Conventional Trapping

We used standard trapping equipment (Voss and Emmons, 1996) to sample the local fauna of small marsupials and rodents near ground level during our 1991 and 1992 field seasons. Most traplines included both Victor rat traps (with push-down wooden bait pedals) and folding aluminum Sherman live traps (measuring 80 × 90 × 230 mm) set at approximately 20-m intervals along existing trails through our study area (fig. 2). Because our objective was to sample the fauna as effectively as possible, no attempt was made to randomize or standardize the placement of traps. Instead, traps were placed to maximize capture success for the widest range of anticipated species as suggested by prior experience in other habitats and at other rainforested localities. Victor rat traps were sometimes set on the ground (often inside the dark cavities of hollow logs), but we usually tied them to lianas, tree trunks, and other woody supports 0–2 m above the ground to capture semiarboreal species (fig. 3, top). Sherman traps were usually placed on the ground next to fallen trunks, under large-leaved vegetation (palms, heliconias, etc.), beneath viny tangles and piled branches, or in similarly sheltered terrestrial situations (fig. 3, bottom).

Victor rat traps were usually baited with a ground-up mixture of peanut butter, raisins, rolled oats, and bacon (in 6:2:2:1 proportion); although this bait effectively attracts many species of small mammals, it also attracts ants and must be renewed daily in most situations. Sherman live traps were baited daily with either cracked corn (sold locally as chicken feed) or with commercial birdseed; both baits were effective when dry, but cracked corn becomes sticky when wet and subsequently spoils unless the trap is washed. Traplines were checked twice daily, usually at dawn and in the late afternoon (between 16:00 and 18:30 hours). All traps were rebaited in the late afternoon. Sherman traps were disassembled, thoroughly washed (without soap), and dried after each capture to remove urine, feces, and old bait. The treadles of Victor traps were scraped clean of old bait at each rebaiting, but the traps themselves were not washed. We marked all trap locations with brightly colored vinyl surveyor's tape, and all Victor traps were tethered to nearby stems or roots with a short (ca. 80 cm) length of braided cotton or nylon cord.

Most trapped trails traversed both well-drained and swampy primary forest, the two principal habitat types sampled by this method. In addition, a few traplines were set for semiaquatic species by wading small streams and setting traps along the banks, and some traps were set in the secondary vegetation bordering our camp clearing and nearby roads. We did not trap in the coastal savannas north of our study area.

We periodically used Tomahawk folding wire live traps (measuring 145 × 145 × 410 mm) to capture large (>300 g) terrestrial marsupials and rodents (fig. 4). Most Tomahawk traps set on the ground were baited with coconut or ripe plantain, and a few unbaited large Tomahawks (measuring 250 × 300 × 810 mm) were set in streams for semiaquatic species. We used small leghold traps and Conibear break-back traps with and without bait or commercial scent lures in 1991, but not in subsequent field seasons; most of these traps were set away from established trails, often in burrows or on logs crossing small streams. In 1992 we also constructed and field-tested two live traps designed for use in armadillo tunnels, but these were unsuccessful and the effort was discontinued in subsequent years.

Arboreal Trapping

Because some species of rainforest marsupials and rodents seldom descend from the canopy or subcanopy, we used a platform-trapping system similar to that described by Malcolm (1991) to capture arboreal species in 1993. Trees were climbed using French pole climbers (see Mori, 1987: fig. I-3A), of which a small pair with a maximum tooth-to-tooth span of 25 cm and a larger pair spanning 35 cm were useful for climbing boles of different diameters. Wooden platforms provided with eyelets for raising and lowering trap assemblies by nylon lines were nailed to tree trunks 7–19 m above the ground in both well-drained and swampy primary forest. Each trap assembly consisted of one Tomahawk folding wire live trap (measuring 145 × 145 × 410 mm) and one piggy-backed Sherman folding aluminum live trap (measuring 80 × 90 × 230 mm) attached to a wooden frame (fig. 5). Sherman traps were baited with the same oatmeal/peanut/raisin/ bacon mixture described previously, and the Tomahawk traps were baited with pieces of coconut and ripe plantain; bait was renewed as necessary, about every three to four days on average. Tree traps were checked daily soon after dawn with binoculars.

Pitfall Trapping

Many species of small terrestrial mammals are not attracted to baited traps, so we used a pitfall trapping design suggested by S. M. Goodman (personal commun.) to supplement our conventional trapping in 1993. We used 15-liter plastic buckets as pitfalls, and sunk these flush with the ground in linear series beneath sheet-plastic drift fences (fig. 6). Each pitfall trapline consisted of 11 buckets spaced 5 m apart, with one bucket at either end, for a total length of 50 m. Drift fences, consisting of a continuous barrier running the entire length of each trapline, were made of 50-cm-wide strips of heavyweight (6 mil) clear polyethylene stapled to vertical stakes hammered into the ground every 3–4 m. A broad (7–8 cm) fold of plastic was heaped with soil to anchor the bottom of the fence, and a flap was cut in this fold where it passed over each bucket.

We perforated the bottoms of the buckets so that rainwater could drain out, but the perforations soon became clogged and we then noticed that undrained buckets containing 10–15 cm of water were much more effective traps than dry buckets. Thereafter, we added water to all of the buckets and tried to maintain this level throughout our pitfall-trapping effort. Most captured animals drowned in pitfalls containing water, but checking the traplines twice daily was sufficient to prevent specimen damage due to decomposition.

Hunting and Sight Surveys

We relied on hunting, sight surveys, and interviews (see below) to census the local fauna of large mammals (≥1 kg) that were difficult or impossible to capture with our trapping equipment. However, we seldom hunted or made sight surveys in the daytime because mornings were occupied by checking traplines, taking down bat nets, or preparing specimens, and afternoons were usually spent rebaiting traps or setting up bat nets. Most of our personal observations of diurnal mammals at Paracou were therefore made haphazardly in the course of other (relatively noisy) activities. Fortunately, local forestry workers were familiar with most species of large diurnal mammals known or expected to occur in the area, so interviews compensated to some extent for the deficiencies of our diurnal census effort.

By contrast, we regularly hunted at night, and we recorded many sightings of nocturnal species encountered by deliberate searching. Equipped with notebook, shotgun, and a nine-volt headlight with focusable reflector, we walked slowly (about 0.5–1.5 km/hour) along established trails scanning the vegetation with the headlight reflector adjusted to produce a moderately broad beam. When eyeshine was detected, we narrowed the beam to give brighter illumination; binoculars were sometimes used to identify animals revealed by eyeshine in the canopy or subcanopy. Most of our nocturnal surveys were made in the early evening (between 19:00 and 24:00 hours), but we also hunted between midnight and dawn on many nights. Some nonvolant species were observed in the course of bat netting, especially large marsupials attracted to the squeals of captured bats.

Specimens were collected by shooting with side-by-side 20-gauge shotguns. We used commercial 20-gauge game loads for large mammals, but small mammals were collected with .410 or .22 caliber shot-shells loaded in removable auxilliary barrels (“auxes”) machined from brass rod stock. Most unvouchered sightings of small marsupials and rodents could not be confidently identified to species.

Interviews

Small teams of day laborers have been continuously employed in forestry research projects at Paracou since the mid-1980s to cut trails, clear experimental plots, and to mark and measure tens of thousands of trees. Among these workers are members of the Saramaka and Boni tribes of so-called bush negroes, descendants of escaped slaves (maroons) who fled into the forests of Surinam and French Guiana and re-created traditional African communities that have resisted acculturation for centuries (Kahn, 1931; Hurault, 1961; Price, 1976). Recruited for their ability to identify commercially valuable timber, these people have observed (and hunted) the Paracou fauna for many years. We periodically discussed our inventory work with them, comparing our own observations with theirs and asking them about species we had not yet seen. Most of our useful interview data, however, was obtained from conversations with their French-born foreman, Pascal Petronelli, himself a gifted naturalist and a local resident since 1983.

Using distributional data summarized for another study (Voss and Emmons, 1996), we compiled a list of the nonvolant mammal species known or expected to occur in French Guiana, and we used this list as the basis for interviewing Mr. Petronelli in 1993. We focused primarily on large species (≥1 kg) that nonmammalogists might reasonably be expected to recognize without special effort, but we also asked about a few smaller species identifiable by obvious external characters. For each species expected to occur in the area but not observed by us, we asked if he had seen it himself, or whether the species had been seen by the forestry workers he supervised. If the species had been seen, we asked for details about where and when the sighting(s) occurred, and about distinguishing morphological or behavioral characteristics; we often used the illustrations in Emmons (1990) to discuss diagnostic differences among related species. For some species, we were able to examine photographs and/or written records (logs of noteworthy sightings on experimental plots). Finally, we asked Mr. Petronelli if any species not mentioned by us had ever been seen in the area.

Previous Mammalogical Research at Paracou

Whereas information about bat diversity at Paracou (summarized by Simmons and Voss, 1998) derived entirely from our own efforts, other researchers had previously worked on the local nonvolant mammal fauna prior to our study (Dubost, personal commun.; Forget, 1991, 1996, 1997; Henry, 1994, 1996, 1999; Forget et al., 1999). Their published results and unpublished observations provided supplementary information about local diversity, including records of three species that neither we nor our resident interviewees had identified. Of particular importance was the very large series of specimens trapped over several years by O. Henry, which we were generously allowed to inspect (through the good offices of L. Granjon) for taxa unrepresented in our voucher collection.

Voucher Preservation

We preserved voucher material of nonvolant species following standard protocols (e.g., Nagorsen and Peterson, 1980; Hall, 1981). As explained previously (Simmons and Voss, 1998: 19–21), our primary motivation in preserving voucher material was to obtain enough material from each species to assess the taxonomic status of the Paracou population with respect to museum samples from other Neotropical localities. Additionally, our voucher collection (now divided between the AMNH and the MNHN) provides a permanent archive of mammalian diversity at Paracou that can be consulted by other zoologists, who may confirm or emend our identifications in the light of future taxonomic revisions.

Nevertheless, we only collected voucher material for small species (few of which could otherwise be identified with confidence) and for some large mammals judged to be sufficiently abundant that removal would not significantly affect the local population. In general, samples consisting of about 10 adult males and 10 adult females are sufficient to obtain meaningful estimates of the mean and range of variation in each sex for characters of taxonomic interest, and our voucher series only exceed these minimal counts for a few of the commonest marsupials and rodents (e.g., Didelphis marsupialis, Philander opossum, Oryzomys megacephalus, Proechimys cuvieri). For open populations of most small nonvolant species, collecting on this scale is probably trivial by comparison with natural demographic processes.1

Marsupialia

Twelve species of marsupials, all traditionally classified in the family Didelphidae,2 are known to occur at Paracou, and two additional species could be expected to occur locally (see appendix 1). Our terminology for most qualitative aspects of marsupial morphology follows Archer (1976a, 1976b) and Hershkovitz (1992, 1997), but we adhere to the traditional interpretation of postcanine dental homologies (Flower, 1867; Luckett, 1993), wherein the replaced molariform tooth of the upper and lower jaw is the deciduous third premolar (dP3/dp3). Thus, the adult didelphid dental formula is I 5/4, C 1/1, P 3/3, M 4/4). We define specimens to be juvenile if dP3 is still in place, subadult if dP3 has been shed but P3 and/or M4 are still incompletely erupted, and adult if the permanent maxillary dentition is complete.

Our quantitative comparisons of marsupial crania are based on the following measurements (fig. 7):

Condylobasal Length (CBL): From the occipital condyles to the anteriormost point of the premaxillae.

Maxillary Toothrow (MTR): Crown length, from the anterior margin of the canine to the posterior margin of M4.

Molar Length (LM): Crown length of M1–4, measured on the labial side of the toothrow.

Palatal Breadth (PB): Measured across the labial extremes of the crowns of the last molars.

Palatal Length (PL): Measured in the midline from the anteriormost point of the premaxillae to the end of the palate.

Nasal Breadth (NB): Measured across the triple-point suture of the nasal, frontal, and maxillary bones on each side.

Least Interorbital Breadth (LIB): Measured at the narrowest point across the frontals between the orbits.

Least Postorbital Breadth (LPB): Measured at the narrowest point across the frontals behind the orbits.

Zygomatic Breadth (ZB): Greatest breadth across the zygomatic arches.

A few other craniodental measurements taken for special purposes are either self-explanatory or are defined in the following species accounts. Because most didelphids exhibit obvious sexual size dimorphism, we summarize morphometric variation separately by gender.

Most of the large opossums in the Paracou fauna (Caluromys, Chironectes, Didelphis, Metachirus, Philander) are easily distinguished, even at a distance, by external characters described by Emmons (1990, 1997). However, all of the “marmosine” opossums (species of Gracilinanus, Marmosa, Marmosops, and Micoureus; formerly placed in the genus Marmosa sensu Tate, 1933) require specimens in the hand for positive identification, and some cannot be confidently distinguished without cleaned cranial material. Husson (1978) provided a useful key to Surinamese marsupials based on craniodental characters, but four species now known to occur in Surinam and/or French Guiana were omitted: Didelphis albiventris, “Gracilinanus” kalinowskii, Marmosops parvidens, and Marmosops pinheiroi. Distinguishing morphological characteristics of these recently reported members of the eastern Guianan fauna are described and illustrated in the accounts that follow.

Caluromys philander (Linnaeus)

Voucher Material: AMNH 266402, 266408, 266409, 267330, 267331, 267333–267337; MNHN 1995.884–1995.886, 1995.894, 1995.902. Total = 15 specimens (not including pouch young).

Identification: Our material agrees closely in external and craniodental characters with Husson's (1978) detailed description of topotypic specimens from Surinam, and measurements of adult Paracou vouchers (table 1) broadly overlap those of topotypic adults (op. cit.: table 1).

In view of the fact that Caluromys has never been revised, it is noteworthy that published measurements of some nominal taxa currently regarded as synonyms of C. philander (see Gardner, 1993) fall outside the known range of variation among Surinamese and French Guianan specimens. In particular, specimens from Trinidad and northern Venezuela formerly referred to C. trinitatis Thomas (type locality: “Botanic Gardens, Trinidad”) are substantially smaller and differ from typical philander in coloration (Thomas, 1894, 1903, 1904; Pérez-Hernández et al., 1994). Furthermore, we have personally observed conspicuous variation in size and pelage traits among populations currently referred to C. philander from Amazonian Brazil. Although Cabrera (1958) recognized C. p. philander (including trinitatis as a synonym), C. p. affinis (type locality: Mato Grosso, Brazil), and C. p. dichrurus (type locality: “Ypanema”, São Paulo, Brazil) as valid subspecies, the empirical basis for a trinomial nomenclature has never been established. The material now available to evaluate geographic variation and subspecies (or species) limits within what might be called the Caluromys philander complex is too large to review in this faunal report, but from the close similarity noted above between Surinamese and French Guianan material it seems clear that the Paracou population is referable either to the nominate subspecies (if a trinomial nomenclature is warranted) or to C. philander sensu stricto (if additional species are recognized in this group).

Field Observations: All of our 15 vouchered records of Caluromys philander at Paracou are from specimens trapped or shot at night in trees in primary forest, at both well-drained and swampy sites. In addition, we recorded three unvouchered nocturnal observations of this species, two of which were sighted in trees in roadside secondary growth and the third on a liana in primary forest. Seven (47%) of our vouchers were obtained by shooting animals sighted in the forest canopy or subcanopy, and eight others (53%) were taken in subcanopy platform traps. Measured heights above the ground for the trapped specimens ranged from 12 to 16 m, whereas visually estimated heights of sightings of free-ranging (untrapped) animals ranged from 3 to 20 m. Figure 8 provides a typical view of the subcanopy habitat of this species at Paracou.

All four of our adult female vouchers were carrying suckling young. The first, captured on 10 August 1991, had three nursing young measuring 36 mm crown-rump; the second, taken on 14 August 1991, had five young measuring 47 mm crown-rump; the third, taken on 16 November 1992, had two young measuring 14 mm crown-rump; and the fourth, taken on 28 August 1993, had four young measuring 54 mm crown-rump. With these exceptions, all of our vouchered and unvouchered records of Caluromys philander are based on solitary individuals; no others were trapped or sighted together.

Chironectes minimus (Zimmermann)

Voucher Material: AMNH 266477, 266478; MNHN 1998.672. Total = 3 specimens.

Identification: Our voucher material conforms closely to most published descriptions of this widespread and distinctive species (e.g., Thomas, 1888; Cabrera, 1919; Krumbiegel, 1940a; Augustiny, 1942; Mondolfi and Medina, 1957; Husson, 1978; Marshall, 1978b; Emmons, 1990, 1997), but a few discrepancies and supplementary observations merit comment. Persistent references (op. cit.) to brownish and yellowish tints in the pelage of Chironectes minimus are probably based on old (faded or stained) museum skins; living animals and fresh skins have clear black-and-gray dorsal markings and pure white venters, with no trace of other hues. According to Cabrera (1919), the digits of the manus are webbed to the ends of the first phalanges, but our specimens (and those illustrated by Augustiny [1942: fig. 14] and Mondolfi and Medina [1957: fig. 1]) have unwebbed manual digits. The “supernumerary facial bristles” mentioned by Marshall (1978b) are the usual superciliary, genal, and interramal vibrissae (Brown, 1971), which are well developed (Augustiny, 1942: fig. 11) but otherwise unremarkable in C. minimus. Apparently, the only postcranial vibrissae in this species consist of a prominent tuft of long carpal hairs at the wrist.

Our only adult voucher (AMNH 266477), an old male, had external measurements of 286 × 345 × 63 × 31 mm and weighed 620 g. Selected craniodental measurements of this specimen fall within the known range of morphometric variation for the species (Marshall, 1978b): condylobasal length, 65.7 mm; length of molars, 14.5 mm; palatal breadth, 23.5 mm; palatal length, 42.7 mm; least interorbital breadth, 13.1 mm; least postorbital breadth, 8.4 mm; zygomatic breadth, 39.0 mm; length of nasals, 31.8 mm. Both of our other vouchers are juveniles.

Remarks: Lutra minima Zimmermann (1780) was based on Buffon's (1776) description of the “petite loutre d'eau douce de Cayenne”, so our specimens are practically topotypes. Krumbiegel (1940a) and Marshall (1978b) both recognized four subspecies, but the necessity for a trinomial nomenclature for water opossums has yet to be demonstrated by any substantive analysis of character data. In the event that any subspecific distinctions are warranted, the Paracou population would obviously be referable to the nominate form.

Field Observations: Water opossums were commonly seen in all of the four named stream systems that have their headwaters in our study area; even the smallest and shallowest creeks were frequented (fig. 9). Because this species has bright eyeshine, is boldly marked, and splashes noisily while swimming or wading, it is not difficult to observe despite its nocturnal habits. Nevertheless, the occurrence of Chironectes at Paracou was previously unsuspected by the forestry workers and local hunters whom we interviewed.

One of our three vouchers was taken in a large (ca. 25 × 30 × 81 cm) wire live trap set in a small (ca. 2 m wide) shallow (ca. 15 cm deep) stream in primary forest; the trap was unbaited, but rows of stakes were driven unto the streambed on either side to funnel animals moving downstream into the trap opening. The other two vouchers were collected by shooting. In addition, we recorded 23 unvouchered observations of water opossums, of which 2 were based on juveniles trapped in small (145 × 145 × 410 mm) wire live traps set in streams, and 21 were sightings of free-ranging individuals.

All of our 26 (vouchered and unvouchered) records of water opossums at Paracou were of animals trapped or sighted while they waded or swam in primary forest streams at night. Most free-ranging individuals alarmed by our presence quickly swam away, but two animals left the water and disappeared in dense streamside vegetation, and another entered a burrow near the water's edge. With the exception of a pair of animals apparently engaged in an aggressive interaction, all of our sightings were of solitary individuals.

Didelphis marsupialis Linnaeus

Voucher Material: AMNH 266456–266460, 266462–266466, 266468, 266470, 266471, 266473, 266475, 267367; MNHN 1995.895–1995.901. Total = 23 specimens (not including pouch young).

Identification: All of our specimens of Didelphis from Paracou are referable to the large, black-eared species D. marsupialis, the type locality of which was restricted by Thomas (1911a) to Surinam. Our material agrees in all qualitative details with Husson's (1978) description of Surinamese topotypes, but three of the four adults measured by Husson are larger than any collected at Paracou. To evaluate possible size differences, we borrowed Surinamese material for side-by-side comparisons with our vouchers; however, only four skin-and-skull preparations of adult specimens (all females) could be located. No differences in external characters were observed between Surinamese and French Guianan exemplars, and the measurement data we obtained (table 2) do not suggest any appreciable morphometric divergence. Considerable ontogenetic size variation is apparently characteristic of Didelphis species (Allen, 1901, 1902; Gardner, 1973), and it seems likely that Husson's large specimens were just old animals that may have been preserved because of their unusual dimensions. The essential identity of the Surinamese and French Guianan material we examined supports the conclusions (hitherto undocumented by published comparisons) of Thomas (1888) and Allen (1902) that D. karkinophaga Zimmermann and D. cancrivora Gmelin (both based on “Le Crabier”, an opossum described by Buffon from Cayenne) are junior synonyms of D. marsupialis.

Julien-Laferrière (1991) reported Didelphis albiventris and D. marsupialis as occurring syntopically in primary forest at Piste St.-Élie (only 14 km WNW of Paracou), the first published record of the former species from French Guiana. Catzeflis et al. (1997) subsequently reported both species from primary forest near Petit Saut (about 28 km SSW of Paracou). In view of these records of sympatry from nearby localities, our failure to record the presence of D. albiventris at Paracou merits comment.

The diagnostic morphological characters of Didelphis marsupialis and D. albiventris are sufficiently striking that collected specimens cannot be misidentified by competent researchers alert to the possible presence of both species. In the Guiana subregion of Amazonia, these taxa are readily distinguished by facial markings (bolder in albiventris than in marsupialis; Mondolfi and Pérez-Hernández, 1984), ear color (the pinnae are usually tipped with white in albiventris but are entirely black in adult marsupialis; op. cit.), caudal pelage (the furred basal portion of the tail is conspicuously longer in albiventris than in marsupialis; M. D. Engstrom, personal commun.), and size (albiventris is smaller; see measurements of the upper molar row of albiventris tabulated by Mondolfi and Pérez-Hernández, 1984). Correlated molecular characters that may be useful for discriminating albiventris and marsupialis were recently discussed by Lavergne et al. (1997).

Although D. albiventris is definitely absent from our voucher material, we did not collect every individual Didelphis that we encountered at Paracou. Because many animals were sighted in dense vegetation where facial markings and other potentially informative details could not be distinguished, it is therefore possible that this species was seen but not recognized. Unfortunately, Didelphis virtually disappeared from our study area from 1992 to 1994, so we had little opportunity to collect specimens after we learned that D. albiventris could be expected to occur locally.

Other Specimens Examined: Surinam—Brokopondo, Brownsberg Nature Park (CM 52697); Coronie, Totness (CM 52702); Saramacca, Bigi Poika (CM 52716, 52724). Listed specimens are adult skin-and-skull preparations; other Surinamese material examined (CM, FMNH) are immature (subadults, juveniles) or incomplete (skull only/skin only) specimens that are not useful for taxonomic comparisons.

Field Observations: All of our unambiguous records of Didelphis marsupialis from Paracou are based on collected specimens. Of these, 15 (65%) were trapped and 8 (35%) were shot. Nine specimens (39%) were trapped or shot in trees or other elevated substrates at heights ranging from 0.3 to 15 m above the ground, whereas the remainder were trapped or shot on the ground. Of 21 specimens accompanied by habitat data, 11 (52%) were taken in clearings, roadside secondary growth, or other disturbed habitats; the other 10 (48%) were taken in primary forest, at both well-drained and swampy sites. All specimens were shot or trapped at night, and all unvouchered sightings of Didelphis were also nocturnal. All encountered individuals were solitary; no collected female was carrying suckling young.

Gracilinanus emiliae (Thomas) Figures 11A, 11B, 12A, 17C, 17D, 18C, 18D

A single specimen from Paracou (AMNH 267006) is the first of this widespread but rarely collected species to be reported from French Guiana. Because the brief accounts by Thomas (1909), Tate (1933), Husson (1978), and Hershkovitz (1992) are inadequate for evaluating morphological similarities and differences with other congeners, we redescribe the species below.

Type Material: The holotype only, a male skin with skull and mandibles (BMNH 9.3.9.10) collected by Emilie Snethlage on 13 February 1909 at “Para” (= Belém, formerly known as Pará), Brazil. The type was described by Thomas (1909: 379) as a “subadult”, by Tate (1933: 189) as a “young adult”, and by Hershkovitz (1992: 33) as a “juvenal”. The animal is, in fact, very nearly adult with P3 and M4 both erupted but still a little below their adult positions in the toothrow.

Pine (1981: 59) suggested that the type locality should be construed as the state of Pará rather than the formerly eponymous city, but Thomas (always scrupulous about type localities) would surely have noted the lack of definite geographic information if he meant “Para” in the sense of a district larger than most European countries. A specimen collected at Capim (ca. 90 km ESE of Belém) provides independent evidence that the species may occur in the environs of the city.

Distribution: Specimens that we examined document the presence of Gracilinanus emiliae in eastern Colombia, eastern Venezuela, southern Guyana, northern Surinam, northern French Guiana, and eastern Brazil (fig. 10). Several other published localities for this species are erroneous or unreliable (see Remarks, below).

Emended Description: Very small murine opossums (table 3) with smooth (not woolly) adult pelage; unruffled dorsal fur dull reddish brown, but basal two-thirds dark gray; ventral fur pure white or cream from chin to groin (the hairs self-colored, not grayish basally). Face marked by mask of black fur extending from mystacial pad to just behind outer canthus of eye on each side, and by narrow midrostral streak of pure orange fur; cheeks (below mask) white or cream-colored like throat; facial vibrissae (including long superciliary, mystacial, and genal hairs) mostly black (but some of the ventralmost genals are usually white); vibrissae of chin and throat (submental and interramal hairs) white. Ears not very large (just covering eye when laid forward over face), apparently naked (a sparse pelage of very short hairs is only visible under magnification), and very thin; opaque and pale basally (possibly yellow in life, but whitish in fresh alcoholic specimens); translucent and darker (brownish or grayish) distally. Gular glands (indicated by a naked or sparsely haired median patch of skin on the throat) present in all specimens examined, including one juvenile.

Wrists, ankles, and dorsal surface of feet covered with short pale (whitish or orange-tinted) hairs. Manus and pes each with six plantar pads (thenar, hypothenar, and four interdigitals); thenar and first interdigital pads of manus separated by at least two rows of minute epidermal tubercles; thenar and first interdigital pads fused on pes of one fluid specimen (AMNH 267006), touching but not fused on pes of two others (ROM 35465, 35466); central palmar surface of manus smooth (not densely tubercular); claws of manual digits II–V small, not extending beyond fleshy apical pads. Scrotal epidermis of holotype entirely unpigmented, of another subadult (AMNH 267006) with dark dorsal blotch surrounding suspensory stalk, of one fully adult specimen (ROM 35466) entirely dark. At least nine (4–1–4) abdominal-inguinal mammae present in one adult female (ROM 35465).

Tail much longer than head-and-body (table 3); less than 1 cm furry at base; uniformly dark (grayish or brownish) without pale blotches, bands, or countershading. Caudal epidermis covered with very small scales in annular or spiral series,3 numbering 40–50 rows/cm at middle of tail (counts from three fluid specimens and two dried skins). Median hair of triplet emerging from posterior margin of each scale about two scale rows long, thicker than lateral hairs, but not grossly flattened or petiolate.

Skull (figs. 11, 12) with slender rostrum, incipiently beaded supraorbital margins, and absence of postorbital processes of frontals; orbits not conspicuously enlarged (delimited posteriorly by well-developed postorbital processes of jugals); braincase not greatly inflated, smooth and unmarked by prominent temporalis scars. Premaxillaries with small but distinct rostral process anterior to incisors; maxillary-premaxillary suture extending to posteriormost incisor alveolus; palate highly fenestrated, with large maxillopalatine and posteromedial (palatine) vacuities; one or more small maxillary palatal vacuities present unilaterally in one specimen (AMNH 203363), bilaterally in three others (BMNH 9.3.9.10, ROM 35466, RMNH 18231). Alisphenoid wing of auditory bulla with well-developed anteromedial process bridging foramen ovale.

Upper incisors 2–5 subequal (not increasing in size from front to back); upper canine with small posterior accessory cusp; lower canine procumbent and premolariform (with flattened blade-like apex and small posterior accessory cusp); deciduous third premolars (dP3/dp3) large and molariform (figs. 17, 18); permanent third upper and lower premolars (P3/p3) slightly smaller than second premolars (P2/p2); metacones much larger than paracones on M1–3.

Comparisons: Gracilinanus emiliae and a sympatric taxon originally described by Hershkovitz (1992) as G. kalinowskii are similar in size and coloration (both have reddish-brown dorsal fur and pure white or cream-colored ventral fur), but they differ in many other characters as explained in the following account. Marmosa lepida (not known to occur at Paracou but reported from other localities in French Guiana and Surinam; appendix 1) is also externally similar to G. emiliae. Although adults of M. lepida are much larger than G. emiliae, immature specimens of the former species might be almost impossible to distinguish from the latter in the field. Cranially, adult M. lepida (see Husson, 1978: pl. 11) can be identified by their (1) distinctively longer premaxillary rostral processes, (2) prominent postorbital frontal processes, (3) lack of posteromedial palatal vacuities, (4) lack of anteromedial bullar processes, and (5) lack of mastoid exposure between the squamosal and parietal bones. Skulls of juvenile M. lepida that lack postorbital frontal processes can still be distinguished from G. emiliae by rostral, palatal, and bullar morphology, and by tooth size (see under Remarks, below).

Remarks: Originally described as Marmosa emiliae (Thomas, 1909), this species was subsequently referred to Didelphis (subgenus Grymaeomys) by Matschie (1916), and to Marmosa (subgenus Thylamys) by Cabrera (1958). The current allocation of emiliae to Gracilinanus follows Gardner and Creighton (1989).

Gracilinanus longicaudus, named by Hershkovitz (1992) from a single specimen (FMNH 87924) collected in eastern Colombia, does not differ significantly from G. emiliae in any external, cranial, or dental character. According to the original description, “… the combination of small size, long tail, whitish underparts, incomplete eye ring, and narrow skull separates [longicaudus] from all other described species [of Gracilinanus]” (op. cit.: 39). However, the black mask of FMNH 87924 does, in fact, completely encircle the eye (a narrow border of black hairs is continuous below the lower lid and around the posterior canthus), and the other characters cited as diagnostic for longicaudus are matched by the type (and other referred specimens) of emiliae. Our side-by-side comparisons of AMNH 203363 (an adult male skin and skull from Capim, Brazil) with the holotype of emiliae in London and, later, with the holotype of longicaudus in Chicago, revealed no differences beyond those that might be expected among individuals from a single local population (for measurements, see table 3). We therefore regard G. longicaudus as a junior synonym of G. emiliae.

Eisenberg (1989) suggested that Gracilinanus emiliae might be conspecific with G. microtarsus from the Atlantic coastal rainforest of southeastern Brazil, but it would be difficult to select two more dissimilar congeners for comparison. Among other differences, Gracilinanus microtarsus is much larger and has a proportionately much shorter tail (see measurements in Tate, 1933; Hershkovitz, 1992), the ventral fur is buffy and gray-based, the supraorbital margins are not beaded, and the crowns of I2–5 increase in size from front to back.

Gardner and Creighton (1989), Hershkovitz (1992), and Gardner (1993) listed Marmosa agricolai Moojen (1943) as a synonym of Gracilinanus emiliae, but Moojen's illustration (op. cit.: fig. 1) and description are difficult to reconcile with this decision: the ratio of tail to head-and-body for agricolai is only 1.28, the maxillary-premaxillary suture (as drawn) does not extend anteriorly to I5, the frontals are shown without supraorbital beads, the zygomatic arches are widely flared, the illustrated bullae appear to lack anteromedial processes, the upper canine appears to have a small anterior accessory cusp, and the third upper premolar (as drawn) is larger than the second. The type of agricolai (in Rio de Janeiro) should be reexamined to evaluate the true status and relationships of this nominal species.

