocephalidaeSeries ZoarciformesFamily Congrogadidae" Cerdalidae" Scytalinidae
" Zoarcidae
" Lycodapodidae" DerepodichthyidaeSeries BrotuliformesFamily BrotulidaeSeries OphidiiformesFamily Rhodichthyidae
" OphidiidaeSeries CarapiformesFamily CarapidaeSuborder HaplodociFamily Batrachoididae
Regan (1912d)
PERCIFORM FISHES?GOSLINE 77Table 2.
?
Basal characteristicsZOARCEOIDAETendency toward elon-gationTendency toward pelvicreduction
NOTOTHENIOIDAETendency toward flat-tening of the headTendency toward spreadof pelvics with reten-tion of 5 soft rays
Tendency toward disin-tegration of circum-orbital chain of bonesParasphenoid and fron-tals always form a stayexcluding the prooticfrom the orbital borderBasisphenoid neverpresentMedial extrascapularalways fused withcraniumPectoral actinosts usually4, broadAll cold water forms Cold water forms Tropical formsOne nostril on each side One nostril Two nos-trils
Tendency toward disin-tegration of circum-orbital chain of bonesParasphenoid and fron-tals never form a stayexcluding the prooticfrom the orbital borderBasisphenoid usuallypresentMedial extrascapularrarely fused withcraniumPectoral radials 3 or 4,broad
Trachinoid-Blennioid-Congrogadoid SeriesVarious, but the headusually not flattenedTendency to use the pel-vies as props under thebody with a strength-ening of the outer raysand incision of themembrane betweenthemTendency toward consoli-dation of circumorbitalbonesParasphenoid and frontalsusually exclude theprootic from the orbitalborderBasisphenoid usuallypresentMedial extrascapularusually fused withcraniumPectoral radials 4, vari-ously shapedAll tropical formsTwo nostrils
78 PROCEEDINGS OF THE NATIONAL MUSEUM vol. 124Table 3.
?
Suborder BlennioideiSuperfamily NotothenioidaeFamily Parapercidae (=Mugiloididae) (Cantwell, 1964)" Trichonotidae (sensu lato) (Schultz, 1960, except Cheimarrichthyidae)" Cheimarrichthyidae (Regan, 1913)
" Bovictidae (Norman, 1957)
" Nototheniidae (Norman, 1957)" Harpagiferidae (Norman, 1957)" Bathydraconidae (Norman, 1957)
" Channichthyidae (Norman, 1957)Superfamily TrachinoidaeFamily Trachinidae (Regan, 1913)" Uranoscopidae (Starks, 1923)" Leptoscopidae (Starks, 1923)
" Dactyloscopidae (Starks, 1923)Superfamily CongrogadoidaeFamily Congrogadidae (Regan, 1912d)
" Notograptidae (Regan, 1912d)? " Peronedyidae (Norman, 1957)Superfamily BlennioidaeFamily Tripterygiidae (Hubbs, 1952)
" Clinidae (Hubbs, 1952)
" Chaenopsidae (Stephens, 1963)Blenniidae (Hubbs, 1952)Superfamily ZoarceoidaeFamily Bathymasteridae (Regan, 1913)
" Stichaeidae (Makushok, 1958)
" Pholidae (Makushok, 1958)
" Anarhichadidae (Makushok, 1958; Barsukov, 1959)
" Ptilichthyidae (Makushok, 1958)
" Zaproridae (McAllister and Krejsa, 1961)
" Cryptacanthodidae (Makushok, 1961a)" Zoarcidae (=Lycodidae) (Norman, 1957, in part)? " Derepodichthyidae (Jordan and Evermann, 1898)? " Scytalinidae (Jordan and Evermann, 1898)? " Lycodapodidae (Jordan and Evermann, 1898)
U.S. GOVERNMENT PRINTING OFFICE: 1968