Gorgas Memorial Laboratory (GML) and the Panamanian Snake Census

Dr. Herbert C. Clark (1877–1960), first Director of the Gorgas Memorial Laboratory, initiated a snake census in Panama and conducted it at varying levels of intensity from January 1929 through 1953 (he retired in 1954). Not a new concept, the census was started as an extension of a then ongoing Central American snake census that used the labor forces on plantations of the United Fruit Company. Clark had earlier supervised the snake census in Tela, Honduras, but, with the resources of GML, Clark expanded the census in Panama well beyond the plantations and continued it for a quarter of a century.

Clark's primary aim was to learn the venomous species in Panama and to assess their relative medical importance (Clark, 1942). Bounties were paid for killed snakes (mostly heads), which were accumulated in formalin in central locations around the country. The GML snake census was initiated in cooperation with the old Antivenin Institute of America, one of whose founders was Thomas Barbour. In 1932, Barbour involved the Museum of Comparative Zoology (MCZ) of Harvard University, which “became actively interested in the studies and supported them in part” (Wright, 1970: 269). Early shipments were sent to the MCZ (e.g., Clark, 1937: 12), from where Barbour turned over the identification and study of the collection to his former student E.R. Dunn, who identified most of the material. Dunn eventually took over the study and reported on the results (Dunn, 1933, 1942, 1949a, 1949b; Dunn and Bailey, 1939).3 Dunn obtained background information on the census methods during his visits to GML starting in 1939, and his 1949a paper is the most complete and easily accessible summary of Clark's GML snake census.

Dunn's 1949a account, however, did not include the census collections from most of western Panama nor from the limited uplands of eastern Panama (see below). But new species, new country records, and other noteworthy specimens from those areas were reported elsewhere (Dunn, 1942; Dunn and Bailey, 1939). Unfortunately, over 90% of the GML specimens were apparently discarded after Dunn's study, except for whole specimens and selected heads especially of rare species that were placed in major museums. Fewer than 1000 specimens with data were saved from the Panama snake census, which spanned a quarter of a century.4

A total of 13,745 snakes, mostly heads only, were accumulated in the overall period 1929–1953 (Wright, l970: 269). Dunn (1949a: 39, 55) reported on nearly 80% of the collection, that is on “10,960”5 specimens from the lowlands of central Panama (“Chagres collection”) and from the Pacific lowlands of west-central and eastern Panama (combined collections from “Coclé-Herrera”, “Sabanas”, “Darién”). Excluding 1073 museum specimens, savanna collections, and those collections from west of the Panama Canal, Dunn's (1949a; table 7) list of 70 species includes 2500 snakes from forested lowlands in the Chagres drainage of central Panama and 3044 snakes from areas of mixed banana plantations and forest in the Darién of eastern Panama.

Neither Atractus nor Geophis is included in the nearly 11,000 lowland specimens obtained by the GML snake census and reported in Dunn's (1949a) summary paper. The absence of Atractus and Geophis from the large lowland collections seems surprising at first blush, inasmuch as there are many Atractus occurring in the South American lowlands and a specimen of Geophis hoffmanni was collected not far from the Panama Canal in 1968. Recent sampling in the central Panamanian lowlands has yielded neither Atractus nor additional specimens of Geophis (Ibáñez et al., “1995” [1997]).

Clark also pursued collections of snakes from upland areas, which are of limited extent in central and eastern Panama. Dunn and Bailey (1939: 3) reported on “268 snakes from the uplands of eastern Panama gathered in 1936 to 1938 through the initiative of Dr. H. C. Clark”. There are no Geophis in this collection, which, however, provided the first two specimens of Atractus from Central America.

There are eight species to be reported in the following pages (maps 1, 2). These eight species are represented by a total of nine central and eastern Panamanian specimens—five specimens of Atractus and four of Geophis. One Atractus was named in 1939; four specimens of Atractus and one of Geophis represent five “new” species that are named herein; the other Geophis specimens represent two species named in the 19th century.

The first two of the aforesaid nine specimens were obtained by the GML census in 1936 and 1938. I collected another two while resident Visiting Scientist at GML during 1964–1967. And five experienced field men (several earlier trained as collectors at GML) each caught one of the other five specimens in the 23-year period 1952–1974.

As a further indication of rarity, it may be noted that five of the eight species (A. depressiocellus, A. imperfectus, G. bellus, G. brachycephalus, G. hoffmanni) are from the region of the Piedras-Pacora Ridge, only about 30 km northeast of Panama City. The Piedras-Pacora Ridge—a section of the continental divide—delimits part of the southeastern side of the Madden Lake6 watershed, and is now included in the Parque Nacional Chagres. A team of Panamanian herpetologists (Ibáñez et al., “1994” [1995]) surveyed the herpetofauna of the Piedras-Pacora Ridge in 1990–1995; they invested 1253 man-hours of collecting in both wet and dry seasons (including 166 man-hours at night in the wet season). Their list is extensive, totaling 131 species of amphibians and reptiles, including 37 species of snakes. Among the small snakes collected was the third known specimen of Coniophanes joanae, but no new specimens of the still elusive Atractus or Geophis.