We examined the two Brazilian specimens identified by Patterson (1992) as Gracilinanus emiliae and found them to be immature examples of Marmosa lepida. Both skulls have very long rostral processes of the premaxillae, lack posteromedial palatal vacuities, and their rounded alisphenoid bullae lack anteromedial processes; the erupted elements of the molar dentition are also larger than those of any specimens of G. emiliae but match the homologous teeth of adult M. lepida from eastern Peru (e.g., AMNH 78001, 98656). Because immature Marmosa lepida can be confused with Gracilinanus emiliae in size and external appearance, literature records of the latter (e.g., Ávila-Pires, 1964) should be regarded as suspect until diagnostic characters can be reconfirmed from specimens.

Other Specimens Examined: Brazil—Pará, Belém (BMNH 9.3.9.10, type), Capim (AMNH 203363). Colombia—Meta, Los Micos (FMNH 87924, type of Gracilinanus longicaudus). Guyana—Upper Takutu-Upper Essequibo, 12 km E Dadanawa (ROM 35465, 35466), no other locality data (ROM 33807). Surinam—Marowijne, Langamankondre (RMNH 18231). Venezuela—Monagas, 47 km SE Maturín (USNM 385066).

Field Observations: Our single specimen from Paracou was shot at 22:18 hours on 21 October 1992 as it perched about 4 m above the ground in dense secondary growth along a dirt road through well-drained forest (fig. 13).

Hyladelphys, new genus

Diagnosis: Very small didelphids distinguished from all other family members by the following combination of traits: four mammae in two abdominal-inguinal pairs; dorsolateral margins of frontals beaded and strongly convergent anteriorly, without postorbital processes; premaxillae short, without rostral process anterior to incisor row; posteromedial (palatine) palatal vacuities absent; tympanic wing of alisphenoid without a well-developed anteromedial strut forming secondary foramen ovale; I2–5 not increasing in size from front to back, their crowns asymmmetrical and nonoverlapping; upper canine without anterior or posterior accessory cusps; P2 much larger than P3; deciduous premolars (dP3/dp3) very small and nonmolariform; molars not highly carnassialized (paracone smaller than metacone on M1–2, but paracone and metacone subequal on M3); lower canine not premolariform.

Type Species: Gracilinanus kalinowskii Hershkovitz, 1992.

Content: Hyladelphys currently contains only the type species.

Etymology: From (wood or forest), sometimes used in its Latinized form (hylaea) in reference to the predominant vegetation of the Amazonian lowlands (e.g., by Ducke and Black, 1953); and (womb), a traditional suffix for New World marsupial genera.

Hyladelphys kalinowskii (Hershkovitz) Figures 11C, 11D, 12B, 15A, 16A, 17A, 17B, 18A, 18B, 19B

Three specimens from Paracou (AMNH 267338, 267339; MNHN 1995.903) and one from Guyana (kindly loaned to us for identification by M. D. Engstrom) represent extraordinary range extensions of this distinctive species, originally described as Gracilinanus kalinowskii by Hershkovitz (1992) on the basis of two Peruvian specimens. We provide an emended description with comparisons and remarks based on this new material and our reexamination of Hershkovitz's type series.

Type Material: The holotype, an adult female skin and skull (FMNH 89991), collected on 9 July 1958 by Celestino Kalinowski at Hacienda Cadena (890 m elevation), Marcapata, Departamento Cuzco, Peru; and the paratype, also an adult female skin and skull (FMNH 65754), collected in October 1948 by J. M. Schunke at Chanchamayo (1100 m elevation), Departamento Junín, Peru.

Distribution: Known only from Amazonian Peru, southern Guyana, and northern French Guiana (fig. 14).

Emended Description: Very small in all external and craniodental dimensions (table 4). Body pelage smooth (not woolly); dorsal fur unpatterned dull reddish brown, but basal two-thirds of hairs dark gray; ventral fur pure white or cream from chin to groin (the hairs self-colored, not grayish basally). Face boldly marked by broad mask of black fur extending from mystacial pad to base of ear on each side, and by prominent median streak of very pale orange fur extending from between eyes to rhinarium; cheeks (below mask) white like throat. Facial vibrissae (including long superciliary, mystacial, and genal hairs) mostly black (but a few of the ventralmost genals are white); vibrissae of chin and throat (submental and interramal hairs) white. Ears very large (covering eyes and extending to mystacial pads when laid forward over face), apparently naked (a sparse pelage of tiny hairs is visible only under magnification), and paper-thin; opaque and bright yellow-orange basally in fresh specimens (this color fading to white after a few months in alcohol), translucent and brownish distally. Eyes very large. Throat glands apparently absent.4

Wrists, ankles, and dorsal surface of metapodials covered with short, orange fur; manus and pes each with six plantar pads (thenar, hypothenar, and four interdigitals), the thenar and first interdigital touching but not fused on either manus or pes; central palmar surface smooth (not densely tubercular); claws of manual digits II–V large, extending beyond fleshy apical pads. Scrotum (of one adult male, AMNH 267338) pigmented, blue. Mammae 2–0–2 = 4, in two abdominal-inguinal pairs (without an unpaired median teat; fig. 15).5

Tail not very long and without furry base; dark (brownish) above and below, but indistinctly banded by absence of pigment over vertebral articulations. Caudal epidermis scaly beneath sparse covering of fine hairs; caudal scales in annular or spiral series, numbering about 25–35 rows/cm at middle of tail (counts from four fluid specimens). Median hair of triplet emerging from posterior margin of each scale usually about three scale rows long, thicker than lateral hairs but not grossly flattened or petiolate.

Skull (figs. 11, 12) with distinctively short, blunt rostrum; beaded and anteriorly convergent supraorbital margins; no postorbital frontal processes; very large orbits (their posterior limits indicated by weakly developed postorbital processes of the jugals); and laterally inflated braincase. Premaxillae without rostral process anterior to incisors; maxillary-premaxillary suture not extending to posteriormost incisor alveolus; maxillopalatine vacuities narrow, discontinuous in some specimens; posteromedial (palatine) and maxillary vacuities absent.6 Alisphenoid wing of auditory bulla without a well-developed anteromedial process bridging the foramen ovale (see Remarks, below).

Upper incisors 2–5 subequal (not increasing in size from front to back), asymmetrical (the crowns with anterior but no posterior styles), and nonoverlapping (fig. 16); upper and lower canines without accessory cusps; lower canine erect and not premolariform; deciduous third premolars (dP3/dp3) extremely reduced, not molariform (figs. 17, 18); permanent upper and lower third premolars (P3/p3) much smaller than second premolars (P2/p2); molars not highly carnassialized (metacone larger than paracone on M1 and M2, but metacone and paracone subequal on M3).

Variation: The four fluid-preserved specimens (with extracted skulls) from Guyana and French Guiana agree in all qualitative characters with the Peruvian holotype despite several inconsistencies between Hershkovitz's (1992) original description and that given above (see footnotes 4–6). The Peruvian specimens are slightly larger than French Guianan material (table 4), but the apparent size difference (perhaps best indexed by molar measurements) is no more than might be expected of conspecific samples collected almost 3000 km apart. The alisphenoid bullae of the holotype and of all four specimens from the Guianas lack anteromedial processes, but weakly developed processes (which are not fused to the floor of the braincase as they normally are in Gracilinanus, Marmosops, and Thylamys) are present on the anteromedial aspect of both alisphenoid bullae of the paratype. We interpret this as intraspecific variation because the paratype is not morphologically remarkable in any other respect.

Comparisons: Hyladelphys kalinowskii appears to differ from all other Recent didelphids in at least two qualitative characters. As far as is known, well-preserved adult females of other didelphid taxa uniformly exhibit odd-numbered mammary counts due to the presence of an unpaired median teat (fig. 15B); by contrast, caenolestids, microbiotheriids, and Old World marsupials have even-numbered mammary counts because their teats are bilaterally paired (Bresslau, 1920; Osgood, 1921, 1943; Tate, 1933, 1947, 1948). All reports of even-numbered didelphid mammary counts that we investigated were either field observations of lactating teats only (e.g., Hershkovitz, 1997: table 4), or were observations from specimens with midventral incisions that might have destroyed the median nipple (e.g., Marshall's [1978a] count of four mammae from an FMNH skin of Glironia venusta; W. T. Stanley, personal commun.). By contrast, our observations of identical (2–0–2 = 4) mammary configurations in two Hyladelphys skins from Peru and two fluid-preserved specimens from French Guiana suggest that these are not rare variants or preservational artifacts. Instead, this taxon appears to be genuinely divergent from other didelphids in a character that displays considerable higher-taxonomic stability among marsupials.

Hyladelphys also differs markedly from other didelphids in the morphology of its deciduous dentition. Whereas previous descriptions of didelphid milk premolars have consistently reported these teeth as large and molariform (Flower, 1867; Thomas, 1888; Bensley, 1903; Tate, 1948; Archer, 1976b), there is in fact some substantial variation in the size and morphology of dP3/dp3 within the family (table 5). However, Hyladelphys is a conspicuous outlier: its milk teeth are much smaller than those of other confamilials, and they are uniquely nonmolariform in occlusal structure (figs. 17, 18). Instead, dP3/dp3 in Hyladelphys fall well within the range of milk-premolar morphologies seen in some Old World marsupial groups (Tate, 1947, 1948; Archer, 1976b; see illustrations in Luckett, 1993, 1994).

From superficially similar “marmosines”, Hyladelphys differs in additional characters. Species of Gracilinanus, among which H. kalinowskii was previously classified, differ by having much smaller caudal scales (> 40 rows/cm), longer and narrower rostrums, less pronounced interorbital constrictions, smaller orbits, less inflated braincases, much more highly fenestrated palates, secondary foramina ovale formed by anteromedial struts of the alisphenoid tympanic wing, more carnassialized molars, and premolariform lower canines. Individual species of Gracilinanus differ from H. kalinowskii in other respects, but only G. emiliae merits explicit comparison here.

Gracilinanus emiliae occurs sympatrically with H. kalinowskii at Paracou, and because these species are similar in size and coloration (both have reddish dorsal fur and self-colored whitish venters), they might be confused in the field. However, numerous external characters permit unambiguous discrimination. Gracilinanus emiliae differs from H. kalinowskii in facial markings (its black facial mask does not extend to the base of the ear, and the streak of orange fur between its eyes is darker and narrower) and has smaller ears, smaller manual claws (not extending beyond the fleshy digital pads), 4–1–4 = 9 mammae, and a much longer tail with smaller scales and no trace of lighter banding over the vertebral articulations. Another external character that might distinguish G. emiliae and H. kalinowskii is the presence/absence of gular glands, but the variability of this trait within species is not well documented and, with so few examples available for comparison, its diagnostic value is uncertain. The skull of G. emiliae differs from that of H. kalinowskii in the cranial characters listed above for Gracilinanus and by the presence of a posterior accessory cusp on the upper canine (a trait not consistently exhibited by other congeners).

Members of other “marmosine” genera differ consistently from Hyladelphys by their larger size and size-correlated proportions (relatively longer rostrums, smaller orbits, less inflated braincases) and in the following qualitative contrasts (in addition to the mammary and milk-dentitional traits mentioned previously): incrassate tails (Thylamys), grossly enlarged central hairs of each caudal-scale triplet (Marmosops), distinct postorbital frontal processes (Marmosa, Micoureus), secondary foramina ovale (Marmosops, Thylamys), I2–5 conspicuously increasing in size from front to back (Marmosops and some Marmosa), P3/p3 larger than or subequal to P2/p2 (Thylamys), more carnassialized molars (all genera), and premolariform c1 (Marmosops). Insofar as can be inferred from polyprotodont character polarities (e.g., as hypothesized by Archer, 1976a, 1976b; Creighton, 1984; Reig et al., 1987; Wroe, 1997), Hyladelphys shows no clear pattern of synapomorphic resemblances with any other didelphid taxon. By the same token, evidence of a closer relationship to nondidelphid clades is also weak. In effect, our assignment of Hyladelphys to the family Didelphidae is based primarily on zoogeography and on morphological traits that are currently interpreted as marsupial plesiomorphies.

Remarks: A published portrait of the head of Hyladelphys kalinowskii reconstructed from dried skins (Hershkovitz, 1992: fig. 14) is misleading in several details. The bulging eyes of our fluid-preserved specimens are proportionately about twice as large as those in the drawing. The pinnae in the illustration are shown bristling with short hairs, but the auricular pelage is actually microscopic and the unmagnified ears appear quite naked; the ears in life are also proportionately much larger than drawn. The portrait does not show the genal vibrissae, but these long black hairs are conspicuous against the short white fur of the cheeks in all of the specimens at hand. The mystacial vibrissae are depicted as fine, inconspicuous hairs that extend only to the outer canthus of the eye, but these robust whiskers actually extend nearly to the tips of the pinnae when laid back along the side of the head. The facial markings in the portrait also lack the vivid definition characteristic of this species: the mask is intensely black in fresh specimens and is boldly accentuated by a broad streak of very pale orange fur down the midline of the rostrum.

We examined the western Ecuadorean specimen (KU 135097) that Hershkovitz (1992: 42) identified as “Gracilinanus sp. (new species)” and that he subsequently (op. cit.: 45) conjectured “… is most nearly like adult Gracilinanus kalinowskii …”. The animal in question is a juvenile male (not a female as originally reported) preserved in fluid with an extracted skull, of which the first upper molar (crown length = 2.26 mm) suggests an adult size far larger than that of Hyladelphys or any known species of Gracilinanus. The large size of its caudal scales, absence of posteromedial palatal vacuities, and absence of an anteromedial alisphenoid strut bridging the foramen ovale are additional traits that cannot be reconciled with Gardner and Creighton's (1989) diagnosis of Gracilinanus. From Hyladelphys, KU 135097 differs conspicuously by its well-developed rostral process of the premaxillae, well-developed postorbital jugal process, upper incisor morphology (I1–5 have large, overlapping, symmetrically rhomboidal crowns that increase in size from front to back), fully molariform dP3, and highly carnassialized molars. Based on these and other attributes, we refer this specimen to the genus Marmosa (sensu Gardner and Creighton, 1989), within which it most closely resembles M. mimetra Thomas (1921) and other nominal taxa currently synonymized (Gardner, 1993) with M. robinsoni.

Other Specimens Examined: Guyana—East Berbice-Corentyne, New River Falls (ROM 34271). Peru—Cuzco, Hacienda Cadena (FMNH 8991 [type]); Junín, Chanchamayo (FMNH 65754 [paratype]); Loreto, Río Gálvez, Nuevo San Juan (MUSM 11031).

Field Observations: Our first example of this species from Paracou (MNHN 1995.903) was shot as it perched on a palm frond about 1 m above the ground in swampy primary forest at 18:35 hours on 25 October 1992. The other two specimens (AMNH 267338, 267339) were taken from the same pitfall trapline, near a small stream in well-drained primary forest, on 21 August 1993.

Marmosa murina (Linnaeus) Figures 7, 19C, 21B

Voucher Material: AMNH 266416, 266417, 267368, 267816; MNHN 1995.904. Total = 5 specimens.

Identification: Our vouchers conform closely in all essential details of external and craniodental morphology with Tate's (1933) and Husson's (1978) authoritative descriptions of this species, the type locality of which was restricted by Thomas (1911a) to Surinam. According to Tate (1933: 95), two subspecies are parapatrically distributed in Surinam and French Guiana: supposedly, Marmosa murina murina ranges “along the narrow coastal strip between the sea and the heavy rainforest”, whereas “in the rainforest it is replaced by the smaller darker [M. m.] muscula [(Cabanis)].” Husson (1978), however, recognized only the former taxon in his Surinamese material, and all of the Surinamese specimens that we examined appear to represent a single recognizable form—closely conforming to Husson's description and resembling our Paracou vouchers in qualitative characters and measurements (table 6)—that we assume to represent typical M. murina. Unfortunately, possible character differences between the types of murina and muscula are now difficult to evaluate.

Thomas (1892) identified two syntypes of Didelphys murina Linnaeus (1758) among the specimens of the Lidth de Jeude collection purchased by the BMNH in 1867. One of these (BMNH 67.4.12.542) was designated as the lectotype by Husson (1978), but his choice was unfortunate: the lectotype is a fluid-preserved adult female from which the skull has been extracted and apparently lost. The skull of the adult male paralectotype (BMNH 67.4.12.541), however, has measurements that fall within the range of variation that we observed among recently collected specimens from Surinam (table 6). Tate (1933) described the pelage color of BMNH 67.4.12.541 in considerable detail, but neither this specimen nor the lectotype are really suitable for subspecific color comparisons because their pigments may have faded after more than two centuries in alcohol.

The holotype of Marmosa murina muscula (type locality: “Caraiben Niederlassung Arrai am obern Pomeroon” [Cabanis, 1848: 778], Pomeroon-Supenaam, Guyana) is a juvenile specimen (with dP3 in place and M4 unerupted) that consists of the skin and skull of an animal originally preserved in fluid. Because of its immaturity, and because the pelage is now faded from preservative and discolored by age, the type of muscula is likewise unsuitable for subspecific comparisons. A small series of skins from Kartabo, Guyana, that Tate (1933) identified as M. m. muscula, however, are dorsally somewhat darker than our Paracou vouchers, but have whiter venters with less extensive lateral zones of gray-based fur. Whereas the dorsal color difference is consistent with Tate's diagnosis of muscula versus murina, the ventral color difference is not. The only qualitative cranial character cited by Tate as diagnostic of muscula, the absence of dorsal grooves along the supraorbital ridges, is apparently useless for defining this taxon (as represented by Tate's own identifications) inasmuch as all the fully adult skulls from the Kartabo series (AMNH 42908, 48135, 99983, 142807) have grooved supraorbital ridges. Given that (1) the other character differences between murina and muscula cited by Tate are indefinite, (2) that the relevant types have discolored pelage and are incommensurate in age, and (3) that the material we examined from Surinam and French Guiana shows no appreciable divergence in size or coat color between coastal and interior populations, it does not seem useful to recognize these subspecies as valid at the present time.

We also examined the types of other nominal taxa from the Guiana subregion of Amazonia currently treated as subjective synonyms of Marmosa murina, including klagesi Allen (1900), chloe Thomas (1907), roraimae Tate (1931), and duidae Tate (1931). All of these are chiefly distinguishable by pelage characters—subtle differences in fur color, length, and texture—that perhaps vary clinally with environmental conditions as suggested by Tate himself (1933, 1939), who ranked them as no more than subspecifically distinct. By contrast, the conspecificity of some taxa from outside the Guiana subregion that are currently referred to M. murina (e.g., by Gardner, 1993) is more problematic. For example, M. quichua Thomas (1899a) from western Amazonia is craniodentally distinctive (Tate, 1933; personal obs.) and no justification for its synonymy with M. murina has apparently been published. In view of the lack of any critical review of the species-level taxonomy of Marmosa (sensu stricto) since Tate's 1933 monograph, that work should still be considered the primary authority for species limits until compelling evidence is provided for alternative classifications.

Other Specimens Examined: French Guiana—Arataye (MNHN 1981.172, 1981.173, 1982.597, 1986.125), Cayenne (MNHN 1986.1024, 1986.1025), Les Nouragues (MNHN 1998.307), Piste St.-Élie (MNHN 1981.417–1981.419, 1981.421, 1981.422, 1982.598), Saül (MNHN 1982.596, 1986.484). Guyana—“Demerara River 29 miles above Georgetown” (BMNH 7.6.20.16 [holotype of chloe]); Cuyuni-Mazaruni, Kartabo (AMNH 42907, 42908, 48135, 99983, 142807); Pomeroon-Supenaam, “Caraiben-Niederlassung Arrai am obern Pomeroon” (ZMB 2331 [holotype of muscula]). Surinam—Brokopondo, Finisanti (FMNH 95315–95319, 95321–95326, 95328); Marowijne, Oelemarie (CM 76729); Para, Zanderij (CM 68346, 68353); Saramacca, Raleigh Falls (CM 68354, 68355, 68356); Suriname, Lelydorpplan (FMNH 95329, 95330, 95332). Venezuela—Amazonas, Mt. Duida (AMNH 76984 [holotype of duidae]); Bolívar, Ciudad Bolívar (AMNH 16121 [holotype of klagesi]), Arabupu (AMNH 75703 [holotype of roraimae]). Without locality data—(BMNH 67.4.12.541 [paralectotype of murina], 67.4.12.542 [lectotype of murina]).

Field Observations: All of our records of Marmosa murina at Paracou are based on specimens collected in secondary growth and other manmade habitats: two were shot at night as they perched 1–2 m above the ground in dense roadside vegetation, one was caught by hand at night in a garden, and two were found killed and partially eaten (probably by domestic cats) near houses in clearings.

Marmosops Matschie

Thirty-five voucher specimens from Paracou are referable to Marmosops parvidens in the sense of Pine (1981), but close examination of pelage characters and correlated morphometric variation in our sample indicates that two species are present. The following paragraphs summarize the evidence supporting this conclusion.

All of the specimens in question are small opossums (21–33 g adult weight) with dark facial masks, dull reddish-brown or grayish-brown dorsal fur, small fore- and hindfeet (each with diminutive claws and six separate plantar tubercles), and long tails. The scrotal sacs of males are entirely white. Female specimens lack any trace of a pouch and appear to have have 3–1–3 = 7 or 4–1–4 = 9 inguinal-abdominal mammae (fig. 15B). Other distinctive attributes include: the grossly enlarged central hair of each caudal-scale triplet; a prominent, spoon-shaped, fleshy tubercle supported internally by bone on the lateral aspect of the wrist of males (fig. 20); smoothly rounded supraorbital margins without distinct beads or processes (fig. 19A); conical alisphenoid bullae with prominent anteromedial processes (fig. 21); absence of posteromedial (palatine) palatal vacuities (fig. 22); upper canines with distinct anterior and posterior accessory cusps; and lower canines that resemble the lower premolars in shape and size, forming a more-or-less undifferentiated series of four subequal teeth (fig. 23). Some of these traits are common to many “marmosines”, others are perhaps diagnostic of the genus Marmosops, and a few may define a distinctive group of species closely related to M. parvidens. For present purposes, however, these attributes serve to distinguish examples of parvidens-like Marmosops from all of the other marsupial taxa that we collected at Paracou.

Within this series, the most conspicuous external variation involves fur color. We scored three pelage characters for statistical analysis. Dorsal coloration was classified as “fuscous” (dusky grayish-brown) or “reddish” (a subtly warmer tone) according to the predominant hue of the unruffled fur. Although this chromatic contrast disappears in material stored for years in alcohol, dried skins and fresh fluid specimens (recently removed from formalin) were readily assigned to one or the other of these two states.7 Ventral coloration was classified as “white” if the self-colored fur and the tips of gray-based hairs lacked any pigmentation, or “cream” if the self-colored fur and gray-based hair tips were pale yellowish. Ventral pattern was classed as “narrow” if self-colored (pure white or cream) fur was confined to the midline (sometimes as a discontinuous streak) by a broad zone of gray-based fur on each side, “broad” if almost the entire ventrum was self-colored, or “intermediate” for specimens with substantial amounts of both self-colored and gray-based ventral fur.

Chi-squared tests of independence provide no evidence for age or sex effects on the expression of these pelage traits. In fact, the largest age-sex class in our sample (16 adult males) includes individuals with fuscous and reddish dorsal fur, specimens with white and cream-colored ventral fur, and examples of all three conditions of ventral pattern. However, pelage characters are not independently distributed inter se (table 7). Most animals with fuscous dorsal fur have self-colored ventral fur narrowly confined to the midline or bordered by extensive lateral zones of gray-based fur; by contrast, many animals with reddish dorsal fur have almost entirely self-colored ventral fur, and none has self-colored fur narrowly confined to the midline. Similarly, almost all animals with fuscous dorsal fur have white (or white-tipped) ventral fur, whereas about one-third of the specimens with reddish dorsal fur have cream (or cream-tipped) underparts.

We used one-way ANOVAs to test for morphometric divergence between individuals with red and fuscous dorsal fur among adult males (the only age-sex class in our sample large enough for such analyses) and found highly significant differences (p < 0.01) in length of molars, palatal breadth, zygomatic breadth, height of the upper canine (measured from the posterior accessory cusp to the tip of unworn teeth), and nasal breadth. Other external and craniodental measurements showed no significant divergence. Not surprisingly, principal components analysis of the log-transformed craniodental measurement data provides clear separation of dorsal fur color classes aligned with the first eigenvector (fig. 24), the coefficients of which (table 8) reflect essentially the same differences as those indicated by univariate statistics: fuscous adult males have taller canines, broader nasals, and wider zygomas than do reddish adult males, with smaller (but still substantial) differences in molar size and palatal breadth. Similar morphometric contrasts between dorsal fur color classes can be seen among the few adult females in our sample, and among immatures.

Close inspection of skulls revealed another difference between the two groups of specimens previously sorted by chromatic and morphometric traits. In specimens with fuscous dorsal fur, taller canines, and broader skulls, the lacrimal bone forms a prominent part of the anteroventral margin of the orbit; the lacrimal foramina then perforate the orbital margin, where they are exposed in lateral view (fig. 25B). By contrast, the lacrimal bone is not a prominent part of the anteroventral orbital margin in specimens with reddish dorsal fur, smaller teeth, and narrower skulls; in these, the lacrimal foramina are always located within the orbit, where they are more-or-less concealed from lateral view (fig. 25A).

We interpret the correlation between pelage characters, morphometric differences, and lacrimal morphology described above as evidence that two species of Marmosops are sympatric at Paracou. Based on our examination of relevant type material (see below), we identify the reddish individuals with shorter canines, narrower skulls, and reduced lacrimals as M. parvidens, and the fuscous individuals with taller canines, broader skulls, and prominent lacrimals as M. pinheiroi. A single Paracou specimen (AMNH 267358), consisting of the skeleton only of a juvenile animal with a smashed skull, is not assignable with certainty to either species.

Marmosops parvidens (Tate) Figures 15B, 22A, 22B, 23A, 25A, 26A

Voucher Material: AMNH 266425, 266426, 267344, 267347, 267348, 267350, 267353, 267359, 267361, 267817; MNHN 1995.927–1995.930, 1995.933. Total = 15 specimens.

Identification: Specimens of Marmosops parvidens can be distinguished from sympatric examples of M. pinheiroi at Paracou by their warmer (more reddish) dorsal fur, broader extent of self-colored fur (which is never discontinuous between chin and anus), smaller teeth (especially upper canines), narrower skull, and by the reduced lacrimal contribution to the anteroventral orbital margin (see above). Because the difference in dorsal pelage color is subtle, and because there is some overlapping variation in the extent of self-colored ventral fur, unambiguous species identifications require cleaned cranial material. In our voucher series, height of canine (HC) affords the clearest discrimination (tables 9, 10), but toothwear and sexual dimorphism must be taken into account in sorting specimens by this criterion. Lacrimal morphology (fig. 25) is perhaps the most reliable cranial character for identifying juveniles with incompletely erupted (and therefore unmeasurable) canines.

Although a revision of what may be called the parvidens group of Marmosops is beyond the scope of this faunal report, we note that our conclusion that two species assignable to this complex are sympatric at Paracou is consistent with recent molecular evidence that M. parvidens sensu Pine (1981) is composite (Mustrangi and Patton, 1997; Patton et al., 2000). In addition to recognizing M. pinheiroi as a valid species, we note that M. juninensis (another “subspecies” of M. parvidens sensu Pine, 1981) is equally distinctive and also merits specific recognition. Originally described by Tate (1931) on the basis of a single specimen (AMNH 63864), M. juninensis (now additionally represented by AMNH 230014–230016, all collected near Tarma in Depto. Junín, Peru) can be unambiguously distinguished from M. parvidens (sensu stricto) and M. pinheiroi by the size and shape of the male carpal tubercle (smaller and not spoon-shaped), by the consistent presence of posteromedial palatal vacuities, and by the absence of distinct accessory cusps on the upper canine (fig. 26). Like parvidens, but unlike pinheiroi, the lacrimal bone in juninensis does not form part of the anteroventral orbital margin, so the lacrimal foramen lies inside the orbit. Unlike any specimens of parvidens, however, the ventral pelage of juninensis is entirely gray-based.

We examined the type of Marmosops parvidens bishopi (USNM 393535), the pelage of which Pine (1981) described as colored essentially like that of M. p. parvidens, but paler. In our opinion, this specimen represents another distinct species that differs from M. parvidens in lacking any trace of an anterior accessory cusp on the upper canine. Specimens from Bolivia (e.g., AMNH 268938) and Peru (e.g., AMNH 67243) with darker (more saturated) dorsal fur than the type may nevertheless be provisionally referred to M. bishopi based on upper canine morphology.

A specimen from northern Venezuela (USNM 371299) that Pine (1981) referred to Marmosops parvidens parvidens lacks an anterior accessory upper canine cusp (like M. bishopi) and exhibits other differences from typical examples of M. parvidens as recognized in this report. The mystacial vibrissae of USNM 371299 appear to be very short, probably not extending much if at all beyond the base of the ear in life, whereas the mystacial hairs reach at least to the posterior margins of the pinnae in typical examples of M. parvidens. The dorsal fur of USNM 371299 is longer than in typical M. parvidens (about 9 mm middorsally versus about 7 mm), is fluffier in texture, and has a faintly marbled appearance that is not characteristic of other specimens that we refer to this species. Taken together, these differences suggest that USNM 371299 may represent an undescribed taxon, but more material should be examined to evaluate this conjecture.

With a single exception (AMNH 97333, see below), all of the specimens that we identify as Marmosops parvidens sensu stricto are from the Guiana subregion of Amazonia (fig. 27).

Other Specimens Examined: Brazil—Amazonas, Boca Rio Piratucu (AMNH 93970), 80 km N Manaus (USNM 579985–579990); Pará, Ilha do Taiuna on lower Rio Tocantins (AMNH 97333). French Guiana—Arataye (USNM 548439). Guyana—Demerara-Mahaica, Hyde Park (FMNH 18545 [holotype]); Upper Takutu-Upper Essequibo, Karanambo (ROM 97938).

Field Observations: Because unvouchered sightings of Marmosops could not be unambiguously identified to species, all of our definite records of M. parvidens at Paracou are from collected specimens. Five specimens of M. parvidens were shot, one was caught by hand, one was caught in a Victor trap, one in a Sherman trap, and the rest (seven) were captured in pitfalls; all were collected at night. Nine specimens were taken on the ground, five were found perching in understory vegetation (usually on vertical stems) 0.2–1.5 m above the ground, and one was trapped on a liana 1.8 m above the ground. Seven specimens were collected in well-drained primary forest, five in creekside primary forest, one in swampy primary forest, one in primary forest of unspecified character, and one in secondary growth.

See the following account for habitat comparisons between Marmosops parvidens and M. pinheiroi.

Marmosops pinheiroi (Pine) Figures 16B, 19A, 20, 21A, 22C, 22D, 23B, 25B

Voucher Material: AMNH 266423, 266424, 267007, 267008, 267341–267343, 267345, 267346, 267349, 267351, 267352, 267354, 267357; MNHN 1995.925, 1995.926, 1995.931, 1995.932, 1995.934. Total = 19 specimens.

Identification: See the preceding account for diagnostic comparisons with Marmosops parvidens.

In addition to the holotype and paratypes of Marmosops pinheiroi, we examined the type series of M. parvidens woodalli (USNM 393529–393532, 393534, 545543), a subspecies that Pine (1981) described from the vicinity of Belém, Brazil. Although these specimens average paler dorsally than examples of M. pinheiroi from north of the Amazon, they are otherwise similar in pelage and craniodental characters, and we provisionally regard them as conspecific. An adult male specimen from the right bank of the lower Rio Xingu (USNM 549294), however, may represent an undescribed taxon. Although most of its qualitative traits match those of M. pinheiroi, it is much paler dorsally and has substantially smaller upper canines (HC = 1.06 mm) for its sex than any example of that species as recognized by us.

