5 4 " =~ 73 “a Fs % . r ted ~ ¥ 7 4 - ' : & % * : y As . . 7 : ~~ Pe! 4 Ww = _% $f - ‘ » [ : - Shs i Sr ae ae t ee taged, oe t ts Sear ee a a a a Se S353 23th as SAA Ee SARS aa oye iO Sd oe oe PV a OU P. H. GREENWOOD . —Ss«éBULLETIN OF e BRITISH MUSEUM (NATURAL HISTORY) IG ee ae oo Supplement 6 e. = LONDON: 10974 | _ THE CICHLID FISHES OF LAKE VICTORIA, EAST AFRICA : THE BIOLOGY AND _ EVOLUTION OF A SPECIES FLOCK BY PETER HUMPHRY GREENWOOD Pp 1-134; 1 Plate, 77 Text-figures — ee BULLETIN OF = THE BRITISH MUSEUM (NATURAL HISTORY) «3=©. ZOOLOGY ee Supplement 6 — LONDON: 1974 | THE BULLETIN OF-THE BRITISH MUSEUM (NATURAL) HISTORY), tmsituied.” tt XO40, 4758 | issued tn five series corresponding to the Departments of the Museum, and an Historical sentes. | 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 Supplement 6 of the Zoological 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.) Suppl. © Trustees of the British Museum (Natural History), 1974 CRUSTERS OF ( | | THE BRITISH MUSEUM (NATURAL HISTORY) | / Issued 11 April, 1974 Price £3.75 ‘ , 9 { SE CICHLID FISHES: OF LAKE VICTORIA, EAST AFRICA : THE BIOLOGY AND EVOLUTION. OF. A’ SPECIES’ FLOCK By PETER HUMPHRY GREENWOOD CONTENTS Page INTRODUCTION . P : : ‘ ‘ : : 3 LAKE VICTORIA, ITS PHYSIOGRAPHY AND HISTORY. : ; : 7, THE FISHES OF LAKE VICTORIA . ‘ : : : : : : II The non-cichlid fishes. : : : : : ; : : II The cichlid fishes . : . : ; ; ‘ , 13 THE HAPLOCHROMIS SPECIES FLOCK : : i : 2 oes 19 Introduction. : : ; ‘ Z ; ; : : ; 19 Basic morphology . : 5 : : ; Ze Feeding habits of the H Diag eine species ‘ ‘ ; : 30 Intralacustrine distribution of the Haplochromis species : : , 44 Breeding biology of the Haplochromis species . 50 INTERRELATIONSHIPS OF THE LAKE VICTORIA HAPLOCHROMIS SPECIES 56 The insectivorous species : : . ‘ ; : : : 58 Phytophagous species . ; : : ; ‘ , 63 Species feeding on benthic One eee : : é : ‘ 67 Scale eating species : , : : : : ; ; : 69 Mollusc eating species. . , : : : 69 Predators on larval and Sao dichia Aches : : : : : 75 Piscivorous predators. ‘ ‘ : : : : 80 Conclusion . . ; . : : : : 93 THE RELATIONSHIPS OF THE MONOTYPIC GENERA . ; , ; 99 THE ANATOMICAL BASIS FOR THE ADAPTIVE RADIATION . : : 103 SPECIATION ; : , ; ; d ; : : TII CONCLUSION . 3 : ‘ : 119 ACKNOWLEDGEMENTS : : : : : : : . : 124 REFERENCES . : ‘ ‘ : : ‘ : . : : T25 INDEX >. : : : ; : : : 3 : : : 12 INTRODUCTION Tue freshwater fishes of the Great Lakes of Africa have long fascinated and perplexed students of evolution. As far back as 1913, Plate suggested that the cichlid fishes in Lake Tanganyika might show a mode of speciation different from that of the usual kind. This view, broadened to include fishes from other lakes, was supported by many subsequent workers, with Rensch (1933) a notable exception, and is still _ echoed today (Trewavas, Green & Corbet, 1972). The problem has, if anything, increased in its complexity (and interest) since _Plate’s time, despite our greatly increased knowledge of these fishes. The publica- tion recently of a book some 500 pages long (Fryer & Iles, 1972) summarizing re- searches into the biology and evolution of but one family of African fishes (the Cichlidae) is a good indicator of the problem’s size. PH. GREENWOOD Attention has always centred on species of the family Cichlidae, perch-like fishes widely distributed in tropical America and Africa, but also occurring in India, Shri Lanka (Ceylon) and Malagasi. This focus on African cichlids has, to a certain extent, distracted attention from other African freshwater fishes, and even from the cichlids of South America (see Lowe-McConnell, 1969), a sad but understandable state of affairs. Each of the larger African lakes is characterized by a strongly endemic cichlid species complex (Table I). In most lakes these species dominate the fish fauna, both in an ecological and a taxonomic sense. Ecologically, the lake cichlids generally show a wider spectrum of adaptive radiation than do all the other families combined. The bare figures of Table I show clearly the high degree of endemism, the extensive speciation undergone by these fishes and also the preponderance of cichlid taxa over those of all other families. Some measure of the morphological differentiation within the various species flocks may be taken from the number of endemic genera recog- nized in each lake. | Much of the literature dealing with African lake fishes has been concerned with these various phenomena, but in particular with the mode or modes of speciation involved, and the reasons for the disparate levels of taxonomic diversity found in different lakes (see Fryer & Iles, 1972). Far less attention has been given to the question of why the Cichlidae more than any other family display this evolutionary potential, a point to which I shall return later (p. 103). One peculiarity of the African Cichlidae which Table I does not bring out is the contrast in diversity and numbers between lake and river dwelling species. The situation in most rivers is quite the reverse of that in the lakes, with species from other families providing the dominant faunal elements. A comparison between Lake Malawi and the Congo river illustrates this point very clearly, especially since both ecosystems have the most speciose fish faunas known from Africa. In Lake Malawi there are ca 200 cichlid species comprising some 78 per cent of the entire fish fauna (ca 242 species), but in the Congo only 7 per cent of the 410 known species are cichlids (Lowe-McConnell, 1969, where comparisons between other African lake and river faunas are clearly demonstrated in text-fig. 2, and comparisons are also made between the African and South American faunas). The contrast between cichlid and non-cichlid species is even greater if one compares Lake Victoria (with ca 170 species of Cichlidae) to any of its affluent rivers. In these, only three cichlid species have been recorded at any distance from the lake itself. Obviously this peculiarity is closely connected with the question of what factors influence and stimulate speciation within the lakes. It is also, of course, associated with the greater number of ecological niches provided by a lacustrine environment. (Niche used here, and elsewhere, in the sense of Elton [1928] rather than that of Weatherley [1963].) These questions will not be considered further and are only mentioned because of their relevance to that of the ancestral species which must have populated the embryo lakes from the preexisting river systems. To return to the lake cichlids. When the evolutionary radiation summarized in Table 11 is seen against a time scale for lake histories, its speed is outstanding. Ee | | | | | | | 4 EVOLUTION OF A CICHLID SPECIES FLOCK ‘eSOIy oye] Surpnpoxy | ‘(pojonb somsy oy} o8ueyo you prnom sorpoq-10}eM Oso} JO UOISNOUT SY} YSnoyy]e) IN eIIOJIA OY} pue eon oYeT BUIPNOX7 » PIEMPH "FLOPNAY ‘CLIOPOIA SOYL'T IO} soyeurjsa { sooIMOs sNOLIeA WOIT POATIOp eye ‘poroystsor ore ( OYE] & WOIF poplooar AyTenzoe sotoads ATU ‘(uorT}eoruNIMOD Teuosied) sap “q pure (zZ61) So] WY JoAI Woy Imepey, oye] IOy osoyy ‘(ad1095) oye] 10g e€L61 ‘poomussiy puv) suoryearosqo Teuosied uo poseq o81005) pue SIOALI JUON Ye $41 SuUIpNyoxo ‘9'1) ‘soye] UvoLIpY (our suo pur) sofetw xis ut ‘soreds pue v1ioud$ ‘SOI[IWIey PITYOr “UOu FO ToquuNU [e}0} oy} pure ‘sotseds sawuosys0jdv FY JO Joquunu oy} YAM 19y3080} ‘e1ouES pue soroeds pr[tyo1o Jo toquinu [e703 oT, O II L 6z Oo on O (ACE fe) IZ O OI I oz oro puyy [eq0T, e1oues JO IoquinN O 14 ce LE Oz Zz oO € O cd O b Vv 8 owe pur [e}OL e1oues JO Ioquinyy fe) al 6 LY L9 €r gz zb éz7+Q ¢ ze VI e 9f €1 Zz Lt. £ OI ge IT orlulapuy [e}0OL soroods Jo Iaquinyy SOI} JO Toquiny SHITINVA AHHLO ¢ 9 ¢ OI Zz (é QzI QzI I 4nq ITV ozI < F 4nq ITV O07 V9 Zz Zz TZ L 4 9 b OI r 4nq [IV PF 4nq TV S3nq qty | or-SE na I 4nq [IV 8 nq TV € 3nq TIV o£1-OS1 v9 ‘dds samosysoidv YY onuopuy “dds samosysoidv yy OIuUepuy [e}O], sorseds jo 1oquinNy AVAITHOID I a1dvV] oqesnqeN eyIAUeSUe T, IMPETeI FTOPNY HOqTV 98.1005)—pIeMp7 { e110791 A, oye] oqesnqen eytAuesury IMP] JOP NS VOT TV 981005)—pIeMpy 4% CITOPOTA ONe'T 6 Poo GREENWOOD of the lake basins is still far from settled, but current estimates suggest that the oldest (Lake Tanganyika) is not more than 10 million years old, that Lake Malawi is about 2 million years, and that Lake Victoria originated during the mid-Pleistocene, about 750 000 years B.P. (see summaries in Fryer & Iles, 1972 ; and Temple, 1969, for Lake Victoria). Lake age must undoubtedly be one of the factors affecting the degree of differentiation within a flock, as must the time and completeness of a lake’s isolation from other major water bodies, be they lake or river systems. Thus the specifically depauperate and ecologically impoverished species flocks of Lakes Rudolf and Albert are a measure of age and lack of isolation. Both lakes are moderately young (early Pleistocene) and both have, or have had until recently, free contact with the Nile. Lake Victoria, the lacustrine background to this essay, is younger, but it has been a closed drainage basin for most of its existence (Doornkamp & Temple, 1966 ; Temple, 1969 ; Greenwood, 1973a). Its cichlid flock is adaptively multiradiate, is highly speciose, but 1s composed of species which, with seven exceptions (two species of Tilapia, and five monotypic genera) are all members of a single genus, Haflo- chromis (but see p. 99). The five monotypic genera are all derived from and closely related to the genus Haplochromis (Greenwood, 1956a, 1959a; alsop.gg below). In this respect its species flock can be considered a relatively simple one when compared with those of Lakes Tanganyika and Malawi (see Table I), as it can also be from the viewpoint of morphological divergence among its Haplochromis species. The situation in Lake Victoria can probably be taken as a real model of a stage through which the flocks of Tanganyika and Malawi have already passed. Its importance in this context is enhanced by the fact that the components of the flock are already well-advanced along different paths of ecological specialization. The comparison with Lake Tanganyika is perhaps less complete because there the species seem to be derived from two major phyletic components of the African Cichlidae (the so-called ‘Tilapia’ and ‘Haplochromis’ lineages; see Regan, 1920 ; also Fryer & Iles, 1972). In Lakes Malawi and Victoria by contrast, the flocks, except for a small group of Tilapia species, are members of the “Haplochromis’ lineage. As was noted earlier, most attention has been paid to the problems of speciation within the lakes, and to purely taxonomic and ecological studies of the fishes. Little has been written about the phylogeny of a particular flock (Fryer’s [1959] analysis of the ‘Mbuna’ generic complex in Lake Malawi excepted), and virtually nothing has been said about the evolution of adaptive characters within a flock. By the latter I mean particularly the anatomical basis for the adaptive trends observed within a phylogenetic framework. I believe that the Haplochromis species flock of Lake Victoria provides very suit- able material for such an analysis, the more especially since it can be linked with a fairly certain physiographical background to speciation within the developing lake basin. I also believe that from such an analysis, it is possible to throw light on one aspect of the question of why cichlids are able to undergo rapid and ecologically successful adaptive radiations when other families remain, by comparison, evolu- tionarily inert. EVOLUTION.OF -A. CICHLID: SPECIES ‘FLOCK i PEAKE VIi€TORIA, ITS-PHYSIOGRAPHY. AND HISTORY Lake Victoria is a large, and by comparison with other African lakes, shallow water body lying across the equator (0°21’N — 3°0’S, 31°39’ — 34°53’E). Its surface area is approximately 69 000 km?, its greatest length and breadth ca 400 and 320 km respectively. The coastline is extremely irregular, and totals some 3300 km in length. Especially in the northern half there are a number of large islands (Text-fig 1) whose shorelines are as varied as those of the mainland. Broadly speaking, the habitat types provided by the shore comprise deeply in- dented, shallow and protected bays, sandy exposed beaches and, occasionally, rocky cliffs or broken rock exposures. [ringing papyrus swamps are common around much of the shore, and many of the bays terminate in broad swamps extending over several square kilometres. | Much of the lake is less than 20 m deep ; the deepest zone (60-90 m) lies somewhat eccentrically towards the eastern shore and occupies a kidney-shaped area (see Graham, 1929). The bottom profile in the deeper waters is not entirely flat, oc- casional ‘hills’ rise well above the general level of the lake floor. Much of the bottom in these deep areas is covered by a thick deposit of organic mud with, here and there, isolated patches of hard substrate (sand, shingle or rock). Organic mud substrata occur inshore as well, especially in sheltered bays, but also along the open coastline in protected regions. In most places where there is a sandy beach, the sand sub- strate grades imperceptibly into mud some few hundred metres from the shore ; rather rarely does the sand extend for more than two or three kilometres offshore and into water over 30m deep. Substrate type appears to be an important factor in limiting the distribution of many species (see p. 46). There is little annual variation in water temperature, the mean surface tempera- ture being ca 24°C (that of deeper water about a degree lower). An annual cycle of thermal stratification (Fish, 1957; Talling, 1963) leads to a marked reduction of dissolved oxygen in deeper parts of the lake. The effects of this relative deoxygena- tion on the biology of fishes living in affected areas has not been satisfactorily in- vestigated. At one time (Greenwood, 1965a) it was thought that no cichlids, and few other species, inhabited depths below about 30m. However, later researches showed that many species of Haplochromis live in water at this depth and deeper. Indeed, there is probably no offshore area of the lake without its populations of _Haplochromis species. What we have yet to discover are the reactions of these fishes during periods of deoxygenation, whether they can temporarily adapt their respiratory requirements to the new environment or whether there is some migration into shallower and better oxygenated zones. In this respect Lake Victoria contrasts strongly with Lakes Malawi and Tangan- yika. In these deep (704 and 1470 m) and trough-like lakes the lower water layers are permanently stratified (Beauchamp, 1964) and no fishes live there (Text-fig. 2). The greatest depth at which fishes have been recorded in Lake Tanganyika, for example, is about 200 m, and in Lake Malawi ca1toom. But these deep-living species are exceptional, and most species (particularly Cichlidae) are restricted to the upper 30 or 40 m),. Pp... GREENWOOD NAPOLEON GULF | Kisumu | GE [Kes Yo NABUGABO SPEKE GULF Mwanza SMITH SOUnr, | | Fic. 1. Sketch map of Lake Victoria. The approximate positions of the 20, 40 and 50 m isobaths are indicated by dotted lines. | | EVOLUTION -OF A CICHLID SPECIES FLOCK 9 Detailed analyses of the chemical composition of the water are given by Tailing & Talling (1965). Lake Victoria is categorized by these authors as a lake with a low total ionic concentration (conductivity between 90 and 145 pmho). There is virtually no short term change in major ionic composition. LAKE TANGANYIKA LAKE MALAW! DEPTH IN METRES N. BASIN (LAT. 5°S ) S. BASIN(7°S) (LAT. 11°S) =a 23-6-26:5°C 93-5-27:5 C 500 1000 PERPETUALLY DEOXYGENATED LAKE VICTORIA (LAT. I°S) ) q 23-8 -26:0°C 100 Fic. 2. Schematic cross-sections through Lakes Tanganyika (north and south basins), Malawi and Victoria to show differences in the shape and depth of their basins, and the extent of permanently deoxygenated water. The horizontal scale is proportional for lake width at the latitudes indicated. The annual temperature range of surface waters in the epilimnion is shown, as is the mean annual temperature of the hypolimnion (modified after Beauchamp, 1964). Probably the major climatic factor affecting overall lake ecology is the biannual rainy season (most marked in the northern half of the lake), in November and December, and from March to May. Apart from any direct effect by rainfall on the water mass itself, the rainy seasons effect the inflow of water through not only the _ larger affluent rivers (Kagera and Nzoia) but also the numerous and often seasonal streams draining the surrounding countryside. 10 Ps er. GRE IN W.@ OD Lake Victoria is, and seemingly always has been, an internal drainage basin (Temple, 1969; Bishop, 1969; Kendall, 1969) fed principally by the four larger rivers (Kagera, Katonga, Nzoia and Mara). Its only outflow, formed late in the lake’s history (Bishop, 1969), established contact with another, much smaller basin, that now occupied by the swampy Lake Kioga. Lake Kioga in turn has a river con- nection with the Nile at the northern end of Lake Albert. Faunistically, Lake Victoria is, however, isolated from the Victoria Nile and Lake Kioga by the now artificially submerged Ripon Falls. (The Owen Falls dam, built downstream of the Ripon Falls is probably an even more effective barrier to faunal interchange than were the falls.) Lake Kioga is cut off from the Nile by the utterly impassable Murchison Falls, some 40 m high. | Since the fishes of Lake Kioga, and especially the Cichlidae, are mostly conspecific with those of Lake Victoria (Greenwood, 1966a) it is very likely that the latter lake was the chief source of fish stocks for Kioga. Regrettably, the cichlids of Kioga have not been investigated at all thoroughly, but at least two endemic Haplochromis species occur there (Greenwood, 1967), and my own studies suggest that more await description. Lake Victoria originated during the mid-Pleistocene, about 750 000 years B.P. (see Bishop, 1969; Doornkamp & Temple, 1966, for summaries of geological and other evidence). At that time the future lake basin was crossed by several westward flowing rivers, of which the Kagera, Katonga, Nzoia and Mara are present-day relicts. These rivers drained the eastern highlands of Kenya, and emptied into what is now the Congo system. A gradual but large-scale warping of the plateau surface between the two arms of the rift valley led to a reversal of river flow and a back-ponding of the western reaches of these rivers. It should be noted that, for some of the rivers, the western upwarp that interrupted and reversed their flows back towards the east was relatively slight, probably in the region of 30-50 m. As a result of upwarping a two-way drainage was established, eastward into the developing Victoria basin, and westward into the proto-Lakes Albert and Edward. This drainage pattern still persists (Doornkamp & Temple, 1966; Greenwood, 1973a). Temple (1969) is of the opinion that backponding of the reversed rivers began earliest in the southern region of the Victoria basin, and progressed northwards along the slowly sinking plateau. As a river valley gradually filled it became a shallow, dendritic lake. Lake Kioga, lying near the head of the now interrupted and re- versed Kafu river valley, may well be a surviving example of such a lake (see especi- ally text-fig. 4 in Bishop, 1969). Eventually, each of the several lakes formed in this way overtopped the inter- vening and low watersheds and joined its neighbour. Developed and developing ~ lakes were gradually linked together to form a single, expansive water body that occupied an area considerably greater than that of the present lake. Raised beaches also indicate that at one time the Pleistocene lake was considerably deeper than it is now. Even after the single water body came into existence the lake basin was subject to periods of tectonic instability. These caused a tilting of the basin and correlated EVOLUTION OR. A CICHLiID-SsPECIES* FLOCK II changes in water level at the raised and lowered areas of its margins. One im- portant consequence of a major tilt to the northeast was the formation of an outlet (at Jinja) to the Kioga basin. Once the outlet was cut (possibly about 20-25 000 years B.P.; see Bishop & Trendall, 1967 ; Bishop, 1969) there was a gradual fall in lake level to that of the present day. It has been thought that during one period of the middle Pleistocene Lake Victoria was in contact, through rivers, with the proto-Lake Edward—George (Trewavas, 1933; Greenwood, 1959d, 1965a; Fryer & Iles, 1972). Evidence for this connection stemmed mainly from overall similarities in the Haplochromis fauna of the two lake systems, and the supposed sharing between them of certain otherwise endemic species (Trewavas, 1933). Recent research on the Haplochromis of Lakes Edward and George (Greenwood, 1973a) suggests, however, that these resemblances are more likely the result of parallel evolution. The supposedly shared endemics are, in fact, specifically distinct and thus endemic each to its own basin. Throughout the early history of Lake Victoria, and particularly during the later stages of tectonic stability, local and more widely spread climatic changes may well have produced fluctuations in lake levels if only by a few metres (Bishop, 1969 : III ; Kendall, 1969). Such changes would undoubtedly lead, first, to the formation of peripheral water bodies, and later to their reunion with the parent lake. A sequence of events like this would have a profound influence on the evolution, particularly the speciation pattern, of the fishes (see Greenwood, 1965a). I shall return to this point later. But, even more important in interpreting the phyletic and evolutionary picture of the Haplochromis species flock, especially its adaptive radiation, were the stages of independent lakes through which the basin passed early inits history. In effect, the present lake must be considered an amalgam of several lakes (Greenwood, 1965a ; and p. 114 below). FER PIshns OF LAKE VICTORIA The non-cichlid fishes Before going on to treat in detail the various elements of the Haplochromis species flock, some attention must be given to the other fishes with which the cichlids share their environment, and with which they may compete for food and living space. Unfortunately it is not possible to discuss the intriguing and fundamental question of competition in any detail because too little precise information is available on the ecology of Lake Victoria fishes, both cichlid and non-cichlid. Likewise, little is known about the invertebrate animals on which so many of the fishes feed. ‘With a few partial exceptions, the 38 species of fishes belonging to families other than the Cichlidae share habitats with the cichlids (Greenwood, 1966a ; and p. 45 below). The partial exceptions are those species whose habitat ranges include zones of the lake where cichlids are absent. For example, certain airbreathing species like the lungfish Protopierus aethiopicus, the anabantoid Ctenopoma muret and at least two species of the catfish Clarias penetrate fairly deeply into papyrus swamps. Only one cichlid (Hemihaplochromis multicolor) is found in this habitat 12 PH. GREENWOOD and even then it is confined to the peripheral and better oxygenated area. The small cyprinid Engraulicypris argenteus is the only truly pelagic fish in the lake. Inshore, where a distinction between pelagic and benthic zones is unrealistic, Engraulicypris does occur with several Haplochromis species, but it alone occupies the surface waters of the open lake. Most non-cichlid species enter rivers and the larger permanent streams at all times of the year, but apparently few cichlids doso. Those that do (Haplochromts nubilus, Hemthaplochroms multicolor and A statoreochromis alluaudt) are all species with a wide distribution in Uganda (Greenwood, 1959a, 1965b, 19732). The greatest intermingling of cichlid and non-cichlid species occurs in the littoral and sublittoral zones of the lake. Beyond a depth of ca 20m cichlid (1.e. Hapilo- chronus species) dominance is clear-cut, and few non-cichlids occur in deeper waters. Only one non-cichlid (the clariid catfish Xenoclarias) is confined to deeper water (10-90 m), a sharp contrast with the forty or more Haplochromis species known only from similar habitats. It is perhaps significant that, excepting Xenoclarias, non-— cichlid fishes inhabiting deepwater areas are all from species whose adults reach a length of over 250 mm, usually over 500 mm. Most non-cichlids do not breed in the lake, but migrate up streams and rivers to spawn during the rainy seasons (Whitehead, 1959 ; Greenwood, 1966a), again con- trasting sharply with the cichlid species. There is, however, some evidence that a few non-cichlids are able to breed in particular lacustrine habitats (Corbet, 1960, 1961; the deepwater clariid Xenoclarias [Greenwood, 1958a] only recognized after Corbet’s studies were completed, should also be included in this category). In- terestingly, the majority of deepwater inhabiting non-cichlids (Bagrus docmac, Mormyrus kannume, Synodontis victoriae and Xenoclarias eupogon) are also those species thought capable of breeding within the lake. These species would, therefore, not have to undertake the extensive vertical and horizontal migrations necessary to reach suitable breeding sites in rivers. It would be interesting to know if the deep- water populations of these species still follow the seasonal breeding patterns of their shallow-water congeners. Of the deepwater non-cichlids not suspected of lacustrine spawning (Protopterus aethiopicus and Claritas mossambicus), one (Clarias) is already known to make long migrations from offshore regions to reach suitable spawning sites (personal observations). The breeding habits of Protopterus (nest construc- tion, parental care etc., see Greenwood, 1958b) and aspects of its larval behaviour (Greenwood, op. cit.) certainly imply that this species is an obligatory inshore breeder. : Feeding habits of non-cichlid species have been studied in some detail (Corbet, 1961). The majority can be classified as insectivores, with chironomid larvae as their principal food organisms. Four species are essentially molluscivorous (feeding on both gastropods and bivalves) and three are predominantly piscivorous (with Haplo- chromis species as the main prey). The small pelagic cyprinid Engraulicypris argenteus mentioned before is the only species that can be considered a zooplankton feeder. Only two species, the characids Alestes sadleri and A. jacksoni are primarily herbivorous, feeding on rooted plants of the littoral region ; both species, however, also eat insects. EVOLULION OF WAY CICHLID SPECIES FLOCK 13 This résumé is perforce oversimplified ; as Corbet (1961) cautions, the feeding habits of all the non-cichlid fishes are to a certain extent facultative (at least within and sometimes beyond the major food categories used here) and vary with the size of the fish. Despite the broadly overlapping food requirements of these fishes, Corbet’s (op. cit.) elegant and detailed analysis led him to believe that there was little inter- specific competition for food ; to quote: ‘The few species with specialized feeding habits appear to enjoy a superabundance of food, whereas the others achieve the same object by remaining mobile and facultative.’ Regrettably, there are no comparably detailed studies on the food of the Lake Victoria cichlids (see Greenwood, 1956-69). However, sufficient is known about these species to indicate the existence of a considerable and broad interspecific trophic overlap, as well as the existence among these fishes of trophic specializations not encountered in the non-cichlids (see pp. 30-44 below). Corbet’s work also shows that non-cichlid feeding habits in lacustrine and fluviatile environments are essentially similar. This fact could be of importance in the ulti- mate evolution of trophic diversity within the Haflochromts species flock (see p. I15). In rivers, the non-cichlid species are the dominant fishes (see p. 4 above). Thus the early colonizers of the embryo Lake Victoria (itself probably a series of river-like lakes) would be mainly non-cichlid species. In the present-day rivers of Uganda, for example, there are only one or two Haplochromis species (and two or three other cichlids) as compared to twenty or more non-cichlids. The restricted environment of a developing lake would probably favour fishes that could exploit unoccupied feeding niches. That the ancestral Haplochromis were able to respond to this selec- tion is manifest in the trophic diversity of their descendant species. Some of the reasons for this cichlid potentiality are discussed in detail below (p. 103 ef seq.). The cichlid fishes The exact number of cichlid species in Lake Victoria is still undetermined ; every new collection yields undescribed species, especially now that fishery research vessels are operating in the deeper waters of the lake (Greenwood & Gee, 1969). It should also be noted, parenthetically, that cichlid species are being added to the fauna as the result of introductions from other lakes. In the following discussion these exotic species will not be considered ; most belong to the genus Tv/apza but inevitably species of other genera have been added unintentionally. There are two endemic Tilapia species (T. esculenta and T. variabilis) and between 150 and 170 species of the genus Haplochromis of which all but one, H. nubilus, are endemic (see Greenwood, 1973a where the presumed occurrence of Victoria species in Lakes Edward and George is discussed fully, and the idea discounted). Not all the Haplochromis species have been formally described, and the taxonomic revision of certain nominal taxa is stillincomplete. The estimated number of Haflo- chromts species is based on collections still unworked, on personal observations in the field, and on information from fishery biologists now engaged on deep- and midwater trawling surveys. In addition to the Haplochromis species sensu stricto there are four endemic monotypic genera (Macropleurodus bicolor, Hoplotilapia retrodens [Text-fig 72], z4 P, H. GREENWOOD Platytaentodus degeni | Text-fig 71] and Paralabidochromts victoriae ; see Greenwood, 1956a). All are derivatives of Hapflochromis species. A fifth monotypic genus, Astatoreochromis alluaud1, again a Haplochromis derivative (Greenwood, 1959, 1965c) is not restricted to Lake Victoria. A sixth species, also of wide distribution outside the lake, is Hemshaplochromis multicolor. This species is probably the only cichlid in the lake that should more properly be considered a fluviatile than a lacustrine one. Until recently Hemi- haplochromis multicolor was classified with Haplochromis. It is now separated from that genus because of its distinctive breeding habits (Wickler, 1963). I have certain reservations about the phyletic soundness of recognizing, as genera distinct from related Haplochromis species, some of the endemic monotypic genera (see Greenwood, 1973a; also p. 99 below). At this point it is interesting to compare the cichlid species flocks of Lakes Vic- toria and Malawi. Lake Tanganyika is not brought into the comparison because its cichlid species flock (or more correctly, species flocks) is not dominated by Haplo- — chroms and Haplochromis derivatives (see Regan, 1920; and Fryer & Iles, 1972 where Regan’s views on the diphyletic origin of these fishes is questioned, I think | justifiably on the evidence now available). There are more Haplochromis species in Lake Victoria than in Lake Malawi (ca 150, cf 105) but far fewer endemic Haplochromis-group genera (four cf about 20 in Lake Malawi). Furthermore, these latter taxa in Lake Victoria are monotypic whereas those in Lake Malawi are mostly polyspecific (Trewavas, 1935; Fryer & Iles, 1972). What interpretation can be derived from these figures? First, and most importantly, it should be stressed that the figures do truly repre- - sent differences between the degree of morphological differentiation existing in the species flocks of the two Jakes. The phyletic conclusions to be drawn are less ob- vious (see p. 99). Possibly the greater age of Malawi and the history of its lake basin (a deep rift valley lake) are or were important factors, as are the different types of habitat provided by a ‘graben’ as opposed to a saucer-shaped ultimate basin. All in all it seems likely that the Lake Victoria Haplochromis flock (including all but one monotypic genus) can be looked upon, at this point in time, as an arrested early stage in the more complex and morphologically more differentiated type of flock seen in Lake Malawi. I believe that the multiple-lake origin of Victoria (see p. 114) combined with its youth (and perhaps the relative rapidity with which it passed through the multiple-lake stage) are reasons why there has been greater speciation and less opportunity for more profound morphological differentiation among the flock. | In certain respects this question of different degrees of morphological divergence between the two flocks is more apparent than real. Considering both flocks from a phyletic viewpoint (i.e. one where propinquity of descent is a more realistic yardstick for measuring taxonomic relationships than are morphological gaps per se), I can find few grounds for elevating some of the Lake Malawi taxa to generic rank. Neverthe- less, there are still some Malawian taxa showing greater anatomical specialization than is seen in Lake Victoria. The Lake Victoria flock is undoubtedly a simpler one with a lower overall level of morphological divergence from the basic fluviatile | | ' EVOLUTION: OF A’ CICHLID<«SPECIES..FLOCK is v Haplochromis type (as represented today in East Africa by Haplochromis bloyeté (Greenwood, 1971)). The two endemic Tzlapia species of Lake Victoria (7. esculenta and T. variabilis) also provide a simpler picture, taxonomically and ecologically, than do the five endemic species of Lake Malawi (Lowe, 1952, 1953). There are good grounds for thinking that the Lake Victoria species were each derived from different ancestral lineages. Thus, strictly speaking, they do not constitute a species flock like the Tilapia of Lake Malawi (Fryer & Iles, 1972). Both Victoria species are specialized phytoplankton feeders (Greenwood, 1953). Tilapia esculenta obtains its food principally from phytoplankton in suspension, but T. variabtlis feeds mainly on the moribund phytoplankton of the bottom deposits. At least partly correlated with the trophic differences are marked interspecific differences in habitat preference (and in breeding biology). Both species are essen- tially from inshore regions, with T. esculenta penetrating into deeper water (Gee, 1968), but rarely to depths greater than 30m. Although the speciescan occupy similar habi- tats, T. esculenta is commoner in sheltered gulfs and bays, and T°. vaviabilis on exposed shores. (The principal references to the biology of Tilapia species in Lake Victoria are Lowe-McConnell, 1956; Fish, 1951, 1955; Garrod, 1957; and Fryer, 1961.) As specialized phytoplankton feeders the two Tilapia species probably occupy a virtually unique trophic niche in the lake. The records of Clarias mossambicus having ingested large quantities of phytoplankton (Graham, 1929; Greenwood, 1966a) only reflect the omnivoracity of this species (Corbet, 1961) if, that is, the ingestion of phytoplankton is not just accidental. Among the Haflochromis species, at least four combine the characteristic gut morphology of a vegetarian with known records of feeding on phytoplankton (see p. 39 below). None, however, shows the specialized pharyngeal dentition of the Tilapia species, nor such relative elongation of the gut. It seems likely that all four Haplochromis species get their food mainly from bottom deposits (i.e. are like T. variabilts in their feeding habits, but not in habitat as none has been captured in an exposed locality). Only these four Haplochromis can be considered in any way trophically competitive with the endemic Tilapia species. The whole question of interspecific relationships between the Tilapia and other species of Lake Victoria is, however, insufficiently studied for there to be any clear-cut indication of their pattern or consequences. About all that can be said is that Tilapia and Haplochromis species occur together in most habitats, and at all sizes, and that some Haplochromis tap the same food sources as does Tilapia variabilis. My recognition of Tilapia esculenta and T. variabilis (and by implication other members of the genus) as specialized species contradicts somewhat the views of Fryer & Iles (1969, 1972). These authors (1972) believe that species of Talapia . are in many respects more generalized than most fishes with which we are con- Bed (other lacustrine cichlids), and even their specialized feeding habits are little removed from those of “‘bottom grubbers”’’. That Tilapia are highly adaptable to a variety of habitats and environmental conditions (a phenomenon well established by field and experimental observations) 16 P. H. GREENWOOD might, I agree, be considered an indication of generalization. Stenotopic species usually evolve from a more generalized ancestor to exploit a particular niche. But, the pharyngeal apparatus of Tapia (especially in the phytoplankton feeders), its musculature and dentition, and the alimentary tract modifications (see Greenwood, 1953) also constitute a specialized condition within the Cichlidae. Attention may be drawn to the differences between these characters in the phytoplankton-feeding Haplochromis species and those in the various Tilapia species. The former only show marked departure from the generalized, omnivorous Haplochromts in the some- what lengthened gut ; the pharyngeal apparatus is nowhere near as specialized as it is in Tilapia. 1 would agree that, anatomically, H. erythrocephalus and other phytoplankton eating Haplochromis in Lake Victoria do not depart greatly from generalized ‘bottom grubbers’, but 77/afza I must consider specialized in its feeding habits. Unfortunately, few comparative data are available for any intergeneric differences in digestive physiology. Recent work by Moriarty and Moriarty:(1973) does not, however, indicate any differences in the ability to digest and assimilate blue-green algae between Tilapia milotica and the phytoplankton-feeding Hapflochroms mgn- pinnts of Lake George. (Haplochromis mgripinnis shows about the same departure from a generalized Haplochromis as does its trophic counterpart, H. erythrocephalus, in Lake Victoria.) If it be accepted that the pharyngeal apparatus, and as a correlate the feeding habits, of Tilapia are specialized it is not surprising that this genus has failed to produce any great radiation in trophic adaptations comparable with that seen in the Haplochromis species flocks. The anatomically generalized fluviatile Haplochromis species (like other generalized animals) have a greater evolutionary potential, as will be discussed later (p. 103 ef seq). What is surprising, is the relatively low level of speciation in Tilapia. That there are about 70 species of Tilapia in Africa compared to nearly 300 Haplochromis species in Lakes Malawi and Victoria alone underlines this point. In Lake Victoria there are only two Tilapia species (each derived from a different ancestral lineage), and in Lake Malawi merely five species (but nevertheless the largest single naturalassemblage — of Tilapia species in any lake). Furthermore, among the fluviatile Tilapia species there are many with a considerably greater geographical range than any fluviatile species of Haplochromis. The question of why Tziplapia did not speciate to the same degree as did the Haplochromis in Lake Victoria (whose developmental history [see pp. 10 & 114] would seem to provide an ideal background for allopatric speciation) is particularly diffi- — cult to answer. There are, among the Haplochromis of this lake, several instances | of species multiplication within any one trophic group. That is, evolutionary change | not involving any change in the particular trophic specializations possessed by the group. . The endemic Haplochromts species of Lake Nabugabo (a but recently isolated bay of Lake Victoria ; see Greenwood 1965b) are another example of speciation without | noticeable adaptive change in feeding habits. The Tilapia isolated in this lake did | not speciate. EVOLUTION OF A CICHLID SPECIES FLOCK 17 It is possible that conditions in the various water bodies which have contributed to, and have been part of, Lake Victoria did not provide sufficient food for isolated Tilapia populations to survive long enough for speciation to take place. This, however, seems a most unlikely explanation. Tilapia, of several species, kept in aquaria and dams are able to survive and breed on the most atypical diets (including carnivorous ones), and their adaptability to adverse environments is well known. About the only conclusion to be drawn at present is that some inexplicable genetical stability, and thus evolutionary conservatism, is an inherent characteristic of Tilapia. By contrast, Haplochromis appears to possess an extreme degree of genetical instability that manifests itself in the repeated explosive speciation of this taxon. The theoretical resolution of this problem is not, unfortunately, quite so simple when it is extended beyond the genus 77/apza to include other parts of the tilapiine* lineage. The supposedly tilapiine derivatives that feature so prominently in the species flocks of Lake Tanganyika (Regan 1920; Poll 1956) seem to show an evolutionary potential at least equal to that of the haplochromines in Lakes Victoria and Malawi and, indeed, in Lake Tanganyika. Doubt has recently been cast on the phyletic integrity of this lineage in Lake Tanganyika (Fryer & Iles, 1972). If the revised and admittedly tentative phylogeny of Tanganyika Cichlidae proposed by Fryer & [les (op. cit. : 502-508) is a better approximation to reality than that implicit in Regan’s (1920) classification, then the evolutionary potentiality of haplochromines is even more outstanding. There is, however, one outstanding example of what seems to be a truly tilapiine radiation, the fishes of Lake Barombi-Mbo (Trewavas, 1972). The 11 endemic cichlid species of this isolated crater lake in northwest Cameroons have been carefully investigated by Trewavas (op. cit.). Four of the five genera present are endemic. Despite their peculiar oral and pharyngeal dentitions, and their unusual body form, the tilapiine affinities of the group are more obvious than are those of most Tanganyika genera. The endemic species of the single non-endemic genus (Saroherodon) depart quite considerably from the ‘typical’ species of that genus. For the moment I would consider that the Barombi-Mbo cichlids provide the most serious challenge to the arguments against tilapiine affinities for the Tanganyika genera mentioned above. The deviation, from their nearest relatives, shown by the Barombi-Mbo fishes is certainly greater than that seen in the four endemic monotypic genera of Lake Victoria. All four are undoubtedly haplochromine derivatives (Greenwood, 1956a) as is the fifth and more widely distributed taxon, Astatoreochromis alluaudt (Green- wood, 1956a, 1959a, 1965b 197384). | That none of these Lake Victoria taxa is polyspecific like so many of their morpho- equivalents in Lake Malawi (e.g. the nine genera and 28 species of the ‘Mbuna’ complex, Trewavas, 1935; Fryer, 1959) may be attributable in part to the greater -* The terms ‘tilapiine’ and ‘haplochromine’ are used here without any formal taxonomic (i.e. sub- familial) connotation, and merely as convenient handles for the supposed two major cichlid lineages. 2 18 P. H. GREENWOOD age of Malawi, and in part to the fact that the Victoria species have not exploited a habitat which was unexploitable by less specialized Haplochromts species. The five monotypic genera have been distinguished from Haplochromis purely on the basis of the morphological gap separating each from any Haplochromis in the lake (Greenwood, 1956a). One genus, Macropleurodus bicolor, is, however, very closely similar to an extant Haplochromis species, H. prodromus (Greenwood, 1957 ; and p. 71 below) ; had more been known about the morphology of Victoria Haplo- chromis at the time of its description (and its later redescription), Macropleurodus might well have been included in that genus. The same reasoning applies to the classification of Paralabidochromis victoriae. This species (Greenwood, 1956a) is known only from a single specimen, whose dentition closely resembles that of the genus Labidochroms from Lake Malawi (Trewavas, 1935), a fact that certainly influenced my decision to place it in a separate genus. The other two genera, Platytaeniodus degeni and Hoplotilapia retrodens, do have very distinctive dental characters (Text-figs 73 and 74 respectively ; also Greenwood, 1956a). Even now, neither can be related to any particular Haplochromis species within or without the lake. They provide, in fact, the only recorded instance of specialized dental or other characters not linked by intermediates to the generalized condition. Astatoreochromts alluaudi is the only monotypic genus differentiated from Haplo- chromis by non-dental characters; its differential characteristics are the higher number of dorsal and anal fin spines, especially the latter (four to six spines cf three in Haplochromis). It also differs from the other genera in having a distribution that extends beyond Lake Victoria to include Lakes Nabugabo, Edward and George, and several small lakes in western Uganda (Greenwood, 1959a, 1965b, 197384). Unlike most Lake Victoria cichlids, A. alluaudi freely enters streams and swamps. Within the lake the species shows a wider habitat tolerance than do the other mono- typic genera and, indeed, many endemic Haplochromts species of the inshore lake regions. But, like these various species it does not penetrate into water more than about 20 m deep. The specialized crushing pharyngeal apparatus of A. alluaudz is identical with that found in two endemic Haplochromis species, H. tshmaeli and H. pharyngomylus (Text-fig. 5). The diet of all three species is identical, viz. gastropod molluscs, especially Melanotdes tuberculata. Despite this shared pharyngeal specialization I do not think that A. alluaudt is closely related to either of the Haplochromis species. Indeed, there are good reasons for considering it to be derived from a different stem to that of the other monotypic genera and the rest of the Haplochromts species flock (Greenwood, 1954, 1959; also p. 100 below). Trophically, all the monotypic genera (except Paralabidochromis whose feeding habits are unknown) can be classed as mollusc eaters. Macropleurodus bicolor feeds on gastropods, its feeding methods are like those of Haplochromis species that remove the snail from its shell before ingesting the soft parts (Greenwood, 1957, and p. 37 below). Hoplotilapia eats, principally, bivalves, crushing the shells between its broad bands of jaw teeth; Text-fig. 74 (also Greenwood, op. cit.). Astatoreo- chromis alluaudi feeds on gastropods which are crushed in the hypertrophied pharyngeal mill (Greenwood, 1959a, 1965c). The feeding habits of Platytaeniodus ere eres EVOLUTION OF A CICHLID SPECIES. FLOCK 19 degent (Text-fig. 73) are largely unknown, but bivalves have been recorded from its gut contents, as have bottom detritus and insect remains (Greenwood, 1956a). It is clear that despite the outstanding dental characters (Text-figs 73-75) of these genera, there is little to indicate that any one has entered an adaptive grade, or a particular habitat, which is not occupied by at least one Haplochromis species (see below). Perhaps the only point of interest in this context is to note that whereas Haplochromis species feeding on bivalves crush the shells by means of variously hypertrophied pharyngeal teeth, the monotypic genera achieve the same ends by using an hypertrophied oral dentition. Four of the monotypic genera are female mouth brooders (as are all Lake Victoria Haplochromis species whose brood care is known) ; nothing is known about breeding in Paralabidochromts victoriae. The monotypic genera of Lake Victoria can be equated with the ‘Mbuna’ generic complex of Lake Malawi (Trewavas, 1935 ; Fryer, 1959) on the basis of their having distinctive dental and oral characteristics. There, however, the comparison ends because the ‘Mbuna’ genera have occupied a habitat (rocky surfaces) to the virtual exclusion of Haflochromis species. In Lake Victoria the monotypic genera are, from the ecological viewpoint, indistinguishable from Haplochromts species. Pie Ae LOCH ROMIS SPECIES FVOCK Introduction Having reviewed what might almost be considered the ‘minor characters’ amongst the fishes of Lake Victoria, attention can now be given to that ecologically and morphologically diverse, closely related and dominant assemblage, the Haplochromis species flock. The estimated total of 150-170 Haplochromis species in Lake Victoria is based partly on the number of species already described (Greenwood, 1956-67 ; Greenwood & Gee, 1969) but also on the undescribed taxa from collections still under study and from information given to me by workers who have sampled areas of the lake in which I have not worked (particularly deep benthic and midwater habitats). In the accounts that follow, data are drawn mainly from the 95 species that have been described or redescribed and which I have studied in the field. It must be emphasized that nothing learned from the still undescribed species in my possession seriously modifies this picture. The term ‘species flock’ ( = species swarm of Mayr, 1963) should, strictly speaking, be applied to a species assemblage of monophyletic origin. A monophyletic origin cannot definitely be established for the Lake Victoria species, although the evidence points in that direction. If the origin was not monophyletic then it was extremely oligophyletic (see below). The existing species are certainly more closely related to one another than to any species outside the lake, and it seems justifiable to refer to the assemblage as a species flock. The Haplochromis of Lake Nabugabo (Greenwood, 1965b), of course, contradict these last remarks. But Nabugabo cannot, in this context, be considered ‘another’ lake ; were it not for a narrow sandbar it would be part of Lake Victoria. 