Jo,

BULLETIN OF

THE BRITISH MUSEUM

(NATURAL HISTORY)

ZOOLOGY

Vol. 29

1976

BRITISH MUSEUM (NATURAL HISTORY) LONDON: 1977

DATES OF PUBLICATION OF THE PARTS

No. i ..... 22 January 1976

No. 2 . . . . .13 April 1976

No. 3 . . . . 13 April 1976

No. 4 . . . .23 April 1976

No. 5 . . . . . -27 May 1976

No. 6 . . . . -27 May 1976

ISSN 0007-1498

Printed in Great Britain by John Wright and Sons Ltd. at The Stonebridge Press, Bristol BS4 5NU

CONTENTS

ZOOLOGY VOLUME 29

PAGE

No. i. A review of the family Centropomidae. By P. H. GREENWOOD . i

No. 2. Fossil reptiles from Aldabra Atoll, Indian Ocean. By E. N.

ARNOLD ........... 83

No. 3. A review of the family Canidae, with a classification by numerical

methods. By J. CLUTTON-BROCK, G. B. CORBETT and M. HILLS . 117

No. 4. The cranial musculature and taxonomy of characoid fishes of the

tribes Cynodontini and Characini. By G. J. HOWES . . . 201

No. 5. Designation of lectotypes of some ostrocods from the Challenger

Expedition. By H. S. PURI and N. C. HULINGS (Pis 1-27) . . 249

No. 6. Some tertiary and recent Conescharelliniform Bryozoa. By P. L.

COOK and R. LAGAAIJ (Pis 1-8) 317

A REVIEW OF THE FAMILY

CENTROPOMIDAE (PISCES, PERCIFORMES)

P. H. GREENWOOD

BULLETIN OF

THE BRITISH MUSEUM (NATURAL HISTORY) ZOOLOGY Vol. 29 No. i

LONDON: 1976

A REVIEW OF THE FAMILY CENTROPOMIDAE (PISCES, PERCIFORMES)

BY

PETER HUMPHRY GREENWOOD

Pp. 1-81 ; 37 Text-figures

BULLETIN OF

THE BRITISH MUSEUM (NATURAL HISTORY) ZOOLOGY Vol. 29 No. i

LONDON: 1976

THE BULLETIN OF THE BRITISH MUSEUM

(NATURAL HISTORY), instituted in 1949, is issued in five series corresponding to the Scientific Departments of the Museum, and an Historical series.

Parts will appear at irregular intervals as they become ready. Volumes will contain about three or four hundred pages, and will not necessarily be completed within one calendar year.

In 1965 a separate supplementary series of longer papers was instituted, numbered serially for each Department.

This paper is Vol. 29, No. i of the Zoology series. The abbreviated titles of periodicals cited follow those of the World List of Scientific Periodicals.

World List abbreviation : Bull. Br. Mus. not. Hist. (Zool.)

ISSN 0007-1498

Trustees of the British Museum (Natural History), 1976

BRITISH MUSEUM (NATURAL HISTORY)

Issued 22 January 1976 Price £5.85

By P. H. GREENWOOD

CONTENTS

Page

INTRODUCTION ........... 4

MATERIALS ........... 5

ABBREVIATIONS USED IN TEXT-FIGURES ...... 8

THE FAMILY CENTROPOMIDAE ........ 10

AN ANATOMICAL AND TAXONOMIC REVIEW OF THE Ldtes AND LutioldteS

SPECIES ........... 12

The neurocranium . . . . . . . . . 14

Hyopalatine arch and the preoperculum . . . . . 27

Circumorbital bones ......... 31

Opercular bones .......... 32

Jaws 34

Branchial skeleton ......... 35

Hyoid arch .......... 38

Pectoral girdle and associated bones ...... 39

Vertebral column ......... 42

Caudal fin skeleton ......... 44

Dorsal and anal fins ......... 45

Swimbladder .......... 47

Baudelot's ligament ......... 47

Lateral line .......... 48

THE INTERRELATIONSHIPS OF SPECIES WITHIN THE GENUS Ldtes, AND THE

TAXONOMIC STATUS OF Luciolates Blgr. ...... 49

A REVIEW OF THE GENUS Psammoperca Richardson . . . . 51

Introduction .......... 51

Osteology and anatomy of Psammoperca waigiensis . . . 52

Neurocranium .......... 52

Hyopalatine arch and the preoperculum . . . . . 53

Circumorbital bones ......... 54

Opercular series .......... 55

Jaws ....... 55

Branchial skeleton ........ 55

Hyoid arch skeleton ........ 56

Pectoral girdle and associated bones . . . . . . 56

Vertebral column ......... 58

Caudal fin skeleton ......... 58

Dorsal, anal and pelvic fins ...... 59

Swimbladder ....... 60

Baudelot's ligament ......... 60

Lateral line ........ 60

THE RELATIONSHIPS OF Psammoperca .... 60

THE RELATIONSHIPS OF Centropomus WITH THE LATINAE . . 62

Neurocranium ........ 62

Hyopalatine arch ...... 63

Preoperculum ....... 64

4 P. H. GREENWOOD

Operculum .... . . . . . . 64

Circumorbital bones ......... 64

Jaws ........... 64

Gill arches ........... 64

Vertebrae ........... 65

Predorsal bones . . . . . . . . . - . 65

Dorsal fins . . . . . " . ' . . 65

Anal pterygiophores ......... 66

Caudal fin skeleton . . . . . . . . 66

Pectoral girdle and associated bones ...... 66

Swimbladder .......... 66

Baudelot's ligament ......... 67

Lateral line .......... 67

FOSSIL CENTROPOMIDAE ......... 68

Lates species .......... 69

Eolates species .......... 70

BlOGEOGRAPHY .......... 71

DIAGNOSES FOR THE CENTROPOMIDAE, ITS SUBFAMILIES, GENERA AND

SUBGENERA ........... 75

Centropomidae .......... 75

Centropominae .......... 76

Latinae ........... 76

ACKNOWLEDGEMENTS ......... 78

REFERENCES ........... 78

INTRODUCTION

THE Centropomidae, a family of tropical estuarine, marine and freshwater percoid fishes, is represented in the New World, Africa and Asia by a total of at least 18 species (the number varying mainly with the taxonomic limits set for the family by different authors). Of the 18 species recognized, 9 occur in the New World, 7 - all freshwater species - occur in Africa and 2 in Asia.

Temporally the family has a good fossil record extending from the Eocene to prehistoric times. Geographically, however, the fossil record is restricted to Africa and Europe, and only one taxonomic division of the family, the closely related genera Lates and Eolates, is represented ; see Sorbini (1973) and Greenwood (1974).

It was, in fact, a new fossil centropomid from the neogene of Africa (Greenwood & Howes, 1975) that led to this revision. Our attempts to identify the new fossil soon made it clear that the African centropomids are more varied anatomically than had been realized previously. Also, it became obvious that the current taxonomic arrangement of the family does not reflect the probable phyletic relationships of its constituent taxa. Indeed, except for Eraser's (1968) analysis of the New World Centropomus species, no fully reasoned attempt has been made to interpret intra- familial relationships along phyletic lines. Also, the presumed relationships of the Centropomidae with the Serranidae (Regan, 1913 ; Berg, 1947 ; Gosline, 1966 ; Greenwood, Rosen, Weitzman & Myers, 1966) appear to be based more on intuitive than on critical taxonomic reasoning, and need reappraisal.

Although it was for these reasons mainly that the present review was undertaken, I also hope that it may provide a step towards the clearing of that taxonomic rag-bag, the 'lower percoid fishes' (see Greenwood et al., 1966 ; Gosline, 1966).

GLAUCOSOMIDAE

Osteological material : Glaucosoma burgeri

CENTROPOMIDAE

Osteological material : Lates calcarifer Lates niloticus Lates niloticus Lates niloticus Lates niloticus

Lates niloticus Lates niloticus Lates niloticus Lates macrophthalmus Lates longispinis Lates microlepis Lates angustifrons

Lates mariae

Luciolates stappersi Luciolates stappersi

Luciolates stappersi Luciolates stappersi

REVIEW OF CENTROPOMIDAE MATERIALS

BMNH 1884.2.26:60 China

BMNH 1873.1.21:2	Fitzroy R.
Unregistered	No locality
BMNH 1971.2.8:186	No locality
Unregistered	'Red Sea'
BMNH 1864.6.29:5	'West Africa'
Unregistered	Lake Rudolf
Unregistered	No locality
Unregistered	Lake No
Unregistered	Lake Albert
Unregistered	Lake Rudolf
BMNH 1900.12.13:37	Albertville
BMNH 1955.12.20:1722	Lake Tanganyika

BMNH 1955.12.20:1667

BMNH 1955.12.20:1672 BMNH 1936.6.15:

1705-6

BMNH 1971.6.23:76-8 BMNH 1975.4.23:2

Lake Tanganyika Lake Tanganyika

293 mm S.L.

360 mm S.L. (skeleton) (alizarin preparation) (disarticulated skeleton) (disarticulated skeleton) (skull and pectoral

girdle)

(alizarin preparation) (alizarin preparations) (alizarin preparation) (skeleton, 102 mm S.L.) (skull)

(disarticulated skeleton) (skull and pectoral

girdle) (skull and pectoral

girdle) (disarticulated skeleton)

Psammoperca waigiensis BMNH 1892.9.2:10-11 Borneo

Psammoperca waigiensis Centropomus undecimalis Centropomus ensiferus Centropomus pectinatus

Dissected specimens : Lates niloticus

Lates niloticus

Lates niloticus Lates macrophthalmus Lates macrophthalmus Lates longispinis

Lates angustifrons Lates mariae

Luciolates stappersi

BMNH 1872.10.18:90 BMNH 1883.12.16:1-2 BMNH 1861.12.12:13 BMNH 1894.12.1:5

BMNH 1907.12.2:

2915-6 BMNH 1907.12.2:

2952-3

BMNH 1931.11.20:1-2 BMNH 1929.1.24:341-4 BMNH 1975.1.18:1 BMNH 1932.6.13:

102-106

BMNH 1906.9.8:87-88 BMNH 1955.12.20:

1628-29 BMNH 1955.12.20:

1669-71

Lake Tanganyika (skull) Lake Tanganyika (alizarin preparation) Lake Tanganyika (disarticulated skeleton from a fish 390 mm S.L.) (skull and pectoral

girdle)

(circumorbital series) (skull)

(skeleton, disarticulated) (skull and pectoral girdle)

Cebu Jamaica No locality Jamaica

Nile

Nile Merowe (Paratype) Lake Albert

(Syntype)

Lake Tanganyika

227 mm S.L.

165 mm S.L. 218 mm S.L. 275 mm S.L. 315 mm S.L.

250 mm S.L. 300 mm S.L.

Lake Tanganyika 220 mm S.L. Lake Tanganyika 215 mm S.L.

P. H. GREENWOOD

Psammoperca waigiensis Psammoperca waigiensis Centropomus undecimalis Centropomus undecimalis Centropomus pectinatus

Specimens .examined : Lates calcarifer Lates calcarifer Lates niloticus Lates macrophthalmus

Lates longispinis

Lates angustifrons Lates mariae Lates microlepis Luciolates stappersi Luciolates stappersi

Psammoperca waigiensis

Psammoperca waigiensis Psammoperca waigiensis Psammoperca waigiensis Psammoperca waigiensis Centropomus undecimalis Centropomus undecimalis Centropomus pectinatus

Centropomus ensiferus

Specimens radiographed : Lates calcarifer Lates niloticus

Lates macrophthalmus

Lates macrophthalmus Lates macrophthalmus

Lates longispinis Lates angustifrons

Lates angustifrons Lates mariae

Lates mariae Lates microlepis Lates microlepis

Luciolates stappersi

BMNH 1892.9.2:10-11	Borneo	150 mm S.L.
BMNH 1872.10.13:90	Cebu	240 mm S.L.
BMNH 1883.12.16:1-2	Jamaica	280 mm S.L.
BMNH 1923.7.30:114	Rio de Janeiro	175 mm S.L.
BMNH 1895-5-27:3-5	Mazatlan	245 mm S.L.
BMNH 1863.2.23:29	Amoy	190 mm S L
BMNH 1936.8.6:43	Queensland	260 mm S.L.
Major part of the entire	collection
BMNH 1929.1.24:
340-344	(Paratypes)	145-283 mm S.L.
BMNH 1932.6.13:
102-106	(Syntypes)	115-270 mm S.L.
Entire collection
Entire collection
Entire collection
Entire collection
RGMARC 129887-889	(Tervuren
	Museum
	specimen)	407 mm S.L.
BMNH 1933.3.11:312	Culion,	no mm S.L.
	Philippines
BMNH 1883.11.28:14	Singapore	165 mm S.L.
BMNH 1891.10.29:66	Ceylon	132 mm S.L.
BMNH 1939.1.17:11	Hong Kong	215 mm S.L.
BMNH 1888.11.6:5	Madras	175 mm S.L.
BMNH 1936.1.31:8	Trinidad	290 mm S.L.
BMNH 1906.6.23:82	Trinidad	180 mm S.L.
BMNH 1920.12.22:	Trinidad	230 & 240 mm S.L.
57-58
BMNH 1903.5.15:3-5	Panama	183-230 mm S. L.
BMNH 1891.11.30:1-8
BMNH 1900.12.2:
2329-38
BMNH 1929.1.24:
340-344
BMNH 1929.4.16:39-41
BMNH 1929.1.24:
341-344
BMNH 1932.6.13:102-6
BMNH 1936.6.15:
1687-97
BMNH 1906.9.8:87-88
BMNH 1955.12.20:1636;
1628-9; 1655-6 ;
1672-86
BMNH 1906.9.6:7
BMNH 1906.9.8:89
BMNH 1955.12.20:
1753-85
BMNH 1955.12.20:
1669-71 ; 1680

REVIEW OF CENTROPOMIDAE

Luciolates stappersi

Luciolates stappersi Psammoperca waigiensis Psammoperca waigiensis Psammoperca waigiensis Psammoperca waigiensis Psammoperca waigiensis Centropomus unionensis Centropomus robalito Centropomus nigrescens Centropomus ensiferus Centropomus undecimalis Centropomus undecimalis Centropomus undecimalis

Fossil material : Eolates gracilis Eolates gracilis Eolates gracilis Eolates gracilis Eolates gracilis Eolates gracilis Eolates gracilis

SERRANIDAE

Osteological material :

Epinephelus tauvina Epinephelus areolatus Epinephelus afer Epinephelus itajara

Dissected specimens :

Epinephalus alexandrinus BMNH 1964.7.14:1 Malta

Specimens examined :

Serranus radialis BMNH 1923.7.30:77-79 Rio de Janeiro

BMNH 1936.6.15:
1705-6
BMNH 1971.6.23:76-78
BMNH 1872.10.18:90
BMNH 1939.1.17:11
BMNH 1892.9.2:10-11
BMNH 1870.12.27:17
BMNH 1888.11.6:5
BMNH 1903.5.15:8
BMNH 1895.5.27:6
BMNH 1883.7.28:14
BMNH 1903-5-15:3-5
BMNH 1895.5.27:2
BMNH 1936.1.31:8
BMNH 1924.2.29:16
BMNH P23803	Monte Bolca	130 mm S.L.
BMNH Pi6i37	Monte Bolca	121 mm S.L.
BMNH P39i8	Monte Bolca	139 mm S.L.
BMNH P23798	Monte Bolca	c. 135 mm S.L.
BMNH Pi6374	Monte Bolca	41 mm S.L.
BMNH Pi 6756	Monte Bolca	29 mm S.L.
BMNH P37225	Monte Bolca	124 mm S.L.
Unregistered	Java	(skeleton, 225 mm	S.L.)
BMNH 1888.12.29:24	Muscat	(skeleton, 280 mm	S.L.)
Unregistered	St Croix	(skeleton, 140 mm	S.L.)
BMNH 1883.12.16:9	Jamaica	(skeleton, 355 mm	S.L.)

243 mm S.L.

115-140 mm S.L.

AMBASSIDAE

Osteological material : Ambassis urotaenia Ambassis commersonii Ambassis wolffii

BMNH 1928.1.17:8-15 BMNH 1855.9.19:359 BMNH 1898.4.2:67

Specimens examined:

The major part of the collections of Ambassis and Chanda

(alizarin preparation) (disarticulated skeleton) (skeleton, 85 mm S.L.)

GERRETDAE

Osteological material :

Gerres oyena Gerres oyena

BMNH 1965.4.4:125-38 BMNH 1960.3.15:670-5

(alizarin preparations) {alizarin preparations)

P. H. GREENWOOD

ABBREVIATIONS USED IN THE TEXT-FIGURES

AA Anguloarticular

ART P Articular process of the pre-

maxilla ASC P Ascending process of the pre-

maxilla AHYF Anterior facet for hyomandi-

bula

Bb 1-3 Basibranchial 'BcF' 'Berycoid foramen'

BOC Basioccipital

BrR Branchiostegal ray

BSP Basisphenoid

Cb 1-5 Ceratobranchial of first to fifth

arch

Ch Ceratohyal

CL Cleithrum

COR Coracoid

D Dentary

Dhh Upper hypohyal

D PROC Dorsal process of the maxilla E Mesethmoid

Ei-E4 Epibranchials of the first to

fourth arches

EaTP, Tooth-plates associated with the

E3TP second and third epibranchials

ECT Ectopterygoid

Eh Epihyal

ENT Entopterygoid

EP Epural

EPI Epioccipital (= epiotic auct.}

EXO Exoccipital

FR Frontal

FRC Frontal crest

FRR Frontal ridge

GF Gill filament

Gh Glossohyal

Gr Gill raker

H1( H5 First and fifth hypurals

Hbi Hypobranchial of first gill arch

HsPU2 Haemal spine of second preural vertebra

Hyomandibula

Intercalar

Interoperculum

Lachrymal (first circumorbital

bone)

LATE Lateral ethmoid

LAT SP Latero-sensory canal openings LC Lateral commissure

LIG Ligament

MET Metapterygoid

NaPU2 Neural arch and spine of second

preural vertebra

OCS Occipito-spinal nerve foramen

OP Operculum

PAL Palatine

PAR Parietal

PARC Parietal crest

Pbi-Pb4 Pharyngobranchials of the first to fourth gill arches

Pb2 TP-

Pb4 TP Tooth plates associated with pharyngobranchials of the sec- ond to fourth gill arches

PCj, PC2 Upper and lower postcleithra

PFr Pectoral fin ray

PH Parhypural

PHYF Posterior facet for hyomandi- bula

PMAXP Posterior maxillary process

PMXP Premaxillary process of the maxilla

POP Preoperculum

PRO Prootic

PS Parasphenoid

PTF Posttemporal fossa

PTO Pterotic

PTS Pterosphenoid

PTSP Pterosphenoid pedicle

PTSS Pterosphenoid spur

PU1 + U1 Fused first ural and preural centra

Q Quadrate

R Radial for pectoral ray

RA Retroarticular

SC Supracleithrum

Sc Scapula

SLP Supralamellar tooth plate

SOC Supraoccipital

SOC S Supraoccipital shelf

SOP Suboperculum

SOS Subocular shelf

SPO Autosphenotic

SY Symplectic

TP Tooth plate

UR Uroneural

URi, URa Upper and lower uroneurals

V Vomer

VHh Lower hypohyal

II -V Second to fifth circumorbital bones

IX Foramen for glossopharyngeal

nerve

X Foramen for vagus nerve ist APTY First anal pterygiophore and Asp Second anal fin spine

REVIEW OF CENTROPOMIDAE

FIG. i. Outlines of : (a) Psammoperca waigiensis, (b) Lates calcarifer, (c) L. niloticus, (d) L. longispinis, (e) L. macrophthalmus , (f) L. angustifrons , (g\ L. mariae, (h) L. wicro- /e^>is, (i) L. stappersi, (j) Centropomus undecimalis.

io P. H. GREENWOOD

THE FAMILY CENTROPOMIDAE

Although in 1955 Matsubara classified several of the genera considered below in the family Serranidae, there is still a consensus of opinion among ichthyologists that these fishes constitute a natural taxonomic group, albeit one closely related to the Serranidae. (See Katayama (1956) for a short taxonomic history of the group.)

There has, however, been rather less agreement on the definition and delimitation of the family Centropomidae in which these various genera are classified, or with which they are thought to be most closely related. In particular there is uncertainty about the affinities of Glaucosoma Temm. & Schl., of Chanda Ham. Buch. (= Ambassis of authors), and of genera related to Chanda. Regan (1913), for example, included both Glaucosoma and Chanda (as Ambassis) in the Centropomidae, as did Norman (1966) who, however, gave Chanda and related genera subfamilial rank (Chandinae) and placed Glaucosoma with Lates Cuv., and Psammoperca Richard- son in the subfamily Latinae. Other views were expressed by Jordan (1923) who gave familial rank both to Chanda and its related taxa (Ambassidae), and to Glauco- soma (Glaucosomidae). Berg's (1947) classification returned Chanda to the Centro- pomidae, but kept Glaucosoma as a monotypic family. Greenwood et al. (1966) followed Berg, as did Lindberg (1971).

There have, of course, been several definitions of the Centropomidae, both sensu latu and stricto (see especially Gill, 1883, and Meek & Hildebrand, 1925, for the family as restricted to species of Centropomus ; Regan, 1913, and Norman, 1966, for the family sensu latu ; Munroe, 1961, for the Chandidae and Centropomidae, and Katayama, 1954, for the only comprehensive definition of the Glaucosomidae). Yet, from none of these definitions is it possible to determine the synapomorph features that could establish the phyletic relationships of the taxa involved, either as a holophyletic assemblage or as two or even three lineages.

With the aim of establishing such relationships I have examined all the characters listed in these various definitions ; as is inevitable in such revisionary work I have discovered other characters which were not taken into account by earlier authors. Most of the characters used by Regan (1913), Norman (1966) and Katayama (1954) are either primitive features widely distributed amongst the lower percomorphs and percoids (i.e. symplesiomorphies), or, if derived ones, are characteristics also shared with several percoid families. In the symplesiomorphic category are the vertebral number, presence of frontoparietal crests, and the dentition and other jaw characters. The derived characters include the presence of an axillary pelvic scale, and the extension of lateral line pore scales onto the caudal fin. This latter character is of interest because, although the lateral line extends some way onto the caudal in several percoid taxa, rarely does it reach or almost reach to the margin of that fin, as it does in Centropomus, Lates and Psammoperca. (Only in the Sciaenidae does the lateral line extend as far posteriorly as in these genera.) This distinction in the degree to which the lateral line extends posteriorly has not been drawn by other workers.

One character not used by previous authors (but mentioned with reference to Centropomus and Lates by Gosline, 1966) is the presence of an anteroposteriorly expanded neural spine on the second vertebra. Indeed, this feature, combined with

REVIEW OF CENTROPOMIDAE n

the extension of the lateral line far onto the caudal fin, provide the only synapo- morph characters at all widely distributed amongst taxa currently classified with the Centropomidae. Because the caudal lateral line character also occurs in the Sciae- nidae, the neural spine character is the sole truly synapomorph feature of the centropomids. Currently recognized genera with such an expanded neural spine are Lates, Luciolates, Psammoperca and Centropomus (Fig. i). Except for Luciolates, these genera also have the caudal lateral line scale row extending or almost extending to the fin margin. (The lateral line in Luciolates is discussed on p. 48.)

Neither Glaucosoma nor Chanda (and its related genera) has either of these features. The lateral line extends only onto the basal third, or less, of the caudal fin, and the second neural spine is no wider than that of the first vertebra (in other words, the usual percomorph condition) .

Thus, on the basis of these characters, and the lack of any other unifying features, it would seem that Glaucosoma and Chanda cannot belong to the same lineage as Centropomus, Lates, Luciolates and Psammoperca. These latter taxa alone are therefore retained in the family Centropomidae.

Questions now arise as to the relationships and status of Glaucosoma and the Chanda-\ike genera, of their relationship to the Centropomidae as here defined, and of the interrelationships of the Centropomidae within the Percoidei.

Nothing I have yet discovered suggests that Glaucosoma is a close relative of Chanda (and its immediate relatives). Both taxa are readily defined by various autapomorphies, but I cannot find any synapomorph characters uniting them. Unfortunately, the sort of detailed information needed for phyletic studies amongst percoid fishes is not yet available for many taxa, and I cannot suggest where the relationships of Chanda and Glaucosoma may lie. For the moment the only course available is to recognize two families, the Glaucosomidae and the Chandidae, and to consider both as of uncertain affinity amongst the Percoidei. The dorsal gill arch skeleton in the Chandidae I have examined (several species of Chanda) is certainly more derived than are those of the Centropomidae and Serranidae (see Rosen, 1973, for a discussion of the gill arches in percoid fishes). In the morphology of the pharyngobranchials, especially the second, Chanda is very similar to Eucinostomus argenteus (Gerridae) as figured by Rosen (op. cit., text-figs 98 & 99). Glaucosoma also shows more derived characters in its gill arch skeleton than does any member of the Centropomidae. I suspect that it will be from the gill arch skeleton that the relationships of these two families will ultimately be determined.