Other Specimens Examined: Brazil—Amapá, Serra do Navio (USNM 461459 [holotype], 461460, 461462–461465); Pará, Belém (USNM 545543), Utinga (USNM 393529–393532, 393534). Guyana—Potaro-Siparuni, Iwokrama Reserve (ROM 108920). Venezuela—Bolívar, Auyan-tepui (AMNH 130521, 130568, 130570), Churi-tepui (AMNH 176352, 176353), 85 km SE El Dorado (USNM 385046).

Field Observations: All of our definite records of Marmosops pinheiroi at Paracou are from collected specimens. Four specimens were shot, three were captured in Victor traps, and the remainder (12 specimens) were caught in pitfalls; all were taken at night. Thirteen specimens were trapped or shot on the ground (one in a hollow log), but six were taken 0.3–1.5 m above the ground on vertical stems and lianas. Four specimens were collected in well-drained primary forest, five in creekside primary forest, six in swampy primary forest, and four in secondary growth.

Marmosops parvidens and M. pinheiroi clearly overlap in habitats at Paracou. Although our ecological classification is coarse and doubtless obscures many subtle differences among capture sites with the same descriptor (e.g., “well-drained primary forest”), we sometimes took both species in close proximity. One line of pitfall traps that traversed 50 m of apparently homogeneous primary forest along a small stream, for example, captured three parvidens and four pinheiroi between 29 July and 13 August 1993. Nevertheless, it is noteworthy that whereas we rarely took parvidens in swampy forest or secondary growth, over half of our pinheiroi specimens were collected in those habitats. Both species appear to occur only in the forest understory: no Marmosops were sighted or trapped at heights greater than about 2 m above the ground.

None of the adult female Marmosops we collected were carrying suckling young.

Metachirus nudicaudatus (E. Geoffroy)

Voucher Material: AMNH 266435, 266439, 266440, 266449, 266450, 266452, 266453, 266455, 267009, 267010, 267362, 267365; MNHN 1995.905–1995.910. Total = 18 (not including suckling young).

Identification: Our voucher material is almost topotypical of this species, which was originally described from a specimen collected at Cayenne (Julien-Laferrière, 1994). The Paracou series agrees closely in qualitative external characters with the description given by Husson (1978), and most craniodental measurements of the type fall within the range of metric variation in our voucher collection (table 11).

Although Metachirus has long been thought to contain but a single valid species (Tate, 1939; Cabrera, 1958; Gardner, 1993), this historical consensus is challenged by recently analyzed mtDNA sequences that suggest deep evolutionary divergence among samples from different Amazonian subregions (Patton et al., 2000). In the absence of any revisionary analysis of morphological specimens, however, it is unclear how mitochondrial haplotypes might correspond with named taxa. Inevitably, nominotypical material from French Guiana will play a pivotal role in any future attempt to resolve this unsatisfactory state of affairs.

Remarks: We agree with Julien-Laferrière (1994) that the name Didelphis nudicaudata is available from Geoffroy's (1803) catalog for the reasons explained by Hershkovitz (1955) and Holthuis (1963).

Other Specimens Examined: French Guiana—Cayenne (MNHN 1990.420 [holotype]).

Field Observations: We recorded 22 observations of Metachirus nudicaudatus at Paracou, of which 18 are based on collected specimens and 4 are unvouchered sightings. Eighteen records (82%) are of animals shot or sighted on the ground, but one specimen (4%) was trapped on the ground in a Tomahawk live trap, and three specimens (14%, all juveniles) were taken in Victor snap-traps tied to lianas 0.5–1.3 m above the ground. All of our records are from animals shot, sighted, or trapped at night. Fourteen individuals (64%) were shot, sighted, or trapped in well-drained primary forest, but one (4%) was encountered in swampy primary forest, four (18%) in primary forest of unspecified character, and three (14%) in secondary vegetation. With the exception of females with nursing young, all shot, sighted, or trapped animals were solitary.

One female shot on 7 July 1991 had seven nursing young measuring 19 mm crown-rump, and another shot on 17 August 1991 had eight nursing young measuring 29 mm crown-rump.

Micoureus demerarae (Thomas)

Voucher Material: AMNH 266428, 266429, 266431–266434, 267370, 267371, 267818; MNHN 1995.911–1995.914. Total = 13 specimens.

Identification: Although our voucher material corresponds closely to Tate's (1933) and Husson's (1978) descriptions of this taxon, the coloration of the ventral pelage and of the tail are variable among Paracou specimens and merits comment. Most of the ventral surface is covered by gray-based fur that is heavily washed with buff, but self-colored (pure buff) hairs cover the groin and throat, and a few specimens have narrow streaks of pure buff fur along the midline of the chest or abdomen. The tails of most Paracou specimens have white tips or are mottled with large white spots distally, but two adults (AMNH 267370, 267818) have entirely dark tails.

The holotype (BMNH 5.11.1.25) and other adult specimens from Guyana that we measured for comparison exhibit broad morphometric overlap with our vouchers (table 12). Additionally, fresh Guyanese skins (e.g., ROM 103370, 104708) are indistinguishable in coloration from our Paracou material. Thus, although several valid taxa may eventually be recognized among the many names currently synonymized with Micoureus demerarae (see below), the Paracou population can be confidently assigned to this species, and to the nominate race if a trinomial nomenclature is warranted.

Juveniles of Micoureus demerarae somewhat resemble Marmosa murina in size and external appearance, and the two species might therefore be confused in the field, even with specimens in hand. Based on our material, the best external characters for discrimination are tail color (most, but not all, examples of M. demerarae have tails blotched or tipped with white, whereas M. murina has consistently all-dark tails), fur texture (longer and woolly in M. demerarae, close and smooth in M. murina), the extent of fur at the base of the tail (conspicuously greater in M. demerarae than in M. murina), and size of the manual claws (extending beyond the fleshy apical pads in M. demerarae but not in M. murina).

Remarks: Originally described as a subspecies of Marmosa cinerea by Thomas (1905), demerarae was treated as a distinct species of the cinerea group in Tate's (1933) monographic revision of Marmosa (sensu lato). Cabrera (1958), however, considered demerarae to be a subspecies of cinerea, citing doubts that Tate (1939: 164, footnote 2) expressed about his own prior classification. The current allocation of Tate's cinerea group to Micoureus follows Gardner and Creighton (1989). Among the many nominal taxa now synonymized with Micoureus demerarae (sensu Gardner, 1993) are several that Tate (1933) recognized as full species, all or some of which may yet prove to be valid (Patton et al., 2000). With the extensive series of specimens now available to evaluate the taxonomy of this widespread complex, the group is ripe for modern revisionary treatment.

Other Specimens Examined: Guyana—“R. Demerara” (BMNH 7.6.20.14); Cuyuni-Mazaruni, Kartabo (AMNH 42887, 64156); Potaro-Siparuni, 5 km SE Surama (ROM 103146), Iwokrama Reserve (ROM 104708); Upper Demerara-Berbice, Comackka on Demerara River (BMNH 5.11.1.25 [holotype]), Tropenbos (ROM 103370); Upper Takutu-Upper Essequibo, Achamere Wan (ROM 34514), Ireng Valley (BMNH 3.4.6.10), Kuitaro River 40 mi E Dadanawa (ROM 35453), Weri More (ROM 33201).

Field Observations: All of our definite records of Micoureus demerarae at Paracou are based on collected specimens; of these, eight were shot, three were caught in Victor traps, and two were caught in elevated platform traps. Most (12) of our specimens were taken 1–17 m above the ground on lianas or in trees, but one juvenile was found climbing among dead branches at ground level in a treefall. Five specimens were collected in well-drained primary forest, one in swampy primary forest, and seven in more-or-less disturbed habitats (roadside secondary growth and selectively logged forest).

Our single adult female specimen, collected on 12 August, was not carrying suckling young.

Monodelphis brevicaudata (Erxleben) Figures 29–31

Voucher Material: AMNH 267000. Total = 1 specimen.

Identification: The genus Monodelphis has never been revised, and many aspects of the currently accepted species-level taxonomy of these short-tailed opossums (summarized by Gardner, 1993) remain untested by substantive analyses of specimen data. To determine the correct identification of our single Paracou voucher, we examined comparative series, types, and original descriptions of all relevant taxa. Our resulting systematic conclusions broadly overlapped those of the late C. O. Handley, Jr., who generously shared with us the unpublished results of his previous research with many of the same specimens that we studied. His suggestions prompted us to reexamine some of our earlier ideas about character variation in the M. brevicaudata complex, and the following account therefore reflects his critical input.

Among the many named forms of red-flanked Monodelphis currently synonymized with M. brevicaudata (see Gardner, 1993) are several readily diagnosable taxa that we provisionally recognize as full species. As understood by us, M. brevicaudata is restricted to the Guiana subregion of Amazonia (fig. 28) and is distinguished from other species of the brevicaudata complex, all of which are allopatric, by the extension of body fur onto the proximal one-third or more of the caudal dorsum; the ventral surface of the tail is just furred at the base (fig. 29). By contrast, only the basal one-sixth or less of the tail is furred, to about the same extent above and below, in M. palliolata (which occurs west of the Orinoco in northern Venezuela and northeastern Colombia), M. glirina (south of the Amazon and west of the Xingu), and in an unnamed form (a subspecies of M. brevicaudata in the view of C. O. Handley, Jr., personal commun.) distributed south of the Amazon and east of the Xingu. In addition, whereas M. palliolata and M. glirina have orange ventral fur (not sharply differentiated from the color of the flanks) and a broad cap of grizzled-grayish fur extending across the crown of the head between the eyes, fully adult specimens of M. brevicaudata have whitish, cream, or buffy ventral fur (sharply differentiated from the reddish flanks) and a narrower cap of grizzled-grayish fur that (when present) is confined middorsally by a broad band of red above each eye. Monodelphis palliolata and M. glirina are externally similar, but differ in size (upper molar row ;le7.9 mm in palliolata, ;le7.9 mm in glirina). The unnamed form south of the Amazon and east of the Xingu resembles geographically adjacent populations of M. brevicaudata (from Amapá and the eastern Guianas) in size and coloration but lacks the dorsal extension of body fur onto the tail.

Thus restricted, Monodelphis brevicaudata includes the following nominal taxa: brevicaudata Erxleben (1777), brachyuros Schreber (1778), touan Shaw (1800), tricolor Geoffroy (1803), hunteri Waterhouse (1841), orinoci Thomas (1899b), and dorsalis Allen (1904). The material we examined exhibits considerable geographic variation in pelage color, the taxonomic significance of which is difficult to evaluate. Possibly, pelage color may reflect species-level divergence that is not apparent in craniodental characters, but the material currently available is insufficient to determine whether chromatic variation is clinal or discontinuous. The following observations are intended to establish which of the above names is applicable to the coat-color phenotype represented by our Paracou voucher, not to revise the nomenclature of the entire brevicaudata complex, a task beyond the scope of this report.

Erxleben's (1777) description of Didelphis brevicaudata and Schreber's (1778) description of Didelphys brachyuros were both based on Seba's (1734: 50) description and illustration (op. cit.: pl. xxxi, fig. 6) of ”Muris sylvestris Americani faemina”. Seba's original specimen (BMNH 67.4.12.540; see Thomas, 1892) is therefore the type of both brevicaudata and brachyuros, so these nominal taxa are objective synonyms (for the priority of Erxleben's name see Thomas, 1888: 356). We examined BMNH 67.4.12.540, an adult female preserved in fluid with an extracted skull. Despite more than two centuries in preservative, this specimen is in remarkably good condition and the apparently unfaded pelage is distinctly bicolored (reddish dorsally and abruptly paler ventrally), exactly as described by Seba, Erxleben, and Schreber; there is no middorsal stripe of grizzled-brownish, -grayish, or -blackish fur. Although Matschie (1916) restricted the type locality of Monodelphis brevicaudata to Surinam, bicolored specimens resembling the type are only known from the interfluvial region between the lower Caroni-Orinoco and the lower Mazaruni-Essequibo in northeastern Venezuela and northwestern Guyana (fig. 28). Because Matschie's restriction was obviously erroneous, we hereby emend the type locality to the vicinity of Kartabo, Cuyuni-Mazaruni District, Guyana (locality 15 in fig. 28), from which a well-preserved bicolored specimen (AMNH 48133) closely resembling the type was collected by William Beebe in 1919.8

By contrast with the limited geographic distribution of bicolored animals, tricolored specimens of Monodelphis brevicaudata have been collected throughout French Guiana, Surinam, Guyana, Guianan Venezuela (south and east of the Orinoco), and Guianan Brazil (north of the Amazon and east of the Rio Negro). In fully adult examples of this coat-color phenotype, a broad middorsal stripe of grizzled-brownish, -grayish, or -blackish fur is sharply set off from the clear (ungrizzled) reddish flanks, which are separated by a similarly abrupt transition from the pale (whitish, cream, or buffy, but sometimes partly gray-based) ventral fur. Whereas tricolored skins from Brazil and the Guianas are often brightly colored (with blackish or grayish middorsal stripes and cream or whitish venters), most tricolored Venezuelan skins have brownish middorsal stripes and buffy venters that exhibit less chromatic contrast with the reddish flanks. Despite Husson's (1978) remark that his Surinamese specimens represented a full range of intermediates between the bicolored and tricolored phenotypes, all of the Surinamese specimens we examined (including Husson's material, obtained on loan from the RMNH) are tricolored.

The oldest available name for any tricolored form of Monodelphis brevicaudata is Viverra touan Shaw (1800), which was based on Buffon's (1789) description of “Le Touan” from Cayenne.9 That the name touan properly applies to the tricolored phenotype is unambiguously supported both by Buffon's and Shaw's explicit mention of a blackish middorsal stripe extending from the rostrum to the base of the tail, and by the fact that only tricolored animals are known from French Guiana. The appropriate trinomial, if one is needed, for our Paracou voucher is therefore M. b. touan. Unfortunately, the application of the name touan has been a persistent source of confusion in the literature. Thomas (1888) and Cabrera (1919) regarded touan as a synonym of brevicaudata, but Cabrera (1958) listed touan (without comment) as a distinct species that included such divergent forms as emiliae, paulensis, and rubida as subspecies. Current usage (Gardner, 1993) recognizes M. emiliae, M. rubida, and M. sorex (including paulensis) as full species that can be distinguished from M. brevicaudata by trenchant craniodental characters (e.g., those described by Pine and Handley, 1984).

In a recent treatment of Venezuelan Monodelphis, Linares (1998) inexplicably reversed the application of touan and brevicaudata by assigning the former name to the bicolored phenotype and the latter name to tricolored animals (including M. palliolata). According to Linares, touan and brevicaudata are distinguished by craniodental characters in addition to pelage color pattern, and occur sympatrically in northeastern Venezuela, the Guianas, and in the Brazilian state of Pará. However, specimens referable to touan and brevicaudata are not craniodentally differentiated in our experience, nor have we seen evidence that two species assignable to the brevicaudata complex are sympatric anywhere. Linares also resurrected Cabrera's (1958) hypothesis that touan and emiliae are conspecific, but our observations support Pine and Handley's (1984) conclusion that emiliae is a distinctive species with no special similarity to touan or to other members of the brevicaudata complex.

In order to definitively resolve these conflicting usages, we hereby designate FMNH 21720 as the neotype of Viverra touan Shaw. The neotype, consisting of the skull (fig. 30) and the tricolored skin (fig. 31) of an adult male, was collected by S. Klages at Cayenne, French Guiana, on 26 February 1917.

Despite the conspicuous geographic variation in pelage color within Monodelphis brevicaudata noted above, we are not persuaded of the necessity for a formal trinomial nomenclature. Although the brevicaudata and touan phenotypes are clearly distinct in Guyana, some bicolored Venezuelan specimens from NE Bolívar (e.g., EBRG 17536, USNM 385004, 385005) have indistinctly grizzled middorsal fur, somewhat resembling the middorsal pigmentation of drab-tricolor skins from Amazonas and southeastern Bolívar, an observation that could be interpreted as evidence that bicolored and tricolored populations intergrade clinally in Guianan Venezuela. Furthermore, although our samples of measurable adults are too small for confident statistical inference (table 13), no morphometric differences are apparent between bicolored and tricolored animals to suggest that these are anything more than local coat-color variants. Larger series of specimens, especially from western Guyana and eastern Venezuela, together with molecular data would be helpful in any future effort to evaluate the taxonomic significance of pelage color variation in this species.

Bicolored Specimens Examined: Guyana—Cuyuni-Mazaruni, First Falls on the Cuyuni River (BMNH 34.6.30.65), Kartabo (AMNH 48133); Essequibo Islands-West Demerara, Buck Hall (BMNH 17.7.7.1). Venezuela—Bolívar, 65 km SSE El Dorado (USNM 385005), 56 km SE El Manteco (USNM 385004), Reserva Forestal Imataca (EBRG 17536). Without locality data—(BMNH 67.4.12.540 [holotype of brevicaudata], 88.1.31.1 [holotype of hunteri]).

Tricolored Specimens Examined: Brazil—Amapá, Serra do Navio (USNM 392050, 392051, 393424, 393428, 393430, 393435, 393436, 393438, 393439, 393441, 461434, 461435); Amazonas, Faro (AMNH 93972–93974, 94161, 94221), 80 km N Manaus (USNM 579976–579979), Santo Antonio da Amatary (AMNH 92879); Pará, Cachoeira Porteira (USNM 546209–546219), “Serra do Tumucumaque” (USNM 392044–392049). French Guiana—Arataye (USNM 578009), Cacao (MNHN 1981.168, 1981.412, 1981.414–1981.416, 1982.599), Cayenne (FMNH 21720 [neotype of touan], MNHN 1990.421 [holotype of tricolor]), Montjoly (MNHN 1994.122), St.-Eugène (MNHN 1995.205, 1995.3216), Sophie (MNHN 1966.1, 1966.2), Tamanoir (FMNH 21793), Trois Sauts (MNHN 1981.413). Guyana—“River Supinaam” (BMNH 10.9.29.26, 10.9.29.27); Cuyuni-Mazaruni, Bartica Grove (BMNH 10.11.10.20), Kamakusa (AMNH 140465, 140466); Potaro-Siparuni, Anundabaru (AMNH 75830, 75831), Minnehaha Creek (AMNH 36317), Potaro (BMNH 12.6.9.9); Upper Demerara-Berbice, Dubulay Ranch (AMNH 267744, 268060, 268061). Surinam—Brokopondo, Brownsberg (CM 52729, RMNH 23403, 23404), Finisanti (FMNH 95338); Marowijne, 3 km SW Albina (CM 76730), Langamankondre (RMNH 18227), 10 km N and 24 km W Moengo (CM 52730), Oelemarie (CM 76731, 76732), Paloemeu Airstrip (FMNH 94018, 94019); Nickerie, Avanavero (CM 68358), Kayserberg Airstrip (CM 68359), King Frederick William Falls (FMNH 48416); Saramacca, La Poule (FMNH 95339), Raleigh Falls (CM 63510, 63511, 68361); Suriname, Cultuurtuin near Paramaribo (RMNH 18076), Jarikaba near Uitkijk (RMNH 20672), Plantage De Morgenstond near Paramaribo (RMNH 17223), Plantation Clevia near Paramaribo (RMNH 21654). Venezuela—Amazonas, Boca Río Ocamo (AMNH 78093–78095), Esmeralda (AMNH 77281, 77282, 77287, 77288), Mt. Duida (AMNH 77283–77285, 77289, 77290–77296), “Río Casiquiare” (AMNH 77286, 78096–78100), Serra de Neblina (AMNH 244469); Bolívar, Arabupu (AMNH 75681–75687), Auyantepui (AMNH 130516, 130560–130565, 130573–130576, 130727), Caicara (BMNH 98.12.1.22 [holotype of orinoci]), Ciudad Bolívar (AMNH 16124–16126 [type series of dorsalis].

Field Observations: The single example of Monodelphis brevicaudata that we collected at Paracou is a partially eaten specimen caught in a Victor trap set under a fallen tree (fig. 32) in well-drained primary forest. Several additional specimens were previously taken by O. Henry in the course of his multiyear trapping study (G. Dubost, personal commun.), but we have not examined his material.

Philander opossum (Linnaeus)

Voucher Material: AMNH 266379–266381, 266383–266387, 266389–266391, 266394, 266395, 266398, 266400, 266994, 266995, 266997–266999, 267014, 267328; MNHN 1995.915–1995.924. Total = 32 specimens.

Identification: The species-level taxonomy of Philander was recently reviewed by Hershkovitz (1997) and by Patton and da Silva (1997), whose taxonomic conclusions differ considerably. Both studies, however, referred Guianan populations of gray four-eyed opossums to P. opossum, the type locality of which (as restricted by Matschie, 1916) is Paramaribo, Surinam. Our Paracou vouchers agree closely with Husson's (1978) description of topotypic specimens of P. opossum in most details, but a few discrepancies merit comment. (1) Whereas Husson described the pelage of his topotypes as usually distinctly darker middorsally than on the sides, only one skin from Paracou (AMNH 266995) has fur that is slightly darker middorsally than on the sides; the rest of the adult and subadult skins from Paracou are a uniform grizzled-gray over the entire dorsum. (2) Husson (1978: 25) described the fur of the head as “of the same dark blackish-brown colour as the median part of the back or even slightly darker”, but the blackish fur of the head contrasts conspicuously with the grizzled-gray back in all adult and subadult Paracou skins, even the aforementioned example with darker middorsal fur. (3) Husson (1978: 26) described the ears as “whitish with a broad black rim”, but the ears of our Paracou specimens are only whitish at the base—most of the pinna is black. (4) Husson's juvenile specimens were described as darker than adults and with less distinct facial markings, but three juvenile skins from Paracou (AMNH 266385, 266391, 266997) are comparable to those of adults and subadults except in pelage texture (the juvenile fur is softer).

Although most of the Surinamese material we borrowed for side-by-side comparison are darker middorsally than on the sides, as Husson described, a few (e.g., CM 52735) are uniformly gray like our Paracou vouchers. The other external differences implied by Husson's description are not apparent in the Surinamese material we examined.10 Cranial measurements of adult females from Paracou and Surinam overlap broadly (table 14), but our few adult males seem exceptionally large. We did not see any consistent qualitative craniodental differences between Surinamese and Paracou specimens.

Patton and da Silva (1997: 98) characterized the phenotype of Philander opossum as “uniformly pale-gray”, a description that better fits our Paracou material than it does the Surinamese topotypes we examined. However, Patton and da Silva's context for taxonomic comparisons included the almost entirely black P. mcilhennyi, as well as the black-striped P. andersoni, both of which contrast strikingly with predominantly grayish animals that those authors referred to P. opossum and P. frenata.

According to Hershkovitz (1997: 39), both gray and “brown” color phases occur throughout the range of P. opossum. We examined the Surinamese specimens that he cited as coat-color exemplars, however, and did not observe any consistent differences between gray and “brown” individuals (e.g., between FMNH 95312 and 95313, both collected by H. A. Beatty in the Wilhelmina Mountains). Yellowish or brownish tints do occur in some Philander skins, but whether these represent true coat-color variants rather than preservational artifacts (e.g., staining by sebaceous secretions or subcutaneous fat) is difficult to determine with the material at hand.

Both Patton and da Silva (1997) and Hershkovitz (1997) referred French Guianan populations of Philander opossum to the nominate form P. o. opossum, a usage consistent with the results of our comparisons. According to these authors, the nominate race occurs throughout the Guianas and the eastern Amazon basin of Brazil, and differs from other subspecies principally by geographic variation in size and pelage color. However, whereas Patton and da Silva recognized four species of Philander (see above), Hershkovitz recognized only P. opossum (including frenata among other subspecies) and P. andersoni (including mcilhennyi as a subspecies). We follow Patton and da Silva in recognizing the gray four-eyed opossum of the Brazilian Atlantic forest as a distinct species, P. frenata, and we interpret their phylogenetic analysis of mtDNA sequences as evidence that additional taxa currently treated as subspecies of P. opossum (e.g., canus and fuscogriseus) might also be recognized as full species. However, many nominal taxa were not represented in Patton and da Silva's molecular analyses, and several important issues of species-level synonymy remain unresolved. Therefore, the geographic limits of P. opossum are still uncertain.

Remarks: We agree with Hershkovitz (1976, 1981) that Philander, not Metachirops, is the correct name for pouched four-eyed opossums (contra Husson [1978] and other authors). Linnaeus's (1758) original description of Didelphis opossum was based on an adult male and an adult female described and figured by Seba (1734). Hershkovitz (1976: 297) designated Seba's female as the lectotype, but the specimen itself was apparently not then known to have survived the breakup and dispersion of Seba's museum (for the history of which, see Boeseman, 1970). Hershkovitz (1997) subsequently stated that the lectotype is still preserved as an alcoholic specimen in the Rijksmuseum van Naturlijke Historie in Leiden.11 According to the current RMNH curator of mammals (C. Smeenk, personal commun.), the lectotype is RMNH 25421, clearly recognizable as the adult female with three pouch young illustrated and described by Seba.

Other Specimens Examined: Surinam—Coronie, Totness (CM 52731, 52733); Marowijne, Moengo (CM 52734), Perica (CM 76736); Nickerie, Avanavero (CM 68363), Kayserberg Airstrip (FMNH 93168), Sipaliwini Airstrip (CM 63517, 76739), Wilhemina Mountains (FMNH 95312, 95313); Para, Zanderij (CM 68365, 76742); Saramacca, Bigi Poika (CM 52735), La Poule (FMNH 95309); Suriname, Clevia (FMNH 95310), Lelydorpplan (FMNH 95308), Plantation Clevia (CM 76743, 76744), Powakka (CM 52741, 52742).

Field Observations: In addition to data accompanying our 32 voucher specimens, we recorded 18 unvouchered observations of Philander opossum at Paracou, for a total of 50 documented records of this species. Of our vouchers, 21 (66%) were shot, 5 (16%) were taken in Conibear traps, 3 (9%) were taken in Sherman live traps, 1 was taken in a Tomahawk wire live trap, and 1 was taken in a Victor snap trap, and 1 was caught by hand. One juvenile captured in a pitfall trap and another taken in a Sherman trap were released. Of all 50 records, 31 (62%) were of individuals shot, trapped, or sighted on the ground, whereas 19 (38%) were of individuals shot, trapped, caught by hand, or sighted in trees (or on other elevated substrates such as inclined trunks, lianas, etc.) from 1.5 to 6 m above the ground. With the exception of females carrying nursing young, all P. opossum that we encountered were solitary, and all were collected or sighted at night. Of 49 records accompanied by habitat data, 32 (65%) were of individuals encountered in primary forest, usually near streams or in swamps, but occasionally at well-drained sites; 17 observations (35%), however, were based on individuals encountered in roadside secondary growth or other more-or-less disturbed habitats.

One female, collected on 9 July 1991, had two pouch young measuring 49 mm crown-rump; another, collected on 24 October 1992, had three pouch young measuring 24 mm; and a third, collected on 11 August 1993, had four pouch young measuring 27 mm.

Xenarthra

Nine xenarthran species have been definitely recorded from Paracou and it is unlikely that any others occur locally (see appendix 1). Because most xenarthran species are easily recognized by external characters (Husson, 1978; Emmons, 1990, 1997), we collected few vouchers.

Bradypus tridactylus Linnaeus

Three-toed sloths are possibly common at Paracou, but they are seldom seen and we did not encounter any in the course of our fieldwork. P. Petronelli (personal commun., 1993) estimated that the species is sighted about once or twice a year at Paracou, much less frequently than two-toed sloths (see below).

Choloepus didactylus (Linnaeus)

Voucher Material: AMNH 265952, MNHN 1995.952. Total = 2 specimens.

Identification: Although the species-level taxonomy of two-toed sloths, genus Choloepus, has never been formally revised, two species are consistently recognized in modern synoptic treatments of edentate classification (e.g., Wetzel and Ávila-Pires, 1980; Wetzel, 1982, 1985); of these, only C. didactylus is known to occur in the Guiana subregion of Amazonia. Our voucher material conforms with the brief description of topotypic specimens of C. didactylus from Surinam provided by Husson (1978), and with the cranial diagnosis of this species provided by Wetzel (1985). However, the fur on the throat of our adult female specimen (AMNH 265952) is distinctly paler and shorter than the pectoral fur, a pelage trait that Wetzel regarded as diagnostic of C. hoffmanni (from western Amazonia and Central America). The throat fur of our juvenile male (MNHN 1995.952) is likewise paler than the chest fur, but the contrast is less marked than in the adult female. For comparison with Surinamese topotypes measured by Husson (1978: table 38), external measurements of our adult female voucher were 698 × 12 × 164 × 29 mm, and it weighed 7.3 kg. Selected cranial measurements of this specimen are: condylobasal length (“greatest length of skull”), 119.1 mm; zygomatic breadth, 73.0 mm; interorbital constriction, 36.3 mm; alveolar length of maxillary toothrow, 44.5 mm. The ratio of the minimal to the maximal interpterygoid width (after Wetzel, 1985) in this specimen is 0.49.

Field Observations: Two-toed sloths are probably common at Paracou. P. Petronelli (personal commun., 1993) estimated that the species is sighted by forestry workers about five or six times a year in our study area, a reasonably high frequency given its cryptic appearance and inconspicuous habits. Of five sightings by inventory personnel from 1991 to 1994, four were nocturnal and one was diurnal: (1) RSV found a lactating female and a juvenile male (the vouchers described above) hanging together motionless in subcanopy vegetation (ca. 20 m above the ground) in well-drained primary forest at 20:15 hours on 30 July 1991. (2) RSV found a large adult of undetermined sex descending a tree head-first (probably to defecate) in well-drained primary forest at ca. 23:00 hours on 6 August 1991; the animal was about 2 m above the ground and remained motionless throughout the encounter (about 10 min), even when touched; its eyeshine was very faint. (3) RSV saw another solitary adult of unknown sex sleeping suspended from a liana ca. 4 m above the ground in old secondary growth at 02:00 hours on 31 July 1993. (4) L. H. Emmons saw a solitary animal of unknown sex climbing rapidly about 20 m above the ground in well-drained primary forest at 10:15 hours on 24 September 1994. (5) L. H. Emmons found an adult of unknown sex hanging motionless under a branch about 20 m above a stream in primary forest at 20:21 hours on 5 October 1994.

Cabassous unicinctus (Linnaeus)

The only record of this elusive armadillo at Paracou is an infant specimen (with unopened eyes) excavated by a bulldozer when a tract of well-drained primary forest was being cleared for a new experimental plantation; forestry workers brought the animal to P. Petronelli (personal commun., 1993), who showed us the photographs he had taken of it.

Dasypus kappleri Krauss

Voucher Material: AMNH 267011. Total = 1 specimen.

Identification: Our single voucher conforms exactly in qualitative external and cranial characters to Husson's (1978) and Wetzel and Mondolfi's (1979) descriptions of this species, the type locality of which is in Surinam. For comparison with quantitative data summarized by those authors, the external measurements of our adult male voucher were 565 × 405 × 125 × 53 mm, and it weighed 10.6 kg. The carapace of this specimen has eight movable bands, of which the fourth has 58 scutes; the condylonasal length of the skull is 128.6 mm, the zygomatic width 53.2 mm, and the mastoidal width 35.3 mm; there are eight paired maxillary teeth and eight paired mandibular teeth (meristic counts and cranial measurements follow Wetzel and Mondolfi's conventions).

Field Observations: Our single voucher of Dasypus kappleri and two other individuals (unambiguously identified but not collected) were all encountered at night, foraging singly on the ground in primary forest. Several other armadillo encounters recorded in our fieldnotes might have been of this species, but positive identification requires a clear and reasonably close view (to accurately judge size or to see the enlarged scutes on the knee), and many animals were only seen fleetingly or at a distance. Because we doubted that the forestry workers and local hunters with whom we spoke reliably distinguished this armadillo from the smaller but otherwise externally similar D. novemcinctus, we did not collect second-hand information about Dasypus species.

