1 OCCASIONAL PAPERS OF THE California Academy of Sciences No. 120, 31 pages, 1 figure, 1 table THE PALATINE-MAXILLARY MECHANISM IN CATFISHES, WITH COMMENTS ON THE EVOLUTION AND ZOOGEOGRAPHY OF MODERN SILUROIDS By William A. Gosline SAN FRANCISCO PUBLISHED BY THE ACADEMY September 11, 1975 COMMITTEE ON PUBLICATION George E. Lindsay, Chairman Diana R. Young, Editor OCCASIONAL PAPERS OF THE California Academy of Sciences No. 120, 31 pages, 1 figure, 1 table THE PALATINE-MAXILLARY MECHANISM IN CATFISHES, WITH COMMENTS ON THE EVOLUTION AND ZOOGEOGRAPHY OF MODERN SILUROIDS By William A. Gosline Museum of Zoology University of Michigan Abstract: The mechanism used by catfishes for ex- tending the maxillary and its barbel was studied in various groups. The objective was to determine the probable structural pathways that have led to the different representations of this mechanism among modern siluroids. Preliminary discussions deal with the identity of certain bones in the catfish suspensorium, notably the ectopterygoid and mesopterygoid, and with the mechanics of the palatine-maxillary system. The results of the study have been interpreted as follows. The palatine-maxillary mechanism is represented in modern catfishes by two basal types: that of Diplomystes with a toothed maxillary, and that of the Bagridae, Ariidae, and several other families in which the mesopterygoid forms a movable link between the palatine and the posterior part of the suspensorium. The amblycipitids have a CALIFORNIA ACADEMY OF SCIENCES [Occ. Papers suspensorial structure that may represent an early deviation from the type found in the Bagridae, but all other catfishes investigated have palatine- maxillary mechanisms which, so far as structure is concerned, could apparently have been derived from the sort represented in the Bagridae. A direction of specialization that has been developed frequently and in various ways is that found in such hill- stream families as the Trichomycteridae , Amphiliidae, and Sisoridae. Other directions of specialization have been followed by the Chacidae, Plotosidae, Mochokidae, and Siluridae, with the palatine- maxillary mechanism of Siturus perhaps the most divergent from the basal bagrid type of all of the catfishes investigated. The implications of these results have been extrapolated into a set of working hypotheses con- cerning catfish evolution and zoogeography. It is hypothesized that primary adaptive radiations from a basal pimelodid-bagrid stock have occurred in South America and Asia, with Africa an important but secondary center of diversification. Introduction All catfishes have a palatine-maxillary mechanism (Eaton, 1948; Alexander, 1965) that extends the maxillary and its barbel, but no other group shows this specialization. A knowledge of the nature and variation of this mechanism may contribute toward our understanding of the inter- relationships of catfishes, a subject upon which records of fossil siluroids as yet throw little light (Lundberg & Case, 1970) . Modern catfishes are the endpoints of a tremendous adaptive radiation (Rossi, 1951), represented by some 2000 species (Bailey, 1971) allocated to about 31 families (Greenwood, et at., 1966; Gosline, 1971). With regard to family classification Tilak (1967b, p. 288) quotes Gar- stang's comment about a "proliferation of pigeonholes." However, Alexander ends his 1965 paper with the statement: "So much parallel evolution has occurred within the catfish that any attempt to reconstruct the phylogeny of the siob- order would seem, in the present state of our knowledge, unprofitable. " One approach to the unraveling of siluroid phylogeny is the intensive and extensive investigation of particular structures or structural systems. Units that have been previously studied in a more or less wide range of catfishes are the caudal skeleton (Lundberg & Baskin, 1969), the pelvic girdle (Shelden, 1937; Tilak, 1968), the pectoral girdle (Tilak, 1963b), the Weberian apparatus and associated structures (Bridge & Haddon , 189 3; Chranilov, 1929; Chardon, 1968), and the otoliths (Frost, 1925; Tilak, 1964b). No. 120] GOSLINE: CATFISHES The present paper deals with the palatine-maxillary mechanism. Inasmuch as this is functionally associated with feeding, it is perhaps especially subject to the parallel changes inherent in adaptive radiation. However, the classification of catfishes has always been more or less heavily based on attributes associated with the mechanism, e.g., the maxillary teeth of Diplomystes . It seems, there- fore, high time that the nature of the variation in the palatine-maxillary mechanism in catfishes be accorded more attention than it has hitherto received. Several levels of structural organization in the palatine-maxillary mechanism seem recognizable among modern catfish groups. Sometimes it is possible to follow in detail the structural transitions from one level to the next. But unless different catfish groups followed demonstrably different morphological pathways it is impossible to dis- tinguish between parallel lines of change. In the palatine- maxillary mechanism it seems that certain potentialities for structural evolution have been followed out repeatedly and, so far as I can determine, over essentially similar pathways. Thus, though it seems "unprofitable" to attempt any "phy- logeny" of palatine-maxillary mechanisms, the information to be presented below does bear rather decisively on certain aspects of catfish evolution. Methods The palatine-maxillary mechanism and structures associated with it consist of bone, cartilage, ligamentous tissue, and muscle that are not easily studied without dissection. In the present investigation, alcohol-preserved specimens provided the primary material. Two dissections starting from different points were usually made. In one, the lacrimal was folded back, the eyeball and cheek muscles were often moved out of the way, and material lateral to the palatine was cleaned off. In the other, the skin of the roof of the mouth was removed through the mouth opening. Prepared skeletons and cleared and stained specimens were examined, but both of these types of material present difficulties in the observation of cartilage, musculature, and ligamentous tissue. BONE NAMES. There are two somewhat different problems concerning bone names in catfishes: one nomenclatural , the other zoological. The zoological problem has to do with the identification of certain bones and will be dealt with at the proper points in the text. Nomenclaturally , mesopterygoid will here be used for the bone often called endopterygoid, and ectopterygoid for that called pterygoid. In catfishes there appears to be no true dermopalatine; the palatine component of the palatine-maxillary mechanism is therefore an autopalatine. CALIFORNIA ACADEMY OF SCIENCES [Occ. Papers FISH NAMES. No reclassification is suggested in this paper, which deals with a single suite of structures and not with whole fishes. The classification adopted is merely one that seems reasonable and intelligible. Except for Diplomystes , no attempt has been made to solve nomenclatural problems regarding family or other names. No effort has been made to check the identifications on the specimen labels . The Cypriniformes (= the ostariophysine fishes) is con- sidered as a single order made up of two suborders: Siluroidei (the catfishes) and Cyprinoidei (the characins, gymnotids, cyprinids, and their allies). The classification of catfish families is basically that of Regan (1911). The Doradidae as here understood includes the Auchenipterinae. The Ariidae is the Tachysuridae of authors. The classifi- cation of the Bagridae is that of Jayaram (1966). The Amblycipitidae comprises the genera Amblyaeps and Liobagrus . The Ictaluridae is the Ameiuridae or Amiuridae of authors, and the Trichomycteridae is the Pygidiidae. Nematogenys is here included in the Trichomycteridae, and the Bunocephalinae in the Aspredinidae (Myers, 1960). MATERIAL USED FOR MORE THAN SUPERFICIAL EXAMINATION. In the following list of specimens, California Academy of Sciences material is listed with CAS preceding the catalog number. All other catalog numbers refer to fishes in the collections of the University of Michigan Museum of Zoology. Diplomystidae : Diplomyste pappilosus , Santiago market, Chile, CAS 13706. (The correct spelling of this generic name is apparently Diplomystes , Bleeker's latinization of Dumeril's French Diplomyste.) Ariidae: Avius felis, Florida, 135912; A. melanopus , Guate- mala, 143459. Doradidae: Traohelyopterus coviaoeus , Bolivia, 66321; Hassar lipophthalmus , Colombia, 185333. Bagridae: Chrysiohthys auvatus , Egypt, 169013; Auohenoglanis ballayi, Cameroun, 191667; Bagrus docmao, Uganda, 187332; Bagroides melaptevus , Sumatra, 155695; Rita vita, Bang- ladesh, 187880; Mystus nemurus , Thailand, 186784. Pangasiidae: Pangasias maovonema , Thailand, 186707. Schilbeidae: Sohilbe mystus, Zambia, 189126. Clariidae: Clarias batrachus , Florida, 190122. Plotosidae: Plotosus avab , Madagascar, 185445. Chacidae: Chaaa chaca, Bangladesh, 189645. Sisoridae: Bagarius bagarius , Sumatra, 155701. Siluridae: Parasilurus asotus , Japan, 180201. Amphiliidae: Amphilius platyohir , Zambia, material on loan from the Royal Ontario Museum. Mochokidae: Synodontis nebulosus , Zambia, 189140. Ictaluridae: lotalurus punctatus , Mexico, 192471; Noturus flavus, Michigan, 56575. Pimelodidae: Pimelodus clarias, Brazil, 147401; Rhamdia guatemalensis , Guatemala, 188074; Pseudopimelodus zungaro , Bolivia, 66312. No. 120] GOSLINE: CATFISHES Trichomycteridae : Nematogenys inermis , Santiago market/ Chile, CAS 12692; Trichomyaterus vivulatus , Peru, 185311. Aspredinidae : Aspredinichthys filamentosus , Georgetown market, British Guiana, CAS 16201. Structure and Mechanics of the Palatine-Haxillary System Nasal, mental, and maxillary barbels are variously developed in catfishes, but only that on the maxillary is constant. These barbels can usually be moved by muscles extending into their bases (Singh, 1967) , but the palatine- maxillary mechanism provides another source of movement for the maxillary barbel. This mechanism has two bony parts derived from different structural systems. The siluroid palatine, more precisely autopalatine , is an endochondral mandibular-arch component of the teleostean suspensorium; the maxillary, an upper jaw element in lower teleosts, is ultimately derived from the dermal skull roof of early actinopterygians (Gregory, 1933). Most of the evolution of the maxillary part of the mechanism can apparently be