Similar problems and lack of data limit the formulation of hypotheses regarding the phylogeny of the Centropomidae. It is generally thought, or implied, that the centropomids are closely related to the Serrandiae (see Regan, 1913 ; Katayama, 1954 ; Gosline, 1966 ; Greenwood et al., 1966 ; Norman, 1966). Again it has so far proved impossible to demonstrate within these families any but symplesiomorph or autapomorph features, none of which provides acceptable information for confirming or refuting this relationship. Thus, for the time being the Centropomidae too must remain as a family incertae sedis amongst the lower percoids.

However, with the limits of the Centropomidae defined (see above) it is now possible to turn to problems of infrafamilial relationship and taxonomy.

12 P. H. GREENWOOD

AN ANATOMICAL AND TAXONOMIC REVIEW OF THE LATES AND

LUCIOLATES SPECIES

The present taxonomic status of several Lates species must be reviewed before considering their anatomy and phyletic relationships. The probably monotypic genus Luciolates Blgr. is also included in this review, although a discussion of its ultimate status is deferred until p. 49.

With one exception, namely Lates calcarifer (Bloch), all extant Lates species are confined to Africa but fossil remains of this genus are known from southern Europe as well as from several areas in Africa (Sorbini, 1973 ; Greenwood, 1974). The extinct taxon Eolates gracilis (Agassiz) from Monte Bolca will be considered later (p. 70), together with the extinct 'species' of Lates.

Lates calcarifer, a coastal and estuarine species, is widely distributed in the Indo-Pacific region (India, Bangladesh, Burma, Malay Peninsula, Java, Sumatra, Borneo, Celebes, Sarawak, Philippines, Papua-New Guinea, northern and western Australia, southern China, and Japan). According to Weber & de Beaufort (1929), this species also occurs in the Persian Gulf ; their reference to L. calcarifer entering the mouths of the Nile, Niger and Senegal, and ascending these rivers, is clearly an error stemming from a confusion of this species with L. niloticus. Although essentially a marine fish, L. calcarifer freely enters and remains in rivers but always returns to estuarine or marine environments for spawning (Dunstan, 1959 ; Lake, 1971).

Lates niloticus (L.) is widely distributed in the rivers and lakes of tropical Africa (Nile, Niger, Senegal, Volta and Zaire [= Congo] rivers ; Lakes Chad, Albert, Rudolf and some of the Ethiopian lakes). Not surprisingly in such a widespread taxon there are indications of some geographically limited morphotypes. As yet there has been insufficient study of these populations to determine the significance of their morphological differences, and none of the fluviatile populations has been given the formal status of a subspecies (see Daget (1954) on Pellegrin's (1922) L. niloticus var. macrolepidotus from Zaire). Worthington (1932), however, has described two subspecies, L. niloticus rudolfianus and L. n. longispinis from Lake Rudolf.

Lates niloticus rudolfianus, a form attaining a large size (up to 148 cm total length) and apparently confined to inshore regions of Lake Rudolf (Worthington, 1932), is acknowledged by Worthington to be morphologically intermediate between L. niloticus of the Nile and populations of that species inhabiting Lake Albert (named L. albertianus by Worthington [1929], but shown by Holden [1967] to be indis- tinguishable from L. niloticus). I have re-examined the type material of L. n. rudolfianus and can find no reason for maintaining the subspecific status of this population. In all morphometric, meristic and gross morphological characters the type specimens lie within the range of variability determined for L. niloticus over its entire range. Thus, at least until larger samples are available from numerous localities in Lake Rudolf, I would consider L. niloticus rudolfianus to be a synonym of the nominate species.

The second subspecies from Lake Rudolf, L. n. longispinis, presents a somewhat different problem. Apparently it is separated ecologically from the other Lates

REVIEW OF CENTROPOMIDAE 13

species in the lake, being a fish of the deeper waters (Worthington, 1929 ; Hopson, unpublished report). Furthermore, it is morphologically differentiate from L. niloticus, and does not attain such a large adult size.

The principal morphometric differences distinguishing the taxon 'longispinis' from L. niloticus are its larger eye (diameter 22-6-39-9 Per cen^ of head in fishes 118-273 mm standard length, cf. 18-3-22-9 per cent in L. niloticus of a comparable size ; in both taxa eye size is negatively correlated with standard length) and longer third spine in the dorsal fin (78-0-84-0 per cent of head, cf. 55-0-70-0 per cent) . The larger eye in 'longispinis' is most clearly manifest when small specimens of both species are compared ; for example the eye is 21-8 per cent of the head in a 107 mm S.L. L. niloticus but is 32-9 per cent in a 118 mm specimen of 'longispinis' '.

Another difference, but one correlated with relative eye size, lies in the less marked posterior extension of the maxilla in 'longispinis'. In specimens of L. niloticus more than 125 mm S.L. the posterior tip of the maxilla lies at a point clearly behind a vertical through the posterior orbital margin ; in 'longispinis' above 125 mm long the maxillary tip lies in or a little anterior to that vertical. In fishes less than 120 mm S.L., the distinction is much less obvious (or even non-existent) because of the relatively larger eye in L. niloticus of that size.

Since, in Lake Rudolf, 'longispinis' and L. niloticus are sympatric (albeit allotopic), and because the two taxa show various and consistent morphological differences, I can find no grounds for considering 'longispinis' to be a subspecies of L. niloticus. The obvious expedient of raising Worthington's (1932) L. n. longispinis to full specific rank, however, requires further consideration when the taxon is compared with L. macrophthalmus Worthington, 1929 (see above ; also Holden, 1967). Lates macrophthalmus is the endemic ecological counterpart in Lake Albert of 'longispinis' in Lake Rudolf (see Holden, 1967), and closely resembles that species as well, sharing with it the presumably derived features of enlarged eyes and elongate third spine in the dorsal fin. The only differential feature I can find is the relatively longer spine of 'longispinis' (78-0-85-0, mean 82-0 per cent of head, cf. 65-0-84-0, m = 74'4 per cent, in L. macrophthalmus). There also appear to be slight differences in the relative proportions of certain head parts, e.g. the vertical limb of the preoperculum lies slightly further forward in 'longispinis'. Detailed comparisons are hampered by the paucity of study material, there being only the five syntypes of L. n. longispinis* and the eleven syntypes of L. macrophthalmus available.

Basically, the problem raised by 'longispinis' in Lake Rudolf and L. macroph- thalmus in Lake Albert is whether each should be considered a distinct and endemic species evolved locally from a population of L. niloticus (the generally accepted hypothesis, see Worthington, 1932, and Holden, 1967) or whether they should be looked upon as sister taxa derived from a common ancestor distinct from L. niloticus. This hypothetical species presumably invaded the developing Lakes Rudolf and Albert alongside L. niloticus. If this latter relationship could be determined it would, on the morphological evidence available, be more realistic to

* The sixth syntype of L. n. longispinis mentioned by Worthington (1932) cannot be located, and neither is it recorded in the Museum's register. This suggests that the word 'six' in the original description is a lapsus for 'five'.

I4 P. H. GREENWOOD

treat ' longispinis' as a subspecies of L. macrophthalmus rather than as a distinct species. Unfortunately I do not have enough material at my disposal to test the two hypotheses, even assuming that anatomical criteria alone would be suitable for this purpose. For the moment then, and without prejudice to an ultimate solution of the taxon's true phyletic position, I propose treating Worthington's subspecies as a full species, namely L. longispinis Worthington (1932).

The three other Lates species, L. angustifrons Blgr., L. microlepis Blgr. and L. mariae Steindachner (see Poll, 1953), require no further comment at this stage. All are morphologically distinct from the other species and from one another.

A fourth Lates-like taxon from Lake Tanganyika is currently placed in the genus Luciolates* Blgr., principally because of the wide separation of the two dorsal fins (Boulenger, 1914 ; Poll, 1953, 1957). Luciolates is closely related to Lates, in particular to L. mariae. As I hope to demonstrate in the next section of this paper I believe that Luciolates should be included in Lates if the principles of phyletic classification are not to be violated.

The anatomy of Lates and Luciolates

The anatomy, and especially the osteology, of Lates and Luciolates has never been subject to a general review encompassing all known species. Gregory (1933) has given a rather superficial account of the syncranial osteology in Lates niloticus,^ and Katayama (1956) a more detailed description of Lates calcarifer which included some details of its soft anatomy.

The account which follows is based on the examination of at least two skeletons of each species, and in the case of L. niloticus on several specimens over a wide size range. Radiographs of several specimens of every species were also examined.

In all intrageneric comparisons made below the conditions found in L. calcarifer and L. niloticus are, with few exceptions, taken to be those primitive for the genus. This conclusion regarding the status of the two species was reached after all the species had been examined and a comparison made with members of other percoid groups apparently related to the Centropomidae (Gosline, 1966 ; Greenwood et al., 1966). Within the Centropomidae as a whole, L. calcarifer and L. niloticus-type cranial osteology should also be taken to represent the primitive condition.

The neurocranium

The overall morphology of the neurocranium in Lates and Luciolates can be judged from Figs 2-8.

Basically, the neurocranium in Lates differs little from that of most serranids (sensu Greenwood et al., 1966). It has, however, well-developed and continuous frontoparietal crests with a sensory canal pore located at or near the junction of the crests, and the exoccipital facets are contiguous (separated in most serranids,

* A second species, Luciolates brevior, has been described (Boulenger, 1914), but is known only from the holotype and has never been recorded again. In all probability L. brevior should be treated as a synonym of Luciolates stappersi Blgr., 1914, and is treated as such in this paper.

f The neurocranium supposedly of Luciolates stappersi, figured by Gregory (1933), is wrongly identified; as far as I can judge it is from a specimen of Lates angustifrons.

REVIEW OF CENTROPOMIDAE

soc

PHYF

SPO

PRO

PAR

PTF

FR

OCS

10mm

SPO

PTO

FIG. 2. Lates niloticus, neurocranium. (a) Left lateral view, (b) Dorsal view. (From

Greenwood & Howes, 1975.)

personal observations ; see also Gosline, 1966). Since continuous frontoparietal crests (usually incorporating a sensory pore) occur in berycoid fishes (see Patterson, 1964), this condition must be considered a primitive one. Likewise, the medially contiguous exoccipital facets are also a primitive feature found in berycoids. The extensive interfrontal penetration of the supraoccipital, however, must be ranked as a derived feature.

The dorsicranium shows some slight interspecific differences in detail but not in basic layout. The supraoccipital extends forward to the level of the median sensory pore of the supraorbital lateral line cross-commissure, and clearly separates the frontals posteriorly. The bone's relative anterior extension appears to be least marked in L. angustifrons, L. mariae and L. microlepis ; this is attributable to the

i6

P. H. GREENWOOD

EPI SOC

FR

LATE

PRO

PS BOC

10mm

FIG. 3. Lates calcarifer, neurocranium, left lateral view.

anteriorly more elongate frontals, a lengthening associated with the elongation of the ethmoid region in these species. In Luciolates stappersi, despite the attenuation of its snout, the supraoccipital extends forward to a point level with the anterior orbital margin ; the median sensory pore has a corresponding anterior displacement (Fig. 8).

These four Lake Tanganyika species also have deeper grooves lying between the median supraoccipital crest and the fronto-parietal ridges on each side of the skull, a consequence, perhaps, of their narrower skulls (see below).

All Lates species have a well-demarcated ledge on either side of the supraoccipital crest, the ledge being confluent anteriorly with the supraocciptal bone itself, and extending backwards almost to the posterior margin of the crest. The ledge is narrower and less conspicuous in Luciolates, and is confined to the anterior part of the crest.

The posttemporal fossa is deep in all species except Luciolates stappersi, and in none do its constituent bones meet at the centre of the fossa ; even in the largest specimen examined the fossa is still open, its aperture closed off from the cranial cavity by a tough membrane. Amongst members of the Serranidae the Lates - Luciolates condition is characteristic of small and apparently juvenile fishes ; in larger individuals (many of which are, nevertheless, considerably smaller than adult Lates} the fossa has a completely bony floor. This interfamilial difference would suggest that the Lates condition is the primitive one.

The wide cephalic lateral line canals of the dorsicranium are completely bone enclosed in all species (including Luciolates} . On each side of the skull the continuous supraorbital- temporal canal opens to the exterior through several pores.

Dorsal and lateral skull outlines are essentially similar in L. calcarifer, L. niloticus, L. macrophthalmus and L. longispinis except for a marked narrowing of the inter- orbital region in the two latter species, and a more forward position of the orbit in L. calcarifer. The ethmoid region is relatively short, and the parasphenoid runs

REVIEW OF CENTROPOMIDAE 17

forward in the same line as the base of the braincase. The preotic skull proportions of the largest L. niloticus examined (neurocranial length 228 mm) are very similar to those in a much smaller L. calcarifer skull (103 mm long, from a fish of c. 40 cm S.L.), and differ from those in smaller L. niloticus skulls. The most noticeable differences apparent when these smaller L. niloticus skulls are compared with the skull of an equal-sized L. calcarifer are the relatively more anterior position of the orbit, and the much longer precommissural skull region in the latter species (see below, p. 20) . The skull proportions of large L. niloticus (i.e. skulls > 150 mm long), however,

PTO

PARC

FR

LATE

LC

PRO

10mm

PAR

PTO

SOC

SPO

LATE

EXO

BOC

PRO

5mm

FIG. 4. (a) Lates longispinis. (b) L. macrophthalmus. Neurocranium in left lateral view. For nomenclature of L. longispinis see p. 12 et seq.

18

P. H. GREENWOOD

come to resemble those of L. calcarifer more closely, the resemblance increasing with the size of the skull. Katayama (1956) figures the neurocranium from a L. calcarifer of 28-6 cm S.L. ; judging from this figure there is little difference between a skull of that size and one from a L. niloticus of comparable length. Seemingly the orbital and precommissural skull proportions change much more rapidly in L. calcarifer ; compare, for example, the 103 mm skull of L. calcarifer (S.L. c. 40 cm) with the 123 mm skull of L. niloticus (S.L. c. 48 cm) in Figs 3 and 2.

Compared with the four species from outside Lake Tanganyika, three endemic Tanganyikan species, L. microlepis, L. mariae and Luciolates stappersi (Figs 5-8) show a distinct narrowing of the skull (particularly the braincase), an elongation of the ethmoid region, and an angling of the parasphenoid relative to the basi- occipital. The slope of the parasphenoid is steepest in L. mariae and least in L. microlepis, with Luciolates occupying an intermediate position in the series.

The fourth Lake Tanganyika endemic, L. angustifrons (Fig. 5), is, in most features of its neurocranial profiles, intermediate between the other endemic species and those from outside the lake. Nevertheless, it is clearly differentiated from the latter by

FRC

PARC

LATE

RO

20mm

SPO

PARC PTO

EPI

SOC PAR FIG. 5. Lates angustifrons. Neurocranium in : (a) left lateral view, (b) dorsal view.

Subgenus Lates (see p. 77)

REVIEW OF CENTROPOMIDAE 19

the elongation of its ethmoid region and by the shape of its ethmoid bones, charac- teristics that unite it with the other endemic species from Lake Tanganyika (see below) .

This elongation of the ethmovomerine skull region in all species of Lates (and Luciolates) from Lake Tanganyika immediately distinguishes the group (see Table i),

TABLE i

Relative length of ethmovomerine region in various Lates spp., and in Psammoperca waigiensis

Ethmovomerine length as % of

Neurocranial1 Ethmovomerine2 neurocranial Species length (mm) length (mm) length

Lates calcarifer 103-0 25-0 24-2%

L. niloticus 16-0 5-0 3I-3%

76-0 21-5 28-3%

124-0 30-5 24-8%

228-0 60-5 26-7%

L. macrophthalmus 32-0 8-5 26-5%

L. longispinis 59-5 16-5 28-7%

L. angustifrons 120-0 43-3 36-8%

L. mariae 26-0 10-0 38-5%

77'5 32-0 4I-3%

L. microlepis 44-0 18-0 4i-o%

L. stappersi 71-0 31-0 43'5%

7*-o 32-5 45-8%

103-5 49-o 46-9%

Psammoperca

waigiensis 43-0 14-5 33'7%

1 Neurocranial length : measured directly from the anterior tip of the vomer to the posterior point on the lower margin of the basioccipital facet for the first vertebra.

2 Ethmovomerine length : measured directly from the anterior point of the vomer to that point on the dorsicranium where the lateral ethmoid-prefrontal passes under lateral margin of the frontal.

and argues strongly for their monophyletic origin. The concave posterior face of the lateral ethmoid in L. angustifrons, L. microlepis, L. mariae and Luciolates stap- persi, as compared with members of the L. calcarifer-L. niloticus complex, has a distinct posterior slope (cf. Figs 2-4 with 5-8). Furthermore, and most strikingly, the lateral margins of the bone are much wider (or, as it appears in lateral view, much deeper) and have a pronounced downward slope. In the L. calcarifer- L. niloticus group the posterior margin of the lateral ethmoid is almost vertically aligned, and its lateral margins are narrow and horizontally aligned (cf. Figs 2-4 with 5-8). Once again it is Luciolates that shows the most profound modifications with, in this instance, L. microlepis showing the least modified condition and L. angustifrons and L. mariae (in that order) occupying the intermediate places in the series.

There is little interspecific variation in the morphology of the lateral ethmoids of L. calcarifer, L. niloticus, L. macrophthalmus and L. longispinis (see Figs 2-4).

Subgenus Luciolates (see p. 71)

P. H. GREENWOOD

FRC

PARC

socs

LATE FR

V

PS

PRO

10mm FIG. 6. Lates microlepis. Neurocranium in left lateral view.

All Lates species, and Luciolates, have three facets on each lateral ethmoid ; two, ventrally placed, are for articulation with the palatine, and the third (situated dorsolaterally above the posterior palatine facet) for articulation with the first circumorbital bone (lachrymal). The facets are less well defined in the Tanganyika species, and are most poorly differentiated in Luciolates.

A noticeable feature of the skull in L. calcarifer, L. niloticus and, to a slightly lesser degree, in L. angustifrons is the way in which the anterior wall of the neuro- cranium (i.e. the prootic, pterosphenoid and ascending arm of the parasphenoid) are extended forward beyond the level of the lateral commissure (Figs 2, 3 & 5) ; in L. calcarifer and L. niloticus the tunnel-like ventral part of this extension surrounds all but the anterior half or more of the basisphenoid. This feature is emphasized when the skulls of these species are compared with those of L. mariae, L. microlepis and Luciolates stappersi, species in which there is only a slight prolongation of the neurocranial wall beyond the level of the lateral commissure (cf. Figs 2, 3 & 5 and 6-8). The situation in L. macrophthalmus and L. longispinis is virtually inter- mediate between those in the other two groups. (See Table 2 and Fig. 4.)

Closer examination of the precommissural extension in specimens of L. calcarifer (neurocranial length, ncl., 103 mm), L. niloticus (ncl., 75 mm and above) and L. angustifrons (ncl., 120 mm) reveals the existence of a pterosphenoid pedicle which, through its contact with the parasphenoid anteriorly and the outer lip of a horizontal groove in the prootic, forms a semi-tubular bridge over the oculomotor and profundus nerves and the internal jugular vein (Figs 9 & 10). Rognes (1973) has called a similar stucture in labrids an internal jugular bridge, and that name will be used here.

REVIEW OF CENTROPOMIDAE 21

TABLE 2

Precommissural skull proportion in Lates and Psammoperca

Precommissural length as % of

Neurocranial1 Precommissural2 neurocranial

Species length (mm) skull length (mm) length

Lates calcarifer 103-0 25-0 24'2%

L. niloticus 76-0 10-0 I3'I%

124-0 22-0 *7'7%

L. longispinis 59-5 6-0 10-1%

L. macrophthalmus 32-0 3-0 9'4%

74-0 10-0 13-5%

L. angustifrons 120-0 14-0 H'7%

L. stappersi 71-0 3-5 4-9%

103-5 6-0 5-8% Psammoperca

waigiensis 43-0 2-5 5*7%

1 See Table i (p. 19).

2 Precommissural skull length: measured directly from the anterior margin of the lateral commissure to the anterior margin of the ascending limb of the parasphenoid.

The relative contributions of the prootic and parasphenoid bones to the internal jugular bridge show marked intraspecific variability, and usually differ on either side (see Fig. 90 -d). The parasphenoid contribution is always the least important, the major part of the ventral wall (and the entire groove) coming from the prootic, and the dorsal and lateral walls from the pterosphenoid pedicle. Except in the small L. niloticus skulls examined (see below) there is always some contact between the three bones at the orbital (i.e. front) margin of the bridge. Since the smallest available skulls of L. calcarifer and L. angustifrons measure 103 mm and 120 mm long respectively, no comment can be made on the interrelationship of these bones in small individuals of those species.

Apparently correlated with the degree of pterosphenoid development and the development of the precommissural braincase is the extent to which the auto- sphenotic is prolonged anteriorly. The correlation is a positive one in species with an extensive precommissural braincase and a well-developed pedicle (i.e. L. calcarifer and L. niloticus). The anterior extension of the autosphenotic is least marked in Luciolates stappersi, L. mariae and L. microlepis, and is of intermediate length in L. angustifrons, L. macrophthalmus and L. longispinis.

Before describing and comparing the precommissural crania for all Lates and Luciolates species, it is necessary to consider the ontogenetic changes involved in the production of an adult L. niloticus-type pterosphenoid pedicle and internal jugular bridge.

The smallest L. niloticus skulls examined (12 mm long) have no noticeable pre- commissural extension of the braincase ; the parasphenoid does not contact the pterosphenoid and there is no bony bridge over the nerves and blood vessel (Fig. ga). At the ontogenetic stages represented by those skulls there is also no obvious

P. H. GREENWOOD

PTO PARC

PRO

EXO

10mm

PARC PTO

SOC FIG. 7. Lates mariae. Neurocranium in : (a) left lateral view, (b) dorsal view.

pterosphenoid pedicle, but a narrow ligament runs from the lower, anterior part of the pterosphenoid to the outer rim of the weakly developed prootic groove lying below the internal jugular vein (Fig. ga) ; in effect, the ligament occupies the position later taken by the pterosphenoid pedicle arm of the internal jugular bridge. At its dorsal base, the ligament is attached to a small spur of bone on the pterosphenoid, which I would interpret as an incipient pedicle.

In progressively larger skulls (i.e. to a length of 76 mm), there is a gradual develop- ment and down-growth of the pterosphenoid pedicle, and of a dorsally directed

a

LATE

REVIEW OF CENTROPOMIDAE

PAR

FR

EXO

PRO

20mm

SPO LC PRO

LATE

BOG

1C

PTO

SPO PTO

EPI

FR

PAR

FIG. 8. Lates stappersi. Neurocranium in : (a) left lateral view, (b) ventral view, (c) dorsal view. For details on the altered generic placement of this species (previously Luciolates stappersi) see p. 50.

P. H. GREENWOOD

spur-like development from the prootic lateral to the internal jugular groove. As a consequence of these growth patterns (and a dorsal extension of the ascending parasphenoid arm) an at first narrow (Fig. gb), but gradually broadening, bony ridge is formed over the internal jugular vein and the associated oculomotor and profundus nerves. Concurrently, there is a gradual forward growth of the pre- commissural region of the skull.

A L. niloticus skull 76 mm long has the pterosphenoid pedicle and precommissural skull developed to an extent comparable with that in a L. angustifrons skull 120 mm long. Growth of the precommissural skull wall in L. niloticus continues beyond this

2mm

1mm

FR

PTS

5 mm

FIG. 9. Outline figures of internal jugular bridge, pterosphenoid pedicle and precommis- sural skull to show in : (a) & (b) growth changes in L. niloticus and in : (c) & (d) variability in the bridge of a single specimen of L. longispinis. (a) Lates niloticus ; left lateral view, neurocranial length 12 mm. Note ligamentous connection between spur of pterosphenoid and process on prootic. (b) L. niloticus ; left lateral view, neurocranial length 16 mm. Note downgrowth of pterosphenoid spur (= pedicle) to join prootic process, (c) & (d) L. longispinis. Left and right sides of skull showing variation in the interrelationships of bones contributing to the internal jugular bridge. Note direct pterosphenoid -para- sphenoid contact in (d).

REVIEW OF CENTROPOMIDAE

LC

PTSP

PRO

10mm

FIG. 10. Lates calcarifer. Outline figure to show relative hyperdevelopment of pre- commissural skull (left lateral view), especially the pterosphenoid pedicle and internal jugular bridge. Compare with Figs g(c) & (d), 5(a), 2(a). From a skull of 10-3 cm neurocranial length.

point ; in a skull 124 mm long it has attained, however, the overall morphology and proportions seen in the largest skull examined (230 mm long).

Unfortunately no L. angustifrons skulls longer than 120 mm could be obtained so it has not been possible to determine the definitive form in that species. However, to judge from the totality of interspecific differences seen in skulls of about the same size it seems unlikely that this region of the neurocranium in L. angustifrons ever attains the proportions found in either L. niloticus or L. calcarifer (see above, p. 21).

As was noted earlier (p. 20), the precommissural skull region in L. macrophthalmus and L. longispinis is less well developed than in adult L. calcarifer and L. niloticus. It must, however, be remembered that members of the two former species reach a much smaller adult size (Worthington, 1929, 1932).