Dasypus novemcinctus Linnaeus

Voucher Material: AMNH 266483, 267012; MNHN 1995.953. Total = 3 specimens.

Identification: Our three vouchers conform in all qualitative external and cranial characters with the descriptions of this species by Husson (1978) and Wetzel and Mondolfi (1979). For comparison with Wetzel and Mondolfi's summary of quantitative data from Dasypus novemcinctus (op. cit.: table 1), the external measurements of our only adult specimen (AMNH 267012, female) were 506 × 380 × 95 × 51 mm, and it weighed 4.8 kg. The shell of AMNH 267012 was not preserved, but the other two specimens (both subadults with unfused basicranial sutures and incompletely erupted dentitions) each have nine movable bands, of which the fourth has 57 (MNHN 1995.953) or 60 (AMNH 266483) scutes. The condylonasal length of the skull of AMNH 267012 is 102.4 mm, the zygomatic width 44.8 mm, and the mastoidal width 29.0 mm; this specimen has seven paired maxillary teeth and eight paired mandibular teeth.

Field Observations: We heard armadillos crashing through the undergrowth or caught brief glimpses of them as they fled almost every night; most were probably Dasypus novemcinctus (usually the commonest rainforest armadillo throughout its extensive geographic range; Emmons, 1990, 1997), but it was impossible to make certain identifications in such cases. Our three vouchers were all shot at night in primary forest, at both well-drained and swampy sites. Numerous unambiguous sightings (many of which were not recorded) indicate that this species is common in all local habitats including primary forest and secondary growth. All of our observations were of solitary individuals. Although most sightings were nocturnal, we occasionally encountered nine-banded armadillos in the late afternoon, usually an hour or less before dusk.

Priodontes maximus (Kerr)

We only saw the giant armadillo once, on the night of 27 October 1994, when A. L. Peffley and RSV encountered a large (ca. 1 m HBL) adult of unknown sex excavating a mound of dead wood beside a rotting log in primary forest near a small stream at ca. 20:00 hours. Alarmed by our headlights, it fled uphill toward a large treefall, where we were unable to follow in the dense undergrowth. Bushnegro forestry workers have sometimes pointed out burrows and excavations said to be made by this species to P. Petronelli (personal commun., 1993); according to them, collared peccaries (Pecari tajacu) use the abandoned burrows of giant armadillos as nocturnal retreats. Only one individual is known to have been shot by local hunters, about ten years before our inventory fieldwork began.

Cyclopes didactylus (Linnaeus)

The only records of the pygmy anteater at Paracou are two sightings by P. Petronelli (personal commun., 1993), each of a single animal of unkown sex crossing a dirt road through our study area.

Myrmecophaga tridactyla Linnaeus

Four sightings of giant anteaters by local forestry workers were reported to us (P. Petronelli, personal commun., 1993). One animal was seen in savanna vegetation, the other three in the forest; all were seen in the daytime. Covering about ten years of human residence at Paracou, these scant observations suggest that this large and conspicuous diurnal species is very uncommon.

Tamandua tetradactyla (Linnaeus)

Although we encountered this species only in 1992, when what was perhaps a single individual was heard twice (tearing apart dead wood or termite nests in trees at night) and sighted once (walking down a road at 07:20 hours) during a two-day interval, tamanduas are apparently not uncommon locally. P. Petronelli (personal commun., 1993) estimated that they are typically seen about five or six times a year by local forestry workers.

Husson (1978) referred the tamanduas of the Guiana subregion to Tamandua longicaudata (Wagner), but we follow Wetzel's (1975) revision of Tamandua in regarding longicaudata as a junior synonym of tetradactyla. Wetzel's revision should be consulted for diagnostic characters, as well as for a discussion of the considerable geographic and nongeographic variation in coat color of this species. Although Wetzel's map (op. cit.: fig. 1) indicates that only partially vested, unvested, or melanistic specimens are known from French Guiana, our single sighting (by DPL in broad daylight at a distance of ca. 10 m) was of a yellow individual with a distinctly blackish vest.

Primates

Six species of primates are definitely known to occur at Paracou, or to have occurred there in the recent past, and a seventh species could be expected (see appendix 1). Unfortunately, the local primate fauna has been decimated by uncontrolled hunting, a process that accelerated following completion of the new asphalt highway from Kourou to Sinnamary in 1992. Even at the begining of our fieldwork in 1991, however, spider monkeys (Ateles paniscus), sakis (Pithecia pithecia), and capuchins (Cebus sp.) were rare. By 1994 (our last field season), only howlers (Alouatta seniculus) and tamarins (Saguinus midas) were commonly heard or seen in the vicinity of our camp, although capuchins could still be found a few kilometers away. Squirrel monkeys (Saimiri sciureus) have probably always been local vagrants, not regular residents.

Thus, we never had the opportunity to census an intact primate community at Paracou, and some of the observations cited below were necessarily recorded at second hand. Fortunately, all Guianan primates are easily identified by obvious external characters (Emmons, 1990, 1997), so the likelihood of mistakes in the second-hand identifications we cite is remote.

Saguinus midas (Linnaeus)

Voucher Material: AMNH 266481, 266482; MNHN 1998.699. Total = 3 specimens.

Identification: Our three vouchers correspond exactly with Husson's (1978) detailed description of this species, which was based on topotypic material from Surinam. In particular, the diagnostic external markings of golden-handed tamarins—bright orange (sometimes reddish or yellow) hands and feet that contrast with the blackish limbs—are conspicuous in our vouchers as they are in all specimens referred to this taxon throughout the Guiana subregion of Amazonia (Hershkovitz, 1977). For comparison with published measurement data (Husson, 1957, 1978; Napier, 1976; Hershkovitz, 1977), selected external and craniodental dimensions (mm) of our two adult female vouchers are: head-and-body length 258, 285; length of tail 425, 440; length of hindfoot 77, 80; ear 36, 37; condylobasal length 38.9, 41.2; orbital breadth 28.7, 29.1; postorbital constriction 24.3, 25.6; zygomatic breadth 33.3, 34.8; maxillary toothrow (crown length C–M2) 12.8, 12.9.

The black-handed tamarins that occur south of the Amazon and east of the Xingu (including Ilha de Marajó) were long considered to be a distinct species from the golden-handed tamarins of the Guiana subregion (e.g., by Elliot, 1912; Cruz Lima, 1945; Hill, 1957; Cabrera, 1961). Hershkovitz (1977), however, treated golden- and black-handed tamarins as no more than subspecifically distinct; according to this authority, the correct name for the golden-handed Guianan tamarins is Saguinus midas midas (Linnaeus), whereas the black-handed tamarins of southeastern Amazonia should be called S. midas niger (E. Geoffroy). Apparently, the only published justification for treating these unequivocally diagnosable taxa as conspecific is the following (Hershkovitz, 1977: 207):

Apparently, Hershkovitz judged the chromatic differences between midas (sensu stricto) and niger to be evolutionarily labile and predicted the existence of an extinct (or undiscovered) form that was (or is) intermediate in coloration and geography. However, the constancy of tamarin markings on opposite sides of the Amazon suggests that the character transformation in question is not evolutionarily labile, nor have populations with intermediate phenotypes yet been reported from any Amazonian islands.

According to Hershkovitz, only the coloration of the hands and feet distinguishes midas from niger, but his monograph contains no explicit comparison of these taxa in nonpelage characters. By contrast, subsequent research has shown that midas and niger have divergent dental measurements (Natori and Hanihara, 1992) and β₂-microglobulin DNA sequences (Canavez et al., 1999). Indeed, parsimony analysis of the β₂-microglobulin data provides compelling evidence that midas is more closely related to another species that occurs north of the Amazon, S. bicolor (Spix), than it is to niger on the opposite bank (Canavez et al., 1999).

In view of (1) the diagnosability of golden- and black-handed tamarins by bold and consistent pelage markings, (2) the existence of correlated divergence in nonpelage characters, and (3) clear indications from phylogenetic analysis that these taxa are not sister taxa, the currently accepted use of midas and niger as subspecies (Hershkovitz, 1977) or synonyms (Groves, 1993) is not defensible. Instead, we recognize the golden-handed tamarins of the Guiana subregion, Saguinus midas, as a species distinct from the black-handed tamarins of southeastern Amazonia historically known by authors as S. niger, S. tamarin, or S. ursulus (see below).

Remarks: Whereas most early authors used the epithets tamarin Link or ursulus Hoffmannsegg for the black-handed tamarin of southeastern Amazonia, Hershkovitz (1977) argued that the oldest applicable name for the zoological taxon in question is niger E. Geoffroy. Confusingly, the holotype of niger was clearly stated to have come from Cayenne (Geoffroy, 1803), far from the known range of black-handed tamarins. Although Hershkovitz reassigned the type locality to Belém, he did not examine Geoffroy's specimen, the identity of which is obviously problematic. Unfortunately, the holotype of niger (No. XXIV in Geoffroy's catalog) is lost: it was not listed in Rode's (1938) catalog of MNHN primate types, and it is not part of the current Paris museum collection (M. Tranier, personal commun.). Whether the original specimen of Geoffroy's niger was a melanistic individual of the golden-handed species collected at Cayenne or was a mislabelled example of the black-handed species is now impossible to determine. Nevertheless, the black-handed species is now widely and consistently known by the epithet niger E. Geoffroy, a usage that should be preserved in the interest of taxonomic stability. For that purpose, we hereby designate as neotype of Sagouin niger E. Geoffroy, 1803, an adult male specimen represented by a well-preserved skin and skull in the American Museum of Natural History, AMNH 96500, collected by A. M. Olalla on 2 November 1931 at Cametá on the Rio Tocantins, Pará, Brazil, from which locality a large series of topotypes is also available.

Other Specimens Examined: Guyana—Cuyuni-Mazaruni, Kartabo (AMNH 65159, 142936).

Field Observations: This is the commonest primate species at Paracou. We saw groups of tamarins daily, in both swampy and well-drained primary forest and in roadside secondary growth. In primary forest, tamarins were invariably sighted in the canopy or subcanopy, but we often saw them descend to within a few meters of the ground in roadside secondary growth; occasionally, groups were seen crossing dirt roads on the ground when the gap between trees on either side was too wide to leap. Unlike other primate species at Paracou, tamarins did not noticeably decline in density from 1991 to 1994, probably because they are not locally hunted for meat.

Alouatta seniculus (Linnaeus)

We heard howlers night and day throughout the course of our fieldwork at Paracou. Despite the audible evidence of their continuous presence, however, we seldom caught more than fleeting glimpses of this locally persecuted species. The few groups surviving in our study area never closely approached our camp, and they seldom ventured within the radius of our daily activities except to visit fruiting fig trees and other transient resources.

Ateles paniscus (Linnaeus)

Spider monkeys, always vulnerable to local extirpation by hunters because of their loud vocalizations, large size, highly prized meat, and low reproductive rate, have been scarce at Paracou for the last decade or more. P. Petronelli (personal commun., 1993) guessed that he had seen solitary individuals, never groups, on only five or six occasions since the early 1980s. We recorded only a single encounter with this species, when DPL saw a solitary individual in the canopy of well-drained primary forest near a fruiting fig tree on 2 July 1991; no noncaptive spider monkeys were seen or heard by us in later years.

Cebus apella (Linnaeus)

One or two species of Cebus were common at Paracou in the early 1980s (P. Petronelli, personal commun., 1993), but capuchins are now rarely seen or heard in the area. The older forestry workers are familiar with both a large and a small species, presumably C. apella and C. olivaceus (see appendix 1), but only C. apella has been definitely identified by sight at Paracou (P.-M. Forget, personal commun., 1994). We heard Cebus on several occasions in 1994 when our inventory activities extended ca. 3 km to the NNW of camp (the limit of our sampling radius), but we caught no more than a brief glimpse of the animals as they fled in the distance and were not able to determine the species.

Pithecia pithecia (Linnaeus)

Although saki monkeys may once have been common at Paracou, P. Petronelli (personal commun., 1993) recalled only one sighting of this species prior to our inventory: a pair that he observed on 31 January 1991. Subsequently, we recorded three observations of Pithecia pithecia in our fieldnotes: (1) DPL saw a solitary individual on 3 July 1991; (2) RSV saw what might have been the same animal in the subcanopy of well-drained primary forest at 07:15 hours on 4 July 1991; and (3) Roland W. Kays saw two individuals leaping from tree to tree as he was walking along a dirt road at 16:30 hours on 6 August 1993. Silent and retiring by disposition, sakis may still linger in remote and seldom-visited parts of our study area, but their numbers have certainly been much reduced by hunting.

Saimiri sciureus (Linnaeus)

Squirrel monkeys have apparently been sighted only once at Paracou, when P. Petronelli (personal commun., 1993) encountered a group of 10–12 individuals on the edge of a small patch of savanna vegetation in the early 1980s. This species is probably a local vagrant that seldom strays far from the tangled growth at the forested margins of nearby savannas and rivers.

Carnivora

Ten species of carnivores are definitely known to occur at Paracou, and it is doubtful that any others occur in our study area except as rare vagrants (see appendix 1). Because all rainforest carnivores can be confidently identified by external characters (Emmons, 1990, 1997), and because many are uncommon and/or elusive, most of the information reported below was obtained by interviewing local forestry personnel.

Speothos venaticus (Lund)

We did not encounter bush dogs during our 1991–1994 fieldwork at Paracou, but P. Petronelli (personal commun., 1993) told us he had seen them twice in previous years: once as a group of four individuals, and another time as a group of three. A visiting photographer also saw a pack of seven bush dogs chase a paca across a road through the forest near our camp. All of these sightings were diurnal.

Herpailurus yaguarondi (Lacépède)

We did not see jaguarundis at Paracou, but P. Petronelli (personal commun., 1993) reported three sightings, all of solitary individuals on the ground in the daytime: one was crossing a road, one was near a stream in primary forest, and another was in secondary growth near the edge of a small patch of savanna vegetation.

Leopardus pardalis (Linnaeus)

Although ocelots are probably not uncommon at Paracou, they are rarely seen. We saw none, but local hunters have killed at least three in the last decade, one of which had the distinctively banded quills of Coendou prehensilis embedded in its neck and shoulders (P. Petronelli, personal commun., 1993).

Leopardus wiedii (Schinz)

Small spotted cats are often seen at night by local forestry workers (P. Petronelli, personal commun., 1993), and we recorded several fleeting encounters in our fieldnotes, but the identification of such observations is problematic. We have only two definite records of margays from our study area. (1) In 1993 we examined and measured an adult female, shot by a local hunter, that measured 563 × 427 × 124 × 54 mm and weighed 3.4 kg; the fur of the nape of the neck was reversed on this specimen, which likewise corresponded in other external characters to the descriptions provided by Pocock (1941), Emmons (1990, 1997), and Oliveira (1998). (2) On 19 September 1994, L. H. Emmons observed an emaciated male at a distance of only 6 m from 19:20 to 19:35 hours; the animal was encountered near a stream in primary forest, and half of its face was bristling with the large white quills of Coendou prehensilis.

Panthera onca (Linnaeus)

In the course of ten years' residence at Paracou, P. Petronelli (personal commun., 1993) told us that he had seen jaguars four times: twice in the daytime and twice at night, all in primary forest. Although we saw the distant eyeshine of large cats on several occasions, our only definite record of this species is based on the pugs of a young adult that L. H. Emmons observed along a dirt road on 13 October 1994.

Puma concolor (Linnaeus)

We did not see this species at Paracou in the course of our fieldwork, but P. Petronelli (personal commun., 1993) told us he had encountered pumas locally on four occasions, twice at night and twice in the daytime. One of the diurnal sightings was of an animal asleep on a high tree limb; another individual was found asleep between the buttresses of a big tree.

Eira barbara (Linnaeus)

Although tayras are common at Paracou, where they are sighted on average about 10–12 times a year by forestry workers (P. Petronelli, personal commun., 1993), we saw only three in the course of our fieldwork: RSV encountered a solitary individual in well-drained primary forest at 14:45 hours on 9 July 1991, DPL saw one crossing a dirt road in the daytime on 6 November 1992, and Nancy A. Voss sighted another in primary forest on an unrecorded date in 1994.

Galictis vittata (Schreber)

The only definite record of grisons within the limits of our study area is a sighting by P. Petronelli (personal commun., 1993), who observed a pair travelling together on the ground in primary forest in 1990. A large adult male that we found dead on a road near Sinnamary (ca. 12 km NNW of Paracou) in 1992 measured 553 × 155 × 92 × 32 mm and weighed 3.8 kg.

Nasua nasua (Linnaeus)

Voucher Material: AMNH 267605; MNHN 1995.959. Total = 2 specimens.

Identification: Our two vouchers, both juveniles, conform closely in pelage characters with the description of Surinamese specimens identified by Husson (1978) as Nasua nasua vittata Tschudi (1845), the type locality of which is in Guyana. Unfortunately, the appropriate trinomial designation for coatis from the Guiana subregion of Amazonia, if indeed a subspecific classification is necessary, remains to be convincingly determined.

Tate (1939) identified the lowland coatis of the Guiana subregion as Nasua phaeocephala J. A. Allen (1904)—apparently overlooking the availability of Tschudi's older name—and proposed a new name, dichromatica, for the montane population on Auyantepui in Venezuela. Subsequently, Cabrera (1958) referred all named forms of South American coatis to N. nasua, of which 11 subspecies were recognized as valid; in his classification, phaeocephala and dichromatica were both treated as subjective synonyms of N. n. vittata. According to Hershkovitz (1959), however, the oldest valid name for a Guianan coati is Viverra quasje Gmelin (1788), a name said to be based primarily on a Surinamese specimen described and illustrated by Seba (1734). Although Seba's specimen (BMNH 67.4.12.447), the presumptive type of quasje, was reported to be extant by Thomas (1892), we have found no evidence that it has been examined by any subsequent author.

Decker's (1991) revision of Nasua did not explicitly recognize any valid subspecies of N. nasua, but Gompper and Decker (1998) listed ten, including N. n. vittata (with phaeocephala and dichromatica as synonyms). Confusingly, Gompper and Decker listed quasje as a synonym of N. n. nasua, thus implying that two valid taxa of coatis occur in the Guianas. Without having undertaken a specimen-based revision of coati taxonomy, we are unable to evaluate the possible significance of any geographic variation in Nasua nasua that might occur in the Guiana subregion of Amazonia. However, it is clear from the literature reviewed above that if the French Guianan and Surinamese populations (as represented by our vouchers and Husson's material) are distinct from the nominate form, the correct name for them may be quasje Gmelin, not vittata Tschudi. Clearly, resolving the application of Gmelin's name and reexamining Seba's original specimen should be a priority in any future revisionary study.

Field Observations: Coatis are uncommon at Paracou. P. Petronelli (personal commun., 1993) told us that he had seen coatis only twice in ten years, once in an experimental plot of disturbed forest (a group of two animals), and the second time in primary forest near a stream (a group of seven or eight); both sightings were in daytime. We encountered coatis only three times in the course of our inventory fieldwork. (1) On 15 August 1993, R. W. Kays sighted a group estimated to consist of about 20 individuals in a large tree at night; the tree was in well-drained primary forest, and the animals (including our two vouchers) were perching among the branches, about 20 m above the ground. (2) L. H. Emmons sighted a group of unknown size in swampy primary forest at 09:25 hours on 13 October 1994. (3) A. L. Peffley encountered a solitary individual of unknown sex in swampy primary forest at 09:00 hours on 18 October 1994.

Potos flavus (Schreber)

Voucher Material: AMNH 265956, 265958, 265959, 266597–266599, 267048, 267050, 267051, 267053, 267607, 267608; MNHN 1995.954–1995.958. Total = 17 specimens.

Identification: Our voucher material corresponds exactly in qualitative characters with Husson's (1978) description of topotypic specimens from Surinam. External and craniodental dimensions of Paracou specimens (table 15) likewise overlap those of topotypes measured by Husson (op. cit.: table 43).

Although partial revisions of Potos flavus by Thomas (1902), Kortlucke (1973), and Hernández-Camacho (1977) each recognized several subspecies as valid, there has been no geographically comprehensive study of kinkajou taxonomy to date. The Paracou population is presumably referrable to P. f. flavus, but the necessity for a trinomial classification remains to be convincingly demonstrated.

Field Observations: Kinkajous are by far the commonest carnivore at Paracou. We heard them squealing and crashing about in the canopy virtually every night throughout the course of our 1991–1994 fieldwork. All of our vouchers were shot at night in trees, at estimated heights ranging from 10 to 30 m above the ground; recorded habitats include well-drained and swampy primary forest and roadside secondary growth. Although many apparently solitary individuals were encountered, kinkajous were also encountered foraging or travelling in pairs and larger groups. Collected specimens accompanied by information about group size include (1) an apparently solitary subadult female, (2) an apparently solitary lactating adult female, (3) a nonlactating and nonpregnant adult female accompanied by at least one other individual, (4) a juvenile male from a group of three individuals, (5) an apparently solitary subadult male, (6) an apparently solitary adult male, and (7) an adult male accompanied by at least one other individual. None of the six adult females we collected (in August and November) were pregnant.

Perissodactyla

Only one perissodactyl species occurs at Paracou, and no others are known from any Amazonian locality. Tapir dung, spoor, and the animal itself are unmistakable.

Tapirus terrestris (Linnaeus)

Although feces and tracks of this species are said to be commonly encountered in the more remote and swampy parts of our study area, the animal itself is seldom seen. During ten years' residence at Paracou, P. Petronelli (personal commun., 1993) encountered only one tapir, near a headwater stream of Crique Paracou in the daytime. Local hunters, however, are known to have killed at least two during the same interval.

Artiodactyla

Four rainforest artiodactyl species are known to occur at Paracou, and no others are expected (see appendix 1). All are unmistakable by external characters (Emmons, 1990, 1997) if a sufficiently good view is obtained.

Mazama americana (Erxleben)

Red brockets are apparently common at Paracou, local hunters having killed dozens in the last decade. According to P. Petronelli (personal commun., 1993), Mazama americana is primarily active at night; occasional daytime sightings are probably of individuals disturbed near their resting places. Although M. americana is believed to be locally more abundant than M. gouazoubira (below), we nevertheless obtained only two unambiguous sightings of red brockets in the course of our 1991–1994 inventory: (1) RSV observed a solitary individual stealthily retreating through the undergrowth of swampy primary forest in the mid-afternoon of 25 June 1991; (2) DPL saw a solitary individual in well-drained primary forest at 10:10 hours on 23 July 1993. Many unidentified deer whose eyeshine was detected in the forest undergrowth at night, or whose alarmed reactions (snorts and foot-stamps) were heard in the darkness, could have belonged to this species, or to the next.

Mazama gouazoubira (G. Fischer)

Voucher Material: AMNH 265960, 265961; MNHN 1995.960. Total = 3 specimens.

Identification: Both of our adult examples, one male (MNHN 1995.960) and one female (AMNH 265961), agree closely with Husson's (1978) qualitative description of Surinamese specimens that he identified as Mazama gouazoubira nemorivaga. Measurements of our vouchers (table 16) likewise correspond closely with morphometric data from M. g. nemorivaga summarized by Husson (1978: table 61) and Bisbal (1991: table II). In qualitative cranial traits, our two adult skulls conform with Medellín et al.‘s (1998) characterization of northern South American populations of M. gouazoubira, except that the mesopterygoid fossae of both specimens have broadly U-shaped (not V-shaped) anterior margins.

Our vouchers are practically topotypes of Cervus nemorivagus F. Cuvier, the description of which was based primarily on specimens from Cayenne (Allen, 1915b; contra Miranda-Ribeiro, 1919). Although this taxon is currently considered a subspecies of Mazama gouazoubira (e.g., by Czernay [1987] and Grubb [1993], presumably following Ávila-Pires [1959]), no revisionary study based on extensive specimen data has, in fact, shown that the small grayish brockets of the Guianas and those of Paraguay (the type locality of gouazoubira) are really conspecific. The character differences that Tate (1939) observed between specimens that he referred to M. nemorivaga and M. simplicicornis Illiger (= M. gouazoubira) should be carefully evaluated in any future taxonomic analysis of these deer.

Remarks: The specific epithet of the gray brocket was originally spelled “gouazoupira” as noted by Grubb (1993). However, gouazoubira is the spelling that has been followed almost universally for many years, and we agree with Gardner (1999) that this usage should be maintained in the interest of nomenclatural stability.

Field Observations: Our three vouchers were all shot at night in well-drained primary forest. The still-nursing fawn (its stomach containing milk only) and the lactating adult female were collected together, whereas the young adult male (with antlers in velvet) was apparently solitary. In addition, L. H. Emmons observed what was probably a single individual on four different nights from 21 September to 10 October 1993 in well-drained primary forest.

According to P. Petronelli (personal commun., 1993), gray brockets are less common at Paracou than are red brockets, but they are nevertheless often sighted by forestry personnel, about 10–12 times per year on average.

Pecari tajacu (Linnaeus)

Collared peccaries were once common at Paracou, but although we often saw tracks and wallows (usually in parts of the forest distant from our normal inventory activities), we rarely saw the animals themselves. According to P. Petronelli (personal commun., 1993), this species is typically encountered locally in groups of 6–7 animals, always in the daytime, about 6–8 times per year on average.

Tayassu pecari (Link)

Groups of about 20 white-lipped peccaries have been sighted at Paracou on at least three occasions, all in the daytime (P. Petronelli, personal commun., 1993). Bushnegro forestry workers say that such groups visit the area about every four years. Although we did not see the animals themselves in the course of our 1991–1994 inventory fieldwork, L. H. Emmons observed fresh tracks of this species in primary forest on 12 October 1994.

Rodentia

Rodents constitute the most diverse group of nonvolant mammals at Paracou, where we documented the occurrence of 22 species in six families: Sciuridae (2 species), Muridae (11 species), Erethizontidae (2 species), Dasyproctidae (2 species), Cuniculidae (1 species), and Echimyidae (4 species). An additional 11 species are known from other localities in French Guiana or Surinam (appendix 1), and some of these could also be expected to occur in our study area. Two species are described as new below.

Predictably, most Paracou rodent identifications proved to involve significant taxonomic problems, the resolution of which occupies the bulk of the following accounts. We define specimens to be adult if the permanent dentition is fully erupted, subadult if the molar dentition is completely erupted but the permanent premolars are not, and juvenile if one or more molars are incompletely erupted. Our quantitative comparisons of rodent crania and dentitions are based on the following measurements (figs. 33, 34).

A few other measurements taken for special purposes are defined as necessary in the text and tables that follow. Because sexual dimorphism is an insignificant source of measurement variation in most rodents (e.g., see Straney [1978] and references cited by Voss [1988: 362]), we do not summarize morphometric data separately by gender.

Most of the larger rodents in the Paracou fauna (all sciurids, erethizontids, dasyproctids, and Cuniculus paca) can be identified at a distance by obvious external characters (Emmons, 1990, 1997), but most of the smaller rodents cannot be confidently identified without specimens in hand, and some closely related species cannot be unambiguously distinguished except from cleaned cranial material. In lieu of keys, we provide tabular summaries of diagnostic traits to facilitate identifications in some speciose genera.

Sciuridae

The two squirrels that we found at Paracou are the only species known to occur in French Guiana, Surinam, and Amapá; therefore, no future additions to the local sciurid fauna are expected. Pending a revision of the complex nomenclature of these animals (M. de Vivo, in prep.), we follow the usages recommended by Husson (1978), who examined type material that we have not seen.

Sciurillus pusillus (E. Geoffroy)

Voucher Material: AMNH 269119. Total = 1 specimen.

Identification: Our voucher is almost topotypical of this species (originally described from specimens collected at Cayenne) and exhibits the reddish head, black ear tips, and white postauricular patches said to distinguish S. p. pusillus from other nominal taxa of South American pygmy squirrels (Anthony and Tate, 1935; Husson, 1978). The external dimensions of AMNH 269119, an adult female, were 109 × 74 × 28 × 14 mm; including two embryos in utero, this specimen weighed 51 g. Because fluid-preserved material of Sciurillus is rare in museum collections, we did not extract the skull of AMNH 269119 for measurement.

Remarks: For the availability of names from Geoffroy's (1803) catalog (rejected by Wilson and Reeder, 1993: 831), see Hershkovitz (1955) and Holthuis (1963).

Field Observations: Our single voucher was shot by L. H. Emmons at 11:45 hours on 21 September 1994 as it fed on something growing on (or concealed beneath) the bark of a large Inga sp. (Mimosoideae) at a height of about 18 m in well-drained primary forest. In addition, we recorded fleeting diurnal observations of this species on five dates from 1991 to 1993; all of these sightings were of solitary individuals in trees in well-drained primary forest.

Sciurus aestuans Linnaeus

Voucher Material: AMNH 266485–266488, 266492, 266493, 267013, 267565; MNHN 1995.989–1995.991. Total = 11 specimens.

Identification: Our voucher material agrees closely in most details with Husson's (1978: 386–387) description of topotypic specimens from Surinam, but several points of comparison merit comment. (1) According to Husson, a few Surinamese examples have “very inconspicuous buffy yellow postauricular patches”, but most do not; there is no trace of a postauricular patch on any specimen from Paracou. (2) Husson described the ventral coloration as “usually pale reddish brown, sharply separated from the colour of the dorsal surface, at least in about the middle of the body, but considerably less so in the anterior and posterior parts”. By contrast, Paracou skins have clear (self-colored) orange fur on the chest, fading to buff or cream on the throat; some clear orange fur extends posteriorly along the ventral midline onto the abdomen, but most of the abdominal fur is gray-based, appearing grizzled like the flanks although much paler. (3) Some Surinamese specimens have substantially smaller measurements (op. cit.: table 62) than our vouchers (table 17), but it is possible that Husson included subadults in his sample.

This species (together with other members of the so-called aestuans group of Sciurus) was referred to the genus Guerlinguetus Gray by Allen (1915a), Tate (1939), Moojen (1942), Moore (1959), and others, but most recent authors have followed Cabrera (1961) in treating Guerlinguetus as a subgenus of Sciurus. Cabrera cited no published analysis of character data to support his opinion, however, and it seems probable that renewed morphological and molecular studies of Neotropical squirrels will advocate a return to the older generic usage.

Although the currently accepted synonymy (Hoffmann et al., 1993) for Sciurus aestuans implies that this species is distributed throughout eastern Amazonia to southeastern Brazil (including such forms as alphonsei Thomas, garbei Pinto, henseli Miranda-Ribeiro, ingrami Thomas, poaiae Moojen, and roberti Thomas; see Cabrera [1961] for bibliographic citations and type localities), we follow the last substantive specimen-based revisionary treatment of Amazonian squirrels (Moojen, 1942) in restricting S. aestuans to the Guiana subregion of Amazonia. In the Amazonian lowlands of southeastern Venezuela (geographically part of the Guiana subregion), however, S. aestuans is replaced by a different species that is usually identified (e.g., by Tate, 1939; Handley, 1976; Linares, 1998) as S. gilvigularis Wagner.

Field Observations: Sciurus aestuans is one of only three common diurnal rodents at Paracou. Most (9) of our 11 voucher specimens were shot in the daytime, and we recorded an additional 11 unvouchered daytime sightings in our fieldnotes; two trapped specimens were found at or near dawn, but might have been captured the preceding afternoon. With the exception of the latter, which were taken near ground level in Conibear and leghold traps set on tree trunks over a small stream, all of our observations of this species at Paracou were of animals perched in trees at heights of 3–30 m above the ground. Most individuals were solitary, but an adult male and an adult female were collected together on 13 August 1991. Habitat data accompanying specimens or sight records include 16 observations in primary forest at both well-drained and swampy sites, and 5 in more-or-less disturbed habitats.