20 P. H. GREENWOOD A close relationship between the species of Lake Victoria and those of Lakes Edward and George (but not Rudolf and Albert) cannot be denied. In fact, if Haplochromis from either of these lakes were put into Lake Victoria, they would not seem at all ‘out of place’ toataxonomist. The same cannot be said for the majority of Haplochromis species from Lake Malawi. For example, in all but one or two species the Malawi Haflochromis have the caudal fin covered by small scales ; only the basal third, rarely the proximal half of this fin is scaled in Victoria species. And again, there are male breeding colours and colour patterns among Malawi Haplo- chromis and related genera that do not occur in the Victoria flock (Regan, 1921 ; Fryer & Iles, 1972 ; personal observations). : The problem of a mono- or polyphyletic origin for the Lake Victoria flock is not a simple one to solve (Astatoreochromis alluaudi excepted since it is manifestly more closely related to species outside the flock, see p. 100). | Tt has long been thought, on morphological evidence, that the flock could have stemmed from a single species (Regan, 1922; Trewavas, 1949). Finding evidence to refute this hypothesis is difficult, and is bound to be so if the mid-Pleistocene rivers of eastern Africa carried a similar Hapflochromis species complement to that of the present-day rivers — one or at most two very closely related species. There seem to be no grounds for assuming that the situation might be any dif- ferent in mid-Pleistocene times. Even if different species did occur in neighbouring rivers, the chances are that they would be closely related. Thus it is not surprising to find that many elements from the species flocks of Lakes Victoria, Edward, George, Albert and Rudolf have a close overall resemblance to one another. The rivers that first drained into these lakes were all part of the east African highland drainage, and presumably carried the same or genetically similar Haplochromis species. The relicts of this drainage system are now populated by a single Haplochromis species. Previously this fish was thought to be H. wingatu (see Trewavas, 1933) but recent work indicates that it is either H. bloyett or a closely related species (Green- wood, 1971). Haplochromis wingatu is restricted to the Nile and Lake Albert, and appears to represent a lineage quite distinct from that of most east African lake Haplochromis species (see Greenwood, 1971, 19738). Lake Malawi is geographically far removed from the rivers of the old east African drainage and it is likely that its ancestral Haplochroms populations would differ from those of Lakes Victoria, Edward and Albert (see p. 99). The modern representative of the basal Haplochromis stock in Lake Malawi is thought to be H. calipterus (Trewavas, 1949). Structurally, this species too is not far removed from H. bloyet. Lake Tanganyika poses a problem. Geographically and hydrographically it lies within the range of present-day H. bloyeti (unpublished observations) yet its cichlid flocks are very different from those of Lake Victoria (Regan, 1920; Poll, 1956; Fryer & Iles, 1972). For one thing, the genus Haplochromis is barely represented in the lake by two species. One of these (H. burtoni) is a generalized species of restricted intralacustrine distribution (Poll, of. cit.; personal observations). The other (H. horet) is a moderately specialized predator also of relatively restricted intralacustrine distribution. Furthermore, nearly half the species flock is composed of genera apparently belonging to the tilapiine lineage (Regan, 1920; but see p. 17 | | | | | | | | EVOLUTION (\OFUVA CICHLID::SPECIES FLOCK 21 above). Those genera belonging to the haplochromine lineage are as distinct from the general Haplochromis morphotype as are the monotypic genera of Lake Victoria, or even more So. Here then we have the anomaly of two geographically distinct lakes, Victoria and Malawi, with basically similar species flocks neither of which resembles superficially or phylogenetically that of the geographically interposed Lake Tanganyika. No immediate explanation is available. Of course the flock in Lake Tanganyika (even restricting comparisons to supposedly haplochromine derivatives) is much further differentiated than that of Victoria — probably a reflection of the lake’s greater age — thereby making an inter-lake assess- ment of relationships very difficult. But, even allowing for differences in lake age, topography and history, there does seem to be a prima facie case for thinking that the ancestral species in Lake Tanganyika were quite different from the progenitor or progenitors of the Lake Victoria flock. Basic morphology Viewed in their entirety the Lake Victoria Haplochromis species show remarkably little diversity in body form, particularly when compared to the species flocks in other lakes (compare Text-fig. 3 with figs. 5-23 in Fryer & Hes, 1972). The piscivorous predators, as a whole, are generally recognizable by their rather elongate form and large mouth (see Text-figs 3, 11-15, 52-56 and 58-64). But, in this trophic group there are several species (the ‘sevranus’ group in particular, Greenwood, 1967) that, apart from their larger size, closely resemble the trophically unspecialized species (cf Text-fig. 9 with Text-fig. 48). Even some of the most specialized predators, the paedophagous species (Greenwood, 1959b ; and pp. 31-37 below), are not particularly outstanding in their superficial appearance (see Text-figs 16-18). There are, of course, exceptions and again most of these are found amongst the piscivores. In these species the outstanding morphological features are as- sociated with the head, like a strongly prognathous lower jaw, an upwardly directed mouth, or the deeply concave dorsal head profile often associated with this mouth form (see Text-figs 14, 56, and 64). The generalized Haplochromis body and head shape (found in more than half of the known species) cloaks a wide variety of trophic specializations, from algal grazing to mollusc eating and even predation on other Haplochromis. These remarks apply to first impressions coming from a superficial examination of the fishes. Closer inspection in many cases shows differences in jaw and tooth morphology among otherwise similar species. Indeed, it is in the head and dentition that the real diversity of these species becomes apparent, and in which lies the evolutionary success of the Haplochromis species flock. Nevertheless, even the most extreme forms (except for two of the monotypic genera ; p. 103) are linked with the generalized type by species showing between them all intermediate stages in the development of a particular characteristic. From the last few remarks it must not be concluded that every species is 1m- mediately distinguishable by its cranial characters. Rather, these should be taken as ‘species group’ characters, shared in many instances by more than half a dozen 22 Pp, H. GREENWOOD H. pallidus Hse ee MUN? i a EA <=<=S . ee = H. chilotes ~~ Macropleurodus bicolor nae MM ify i “Za seamen i (@) = @) . fos Rog SS oe eE SS SS H. saxicola & H. empodisma WS Ss Outline drawings (not to scale) of various Lake Victoria Haplochromis species (and one monotypic genus, Macropleurodus bicolor) to show range of body form in this species flock. Fic. 3. EVOLUTION:-OF A CICHLID SPECIES. FPLOCK 23 H. ishmaeli De oy VG p74 Mh foefff B= {7 {ff ff fi, Yf AY y P oo Z Yi V4 or LF > LEE H. dentex H. obesus SS SS H. longirostris H. parvidens Mtl Wipes H. guiarti. H. macrognathus 24 P. H. GREENWOOD species. The two mollusc-eating species groups (Greenwood, 1957, 1960), the ‘tvidens’ group of species feeding on benthic Crustacea in deep water (Greenwood & Gee, 1969) and the several morphogroups of piscivorous predators (Greenwood, 1962, 1967) typifying this situation. Within a group the morphological characters separating species are mainly slight proportional differences, squamation patterns, differences in the number or disposition of teeth (and less commonly tooth shape) and, most clearly, differences in male breeding coloration (see p. 52 below and Plate I). In most of what may be termed ‘specific characters’, except male coloration, there is a high level of individual variability. To exemplify this last point, and at the same time to emphasize the relative in- variability of some morphological features, one may consider the meristic characters of the flock. The range of fin ray numbers, lateral line and most other scale counts, and the number of vertebrae is such that the range for the entire flock can be en- countered in one species. There are a few exceptions to this generalization ; all are species with an elongate body form. In the exceptional species, the modal number ~ and upper limits for the range of vertebral and lateral line scale counts may lie beyond the general range, but the lower limits lie within that range. Few species show any external evidence of modifications associated with a par- ticular habitat. Exceptional in this regard are some, but not all, species from deeper water (i.e. > 50m). These fishes have relatively larger eyes, and the cephalic laterosensory canals and openings are slightly enlarged, presumably as adaptations to a dimly lit environment (Greenwood & Gee, 1969 ; Greenwood, 1973a). Adult individuals of most species are small fishes, between 70 and 110 mm long, but with some of the piscivorous predators growing to lengths of 180-220 mm. In all the morphological features discussed so far, the Lake Victoria Haplochromis flock (including the monotypic genera) shows less diversity than does the flock in Lake Malawi (even if the derivative genera are excluded). Because few morpho- metric data are available for individual species of Malawi Haplochromis, it is im- possible to compare the levels of intraspecific variability in the two flocks. Contrasting with the relative uniformity of body shape, the jaw teeth and dental patterns in Lake Victoria Haplochromis show a much wider range of diversity (see Text-figs 4, 36, 39 and 41). It is this particular diversity, more than any other factor, that has contributed to the success of these fishes in the lake (and in other lakes). There are, of course, correlated changes in syncranial architecture, especially neurocranial shape, and many of these are probably to be considered the primary changes involved inanadaptiveradiation. Certainly, takenin concert, skull form and dentition are basic to the trophic radiation so characteristic of all cichlid species flocks. The well-developed pharyngeal apparatus of the cichlids (i.e. the toothed and separate upper pharyngeal bones and the toothed but suturally united lower ele- ments, together with the associated musculature) are effectively a second pair of jaws. In many species these ‘jaws’ are of greater importance than the true jaws, or at least have given the species an enhanced potential for exploiting a wider variety of food sources. Examples of this potential realized are found in the phytoplankton feeders (Green- wood, 1953), the mollusc eaters (Greenwood, 1959a, 1960) and in the piscivorous EVOLUTION OF A CICHLID SPECIES FLOCK 25 Fic. 4. Jaw teeth of various Haplochromis species. A: H. macrops (an omnivore) ; bicuspid outer and tricuspid inner tooth from the dentary. These teeth may be considered as representative of the typical generalized tooth form. B: Unicuspid, caniniform outer teeth from the premaxilla of H. dentew (a piscivore), showing different degrees of curvature. C: Stout, strongly recurved, bicuspid outer row tooth from the dentary of H. welcommei (a scale scraper) in buccal (left) and lateral (right) views. D-G: Teeth of four vegetarian species showing morphological stages in the evolution of specialized, obliquely cuspidate cuter teeth in the periphyton grazer, H. obliquidens (G). The species represented are: (D) H. evythrocephalus (phytoplankton eater), (E) H. nuchi- squamulatus, (F) H. lividus and (G) H. obliquidens, all periphyton grazers (see p. 39). predators (Greenwood, 1962). Phytoplankton feeders have the pharyngeal teeth fine and hooked, thus enabling the fish to comb aggregates of mucus and phyto- plankton into the oesophagus. In mollusc eaters the molariform pharyngeal teeth (Text-fig. 5), strong pharyngeal bones and powerful upper branchial musculature enable the fish to crush gastropod shells. In piscivores (Greenwood, 1962 : 211) there is a macerating action of the strong but fine upper and lower pharyngeal teeth moving against the prey caught between them. This allows the predator to ingest much larger prey than would be possible if the food was bolted whole (as is the usual -way in most non-cichlids). Such examples are, in some respects, extreme cases. Most species have what can be described as a ‘general purpose’ pharyngeal apparatus. Consequently there is, 26 P. H. GREENWOOD on the whole, rather less diversity in bone form and tooth shape than is seen in the morphology and dentition of the jaws. In the jaws (premaxillary and dentary bones) there is always an outer row of large teeth, followed by from one to five (sometimes more, but usually two or three) rows of much smaller teeth (Text-fig. 41A). Teeth in the outer row show greater vals in shape and size (Text-fig. 4) than do those of the inner series. Fic. 5. Lower pharyngeal bones (in occlusal view) of four species to show increasing molarization of the teeth associated with an increasingly molluscivorous diet. Haplo- chromis empodisma (extreme right) has a mixed diet of insects and bottom detritus while H. pharyngomylus (extreme left) is exclusively molluscivorous when adult. Drawings not to scale. B: H. obtusidens. C: H. humilior. Inner teeth are generally tricuspid or unicuspid ; their main contribution to dental diversity is in the number and pattern of the rows present. For example, there are the broad bands of fine, tricuspid inner teeth in those species that scrape epiphytic and epilithic algae (e.g. H. obliquidens, H. lividus, H. mgricans and H. nuchisqua- mulatus ; Greenwood, 1956b) or, ike H. welcomme1, which scrapes scales from the caudal fins of other cichlids (Text-fig. 39 ; and Greenwood, 1966b). Broad bands, this time of stouter teeth (Text-fig. 41), are found in certain of the species feeding on molluscs, either by wrenching the snail from its shell or by crushing the shell between the jaws as in H. prodromus, H. granti and H. xenognathus ; see Greenwood, 1957. The ultimate development of this trend is seen in the broad and posteriorly expanded inner tooth rows of Hoplotilapia retrodens and Platytaeniodus degen (Text-figs 73 and 74). Teeth from the outer row in the premaxilla and dentary are, despite their diversity in form, variants of two basic types (Text-fig. 4). In one, the crown is compressed and unequally bicuspid, in the other the crown is conical and protracted into a single sharp point (i.e. it is unicuspid). Occasionally, tricuspid teeth are found in the outer rows, but with few exceptions (e.g. H. tvidens and related species ; Greenwood & Gee, 1969) are never common anteriorly in the jaws. Teeth morphologically intermediate between the fully bicuspid and the unicuspid | type are known, but never seem to occur in the same species or in the same individual (except as the result of wear on a bicuspid). The bicuspid tooth is, by analogy with teeth in generalized and fluviatile species like H. bloyeti, taken to be the basic and generalized tooth form (Text-fig 30B). It also seems to be the juvenile tooth form irrespective of the definitive tooth shape for a EVOLULION-OF A CICHLID SPECIES. FLOCK 27 particular species. This temporal succession is seen, albeit in muted form, among species whose definitive dentition is that of unequally bicuspid teeth ; in some large antividuals of these species there is often an admixture of bi- and unicuspids in the inderior and lateral parts of the dental arcade. One or two relatively enlarged unicuspids are generally developed at the posterior end of the outer tooth row in the premaxilla, irrespective of tooth shape elsewhere in this jaw ; exceptional are the few species like H. obliquidens (Text-fig 4G) where crown form is very greatly modified from the usual condition. Tooth succession is vertical, the replacement teeth developing in alveoli beneath the erupted and functional teeth. As far as can be determined, the replacement of outer teeth is more regular than that of inner ones (personal observations). The basic bicuspid tooth has one cusp (the major cusp) noticeably larger than the other (Text-fig. 4A and D). Both major and minor cusps are triangular in outline, and lie in the same plane. Variations on this pattern involve changes in the size of the apical angle of the cusps, changes in the relative sizes of the two cusps or the inclina- tion of one cusp away from the plane of the other. The minor can be as large as the major (Text-fig. 36) or the latter can be drawn out obliquely into a scraping blade many times larger than the minor cusp (Text-fig. 4G). Another but rarer type of variation involves a coarsening of the whole tooth with a consequent increase in the breadth of the crown ; sometimes this trend is combined with one of those noted earlier, and results in the powerful teeth of Macropleurodus bicolor, an admittedly extreme example of this trend (Text-fig. 75). The length of a tooth’s neck also varies interspecifically, as does its circumference. In consequence, teeth can be short and stout, long and slender, or of a intermediate type, but all with a similar cusp shape. Unicuspid teeth differ in relative size, degree of curvature (usually in the neck of the tooth, but sometimes in the angle between the crown and neck, as in HZ. obesus, H. maxillaris and H. melanopterus ; see Greenwood, 1959b), or in the angle at which the whole tooth is implanted on the jaw bone. These differences are well exemplified by the more specialized piscivorous species, and by some of the specialized insectivores and certain mollusc eaters. In the former group (Greenwood, 1962, 1967), the teeth are large and strongly recurved, thereby providing a means for gripping the prey while the pharyngeal teeth macerate it preparatory to swallowing. In certain specialized insectivores and mollusc eaters the teeth are also relatively enlarged, and although the tips are recurved, the teeth are implanted so as to project forward (Greenwood, 1957, and 1959b : 207-211). The procumbent teeth in the insectivores (especially H. chilotes, Text-fig. 33) create an effective ‘forceps’ used to remove burrowing insects from their holes. In the molluscivores, the teeth are used either to hold a snail while it is eased from its shell, or actually to crush the shell away from the body. Lake Malawi Haplochromis (and especially the derivative genera) show an even greater range of dental morphology and arrangement. The difference is clearly seen by comparing Text-figs 4, 36 and 73-75 here with those for Malawi species published by Fryer (1959) and Fryer & Iles (1972) ; similarly for the haplochromine species of Lake Tanganyika (Poll, 1956 ; Fryer & Iles, op. cit.). 28 P. H. GREENWOOD The whole subject of cranial and dental morphology in the Lake Victoria Haplo- chromts flock is dealt with in greater detail below (pp. 56-99). For the moment it must suffice to note that although species showing specialized dental, pharyngeal and syncranial characters are very distinctive (and readily advertise their feeding habits) the majority are linked, through species with these characters at an intermediate stage, to the generalized type (see Text-figs 4, 5 and 65-60). : Many of these morpho-lineages appear also to be truly phyletic ones because the component taxa can be related, primarily, though shared specializations, and secon- darily by degrees of specialization in the same characters. There are a few instances where intraspecific variability in a complex of specialized characters is such that, were only the extreme individuals known, they would probably be classified as a separate genus (for example, H. xenognathus, Greenwood, 1957; and H. welcommei, Greenwood, 1966b). Surprisingly, amongst species with such diverse feeding habits, there is little diversity in the shape, length or number of gill rakers; the modal numbers of rakers aregorio. The plankton-feeding species (e.g. H. erythrocephalus ; Text-fig. 6) have A Why, ~. “., vorous—piscivorous species) and B: H. erythrocephalus (a phytoplankton feeder). rakers that are but marginally longer than those of an insectivore, are only a little more closely spaced, and are more numerous by one or two rakers. No explanation — can be offered, except that the feeding mechanism of phytoplankton feeders may not require the evolution of close-set and fine rakers. These species entangle the plank- ton in mucus boli that are too large to pass through even the relatively wide spaces between rakers of the first row ; subsequent rows of rakers interdigitate and have even smaller interspaces (Greenwood, 1953 ; and unpublished observations on several Haplochromis species). EVOLUTION OF A CICHLID SPECIES FLOCK “2g The theory of character displacement (Brown & Wilson, 1956) may be discussed here because it has been largely established on morphological characters, and because it would seem to be of particular relevance to a species flock situation. Its ecological counterpart, the concept of competitive exclusion (Gause’s Law), will be considered later (p. 48) when discussing species’ interrelationships in the lake. In essence, the theory of character displacement postulates that closely related sympatric species will generally show more differences than similarly related allopatric ones. In other words, selection will sharpen differences between sympatric species if these differences lead to reduced interspecific competition. A species flock should, therefore, provide good material against which to check the first premise of the theory, although the second premise cannot be tested because none of the species ever occurs allopatrically with a close relative. The general effect of character displacement seems to be discernible within the flock for many morphological features (and, by implication, even more clearly for ethological characters, see p. 51). A most striking example is provided by the dif- ferences in male breeding coloration (see p. 52). This example is particularly interesting because interspecific colour differences are most marked among species that occur syntopically and are less obvious between allotopic species. The picture for other morphological features, especially anatomical ones, is far less obvious, and at first glance might even seem to contradict the idea of enhanced differentiation between sympatric species. To illustrate this difficulty one can take the several anatomically and morphometri- cally stmilar species clustered around any one adaptive peak in a phyletic lineage. This is especially well demonstrated among such disparate groups as the generalized insectivores (Greenwood, 1960), in the specialized mollusc crushers (Greenwood, of. cit.), and particularly in the piscivores (Greenwood, 1962, 1967). However, traces of character displacement are evident, not so much at the level of individual species but at the level of species groups (i.e. phyletic lineages.) The characteristic features of these groups are, almost without exception, trenchantly defined. Proximately, of course, what are now recognizable as phyletic groups must have originated as species. It was presumably at that level and at that time that selec- tion pressure, and hence character displacement, was most intense. The mosaic of interspecific similarities and dissimilarities in the Hapflochromis of Lake Victoria could well mirror the way in which the flock evolved. Character displacement (and the origin of phyletic lines) would be most marked during early phases of lake development when a high premium might be placed on trophic speciali- zation. Character replication, through simple speciation without obvious adaptive change, is likely to be a feature of later lake development, with the isolation and re- incorporation of peripheral, Nabugabo-like lakes (see p. 112). _ That intragroup character replication should exist on the scale it does among these fishes is certainly unusual (see the numerous contrary examples cited by Brown & Wilson, 1956). It seems to suggest two possible explanations, either a lowered level _ of competition and selection during certain phases of lake evolution, or the involve- ment of adaptive characters other than those reflected in morphological features. A temporal factor may also be involved. Ona purely subjective assessment there seems 30 Po oe GREENWOOD to be greater interspecific differentiation among the Haplochromis species of Lake Malawi than those of Lake Victoria. The Malawi flock is older (see p. 6), and during its longer life perhaps some of the less distinctive taxa (1.e. less particularly specialized species) have been eliminated at times of more rigorous selection. Feeding habits of the Haplochromts species The importance of trophic specializations in the adaptive radiation of Lake Vic- toria Haplochromis species has long been recognized (see Regan, 1921, in the pre- factory remarks to the first systematic revision of these fishes). The full magnitude of this radiation only became apparent, however, when field studies were made (Graham, 1929; Greenwood, 1959c, 1905a, 1973b, and in taxonomic papers dealing with these species, 1956-69). Lake Victoria Haplochromis species are by no means unique in this respect, and the phenomenon can be considered a characteristic of cichlid fishes in many African lakes (and elsewhere in the world). Outlines of these different radiations were published, for Lake George by Greenwood, 1973a ; for Lake Malawi by Fryer, 1959 and especially by Fryer & Iles, 1972; for Lake Tanganyika by Poll, 1956 (also in Fryer & Iles, op. cit.), and for Lake Albert by Trewavas, 1938. When discussing Haplochromis feeding habits in Lake Victoria certain points must be borne in mind. First, the food of small fishes (i.e. < 20 mm long) is unknown for the majority of species, partly because small individuals were rarely caught, but mainly because it is impossible to identify fishes of this size. Second, fishes caught by trawling in deep water usually have everted guts when brought to the surface, and much of the food is lost in this way. Finally, there has been little detailed research (like that of Corbet’s [1961] on non-cichlids) into the food of Haplochromis ; what data there are, derive chiefly as an offshoot from my own taxonomic studies. Within these limitations it is possible, however, to say that every major food source in the lake, except for zooplankton, has been exploited by one or several Haplochromis species (Text-fig. 7). Insects, especially larval and pupal chironomids and Ephemeroptera, are prob- ably the most important food organisms. Besides the 11 species* feeding principally on insects many others include some insect material in the diet, either regularly or opportunistically. Most insectivores are morphologically (Text-figs 3, 8-g and 42) and dentally generalized fishes, with bicuspid jaw teeth, generalized jaw structure (Text-figs 31 * Only the 95 species so far dealt with in my revision are used in these analyses. Fic. 7. Diagrammatic representation of feeding habits in the Lake Victoria Haplochromis species studied to date (see p. 19). For details of species in the H. macrops, H. serranus, H. prognathus and H. tridens species complexes see text, pp. 58; 81; 85; and 67. Where more than one food is represented, the lengths of the bars are not strictly pro- portional ; they should be considered merely as indicative of the relative importance of the particular foods in the diet of that species. Only the principal types of food organisms are shown. ‘The species are arranged in phylogenetic groups (see Text-fig. 70). TZ FO H. brownae H. cinereus TILL LL H. pharyngomylus =n” An sei 8 snuajdourjew) ‘H H. bloyeti — like ancestor mH salineds ¢ one Hadios Sd @ H. Rrogna (eo) s e Farban, e@ cr, 14 e Ploda, LY, iy N NO) “roy N NA i V N Q % mm ‘A. \ IS on \ * Vo @ ; % V \ 4 “yg z %, om & @ % eS : a 6. Je % N Insects ig Mollusca (w % —< Bg Phytoplankton ij Fish - H, serranus complex (IN species) H. parorthostoma thy. * (20 speqi ote, es) Detritus < 1! eee un 8 e Periphyton Macrophytes Crustacea Fish eggs & larvae Rie Br sano ye + ‘ EVOLUTION OF A CICHLID SPECIES FLOCK 31 and 33) and a pharyngeal apparatus as unspecialized as the primary jaws (Text-figs 5D and 30). The indications are that these species feed by sucking or picking their prey from the bottom or other substrata (plants, rocks, etc.). The most specialized insectivore, H. chilotes (Text-fig. 10), is also, by virtue.of its hypertrophied lips, the most easily recognized species in the whole flock. Anatomi- cally, H. chilotes has a forceps dentition of procumbent, unicuspid and slightly curved teeth (Text-fig. 33 also see p. 27 and p. 61 