(Family Mullidae)!

ewe

Structure, Function, and Ecology in the Goatfishes (Family Mullidae)!

WILLIAM A. GOSLINE2

ABSTRACT: The Mullidae differ from other percoid families in a number of structural features. The most notable ofthese is a pair ofhighly developed hyoid barbels used in feeding. Many ofthe other goatfish specializations seem to be as sociated in one way or another with the use made of these barbels. Structural peculiarities of the Mullidae are described and their functional and ecological im plications suggested where possible. The hypothesis is made that the goatfishes have evolved a distincti ve ecological niche for themselves based on their use of the barbels in hunting.

MOST OF TH E50 ORSOfamilies of percoid fishes Upeneus tragula, shown in Figure 2, use their are separated from one another by minor barbels as an excavating device (see also the structural features (Regan 1913). The goat comments of Hiatt and Strasburg [1960: 84] fishes (Family Mullidae) differ dra stically concerning Parupeneus cyclostomus). In from this general pattern. They have developed Hawaii Parupeneus cyclostomus (formerly P. a structural peculiarity unique among per chryserydros) pokes its barbels into holes and coids, the hyoid barbels, and a number of other crevices in hard substrates to dislodge small specializations, several of which appear to be animals (Hobson 1974). Mulloides fiavo functionally or structurally associated with the lineatus (Mulloidichthys samoensis auctorum) barbels.The adaptations of the Mullid ae seem may back off from an item discovered by the to have been rather successful, for goatfishes barbel s (Figure 1) and " blow" away sand to are represented circumtropically in inshore uncover it (pers. obs.); the same species may areas , often by large numbers of individual s. also burrow into the sand with its snout to re Within the family, there appears to have been trieve a food item (Hobson, pers. comm .). a minor adaptive radiation based on the ex Mullus surmuletus on occasion ploughs up the ploitation of the barbels for various types of mud with its snout in feeding and apparently feeding because the principal differences be expels water from its mouth while doing so tween the six otherwise quite similar genera of (LoBianco 1907). These feeding methods are the family are in dentition. only used by half-grown and adult goatfishes , A pair of well-developed barbels is present for all mullids so far as known have pelagic in all adult goatfishes. The two members of eggs, larvae, and juveniles (Caldwell 1962). each pair are independently movable and bear Because of the use that goatfishes make of numerou s sense organs (Sato 1937). When a their barbels and functionally associated char goatfish is not feeding, the barbels are folded acters, the interrelationship between struc back under the rims of the chin and gill covers. ture, function, and ecology in the Mullid ae is In feeding, the barbels are lowered to touch particularly close and is stressed here. On the the bottom under the fish. Probably all goat other hand, the distinctiveness of the family fishes use their barbels to detect food items on seems to have obscured evidence for relation or slightly below the surface of the substrate ship between the Mullidae and other percoids, (Figu re 1). When an area with such items is and an attempt to determine these relation located , different goatfishes adopt various ships has not been very successful. In the pre methods for obtaining the prey. Some, such as sent pap er the various peculiarities of the goa tfishes are described beginning with the barbels. These peculiarities are compared with I Manuscript accepted I July 1984. 2 University of Michigan, Museum of Zoology, Ann similar developments elsewhere in teleosts Arbor , Michigan 48109. and where possible are interpreted in terms of 312 Structure, Function, and Ecology in Goatfishes-GosLINE 313

FIGURE 1. Mulloidesflavolineatus probing for food (Kona, Hawaii). Photograph by E. S. Hobson.

FIGUR E 2. Upeneus tragula using its barbels as an excavating tool (Ujung offPandang, Sulawesi, formerly Makassar, Celebes) . Photograph by J. E. Randall. function. Comments on mullid genera, family Parupeneus should be synonymized with the relationships, and ecology conclude the paper. closely related Pseudupeneus is left open . The material on which the article is based consists of partly dissected and whole speci mens ofall six mullid genera in the University THE BARBEL APPARATUS of Michigan collections. The scientific names used in the paper are There are two mechanically separate com those suggested by Randall (pers. comm., ponents to barbel manipulation in goatfishes March 4, 1984). The name changes of concern (Figure 3 A, B). One has to do with the move here are as follows: Parupeneus chryserydros is ments of the barbels (A, AI' and A z) relative placed in the synonymy ofP. cyclostomus and to the hyoid bars with which they articulate. Upeneus arge in that of U. taeniopterus; Mul Each of the two barbels has its own muscula loides is considered a valid generic name, re ture that controls such movements, thus en placing the more widely used substitute Mul abling the two to move independently. The loidichthys. The question ofwhether the genus second mechanical component (B, BI , and 314 PACIFIC SCIENCE, Volume 38, October 1984

---

FIGURE 3. Diagram to illustrate goatfish barbel mechanics, Lateral view from right side. Stippled areas represent bone. Arrows indicate direction of movement. A, barbel with the ligaments that (A I) retract and (At) lower it. B, hyoid bar with the urohyal (B I ) that lowers and the M. prot ractor hyoideus (Bt ) that raises it.