The Piedras-Pacora Ridge may be a premontane forest refuge for some small part of the fauna that is barely surviving the climatic-vegetational vicissitudes of the Pleistocene. The mingling of several very rare species in one place, including three species known only from that place, leads me to fear that it may be their “last stand” on the way to extinction. Should this be so, their Latin names may be their only epitaphs.

Methods of Study

Total length and tail length were determined by moderately stretching preserved specimens along a metric rule. Head and snout lengths, eye length, and head plates and their sutures were measured to the nearest 0.1 mm with an ocular micrometer in a dissecting microscope. Some measurements are used for comparisons, but most were converted to proportions for purposes of descriptions. It bears emphasizing that snout length (tip of snout to eye) was not measured on the oblique angle but on the sagittal plane by ocular micrometer, whereas head plates and the distance from eye to lip were determined along the plane of the greatest dimension, meaning that the head had to be differently inclined for different measurements. Therefore anyone comparing the descriptions with the drawings will have to be alert to the usual problems of parallax: In no case, for example, will a proportion such as nasal plate/loreal plate be retrievable from the figures, whereas one such as eye length/eye-to-lip may or may not match a figure, depending on perspective of the drawing (e.g., dorsolateral or lateral) and degree to which the lip is flared.

The drawings are my own, made with the aid of a camera lucida fitted to a Wild dissecting microscope. I photographed specimens positioned under glass on a raised plastic platform to eliminate shadows, and under alcohol to reduce glare from reflective surfaces of scales.

Ventral plates were counted by the Dowling (1951) method, starting with the first plate bordered on each side by the first dorsal scale rows; gular plates anterior to this, if wider than long, were termed preventrals (Myers, 1974: 37) for purposes of comparing type specimens with old original descriptions in which these plates were included in ventral counts.

Dowling's method has the advantage of giving a standard starting place for ventral counts that was said to correspond (in some snakes) in position with the atlas-axis articulation of the vertebral column. But the first ventral is not always quickly ascertained in the case of small, poorly preserved snakes with shiny scales, and I am sorry to have adopted this method some years ago. I seldom notice more than a few “preventrals”, but comparisons with older literature are less accurate if these are not counted and added to the total. Therefore, I would not be critical of anyone refusing to adopt the Dowling method over the older one summarized by Schmidt and Davis (1941: 26):

Except when a dentigerous bone was dissected out for illustrating, maxillary teeth were counted in situ (Myers, 1974: 27); the once common practice of excising maxillae from rare snakes for the sole reason of counting teeth is neither necessary nor justified.

Hemipenes were treated as described in Myers (1974), except that everted organs were inflated with carmine-dyed petroleum jelly (rather than wax or latex). A few hemipenes in preserved specimens were manually everted from either the retracted state (fig. 17) or from partially everted organs (figs. 14, 20B), after they were dissected out and then soaked first in glycerin and finally in a saturated solution of trisodium phosphate to restore elasticity. Although hemipenial eversions made from preserved specimens may be fully everted with all structures showing, some such organs may be less completely inflated (i.e., less robust) than fresh preparations. Techniques are reviewed by Myers and Cadle (ms).

All specimens were collected long before the availability of Global Positioning System (GPS) satellite receivers. The detailed 1:50,000 map series of Panama is still incomplete and, with one exception (see Atractus hostilitractus), topographic maps of this scale are still unavailable for the type localities in this paper. Therefore, approximate geographic coordinates were determined mainly from the maps Canal Zone and Vicinity 1:100,000 (1957 edition, U.S. Army Map Service, Washington, DC) and the 12-sheet Mapa General de la República de Panamá 1:250,000 (1st ed., circa 1966–1967, Dirección de Cartografía, Panama City). Some coordinates differ slightly from those in Fairchild and Handley (1966), whose important gazetteer of GML and other collecting stations was based on an earlier 3-sheet 1:500,000 map of the Republic.

Atractus clarki Dunn and Bailey Figures 1A, 2A, 3A, 4, 5; map 1

Redescription of Holotype

The specimen is a female, shown to be adult by condition of the oviducts, which are large and convoluted. The neck was nearly severed from the body by the collector, but otherwise it is in good condition.

A Second Specimen of Atractus clarki

A Colombian specimen (MCZ 13301) from the Chocó, without further locality data, has for many decades been misidentified as Atractus multicinctus, which is a better known Chocoan species. This “new” specimen (fig. 4) is a male, judged adult because the hemipenial spines are calcified. It was collected by M.V. Campbell in 1919.

Remarks

Dunn and Bailey (1939) named this species after Herbert C. Clark, first director of the Gorgas Memorial Laboratory and instigator of the Panamanian snake census. Santa Cruz de Cana, the type locality of Atractus clarki, is situated at 500 m elevation (my altimeter reading) on the eastern flank of the Serranía de Pirre—roughly at 7°46′N, 77°41′W (map 1). It is the site of gold mining that has been carried on periodically since the end of the 16th century, being usually deserted and reverting to jungle in the intervening periods (Myers, 1969: 22).