below). With these teeth H. chilotes is able to extract the larvae of a boring mayfly (Povilla adusta) from their burrows ; prawns are also probably removed in the same way from cracks and holes in rocks. The hypertrophied lips (which are highly variable in their degree of development) seem to serve merely as shock absorbers when the fish forces its mouth against the rock or wood face ; histological examination of lip tissue does not show any particular increase in the number of sensory cells as compared to lip tissue from other species. Free-living insect larvae are also eaten by H. chilotes, and presumably taken, like the food of less specialized species, from the lake bottom. Another species, H. chromogynos, has teeth and jaws closely similar to those of H. chilotes (see Text-fig. 33), but there is no suggestion from its known food (larval Diptera) of specialized feeding habits. Next to insects, fishes are probably the most important source of food for Haplo- chromis (Greenwood, 1962, 1905a, 1967). I have estimated (Greenwood, 1959¢, 1967) that about 40 per cent of Haplochromis species in Lake Victoria are piscivorous predators. With more knowledge of deep water species (Greenwood & Gee, 1969 ; unpublished observations) I would now reduce that figure to one of about 30 per cent, still a high proportion. Most of the fishes eaten by the piscivores are other Haplochromis, although small cyprinids (especially Engraulicypris argenteus) constitute part of the diet in the larger predators. ‘The fish remains recovered from a predator’s guts are so macerated that it is generally impossible to identify the species of the Haplochromis prey consumed. Generic identification, however, is certain. No Tilapia remains have ever been iden- tified with certainty ; probably this is because the young TiJapza (of both species) are ‘protected’ in their nursery zones (where Haplochromts species do not penetrate) until they reach a size where they are too large to be eaten by a predatory Haplo- chromis. The clearly recognizable ‘predatory’ facies of fish-eating Haplochromis (see Text- figs 3, 12-15 and 52-63) was noted before. There are, however, a few species which superficially and even in their dentition and mouth size could well be mistaken for generalized insectivores (see Text-figs 48 and 50). The principal components of the ‘predatory facies’ are a large, moderately distensible mouth, strong unicuspid jaw _-teeth (Text-fig. 4B), and pharyngeal teeth, also strong, with the crowns so orientated and shaped as to form a macerating mill. Some piscivorous Haplochromis are -among the largest in the lake, reaching, when adult, lengths of between 200 and 300 mm. __ Also to be included in the piscivore category is a group of eight species that feed exclusively (or principally) on the embryos and yolk-sac larvae of other cichlids (Greenwood, 1959b, 1967). The feeding habits are definitely established for six of P. H. GREENWOOD OS aes he Fic. 9. Haplochromis pallidus, an insectivore. (Natural size.) Fic. 10. Haplochromis chilotes, a specialized insectivore ; see p. 31. (Natural size.) EVOLUTION. OF A CICHLID SPECIES FLOCK 33 these species, strongly indicated for the seventh and assumed on anatomical grounds, for the eighth species. Apparently none of these paedophagous species preys on the embryos or larvae of non-cichlid fishes. This is not surprising when one recalls that most non-cichlids breed outside the lake (Greenwood, 1966 ; Corbet, 1960) and then only biannually. Cichlids, on the other hand, spawn throughout the year (see p. 51). However, all the Lake Victoria Haplochromis (and other cichlids) for which data are available brood the embryos and larvae in the female parent’s mouth ; there are no indications whatsoever that any cichlid in the lake does not practise mouth-brooding. One might reasonably conclude, therefore, that the paedophages have evolved some method of obtaining their food from the mouth of a brooding female (but see below). Two lineages are involved in this trophic specialization (see p. 75), with both showing a certain degree of convergence (pace Fryer & Iles, 1972: 101). For in- stance, there is a marked trend towards increased distensibility and protrusibility of the mouth, and a tendency for the teeth to be buried in a thickened oral mucosa (so much so that superficially the more specialized species in each lineage appear to be edentulous). The least specialized species in each lineage, H. cronus (Text-fig. 16) and H. barbarae (Text-fig. 29), differ little from the trophically generalized species at their particular level of overall specialization (see Greenwood, 1959b and 1967 for the species respectively). That is, H. cronus resembles certain species in the benthic omnivore-insectivore grade, and H. barbarae species in the near basal piscivore grade. Precisely how the paedophagous predators obtain their food has not been estab- lished. The turbid water of Lake Victoria effectively prevents underwater observa- tions, and aquarium studies have yielded equivocal results. In an aquarium #H. parvidens (Text-fig. 18) did show behavioural patterns suggesting that it might engulf the snout of a brooding female, and in this way force it to disgorge its brood. Cer- tainly the very distensible mouth characterizing all but the two most generalized paedophages lends support to the idea of snout-engulfing. The deeply embedded teeth also give support to this idea, since hidden in this way there would be less chance of the attacker becoming too firmly attached to the brooding female. If the predator has to struggle to free itself there is a danger of it losing its meal in the process. On the other hand, a greatly distensible mouth could be an adaptation for sucking in the small, dispersing prey should the mother be induced by some other means to jettison her brood. In either eventuality it must be emphasized that brooding female H aplochromis are extremely tenacious of their broods, even when under great stress, like capture in a seine-net, handling after CONC or being chased around an aquarium (personal observations). Fryer & Iles (1972) do not think it likely that paedophagous species get their food by directly attacking or threatening the parent fish. Instead, they believe that the predators rely on brooding females jettisoning a brood without direct interference from external sources. That females should voluntarily abandon their broods is thought by Fryer & Iles to be one of several homeostatic population density control mechanisms operating within the flock. 3 34 Fic. 11. Haplochromis serranus, a piscivore. hiGs 12 HIG: 13; P. He GREENWOOD Haplochromis spekii, a piscivore. Haplochromis prognathus, a piscivore. (Half natural size.) (Half natural size.) — EVOLUTION OF A CICHLITD SPECIES FLOCK 35 Fic. 14. Haplochvomis estoy, a piscivore. (Half natural size.) Fig. 15. Haplochromis mento, a piscivore. (Half natural size.) Fic. 16. Haplochvomis cronus, a paedophage ; see p. 33. P. H. GREENWOOD Fic. 18. Haplochromis parvidens, a paedophage. (Half natural size.) CSE Z Xela 1 EONS aie i SS Fic. 19. Haplochromis sauvagei, a mollusc eater (oral sheller, see p. 69). (Natural size.) EVOLUTION OF: A CICHLID SPECIES FLOCK 377 With no direct evidence to support either suggested mode of food gathering, further speculation is pointless. However, I cannot agree with some of the evidence (or interpretation of observations) that Fryer & Iles (op. cit. : 101) use in support of their thesis for homeostatically induced jettisoning of the brood. As this evidence is concerned mainly with anatomical features and questions of physiology, discussion is deferred to a later page (p. 75 ef seq.). As to the origin of paedophagy, I would agree with Fryer & Iles (op. cit.) that it, like other trophic specializations, stemmed from an originally facultative habit later selected for, and further developed by, the evolution of more specific anatomical specializations. Certainly the existing ‘basic’ paedophages, H. cronus and 4H. barbarae, could not be excluded from other trophic groups on anatomical grounds alone. Haplochromis species feeding on Mollusca form a fairly well-defined group which does, however, intergrade with the insectivorous one ; some species have a mixed insect and mollusc diet. . Ten species can be classed as predominantly molluscivorous. In this group there is a clear-cut dichotomy into species that crush the prey in the pharyngeal mill (six species) and those either wrenching the snail from its shell or crushing the shell be- tween the jaws before swallowing the soft parts (four species). It will be recalled that at least two monotypic genera (Hoplotilapia and Macropleurodus) belong to the second group, a third (Platytaeniodus) may do so, and that a fourth (A statoreochromis) belongs to the first category. Those species crushing or dismembering molluscs orally form a distinct phyletic assemblage (Greenwood, 1957; and p. 69 et seg. below) which includes one of the monotypic genera, Macropleurodus bicolor. Snails are the chief food organisms of this group. The species that crush molluscs (both bivalves and gastropods) intrapharyngeally are of diverse relationships (see p. 72). A common feature in all, however, is some degree of hypertrophy in the pharyngeal bones, particularly the lower one, and some molarization of the dentition on these bones (again, particularly the lower one, see Text-fig. 5). In the virtually exclusive mollusc eaters (H. ishmaeli and H. pharyn- gomylus, Text figs 3 and 23), these bones and their dentition reach the peak of development seen in this trophic group (see Text-fig. 5 and Greenwood, 1960). Other members of the group (e.g. H. theliodon, H. humilior [Text-fig. 22], H. riponianus [Text-fig. 42] and H. obtusidens) have the lower pharyngeal bone but slightly stouter than that in generalized species ; usually only the median tooth rows show any molarization. It is these species which, to a certain extent, are the mixed mollusc- insect eaters, and in which bivalves predominate over gastropods in the diet (Green- wood, 1960). In terms of a restricted diet, and in associated anatomical specializations, the mollusc-eating Haplochromis (particularly the pharyngeal crushers) clearly show species at every major stage in the development of the trend (Text-fig. 5A, B,C). It seems likely that the anatomical ability to crush mollusc shells has evolved, indepen- dently, on a number of occasions. But at least one phyletic line can be detected, with representatives of its insect—-mollusc eating species, bivalve-gastropod eating species and almost exclusively gastropod eating species, still extant (see pp. 69-75 below). 38 FIG. 20. P. H. GREENWOOD Haplochromis granti, a mollusc eater (oral sheller). (About natural size.) ea any ryt a ae Ss ™ >, SAE (Natural size.) FIGs 22° Haplochromis humilioy, a mixed mollusc—insect eater, the molluscs are crushed in the pharynx. (Natural size.) EVOLUTION :OF .A{ CICHLID SPECIES FLOCK 39 Small Crustacea, especially ostracods and, to a lesser extent, prawns of the genus Caridina, contribute to the diet of several Haflochromis species in the insectivore and the mollusc-insectivorous trophic groups. There is, however, a small group of five species that appear to feed equally on insects and Caridina (Greenwood & Gee, 1969; Greenwood, 1967 for H. tridens). All are bottom feeders and most occur in water 20-35 m deep (with one species extending into shallow water). Almost certainly more species will be added to this group when undescribed material from the deep waters of the lake is analysed. No obviously adaptive anatomical features are associated with the insect—crustacean diet ; all the species in this group, with one possible exception, appear to be closely related (Greenwood & Gee, I969 ; and p. 67 below; Text-figs 24 and 25). Plant material (especially algae) is frequently found in the guts of bottom feeders whose food is otherwise of animal origin. The nutritive value of the plant material in the diet of such species is unknown ; usually the quantity ingested is small and rarely does it appear to be digested completely. Eight species, however, show the characteristically long, much coiled intestine and the highly distensible stomach, associated with a truly vegetarian diet. The gut contents of all eight species confirm their placement in this trophic category. Once again, there are indications from the unstudied material at my disposal that the number of phytophagous species will be increased eventually. The plant eating species can be split into three subgroups: those feeding on epi- phytic and epilithic algae, those feeding on phytoplankton and those eating macro- phytic plants. Since only one species (H. phytophagus ; Greenwood, 1966b) is known from the latter subgroup it can be discussed briefly. Haplochromis phytophagus has the overall structural and dental characters of a generalized species, except that the teeth (both oral and pharyngeal) seem adapted for biting off and then macerating the relatively tough tissues of higher plants. Analyses of gut contents suggest, however, that the principal source of food is not this tissue itself, but the associated flora of attached diatoms. Only the ruptured leaf and stem cells appear to be digested. The four species feeding directly on epilithic and epiphytic algae (Greenwood, 1950b) obtain their food in a different way from that of H. phyiophagus. They graze the algae from the substrate and little higher plant tissue is ingested (and, as in H. phytophagus, this is poorly digested). One of the four species (H. nigricans ; Greenwood, op. cit.) is predominantly a grazer on epilithic algae and is rarely caught away from rocks and rocky shore lines. The other three species (H. obliquidens, H. lividus and H. nuchisquamulatus) some- times extend into the habitat range of H. nigricans and also graze off rocks. Usually, however, these species are found in or near stands of rooted plants. All four species feeding on periphyton show some dental modifications (Text-figs AD-G and 36), with H. obliqguidens the most obviously specialized in this respect (Text-fig. 4G). The phyletic-structural story here is interesting because, despite the different degrees of specialization in tooth shape (from near generalized in H. nuchisquamulatus, through H. lividus to highly specialized in H. obliquidens), there is apparently little difference in their effectiveness as scrapers. The teeth of H. P. H. GREENWOOD 40 (Natural size.) (pharyngeal crusher). a mollusc eater Haplochromis ishmaelt, ZZ Fic. (| \ \ \ \ \ il { \ WNIT ———— —S _——= SS SS = SS SSS = SSS, ———— Haplochromis dolichorhynchus, a benthic crustacean eater. 1G.24: F is tyrianthinus, a benthic crustacean eater. Haplochrom Fic. 25. EVOLUTION OF A CICHLID SPECIES FLOCK 41 nigricans (Text-fig. 36) are quite different, and probably are associated with its diet of epilithic periphyton. The three phytoplankton feeders (H. erythrocephalus [Text-fig. 26], H. paropius and Hf. cinctus | Text-fig. 27]) are in all respects, except the long gut, typical members of the structurally generalized species groups. None shows any obvious superficial characters correlated with feeding on finely particulate matter. The gill rakers, for example, are neither longer nor more closely arranged than in an insectivore (Text- fig. 6), although in HZ. erythrocephalus there is a slight increase in number (12 or 13 cf 8-10 for the other phytoplankton feeders and most insectivores). Judging from the nature of the algal and other material found in the stomachs of these fishes, food is taken from the bottom and not while it is in suspension (Green- wood & Gee, 1969). Finally, there is the enigmatic H. acidens (Text-fig. 28), an apparent phytophage with a distinctly predatory facies. The gut contents of the 16 fishes examined (from different localities) consist mainly of finely macerated phanerogam tissue, but also a few insect remains, some bones of small fishes and in one specimen the soft parts of a mollusc. The intestine of H. acidens is proportionately somewhat longer than that in omnivores, but is shorter than in other vegetarian species. To confuse the issue further, the jaws, skull and teeth of this species are typically those of a piscivorous predator (see Text-fig. 37; and Greenwood, 1967; also p. 67 below). It was with a piscivorous lineage that I originally suggested H. acidens be related (Greenwood, op. cit.). Fryer & Iles (1972) call H. acidens an ‘. . . adventurous species, formerly pisci- vorous, which made experimental sorties into plant eating . . . an almost vacant niche in Lake Victoria’. They go on to note that such a trophic shift would be one from a way of life in which specialization was at a premium to one in which the reverse is true, thus explaining what might be considered ‘imperfections’ for phytophagous habits in the jaws and teeth of H. acidens. This idea is intriguing but it must be remembered that there are trophically generalized species not far removed anatomi- cally from the basal species in the piscivore radiation with which H. acidens was first associated (but see p. 67 below). Recent studies on this species, however, indicate that it could as well be related to a generalized and omnivorous species in a lineage that produced no predators but which did produce another herbivore (H. erythro- cephalus). The ‘predatory facies’ of H. acidens is thus all the more difficult to explain. | The last trophic group to be considered comprises a single species, H. welcommet (Text-figs 3 and 39), the only known lepidophagous species in the lake. The guts of all specimens examined, contained numerous small fish scales (like those covering the caudal fin base of Haplochromis), together with a few fragmentary fin rays and bits of skin. (The number of guts checked is now considerably greater than when the species was first described [Greenwood, 1966b], but is still too small for satisfactory generalizations to be drawn.) Scale eating is practised by many cichlid species in Lakes Malawi and Tanganyika (reviews in Greenwood, 1966b and Fryer & Iles, 1972). Some of these species, like H. welcommei, scrape the caudal fin squamation from other cichlids ; others nip 42 = OO WAAR ie ue DES SOOT Nei \': ' > of) 4 y Ai \o, 1 Dice 3 Sas YY asi) IRA Cc) 4 f “5 4 Bp Fie. 28." i aplochromis acidens, probably a feeder on macrophytes, see p. 41. EVOLUTION OF A CICHLID SPECIES FLOCK 43 scales from the body of a variety of fishes. The diet of H. welcommez is, however, not entirely one of scales and fin rays ; most guts also contain a quantity of the dia- tom Melosiva, and in some there are fragments of macerated plant epidermis. This plant material suggests that H. welcommei may also feed by grazing on the epiphytic flora of rooted plants, for which habit its dentition would seem well adapted (Text- fig. 39 ; also p. 69 below). If this is so, then the species shares a trophic niche with H. obliquidens and the other grazers (see above, p. 39). Interestingly, one of the Lake Malawi lepidophages, Genyochromis mento, also feeds on periphyton (Fryer, 1959) which it browses rather than grazes from rock surfaces. Fryer (op. cit.) suggests that G. mento could have evolved from the same specialized stem as some of the epilithic algal grazers in that lake. Haplochromis welcommet, on the other hand, does not seem to be at all closely related to the grazing species of Lake Victoria. Its relationships are apparently with the benthic, crustacean-eating species discussed on p. 67 (see also Greenwood, 1966b). There are still several of the described Haplochromis species whose feeding habits are unknown, or for which so few and contradictory data are available that the species cannot be put in any particular trophic category. Furthermore, there are at least thirty undescribed species collected recently from several deepwater habitats in the centre lake region. Preliminary studies indicate that most belong to the insectivore and crustacean-eating groups, with two or three other species referable to the piscivores. The absence of a pelagic zooplankton eating Haplochromis species has already been commented upon. No reason for this obvious gap in the trophic radiation is im- mediately apparent. There is, of course, a non-cichlid Engraulicypris argenteus occupying this niche, but then there are non-cichlid species in all the other trophic niches occupied by Haplochromis species living in the same habitat. Possibly there isa heavy predation of zooplankton by young fishes (and other animals) in the littoral and sublittoral lake regions, so that the niche is, in fact, fully exploited. This purely speculative suggestion would not seem to apply to the offshore, open-water areas of the lake (except for at least their partial occupancy by shoals of E. argenteus ; see Graham, 1929). Here again one is hampered by a lack of information on zooplank- ton density and distribution, and by a complete ignorance of what trophic categories of cichlids there are in the pelagic and midwater zones of the lake. Before moving on to consider other aspects of Haplochromis biology one should recall that the adaptive radiation described above has taken place in a single genus. To put this phenomenon in perspective one should also remember that in any major tropical marine biotope (say a coral reef) there might be as much trophic diversity among its fishes, but several phyletically distinct families would be involved (Green- wood, 1965a). Particularly instructive in this respect, since it involves a lacustrine community (albeit a cold temperate one), is the work by Keast & Webb (1966) on the feeding habits of fishes in Lake Opinicon, Ontario. These authors analysed the mouth and body forms, and the feeding habits of 14 species belonging to seven families (represent- ing, incidentally, five orders). In terms of their overall function and gross mor- phology (except fin position) the characteristics of these different taxa are all 44 P. H. GREENWOOD apparent in various Lake Victoria Haplochromis species. - What is more, the Haplo- chromts have exploited more trophic niches than have the fishes of Lake Opinicon. Keast & Webb (of. cit.) conclude that mouth and body structure combine with food specializations and habitat preferences to reduce greatly interspecific competition. By contrast, in Lake Victoria it is usual for more than one Haplochromis species to occupy the same trophic niche in the same habitat. There is, however, a similarity in the two lake situations, for the Hapflochromis species, like the fishes of Lake Opinicon, retain a measure of flexibility in their feeding habits (see data in papers by Greenwood, 1956-69). Keast & Webb (op. cit. : 1871) consider such flexibility an important feature for survival in cold temperature lakes ; it may well be equally important in warm tropical lakes, and have played a part in the evolution of the Haplochromis species flock (see p. 115 below). The Lake Victoria Haplochromis are, of course, not unique in these respects. The Lake Malawi species are even more adaptively multiradiate, and occupy some niches not even represented among the Victoria flock (although paedophagy seems - to be a unique feature of that flock). For a detailed account of the situation in Lake Malawi the reader is referred to the excellent account given by Fryer & Iles (1972). Intralacustrine distribution of the Haplochromis species When considering the distribution, habitats and vertical range (i.e. depth distribu- tion) of the Lake Victoria Haplochromis species the same restrictions must be applied as were applied to data on feeding habits (see p. 30 above). Furthermore, the murky waters of the lake (transparency 1:3-8-2 m, depending on habitat, lowest in sheltered bays) have greatly impeded field observations on the fishes. Thus, information on the distribution of a species within a major habitat has to be got indirectly, and its precision is largely dependent on the type of collecting gear used. For example, a trawl or seine net could pass through clumps of plants, could traverse several types of bottom and could take fishes at a variety of depths. Lack of rigorously quanti- fiable data on the number of individuals in any one habitat is also a hindrance in assessing species’ distributions and in determining habitat preferences. Major habitat types within the lake tend to intergrade with one another, but are recognizable as: (i) Exposed beaches dit some wave action. The substrate here is generally sandy, with small rocks and rocky outcrops ; some rooted plants occur, and there are often marginal stands of emergent vegetation. The sandy substrate extends offshore for variable distances, grading gently into the organic mud substrate charac- teristic of sheltered bays and the deeper parts of the lake. (ii) Sheltered bays, usually fringed with papyrus swamps ; often the inshore end of a bay is an extensive papyrus swamp. In most sheltered bays the bottom is composed of organic mud extending right up to the marginal papyrus ; sometimes, however, there are narrow sand beaches, particularly if the papyrus margin is in- terrupted. (111) and (iv) Sub-littoral regions (water depth 6-20 m), and deepwater regions (more than 20m deep). These are rather more artificial than natural categories, and are used in order to divide the offshore zones into ‘shallower’ and ‘deeper’ es EVOLUTION::OF A CICHLID SPECIES FLOCK 45 waters. The sublittoral substrate is either organic mud, shingle, or mud with low outcrops of lateritic rock that extend for several metres as flat sheets rising one or two metres above the prevailing substrate. This zone grades imperceptibly into the ‘deepwater’-one where the substrata are also similar. Islands of various sizes (from Buvuma island, about the size of the Isle of Wight, to mere rock pinnacles, ca 10-20 m across) are common in the sublittoral zone. Depending on the island’s size, its shoreline can be as varied as that of the main- land. Larger islands also provide the chief areas of truly rocky habitats. (v) Coastal littoral other than sandy beaches. Apart from a few regions of the lake where there is a rocky coastline, much of this littoral region is adjacent to fringing papyrus and often includes a water-lily swamp zone. The bottom can be hard (sand or shingle) or soft (organic mud). The papyrus swamps (even narrow fringing ones) rarely harbour cichlids, except at the perimeter where, unlike water further into the swamp, the dissolved oxygen level is adequate for these fishes. Otherwise the chief occupants of such swamps are air-breathing non-cichlids (see . LI). The large gulfs and sounds (e.g. Kavirondo Gulf, Napoleon Gulf, Smith Sound) are not in themselves particular habitats ; each contains most of the habitats detailed above. An indication of the number of Haplochromis species occupying a particular habitat can be derived from an exposed beach near Jinja (Nasu Point beach) and a sheltered, papyrus fringed bay about 8 km south of Jinja. In both places I was able to carry out regular sampling over a period of six years. The beach habitat is occupied by 35 Haplochromis species, the bay by 21 species. In the former habitat few non-cichlids (Barbus spp. and Clarias mossambicus) appear to be regular inhabitants, and then only in very small numbers. Occasion- ally, however, the beach was invaded by large shoals of the characin Alestes sadlert, or by shoals of juvenile Tilapia variabilis (otherwise not a regular component of the cichlid faunas in that habitat). Ekunu Bay, by contrast, carries a varied non-cichlid fauna including Protopterus aethiopicus, Synodontis victoriae, Clarias mossambicus, Bagrus docmac, Mormyrus kannume, Gnathonemus spp., Labeo victortanus and, occasionally, the large Barbus altianalis. Both Tilapia esculenta and T. variabilis are resident inhabitants. The two habitats differ quite markedly in the species of Haplochromis present ; probably not more than two species are shared, apart from the ubiquitous A statoreo- chromis alluaud1. There are differences too in the trophic specializations shown by the Haplochromis of the two habitats. In this respect Nasu Point beach is somewhat more varied, but there is a higher proportion of piscivorous predators in Ekunu Bay (13 of 21 Haplochromis species, cf 13 of 35 at Nasu Point). Surveys made in other beach and bay habitats seem to confirm the impression of greater diversity of Haplochromis species from beaches, and the virtual exclusion of non-cichlids from this habitat. Some idea of population density in the Haplochromis species occupying a beach and its sandy sublittoral region can be gained from the commercial catches made in these areas. Over the period 1951-58 the average annual catch from seine nets 40 P. H. GREENWOOD operated at a beach near Majita (Tanzania) was 2-5 million individuals ; the highest annual seine net catch ever recorded, also from a beach in Tanzanian waters (at Nyamwikumulu in 1957), was 7 589 599 individuals in a total of 317 hauls (figures from the Lake Victoria Fisheries Service Annual Reports). Both these figures in- clude the monotypic genera (which were not distinguished from Haplochromts by the counters), but these species contribute only a small fraction of the whole (personal observations on these beaches). For comparison, the catch of non-cichlid fishes in the same number of seine net hauls at the same beach, Nyamwikumulu, in 1957 was 10 974. 