PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON 116(2):356-379. 2003. Geographic variation in Poecilia Bloch and Schneider, 1801 (Teleostei: Poeciliidae), with descriptions of three new species and designation of lectotypes for P. dovii Giinther, 1866 and for P. vandepolli van Lidth de Jeude, 1887 Fred. N. Poeser Zoological Museum of Amsterdam, Department of Ichthyology, P.O. Box 94766, 1090 GT Amsterdam, The Netherlands — Abstract. The South American cyprinodontiform fish species with the ver- nacular name "mollies" are analyzed and three new species of the genus Poe- cilia are described and figured, viz., P. boesemani, n. sp. from Port of Spain, Trinidad, P. koperi, n. sp. from coastal areas of Venezuela and Colombia, and P. wandae, n. sp. from the Zulia district, Venezuela, west of Lake Maracaibo. Thirteen species of Poecilia are presently recognized from the northeastern part of South America and adjacent islands, viz., six mollies (including P. vivipara and P. caucana), five micropoeciliids, P. heterandria and the guppy {P. retic- ulata). Two different populations in both P. koperi and P. wandae are recog- nized, based on variation in pigmentation. From the type series of P. dovii Giinther, 1866 and P. vandepolli Van Lidth de Jeude, 1887 lectotypes are se- lected. Four subspecies of P. gillii are recognized as morphologically and geo- graphically distinct populations. Poecilia reticulata, P. heterandria, P. bifurca, P. parae, P. picta, P. branneri and P. minima are not considered herein, whereas P. laurae remains a species A inquirendae. geocline in characters of the species of Poecilia is recorded and an evolutionary scenario is discussed. This paper seeks to extend our biogeo- sphenops Valenciennes, 1846 (Regan 1913: graphic knowledge of the genus Poecilia 1013; Price 1955:18; Boeseman 1960:122); & Bloch Schneider, 1801 and describes P. vandepolli van Lidth de Jeude, 1887 (de three new species from the southern end of Beaufort 1940:111; Schultz 1949:84, 97- the range of this genus. The present study 99; Feltkamp and Kristensen 1969); or records seven species from the northeastern mentioned as 'unidentified species' (Rosen & part of South America. Whereas the number Bailey 1963:48). of poeciliid species constitutes a dominant Regan (1906-08) examined variation part of the Central American fish fauna, this within the species of Poecilia and extended is not the case in South America. Miller our knowledge of the expression of mor- (1983) constructed a key and checklist for phology and pigmentation (Regan 1906-08, the Mexican species, but there are no com- fig. 13), as well as the geographic range prehensive lists for the species of Poecilia (from northern Mexico to the Leeward Is- from Central and South America. Speci- lands) of the P. sphenops complex, mens of the species herein described as With the discovery of the structures of new, viz., P. koperi, P. boesemani and P. the gonopodial tip as a taxonomic tool (Ei- wandae, were previously identified as either genmann 1907), Regan (1913) reorganized & P. vivipara Bloch Schneider, 1801; P. species groups in the subfamily Poeciliinae. VOLUME NUMBER 1 16. 2 357 With respect to the P. sphenops complex, geographically, and environmentally and, he allocated P. sphenops to the genus Mol- therefore, they are of little or no value in lienesia on the basis of identical gonopodia distinguishing species (except in relative fin ofM. latipinna and M. sphenops. Moreover, position)." Extensive and detailed investi- the number of dorsal fin rays ofM. formosa gations (Schultz and Miller 1971, Menzel was intermediate between the numbers and Darnell 1973, Miller 1975) resulted in found in M. latipinna and M. sphenops, a checklist for the Central American mollies which therefore rendered the character in- (Miller 1983), which included eight species valid. Although Regan did not mention P. of Central American short-finned mollies. vandepolli, the geographic range of his M. No species were mentioned from outside sphenops still included the Leeward Is- this range. Comparisons of the Mexican lands, i.e., the Lesser Antilles. species to those of Central and South Amer- The importance of the gonopodial tip ica (Poeser 1992, 1995, 1998) have led to was also recognized by Hubbs (1924), who the present study. used the gonopodium as "the chief distinc- tive feature of the Poeciliidae (as here de- Methods limited)." The M. sphenops group was de- fined by Hubbs (1926) as "the multitude of Some 17 lots of Poecilia vandepolli, in- races allied to, or inseparable from M. cluding the type series ofP. vandepolli