followed among living catfishes. The development of its palatine part, however, seems to have been made possible by earlier changes in suspensorial structure that require some discussion. In most lower teleosts, as in Flops (Ridewood, 1904, fig. 10), the suspensorium is a single firmly-knit structural unit of composite origin, made up partly of endochondral ossifications and partly of dermal supporting elements. This suspensorium has movable abutments against the skull at its anterior (palatine) and posterior (hyomandibular) ends. Thus the mandible in such fishes is propped away from the skull mostly by the hyomandibular but partly by the auto- palatine. In catfishes, which in this respect appear to represent an endpoint in an evolutionary trend already pre- sent in ancestral cypriniform fishes (Gosline, 1973), most of the dermal elements seem to have 'come loose' from the suspensorium, and the endochondral ossifications are in two sections that are, at most, ligamentously interconnected. Anteriorly, the autopalatine has become a part of the palatine-maxillary mechanism. Posteriorly, the hyomandi- bular-quadrate region has taken over the whole propping function for the mandible. The mesopterygoid-ectopterygoid area, which in most lower teleosts forms a firm strut bet- ween the anterior and posterior parts of the suspensorium, has in catfishes lost this function, and its reduced com- ponents are variably represented or absent. The rearrangements in the posterior area of the siluroid suspensorium may be briefly summarized and dismissed from further consideration. So long as the suspensorium had a long, firm horizontal axis, it was equipped to withstand a longitudinal pull from the contracting M. adductor mandi- bulae. But in catfishes this horizontal suspensorial axis CALIFORNIA ACADEMY OF SCIENCES [Occ. Papers has been fragmented. Probably in association with this fragmentation, the main axis of contraction of the M. adductor mandibulae in catfishes has shifted to a more vertical plane (compare Takahasi, 19 25, figs. 1-10; or the insertions of the M. adductor mandibulae in Alexander's, 1965, fig. 7). In connection with the closer alignment between the M. adductor mandibulae and the hyomandibular-quadrate axis of the suspensorium, that axis has been simplified and pre- sumably strengthened. The hyomandibular of catfishes has been brought into direct contact with the quadrate; the metapterygoid has, so to speak, been extruded forward from its original position between the anterior part of the hyomandibular and the quadrate; and posteriorly the symplectic has disappeared as a separate element. The preopercle forms a strengthening strut along the posterior border of the suspensorium. (All of these features are found again in eels which, like the catfishes, have frag- mented suspensoria. ) The change toward a closer alignment between the M. adductor mandibulae and the hyomandibular-quadrate axis of the suspensoriiom has also led to certain changes in the areas of origin and insertion of the muscle (again paralleled in eels). As to insertion, the coronoid process of the cat- fish mandible is often low or absent. With regard to origin, a portion of the M. adductor mandibulae frequently extends, as in Diplomystes , onto the dorsal surface of the cranium. Whether or not such an extension occurs in catfishes appears to be associated with mandibular structure. Thus, catfishes with part of the M. adductor mandibulae originating on the top of the skull are mostly forms with relatively large, horizontal mouths, e.g., ictalurids, plotosids, some bagrids. In catfishes with small mouths, e.g., Synodontis (Stix, 1956) , the M. adductor mandibulae does not extend over the cranium. Nor does it extend on to the skull in the large- mouthed Silurus . However, in Silurus the lower jaw is oblique, and fibers of the M. adductor mandibulae are more horizontally aligned than is usual in catfishes (compare Takahasi's pi. 1, fig. 8 of Parasiturus with pi, 1, fig. 10 of Plotosus ) . In lower teleosts such as Slops (Nybelin, 1968) the dermal ectopterygoid and mesopterygoid bones have a double role as tooth-bearing plates and as a structural bridge between the anterior and posterior parts of the suspensorium. In characins and cyprinids this bridge is usually represented as a strut movable at both ends (Gosline, 1973); there are no teeth on the mesopterygoid, but teeth may be present on the characin ectopterygoid. In catfishes, these two bones are not only variable, if present, but the names applied to them in the literature are so confused that some attempt at clarification is obligatory. A true ectopterygoid does not seem to me to be identi- fiable with certainty in catfishes. Usually in teleosts the ectopterygoid adjoins the mesopterygoid for much of the No. 120] GOSLINE: CATFISHES length of both; in catfishes the only bone that ever adjoins the mesopterygoid is the metapterygoid which, in siluroids, is forward of its usual position. In catfishes there is often, as in Diplomystes (Alexander, 1965, fig. 4), a bony projection extending forward from the lateroventral part of the metapterygoid; such a projection is usually attached to the vomerine part of the skull by ligamentous tissue. Some- times, as in the bagrid Rita or on one side of a skeleton of the ariid Arius assimilis (UMMZ 190074-S) , this pro- jecting area is represented as a separate ossification. Such an ossification, when it exists, has the position of the cyprinid ectopterygoid, but it seems to have become separated from the metapterygoid, which is not a bone the cyprinid ectopterygoid could have fused with. To me a more satisfactory explanation for the ossification under con- sideration is that it represents a fragmented part of the metapterygoid. Often in catfishes there is a tooth-bearing plate on the oral surface of the ligamentous tissue mentioned above. Such plates have frequently been identified as ectoptery- goids , but again the identification seems questionable. Dentition in the roof of the catfish mouth is very variable. Teeth are usually present on the premaxillary and often on the vomer, but they may also be associated with such bones as the mesethmoid (Starks, 1926, fig. 16) and parasphenoid (Starks, 1926, fig. 12). Tooth-bearing plates may also develop in various areas of the front of the mouth roof where adjacent structural support is present. For example, dental plates may occur just lateral to the premaxillaries over the ligamentous tissue between the premaxillary and the lower jaw (Tilak, 1961, fig. 4). Functionally, such dental plates provide extensions for the premaxillary band of teeth just as dental plates on the metapterygoid-vomerine ligament provide extensions of the vomerine tooth band. Tooth-bearing plates on the metapterygoid-vomerine ligament may be separate from both the vomer and the metapterygoid (Starks, 1926, figs. 11, 12). Sometimes such plates may apparently fuse with the vomer with age (Eigenmann ^ Eigenmann, 1890, p. 64). Sometimes part or all of such a tooth plate becomes firmly attached to the metapterygoid as in the schilbeid Eutro- piichthys (Tilak, 1961, figs. 7, 8) or even fuses with the metapterygoid as in the bagrid Chvysiohthys auratus (though not in the other species of Chrysiohthys examined) . The question arises of which, if any, of these tooth plates represent the usual teleostean ectopterygoid. (Perhaps the phrase "ectopterygoid teeth" could usefully be continued as a regional designation for dentition on the roof of the cat- fish mouth without prejudice to the question of ectopterygoid homology. ) Still another element that has been identified as an ectopterygoid is the small bony ossicle embedded on the lower surface of the autopalatine of ariids (Starks, 1926, fig. 11, p2; see also Tilak, 1965a, figs. 2, 5) and in the bagrid Bagroides (see below) . Lundberg (in litt. ) states CALIFORNIA ACADEMY OF SCIENCES [Occ. Papers that such an ossicle occurs in a cleared and stained speci- men of Diplomystes , but I have been unable to find it in the unstained fish available to me. In the present paper this ossicle is provisionally identified as a fragmented part of the mesopterygoid. A final and different problem arises from the fact that the (unfragmented) siluroid mesopterygoid has frequently been misidentif led as an ectopterygoid (see below) . The mesopterygoid is frequently absent in catfishes. When present it is a toothless bone with a strong antero- medial ligamentous attachment to the lower surface of the skull just behind or beside the vomer. (The mesopterygoid is most easily located by opening the mouth and removing the skin from the mouth roof. ) The catfish mesopterygoid may extend laterally to below the posterior part of the autopalatine. When present, it is joined to the metaptery- goid behind it either directly or by ligament (except in amblycipitids) . Regan (1911) has usually, though by no means always, identified the mesopterygoid correctly; for example, he calls the mesopterygoid of pimelodids a pterygoid. The bone named ectopterygoid in most of Tilak's catfish papers is the mesopterygoid of the present paper; however, the entopterygoid of his 1961 work and of figures 2, 7, 9, and 18 of his 1964(a) paper is the mesopterygoid of this one. The use of the name mesopterygoid is also confused by Jayaram (1966 ) . The basic components of the palatine-maxillary mechanism are as follows (fig. lA) . There is always a hinge-joint between the anterior end of the autopalatine (or its carti- laginous extension) and the posterior face of the median end of the maxillary. The lateral ethmoid has a flange extending out to or over the autopalatine, which slides along and/or rocks around this flange. Movement of the autopalatine is brought about by contraction of an anterior part of the M. adductor arcus palatini (Takahasi , 1925). This muscle originates on the skull and usually inserts along the medial rim of the posterior part of the autopalatine. There is generally a nondistensible ligament extending medially from the anterior face of the median part of the maxillary to the premaxillary or sometimes {Synodontis) to the mesethmoid. The maxillary rocks around this ligamentous attachment when autopalatine movement causes lateral or posterior displace- ment of the palatine-maxillary hinge. The result is to force the distal end of the maxillary and its barbel forward (except in plotosids) . Almost all of the palatine components of the siluroid palatine-maxillary mechanism can be located in one or another of the cyprinoid fishes (though usually