The second specimen and first male of Atractus clarki, from somewhere in the Colombian Chocó, corroborates the diagnostic features of the species but might be perceived to call into question its generic status. Savage (1960: 29) stated that the hemipenes of Atractus are “bilobed at tips” and “never calyculate.” However, the organ of A. clarki is bilobed for a third its length and the lobes are markedly calyculate. Hoogmoed (1980) explicitly mentioned the existence of calyces in several species of Atractus in Surinam, although in one instance (p. 23) he described them as “scalloped calyces (Savage, 1960, fig. 4A)”—this reference being to structures described by Savage (1960: 24) not as calyces but as “scalloped transverse flounces” (which conceivably might be derived from calyces by loss of the vertically aligned tissue ridges). Fernandes (“1995” [1996]: 43, 50, fig. 1) added the trait of hemipenial capitation with calyculation for the Brazilian A. reticulatus and some other southern Atractus, and Fernandes et al. (2000: 3, 5) described the Brazilian A. maculatus and A. zebrinus as having hemipenes “capitate, bilobate … the distal portion covered by spinulate calyces”.

The taxonomic distribution and nature of both capitation and calyces/flounces within Atractus need to be revisited (male genitalia have been comparatively examined for less than half of the currently recognized species [mainly by Savage, 1960, and Hoogmoed, 1980] and illustrations are regrettably rare). Calyces are probably symplesiomorphic for both xenodontine and dipsadine colubrids, with absence being a presumed secondary loss, as in the tribe Xenodontini (Myers, 1986: 6) and in many or most Atractus. When Savage's seminal study is extended to cover the entire genus, it may be found that Atractus clarki is simply primitive in the retention of relatively long, hemipenial lobes bearing typical calyces. See page 9 for a comment on the bifurcation of the sulcus spermaticus.

Atractus clarki also seems primitive in its relatively large eye, which is sufficiently protuberant so as to be visible (holotype) or almost visible (MCZ 13301) in ventral view, more like a generalized terrestrial colubrid than the majority of Atractus, most of which seem to be semifossorial on morphological grounds.

Atractus darienensis, new species Figures 1B, 2B, 3B, 6, 9A; Map 1

Description of Holotype

The holotype is a sexually mature female as shown by enlarged, flabby (nonconvoluted) oviducts that appear recently spent.

Remarks

I found the holotype of Atractus darienensis at night in dense forest (fig. 8), as it was crawling across a small bare spot on the forest floor. The species is quite distinct from its geographically nearest neighbor, A. clarki, whose type locality is only about 25 km to the south. Both type localities lie at 500 m elevation on the lower slopes of the Serranía de Pirre (map 1), in an evergreen seasonal forest (monsoon rain forest) having a January–March dry season during which many trees are deciduous (Myers, 1969, 1982).

Atractus depressiocellus, new species Figures 1C, 2C, 3E; map 1

Description of Holotype

As suggested by its large size and proved by large ovaries and large convoluted oviducts, the specimen is a sexually mature female. It is rather soft, with internal organs poorly preserved; color pattern and other external features are clearly evident except that the rostral region was rubbed raw or injured in life. The rostral plate and anterior parts of the nasal plates consequently are lacking, and presence of internasals cannot be determined, although the internasals would be exceptionally small if present (fig. 3E).

Remarks

The collector was a former field assistant of the Gorgas Memorial Laboratory, which is relevant to placing the type locality of Atractus depressiocellus. “Cerro Azul” of GML field parties was essentially equivalent to Cerro Jefe, the 980-m high point on the Piedras-Pacora Ridge about 37 km NE of downtown Panama City (Myers, l969: 27). I have added the word “region” to the type locality because of a belief that specimens labeled “Cerro Azul” might have been taken anywhere from about 5 km southwestward of Cerro Jefe to about 5 km to the north. The likely elevational range for most reptiles collected is about 200–800 m in my estimation.

Atractus depressiocellus is morphologically quite divergent from the other Panamanian species, particularly in the depressed and very small eye and in the internasals, which are either lacking or much reduced in size. The dentition is also different from the other four Panamanian species, which have the maxilla strongly arched and bearing well-spaced, firmly set teeth that decrease in size posteriorly. In contrast, the maxilla of A. depressiocellus is not strongly arched and the anterior teeth are not as widely spaced and are subequal in size; also, the maxillary teeth of depressiocellus are not as firmly ankylosed as in the other species and are easily dislodged from their sockets (due partly to the soft state of preservation of the holotype?).

With a total length of 750 mm (673 mm SVL), Atractus depressiocellus ranks among the largest of Atractus, and it seems extraordinary that it has escaped detection for so long.

Atractus hostilitractus, new species Figures 1D, 2D, 3C, 9B; map 1

Description of Holotype

It is a sexually mature female, with ova >4 mm and enlarged convoluted oviducts. Part of the posterior body was damaged during capture, but the specimen is well preserved.

Atractus imperfectus, new species Figures 3D, 9C, 10; map 1

Partial Description

Dunn and Bailey (1939) gave some scale counts for this specimen but curiously provided nothing about the color pattern save mention of a “black postocular streak across the last upper labial” (but see below and figs. 9C, 10). Although the posterior color pattern, number of ventrals etc. are unknown, the following description should allow it to be matched whenever a second Panamanian specimen is obtained.