van- sphenops.'' This group was separated from depolli and P. V. arubensis, have been re- M. latipinna and allies by the number of examined. From the type series, a lectotype dorsal fin rays. For M. sphenops, Hubbs is selected. Meristic data and innerjaw den- was "unable to delimit, in either distribu- tition (cf. Schultz and Miller 1971) received tion or in characters, any multitude of ele- special attention. Vernier calipers were used mentary species making up the sphenops to record distances to 0.1 mm. The average complex of Middle America. The problem of the measurements and counts are given of determining the relationships of the di- verse types, many of which have received and compared to a similar study of Felt- kamp and Kristensen (1969) and to the data specific names, is in prospect a most fas- of P. vivipara and of P. gillii. cinating study." In the description of the new species, The situation remained promising, even after the revision of Rosen and Bailey proportional morphometric values (Table 2) (1963), who listed a total of 35 nominal are recorded in thousandths of the standard species in their synonymy of P. sphenops. length (SL), following Miller (1975). The Schultz and Miller (1971) mentioned: ". small size of P. wandae made measure- . . a thorough study of the whole complex and ments difficult to impossible and therefore the type specimens will be required before some were omitted. Meristic characters fol- systematic units and the nomenclature can low Hubbs and Lagler (1947). In the type be convincingly coordinated." Species rec- series of P. wandae, all specimens have ognition was aided by dental analyses damaged caudal fins; therefore, the caudal (Schultz and Miller 1971), i.e., unicuspid fin ray count is estimated. Terms concern- and tricuspid species were recognized. Oth- ing the shape of inner jaw dentition follow er alpha-taxonomic features (meristic and Garman (1895). Melanophore pigmentation morphologic characters, body and fin color, of preserved specimens and gonopodial even allozyme data) remained confusing. structures are also recorded (Figs. 1 to 4). Rivas (1978) remarked that "proportional The combined data are used to evaluate body measurements are [to be] omitted relationships between the newly described . . . , A there is considerable variation in characters species. key to the species is provided in individually, ontogenetically, seasonally. Appendix 1. 358 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON The following abbreviations are used in ciennes, 1821:158 (type locality: Brazil, the species diagnosis provided below. Rio de Janeiro). A = anal fin rays Poecilia schneideri Humboldt & Valenci- BS = scales around the body ennes, 1821:159 (type locality: Surinam). CPD = least depth of the caudal peduncle Neopoecilia holacanthus Hubbs, 1924:11 CPS = scales around the caudal peduncle (type locality: Puerto Rico, introduced). D = number of dorsal fin rays — ZMA G = Material West Indian Antilles: gonopodial ray LLS = 120.384 (22), St. Lucia, freshwater pool, lateral scales P = pelvic fin rays ±1 km west from Fort Vieux, coll. J. H. PL = predorsal length Stock, 20-n-B1M97N4.H PS = predorsal scales Guyana: 1974. 10. 10. 527-619 SL = standard length (7 out 93), Georgetown, Seawall trench, no date; CAS 59364 (12), Georgetown trench- es, C. H. Eigenmann 1908; CAS (SU) Systematics 21784 (2), same data as CAS 59364; The genus Poecilia is diagnosed by the USNM 66278 (2), same data as CAS shape of gonopodial ray 4p, which has six 59364; ZMA 100.629 (2), same data as to fourteen unserrated distal segments, fol- CAS 59364; ZMA 119.912 (3), East Coast lowed by dorsally serrated subdistal seg- Demerara, Turkeyen, drainage canal, coll. ments (cf. Miller 1975). The genus consists M. Tamessar, 28-M987; ZMA 119.913 (6), of several ill-defined subgenera. A more de- East Coast Demerara, Bel Air, stagnant ZMA tailed analysis is in progress. ditch, coll. M. Tamessar, 28-L1987; In the subgenus Poecilia, ventral spine- 119.917 (4), East Coast Demerara, Indus- like serrae are present on the third gono- try, shallow pool, coll. M. Tamessar, 28-1- podial ray. Gonopodial rays 4a and 4p are 1987; ZMA 121.005 (5), Guyana, coll. E very similar, as are rays 5a and 5p. The P. Vermeulen, no date; ZMA 121.007 (15), sphenops species group has terminal seg- Guyana, coll. F Vermeulen, 1992; ZMA ments of gonopodial ray 4a and of gono- 121.009 (5), Guyana, coll. E Vermeulen, podial ray 4p close to each other, giving the 1992. gonopodium a sharp appearance. The mem- Surinam: MNHN B. 932 (5 syntypes of bers of this species group are