associated with a different type of maxillary movement) . Certain similarities between the cyprinid palatine and that of Diplomystes are striking. Thus, the cyprinid autopalatine is two-headed anteriorly, usually with one head extending medially and articulating with the skull and the other projecting forward to an abutment against the posterior face of the maxillary. No. 120] GOSLINE: CATFISHES However, in the cyprinid Ptyohooheilus , somewhat as in Diplomystes , the medial head has moved anteriorly and both heads approach the maxillary. In cyprinids the posterior part of the palatine has a movable articulation with the mesopterygoid below the later- al flange of the lateral ethmoid, and the M. adductor arcus palatini does not extend forward onto the autopalatine (Takahasi , 1925). In catfishes the posterior end of the autopalatine is free to move in a fore and aft direction; it usually extends well behind the lateral ethmoid flange and is almost always moved directly by a part of the M. adductor arcus palatini that inserts on it. I know of no cyprinoids with the autopalatine free posteriorly. However, in some cobitids, e.g., Cobitis taenia and Misgurnus anguillioaudata , Takahasi (1925, p. 26) describes the M. adductor arcus palatini as extending forward onto the posterior end of the autopalatine. In the specimen of Misgurnus fossilis dis- sected, the M. adductor arcus palatini does not extend for- ward quite as far as the autopalatine, but some of its fibers extend almost directly back from the anterior part of the mesopterygoid; contraction presumably pulls the mesopterygoid- autopalatine joint, and hence the autopalatine, backward, thus, but indirectly, having the same effect on the autopalatine as does contraction of the M. adductor arcus palatini in Diplomystes . By way of background to the functioning of the palatine- maxillary system, two matters deserve notice. The first of these is the evolution of muscular coordination in maxillary movement. Siluvus is said to be able to flutter its maxillary barbels (Juge, 1899). This implies a complete independence between maxillary movements and mouth movements. However, Siluvus, so far as its palatine-maxillary mechanism is concerned, is one of the most specialized of all catfishes, and its maxillary movements are probably also specialized. The muscular coordination of fishes with more generalized palatine-maxillary mechanisms cannot be determined from preserved specimens, but some circumstantial anatomical evi- dence is available. Though the system of retracting the extended maxillary barbel varies greatly from species to species, maxillary retraction in many catfishes appears to be at least partly coordinated with raising of the mandible. This may be brought about by ligamentous tissue of various types between the maxillary and the mandible, as in Diplomystes or Pimelodus (see below). In many catfishes a maxillary- mandibular ligament is absent, but there is a M. retractor tentaculi. This again may be of various sorts. In Plotosus the M. adductor mandibulae serves as a retractor muscle for the maxillary. Here maxillary retraction and raising of the lower jaw are presumably coordinated. Most catfishes with a M. retractor tentaculi have this muscle completely separate from the M. adductor mandibulae. But even here, at least in lotalurus , the M. retractor tentaculi "is supplied by a branch of the same nerve that supplies the deeper 10 CALIFORNIA ACADEMY OF SCIENCES [Occ. Papers portions of the add. mand. " (McMurrich, 1884, p. 314). It appears that, whatever independence of movement has been ultimately attained, retraction of the maxillary and its barbel in catfishes was originally coordinated with raising of the mandible, as in teleosts generally. Extension of the maxillary and its barbel is brought about in a more uniform manner throughout the siluroids (except plotosids), i.e., by contraction of the palatine portion of the M. adductor arcus palatini (Takahasi, 1925; Stix, 1956). However, extension of the maxillary in most catfishes does not seem to be coordinated with mandibular lowering, as is usual in teleosts, but rather with the reduction in the size of the oral cavity that results from contraction of the posterior part of the M. adductor arcus palatini. McMurrich (1884), Takahasi (1925) and others have shown not only that in catfishes the palatine and more posterior section of the M. adductor arcus palatini are parts of the same muscle, but also that they have the same innervation. Indeed, in a few catfishes, e.g., Chaca (see below) , the palatine and more posterior section of the M. adductor arcus palatini form a single continuous muscle. These items of structural evidence regarding coordination of maxillary movements in catfishes appear to lead to certain operational problems (reduction of the oral cavity presumably follows closely on raising of the mandible) that are un- doubtedly more confusing to this author than to the catfishes. Suffice it to say only that many if not all catfishes seem able to move their maxillary barbels without noticeable dis- location of their mouth structures, and that the development of a separate palatine section of the M. adductor arcus palatini is apparently the basis for doing so. With regard to more mechanical matters, the autopalatine may, as noted, slide over the lateral ethmoid flange, rock around it, or both. Judging from preserved specimens, certain criteria of effectiveness determine to at least some extent which type of autopalatine movement is adopted. The palatine part of the M. adductor arcus palatini normally extends in a medial direction from the autopalatine to the skull (fig. lA-C) . Contraction of such a muscle will tend to rock the autopalatine around the lateral ethmoid fulcrum, and the autopalatine-maxillary hinge will be displaced laterally. Insofar as there is such a rocking motion, the contraction of the M. adductor arcus palatini, the abutment of the lateral ethmoid against the autopalatine, and the arc of rotation of the maxillary and its barbel will all tend to be in the same plane because "Muscles which act obliquely generate useless forces at joints" (Alexander, 1965, p. 139). The question perhaps arises of why a sliding motion of the autopalatine over the lateral ethmoid flange is developed by catfishes at all. There seem to be two different mechani- cal reasons for this. One is exemplified by Diplomystes . The maxillary of this fish swings through an almost vertical arc. To avoid the generation of useless forces at joints, the fibers of the palatine part of the M. adductor arcus No. 120] GOSLINE: CATFISHES 11 palatini could theoretically be extended directly upward and a rocking motion of the autopalatine adopted to accom- plish maxillary extension. Actually Diplomystes has developed a different mechanical system with an autopalatine that slides and with a M. adductor arcus palatini that extends almost directly backwards from the posterior end of the autopalatine. Under such a system the palatine-maxil- lary hinge is pulled directly back, and the arc of rotation of the maxillary and its barbel would seem to be equally effective in any plane (just as the spokes of an umbrella open out equally effectively in various planes) . A different mechanical reason for a sliding autopalatine seems to apply particularly to those forms with a very wide arc of maxillary rotation. Examples are those catfishes with very long maxillary barbels which are held back along the sides (fig. lA) when the fish is at rest, but which are extended almost directly forward when the fish is hunting. In such forms the maxillary extends into the base of the barbel and hence, when the barbel is retracted, the maxillary makes an acute angle with the autopalatine. If in such fishes the anterior end of the palatine were displaced laterally, the palatine-maxillary hinge would tend to close, not open. It seems that to open such a hinge (fig. lA) the first movement of the autopalatine must be in a posterior rather than a lateral direction (which necessitates a sliding autopalatine-lateral ethmoid articulation) , although once the hinge opening becomes oblique, lateral movement of the autopalatine head may be the more effective way of swinging the maxillary anteriorly. Forms in which the resting maxillary already extends out at a relatively wide angle from the autopalatine, e.g., catfishes with the maxillary included in the gape, often are restricted in the amount of sliding autopalatine movement. This is probably most easily accomplished by reducing the amount of flexibility in the membranous attachment between the lateral ethmoid and the autopalatine. But many cat- fishes go one or more steps beyond this , presumably for mechanical reasons. If the extended or partly extended maxillary meets any force from the front, this force will push the distal part of the maxillary back and pull a sliding autopalatine forward. If, however, the lateral ethmoid flange extends into a secure socket in the side of the auto- palatine (fig. IB) , forward sliding of the autopalatine will be prevented. In short, forms with such blocks against autopalatine sliding seem better able to force forward maxillaries that form part of the upper border of the mouth, e.g., many ground-feeding catfishes with subterminal mouths. Indeed, in such fishes as Synodontis and many others (fig. IB) the forwardly moved maxillaries are used to pry the lateral ends of the premaxillaries forward, thus approaching the protrusile upper jaw of cyprinids. Such a forcing system for the maxillaries is represented in catfishes by several stages of development. The first of these, already mentioned, is that in which a more or less 12 CALIFORNIA ACADEMY OF SCIENCES [Occ. Papers centrally notched autopalatine rocks around a lateral ethmoid flange, like a teeter-totter. A further stage is represented, in amphiliids for example, where the flange projects forward behind a shelf on the autopalatine and a more backwardly directed pull on the posterior part of the autopalatine causes it to rock around the shelf (fig. IC) . Such a system requires a firmer posterior abutment for the lateral ethmoid flange than does the teeter- totter stage. However, along this line of development the postarticular part of the autopalatine may become progressively reduced in size, as in loricariids. The Palatine-FIaxillary Mechanism in Various Catfish Groups In the palatine-maxillary mechanism, as in so many other structures (see, for example, Regan, 1911; Alexander, 1965; Chardon , 1968), Diplomystes stands apart from all