Muridae

All of the genera of rainforest murids known to occur in the Guiana subregion of Amazonia are documented by vouchers collected at Paracou, including species of Neacomys, Nectomys, Neusticomys, Oecomys, Oligoryzomys, Oryzomys, and Rhipidomys. Although most Guianan murids can be provisionally identified to genus in the field by external characters described by Husson (1978) and Emmons (1990, 1997), we provide illustrations and supplementary information to facilitate field determinations of problematic taxa. Our anatomical terminology for muroid morphological characters follows that referenced or defined by Reig (1977), Voss (1988, 1993), Carleton and Musser (1989), Voss and Carleton (1993), and Musser et al. (1998).

Neacomys Thomas

The Neotropical spiny mice of the genus Neacomys have never been revised, and many aspects of their species-level taxonomy have long been problematic. The identification of spiny mice from the Guiana subregion of Amazonia is a case in point: although these have traditionally been identified as N. guianae (e.g., by Anthony, 1921a; Tate, 1939; Carvahlo, 1962; Husson, 1978; Genoways et al., 1981; Guillotin, 1982; Malcolm, 1990; Voss and Emmons, 1996), the diagnostic morphological characters and geographic range of this species are not documented in the literature. In his original description, Thomas (1905: 310) compared N. guianae only with N. spinosus (Thomas, 1882), stating that the new species was very similar but “conspicuously smaller”. However, size does not distinguish guianae from such other diminutive forms as tenuipes Thomas (1900), pusillus Allen (1912), and pictus Goldman (1912). Musser and Carleton (1993) listed N. guianae, N. pictus, N. spinosus, and N. tenuipes (including pusillus) as valid species, but the recent description of two additional species from western Amazonia (N. minutus and N. musseri), together with sequence comparisons showing high levels of genetic differentiation among several undescribed mtDNA clades of spiny mice, suggests that the genus is much more diverse than previously recognized (Patton et al., 2000).

In order to identify our Paracou vouchers, we examined original descriptions of all nominal taxa of Neacomys, and we examined holotypes or paratypes of all the smaller named forms (guianae, minutus, musseri, pictus, pusillus, tenuipes). We tried to locate every Guianan Neacomys specimen currently housed in North American and European museums, and we measured representative series to document morphometric variation within and among species. The results of our comparisons indicate that at least three distinct species are present in the Guiana subregion of Amazonia, of which two are new and occur sympatrically at Paracou. Because the very brief diagnoses of Neacomys provided by Thomas (1900), Gyldenstolpe (1932), and Ellerman (1941) are now insufficient as a basis for taxonomic inference, we rediagnose the genus here.

Emended Diagnosis of NEACOMYS: Small oryzomyines (sensu Voss and Carleton, 1993: 31) with coarsely grizzled yellowish-, reddish-, or grayish-brown dorsal fur containing short, grooved spines in addition to conventional guard hairs and underfur; ventral fur similar in composition to dorsal fur, but shorter and always contrastingly colored; pinnae small, dark, and sparsely haired; mammae eight in inguinal, abdominal, postaxial, and pectoral pairs; hindfoot with outer digits (I and V) much shorter than three middle digits (claw of dI not extending beyond middle of first phalange of dII, claw of dV not extending beyond first interphalangeal joint of dIV); claws of pedal digits II–V provided with ungual tufts of long whitish or silvery hairs that exceed the claws in length; tail sparsely haired (appearing naked except under magnification) with prominent epidermal scales in annular series, sometimes with a thin terminal pencil but never with a conspicuous tuft of long hairs at tip. Skull with prominently beaded supraorbital margins; interparietal large; palate long and wide, with prominent and often complex posterolateral pits flanking anterolateral margins of mesopterygoid fossa; parapterygoid fossae shallow (never deeply excavated above level of palate); alisphenoid strut absent (except as rare, usually unilateral variant); carotid circulation includes large stapedial artery (pattern 1 or 2 of Voss, 1988); tegmen tympani not overlapping posterior margin of squamosal (posterior suspensory process of squamosal absent); subsquamosal fenestra sometimes small but always present and usually patent. Upper incisors small, narrow, and opisthodont (never orthodont or proodont); lower incisor root contained in prominent capsular process on lateral surface of mandible; molars small and pentalophodont (mesolophs on M1 and M2 always well developed and fused to mesostyle on labial cingulum); M1/m1 without accessory roots.

Neacomys dubosti, new species Figures 36, 37, 38B, 39A, 39C, 43

Type Material and Type Locality: The holotype, AMNH 267569, an adult female preserved as a fluid specimen with the skull extracted and cleaned, was collected at Paracou by R. W. Kays (original number: RWK 9) on 7 August 1993. No other material is known from the type locality, but all of the additional specimens we examined from French Guiana and Amapá are hereby designated as paratypes.

Distribution and Sympatry: Based on specimens we examined, Neacomys dubosti occurs in French Guiana, Amapá (Brazil), and southeastern Surinam (fig. 35). In Surinam, N. dubosti has been collected sympatrically with N. guianae (at the Sipaliwini Airstrip, Nickerie District), and it occurs sympatrically with another new species in French Guiana and Amapá (see the next account, below).

Etymology: The specific epithet honors Gérard Dubost for his many contributions to knowledge of mammalian ecology and natural history in the lowland rainforests of French Guiana and Gabon. We are also grateful for his original suggestion of Paracou as the site for our mammal inventory, and for his subsequent support and advice throughout the course of our fieldwork there.

Diagnosis: A small species of Neacomys (measurements in table 18) distinguished from other diminutive congeners by its short, usually unicolored tail; moderately short rostrum flanked by relatively shallow zygomatic notches; broad and strongly convergent interorbital region with highly developed, shelf-like supraorbital beads; broad and distinctly inflated braincase; short, convex-sided incisive foramina; carotid circulation pattern 1; M1 with undivided anterocone; mesoloph of M1 with more-or-less symmetrical connections to protocone and hypocone; persistently tubercular molar cusps; and a distinctive range of craniometric variation.

Morphological Description: Dorsal pelage coarsely grizzled tawny- or reddish-brown, somewhat paler along sides due to middorsal concentration of dark-tipped spines; ventral fur abruptly paler, sometimes pure white from chin to anus (e.g., CM 76840, MNHN 1983.412, USNM 46182), but more commonly suffused to a greater or lesser extent with buff or orange; ventral hairs usually pale to roots, very rarely with indistinctly gray bases (e.g., USNM 461571, 461590); broad lateral line of clear buff or orange separating dorsal and ventral pelage present in all specimens examined. Superciliary, genal, and some mystacial vibrissae extend behind pinnae when laid back against head. Dorsal surface of manus and pes covered by short pale fur in most specimens, but hairs over central metapodials sometimes indistinctly darker than those on digits and outer metapodials; claw of pedal digit I extending about one-half length of phalange 1 of adjacent digit II; claw of pedal digit V extending to but not beyond end of first phalange of adjacent digit IV; small but distinct hypothenar (lateral tarsal) plantar pad present on hindfoot of one fluid specimen (this trait is difficult to score reliably on dried skins). Tail about as long as combined length of head-and-body; almost always unicolored (dark above and below), but occasionally indistinctly paler ventrally at base (e.g., CM 76842, MNHN 1972.641); with small caudal scales (21 rows/cm near base of tail in one fluid specimen) forming relatively narrow annulations.

Skull with moderately short, tapering rostrum flanked by relatively shallow zygomatic notches; interorbital region broad, with strongly convergent lateral margins; supraorbital beads highly developed, projecting as small shelves over posterior orbits and continuing onto braincase as low temporal crests; braincase inflated, conspicuously domed, and very broad behind squamosal zygomatic processes. Incisive foramina relatively short (averaging about 57% of diastemal length), usually with distinctly convex lateral margins; zygomatic plate relatively narrow; carotid circulation with well-developed supraorbital ramus of stapedial artery (occupying squamosal-alisphenoid groove and sphenofrontal foramen; pattern 1 of Voss, 1988); subsquamosal fenestra smaller than postglenoid foramen but always distinct and patent; auditory bullae usually flask-shaped, tapering gradually from tympanic ring to unconstricted bony eustacian tubes.

First maxillary molar with undivided anterocone; anteroloph of M1 seldom distinct, usually fused labially with anterocone (anteroflexus usually distinguishable only as persistent internal fossette); mesoloph of M1 straight and slender, projecting labially from symmetrically Y-shaped junction with median mure, without disproportionate connection to hypocone; principal labial cusps (paracone, metacone) slightly reduced in size relative to lingual cusps (protocone, hypocone); principal cusps persisting as distinctly tubercular elements with moderate wear.

Karyotypes: Two specimens from Amapá (MNHN 1972.640, 1972.641) karyotyped by M. Tranier had diploid counts of 2N=62 chromosomes (as recorded on skin labels).

Variation: The three geographic samples at hand, from Surinam, French Guiana, and Amapá (table 18), are very similar in most qualitative and quantitative characters. Instead, most of the variation in the material we examined (e.g., as noted parenthetically in the preceeding description) occurs as individual differences within local populations. However, resemblances are strongest between French Guianan specimens and a large series from the Serra do Navio in Amapá, Brazil. By contrast, our few Surinamese examples have slightly narrower molars and interorbital regions, and their supraorbital beads appear somewhat less developed as projecting shelves.

Comparisons: Neacomys dubosti could potentially be confused with two previously described congeners from northern South America—N. tenuipes and N. guianae—in addition to N. paracou, another new species described below. Selected qualitative contrasts among these four taxa are summarized in table 19, descriptive univariate statistics for measurements of representative series are provided in table 20, and the results of multivariate morphometric analyses are represented in figure 40 and table 21. More detailed, character-by-character comparisons are deferred to the next account. Patton et al (2000) recently described additional species of small-bodied Neacomys from western Brazil, but those bear no close resemblance to either N. dubosti or N. paracou and so are not treated in these accounts.

Remarks: At least some of the specimens previously reported in the literature as Neacomys guianae by Carvalho (1962), Genoways et al. (1981), and Guillotin (1982) are probably referable to N. dubosti, but those authors did not provide the museum catalog numbers of relevant voucher material and we are therefore unable to associate confident species identifications with their observations.

Other Specimens Examined: Brazil—Amapá, Serra do Navio (USNM 461563–461569, 461571, 461572, 461574–461576, 461579–461582, 461584, 461588, 461590–461595, 461601, 461604, 461612); no other locality data (MNHN 1972.640, 1972.641). French Guiana—Cacao (MNHN 1983.426, 1986.534, 1986.535, 1986.537, 1986.538, 1986.541–1986.545), Camopi (MNHN 1983.403), Iracoubo (MNHN 1983.409), Piste St.-Élie km 16 (MNHN 1986.876, 1986.877), Saül (MNHN 1983.405–1983.407, 1983.422, 1983.423, 1983.425), St.-Eugène (MNHN 1995.3226–1995.3229, 1998.1835, 1998.1839), Trois-Sauts (MNHN 1982.629, 1982.630, 1983.410, 1983.412, 1983.414–1983.416). Surinam—Marowijne, Oelemarie (CM 76835–76837, 76839–76843); Nickerie, Sipaliwini Airstrip (CM 76846).

Field Observations: Our single specimen of Neacomys dubosti from Paracou was taken in a pitfall trap in creekside primary forest.

Neacomys paracou, new species Figures 36, 37, 39B, 39D, 42B, 43

Type Material and Type Locality: The holotype, MNHN 1995.1020, an adult male preserved as a complete skeleton, was collected at Paracou on 23 August 1993 by Roland W. Kays (original number: RWK 30). All of the additional specimens we examined from Paracou (see Specimens Examined, below) are hereby designated as paratypes.

Distribution and Sympatry: Specimens that we refer to Neacomys paracou are from French Guiana, Surinam, Guyana, eastern Venezuela (Bolívar state), and Guianan Brazil (north of the Amazon and east of the Rio Negro) (fig. 41). Based on these records it would be reasonable to expect that the species occurs throughout the Guiana subregion of Amazonia, but we have not seen any material from the Amazonas federal territory of Venezuela. Neacomys paracou has been collected sympatrically with N. guianae in Guyana (at Kartabo), and with N. dubosti in Surinam (Oelemarie), French Guiana (Cacao, Paracou, Saül, St.-Eugène), and Amapá (Serra do Navio).

Etymology: The species is named for our study area, treated as a noun standing in apposition to the generic name.

Diagnosis: A small species of Neacomys (measurements in table 22) distinguished from other like-sized congeners by its very short outer pedal digits; short, unicolored tail; short rostrum flanked by relatively deep zygomatic notches; broad and usually strongly convergent interorbital region with well-developed and often shelf-like supraorbital beads; narrow, uninflated braincase; long, parallel-sided incisive foramina; carotid circulation pattern 1; M1 with narrow, undivided anterocone; mesoloph of M1 stout, often curving from and disproportionately connected to hypocone; principal molar cusps quickly worn to enamel loops, not persistently tubercular; and a distinctive range of morphometric variation.

Morphological Description: Dorsal pelage coarsely grizzled tawny- or reddish-brown, somewhat paler along sides due to middorsal concentration of dark-tipped spines; ventral fur abruptly paler, often pure white from chin to anus, but sometimes with orange pectoral markings (e.g., AMNH 266548), or broadly suffused with orange (e.g., CM 76845); ventral hairs pale to roots in most specimens (rarely with indistinctly gray bases between the fore- and hindlegs; e.g., AMNH 266545); broad lateral line of clear buff or orange separating dorsal and ventral pelage in many specimens, but lateral line narrow or absent in others (e.g., AMNH 266542, CM 76844, MNHN 1986.285). Superciliary and genal vibrissae extending behind pinnae when laid back alongside head, but mystacial vibrissae consistently shorter, not extending much if at all behind pinnae on properly made-up skins. Dorsal surface of manus and pes covered with short pale fur, often with indistinctly darker markings over central metapodials; claw of pedal digit I extending less than one-half length of phalange 1 of adjacent digit II; claw of pedal digit V extending no more than three-fourths length of phalange 1 of adjacent digit IV; hypothenar (lateral metatarsal) plantar pad small but distinct in some specimens, indistinct or absent in others. Tail about as long as, or a little shorter than, combined length of head-and-body; unicolored (dark above and below), rarely indistinctly paler ventrally at base (e.g., AMNH 266548); with large caudal scales (15–18 rows/cm near the base of the tail in nine fluid specimens) forming coarse and conspicuous annulations.

Skull with very short rostrum flanked by moderately deep zygomatic notches; interorbital region broad and usually strongly convergent; supraorbital beads well developed, often projecting as small shelves over posterior orbits and continuing onto braincase as low temporal crests; braincase relatively narrow in most specimens and not conspicuously inflated. Incisive foramina relatively long (averaging about 65% of diastemal length) and narrow, usually with more-or-less parallel lateral margins; zygomatic plate relatively broad; carotid circulation with well-developed supraorbital ramus of stapedial artery (occupying squamosal-alisphenoid groove and sphenofrontal foramen; pattern 1 of Voss, 1988); subsquamosal fenestra often very small and sometimes occluded by internal flange of petrosal; auditory bullae usually globular, with spherical tympanic capsules and abruptly constricted bony eustacian tubes.

First maxillary molar typically with very narrow, undivided anterocone; anteroloph of M1 usually indistinct (fused with anterocone) even on unworn teeth; mesoloph of M1 very prominent (mesoloph/mesostyle complex sometimes rivalling paracone and/or metacone in size) and disproportionately connected to hypocone by median mure, arising anterolabially from that cusp in an uninterupted curve on most unworn teeth; principal labial cusps (paracone, metacone) distinctly smaller than lingual cusps (protocone, hypocone); all principal cusps quickly worn down to enamel loops, not persisting as distinctly tubercular elements in most adult dentitions.

Karyotypes: Two specimens of Neacomys paracou karyotyped by M. Tranier from “Cayenne, Rte. de Cacao”, French Guiana (MNHN 1983.419, 1983.420) and another from Saül (MNHN 1983.418) had diploid counts of 2N=56 chromosomes (recorded on skin tags). The same diploid counts were obtained by E. Bach from chromosomal preparations of two Venezuelan specimens (AMNH 257270, MNHLS 8064) that were part of the series from San Ignacio Yuruaní originally misidentified by Voss (1991: table 23) as N. tenuipes.

Variation: Samples that we refer to Neacomys paracou are remarkably similar in morphological characters across a very large geographic range. The most metrically divergent series consists of three Venezuelan examples (from San Ignacio Yuruaní in eastern Bolívar state; table 22), which have longer hindfeet, slightly larger molars, and slightly longer rostrums than most Paracou specimens; broad overlap between these samples in most measured dimensions (together with the lack of other distinguishing characters), however, suggest that they are not specifically distinct. Specimens from two other geographically outlying samples (in the Brazilian states of Amazonas and Pará; measurements not tabulated) have less well-developed supraorbital beads than most topotypical specimens but do not appear to be morphometrically divergent or remarkable in other qualitative respects.

Comparisons: Neacomys paracou requires close comparisons with three other small species from northern South America, N. dubosti, N. guianae, and N. tenuipes. Of these, dubosti, guianae, and paracou occur in the Guiana subregion of Amazonia, where they have been collected sympatrically in all pairwise combinations (but never all three together). Neacomys tenuipes does not occur in the Guiana subregion, but its nomenclaturally crucial status as the oldest named species of small spiny mice compels us to include it in this comparative analysis.

Although all small species of spiny mice are similar in most external characters, the morphology of the hindfoot and the tail can be used in combination to provide tentative field identifications. In tenuipes and dubosti, the outer pedal digits (dI and dV) are relatively long: the claw of dI extends almost or fully half the length of the first phalange of dII, and the claw of dV extends almost or fully to the end of the first phalange of dIV. By contrast, the claw of dI does not extend much beyond the base of the first phalange of dII in guianae and paracou, and the claw of dV in these two species does not extend more than about half the length of the first phalange of dIV. These proportional differences are easiest to see in fresh material, or in fluids, where the digits can be straightened and freely manipulated; in carelessly made-up skins (with twisted or bent toes), however, digital proportions can be hard to evaluate.

In specimens measured by the American method (total length and tail length [dorsal flexure to fleshy tip] measured in the field; head-and-body length calculated by subtraction), the tail is consistently longer than the head-and-body by a substantial amount (the ratio LT/HBL averaging about 115%) in tenuipes, but in the other three species the tail is about the same length as the head-and-body, on average. Unfortunately, the ratio of tail to head-and-body cannot be compared meaningfully among specimens measured by different protocols, and many specimens of Neacomys are captured with bobbed tails. Nevertheless, the contrast in tail length is visually obvious when comparing series of skins of tenuipes with those of dubosti, guianae, and paracou.

The tail is distinctly bicolored (dark above, pale below), at least near the base, in most specimens of tenuipes, but most specimens of dubosti and paracou have unicolored (all dark) tails. Unfortunately, the few available skins of guianae are too variable to characterize the species with confidence for this trait: whereas the type and two Surinamese specimens have tails that are distinctly bicolored at the base, four other specimens have indistinctly bicolored or unicolored tails.

In visual comparisons of dried skins, the caudal scales appear to be larger and to form coarser annulations in paracou than in the other three species, but this difference is hard to quantify because tails are stretched to varying degrees when skins are stuffed. Although we counted the number of scale rows per centimeter near the base of the tail on fluid specimens, only a few fluids were available for most species. Nevertheless, our data suggest that this character might be useful for field identifications: whereas nine adult fluid specimens of paracou from the type locality had 15–18 (mean = 16) scale rows/cm, the fluid holotype of dubosti had 21 rows/cm. We were not able to examine any fluid specimens of tenuipes or guianae.

We assessed species differences in cranial morphology by visual comparisons supplemented by measurements of representative samples (table 20). Although most statistical details of our morphometric analyses are necessarily omitted from this faunal report, the scatter plots in figure 40 depict patterns of multivariate divergence revealed by six pairwise principal components ordinations, and the matrix in table 21 summarizes the outcome of a linear discriminant function analysis with all species treated simultaneously. Both methods indicate that these taxa are craniometrically distinct in all pairwise combinations with the exception of dubosti and tenuipes, which have partially overlapping multivariate distributions. The following are the principal points of quantitative and qualitative cranial difference based on our visual and analytic comparisons.

When samples of skulls are lined up in comparative series, each species has a distinctive dorsal gestalt as a consequence of taxonomic variation in four anatomically adjacent and visually juxtaposed structures: (1) The rostrum varies in absolute and relative length, being longest on average in tenuipes and shortest in paracou; the rostrum is of intermediate length in guianae and dubosti. (2) The zygomatic notches (dorsal emarginations of the maxillary bone flanking the base of the rostrum) vary in depth as a correlate of variation in the width of the zygomatic plate; the zygomatic notches are deepest and the zygomatic plates widest in paracou, whereas tenuipes, guianae, and dubosti have shallower zygomatic notches and correspondingly narrower zygomatic plates. (3) The interorbital region is relatively narrow, and the supraorbital beads are relatively weakly developed (seldom produced as shelf-like projections over the posterior orbits) in tenuipes and guianae; in these species, the modal interobital morphology could be described as weakly convergent (fig. 38A). By contrast, dubosti and paracou have relatively broader interorbits, and the supraorbital beads are more frequently developed as projecting shelves; their modal interorbital morphology is strongly convergent (fig. 38B). (4) The braincase is relatively broader and more inflated in dubosti than in any of the other three species.

Taxonomic variation in other quantitative and qualitative cranial traits also contributes to species recognition. The incisive foramina of paracou are longer in relation to the diastema (LIF averaging about 65% of LD) than those of tenuipes, guianae, and dubosti (in which this proportion averages about 56–57%), and subtle taxonomic differences in the shape of these diastemal perforations are also present. Thus, the foramina are relatively narrow in proportion to their length and usually have subparallel lateral margins in paracou, whereas the foramina are relatively broader with more convex or anteriorly convergent lateral margins in the other species.

Although the shape of the auditory bullae exhibits individual variation within most population samples, the bullae of paracou are more consistently globular in form, each consisting of a roughly spherical tympanic capsule that is usually abruptly constricted anteromedially to form narrow bony eustacian tubes (fig. 42B). By contrast, the bullae of tenuipes, guianae (fig. 42A), and dubosti are usually flask-shaped, each tapering gradually from the tympanic annulus to a relatively broader eustacian tube. Insufficient in itself for species diagnosis, this character is nevertheless useful for corroborating identifications when used in conjunction with other traits.

The morphology of the first maxillary molar also differs significantly among the four species (fig. 43). In tenuipes, M1 is more-or-less rectangular in outline because the anterocone is almost as broad as the paracone-protocone cusp-pair behind it. In many specimens with unworn dentitions (especially from the central Andean cordillera of Colombia; e.g., FMNH 70126, USNM 499555), the anterocone is deeply divided into anterolingual and anterolabial conules by an anteromedian flexus, and the anteroloph is large and distinct. The occlusal organization of the tooth is strikingly symmetrical, with subequal lingual and labial cusps that remain persistently tubercular with moderate wear. The mesoloph is a slender crest of enamel, perpendicular to the long axis of the tooth, that forms a Y-shaped junction with the median mure and lacks a disproportionate connection to either of the two principal lingual cusps (protocone and hypocone).

The modal morphology of M1 in guianae and dubosti is essentially similar to that seen in tenuipes, but differs in certain details. Thus, the anterocone is undivided and usually distinctly narrower than the protocone-paracone cusp-pair, giving the tooth a less rectangular and more egg-shaped outline, and the anteroloph is seldom distinct (the anteroflexus usually persisting, if at all, only as a small internal fossette). There is also a tendency, that is more marked in some specimens than in others, for the labial cusps (paracone and metacone) to be reduced in size relative to their lingual counterparts (protocone and hypocone), resulting in a less symmetrical occlusal design. In addition to these shape differences, the toothrow is absolutely shorter in guianae than in either tenuipes or dubosti.

The typical morphology of M1 in paracou differs in several respects from that seen in the other three species. The tooth is visibly narrower in relation to its length, on average, and the undivided anterocone is usually much narrower than the protocone-paracone cusp-pair behind it. In most specimens, this tooth exhibits a striking departure from bilateral symmetry, with the labial cusps being much reduced in size relative to their lingual counterparts, and with an enlarged mesoloph that runs obliquely and disproportionately from the hypocone to the labial cingulum. In addition, the principal cusps are not persistently tubercular because they are quickly worn down to enamel loops; thus, even moderately worn dentitions are essentially flat-crowned, lacking any significant occlusal relief.

Remarks: Specimens that we examined and determined to be Neacomys paracou include at least some of the material previously identified as N. guianae by Anthony (1921a), Husson (1978), Guillotin (1982), Malcolm (1990), and Voss and Emmons (1996: appendices 4 and 5). It is probable that other literature records of N. guianae are also based partly or entirely on specimens of N. paracou, which appears to be the commonest and most widespread of the three Neacomys species now known from the Guiana subregion of Amazonia.

Specimens Examined: Brazil—Amapá, Serra do Navio (USNM 461570, 461577, 461578, 461583, 461585, 461586, 461596, 461597–461600, 461603, 461605, 461608, 461609); Amazonas, 80 km N Manaus (USNM 580008–580011); Pará, Cachoeira Porteira (USNM 546277–546281). French Guiana—Arataye (MNHN 1986.284–1986.286, 1986.870–1986.875), Cacao (MNHN 1986.536, 1986.539), Cayenne (MNHN 1983.419, 1983.420), Mont St.-Michel (MNHN 1983.411), Paracou (AMNH 266542, 266544–266546, 266548–266550, 266552–266557, 267570, 267572, 267574–267577; MNHN 1995.1013–1995.1022 [type series]), Saül (MNHN 1983.405, 1983.418, 1983.421, 1983.424), St.-Eugène (MNHN 1998.1834, 1998.1836–1998.1838). Guyana—“River Supinaam” (BMNH 10.5.4.22); Barima-Waini, Baramita (ROM 100947); Cuyuni-Mazaruni, Kartabo (AMNH 42893, 64146, 64147, 142821, 245037); Potaro-Siparuni, Kurupukari in Iwokrama Reserve (BMNH 1997.44, 1997.46); Upper Takutu-Upper Essequibo, Nappi Creek in Kanuku Mountains (ROM 31760). Surinam—Brokopondo, 18.5 km W Afobakka (CM 54016), Locksie Hattie on Saramacca River (FMNH 95642, 95643); Marowijne, Oelemarie (CM 76838, 76844), Perica (CM 76845). Venezuela—Bolívar, San Ignacio Yuruaní (AMNH 257269–257271).

Field Observations: All of our inventory records of Neacomys paracou are based on collected specimens (N = 29), of which 17 (62%) were taken in Sherman traps, 8 (28%) were taken in pitfalls, 3 (10%) were taken in Victor rat traps, and 1 was shot. Most trapped specimens were found at dawn, but a single specimen was found in the late afternoon in a pitfall that had been checked earlier on the same day. Fourteen specimens (48% of the total) were shot or trapped in secondary vegetation, 10 (35%) were trapped in well-drained primary forest, and 5 (17%) were trapped in swampy primary forest. All specimens were collected at or near ground level. Of the 18 Sherman- or Victor-trapped specimens, most were taken in dense undergrowth near woody shelter: under logs (6 specimens), on top of logs (3), among stilt roots (3), beside logs (2), under piled branches (2), inside a hollow log (1), and at the base of a tree (1).

Nectomys melanius Thomas Figures 44, 45B, 46B, 47B

Voucher Material: MNHN 1998.680, 1998.681. Total = 2 specimens.

Identification: The genus Nectomys was last revised by Hershkovitz (1944), who recognized all of the material he examined as belonging to one or the other of two polytypic species assigned to different subgenera, Nectomys (Nectomys) squamipes (Brants) and N. (Sigmodontomys) alfari (J. A. Allen). Current usage (summarized by Musser and Carleton, 1993), however, recognizes Sigmodontomys and Nectomys as full genera, the former with two valid species (S. alfari and S. aphrastus) and the latter with three (N. palmipes, N. parvipes, and N. squamipes). Only Nectomys (sensu stricto) is known to occur in Amazonia, Sigmodontomys being restricted to trans-Andean and Venezuelan coastal rainforests (Voss and Emmons, 1996: table 1).

Nectomys squamipes melanius was originally described by Thomas (1910: 185–186), who considered it “the Guianan representative of the Brazilian water rat, N. squamipes, but … distinguishable by its darker dorsal color and smaller skull and teeth.” Thomas's account was based on a small series of specimens from Guyana and Surinam, but additional material identified as melanius was subsequently described by Hershkovitz (1944) and Husson (1978). Whereas both Hershkovitz and Husson followed Thomas in treating melanius as a valid subspecies of N. squamipes (Brants, 1827), Tate (1939) synonymized melanius with N. s. palmipes, a taxon that was originally described (as a full species) by Allen and Chapman (1893) from Trinidad. Petter (1979) subsequently described N. parvipes from French Guiana and compared it with sympatrically collected material that he identified as N. s. melanius. The identification of our voucher material therefore involves each of the three species of Nectomys regarded as valid by Musser and Carleton (1993).

All of the French Guianan material we examined agrees closely with the descriptions of Nectomys squamipes melanius provided by Thomas (1910), Hershkovitz (1944), and Husson (1978). The French Guianan holotype of Nectomys parvipes, raised in the laboratory from a wild-caught nestling (Petter, 1979), appears to be no more than an unusually small individual (table 23), perhaps stunted by an inadequate diet or other captive conditions. We examined this specimen (MNHN 1979.345) and determined that none of its qualitative characters diverge from the range of variation exhibited by other specimens of French Guianan Nectomys with substantially larger measurements. Although the range of variation in molar length (LM) in our French Guianan series is considerable (5.9–7.1 mm), there is no hint of bimodality in the frequency distribution of this measurement, and there is no correlated variability in other characters to suggest that our sample is composite. French Guianan skins are not quite as dark, on average, as topotypical melanius from Guyana, but this appears to be the only point of external difference. Although measurements of Guyanese exemplars suggest that there may be a modest east-to-west increase in average molar size in this taxon (table 23), our side-by-side comparisons of French Guianan and Guyanese specimens revealed no qualitative craniodental differences. Based on specimens we examined, the same phenotype apparently extends westward into the Venezuelan state of Amazonas, and southward into the Brazilian state of Pará on the north bank of the Amazon.

We provisionally recognize Nectomys melanius as a distinct species based on geographical, morphological, and cytogenetic comparisons with both of the taxa that have previously been considered to be senior synonyms (table 24). From N. palmipes, its nearest neighbor, melanius differs conspicuously in diploid chromosome counts (2N = 16–17 versus 2N = 52–56; references in footnotes to table 24) and in several morphological traits that can be used to identify museum specimens unaccompanied by karyotypes. (1) Whereas the lateral margins of the nasal bones taper gradually from front to back in melanius without a sharp change in angle at the premaxillary-frontal suture, the nasals of palmipes are more abruptly constricted behind the premaxillae (fig. 44) in most of the specimens we examined.12 (2) The interparietal bone is a shallow and wide element in melanius (fig. 45B) versus deeper and narrower in palmipes (fig. 45A), a difference that is correlated with a marked dorsolateral expansion of the exoccipital in the latter species. (3) The nasolacrimal capsules on the sides of the rostrum are mostly exposed to lateral view in melanius (fig. 46B), but these structures are partially concealed by the zygomatic plate in palmipes (fig. 46A). (4) The tegmen tympani is usually inconspicuous in melanius (fig. 47B), but a large anterior process of the tegmen tympani is always present in palmipes (fig. 47A).

Based on museum specimens that we sorted by these four morphological characters, Nectomys palmipes occurs throughout the island of Trinidad, where the type (AMNH 5928/4658) was collected at Princestown. This species also occurs on the adjacent Venezuelan mainland, from which we examined several specimens including the type of Nectomys squamipes tatei Hershkovitz (1948a). Collected at San Antonio in the Venezuelan state of Monagas, the type of tatei (AMNH 69899) is craniodentally indistinguishable from Trinidadian material, as are two additional specimens from Monagas (AMNH 142608, USNM 415009); all were collected within the mapped distribution of karyotyped individuals with 2N = 16–17 chromosomes (in the states of Anzoategui, Delta Amacuro, Monagas, and Sucre; Barros et al., 1992). However, a single specimen with the same morphological characters (AMNH 16964) is from El Llagual (ca. 7;dg25′N, 65;dg10′W) in the northern part of Bolívar state. Because material from southern Bolívar (e.g., AMNH 75634, 75635, 130733, 130784) is unambiguously referable to melanius, specimens of water rats from intermediate localities in that state should be examined carefully to determine whether melanius and palmipes (the latter including tatei as a subjective junior synonym) are parapatrically or sympatrically distributed in eastern Venezuela.