moderately P. surinamensis), Surinam River, Levail- sized to large, with mnumptial specimens usu- lant, no date; MNHN 4391 (48 syntypes of ally larger than 35 SL. The members P. surinamensis), Cayenne, Rousseau, no of the P. caucana species group are defined date; MNHN B. 918 (60), without exact lo- by the relative thickness of gonopodial ray cality, Duvernoy, no date; RMNH 18516 4p versus 4a. The tips of these rays are (2), Surinam, coll. J. Th. Noordijk, August split, making the gonopodium bmlumnt. Nup- 1887; ZMA 105.332 (2), Paramaribo-west, tial specimens are less than 35 SL. garden canals, coll. H. Nijssen, 12-XII- ZMA & 1966; 106.757 (16), Marowijne Riv- Poecilia vivipara Bloch Schneider, er, coastal plain on Tijgerbank, west of 1801 ZMA mouth, coll. H. Nijssen, l-IV-1966; Poecilia vivipara Bloch & Schneider, 1801: 115.118 (45), Surinam district. Lagoon 7 & 452, pi. 86, Fig. 2 (type locality: Suri- 8, 05°59'N, 54°49'W, coll. M. P Panday, ZMA nam). 22-IV-1974; 115.121 (20), same data & ZMA ZMA Poecilia surinamensis Humboldt Valen- as 115.118; 119.908 (6), Suri- ciennes, 1821:158 (type locality: Suri- nam River, pool near beach at Braamspunt, ZMA nam, French Guyana, Brazil). coll. I. Kristensen, 14-in-1960; Poecilia unimaculata Humboldt & Valen- 121.000 (4), Paramaribo, Mason Street, pol- VOLUME NUMBER 116, 2 359 & luted trench, coll. F. Vermeulen W. Suy- holocanthus, correcting his misidentifica- ker, 24-III-1991. tion. USNM French Guiana: 121833 (9), Cay- enne, Carriera Guila, S. F YoUes 1945; Poecilia mexicana Steindachner, 1863 USNM 121834 Cayenne, Pont Magie, (2), USNM Poecilia mexicana Steindachner, 1863:178, S. F Yolles 1945; 121835 (9), Cay- pi. 4, fig. 1, la (type locality: Mexico, enne, Ruisseau de I'institute-eau douce, S. USNM Orizaba). F Yolles 1945; 149938 (4), Cay- Poecilia thermalis Steindachner, 1863:181, enne, S. F Hildebrand 1945. RMNH plate 4, fig. 3, 3a (type locality: Central Brazil: 2724 (2 syntypes of P. America, warm springs). unimaculata), Rio de Janeiro, coll. Dela- ZMA IGambusia (?) modesta Troschel, 1865:105 lande, no date; 100.628 (1), Brazil, ZMA (type locality: Mexico). coll. Moesch, 1884; 116.210 (14), Rio Grande do Norte, Rio Agu (Rio Assu), IGambusia {!) plumbea Troschel, 1865:106 (type locality: Mexico). downstream of Macau, coll. R. Boddeke, Poecilia chisoyensis Giinther, 1866:342 10-1-1979. Diagnosis.—A == 8; D = 7; C = 16—18; (type locality: River Chisoy, Vera Paz). LLS = 24—26; CPS =16. Poecilia vivipara Poecilia dovii Giinther, 1866:344 (in part; may have a midlateral spot or blotch in both type locality: Mexico; lectotype is P. gil- lii). sexes, a unique character in the subgenus. & Poecilia limantouri Jordan Snyder, 1900: The gonopodium has no extruding hooks or 116-117, 129-131 (type locality: Tam- spines, gonopodial ray 4a with serrae on pica, Tamaulipas, Mexico). dorsal surface. — Mollienesia sphenops vantynei Hubbs, Description. Medium sized species, mm 1935:11, plate 2, fig. 1 (type locality: mature males smaller than 50 and fe- Guatemala, Uaxactum, Rio Hondo). males smaller than 60 mm. The body is Mollienesia sphenops macrura Hubbs, truncate and displays dark stripes on the 1935:12, plate 2, figs. 2-3 (type locality: sides. The caudal and dorsal fins have black Guatemala, Rio San Pedro de Martir). markings, with broad yellow margins. The Mollienesia sphenops altissima Hubbs, caudal fin has black margins, at the base 1936:239, plate 9, figs. 1-3 (type locality: broader than at the terminal end. There is Mexico, Yucatan Peninsula, Miramar some variation in the presence or absence Spring). of the spot at the side of the body. No ex- Mollienesia sphenops melanistia Hubbs, tensive records are available of the extent 1937 (type locality: Mexico, Tamaulipas, of this variation. In the populations in Arroyo Marmoleyo). which it is present, the blotch is also prom- BMNH inent in young specimens. The gonopodium Mar^r/a/.—Barbados: 1970.1.29: is figured in Miller (1975), the presence of 1 (1) Barbados, coll. R. Heath, no date; RMNH dorsal serrae on ray 4a are unique within 24814 (55), rivulet near Three Mills the subgenus. — (13°10'N, 59°27'W), coll. I. Kristensen, RMNH Distribution. Poecilia vivipara is found 13.6.1961; 24804 (6), rivulet at in coastal habitats from Venezuela to Ar- Three Mills (13°10'N, 59°27'W), coll. I. RMNH gentina. It is also found on some islands of Kristensen, 13.6.1961; 24809 (5), the Lesser Antilles. rivulet near Three Mills (13°10'N, — Remarks. Garman (1895) diagnosed 59°27'W), coll. I. Kristensen, 13.6.1961. this species and provided a list of syno- Diagnosis.—A = 9; D = 9-10; C = 18- nyms. He also explained the obvious aber- 22; LLS = 26-—27; CPS = 18. rant figure in the original description. Description. Poecilia mexicana is rath- Hubbs (1926) synonymized Neopoecilia er variable in its morphology. This is a trun- 360 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON cate, torpedo shaped species in northern Poecilia mexicana [non Valenciennes, Mexico but ranges to high, laterally flat- 1863]; Poeser, 1992:86 (misidentifica- tened specimens at the eastern end of the tion). range in the Yucatan Peninsula. While the Poecilia thermalis [non Steindachner, 1863] northern populations possess deeply blue Giinther, 1866:341 (type locality: San pigmented bodies, alpha-males of the Yu- Salvador, warm springs). catan populations have a more copper-red Poecilia dovii Giinther, 1866:344 (in part: body. The fin pigmentation is also varying, type locality: Lake of Nicaragua). from totally black at the base of the fin with Platypoecilus mentalis Gill, 1876:335 (type an orange outer area (in northern popula- locality: Isthmus of Panama). tions) to a black and yellow reticulate pat- Poecilia Boucardi Steindachner, 1876:386 tern (on the Yucatan Peninsula). Between (type locality: Colon, Panama). these extremes clinal variation is noted. Poecilia cuneata Garman, 1895:179, plate V In the Barbados material, both males and (type locality: Colombia, Gulf of Ur- females exhibit considerable variation in aba. Turbo); Poecilia sphenops cuneata\ their pigmentation, especially the speci- Hubbs, 1926b:77. RMNH mens in 24804 have many spots on Poecilia salvatoris Regan, 1907:65, plate their body (perhaps a basis for why they 14, figs. 2-3 (replacement name for P. were separated from the other lots). Fe- thermalis Giinther, 1866). males have nine anal fin rays, some females Platypoecilus tropicus Meek, 1907:146 have 10 dorsal fin rays. The membranous (type locality: Costa Rica, Turrialba). hook on gonopodial ray 3 is extremely Poecilia tenuis Meek, 1907:147 (type lo- small, without a bony spine. Several spec- cality: Costa Rica, Tiribi). imens have 20 caudal fin rays, whereas all Poecilia caudata Meek, 1909:209-210 specimens have 18 scales around the caudal (type locality: Costa Rica, Turrubares). peduncle. Poecilia spilonota Regan, 1908:460 (type The diagnostic features of P. mexicana locality: San Jose, Costa Rica). generally most closely agree with the nom- Lembesseia parvianalis Fowler, 1949:267— inal subspecies P. gillii gillii (see below), 269 (type locality: Africa, Congo system. with the exception of the number of scales Oka). around the caudal peduncle (18 versus 16 Mollienesia sphenops petersi Schindler, in P. gillii). — 1956:1-4, fig. 1 (type locality: Honduras, Distribution. Poecilia mexicana occurs lake Yojoa). on the Atlantic coast of Central America, — from the Texas border, through Yucatan into Material. Nicaragua: Lectotype of P. Guatemala and Costa Rica. The population dovii (BMNH 1863.12.16.77, Lake of Nic- found on Barbados is probably derived aragua, coll. captain J. W. Dow, no date). 5 (BMNH from escaped aquarium specimens, which is Paralectotypes of P. dovii, BMNH confirmed by their abnormal variability in 1863.12.16.78-92, same data as body pigmentation. 1863.12.16.77). Costa Rica: 4 Syntypes of P. spilonota Poecilia gillii (BMNH 1907.2.11.44-50, San Jose, coll. P (Kner & Steindachner, 1864) Bidley, no date). NWM Panama: 21608 (1 syntype of Xiphophorus Gillii Kner and Steindachner, Xiphophorus Gillii), Rio Chagres, Panama, & MCZ 1864: in Kner Steindachner, 1865:25 no further data. 29433 (10), 1 mile (type locality: Panama, Rio Chagres); south of Panama City, coll. USFC Steamer Poecilia gillii', Giinther, 1868:395; Poe- Albatross (Alex. Agassiz), 23-X-1904; MCZ cilia sphenops gillii; Hubbs, 1953:145. 33847 (10), ca. 9°43'N, 79°43'W, be- VOLUME NUMBER 116, 2 361 tween Gorgona and Matachin, collected be- Diagnosis.—A = 9; D = 9-10; CPS = fore 1930; MCZ 54068 (10), small tributary 16; LLS = 26-29. In his checklist. Miller of main stream on left about 0.75 km up- (1983) was uncertain if P. gillii was differ- river of bridge, Bayano drainage, probably ent from P. mexicana. Therefore, I have in- & Rio Canita, Panama, coll. W. L. Fink K. cluded a discussion of this species in the USNM E. Hartel, l-IV-1978; 050368 (14), remarks section. — USNM Panama, coll. C. H. Gilbert; 64764 Description. This species is very simi- (17), Folks R. Swamp, Cristobal, coll. A. lar to P. mexicana, with the exception of USNM H. Jenning, 4-VIII-1909; 65618 the number of scales around the caudal pe- (15), 1 mile south of Panama