Remarks

Obtained in 1936, this partial specimen was the first Atractus collected in Panama. Dunn and Bailey (1939) compared it with topotypes of Atractus crassicaudatus from Bogotá, Colombia. They (1939: 8) stated that, except for a few minor differences (i.e., number of labials in contact with genials, rarity in crassicaudatus of an elongated upper secondary temporal, and position of a postocular marking), the “characters of scalation and markings are within the range of variation of Colombian crassicaudatus.” Comparison of a large series of Bogotá crassicaudatus in the American Museum fails to substantiate a close resemblance: For example, the Colombian species differs in having a larger eye relative to a shorter loreal and in having noticeably shorter genials (the main reason why the Panamanian specimen has 4 infralabials in contact with the genials, whereas Dunn and Bailey (loc. cit.) were “unable to find a Colombian specimen with more than three in contact”).

I had originally hoped that this specimen could be assigned to one of two other new Atractus described herein, namely darienenis, which has a similarly shaped head, vague nuchal collar, and white labials, or hostilitractus, which is superficially similar in having a ringed neck and extensive black pigmentation under the throat. But I decided otherwise after direct comparison of snout shapes (see fig. 9), details of anterior color patterns, eye size relative to loreal and distance to lip, and consideration of details of head plate sizes and proportions. I concluded that the species represented by only a head and neck is distinct—as also suspected by Savage (1960: 32), who mentioned “an undescribed population [in Panama] allied to Atractus crassicaudatus”.

The closest resemblance seems to be with Atractus darienensis, which differs from A. imperfectus in a few details of coloration and, more convincingly, in proportions. The first two light brown interspaces (between black blotches) on the neck of darienensis (fig. 1B) seem somewhat comparable to the pale broken ringlike markings (on black) in imperfectus (fig. 10). The anterior interspaces in darienensis are brown to the top of the first scale row, which is white like the anterior ventrals; the paler broken “rings” in imperfectus have some slight suffusions of pale brown, but are mainly yellowish white like the adjacent edges of the ventrals. The supralabials in darienensis have little encroachment of dark pigment from above, whereas those of imperfectus are boldly serrated (compare fig. 9A and 9C).

The holotypes of A. imperfectus and A. darienensis seem roughly comparable in size, with the latter having a slightly shorter head that is, however, slightly wider across the temporal region than the head of imperfectus (table 1). But direct comparison of the specimens shows imperfectus to have a relatively broader snout in dorsal view, with this being reflected in proportions of the paired prefrontal plates, which are anteriorly tapered to a greater extent in darienensis (fig. 3B) than in imperfectus (fig. 3D). The width across the posterior margin of the paired prefrontals is the same (4.3 mm) in both specimens, but the anterior margin narrows to 2.0 mm in darienensis vs. 2.9 mm in imperfectus. Closer to the snout tip, the width across the internasals is 1.9 mm in darienensis vs. 2.1 mm in imperfectus. The rostral plate extends more dorsad in imperfectus than in darienensis. The somewhat smaller head of darienensis is slightly longer in snout length (table 1), which is reflected in length of the loreal plate (2.6 mm in darienensis, 2.3 mm in imperfectus) and also in relative eye size. The eye is relatively larger in imperfectus (36% of snout length, 56.5% of loreal length) than in darienensis (30% of snout, 46% of loreal). The eye of imperfectus also is relatively larger compared with its distance to the edge of the lip (eye length/eye-to-lip distance = 1.08 in imperfectus, 0.857 in darienensis).

All the differences elaborated above convince me that Atractus imperfectus is a species different from A. darienensis. Nonetheless, one can scarcely predict the extent of variation in diagnostic characters derived from two specimens from widely separated populations, particularly when one specimen is only a fragment! If a whole specimen of ”imperfectus” ever turns up and disproves this hypothesis, I would urge that the species name darienensis be given priority simply because that name is based on a complete holotype from a more precise type locality.

In a first draft of this description written a decade ago, I treated Atractus imperfectus as a “species inquirenda”, even stating in manuscript that it “would seem poor judgement to add a name to the literature without being able to describe the posterior color pattern and other missing details.” But, considering the real rarity of this apparently diagnosable species, I now join my paleontological brethren in recognizing that specimens ofttimes are not perfect. Atractus imperfectus therefore joins several other snakes named from incomplete specimens gathered by Clark's Panama snake census: Sibynomorphus nicholsi Dunn, 1933 (= Dipsas nicholsi); Hydromorphus clarki Dunn, 1942; and Micrurus nigrocinctus yatesi Dunn, 1942 (= M. alleni yatesi).

Records Excluded from Central Panama

Geophis brachycephalus (Cope): This is the most common Panamanian Geophis in collections, but Downs' (1967: 146) statement that it ranges from “Costa Rica southward through Panama to Colombia” needs correcting. South America was included in the range because of Downs' synonymizing of the Colombian Geophis nigroalbus, which is shown herein to be distinct based on proportional and hemipenial characters (see Comparisons under Geophis bellus, new species).13

The only Panamanian specimens of G. brachycephalus with definite locality seen by Downs (1967: 153) are from far western Panama. Central Panama was included on the basis of a series of specimens (ANSP 24723–24734) from “Panama Sabanas”. These specimens, consecutively numbered with a partial specimen (ANSP 24723) of G. godmani (see below) having the same locality data, are from the GML snake census (see Introduction).