In a L. macrophthalmus skull 74 mm long (from a fish of 275 mm S.L.) the internal jugular bridge and the pterosphenoid pedicle have about the same degree of develop- ment as in a 76 mm long skull of L. niloticus (S.L. c. 290 mm) or a 120 mm long skull of L. angustifrons (S.L. c. 350 mm) ; the situation is similar in a slightly larger individual of L. macrophthalmus (320 mm S.L., neurocranial length no mm) Both specimens have a narrow parasphenoidal contribution to the bridge which is otherwise formed mainly from the pterosphenoid pedicle and the prootic spur. The smallest L. macrophthalmus skull examined (32-5 mm long, from a fish of no mm S.L.) shows a degree of development comparable with that in a L. niloticus skull only 16 mm long ; namely, a ligamentous bridge, and the pterosphenoid pedicle manifest only as a small spur of bone (Fig. ga) .

Conditions in L. longispinis, as seen in a skull 59 mm long (from a fish c. 250 mm S.L.), are close to those in the 74 mm skull of L. macrophthalmus, but the bridge is a little narrower. In a larger skull (70 mm long from a fish 275 mm S.L.) the bridge and pedicle, and the precommissural skull proportions are similar to those in the

26 P. H. GREENWOOD

76 mm skull of L. niloticus described above (p. 24), with a distinct pedicle and, at least on one side of the skull, a parasphenoidal contribution to the internal jugular bridge (Fig. gd) ; on the left side of this specimen, the ascending parasphenoid limb fails to reach the level of the upper lateral margin of the prootic (Fig. gc).

It would seem, then, that the internal jugular bridge and the precommissural skull in both L. longispinis and L. macrophthalmus are comparable with those in similar-sized skulls of L. niloticus, or are perhaps a little less advanced in some individuals. In other words, the adult skull of L. macrophthalmus and L. longispinis retains at least some of the pre-adult features of L. niloticus.

A really marked reduction in the adult precommissural braincase and in the pterosphenoid pedicle and internal jugular bridge is seen in the skulls of three Lake Tanganyika taxa, namely L. mariae, L. microlepis and Luciolates stappersi. (This region of the skull is also relatively reduced, as compared with L. niloticus, in the fourth Tanganyika species, L. angustifrons , see pp. 20-25 above.)

In none of these three species does the parasphenoid contact the pterosphenoid, always being separated from that bone by the prootic (Figs 6-8). No trace of a pterosphenoid pedicle, even as a low ridge, is detectable in the three Luciolates stappersi skulls I have examined (neurocranial lengths 71 (f. 2) and 113 mm), but a low ridge was found in the largest of the three L. mariae skulls (26-0, 77-5 and 104-0 mm long).

A similar ridge is developed on the right but not the left pterosphenoid of a 44 mm long skull of L. microlepis. A larger skull (95 mm) of L. microlepis, however, has a well-developed, broad-based but distally narrowed pedicle which reaches almost to the level of the prootic spur (Fig. 6) . It is connected to the prootic spur by a short section of what appears to be ossified ligament.

Thus, of these three species, L. microlepis is the only one in which the ptero- sphenoid pedicle makes a significant contribution to the internal jugular bridge. Even in the largest skulls of L. mariae and Luciolates stappersi there is only a liga- mentous bridge, a condition directly comparable with that in the smallest specimens of L. niloticus, except that in the Tanganyika fishes the ligament appears to be ossified. In other words, the precommissural braincase in large specimens of L. mariae and Luciolates stappersi (standard lengths 390 and 415 mm respectively) is like that in L. niloticus of about 60 mm standard length, while that of a L. micro- lepis 390 mm standard length is comparable with a L. niloticus of about 130 mm S.L.

The pterosphenoid -prootic ligament found in juvenile L. niloticus and adults of Tanganyika taxa described above is readily separated from both its bones of attach- ment. Thus it seems unlikely that it is truly part of the pterosphenoid pedicle. Presumably the ligament is replaced by the pedicle as it grows down to meet the spur from the outer rim of the prootic groove. The large L. microlepis specimen noted above represents a late phase in this developmental sequence, the small L. macrophthalmus (ncl., 32-5 mm ; p. 25) an early phase, and the adult condition in L. niloticus, L. calcarifer and L. angustifrons the terminal state.

An internal jugular bridge is of sporadic and phyletically widespread occurrence amongst living teleosts. Rognes (1973) gives detailed accounts of the bridge in labrine Labridae, and reviews records of its occurrence in other groups. I can

REVIEW OF CENTROPOMIDAE 27

confirm its presence in certain ostariophysans (Alburnm ; see also Holmgren & Stensio, 1936, for Abramis), certain scorpaeniforms (Enophrys bison, Scorpaena scrofa, Trigla hirudo, see also Allis, 1909 ; Allen [1905] describes a bridge in Ophidian [Hexagrammidae]), and in several percoids (Epinephelus species [but not other serranids], Stizostedion volgensis, Perca fluviatilis [but not, apparently, in Gymno- cephalus}}, and in some sphyraenoids (Sphyraena sp.).

In the majority of cases where a bridge is present, it is of the type found in juvenile L. niloticus, namely a ligament (generally ossified) joining a reduced pterosphenoid pedicle to a process developed on the prootic (see above, p. 21). Only in Enophrys bison is a bridge of the L. angustifrons type present.

This list, based on samples taken from the families represented in the dry skeleton collection of the British Museum (Natural History), cannot by any means be con- sidered complete, especially since the bridge is not always preserved in dry skeletons. Nevertheless, it is interesting to find that in none of the beryciform skeletons at my disposal is there any indication of a bridge nor even of the pterosphenoid pedicle (which is usually obvious even if the ligamentous part of the bridge is missing). Neither a bridge nor a pterosphenoid pedicle was noted in any of the Mesozoic beryciforms described by Patterson (1964).

Superficially, Salmo trutta has what appears to be an internal jugular bridge, but closer inspection shows that it is formed entirely within the prootic. Thus it would seem to be homologous with the 'prelateral commissure' described by Rognes (1973) in the labrid Ctenolabrus exoletus (see Rognes, op. cit., fig. 59).

The pterosphenoid pedicle has a long history in actinopterygian fishes, being well developed in some leptolepids and pholidophorids, in Amia and its fossil relatives Sinamia and Ellenes, and at least partially developed in some palaeoniscids (Patter- son, 1975). As Patterson (op. cit., p. 409) observes : ' ... It is therefore likely that a pterosphenoid pedicle of some sort, or at least the potentiality to develop such a structure is a primitive actinopterygian feature.'

Since the pterosphenoid pedicle is an integral part of the internal jugular bridge (see above) and because this bridge is of widespread occurrence among teleosts, one may conclude that the bridge too is a primitive feature.

The absence or great reduction of the bridge and pedicle in certain Lates species can, therefore, be interpreted as an apomorphic feature, at least when individuals of these species attain a size at which the bridge would otherwise be present in related taxa. Lates macrophthalmus and L. longispinis (both species with reduced bridges) are examples of the situation where maximum adult size is about equal to that in preadult L. niloticus and L. calcarifer ; at that size, specimens of L. niloticus (and presumably L. calcarifer} have a poorly developed bridge. Thus, it is probably correct to consider L. macrophthalmus and L. longispinis as plesiomorphic with respect to the bridge character.

Hyopalatine arch and the preoperculum (Figs n & 12)

Apart from slight proportional changes in, particularly, the length of the palatine and ectopterygoid bones of the Tanganyika species, there is little interspecific variation in the hyopalatine arch of Lates species (see Figs n & 12).

P. H. GREENWOOD

PAL

POP.

FIG. ii. Hyopalatine arch, right (including preoperculum) in lateral view of (a) Lates mariae, (b) L. niloticus.

The hyomandibula has two well-defined articulatory facets interconnected by a thin lamina of bone.

The metapterygoid has a strong sutural union with the hyomandibula and with the posterior tip of the expansive endopterygoid. There is no true metapterygoidal lamina (sensu Katayama, 1956, and Gosline, 1966) but a slight ridge is detectable in the position where a lamina would occur ; also, in many species there is a small foramen (or fenestra) in the metapterygoid at the postero-dorsal end of the ridge. I would interpret these structures as the remnants of a greatly reduced metaptery- goidal lamina.

Fine viliform teeth cover the entire ventral surface of the palatine. A similarly shaped (i.e. elongate ovoid) tooth patch occurs on the medial aspect of the anterior

REVIEW OF CENTROPOMIDAE

29

arm of the ectopterygoid, sometimes extending a short way onto the vertical arm of that bone as well.

The autopalatine is a fairly stout bone. Anteriorly, on its medial face are two well-defined articulatory surfaces for contact with the ethmoid ; dorsally there is a weakly demarcated facet for articulation with the lateral ethmoid. A panhandle- like, cartilage-tipped projection from the upper surface of the palatine provides articulation between this bone and the maxilla.

In most details, including the presence of a reduced metapterygoidal lamina, the hyopalatine arch of Luciolates resembles that of Lates, particularly the Lake Tangan- yika species of the genus. However, all the bones (especially the endopterygoid) are thinner and the palatine is less robust, with poorly demarcated articulatory facets. The palatine tooth patch is much narrower in Luciolates, and there is a great

10mm

PAL

POP

10mm

FIG. 12. Hyopalatine arch, left side (including preoperculum) in lateral view of (a) Lates stappersi, (b) L. angustifrons.

30 P. H. GREENWOOD

reduction in the area of the ectopterygoid teeth, the tooth-patch being either reduced to a small oval near the ectopterygoid-palatine articulation or it is completely absent. In one of the three skeletons examined the tooth-patch was present on one side only.

Like the hyopalatine arch, the preoperculum in Lates shows little interspecific variation, although it does show some intraspecific variability. The entire posterior margin of the vertical limb, except for a short length near its ventral angle, is finely serrate, the individual serrae are slender, sharp-pointed and tall. In very large specimens of L. niloticus (> 150 cm S.L.), the serrations are considerably reduced in height, and consequently the posterior margin of the bone is merely irregular (see also Sorbini, 1973).

At the posterior angle between the horizontal and vertical preopercular limbs there is a large, posteriorly directed and triangular spine (Fig. n) ; very rarely this spine is subdivided almost to its base, resulting in two narrower but still triangular spines. On the horizontal limb there are generally three triangular spines, each slightly shorter and narrower than the spine at the bone's posterior angle. In larger L. niloticus the spines become irregular in outline, relatively shorter, and may have rounded rather than acute points.

Although three preopercular spines are modal for all species but L. macrophthalmus, a fourth spine is sometimes developed either on one or both sides. Usually the extra spine is a distinct entity, but sometimes it appears merely to be a subdivision of one of the other spines. Lates macrophthalmus is apparently exceptional in having a high proportion of individuals with four spines (seven of the eleven specimens examined). The proportion of four-spined fishes amongst samples of the other species is : L, calcarifer, none out of 18 ; L. niloticus, 7 out of 31 ; L. longispinis, 3 out of 6 (a high proportion, approaching that of L. macrophthalmus, which may be a related taxon, see p. 13) ; L. angustifrons, none out of 14 ; L. mariae, 4 out of 20 ; L, microlepis, 2 out of 27.

The occurrence of four-spined individuals may be a population feature, hence my reservations about the seemingly unusual condition in L. macrophthalmus. All but one of the L. niloticus specimens with four spines came from a single sample (incidentally, the largest available for L. niloticus and one much larger than was available for any other species).

Luciolates stappersi (Fig. I2a) also has a serrated posterior margin to the vertical preopercular limb, but here the serrations are lower and less well defined (in this respect resembling the condition in 16-20 mm standard length L. niloticus}. The spine at the preopercular angle is always present and prominent, although it is somewhat finer than in any Lates species. The horizontal limb may have three large and relatively short spines, but specimens with two or three groups of very small spines, or even what amounts to a crenellated border, are common. The incidence of bilateral asymmetry in the type of spination is also high.

In both Lates and Luciolates the preopercular lateral line canal is completely bone enclosed, with its pores confined to the horizontal limb.

Although a serrated or otherwise ornamented vertical preopercular limb is of common occurrence amongst the lower percoids (e.g. in the Serranidae), the presence

REVIEW OF CENTROPOMIDAE

of large and discrete spines on the horizontal limb and at its angle is extremely rare (Percalates and Siniperca [Serranidae] are, as far as I can determine, the only taxa having the same type of preopercular ornamentation as Lates). A similar generali- zation can be made for the lower percomorphs (sensu Rosen, 1973 ; for example, the 'Beryciforms'). Thus, it seems reasonable to conclude that the ventral pre- opercular ornamentation in Lates (and probably other centropomids as well, see below) is a derived condition (see also Rosen, 1973 : 469). Luciolates too can be included in this generalization, the condition here being interpreted as the secondary simplification of a derived condition mimicking a plesiomorphic one.

Circumorbital bones (Figs I3a-d)

The greater part of the ventral margin to the first circumorbital bone (the lachry- mal) is finely serrated in all Lates species ; only a short anterior part is smooth. In all species the entire margin of the second circumorbital is also serrated.

The infraorbital lateral line canal in Lates and in Luciolates is enclosed throughout its length, communicating with the exterior through five pores in the lachrymal, one anteriorly on the third circumorbital bone, and through other pores found between successive bones in the series.

LAC

SOS

LAC

LAC

L_LAC

FIG. 13. Circumorbital bones (right side) in : (a) & (b) Lates niloticus, and in : (c) & (d) L. stappersi ; (a) and (c) lateral view, (b) and (d) viewed dorsally and somewhat anteriorly.

32 P. H. GREENWOOD

All species (including Luciolates stappersi) have a well-defined facet developed at about the middle of the upper lachrymal margin ; it articulates with a similar facet on the lateral ethmoid.

There is a general similarity in the shape of the first two circumorbital bones in all Lates species, although the three species from Lake Tanganyika (L. angustifrons, L. mariae and L. microlepis} have a slightly more elongate lachrymal. These species (except L. angustifrons) also differ from L. niloticus, L. calcarifer, L. macroph- thalmus and L. longispinis in having a relatively more elongate fifth circumorbital, and in having much narrower bony flanges developed from the ventral contours of the cylindrical canal-bearing portions of the third, fourth and fifth bones.

Greatest departure from the L. niloticus -L. calcarifer situation is seen in the re- duced size of the subocular shelf in the Tanganyika species, again excepting L. angustifrons where the shelf is like that in L. niloticus and L. calcarifer, viz. a thin but broad bony plate that curves upwards from the third circumorbital to He along the entire length of the fourth bone. In L. microlepis the subocular shelf is reduced in width, and just reaches upwards to the level of the articulation between the third and fourth circumorbitals ; in L. mariae there is a further and marked reduction in width, and the shelf barely reaches to the level of the articulation between the bones. Both species have the ventral flange on the third and fourth circumorbitals reduced to a thin flange.

These reductional trends are carried further in Luciolates, where the serrations on the lachrymal are very weak and are confined to about the posterior third of the bone ; serrations are completely absent from the second circumorbital. The facet for articulation with the lateral ethmoid is weakly developed, and its origin from the lachrymal is far less well defined than in the other species. The subocular shelf is, relatively, a little narrower than in L. mariae, but it does extend further up the fourth circumorbital (along about its lower third) ; see Fig. I3c-d. The depth of the ventral flange on the second to fourth circumorbitals is almost comparable with L. mariae, as is the flange on the fifth bone. In their gross morphology, the circum- orbital bones in Luciolates stappersi are noticeably more elongate than those in any Lates, including the other Lake Tanganyika species. Apart from differences in the overall proportions of the first, third and fourth bones, the morphology of the entire series in an adult Luciolates of 105 mm standard length closely resembles that in a juvenile Lates niloticus 32 mm long.

Opercular bones (Figs I4a-b)

There is little variation in the operculum, suboperculum and interoperculum of Lates and Luciolates, apart from a slight relative elongation of the interoperculum in the Tanganyika species, especially Luciolates. In all taxa there is a well-defined, curved ridge on the medial face of the interoperculum against the upper, concave surface of which the proximal end of the epihyal articulates.

The operculum (Fig. I4a-b) is armed with a single stout spine formed from the posterior tip of the near-horizontal strut which runs backwards from the hyo- manibular facet of the bone.

REVIEW OF CENTROPOMIDAE OP

33

SOP

10mm

10mm

IOP

OP.

SOP

IOP

FIG. 14. Opercular series (medial aspect of bones from right side) in : (a) Lates angustifrons, (b) L. stappersi, (c) Psammoperca waigiensis.

A characteristic feature in all taxa is the thinness of the sub- and interopercular bones.

Jaws (Figs 15 & 1 6)

Both the maxilla and the premaxilla show little interspecific variation amongst Lates species, and are of the generalized percoid type. There is also little difference between Lates and Luciolates in the morphology of these bones. However, in Lates species the ascending process of the premaxilla is from 30 to 60 per cent higher than the articular process (apparently being lowest in the Lake Tanganyika species) ; it is only a little higher than the ascending process in Luciolates stappersi.

The premaxillary dentition in all Lates species is composed of numerous close-set rows of small conical to subconical teeth which form a villiform covering to the complete width of the bone over almost its entire length (Figs I5b-c).

34

ASCR

P. H. GREENWOOD

10mm

PMAXP

FIG. 15. Lates niloticus. (a) Premaxilla (left) lateral view, (b) Dentary (left) lateral view, (c) Dentary (left) occlusal view. (All from Greenwood & Howes, 1975.)

5mm

ASCP

FIG. 1 6. Lates stappersi. (a) Premaxilla (right), occlusal view, (b) Premaxilla (right), lateral view, (c) Dentary (right), lateral view.

REVIEW OF CENTROPOMIDAE

35

Although most premaxillary teeth in Luciolates stappersi are like those in Lates, the species is noteworthy for the presence of at least one greatly enlarged and two slightly smaller caniniform teeth adjacent to the symphysial surface of the pre- maxilla ; a few neighbouring teeth may also be somewhat enlarged. In general the larger teeth are linearly arranged, with the largest one situated lingually.

The upper jaw elements in Lates and Luciolates show no derived characteristics and, of course, both genera retain the supramaxilla. The enlarged median teeth of Luciolates, however, would seem to be a derived feature.

Like the upper jaw, the lower jaw elements (dentary, anguloarticular and retro- articular) show little interspecific variation. In Luciolates the anguloarticular is relatively shallow, but otherwise has a typical 'Lates' form.

The dentition of the dentary mirrors that on the premaxilla, except that in Luciolates the outermost tooth row is composed of noticeably larger and more clearly caniniform teeth, and there are no enlarged symphysial teeth.

Branchial skeleton (Figs 17-19)

The branchial skeleton in both Lates and Luciolates is of a generalized percoid type (see Rosen, 1973), and it shows few interspecific differences, apart from a relative

E2TP

E2

E4 E3

E2

E3TP

1mm

FIG. 17. Lates niloticus. Branchial skeleton, dorsal part (drawn from an alizarin pre- paration, 40 mm S.L.). (a) Dorsal aspect of left side, (b) Ventral aspect (left side) to show upper pharyngeal teeth and tooth plates.

34

ASCR

P. H. GREENWOOD

FIG. 15. Lates niloticus. (a) Premaxilla (left) lateral view, (b) Dentary (left) lateral view, (c) Dentary (left) occlusal view. (All from Greenwood & Howes, 1975.)

ASCP

5mm

FIG. 1 6. Lates stappersi. (a) Premaxilla (right), occlusal view, (b) Premaxilla (right), lateral view, (c) Dentary (right), lateral view.

REVIEW OF CENTROPOMIDAE

35

Although most premaxillary teeth in Luciolates stappersi are like those in Lates, the species is noteworthy for the presence of at least one greatly enlarged and two slightly smaller caniniform teeth adjacent to the symphysial surface of the pre- maxilla ; a few neighbouring teeth may also be somewhat enlarged. In general the larger teeth are linearly arranged, with the largest one situated lingually.

The upper jaw elements in Lates and Luciolates show no derived characteristics and, of course, both genera retain the supramaxilla. The enlarged median teeth of Luciolates, however, would seem to be a derived feature.

Like the upper jaw, the lower jaw elements (dentary, anguloarticular and retro- articular) show little interspecific variation. In Luciolates the anguloarticular is relatively shallow, but otherwise has a typical 'Lates' form.

The dentition of the dentary mirrors that on the premaxilla, except that in Luciolates the outermost tooth row is composed of noticeably larger and more clearly caniniform teeth, and there are no enlarged symphysial teeth.

Branchial skeleton (Figs 17-19)

The branchial skeleton in both Lates and Luciolates is of a generalized percoid type (see Rosen, 1973), and it shows few interspecific differences, apart from a relative

E2TP

E4 E3

E2

E3TP

1mm

FIG. 17. Lates niloticus. Branchial skeleton, dorsal part (drawn from an alizarin pre- paration, 40 mm S.L.). (a) Dorsal aspect of left side, (b) Ventral aspect (left side) to show upper pharyngeal teeth and tooth plates.

38 P. H. GREENWOOD

elongate so that it overlaps the joint between the two bones. The plates associated with the third arch are generally the largest of the series, approach one another medially and cover a great deal of the third basibranchial.

In the one available branchial skeleton of L. angustifrons the individual plates appear to have fused together on the first two gill arches to form a long tooth plate on each side of the arch. A similar arrangement is seen in the alizarin preparation of a small (96 mm S.L.) Luciolates stappersi, but this specimen differs in other respects, especially in having a single, median plate on the third basibranchial and a small plate on each hypobranchial of that arch. The arches dissected from a much larger specimen (270 mm S.L.) have the plates of the third basibranchial narrowly separated medially, a long plate at the base of the second gill arch and a small plate intercalated between it and the basal plate of the first arch. Clearly, at least in this species, there can be quite considerable individual variability in the pattern of tooth plate distribution (see Nelson, 1969 : 500-501, for a description of variation in another percoid, Pomatomus saltatrix [Pomatomidae]).

Hyoid arch (Figs 20 & 21)

The hyoid arch in Lates and Luciolates is of a basal percoid type, with dorsal and ventral hypohyals, a large and complete 'berycoid' foramen and seven branchio- stegal rays. There is remarkably little interspecific variability in the shape of this

Eh

'BcF'

BrR

10mm

10mm

FIG. 20. Hyoid arch and branchiostegal rays (right side), viewed laterally, in (a) Lates stappersi, (b) L. mariae, (c) L. niloticus.

REVIEW OF CENTROPOMIDAE

39

10mm

FIG. 21. Urohyal. (a) & (b) Lates stappersi (left lateral and ventral views respectively), (c) & (d) L. angustifrons (left lateral and ventral views respectively).

arch, without even, as might be expected, clear-cut proportional differences in the arches from species with elongate skulls (i.e. the Tanganyika species).

The first four branchiostegal rays articulate with the ceratohyal, the fifth with either the ceratohyal or at the cerato-epihyal suture, and the last two rays (the stoutest and broadest of the series) articulate with the epihyal. The first three rays contact the ventral face of the ceratohyal, the other four lie on the lateral aspect of the cerato- or epihyal. These latter rays have progressively broader heads, with the dorsal outline of the head on the last two, or occasionally three rays somewhat indented.

The basihyal is an elongate bone, spatulate in dorsal outline, and does not carry a tooth plate.

The urohyal (Fig. 21) is similar in all species, but is markedly more elongate in Luciolates, even when it is compared with the urohyal in the Lates species of Lake Tanganyika.

Pectoral girdle and associated bones (Fig. 22)

The pectoral girdle shows few interspecific or intergeneric differences, either in its overall proportions or in the shape of its individual bones. Judging from the only

4o P. H. GREENWOOD

sc.

CL CL

COR

COR

FIG. 22. Pectoral girdle (right half) in : (a) Lates calcarifer, (b) L. stappersi. (The supracleithmm is removed from this specimen.)

available skeleton of Luciolates stappersi the horizontal limb of the cleithrum is somewhat narrower than it is in Lates, and has less ventrolateral curvature ; the scapula and coracoid are also noticeably deeper in this species and the foramen enclosed between the coracoid and the medioventral margin of the cleithrum is larger (cf. Figs 22a & 22b).

The posterior angle of the cleithrum in both genera is expanded and slightly protracted, and its hind margin is serrated. These serrations are most numerous in L. calcarifer, L. niloticus and L. angustifrons (6-10 serrae, the uppermost often ill-defined), are fewer in L. macrophthalmus (5-7) and fewest (3 or 4) in L. microlepis and L. mariae. Judged on the size range of available material for any one species, the number of serrae is not obviously correlated with the fish's size, and the number may differ on either side of an individual.

In Luciolates the cleithral projection can have a smooth posterior border or be ornamented with from one to three weak serrations ; as in Lates there are lateral discrepancies in the number of serrae.

The three upper radials articulate with the scapula, and the fourth either articu- lates with the coracoid or partly with the coracoid and partly with the scapula.

The supracleithrum in both Lates and Luciolates is a slightly curved, dagger- shaped bone showing no interspecific variability in shape or size.

The first postcleithrum is a flat, scale-like bone, the second is elongate and spini- form (Fig. 23). No obvious interspecific or intergeneric differences were detected in either element.