Bz) causes the lowering and raising of the whereas contraction ofthe muscle to the lower anterior ends ofthe pair of hyoid bars relative ligament (A 1) retracts the barbel. The muscles to the head of the fish. to both ligaments appear to be modified parts The barbel itselfconsists ofa fleshy, sensory of the hyohyoideus system (Winterbottom part that encloses a long, tapering, distally 1974a). flexible branchiostegal ray. A large nerve en The second component of the barbel ap ters the base of the fleshy part and innervates paratus governs the movement of the anterior the numerous sense organs on the surface of ends of the hyoid bars. Generally in teleosts the barbel. The modified branchiostegal ray the hyoid bars are movable at both ends. (La Bianco 1907) provides a strengthening Anteriorly they are connected to the sym support for the barbel. It originates in a cap of physeal area of the chin by musculature (the fibro-cartilage which forms a socket over the M. protractor hyoideus of Winterbottom tip of a forward projection from the anterior 1974a; the M. genioh yoideus of Osse 1969); ceratohyal of the hyoid bar (Figure 4). Two posteriorly to the inner surface of the sus ligaments insert on the cap of fibro-cartilage pensorium by the intercalation of a movable at the base of the barbel, one above the interhyal bone. In the Mullidae the freedom of ceratohyal projection and one below. These movement ofthe anterior ends ofthe hyoid of ligaments lead back into musculature the hyoid bars has been increased and that of originating along the hyoid bar (not shown in their posterior ends reduced to rotation in a Figure 3). Contraction of the muscle to the firm socket. upper ligament (Figure 3A z) lowers the barbel The structural changes associated with the .,Ui P lIP! rnzmr

Structure, Function, and Ecology in Goatfishes-e-Gosuxs 315

Teleosts have evolved ventrally directed projections ofvarious types, usually or always equipped with sense organs. Such projections range from barbels on the chin to extensions ofthe pectoral rays. However, sensory probes Uh Lh developed in association with branchiostegal A rays are rare and occur, to my knowledge, Ac only in the Mullidae and in the beryciform genus Polymixia. That the hyoid barbels in these two taxa have evolved independently is clear from a number of structural differences between them. Furthermore, the barbel ap B2Hy B1 Uh Lh GI Ac paratus of the Mullidae is much more spe cialized than that of Polymixia. In Polymixia three modified branchiostegal rays support the barbel rather than the one ray in the Mullidae (Starks 1904). In Poly mixia the hyoid bar has the usual movable attachment at both ends rather than the mod ifications described above in the Mullidae. There is also a considerable difference be tween the two taxa in the basal articulation of the barbel and in the bones at the anterior end of the hyoid bar. FIGURE 4. Bones associated with the forward end of In Polymixia the hypohyal bones hold their the hyoid apparatus in Pseudupeneus macula/us. A, lateral usual position, one under the other, at the and B, dorsal view. Only the basal part of the branchi ostegal ray is shown. Ac, anterior ceratohyal; BI, first anterior end of the hyoid bar, and the three basibranchial; B2, second basibranchial; Br, branchios branchiostegal rays supporting the barbel tegal; GI , glossohyal; Hy, first hypobranchial; Lh, lower have simply moved forward to an articulation hypoh yal; and Uh, upper hypohyaI. with the under surface of the lower hypohyal (Starks 1904: fig. 4). In the larval goatfish hyoid bar movements of goatfishes are as fol Mullus surmuletus the branchiostegal rays lows. The posterior end of each hyoid bar is have their normal serial arrangement restricted to a socket formed primarily by a (LoBianco 1907), but during growth the ante semicircular, medianly projecting flange on rior branchiostegal moves forward along the the inner surface of the interopercle (Figures hyoid bar and becomes widely separated from 3, 5). In front of this flange, the posterior end the other branchiostegal rays. This forward of the bar is propped from above by the inter movement of the anterior branchiostegal ray hyal bone. Anteriorly, the forward end of the is apparently accompanied by the develop pair of hyoid bars is connected with the chin ment of a forward projection from the lower only by flexible membrane. The protractor part ofthe anterior ceratohyal, because in the hyoideus muscle that is usually in this area has adult goatfish, this projection extends anterior moved its posterior attachment ' well back to the two hypohyals which themselves have

along the shaft ofthe hyoid bar (Figure 3B 2 ) . become horizontally aligned (Figure 4; see It still serves to raise the hyoid bar, but the also Starks 1904: fig. 5). The result of these posterior displacement of its attachment per changes is that in goatfishes the barbel articu mits considerably greater movement of the lation has moved forward to about the anterior ends ofthe bars than in other teleosts. level of the tip of the glossohyal (concealed The cause of such lowering is, as usual , con under the membrane of the floor of the traction of the sternohyoideus muscles acting mouth) and forward of the hypohyals (Fig ) . via the urohyal bone (Figure 3B 1 ure 4). 316 PACIFIC SCIENCE, Volume 38, October 1984

---+--t--0 p

---,~--\--,t----+--p 0

/I..;.-----,,~---+------j'-- Ih

-+---80

Qu 8y Hy 10

F IGURE 5. Internal view of the suspensorium of Mullus barbatus with the posterior end of the hyoid bar and the interhyal included . A w , part of the A w section of the M. adducto r mandi bulae; ca, cartilage; Ec, ectopterygoid; Hm , hyomandibular; Hy, posterior part of the hyoid bar; Ih, interhyal; 10, interopercle; Ms , mesopterygoid; M t, metaptery goid; Op, opercle; Pa, palatine; Po, preope rcle; Qu, quadrate; So, subopercle; st, strut on the mesopterygoid that abuts against the back of the lateral ethmoid; and Sy, symplectic.

There are two points to be made about the THE LENGTHENIN G OF THESNOUT REGION forward articulation of the barbel in goat fishes. First, the greater the distance of the ar The snout length in various goatfishes is ticulation from the fulcrum at the posterior quite different , ranging from short and blunt end of the hyoid bar, the greater the amount in Mullus to con siderably extended in Paru the barbel will be lowered when the bar itself is peneus and Pseudupeneus. Nevertheless two rotated through a given angle. Presumably, osteological features present in all goatfishes then, the extent to which the hyoid bar can suggest that the preorbital part of the head has move the barbel do wnward has been in been lengthened in the mullid ancestral stock. creased by the forward displacement of the One is in the superficial circumorbital series of barbel articulation. The second point con bon es. Usually in teleosts, the anteriormost cerns the possible structural relationship be member of the series, the lacrimal, articulates tween the anterior elongation ofthe hyoid bar with the lateral ethmoid in fron t of the orbit, and the lengthening of the preorbital part of and the infraorbital bone behind it is well the head that appears to have occurred in the separated from the lateral ethmoid. In the Mullidae. In goatfishes the snout region above Mullidae, the lacrimal seems to have shifted the mouth has apparently become extended as forward, for it has only a slight contact with have the hyoid bars below the mouth, and the the lateral ethmoid, wherea s the infraorbital possibility seems high that the two features are behind it has the main contact with the lateral related. ethmoid (Smith and Bailey 1962). Structure, Function, and Ecology in Goatfishes-GosLINE 317

The other apparently forward shift in the region appear to be related to their bottom anterior head structures occurs in the sus feeding. pensorium. Normally, the anteriormost bone The lacrimal bone, as in many bottom in the suspensorium, the palatine, provides feeding percoids, (e.g., the lutjanoid-sparoid the only articulation between the forward part haemuloid series, Johnson 1980), is extended of the suspensorium and the skull. As with downward so as to overlap a large part of the the circumorbital series, in the Mullidae a maxillary shaft. Also, as in the same and other bone behind the palatine, the mesopterygoid, percoid groups, the originally movable nasal has developed an articulation with the skull bone has in goatfishes a rigid posterior attach (Figure 5). ment to the skull. LoBianco (1907) suggests The apparent forward dislocation of the that in Mullus this increased rigidity of the lacrimal and palatine bones in the Mullidae is anterior part ofthe skull is associated with the almost exactly duplicated in another very habit in this genus of ploughing up the bottom different group of percoids, the Pentacerotidae with its snout. (Quinquarius examined). Some pentacerotids, Goatfishes have gone one step further along like mullids, also have barbels, but those of this same line of development than many the pentacerotids are on the chin , not the fishes with rigidly attached nasal bones. They hyoid bars. have lost the usual ethmoid-maxillary liga In Quinquarius the anterior (palatine) part ment and have functionally replaced it with a of the suspensorium appears to have a ligament that passes anteriorly from the normal, firm attachment to the rest of the lateral border of the nasal bone over the pala suspensorium, and the whole compound tine prong to an attachment on the forward structure seems to move as a single unit. In the rim of the maxillary (Figure 6A). This same Mullidae, however, the upper part of the substitution of ligaments occurs elsewhere in palatine is joined to the mesopterygoid strut percoids in such fishes as the pentacerotid behind it by a flexible area of cartilage (Figure Quinquarius and in lethrinids (pers. obs. ; see 5), providing ajoint between the anterior and also Johnson 1980). posterior components of the suspensorium. In goatfishes the premaxillae protrude in a Such a joint is common in percoids. Further more or less downward direction, and a num more, in the Mullidae the M. adductor arcus ber of specializations seem to be associated palatini, or a more or less separate section of with this. The anterior face ofthe vomer has a it, extends forward along the inner surface of raised, median, cartilaginous ridge (Starks the palatine. I have not seen such a forward 1926) over which the ascending premaxillary extension elsewhere. It may have to do with processes slide. There is a similar ridge in, for the expulsion ofajet of water from the mouth example, Menticirrhus, a sciaenid with a in these fishes. downwardly projecting mouth opening. In goatfishes the shaft of the maxilla is about equally deep throughout, the primordial liga ment does not pass across it to an attachment OTHER STRU CTURES ASSOCIATED on its outer rim, and the palatine prong ex WITH THE MOUTH tends down over the maxilla and has a more or In addition to the goatfish specializations less expanded tip (Figure 6A). These features already discussed, there are a number ofother are all present in the sparoid genus Lethrinus. peculiarities associated with the mouth that Other peculiarities associated with the range in distribution from one which is widely mouth ofgoatfishes cannot readily, if at all, be held among percoid fishes to at least one fea functionally associated with the downward ture that appears to be entirely restricted to protrusion of the premaxillae. The most dis the Mullidae. tinctive of these is what will here be called a Adult goatfishes feed on benthic animals supramaxillary scale (Figure 6A). Like the and their mouths open forward and down supramaxillary bone of many lower percoids ward . Some ofthe specializations in the snout this specialized scale, present in all Mullidae, 318 PACIFIC SCIENCE, Volume 38, October 1984

+- nm --\-\-..--- Po

--~~-'--- Mx

- __-:::j",-....,...... :r-----,--7'------sm

Ay---lr-

~_~=::::..---____:;~=------Md- ~ rb

__--rm

Mx----_+_

Md----1~

ph cp

.FIGURE 6. The adductor mandibulae muscle and topographically associated features in Upeneus moluccensis. A,

externa l and B, interna l view. A I'A 3 , and A w ' three of the sections of the M. adductor mandibula e; Ax and Ay , the two parts of the usual A 2 section; ap, aponeurosis; Md, mandible; Mx, maxilla; nm, nasal-maxillary ligament; Pa, palatine; ph, upper end of the M. protractor hyoideus; rb, lower branch of the ramu s mandibularis nerve; rm, ramus mandibu laris; and sm, supramaxillary scale.

extends upward from the distal end of the particularly large supramaxillary scale, there maxillary bone. It frequently takes on some of ·are circuli around its anterior surface . the characteristics of the distal end of the Usually in goatfishes there are a number of maxillary bone (e.g., sculpturing). Further scales on the maxilla, but in some species of more , it may underlie other, more normally Upeneus only the supramaxillary scale ap developed scales. It differs from the supra pears to be present. I do not know ofa scale of maxillary bone in that its base overlaps the this type elsewhere, butin a number of percoids maxilla rather than riding on its upper rim. with a scaleless maxilla there is a cartilagelike Furthermore, in Upeneus parvus, which has a upward extension along the dorsal rim of the Structure, Function, and Ecology in Goatfishes-GosLINE 319

distal end of the maxilla (e.g., in the sciaenid (Johnson 1980), but I do not know of such Menticirrhus). an attachment elsewhere. The configuration of the adductor mandi bulae muscle, though inadequately investi gated in percoids up to now, can be a good THE SPH ENOTI C PROJECTION systematic character (see, for example, John Goatfishes all have one peculiarity in the son 1980); it certainly is within the Mullidae. bony postorbital rim. In percoids with more As in so many small-mouthed higher teleosts or less cylindrical heads (e.g., Perea , Osse (see, for example, Winterbottom 1974b) the 1969), the M. levator arcus palatini originates . M. adductor mandibulae of all goatfishes is on the under surface of the postorbital part of complex (Figure 6). In Upeneus molueeensis, the skull. In fishes with high, compressed which seems to represent the basal goatfish heads this muscle usually originates on the condition in this feature, there are five more or posteroventral surface of a lateral projection less separate sections of this muscle passing from the sphenotic which, with the frontal between the suspensorium arid the jaws. The bone above it, forms a continuous rim around uppermost section in external view (Figure the posterodorsal part of the eye. In goatfishes 6A) is undoubtedly an extended maxillaris there is a gap, crossed externally by the der (A 1) component. It originates on the posterior mosphenotic, between the sphenotic projec part of the suspensorium and passes forward tion and the frontal borderofthe orbit above to an insertion on the inner surface of the it. In some goatfishes (e.g., Mullus, Mulloides, maxilla just behind the proximal head of that Upeneiehthys) the M. dilatatlor operculi

bone. The usual A 2 component of the M. ad passes forward through this gap to an attach ductor mandibulae appears to be represented ment on the inner surface ofthe orbital wall. A by two sections (Figure 6 Ax, A y). The upper of similar gap in the postorbital rim occurs in the two in external view passes forward to an some sparoid fishes (e.g., Calamus, Archosar insertion along the inside of the mandible gus) but by no means all. The functional sig

above A w (Figure 6B). The lower of the A 2 nificance ofthis gap is unclear, and its system sections in external view almost completely atic significance, if any, has been too little underlies the upper. It passes forward into the investigated to determine. aponeurosis with A w and then is continued forward by A on the inside of the mandible. w STRUCTURES ASSOCIAT ED WITH TH E FINS The mandibular ramus of nerve V passes for ward medial to A 1 and the upper section of Goatfishes always have two, well-separated , A 2 then forward into the mandible between dorsal fins. They also have a single, minute the two sections of A 2 , sometimes , as in anal spine and a well-developed scaly process Figure 6A, with a branch that passes down in the axil of the pelvic fin. These features , held across the base of the lower section of A 2 ' A 3 in common with other percoids, mayor may is the usual small muscle visible only in inter not have value as indicators of relationship. nal view. In addition, A w , normally limited to In two other features associated with the the mandible, has a part attached by ligament fins, the goatfishes are highly peculiar. The to the inner surface of the quadrate bone of scapula of the pectoral girdle contains either the suspensorium (Figure 5). two or three foramina (Starks 1930). Most Pseudupeneus (see Smith and Bailey 1962) other teleosts have only one (Starks, 1930, and Parupeneus differ notably from other record s two in cirrhitids) through which a goatfishes in that most or all of the A 1 section major nerve passes (see, for example, Frei ofthe adductor mandibulae is attached to the hofer 1963). Examination of the scapular fora under surface of the infraorbital bones. Such mina of goatfishes indicates that the anterior an attachment also occurs in the percoid most corresponds to that of other fishes, and genus Lutjanus (Smith and Bailey 1962) that the other one or two transmit nothing. and certain other members of the Lutjanidae Their significance is unclear.

I I j ."....,-,----",,,,,..,.:-:=:======::;:-;;= = = :::== = = =:-:======..:-:...:::...c::_.-:::__=. _ .~" ."" 320 PACIFIC SCIENCE, Volume 38, October 1984

Goatfishes have 13 instead of the usual 15 all. The teeth on the palatines, on the other branched caudal rays ofpercoids. A reduction hand, are always well represented. Further in branched ray count occurs sporadically more, the palatine patch seems always to be elsewhere in higher teleosts, for example in continued back on to the ectopterygoid to a anabantoids and Scatophagus. More notable greater or lesser extent. is the pattern of fusion in the caudal skeleton. Mullus is usually described as lacking teeth Externally the caudal fin ofgoatfishes is essen in the upper jaw. This, as Caldwell (1962) tially symmetrical but the caudal skeleton is notes, is only true of adults. The vomer of more than usually asymmetrical (see figures in Mullus is expanded into a broad, medially Monod 1968). The hypurals to the upper grooved plate bearing pebblelike teeth. caudal lobe are fused to one another and to Though the palatines are sometimes said to the urostylar process, but the hypurals to the provide part of the patch of teeth on the roof lower lobe remain separate from one another of the mouth (e.g., Lachner 1960), they are and from the urostyle. I have not seen this type toothless and quite separate from the vomer. of fusion in the caudal skeleton elsewhere. Lachner (1960: 3)states that Pseudupeneus The reasons behind the differential fusion in differs from Parupeneus in the presence on the the upper and lower caudal lobes of goatfishes upper jaw of a second, outer row of one to can only be a matter for speculation. How several teeth. These may be somewhat rever ever, the following possibility seems worth ted and tusklike. What apparently happens suggesting. Goatfishes have the forked tails of here is that the teeth in the single row of strong swimmers, which they undoubtedly juveniles develop irregularly as the fish grows, are. Both lobes of the caudal fin are swung for I can find no second row of teeth in avail from side to side during rapid locomotion, able specimens smaller than 164mm in stan and the fusion in the parts of the caudal skele dard length . Specimens available seem to have ton associated with the upper caudal lobe may a single row of teeth in the lower jaw, not two increase the strength of the tail beat. However, or more as stated by Lachner (1960). during hunting (Figure 1) or feeding (Figure Two other aspects of differentiation be 2) it is necessary to adjust the head precisely tween goatfish genera deserve comment. with reference to the bottom. The relative Upeneichthys, by comparison with other flexibility of the lower caudal lobe may be genera , appears to have emphasized percep used for this purpose by enabling differential tion by the lateralis system, as the old name movements between the upper and lower porosus applied to one of the species indicates. caudal lobes during slow forward movements In U. lineatus the whole snout region is or for lowering or raising the head during covered by a series of moderately large pores feeding. and lacks the embedded scales of other goat fish genera . Parupeneus and Pseudupeneus, two closely related genera (or subgenera), differ more COMMENTS ON GENERA radically from other goatfishes and appear to Dentition is the principal character used for be the most specialized genera in the family. differentiating goatfish genera (see, for The following specializations may all be as example, Lachner 1960). Certain rectifica sociated with their relatively elongate snouts. tions in descriptions of this feature may be In addition to the attachment ofthe A 1 section noted here. ofM. adductor mandibulae to the infraorbital Upeneus has the most extensive dentition bones noted above , they have the epaxial body among goatfishes and one that is of a gen musculature extending farther forward over eralized percoid type. The genus is said to the top of the head than in other goatfishes have teeth on the jaws, vomer, and palatines (except possibly for some species of Upeneus) (e.g., Lachner 1954).Though there are norm and have lost the superficial connection be ally teeth on the vomer, they may be few. In a tween the preopercular and temporallateralis specimen of U. sundaicus I cannot find any at canals present in other mullids. Structure, Function, and Ecology in Goatfishcs-e-Gostrxs 321

FAMILY RELATIONSHIPS and all sparoids, or even with the most gen eralized sparoids, but rather with now one The systematic position of the Mullidae segment and now another of the sparoid among percoid families is obscure. As already group (Johnson 1980). To start with the least noted, several goatfish specializations appear diagnostic, the features held in common are: to be unique among percoids. One of these, 1. a well-developed scaly process in the axil the increase in the number of scapular fora of the pelvic fin; mina, seems to offer no indication ofrelation 2. 24 vertebrae; ship. Two others suggest something about 3. two widely separated nostrils on each mullid ancestry by implication; Morphologi side ofthe head with the posterior one a cally, the hyoid barbels have no counterpart, valved slit near the orbit as in such rudimentary or well-developed, elsewhere sparids as Calamus; among percoids, but their position and the use 4. the lacrimal bone broadly overlapping made of them suggest that they originated in a the maxilla; bottom-feeding fish. The supramaxillary scale 5. the ethmoid-maxillary ligament func again seems to have no counterpart elsewhere, tionally replaced by a ligament from the but its position suggests a substitute develop nasal bone to the maxilla as in many ment for the supramaxillary bone of many sparoids; lower percoids and lower teleosts generally . If 6. the maxillary blade ofabout equal width this interpretation is correct, two implications for most of its length and overlapped are at least worth considering: that goatfishes anteriorly by a palatine prong that tends evolved from a stock that (1) had lost its to be expanded at its tip as in the lethri supramaxillary bone and (2) had scales on the nid sparoids; maxilla. 7. a gap between the sphenotic projection Other goatfish specializations, or even and the frontal border of the orbit above suites of them , seem to have evolved inde it as in such sparid genera as Calamus; pendently in other percoid groups. This is true and of the attachment of the A 1 section of the 8. the distally expanded symplectic. M. adductor mandibulae to the infraorbital bones in certain mullids and lutjanids. It also There are of course numerous differences appears to be true ofthe several similarities in between the mullids and the sparoids. Perhaps the anterior part of the head between mullids particularly significant among these, so far as and the pentacerotid genus Quinquarius: the possible relationships are concerned, are two articulations of both the first infraorbital characters of the Mullidae that might be ex and the mesopterygoid with the lateral eth pected in a related group but which do not moid , and the loss of the ethmoid-maxillary occur in sparoids, namely the two well ligament. separated dorsal fins and the squamation of Two suggestions concerning mullid rela the maxilla. To me, a real relationship be tionships have been made. One is with the tween mullids and sparoids seems far from Lutjanidae by Regan (1913) and the other established, but I am unable to suggest a more with the Sparidae by Boulenger (1904). There convincing relative for the Mullidae. are far more similarities between goatfishes and the sparoid fishes than between mullids and lutjanids. These similarities are listed STRUcrUR E AND ECOLOGY below for what they are worth, but to me they do not furnish very convincing evidence of The structural peculiarities of goatfishes relationship. The possibility of their inde seem to be reflected at several points in their pendent development in two different bottom ecology, or vice versa. Goatfishes appear to feeding percoid groups seems strong. A caveat have evolved a particular feeding niche for in this regard is that some of the more notable themselves. This niche depends on a special similarities do not occur between goatfishes ized method of hunting, not on the exploita- PO

322 PACIFIC SCIENCE, Volume 38, October 1984

tion of any particular food source. To judge depends on rapid acceleration and sustained from reports of goatfish stomach contents speed among other things. The horizontal po (see, for example, Hiatt and Strasburg 1960, sition of the body above the substrate while Hobson 1974), mullids eat a wide variety of hunting or excavating appears to be advan small to medium-sized animals. No goatfish tageous here. In moving off, goatfishes do not species appears to be a food specialist. (In this have first to leave the substrate or to shift from .regard the gill rakers and pharyngeal den a head-down bod y position (except when they tition of all goatfishes seem to be of normal, are actually feeding rather than hunting). percoid type.) The same horizontal position of the body The highly developed hyoid barbels seem to above the substra te while locating food also provide the base from which goatfishes have appears advantageous as a method ofhunting evolved their ecological peculiarities. Many over wide areas with occasional sampling of fishes have sensory probes by means ofwhich the bottom for food. Indeed, one goatfi sh they locate food on or in the bottom. The uses, at least on occasion , a roving method of goatfish barbel apparatus, however, has hunting in which the barbels are not used at evolved beyond analagous structures in other all. Hiatt and Strasburg (1960 :87) state re fishes in a number of respects. The hyoid bar garding Upeneus arge (now U. taeniopterus): bels of the Mullid ae are relatively long, " It does not probe the sand with its barbels as strong, and highly movable. The y can be and do oth er goatfishes but rather seeks out ben are used as an excavating device (Figure 2) or thic crustaceans living exposed on the bottom. for harassing and dislodging prey from holes Perhaps because it does not probe, it seems and crevices in hard substrates. Indeed, other swifter than oth er goatfishes." However, U. fishes often take advantage of these mullid taeniopterus has welI-developed barbels which barbel activities. Hobson (1974: 962) notes it probably uses at least on occasion, as U. that in Hawaii Parupeneus chryserydros (now tragula (Figure 2) certainly does. P. cyclostom us) is frequently followed around To summarize briefly, goatfish feeding over by the jack, Caranx melampygus, "probably sand or mud seems to combine attributes of as a tactic to capture prey driven from cover as roving predators such as the bonefish (A lbula) the foraging goatfish disturb the substratum," and the pompano (Trachinotus) with some of and Randall has a photograph ofajack inves those of excavators such as ophichthid eels tigating the possibilities of Mulloides foraging and flatfishes. in the West Indie s. As noted in the introduction, there appears Equally or more important is the goatfish to have been a minor adaptive radiation abilit y to raise or lower the barbel appa ratus within the Mullidae. Morphological differ (Figure 3) while the fish maintains a hori zon ences between genera, together with what tal position abo ve the substrate (Figures 1, 2). is known of their ecology, indicate variations Though members of the relatively special in feeding methods. The relatively extensive ized Parupeneus forage in reef areas, the dentition of Upeneus suggests a generalized more generalized Upeneus, Mullus, and Mul carnivore from which the genera Mullus on loides feed over sand or mud . Sandy or muddy the one hand and Parupeneus and Pseudu bottoms may occur as pockets in hard sub peneus on the other have become specialized strates but are more frequently present as in two different directions. In Mullus the bro ad, unprotected areas. There appear to be absence of teeth in the adult upper jaw two aspects of the relationship between the suggests a suction feeder, and its habit of goatfish barbel apparatus and feeding in such ploughing up the mud is in line with such a areas. One has to do with escape from pre feeding method. At the other extreme, the dators. Goatfishes do not have effective strong jaw teeth near the end of a relati vely spines; they do not bury themselves in the long snout in Parupeneus and Pseudup eneus substrate; and they are often far from holes in suggest that their teeth are used for immedi which they can take refuge. To escape, they ately seizing active prey that might otherwise simply swim off into open water, a system that escape or be eaten by other fishes. That at least . ,

Structure, Function, and Ecology in Goatfi shes-GosLINE 323 some of the prey of these two mullid genera goatfish genus Upeneus with descriptions of are relatively large, active animals is indi two new species. Proc. U.S. Natl. Mus. cated by the higher proportion offishes in the 103:497-532,pls.13,14. stomachs of Parup eneus chryserydros (now - -- . 1960. Family Mullidae: Goatfishes. P. eyclostomus) than in the stomachs of any Pages 1-46, pis. 75-78 in L. P. Schultz et aI., of the Hawaiian goatfishes Hobson (1974) Fishes of the Marshall and Marianas investigated. Islands, vol. 2. U .S. Natl. Mus. Bull. 202. LoBIANCO, S. 1907. L'origine dei barbigli tat tili nel genere Mullus. Atti della Reale Ac cademia dei Lincei, Ser. 5, Rendiconti, ACKNOWLEDGMENTS Classe di Scienzi fisiche , matematiche e I am greatly indebted to three people for naturali, 16: 577-586, 8 figs. help with this paper: John Paxton supplied MONOD, T. 1968. Le complexe urophore des specimens of Upeneichthys and Quinquarius poissons teleosteens, Memoires de l'Institut hendeeaeanthus from the Australian Museum; fondamental d'Afrique noire 81: 1-702, Jack Randall brought me up to date on mullid 989 figs. nomenclature and provided Figure 2 as well as OSSE, J. W. M . 1969. Functional morphology general information; and to Ted Hobson I am of the head of the perch (Perea fluviatilis indebted for Figure I and a discussion of L.): An electromyographic study. Nether goatfish ecology. lands. J. Zool. 19:289-392,15 pis., 24 text figs. R EGAN, C. T. 1913. The classification of the LITERATURE CITED percoid fishes. Ann. Mag. Nat. Hist., ser. 8, 12 :111-145. BOULENGER, G . A. 1904. Fishes (Systematic SATO, M . 1937. Preliminary report on the bar account of Teleostei). Pages 541-727, figs. bels ofa Japanese goatfish, Upeneoides ben 325-440 in The Cambridge natural history, sasi (Temminck & Schlegel). Sci. Repts. vol. vii. Macmillan, London. Tohoku Imperial University, 4th Ser. (Bio CALDWELL, M. K. 1962. Development and dis logy), II: 259-264, 3 figs. tribution oflarval and juvenile fishes of the SMITH, C. L., and R. M. BAILEY. 1962. The family Mullidae of the western North At subocular shelf of fishes. J. Morph. 110: lantic. Fish. Bull. Fish and Wildl. Ser. 1-18,3 pis. 62: 403-457, 43 figs. STARKS , E. C. 1904. The osteology of some FREIHOFER, W. C. 1963. Patterns of the ramus berycoid fishes. Proc. U.S. Nat. Mus. lateralis accessorius and their systematic 27:601-619,10 figs. significance in teleostean fishes . Stanford - --. 1926. Bones of the ethmoid region of Ichthyol. Bull. 8: 81-188, 29 figs. the fish skull. Stanford Univ. Pub!. , Univ. HIATT, R. W., and D. W. STRASBURG. 1960. Series , BioI. Sci. 4: 139-338, 58 figs. Ecological relationships ofthe fish fauna on - -- . 1930. The primary shoulder girdle of coral reefs of the Marshall Islands. Eco!' the bony fishes. Stanford Univ. Publ., Univ. Monogr. 30:65-127, 7 figs. Series , BioI. Sci. 6 : 149-239,38 figs. HOBSON, E. S. 1974. Feeding relationships of WINTERBOTTOM , R . 1974a. A descriptive syn teleostean fishes on coral reefs in Kona, onymy of the striated muscles of the Hawaii. Fish. Bull. Fish and Wildl. Ser. Teleostei. Proc. Acad. Nat. Sci. Philadel 72:915-1031,42 figs. phia 125: 225-317, 56 figs. JOHNSON, G . D. 1980. The limits and relation -- 1974b. The familial phylogeny of ships of the Lutjanidae and associated the Tetraodontiformes (Acanthopterygii: families. Bull. Scripps In st. Oceanogr. Pisces) as evidenced by their comparative 24 : I-Il4, 37 figs. myology. Smithsonian Contr. Zool. 155, LACH NER , E. S. 1954. A revision of the 201 pp ., 185 figs.