3 As far as I can detect, the nature of the substrate is the major factor influencing the distribution of many Haplochromis species, at least within the depth limits of the various species (see below). Some species are virtually confined to habitats (or parts of a habitat) where the bottom is muddy. Others are restricted to places where the substrate is hard and composed of sand, rock or shingle. Unfortunately, the data available are not sufficiently refined to be able to tell if the exact nature of the substrate, that is its constitution, or its associated flora and fauna, has the more particular influence on species distribution. Of the 69 species apparently restricted to water less than 20 m deep (i.e. inshore Species), 27 are found only over hard substrates, and five only over organic mud or other soft substrates. The remaining 37 species have been caught over both hard and soft bottoms, but it must be stressed that of these species, nine rarely occur over soft substrates and seven are rare over hard bottoms. In other words, 32 of the 69 species have a restricted substrate preference, and 21 occur over either substrate type with sufficient frequency for one to consider them as being free from substrate restraint (data from Greenwood, 1956-69). Correlating specific substrate preferences with the trophic groups to which the species belong is difficult, and because of insufficient data also rather subjective and imprecise. Some species in all trophic groups, except the algal grazers, show no marked substrate preferences. Molluscivorous species of the group crushing gas- tropods orally (or wrenching them from the shell) are virtually confined to hard substrata, as are the paedophagous species (see pp. 31-37 above). All other trophic types are found over every type of substrate. Likewise it seems that representatives of all trophic categories are found in each of the major habitats. However, the algal grazers and the species feeding on higher plants (p. 39 above) are restricted to habitats or parts of a habitat where there are rooted plants or, in the case of the grazers, suitable substrata for algal growth. Haplochroms nigricans of this group seems to have the most restricted distribution, rarely being found far from rock surfaces. The paedophages as a group are relatively restricted to littoral habitats, probably because it is mainly in these regions that brooding female Haflochromis are concentrated. Particular interest attaches to six specimens of H. obesus (or a taxon anatomically very like that species) caught over a mud bottom at a depth of 27-30 m (Greenwood & Gee, 1969). Collections of Haplochromis from this and similar habitats rarely produce brooding females (or individuals showing the characteristic buccal distortion that goes with brooding and which persists for a short while after the young have been jettisoned). Thus it is Se EVOLUTION. OF A CICHLID SPECIES FLOCK 47 especially noteworthy that these deep-living H. obesus had all fed exclusively on cladocerans and copepods. It is clear from field observations that more than one Haplochromis species of a particular trophic group can be found in any one habitat. The extent of such inter- specific overlap, and hence possible trophic competition, may be less than is implied by the mere record of several species seeming to feed on the same food source. Two algal grazers, H. lividus and H. obligudens, occur together in the sandy beach habitat ; but the former species extends its range into deeper water than does H. obliquidens. Haplochromis mgricans, with similar food requirements, also occurs in this habitat but it feeds principally from submerged rocks and stones. A similar situation exists with two mollusc eaters in this habitat, H. sauvage and H. prodromus, where H. prodromus extends into the sublittoral beach zone. Further- more, the two species have different feeding methods, H. sauvagez levering the snail from its shell, and H. prodvomus crushing the shell between its jaws. These dif- ferences may influence the selection of prey species by the predators. The monotypic genus Macropleurodus bicolor also enters the picture since its habitat range overlaps in part those of H. sauvage: and H. prodromus, being very like that of the latter species (Greenwood, 1956a, 1957). Its feeding habits, too, are more like those of H. prodromus, although it is known to feed on insects as well as snails (Green- wood, op. cit.), and probably to a greater extent than H. sauvager, also a mixed feeder. The whole situation is further complicated by the presence of two other mollusc eaters of the same trophic complex, H. granti and H. xenognathus (see p. 69 et seq.). Both these species occur in the habitat under discussion, and both share the same food sources and intrahabitat preferences as the species already considered. Nor are these the only mollusc eaters present. Haplochromis humilior and H. pharyngomylus also live in beach habitats (Greenwood, 1960). These two species, unlike the others, crush their prey in the pharyngeal mill; the food of H. pharyngomylus (chiefly the snail Melanoides tuberculata) is virtually identical with that of H. prodromus, H. xenognathus, H. granti and M. bicolor. The food of H. humilior, a smaller fish, overlaps that of these species but does include a greater proportion of other organisms, especially insects (Greenwood, 1960). Other partly molluscivorous species could be mentioned but I think the problem is sufficiently obvious without introducing additional complexities. An equally complicated picture can be compiled for the insectivorous species in a habitat, and to a lesser extent for the phytoplankton eating species as well. The piscivorous predators provide perhaps the strongest prima facie case for interspecific trophic congruence ; yet even here interspecific competition cannot definitely be established from the data available. In the habitats I have studied there seems always to be a group of piscivorous species with completely overlapping ranges, as well as a few species showing a restricted distribution. All these fishes feed on other Haflochromis species. Since it is impossible to identify prey recovered from the guts to more than the generic level, it is equally impossible to tell what degree of competition (if any) there is for prey species. On a subjective impression there would seem to be a sufficiency of prey species for competition between the 48 P. H. GREENWOOD predator species not to develop, and the situation is further ameliorated by some of these having a limited intrahabitat distribution. The extent to which some species may move between habitats, or from one type of substrate to another, has not been established. Thus I would be chary of sug- gesting, as have Fryer & Iles (1972 : 308), albeit tentatively, that in Lake Victoria ‘.. . It seems probable that some species will at times utilize this ability, (to shift habitat), at least as a temporary expedient for avoiding adverse conditions’. I would, however, agree with the general conclusion reached by these authors, that in Lake Victoria there is, as compared to Lake Malawi, less obvious stenotopy amongst members of the cichlid species flock (always bearing in mind that far less is known about the generality of Haplochromis species in Lake Malawi than about the ‘Mbuna’ generic complex and the few rock-haunting Haplochromis species that coexist with them, a point not explicitly made by Fryer & Iles). None of the Lake Victoria Hapflochromis or related species shows any geographical restriction within the lake, nor have any morphologically distinguishable populations been discovered. In both these respects the Victoria species differ from those of Lake Malawi, and especially Lake Tanganyika (see summary in Fryer & Iles, 1972). The question of interspecific overlap in feeding habits and habitat requirements among Lake Victoria Haplochromis species has direct bearing on the principle of “competitive exclusion’ (the so-called Gause’s principle, but see Mayr, 1963). This postulated that no two species can exist at the same locality if they have identical ecological requirements. As Wynne-Edwards (1962) has demonstrated, the principle is logically and ecologically unsound. If two species have truly identical require- ments they will be, in effect, ecologically one species. Competition will then be on an intraspecific rather than an interspecific level ; that is, between individuals. From what is known about the inshore Haplochromis species considered so far, there are relatively few, if any, aggregates with absolutely identical ecological re- quirements (i.e. a condominium, sensu Wynne-Edwards). Instead there are many species with a certain degree of overlap. Following the spirit of Professor Wynne- Edward’s terminology, perhaps such groups should be thought of as a ‘common- weal’? The vertical movement of inshore Haplochromis species is very poorly known, and their apparent restriction to water less than 20 m deep may prove illusionary. Taking the latter point first, most available depth records were obtained before the deeper, offshore areas of the lake were studied intensively. In recent years there has been a great deal of exploratory trawling in these areas. To a certain extent this research has confirmed the earlier depth distribution records, but it has also provided some surprises. The extended range of H. obesus has already been noted (p. 46 above). During a recent trawl survey a haul made over sand at a depth of ca 50 m in midlake (opposite Entebbe) contained, in addition to deepwater species, several other species that were previously known only from inshore beach habitats (personal observations). | Little can be said about the movement of inshore species through the water — column. The indications are that species in this zone remain near the bottom, at | least during daylight (personal observations). In water less than 2 or 3m deep, | EVOLUTION: OF A CICHLID SPECIES FLOCK 49 many species may range from surface to bottom, but in progressively deeper water their behaviour pattern becomes increasingly benthonic. Judging from the presence of bottom debris associated with food in the guts, most species feed on or near the bottom. For the piscivorous species, such indirect evidence is not available; the occasional field observation in clear water suggest that piscivores range rather freely through the water column, at least in shallow parts of the lake. As yet the taxonomy of deepwater species is very incompletely known, and the ecology of even the described species is still imperfectly documented (Greenwood & Gee, 1969). Apart from a few inshore species whose range extends beyond ca 20 m, 17 described species are known from deeper waters (Greenwood & Gee, op. cit. ; Greenwood, 1967). I estimate that at least 30 more species will be described from collections already obtained. Only two species, H. dolichorhynchus and H. ery- throcephalus, have been recorded from depths less than 15m; these species have depth ranges of 10-30 and 10-35 m respectively. Depth ranges for the remaining species are between 17 and 70m. Most have a wide range within the limits of 25- 35 m, but there are indications of some species being restricted to depths near the upper and lower limits of the total range. None extends into the truly littoral zone, but several species may live at depths greater than those recorded here (personal observations). Deepwater species are necessarily confined to the strictly offshore regions of the lake, but at least three species (fH. erythrocephalus, H. dolichorynchus and H. crypto- gramma) have been caught in the deeper areas of large bays (Greenwood & Gee, 1969). It is impossible to define habitats in the deeper water, and the collecting gear employed there (trawl nets) means that several habitats may be sampled in one trawling period (half an hour at a speed of about 2 knots). That all but two of the known species were caught over a soft bottom is probably not a true reflection of the species’ substrate preferences ; more trawling has been done over mud than over hard substrata. Nor, because of limits imposed by the collecting methods, is it possible to determine the species structure of any particular area. All one can say is that several species are caught in a trawl, that these species will be from several trophic groups, and that any particular trophic group is usually represented by more than one species. In other words, a replication of the situation in shallower waters of the littoral and sublittoral. The structute of deepwater communities may, however, be simpler because certain trophic groups are not represented there. So far, insectivores, bottom detritus feeders, predators on small Crustacea and piscivorous predators have been identified (Greenwood & Gee, 1969). No mollusc eaters have been found, and the sole paedo- phage (or presumed paedophage, H. obesus, see p. 46 above) had fed on Crustacea. These remarks must, however, be qualified by reference to the ‘inshore’ species complex caught in midlake (see p. 48); at least two mollusc eaters (including Hoplotilapia retrodens) were amongst the species then caught. The apparent paradox of that deep-living ‘shallow water’ community shows clearly the difficulty of generalizing about the ecology of deepwater regions in the lake. If sandy bottom, inshore species can adapt to deep water, why is it that none of the 4 50 P. -H. GREENWOOD soft bottom inshore species has done likewise, especially since the nature of the soft substrata in both regions is apparently identical? Much more will have to be learned about deepwater fishes, and particularly if there are species confined to sand at those depths, before an attempt can be made to answer this question. It is assumed from the evidence available (par ticularly the nature and composition of gut contents) that all deepwater species are benthic in habit (Greenwood & Gee, 1969). One species, H. laparogramma, is suspected, however, of feeding away from the bottom (Greenwood & Gee, op. cit.), and other members of the insect—crustacean eating trophic group may have similar habits, or at least be facultative in this respect because their prey are not necessarily confined to the benthos. Echo-sounding has demonstrated the presence of fish in midwater at a variety of depths (Gee, 1968). So far it has been impossible to identify these fishes ; that they may be species of Haplochromts is indicated by some rather inconclusive midwater experimental trawling. (Personal communication from staff members of the E-ALE.E-R:-O., 1071.) No counts are available for the number of individual fish caught per trawl in deep waters. An estimate by eye suggests that the density of Haplochromis in these waters is not significantly less than from inshore habitats (see p. 46 above). Brief mention may be made here to the habit of shoaling. In Lake Malawi several of the offshore, zooplankton-eating Haplochromis species (the ‘Utaka’ group) exhibit shoaling behaviour, as does the rock frequenting H. kiwimgi during the zooplankton-eating phase of its life cycle (see Fryer & Iles, 1972). The Lake Victoria Haplochromis species do not appear to furnish a single example of shoaling, at least during adult life. Since adult shoaling is associated with feeding on suspended, particulate food (zooplankton or phytoplankton) the absence of this behavioural trait is not altogether surprising. However, it should be remembered that direct observation on fishes in Lake Victoria is well nigh impossible, and the recognition of a shoal would be from the extreme abundance of a particular species inacatch. Thus, with trawling, when several thousand individuals are caught from over a large area, a small shoal could pass undetected in the catch. Species of the shoaling ‘Utaka’ Haplochromis group in Lake Malawi have charac- teristic ‘shoaling marks’, in the form of spots, or a well-defined stripe, on the flanks. Few Lake Victoria species show such coloration, at least as fixed patterns (see p. 52). Of the species that do, H. martini and H. michael: (lateral stripe) provide no evidence for shoaling, and neither species is a plankton eater. The possibility remains open for two deepwater species with marked patterns (H. laparogramma [a lateral stripe] and H. cryptogramma [an interrupted lateral stripe]). Both species are caught only in trawls, thus making an estimate of their relative abundance in a catch particularly difficult. Since H. cryptogramma feeds on the pupae of chironomid flies (which occur in dense aggregates), and seems to feed away from the bottom, it is the most likely suspect for shoaling behaviour. Breeding biology of the Haplochromis species To conclude this review of Haplochromis bionomics the rather scattered informa- tion on their breeding biology will be brought together and reviewed. | ; EVOLUTION-.OF A. CICHLID SPECIES FLOCK 51 Many species are known to be female mouth brooders. There is no indication, either direct or by implication, that any species does not practise this form of parental care. For the several species where females actually carrying young have not been caught there is indirect evidence of mouth brooding ; namely, spent females with the characteristic deformation of the buccal cavity (a deeply depressed hyoid arch) associated with mouth brooding, and the presence of few but large ova in the mature ovary. These ova also fail to show the modifications to the zona radiata which are invariably associated with substrate spawning habits in the Cichlidae (see Fryer & Iles, 1972 for review). No records are available for the breeding activity in any particular species. However, repeated observations made at a beach near Jinja, and from other localities in northern waters of the lake, strongly suggest that at least part of a population is breeding at any one time. That is to say, for the majority of known inshore species breeding is continuous. (Unpublished personal observations.) Nothing (except indirect evidence for mouth brooding) is known about the reproductive behaviour of species from deeper waters. Regrettably the opaque water of Lake Victoria does not often permit direct observation of the fishes, and even when this is possible it is not easy to identify the species seen. Thus the exact spawning sites for inshore species, or the mode of their spawning, is not known. On occasion I have seen males guarding simple pit nests in the clear sand bottom near stands of emergent swamp grass. If nest building (or at least the preparation of a substrate for oviposition) is a common feature for all species, the question is raised of where do species living over soft substrate spawn. The flocculent organic mud covering so much of the lake floor would seem to be a most unsuitable substrate for this purpose. This leads one to consider the possibility of species which live over soft mud during non-breeding phases moving to areas more suitable for spawning. Brooding females of some of these species are, however, found in the same areas as non- breeding fishes. The same question is raised with regard to the deepwater species. None has been found inshore so presumably breeding takes place offshore, possibly over those areas where the bottom is hard. There are numerous other unanswered problems associated with the breeding habits of these species. For instance, if there is a migration to hard substrata, is there competition for breeding space with species normally resident in such places, or at least conflict between breeding immigrants and feeding residents? Then there is the question of light and vision in deep water. Visual signals and stimuli play an important part in cichlid courtship and reproductive behaviour (see extensive summary of researches in this field given by Fryer & es [1972], and papers by Baerends & Baerends van Roon [1950], Wickler [1966] and Neil [1964]). Light penetration in Lake Victoria is low even in water less than 10 m deep ; it must be even less at greater depths. For the moment we are nowhere near supplying an answer to these problems, and their resolution will be difficult. All known Haflochromis species and the monotypic genera show clearly defined sexually dimorphic coloration. With two exceptions, it is the male fish that has the 52 P. H. GREENWOOD brighter and more colourful livery, even in non-breeding periods. The exceptions are provided by H. dichrourus (Greenwood, 1967) and H. chromogynos (Greenwood, 1959b). In H. dichrourus, although the sexes are differently coloured, the female is more polychromatic than the male (see Greenwood, 1967 ; these observations have since been confirmed and extended by additional specimens). In H. chromogynos, too, there is dimorphic coloration. Males have what may be termed ‘typical male coloration’ but the females are distinctly marked by a black-and-white piebald coloration that is otherwise found as a sex-limited polymorph in females of certain species (where its frequency is never in excess of ca 30 per cent [Greenwood, 1956a and b, and p. 53 below; also discussion in Fryer & Iles, 1972]). The importance of male coloration in species recognition amongst many cichlid genera is now well established (see Baerends & Baerends van Roon, 1950). Thus itis not surprising to find male breeding coloration is, at least to the human eye, species specific. (Personal observations ; see also descriptions in Greenwood, 1956-69.) With the relatively limited chromatophore pigments and interference colours available it is also not surprising to find these differences not always sharply marked. That is, there is not always a gross interspecific difference in coloration or colour pattern. Instead, apart from differences in ground coloration, the specific differ- entiae involve differently coloured suffusions (red, orange, coppery) over various parts of the flanks, belly, chest or head, and differently patterned colour flushes (especially red or pink) on the median fins (particularly the anal and caudal fins). In addition there may be coloured lappets (predominantly red or orange) on the dorsal fin, or these may be colourless (Plate 1). It is difficult to create a word picture for the ground colour of the body. Reduced to basic colours, four principal types are found, viz: blue-grey (the commonest), shades of green and blue (from turquoise to malachite), yellow to yellow brown (with beige as an extreme) and, least common, uniformly black. Ground coloration is most intense on the dorsal half of the body, shading to silver, grey or yellow ventrally ; in other words the fishes are basically countershaded. Most species show patterns of dark vertical bars and horizontal stripes on the body, and various bars and stripes on the head and opercular region (see Text-fig. 8). These patterns are, with few exceptions, under emotional control and seem to play an important part in the fishes’ repertoire of signals (see Wickler, 1964). The exceptional species (e.g. H. percoides [Text-fig. 52A], H. flavipinnis [Text-fig. 52B], H. martim [Text-fig. 50] and H. squamulatus [Text-fig. 49] and some others) have fixed, but individually variable, patterns which are present in both sexes. Species from deepwater habitats also show bright male coloration (Greenwood & Gee, 1969). Indeed, several species have colours not seen among inshore species, PLAGE «1 Above. Haplochromis riponianus. Sexually active male. Note the prominent ‘egg dummies’ on the anal fin. Below. Haplochromis brownae. Adult male, showing almost complete development of breeding coloration. Photographs by André Roth. Bull. Br. nat. 5) BEAL Bain EVOLUTION OF A CICHLID SPECIES FLOCK 53 for