The posttemporal is characterized, in both genera, by a deep and dorsally directed oval pocket formed in the body of the bone immediately lateral to the base of its

REVIEW OF CENTROPOMIDAE

PC2

FIG. 23. Lates calcarifer. Postcleithra (left).

intercalar limb and a little anterior to the facet for articulation with the supra- cleithrum (Fig. 24). The pocket opens dorsally into the lateral line canal, and its lateral wall bulges slightly outwards ; in alizarin preparations of a young L. niloticus this wall has a pitted, 'strawberry-skin' appearance similar to that of the auditory bulla in many clupeomorph fishes. The pit is occupied by the distal end of the ligament which runs from the posttemporal to the tunica externa of the swimbladder (see p. 47 below).

Posteriorly, the shield-like body of the posttemporal is serrated, the extent and size of the serrations apparently not differing between the various species.

The extrascapula is a small Y-shaped bone, largely occupied by the lateral line sensory canals it carries (i.e. the supratemporal and temporal lines), and shows little

FIG. 24. Posttemporal in : (a) & (d) Lates niloticus (right bone), (a) lateral, (d) medial aspect, (b) & (e) Psammoperca waigiensis (left bone), (b) lateral, (e) medial aspect, (c) & (f) Centropomus undecimalis (left bone), (c) lateral, (f) medial aspect.

42 P. H. GREENWOOD

interspecific variability. It articulates closely with the posttemporal, the two bones together partially covering the posterior part of the posttemporal fossa.

Vertebral column (Figs 25 & 26)

The total count in all Lates species and in Luciolates stappersi is 25, viz., n ab- dominal, 13 caudal and the fused first ural and preural centra.

There are nine pairs of pleural ribs, the first pair associated with the third vertebra. On those vertebrae with parapophyses, (the eighth and subsequent abdominals have obvious parapophyses but a small projection is visible on the seventh), the ribs articulate with the posterior face of the parapophysis ; at least in Lates the rib articulation on the preceding centra is through a shallow facet whose ventral lip is slightly produced laterally.

The parapophyses in Luciolates differ from those in Lates in being almost vertically aligned, and by having, in all bar the first pair, a horizontal strut joining the para- pophyses of each centrum near their distal tips. Also, in this genus the articulatory pit on the first three rib-bearing centra has no ventral lip, but on the fourth rib- bearing centrum (i.e. the sixth abdominal vertebra) the lip is sufficiently produced to resemble a very short parapophysis.

I have been able to check the dorsal ribs in only two species (L. niloticus and Luciolates stappersi). Lates niloticus has epineural ribs associated with the first

FIG. 25. First three abdominal vertebrae in : (a) Lates angustifrons, (b) Psammoperca waigiensis, (c) L. stappersi, (d) Centropomus ensiferus.

REVIEW OF CENTROPOMIDAE 43

two vertebrae, and epipleurals present on the first six pairs of pleural ribs ; Luciolates stappersi has epineurals as in Lates, but the epipleurals apparently are confined to the first three pairs of ribs only.

The first three abdominal vertebrae are the most individually distinctive elements in the entire column (Fig. 25). The second vertebra is characterized by the great expansion anteroposteriorly of its neural spine, which is 2-3^ times broader than the spine of the first vertebra and about twice as broad as the spine of the third vertebra. Not only is the spine expanded but it has a characteristic outline. The anterior and posterior margins run almost parallel to one another for most of the spine's height (rather than converging with one another), and the spine narrows smoothly at a point about three-quarters of its height above the centrum. At this point, the front margin curves backwards to meet the posterior margin which may be almost vertical or, and more generally, it may have a slight posterior curvature. The hind margin of the first neural spine is closely applied to the front of the second spine, but the third neural spine slopes away from the second at a marked angle.

Neither the first nor the third neural spine has parallel margins except basally ; the margins slope towards one another over most of their height, giving the spine a narrowly triangular outline.

A variety of skeletons covering a wide size range of individuals (c. 16 to 1000 mm) is available only for L. niloticus. These skeletons indicate that the relative antero- posterior expansion of the second neural spine may at first show a positively cor- related increase with increasing standard length, but that in very large fishes the spine becomes relatively narrower.

There are quite marked interspecific differences in the length -height proportions of certain centra, particularly those in the abdominal region of the column. In the descriptions that follow the first three abdominal vertebrae are excluded since those are not affected by proportional changes ; all measurements are maxima.

In L. niloticus, L. macrophthalmus and L. longispinis, the abdominal and caudal centra are of approximately equal length and depth or are only a little longer than deep (the latter proportions applying especially to caudal vertebrae) . Lates calcarifer has caudal centra like those in the former species but its abdominal centra are slightly more elongate.

The abdominal centra in L. angustifrons have proportions similar to those in L. calcarifer, as do the first five or six caudal centra. Beyond that point, however, the caudal centra are noticeably more elongate (i.e. they are about i| times longer than deep). Lates mariae shows slightly greater elongation of its abdominal centra (c. 1 1 times longer than deep), but the caudal centra are similar to those in L. angustifrons. This trend is accentuated in L. microlepis where, although the abdominal centra have proportions like those of L. mariae, the posterior caudal elements are from if to twice as long as deep ; the anterior caudals, however, are still about i \ times longer than deep. Finally, in Luciolates, all the centra are clearly elongate (c. if to twice as long as deep) and there is no difference in pro- portions between the caudal and abdominal elements of the column.

There are three predorsal bones in all Lates and Luciolates (pace Fraser, 1968), the proximal end of the first lying just anterior to the first neural spine, the ends of

44

P. H. GREENWOOD

the second and third bones lying, respectively, in front of and behind the second neural spine.

Caudal fin skeleton (Fig. 26)

There is but slight interspecific variation in the caudal skeleton of Lates (Fig. 26). All species have two epurals, two uroneurals and five hypurals ; the first, second and fifth hypurals are autogenous (as is the parhypural and the haemal arch and spine of the third preural centrum). The hypurapophysis is weakly to moderately developed. The neural spine on the second preural vertebra is reduced to a low crest in all species.

The principal caudal fin ray formula for all species is 1,8 + 7,1.

Although the caudal skeleton in Luciolates is basically similar to that in Lates, it differs in having the first to fourth hypurals fused into a single plate except for a narrow proximal gap between the fused first and second, and the fused third and fourth hypurals ; the fifth hypural is free and is autogenous basally.

One small specimen (96 mm S.L.) of Luciolates stappersi, an alizarin preparation, has a small and free sixth hypural, the fifth hypural in this specimen being fused in with the third and fourth. Unlike the other Luciolates examined (by dissection and radiographically) the second and third hypurals in this fish are not apposed over their distal halves but are fused proximally instead.

As in Lates, there are 1,8 + 7,1 principal caudal rays in Luciolates.

The caudal fin margin in adult L. calcarifer, L. niloticus, L. macrophthalmus , L. longispinis and L. angustifrons is weakly truncate to markedly subtruncate (nearly rounded), in L. mariae it is truncate to weakly emarginate, but in L. micro- lepis it is so strongly emarginate as to be almost crescentic. (In juveniles, however,

NaPU2

PU-i+U-i

UR

Caudal fin skeleton in : (a) Lates niloticus, (b) Eolates gracilis (After Sorbini, 1973).

REVIEW OF CENTROPOMIDAE 45

the margin is distinctly truncate [see Poll, 1953] or weakly subtruncate [see Boul- enger, 1915].) A crescentic margin is also developed in Luciolates stappersi, and is deepest in fishes over 150 mm standard length.

Dorsal and anal fins

The number of pterygiophores supporting the rays of the dorsal fin (or fins) shows some slight interspecific variation ; viz : L. calcarifer 18, L. niloticus 18 or 19, L. microlepis 19 (rarely 20), L. macrophthalmus 18 (rarely 19), L. longispinis 18 or 19, L. angustifrons 19 and L. mariae 19. Each of the first eight or nine pterygiophores carries a single spine, and no medial radials are associated with these bones. An examination of the dorsal fin ray supports in alizarin preparations of small (16-20 mm S.L.) L. niloticus suggests that the medial radial fuses with the proximal one (the pterygiophore) to form the elongate head of that bone. Distinct medial radials are also absent from those pterygiophores carrying the branched dorsal fin rays.

Luciolates stappersi has 19 dorsal pterygiophores, the first nine of which bear a single spine (again without the interposition of a medial radial). The seventh and eighth pterygiophores have markedly elongate heads, and each carries a short weak spine which is largely embedded in the epaxial body musculature. Superficially, these spines are well separated from each other and from the first and second dorsal fins. The ninth pterygiophore carries a longer and somewhat stouter spine which is the first ray of the second dorsal fin. Unlike Lates, the posterior branched rays of the dorsal fin in Luciolates do have distinct medial radials, even in the largest individuals examined.

The wide gap between the dorsal fins of Luciolates was, and in published accounts of this taxon still is, the principal diagnostic feature for the genus. It is therefore of some importance to reconsider the relative positions of the dorsal fins (or of its two sections where the fin is apparently a single unit, as in L. niloticus}. As Poll has described (Poll, 1953) and I have been able to confirm, the Lake Tanganyika species of that genus show ontogenetically correlated changes in relative fin position. However, my observations also indicate that the definitive fin positions in these species are reached well before the cessation of obvious growth in body length.

Lates calcarifer (as compared with L. niloticus} has a distinctly greater interval between the last and first spines of the two fins than that existing between the penultimate and last spines of the first fin ; the gap is bridged by a low membrane. In L. niloticus the spacing between these three rays is almost equal, and the inter- connecting membrane appears to be slightly deeper. The condition in L. macroph- thalmus and L. longispinis approaches that in L. niloticus but with a slightly greater distance between the spines in L. macrophthalmus.

The condition in L. angustifrons (the seemingly most generalized of the Tanganyika species) is either comparable with that in L. calcarifer or, in some individuals, the inter-fin spacing may be a little greater. Some specimens I have examined (up to 345 mm S.L.) have no membrane connecting the two fins and in a few the 'last' spine of the first dorsal is not connected with the rest of the fin ; it is impossible to

46

P. H. GREENWOOD

tell whether this latter condition is the result of damage. The smallest fish measured (90 mm S.L.) has a distinct but membrane-spanned gap between the fins.

None of the L. microlepis examined has a membrane connecting the fins, and in several there is an isolated spine in the gap.

The usual condition in L. mariae (except in fishes < 70 mm S.L.) is a distinct gap between the fins, with a single, isolated spine lying at about its midpoint. This species is unusual in having a modal dorsal spine count of nine (eight is the mode in other Tanganyika species, although occasional individuals with nine spines are recorded ; see Poll, 1953).

Within the Lates species of Lake Tanganyika then, one finds a complete inter- gradation between a continuous, albeit deeply notched dorsal fin, and two separate fins with an isolated spine interposed. The condition in Luciolates differs from the latter state only in the greater width of the gap and the occurrence of two spines within it. Luciolates is, however, unusual in having only six spines in the first fin

2ndASp

2ndASo

FIG. 27. First anal pterygiophore, and abdominal -caudal vertebral transition, in : (a) Lates niloticus, (b) Centropomus ensiferus (drawn from radiograph 1903.5.15:3-5 and dry skeleton 1861.12.12:13).

REVIEW OF CENTROPOMIDAE 47

and one spine with only 9 or 10 branched rays in the second (compared with the usual seven spines and one spine plus 11-13, rarely 10, branched rays in the fins respectively). The two isolated spines in Luciolates may therefore represent the detached 'ultimate' and 'first' rays respectively of the ancestral type fin, with what we now consider to be the first spine of the second dorsal fin a neomorphic develop- ment from a branched ray. Alternatively, and as would seem more probable, the ancestral species could have had seven spines in the first dorsal fin, an isolated spine between it and the second dorsal, and the latter comprising one spine and 10 branched rays (a condition found in some specimens of L. mariae).

The anal fin skeleton is similar in all Lates species. There are nine, rarely eight, pterygiophores, the first a large double structure carrying two spines (Fig. 2ya) ; it articulates with the cross-bar on the haemal arch of the first abdominal vertebra. All other anal pterygiophores, except the last, carry a single ray (that on the second a spinous one) ; the last pterygiophore carries two rays. Medial radials are absent except on the last three or four pterygiophores.

Morphologically the anal fin skeleton of Luciolates is like that in Lates, although the first pterygiophore is less robust and there are nine others in the series (i.e. a total of 10). A medial radial is present in the last four pterygiophores.

Swimbladder

One outstanding feature of the swimbladder in Lates and Luciolates is the presence of a tough connective tissue strap running from a point anterodorsally on the tunica externa to the posttemporal, which has a well-defined ventrolateral recess for the reception and anchorage of the strap (see above, p. 41). Katayama (1956) does not describe this connection in L. calcarifer but I have been able to confirm its presence in that species.

Apart from Psammoperca (see p. 60 below), I know of no other percoid species in which a similar swimbladder -posttemporal connection has been described, nor indeed of any connection between those two points. The functional significance of a swimbladder -posttemporal linkage is not readily apparent.

The anterior end of the swimbladder in all Lates species and in Luciolates stappersi has a deep median indentation which gives that end of the swimbladder a distinctly bilobed appearance.

Baudelot's ligament

This ligament is well defined in Lates and Luciolates, and originates from a deep pit on the basioccipital. In L. niloticus and Luciolates stappersi (the two species dissected) little or no epaxial body musculature runs below the ligament medially ; laterally, however, there is a broad muscle band passing below and above it to insert partly on the anterolateral aspect of the basioccipital but mainly on the exoccipital. Thus at least the distal half of Baudelot's ligament is embedded in muscle.

The relationship of the ligament to the epaxial musculature seems to combine certain features of both the percichthyid and serranid types described by Gosline (1966), but is more akin to the serranid type.

48 P. H. GREENWOOD

Lateral line (Fig. 28)

In all Lates species the pored lateral line scales of the body continue onto the caudal fin where they extend, or almost extend, to the posterior margin of the fin. Since the posterior margin is generally abraded or damaged it is difficult to tell in the latter cases whether the absence of scales from the immediate marginal area is artef actual or not. Two other rows of pore-bearing scales are present on this fin, one lying above and the other below the median row (from which they are separated by a space usually equal to that between two fin rays). These upper and lower scale rows generally do not quite extend to the posterior fin margin. Because of their small size the scales in these rows are difficult to see in fresh and spirit-preserved specimens unless the fin is allowed to dry out completely.

Superficially there does not seem to be any linkage between the median and the other caudal lateral line scale rows ; scales in the latter rows cease to be pored at the base of the fin. Dissection of adult specimens does not reveal any deeper-lying connecting channels.

The presence of a triple lateral line on the caudal fin in Lates has not been recorded before, and to the best of my knowledge has not been described in any other percoid species. Since it is clearly a derived condition it is a useful indicator of the mono- phyletic origin of these species.

The posterior extremity of the lateral line in Luciolates is also triradiate, but here the three branches are interconnected by pore-bearing scales (Fig. 28b). The median row extends onto the caudal fin, but the line of pored scales is interrupted by the presence of poreless ones, and it never extends to the margin of the fin. The upper and lower lines do not extend for more than one or two scales beyond the limits of the body scales. However, in a few specimens an occasional pored scale is found some distance onto the fin membrane and in the same line as a basal branch.

Although the condition of the caudal lateral line in Luciolates does differ from that in Lates it is still a triradiate one and the two taxa can reasonably be thought to

FIG. 28. Caudal fin, showing : (a) lateral line pore scales in three rows (drawn from Lates niloticus, but typical for all species except L. stappersi), (b) L. stappersi showing 'trident' arrangement of lateral line pore scales at body -caudal fin junction ; note that pores do not continue onto membrane of fin.

REVIEW OF CENTROPOMIDAE 49

share a derived character. It is difficult to tell from the evidence available whether the Luciolates condition should be considered a further derivative - albeit a reduc- tional one - of the Lates type, or whether it represents an early stage in the evolution of the Lates type.

THE INTERRELATIONSHIPS OF SPECIES WITHIN THE GENUS LATES, AND THE TAXONOMIC STATUS OF LUCIOLATES BLGR.

An analysis of the osteological and other anatomical features described in the previous sections shows that all seven Lates species share two derived characters, viz. (i) three rows of pored lateral line scales on the caudal fin and (ii) the ventral (i.e. horizontal) arm of the preoperculum has three or more large serrae.

There are three other derived characters (the swimbladder-posttemporal ligament, the anterior extension of the supraoccipital, and the presence of two epurals in the caudal fin skeleton), but as these are shared with Psammoperca (see below, p. 61) they are of no value in establishing the monophyletic origin of the genus Lates on the basis of synapomorph characters occurring within its constituent species. Since, however, the first two apomorph features noted above do not occur in any other members of the Centropomidae except Lates species, they argue strongly for the monophyly of the genus.

It is possible to subdivide the genus Lates by grouping together three species sharing one clear-cut apomorphy and at least four apomorph trends. Such a subdivision would bring together L. angustifrons , L, mariae and L. microlepis, species with an elongate ethmovomerine region in which the posterior face of the lateral ethmoid slopes backwards at a pronounced angle and the dorsolateral aspects of that bone slope sharply downwards ; this characteristic appearance of the snout region is clearly seen in Figs 5, 6 & 7. The apomorph trends shared by these species are an elongation of the caudal and posterior abdominal vertebrae (most marked in L. mariae and L. microlepis ; see p. 43), a division of the dorsal fin into two separate parts (reduced interconnecting membrane in L. microlepis, actual separation of the fins in L. mariae ; see p. 45), reduction of the pterosphenoid pedicle and internal jugular bridge (slight reduction in L. angustifrons, progressively greater reduction in L. microlepis and L. mariae ; see pp. 20-27) and, lastly, an elongation and narrowing of the entire skull (a trend not necessarily correlated with the former which is also manifest in species with broad skulls, e.g. L. macrophthalmus ; see pp. 17-19). Finally, and no doubt of significance, it may be noted that the three species are all endemic to Lake Tanganyika.

In view of these characteristics, especially the changes in lateral ethmoid mor- phology, it would seem phyletically proper to recognize the species as more closely related to one another than to any other Lates species still extant. This topic will be taken up again later (p. 51).

It is difficult to establish any well-founded scheme of interrelationships for the remaining species, L. calcarifer, L. niloticus, L. macrophthalmus and L. longispinis. Part of this difficulty stems from the problematical relationships of L. longispinis and L. macrophthalmus, as was discussed above, p. 13. These two species alone in the

50 P. H. GREENWOOD

group show and share definite apomorph characters* (enlarged eyes and long dorsal fin spines ; for a discussion of the reduced pterosphenoid pedicle see p. 25). All four species otherwise exhibit only the synapomorph features of the genus, and are distinguished from each other by slight meristic and morphometric differences.

The taxonomic status of Luciolates has never been reviewed critically since Boulenger (1914) first differentiated the genus from Lates on the grounds of its having ' . . . corps plus allonge, nageoires dorsales largement separees 1'une de 1'autre, et ventrales inserees en arriere de la base des pectorals'.

It will be recalled (p. 47) that the condition of the dorsal fin in Luciolates represents a slight exaggeration of that existing in L. mariae. In turn, the L. mariae fin condition is a development of that in L. microlepis which is a further slight deviation from the condition found in the basic L. calcarifer-L. niloticus type. In other words, the apparently characteristic dorsal fin of Luciolates is in fact linked by intermediates with that of the most generalized Lates species.

Amongst the various Lates species similar intermediate character states can be found for most of the features which, at first sight, might seem to distinguish Lucio- lates from a generalized Lates species. As examples of these 'distinguishing' features one can cite the relative elongation of the vertebral centra, the protraction of the snout (especially the ethmovomerine skull region) and the general elongation and narrowing of the neurocranium. But, all are features shared with the Lates species of Lake Tanganyika, especially the peculiarly shaped ethmoid (cf. Figs 8, with 5-7). Even the supposedly distinctive position of the pelvic fins in Luciolates is closely approached by L. mariae.

There are, of course, certain characters in which Luciolates does differ trenchantly from all Lates species, and these features must be given particular attention.

No Lates species has enlarged caniniform teeth such as occur, in small numbers, near the symphysis of the upper jaw in Luciolates (see p. 35), none shows such a high degree of hypural fusion (see p. 44), and Luciolates is unique in having the three caudal extensions of the lateral line restricted to the proximal part of the fin and visibly interconnected with each other.

One may, I think, rate the dentition and fused hypural plates of Luciolates as derived characters. The condition of the lateral line may be primitive or it could be a secondary reduction of the Lates type (i.e. a derived character), although the interconnection of the lines might argue against such a conclusion. But, even if all these character states are derived ones, they are autapomorphies ; on the basis of synapomorphies Luciolates still has as its nearest relatives the three Lates species of Lake Tanganyika. Furthermore, Luciolates shares with these species one apomorph character (the morphology of the lateral ethmoid) which distinguishes the group as a whole from all other African species of Lates, as well as from the Indo-Pacific marine species L. calcarifer.

For these reasons I propose that Luciolates should be united with its nearest rela- tives in the genus Lates. At the same time I propose placing the Lake Tanganyika

* On the evidence currently available, L. macrophtkalmus (from Lake Albert) and L. longispinis (from Lake Rudolf) could either be sister taxa derived from a common ancestor (itself a sister species of L. niloticus) or each could have been derived locally, in late Pleistocene times, from the population of L. niloticus then inhabiting these lake basins.

REVIEW OF CENTROPOMIDAE 51

Lates species, that is L. angustifrons , L. microlepis, L. mariae and now Lates stappersi in one subgenus (for which the name Luciolates Blgr. is available), separate from L. calcarifer, L. niloticus, L. macrophthalmus and L. longispinis which species comprise the subgenus Lates. Definitions and synonymies for these taxa are given on pp. 77-78.

Interrelationships within the subgenus Luciolates may be delimited on the basis of vertebral morphometry, the division of the dorsal fin, the morphology of the lateral ethmoid, and on neurocranial anatomy and morphology (see relevant sections on pp. 14-45).

Lates angustifrons is clearly the plesiomorph sister species of all others in the subgenus. Lates mariae and L. microlepis show generally similar degrees of speciali- zation in all the characters noted above, and can thus be considered sister species ; since in some features (e.g. the lateral ethmoid) L. microlepis is less specialized than L. mariae it can be considered the plesiomorph member of the pair.

The greatest level of specialization is seen in Lates stappersi which is therefore ranked as the apomorph sister species of L. mariae and L. microlepis combined (see Fig. 37).

The difficulties of ranking species within the nominate subgenus have been discussed above (see pp. 49-50). Indeed, it is not even possible to show that this subgenus is monophyletic since its 'diagnostic' features are those plesiomorphic for the genus as a whole.

A REVIEW OF THE GENUS PSAMMOPERCA RICHARDSON

Introduction

There are no published accounts of the osteology and anatomy of Psammoperca. A brief outline of the osteology of P. waigiensis (Cuv.) is given here, together with some notes on various aspects of the soft anatomy in this species, particularly those features which have some bearing on the phyletic relationship of the taxon.

Fishes of the genus Psammoperca (Richardson, 1844) occur in coastal waters from the Bay of Bengal, the Indo-Australian archipelago, northern Australia, the Philip- pines and the China Sea. To a considerable extent, this distribution overlaps that of Lates calcarifer (see above, p. 12 ; also Fig. 36, and Weber & de Beaufort, 1929). Two nominal species, P. waigiensis (Cuv.) and P. macroptera Gunth. are recognized, the latter restricted to Australia and known only from the holotype. The material I have examined is entirely of P. waigiensis, but the individual variability rep- resented in these samples certainly indicates that P. macroptera should be con- sidered a synonym. The question could be solved if large samples from the type locality and other regions of Australia were examined.

Superficially, P. waigiensis is much like L. calcarifer (Fig. i), but is distinguished by its widely separated nostrils, smooth lower border to the preoperculum and to the lachrymal, and by the more extensive squamation of the dorsal and anal fins.

52 P. H. GREENWOOD

Osteology and anatomy of Psammoperca waigiensis Neurocranium (Figs 2ga-b)

The proportions and general appearance of the neurocranium closely resemble those of Lates macrophthalmus (cf. Figs 29 & 4b) ; that is to say, a member of the subgenus Lates in which there is a reduction in the length of the precommissural neurocranium without elongation of the ethmoid skull region.

The ethmovomerine region is exactly like that in members of the subgenus Lates ; the posterior wall of the lateral ethmoid is slightly concave and rises steeply to meet

LATE

FR

SPO

BOC

10mm

PAR

PTF

EPI

SOCS

SPO

PTO

FIG. 29. Psammoperca waigiensis. Neurocranium in : (a) left lateral view,

(b) dorsal view.

REVIEW OF CENTROPOMIDAE 53

the frontal, and there are two palatine and one lachrymal articulatory facets on this face. A slight intergeneric difference lies in the strongly concave posterior margin of the vomer, which gives the tooth patch in Psammoperca a distinct arrowhead outline in ventral view.

The precommissural region (see p. 20) of the braincase in Psammoperca differs in certain details from that in L. macrophthalmus. The pterosphenoid is about the same relative size and the ascending limb of the parasphenoid meets the prootic to create a generally similar appearance for this region of the skull. However, in Psammoperca there is no trace of a pterosphenoid pedicle and there is no bridge, not even a ligamentous one, across the internal jugular vein and its associated nerves (see pp. 20-26 above). In this respect the skull of Psammoperca resembles, most closely, that of Lates (Luciolates) stappersi.