As summarized by Dunn (1949a: 47), the GML “sabanas collection” was accumulated from various localities in the central Pacific lowlands, “from Capira to the west to Canita [Cañita] in the mid-basin of the Bayano to the east”—a straight-line distance of 120 km—although “the majority of the specimens come from the true ‘Panama sabanas' which lie between Panama City and Chepo.” Dunn (1949a: 47, table 7) reported on over 3900 snakes from this collection, with no mention of Geophis in this or in any other part of the GML lowland collection, even though Dunn presented the “sabanas” Geophis to the Academy of Natural Sciences in 1942, seven years before publication of his summary analysis.

The “sabanas” locality seems wrong on the face of it. Neither G. brachycephalus nor the rare G. godmani are lowland savanna snakes, and the thought that both might have been found in sympatry in such an unlikely habitat is hard to swallow and harder to digest. Furthermore, several specimens of the little montane snake Trimetopon slevini (ANSP 24717–24719) also were presented by Dunn in 1942 and were cataloged as “Panama Sabanas”; neither were these mentioned in Dunn's 1949a paper. I surmise that Dunn either knew that the locality had to be wrong for these specimens and ignored them, or there was a cataloging error that he either was unaware of or never got around to correcting.

Although the “sabanas” record for Geophis brachycephalus must be discarded as erroneous, this species does occur in upland wet forest in east-central Panama, as demonstrated later in this paper.

Geophis godmani Boulenger: This is a rare snake in Panama, and it evidently is a montane species throughout its range. Downs (1967: 72), followed by Peters and Orejas-Miranda (1970: 119), gave its distribution as “Caribbean and Pacific slopes of central Costa Rica southward to the Canal Zone of Panama [at known localities] between 1300–2100 meters above sea level.”

The three Panamanian specimens of G. godmani seen by Downs (1967: 75) are all from the GML snake census—two from Finca Lérida in extreme western Panama, the third (ANSP 24722) purportedly from the “Panama Sabanas” in central Panama. This last specimen is cataloged consecutively with the series of G. brachycephalus and in proximity with the series of Trimetopon slevini discussed above, and the godmani record must be discarded for the same reasons. The most likely correct locality for the “sabanas” Geophis brachycephalus, G. godmani, and Trimetopon slevini is Finca Lérida in Chiriquí Province, which provided some census material and where all three species occur.14

Geophis bellus, new species Figures 11–14, 15C; map 2

Description of Holotype

Despite its small size and slender habitus, the male holotype is unambiguously sexually mature because (1) the hemipenial spines had hardened, (2) the relatively large testes (about as long as the snake's head) are strongly tubular, (3) the slender vasa deferentia are convoluted, and (4) sperm were present in a small fragment of testis examined under a compound microscope. The anterior (right) testis is 6.8 mm long and the posterior is 6.5 mm long, both being about 2.1 mm wide as measured in situ.

Comparisons

Geophis bellus is a member of Downs' (1967: 137) sieboldi species group, with which it generally agrees in features of scutellation, small eye, long projecting snout, and, especially, in the simple (unforked, unexpanded) distal end of the ectopterygoid process and in maxillary features. Pertinent maxillary characters include a small number of subequal teeth, dorsal and ventral flattening of the maxilla, a toothless anterior projection, and a ventrally curved, bluntly pointed posterior end (compare fig. 13 with that of G. brachycephalus [Downs, 1967: 26]). The presence of an anterior temporal plate—on the right side of the holotype of G. bellus—is a rare condition in the sieboldi group. It seems quite clearly to have arisen by division of the fifth supralabial (compare lateral views of head in fig. 12), thus providing a clue to the method of original loss of the primary temporal in the sieboldi group (see Downs, 1967: 14).

The single specimen of Geophis bellus was shown to F. L. Downs shortly after he had completed his monograph of the genus. It is the specimen mentioned in a footnote (Downs, 1967: 146) as leading him “to doubt that my inclusion of G. nigroalbus Boulenger in the synonymy of G. brachycephalus is justified.” Downs (1967: 152) had not seen the Colombian type of Boulenger's nigroalbus, but direct comparison of the holotypes of bellus and nigroalbus convinces me that they are different species (see below). I also closely compared the type of bellus with the only central Panamanian specimen of brachycephalus, which probably was collected within a few kilometers of the former. Measurements of these specimens are given in table 2, with further comparisons following.

Geophis bellus is a much smaller snake than the Colombian G. nigroalbus, in which males attain a total length of at least 370 mm, and the slender bellus is relatively less robust. The juvenile holotype of G. nigroalbus, although 67 mm shorter than the adult holotype of G. bellus, nonetheless has a slightly larger head (figs. 15A, 16, and table 2). The holotype of nigroalbus resembles bellus in having a dark body and a similarly positioned and vivid pale band across the head and nape (absent or vestigial in other specimens of nigroalbus seen), but differs in the uniformly pale venter (adult nigroalbus may acquire dark transverse banding ventrally).16

Although the adult male holotype of G. bellus is essentially identical in total length to the nearly sympatric juvenile male of G. brachycephalus, the latter specimen (figs. 15B, 18) is a more robust snake with a longer and wider head and noticeably larger eye; most of the various head plate and suture measurements are smaller in bellus, except for slightly longer internasals and parietals. The central Panamanian specimen of brachycephalus also differs from the probably sympatric bellus in having a pale-spotted dorsum and a pale venter (fig. 18).