Nectomys melanius closely resembles N. squamipes in all of the morphological traits by which N. palmipes differs from both (table 24), but we are persuaded by the karyotypic data and breeding experiments reported by Bonvincino et al. (1996), which suggest that the water rats of southeastern Brazil represent a distinct species from Amazonian populations. Based on the restricted type locality of squamipes (São Sebastião, São Paulo state; see Hershkovitz, 1944), this would appear to be the correct name for Bonvincino et al.'s southeastern Brazilian taxon. We agree with Patton et al. (2000) that aquaticus Lund (type locality: near Lagoa Santa, Minas Gerais) and olivaceus Hershkovitz (type locality: Therezopolis, Rio de Janeiro) are probable synonyms of N. squamipes, and that this species probably extends into northern Argentina. The observations of Peters (1861) and Hershkovitz (1944), together with our own examination of southeastern Brazilian material suggest that the hindfeet of N. squamipes usually have six plantar tubercles, whereas the hindfeet of most specimens of N. melanius lack a distinct hypothenar (lateral tarsal) pad.

Patton et al. (2000) used the name apicalis Peters (1861) for western Amazonian populations of Nectomys with low diploid numbers (2N = 38–42 chromosomes), large teeth (LM = 7.0–7.4 mm), and deep-narrow interparietals.13 Based on the difference in diploid numbers alone, it seems unlikely that apicalis and melanius intergrade in western Amazonia (contra Hershkovitz, 1944: 51–52), but without having seen the type of apicalis and without undertaking an extensive analysis of morphological variation in western Amazonian populations of Nectomys, we are unable to rule out this possibility.

Remarks: It seems probable that Nectomys rattus, originally described by Pelzeln (1883) based on a single immature specimen collected at Marabitanas (0;dg58′N, 66;dg51′W) on the upper Rio Negro in the Brazilian state of Amazonas, is a senior synonym of melanius. Although we have not seen Pelzeln's problematic type,14 the geographic proximity of Marabitanas to Venezuelan localities from which we have examined specimens referable to melanius tends to support this synonymy. Clearly, a comprehensive revision of Nectomys based on first-hand examination of relevant types and a critical analysis of morphological variation among the many hundreds of museum specimens now available for study will be essential for resolving this and other taxonomic enigmas. In the meantime, N. melanius is the oldest available name that we can confidently apply to the material at hand.

Other Specimens Examined: Brazil—Pará, Cachoeira Porteira (USNM 546290, 546291). French Guiana—Arataye (MNHN 1981.162), Awara (MNHN 1986.271), Cacao (MNHN 1979.345 [holotype of parvipes], 1981.1303, 1981.1304, 1986.272, 1986.273), Cayenne (MNHN 1970.224, 1981.1298, 1981.1299, 1986.274, 1986.275), Rorota (MNHN 1981.1305), Ouanary (MNHN 1981.1297), Piste St.-Élie (MNHN 1981.184), Saül (MNHN 1980.407, 1981.1296, 1986.270). Guyana—Cuyuni-Mazaruni, Kartabo (AMNH 42332, 42333, 42882, 42885, 42891, 64140), Oko Mountains (USNM 46216); Upper Demerara-Berbice, Rockstone (AMNH 34651). Venezuela—Amazonas, Acanaña (USNM 406237), Boca Mavaca (USNM 374662, 374664, 374665, 406062, 406063, 406233), Cerro Neblina Base Camp (USNM 560824), Esmeralda (AMNH 77303), Mt. Duida (AMNH 77306), Río Casiquiare (AMNH 78080), San Carlos de Río Negro (USNM 560650); Bolívar, Auyantepui (AMNH 130733, 130784), Mt. Roraima (AMNH 75634, 75635).

Field Observations: Both of our vouchers of Nectomys melanius were collected by O. Henry, whose field notes indicate that one was trapped “près de la crique” on 4 November 1989, and the other “sur la piste” on 20 April 1990.

Neusticomys oyapocki (Dubost and Petter) Figures 48, 49B, 49D

Voucher Material: AMNH 267597; MNHN 1995.992. Total = 2 specimens.

Identification: Our two vouchers and another specimen subsequently collected at nearby St.-Eugène provide an opportunity to reevaluate the characters of this obscure taxon. Previously known from a single specimen (MNHN 1977.775) from Trois Sauts in southeastern French Guiana, Daptomys oyapocki was initially diagnosed only by the absence of upper and lower third molars, and by the small size of its remaining cheekteeth (Dubost and Petter, 1978). Other distinctive attributes of the holotype were reported by Voss (1988), who treated Daptomys as a junior synonym of Neusticomys. The new material closely resembles the type and corroborates the status of N. oyapocki as a valid species.

The Paracou examples are both males. The smaller specimen (MNHN 1995.992) we judge to be subadult because of its uniformly blackish pelage, undescended testes, conspicuous metapodial epiphyses, and open basicranial sutures; although its molar dentition (see below) is completely erupted, the animal is obviously immature. The larger specimen (AMNH 267597; figs. 48, 49B, 49D) appears to be a young adult; its pelage is also dark, but the color is distinctly brownish and finely ticked with tawny-banded hairs, the testes are scrotal, the metapodial epiphyses are inconspicuous (but perhaps not completely absorbed), and the basicranial sutures are closed (although not completely fused). In fact, AMNH 267597 seems to be nearly the same age as the holotype and compares with it closely in external and cranial dimensions (table 25). The specimen from St.-Eugène is a fully adult male, with fur like the larger Paracou specimen, well-worn molars, and fused basicranial sutures.

Like the holotype, all of the three new specimens of Neusticomys oyapocki lack M3/m3, and the remaining molars are small by comparison with their homologs in the only other Guianan congener, N. venezuelae (fig. 49). As noted by Voss (1988), loss of the third molar is accompanied by morphological changes in the second, now the most posterior element in the toothrow. In all specimens of N. oyapocki, the hypocone/metacone lobe of M2 is greatly reduced by comparison with that of N. venezuelae, and there is no trace of a posteroloph.

Neusticomys oyapocki can be distinguished from other lowland congeners (formerly classified as Daptomys; fig. 50) by additional characters: (1) The ears and feet of N. mussoi and N. peruviensis are cream-colored and contrast with the brownish dorsal body pelage, but the ears and feet of N. oyapocki and N. venezuelae are dark brown and do not contrast with the dorsal fur. (2) In N. venezuelae and N. mussoi, the posterior edge of the inferior zygomatic root (zygomatic plate) lies above or just anterior to the anterocone of M1; in fully adult examples of N. oyapocki and N. peruviensis, however, the posterior edge of the inferior zygomatic root is located well anterior to the toothrow (fig. 48; also see illustrations in Musser and Gardner [1974], Voss [1988], and Ochoa and Soriano [1991]). (3) A small orbicular apophysis of the malleus is present on the type (and only known fully adult specimen) of N. peruviensis, but this structure is absent in both N. oyapocki and N. venezuelae; the character has not been described or illustrated for N. mussoi. Table 26 summarizes these and other relevant comparisons.

Two peculiarities of the holotype of Neusticomys oyapocki noted by Voss (1988) are apparently not diagnostic for the species. The type lacks masseteric tubercles, the bony processes from which M. masseter superficialis originates in ichthyomyines, but a distinct masseteric tubercle is present at the base of the inferior zygomatic root on each side of the skull in the larger Paracou specimen (AMNH 267597) and in MNHN 1995.3234 (from St.-Eugène). The type of N. oyapocki also appears to have an unusually narrow interorbital constriction by comparison with both of the other conspecific adults at hand, and with specimens of N. venezuelae (table 25).

Other Specimens Examined: French Guiana—St.-Eugène (MNHN 1995.3234), Trois Sauts (MNHN 1977.775 [holotype]).

Field Observations: Both of our specimens of Neusticomys oyapocki from Paracou were taken in pitfall traplines in primary forest. The first example (AMNH 267597) was collected on 14 August 1993 about 5 m from a small (ca. 1.4 m wide), shallow (ca. 20 cm deep), clear, sandy-bottomed stream; the habitat at this site is perhaps best characterized as moist creekside forest on level sandy soil (fig. 51). The second animal (MNHN 1995.992) was taken on 8 September 1993 at approximately the same distance from a slightly smaller stream, but on sloping, well-drained ground. The remains of small crabs (fig. 52) found along other streams in our study area suggest that this species is not uncommon locally, but intensive trapping with Victors and Tomahawks set at streamside and baited with crabs produced no additional specimens.

Oecomys Thomas

For most of the last three decades, Neotropical mammalogists have identified small specimens of Oecomys (formerly considered a subgenus of Oryzomys) as O. bicolor and large specimens as O. concolor following Hershkovitz (1960). Unpublished revisionary research, however, suggests that at least 13 valid species of Oecomys are represented among the many nominal taxa that Hershkovitz lumped into bicolor and concolor (see Musser and Carleton, 1993). Four morphologically diagnosable species are known to occur in French Guiana (table 27), of which two are represented among our vouchers. Pending the publication of a comprehensive revision of this difficult genus, we offer preliminary descriptions and diagnostic comparisons of both Paracou species to document our identifications.

Species of Oecomys are small to medium-sized murids, ranging in average adult body weight from about 20 to 60 g. The dorsal fur is soft (not spiny) and, in adults with fresh pelage, usually some shade of reddish brown. The ventral fur can be either self-colored (pure white) or gray-based with a superficial wash of white, buff, or orange. The mystacial vibrissae are long, extending well behind the pinnae when laid back alongside the head. The dorsal pelage of the hindfeet is sometimes indistinctly darker over the metatarsals than on the digits, but sharply defined metatarsal spots or bands are absent; the plantar surface of the hindfoot is either unpigmented (whitish in preservative, pink in life, brown or amber in dried skins) or lightly pigmented (grayish), but apparently never blackish. Structurally, the hindfeet are short and broad, with large plantar pads and semi-opposable fifth digits (fig. 53B). Tails are unicolored in most species (dark above and below), and they are usually longer than the combined length of head-and-body; a terminal tuft of long hairs is present in some, but not all species.

Amazonian species of Oryzomys are sometimes misidentified in the field as Oecomys (and vice versa), but differ externally by their much shorter mystacial vibrissae (not extending behind the pinnae), distinctive hindfeet (fig. 53A), and shorter tails that are often bicolored (at least basally) and never have terminal tufts of long hairs (see the account of Oryzomys below for more detailed descriptions of external traits).

Species of Rhipidomys resemble Oecomys externally in possessing long mystacial vibrissae; short-broad hindfeet with unpigmented soles, large plantar pads, and semi-opposable fifth digits; and long, usually unicolored, tufted tails. However, the hindfeet of Rhipidomys are distinctive, with darker metatarsal markings, larger plantar pads, and relatively longer fifth pedal digits (fig. 53C). In addition, whereas female Oecomys have eight mammae, female Rhipidomys have only six. More detailed comparisons between like-sized species of Oecomys and Rhipidomys that might be confused in the field are provided in the account that follows.

Oecomys auyantepui Tate Figures 53B, 55, 56, 57A, 62B

Voucher Material: AMNH 266560, 266564, 267593, 267595, 267596; MNHN 1995.1027, 1995.1028. Total = 7 specimens.

Identification: Oecomys auyantepui was originally described by Tate (1939) from two specimens collected at 1100 m elevation on Auyantepui in Estado Bolívar, Venezuela. Hitherto regarded as a junior synonym of O. concolor (by Hershkovitz, 1960), O. trinitatis (by Cabrera, 1961), or O. paricola (by Musser and Carleton, 1993), auyantepui is unambiguously diagnosable from other named forms of Oecomys and merits recognition as a distinct species. Distinguishing traits include its predominantly gray-based ventral fur, a distinctly tufted tail, lack of broadly shelved supraorbital margins and postorbital processes, a primitive carotid arterial circulation, presence of an alisphenoid strut, a large postglenoid foramen, complete closure of the subsquamosal fenestra, and a distinctive range of morphometric variation (see table 27 and below). All of the specimens we refer to O. auyantepui are from the Guiana subregion of Amazonia (fig. 54).

In the hand, Oecomys auyantepui is an attractive mouse with soft reddish-brown fur that is much brighter in mature adults with fresh glossy pelage than in juveniles, subadults, or specimens with obviously worn, dull coats. The small ears are covered with a short but macroscopically visible pelage that is colored essentially like that of the head and nape (not contrastingly darker). The ventral fur, superficially whitish, cream-colored, or pale buff, is sharply set off from the reddish-brown fur of the sides and back. The ventral fur is predominantly gray-based in most of the specimens at hand, but the fur of the chin and throat is usually self-colored (all pale), and a few specimens have self-colored fur extending caudally along the ventral midline to the groin. The hindfeet are either covered uniformly with pale buffy hairs or the metatarsus is indistinctly darker than the toes, but a distinct metatarsal band of blackish fur is apparently never present. Undamaged tails are uniformly dark (almost blackish in some specimens) with a terminal tuft of hairs that are distinctly longer (6–10 mm) than the short (<2 mm) hairs on the proximal part of that organ.

The skull (figs. 55, 56) is unremarkable in general aspect, with the short rostrum, shallow zygomatic notches, convergent interorbit, beaded supraorbital margins, wide-long palate, and small bullae characteristic of this oryzomyine genus. The incisive foramina are of average length relative to the diastema (neither very short nor very long by oryzomyine standards), with wide and more-or-less evenly convex lateral margins. The roof of the mesopterygoid fossa is completely bony, with no trace of sphenopalatine perforations in most of the specimens at hand. A large stapedial foramen on the medial surface of the bulla, a translucent squamosal-alisphenoid groove on the internal surface of the braincase, and a distinct sphenofrontal foramen in the rear of the orbit indicate that the pattern of carotid arterial supply is primitive (pattern 1 of Voss, 1988). A robust diagonal strut of the alisphenoid bone separates the foramen ovale accessorius from the buccinator-masticatory foramen on both sides of the skull in most specimens, and the subsquamosal fenestra is invariably absent (fig. 57A; see table 29).

Morphometrically, series of Oecomys auyantepui that we measured from Guyana, French Guiana, and Brazil (Amapá) are remarkably similar (table 28). In fact, the most divergent specimen we examined is the Venezuelan type (AMNH 131156), which is larger than most of the other material at hand but does not differ in any qualitative external or craniodental character. Additional Venezuelan material would be useful to determine whether the type represents a western population characterized by large size, or is merely an unusually large individual.

Oecomys auyantepui is intermediate in size to other congeneric species that occur in the Guiana subregion of Amazonia, two of which (O. bicolor and O. rutilus) are substantially smaller, and four of which (O. concolor, O. rex, O. roberti, O. trinitatis) are larger. Additionally, other Guianan subregion species differ from auyantepui by having entirely self-colored (pure white) ventral fur (bicolor, rutilus), untufted tails (concolor, rex, roberti, trinitatis), more-or-less bicolored tails (trinitatis), broadly shelved supraorbital margins (rex), a derived pattern of carotid arterial supply (concolor), confluent accessory oval and buccinator-masticatory foramina (most bicolor specimens), and/or patent subsquamosal fenestrae (bicolor, concolor, rutilus, trinitatis). Instead, Oecomys auyantepui is morphologically most similar to O. paricola, a species that we do not recognize as occurring in the Guiana subregion.

Oecomys paricola (Thomas, 1904) was originally described on the basis of a very young female specimen (BMNH 4.7.4.63) collected at “Igarapé-Assu” (= Igarapé Açu at 1;dg07′S, 47;dg37′W; Paynter and Traylor, 1991) near Belém. To evaluate the hypothesis that O. paricola and O. auyantepui are synonyms (Musser and Carleton, 1993), we examined the type of paricola and 21 additional specimens collected in the same interfluvial region (south of the Amazon and east of the Rio Tocantins).15 Because it was not our intention to evaluate the taxonomic status of all museum material currently identified as O. paricola, we did not include specimens so determined from other interfluvial regions, some of which exhibit characters not shown by the type.

Oecomys paricola is about the same size as O. auyantepui, and these two species are perhaps indistinguishable in pelage color and external morphology. Cranially, however, two qualitative characters (table 29, fig. 57) permit unambiguous identification. (1) A well-developed alisphenoid strut almost always separates the buccinator-masticatory and accessory oval foramina in auyantepui, but an alisphenoid strut is missing and these foramina are consistently confluent in paricola. (2) Whereas the subsquamosal fenestra is consistently absent in auyantepui, a distinct subsquamosal fenestra (separated from the postglenoid foramen by the hamular process of the squamosal bone) is always present in paricola. Although their close resemblance in other respects suggest that auyantepui and paricola are likely to be sister taxa, we recognize them as valid (diagnosable) species allopatrically distributed north and south of the Amazon.

Other Specimens Examined: Brazil—Amapá, Serra do Navio (USNM 393820, 393821, 394239–394243, 394246–394249, 461521); Amazonas, 80 km N Manaus (USNM 579996–580001). French Guiana—Arataye (MNHN 1986.865; USNM 578015, 578019), Iracoubo (1983.394, 1983.395), St.-Eugène (1994.124, 1995.3235, 1998.1844), Trois Sauts (MNHN 1983.398). Guyana—Cuyuni-Mazaruni, Kartabo (AMNH 64135); Potaro-Siparuni, 5 km SE Surama (ROM 102944, 103051, 103052, 103244, 103288); Upper Demerara-Berbice, 18 km SW Kwakwani (AMNH 269829, 269830), Tropenbos (ROM 103433, 103502). Venezuela—Bolívar, Auyantepui (AMNH 131108, 131156 [holotype]).

Field Observations: All of our definite records of Oecomys auyantepui from Paracou are based on collected specimens. Of these, one was taken on the ground in a Sherman trap, three were taken in Victor snap traps tied to lianas 1.2–1.5 m above the ground (fig. 58), and three were taken in arboreal platform traps 7.2–10.5 m above the ground. Three specimens were trapped in creekside primary forest, two in well-drained primary forest, and two in swampy primary forest. All specimens were found in the traps at or soon after dawn.

Oecomys rutilus Anthony Figures 55, 56

Vouchers: AMNH 266561, 267584, 267586, 267588–267591, 269121; MNHN 1995.1023–1995.1026. Total = 12 specimens.

Identification: Oecomys rutilus was originally described by Anthony (1921b) based on a single specimen collected by W. Beebe at Kartabo, Cuyuni-Mazaruni District, Guyana. Anthony (p. 4) characterized rutilus as “A small, brightly colored species, with very short tail and clear white under parts”, and he remarked that it was quite distinct from another species originally described from Guyana, Oecomys nitedulus Thomas (1910):

Despite Anthony's explicit statement that rutilus and nitedulus were valid sympatric species, Hershkovitz (1960) listed both names as synonyms of Oecomys bicolor, a taxon originally described (Tomes, 1860) from eastern Ecuador. Hershkovitz (p. 539) admitted that he had not seen Anthony's type of rutilus, but remarked that “Judged by the original description, it is a subadult of the same Kartabo series identified by Anthony as nitedulus.”

Oecomys rutilus was listed as a valid species by Musser and Carleton (1993), but no account of its diagnostic characteristics has yet been published to supplement Anthony's preliminary observations. To identify our Paracou material, we examined every available specimen of small Oecomys from the Guiana subregion of Amazonia, including the types of rutilus and nitedulus. In the following account we provisionally accept the hypothesis that nitedulus and bicolor are conspecific (Hershkovitz, 1960; Musser and Carleton, 1993), but we note that this synonymy remains untested by published analyses of character data, and that we have not made a careful study of typical bicolor (from eastern Ecuador). Instead, our taxonomic comparisons are based exclusively on Guianan material.16

The frequency distribution of the length of the upper molar row (LM) for small Oecomys collected in the Guiana subregion of Amazonia (N = 66) is distinctly bimodal (fig. 59), with one peak in the interval 3.2–3.4 mm (including the type of rutilus with LM = 3.26 mm) and another in the interval 3.6–3.9 mm (including the type of nitedulus with LM = 3.79 mm). Although this univariate graph does not of itself show any morphometric discontinuity that could be used to sort specimens into discrete size classes (all frequency intervals between 3.0 and 4.0 mm are occupied), the bimodal pattern clearly indicates that our sample is heterogeneous. Fortunately, variation in other characters is correlated with molar toothrow length and provides unambiguous evidence for species diagnosis. In fact, the specimens we measured of the smaller species, O. rutilus, have molar toothrows shorter than 3.5 mm, whereas our measured specimens of O. bicolor have toothrows longer than 3.5 mm. With larger samples of both species, however, it seems inevitable that the observed ranges of variation in this dimension will eventually be found to overlap.

Oecomys rutilus and Guianan specimens of O. bicolor (= nitedulus) are similar in external appearance: both are small, usually reddish mice with dark, tufted tails and pure white venters (table 27). Although rutilus averages smaller than bicolor in all standard external dimensions (table 30), there is sufficient morphometric overlap between them that no measurement is diagnostic. Most of the material at hand is preserved in fluid, so it is difficult to evaluate Anthony's statement that these species differ in fur color; however, no consistent differences were apparent in the few dried skins we examined. The species difference in tail length mentioned by Anthony (presumably meant to be considered in relation to the head-and-body) is also difficult to assess because we have not measured any specimens of bicolor ourselves (minor differences in measurement methodology can produce substantial artifactual divergence in computed ratios). Instead, two other characters are useful for field identification.

As observed by Anthony, the dorsal fur is longer in Oecomys rutilus than it is in O. bicolor, and the two species also differ in fur texture. In rutilus, the fur averages about 6–8 mm middorsally near the rump, and it feels soft and lax when ruffled because the unresistant hairs return slowly to their normal (unruffled) condition. By contrast, the dorsal fur of bicolor is only 4–5 mm long in most specimens, and because bicolor is the larger species this absolute difference makes the pelage appear relatively much shorter. Also, the short fur of bicolor feels “crisp” to the touch because the stiffer hairs are more resistant to ruffling.

The tuft of hairs at the tail-tip is significantly longer in Oecomys rutilus, averaging 7.7 ± 1.4 mm (observed range: 5–11 mm, N = 19). In O. bicolor, this tuft measured 4–5 mm in all nine specimens we examined with intact tail tips. Since the long hairs of the tail tuft are often exposed to bending and compression in museum trays, it is probable that some of the tufts we measured were broken short, so the mean tuft length determined from fresh specimens with undamaged tails might be larger for both species than the values reported here. Nonethless, the difference is visually conspicuous and is useful for sorting skins in combination with other traits.

Skulls of Oecomys bicolor and O. rutilus differ in size (figs. 55, 56; table 30) and in the usual size-correlated proportions, but the incisive foramina are notably longer relative to the diastema in bicolor than in rutilus, a difference that is not attributable to standard patterns of muroid craniodental allometry. Although bicolor and rutilus are similar in most qualitative osteological traits, they differ significantly in the frequency of occurrence of the alisphenoid strut, an ossification that occurs bilaterally in most rutilus, but is bilaterally absent in most bicolor (table 31). Otherwise, these species are craniodentally similar, both having primitive carotid circulations (pattern 1 of Voss, 1988), beaded but unshelved supraorbital margins, and consistently large postglenoid foramina and subsquamosal fenestrae.

Although Oecomys rutilus and O. bicolor are widely distributed in the Guiana subregion of Amazonia, they have been collected sympatrically only at Kartabo (Cuyuni-Mazaruni District, Guyana) and Les Nouragues (French Guiana). Musser and Carleton (1993) gave the range of rutilus as restricted to Guyana, Surinam, and French Guiana, but material from San Ignacio Yuruaní (Venezuela) and from 80 km N Manaus (Brazil) previously misidentified as bicolor (e.g., by Voss, 1991: table 23; Voss and Emmons, 1996: appendix 7) extend the range of this species east and south of the Guianas proper (fig. 60). Although we have not examined any material of rutilus from outside the Guiana subregion of Amazonia, the small unnamed Oecomys that Patton et al. (2000) reported from the Rio Juruá appears to be similar in some respects and merits close comparison in any future revisionary study.

The material we examined of Oecomys rutilus is remarkably uniform with little indication of significant variation among samples collected at widely separated localities. Thus, measurements of the Guyanese type (AMNH 42910), a mature adult female (not a subadult as conjectured by Hershkovitz, 1960), are all within the range of variation exhibited by French Guianan specimens (table 30). A few measurements of Brazilian and Venezuelan specimens exceed the observed range of variation for homologous measurements of French Guianan material, but the discrepancies are small in all cases.

Other Specimens Examined: Brazil—Amazonas, 80 km N Manaus (USNM 579992–579995). French Guiana—Cacao (MNHN 1983.400), Les Nouragues (AMNH 269822, V-889, −892, −899, −900, −906, −913), St.-Eugène (MNHN 1995.3236, 1995.3237, 1998.1845, 1998.1846). Guyana—Cuyuni-Mazaruni, Kartabo (AMNH 42910 [holotype], 142820); Upper Demerara-Berbice, Dubulay Ranch (AMNH 267745), 18 mi SW Kwakwani (AMNH 269828), Tropenbos (ROM 103482). Surinam—Suriname, Carolinakreek (FMNH 95591). Venezuela—Bolívar, San Ignacio Yuruaní (AMNH 257268, USNM 448576).

Field Observations: Although we saw small reddish mice racing along branches or lianas at night on many occasions, all of our definite records of Oecomys rutilus at Paracou are based on collected specimens. Of our 12 vouchers, 7 (58%) were taken in pitfall traps, 2 (17%) were shot at heights of 4–5 m in trees, 1 (8%) was taken in a Victor snap-trap tied to a liana 2 m above the ground, 1 was taken in a Sherman trap placed on a liana 30 cm above the ground, and 1 was taken in a platform trap 15.9 m above the ground. Six specimens (50%) were taken in well-drained primary forest, 4 (33%) in swampy primary forest, 1 (8%) in creekside primary forest, and 1 in roadside secondary growth. Both shot specimens were taken at night, and all of the other specimens were found in the traps at dawn.

Oligoryzomys fulvescens (Saussure)

Voucher Material: AMNH 267022, 267023; MNHN 1998.673. Total = 3 specimens.

Identification: Although the genus Oligoryzomys (formerly a subgenus of Oryzomys; see Carleton and Musser, 1989) has never been comprehensively revised, several publications have at least partially clarified the species-level systematics of Oligoryzomys in certain regions, notably Paraguay (Myers and Carleton, 1981), Bolivia (Olds and Anderson, 1987), and Central America (Carleton and Musser, 1995). Unfortunately, the Oligoryzomys of northern South America have received no revisionary attention to date. In a preliminary review of the genus, however, Carleton and Musser (1989) hypothesized that a single widespread polytypic species—O. fulvescens—extends from Mexico throughout most of Central America, thence southward into Colombia and northern Ecuador and eastward throughout Venezuela, Guyana, and Surinam. Included as subjective synonyms in a subsequent synopsis of O. fulvescens (see Carleton and Musser, 1995) were costaricensis J. A. Allen, delicatus J. A. Allen and Chapman, navus Bangs, messorius Thomas, tenuipes J. A. Allen, munchiquensis J. A. Allen, lenis Goldman, mayensis Goldman, engraciae Osgood, and pacificus Hooper. As defined geographically by Carleton and Musser (1989), O. fulvescens was not known to occur east of Surinam or south of the Amazon. Instead, populations of Oligoryzomys extending along the entire south bank of the Amazon from near the headwaters of that river to its mouth (near Belém) were identified as comprising a distinct species, O. microtis (J. A. Allen).

The taxonomic status of Oligoryzomys populations in French Guiana (first reported as O. delicatus by Charles-Dominique, 1993) and in the Brazilian state of Amapá (tentatively identified as O. navus by Carvalho, 1962) has yet to be critically evaluated. Because these regions constitute a geographic hiatus between the known ranges of O. fulvescens and O. microtis (as delimited by Carleton and Musser, 1989; see above), we compared our Paracou vouchers with typical material of both species.

Craniodental measurements of our three Paracou specimens (table 32) fall almost entirely within the range of morphometric variation observed by Carleton and Musser (1995) for a large sample of typical Oligoryzomys fulvescens (topotypes and other specimens from Veracruz, Mexico); the few exceptions are Paracou values (of BM1, BIF, BZP, and BB) that only exceed the observed range in homologous dimensions of typical fulvescens by 0.1 mm. Our vouchers are larger than average fulvescens from Mexico, but Carleton and Musser (1995) documented a southward cline of increasing size among their Central American samples; in craniodental measurements, our vouchers more closely resemble Costa Rican and Panamanian populations that Carleton and Musser referred to O. f. costaricensis than they do typical (Mexican) material.

The only noteworthy morphometric contrast between our Paracou vouchers and typical Oligoryzomys fulvescens appears to be the ratio of tail length to head-and-body length. That ratio is about 1.16 in our two specimens with intact tails, whereas the ratio of mean tail length to mean head-and-body length calculated from Carleton and Musser's (1995) data for Mexican fulvescens is 1.31. The same ratio for Costa Rican and Panamanian samples of fulvescens measured by those authors ranges from 1.27 to 1.44. In side-by-side comparisons, Mexican and Central American skins of fulvescens appear visibly longer-tailed than our Paracou vouchers.

Subtle qualitative cranial differences are present between our three specimens and the Mexican and Costa Rican exemplars of fulvescens with which we compared them. For example, the frontal sinuses appear slightly more inflated in the Mexican and Central American samples, producing a noticeably larger swelling behind the lacrimal bone in the front of the orbit than that seen in our French Guianan material. The rostrum also appears relatively longer and more slender, and the upper incisors perhaps less strongly opisthodont in Mexican and Central American specimens than in our vouchers. These are not conspicuous contrasts, however, and their possible taxonomic significance is difficult to assess without a careful study of the many hundreds of museum specimens that are now available from dozens of geographically intermediate localities. In other qualitative characters, including those that have previously been used to diagnose species of Oligoryzomys (e.g., the dorsal projection of the capsular process of the lower incisor alveolus, position of the incisive foramina relative to the toothrows, pelage color and texture), our vouchers appear to be indistinguishable from Mexican and Central American examples of fulvescens.

Confusingly, the Paracou material is also morphometrically similar to typical Oligoryzomys microtis as represented by Allen's (1916a) original series (and several topotypes, table 32) collected by Leo E. Miller on the “Lower Rio Solimoens (fifty miles above mouth)”, a locality that can now be restricted with some confidence to a specific site on the north bank of the river.17 Apparent mensural differences between these very small samples (3 and 10 specimens, respectively) are unimpressive, especially when the considerable distance between collection localities (roughly 1300 km) is taken into account. Length of the molar toothrow is the only nonoverlapping measurement, but the mean sample difference (0.2 mm) is trivial. The few specimens at hand from immediately south of the Amazon (e.g., AMNH 95983, 188964) have toothrow measurements that do overlap those of our vouchers, so very large samples will probably be required to demonstate any significant divergence in this dimension between Oligoryzomys populations on opposite banks of the river, if any such difference in fact exists.