The otic skull region in Psammoperca is like that in Lates (Lates} macrophthalmus as far as the relative sizes and relationships of the constituent bones are concerned, but the posterior half of the prootic is noticeably inflated and is thinner in Psammo- perca.

The posttemporal fossa, like that in all extant Lates species, is large and deep, and does not have a complete bony floor. The exoccipital facets meet medially.

The autosphenotic does not extend far into the orbit medially or dorsolaterally ; again the resemblance is more to L. macrophthalmus than to other members of the subgenus Lates, and there is some resemblance to species of the subgenus Luciolates.

The dorsocranium is, in all respects save one, like that in L. macrophthalmus, with the supraoccipital extending forwards to separate the frontals, high frontoparietal crests, deep excavations between these crests and the supraoccipital, and a clearly demarcated lateral shelf on the supraoccipital where the crest extends posteriorly beyond the epioccipitals. The one difference I can detect is the absence of a bone- enclosed supraorbital transverse commissure in Psammoperca. The cephalic lateral line system in other respects, however, is like that in Lates.

The parasphenoid resembles closely that in Lates but is more sharply angled upwards from the level of the ascending limb ; in this respect Psammoperca resembles species of the subgenus Luciolates.

Hyopalatine arch and the preoperculum (Fig. 30)

Again, it is only in certain details that the hyopalatine arch of Psammoperca differs from that arch in Lates.

Psammoperca has no tooth patch on the ectopterygoid, and the dermopalatine tooth patch is very narrow. According to Weber & de Beaufort (1929), ectoptery- goid teeth are present in P. waigiensis but I have been unable to detect any on the specimens I have examined. Ectopterygoid teeth are absent in some specimens of Lates (Luciolates} stappersi, and it is interesting to recall that the dermopalatine tooth patch is narrowed to an extent comparable with that in Psammoperca. The palatine in P. waigiensis has a distinct dorsal ridge on the autopalatine immediately in front of the facet for articulation with the lateral ethmoid. This ridge is absent in all members of the subgenus Luciolates and is but weakly developed in species of the nominate subgenus.

54

P. H. GREENWOOD

ENT

.PAL

POP.

FIG. 30.

Psammoperca waigiensis. Hyopalatine arch, including preoperculum (right side), viewed laterally.

As in Lates, the vertical limb of the preoperculum has a finely serrated posterior border, and the mandibular-preopercular sensory canal is bone enclosed (but opening through three ventrally directed and elongate pores on the horizontal arm) . The margin of the ventral limb, however, is entirely smooth except for a stout, posteriorly directed spine at the angle between the vertical and horizontal limbs (Fig. 30) . A similar spine is, of course, present in all Lates species but Psammoperca lacks the three or four stout and ventrally orientated spines on the horizontal limb. Lates (Luciolates} stappersi, it will be recalled (p. 30), often shows some reduction in the size of these spines, but in no individual are they entirely wanting.

Circumorbital bones (Fig. 31)

The five circumorbital bones are very similar to those in Lates ; the relative elongation of the lachrymal and of the fifth circumorbital in Psammoperca is more like that seen in members of the subgenus Luciolates.

The lateral line canal is bone enclosed but opens to the exterior through five pores on the lachrymal, a pore between each articulation of the individual bones, and a ventral pore on the third circumorbital bone.

The suborbital shelf (on the third bone) is well developed to an extent almost equalling that found in members of the subgenus Lates ; it extends dorsally to about the upper end of the fourth circumorbital bone.

The most marked difference between the circumorbital series in Psammoperca and Lates lies in the completely smooth ventral margin to the lachrymal and second circumorbital bones. These bones are strongly serrated in all Lates species, except L. stappersi, but even in that species some definite trace of the serrations does remain on the posterior part of the lachrymal (see p. 32 and Fig. i3c-d).

REVIEW OF CENTROPOMIDAE

55

Opercular series (Fig. 140)

The operculum of Psammoperca, like that in Lates, is armed with a single stout spine developed at the posterior end of the stay running from the articular facet for the hyomandibular boss. Indeed, the entire opercular series is like that of Lates, the relatively elongate interoperculum having the proportions of that bone in L. (Lucio- lates) mariae, L. (Luciolates) microlepis and L. (Luciolates} stappersi rather than that in other species of the genus. As in Lates, the sub- and interopercular bones of Psammoperca are thin.

Jaws (Figs 32a-c)

The maxilla, supramaxilla and premaxilla are, except for the coarser teeth on the latter bone, identical with those elements in species of the subgenus Lates.

The bones of the lower jaw (dentary, anguloarticular and retroarticular) are also like their counterparts in members of that subgenus ; again, the teeth are stouter than in Lates.

Branchial skeleton

In its basic morphology and in the details of its upper pharyngeal dentition the gill arch skeleton of Psammoperca is identical with that of Lates niloticus (see p. 35). The only difference I can detect from the one Psammoperca skeleton studied is that the regularly arranged, small, rectangular tooth plates lying laterally on the gill arch above the filaments (the supralamellar plates, see p. 37) are restricted to the outer side of the first four gill arches (in Lates plates are present on both aspects of an arch).

This reduction in tooth plates should be considered as a derived condition since a marked reduction or even the complete loss of free dermal tooth plates is a feature of the more specialized percomorph groups.

SOS

LAC

FIG. 31. Psammoperca waigiensis. Circumorbital bones (right) in : (a) lateral view, (b) viewed dorsally and somewhat anteriorly.

P. H. GREENWOOD PMAXP

ASCP

SMX

DPROC PMX P

AA

10mm

FIG. 32. Psammoperca waigiensis. (a) Premaxilla (right) in lateral view, (b) Maxilla (right) in a slightly oblique dorsal view to show supramaxilla. (c) Dentary (right), with anguloarticular and retroarticular, in lateral view.

Hyoid arch skeleton (Fig. 33)

The only marked difference between the hyoid skeletons of Psammoperca and Lates (especially members of the subgenus Lates} is the presence in the former of a moderately large, ovoid tooth-patch firmly attached to the broadly spatulate basihyal.

Psammoperca has seven branchiostegals, the posterior two of which articulate laterally with the epihyal, the next two with the ventrolateral face of the ceratohyal, and the first three with the ventral margin of that bone.

The presence of a basihyal tooth plate must be considered a plesiomorph character for the genus, the only living member of the Centropomidae in which it has persisted.

Pectoral girdle and associated bones (Fig. 34)

The one obvious difference between the pectoral girdles (i.e. supracleithrum, cleithrum, scapula and coracoid) of Psammoperca and Lates is the absence of serra- tions on the posterolateral angle of the cleithrum. In all other respects the girdles

REVIEW OF CENTROPOMIDAE

57

2

mm

FIG. 33. Psammoperca waigiensis. Dorsal view of basihyal, showing tooth plate.

in the two genera are similar, but with a greater resemblance in overall proportions between the girdle of Psammoperca and that in the subgenus Lates.

Although in Psammoperca there are no serrations at the posterior cleithral angle, the bone in that region is drawn out into a short but well-demarcated spine.

As in Lates, the three upper fin radials articulate with the scapula, and the lowest with the coracoid.

There are two postcleithra, but in Psammoperca the upper member of the pair is less expansive than in Lates.

The posttemporal and extrascapula are similar in both genera, the posttemporal in Psammoperca even having the same kind of pit for the reception of the swimbladder ligament (see p. 41), but it does lack serrations on its hind margin.

SC

PFr

COR

10mm

FIG. 34. Psammoperca waigiensis. Pectoral girdle (right half) with supracleithrum in situ and post-cleithra displaced posterodorsally. Lateral view.

58 P. H. GREENWOOD

Vertebral column

The total vertebral count in P. waigiensis is 25, comprising n abdominal verte- brae, 13 caudal, and the fused first preural and ural elements of the caudal fin skeleton.

There are nine pairs of pleural ribs, the first pair carried on the third vertebra, the last pair on the eleventh abdominal vertebra ; this latter pair of ribs, instead of sloping gently backwards parallel with the preceding pair (as in Lates), runs almost horizontally and generally overlaps the proximal tip of the first anal pterygiophore.

The first definite parapophyses appear on the seventh vertebra, and are but a little shorter than those on the succeeding centra, although there is a slight and posteriorly progressive elongation of these processes. Anterior to the seventh vertebra, the ribs articulate with a shallow pit on the centrum. Where parapo- physes are present, the rib articulates with the posterior face of the process.

In all these features, except for the better developed first and second parapophyses, and the angling of the last pair of ribs, Psammoperca is like Lates (see p. 42).

Epineural ribs are present on the first three vertebrae, and epipleural ribs on at least the first four pleural ribs. (These data were obtained from radiographs.)

The first three vertebrae are shown in Fig. 25b ; their close resemblance to those in Lates is obvious (cf. Fig. 25a). One slight difference is in the development of a low median ridge on the ventral face of the second centrum of Psammoperca.

As in Lates, the neural spine of the second vertebra is much broader than the spine of the first and third centra, has its anterior and posterior margins parallel over much of their lengths, and tapers rather abruptly to form a slightly hooked tip. The angle between the posterior face of the second spine and the anterior face of the third spine is from 20° to 25 °.

Except in the first four vertebrae, all centra are a little longer than deep, the relative length of the centrum increasing somewhat in the posterior abdominal vertebrae, which have about the same proportions as the caudal vertebrae. In this respect the centra in Psammoperca are rather more like those in Lates (Luciolates) angustifrons than in other species of that subgenus or in species of the nominate subgenus.

There are three predorsal bones, the first lying immediately anterior to the first neural spine, the second and third situated immediately before and behind the tip of the second neural spine.

Caudal fin skeleton (Fig. 35)

The caudal skeleton in Psammoperca differs from that in Lates in one important respect, namely the presence of a single uroneural (see p. 44). Otherwise there is great intergeneric similarity in this structure (viz. 2 epurals, 5 hypurals, 1,8 + 7,1 principal fin rays and a low neural crest on the second preural vertebra [lower, in fact, than in Lates]).

There is, as far as can be detected from radiographs, probably no fusion between any of the hypurals, although in one fish (240 mm S.L.) of the eight examined, hypurals 3 and 4 are so closely apposed as to appear fused. The first and fifth hypurals are autogenous, the others are fused to the underlying vertebral support.

REVIEW OF CENTROPOMIDAE

59

UR

NaPU2

PH

HsPU2

5mm

FIG. 35. Psammoperca waigiensis. Caudal fin skeleton (drawn from specimen 1872.9.2:10-11).

The posterior margin of the caudal fin is rounded.

A single uroneural must be considered a derived feature, and in this respect the caudal fin skeleton in Psammoperca is, relative to that in Lates, more specialized. Indeed, since there are in Psammoperca two and not three epurals the caudal fin skeleton is more specialized than that in any member of the Serranidae (where there are, invariably, three epurals - but one uroneural - in the fin skeleton ; see Gosline, 1966).

Dorsal, anal and pelvic fins

There are 19 pterygiophores (proximal radials) in the dorsal fin skeleton, each one except the last supporting a single fin ray. Although an occasional member of the subgenus Lates may have 19 pterygiophores (see p. 45), the usual number in that taxon is 18. Nineteen, however, is the modal number of pterygiophores (20 the unusual one) in species of the subgenus Luciolates.

Unlike Lates, Psammoperca has some of the dorsal fin spines (the fifth through the eighth) associated with discrete medial radials ; a medial radial is also associated with the last branched ray in this fin. Lates (Luciolates} stappersi alone amongst the Lates species has medial radials (associated with the posterior four or five branched rays).

60 P. H. GREENWOOD

The dorsal fin is deeply indented to form an anterior part with seven spines, and a posterior portion with one spine and 12 branched rays. A continuous but low membrane connects the two parts of the fin. The spacing between the spines of the two fin divisions is fairly even (cf. Lates species, p. 45).

The anal fin skeleton comprises nine pterygiophores, of which the first is a double structure and carries two spines. All other pterygiophores, except the last, carry a single spine or ray. (Total fin ray count 111,8.) Medial radials are absent from all pterygiophores except the last. (In Lates species medial radials are present on the last three or four pterygiophores, see p. 47.)

The first of the nine anal pterygiophores, like that in Lates, is a stout and elongate bone and is in contact proximally with the haemal spine of the first abdominal vertebra.

The origin of the pelvic fins lies slightly behind that of the pectoral fins ; in other words the fins have the same positions as in Lates (Luciolates] stappersi.

Swimbladder

As in Lates, so in Psammoperca there is on each side of the swimbladder anteriorly a tough connective tissue strap extending from the tunica externa to the posttemporal. The position, shape and size of the strap are identical in both genera (as are the modifications to the posttemporal, see pp. 41 and 47).

The gross morphology of the swimbladder resembles that in Lates. The tunica externa is thick, and a pair of short blunt processes extends forward on either side of a median invagination of the swimbladder. Psammoperca does differ, however, in having a narrow posterior diverticulum extending outside the visceral cavity. In a single specimen dissected, this caudal swimbladder prolongation lay on the left side of the first anal pterygiophore ; it is embedded in the body musculature of that region and does not penetrate into the haemal arches of any caudal vertebrae.

Baudelot's ligament

The ligament is well developed and its relationships with the body musculature in the cervical region are like those described for Lates niloticus on p. 47 ; that is, it closely approximates to the serranid type described by Gosline (1966).

Lateral line

Unlike Lates there is only one series of pore-bearing scales on the caudal fin of Psammoperca. These small scales are an uninterrupted continuation of the body lateral line scales ; they extend almost to the posterior margin of the caudal fin. In one of the nine specimens available, a few widely separated pore scales were found on the fin membrane between a pair of rays on the lower part of the fin.

THE RELATIONSHIPS OF PSAMMOPERCA

The close overall resemblance between Psammoperca and Lates has long been recognized (Regan, 1913), and has even resulted in a false record of Psammoperca for the Japanese fauna (see Katayama, 1956).

REVIEW OF CENTROPOMIDAE 61

.A detailed examination of the characters shared by the two taxa shows that many must be ranked as primitive features (i.e. symplesiomorphies) and therefore of little value in estimating relationships. Included amongst the symplesiomorphies are the vertebral count, the presence of a single opercular spine, similarities in gill arch anatomy and dentition, the single spine carried on the first pterygiophore of the dorsal fin, and many details in syncranial morphology and anatomy.

There are, however, four derived characters shared by Psammoperca and Lates which are not present in any other members of the Centropomidae. These syna- pomorphies are :

(i) A swimbladder-posttemporal ligament (and correlated modifications to the

posttemporal bone) ; see p. 60. (ii) A large spine at the posterior angle of the preoperculum.

(iii) Two epurals in the caudal fin skeleton ; see p. 58.

(iv) An anteriorly extended supraoccipital which separates the posterior parts of the f rentals.

On the basis of these characters it is reasonable to conclude that Lates and Psammo- perca are members of the same lineage, a lineage distinct from that of Centropomus (see below, p. 62). To indicate this relationship I propose placing Psammoperca and Lates together in one subfamily, the Latinae. Members of the genus Centro- pomus would thus constitute a second subfamily, the Centropominae, which can be readily defined on the basis of several specialized characters (see p. 67 below).

Psammoperca waigiensis (and, where these features can be checked, also P. macroptera ; see p. 51 above) differs from all or most species of the genus Lates in at least 14 features. In the list that follows, the condition of these features in Lates is given in parentheses.

1 . A single series of lateral line scales on the caudal fin. (Three series.)

2. A single uroneural. (Two uroneurals.)

3. Some spine-bearing dorsal fin pterygiophores with a median radial. (None.)

4. No tooth patch on the ectopterygoid. (Present, but reduced in L. stappersi.)

5. No spines on the ventral (horizontal) limb of the preoperculum. (Three or four spines.)

6. Ventral margin of the first infraorbital bone (lachrymal) smooth. (Serrated, strongly so in most species.)

7. No pterosphenoid pedicle or internal jugular bridge. (Pterosphenoid pedicle present in all species except L. stappersi and L. mariae ; internal jugular bridge present in all species, even if reduced to a ligament.)

8. Transverse commissure of supraorbital lateral line system absent or poorly developed. (Present and well developed.)

9. Dermal tooth patch fused with basihyal. (Absent.)

10. Supralamellar tooth plates (p. 37) present only on the outer side of each gill arch. (Pre- sent on both sides.)

11. Posterior margin of the posttemporal smooth. (Serrated.)

12. A single short spine at the posterior angle of the cleithrum. (One large and two smaller spines.)

13. Posterior extravisceral extension of the swimbladder. (None.)

14. Second dorsal and anal fin entirely covered by small but densely arranged scales. (Squamation restricted to about the proximal two-thirds of the fin.)

In some of these characters (e.g. i, 3, 5 and 9) Psammoperca is more primitive than any Lates species ; in others (2, 4, 7, 10 and 13) it shows derived characters. The

62 P. H. GREENWOOD

status of characters 6, n, 12 and 14 is at present indeterminable. (See discussion on pp. 30-32.)

It is on the basis of the unique derived characters (i.e. autapomorphies) found in each of the two taxa that I would maintain them as distinct genera, the implication being that Psammoperca split off from the common latine lineage before the evolution of a serrated preoperculum and the tripartite lateral line extension onto the caudal fin. The derived characters seen in Psammoperca (especially the loss of a ptero- sphenoid pedicle, the presence of a single uroneural, the loss of certain branchial arch tooth plates, and the loss of ectopterygoid teeth) must have evolved after this split occurred. In these features Psammoperca is certainly more 'advanced' than is Lates.

It is interesting to note that a reduction and ultimate loss of the pterosphenoid pedicle is seen in certain Lates species of the subgenus Luciolates (see pp. 20-27), and that Lates (Luciolates} stappersi also shows a considerable reduction in, and oc- casionally the loss of, ectopterygoid teeth. Furthermore, this species also shows a marked weakening of the serrations on the lachrymal. Similar parallel trends in all three characters are found amongst the species of Centropomus (see below) thus suggesting that this is the manifestation of a potentiality possessed by the common ancestor of all living centropomids.

THE RELATIONSHIPS OF CENTROPOMUS WITH THE LATINAE

Fraser (1968) has given a good account of the osteology of five species of Centro- pomus but he was unable, through lack of published information, to compare fully these species with members of the genera Lates and Psammoperca. He did, however, list a number of differences between Lates and Centropomus and these will be com- mented upon below.

In my comparison of Centropomus and the Latinae I have drawn on Fraser's (op. cit.) information and supplemented it from dissection, radiographs and dry skeletons of C. parallelus Poey, C. pectinatus Poey, C. armatus* Gill, C. unionensis* Bocourt, C. robalito* Jordan & Gilbert, C. nigrescens* Giinther, C. ensiferus Poey and C. undecimalis (Bloch) ; species not described by Fraser (op. cit.) are marked with an asterisk.

The neurocranium in all Centropomus species is narrow and elongate, with a pronounced relative lengthening of the ethmovomerine region. In these respects it resembles the neurocranium of Lates (Luciolates} mariae and L. (Luciolates} stappersi, but it does differ in having only a gently angled parasphenoid (or even a straight one ; cf. C. pectinatus, text-fig. 4, and C. undecimalis, text-fig. 5, in Fraser (1968), with Figs 7 & 8 above), and in having the postotic region of the skull relatively longer.

Within the Centropomus species I have examined, there is a trend of neurocranial elongation which closely parallels that found in members of the Lates subgenus Luciolates.

Another parallelism with Lates is seen in the reduction of the pterosphenoid pedicle and internal jugular bridge.

The pedicle and bridge are best developed in C. ensiferus (see text-fig. 6 in Fraser, 1968) where the condition of the bridge is like that in L. (Luciolates} angustifrons

REVIEW OF CENTROPOMIDAE 63

(see p. 24 and Fig. 5a). A noticeable difference, however, is that, in C. ensiferus, there is no ascending arm of the parasphenoid and the bridge is formed by contact between the pterosphenoid and prootic. (The prootic in all Centropomus contributes to the posteroventral margin of the orbit ; in Lates this rarely happens because part of the ascending parasphenoid arm usually rises in front of the prootic. This tongue of parasphenoid is, however, very narrow in the more specialized species of the subgenus Luciolates.)

In all other Centropomus species I have examined or which are figured by Fraser (1968), excepting C. undecimalis, the pterosphenoid pedicle is either reduced (e.g. C. pectinatus) or is greatly reduced to a small bony knob (that is, to conditions com- parable with those in very small L. (Lates) niloticus or those in adult L. (Luciolates} microlepis ; see p. 26). In C. undecimalis there is no trace of a pterosphenoid pedicle ; in other words a situation directly comparable with that in L. (Luciolates) stappersi and in Psammoperca waigiensis.

Correlated with this reduction in the pedicle, the internal jugular bridge is reduced from a narrow bony strut in C. ensiferus to a ligament in the other species (except C. undecimalis), again paralleling exactly the trend seen in Lates (pp. 21-26). In C. undecimalis even the ligament has disappeared (at least in the specimen of 175 mm S.L. I dissected) ; this, it will be recalled, is the condition also found in Psammoperca (p. 53).

Probably as a correlate of the lengthening ethmoid-vomerine skull region, the shaft of the vomer in all Centropomus species is much broader anteriorly and has a closer sutural union with the lateral ethmoid than it does in any latine species. In other details, however, this region of the skull is generally similar in both Centro- pomus and the Latinae.

The otic region in Centropomus is bullate, more markedly so in some species than in others, but always more inflated than in any Lates species and rather more so than in Psammoperca.

An outstanding inter-subfamilial difference is found in the lateral line system of the dorsicranium. In Lates and in Psammoperca all three major canals are bone enclosed. In Centropomus the canals are in the form of laterally orientated open gutters, with only the posterior part of the supraorbital line completely tubular. The frontal cross-commissure is also open (with the gutter directed medially), as is the entire length of the frontoparietal branch (whose gutter is directed laterally).

Fraser (1968) has corrected Regan's (1913) erroneous observation that parietal crests are absent in Centropomus, but as compared with Psammoperca and Lates, the parietal crests, and their counterparts on the frontals, are low and very poorly defined, and do not extend to the posterior margin of each parietal (often being confined to the anterior half of that bone) .

The supraoccipital in Centropomus does not extend so far anteriorly as it does in Lates and Psammoperca, its tip barely separating the frontals and only reaching a level with a vertical through the middle or the posterior third of the prootic.

There are few noteworthy differences in the hyopalatine arches of Centropomus and the Latinae. As Fraser (1968) noted (pace Regan, 1913), ectopterygoid teeth are present in Centropomus. From Fraser's drawing (op. cit., text-fig, n) one gains

64 P. H. GREENWOOD

the impression that a metapterygoidal lamina is present in at least some Centropomus species, but I cannot confirm this from the dry skeletons I have examined.

There are several inter-subfamilial differences in the morphology of the pre- operculum. First, the lateral line canal in Centropomus is an open gutter, the opening orientated posteriorly on the vertical limb of the bone and ventrally on its horizontal limb. It is the upper rim of this gutter that has been described as a 'ridge' on the preoperculum in Centropomus species (see Fraser, 1968). In the Latinae, where the canal is bone enclosed and tubular no 'ridge' is, of course, de- tectable.

A second and pronounced difference is in the ornamentation of the preoperculum, a difference most obvious when Centropomus is compared with Lates. In Centro- pomus, as in Lates and Psammoperca, the posterior margin of the vertical limb is serrated (less regularly so in Centropomus), but the horizontal limb in that genus has a number of low, rather irregular serrations that are enlarged posteriorly at the angle of the bone. In no Centropomus species is there any indication of the three (or four) large triangular spines that characterize all Lates species ; likewise, no Centropomus has the completely smooth horizontal preopercular arm of Psammoperca. Also, unlike both Lates and Psammoperca, there is no single, stout spine at the pos- terior angle between the two preopercular arms ; instead, in Centropomus there are a variable number of spines, all of which are somewhat larger than those preceding and succeeding them on the arms of the bone, but none is as large nor as distinctive as the single spine of the latines.

The operculum in Centropomus lacks a spine at its posterodorsal angle (see p. 55) but otherwise the opercular series shows no marked departure from its counterpart in the Latinae.

The open lateral line gutters of the circumorbital bones, the reduced serration of the ventral lachrymal border, and the relatively short fourth and fifth circumorbitals are the most obvious inter-subfamilial differences noted in this region of the skull. Apparently the subocular shelf in Centropomus is like that in Psammoperca and members of the latine subgenus Lates, but I have been unable to check this point in all Centropomus species, and in particular those with narrower and more elongate heads.

Apart from some slight proportional differences, the major feature differentiating jaw elements in the two subfamilies is the much shorter ascending process of the premaxilla in Centropomus. In the Latinae the ascending process is at least one and a half times the height of the maxillary process (see p. 34) but in Centropomus the two processes are of equal height (cf. text-fig. 12 in Fraser, 1968, with Figs 15, 16 & 32 above).