Comparisons of the holotype of G. bellus with specimens of G. brachycephalus from western Panama lead to similar conclusions: Geophis bellus is a much smaller, more slender species. The striking nape band of bellus seems to be a retention of a variably present-or-absent juvenile feature in brachycephalus and nigroalbus. Downs (1967: 150) noted that a pale collar, if present, is usually lost in brachycephalus by 150 mm SVL (adult holotype of bellus = 169 mm SVL), with traces of the collar only occasionally persisting to a larger size. According to Downs, brachycephalus attains a maximum total length of 418 mm in males and 460 mm in females.

I also compared the holotype of Geophis bellus directly with the Costa Rican holotype of G. talamancae Lips and Savage (1994), which is similar in size, relative tail length, and in numbers of ventrals and subcaudals.17 The juvenile female holotype (the only reported specimen) of G. talamancae, is 11 mm longer than the adult male specimen of bellus, but it obviously is a more robust snake, with a much stockier body and a relatively larger head (greatest body width in talamancae = 7.2 mm vs. 4.8 mm in bellus; greatest head width × head length to end of parietals in talamancae = 6.0 × 8.5 mm vs. 4.5 × 7.0 mm in bellus). Geophis talamancae appears to be close to G. brachycephalus and perhaps G. nigroalbus. Extent of keeling probably does not distinguish it from brachycephalus as thought by Lips and Savage (1994: 411), but the ventral color pattern (dark pigmentation across bases of ventrals causing a fuzzy pattern of alternating dark and white cross-bands) distinguishes talamancae at least from Panamanian G. brachycephalus and Geophis, species inquirenda. This type of ventral pattern (suggested in one case to mimic noxious millipedes [Leonard and Stebbins, 1999]) is approached in some specimens of the Colombian G. nigroalbus (FMNH 54882, LACM 136675). Although I do not suggest conspecificity, it needs noting that G. nigroalbus is not distinguished from talamancae by separation of postocular and supraocular by an anterior projection of the parietal (Lips and Savage, 1994: 411).18

Finally, a word on the Colombian Geophis betaniensis Restrepo and Wright (1987), which is known from two specimens collected in sympatry with the similar-sized G. nigroalbus. It is a distinctive species, separated from nigroalbus at a glance by dark ventrolateral stripes that sharply confine and set off the pale midventral area. Geophis betaniensis may be more closely related to G. nigroalbus than initially thought. Restrepo and Wright (1987: 191) provisionally assigned G. betaniensis to the championi species group, but Lips and Savage (1994: 414) later moved it to the sieboldi group based on “a complete concordance” in head shape and scutellation. Cursory examination of the paratype of G. betaniensis shows that it lacks certain derived maxillary characteristics shared by most members of the sieboldi group; the maxilla is not ventrolaterally compressed and lacks a toothless anterior projection (the first tooth springs from the anterior tip of the bone). Nonetheless, a Costa Rican member of the sieboldi group—Geophis zeledoni—also lacks a “distinct toothless tip” on the maxilla, which is dorsoventrally flattened only along its posterior half (fide Downs, 1967: 145, 175). But the maxilla of Geophis betaniensis is relatively high and I see no sign of dorsoventral compression even posteriorly (right maxilla examined in situ). Tentatively accepting Lips and Savage's (1994: 414) reassignment of betaniensis (as a species with presumably plesiomorphic maxillary features) makes the sieboldi group of Geophis the only one known from South America.

Geophis betaniensis was described as having “a small tentacle-like projection from the posterior margin of the nostrils” (Restrepo and Wright, 1987: 191–192, fig. 3). A photograph shows the “hair-like ‘tentacle’” on the right side of the holotype. A tentacle was said to be present only in the left nostril of the paratype (LACM 136189), but I could see no sign of it when I examined the specimen in early 2002. My own guess is that these structures, initially present in three of four nostrils, were foreign to the two snakes—perhaps defensive arthropod urticating hairs that might glance off the snakes' head scales but temporarily lodge in their nostrils.

It is of course impossible to determine at this time whether there is significant color and size variation in Geophis bellus, but its specific status, as first suggested by the striking coloration and small adult size, is strongly corroborated by hemipenial differences: In G. bellus the hemipenis (fig. 14) is bilobed for about a fourth of its length, with weak calyces only on the distal lobes; the large capitulum otherwise is sparsely covered with small spines (or large papillae). In G. nigroalbus (fig. 17) and G. brachycephalus (fig. 20) the organ is barely or not bilobate and the entire capitulum is covered with well-developed calyces.