example bright purple and pastel green. Although various shades of purple seem to predominate in the species so far described, this colour does not appear to be at all common in the species still awaiting description. Females from all habitats, in sharp contrast to males, are drably coloured, with greyish silver or sandy ground colours predominating (H. chromogynos and H. dichrourus, excepted ; see p. 52 above). Unlike males there are few marked inter- specific differences in female coloration, but slight differences do exist between some species. Cephalic and somatic bars and stripes are developed as in males, and also seem to be under emotional control. A very prominent and characteristic feature in the males of all Haplochromis species (and related genera) are the ocelli or ‘egg dummies’ on the anal fin. From one to several “egg dummies’ are located on the posterior quarter of this fin. Each is a roughly ovoid colour patch, usually yellow or orange-yellow, surrounded by a completely colourless ring of fin membrane. In life this clear zone gives the spot a three dimensionality such that it closely resembles a Haplochromis egg. The func- tional importance (and indeed the name) of these ‘egg dummies’ was discovered by Wickler (1962a and b, 1968). He was able to demonstrate that, after the female has oviposited and taken the ova into her mouth, the male displays his anal fin in such a way as to bring these spots into prominence. The female then attempts to pick up the spots, presumably identifying them as newly deposited ova. Whilst she is in the close proximity of the male’s anal fin, and thus near his genital papilla, he ejaculates (Wickler, of. cit.). Clouds of sperm are then drawn into the female’s mouth, and the ova are fertilized. In other words, these fishes have evolved a kind of secondary internal fertilization. “Egg dummies’ have not been found in any female, but there are often colour spots on the anal fin of females in the same position as the ‘egg dummies’ in males. However, there is never a clear zone around the spots and they certainly do not have the appearance of a true ‘egg dummy ’, neither are they so large nor so clearly defined. Sex-limited female polychromatism occurs in at least eight species of the Lake Victoria Haplochromts, and in two of the endemic monotypic genera (Macropleurodus bicolor and Hoplotilapia retrodens). In none of these species does the occurrence of polymorph individuals exceed ca 30 per cent of females in a population. The commoner polymorph coloration found in all species showing the phenomenon is an irregular black piebald-on a silver or yellow background (Text-fig. 29) ; no inter- grades between the piebald and normal female coloration have been found. This piebald pattern does not vary with the emotional state of the fish, nor with the type of habitat in which the fish is living. A second morph is also known from Hoplo- tilapia retrodens, and in at least one undescribed Haplochromts species ; it comprises a blotched yellow and orange ground coloration on which is superimposed small black blotches or merely a fine but distinct peppering of melanophores (Greenwood, 1956a). The eight Haplochromis species showing piebald coloration belong to at least four different phyletic groups, none of which shows especially close interrelationships within this admittedly closely related species flock. Algal grazing species (H. nigricans), paedophages (H. cronus, H. obesus and H. barbarae), mollusc eaters (H. 54 P. H. GREENWOOD sauvaget), specialized insectivores (H. chilotes) and a piscivore (H. altigents) all exhibit the phenomenon. Selective advantages associated with the maintenance of this balanced poly- morphism are unknown (Greenwood, 1956a; see also discussion in Fryer & Iles, 1972), but must be fairly substantial considering both the frequency of polymorph individuals in a population, and the fact that 1t occurs in so many species. LS Boats Ear) u SSE SEE Fic. 29. Haplochromis barbarae female, showing piebald coloration (see p. 53). The usually complete linkage of piebald coloration to the female sex is disturbed in Macropleurodus bicolor, where two piebald males have been found (Greenwood, 1950a). In one of these fishes the pattern is identical with that of females, but in the other it is less intense and the background coloration is darker. As far as I can determine, the gonads of these atypical males contained only testicular tissue, and provided no evidence to suggest possible protandry in the species (as is suspected for a Malawian genus Labeotropheus fuelleborm [report by Professor H. Peters, quoted in Fryer & Iles, 1972: 172]). A male showing partial female-type piebald colours is also recorded for Hoplotilapia retrodens (Greenwood, 1956a), and I have seen a live Haplochromis chilotes with piebald coloration and well-defined ‘egg dummies’ on the anal fin. Unfortunately it was not possible to dissect the latter specimen, an aquarium fish. The genetic basis for this almost completely sex-limited polychromatism, and its occasional breakdown, has not been determined. Fryer & Iles (1972) argue cogently in favour of its being the manifestation of a potency balance between the expression of alleles on autosomes and ‘suppressor’ alleles on the sex chromosomes. On the assumption that female cichlids are the heterogametic sex (which is also assumed in my explanation of the phenomenon ; Greenwood, 1956a), this interpretation seems at least to provide a working hypothesis. I am less sanguine about their view that the occurrence of atypically coloured (i.e. piebald) males violates the sanctity of male coloration as an important factor in species recognition. That the proportion of piebald males is very much lower than that of polymorph females suggests that alleles with a potency sufficient to overcome the ‘suppressor’ alleles are extremely rare. EVOLUTION:-OF A CICHLID SPECIES FLOCK 55 Could not this rarity be attributable to adverse selection pressure on those in- dividuals possessing them? That is, the chances of such males being ‘recognized’ and ‘chosen’ by a female (an obligatory prelude to spawning) are greatly reduced ; this in turn would lead to a great reduction in the number of these ‘ powerful’ alleles in the population. Is it not even possible that piebald males do not mate at all, and that the alleles in question are maintained purely by mutation? To me, a far more telling point against the importance of male coloration as the prime means of species recognition in courtship derives from the Haplochromis species flock in Lake George. In this lake the euphotic zone extends only to a depth of 50-60 cm (Greenwood, 1973a). Most Haplochromis species live and apparently spawn below this depth, or at least in places with extremely low light intensities. To the human observer underwater in these lighter areas, a fish’s shape, but not its colour, is just apparent. Thus in these species there seems to be a prima facie case against the overriding importance of male coloration. Yet, males are brightly coloured, and their coloration (with one exception, see Greenwood, 19734) is species specific, and does not show any greater degree of intraspecific variability than do the colours of the Lake Victoria species. Finally, in dealing with colour aberrancies in Lake Victoria species, mention must be made of apparently melanic Haplochromis obesus. Atypically dark individuals of both sexes are known, but melanic males are commoner. The degree of darkening shown by these fishes is somewhat variable ; extreme individuals are uniformly black, but others are much brighter ventrally and do in fact intergrade with the normal male coloration. No correlation has been observed between locality or environment and the presence of melanic individuals, nor with the degree of melanism displayed. In a large and complex species flock like that of Lake Victoria, with many species occupying the same habitats, one might well expect a number of interspecific hybrids. Detecting hybrid individuals on purely morphological grounds would be difficult because of the close similarity between putative parent species. Nevertheless, despite constant awareness of this possibility, I have found only one specimen that might bea hybrid. Whatever the barriers to interspecific mating may be, they seem to be very effective. Comparisons between the coloration of Haplochromis species from Lakes Victoria and Malawi are difficult to make, partly because of difficulties in verbalizing colours, and partly because colour descriptions are not available for many Lake Malawi species. Nevertheless, the views of those who know both faunas strongly indicate a fundamental difference in colour patterns and dominant colours (see remarks in Fryer & Iles [1972]; also personal comments by Drs Fryer, Lowe-McConnell and Trewavas). : The reasons for these differences are probably manifold (for example, the greater number of pelagic species in Lake Malawi, the existence of shoaling species in that lake etc.), and must be, in part, of phylogenetic origin. In this context it is important to emphasize the basic similarity in coloration between species of the Lake Victoria and Lake Edward—George fishes, where phyletic integrity is ascertainable on other grounds (Greenwood, 1973a). 56 P. H. GREENWOOD INTERRELATIONSHIPS OF THE LAKE VICTORIA HAPLOCHROMIS SPECIES Having outlined the main morphological and biological features of this Haplo- chroms species flock, attention can now be turned to four major evolutionary ques- tions. These are, the origin of the flock, its internal phylogeny, the factors under- lying its explosive speciation and adaptive radiation and, finally, why it is that, under such conditions the Cichlidae, more than any other family, react in this fashion (see p. 4 above and pp. 103-111 below). The phylogenetic problem can be considered first (Text-fig. 70). Any attempt to analyse this flock is fraught with difficulties. The very basis for a phyletic study, the flock’s monophyletic origin, cannot be established unequivocally. Reasons for assuming a mono- or oligophyletic origin for the endemic species have already been discussed (p. 20 above). To these may be added one other, namely, that a careful survey of most known species indicates a basic morphological homogeneity that would be unlikely if several unrelated species were implicated as ancestors. In other words, if the flock is not strictly monophyletic in origin, its stem species were likely — to have been no more distantly related than sister species (sensu Hennig, 1966). Be that as it may, it is certainly possible to establish monophyletic groups within the flock (Text-fig. 70). At a gradal level, the Victoria Haplochromis species present an interesting and, for extant animals, an unusual picture of virtually complete morphological intergradation between the generalized and the specialized in any one adaptive radiation. Two trenchant examples are found in the mollusc-eating species and a third in the pis- civorous predators (see pp. 31 and 37 above). Graded anatomical stages, involving changes in dentition and dental pattern among species that shell molluscs orally are all represented in the lineage H. sauvagen, H. prodromus and H. xenognathus, with H. grant a slight deviant from the Z. prodromus stem (see Text-fig. 41 ; and Greenwood, 1957). The genus Macropleuro- dus 1s an extreme modification of the same morphotype (see Text-figs 75A and B; also Greenwood, 1957). Stages in the hyperdevelopment of a pharyngeal crushing mill amongst mollusc eaters are well represented by seven species, H. pallidus, H. riponianus, H. saxicola, H. humilior, H. theliodon, H. obtusidens and H. ishmael1, the species listed in approximate order of increasingly molarized lower pharyngeal teeth (see Text-fig. 5 and Greenwood, 1960 ; also p. 37 above). The piscivorous predators present an even more complex and just as complete picture of bridged gaps between specializations (see Greenwood, 1962, 1967, 1973b). The other trophic groups all show similar, but often shorter, trends in specialization. The question now to be asked is whether all the constituent species of each gradal - complex represent truly monophyletic lineages within the flock, or whether we are confronted by a web of parallelisms. Detailed studies of the species comprising the different radiations suggest that the latter explanation is the more likely one. The — task then is to untangle the gradal groups by seeking out combinations of specialized characters (apomorph characters of Hennig, 1966) that will link together phyletic lineages within each grade. The student of the Lake Victoria flock is in the position of a palaeontologist who has a nearly complete phyletic record for several branches of a monophyletic lineage, EVOLUTION. OF A CICHLID SPECIES FLOCK 57 the extant end-products of which are clearly defined taxa, but some of whose an- cestors tried out, as it were, specializations now characterizing living taxa in a sister lineage. Unlike the palaeontologist, the student of the Lake Victoria Haflochromis has no temporal sequence to help him sort out branching points. Historically, some of these difficulties could spring from the flock’s relative youth. In an older assemblage a number of species now seen as representing an intermediate stage of specialization could well have been eliminated as unsuccessful competitors with more specialized species of their own or related lineages. Whether or not an older cichlid species flock would be an easier subject for phyletic analysis, and whether the results would be more precise, are moot points. Un- fortunately no such test has been made on the flocks of either Lake Tanganyika or Lake Malawi in their entirety. Fryer (1959; also repeated in Fryer & Iles, 1972) has, however, attempted to work out the interrelationships of the ‘Mbuna’ generic complex in Lake Malawi. His criteria of relationship are not explicitly defined and the resulting ‘tree’ seems more an indicator of parallel radiation than of phylogenesis. A feature shared by Haplochromis in both the older flock of Lake Malawi and the younger one of Lake Victoria is the existence of two or more very similar species clustered around any one level of adaptation. It is these species in particular that have given the Cichlidae a reputation among taxonomists of being a difficult group. Examples of species knots from Lake Victoria would include five species in the insectivore lineage (H. macrops, H. megalops, H. pallidus, H. lacrimosus and H. piceatus ; see Text-figs 3, 8 and g and p. 30 above), the five species of the crustacean eating, deepwater, H. tvidens group (see p. 39 above ; also Greenwood & Gee, 1969), H. sauvage: and H. prodromus in the mollusc eating grade (Text-figs 3 and 19-21 ; see also p. 37 above ; and Greenwood, 1957), H. wshmaelt and H. pharyngomylus in the other molluscivore grade (see p. 37 above ; and Greenwood, 1960), H. riponianus, and H. saxicola (Text-figs 3 and 42) also mollusc crushers (Greenwood, op. cit.), H. obesus, H. maxillaris and H. melanopterus among the paedophages (Text-figs 3 and 17; see also p. 33 above; and Greenwood, 1959b), and many other examples of twin, triplet and even quintuplet species among the piscivorous predators (Green- wood, 1962, 1967). These species groups (if they are truly phyletic, and some certainly seem to be that) may reflect the physiographical background to speciation in the lake, and be the products of simultaneous multiple speciation each from the same common ancestor isolated in different water bodies (see p. 10 above and p. 114 below). Relatively few characters can be used to construct phylogenies. It will be recalled that meristic characters are almost uniform throughout all species. Thus, interest centres on cranial and dental features which, on the whole, clearly show levels of specialization or generalization. . _ As the base line for this study I have identified as generalized those characters shared by Victoria species and a fluviatile species widespread in the rivers of Uganda (i.e. a taxon anatomically identical with, if not actually H. bloyeti ; see Greenwood, 1g7I). Skull shape, jaw form and tooth morphology in this species are illustrated in Text-figs 30 and 31. Characteristic features of this skull type are the relatively 58 P. H. GREENWOOD decurved dorsal profile to the preorbital face of the neurocranium (giving the orbital margin a near-rounded circumference), the high cranial vault and the relatively short ethmovomerine region. The premaxillary ascending process is shorter than the dentigerous arm of the bone, and the lower jaw (dentary plus articular) is neither foreshortened nor elongate. The lower pharyngeal bone is not noticeably thickened, and its equilateral denti- gerous surface is covered by fairly well-spaced rows of cuspidate and laterally com- pressed teeth. Teeth in the median rows are usually a little coarser than those situated laterally (Text-fig. 30A). 3mm Fic. 30. Haplochromis bloyeti. A: Lower pharyngeal bone in occlusal view. B: Outer row teeth from the premaxilla; viewed anterolaterally. Scale = Imm.) The outer jaw teeth in H. bloyeti are unequally bicuspid, the cusps triangular in outline (Text-fig. 30B) ; the inner teeth are small and tricuspid, and are arranged in not more than three rows in either jaw. Usually a few unicuspid teeth occur pos- teriorly in the outer row of the premaxilla, and in larger fishes (> 80 mm long) a few unicuspids are intercalated with the bicuspids anteriorly and anterolaterally in both jaws. Among extant Lake Victoria endemic species, H. pallidus has a syncranium and dentition virtually identical with that of H. bloyets (see Text-figs 31-33), and there are several other similarly generalized species known. Having established the anatomical nature of a generalized species, each trophic group (see pp. 58-80) will be considered in turn, probable phyletic lineages (based © on shared specialized characters, i.e. synapomorphy) will be delimited and, where possible, interrelated. The insectivorous species Typically generalized skull, jaws and dentition are found in five species (#. pallidus [Text-figs 32 and 33], H. macrops [Text-figs 32 and 33], H. lacrimosus, H. megalops and H. piceatus ; see Greenwood, 1960 and Greenwood & Gee, 1969). In all except H. pallidus the lower pharyngeal dentition is also generalized; in H. EVOLUTION OF A CICHLIID SPECIES FLOCK 59 Fic. 31. Haplochromis bloyeti. Neurocranium and lower jaw, in left lateral view. (Scale := 3 mm.) pallidus some median teeth are slightly enlarged. A sixth species, H. cinereus, is osteologically one of this group, but differs in having a predominance of unicuspid over bicuspid outer jaw teeth, even in fishes at a size where bicuspids predominate in other species (Greenwood, 1960). Because of their generalized cranial anatomy it is impossible to determine the interrelationships of these six species ; all could be cognate. However, except for H. pallidus all species have the preorbital skull face slightly more decurved than in the presumed ancestral type, with which H. pallidus is virtually identical. Anatomically, it is interesting to see amongst these fishes a representation, in embryo as it were, of many characters that are developed in various characteristic ways among more specialized trophic groups. For instance, there are the unicuspid teeth of H. cinereus (but see below), the incipient development of enlarged pharyngeal teeth in H. pallidus, and a basic syncranial ‘bauplan’ that, through differential growth of certain elements (or regions in the case of the neurocranium), provides the starting point for the different types found in other trophic groups. None of the species considered so far has adults exceeding a length of more than 105 mm. Haplochromis saxicola is, superficially, like the foregoing species (cf Text-figs 3 with 8 and 9). However, its neurocranial morphology departs from the basic type towards that of some piscivorous species (Text-figs 34 and 68). That is, the pre- orbital region is more elongate (as is the entire skull anterior to the brain case), its dorsal profile is not noticeably decurved and it slopes downwards less steeply, and there is a reduction in the depth of the brain-case region. The jaw dentition too differs somewhat from the generalized type because unicuspid teeth predominate ; since the smallest H. saxicola known is larger than the largest members of the general- ized group, this observation may be of doubtful significance (see p. 106). Thus, H. saxicola, although still an insectivore, does depart from the presumed basal members of that trophic group in several important anatomical features, and in reaching a larger adult size, vzz 125 mm. The nearest living relative of H. saxicola is H. riponianus (Text-figs 34 and 42), here classified as a mollusc eater (see below). Nevertheless, as the representative of a specialized structural level, the H. saxicola condition could be ancestral to the pisci- vorous predator radiation (see p. 82). 60 P.. Ho. GREENWOOD = H. macrops H. pallidus H. empodisma H. chilotes Fig. 32. Neurocranial form in insectivorous Haplochromis species (see also Text-fig. 33). (Scale+— 93 ams) EVOLUTION OF A’ CICHLID SPECIES FLOCK 61 Also related to H. saxicola, at least on the basis of jaw and neurocranial morphology (Text-fig. 34), is H. aelocephalus, a predominantly insectivorous species that also feeds on other invertebrates, including molluscs. Externally, H. aelocephalus is readily distinguished by its protracted snout which, in extreme individuals, is almost tubular (see Greenwood, 1959b). The outer jaw teeth in H. aelocephalus are simple unicuspids like those in H. saxicola, but the inner teeth are arranged in broad bands, more like those in the H. sauvagei—H. chilotes lineage (see p. 72 below). There is, however, no reason to suppose that H. aelocephalus is at all closely related to that lineage, from which it differs in neurocranial morphology, the shape of the outer jaw teeth, and in having an elongate and slender lower jaw. H. macrops H. pallidus H. chilotes H. chromogynos Fic. 33. Lower jaw form in insectivorous Haplochromis species (see also Text-fig. 32). (Scale = 3 mm.) Even superficially, H. chilotes, with its hypertrophied and lobate lips (Text-fig. 10), would qualify as a specialized species, a rating confirmed by its feeding habits (see p. 31). The oral dentition of H. chilotes departs markedly from that in all the species so far considered (Text-fig. 33). The outer teeth are stout unicuspids with the crown strongly incurved ; such teeth occur in only one other group of species (see p. 72 below). The lower jaw is short and stout, but the ascending process of the premaxilla is relatively longer than in the generalized type of bone. The neurocranium of H. chilotes could be classified as a derivative of the generalized type in which the preorbital region has become strongly decurved. Adult H. chilotes reach a length of ca 150 mm. 62 P. H. GREENWOOD H. saxicola H. aelocephalus H. aelocephalus H. saxicola Fic. 34. Neurocranial and lower jaw form in insectivorous Haplochromis species. (Scale = 3 mm.) Virtually identical syncranial morphology and dentition are found in H. chromo- gynos (Text-figs 32 and 33), but the lips are not hypertrophied in this species ; adult - fishes are apparently smaller (maximum length 110 mm) than in H. chilotes. I would consider H. chilotes and H. chromogynos to be sister species, with H. chilotes the derived (apomorph) member of the pair. Similarly specialized skull and © jaw forms and outer tooth morphology are seen in one lineage of mollusc eaters, and it is thought that H. chilotes and H. chromogynos are members of the same lineage (see p. 72, and Text-fig. 66). Finally in the insectivorcus group there is Haplochromis empodisma, a species which could be described as a deep-bodied and larger form of the five species described first EVOLUTION. OF A -GICHLID SPECIES FLOCK 63 in this section (Greenwood, 1960) ; adults of H. empodisma reach a length of 120 mm, compared to 100 mm in the other species. The oral and pharyngeal dentition of H. emposdisma are of a basic type, but the neurocranium departs quite noticeably both from the generalized type (Text-fig. 32) and, especially, from that of H. chilotes and H. chromogynos. The preorbital profile is straighter and slopes less steeply because the brain-case depth is somewhat shallower, giving the whole skull a more linear and less rounded appearance. In fact, the skull of H. empodisma is almost intermediate between the H. saxicola type and that of the more generalized species (e.g. H. pallidus and H. macrops). Admittedly the H. empodisma neurocranial form differs less markedly from the basic form (e.g. H. pallidus) than does the skull in H. saxicola or, more especially, H. chilotes. But it does seem to represent a distinct type and one which appears elsewhere within the flock. One near relative of H. empodisma is the mixed mollusc—insect eating H. obtusidens (see p. 73 below ; and Greenwood, 1960). To summarize. Among the insectivorous species there are four distinct lineages (Text-figs 65, 66,68 and 70). One, the most generalized, departs little from the pre- sumed ancestral and fluviatile species represented by H. bloyeti. The second (H. saxicola) shows an elongation of the skull and jaws approaching that found in certain piscivorous and paedophagous species. The third lineage (H. chilotes and H. chromogynos) exhibits strong decurvature of the skull anteriorly, strengthening of the jaws (especially the lower) and a distinctive dentition, all features shared with a specialized branch of the mollusc-eating trophic group. The fourth lineage (H. empodisma) represents a slight departure from the generalized H. pallidus-like condi- tion towards a type seen in certain piscivorous predators and other trophic groups. Apart from their dentition, the differentiation of these lineages is manifest in a changed shape to the anterior moiety of the neurocranium and, to a lesser extent, in the relative length of the lower jaw. In other words, a change in relative growth patterns. It is probably significant in this connection that there are marked dif- ferences in the maximum adult size attained by members of the various lineages, with those of the more specialized lines growing to a larger size. Phytophagous species In most respects the morphologically least differentiated member of this trophic group is H. phytophagus (see Greenwood, 1966b), one of the two species known to feed directly on macrophytes ; see pp. 39 & 41. Syncranial organization in H. phyto- phagus is like that in any generalized insectivore. Specialization is seen in the coarser jaw teeth, the coarser, less numerous teeth on the pharyngeal bones, and in the lengthened intestine. The dental modifications, however, are but slight variants of the basic bicuspid oral dentition, and even slighter changes in the pharyngeal denti- tion ; both are adaptations for biting and then macerating the plant tissues eaten (particularly leaves). _ Three of the four algal grazing species (H. nuchisquamulatus, H. lividus and H. obliqguidens) also have a generalized syncranium. All four species (i.e. including H. nigricans, see p. 39) have a greatly elongate intestine (Greenwood, 1956b) and rather 64 PY oH. vGREEN WOOD BIG. 35. H. erythrocephalus H. paropius LS TRESS H. obliquidens Neurocranial and lower jaw form in phytophagous Haplochromis species. (Seale = 3 muz:) EVOLUTION JOE A-*CICHELD SRECIES FLOCK 65 fine and numerous pharyngeal teeth, but H. nigricans does show certain syncranial pecularities not shared by the others. As far as tooth shape 1s concerned, H. nuchisquamulatus has the least specialized dentition (Text-fig. 4) since it is of the basic bicuspid type with acutely pointed cusps ; compared with the generalized insectivore, however, H. nuchisquamulatus does show an increase in the number of inner tooth rows, a wider area of implantation for these teeth, and all the teeth are moveably implanted. Hapflochromts lividus, in comparison, has many outer jaw teeth in which the major cusp is relatively broader and obliquely truncate (not acute), see Text-fig. 4F ; a few teeth retain a generalized shape. Inner jaw teeth and the pharyngeal dentition in H. lividus do not differ from those of H. nuchisquamulatus. The morphological trend apparent in the teeth of H. lividus is accentuated in H. obliquidens. Here the minor cusp is suppressed, and the major cusp is drawn out and obliquely truncate (Text-fig. 4G) ; occasionally some teeth have a vestigial minor cusp, and a few teeth of the H. lividus-type may be present posteriorly in the pre- maxilla. Generally the inner teeth are typical tricuspids, but there is often an ad- mixture of these and teeth differing from those of the outer row only in their smaller size. (It may be noted that larval H. obliquidens have the fine, setiform teeth that probably occur in all Haplochromis species irrespective of their definitive adult dental morphology ; Greenwood, 1956b, and further, unpublished observations.) It seems reasonable to consider H. obliquidens, H. lividus and H. nuchisquamulatus as members of a phyletic lineage ; certainly H. obliquidens and H. lividus are sister species. Although it is difficult to be certain of the relationship between these species and H. nuchisquamulatus, a relationship seems to be indicated by the slight departure of tooth form in the latter towards the H. lividus type (i.e. moveable implantation, expansion of the crown and its less acutely pointed tip). The base of this lineage lies undetectably within the generalized insectivore species group. The relationships of H. nigricans, the fourth species feeding on phytoplankton, provide something of a puzzle. Its oral dentition is of a modified bicuspid type (Text-fig. 36), but does not show the obliquely cuspidate major cusp typifying the Saree Fic. 36. Outer row teeth from the dentary of H. nigricans (seen in lateral view). (Scale = 1 mm.) other species. Indeed, many teeth are subequally cuspidate, the cusps rather bluntly pointed, although the whole crown, as in the grazers, is expanded relative to the neck of the tooth (Text-fig. 36). The inner teeth are tricuspids and are arranged in numerous rows with a reduced space between them and the outer tooth row (again, 5 66 P. H. GREENWOOD like the other species). Haplochromis migricans differs in the shape of its neuro- cranium, in which the preorbital face is somewhat decurved (Text-fig. 35), and in having a stouter lower jaw. In both these characteristics, H. nigricans shows some approach to the H. chilotes-H. chromogynos condition. The question raised by these contrasting features is whether the neurocranial and jaw characters indicate relationship to the H. chilotes lineages, or whether they are to be interpreted as parallelisms. I incline towards the latter interpretation mainly because of the very different morphology of the teeth in H. chilotes and other members of its lineage. Haplo- chromis mgricans, unlike the other periphyton grazers, feeds principally from rocks and not plants, and it has a rather different feeding method. The epiphytic feeders scrape algae mainly from leaves which are taken and held between the jaws. Haplo- chromis mgricans, on the other hand, nibbles algae from a rigid substrate, for which the stout jaws, downward protrusion of the upper jaw (because of the decurved surface on which the premaxillae slide) and the subequally bicuspid outer teeth would appear to be adaptations. , The three remaining phytophages (H. dps [Text-fig. 27], H. pavopius and H. erythrocephalus {Text-fig. 26] ; Greenwood & Gee, 1969) all feed on phytoplankton apparently gathered from the bottom and not while in suspension ; all have rela- tively long intestines. Unfortunately, lack of material precludes a detailed knowledge of cranial anatomy in H. cinctus ; however, its dentition, both oral and pharyngeal, is of the generalized bicuspid type (Greenwood & Gee, 1969). Haplochromis paropius has the syncranial architecture and oral dentition of a generalized insectivore like H. macrops (see Text-figs 32, 33 and 35), but its pharyngeal dentition is of the finer type found in the periphyton feeders (Greenwood & Gee, op. cit.). The neurocranium in H. erythrocephalus differs from this generalized type in exactly the same way as does the neurocranium of H. empodisma (see p. 63 above ; also Greenwood & Gee, op. cit.). The pharyngeal and oral dentition of these two species is also similar. No immediately obvious adaptive features are apparent in this type of syncranial architecture. Possibly the larger mouth and deeper oro- pharyngeal cavity (relative to those of the generalized type) are of advantage to a species which, when feeding, must pass a considerable volume of water and par- ticulate material through this cavity in order to obtain its food. Because of their generalized cranial characters and dentition, H. cinctus and H. paropius cannot be related to any particular species. But because of their specialized alimentary characters the species must be considered as apomorph derivatives of the | generalized insectivore stock, and are probably related to one another (i.e. sister species). Haplochromis erythrocephalus, on the other hand, can be related to H. empodisma also as a derived species (Text-figs 65 and 70). To summarize: Three phyletic lines are represented among the phytophages so far considered. The periphyton grazers comprise a single lineage in which most of the adaptational stages of the radiation are still extant, together with a differently specialized offshoot, H. nigricans. The phytoplankton feeders comprise two line- ages; one (H. cinctus and H. paropius) is little different from the basal insectivore EVOLUTION OF A CICHLID SPECIES FLOCK 67 stem, the other (H. erythrocephalus) is part of a lineage whose living basal representa- tive is probably H. empodisma. Haplochromis phytophagus, a browser on macro- phytes, could belong to the same lineage as H. cincius ; like those species its dentition shows little departure from the generalized type, but it does have the same alimentary canal specializations. Possibly the three species are an offshoot of the lineage cul- minating in the periphyton grazers (Text-figs 65 and 70). There remains one other species, H. acidens (Greenwood, 1967), which, like H. phytophagus, feeds directly on macrophytes. The skull, jaws and unicuspid dentition (Text-fig. 37) of this species conform with the type found among a group of specialized piscivorous predators (Greenwood, op. cit.), as does the pharyngeal dentition. Furthermore, despite the vegetarian habitats of H. acidens, its intestine is relatively shorter than in other plant-eating species, although it is longer than in a piscivore. Fic. 37. Neurocranium and lower jaw of the enigmatic H. acidens (see p. 41). (Scale = 3 mm.) I have suggested elsewhere (Greenwood, 1967) that, anatomically, H. acidens could be a member of a particular piscivore lineage (the ‘sevranus’ group, see Green- wood, 1962). It could also be a derivative from an H. empodisma-like species ; the principal change involved would be an increase in the relative number of uni- cuspid teeth so that these teeth predominate over bicuspids in both jaws (there is an admixture of bi- and unicuspids in large H. empodisma [i.e. > 95 mm long] and in H. acidens < 90 mminlength). In both species the teeth, irrespective of cusp form, are slender in comparison with those of the generalized type. Considering the rather indefinite feeding habits of H. empodisma (a benthic insectivore also ingesting quanti- ties of plant debris) and the interspecific similarities in syncranial architecture, I would now consider H. acidens to be a derivative of the H. empodisma lineage, rather than of the ‘seyvanus’ piscivore group (see also p. 41). Species feeding on benthic Crustacea This small and trophically rather ill-defined assemblage of five species, the ‘trzdens’ _ group (H. dolichorynchus, H. tyrianthinus, H. chlorochrous, H. cryptogramma and H. trvidens [Text-figs 24 and 25 ; Greenwood & Gee, 1969 ; Greenwood, 1967] is, how- ever, well-defined morphologically. Group syncranial architecture (Text-fig. 38) is 68 Ps... GREENWOOD essentially of the H. saxicola type (Text-fig. 34; see also p. 59 above), but the oral definition is peculiar. There is a high proportion of tricuspid teeth in the outer_ tooth rows of both jaws, but especially in the lower one. The other teeth in these series are either slender bicuspids or unicuspids ; that is, like the teeth of H. saxicola. Unlike H. saxicola, none of the ‘tvidens’ group shows any enlargement of the median teeth on the lower pharyngeal bone (and never, as in some H. saxicola individuals, an enlargement of the bone itself ; see Greenwood, 1960). Fic. 38. Typical neurocranial form in a member of the H. tvidens lineage (ex H. iridens). (Scale = 3 mm.) Neither the ‘tvidens’ group nor H. saxicola exhibits any specialized characters obviously associated with their feeding habits (unless the slightly enlarged median pharyngeal teeth of H. saxicola are considered thus; but the known diet for this species does not indicate any increase in the number of molluscs eaten as compared with generalized insectivores). It is, therefore, impossible to consider either the ‘tvidens’ species or H. saxicola as the apomorph (1.e. derived) sister group of the other. That both should be considered part of the same phyletic lineage seems likely from their shared syncranial specializations. At least one other species, H. melichrous, might be included in this trophic group (Greenwood & Gee, 1969). It has a rather generalized dentition of mixed bi- and unicuspid teeth, but the occasional tricuspid does occur in the posterior part of the lower jaw. Haplochromis melichrous differs from members of the ‘tvidens’ group in its neurocranial morphology, and in these characters resembles members of the ‘serranus’ group piscivores (Greenwood, 1967). The dentition of H. melichrous 1s sufficiently generalized (despite the occasional tricuspid teeth) for it to be of little - value as a phyletic indicator. Skull and jaw characters are also not particularly clear-cut in this context, although they do seem to exclude the possibility of relation- ship with species in the ‘tvidens’—‘ saxicola’ lineage. For the moment, H. melichrous — can only be associated with the ‘sevvanus’ group, to be discussed on p. 81 ef seq. In summary, the benthic crustacean-eating group seems to comprise two lineages. The larger one is a derivative of the H. saxicola line, the smaller (H. melichrous only) is of uncertain affinity but could possibly be related to the ‘servvanus’ piscivore lineage (Text-fig. 70). EVOLUTION: OF A CICHLID SPECIES FLOCK 69 Scale eating species Comparatively little is known about the anatomy of H. welcomme: (Greenwood, 1966b) the unique lepidophage in Lake Victoria. The slender, moderately elongate body (Text-fig. 3), and the superficial head shape of this species are not unlike those of species in the ‘ividens’ group discussed above. The broad bands of inner teeth in H. welcomme: (Text-fig. 39) are a specialization not found in that group, and are in fact not approached by the dental pattern in any other species (except, perhaps, OOOO Pose e@ OAR OLE Pre 222 PL. On poe Gan (4g RISE RE vi Ona a hd (A 95 %q Oa Ante p29 3am “aP3of404 age Og fg 4 noe vy fi boreal eee ; a GasaaSa pe Bp agcyga UG 99 OODe L BAD 06 3B 2Pc4 Oak GoD OP a6 a Oe S800” 5° 0.0 GO Op fa) 0 %5% (Ponce Gogo? OOS Imm 0 0, ne raared i “2 a9 ; on & Doe > D0? gH D 3 PG > 005,30, % ) 2 oO poe BE Uy 2D OLD ee eH pla Dp PE D oe OD eee lp 2580 oD 5°, py Fe ROI gin 2929p BeooP LO oy) > &° be 5? o & 220 Ss? Roe IE 5, yo Opa Soe OL beg S$ ORG SL OLE GO op Seegee Fic. 39. Jaw tooth pattern in H. welcommei (a scale scraper) ; premaxilla above, dentary below. Only one half of each jaw is shown. the monotypic Platytaeniodus degent, see p. 103 and Text-fig. 73). Judging from radiographs of the head (no skeletons are available), H. welcomme: does have a neurocranium like that of a ‘tvidens’ group species, although it also resembles H. paraguiarii, one of the piscivorous predators (see p. 89 below). Until good osteo- logical material is available the relationships of H. welcomme: must remain enig- matic, although the indications are of affinity with the ‘tvzdens’ lineage (Text-fig. 70). Mollusc eating species _ The clear-cut dichotomy in feeding methods among species in this trophic grade has been discussed in detail on p. 37. | Those species that wrench snails from their shells (or crush the shell orally) con- stitute one of the best defined phyletic lineages within the whole flock. Specializa- tions shared by all four species (H. sauvagei [Text-fig. 19], H. prodromus, H. grants [Text-fig. 20] and H. xenognathus [Text-fig. 21]) are the strongly decurved, in some species almost vertical, preorbital skull, a stout lower jaw, the stout unicuspid teeth 70 P. H. GREENWOOD H. prodromus H. xenognathus H. prodromus Fic. 40. Neurocranial and jaw form in two species of mollusc-eating Haplochromis, both oral-shellers. (Scale = 3 mm.) with strongly incurved crowns, and the broadly arranged rows of unicuspid inner teeth (see Text-figs 40, 41 and 66; also Greenwood, 1957). Smaller members of these species (that is fishes between 80 and 100 mm long, depending on the maximum size reached in the species) do have bicuspid outer teeth, or a mixture of bi- and unicuspids. The bicuspids, like the unicuspids, are strongly recurved, thus contrast- ing with the barely recurved generalized type of bicuspid. The lower jaw is shorter than the upper in all species but H. granti, where it is sometimes a little longer. Haplochroms grant: also differs from other members of the group in having the mouth inclined upwards at a small angle ; it is horizontal in the others. Anatomically, H. xenognathus is the most specialized species, H. sauvage: and H. prodromus the least specialized (cf Text-figs 40 and 41); but these degrees of dif- ference are not apparent in feeding habits, which are identical (Greenwood, 1957). Haplochromis xenognathus differs chiefly in having some lower jaw teeth implanted horizontally, the inner tooth rows in much broader bands, with those of the dentary presenting a convex occlusal surface (see Text-fig. 41C). Haplochromis sauvager and H. prodromus are virtually identical in their oral, dental and neuro- cranial morphology, but individuals of the latter species reach a larger size (130, cf 105 mm). EVOLUTION OF A CICHLID SPECIES FLOCK 71 The monotypic genus Macropleurodus bicolor is, on the basis of its syncranial architecture, a member of the same species group. The jaw teeth of M. bicolor (Text-fig. 75) differ from those of other species in being coarser, more strongly in- curved and in generally retaining traces of a minor cusp (Text-fig. 75B). This cusp persists as a slight, vertical hump on the near-horizontal occlusal surface presented by the major cusp. The dentigerous arms of the premaxillae are bowed (usually more so on one side than the other) ; consequently one side of the mouth is slightly | a ir ig n y (A a9 nee \ 4] / 7 U pe esis Sh cea 3, SUPA ENITAN PPO 2 J) K anil (du WG WE £0 Ye UNV EMA ye, D ON NGO S : a )) CSAS tL oz A ra DSA 43 is “ = D\ q4 TIL yy IW cA ms ee 3 xX Fy : milks { WS) Toll Wy (Wig t A he Py, ia ene Gyo BRR GS gate e Rad Uf Gal By, Iu" BUSY Fic. 41.. Jaw tooth shape and pattern in two species of mollusc-eating Haplochromis, both oral shellers. A: H. prodvomus, occlusal view, premaxilla uppermost. Band C: H. xenognathus, occlusal view of both jaws, and lateral view of lower jaw respectively. open even when the jaws are closed. This peculiar arrangement could be of adaptive significance in a species whose feeding involves grasping a snail so that it cannot withdraw into its shell. In other words, because of its arched upper border the gape is increased without an increase in the extent to which the dentary would otherwise _ have to be lowered. The near vertical ethmoid region (over which the ascending processes of the pre- maxillae run) in all H. sauvagei-group species (see Text-figs 40 and 66) produces an 72 P. H. GREENWOOD unusual jaw opening movement, namely downward and slightly forward rather than mostiy forward ; the short and deep adductor mandibulae muscles ensure a powerful and rapid closing action for the dentary. The strong jaws and teeth, coupled with these jaw movements, all indicate a high level of adaptation for the unusual feeding methods practised by members of the lineage. It is interesting to see that the ethmo- vomerine bloc in the skull of these fishes is slightly longer than the near-horizontal ethmo-vomer of a generalized skull. (In fact, it is relatively as long as in many piscivorous species groups; see pp. 80-93.) Presumably the lengthened ethmo- vomer gives added support and backing to the premaxillary processes when the upper jaw is protruded downwards and the fish is wrenching a snail from its shell. A possible phyletic relationship between the H. sauvagei group and the insecti- vorous species H. chilotes and H. chromogynos has already been suggested (p. 61). The skull in these latter species is rather less strongly decurved (see Text-figs 32, 40 and 66), and this coupled with their relatively unspecialized feeding habits (except, on occasion, in H. chilotes, see p. 31), leads me to rank the H. sauvager — lineage as the apomorph sister group of the H. chilotes-H. chromogynos pair. Phylo- genetically it may be significant that black-and-white piebald female polymorphs (see p. 53) are recorded in four of the six species in this lineage (H. sauvager, Macro- pleurodus bicolor, H. chilotes and H. chromogynos [where it is, indeed, the ‘normal’ female coloration]). Similar polychromatism occurs in several other species (none closely related to the “sauvagez’ lineage) but never elsewhere is it known from so many species in a presumed lineage (see pp. 53-54 above). When more is known about certain Lake Victoria species, for example H. crassila- bris (see Regan, 1922), the H. sauvagei lineage may have to be expanded. Hopefully it will then prove possible to find the group’s plesiomorph relatives. The second mollusc eating group is made up from those species in which the prey is crushed between the upper and lower pharyngeal teeth and bones. Two species, H. tshmaelt and H. pharyngomylus (Greenwood, 1960; also p. 37 above), have the most highly developed pharyngeal mills (see Text-fig. 5A). The lower bone is greatly enlarged and most of its teeth are molariform ; the upper bones are correspondingly modified. The skull is essentially like that of the generalized type, except for the better-developed ventral apophysis on which the upper pharyngeal bones articulate, see Text-figs 43 and 44. This apophysis is formed, as in all Haplo- chromis species, mainly from the parasphenoid and basioccipital bones, but there is also a small contribution from the prootics as well (Text-fig. 44D). The jaws and oral dentition are essentially those of a generalized species like H. pallidus (see Text- figs 33 and 43). What differences there are lie within the limits of variation resulting © from allometric growth; individuals of H. ishmaelt and H. pharyngomylus reach a much larger size (125-135 mm) than do those of the ‘pallidus’ insectivore group (ca 100-105 mm). Superficially, H. obtusidens closely resembles H. ishmaeli and H. pharyngomylus (see Greenwood, 1960), but its pharyngeal mill is at an intermediate level of hyper- trophy, the lower bone moderately stout and its dentition partly molarized ; see Text-fig. 44B. The paryngeal apophysis on the skull is also at an intermediate level of specialization. EVOLUTION OF A CICHLID SPECIES FLOCK 73 Fic. 42. Haplochromis riponianus. (Natural size.) The skull of H. obtusidens taken as an entity is, however, more like that of H. empodisma (cf Text-figs 32 and 43), an insectivorous, bottom-detritus eating species (see p. 63 above). Thus, although I have previously indicated a close relationship between HZ. obtusidens and the H. tshmaeli pair (Greenwood, 1960) and therefore a possible phyletic lineage of increasing specialization, I now suspect that this was an error. Recent analysis of various character complexes suggests that although H. empodisma and H. obtusidens are related (and are members of the ‘empodisma’ lineage, see p. 63) and although H. ishmaeli and H. pharyngomylus are also inter- related, the latter species pair belong to a distinct lineage (Text-fig. 70). The most likely plesiomorph sister species for H. tshmaelt and H. pharyngomylus is H. humilior (Text-figs 5C and 43; see also Greenwood, 1960). In H. humulior the pharyngeal mill is moderately developed (and shows a high degree of intraspecific variability, see Greenwood, op. cit.) and the pharyngeal apophysis shows some depar- ture towards the ‘zshmaelt’ condition (an increased area of basioccipital in the facet ; see Text-fig. 44C). But the overall form of the neurocranium in H. humilior is of the basic insectivore type, with a short and decurved preorbital face (see Text-fig. 43). Apart from the larger pharyngeal apophysis, the neurocranium in H. zshmaeli is identical with that of H. humilior (Text-fig. 43). Haplochromis humlior could well be derived from a species like H. pallidus, a taxon in which there is already some slight hypertrophy of the lower pharyngeal teeth (see p. 59). The close correlation that can exist between the relative massive- ness of the pharyngeal bones (and their teeth) and the degree to which the neuro- cranial apophysis is developed has been clearly demonstrated in both natural and experimental populations of the mollusc-crushing Astatoreochromis alluaudi (Green- wood, 1959a, 1965c). It is also apparent in ontogenetic series of H. tshmaeli and H. pharyngomylus (personal observations). Thus the interspecific differences seen in that region of the skull in H. pallidus and H. humilior on the one hand and H. ishmaeli and H. pharyngomylus on the other would seem to be associated with in- creasing hypertrophy of the pharyngeal mill. 74 P. H. GREENWOOD Zane if H. humilior peo ( \ d \ | H. ishmaeli H. obtusidens H. obtusidens Fic. 43. Neurocranial and jaw form in some mollusc-eating species (all pharyngeal crushers). Arrow indicates the apophysis (formed from the parasphenoid and _ basi- occipital bones) on which the upper pharyngeal bones articulate (see Text-fig. 44). (Scale =3 mm.) All the species considered so far are those in which Mollusca (especially gastropods) are the chief food source. Two other species, H. riponianus and H. theliodon (Greenwood, 1960), are mixed insect—mollusc eaters. Haplochromts riponianus shares the syncranial (Text-fig. 34) and dental charac- teristics of H. saxicola but with a more consistent hypertrophy of the median tooth EVOLUTION: OF A CICHLID SPECIES FLOCK 75 rows on the lower pharyngeal bone, and generally some enlargement of the bone as well. As in all other species with similar characteristics of the pharyngeal mill, the degree of bone and tooth hypertrophy is size correlated. Phylogenetically, H. viponianus is clearly the apomorph sister species of H. saxicola. Lack of sufficient material for anatomical studies makes it impossible to suggest the relationships of H. theliodon (see discussion in Greenwood, 1960). In brief, at least four distinct lineages are involved in the mollusc-eating adaptive radiation ; each can be traced back to an essentially insectivorous level still repre- sented by different species in the flock (Text-figs 65 (3), 66, 68 (2) and 70). Predators on larval and embryo cichlid fishes This, the paedophage radiation, was described at length on pp. 31-37. There is a clear-cut diphyletism in the group, the branches comprising H. obesus, H. Fic. 44. Ventral view of the neurocranial apophysis for the upper pharyngeal bones, showing the relative increase in the size of the basioccipital facets correlated with increasing stoutness of the pharyngeal bones and teeth (see Text-fig. 5). (Scale = 3 mm.) A: H. empodisma (typical of condition found in most species). B: H. obtusidens (a species with slightly enlarged pharyngeal bones and teeth). C: H. humilior (with moderately enlarged pharyngeal bones and teeth). D: H. ishmaeli (massive pharyngeal bones and teeth). 