The basic gill arch morphology and dentition are similar in Centropomus and the Latinae, although the tooth plates associated with the basibranchials are slightly more elongate in at least some members of the Centropominae. The supralamellar tooth plates in most Centropomus species which I have dissected (C. ensiferus, C. parallelus, C. pectinatus, C. undecimalis and C. armatus) show a unique arrangement not found in any member of the Latinae. The plates are present only on the outer aspect of the second to fourth gill arches, and are fused, in pairs, with the bases of

REVIEW OF CENTROPOMIDAE 65

the gill rakers on that aspect of the arch ; occasionally a single plate may occur between a pair of gill rakers. An exception is provided by a small (160 mm S.L.) specimen of C. parallelus in which the plates are serially arranged like those in Psammoperca.

A slightly larger fish (220 mm S.L.) shows a condition intermediate between that in the smaller specimen and that of the other species (and specimens of C. parallelus) I examined. Possibly this change in plate arrangement is a growth phenomenon.

All Centropomus species have 24 vertebrae (including the fused first ural and preural centra), comprising 9 abdominal and 15 caudal elements. (Eraser, 1968, gives a count of 10 + 14, indicating that we differ in our interpretation of which vertebra constitutes the first caudal element ; I identify it as that vertebra with which the first anal pterygiophore articulates.) All members of the Latinae, in contrast, have 25 vertebrae (n abdominal and 14 caudal elements).

There are seven or eight pairs of pleural ribs in centropomines (nine in the Latinae) , the number apparently showing some intraspecific variability. The first rib articu- lates directly with the third vertebra. Definite parapophyses are developed on the seventh abdominal centrum but a low process occurs on the sixth vertebra. Anterior to these centra the ribs articulate as in the Latinae, that is, with a pit in the centrum. Also as in the Latinae, the ribs articulate with the posterior face of the parapophysis when these are present.

As far as I can tell from radiographs, the shape and proportions of the centra in all Centropomus species are like those in Lates calcarifer. That is, posterior to the third vertebra all the centra are slightly longer than deep, with little difference in pro- portion between abdominal and caudal elements. Apart from the neural spine on the second vertebra the first three vertebrae are like their counterparts in latine fishes. The second vertebra, however, has a very greatly expanded neural spine (see Fig. 25d) into the anterior face of which the entire posterior margin of the first neural spine is fitted. Fraser (1968) has shown that the proportions of the second neural spine change with age in at least some species of Centropomus ; the spine in young fishes resembles that in adult Lates and Psammoperca (see Fraser, op. cit., text-fig. 14, and pp. 454-5 ; and cf. Figs 25a-c above).

All Centropomus species have three predorsal bones, the first situated above the tip of the first neural spine, the second at about the middle of the expanded second spine, and the third lying immediately behind that spine. Fraser (1968) states that there are only two predorsals in Lates, but this is not so (see p. 43 above) ; there are, in fact, no intergeneric differences in this feature.

A distinct gap separates the two dorsal fins in all Centropomus species ; the size of the gap, however, shows some specific variation. Unlike those Latinae with separate dorsals (members of the subgenus Luciolates ; see p. 45), the centropomines have no isolated spines between the fins. The head of the seventh pterygiophore is drawn out posteriorly so that it effectively underlies the gap between the fins ; the spine which this pterygiophore carries thus becomes the first (and only) spine of the second dorsal fin.

There are 16 or 17 dorsal pterygiophores in Centropomus (cf. 18 or 19, rarely 20 in the Latinae), none of which, as far as I can determine, has a separate medial

66 P. H. GREENWOOD

radial (see p. 45). The first dorsal pterygiophore, unlike that of the latines, carries two spines ; except for the last dorsal and first anal pterygiophores, all the others carry a single spine or ray.

The dorsal fin ray counts in Centropomus are VII or VIII and 1,8 or 9 (cf. VI-VIII and 1,10-13 in all Latinae except Lates (Luciolates} stappersi which has VI + I + I and 1,9 or 10). Thus in the centropomines there has been not only a trend towards separation of the dorsal fins (a trend also apparent in the latines, see p. 46 above) but also a reductional trend in the number of dorsal fin rays, particularly the bran- ched rays. Interestingly, if the two independent and much reduced fin spines are 'removed' from the fin formula of Lates (Luciolates) stappersi, the result - save for an extra branched ray -is the formula of a Centropomus (i.e. VI + I + I and I,io — > VI and I,io).

A most characteristic feature of all Centropomus, and one not even approached by any member of the Latinae, is the very strong and long first anal pterygiophore (see Fig. 27b) ; in many species there is also hypertrophy of the second anal spine. Despite the length of this pterygiophore it extends only a little further distally (i.e. towards the vertebral column) than does its counterpart in the Latinae. The greater length of the bone in Centropomus is accommodated by the bone sloping obliquely backwards so that the articulation for the spines lies in a vertical below about the seventh rather than the second or third abdominal vertebra as is the case in Lates (Fig. 2ya) and Psammoperca.

The caudal fin skeleton in Centropomus differs from that in the Latinae in either one (Lates) or two (Psammoperca) characters and is of a more primitive kind. Primi- tive features in Centropomus are the three epurals (two in Latinae) and the two uroneurals. Lates also has two uroneurals but only one is present in Psammoperca (the fin skeleton in that genus being the most evolved within the Centropomidae) .

All Centropomus species have a deeply forked caudal fin whereas in the Latinae the fin is usually rounded or truncate, although it is weakly forked in some species of the Lates subgenus Luciolates. Like the Latinae, the caudal fin formula of the Centropominae is 1,8 + 7,!.

The pectoral girdle and fin skeleton are basically alike in the Centropominae and Latinae except for slight differences in the postcleithral elements.

The posttemporal in Centropomus lacks the cavity and associated bullation that characterize this bone in Lates and Psammoperca, a consequence of there being no swimbladder- posttemporal ligament in Centropomus (see below). Otherwise the posttemporal is similar in both subfamilies. The extrascapula in Centropomus is also basically like that in the Latinae, but it is characterized by having the lateral line canals situated in open gutters and not enclosed in bony tubes (see p. 41 above).

In those Centropomus species which I have been able to dissect, the anterior end of the swimbladder has no medial invagination (see p. 60 above). However, in some species there are a pair of short horns arising from the dorsolateral aspect of the bladder and extending part way towards the skull ; in none could I find any direct connection between the skull and the horns, and neither could I find any trace of a swimbladder -posttemporal ligament such as occurs in all members of the Latinae. The development of the swimbladder horns seems to be restricted to certain species.

REVIEW OF CENTROPOMIDAE 67

When present these appendages may be short, simple and anteriorly directed, or, as in C. imdecimalis, they may be elongate and curved backwards to lie laterally along the swimbladder (see Meek & Hildebrand, 1925). This latter condition is reminiscent of that found in certain species of Sciaenidae (a family in which there is also an extension of the lateral line onto and reaching the margin of the caudal fin ; the possibility of there being some phyletic relationship between sciaenids and centropomids is under review).

Baudelot's ligament is present in Centropomus and is moderately well developed. The relationships between this ligament and the body muscles are like those in the Latinae (see p. 60 above), with little or no muscle passing medially below the liga- ment, but with a broad band passing underneath it laterally to insert on the basi- occipital and exoccipital.

A single extension of the body lateral line scale series onto the caudal fin is found in all species of Centropomus (there are three extensions in Lates but only one in Psammoperca). In Centropomus, as in the Latinae, the caudal extension of the lateral line is continuous to the margin of the fin or almost so.

When all these characters are taken into account, it is clear that the Centro- pominae (i.e. Centropomus species) differ from the Latinae (Lates and Psammoperca species) in a number of features. Some of these differences involve the retention of characters primitive for the family whilst others represent the development of unique specializations shared only by Centropomus species. In the former (i.e. plesiomorph) category may be listed the caudal fin skeleton, the short supraoccipital bone, the single lateral line extension onto the caudal fin, and the absence of a swimbladder -posttemporal ligament. The autapomorphic features of the Centro- pominae are more numerous and include the open cephalic lateral line canals, the separation of the dorsal fin, the hypertrophy of the first anal pterygiophore (and at least relative hypertrophy of the second anal spine), the absence of medial radials throughout the dorsal and anal fins, the development of a curved and posteriorly protracted head on the seventh pterygiophore of the dorsal fin (see p. 65), the development in most species of anterior horns on the swimbladder, the incorporation of the supralamellar tooth plates into the gill rakers, and the elongation of the skull, especially its ethmovomerine region (with which feature may be correlated changes in the shape of the ethmoid and vomer).

There are other differences, like the absence of an opercular spine, the forked caudal fin, and the markedly reduced squamation of the dorsal, caudal and anal fins, whose apo- or plesiomorph status is uncertain.

On the basis of those characters that are clearly synapomorphic the Centropomus species can be recognized as a monophyletic group and one which, although sharing a common ancestry with the Latinae, is clearly distinct from that lineage. It is for this reason that I propose giving the Centropomus species-group coordinate ranking (as the subfamily Centropominae) with the Latinae (see also above p. 61 ; and p. 75 below for diagnoses).

When the mosaics of apo- and plesiomorph characters within the two subfamilies are compared it becomes impossible to decide which taxon should be considered the plesiomorph sister group of the other. However, it does seem that what we are now

68 P. H. GREENWOOD

observing is the product of vicariant differentiation from a once widespread basic centropomid taxon, a differentiation that produced the Centropominae in America, and the Latinae in Asia and the Mediterranean region (including Africa), leaving each group with its own association of primitive and derived features.

Fraser (1968) noted certain shared characters amongst the various species of Centropomus, and from their pattern of occurrence concluded that three phyletic lineages are represented amongst the living species. Unfortunately, Fraser does not give a really critical analysis of the characters on which his phylogeny is constructed and it is thus impossible to test the supposed interrelationships of the three lineages he hypothesizes. In particular it would seem that his monotypic lineage comprising C. pectinatus is more likely a member of the C. ensiferus - C. robalito lineage, and not, as is expressed in Fraser 's diagram, one distinct from the other two lineages and occupying an equal phyletic relationship with both of them (see Fraser, 1968, text-fig. 9).

Although Fraser's analysis is not documented in terms of synapomorph and symplesiomorph characters it obviously shows that similar trends can be found within the Centropominae and the Latinae. This aspect is particularly well demon- strated in the neurocranial morphology and in the reduction of the pterosphenoid pedicle and internal jugular bridge. There is also inter-subfamilial similarity in the trend towards greater separation of the two parts of the dorsal fin. In this trend the Latinae appear not to have evolved much beyond the early phases, whereas the centropomines have carried the trend further and no longer preserve traces of its earlier stages within their numbers.

FOSSIL CENTROPOMIDAE

Apart from identifications based solely on otoliths,* all species of fossil centro- pomids so far discovered are currently referred either to Lates or to Eolates Sorbini (see Sorbini, 1973), that is, to the subfamily Latinae.

The time range of these fossil taxa extends from the Eocene to the Holocene, and their geographical range from the Paris Basin, through Austria, Portugal, northern Italy and Croatia to Egypt, the Sahara and eastern Africa (Sorbini, 1973 ; Green- wood, 1974 ; Greenwood & Howes, 1975).

With the exception of some material from Europe (Sorbini, 1973) the majority of fossils are from Africa and are in the form of disarticulated and damaged bones. The problems of specific (or, indeed, generic) identification when dealing with material of this nature need not be stressed. In most instances the fossil bones have been compared with their counterparts in Lates calcarifer and L. niloticus. If the fossils are from Africa, and the bones are not noticeably different from their counter- parts in L. niloticus the material was either referred to that species or, and probably more accurately since diagnostic features are rarely preserved, merely to Lates sp.

* Psammoperca sheppeyensis Frost 1934, Centropomus superpendens Frost, 1934 an(l C. excavatus Stinton, 1966, all from the London Clay (Eocene), are species described from otoliths only. Since so little is known about otolith morphology in living centropomids and because no skeletal material is available for the species, these records can at present contribute little to our understanding of centro- pomid phylogeny and biogeography.

REVIEW OF CENTROPOMIDAE 69

When obvious morphological differences could be detected the remains have been taken to represent different species (e.g. L. fajumensis Weiler, 1929 ; L. karungae Greenwood, 1951 ; L. rhachirhinchus Greenwood & Howes, 1975).

Because these species are based on fragmentary, disarticulated bones it is impos- sible to determine their phyletic relationships with each other or with the extant species of Latinae (see discussions on L. rhachirhinchus in Greenwood & Howes, 1975). About all that can be said with any certainty is that latine centropomids had, by late Miocene times, a distribution that included Egypt, Tunisia and eastern Africa (Lakes Victoria and Albert regions) and that at least one species, L. rhachi- rhinchus, showed several derived characters even when compared with extant species of that genus (Greenwood & Howes, 1975). All these remarks are, of course, based on the assumption that the taxa are correctly placed in the genus Lates ; in no case is it possible to check on the autapomorph characters used here to define the genus (see p. 77), the generic identity being based on an overall similarity between the fossil bones and their counterparts in extant Lates species.

The situation is little better for the three European species in which the entire skeleton is preserved, viz. L. partschii Heckel, 1855 (Miocene, Vienna Basin) ; L. croaticus Kramberger, 1902 (Miocene, Croatia) and L. macropterus Bassani, 1899 (Oligocene of Vicenza).

I have not been able to examine any material of L. croaticus, and the only pub- lished description and figures of this species are inadequate for critical interpretation, although Kramberger (1902) does give a vertebral count of 27, that is two more vertebrae than in any other member of the Latinae for which the count is available. The status and relationships of this nominal species must therefore remain incertae sedis.

Sorbini (1973) has re-examined L. macropterus, but was unable to draw any definite conclusions about its relationships. Again it is impossible to check on any diagnostic characters of phyletic importance.

Lates partschii (Miocene of Vienna) has been thoroughly redescribed by Sorbini (1973), who also published a photograph of the holotype, and a close-up picture of its caudal skeleton. But once again certain critical details are either not preserved, are obscured, or are damaged. For instance, there seems to be only a single and median row of lateral line scales on the caudal fin but one cannot be certain that dorsal and ventral scale rows were not present. There are certainly only two epurals in the caudal fin skeleton, and there are, apparently, two uroneurals, both features which are characteristic of Lates (see pp. 44 and 77). From this and other circumstantial evidence given in Sorbini's account, it seems likely that partschii can be placed in Lates, but it is impossible to determine its relationships with any extant species of that genus.

Fortunately, many important features are preserved in the extensive material of Eolates that is available for study (Sorbini, 1973 ; personal observations on specimens in the collections of the British Museum (Natural History))*.

* Two nominal species are recognized, Eolates gracilis (Agassiz) 1833, and E. macrurus (Agassiz) 1833. According to Sorbini (1973), E. macrurus may yet prove to be a synonym of E. gracilis. For this reason, and because the osteology of E. gracilis is much better known, only that species is taken into account in the discussions that follow.

7o P. H. GREENWOOD

Eolates gracilis is distinguished from all Lates species by at least one character complex (the caudal fin skeleton), and probably by two other characters as well (the absence of upper and lower lateral line scale rows on the caudal fin, and the disposition of the branchiostegal rays).

I have examined eight specimens of E. gracilis (from the BMNH collections) in which the caudal fin is well preserved. In all, the median lateral line scale row is clearly developed and it is also possible to see other scales, often still in their rows, on other parts of the fin. None of these other scales is perforated and I am confident that only one lateral line scale row (the median one) is preserved. My colleague, Dr K. Banister, has recently examined E. gracilis holotype in the Paris Museum and reports that only a median row can be detected in this specimen as well.

The caudal fin skeleton in E. gracilis (Fig. 26b) shows a well-developed neural spine on the second preural centrum (spine greatly reduced in Lates), three epurals (two in Lates) and two uroneurals (two uroneurals also present in Lates). In other words, the caudal fin skeleton is of a more primitive type than that in Lates. (It will be recalled that Centropomus also has three epurals, but the second preural arch and spine are reduced and resemble those in Lates.)

A possible third intergeneric difference concerns the number and disposition of the branchiostegal rays, but this requires confirmation since it is based on data available from only one of the E. gracilis specimens examined by Sorbini (1973). In the sole specimen from which a branchiostegal ray count could be made Sorbini (op. cit.) records, with some uncertainty, a total of eight rays (seven in Lates and other centropomids). Judging from the photograph of this specimen (Sorbini, 1973, Plate IV, fig. i), I should doubt that the fragment at the anterior end of the ceratohyal is indeed part of a branchiostegal ray.

There is, however, no doubt that in this specimen all the branchiostegal rays are associated with the ceratohyal. According to McAllister (1968) this condition is not found in any living percoid fish ; there is always at least a half articulation between a ray and the epihyal. The rays in the E. gracilis specimen are in no way disarranged, and the posterior one is well forward of the epi-ceratohyal junction. Clearly no decision can be made on the validity of this apparent intergeneric differ- ence (or its apparent uniqueness amongst percoids) until further specimens can be examined.

Like the preoperculum in Lates this bone in Eolates has three large ventral spines on its horizontal limb, and an enlarged spine at the posterior angle of the bone. Also as in Lates, there is a single, large spine on the posterodorsal margin of the operculum in Eolates. Ornamentation of the cleithrum and on the first infraorbital bone (lachrymal) is similar in Eolates and Lates, but the phylogenetic importance of these latter characters is probably not great.

Regrettably, little detailed information can be obtained about the morphology of the neurocranium in Eolates. Sorbini (1973) gives no description of the posterior orbital region of the skull, presumably because in his material, as in that of the BMNH, this area of the head is either crushed or obscured by other bones overlying it. Thus it is impossible to determine what type of pterosphenoid pedicle and inter- nal jugular bridge is present.

REVIEW OF CENTROPOMIDAE 71

The ethmoid region is generally well preserved, and resembles that found in members of the subgenus Lates (see p. 19 above).

In a few E. gracilis specimens the posttemporal is well preserved ; it seems to show the slightly bullate outer surface that, in living centropomids, is associated with the insertion point of a swimbladder- posttemporal ligament (see p. 41 above), a derived feature characterizing members of the subfamily Latinae (see above, p. 66).

In brief, Eolates (as represented by E. gracilis) is clearly a member of the sub- family Latinae and shares at least one derived character (the ventral preopercular spines) with the genus Lates (see p. 31). Eolates differs from Lates in having only one series of lateral line scales on the caudal fin (presumed in this context to be a primitive feature, see p. 48 above), and in having a caudal fin skeleton that is primitive in relation to this skeleton in Lates (see above, p. 66). A third intergeneric difference is in the less deeply indented dorsal fin of Eolates, a feature with which may be cor- related the equal spacing between the 'last' (i.e. shortest) spine of the anterior part of the fin and the longer 'first' spine of the fin's posterior half. This character too should be considered a plesiomorph one because a deeply divided fin is a basal condition in the centropomid trend leading towards completely separate first and second dorsal fins (see above, pp. 46 and 65).

All the features discussed so far indicate that Eolates should be considered more primitive than Lates. In phyletic terms it should be ranked as the plesiomorph sister group of that taxon. The relationship of Eolates within the subfamily Latinae is, therefore, best indicated by uniting Eolates with Lates in a single tribe (Latini, new tribe) which would then become the sister taxon of the tribe containing only the genus Psammoperca (tribe Psammopercini nov.).

Sorbini (1973) also recognizes the affinity of Lates and Eolates, but he would regard the relationship as an ancestor -descendant one (op. cit. : 41) rather than that of recent shared common ancestry as is proposed here.

Sorbini's claim that ' . . . The living marine species L. calcarifer presents the greatest relationship to fossil Tertiary species, which lived in a similar habitat' (Sorbini, 1973 : 41) certainly cannot be substantiated by the meristic and morpho- logical data available from these fossils. For example, as interspecific similarities between L. calcarifer and E. gracilis Sorbini lists (op. cit. : 36) ' . . . disposizione delle vertebre, n. raggi branchiostegi, habitat . . .'. The habitat is similar, but what importance can be attached to this feature in a family with several euryhaline species? The arrangement of the vertebrae in Eolates is like that in several Lates species, while the reference to the number of branchiostegal rays is, I presume, a lapsus for 'spine branchiali'. Eolates has either seven or eight branchiostegal rays (there are seven in all other centropomids ; see above, p. 70), but nine gill rakers (the same number as L. calcarifer}. However, a low gill raker count (8-12) is com- mon to several Lates species, and is apparently the primitive state for the family as a whole.

BIOGEOGRAPHY

The contemporary world distribution of the Centropomidae (Fig. 36) strongly suggests a Tethyan distribution for the common ancestor of its two subfamilies, the

P. H. GREENWOOD

FIG. 36. World distribution of extant species of Centropomidae. Stippled areas : Lates species (outside Africa = L. calcarifer). Black spots : Psammoperca species (probably only one, P. waigiensis). Black area ; Centropomus species.

Centropominae (America) and the Latinae (Mediterranean, African freshwaters and Asia).

The Centropominae, on this hypothesis, have evolved in the tropical New World, probably in estuarine and marine habitats, and the Latinae principally in African freshwater habitats. There is, however, a major dichotomy in the Latinae, between the tribes Latini and Psammopercini, which must have taken place before the Latini invaded Africa.

As evidenced by various European fossil species (see above, pp. 68-71, and Sorbini, 1973) and by the wide dispersal of Lates calcarifer (see p. 12), the Latini were and still are successful coastal fishes. The greater diversification of the tribe in African freshwaters can probably be attributed to the greater opportunities for speciation provided by the developing tropical lakes and river systems of later Tertiary and Quaternary Africa. (For a summary of these historical factors see Beadle, 1974.) It will be recalled that there are seven extant and at least three extinct Lates species in Africa, compared with the single extant (L. calcarifer} and three extinct marine or estuarine species (see above, p. 69 ; also Sorbini, 1973 ; Greenwood, 1974 ; Greenwood & Howes, 1975).

There are, of course, at least nine species of Centropomus (Centropominae) all of which are essentially marine species (although some freely enter freshwater ; Meek & Hildebrand, 1925). Trans-isthmian isolation could account for four of these species (Eraser, 1968) but there still remain the other five species to contrast with the single marine Lates species (L. calcarifer} of the Indo-Pacific region. The causal factors involved in this aspect of Centropomus speciation are not apparent.

It is interesting to compare the morphological radiation undergone by the Centro- pominae and Latinae, and to notice the marked parallelism apparent in the two

REVIEW OF CENTROPOMIDAE 73

groups. For example, in both subfamilies there are trends of specialization leading to the reduction and loss of the pterosphenoid pedicle and internal jugular bridge (p. 63), to elongation of the skull through differential lengthening of the ethmoid region (p. 62), to an increase in the number of gill rakers on the first gill arch, to a reduction in the number of supralamellar tooth plates (p. 64), and towards the division and then separation of a primitively continuous dorsal fin (see p. 66). So similar are all the features involved in any one of these trends that one can eliminate any possibility of convergence. The similarities must reflect shared genotypic factors stemming from common ancestry.

The absence of Lates (or some related latine fish) from the present-day Medi- terranean Sea may, as Sorbini (1973 : 40) suggests, be due to climatic changes adversely affecting the one or more species that were present in the Mediterranean basin during parts of the Tertiary (Sorbini, op. cit., especially text-fig. 10). In- creasing salinity in the developing Mediterranean may also have had its effect on local populations.

During the Eocene and Miocene, species of Lates were also present in Africa (Sorbini, 1973 ; Greenwood, 1974). The Eocene fishes from the Fayum in Egypt may have been estuarine and marine (Weiler, 1929), as may have been the Miocene species from Tunisia (Greenwood, 1973). However, Miocene records of Lates from the equatorial regions of Lake Victoria (Greenwood, 1951) and Lake Albert (Green- wood & Howes, 1975) show that some latine species had adjusted fully to freshwater environments, and that enough time had elapsed since the first invasion for latine species to have reached areas some 3750 km inland from the Mediterranean coast.

Like all other fossil Latinae from Africa, the Miocene species are referred to Lates solely on the overall similarity between the preserved fossil bones and their counter- parts in extant Lates species. Such critical features as the nature of the lateral line scales on the caudal fin and the morphology of the posttemporal bone are unknown for any one of them. One Miocene species, L. karungae Greenwood, 1951, from Rusinga Island, Lake Victoria, is represented by only a few vertebrae ; the specific diagnostic features for this taxon relate to the morphology of the third vertebra (Greenwood, 1951). The other taxon, L. rhachirhinchns, from the Lake Albert-Lake Edward region of Zaire is better represented by numerous skeletal parts (Greenwood & Howes, 1975). It differs from all other Lates species in several features, many of which can be considered as derived, and one of which (vertebral proportions) is shared with certain members of the endemic subgenus Luciolates from Lake Tangan- yika (see p. 43 above, and Greenwood & Howes, op. cit.). Even though it is im- possible to identify specifically the Lates remains from the Miocene and Pliocene deposits in North Africa and Egypt, L. rhachirhinchus is morphologically quite distinct from those taxa.

Thus, one may conclude from this situation either that more than one taxon invaded Africa or that, by Miocene times, the population of Lates in the Lake Albert -Lake Edward region had undergone marked morphological differentiation, presumably in isolation from its parental stock. The same arguments could be applied to L. karungae although in this instance there is less evidence for the extent to which the morphological differentiation had progressed.

74 P- H. GREENWOOD

Sorbini (1973) postulated certain time sequences and migration routes to explain the present-day distribution of Lates species in Africa. Basically the problem Sorbini sets out to explain is the widespread occurrence of one species, L. niloticus, in the Nile, Niger, Zaire, and Senegal river systems, and in Lakes Rudolf, Albert and Chad, in contrast to the occurrence of four endemic species (one supposedly a distinct genus) in Lake Tanganyika. He notes the former occurrence of Lates in other lakes (Edward and Victoria) but is not concerned with the factors that led to these local extinctions, and neither does he take into account the endemic species that coexist with L. niloticus in Lakes Rudolf and Albert.

There are two basic tenets in Sorbini's hypothesis, first that the various invasions he postulates originated in Egypt, and second that fossils identified as L. niloticus are indeed representatives of that species. As I have discussed above the latter assumption is not necessarily acceptable, and neither can I find any a priori grounds for postulating repeated and temporally extended invasions from a single area (in this argument, Egypt).

That a species of Lates had reached the regions of Lake Victoria and Lake Albert - Lake Edward by Miocene times is not disputed (see above), and Sorbini's argument for the contemporaneous presence of a Lates species in the Lake Tanganyika basin is also acceptable. Why, then, should Lates not have occurred in other Miocene rivers and water bodies, environmental conditions, of course, permitting such colonization? To the best of my knowledge there is no evidence to show that suitable conditions were confined to the regions from which Miocene fossils have been re- covered. Thus I find it difficult to understand why, in order to explain the present distribution of L. niloticus, Sorbini should postulate two invasions, each following different routes, but both originating from Egypt during the Pliocene and continuing through the Pleistocene. Presumably a major reason for putting forward this hypothesis is the fact the fossils identified as L. niloticus are first recorded from the Pliocene of Egypt, thereby implying the origin of that species in Egypt at a later date than the one at which another species (L. karungae] was already present in the Lake Victoria area (and, had he known it, a second species L. rhachirhinchus was present in the area of Lake Albert-Lake Edward ; Greenwood & Howes, 1975).

In view of the known distribution for Miocene Lates and because of the uncertain- ties associated with the specific identification of most fossil Lates remains, a simpler hypothesis can be made, viz. :

At some stage prior to the late Eocene a species of Lates invaded Africa, possibly through more than one entry point, but almost certainly from the north. In the course of time this species gradually dispersed through the various river systems with some isolated populations evolving into distinct species now extinct (e.g. L. rhachirhinchus and L. karungae, possibly also L. fajumensis), and others or their descendants (like the endemic species of Lake Tanganyika) still surviving. A little modified descendant of the original invader, the species now recognized as L. nilo- ticus, continued to spread (by such means as river capture or lake extension) until it came to have its present distribution. The L. niloticus-like fossils of Pleistocene times (Greenwood, 1959, 1974 ; Sorbini, 1973) stand witness to a much wider area

REVIEW OF CENTROPOMIDAE

75

for the distribution of Lates and even probably for the species L. niloticus (but of that point we must remain uncertain).

It is unnecessary here to discuss the history of L. niloticus in lakes such as Rudolf and Albert which may, at some time in their histories, have dried out completely, and which have had complex relationships with the River Nile and other lakes (see discussions in Greenwood, 1959, 1974 ; also Beadle, 1974). There is still, however, the problem of the endemic Lates species in the two lakes, L. longispinis in Lake Rudolf and L. macrophthalmus in Lake Albert. In brief, on morphological criteria (p. 12) these species are apparently more closely related to one another than either is to L. niloticus, the species from which each was supposed to have been derived at some time during the Pleistocene (Worthington, 1932 ; Holden, 1967). On the evidence currently available it is impossible to determine whether L. longis- pinis and L. macrophthalmus do in fact represent survivors of a distinct lineage or if, as Worthington (1932) postulated, they are offshoots of earlier L. niloticus popula- tions that once inhabited the two lakes (see discussions on pp. 13 and 14).

DIAGNOSES FOR THE CENTROPOMIDAE, ITS SUBFAMILIES, GENERA AND SUBGENERA

CENTROPOMIDAE Poey, 1868

Poey, F., 1868, Repertorio Fisico-Natural de Cuba, 5, no. 2 : 280. (See also Gill, T., 1883, Proc. U.S. natn. Mus., 5 : 484-485).

TYPE GENUS : Centropomus Lacepede, 1802.

DIAGNOSIS. Percoid fishes, some attaining a large size (up to 2m), with the neural spine of the second vertebra markedly expanded in an anteroposterior

CENTROPOMINAE

LATINAE

Eolatest Lates(Luciolates) Lates( Lates)

FIG. 37. Cladogram to illustrate phyletic relationships within the Centropomidae.

76 P. H. GREENWOOD

direction, and the pored scales of the body lateral line continued onto the caudal fin, reaching the posterior margin of that fin in all but one species. Twenty-four or 25 vertebrae (including the fused first ural and preural centra of the caudal skeleton) ; pleural ribs associated with parapophyses except on the first three to five rib-bearing vertebrae (the first two vertebrae of the column are without ribs) ; 3 predorsal bones. Dorsal fin either deeply divided, the first part entirely spinous (7 or 8 spines), the second of one spine and 8-n branched rays, or the two parts of the fin separated by a distinct gap ; anal fin with 3 spinous and 6-9 branched rays ; caudal fin with 17 principal rays, its posterior margin rounded, truncate or forked. Scales ctenoid, small to moderate in size, dorsally not extending forward on to the head beyond the level of the midpoint of the eye (usually only to the level of the posterior margin of the orbit) but present on the cheek and operculum ; scaly sheath at the base of the anal and soft dorsal fins, but squamation extending onto all fin membranes (including that of the caudal). No scales on the maxilla ; a small supramaxilla present. Teeth on the premaxilla, dentary, vomer, palatine and, in most species, the ectopterygoid ; teeth absent, except in Psammoperca, from the glossohyal. Jaw teeth generally small, viliform or conical, and arranged in several rows. Pterosphenoid pedicle and internal jugular bridge present in all but three or four species, although variously developed ; frontoparietal crests present. Seven branchiostegal rays. Pseudobranch present. About 20 extant species from marine, estuarine and freshwater habitats in the tropical New World (Atlantic and Pacific coasts), tropical Africa (predominantly fresh- or brackish water species), and from Indo-Pacific coastal waters. Six extinct species (some from Europe), the earliest being from the Eocene of Monte Bolca.

Subfamily GENTROPOMINAE

Centropomid fishes with 24 vertebrae ; the cephalic lateral line canals not enclosed in bony tubes but carried in skin-covered bony gutters ; the supraoccipital barely separating the frontals ; the first anal pterygiophore hypertrophied and inclined backwards at an oblique angle ; no opercular spine but three or four enlarged spines at the posterior angle of the operculum ; no swimbladder-posttemporal ligament developed ; no isolated spine or spines situated between the first and second dorsal fins (these fins always separated by a distinct gap) ; pseudobranch superficial.

TYPE GENUS : Centropomus Lacepede, 1802.

A single genus Centropomus (type species Sciaena undecimalis Bloch, 1792), generic diagnosis as for the subfamily with, additionally, caudal fin skeleton having three epural and two uroneural bones. The genus is confined to the tropical waters of North, Central and South America, and occurs on both the Pacific and Atlantic coasts. A key to the species of Centropomus is provided by Meek & Hildebrand (1925), and supplementary information by Chavez (1961) and Rivas (1962).

Subfamily LATINAE Jordan (1923)

Centropomid fishes with 25 vertebrae ; the cephalic lateral line enclosed in bony tubes ; the supraoccipital extending far forward between the frontals ; the first

REVIEW OF CENTROPOMIDAE 77

anal pterygiophore not hypertrophied, and inclined backwards at only a slight angle ; a single well-developed opercular spine and a single, enlarged spine at the posterior angle of the preoperculum ; a stout ligament connecting the swimbladder with the posttemporal (which is itself modified to receive the ligament) ; dorsal fin deeply indented or separated into two fins (between which there are one or two isolated spines) ; pseudobranch covered.

TYPE GENUS : Lates Cuvier & Valenciennes, 1828.

Three genera, two extant and one extinct. The two extant genera are :

PSAMMOPERCA Richardson, 1844

TYPE SPECIES : Ldbrax waigiensis C. & V., 1828.

Latine fishes with : a smooth horizontal limb to the preoperculum, a basihyoid tooth plate, supralamellar tooth plates present on the outer face of the first four gill arches only ; with a single series of lateral line scales on the caudal fin, with the nostrils of each side widely separated, and a caudal fin skeleton in which there are two epural bones and a single uroneural.

Probably only one species, P. waigiensis (a second nominal species P. macroptera Giinther, 1859, is almost certainly a synonym), widely distributed in the coastal waters of the Indo-Pacific.

LATES C. & V., 1828

TYPE SPECIES : Perca nilotica, L., 1758.

Latine fishes with the horizontal limb of the preoperculum produced into three or four (rarely more) large, flattened and triangular spines, no basihyoidal tooth plate but supralamellar tooth plates present on both the anterior and posterior faces of the first four gill arches, with three series of lateral line scales on the caudal fin, with the nostrils of each side close together, and a caudal fin skeleton with two epurals and two uroneurals.

Eight extant species (seven of which are African and confined to freshwaters, and one marine or estuarine and widely distributed in Indo-Pacific coastal waters) arranged in two subgenera :

LATES (LATES)

TYPE SPECIES : L. niloticus (L).

Species of the genus in which the posterior face of the lateral ethmoid has only a slight slope posteriorly, the dorsolateral parts of that bone are almost horizontally aligned, and the entire ethmovomerine region of the skull is not noticeably elongate. Four species : L. calcarifer (Indo-Pacific), L. niloticus (rivers of northern and western tropical Africa, and also in Lakes Chad, Albert and Rudolf [introduced into Lakes Victoria and Kioga]), L. macrophthalmus (Lake Albert only) and L. longispinis (Lake Rudolf only).

78 P. H. GREENWOOD

LATES (LUCIOLATES)

TYPE SPECIES : Luciolates stappersi Boulenger, 1914.

Species of Lates having a characteristically shaped and elongate ethmovomerine skull region in which the posterior face of the lateral ethmoid slopes backwards at a pronounced angle, and the dorsolateral aspects of that bone are directed ventrally at a steep angle. Four species, all endemic to Lake Tanganyika : L. angustifrons, L. microlepis, L. mariae and L. stappersi.

The single extinct genus is :

EOLATES Sorbini, 1970 TYPE SPECIES : Lates gracilis Agassiz, 1883.

See Sorbini, 1973, for full description, synonymies, etc.

Eolates, with one species (E. gracilis} and possibly a second, E. macrurus (Ag.), 1833, is known only from the lower Eocene deposits of Monte Bolca, northern Italy.

Eolates differs from Lates in the structure of its caudal fin skeleton (three epurals ; a well-developed neural spine on the second preural vertebrae), in having only a single series of lateral line scales (the median one) on the caudal fin, and in having a less deeply indented dorsal fin (see p. 70 above).

The phyletic relationships of Eolates within the Latinae are discussed on p. 71, where it is suggested that Lates and Eolates are sister taxa and should be placed in the Tribe Latini nov., the sister group of the Tribe Psammopercini nov. (a taxon containing only the genus Psammoperca) .

ACKNOWLEDGEMENTS

Again, it is a pleasure for me to thank Drs Colin Patterson, Donn Rosen and Gareth Nelson for the numerous and illuminating discussions we have had on the centropomid fishes and on aspects of their anatomy in a broader context. To my colleague, Gordon Howes, goes my gratitude for the great amount of work and skill he has put into producing the figures illustrating this paper, and, as always, for his invaluable assistance in innumerable other ways. To Dr Thys van den Audenaerde go my thanks for lending me specimens of Luciolates stappersi from the Musee Royal de 1'Afrique Centrale, Tervuren, to Dr Mary Burgis (the City of London Poly- technic) for her considerable efforts in getting material from Lake Tanganyika, and to my colleague, Dr Keith Banister, for examining the type of Eolates gracilis on my behalf.

REFERENCES

ALLEN, W. F. 1905. The blood vascular system of the Loricati, the mail-cheeked fishes.

Proc. Wash. Acad. Sci. 7 : 27-157. ALLIS, E. P. 1909. The cranial anatomy of the mail-cheeked fishes. Zoologica, Stuttg. 22

(57) : 1-219.

REVIEW OF CENTROPOMIDAE 79

BASSANI, F. 1889. Ricerche sui pesci fossili di Chiavon. AM Accad. Sci. fis. mat. Napoli

2 (6) : 1-102.

BEADLE, L. C. i'>74- The inland waters of tropical Africa, viii + 365 pp. London. BERG, L. S. 1940. Classification of fishes, both recent and fossil. Trav. Inst. zool. Acad. Sci.

U.S.S.R. 5 (2) : 87-517. (Russian and English texts. Reprint Edwards, Ann Arbor,

Michigan, 1947.)

BLOCK, M. E. 1792. Naturgeschichte der ausldndischen Fische, 6. Berlin. BOULENGER, G. A. 1914. Mission Stappers au Tanganyika-Moero. Diagnoses de poissons

nouveaux. I. Acanthopterygians, Opisthomes, Cyprinodontes. Revue zool. afr., 3.

442-447. 1915- Catalogue of the freshwater fishes of Africa in the British Museum (Natural History),

3:xii + 526. London. CHAVEZ, H. 1961 . Estudio de una nueva especie de robalo del Golfo de Mexico y redescripcion

de Centropomus undecimalis (Bloch) (Pise., Centropom.). Ciecia Mex., 21 (2) : 75-83. CUVIER, G. L. C. F. G. & VALENCIENNES, A. 1828. Histoire naturelle des poissons, 2 : 490 pp.

Paris.

DAGET, J. 1954. Les poissons du Niger superieur. Mem. Inst. fr. Afr. noire, 36 : 1-391. DUNSTAN, D. J. 1959. The Barramundi Lates calcarifer (Bloch) in Queensland waters.

Tech. Pap. Div. Fish. Aust. no. 5 : 1-22. FRASER, T. H. 1968. Comparative osteology of the Atlantic Snooks (Pisces, Centropomus).

Copeia, 1968 : 433-460. FROST, G. A. 1934. Otoliths of fishes from the lower Tertiary formations of southern England.

Ann. Mag. nat. Hist. (10), 13 : 426-433.

GILL, T. N. 1883. On the family Centropomidae. Proc. U.S. natn. Mus. 5 : 484-485. GOSLINE, W. A. 1966. The limits of the fish family Serranidae, with notes on other lower

percoids. Proc. Calif. Acad. Sci. 33 : 91-112. GOSSE, J.-P. 1956. Dispositions speciales de 1'appareil branchial des Tilapia et Citharinus.

Ann. Soc. zool. Belg. 86 : 303-308. GREENWOOD, P. H. 1951. Fish remains from Miocene deposits of Rusinga Island and Kavi-

rondo Province, Kenya. Ann. Mag. nat. Hist. (12), 4 : 1192-1201.

- 1959- Quaternary fish-fossils. Explor. Pare. natn. Albert Miss. J. de Heinzelin de Brau- court, Brussels, 4:i-8o.

— 1973. Fish fossils from the late Miocene of Tunisia. Notes Serv. geol. Tunis, 37 : 41-72.

- 1974- Review of Cenozoic freshwater fish faunas in Africa. Ann. geol. Surv. Egypt, 4 : 211-232.

- & HOWES, G. J. 1975. Neogene fossil fishes from the Lake Albert-Lake Edward Rift (Zaire). Bull. Br. Mus. nat. Hist. (Geol.), 26 : 69-124.

- ROSEN, D. E., WEITZMAN, S. H. & MYERS, G. S. 1966. Phyletic studies of teleostean fishes, with a provisional classification of living forms. Bull. Am. Mus. nat. Hist. 131 :

339-455- GREGORY, W. K. 1933. Fish skulls : A study of the evolution of a natural mechanism.

Trans. Am. phil. Soc. 23 : 75-481.

GUNTHER, A. 1859. Catalogue of the fishes in the British Museum, 1 : xxi + 524. London. HECKEL, J. J. 1855. Neue Beitrage zur Kenntniss der fossilen Fische Osterreichs. Denkschr.

Akad. Wiss. Wien, 11 : 187-274. HOLDEN, M. J. 1967. The systematics of the genus Lates (Teleostei: Centropomidae) in Lake

Albert, East Africa. /. Zool., Lond. 151 : 329-342. HOLMGREN, N. & STENSIO, E. A. 1936. Kranium und Visceralskelett der Akranier, Cyclosto-

men und Fische. In : Bolk, L., Goppert, E., Kallius, E. and Lubbock, W. (Eds), Handbuch

der vergleichenden Anatomie der Wirbeltiere, 4 : 233 — 500. Berlin and Vienna. JORDAN, D. S. 1923. A classification of fishes including families and genera as far as known.

Stanford University Publications, University Series, Biological Sciences, 3, no. 2 :

i-x + 77-243-

8o P. H. GREENWOOD

KATAYAMA, M. 1954. Systematic position of the genus Glaucosoma. Bull. Fac. Ed. Univ. Yamaguchi, 4, no. i : 23-29.

1956. On the external and internal characters of Lates calcarifer (Bloch), with its systema- tic position. Bull. Fac. Ed. Univ. Yamaguchi, 6, no. i : 133 — 140.

KRAMBERGER, K. G. 1902. Palaeichthyologische Beitrage. Mitt. Jb. K. ung. geol. Anst. 14 : 1-21.

LAKE, J. S. 1971. The freshwater fishes and rivers of Australia. 61 pp. Melbourne.

LINDBERG, G. U. 1971. Opredelitel' i kharakteristika semelstv ryb mirovolfauny. Izdatel'stvo 'Nauka' Leningrad. (Translated as : Fishes of the world. Halstead Press, New York,

1974- 545 PP-) MCALLISTER, D. E. 1968. The evolution of branchiostegals and associated opercular, gular,

and hyoid bones and the classification of teleostome fishes, living and fossil. Bull. natn.

Mus. Can. no. 221 : xiv-239pp. MATSUBARA, K. 1955. Fish morphology and hierarchy. Pts I-III. 1605 pp. Tokyo,

Ishizaki-Shoten . MEEK, S. E. & HILDEBRAND, S. F. 1925. The marine fishes of Panama. Pt II. Publs

Field Mus. nat. Hist. no. 226, Zool. Ser., 15 : 331-707. MUNROE, I. S. R. 1961. The marine and freshwater fishes of Ceylon. 351 pp. Dept. of

External Affairs, Canberra. NELSON, G. J. 1969. Gill arches and the phylogeny of fishes, with notes on the classification

of vertebrates. Bull. Am. Mus. nat. Hist. 141 : 475-552. NORMAN, J. R. 1966. A draft synopsis of the orders, families and genera of recent fishes and

fish-like vertebrates. 649 pp. British Museum (Nat. Hist.), London. PATTERSON, C. 1964. A review of Mesozoic acanthopterygian fishes, with special reference to

those of the English Chalk. Phil. Trans. R. Soc. B, 247 : 213-482. 1975. The braincase of pholidophorid and leptolepid fishes, with a review of the actino-

pterygian braincase. Phil. Trans. R. Soc. B, 269 : 275-579. PELLEGRIN, J. 1922. Poissons de 1'Oubanghi-Chari recueillis par M. Baudon. Description

d'un genre de cinq especes et d'une variete. Bull. Soc. zool. Fr. 47 : 64-76. POLL, M. 1953. Poissons non Cichlidae. Result, scient. Explor. hydrobiol. Lac Tanganyika

(1946-1947), 3 (sA) : 1-251. I957- LCS genres des poissons d'eau douce de 1'Afrique. Annls Mus. r. Congo Beige

Ser. 8°, Sci. Zool. 54: 1-191. REGAN, C. T. 1913. The classification of the percoid fishes. Ann. Mag. nat. Hist. (8),

12 : 111-145. RICHARDSON, J. 1844. Ichthyology. In : Richardson, J. & Gray, J. E. (Eds), The zoology

of the voyage of H. M.S. Erebus and Terror, under the command of C apt. Sir J. C. Ross during

1839-43. 139 pp. London. RIVAS, L. R. 1962. The Florida fishes of the genus Centropomus, commonly known as snook.

Q. Jl Fla Acad. Sci. 25 : 53-64. ROGNES, K. 1973. Head skeleton and jaw mechanism in Labrinae (Teleostei: Labridae) from

Norwegian waters. Arbok Univ. Bergen Mat.-Naturv. Ser. no. 4 : 1-149. ROSEN, D. E. 1973. Interrelationships of higher euteleostean fishes. Zool. J. Linn. Soc.

53 (Suppl. i) : 397-5I3- SORBINI, L. 1973. Evoluzione e distribuzione del genere fossile Eolates e suoi rapporti con

il genere attuale Lates ( Pisces -Centropomidae). In : Studi e ricerche sui giacimenti

terziari di Bolca. Memorie Mus. civ. Stor. nat. Verona (fuori Ser. 2), 2 (i) : 1-43. STINTON, F. C. 1966. Fish otoliths from the London Clay. In : Casier, E., Faune ich-

thyologique du London Clay : 404-496. British Museum (Nat. Hist.), London. WEBER, M., & DE BEAUFORT, L. F. 1929. The fishes of the Indo-Australian Archipelago,

5:xiv + 458. Leiden. WEILER, W. 1929. Die mittel- und obereocane Fischfauna Agyptens mit besonderer Beriick-

sichtigung der Teleostomi. Abh. bayer. Akad. Wiss., Munich, n.f. 1 : 1-57.

REVIEW OF CENTROPOMIDAE 81

WORTHINGTON, E. B. 1929. New species of fish from the Albert Nyanza and Lake Kioga.

Proc. zool. Soc. Lond. 1929 : 429-440. 1932. Scientific results of the Cambridge Expedition to the East African Lakes, 1930-1.

-2. Fishes other than Cichlidae. /. Linn. Soc. (Zool.), 38 : 121-134.

P. H. GREENWOOD, D.Sc.

Department of Zoology

BRITISH MUSEUM (NATURAL HISTORY)

CROMWELL ROAD

LONDON, SWy 5BD

A LIST OF SUPPLEMENTS TO THE ZOOLOGICAL SERIES

OF THE BULLETIN OF THE BRITISH MUSEUM (NATURAL HISTORY)

1. KAY, E. ALISON. Marine Molluscs in the Cuming Collection British Museum (Natural History) described by William Harper Pease. Pp. 96 ; 14 Plates. 1965. (Out of Print.)

2. WHITEHEAD, P. J. P. The Clupeoid Fishes described by Lacepede, Cuvier and Valenciennes. Pp. 180 ; n Plates, 15 Text-figures. 1967. £4.

3. TAYLOR, J. D., KENNEDY, W. J. & HALL, A. The Shell Structure and Mineralogy of the Bivalvia. Introduction. Nuculacea-Trigonacea. Pp. 125 ; 29 Plates, 77 Text-figures. 1969. £4'5o.

4. HAYNES, J. R. Cardigan Bay Recent Foraminifera (Cruises of the R.V. Antur) 1962-1964. Pp. 245 ; 33 Plates, 47 Text-figures. 1973. £10-80.

5. WHITEHEAD, P. J. P. The Clupeoid Fishes of the Guianas. Pp. 227; 72 Text-figures. 1973. £9-70.

6. GREENWOOD, P. H. The Cichlid Fishes of Lake Victoria, East Africa : the Biology and Evolution of a Species Flock. Pp. 134 ; i Plate, 77 Text-figures. 1974. £3-75. Hardback edition £6.

Printed in Great Britain by John Wright and Sons Ltd. at The Stonebridge Press, Bristol BS4 5NU

fSAPRf

> UHUURY

FOSSIL REPTILES FROM ALDABRA ATOLL, INDIAN OCEAN

E. N. ARNOLD

BULLETIN OF

THE BRITISH MUSEUM (NATURAL HISTORY) ZOOLOGY Vol. 29 No. 2

LONDON: 1976

15 APR

FOSSIL REPTILES FROM ALDABRA ATOLL, \^AJ INDIAN OCEAN

BY

EDWIN NICHOLAS ARNOLD

Pp. 83-116 ; 3 Text-figures, 4 Tables

BULLETIN OF

THE BRITISH MUSEUM (NATURAL HISTORY) ZOOLOGY Vol. 29 No. 2

LONDON: 1976

THE BULLETIN OF THE BRITISH MUSEUM

(NATURAL HISTORY), instituted in 1949, is issued in five series corresponding to the Scientific Departments of the Museum, and an Historical series.

Parts will appear at irregular intervals as they become ready. Volumes will contain about three or four hundred pages, and will not necessarily be completed within one calendar year.

In 1965 a separate supplementary series of longer papers was instituted, numbered serially for each Department.

This paper is Vol. 29 No. 2 of the Zoology series. The abbreviated titles of periodicals cited follow those of the World List of Scientific Periodicals.

World List abbreviation Bull. Br. Mus. nat. Hist. (Zool.).

ISSN 0007-1498

Trustees of the British Museum (Natural History), 1976

BRITISH MUSEUM (NATURAL HISTORY)

Issued 13 April, 1976 Price £2-25

FOSSIL REPTILES FROM ALDABRA ATOLL, INDIAN OCEAN

By E. N. ARNOLD

CONTENTS

Page

SYNOPSIS ........... 85

INTRODUCTION .......... 86

Family TESTUDINIDAE - LAND TORTOISES ...... 88

Geochelone ........... 88

Family CROCODYLIDAE - CROCODILES ....... 89

Crocodylus ........... 89

Family GEKKONIDAE - GECKOES ....... 90

Paroedura ........... 90

Geckolepis ........... 91

? Phelsuma ........... 94

Family IGUANIDAE - IGUANID LIZARDS ...... 95

Oplurus ........... 95

Family SCINCIDAE - SKINKS ........ 98

'Scelotes' ........... 98

Mabuya ........... 100

ORIGIN OF THE POINT HODOUL LIZARD DEPOSIT ..... 101

COMPARISON OF ALDABRA REPTILES WITH THOSE OF NEIGHBOURING AREAS 104

HABITAT REQUIREMENTS OF EXTINCT ALDABRA REPTILES . . . 106

POSSIBLE RESOURCE PARTITION BY EXTINCT ALDABRA REPTILES . . 107

BODY SIZE AND 'ISLAND GIGANTISM' ....... 109

POSSIBLE CAUSES OF EXTINCTION . . . . . . . in

COMPETITIVE EXCLUSION . . . . . . . . in

EXTERMINATION BY AN INVADING PREDATOR . . . . . 112

LOSS OF ECOLOGICAL RESOURCES . . . . . . . 113

COLONIZING ABILITY OF ALDABRA REPTILES . . . . . 113

ACKNOWLEDGEMENTS . . . . . . . . . 114

REFERENCES ........... 114

SYNOPSIS

The present reptiles of Aldabra comprise only the giant tortoise, Geochelone gigantea, two geckoes, Phelsuma abbotti and Hemidactylus mercatorius, and the skink, Cryptoblepharus boutonii, but a richer and quite different fossil fauna has recently been discovered. Giant tortoises occur in most terrestrial deposits on the atoll and remains of crocodiles and lizards have been found in two : the Bassin Cabri Calcarenites (undated, but considerably older than 125 ooo years B.P.) and cavity fillings in the Aldabra limestone at Point Hodoul formed since 100 ooo years B.P. The Bassin Cabri Calcarenites contain remains of a crocodile similar to Crocodylus niloticus and of an iguanid lizard of the genus Oplurus. These also occur in the Point Hodoul deposits together with five other kinds of lizards, two or three of which are geckoes (a Paroedura similar to P. sanctijohannis and P. stumpffi, a Geckolepis close to G. maculata and what is possibly a Phelsuma) and two skinks (a 'Scelotes' similar to 'S.' johannae and a Mabuya very like M. maculi- labris). The Point Hodoul lizard remains may be the food residue of a predator, perhaps an owl. The reptile fauna of Aldabra is much more similar to that of Madagascar and the Comores

6*

86 E. N. ARNOLD

Islands than to that of East Africa or of the Seychelles. Composition of the Point Hodoul fauna and the presumed requirements of some of its members suggest that conditions on Aldabra at that time may have been rather similar to those now occurring on the Comores. It is likely that the ecological requirements of the fossil forms were sufficiently different for them all to be able to coexist. Some of the fossils clearly differ in size from their closest modern relatives, both the Geckolepis and the Opiums being very large ; possible reasons for this are discussed. Aldabra was completely submerged after the laying down of the Bassin Cabri Calcarenites, but not since the Point Hodoul deposits were formed. Extinction of the reptiles found in the latter may have been largely caused by transient or permanent loss of ecological resources, although competition from the species existing on the atoll today could have been a minor factor. The possible importance of invading predators is difficult to assess. All the reptiles known from Aldabra seem to have been well adapted to the problems of transmarine colonization. There is evidence that the giant tortoises reached the island three times and the crocodile and Oplurus at least twice.

INTRODUCTION

THE study of island faunas has had a long history and is again fashionable, one of the most influential events in bringing this about being the appearance of The theory of island biogeography by MacArthur and Wilson in 1967. Among the topics that have recently received attention is the problem of what factors limit the number of species on islands and the importance of extinction rates in this process. A restriction on such investigations is the paucity of direct evidence of natural faunal change on islands. This is especially true in the case of reptiles. Instances are known where late Pleistocene or more recent island reptiles have become extinct (see, for instance, Etheridge, 1964, 1965, 1966 for the West Indies, Bravo, 1953 for the Canary Islands and Vinson & Vinson, 1969 for the Mascarenes), but often only part of the previous fauna has disappeared and there is frequently circumstantial evidence that human influences have been important in bringing such changes about. Situations where there has been an extensive faunal turnover in probably more natural conditions would consequently be of interest. Aldabra appears to be a case in point.

Aldabra is situated about 640 km east of the African mainland, about 380 km northeast of the Comores Islands and some 420 km northwest of Madagascar (see Fig. i). It is a low atoll, being only about 10 m above sea level at its highest point, and is some 34 km long by 14-5 km wide. There is a large central lagoon and the total land area is 155 km2. Much of the present surface consists of coral limestone, which is often covered by scrub. There are few natural large trees and the lagoon is fringed by mangrove. The present reptile fauna consists of giant tortoises (Geochelone gigantea Schweigger, 1812), two geckoes (Hemidactylus mercatorius Gray, 1842 and Phelsuma dbbotti Stejneger, 1893) and the skink, Cryptoblepharus boutonii (Desjardins, 1831). None of these species is confined to the atoll and all of them have relatively wide distributions in the West Indian Ocean.

The island has been the subject of considerable scientific research by a series of expeditions organized by the Royal Society since 1966. Some of the results of this work have been published and a number of papers on the atoll form volume 260 of Philosophical Transactions of the Royal Society of London, series B (1971). Braith- waite, Taylor & Kennedy (1973) have given an account of the depositional and

FOSSIL REPTILES FROM ALDABRA

87

SEYCHELLES . '•

AMIRANTES -.-

AFRICA

ALDABRA

COMORES

MADAGASCAR

FIG. I. West Indian Ocean, showing position of Aldabra and other nearby islands mentioned in text.

erosional history of Aldabra. During investigations on the island, Dr J. D. Taylor of the British Museum (Natural History) collected substantial amounts of fossil material including the reptiles that form the basis of this paper. Large numbers of tortoise remains were also observed in situ but could not be extracted for detailed examination.

Tortoises occur in many of the terrestrial deposits of Aldabra (Braithwaite et aL, 1973). Lizards and crocodiles have also been found, but only in two deposits, the Bassin Cabri Calcarenites and cavity fillings in the Aldabra Limestone at Point Hodoul (see Fig. 2). The Bassin Cabri Calcarenites are of unknown age, but they are older than the Aldabra Limestone which has yielded dates by the 23<rr;h/234u method of 118000 to 136000 + 9000 years B.P. (Thompson & Walton, 1972). The Point Hodoul deposits were probably formed since 100 ooo years B.P. (Braithwaite et aL, 1973). Most of the reptile fossils consist of dissociated bones, few of which are intact. However, although many are broken, the majority are not obviously eroded and are therefore identifiable. This is especially true of the lizard bones, over 1000 of which have been determined.

88

E. N. ARNOLD

FIG. 2. Aldabra Atoll, showing the two main localities where fossil reptiles were collected.

Because of the relatively recent age of the deposits, it was assumed as a working hypothesis that the species represented would probably have close relatives amongst the extant reptiles of the West Indian Ocean, or of the Ethiopian region as a whole. Initial identification was consequently attempted by comparison with the present fauna of that area and it proved possible to find good matches for nearly all the material by this means. Two difficulties in assessing the fossils were the lack of really adequate comparative material in museum collections and the need for taxonomic revision of some of the groups concerned (for instance, the scincine lizards of Madagascar and nearby islands -see Greer, igyob). These constraints have limited the preciseness of identification in some cases.

As nearly all the fossils are generally like known modern forms, they are not described in detail below. Instead, comment is restricted to diagnostic characters and to any features in which the fossils differ from their modern counterparts. In most cases, the names used for lizard skull elements follow Oelrich (1956). Because many of the fossils are broken, numbers of particular bones, given in the lists of material examined, tend to be minima.

Order TESTUDINES Chelonians

Family TESTUDINIDAE

Land tortoises

GEOCHELONE Fitzinger, 1835

MATERIAL REFERRED. Point Hodoul cavity fillings. Registered number : R8762. Numerous fragments, especially eroded limb bones.

Tortoise remains were observed in nearly all the terrestrial deposits on Aldabra.

IDENTIFICATION. This material has not been examined in detail, but it resembles Geochelone gigantea (Schweigger, 1812), the giant tortoise that occurs on Aldabra at the present time.

FOSSIL REPTILES FROM ALDABRA 89

Order CROCODYLIA Crocodilians

Family CROCODYLIDAE

Crocodiles

Crocodylus Laurenti, 1768

MATERIAL REFERRED. Bassin Cdbri Calcarenites. Registered number : R8885. Tooth : i.

Point Hodoul cavity fillings. Registered number : R8763~98. Premaxillae : right - 2, left - i. Maxilla : i (fragment). Jugals : right - 5, left - 6. Frontals : 3. Frontal + parietal : i. Parietal: i. Parietal + supraoccipital : i. Squamo- sals : right - 4, left - i. Pterygoids : right - i, left - 1. Ectopterygoids : right - i, left -2. Dentary fragments : n. Isolated teeth : 6. Vertebrae: 2 (fragments). Caudal chevrons : 2. Osteodermal scutes : 45.

IDENTIFICATION. The Point Hodoul remains include premaxillae. These show a clear lateral notch that in life would have received the enlarged fourth mandibular tooth. The bones also suggest that the animals from which they came had snouts that were neither conspicuously narrowed nor laterally expanded. Amongst recent crocodilians, this combination of features is restricted to Crocodylus. Comparison of the two extant African species of the genus shows that the fossils are very similar to modern C. niloticus Laurenti, 1768, although some elements, e.g. the jugals, appear more robust than those of equivalent-sized mainland African animals with which they were compared. On distributional grounds, C. niloticus is the most likely modern crocodile to have reached Aldabra, for it is the only species in eastern Africa. It is also known to have been present in the Seychelles and exists on the Comores and on Madagascar.

ESTIMATED NUMBER OF INDIVIDUALS IN SAMPLE. On the basis of the commonest element, viz. the jugal, there must have been at least six individuals represented in the Point Hodoul sample.

ESTIMATED BODY SIZE. Comparison of the fragments with intact skulls of modern African specimens suggests that the largest came from crania about 310 to 320 mm in length (snout-tip to medial posterior border of the supraoccipital bone) . Bellairs (1969) has found that the ratio of skull length to total length is about i : 7-5 in C. niloticus and that this holds over a wide variety of sizes. On this basis, the largest Aldabra fragments would have belonged to an animal 2-33 to 2-40 m long. This is much smaller than the known maximum for mainland African populations, although obviously the small sample available cannot exclude the possibility that the Aldabra animals grew bigger. However, the robustness of some of the bones could indicate maturity and if this is so, the atoll population may have been characterized by small body size. This would not be unexpected, especially as 'dwarf populations of this species are known from some areas of East Africa (Cott, 1961).

9o E. N. ARNOLD

Suborder SAURIA Lizards

Family GEKKONIDAE Geckoes

Paroedura Giinther, 1879

MATERIAL REFERRED. Point Hodoul cavity filling. Registered number : R884Q-77. Premaxillae : 18. Maxillae: right - 25 (one complete), left -21. Conjoined nasals - 4. Prefrontals : right - 18, left - 16. Frontals : 28. Postfrontals : right - 23, left -15. Parietals : right - n, left - 17. Pterygoids : right - n, left - 13. Quadrates : right - 21, left - 19. Dentaries : right - 21, left - 18. Coronoids : right - 16, left -12. Proximal jaws : right - 14, left - 14. Axis: i. Cervical vertebrae : 10. Dorsal vertebrae : 86 approx. Sacrum : i. Basal caudal vertebrae : 22. Autotomic caudal vertebra : i. Scapulocoracoids (incom- plete) : right - i, left - 3. Scapulae : left - i. Pelvic girdles (incomplete) : right - 9, left - 7. Humeri : more or less intact, right - i, left - i ; proximal sections, right - 13, left - 6 ; distal sections, right - 12, left - 6. Femora : proximal sections, right - 18, left - 21 ; distal sections, right - 7, left - 7. Tibiae : right - 6, left -3.

IDENTIFICATION. Certain elements of this material are clearly of gecko origin. The frontals, which are undivided, have lateral downgrowths that meet and fuse ventrally to form a cylinder and the vertebrae are amphicoelous. Amongst lizards, both these features are confined to the Gekkonidae, although not universal in them.

Other striking characteristics of these fossils include the following : heavy ossi- fication of the dorsal skull bones ; presence of a well developed, 'sculptured' orna- mentation of their outer surfaces ; conjoined nasal bones and prominent notching of the anterior lateral edges of the parietals to take the posterior section of each postfrontal. Within the geckoes, this combination of characters seems to be limited to nine nominal species now confined to Madagascar and the Comores Islands. These have usually been included in Phyllodactylus Gray, 1828, but Dixon & Kroll (1974) have recently transferred them to a separate genus, Paroedura. As the species concerned do appear to form a natural assemblage, this course is followed here, although the authors' implication that Paroedura may have had a separate origin from Phyllodactylus, as they understand it, and that the two groups evolved similar foot structure independently is unproven. It seems just as likely that Paroedura is a derivative of the more typical members of Phyllodactylus.

Paroedura is a quite tightly knit group of species, the interrelationships of which have not yet been fully worked out. This makes it difficult to decide with certainty which of the modern forms are most closely related to the fossil Aldabra material, but the latter has most superficial likeness to P. sanctijohannis Giinther, 1879 of the Comores Islands and P. stumpffi (Boettger, 1878) of North Madagascar and Nossi Be. It resembles the former closely in the degree and pattern of sculpturing on the dorsal skull bones, in parietal shape (distinctly curved transverse section that is convex above and a distinctly sinuous posterior border) and in having very broad quadrate

FOSSIL REPTILES FROM ALDABRA 91

bones. However, the f rentals are less obviously concave above than in the modern specimens of P. sanctijohannis examined (from Anjouan and Grande Comore) and in this respect are more like P. stumpffi, although the fossils resemble this species less in details of parietal shape and sculpturing. P. gracilis (Boulenger, 1896) is also rather similar, but its parietal has a straighter posterior border and its frontal is more deeply concave above. P. picta (Peters, 1854) has a much narrower quadrate and rather coarser 'sculpturing'. The 'sculpturing' in P. androyensis (Grandidier, 1867) also seems to be coarser than in the fossils and the parietal is more rectilinear with a flatter transverse section. P. bastardi (Mocquard, 1900) shares the latter feature and its 'sculpturing' is coarser still ; also the crest on the proximal antero ventral face of the humerus, for the pectoralis muscle, is set almost at right angles to the ventral face of the bone and its anterior face is slightly concave with a well developed anterodorsal ridge forming its upper border. In P. oviceps (Boettger, 1881) the parietal is flatter than in the Aldabra material with a less sinuous posterior border and its surface sculpturing is weaker.

Of the two remaining nominal species of Paroedura, the unique type of P. homalo- rhina (Angel, 1936) no longer exists (J. Guibe, personal communication) and P. guibae Dixon & Kroll, 1974 has not been examined by me.

ESTIMATED NUMBER OF INDIVIDUALS IN SAMPLE. There are 28 recognizable frontal bones of Paroedura, so at least this number of individuals must be represented in the sample.

ESTIMATED BODY SIZE. On the basis of comparison of limb bones and frontals, it seems that the Aldabra Paroedura grew to about 60 mm from snout to vent, or a little more. This compares with a maximum length of 70 mm for modern P. and 67 mm for P. sanctijohannis (see Angel, 1942).

Geckolepis Grandidier, 1867

MATERIAL REFERRED. Point Hodoul cavity filling. Registered number : R8828-48. Premaxillae : 3. Maxilla : left - i. Pref rentals : left - 3. Frontals : left - i (anterior section only). Parietals : right - 2 fragments (anterior lateral section, central lateral section), left - 3 fragments (2 central lateral sections, posterior lateral section). Palatine : left - i. Pterygoids : right - i fragment, left - 4 fragments. Quadrates : left - 2. Dentaries : right - i (anterior section only), left - i (anterior section absent). Coronoids : right - i, left - i. Vertebral centra : 5. Sacrum : i. Scapulocoracoids : right - 2, left - 2. Pelvic girdles : right - 2, left - i. Humeri : more or less intact, right - i ; shafts, right - 4, left - 2 ; proximal section, left - i ; distal section, left - 2. Femora : proximal section, left - 2 ; distal section, right - 2, left - 2.

IDENTIFICATION. These fragments come from a gecko considerably larger than the Aldabra Paroedura discussed above. Initial tentative allocation to the Gek- konidae was based on a number of suggestive, but not fully diagnostic features, for instance the dentary is lightly built with a closed Meckel's groove and a very rounded

92 E. N. ARNOLD

lingual face. But, because the remains are so fragmentary, definite family assign- ment really depends on their detailed resemblance to a particular gecko species. The more distinctive features of the material are as follows :

1. The largest premaxilla has an estimated 13 teeth. The nasal process is rather broad and is slightly constricted at the level where it would pass between the external nasal openings of the skull, but it expands again above this. In the most complete premaxilla, the process then rapidly tapers to an obtuse point.

2. The maxilla has a broad palatal shelf and the nasal process (i.e. the dorsally directed lamina that forms part of the side of the snout) rises smoothly from the tooth-bearing body of the bone with no pronounced inward 'step' just anterior to the orbit.

3. The anterior lateral fragment of the parietal is peculiar in that the angle between the fronto-parietal suture and the anterior section of the lateral margin of the bone is more than a right angle, whereas in most geckoes it is less (see Fig. 3).

4. From the fragments available, it is apparent that the upper surface of each parietal is distinctly convex above the descending flange of this bone, but lateral to this it is slightly concave. The convexity extends as a poorly defined ridge that curves inwards and backwards towards the posterior midline of the skull. From

FIG. 3. Dorsal views of the left posterior areas of two gecko skulls, showing difference in shape of the anterolateral region of the parietal (arrowed). A. Geckolepis maculata (based on BM 91.6.30.1). B. Homopholis fasciata erlangeri (based on BM 1931.7.20.255).

FOSSIL REPTILES FROM ALDABRA 93

this ridge, an even less prominent one runs along the proximal upper surface of the supratemporal process of the parietal.

5. The outer surfaces of all the fragments of the superficial bones of the skull are either smooth and unornamented, or they have a weak pattern of shallow, thin grooves.

6. The quadrates have a very slender posterior crest. This is quite strongly and evenly curved (more so than in Aldabra Paroedura).

7. The anterior upper border of the ilium does not rise very abruptly, or recurve forwards, as it does in many geckoes.

8. The humerus is more robust than in the majority of geckoes. The crest on the proximal anteroventral surface of this bone is set almost at right angles to the ventral surface and its anterior exposure is strongly concave.

Examination of a wide range of Ethiopian and West Indian Ocean gekkonid lizards shows that this combination of features is approached closely only in the genus Geckolepis, which is now restricted to Madagascar and the Comores. The apparently related genus, Homopholis Boulenger, 1885, differs in having a distinct 'step' in the outer surface of the maxilla, just anterior to the orbit. Also, the fronto-parietal suture of Homopholis forms an acute angle with the anterior lateral edge of the parietal bone and there is no concavity on the forward surface of the proximal anteroventral crest of the humerus (all features confirmed on specimens of H. walbergii (Smith, 1849) and on H.fasciata erlangeri Steindachner, 1906 ; presence of a maxillary 'step' and the shape of the parietal were also checked on the type of H. heterolepis Boulenger, 1896). The more distantly related, widespread West Indian Ocean genus Phelsuma Gray, 1825 also differs in parietal shape. Furthermore, the premaxilla usually has a slender nasal process, the palatine shelf of the maxilla is typically rather narrow, the quadrate is usually more robust and the anterior upper border of the ilium rises abruptly or is recurved.

Within Geckolepis, the Aldabra material was compared with the largest extant species, G. maculata Peters, 1880 of Madagascar and the Comores, and with three of the four remaining species, all of which are restricted to Madagascar, viz. G. anomala Mocquard, 1902, G. polylepis Boettger, 1893 and G. typica Grandidier, 1867. Of these, it resembles G. maculata most closely. However, the fossils exhibit some mainly minor differences from the small sample of G. maculata available for com- parison (n = 3) ; these are listed below :

1. The Aldabra fragments are distinctly larger than corresponding elements in modern G. maculata.

2. The nasal process of the premaxilla is slightly more constricted at the level where it would pass between the external nasal openings of the skull.

3. In modern material examined, the inner surface of the dentary swings slightly, but clearly, upwards from the base of the tooth-row before curving downwards to form the lingual surface of the bone. In the fossil dentary sections, the upward swing is less marked.

94 E. N. ARNOLD

4. The anterior upper edge of the ilium curves upwards rather less than in compared modern material.

5. The crest on the proximal antero ventral surface of the humerus is better de- veloped and its anterior exposure is more strongly concave. Also, the bone is more precisely moulded.

6. This is true of the proximal section of the femur as well. In addition, the femoral head has a more pronounced neck which has better developed ridges on its posterior and ventral surfaces.

It is possible that some or all of differences 2-6 are merely allometric concomitants of larger size. If they were due to allometric change, it might be expected that trends towards the conditions found in the Aldabra material would be apparent in a series of modern Geckolepis of increasing size, although absence of such trends could merely mean that the changes would only begin to be apparent at sizes larger than those attained today. In fact, only the expected changes in the humerus can be detected in the small series of G. maculata and G, anomala available. Another less direct indication that allometric growth might be responsible for some of the charac- ters of the fossils would be the presence of similar differential growth processes in a related form that grows as large as the Aldabra animals. When Homopholis walbergii of assorted sizes were examined, changes in the femur similar to those expected in Geckolepis were found.

Although the features of the humerus and femur that distinguish the fossil Gecko- lepis of Aldabra from modern G. maculata may be due to allometric change, there is no evidence that the other characteristics of this material (i.e. differences 2-4) are due to this factor. However, they are relatively trivial and on their own, or even in conjunction with greater body size, do not seem to provide adequate grounds for recognizing the Aldabra material as a distinct form, especially as the sample of modern G. maculata available is so small and the Aldabra remains are so fragmentary.

ESTIMATED NUMBER OF INDIVIDUALS IN SAMPLE. There are five right humeri, the commonest element present. But there is also a distal section of a left humerus that is smaller than any of the right ones, so at least six individuals must be represented.

ESTIMATED BODY SIZE. Calculations based on the sizes of fossil humeri and femora, compared with those of modern animals, suggest that the largest fossil Geckolepis were 90 to 100 mm from snout to vent. This contrasts with a present apparent maximum of 70 mm (Angel, 1942). .

? Phelsuma Gray, 1825

MATERIAL REFERRED. Point Hodoul cavity filling. Registered number : R8878. Maxilla : left - 2 fragments (anterior section, section from just below lachrymal area). Dentaries : right - I fragment, left - i fragment (both from middle of bone).

IDENTIFICATION. The dentary fragments are lightly built and pleurodont with closed Meckel's grooves. The cylindrical body of the bone shows little dorsoventral

FOSSIL REPTILES FROM ALDABRA 95

taper along its length, its lower edge is almost straight and the lingual face is strongly rounded. The labial wall extending upwards from the dentary cylinder is very high compared with the depth of the latter and the labial face of the bone has a strongly curved transverse section. On this face, the mental foramina are situated considerably nearer the upper than the lower border of the dentary. The maxillary fragments indicate that the palatine shelf is narrow and concave beneath, at least in the area ventral to the lachrymal bone.

These fossils are not like the bones of the other species represented in the Point Hodoul deposit. The general facies of the dentary (light build, closed Meckel's groove, rounded lingual surface and straight lower border) suggests gecko origin. Certainly, it does not resemble the dentary of any of the lygosomine skinks found in the area that were examined, although these also have closed Meckel's grooves. Among those checked was Cryptoblepharus boutonii (Desjardins, 1831), which occurs on Aldabra today. In this species, the dentary is considerably narrower latero- medially and the lingual face is more flattened than in the fossils.

Although the bones could not be precisely matched with any of the many West Indian Ocean and Ethiopian geckoes compared with them, they do have a general resemblance to the smaller species of Phelsuma. However, none of the members of this genus examined* have the mental foramina placed so close to the upper margin of the dentary. This feature and the very restricted taper of the cylindrical body of the bone distinguish the fossils from P. abbotti, which is now present on Aldabra.

ESTIMATED NUMBER OF INDIVIDUALS IN SAMPLE. Possibly only one individual is represented.

ESTIMATED BODY SIZE. The very fragmented nature of the material makes this difficult to assess, but the remains probably come from a lizard not more than 50 mm from snout to vent.

Family IGUANIDAE Iguanid lizards

Opiums Cuvier, 1829 Bassin Cdbri Calcarenites

MATERIAL REFERRED. Registered number : 1^8883-84. Maxilla : right - 3 in- complete sections (complete dental arcade, but nasal and palatine processes absent ; section lacking most of