Remarks

I found the unique specimen of Geophis bellus at night, while walking on a muddy road through cut-over evergreen seasonal forest, at 700 m above sea level. The type locality is on the “Piedras-Pacora Ridge”, the low continental divide that separates the watershed of the Río Pacora (Pacific versant) from that of the upper Río Chagres (Atlantic). Geophis bellus shares the Piedras-Pacora Ridge with G. brachycephalus and G. hoffmanni (see below), and also with the new species Atractus depressiocellus and A. imperfectus of this paper. The one new species of Geophis and two new Atractus are known only from single specimens collected over a period of 29 years. The two previously named species of Geophis seem to be just as rare on the Piedras-Pacora Ridge (although common in other parts of their ranges), since each is known there from a single specimen, as follows.

Geophis brachycephalus (Cope) Figures 15B, 18, 20A; map 2

Taxonomic Comment

Downs (1967: 149) commented that “The color pattern is the most perplexing variable in G. brachycephalus, and is at least partly responsible for the lengthy synonymy”. Some populations appear to be highly polymorphic in color pattern, as exemplified in Panama by Slevin's (1942: 474–476) description of a large population sample from the vicinity of Boquete in the Chiriquí highlands:

However, I strongly suspect that some snakes with uniformly dark dorsa from along the Atlantic versant of western Panama (and Costa Rica?) are currently masquerading in collections under the name Geophis brachycephalus. Although I am not prepared to give adequate attention to this problem, it may be helpful to call attention to it by segregating the possible sibling species as follows.

Geophis, species inquirenda [G. brachycephalus, auctorum] Figures 19, 20B

Commentary

The earliest specimens, from La Loma in Bocas del Toro, were collected and identified as Geophis brachycephalus by Dunn (1942: 4). Downs (1967: 154) had no additional specimens from Bocas del Toro, but he did mention a few small Costa Rican samples lacking blotches. I have not seen these, but I have collected and examined additional Panamanian specimens that may be conspecific with the La Loma specimens and distinct from brachycephalus.

These specimens are uniformly black or brown in preservative (not bluish as in some albeit not all preserved G. brachycephalus) and all lack pale spots or partial ringlike markings (a vague pale nape band present or absent). The hemipenis of an Atlantic-versant specimen differs from that of a highland specimen of G. brachycephalus in having a relatively long, slender capitulum with no trace of bilobation (fig. 20). However, part of the difference (especially relative slenderness—but not relative length—of the capitulum) conceivably might be an artifact of eversion technique (see comment under Methods of Study).

Several of the newer specimens are from low and intermediate elevations (40–830 m) in Bocas del Toro, but four are from a collection that I made in the highland valley of the Río Chiriquí in Chiriquí Province (1000–1120 m). This valley may be considered as delimiting the eastern end of the Cordillera de Talamanca and the western end of the Serranía de Tabasará. This high Pacific-side valley is influenced by moisture-laden air spilling southward over the low continental divide, providing access to animals and plants from the Caribbean versant (the herpetofauna also has a strong Talamancan component, but only two of 50 species collected have primarily Pacific-drainage distributions [Myers and Duellman, 1982: 14]). Unfortunately, the Geophis collected in the valley of the upper Chiriquí are all females, and their assignment to the Atlantic-versant population is based on segmental counts and absence of pale spots or partial rings in all four specimens.

There is a pronounced difference in numbers of ventrals and subcaudals between Panamanian populations with color polymorphism (brachycephalus) and those apparently without (species inquirenda). For purposes of comparison, I calculated population statistics from the counts given by Slevin (1942: 474–476) for his large collection of Geophis brachycephalus from the vicinity of Boquete, where it was the most abundant snake in the cafetals. As shown in table 3, Boquete specimens of brachycephalus have significantly higher numbers of ventrals and subcaudals in both sexes. There is no overlap in female ventral counts. Overlap in male ventrals and subcaudals is caused by one brachycephalus (CAS 78983) with exceptionally high counts of 142 + 46, without which the Boquete ranges would be 121–129 ventrals and 37–41 subcaudals. The high ventral count, however, was confirmed by Downs (1967: 148), who gave 140 ventrals for this specimen. The differences are maintained when adjustments are made for somewhat dissimilar methods of counting ventrals.19

Downs (1967: 148) remarked that ventral counts are somewhat lower in Panamanian G. brachycephalus from Finca Lérida to the west of Boquete, and that both these Panamanian populations have considerably lower counts than in Costa Rican populations. The ranges and means for all his Costa Rican specimens combined were given as 131–148 (138.5) in 54 males and 135–145 (140.7) for 70 females. These ranges encompass those for species inquirenda in Panama, but little more can be said without detailed examination of the situation in Costa Rica, from where considerably more material is now available. See Remarks below.

Color in Life

In life, the dorsa are uniformly black; preserved specimens (MCZ 19325–19326, KU 110701 [see below]) may fade to brown with the head becoming lighter than the body. Two specimens from the Río Changena (FMNH 130609, KU 110702) in preservative have pale but vague nape bands, which may have been more pronounced in life (grayish white in KU 110702). My field descriptions for several specimens follow:

Natural History

The habitat is lowland and lower montane rain forest and cloud forest in an elevational range of 40–1120 m. One (KU 110702) was under a log by day. I found a lowland specimen (KU 110701) as it was crawling on a log across a rocky forest steam, at night in the rain. The highest elevation specimen (AMNH 124015) was active at night, on the ground in cloud forest, at the edge of a small seepage pool in which Hyla graceae was calling. Three specimens (AMNH 114317–114319) from the area of the Fortuna Dam site were found on a second-growth hillside; a workman found one, I caught another at night on a trail near the river, and I found the third by day—on the ground near a rocky stream, at about 10 a.m. on a sunny morning.