Allen's (1916a) original description of Oligoryzomys microtis emphasized the diagnostic value of pale coloration, relatively small ears, and short tail for distinguishing this species from other congeners. In coloration, however, Allen's material of microtis appears to be indistinguihable from typical fulvescens based on our side-by side comparisons of skins. Small differences in ear size are hard to evaluate without measurements taken in the flesh, which Leo E. Miller (the collector of Allen's specimens) did not record, and our visual comparisons of dried ears between typical material of microtis and fulvescens revealed no obvious size contrast. The reputedly diagnostic short tail of microtis is also hard to assess. Although Miller's field measurements of the type (AMNH 37091) indicate that its tail was slightly shorter than the head-and-body,18 three paratypes (AMNH 37088, 37089, 37097) have an average ratio of tail to head-and-body of 1.20. Because Miller did not indicate whether his measurements were of complete or bobbed tails (AMNH 37096 has an obviously bobbed-tail measurement of 20 mm with no accompanying notation to that effect), and because most of his skins no longer have intact tail-tips (the result of bending and compression in museum trays), it is now impossible to evaluate the ratio of tail to head-and-body length in Allen's series. Other specimens of microtis (sensu Carleton and Musser, 1989), however, are not very short-tailed: one specimen from the south bank of the lower Amazon (AMNH 203400) has a tail:body ratio of 1.17, whereas another (AMNH 188964) has an improbably large ratio of 1.57. Given the usual methodological inaccuracies associated with muroid tail measurements (Howell, 1924), any taxonomic inferences based on this dimension in the small samples at hand would be premature.

We found no obvious qualitative characters to distinguish Oligoryzomys fulvescens (as represented by our Mexican and Central American exemplars) from typical O. microtis, and in view of the negligible morphometric differences indicated by table 32 we are unable to confidently assign our Paracou material to one or the other taxon. Although an obvious implication of the preceding discussion is that fulvescens and microtis are conspecific, it is also possible that multivariate analyses of larger samples could detect morphological discontinuities that are not now apparent; karyological and biochemical comparisons among key geographic populations could likewise contribute to an assessment of these and other phenotypically similar nominal taxa of Oligoryzomys as valid species. Lacking the time to undertake such a critical revisionary study for this faunal report, we simply apply the older name to our Paracou material.

Other Specimens Examined: Brazil—Amazonas, near Manacaparú on north bank of lower Rio Solimões (AMNH 37088, 37089, 37091 [holotype of microtis], 37092–37097, 37157); Pará, Capim (AMNH 188964, 203400), Pôrto de Moz (AMNH 95983), Vilarinho do Monte (AMNH 95984–95986, 95997). Costa Rica—Puntarenas, Cañas Gordas (AMNH 142440–142458, 142490–142495, 142500). Mexico—Veracruz, Jalapa (AMNH 12536/10846–12541/10851, 12543/10853–12549/10959, 12583–12585).

Field Observations: All of our definite records of Oligoryzomys fulvescens at Paracou are based on collected specimens. All three of our vouchers were caught by hand at night in roadside secondary growth (fig. 13), where two were encountered running on the ground through sparse weeds and the third was climbing among the dried leaves of a felled tree.

Oryzomys Baird

The Amazonian species of Oryzomys were recently revised by Musser et al. (1998), whose species-level taxonomy is followed herein. As a group, Amazonian species of Oryzomys are easily distinguished from other sympatric rodents by external characters. These are medium-sized murids, ranging in average adult body weight from about 40 to 80 g. The dorsal fur is soft (not spiny) and varies in color (depending on species, age, and stage of molt) from rich reddish or tawny hues to drab brown. The ventral fur is superficially whitish or whitish gray, contrasting abruptly in color with the dorsal fur (except in juveniles), but the ventral hair bases are always dark gray. The mystacial vibrissae are short, not extending beyond the tips of the pinnae when laid back alongside the head. The dorsal surface of the hindfoot is unmarked by dark metatarsal spots or bands, and the naked plantar surface is pigmented (grayish in life and in fluid-preserved material, blackish on dried skins). The hindfoot (fig. 53A) appears narrow because the three central pedal digits (II, III, and IV) are much longer than the outer digits (I and V); the fifth digit is not semi-opposable (its claw not extending to the end of the first phalange of dIV). Tails are usually at least partially bicolored (at least near the base), average about as long as (never much shorter or longer than) the combined length of head-and-body, and appear quite naked (a sparse caudal pelage is visible only under magnification). Contrasting external characters of Oecomys species, which are sometimes misidentified as Oryzomys (and vice versa) by fieldworkers, are given in the introductory account for that genus (above).

Three species, Oryzomys macconnelli, O. megacephalus, and O. yunganus, occur sympatrically at Paracou as they probably do throughout the Guiana subregion of Amazonia (see range maps in Musser et al., 1998). Whereas adult specimens of O. macconnelli are easily recognized by external characters, accurate field identifications of O. megacephalus and O. yunganus are more difficult. Juvenile Oryzomys, which lack the diagnostic external dimensions and coloration of adults, cannot be reliably identified in the field.

Oryzomys macconnelli Thomas

Voucher Material: MNHN 1998.674–1998.676. Total = 3 specimens.

Identification: Our three Paracou vouchers (collected by O. Henry, see below), together with additional material that we examined from other French Guianan localities, agree with Musser et al.'s (1998: 225–232) description of Oryzomys macconnelli, an identification that we confirmed by direct comparison with Thomas's (1910) type series from the Supenaam River, Guyana. Although the French Guianan material averages smaller than the type series in many dimensions (table 33), most French Guianan specimens are young adults (with lightly worn molars) whereas the type series is composed of older specimens (with more advanced toothwear). In the absence of other noteworthy differences between the two series, most of the observed measurement divergence could be attributed to sample age composition. The longer molar rows of the Guyanese series, however, cannot be attributed to advanced age, and the observation of even longer toothrows in some Venezuelan series (LM averages 5.2 mm in 12 specimens that we measured from Estado Bolívar) suggests that a real east-to-west size gradient may exist among O. macconnelli populations from the Guiana subregion of Amazonia. Western Amazonian samples are even more divergent morphometrically (Musser et al., 1998: fig. 106), and available karyotypes from western Amazonia differ dramatically from those of the single population sampled for chromosomes in the Guianan subregion (op. cit.: fig. 105). In the event that western Amazonian populations currently referred to O. macconnelli merit formal taxonomic recognition, the name O. mureliae J. A. Allen, based on a specimen collected in eastern Colombia, is available (Musser et al., 1998: 278–280). The Paracou population, however, is unambiguously assignable to O. macconnelli, or to the nominate race if a trinomial nomenclature is adopted.

Based on the samples at hand from Paracou and other localities in French Guiana, adult specimens of Oryzomys macconnelli can be readily distinguished in the field from O. megacephalus and O. yunganus by their larger external dimensions (especially hindfoot length: tables 33, 34), brighter pelage colors (redder dorsally and whiter ventrally versus drab brown dorsally and grayer ventrally in megacephalus and yunganus), longer dorsal fur (12–15 mm versus <10 mm in megacephalus and yunganus), sharply bicolored tails that are slightly longer than heads-and-bodies (versus indistinctly bicolored or unicolored-dark and shorter in megacephalus and yunganus), and six plantar pads on the hindfoot (versus five in most yunganus, see below). Oryzomys macconnelli also has a longer rostrum than either of the other species with which it is sympatric in French Guiana, a contrast that is obvious in cranial comparisons (Musser et al., 1998) but can also be seen in living specimens.

Other Specimens Examined: French Guiana—Arataye (MNHN 1983.371–1983.373, 1986.276), St.-Eugène (MNHN 1994.126, 1994.127, 1995.208, 1998.1842–1998.1844), Saül (MNHN 1983.365, 1983.367). Guyana—“River Supinaam” (BMNH 10.5.4.29–10.5.4.34 [type series]). Venezuela—Bolívar, San Ignacio Yuruaní (AMNH 257236–257238; MHNLS 7831, 7836, 7880, 8075, 8076, 8088; USNM 448584–448586).

Field Observations: Our only records of Oryzomys macconnelli at Paracou are based on three specimens trapped by O. Henry, none of which were accompanied by ecological information.

Oryzomys megacephalus Fischer Figure 53A

Voucher Material: AMNH 266494, 266497, 266498, 266501, 266502, 266504, 266508, 266514, 266515, 266518, 266521, 266523, 266525, 266527–266530, 266533, 266535, 266538, 266539, 266541, 267018, 267566; MNHN 1995.999–1995.1010. Total = 36 specimens.

Identification: At Paracou, and apparently throughout most of Amazonia, Oryzomys megacephalus (formerly O. capito, see below) occurs sympatrically with another morphologically similar species, O. yunganus (see range maps in Musser et al., 1998). Both are drab-colored Oryzomys with brownish dorsal fur, whitish-gray ventral fur, and indistinctly bicolored or unicolored-dark tails that average a little shorter than heads-and-bodies. Comparisons of external measurements from our voucher material suggest that French Guianan O. megacephalus have slightly longer tails than sympatric O. yunganus, but the difference in average values for the ratio LT/HBL is small (0.94 versus 0.89) and insufficient for field identification because of overlapping variation (the observed range in this ratio among our vouchers is 0.78–1.03 for megacephalus, 0.69–1.13 for yunganus). Our measurement data suggest no appreciable species difference in absolute or relative size of the hindfoot. On average, megacephalus has somewhat brighter adult dorsal pelage than yunganus, but the difference is subtle and not useful for field identification.

The only external character potentially useful for distinguishing Oryzomys megacephalus from O. yunganus in the field is the number of plantar pads on the hindfoot (for illustrations, see Musser et al., 1998: fig. 17). Almost all specimens of megacephalus have six plantar pads: thenar, hypothenar, and four interdigitals. By contrast, most French Guianan yunganus have five pads because the hypothenar is absent (table 35). One Paracou specimen of megacephalus (AMNH 266527), however, lacks the hypothenar completely on one foot and has only an indistinct hypothenar on the other. Similarly, a few yunganus from Paracou (e.g., AMNH 266495, MNHN 1995.994) have distinct hypothenars on one or both hindfeet. Based on our tabulation of trait frequencies, a conservative approach to identifying these species in the field in French Guiana (classifying animals with six distinct plantar pads on both hindfeet as megacephalus and those with only five pads on both feet as yunganus) would leave an estimated 3% of specimens undetermined (i.e., those with asymmetrical numbers of pads, plus those with indistinct hypothenars on both feet) and would result in an expected error rate of about 5% (mostly yunganus misidentified as megacephalus).

The only truly reliable basis for identifying megacephalus and yunganus is cleaned cranial material, from which diagnostic molar characters can be determined. All of the Paracou vouchers we identify as megacephalus have (1) the long paraflexus and single fossette on M2, and (2) the long hypoflexid, diagonal median murid, and lack of fossetid on m2 described and illustrated by Musser et al. (1998). By contrast, specimens we identify as yunganus have (1) a short paraflexus and two fossettes on M2, and (2) a short hypoflexid separated from a lingual fossetid by a less oblique median murid on m2 (op. cit.). Our multivariate statistical analyses of cranial measurement data from Paracou vouchers identified by these molar traits reveal that specimens of megacephalus have, on average, wider incisive foramina and narrower zygomatic plates than like-sized examples of yunganus. However, measurement variation is too extensive for cranial proportions to be used as the sole basis for species identification at this locality. Although molar measurements differ significantly between sympatric samples of megacephalus and yunganus from some parts of Amazonia (Musser et al., 1998: table 8), these species do not diverge in dental dimensions at Paracou.

Husson (1978) and most other recent authors have called this species Oryzomys capito (Olfers), but Tate (1939) referred Guianan populations to O. laticeps (Lund). Both capito Olfers and megacephalus Fischer are based on Azara's (1801) description of the “Rat Seconde ou Rat à Grosse Tête” from Paraguay. Musser et al. (1998) designated a Paraguayan neotype for megacephalus and explained why this name should replace capito (a junior objective synonym). Musser and his colleagues also diagnosed Oryzomys laticeps as a distinct species restricted to the Atlantic rainforest region of southeastern Brazil.

As understood by Musser et al. (1998), Oryzomys megacephalus occurs throughout Amazonia and extends southward into the Paraná basin of eastern Paraguay. Morphometric, karyotypic, and molecular data summarized by Musser and his colleagues, however, strongly suggest that western Amazonian populations (characterized by large size, diploid counts of 52 chromosomes, and distinctive mtDNA haplotypes) are genetically and evolutionarily distinct from Paraguayan and eastern Amazonian populations (characterized by small body size, diploid counts of 54 chromosomes, and different mtDNA sequences). Although Musser et al. (1998) recognized this dichotomy, they emphasized the difficulty of identifying geographically intermediate samples as belonging to either the eastern or western clades by morphological criteria and provisionally regarded all of their megacephalus-like Amazonian material as conspecific. Patton et al. (2000) subsequently recognized the western Amazonian form as a distinct species, O. perenensis J. A. Allen, a decision with which we concur.

In fact, Oryzomys megacephalus may be composite even in the restricted sense of Patton et al. (2000) because their cytochrome-b sequence analyses (op. cit.: fig. 97) suggest that samples of the small 2N = 54 taxon from north and south of the Amazon form reciprocally monophyletic groups. In the event that these geographic moieties should prove to be diagnosably different by additional criteria, the oldest available name for the northern form (to which our Paracou sample is presumably referable) is velutinus J. A. Allen and Chapman (1893), based on a holotype collected at Princestown, Trinidad.

Other Specimens Examined: French Guiana—Arataye (MNHN 1986.287–1986.293, 1986.295, 1986.296, 1986.298–1986.301, 1986.314, 1986.878–1986.880, 1986.882), Cacao (MNHN 1980.275, 1983.370, 1986.278, 1986.279, 1986.316, 1986.493–1986.499, 1986.501, 1986.506, 1986.511, 1986.514, 1986.531, 1986.533), Camopi (MNHN 1982.600, 1982.620), Cayenne (MNHN 1970.225, 1986.317, 1986.319, 1986.952, 1986.953), “Marais de Kaw” (MNHN 1986.1106, 1986.1107), Rorota (MNHN 1986.954, 1986.955), Piste St.-Élie km 16 (MNHN 1986.884–1986.886), Saül (MNHN 1981.181, 1983.368, 1983.369, 1986.489, 1986.492, 1986.518, 1986.521, 1986.525), Sauts de l'Itany (MNHN 1962.1024–1962.1028), Trois Sauts (MNHN 1981.150, 1982.616–1982.619, 1982.623–1982.627, 1986.283, 1990.908–1990.910).

Field Observations: All of our definite records of Oryzomys megacephalus at Paracou are based on collected specimens. Of our 36 vouchers, 23 (64%) were taken in Sherman or Victor traps set on the ground, 7 (19%) were taken in Sherman or Victor traps tied to lianas 0.3–1.2 m above the ground, 5 (14%) were shot on the ground, and 1 (3%) was taken in a pitfall. Microhabitat notes accompanying 26 specimens shot or trapped on the ground record 18 captures under or beside logs, 4 captures under the roots or buttresses of fallen trees, 2 captures under tangled branches of fallen trees, 1 capture at the base of a buttressed tree, and 1 capture at the entrance to a hollow log. All of the shot specimens were encountered at night, and all of the other specimens were found in traps at or near dawn. Twenty-two specimens (61%) were taken in well-drained primary forest, 3 (8%) in swampy primary forest, 2 (6%) in creekside primary forest, 5 (14%) in primary forest of unspecified character, and 4 (11%) in secondary vegetation. See table 36 and the following account for capture-habitat comparisons with O. yunganus.

Oryzomys yunganus Thomas

Voucher Material: AMNH 266495, 266496, 266503, 266510, 266511, 266513, 266516, 266517, 266520, 266532, 267017, 267567; MNHN 1995.993–1995.998. Total = 18 specimens.

Identification: See the account above for morphological comparisons with Oryzomys megacephalus, the only species with which O. yunganus could plausibly be confused.

Long unrecognized as a member of the Guianan fauna, O. yunganus was recently revised by Musser et al. (1998), who documented the extensive Amazonian distribution of this species by mapping all known collection localities (op. cit.: fig. 14). As noted by Musser and his colleagues, the geographic samples they refered to O. yunganus exhibit considerable divergence in body size. Especially notable are specimens from Guyana, Surinam, French Guiana, and eastern Amazonian Brazil, which are diminutive by comparison with specimens from Venezuela, Colombia, Ecuador, Peru, and western Brazil. For example, the observed range of variation in crown length of the upper molar series (LM) among our vouchers (4.3–4.6 mm; table 34) does not overlap with the observed range of variation among 52 specimens (including the type) that we measured from Colombia, Ecuador, Peru, and Bolivia (4.8–5.6 mm). Additionally, French Guianan samples of O. yunganus differ conspicuously from western Amazonian samples in the frequency of occurrence of the hypothenar pad on the hindfoot (Musser et al., 1998: table 7). The currently accepted provisional hypothesis, that these and other differences among samples currently referred to O. yunganus represent intraspecific geographic variation, merits testing by additional collecting at intermediate localities (op. cit.: p. 109), and by analyzing molecular sequence data from Guianan and western Amazonian populations (as by Patton et al., 2000).

Other Specimens Examined: Bolivia—Cochabamba, Charuplaya (BMNH 2.1.1.39 [holotype]). Colombia—Caquetá, Tres Troncos (FMNH 72036, 72051, 72066); Meta, La Macarena (FMNH 58778, 58779, 87969, 87970); Putumayo, Río Mecaya (FMNH 72067). Ecuador—Pastaza, Río Capahuari (FMNH 43268, 43271), Río Yana Rumi (FMNH 43265). French Guiana—Arataye (MNHN 1986.294, 1986.297, 1986.313, 1986.881, 1986.883), Cacao (MNHN 1986.490), Cayenne and Rorota (MNHN 1986.322, 1986.324, 1986.326, 1986.327, 1986.800–1986.803), Kaw (MNHN 1986.1105). Peru—Cuzco, Hacienda Cadena (FMNH 65704, 66399, 66401, 68630, 68631), Quincemil (FMNH 75242, 75253, 75254, 75257, 75259, 75261–75264, 75272; Huánuco, Chinchao (FMNH 23721, 23722), Hacienda Buena Vista (FMNH 24544, 24547, 24548); Loreto, Río Pastaza (BMNH 54.421, 54.422, 54.425, 54.429, 54.430); San Martín, Moyobamba (FMNH 19376, 19387, 19392), Puca Tambo (BMNH 26.5.3.31–26.5.3.38, 26.5.3.40–26.5.3.42; FMNH 19787); Surinam—Nickerie, Kayserberg Airstrip (FMNH 93284, 93286).

Field Observations: All of our definite records of Oryzomys yunganus at Paracou are based on collected specimens. Our 18 vouchers represent only 16 sampling events, however, because pairs of juvenile individuals were taken in the same trap on two occasions. Of these 16 independent captures, 15 (94%) were in Sherman traps set on the ground, and one was in a pitfall. All of our specimens were found in the traps at or near dawn. Seven captures (44%) were in well-drained primary forest, another 7 were in swampy primary forest, 1 (6%) was in creekside primary forest, and 1 was in primary forest of unspecified character. Microhabitat notes accompanying 15 specimens record 4 captures under or beside logs, 3 captures at the bases of trees, 3 captures in dense undergrowth unsheltered by woody objects, 2 captures under tangled dead branches, 2 captures under fallen palm fronds, and 1 capture among the stilt roots of a standing tree.

Our capture data from Paracou are broadly consistent with specimen counts from other Amazonian localities (summarized by Musser et al., 1998) in suggesting that Oryzomys yunganus is less abundant than O. megacephalus wherever these species occur sympatrically. Although we sometimes caught both species in the same trapline on the same date, statistical comparisons of capture frequencies by habitat (table 36) suggest that O. yunganus prefers moister primary forest habitats (swamp or creekside formations) than does O. megacephalus. The latter species was also trapped above ground level on lianas, and in secondary vegetation, situations in which O. yunganus was not encountered. Future ecological studies of Amazonian rodent communities should test the hypothesis that O. yunganus is a habitat specialist by comparison with O. megacephalus, but we caution that destructive sampling (or molecular typing, as by Lavergne et al., 1997; Steiner et al., 2000) will be necessary in order to obtain reliable taxonomic identifications for this purpose.

Rhipidomys nitela Thomas Figure 53C, 61B, 62A

Voucher Material: AMNH 267021, 267580, 267582, 267583, 267594; MNHN 1995.1011, 1995.1012. Total = 7 specimens.

Identification: The ten nominal taxa of Rhipidomys based on type material collected in the Guiana subregion of Amazonia appear to represent four valid species that can be readily distinguished by external and craniodental characters. Rhipidomys macconnelli de Winton (1900) (including subnubis Tate, 1939) and R. wetzeli Gardner (1989) have long, soft fur and a primitive carotid arterial circulation (pattern 1 of Voss, 1988); they occur in montane and premontane vegetation associated with rocky outcrops of the Pantepui complex (Tate, 1939; Handley, 1976; Gardner, 1989) and neither is known from French Guiana (where such habitats are absent). The other two species have shorter, coarser fur, a derived carotid arterial morphology (pattern 3 of Voss, 1988), and occur in lowland rainforest. One of the lowland rainforest species (represented by the Guianan holotypes of sclateri Thomas [1887], bovallii Thomas [1911b], and aratayae Guillotin and Petter [1984]) is large (HF = 32–36 mm; LM = 5.9–6.8 mm) with gray-based ventral fur, whereas the other (represented by the Guianan holotypes of nitela Thomas [1901], fervidus Thomas [1904], milleri Allen [1913b], and yuruanus Allen [1913b]) is small (HF = 24–28 mm, LM = 4.1–4.8 mm) with (usually) pure white ventral fur.

Although the oldest Guianan name for the large lowland species is Rhipidomys sclateri, Musser and Carleton (1993) and Tribe (1996) treated sclateri as a subjective junior synonym of leucodactylus Tschudi (1844), the type locality of which is in eastern Peru. Only a single specimen, the type of R. leucodactylus aratayae, is currently known from French Guiana.

The oldest Guianan name for the small lowland species is Rhipidomys nitela, but nitela has often been treated (e.g., by Cabrera [1961] and Husson [1978]) as a junior synonym of R. mastacalis (Lund, 1840), and some nominal taxa referable to nitela were originally described as subspecies of R. venezuelae Thomas (1896). Based on our examination of types and other material, we agree with Musser and Carleton (1993) and with Tribe (1996) that nitela, mastacalis, and venezuelae represent three valid species with diagnostic morphological and karyotypic attributes, and with discrete geographic ranges (table 37). Our six Paracou vouchers together with 13 additional specimens subsequently collected at Les Nouragues (by F. Catzeflis and his colleagues from Montpellier) are apparently the only examples known from French Guiana.

For the most part, external and craniodental measurements of French Guianan specimens of Rhipidomys nitela agree closely with those of the Guyanese type series (table 38). Although relative tail length appears to be divergent in the two samples (averaging about 130% of head-and-body length in French Guianan material versus 115% in the type series), methodological artifacts might explain this proportional difference between small series of specimens measured in the field by different collectors. By contrast, measurements of the hindfoot and of the molars (both measured by us) suggest that the populations in question do not differ much, if at all, in size. Because the French Guianan material additionally resembles the type series in qualitative characters, we interpret these samples as representing populations of a single species.

Some of Husson's (1978: table 71) measurements of the hindfeet of Surinamese specimens that he identified as Rhipidomys mastacalis nitela are smaller than any that we or Tribe (1996) observed for this species. If Husson's material was correctly identified, which we do not doubt, it is likely that the feet were simply mismeasured. The single weight datum that Husson tabulated for this species (150 g) was obtained from a female with three near-term embryos and is obviously not comparable with our weights of nonpregnant animals.

All of the specimens that we refer to Rhipidomys nitela (see below) appear to represent a morphologically cohesive taxon that is geographically limited to the Guianan and Southeastern subregions of Amazonia. Tribe (1996), however, recognized two isolated records of R. nitela from outside Amazonia. One of these records, consisting of the type series of R. nitela tobagi Goodwin (1961) from Little Tobago Island, is equivocal in our judgment because the diagnostic presence of a conspicuous caudal tuft (fig. 61B) cannot be confirmed from the two partially decayed fluid specimens of this taxon; possibly, these represent an insular form of the adjacent mainland species R. venezuelae. The other nonAmazonian record of R. nitela mapped by Tribe (1996: fig. 7.6) is improbably isolated on the northern Caribbean coast of Colombia. We examined the voucher in question (MHNG 1706.75, from Bonda, Departamento Magdalena), which differs from R. nitela by its pale bicolored tail and short (6 mm) caudal tuft; in our opinion, this specimen represents another taxon, perhaps allied to R. venezuelae despite its short (4.6 mm) toothrow.

Recently, Anderson (1997) reported Rhipidomys nitela from eastern Bolivia, a considerable range extension that we attempted to confirm by examining his material. Of the two specimens that Anderson cited by number (AMNH 119406, UMMZ 156298), however, we were only able to locate one. That specimen, UMMZ 156298, is an example of Thomasomys resembling T. oreas Anthony (1926). Therefore, as far as we have been able to determine by direct examination of museum specimens, R. nitela appears to be an Amazonian endemic that does not occur west of the north-south zoogeographic axis represented by the Rio Negro and the Rio Madeira.

Species of Oecomys externally resemble Rhipidomys by their large eyes, long vibrissae (extending well behind the pinnae when laid back alongside the head), short-broad hindfeet with semi-opposable fifth digits, and long tufted tails. At Paracou, O. auyantepui and R. nitela are of similar size and might be confused in the field. However, R. nitela and O. auyantepui differ in many external characters that are potentially useful for identification of specimens in hand. (1) Whereas the pinnae of R. nitela are blackish (contrasting in color with the fur of the head) and appear quite naked (a very sparse auricular pelage is visible only under magnification), the pinnae of O. auyantepui are not contrastingly colored and have a visible pelage of reddish-brown hairs. (2) The dorsal body pelage of R. nitela is very short (5–6 mm), somewhat coarse to the touch, and dull grayish-brown, but the dorsal fur of adult specimens of O. auyantepui is much longer (≥10 mm), very soft to the touch, and lustrous reddish-brown in appearance. (3) The ventral pelage of R. nitela is almost entirely self-colored (pure white or cream to the roots of the hairs), except immediately along the flanks (where some hairs have gray bases and white tips); by contrast, the ventral fur of O. auyantepui is mostly gray-based, except on the throat and along the midline (where some hairs are pure white). (4) The hindfeet of R. nitela are always prominently marked by a dark band of brownish hairs that extends from the ankle over all or part of the metatarsus to the base of the toes, which are usually white (pigmented hairs extend onto the proximal phalanges of the middle toes in a few specimens, but the outermost digits are always completely white); by contrast, pedal markings are not conspicuous in most specimens of O. auyantepui, some of which have uniformly pale feet. (5) Adult female specimens of Rhipidomys have six mammae in postaxial, abdominal, and inguinal pairs (muroid mammary loci are illustrated in Voss and Carleton, 1993: fig. 8), but female Oecomys have eight mammae (the additional teat-pair is pectoral).

Extended craniodental comparisons between Rhipidomys and Oecomys are unnecessary for fieldworkers, but it is relevant to note that these genera superficially resemble one another by their shallow zygomatic notches, convergent and beaded supraorbital margins, and pentalophodont molars. Nevertheless, skulls found in stomachs or scat can be easily identified by palatal architecture. The bony palate of Rhipidomys (fig. 62A) is “short” because it does not extend behind the molar rows, and the posterior palatal margin is biconcave because a small median palatal process is present; the posterior palatal pits are small, simple perforations. By contrast, the bony palate of Oecomys (fig. 62B) is “long” (extending behind the molar rows) with a typically arch-shaped posterior margin, and the posterior palatal pits are larger and often more complex. The only French Guianan species of Rhipidomys and Oecomys that are at all likely to be confused, R. nitela and O. auyantepui, can also be distinguished by carotid arterial morphology (pattern 3 versus pattern 1 [of Voss, 1988], respectively), but carotid morphology does not consistently differ between other representatives of these genera.

Other Specimens Examined: Brazil—Pará, Aramanay on Rio Tapajos (AMNH 94810–94813). French Guiana—Les Nouragues (AMNH 269821; V-824, −825, −826, −831, −876, −886, −890, −891, −893, −905, −914). Guyana—Potaro-Siparuni, Minnehaha Creek (AMNH 36331–36336 [type series of milleri]); Upper Takutu-Upper Essequibo, Quatatat (BMNH 1.6.4.81–1.6.4.86 [type series of nitela]). Venezuela—Bolívar, La Unión (BMNH 4.5.7.34, 4.5.7.35 [type series of fervidus]), La Vuelta (BMNH 4.5.7.36), Río Yuruán (AMNH 30727–30735, 30737 [type series of yuruanus]), San Ignacio Yuruaní (AMNH 257273–257275; MHNLS 7845, 7846, 7848–7850, 7891, 7892, 7895, 7896, 7898, 7899, 8072, 8074; USNM 448613–448616, 448618–448623, 448625–448628).

Field Observations: All of our definite records of Rhipidomys nitela at Paracou are based on collected specimens. In 1992, one adult female was shot in the daytime as it perched several meters above ground level in the dark interior of a hollow tree (fig. 63). In 1993, the same tree cavity contained four individuals, of which one adult female and two juveniles were shot and one juvenile escaped. One specimen was captured at ground level in a pitfall, and two others were taken in platform traps 14.5–15.2 m above the ground. All of our vouchers were shot or trapped in well-drained primary forest.

Erethizontidae Coendou melanurus (Wagner) Figures 64, 65, 67, 68, 69, 70A, 71

Only two specimens documented the occurrence of this distinctive porcupine in French Guiana prior to fieldwork at Paracou: the mounted skin of an immature animal from “Guyane” (MNHN 1909.241), and an adult skin-and-skull from St. Laurent du Maroni (MNHN 1909.242). Just a single individual (AMNH 266565) was encountered in our inventory, but simultaneous faunal rescue operations at the Petit Saut hydroelectric dam site resulted in 59 captures (Lemercier, 1998; Vié, 1999), from which three specimens were salvaged as vouchers (MNHN 1997.640, 1997.641, 1999.1080; F. Catzeflis, personal commun.). Evidently, the species is not rare, but cryptic. Because the morphological and geographic limits of Coendou melanurus are not adequately documented in the literature, we redescribe the species below, discuss its morphological variation, compare it with other congeners, and provide new information about its geographic range.

Type Material: Wagner's (1842) original material of Cercolabes melanurus consists of two specimens in the Naturhistorisches Museum Wien collected by Johann Natterer at Barra do Rio Negro (= Manaus), Estado Amazonas, Brazil.19 Both are skins, originally stuffed and mounted for exhibition, with skulls and mandibles subsequently extracted.

In order to clarify the application of Cercolabes melanurus, we select as lectotype NMW 42010, an adult female collected in February 1834. The skull (fig. 64) is that of a fully mature animal, with slightly swollen frontal sinuses and most cranial sutures fused; the cheekteeth are worn flat, but all essential details of the occlusal morphology remain. The squamosal root of the right zygomatic arch is broken, as is the right pterygoid process; both occipital condyles and part of the basioccipital are missing. The skin (fig. 65) is essentially intact, but the tail is partially broken away at the base and secured to the body with thread. The paralectotype (NMW B-1017) is a subadult female with deciduous premolars and unfused cranial sutures. Like the lectotype, the tail of this specimen is partially broken away and tied to the body with thread.

Distribution: Specimens that we examined and literature records that we judge to be reliable suggest that Coendou melanurus ranges throughout the Guiana subregion of Amazonia (fig. 66). Although Emmons (1990, 1997), Alberico et al. (1999), and Tirira (1999) described the range of C. melanurus as extending into western Amazonia, all extralimital records that we investigated were found to be based on misidentified material of other species (see Comparisons, below).

Description: Small long-tailed porcupines (see measurements in table 39) with dorsal pelage composed of short quills more-or-less concealed beneath a thick coat of long blackish fur coarsely streaked with yellow. Quills bicolored (yellowish basally with dark-brown tips), about 3 cm or less in length, densely covering dorsal surface of head, neck, trunk, and upper limbs, but only exposed on cheeks and crown of head (except where fur has come away in patches due to faulty preservation). Dorsal fur long and abundant, concealing quills from crown to rump; individual hairs pale basally (among the quills), but emergent fur black heavily streaked with pale yellow (a mass effect produced by coarse guard hairs with very long yellow tips scattered abundantly among finer hairs that are entirely black or have only inconspicuous pale tips). Ventral surface of body without offensive quills, covered sparsely with short, coarse, brownish hairs (usually frosted with whitish bands or tips) from chin to anus. Face with very short quills on forehead and cheeks but otherwise almost naked, especially around eyes; facial vibrissae consisting of long, black mystacial, genal, and superciliary hairs; submental vibrissae short and black. Pinnae rudimentary and almost naked. Hands and feet covered dorsally with very coarse blackish hairs. Tail probably about as long as head-and-body on average; dorsum of proximal third with body pelage (quills and yellow-streaked black fur), but remainder of tail (except for naked prehensile surface) densely covered above and below with stiff black bristles.