other living catfishes. The distal part of the maxillary of Diplomystes , like that of most cyprinids , characins , and other lower teleosts, is expanded into a flat blade which projects down over the side of the mandible. A flat, broad-based, tapering barbel extends from the end of the maxillary. The teeth on the maxillary of Diplomystes do not, as they do in characins, extend directly downward, but rather down and in. Most of the maxillary teeth, like those on the premaxillary and dentary, are long, depressible, somewhat incurved, and flat- tipped. The anterior (medial) end of the maxillary patch has the teeth about three deep, the anteriormost of which extend down and in from the lower rim of the maxillary. There is a large, roundish patch of teeth on the vomer. The vomerine teeth, like those elsewhere in the mouth, are depressible and have flattened tips, but are shorter, stouter, and more widely spaced than elsewhere. Though the toothed maxillary of Diplomystes is certainly primitive for catfishes, the depressible, flat- tipped teeth on the maxillary and elsewhere provide a highly specialized dentition. Also, Diplomystes is one of the relatively few catfishes with larger teeth on the vomer than on the jaws. This emphasis on vomerine teeth occurs again in the bagrid Rita and in the Plotosidae and is presumably associated with feeding specialization. By contrast, the usual siluroid dentition is made up of bands of small, presumably grasping teeth on various bones of the mouth roof, and when speciali- zation occurs it is usually in the jaw teeth, with a reduc- tion or total loss of dentition elsewhere. Compared with the rodlike autopalatine of so many other catfishes, that of Diplomystes has a rather complicated shape. The anterior part has a horizontally expanded, doughnutlike form with a hole passing vertically through it, at least in the rather large (180 mm. S.L.) specimen examined. The floor of the nasal cavity appears to extend down partway into the hole. The part of the autopalatine anterior to the No. 120] GOSLINE: CATFISHES 13 hole is cartilaginous, and the anterior rim of the auto- palatine is thus a transverse cartilaginous bar. The arti- cular surface of the maxillary spreads broadly across the front of this cartilaginous area in such a way as to form a broad-based autopalatine-maxillary hinge that restricts the swinging of the distal end of the maxillary to a single, almost vertical plane. Behind the doughnutlike expansion, the autopalatine slides under a lateral flange of the lateral ethmoid. In the area under which the autopalatine passes, the lateral ethmoid has a lateral projection from which membranous tissue extends to an attachment on the outer sur- face of the autopalatine. Judging from the preserved speci- men, the membranous lateral ethmoid-autopalatine attachment permits a certain amount of fore-and-aft sliding of the autopalatine under the lateral ethmoid. The articular sur- face between the autopalatine and the lateral ethmoid is somewhat oblique, so that as the autopalatine moves backward it is forced slightly downward. The autopalatine is entirely free from the posterior part of the suspensorium. Behind its sliding articulation it narrows to a strut that is continued posteriorly as a long, tapering, cartilaginous point (labeled oc by Alexander, 1965, fig. 4). The palatine portion of the M. adductor arcus palatini inserts along the cartilaginous extension and strut and extends almost straight back to an origin on the skull. When the palatine part of the M. adductor arcus palatini contracts, the forward end of the autopalatine appears to be displaced primarily backward, but also somewhat downward and a little laterally. The effect in the preserved specimen is to rock the distal end of the maxillary downward and forvard. The reason why the distal end of the maxillary swings forward when its proximal end is pulled backward is not altogether clear. I can find no particular ligamentous attachment bet- ween the maxillary and the fixed premaxillary that might serve as a fulcrum. Rather it seems that the membranous tissue between the side of the lower jaw and the inner sur- face of the mid-portion of the maxillary serves as the ful- crum around which the maxillary rocks. If the proximal part of the maxillary is pulled back with tweezers, the ligamen- tous tissue from the mandible holds the central part of the maxillary in place and its distal end rocks forward. But if, in this specimen, the autopalatine is held in position and the mandible is lowered, the distal end of the maxillary again swings forward, presumably also because of the mandible-maxillary ligamentous attachment. I can find no adductor muscle for raising the lowered maxillary and its barbel; however, raising the lowered mandible brings this about in the preserved specimen. Thus, so far as can be determined from a preserved specimen, it appears that in Diplomystes the palatine-maxillary mechanism serves as an additional system to mandibular lowering for swinging the distal end of the maxillary and its barbel downward and forward, but that retraction of the lowered maxillary is dependent on raising the mandible. 14 CALIFORNIA ACADEMY OF SCIENCES [Occ. Papers At first sight, the rather complicated shape of the autopalatine of Diplomystes seems specialized. However, this shape, as already noted, shows considerable resemblance to certain autopalatine types in cyprinids and cobitids. The question of primitiveness versus specialization in the Diplomystes palatine, therefore seems best left unanswered. Unlike the situation in many catfishes, there is, at least in the specimen of Diplomystes dissected, no mesoptery- goid link between the autopalatine and the posterior part of the suspensorium. Indeed, I can find no separate mesoptery- goid of any sort, either in the area labeled ms in Alex- ander's (1965) figure 4 or elsewhere, though Lundberg {in litt. ) reports a mesopterygoid in cleared and stained speci- mens. The posterior part of the suspensorium of Diplomystes is peculiar in several respects. The hyomandibular articu- lation with the skull is unusually long, as Alexander (1965) has noted. The metapterygoid has two well-developed ante- rior prongs. One extends medially below the cranial nerves. The other passes forward ventrolaterally to beside the autopalatine; the anterior end of this prong provides an origin for a ligament attached anteriorly to the side of the vomerine plate (and is hence reminiscent of the condition in the bagrid genus Rita). The question arises of why Diplomystes alone among living catfishes has retained a toothed maxillary of generalized lower teleostean type. A partial answer to this lies, I think, in certain differences between the bottom- feeding adaptations of Diplomystes and those of other siluroids. In other catfishes the maxillary is part of a complex barbel apparatus that moves more or less independently of the jaws. In Diplomystes the maxillary remains part of the jaw apparatus, and certain nonbarbelled sensory specializations have been developed which are apparently used in combination with the maxillary barbel in locating food. In Diplomystes tremendous cranial nerve rami pass forward into the fleshy upper lip, which is directed downward in front of the mouth. Similar rami, less extensively developed, occur in certain other catfishes, e.g., Triohomyaterus . Alexander (1965, p. 97) has stated regarding Diplomystes : "A very large foramen, between parasphenoid, pterosphenoid and prootic, corresponding to the sphenoid fissure of the chondrocranium, must have given passage to the optic, trigeminal and facial nerves. In all other catfish, the parasphenoid and ptero- sphenoid meet immediately posterior to the optic nerve." In Diplomystes the large nerve rami to the upper lip issue from this foramen, and its enlargement seems to be to accomodate them. I assume that these rami are associated with taste perception in the upper lip. If this is correct, it seems to follow that the upper lip is closely applied to the bottom during feeding. Certain other structural features in Diplomystes are in accord with such an assumption. The dentition is of a specialized, perhaps scraping type. Diplomystes , unlike most catfishes, has no mental (or nasal) barbels. Instead, the lateral line pores on the chin open No. 120] GOSLINE: CATFISHES 15 from small hiammocks in clearly distinguished bare areas. Finally, forward movement of the maxillary carries its bar- bel downward and forward to below the chin as in cyprinids , not more or less laterally as in other catfishes where mental barbels perhaps serve as bottom probes. If the maxillary part of the palatine-maxillary mechanism of Diplomystes represents a more primitive stage of struc- tural evolution than that of other catfishes, the palatine part of the same mechanism in certain other catfishes seems to be more generalized than that of Diplomystes . In cyprinids and cobitids the anterior end of the mesopterygoid has a membranous attachment to the under surface of the skull and, more laterally, a movable articulation with the posterior end of the autopalatine. In a number of catfish families a rather similar mesopterygoid arrangement occurs, and in such families the mesopterygoid seems to retain its function, considerably modified, of a movable link between the autopalatine and the posterior part of the suspensorium. The most general level of palatine-maxillary organization, other than that of Diplomystes , is here considered to be that in which the mesopterygoid forms such a link. Among catfishes with this type of mesopterygoid, the greatest variability in palatine-maxillary mechanisms is found in the Bagridae, which will be discussed first. Indeed, it seems possible, so far as structural change is concerned, to trace all of the types of catfish palatine-maxillary mechanism except that of Diplomystes and perhaps of the Amblycipitidae to one or another of the forms of this structural complex found in the Bagridae. Variation in bagrid suspensorial structure was used as a basis for classifying genera by Regan (1911) . His first character for distinguishing his two siobfamilies Bagrinae and Chrysichthyinae is a difference in mesopterygoid- pterygoid relationship. In the bagrids I have examined I have not been able to follow the distinction Regan makes, nor have I seen any bagrid that corresponds at all well with the figure (IB) Regan gives for the suspensorium of Clarotes , a genus unavailable to me. The great variability in bagrid suspensorial structure, even within a single