Remarks

A satisfactory resolution of the taxonomic problem presented by this species inquirenda is outside the scope of the present paper, which I had intended to confine to problems in eastern Panama. Costa Rican specimens and names currently in the synonymy of Geophis brachycephalus will need to be considered. There especially needs to be examination of more hemipenes to determine if the differences seen in figure 20 are correlated with presence or absence of color-pattern polymorphism.

Geophis hoffmanni (W. Peters) Map 2

Downs (1967: 155) gave the range of this species as “low and moderate elevations in Honduras … and Nicaragua, southward along both Caribbean and Pacific versants of Costa Rica into Panama”. This little snake evidently is common in parts of Costa Rica and has long been known to occur in extreme western Panama; Downs' easternmost station in Panama was El Valle de Antón in Coclé Province, based on AMNH 76016; a second specimen (KU 116902) has subsequently been collected there. El Valle de Antón, an old volcanic crater, is shown as the westernmost station in map 2.

As listed below, two additional specimens extend the range eastward in Panama from El Valle to the central lowlands and thence to the Piedras-Pacora Ridge (map 2), and a third specimen indicates the occurrence of G. hoffmanni in Colombia. All three specimens are dark brown above, whitish below, with dorsal scales smooth or mostly smooth, in 15–15–15 rows, supralabials 5 (3–4 in eye); temporals absent (see Remarks); infralabials 6 (1–3 in contact with first genials).

As discussed elsewhere (Cadle and Myers, 2003),

Remarks

As discussed by Downs (1967: 159), the reduction to five supralabials in Geophis hoffmanni appears to be the result not of scale fusion but of a shortening of the jaw. A postlabial (i.e., the first scale not bordering the free edge of the lip) appears to be a homologue of the sixth supralabial in related species. And a scale dorsal to the postlabial and ventral to the posterior part of the parietal appears to represent the original posterior temporal—temporals being pragmatically defined simply as scales positioned mainly or completely between the parietals and posterior supralabial(s).

The easternmost record of Geophis hoffmanni in Panama is in the Piedras-Pacora ridge system (map 2) and is of particular interest in the context of the present paper. Of the snakes from the eastern half of Panama that are treated herein, five of the eight species occur in that relatively small area. The following remarks document the locality for Geophis hoffmanni, the most widely occurring of the five species.

After the event of several fatal human cases of sylvan yellow fever in 1948–1949, the Gorgas Memorial Laboratory established temporary field stations in the forest at Cerro La Victoria in Panamá Province east of Panama City. Primary objectives were to collect mosquitoes simultaneously from the forest floor and from platforms up into the upper canopy, and to survey arboreal forest mammals in order to discover natural reservoirs of the yellow fever virus.

Cerro La Victoria is not shown on standard maps nor is it listed in Fairchild and Handley's (1966) gazetteer of GML collection sites, but it was an important GML collecting area during the early 1950s. I know Cerro La Victoria only through the publications of Galindo et al. (1950, 1951, 1956), the first two of which contain outline maps showing a series of collecting stations aligned northwest of the Pacific lowland town of Pacora, about midway between Pacora and the mouth of the upper Río Chagres at Madden Lake.

From the above information, it is evident that Cerro La Victoria is a high, forested ridge above the west side of the south-flowing Río Cabra, which drains the southern slopes of the Piedras-Pacora Ridge (i.e., the continental divide) to the west of the headwaters of the Río Pacora. It is an area of premontane rain forest. The La Victoria Ridge connects to the main divide probably in the general vicinity of Cerro Tobova (9°12′N, 79°23′W).21

Therefore, the northern end of Cerro La Victoria on the Pacific versant is only several km south of the better known Cerro Jefe in the Atlantic versant (shown as “El Jefe” in the Madden Lake Watershed inset in the old 1:100,000 map Canal Zone and Vicinity published by the U.S. Army Map Service). Cerro Jefe lies just to the north and higher than the continental divide, at 9°14′N, 79°23′W, with an elevation of 3254 ft (982 m) according to the aforesaid map and as confirmed by my altimeter reading of 980 m (Myers, 1969: 27). From the summit of Cerro Jefe, during the occasional clear period, the Cerro La Victoria region (including Cerro Tobova) and the Pacific coastal plain are visible to the south.

Trapido and Dunn's specimen of Geophis hoffmanni, from 2200 ft. (670 m), apparently was collected above station D near the crest of Cerro La Victoria (Galindo et al., 1950: fig. 7 [aerial photo showing location of station D]). The specimen was said to have been found “in bromeliads” (presumably ground bromeliads) with a curious assemblage of other small reptiles.22 Half a century ago, the Pacific-versant forest in the immediate region was already being encroached by road building and slash-and-burn agriculture (Galindo et al, 1950: figs. 2–3, 7–8), and the forest by now has either disappeared or reverted to second growth.

MUSEUM ABBREVIATIONS AND ACKNOWLEDGMENTS

Abbreviations: The following collection abbreviations are used in this paper: