WASHINGTON ET AL.: SCORPAENIFORMES 



445 



Cottidae, until further study can define the phylogenetic rela- 

 tionships or monophyletic nature of these groups (Howe and 

 Richardson, 1978; Washington and Richardson, MS). 



Yabe (1981) recognized the family Ereuniidae for the Jap- 

 anese "cottid" genera Ereunias and Marukawichthys. He used 

 derived characters such as free pectoral fin rays and associated 

 pectoral giidle modifications to define the family. Yabe con- 

 cluded that the genus hclus belonged in the Cottidae. Previous 

 workers (Matsubara, 1936; Berg. 1940; Nelson, 1976) have placed 

 Marukawichihys and Ereunias in the family Icelidae with mem- 

 bers of the genus Icelus. 



Nelson (1982) has revised the "family' Psychrolutidae which 

 includes two subfamilies (Psychrolutinae and Cottunculinae). 

 Nelson could not define the family as monophyletic on the basis 

 of unique, derived characters and stated that the question of 

 whether to include the psychrolutids in the Cottidae was sub- 

 jective at this time. He rejected a close affinity between the 

 psychrolutids and liparidids as suggested by early workers. 



The families Comephoridae and Cottocomephoridae are en- 

 demic to the Lake Baikal basin (U.S.S.R.). Berg (1940) recog- 

 nized each as separate families within the superfamily Cotto- 

 idae. Taliev (1955), after detailed study of the two groups, 

 suggested that they had originated from cottid ancestors and 

 cited as evidence their similarities to two cottid genera, Meso- 

 cottus and Trachydenmis. Both Taliev (1955) and Kozhov ( 1 963) 

 placed the cottocomephorids in the Cottidae while the vivipa- 

 rous comephorids were recognized as a separate family. 



The family Agonidae has been reviewed only by Freeman 

 (1951) who suggested that the agonids were most closely related 

 to the cottids. The family is distinct in having fused, bony plates 

 covering the body. 



Nonnanichthys crocken. the sole member of the Normanich- 

 thyidae, occurs off the coasts of Peru and Chile. Its relationships 

 are obscure. Norman (1938b) considered it to be a primitive 

 cottid, while others (Berg, 1940; Quast, 1965) have placed it in 

 its own family, in the superfamily Cottoidae. In addition to a 

 different body plan, the suborbital stay of Nonnanichthys is 

 quite distinct from other scorpaeniform fishes (Poss, 1975). Its 

 relationships to cottoids have yet to be established. 



relatronships based on 

 Larval Characters 



Larvae of only about 20% of the sc&rpaeniform genera are 

 known, and only recently have larvae been used in systematic 

 studies (see Richardson, 1981a; Washington, 1981; Kendall and 

 Vinter, 1984; Washington and Richardson, MS). The most ex- 

 tensive information dealing with systematic characters of scor- 

 paeniform larvae is presented in a recent study by Washington 

 and Richardson (MS). This work dealt with over 100 osteolog- 

 ical characters of larval and juvenile cottids and their allies. 

 About half of the 70 characters used in their analysis were re- 

 stricted to the larval period. In general, larval characters were 

 most useful in defining groups below the subordinal level. 



Larvae of many scorpaenoid families are not yet known. Char- 

 acters such as head and preopercular spination and pectoral fin 

 length and pigmentation may be useful in future systematic 

 analyses; however, at present, larvae of too few taxa are known 

 to suggest relationships within the suborder Scorpaenoidei. 



The results of Washington and Richardson's (MS) study, agree 

 with those of past studies which propose that a scorpaenid-like 

 stock was ancestral to the Scorpaeniformes and was derived 



from a "generalized" perciform. Larvae ofthescorpaenid genera 

 Sebastes. Sebastolobus, and Scorpaena possess some characters 

 which are but slight modifications of those possessed by some 

 generalized percoids. In contrast, other scorpaeniform larvae 

 examined possessed considerable modifications of these char- 

 acters. These generalized scorpaenid characters include among 

 others: presence of predorsal bones; large, fused first anal pte- 

 rygiophore with three, stout anal spines; pleural ribs on abdom- 

 inal vertebrae; epipleurals attached to pleural ribs; hypurals 1 -I- 

 2 partially fused; hypurals 3-1-4 partially fused; presence of a 

 fifth hypural and parhypural; all hypural elements autogenous 

 and a specialized neural spine on preural centrum 2. Without 

 the suborbital stay, larvae of a scorpaenid such as Sebastes cou\d. 

 easily be mistaken for those of a generalized percoid. We con- 

 sider these character states to represent the plesiomorphic con- 

 dition in the Scorpaeniformes. 



Washington and Richardson's study focused in detail on cot- 

 toid and hexagrammoid fishes where larvae of many taxa are 

 fairly well known. They found that the hexagrammoids exhibit 

 many character complexes which are derived relative to the 

 scorpaenids. These include: 1 ) reduced anal spines and first anal 

 pterygiophore; 2) the pleural and epipleural ribs inserted to- 

 gether on the vertebral parapophyses; and, 3) the pectoral radials 

 broadened and anvil-shaped, but with distinct foramina be- 

 tween them. None of these characters is unique to the larval 

 period. 



Within the taxa traditionally assigned to the Hexagrammoidei 

 (Nelson, 1976), two monophyletic groups are recognized by 

 Washington and Richardson (MS). The first includes the hex- 

 agrammid genera Hexagrammos, Pleurogrammus. and Ophio- 

 don and the anoplopomatid genus Anoplopoma. This group is 

 defined by seven autopomorphies: 1) reduced head spination; 

 2) prolonged chondrification; 3) a unique (within Scorpaeni- 

 formes) sequence of ossification of the vertebral centra; 4) paired 

 first dorsal fin elements; 5) five preural centra involved in caudal 

 fin support; 6) anterior insertion of principal caudal rays; and, 

 7) a high number of vertebrae and ribs. The first four characters 

 are restricted to the larval period. 



In contrast, larvae of the second group, Oxylcbius and Zan- 

 iolcpis, do not possess any of the synapomorphies of the first 

 group. They do share one derived character— an unfused neural 

 arch and spine of the first vertebral centrum. The arms of the 

 first neural "arch" and spine remain unfused for a brief time 

 during larval development, a unique condition among known 

 scorpaeniform larvae. Other larval characters support the sep- 

 aration of these groups, but we are cautious in the interpretation 

 of these characters. They include: 1) large versus small size at 

 hatching; 2) neustonic versus planktonic larvae; and 3) long, 

 slender versus deep body shape. 



Washington and Richardson (MS) concluded that the first 

 group of hexagrammoids is very distinctive and differs from all 

 other scorpaeniforms so far examined, particularly in the mode 

 of ossification of the vertebral column and in the number of 

 preural centra involved in the caudal fin support. Because of 

 the uniqueness of these characters, Washington and Richardson 

 (MS) suggest that members of this hexagrammoid group prob- 

 ably comprise a separate lineage within the order, distinct from 

 Oxylebius and Zaniolepis and the other cottoids. The second 

 group, Oxylebius and Zaniolepis. is distinctive but appears to 

 be closer in many characters to the scorpaenids than to other 

 hexagrammoids. 