genus (Mystus), is illustrated by Tilak (1965b). Jayaram (1966) recognizes five bagrid subfamilies. A member of each of these five subfamilies was dissected. These five species, together with material from other bagrids examined, are discussed below (table 1) . In all of the bagrids examined except Rita a mesoptery- goid is movably associated with the autopalatine laterally and, by ligament, with the metapterygoid posterior to it. Insofar as the nature of this mesopterygoid link can be shown on a flat surface, it is well illustrated in Starks ' (1926, p. 178) figure of the ariid Feliohthys . The medial end of the mesopterygoid is ligamentously attached to the skull anteriorly and to the metapterygoid posteriorly. The mesopterygoid extends distally to below the posterior end of the autopalatine. Here, it usually has an anteriorly- 16 CALIFORNIA ACADEMY OF SCIENCES [Occ. Papers directed projection which extends forward below the auto- palatine, to which it is membranous ly attached. In most bagrids the mesopterygoid is a single structure of complex shape, but in Bagvoides , as in ariids (Starks , 1926, fig. 11), it consists of two separate ossifications, one extending from the skull to the autopalatine , and the other a small ossicle firmly lodged on the lower surface of the auto- palatine. In Rita, unlike the other bagrids I have examined, there is no mesopterygoid association with the autopalatine. Hashmi (1957) and Jayaram (1966, fig. 1, msp) have recorded a mesopterygoid in this genus, but the only bone I have found that might be identified as such seems to me to be more probably a fragmented part of the metapterygoid (see above). Rita, like Diplomystes , has a vomerine tooth plate expanded into the area in which a mesopterygoid-autopalatine link would ordinarily occur. The dissociation between the mesopterygoid and the autopalatine may also be represented in Clarotes (Regan, 1911, fig. IB) , a bagrid genus I have not seen, and certainly occurs in a whole series of more advanced catfish types (see below) . In bagrids the maxillary is sometimes included in the upper lip, i.e., it forms part of the upper border of the gape, and sometimes the upper lip extends up as a separate structure to below the maxillary base so that maxillary movement is independent of changes in the shape of the mouth opening. The two bagrid genera examined in which the maxillary is included in the gape {Chrysichthys and Auchenoglanis) have certain other palatine-maxillary characteristics in common. In the first place, their maxillaries are larger than those of other bagrids (Jayaram, 1966). In the second, their autopalatines rotate around a protruding cartilaginous-tipped articular facet on the lateral face of the lateral ethmoid which blocks any sliding of the autopalatine. In certain other respects, however, the palatine-maxillary mechanisms of these two genera differ greatly from one another. At least some of these differences seem to be associated with the fact that Chrysichthys is round-headed whereas Auchenoglanis is flat-headed. Thus, the maxillary of Chrysichthys extends back over a somewhat arched gape. For- ward rotation of its distal end and barbel are consequently through a ventrolateral arc. Also, the articular facet on the lateral ethmoid is more ventrolaterally directed and the M. adductor arcus palatini pulls dorsomedially. In Auchenoglanis the maxillary extends laterally over the flat gape, the articular facet of the lateral ethmoid is laterally directed, and the M. adductor arcus palatini pulls backward. The main difference in the palatine-maxillary mechanism of Chrysichthys and Auchenoglanis , however, is that in Chrysichthys , as in most catfishes, the resting maxillary and its barbel are held back along the body and the main muscular effort seems to be exerted in the extension of these structures, whereas in Auchenoglanis the maxillary is No. 120] GOSLINE: CATFISHES 17 held out at an angle to the body and the principal muscula- ture seems to cause its retraction. The maxillary barbel structure appears to be associated with the small amount of maxillary extension in this fish; the barbel is thick-set with musculature entering its base which extends out at an angle to the maxillary bone. That this barbel can be moved forward separately from the maxillary bone is demonstrated by the position of the barbel in Sterba's (1959, fig. 494) photograph of the living fish. This relationship between the barbel and the maxillary bone of Auohenoglanis is paralleled in such other catfishes as Chaca and Plotosus . So long as the maxillary is included in the gape, its arc of movement is limited by the upper lip, but once the maxillary becomes free of the gape, the potential arc of movement is much wider. For example, in Mystus the maxillary extends into the base of a long, relatively stiff barbel that can apparently be swung through an arc of nearly 180 degrees. Other fishes with this type of barbel, and presumably Mystus, normally hold the barbel back along the sides but extend it almost directly forward when hunting. As already discussed, this type of barbel apparently requires an autopalatine that can be moved in an anteroposterior direction, i.e., one that slides over the lateral ethmoid flange. In Mystus, Bagroides , and apparently to some extent in Rita (the bagrids investigated in which the maxillary is more or less excluded from the gape) the rodlike palatine appears to be capable of some sliding movement. That retraction of the autopalatine also results in some rocking around, as well as sliding over, the lateral ethmoid flange seems to be indicated by the M. adductor arcus palatini, which extends medially into its origin on the skull. There are a number of catfish families in addition to the Bagridae in which at least some members have a mesoptery- goid link between the autopalatine and the posterior part of the suspensorium. This link is usually, as in bagrids, between the autopalatine and the metapterygoid. But in the amblycipitids Liobagvus (Regan, 1911, fig. IC) and Ambtyaeps (Tilak, 1967a, fig. 2) the mesopterygoid seems to have the usual anteromedial ligamentous attachment to the skull as in bagrids and has a lateral prong extending under the auto- palatine, but instead of a ligamentous connection between the mesopterygoid and the metapterygoid, the mesopterygoid has a long extension passing back alongside the small metapterygoid nearly to the hyomandibular. The only remote resemblance to this amblycipitid mesopterygoid that I know of in other catfishes is the bone labeled AB in Tilak' s (1963a) figure 42 of the sisorid Glyptosternum and this bone has a projection extending forward, not backward, from the lateral arm. Certain Old World catfish families have at least some members that resemble such bagrids as Mystus not only in having a mesopterygoid ligamentously attached posteriorly to the metapterygoid and extending laterally to just below the autopalatine, but also in having a rodlike sliding auto- 18 CALIFORNIA ACADEMY OF SCIENCES [Occ. Papers palatine and the maxillary excluded from the gape. Such families include the Pangasiidae (Tilak, 1964a, figs. 7, 8) and the Schilbeidae {Schilbe examined). The semimarine circumtropical family Ariidae also has most of the characteristics mentioned above but is closer to Bagvoides than to Mystus in having the mesopterygoid repre- sented by two separate parts, the anterior of which is firmly attached to the under side of the autopalatine (Starks, 1926, fig. 11; Tilak, 1965a). In the South American family Doradidae, Tvachelyopterus has all of the palatine-maxillary features of Mystus listed above. However, other members of the family, such as Hassar, have subterminal mouths with the maxillaries included in the gape, though they retain the long barlike sliding auto- palatine. The mesopterygoid condition in the South American family Pimelodidae appears to be more varied. In Pimelodus the mesopterygoid is a long curved, wirelike bone that has the usual basal ligamentous attachment to the skull; it extends around the palatine portion of the M. adductor arcus palatini to below the autopalatine. Regan (1911) records an apparently similar mesopterygoid (under the name pterygoid) from the pimelodid genera Callophysus and Sciades , as well as from Pimelodus . In Pseudopimelodus the mesopterygoid is not a wirelike ossicle extending around the back of the palatine part of the M. adductor arcus palatini but rather a broad, flat plate extending laterally below this muscle; its distal rim is just below and membranously attached to the autopalatine. In Rhamdia the mesopterygoid, as in Pseudopimelodus , is a flattened plate extending laterally, but its distal rim is well below and completely separated from the autopalatine. Thus, there is a transition within the Pimelodidae from types of mesopterygoid in Pimelodus and Pseudopimelodus , essentially similar to those in the Bagridae, to a condition in Rhamdia which closely approaches that of the mesopterygoids of such families as the Ictaluridae, Plotosidae, and Siluridae. This transition is accompanied by certain changes in palatine-maxillary musculature. In Rhamdia there is a well-developed M. retractor tentaculi originating on the metapterygoid and inserting on the poste- rior surface of the maxillary lateral to the palatine- maxillary hinge; the metapterygoid apparently serves as a fixed point toward which contraction of this muscle pulls the distal end of the maxillary. In Pimelodus there is no M. retractor tentaculi. Instead there are two ligaments inserting on the maxillary lateral to the palatine-maxillary hinge. The more lateral of the two originates on the coronoid process of the mandible; it appears to be slack except when the mouth is almost completely or completely closed. The medial of the two ligaments extends back to an origin on a point of bone projecting laterally from the metapterygoid. In Pimelodus the retraction in part, like the extension of the long maxillary barbel, appears to be coordinated with movement of the posterior part of the No. 120] GOSLINE: CATFISHES 19 suspensorium. Pseudopimelodus has the same two ligaments extending from the maxillary as does Fimelodus , but the medial of the two extends farther posteriorly than in Pime lodus . Another South American family in which a mesopterygoid link, but of a different type, appears to be present between the autopalatine and the metapterygoid is the Trichomycteri- dae. In Nematogenys the autopalatine, mesopterygoid, and metapterygoid are all on the same plane, with flat surfaces just above the fleshy roof of the mouth. The mesopterygoid is a small plate of bone between the posterior end of the autopalatine lateral to it, the skull medial to it, and the metapterygoid posteriorly. The autopalatine in Nematogenys , as in all of the trichomycterid series of families, has a firm posteromedial abutment against the lateral ethmoid flange (not shown in Regan's, 1911, fig. 2). In Tricho- myotevus the general configuration of the palatine-maxillary mechanism is about as in Nematogenys , but the mesopterygoid is absent. Chardon (1967, 1968) has established the relationship between the South American families Trichomycteridae, Aspredinidae, Callichthyidae , Astroblepidae , and Loricariidae. Aside from the Trichomycteridae, the only family in the group from which a mesopterygoid has been reported is the Aspredinidae. Regan (1911, p. 575) says: "...mesopterygoid, when present, small, attached to the lateral ethmoid." No mention is made of which aspredinids have mesopterygoids. I have dissected a specimen of Aspvedinichthys filamentosus . In this specimen there is a band of ligamentous tissue extending forward from the quadrate. In this band is a bone free at both ends from other ossifications. I interpret this bone as a metapterygoid, not a mesopterygoid; if it is a mesopterygoid, then there is no metapterygoid. The sheath in which it is embedded continues forward and divides into two parts. The larger upper part is attached to the lower surface of the autopalatine and the smaller lower part to the outer end of the movable premaxillary. For geographic reasons and because of certain superficial similarities between some members of the trichomycterid- aspredinid-callichthyid-astroblepid-loricariid series and Diplomystes , the possibility that the series evolved from a Diplomystes-like ancestor and not from the bagrid stock was considered and explored. The palatine-maxillary mechanism does not support such a possibility. However, the palatine- maxillary mechanisms of Nematogenys , Asprediniohthys , and other members of the series appear to have evolved in a way that is not closely paralleled in other catfishes I have investigated. With regard to the Chacidae Regan (1911, p. 45 7) states: "...the small mesopterygoid is attached to the lower surface of the palatine." Again I have not been able to find this bone, this time perhaps because of the small size of the specimen of Chaca dissected (78 mm.). In this specimen the metapterygoid extends forward as a flat plate under the 20 CALIFORNIA ACADEMY OF SCIENCES [Occ. Papers autopalatine (about as indicated in Regan's 1911, fig. 2D). From the upper surface of the anterior part of the metaptery- goid a short, strong ligament passes upward to the under surface of the autopalatine. The palatine and posterior parts of the M. adductor arcus palatini in Chaaa form a single continuous sheath of musculature passing medially to an origin on the skull. In the rest of the catfishes to be considered the meso- pterygoid is not associated with the autopalatine; it is either a platelike bone more or less closely attached to the front of the metapterygoid or it is absent. In most of the families to be mentioned the lateral ethmoid flange extends into a more or less we 11 -developed socket on the autopala- tine which restricts or prevents sliding of the autopalatine. The exception is the family Plotosidae. In the Plotosidae the mesopterygoid is a small platelike bone attached to the skull anteriorly and to the metaptery- goid posteriorly. The autopalatine is a long rod the move- ment of which is blocked posteriorly by a semicircular ridge rising from the upper surface of the metapterygoid. The M. adductor arcus palatini extends anteromedially from the back of the autopalatine to the skull. The maxillary is a rather large bone included in the gape, and the maxillary barbel extends out at an angle from the long axis of the maxillary. From manipulation of preserved specimens it appears that contraction of the M. adductor arcus palatini pulls the autopalatine forward, rather than backward as in other cat- fishes. Since the movable premaxillaries of Plotosus extend laterally in front of the maxillaries, it seems that forward movement of the autopalatine and palatine-maxillary hinge has the effect of forcing the distal parts of the premaxil- laries forward and downward and of rotating the large pre- maxillary teeth forward. Other catfishes use autopalatine movement as a means of moving the premaxillaries, but they do not do so by pulling the autopalatine forward. The reversal of usual maxillary movement seems to be reflected not only in the position of the maxillary barbels but in the maxillary retractor mechanism. This consists of ligamentous tissue extending from the back of the maxillary to the outer surface of the M. adductor mandibulae. Takahasi (1925) pointed out that this arrangement resembles the retractor system for the maxillary in cyprinids, which he considers a maxillary component of the M. adductor mandibulae. Because of this similarity between Plotosus and cyprinids Takahasi postulated an origin of the M. retractor tentaculi of cat- fishes in the maxillary part of the M. adductor mandibulae of cyprinids and other teleosts. Perhaps so, but the retractor system for the maxillary of Plotosus is a special case; in other catfishes with a M. retractor tentaculi, this muscle extends deep to and is separate from the M. adductor mandibulae (McMurrich, 1884). The family Ictaluridae has usually been placed with the Bagridae and Pimelodidae, as by Regan (1911). In general appearance the palatine-maxillary mechanism of ictalurids No. 120] GOSLINE: CATFISHES 21 supports such a placement, but the mechanism differs in particular features from that of either family. The plate- like mesopterygoid ahead of the metapterygoid and separate from the autopalatine in ictalurids is essentially of the type found in such pimelodids as Rhamdia. The differenti- ation of this type of mesopterygoid from that usually found in bagrids has occurred within the Pimelodidae (see above) and may have evolved repeatedly, for it is present in a number of catfish families. The maxillary of ictalurids, unlike that of pimelodids, is included in the gape as in the bagrids Chrysiohthys and Auohenoglanis , and, as in those two genera, the lateral ethmoid flange of ictalurids extends into a rounded socket on the autopalatine. In its major features the palatine-maxillary mechanism of Clavias is essentially similar to that of ictalurids. Certain differences are probably associated with the flat- headedness of Clavias . In such fishes there seems to be a tendency to move the corners of the mouth anteriorly so that the arched gape of high-headed forms becomes transverse in flat-headed catfishes. Such a change is often accompanied by enlargement of the laterally extended maxillary, as in Auohenoglanis and Chaca. In Clavias this development seems to have been carried a step farther by incorporating the corner of the mouth into the laterally extending maxillary barbel, for the lower lip is attached laterally to the lower surface of the barbel. The Sisoridae and Amphiliidae are primarily hill-stream catfishes of Africa and Asia respectively. As might be expected, both have subterminal mouths with the maxillaries included in the gape. However, the palatine part of the palatine-maxillary mechanism is quite different in the two families. In the Sisoridae (Tilak, 1963a; Mahajan, 1966), as in the Clariidae (Burne, 1909, p. 624, fig. 196A) and to a lesser extent in the Ictaluridae, the autopalatine is a rodlike structure which rocks around a barlike extension of the lateral ethmoid in teeter-totter fashion (fig. IB) . This type of autopalatine mechanism appears to be fore- shadowed in the Asiatic bagrid genus Leiocassis , which has an essentially similar autopalatine-lateral ethmoid arti- culation. In the African Amphiliidae (Harry, 1953) , as in the Malapteruridae (Burne, 1909, p. 629, fig. 197) and Mochokidae (see belov;) , the autopalatine has at most a short, tapering projection behind the lateral ethmoid articulation (fig. IC) . This autopalatine type is essentially that of the African bagrid genera Chrysiohthys and Auohenoglanis . The Mochokidae and Siluridae are the only other families that will be mentioned. Both of these families have highly specialized palatine-maxillary mechanisms and both include forms in which the palatine has moved entirely to the front of the lateral ethmoid flange (as in fig. ID) . In other respects the palatine-maxillary mechanisms of the two families are wholly different. In Silurus (Jobert, 1872), the maxillary rocks around a ligamentous attachment to the firmly fixed premaxillaries (as in fig. lA) , and there is a 2 2 CALIFORNIA ACADEMY OF SCIENCES [Occ. Papers sliding articulation between the short, nodular palatine and the ethmoid region of the skull. In the Mochokidae, the palatine rocks around the ethmoid flange and maxillary extension moves the premaxillary forward (fig. IC) . Also, in at least the mochokids Euchiliichthys (fig. ID) and Synodontis there is a maxillary extensor muscle between the anterior tip of the palatine and the front surface of the maxillary that I have not noted elsewhere in catfishes. Conclusions Regarding the Palatine-Maxillary Mechanism Only the basic features of the evolution of the palatine- maxillary mechanism in modern catfishes will be discussed here. A toothed, distally expanded maxillary and a mesoptery- goid link between the autopalatine and the posterior part of the suspensorium have been postulated as characters of cer- tain modern catfishes that have been inherited from an ancestral siluroid stock. These two characteristics do not occur in any one catfish today. Diplomystes is the only modern form with a toothed maxillary. It also has a number of other characteristics, some presumably primitive and others apparently specialized, that do not occur in other siluroids. Diplomystes is therefore considered as a relict form well separated from the main stem of catfish evolution. A mesopterygoid link between the autopalatine and the posterior part of the suspensorium occurs in a number of catfish families. It appears in two basal forms. One is represented only in the Amblycipitidae , in which the meso- pterygoid passes back alongside the small metapterygoid to or almost to the hyomandibular. This type of link may represent a specialization from the form that occurs else- where, but it may be a separate development. In all other catfishes with a mesopterygoid, that bone is attached by ligament or directly to the anterior end of the metaptery- goid. Of the modern catfish families with a mesopterygoid link present between the autopalatine and the metapterygoid, the Bagridae shows the greatest palatine-maxillary variation. This variation is divisible into three basic categories. In one, represented only by Rita, expansion of the vomerine tooth plate appears to have led to the loss of the meso- pterygoid link (a similar loss for a perhaps similar reason has occurred in Diplomystes) . A second type of bagrid palatine-maxillary structure occurs in all of the Asiatic bagrids I have examined (except Rita) and in the African Porous (or Bagvus) . Here, the maxillary is free from the gape and the autopalatine is a rod-shaped bone that usually slides over the lateral ethmoid flange. Other families with a sliding autopalatine in which at least some members have the maxillary free from the gape and a mesopterygoid link between the autopalatine and the metapterygoid are the Pimelodidae, Doradidae, Ariidae, I No. 120] GOSLINE: CATFISHES 23 Schilbeidae, and Pangasiidae. This group of families, to- gether with the Bagridae, contains to my knowledge all those catfishes with the maxillary extending into a long, rela- tively stiff barbel that can be rotated through an arc of nearly 180 degrees. So far as change from ancestral teleosts is concerned this barbel type represents about the greatest differentiation that occurs in catfishes. Yet it is present in a group of families placed {Diplomystes aside) at the base of the modern siluroid series, whatever point of view is considered. There seems no reason to believe that the type of palatine-maxillary mechanism represented in these families has been developed more than once. The other main type of palatine-maxillary mechanism in bagrids is that of the African genera Chrysiohthys and Auchenoglanis . In these the mesopterygoid link is present as usual in bagrids, the maxillary is included in the gape, and the autopalatine is propped more or less securely against the lateral ethmoid in a socketed articulation. A rocking articulation of a different type seems to be in the process of development in the Asiatic bagrid Leiooassis , and one of still another sort is present in the South American trichomycterid series of catfishes. It seems clear that a rocking autopalatine articulation has developed a number of times in catfishes and that, at least in Leiooassis , this development can be traced back to forms with sliding auto- palatines of the Bagrus type. Though the source of the trichomycterid palatine-maxillary mechanism is not clear, at least to me, there seems no reason to believe that it, or the palatine-maxillary mechanism of any other modern catfish, has evolved directly from that of Diplomystes . Evolution and Zoogeography of Hodern Siluroids In this section the palatine-maxillary mechanism will be used as a primary basis for suggesting some working hypo- theses concerning the evolution and zoogeography of modern catfish groups. Within the ostariophysine fishes, which are undoubtedly of monophyletic origin, it is generally agreed that the earliest split is that between the catfishes on the one hand and the characins , gymnotids , cyprinids, and their allies on the other. The differences in distribution patterns between these two major divisions (Siluroidei and Cyprinoidei) have been the subject of considerable discussion, most recently by Gosline (1975). Because of a possible relevance to what will be said below, the distribution of the Cyprinoidei will be briefly stated. The gymnotoid families are restricted to South America (with the exception of certain extensions into Central America) . The characins occur today in South (and Central) America and Africa, though fossil characin teeth have been recently recorded from France (Cappetta, et al . , 1975). The cyprinids and their allies are most diversified in eastern Asia, but also occur in Europe, North and Middle 2 4 CALIFORNIA ACADEMY OF SCIENCES [Occ. Papers America, and Africa. Unlike the Cyprinoidei, the Siluroidei is not divisible into diversified, distinct sections. Rather, the main division lies between Diplomystes , a relict South American genus, on the one hand and all the rest of the siluroids on the other. The question arises of how to divide the non- Dip Z-omi/stes catfishes for purposes of further analysis. The following discussion adopts one possible basis (for a very different one, see Chardon, 1968). For reasons that have been discussed in the body of the paper, those catfishes with a mesopterygoid link between the metapterygoid and a bar-shaped, sliding palatine, the maxillary usually free from the lip, and often with long, stiff maxillary barbels will be considered closest to the basic stock of modern siluroids so far as palatine-maxillary structure is concerned. Families in which at least some members have these characteristics are: Ariidae, Doradidae, Pimelodidae, Bagridae, Schilbeidae, and Pangasiidae. The Schilbeidae and Pangasiidae are generally thought to be bagrid derivatives, e.g., by Chardon (1968), and will not be considered further. Of the other families, the Ariidae is circumtropical with mostly marine and estuarine forms; the rest are almost completely freshwater catfish groups. The doradids and pimelodids are restricted to South and Central America, and the bagrids are found in Asia and Africa. All of the families mentioned above contain fishes that live primarily in large, unobstructed waters. Gosline (1973) on anatomical grounds suggested that the ostariophysine fishes originated as small, upland stream forms, but that the catfishes may have become the first of the ostariophysines to have developed large-river and thence estuarine types (Gosline, 1975; see also Rossi, 1951). If the above hypotheses are correct, a relatively early catfish dispersal across lowland and/or marine barriers impassable at that time to the Cyprinoidei may in part explain why the basal pimelodid-bagrid group of catfishes is represented today from South America through Asia whereas none of the groups of the Cyprinoidei have so wide an intercontinental range. In any event, the siluroids are the only ostario- physine group that has developed fully adapted and well- differentiated marine families (Ariidae and Plotosidae) . To return to the ariid-doradid-pimelodid-bagrid series, the Ariidae not only is a family of circumtropical dis- tribution made up mostly of marine or semi-marine forms, but is recorded as far back in the fossil record (Eocene) as any catfish family. Nevertheless, in a number of char- acters outside the palatine-maxillary system the ariids are highly specialized. Thus, ariids are the only catfishes other than doradids with long backward extensions of the epiotics (see, for example, Chardon, 1968). It therefore seems highly improbable that any catfish group, other than perhaps the Doradidae, has been derived from the Ariidae. So far as relationships are concerned, the Doradidae is an enigmatic family with some specialized characters held No. 120] GOSLINE: CATFISHES 25 in common with the ariids (but perhaps developed indepen- dently) and other features that suggest a pimelodid deri- vation. The two basal groups that remain to be discussed are the South American pimelodids and the Old World bagrids. Though the palatine-maxillary data provide weak supporting evidence at best, they in no way contradict Regan's (1911) view that the various South American catfish families other than the Diplomystidae , Ariidae, and Doradidae have been derived from the Pimelodidae. In the Old World the palatine-maxillary data suggest that the Bagridae is divisible into two primary groups: a basal Bagrinae that includes all of the Asiatic forms plus the single African genus Bagrus (or Porous), and a deriv- ative African Chrysichthyinae. The Chrysichthyinae could well have given rise to the African endemic families Mochokidae, Amphiliidae, and Malapteruridae so far as pala- tine-maxillary structures are concerned. The Bagrinae, in regard to these same structures, could have provided the base not only for the Eurasian catfish families, except possibly the Amblycipitidae , but for the North American Ictaluridae as well. If these conclusions concerning catfish evolution are correct, they give rise to the zoogeographic difficulty of having two primary centers of catfish diversification (South- east Asia and South America) separated by a continent (Africa) that is a secondary center of catfish evolution. The fact that the Cyprinoidei can well be interpreted as having the same two primary centers of diversification (see, for example, Banarescu, 1971) does nothing to remove the incon- venience of Africa's geographic position for such interpre- tations . It seems well to stress once again the provisional nature of the working hypotheses presented in this discussion. The zoogeographical problem of Africa raised in the last para- graph, for example, may well turn out to be a 'pseudo- problem. ' If it really exists, it was already solved in pre-Eocene times because, as Regan (1922) pointed out, the present intercontinental distribution of the major ostario- physine groups was already pretty well established in the Eocene. Acknowledgments For the loan of certain specimens, notably of Diplomystes and Nematogenys , I am very grateful to Dr. William N. Eschmeyer of the California Academy of Sciences. Donald J. Stewart of the University of Michigan has been good enough to obtain on exchange specimens of Euchiliichthys . I sincerely thank Dr. Michel Chardon of the University of Li^ge and Dr. John L. Lundberg of Duke University for their comments on the manuscript. 26 CALIFORNIA ACADEMY OF SCIENCES [Occ. Papers References Cited ALEXANDER, 1965. BAILEY, R. 1971. BANARESCU, 1971. BRIDGE, T, 1893. function in the catfish, vol. 148, pp. 82-152, 19 Journal figs . BURNE, R. 1909. H. R. McN. Structure and of Zoology, M. Pisces (Zoology). In McGraw Hill Encyclopedia of Science and Technology, 3rd ed. , vol. 10, pp. 281, 282. P. Competition and its bearing on the fresh-water faunas. Revue Romaine de Biologie, Serie de Zoologie, vol. 16, pp. 153-164, 4 figs. W. , and A. C. HADDON Contributions to the anatomy of fishes. II. The air-bladder and Weberian ossicles in the silu- roid fishes. Philosophical Transactions of the Royal Society of London, ser. B, vol. 184, pp. 65-333, pis. The anatomy of the olfactory organ of teleostean fishes. Proceedings of the Zoological Society of London, May-Dec. 1909, pp. 610-663, figs 188-213. CAPPETTA, H., D. E. RUSSELL, and In press. Sur la decouverte Cypriniformes) dans CHARDON, M. J. BRAILLON de Characidae (Pisces, 1' Eocene inferieur Francais. 