76 P. H. GREENWOOD maxillaris and H. melanopterus (Text-figs 3 and 17), and H. cryftodon, H. microdon and H. parvidens (Text-figs 3 and 18) respectively (Greenwood, 1959b). There are two other paedophagous species, H. cronus (Greenwood, op. cit.) and H. barbarae (Greenwood, 1967) whose lack of anatomical specializations relative to the main lineages make their relationships difficult to determine (Text-figs 16 and 29). In both lineages the anatomical specializations involve, particularly, development of a large mouth that is both protrusible and laterally distensible, a reduction in the number of jaw teeth, together with their restriction to the anterolateral parts of the H. obesus H. barbarae SES H. microdon Fic. 45. Neurocranial form in paedophagous species (see text p. 75; also Text-fig. 46). (Seale)—=) 3mm.) EVOLUTION OF A-.CICHLID SPECIES FLOCK 77 jaw bones, and their almost complete burial in the thickened oral mucosa. (In one species, H. maxillaris the teeth are further covered by an inward curvature of the upper lip tissue.) The functional anatomy of the jaws in these species has not been adequately analysed. Dissection does not reveal any marked departure from a typical arrange- ment and interrelationship of the jaw bones, ligaments and muscles. However, in all species the ascending premaxillary processes are long, the maxilla is rather bullate distally, and in all except H. obesus and H. maxillaris, the lower jaw shows a marked narrowing at about the middle of its length. As a result of this latter peculiarity, the lower jaw closes within the upper jaw. Gauged from the effects of manipulation on freshly dead fishes, the mouth is more protrusible, and there is greater lateral displacement of the upper jaw in members of the H. cryptodon lineage than in those of the H. obesus line. Haplochromis obesus, H. maxillaris and H. melanopterus are characterized by all or most of the outer dentary teeth having the cusp inclined outwards (Text-fig. 460A). This tooth shape is found in no other Lake Victoria Haplochromis species. Haplo- chromts obesus has a neurocranium rather like the generalized type (Text-fig. 45). That of H. maxillaris (Text-fig. 45) has a slightly straighter preorbital face, in this respect approaching the H. empodisma skull type (see Text-fig. 32). Despite this difference in neurocranial shape, the shared peculiarity in lower jaw tooth shape seems to exclude the possibility of there being closer relationships between H. empodisma and H. maxillaris than between the latter species and H. obesus or H. melanopterus. Judging from the relative extent of jaw protrusibility in H. maxillaris (possibly correlated with the flatter skull profile) this species is more specialized than H. obesus. Regrettably no skeletal material of H. melanopterus is available, and the species is included in the H. obesus lineage mainly on the characteristic shape of its lower jaw teeth (see Greenwood, 1959b). Neurocranial shape in the H. cryptodon and H. parvidens lineage (Text-figs 45 and 68) is essentially of the type found in the less extreme members of the ‘prognathus’ group of piscivorous predators (see Greenwood, 1967, and pp. 85-89 below). Haplo- chromis parvidens (Text-fig. 45) has the most ‘prognathus’-like skull, H. microdon and H. cryptodon the least modified in that direction (Text-fig. 45). Dentary shape is correlated with neurocranial shape, H. parvidens having the most extreme degree of anterior narrowing (Text-fig. 46C). It is this unusual and unique shape of the dentary, combined with overall neurocranial morphology, that provide the principal evidence for the presumed monophyly of the group. Morphologically, the oral teeth in these fishes are like those of typical piscivorous predators in the ‘serranus’ and ‘prognathus’ groups (see Greenwood, 1960, 1967 ; and pp. 82-89 below). The maximum adult size reached by members of both paedophage lineages is in the range of 130-170 mm ; that is, these are among some of the larger Haplochromis species in Lake Victoria. Haplochromis barbarae (Greenwood, 1967) and H. cronus (Greenwood, 1959b) are included as members of the paedophage group solely on the basis of their feeding habits. Neither species shows any of the dental or oral specializations of either the H. obesus or H. cryptodon lineages. 78 P. H. GREENWOOD Fic. 46. Lower jaw in two paedophagous species. A: H.obesus. BandC: H. parvidens. A and B in lateral view, C in ventral view to show characteristic outline of the dentary in H. parvidens-group paedophages. (Scale = 3 mm.) In H. cronus the outer jaw teeth are moderately stout unicuspids, not noticeably reduced in number or size, nor restricted in their distribution nor deeply embedded in the mucosa. Little is known about the syncranial architecture of the species, except from partial dissection of one fish. Judging from information gleaned in this way, H. cronus does not have the syncranial characteristics of any piscivorous © group. Like H. tshmaeli, which it resembles in many ways, H. cronus seems to share the syncranial morphology of a generalized insectivore, but it is larger than any of these fishes (adult sizes range to at least 135 mm). Unlike the majority of Lake Victoria species, irrespective of their trophic associations, H. cronus has almost the entire surface of the caudal fin densely covered with small scales (Greenwood, 1959b). As H. cronus seems to have the syncranial architecture and dentition of a general- ized Haplochromis, it could be ranked as the plesiomorph sister group to the H. EVOLUTION OF A CICHLID SPECIES FLOCK 79 obesus lineage. These fishes, more than those of the H. cryptodon lineage, have retained a near-generalized skull and jaw morphology (despite the slight but obvious specializations of the latter). For these reasons, the relationships of the H. obesus lineage cannot be defined readily. It could be considered either an independent off- shoot from the basal complex of insectivorous species (p. 63 above) or as a derivative of the A. cinereus—H. squamulatus line. Haplochromts barbarae, on the other hand, has the syncranial architecture (Text- fig. 45) of the ‘altigenis’ predator group (see p. 82), and the type of oral dentition found in small individuals of that species assemblage (adult H. barbarae are 85-106 mm long). Basically, skull morphology in members of the ‘altigenis’ group is not greatly different from that in less specialized species of the ‘prognathus’ group (Greenwood, 1962, 1967 ; also pp. 82 & 85 below). It will be recalled that the neurocranial form in the H. crypiodon lineage is also like that among basal members of the “prognathus’ group (see above, p. 77). Thus, in view of its feeding habits and of its anatomical specializations, H. barbavae could be included in the H. cryptodon lineage as the plesiomorph sister group of the other species combined. Fryer & Iles (1972) believe that there is ‘. . . lack of striking diagnostic indications of feeding habits among these species .. .’ and that this ‘. . . is not really surprising’. They go on to argue that the collection of such ‘soft morsels’ as fish embryos and larvae does not require the evolution of a specialized anatomy in the predator. Furthermore, Fryer & Iles believe the morphological variation found in paedophage species suggests ‘. . . dissociation of structure and function ...’. I fail to find any evidence supporting a single one of these ideas. In both paedophage lineages there is ample evidence for anatomical differentiation in jaw morphology, especially in the H. cryptodon lineage (see above), and in the buried dention of all species (a specialization that Fryer & Iles themselves acknow- ledge). Finally, the manifestation of these various anatomical traits is very obvious to anyone who has handled fresh specimens and compared the extent of their jaw protrusion with that in most other Lake Victoria Haplochromis species. (Some predator species are excluded from this generalization, but these belong to lineages whose basic cranial specializations are greater than those of the paedophages.) As for the postulated dissociation of structure and function, surely the anatomical evidence negates any such idea in its entirety? I find it surprising that Fryer & Iles should think that there are no diagnostic indications of feeding habits among the paedophages. Once one has associated the morphology of any paedophage species with its diet, it is remarkably easy to recognize any other paedophage at sight, or so has been my experience and that of others working in the field, even when dealing with the species of another lake. The intraspecific variability seen in paedophagous species (Greenwood, 1959b) is high, but no higher than that in many other species. It is perhaps the generally bizarre cranial morphology of these fishes that accentuates this variability. I can find nothing to suggest, as Fryer & Iles imply, that paedophage anatomy is not directly correlated with feeding habits (i.e. is non-adaptive) and that there has been a consequent relaxation of selection pressure leading to greater deviation from the mode in these characters. Fryer & Iles’ views, I suspect are coloured by their 80 Pon GREENWOOD belief in the idea that paedophages do not actively obtain their prey from a parent fish, but snap up voluntarily jettisoned young and embryos (see p. 33 above). I find myself more in agreement with Fryer & Iles (op. czt.) when it comes to the question of how paedophagous habits originated. The habit probably stems from a facultative response to the appearance of young liberated by the parent during the normal course of brood care. That many Hapflochromis species, irrespective of their usual feeding habits, respond to the sudden appearance of small objects in the water is often demonstrated in Lake Victoria. On several occasions at Nasu Point beach near Jinja, every species present was found to be gorged on termites after a heavy hatch of these insects had been carried into the lake. On other occasions all species (including specialized piscivores) had ingested large quantities of colonial blue-green algae that were floating in the water after a period of rough weather (unpublished personal observations). Whatever the origins of paedophagy, there are now two phyletic lines of paedo- phagous species in Lake Victoria (the only lake, apart from Lakes Edward and — George, in which the habit has evolved [Greenwood, 1973a]). One lineage (H. obesus and related species) evolved from a basic piscivore stem, the other (H. cryptodon and allies) from a derived, piscivore line (Text-figs 67, 68 and 70). Piscivorous predators The 35 or more species in this trophic group present the most complicated phyletic puzzle in the whole flock. In part this is attributable to the larger number of species involved, but mainly it is because of the few characters that can be used to determine relationships. Whereas in other groups oral and pharyngeal dentition provide indications of relationship, among predators the dentition is essentially a uniform one of strong, somewhat recurved, unicuspids (Text-figs 4B and 57). Even small individuals have this type of dentition, although in juveniles the teeth are bicuspid and of the generalized type. Body form, too, is of little value. The only character complex seemingly of phyletic value is the syncranium, especially the shape of the neurocranium (Greenwood, 1962, 1967). Difficulties are also encountered when interpreting these neurocranial features, and it seems unlikely that a satisfactory scheme of phyletic interrelationships can be achieved. To start with, it is impossible to tell whether the main groups of predators are of monophyletic origin, even if the concept of monophylogeny is broadened to include origin from more than one species provided the species are themselves sister taxa. The chief diagnostic features of the piscivorous predator grade concern adaptations. for feeding on larger and faster moving objects than are utilized as prey by the other trophic groups. This has involved elongation of the preotic part of the skull, elongation of the jaws (often associated with a marked upward inclination of the gape), and the development of a strong, unicuspid dentition adapted for holding a struggling prey fish. In short, the production of a large-mouthed, streamlined fish of a somewhat greater size than the members of species in other trophic groups. In previous papers (Greenwood, 1962, 1967) three principal groups of piscivorous predators were defined, chiefly on the grounds of neurocranial shape and proportions. EVOLUTION: OF: A CICHEID SPECIES FLOCK 81 H. guiarti H. squamulatus H. martini H. serranus Fic. 47. Neurocranial form in piscivorous predators. The species represented are from the H. guiarti-H. squamulatus and the H. serranus lineages. (Scale = 3 mm.) The first or “servranus’ group (comprising eight species) has a neurocranial form nearest that of the generalized type, but one already showing some elongation of the preotic skull and a marked flattening of the dorsal outline to the preorbital face (Text-figs 47 and 57). The second or ‘altigenis’ group (also of eight species) has a more elongate neurocranial outline, with the preotic face slightly longer and sloping less steeply than in the ‘sevvanus’ type (Text-figs 51 and 57). The third or ‘frogna- thus’ group (11 species, Text-figs 51, 57 and 62) shows further accentuation of the ‘altigenis’ trend, together with an overall narrowing of the skull. A number of 6 82 P. H. GREENWOOD smaller and probably polyphyletic groups, all of uncertain affinities with the major ones, were also recognized (Greenwood, op. cit.). Originally I thought the three major groups were each of monophyletic origin, but with a strong possibility of the ‘altigenis’ and ‘prognathus’ groups stemming from an ancestor not shared with the ‘sevrvanus’ group. However, species discovered recently (Greenwood & Gee, 1969) and further study of the known species have caused me to modify somewhat the views expressed before. To begin with, it seems that the ancestral skull type 1s not of the kind found in H. guiarti and H. brownae (pace Greenwood, 1962). These species have a skull form very like the generalized insectivore type (see pp. 57-59), whereas the basic predator skull type (‘serranus’ group) is, in fact, more specialized and akin to that found in H. saxicola (see p. 59). Haplochromis saxicola is, of course, an insectivore but its whole level of syncranial organization departs from the generalized H. pallidus type toward that of the piscivorous predators (including preotic elongation of the neurocranium, marked flattening of the dorsal preorbital profile and a preponderance of uni- over bicuspid teeth). Individuals of. H. saxicola also reach a larger adult size than do those of the H. macrops-like species ; but adult size may be of secondary importance because H. brownae, an insectivore—piscivore is no larger than H. macrops. The possible relationships of H. brownae and H. guzartt will be discussed later (p. 85). Since H. saxicola (despite its insectivorous habits) has a neurocranial form already specialized towards that of the piscivorous predators, it seems more reasonable to consider some HZ. saxicola-like species as the ancestor of almost the entire piscivorous predator radiation. Of the three predator groups originally defined, I would now consider the ‘alt- gems’ and ‘prognathus’ lines to represent a single phyletic assemblage (henceforth called the prognathus group). In other words, species of the ‘frognathus’ group could be derived both from species at an ‘altigenis’ level of specialization and from species that had already reached the ‘prognathus’ level. This conclusion was reached after reexamination of the two ‘lineages’ had shown not only their great similarity, but also the impossibility of demonstrating, unequivocally, that a species in the more specialized lineage could only be derived from an equally specialized taxon rather than as a somewhat more apomorphic derivative of a species at the ‘altigents’ level. The phyletic integrity of the ‘serranus’ group, on the other hand, remains un- changed, although its membership has to be increased (see p. 84 below). There is no evidence to suggest that any member of the ‘prognathus’ lineage might have been derived from a ‘serranus’-like species. The specialized neurocranial form in the ‘pbrognathus’ lineage is more readily derived from that of the presumed H. saxtcola- like common ancestor of the two lineages than through a ‘sevranus’-like form. Neurocranial shape in the ‘servranus’ lineage is less variable interspecifically than it is in its apomorph sister group (‘prognathus’), and the general facies of its con- stituent species is also more uniform (cf Text-fig. 68 (3) with Text-figs 68(4) and 69). Determining the intrarelationships of species within the two lineages is very difficult, and the results presented here must be considered extremely tentative. EVOLUTION OF A CICHLID SPECIES FLOCK 83 Peat Fic. 48. Haplochromis guiarti. (About % natural size.) Fic. 49. Haplochromis squamulatus. (About half natural size.) Fic. 50. Haplochromis martini. (About natural size.) 84 P. H. GREENWOOD H. paraguiarti H. flavipinnis H. xenostoma Fic. 51. Neurocranial form in piscivorous predators. (Scale = 3 mm.) To the eight species originally placed in the ‘sevvanus’ group (H. serranus [Text-fig. 11], H. victorianus, H. nyanzae, H. speki [Text-fig. 12], H. maculipinna |Text-fig. 47], H. boops, H. thuragnathus and H. pachycephalus ; see Greenwood, 1967), three others should be added, viz H. cavifrons [Text-fig. 3], H. plagiostoma [Text-fig. 47] and H. decticostoma (see Greenwood, 1967; Greenwood & Gee, 1969). Neurocranial morphology in all three species (Text-figs 47 and 57) is typically that of the ‘serranus’ lineage, despite the rather atypical general facies of H. cavifrons and H. plagiostoma (Text-fig. 3). Indeed, apart from these two species the ‘sevvanus’ lineage has a EVOLUTION OF -A-CIGHLID SPECIES FLOCK 85 remarkably uniform facies. Haplochromis cavifrons and H. plagiostoma not only differ in their gross appearance (from one another as well as from other species) but also have coloration that is outstandingly different. Because of the few noticeably specialized features shown by its members, intra- lineage relationships are not at all distinct. Once again A. cavifrons and H. plagio- stoma are outstanding (because of their oblique jaws), but there is little else to indi- cate a particularly close relationship between them. Certainly their coloration would belie any such suggestion (see Greenwood, 1962). As mentioned earlier, Haplochromis guiartt (Text-fig. 48), formerly considered a member of the ‘altigenis’ group (see Greenwood, 1967), has a neurocranial shape little different from that of the generalized H. pallidus type (see Text-fig. 47 ; and pp. 58-59 above). What specialization there is, is manifest in the somewhat more elongate preotic skull, and the slightly less curved preorbital skull roof. The denti- tion in H. guzarti is, however, predominantly unicuspid in fishes of a size that in H. pallidus-like species would have bicuspid teeth only. In this respect H. guzarti is a more specialized (and larger) form of H. brownae (Greenwood, 1962), a species little removed in neurocranial form and in dentition from H. pallidus. Thus, H. gwiarti and H. brownae could be sister species, apomorph derivates from the H. pallidus—H. macrops species complex. Three other piscivorous (or predominantly piscivorous) species, H. michaeli, H. squamulatus (Text-fig. 49) and H. martini (Text-fig. 50), have a neurocranial shape close to that of H. gwiarii (see Text-figs 47 and 67 ; also Greenwood, 1962, 1967, 1960 for the species respectively). In H. martini (maximum length 104 mm) the outer series of jaw teeth are mostly bicuspids, with a few unicuspids intercalated ; in H. squamulatus (a larger fish, maximum size 198 mm) a similar admixture of teeth is found in fishes less than 115 mm long, but unicuspids predominate in larger fishes, while in H. michael1 (maximum size 145 mm, smallest known specimen 117 m) all the outer teeth are unicuspid. All three species have very small scales on the pectoral region (a character rarely encountered in the Lake Victoria species), and a well-developed, apparently in- variable, midlateral stripe from the opercular margin to the caudal fin origin or slightly beyond (another unusual feature for this species flock). Furthermore, both H. martim and H. squamulatus have a distinctly yellow ground coloration, also an uncommon feature. (The live colours of H. michaels are not known.) Although neurocranial shape (Text-fig. 47) in these species (and in H. guzarti) 1s more like the generalized than the ‘serranus’ type, it is nevertheless a derived form (see above). Thus H. guzarts and the species discussed above may represent a true phyletic lineage, with H. squamulatus, H. martint and, H. michaeli more closely related to one another than any one is to H. guiarti (Text-figs 67 and 70). The origin of this presumed lineage, unlike that of the ‘serranus’ and ‘prognathus’ lineages, is probably from an H. macrops- or H. pallidus-like species, for example a species akin to H. brownae which is part insectivore and part piscivore (Greenwood, 1962). , _ The 20 species comprising the ‘ prognathus’ lineage (i.e. the combined ‘ prognathus— alugents’ groups of Greenwood, 1962 and 1967; see p. 82 above) all have a more GREENWOOD Peon 86 (About half : HA. flavipinnis. B (About natural size.) : H. percoides. A Fic. 52. ) 1Ze natural s Haplochromis paraguiarti. 53° FIG EVOLUTION OF A CICHLID SPECIES FLOCK ae ei Tire Fic. 54. Haplochromis pseudopellegrini. Fic. 55. Haplochromis gilberti. Fic. 56. Haplochromis xenostoma ; a juvenile fish. (About half natural size.) 87 P. H. GREENWOOD H. gowersi H. mento Fic. 57. Neurocranial and lower jaw form in piscivorous predators. The neurocrania are from members of the H. prognathus lineage, the lower jaws from members of the H. sevvanus (left) and H. prognathus (right) lineages. (Scale = 3 mm.) Fic. 58. Haplochromis pellegrint. (About natural size.) EVOLUTION OF A CICHLID SPECIES FLOCK 89 specialized skull form than that occurring in the other two predator lineages (see Text-figs 51, 57, 68(4) and 69). The preotic length of the neurocranium is further elongate (65-68 per cent of neurocranial length, cf 60-65 per cent for the other lineages), the brain-case is shallower (28-32 per cent neurocranial length, cf 34-40 per cent), and the preorbital profile slopes at a much smaller angle (cf Text-figs 68(3) and 68(4) ; also Text-figs 47 and 51). Within the lineage a gradual intensifica- tion in this trend can be detected. Two species (H. bayont and H. dentex [Text-fig. 63]; see p. 92 below) in the ‘prognathus’ lineage can be grouped together on the basis of their having a skull in which the ethmo-vomerine region is noticeably decurved, although the rest of the skull retains the form typical for this lineage (Text-fig. 62). Apart from these two species, all the others can, on the basis of increasingly specialized skull form, be collected into four subgroups. The phyletic interrelation- ships of the subgroups are but vaguely discernible. Indeed these categories show in cameo the difficulties involved in studying the phylogeny of the whole flock. Essen- tially, the problem is whether these intralineage groups are true, hierarchically evolved, sister groups (showing increasing apomorphy), or whether they represent gradal assemblages of polyphyletic ancestry (the ancestral species of each having, of course, reached the ‘prognathus’ level of apomorphy). For the moment, this prob- lem seems insoluble. Species in each subgroup show some variation in gross morphology, although most are elongate, relatively slender fishes (the adult size range 140-200 mm), with a large mouth in which the inner and outer teeth are strong and unicuspid (even in small fishes, although some bicuspid outer and tricuspid inner teeth are found in the smallest fishes where these are known). The first subgroup (‘pavaguiartt’) has a neurocranial shape nearest that charac- terizing the ‘servranus’ lineage (Text-figs 51 and 68(4); see p. 84). Of its seven constituent species, four, H. paraguiarti (Text-fig. 53), H. gilberti (Text-fig. 55), H. pseudopellegrimt (Text-fig. 54) and H. altigenis are fairly similar in appearance, having a slightly oblique mouth and moderately slender body (Greenwood, 1967 ; Green- wood & Gee, 1969). Other species depart from this morphotype. Haplochromis artaxerxes has greatly elongate pectoral fins and a near horizontal mouth (Greenwood, 1962), H. xenostoma (Text-fig. 56) has a markedly oblique mouth (40-45 degrees with the horizontal) and a rather deeper body (Greenwood, 1967), while H. favipinnis (Text-fig. 52B) also has an oblique mouth but with a deeply concave dorsal head profile (convex or straight in the other species ; Greenwood, 1962). Neurocranial form in members of the ‘faraguzarti’ subgroup is shown in Text-fig. 68(4). ; - Morphologically, species of the second subgroup (‘prognathus’, containing five species) are somewhat less uniform than those of the ‘faragwiarti’ subgroup. Haplochroms prognathus (Text-fig. 13 ; Greenwood, 1967) and H. barton (Greenwood, 1962) are closely similar species, but H. mandibularis (Greenwood, 1962) has a nar- rower head, deeper cheek and more prognathous lower jaw, giving it a distinctive appearance seen again in a species of the following subgroup. Hapflochronus pellegrint (Text-fig. 58 ; Greenwood, 1962) is of interest because of its small adult P. H. GREENWOOD Fic. 59. Haplochromis macrognathus. (About half natural size.) Fic. 60. Haplochromis longivostris. (About half natural size.) GL ee & = Fic. 61. Haplochromis argenteus. EVOLUTION OF A CICHLID SPECIES FLOCK gl size (7I-105 mm). Inits gross morphology H. fellegrinz is rather like H. paraguiarti (of that nominal subgroup) but its neurocranial shape is virtually identical with that of H. prognathus ; in other words, of a more specialized kind. The fifth member, H. percoides (Text-fig. 52A ; Greenwood, 1962) also has small adults (7o-95 mm), but its body shape and head form are quite unlike those of H. pellegrini ; in these characters H. percoides closely approaches H. flavipinnis (of the ‘paragumiarti’ subgroup), a species with which I previously considered it to be closely related (Greenwood, 1967). Like H. flavipinnis, H. percoides has a distinctive and unique colour pattern (cf Text-figs 52A and B). Neurocranial form in members of the ‘prognathus’ subgroup is shown in Text-figs 51 and 60(1). The five species comprising the ‘estoy’ subgroup are perhaps the most striking and obviously predatory Haplochromis in the ‘prognathus’ lineage. All have what may be considered a ‘full’ development of the skull type characterizing the lineage (Text-figs 57 and 69(2)), and all except H. longirostvis (maximum adult size 145 mm) are among the larger piscivorous Haplochromis species in Lake Victoria. Three species, H. gowersi, H. mento (Text-fig. 15) and #. estor (Text-fig. 14) are rather similar in appearance (Greenwood, 1962).