city, coll. Str. duncle. In P. mexicana, 18 scales around USNM Albatross, 23 X-1904; 78837 (27), the caudal peduncle are common. — Upper Trinidad, coll. Meek & Hildebrand, Remarks. All examined type material is USNM 7-III-1911; 247529, (24) canals in consistent with the above meristic data. banana field, Bocas Province, California, Over its extensive range, P. gillii shows USNM coll. Loftin, 2-IX-1962; 247531, considerable variation in body shape and in (44) Rio Gaurumo, Bocas Province, coll. dorsal fin color, and moderate variation USNM Loftin, 18-IV-1963; 247432 (10 of gonopodial features. These variations ap- 50), Canal zone, behind Fort Clayton, res- pear to reflect intraspecific diversification, & identialUSarNeaM, coll. Loftin Tyson, 19-IV- justifying taxonomic separation, i.e., divi- 1962; 247436, (6) Bocas del Toro, sion in subspecies. Carr & Giovannoli Esendo de Verequas island, coll. J. Legler, (1950:17-18) reported the live colors of P. USNM 14-V-1962; 247548, (35) San Bias, gillii and the sympatric P. marcellinoi Poe- small river op&posite to Mulatupo island, ser, 1995 from Honduras: "One male with coll. Loftin Evermann, 2-XII-1962; USNM yellow spots, a black blotch at the caudal 247550, (18) San Bias, Rio Ada, USNM base and spotted dorsal and caudal fin, and coll. Loftin, 16-Xn-1962; 293473, one female with black spotted sides (=P. (10 of 51) 9°14'N, 78°58'W, Rio Tearbles, marcellinoi). One female with yellow spot- Bayano drainage, Panama province, coll. USNM ted sides, and a male with golden-orange W. C. Sternes et al., 25-11-1985; blotches, with a black blotch on the dorsal 293476, (23) 9°28'N, 79°3'W, Comarca Kuna Yala, Rio Mandinga, coll. W. C. Ster- base and rest of fin orange (=P. gillii).'' nes et al., 5-III-1985; USNM 293494 (42), Poecilia gillii salvatoris is reported from El Salvador, where males from most pop- Panama province, Rio Frijoles, above pipe- line rd., N. of Gamboa (Rio Chagres drain- ulations are reported to have red dorsal fins (Hildebrand 1925). This prompted Miller age, Atlantic side), coll. W. C. Sternes et (1994) to redescribe P. salvatoris Regan, al., 26-11-1985. Colombia: 2 Syntypes of P. cuneata 1907, although all other characters are as (MCZ 6458, Turbo, Gulf of Uraba, Colom- found in the present study for P. gillii. Poe- bia, coll. T. Barbour, no date); 3 Syntypes ser (1995) in his redescription of Poecilia of P. cuneata (USNM 120285, same data salvatoris agreed with the diagnosis of P. as MCZ 6458); 5 Syntypes of P. Boucardi salvatoris by Regan (1908), with the excep- (MCZ 32959, San Pablo (Aspinwall (Co- tion of the number of anal fin rays. Regan lon), Quebrada San Pablo), coll. Hassler (1908:104) mentioned 8-9 anal fin rays; & Expedition (Steindachner Mr. Boucard), however, examination of 14 of his syntypes BMNH VII-1872); 99.3.15.27-29 Mon- did not yield any specimen with eight anal (3), key Hill, Colon, coll. Dr. H. Festa, no date. fin rays, so Regan's account is judged er- Additional material from Nicaragua and roneous. In the preserved material, no trace Panama in Poeser (1992), from El Salvador of red was found in the dorsal fins. The red in Poeser (1995). finned western Central American popula- 362 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON tions are here considered as a subspecies, (with 18 scales around the caudal pedun- viz., P. gillii salvatoris. cle), as well as from P. gillii (with 16 scales Villa (1982) provided a key to the genus around the caudal peduncle), it is only par- Poecilia in Nicaragua, in which he recog- tially alluded to the synonymy of P. gillii. nized three species, viz., P. gillii, P. sphen- In honor of captain J. W. Dow, the lectotype ops, and "an undescribed species with uni- is selected from the lot he collected himself cuspid teeth". He mentioned (p. 134, trans- in Nicaragua, which is P. gillii. Giinther lated from Spanish): (1866) mentioned that P. dovii occurred P. gillii: "Unicuspid inner teeth, only in Guatemala and Mexico. However, incomplete supraorbital system, preorbital since the type locality includes Nicaragua, pores free, 26-28 (modally 27) lateral this country should be included in the range scales, 16 scales around caudal peduncle, of distribution. The specimens from Lake 28-30 (modally 30) vertebrae. Guatemala, Amatitlan (BMNH 1865.6.10.13-15) are El Salvador, to Panama". not registered in the British Museum as P. spec: "Unicuspid inner teeth, com- syntypes, and these were not studied. The plete supraorbital system, free preorbital specific status of these types remains to be pores, 28-30 (modally 29) lateral scales, 18 determined. scales around caudal peduncle, 28-30 Poecilia gillii appears to be nearly the (modally 30) vertebrae. Rio Ulya (Hondu- only species of Poecilia present in Costa ras), Lagunas Apoyeque and Xiloa, lake Rica. Meek (1914:116-117) placed all Cos- & Managua and Nicaragua, and Rio Sapoa ta Rican taxa in synonymy with Platypoe- Frio, Tilaran region, Costa Rica". cilus tropicus i=P. gillii) as follows: "The P. sphenops: "Tricuspid inner teeth, in- inland or fresh-water forms of this species complete supraorbital pores, preorbital or variety found in Costa Rica are very var- pores covered, 26-28 (modally 27) lateral iable, and as a result several species have scales, 16-18 scales around caudal pedun- from time to time described. The fe- cle, 28-30 (modally 29) vertebrae. Guate- . . . males and many males of these inland mala and El Salvador, Atlantic side of Hon- forms usually have a black spot at the base duras, Nicaragua including the Great of the middle dorsal rays, .... On many Lakes". specimens from salt and brackish water Villa's diagnosis of P. sphenops fits the some of the scales have a dark spot which description of P. marcellinoi (tricuspid in- forms lines along the rows of scales. On ner teeth, 16 scales around the caudal pe- . . . many of the males, especially from larger duncle). While his description of P. gillii is accurate, his figure of P. gillii shows 9 streams, the basal half of the caudal is black, or with black blotches; on some of transverse scales on the caudal peduncle, which is a diagnostic character of P. mexi- these the basal half of the dorsal is also cana. I suspect the figures were switched. black." Specimens with a black basal half His undescribed species might very well be of the caudal fin, formerly described as P. P. mexicana. The apparently aberrant num- caudata Meek, 1909, are herein recognized ber of lateral scales (also mentioned by as a subspecies, viz., P. gillii caudata. Bussing [1987]) is explained by character Bussing (1987) identified most of the displacement (cf. Poeser 1995). The only Costa Rican populations as P. gillii. He synonym of P. sphenops in Rosen & Bailey gave the following description (translated (1963) from Nicaragua is P. dovii Gtinther, from Spanish): "Body with yellow spots, in 1866. Poecilia dovii was considered syn- some females these spots are black. Scales onymous with P. sphenops by Regan in a lateral series 26-28, mostly 27. Three (1908). Since the type material of P. dovii orbital pores in one line. In the males some- contained specimens from P. mexicana times a large dorsal fin, with black spots or VOLUME NUMBER 116, 2 363 blotches at the base. Caudal fin with spots, counts suggests that all populations belong blotches or solid black pigmentation. Some to the same species. males with orange in their caudal fin. Other Because the missing spine is oftaxonom- fins yellow, head and body bluish. Note: ic significance, this form was named P. gil- some males have a red dorsal. Large spe- lii cuneata Garman, 1895. The populations cies, up to 105 mm." Males possessing a of P. gillii from Panama examined in the red dorsal fin are P. gillii salvatoris. Bus- present study are partly sympatric with P. sing also recorded P. mexicana, in a much marcellinoi like in El Salvador (Poeser lower frequency, i.e., three populations 1995). Poeser (1992) reported a male with from over fifty in total, widely apart from an aberrant gonopodium from Nicaragua each other (Bussing 1987:144, map 20). His (GCRL 6697), identified then as P. mexi- diagnosis for this species is: "Very much cana mexicana. Re-examination of my like P. gillii. Yellow spots on body, black notes proved that the identification and lo- in some females. 28-30 (modally 29) Lat- cality was wrong. The sample containing eral scales, and three orbital pores, forming this male was GCRL 8748, viz., P. gillii a triangle. Dorsal and caudal spotted, rest cuneata from Panama. & of fins yellow. It is also a large species, up Rosen Bailey (1963) considered Lem- to 110 mm." Bussing (1987) illustrated P. besseia parvianalis a synonym of P. sphen- gillii and P. mexicana in his paper, and his ops. However, since L. parvianalis has uni- diagnoses of P. gillii and P. mexicana cor- cuspid inner teeth. Miller (1983) placed it respond with the account of Villa (1982) in the synonymy of either P. mexicana or (see above). Surprisingly, Bussing did not P. gillii. Based on the original description, record any populations of P. marcellinoi. in which 8 transverse scales on the caudal Hubbs (1926, 1953) considered P. sphen- peduncle are illustrated, I assign it to the ops in Panama either P. sphenops cuneata synonymy of P. gillii. Fowler's (1949) re- (cf. Hubbs 1926), or P. s. gillii (cf. Hubbs cord is important for its taxonomic value. 1953). Examination by me of Panamanian The P. sphenops group, if raised to generic populations confirms earlier findings of level, would become Lembesseia (with L. profound morphometric differences, as well surinamensis (=P. sphenops) as type spe- as constant meristic data, in all Panamanian cies). — A populations. conspicuous reduction ofthe Distribution. Poecilia gillii is recorded gonopodial spine on ray 5 is noted. In pop- from the Pacific coast of Guatemala to the ulations near the Costa Rican border (Bocas Atlantic coast of Colombia. In northern re- del Toro district of Panama) the spine is gions, from Guatemala to Costa Rica, it is distinct. In populations near Colombia, i.e., represented by the subspecies P. gillii sal- the San Bias area, it is reduced or absent, vatoris. The subspecies P. gillii caudata oc- as is found in P. koperi (Fig. 4a). This shift curs in Costa Rica, whereas P. gillii gillii in character expression from west to east is is present in Panama. The South American not accompanied by meristic changes. populations, as well as adjacent populations Although some populations contain poor- in Panama, are considered to be P. gillii ly pigmented, slender specimens, other pop- cuneata. The type locality of P. cuneata ulations manifest heavily pigmented, stout was designated "Turbo, Gulf of Darien" specimens. Nevertheless, all specimens (Garman 1895). Rosen and Bailey (1963) have nine anal fin rays, nine (or rarely ten added 'Panama' to this locality. However, in the San Bias district) dorsal fin rays, 16 study of several maps did not confirm this scales around the caudal peduncle, 18 addition. The nearest village with that name scales around the body, and 26 to 29 scales is near the Gulf of Uraba, an extension of, in a lateral series. The consistency of these and sometimes also so-called, the Gulf of 364 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON Darien, in Colombia. I presume that this is Poecilia vandepolli van Lidth de Jeude, the correct type locaUty. 1887 Poecilia Vandepolli van Lidth de Jeude, Poecilia caucana (Steindachner, 1880) 1887:137, PI. 2, Figs. 4 and 5 (type lo- cality: Curagao). Girardinus caucanus Steindachner, 1880: Poecilia Vandepolli arubensis van Lidth de 87, plate 4, figs. 4, 5 (type locahty: Co- Jeude, 1887:138, PI. 2, Figs. 6-10 (type lombia, Caceres); Allopoecilia caucana locality: Aruba). Hubbs, 1924:11. — RMNH Lectotype. Curasao: 5155, UMMZ Ma^^r/^/.—Colombia: 186317 adult male, Curasao, coll. Neervoort and v. (51), Depto. Cordoba, Tierra Alta, Rio d. Poll, no further data. & Sinu, coll. W. Moberley K. Adler, 1965. Paralectotypes.—KMnn 33843 UMMZ (4), Venezuela: 186931 (54), Rio same data as RMNH 5155. UMMZ — Monay, coll. F. F. Bond, 1938; Material. Aruba: RMNH 5156 (6 syn- 186934 (24), Quebrada Goajira, F F Bond, types of P. Vandepolli arubensis), Aruba, UMMZ 1938; 186937 (44), Rio Bucares, coll. Neervoort and v. d. Poll, no further coll. F F Bond, 1938; USNM 86264 (4), data. See also tables below, details in Poe- Valera, Trujillo, coll. H. Pitteir, 1923; ser (1992). USNM 121677 (33), Rio Motatan, 4 km Diagnosis.—A = 8-9; D = 7-8; C = 16; above Motatan, coll. L. R Schultz, 25-III- LLS = 26—27; CPS = 16. Several speci- 1942. mens have humeral blotches anterior to the USNM Panama: 293444, (7), Rio Meteti, position of the lateral side spot as occurs in NNW Darien Province, 40 km of Yaviza, P. vivipara. The gonopodium lacks extru- USNM coll. W. C. Sterns et al., 24-11-1985; sions on ray 3—and 5p. 293574, (10), Rio Peresinico, Darien Prov- Description. Poecilia vandepolli is re- ince, coll. B. Chernoff, J. Lundberg, L. ported to have specimens with orange on McDade, 23-11-1985. the ventral side of the body, yellow speci- Diagnosis.—A = 8; D = 7-8; C = 18- mens with blue sides and grayish brown 22; LLS = 26-27; CS = 16; CPS = 14. specimens in the same population (Felt- Poecilia caucana is a small to medium kamp and Kristensen 1969). Speckled spec- sized species, characterized by a black band imens, allegedly only occurring in Vene- in its dorsal fin. The gonopodium has ray zuelan populations (Feltkamp and Kristen- 4p with broade—r rays than ray 4a. sen 1969), also appear on Curasao. The pig- Description. This is one of the smaller mentation at the base of the dorsal fin forms species of the genus. The largest specimens a blotch, in addition to dark spots. The body mm examined are a female of 37.4 SL and may have a humeral