1967. 1968. CHRANILOV, 1929. EATON, T. 1948. EIGENMANN, 1890. Reconnaissance d'un groupe naturel de six families de siluriformes sud-americains grace a 1' etude anatomique de I'appareil de Weber au sens large. Annales de la Societe Royale Zoologique de Belgique, vol. 97, pp. 35-58, 7 figs. Anatomie comparee de I'appareil de Weber et des structures connexes chez les siluriformes. Annales du Musee Royal de I'Afrique Centrale, ser. in 8°, no. 169, 277 pp., 3 pis., 205 text figs. N. S. Beitrage zur Kenntnis des Weber' schen Apparates der Ostariophysi. 2. Der Weber' sche Apparate bei Siluroidei. Zoologische Jahrbticher 51, pp. 323-462. vol (Anatomie) , H. Form and function in the head of the channel cat- fish, Ictalurus lacustris punctatus. Journal of Morphology, vol. 83, pp. 181-194, 6 figs. C. H. , and R. S. EIGENMANN A revision of the South American Nematognathi or cat-fishes. Occasional Papers of the California Academy of Sciences, no. 1, 508 pp., 1 map, 57 text figs. No. 120] GOSLINE: CATFISHES 27 FROST, G. A. 19 25. A comparative morphology of the otoliths of the neopterygian fishes (continued): II. Ostario- physi, B. Siluroidae. Annals and Magazine of Natural History, ser. 9, vol. 16, pp. 433-446. GOSLINE, W. A. 1971. Functional morphology and classification of Teleostean fishes. University of Hawaii Press, Honolulu: 208 pp., 29 figs. 19 73. Considerations regarding the phylogeny of cyprini- form fishes. Copeia, 1973, pp. 761-776, 1 fig. 19 74. A reexamination of the similarities between the freshwater fishes of Africa and South America. In press. A reexamination of the similarities between the freshwater fishes of Africa and South America. Museum National d'Histoire Naturelle, Paris. GREENWOOD, P. H., D. E. ROSEN, S. H. WEITZMAN, and G. S. MYERS 1966. Phyletic studies of teleostean fishes, with a provisional classification of living forms. Bulletin of the American Museum of Natural History, vol. 131, pp. 345-455, pis. 21-23, 9 text figs. GREGORY, W. K. 19 33. Fish skulls; a study of the evolution of natural mechanisms. Transactions of the American Philosophical Society, new ser., vol. 23, pp. 75-481, 302 figs. HARRY, R. R. 1953. A HASHMI, T 1957. JAYARAM, K, 1966. JOBERT, C 1872. JUGE, M. 1899 contribution to the classification of the African catfishes of the family Amphiliidae , with description of collections from Cameroon. Revue de Zoologie et de Botanique Africaines, vol. 47, pp. 177-232, 11 figs. A. The skeleton of Rita vita (Hamilton) (Teleostei, Siluridae) . Biologia, Lahore, vol. 3, pp. 73- 121, 37 figs. C. Contributions to the study of the fishes of the family Bagridae . 2. A systematic account of the African genera with a new classification of the family. Bulletin de I'Institut Fondamental d'Afrique Noire, ser. A., vol. 28, pp. 1064- 1139, 4 figs. Etudes d'anatomie comparee sur les organes de toucher chez divers mammiferes, oiseaux, poissons et insectes. Annales des Sciences Naturelles, Zoologie, ser. 5, vol. 16, art. 5, 162 pp. , 8 figs . Recherches sur les nerfs cerebraux et la muscu- lature cephalique de Silurus glanis. Revue Suisse de Zoologie, vol. 6, pp. 1-171, 3 pis. 28 CALIFORNIA ACADEMY OF SCIENCES [Occ. Papers LUNDBERG, 1969. LUNDBERG, 1970. McMURRICH 1884. MAHAJAN, 1966. MYERS, G. 1960. Proceedings of the vol. 2, pp. 311- NYBELIN, 1968. O REGAN, C. 1911. 1922. RIDEWOOD, 1904. ROBERTS, ' 1972. ROSSI, L. 1951. SHELDEN, 1937, J. G. , and J. N. BASKIN The caudal skeleton of the catfishes, order Si luri formes. American Museum Novitates no. 2398, 49 pp. , 9 figs. J. G. , and G. R. CASE A new catfish from the Eocene Green River For- mation, Wyoming. Journal of Paleontology, vol 44, pp. 451-457, pis. 81, 82. , J. P. The miology of Amiurus oatus . Canadian Institute, Toronto, 351, pi. 3. Sisor rabdophopus - a study in adaptation and natural relationship. I. The head skeleton. Journal of Zoology, vol. 149, pp. 365-393, 24 figs. S. The genera and ecological geography of the South American banjo catfishes. Family Aspredinidae. Stanford Ichthyological Bulletin, vol. 7, pp. 132-139. The dentition in the mouth cavity of Flops, In T. 0rvig, ed. , Current Problems of Lower Verte- brate Phylogeny, Nobel Symposium 4, pp. 439- 443, 3 figs. T. The classification of the teleostean fishes of the Order Ostariophysi. II. Siluroidea. Annals and Magazine of Natural History, ser. 8, vol. 8, pp. 553-557, 3 figs. The distribution of the fishes of the Order Ostariophysi. Bijdragen tot de Dierkunde, vol. 22, pp. 203-208. W. G. On the cranial osteology of the fishes of the families Elopidae and Albulidae, with remarks on the morphology of the skull in the lower teleostean fishes generally. Proceedings of the Zoological Society of London, 1904, pt. 2, pp. 35-81, figs. 8-18. [". Ecology of fishes in the Amazon and Congo basins. Bulletin of the Museum of Comparative Zoology, vol. 143, pp. 117-146. Radiazione adattativa e distribuzione geographica dei pesci nematognati. Bolletino di Zoologia, vol. 18, pp. 235-244, 3 figs. ?. F. Osteology, myology and probable evolution of the nematognath pelvic girdle. Annals of the New York Academy of Sciences, vol. 37, pp. 1-96, 62 figs. No. 120] GOSLINE: CATFISHES 29 SINGH, B. 1967. R. fishes. 402-412, STARKS, E. 1926. STERBA, G. 1959. STIX, W. 1957. TAKAHASI, 1925. TILAK, R. 1961. N, 1963a. 19 6 3b. 1964a, 1964b. 1965a. 1965b. Movements of barbels of some siluroid Zoologischer Anzeiger, vol. 178, pp 11 figs. C. Bones of the ethmoid region of the fish skull. Stanford University Publications, University Series, Biological Sciences, vol. 4, pp. 139- 338, 58 figs. Stisswasserf ische aus aller Welt. Verlag Zimmer und Herzog, Berchtesgaden, 638 pp., 1193 figs. Vergleichende Untersuchungen an der Trigeminusmu- skulatur der Siluridae (Teleostei) . Gegenbaurs Morphologisches Jahrbuch, vol. 97, pp. 45-76, 18 figs. On the homology of the cranial muscles of the cypriniform fishes. Journal of Morphology, vol. 40, pp. 1-109, 3 pis., 16 text figs. The osteocranium and the Weberian apparatus of Eutropiiohthys vacha (Ham. ) and Eutropiichthys murius (Ham. ) : a study of inter-relationship. Zoologischer Anzeiger, vol. 167, pp. 413-430, 19 figs. The osteocraniiam and the Weberian apparatus of the fishes of the Family Sisoridae (Siluroidea) : a study in adaptation and taxonomy. Zeitschrift fUr Wissenschaftliche Zoologie, vol. 168, pp. 281-320, 75 figs. Studies on the nematognathine pectoral girdle in relation to taxonomy. Annals and Magazine of Natural History, ser. 13, vol. 6, pp. 145-155, 23 figs. The osteocranium and the Weberian apparatus of the fishes of the Family Schilbeidae (Pisces: Siluroidea) . Proceedings of the Zoological Society of London, vol. 143, pp. 1-36, 85 figs. Studies on the comparative morphology of the otoliths of Indian siluroids. Zoologischer Anzeiger, vol. 173, pp. 181-201, 10 figs. The comparative morphology of the osteocranium and the Weberian apparatus of Tachysuridae (Pisces: Siluroidei) . Journal of Zoology, vol. 146, pp. 150-174, 77 figs. The osteocranium and the Weberian apparatus of the fishes of the Family Bagridae (Pisces: Siluroidei). Morphologisches Jahrbuch, vol. 107, pp. 415-443, 48 figs. 30 CALIFORNIA ACADEMY OF SCIENCES [Occ. Papers 1967a. The osteocranium and the Weberian apparatus of Amblyceps mangois (Hamilton) (Pisces: Siluroidei) in relation to taxonomy. Zoologischer Anzeiger, vol. 178, pp. 61-74, 5 figs. 1967b. Studies on the osteocranium and the Weberian apparatus of Indian siluroids in relation to taxonomy. Bulletin of the National Institute of Sciences of India, no. 34, pp. 288-295. 1968. Studies on the osteology of the nematognathine girdle in relation to taxonomy. Journal of the Zoological Society of India, vol. 19, pp. 101- 110, 30 figs. [ I bone rotating articulation FIGURE 1. Diagrammatic representation of different types of palatine-maxillary mechanisms in catfishes (dis- cussed in text) . The paired dashed line on the left side of A indicates the maxillary barbel, which is not shown else- where. A: sliding type of palatine-skull articulation; B-D: different types of rocking articulations. In A the premaxillary is firmly united to the skull and the maxillary rocks around the maxillary-premaxillary ligament; in B and C the premaxillary is membranously attached to the skull and its lateral end moves with the maxillary. In A and B the retracted condition of the maxillary (with its barbel) is shown on the left side and the extended condition on the right. In D the premaxillary is not indicated. le , Lateral ethmoid; Imp, maxillary-premaxillary ligament; map, M. adductor arcus palatini; mrt, M. retractor tentaculi; mx, maxillary; pal, autopalatine; and pmx, premaxillary. No. 12 0] GOSLINE: CATFISHES 31 (U 4J c u ■H >0 •^^ -ri -l "SOW -O m O nS >, ra ^3 O 0113T3 (D CP ^ 0-^ )-IM(0rHO3 <» 5'^ (DTI C nJ (-1(0 0(UOa'0>iCr' « ™? ''3 3 ^ ■'^ >-i iM ••^ii coi cp .-) o a3^io KCntoOJ-a C 10 U 1010103 ei03.p3.H-H (-1 (u ^a tjcr -Sffl-w-UrHrHTD 0) -P j:;ija)3 0 tolOOlO-H +) 01 4-1iOji;4JO era a ^^ +^ O C •fi a) « iH CO H « S, 10 I 0) O -ri-ri e c s T) o g S>-u E-HC ■:-> O 0) u O i 04-ix;a)4Ji/ia)ra o) ^c-m ™fi«' ec-PCG-H3> MC -HSiO K -Psi j3-H-H(l)C0rs;rH (U-H -U-r-iM •'^4^'!^ 4JC4-'>-Ha)+>iB Ti4J M-a ■*i O+i djoioifl &.« -H C 10 10 i-vi Q) TJ 10 iO(UQ)Q,io>j^4tog -da io-h cu yi 'a TS a u u ■!-.>; ov4ii)C)s>o cuo Xj3 01 US< 3 O TI +JH03.PiO-P^M 03 tirH-H •S .«; O C O lOaoiiOiOrHOOO OiO •nt-i'i ■w os^.HgH H -a Si 3 .c +i 0) 0) «> « 3 OH ■p oj +^ H 0) la e ID -^ o n H •^ C S, X O 9) 10 U -H 0) +J 4J .^ o -P a Id T3 to -H ■»'«>, Cn-P rH 0) ojca-^ajH cc o g 4j e ccTJoiH-HO) M 0 c CCfrHTD-P -P m OJ <0 3S-II0-HXC U 10 u GiogHajo) o X! ■'^ U r-{ U) U) > T) Ifl 01 s^ -<_ o' -p-po a-HTSci -o •H O 01 >1 u 0) ■M a 10 4J U •o fl 0) H p 3 •H ja 0) § -H -p -d 10 c u >i fu T3 H e 10 H 0 D 10 >i rr M s: 3 T) -P 0 C 3 -p 10 tn c c -o 4-1 W 0 0 •H U m •H H 0 10 1-1 0 4J JJ g a, 0 S-l 10 10 s: C H p 0 H IH H 4-) 0 0) Ti H dj 0 0) 0 3 0) ■H x: -H 3 C 3 ■H c >-l 0 4-> +J 0 a-H T3 c H a 0) ■H H •H E 4-1 T3 ■H T3 " — ■ C -H N 4-1 10 in 0 x; M-l 10 10 p •H -H -H •H 1-1 1-1 0 P 4-1 0 rH 10 0 0) W >i 10 OJ a 0) 10 . H 01 Ol- >, H 1-1 ■p 0) c as 10 >, io C V4 OJ 0) 10 0) 10 (1) > H 0 Oj M x: 0 10 a > H c H C -H 10 ■ri d) d) (U in H rH a ■H H -H H P ^4 p x; x; (11 4J p -H 3 -P ■H p £ 4-1 10 (1) 0 4-14-1 3 a T3 0 H 10 X 10 4J 10 H 4-1 0) o 0 10 0) X 0 H 10 fH ■H H QJ 10 IH IM 1-1 14 01 (1) w 10 -P 01 g 10 3 10 1-1 H H 0 0 10 cu a: s a a, a. Q 2 U)H I'^Gl ^