blotch, positioned mm a male of 27.2 SL. The dorsal fin has more anteriorly than a similar such patch of a black transverse band at the base. This fin color in P. vivipara. The females tend to be is usually milky white or yellow, although paler than males. also red fins are reported in aquarium lit- The gonopodium, figured in Poeser erature. The other fins are unmarked. (1992), is like that of P. vivipara, with the The gonopodial characters figured in Ro- exception of the serrae on ray 4a. When sen and Bailey (1963) are of considerable fully developed, a little membranous bulge interest because gonopodial ray 4p is thick- is found on gonopodial ray 3, which covers er than ray 4a. In addition, the hook on gon- extruding serrae. — opodial ray 5 is—weakly developed. Description of the types. The lectotype mm Distribution. Atlantic slopes of Pana- is a mature male, 27.3 SL. It has 8 ma, Colombia and Venezuela. dorsal fin rays and 16 scales around the VOLUME NUMBER 116, 2 365 caudal peduncle. Several scales in the lat- Poecilia sphenops (non Valenciennes, eral series are missing on this specimen and 1846); Regan, 1913:1013 (in part). the caudal fin is damaged, so no further me- Poecilia sphenops vandepolli (non Van ristic data can be given. Gonopodial ray 3 Lidth de Jeude, 1887); Schultz, 1949:84, has a long terminal segment, the hood is 97-99 (in part). short. An extremely short extrusion similar Poecilia sphenops cuneata (non Garman, to a small hook is present. Gonopodial ray 1895); Hubbs, 1926:77. 4a is unmodified, ray 4p has eight unmod- — UMMZ ified distal segments followed by serrated Holotype. Adult male, 223343, segments. Gonopodial ray 5a is one seg- Venezuela, Rio Curipe at Higuerote, coll. F. ment longer than 5p. Subdistal ventral mod- F Bond, 2-V-1938. — UMMZ ifications occur on ray 5a. The largest fe- Allotype. Adult female, mm UMMZ male paralectotype, 47.5 SL, has about 223344, same data as 223343. 25 scales in a lateral series (this number is Material—Bymn 1909.2.25: 53-56 (5), not accurate because some scales are miss- Venezuela, coll. Arnold, no date; UMMZ ing). It has nine dorsal fin rays and nine 200738 (15 of 123), 2 km N of Ocumare, anal fin rays, the caudal fin is damaged. lagoon on flats near mouth of Rio Cumboto, The specimens in the type series from coll. F F Bond, 5-M938; UMMZ 200740 Aruba are much smaller. The largest male, (15 of 57), Rio Guaiguaza, 3 km W of Por- mm 22.0 SL, has a gonopodium like the to Cabello, 2 km from mouth of river, coll. lectotype, with seven distal unmodified seg- F F Bond, 15-1-1938; UMMZ 200744 (12), ments on gonopodial ray 4p and no trace of Rio Sanchon, 5 km W of Tavorda, 10 km a hook on ray 3. It has 8 dorsal fin rays, 16 W of Porto Cabello, coll. F F Bond, 26-1- scales around the caudal peduncle, and UMMZ 1938; 200753 (15 of 430), Lagun- about 24 scales in a lateral series (some km W ita, 5 of Coro, Estado Falcon, coll. scales are missing). The caudal fin is dam- F F Bond, 19-in-1938: UMMZ 200755 (15 aged. This male has a pigmented band of 1 15), Falcon, Laguna del Rio Capatarida, proximally on the dorsal fin and a pig- at mouth, 5 km N of Capatarida, coll. F F mented humeral blotch. The largest female UMMZ mm Bond, 2-III-1938; 200760, (15 of (26.5 SL) has eight anal fin rays and km N 152), Falcon, Coastal lagoons, 15 of nine dorsal fin rays. The caudal fin is dam- Maracaibo, coll. F F Bond, 6-IV-1938; aged. Sixteen scales are found around the UMMZ 200761 (13 of 202), same data as caudal peduncle and 27 in a lateral series. UMMZ UMMZ 223343; 200762 (15 of To establish morphometric and meristic 80), Estado de Miranda, Lagunita de Taca- variations a detailed examination of 15 pop- ulations was carried out. Some 12 lots were rigua, at Tacarigua, 85 km E of Caracas on examined from Aruba, and three lots from the coast,UnMeaMrZthe boca, coll. F F Bond, 3- Cura9ao (Table —1). Cn-u1m9b3o9t;o, 2 km2N00W764of(28O)c,uBmaorcea;deZl MRiAo Distribution. Poecilia vandepolli oc- curs naturally in all kinds of waters of the 109.206 (6), La Goajira, Rio Calancala, San Netherlands West Indies (Aruba, Curasao, Antonio, coll. P. Wagenaar-Hummelinck, ZMA and Bonaire). It is introduced on St. Maar- 17-M937; 119.909 (30), Paraguana ten/St. Martin (Poeser 1992). Estangue de Moruy, coll. P. Wagenaar- ZMA Hummelinck, 18-n-1937; 119.910 Poecilia koperi, new species (15), Paraguana, Estangue de Santa Ana, Fig. 1, Table 2 ZcolMl.AP. Wagenaar-HummkelminNck, 16-11-1937; 120.885 (4), 2 of Barcelona, Poecilia vivipara (non Bloch & Schneider, Rio Guanta, coll. P. Wagenaar-Humme- 1801); De Beaufort, 1940:111. linck, 1936.