Frontal and nasal sinuses uninflated (resulting in a flattened dorsal profile from midparietal region to nasal tips) or weakly inflated (resulting in a noticeable bulge over the orbits). Rostrum usually short and very broad, not conspicuously excavated laterally for origin of infraorbital muscle; nasals parallel-sided or weakly convergent posteriorly, with rounded posterior margins that extend well behind the premaxillae. Zygomatic arches (viewed from above) usually with rounded lateral deflection at orbits, but sometimes convergent anteriorly with no orbital deflection (e.g., AMNH 266565); jugals moderately expanded dorsoventrally behind maxillary suture. Dorsolateral contours of braincase weakly sculpted by bony scars of M. temporalis in most specimens (but temporalis scars well marked in AMNH 70120), the right and left scars always widely separated (never joined to form a sagittal ridge or crest).

Incisive foramina short and completely contained by premaxillae (e.g., AMNH 94174), or longer and bordered by maxillae posteriorly; left and right foramina recessed in a common fossa and incompletely separated (e.g., AMNH 94174) or completely separated by a stout bony septum and not recessed (e.g., AMNH 142955). Posterior diastema usually distinctly trisulcate; palatal bridge usually without a distinct median keel, or keel weakly developed (never a high crest flanked by deep lateral gutters). Anterior margin of mesopterygoid fossa a broad, blunt wedge penetrating between third molar crowns; bony roof of mesopterygoid fossa usually completely ossified (sometimes with tiny perforations but never large vacuities). Auditory bullae very large and antero-posteriorly elongated; roof of external auditory meatus with conspicuous bony ridge extending from dorsal lip of bulla to malleus.

Variation: This species is extraordinarily uniform in qualitative external characters despite the apparent plasticity of some morphometric and osteological traits. High variability in external measurements (table 39) is common in erethizontids because of the difficulty of handling animals protected by sharp quills; additionally, porcupine feet do not resemble those of other rodents, and some field collectors may have measured the hindfoot erroneously, from the rounded posterior margin of the plantar callosity instead of the heel. As in other caviomorphs, qualitative cranial and dental characters of erethizontids are annoyingly variable. Some differences among the specimens of Coendou melanurus that we examined may be correlated with age and sex, but with miniscule samples from widely scattered localities, interpretation is difficult.

Comparisons: Many faunal accounts and checklists (e.g., Cabrera, 1961; Husson, 1978; Woods, 1993) have treated Coendou melanurus as a subjective junior synonym of C. insidiosus (Olfers, 1818), a taxon endemic to the Atlantic rainforests of southeastern Brazil. The history of this erroneous usage was reviewed by Voss and Angermann (1997), who rediscovered and redescribed the holotype of C. insidiosus, compiled geographic data from other known specimens, and provided diagnostic comparisons with C. melanurus. Morphological differences between these highly distinctive species are here illustrated for the first time (figs. 67–70).

Other species of South American porcupines with long fur are easily distinguished from Coendou melanurus by conspicuous external characters. For example, Coendou vestitus Thomas (1899c) and C. pruinosus Thomas (1905) from Venezuela and Colombia are smaller, shorter-tailed porcupines that have long wire-like bristles mixed among the quills and woolly fur of the dorsal pelage. We have also examined several Ecuadorean specimens previously misidentified as C. melanurus and determined that they represent an undescribed species of the vestitus group; this taxon superficially resembles melanurus because it has long yellow-tipped bristles that contrast with the black-tipped quills to produce a similarly streaked pelage pattern. However, the new species is substantially smaller (HF, 58–59 mm; CIL, 58.8–64.4 mm; MTR, 14.1–15.2 mm), lacks a dense coat of long adult fur, and has more acutely angled mesopterygoid margins that penetrate more deeply between the toothrows (Voss and da Silva, submitted).

Throughout the Guiana subregion of Amazonia, Coendou melanurus is sympatric with another congener, C. prehensilis (see below). Although melanurus is smaller than prehensilis, there is some overlapping variation in weights and linear dimensions from large sympatric samples (Richard-Hansen et al., 1999). Fortunately, the two species are unmistakeable in qualitative external characters (fig. 71). Diagnostic craniodental comparisons between Surinamese samples of melanurus and prehensilis were discussed by Husson (1978), who misidentified his material of the former species as Sphiggurus insidiosus (see above and Remarks, below)

Remarks: Coendou melanurus and other long-furred Neotropical porcupines have often been referred to the genus Sphiggurus F. Cuvier (most recently by Husson, 1978; Concepción and Molinari, 1991; Woods, 1993; Eisenberg and Redford, 1999). However, we agree with Handley and Pine (1992) that Coendou and Sphiggurus are not meaningfully diagnosable based on current knowledge of morphological character variation, and that the latter name should be treated as a subjective junior synonym of the former.

Sphiggurus melanurus Gray (1842), published in the same year as Cercolabes melanurus Wagner, is based on a skin in the Natural History Museum (London) bearing the number 86a on a cardboard tag attached to the hindfoot. This specimen agrees in every essential respect with Wagner's type series and with the other material herein referred to Coendou melanurus. Waterhouse (1848: 425), the first reviser in the sense of the International Code of Zoological Nomenclature (ICZN, 1999), can be considered to have chosen Wagner's name to have precedence over Gray's (a junior subjective synonym).

Specimens Examined: Brazil—“Brazil” (BMNH specimen numbered 86a in J. E. Gray's manuscript catalog [type of melanurus Gray]); Amapá, Serra do Navio (USNM 394732); Amazonas, Manaus (NMW 42010, B-1017); Pará, Lago Claimy on Rio Jhamunda (AMNH 94174). French Guiana—“Guyane” (MNHN 1909.241), Paracou (AMNH 266565), St.-Laurent du Maroni (MNHN 1909.242). Guyana—Cuyuni-Mazaruni, Kartabo (AMNH 70120, 70131, 142955); Demerara-Mahaica, Georgetown (FMNH 17762); Upper Takutu-Upper Essequibo, 25 mi E Dadanawa (ROM 31984), Nappi Creek near Letham (ROM 31683, 31783, 31801). No locality data—(NMW B-1006).

Field Observations: Our single voucher is the only definite record we have of Coendou melanurus from Paracou; surprisingly, none of the forestry personnel whom we interviewed were aware that this species occurred in the area. This specimen was shot at night as it perched 15–20 m above the ground in the subcanopy of well-drained primary forest; its stomach was completely filled with a homogeneous, finely masticated, bright-green pulp. We surmise that the animal had been eating the new leaves of the tree in which it was shot because these had the same color and odor when crushed as the freshly dissected stomach contents.

Coendou prehensilis (Linnaeus) Figure 71

Although we did not directly observe this species at Paracou, we found one of its distinctively large, tricolored quills on a dirt road near our camp. P. Petronelli (personal commun., 1993) told us that he had also found quills of this species lying on the ground in our study area, and that an ocelot (L. pardalis) killed at Paracou by a local hunter several years before our inventory work began had quills of C. prehensilis embedded in its neck and shoulders. The species has occasionally been observed in roadside secondary growth by visiting researchers (G. Dubost, personal commun.), but accurate counts of such observations are unavailable.

Dasyproctidae

The Paracou fauna contains two dasyproctids, one species each of Dasyprocta and Myoprocta, the usual number known to occur sympatrically throughout most of Amazonia. Although Guianan dasyproctids are easily identified in the field and in the museum, their technical names are still controversial due to unresolved issues of usage and synonymy. Large, edible, and diurnal, dasyproctids were among the first mammals of the Guianan fauna to be reported by European travellers, whose inadequate published descriptions are at the root of several nomenclatural problems.

Dasyprocta leporina (Linnaeus)

Voucher Material: AMNH 265955; MNHN 1998.677. Total = 2 specimens.

Identification: The genus Dasyprocta has never been revised and the current species-level taxonomy (e.g., as summarized by Cabrera, 1961; Emmons, 1990, 1997; Woods, 1993) is sorely in need of critical attention. Traditionally, the red- or yellow-rumped agoutis of Amazonia have been referred to D. aguti (Linnaeus, 1766), but the oldest available name for these animals is unequivocally D. leporina (Linnaeus, 1758) (see Husson [1978] and Remarks, below).

Our voucher material closely resembles Husson's (1978) description of Dasyprocta leporina, which was based on the neotype and other specimens from Surinam. The only noteworthy point of difference is the color of the long nape hairs, which are blackish in our vouchers, whereas Husson (p. 459) stated that “[t]he anterior part of the dorsal surface of the body including the neck is olivaceous gray speckled with yellowish”, the nuchal hairs apparently not being distinctively colored in the neotype. Husson's other specimens, however, were described (op. cit.) as having darker necks, and most of the Surinamese agoutis we examined had blackish napes like our vouchers. As noted by Husson, the rump color in this species is also variable, ranging from clear yellow-orange to grizzled reddish-brown in the specimens we examined. Although our single adult voucher (AMNH 265955) is larger than any of the 12 Surinamese specimens measured by Husson (1978: table 82), the differences between homologous dimensions of AMNH 265955 and his largest example (RMNH 18235) are proportionately small (e.g., about 4% in maxillary toothrow length), and we do not regard them as taxonomically significant. Morphometric variation among nine FMNH specimens from Surinam (table 40) likewise suggests that our voucher is large but probably not outside the normal size range of typical D. leporina. For future revisionary work, a large series of specimens collected by H. A. Beatty in the Wilhelmina Mountains of Surinam (see Specimens Examined, below) provides a useful sample of individual variation from a single local population of this nomenclaturally important species.

Both Cabrera (1961) and Ojasti (1972) recognized several subspecies of Dasyprocta aguti (= D. leporina; see above and Remarks, below) as valid, but the necessity for a trinomial classification remains to be convincingly established by a comprehensive study of geographic variation. From the close similarity between our voucher material and the Surinamese specimens with which we compared them, the Paracou population would be unambiguously referable to the nominate form if other subspecies were to be recognized in a future revisionary study. Alternatively, it is possible that red-rumped agoutis include two or more separate species (as suggested by Krumbiegel, 1941), another hypothesis that remains to be effectively tested by critical analyses of specimen data.

The type of Dasyprocta cristata (Geoffroy, 1803) and another specimen so identified in the RMNH are both zoo animals said to have been collected in Surinam (Husson, 1978). A single 19th-century RMNH specimen of D. fuliginosa is also said to have come from Surinam (op. cit.). None of this material, however, is accompanied by names of collectors, dates of collection, or any additional geographic information. Because we have not seen any material referable to D. cristata or D. fuliginosa accompanied by definite evidence of origin from any of the Guianas, we assume that these old and poorly documented Surinamese records are erroneous. As discussed by Goeldi and Hagmann (1904), Thomas (1917), and Cabrera (1961), the type of D. prymnolopha, said to be from “Guiana” (Wagler, 1831: 619), was probably collected somewhere along the Atlantic coast of Brazil between the mouth of the Rio Tocantins and Bahia. Apparently, D. leporina is the only agouti species validly known from Guyana, Surinam, French Guiana, and Guianan Brazil. The black agouti (Dasyprocta fuliginosa), however, occurs allopatrically in the headwaters of the Orinoco (geographically part of the Guiana subregion of Amazonia) in southernmost Amazonian Venezuela (Tate, 1939; Ojasti, 1972; Handley, 1976).

Remarks: Although Dasyprocta aguti has long been used as the technical name for one or more geographic forms of red- or yellow-rumped agoutis (e.g., by Waterhouse, 1848; Thomas, 1917; Cabrera, 1961), the basis for this traditional usage is problematic. Because Linnaeus's (1766) original description of Mus aguti did not mention rump color, the identification of aguti has been justified primarily by reference to the bibliographic sources of his account (Marcgraf, 1648; Piso, 1658; Ray, 1693; Brisson, 1756). Thomas (1898) argued that all of Linnaeus's sources for Mus aguti could be traced back to the Brazilian animal described by the Dutch explorer-naturalist Georg Marcgraf. According to Thomas, Marcgraf's agouti was a yellow-rumped animal, but this inference is not consistent with what is known about the South American travels of that author.

Marcgraf, an employee of the Dutch West India Company, principally resided at Recife in the northeastern state of Pernambuco, from which base he explored the region effectively controlled by the 17th-century Dutch millitary occupation of northeastern Brazil (Whitehead, 1979). However, red- (or yellow-) rumped agoutis are not known to occur in Pernambuco (see range map in Emmons, 1990, 1997), nor anywhere else within the geographic limits of Dutch Brazil (as mapped by Boxer, 1973). Instead, Marcgraf's agouti was almost certainly the black-rumped species of the northeastern Brazilian coast, a taxon currently known as D. prymnolopha (Wagler, 1831). Marcgraf's (1648) brief color description and an accompanying woodcut of his “Aguti vel Acuti Brasiliensibus” are too ambiguous to support or refute this inference, but a painting by one of Marcgraf's contemporaries in Dutch Brazil (reproduced by Teixeira, 1995: vol. 5, p. 28) shows the local species as a more-or-less reddish animal with a blackish middorsal stripe over the rump, closely resembling D. prymnolopha of current usage.

Although this line of reasoning suggests that Dasyprocta aguti is a senior synonym of D. prymnolopha (as suggested by Carvalho and Toccheton, 1969) and not a junior synonym of D. leporina (contra Husson, 1978), our examination of Linnaeus's (1766) bibliographic sources for Mus aguti does not support Thomas's (1898) statement that all were based on Marcgraf's species. Whereas it is true that Piso's (1658: 102) and Ray's (1693: 226) accounts were obviously extracted from Marcgraf (1648), Brisson's (1756: 143) agouti description was based on an examined specimen (as previously noted by Tate, 1935). Unfortunately, Brisson's description is taxonomically uninformative (like Marcgraf's), and the geographic origin of his specimen is not stated. However, Brisson was employed in the natural history cabinet of the French naturalist Réaumur (Taton, 1970), so it is probable that all of the Neotropical specimens seen by him were from French Guiana. Additionally, Brisson (1756: 144) gave the geographic range of his agouti as “Guiania & Brasilia” and he cited Barrère's (1741) description of the “Agouty” of Cayenne. After Réaumur's death, his collection was transferred to the Cabinet du Roi (Taton, 1970), which subsequently became part of the Muséum National d'Histoire Naturelle. Geoffroy (1803), however, recorded no MNHN agouti material from the “ancien cabinet”, nor does that museum now contain any 18th-century specimen of Dasyprocta that might have been the one described by Brisson (L. Granjon, personal commun.).

Finally, Linnaeus's (1766: 80) account of Mus aguti gave the geographic range as “Brasilia, Surinamo, Guiania” despite the fact that none of his cited references for this species mentioned Surinam. It is therefore probable that Linnaeus himself either had seen specimens of Surinamese agoutis or had reliable reports of them from his many Dutch colleagues and aquaintances. Regrettably, no material of Dasyprocta that is certainly known to have been examined by Linnaeus has apparently been identified in the literature, nor have we succeeded in locating any nomenclaturally useful 18th-century specimens in museums known to contain Linnaean mammals. However, any Surinamese agoutis seen by or reported to Linnaeus were necessarily red-rumped animals.

Linnaeus's Mus aguti is therefore composite, having been based directly or indirectly on lost specimens of both the red- and the black-rumped species. Because Thomas's (1898) proposal to restrict aguti to Marcgraf's Brazilian animal was not equivalent to a lectotype designation, the application of the name aguti is still an open question. In our judgment, it would not be desirable to replace prymnolopha, the name by which the black-rumped agouti of the Brazilian Atlantic coast has been long and consistently known, with aguti, the traditional name for the red-rumped species. We therefore select as the neotype for Mus aguti Linnaeus (1766) the same specimen (RMNH 20752) that Husson (1978) designated as the neotype of Mus leporinus Linnaeus (1758). By this action Dasyprocta aguti becomes an objective junior synonym of D. leporina and current usage (Woods, 1993) is preserved.

Other Specimens Examined: Surinam—Brokopondo, Locksie Hattie on the Saramacca River (FMNH 95757, 95758, 95760); Marowijne, Paloemeu Camp (FMNH 95763, 95765, 95767–95771); Nickerie, Wilhelmina Mountains on West River (FMNH 95772–95778, 95790–95792); Para, Zanderij (BMNH 1952.1152–1952.1154); Paramaribo, Paramaribo (BMNH 1952.1155, 1952.1156); Saramacca, Dirkshoop (FMNH 95761); Suriname, Carolina Kreek (FMNH 95756).

Field Observations: We saw or heard Dasyprocta leporina almost every day that we were in the forest at Paracou, often in the early morning or late afternoon, less frequently in the middle of the day, and only rarely after dark (most nocturnal sightings may have been of individuals frightened from their resting places in dense vegetation). Most individuals were encountered singly, but a few groups of two or three individuals were also seen. We recorded sightings in well-drained primary forest, swampy primary forest, creekside primary forest, and roadside secondary growth. Both of our voucher specimens were shot.

Myoprocta acouchy (Erxleben) Figures 72A, 74, 75

Voucher Material: AMNH 266566; MNHN 1998.678. Total = 2 specimens.

Identification: Members of the genus Myoprocta—commonly known as acouchies—are restricted to Amazonia, where most authors have recognized two species, one “red” (or “reddish”) and the other “green” (or “greenish”), based primarily on coat-color differences. Despite this consensus, the diagnostic morphological characters and geographic distribution of red and green acouchies have yet to be convincingly documented by any published revisionary study based on museum specimens, and the nomenclature of Myoprocta species is currently confused.

Red and green acouchies can be distinguished unambiguously by a combination of external and cranial characters whose diagnostic value we tested by examining all of the Myoprocta specimens (including types) in five American and European museums (AMNH, BMNH, FMNH, MNHN, and USNM). In external appearance, most red acouchies are rich reddish-brown dorsally, with uniformly orange or reddish underparts, whereas most green acouchies are drab yellowish- or grayish-brown dorsally with yellowish underparts that are usually marked by white midventral streaks. Occasional skins of both the red and green species, however, have somewhat intermediate pigmentation. Thus, a few red acouchies (especially zoo specimens) have rather drab fur (e.g., AMNH 130148, BMNH 5.11.1.19, FMNH 21786), and a few green acouchies (e.g., AMNH 68243, BMNH 54.608) have warmer pelage tones than usual. Some coat-color variation is geographic in origin (e.g., red acouchies from the Guianas have generally less saturated pigments than Brazilian specimens from the north bank of the Amazon), but individual tonal differences also exist within most large series (e.g., 16 AMNH skins of green acouchies from San José Abajo, Ecuador). Nevertheless, the difference in coloration between red and green acouchies is conspicuous when typical exemplars of both kinds are viewed side-by-side.

Another obvious external difference is the presence in all red acouchies of a distinct rump patch of very long (60–80 mm) highly polished hairs that are typically much more heavily pigmented than the fur of the sides, middle back, and forequarters; in life, these hairs form a dark, glossy fringe that actually extends beyond the rump to overlap the base of the tail (Emmons, 1990, 1997). Some specimens (e.g., AMNH 93043) have blackish rump hairs, but the usual color is a deep mahogany brown with or without inconspicuous basal bands of red. In a few Brazilian specimens (e.g., AMNH 94068, 94071; BMNH 20.7.1.23), the rump hairs are not substantially darker than the hairs of the sides and middle back, but they are still distinguishable from the fur of the latter parts by their length and high polish. By contrast, green acouchies never have a distinct rump patch, the fur over the hindquarters being essentially similar in length, color, and texture to the rest of the dorsal pelage.

Red acouchies are, on average, larger than green acouchies, a comparison that is best appreciated by craniometric comparisons of representative series (table 41). Visually, most red acouchy skulls have noticeably larger toothrows and bullae, broader interorbits, and longer nasals than most green acouchy skulls. Although no univariate measurement is diagnostic, red and green acouchies have nonoverlapping morphometric distributions in multivariate ordinations (e.g., by principal components analysis; not shown). Of the several qualitative cranial traits that Tate (1939: table IV) used to distinguish red and green acouchies, the most consistently useful is the size of the “palatal foramina” (= sphenopalatine vacuities), which perforate the bony roof of the mesopterygoid fossa; these are very narrow slits (<1 mm wide) in most specimens of red acouchies (fig. 72A), but they are wider (>1 mm) teardrop-shaped openings in most specimens of green acouchies (fig. 72B).

Based on specimens that we examined, red acouchies occur throughout Guyana, Surinam, French Guiana, and Brazil north of the Amazon and east of the Rio Branco (fig. 73). We are aware of only two vouchered records of red acouchies outside the Guiana subregion of Amazonia: (1) one specimen (AMNH 37123) collected by Leo E. Miller on 21 April 1914 on the “Lower Solimões”, a locality that is now believed (see footnote 17) to correspond to Manacaparú, a settlement on the north bank of the upper Amazon just west of its confluence with the Rio Negro (locality 5 in fig. 73); and (2) two specimens (FMNH 50895, 50896) collected by A. M. Olalla on 9 August 1936 at Lago do Baptista on the south bank of the Amazon between the Rio Madeira and the Rio Tapajos (locality 3 in fig. 73). Based on the latter record, we assume that unvouchered reports of unidentified Myoprocta from other sites along the south bank of the Amazon between the Madeira and the Tocantins (e.g., George et al., 1988; Voss and Emmons, 1996: appendix 8) were probably of red acouchies. Acouchies are apparently unknown east of the Rio Tocantins (Carvalho and Toccheton, 1969; Pine, 1973).

By contrast, green acouchies are extensively distributed in western Amazonia (west of the zoogeographic axis defined by the Rio Negro and the Rio Madeira), and they also occur in the headwaters of the Orinoco in southernmost Venezuela (Tate, 1939; Handley, 1976; Linares, 1998). Although a few published sources imply that red and green acouchies occur sympatrically in eastern Ecuador (Lönnberg, 1925) or eastern Colombia (Emmons, 1990, 1997), our specimen data suggest that red and green acouchies are allopatrically distributed. At least some of the historical uncertainty about the geographic ranges of red and green acouchies is due to the confused technical nomenclature for these animals.

Erxleben's (1777) original description of Cavia acouchy mentioned only small size, presence of a tail, and olivaceous coloration as characters distinguishing this species from other terrestrial hystricognaths then referred to the genus Cavia. Apparently, Erxleben did not examine any specimens himself, but instead based his description of acouchy on the earlier accounts of des Marchais (1730), Barrère (1741), Buffon (1767), and Pennant (1771) that he cited as references. We examined all of these early works and determined that Buffon and Pennant contain nothing more than rephrased versions of des Marchais' and Barrère's very brief reports about the “Agouchi” or “Akouchy” of Cayenne. By way of description, des Marchais stated only that the Agouchi is smaller and tastes better than the Agouti (Dasyprocta leporina), but Barrère (quoted verbatim by Tate [1935: 331] and Husson [1978: 471]) mentioned a tail and olive coloration.

There has never been any question that Erxleben's sources were describing a Myoprocta from French Guiana, but the reference to olive coloration has led to conflicting taxonomic applications of the epithet acouchy. Thomas (1926: 639) argued that this name applies to the reddish species based on the geographic origin of des Marchais' and Barrère's observations, noting that “many specimens of this animal are of a somewhat olivaceous tone, which, in the absence of [M.] pratti [the green acouchy], might easily justify the word being applied to them.” Cabrera (1961) agreed, noting that Erxleben's original sources were not professional naturalists and might have been describing the species inaccurately from memory. By contrast, Tate (1939), Carvalho (1962), and Husson (1978) applied the name acouchy to the green species based on Barrère's color description; according to these authors, the red acouchy should be called M. exilis (Wagler, 1831).

The crux of this disagreement is whether geography or color is to be given greater importance in applying the name acouchy. In the absence of an extant type,20 the issue cannot be definitely resolved, and both of the conflicting usages mentioned above are current in the literature (e.g., Woods, 1993; Emmons, 1990, 1997). Since the geographic datum is definite whereas the color description is subject to interpretation, we favor Thomas's (1926) and Cabrera's (1961) usage, which also avoids the absurdity of making French Guiana—where only red acouchies are known to occur—the type locality for the green species. In order to fix this application of Myoprocta acouchy and thereby stabilize the species-level nomenclature of Myoprocta, we select as the neotype of Cavia acouchy Erxleben (1777) our adult Paracou voucher, AMNH 266566, consisting of a well-preserved skin (fig. 74), skull (fig. 75), and postcranial skeleton; measurements of this specimen are provided in table 42 along with those of other conspecific adults from French Guiana.

Our assignment of nominal taxa in the genus Myoprocta to either the red or green species groups (table 43) is based on first-hand examination of types and/or original descriptions. Although some Brazilian specimens of red acouchies are larger and redder than most specimens from Guyana, Surinam, or French Guiana, all red acouchies closely resemble one another and we see no compelling evidence that more than a single species is represented. Green acouchies, however, appear to exhibit significant geographic variation and may eventually prove to be a complex of closely related species; nevertheless, all can be provisionally referred to M. pratti pending a comprehensive study of this group.

Other Specimens Examined: Brazil—Amapá, Serra do Navio (USNM 546313); Amazonas, Boca Rio Paratucu on Rio Jamundá (AMNH 94073–94075), Lago do Baptista on Rio Amazonas (FMNH 50895, 50896), Lago do Serpa on Rio Amazonas (FMNH 50897), “Lago do Taraci on Rio Negro” (BMNH 27.8.11.52), Lower Solimões (AMNH 37123), Santo Antonio do Amatari (AMNH AMNH 92886–92888, 93043); Pará, “Castanhal on Rio Jamundá” (AMNH 94076), Colonia do Veado (BMNH 12.5.11.9), Faro (AMNH 94068–94072), Monte Alegre (BMNH 20.7.1.23), Cachoeira Porteira (USNM 546296, 546297), San José on Rio Jamundá (AMNH 94077); Roraima, Conceição (FMNH 20007), Serra Grande (FMNH 20019). French Guiana—Saut Macaque (MNHN 1962.1329), St.-Laurent du Maroni (MNHN 1909.243), Tamanoir on Mana River (FMNH 21783, 21785–21787), no other locality data (MNHN 1962.1330, 1974.268). Guyana—“Bonasica on Essequibo River” (AMNH 36493 [type of demerarae], BMNH 12.6.5.28, 13.5.23.6–13.5.23.9), “Kuitaro River” (USNM 338969–338971), “Manarica Creek on Essequibo River” (BMNH 13.6.8.11), “Moon Mountains” (BMNH 11.6.7.45), “Supinaam River” (BMNH 20.7.1.20–20.7.1.22); Cuyuni-Mazaruni, Kartabo (AMNH 8178); Potaro-Siparuni, Potaro (BMNH 3.4.6.6, 3.4.6.7); Upper Demerara-Berbice, “Comackka on Demerara River” (BMNH 5.11.1.19), Moraballi (BMNH 34.6.30.57–34.6.30.61); Upper Takutu-Upper Essequibo, Dadanawa (USNM 339670). Surinam—Brokopondo, Locksie Hattie (FMNH 95753 [two specimens with this number]); Marowijne, Paloemeu Camp on Tapahoni River (FMNH 95593, 95754, 95787); Nickerie, Kaiserberg Airstrip on Zuid River (FMNH 93270–93277), Wilhelmina Mountains on West River (FMNH 95755).

Field Observations: We frequently heard the alarm calls of Myoprocta acouchy and we caught glimpses of fleeing individuals on many occasions, but we seldom obtained an unobstructed view of the animal in repose. Our single adult voucher was taken on the ground in a leghold trap in swampy primary forest, and a juvenile specimen was subsequently taken in a Victor trap tied to the base of a broad liana 50 cm above the ground in well-drained primary forest. Four unvouchered observations recorded in our fieldnotes document the presence of this species in well-drained primary forest, swampy primary forest, creekside primary forest, and secondary vegetation. Most of our sightings were diurnal, usually in the very early morning or late afternoon, but a few animals were flushed from their hiding places at night.

Cuniculidae

Only a single cuniculid species occurs at Paracou. The lowland paca is unmistakable in external and craniodental characters and cannot be confused with any other species of Guianan mammal.

Cuniculus paca (Linnaeus)

Voucher Material: AMNH 265954, 266567, 266569; MNHN 1998.679. Total = 4 specimens.

Identification: Our specimens are topotypes of this species (the type locality of which was restricted to French Guiana by Tate, 1935) and agree closely with Husson's (1978) detailed description and illustrations of Surinamese material. Selected measurements (mm) and weights (kg) of two adult male vouchers (AMNH 265954, MNHN 1998.679) are: HBL 650, 739; LT 0, 11; HF 115, 121; Ear 52, 53; CIL 138.1, 134.4; LD 54.3, 52.7; MTR 27.3, 30.5; LN 52.4, 51.0; LIB 41.2, 42.4; ZL 72.3, 77.0; ZB 98.4, 105.2; Wt 9.5, 9.2.

The lowland paca ranges from southern Mexico to Paraguay, and several subspecies have been recognized as valid by authors (e.g., Krumbiegel, 1940b; Cabrera, 1961). Although Krumbiegel (1940b) cited pelage and cranial characters to justify his trinomial distinctions, no comprehensive review of geographic variation in this species based on the very large museum collections now available for study has yet been attempted. In the absence of such a critical undertaking, a subspecific nomenclature does not seem warranted at present; however, our material would obviously represent the nominate race if trinomials were to be recognized in any future revision.

In a recently published opinion, the International Commission on Zoological Nomenclature (ICZN, 1998) ruled that Cuniculus was available from Brisson (1762), thus replacing Agouti Lacépède (1799) as the oldest valid generic name for the lowland paca.

Field Observations: We collected four specimens of Cuniculus paca at Paracou and recorded 24 unvouchered observations in our fieldnotes. Habitat information accompanying 26 records include 7 encounters (27%) in well-drained primary forest, 4 (15%) in swampy primary forest, 6 (23%) in creekside primary forest, 1 (4%) in primary forest of unspecified character, and 9 (35%) in secondary vegetation; all encounters were nocturnal. Pacas were encountered singly, except in rare cases when we witnessed agonistic encounters between two adults or saw an adult female accompanied by a juvenile. Most individuals were sighted on the ground, but one individual was submerged in a stream under a fallen tree, apparently attempting to hide underwater.

Echimyidae

We captured or observed four echimyid species at Paracou, including representatives of the genera Makalata, Mesomys, and Proechimys. Based on collections from other localities in French Guiana and Surinam (appendix 1), it seems probable that two additional echimyids could occur locally. Although species of Proechimys can only be distinguished from one another with adult specimens in the hand, other French Guianan echimyids can be identified at a distance by well-marked external characters (Emmons, 1990, 1997) if a sufficiently clear view is obtained. Unfortunately, although the taxa in question are readily distinguished, their nomenclature is still problematic in some cases.

Makalata didelphoides (Desmarest)

Our only record of this species at Paracou is an unambiguous sighting by DPL of a solitary adult perched on a tree trunk 3 m above a small stream in primary forest at 10:30 hours on 27 October 1992.

In most of the older literature (e.g., Tate, 1935, 1939; Cabrera, 1961) this species is called Echimys armatus (I. Geoffroy), but Husson (1978) proposed the new genus Makalata with armatus as type species in recognition of the well-marked craniodental differences between this taxon and Echimys chrysurus (the type species of Echimys G. Cuvier). For the use of didelphoides to replace armatus as the oldest available name for the red-nosed tree rats of the Guiana subregion of Amazonia, see Emmons (1993b).

Mesomys sp

Voucher Material: AMNH 266596.

Identification: The genus Mesomys has never been revised and the identification of its constituent species has long been problematic. Woods (1993) recognized M. didelphoides (Desmarest), M. hispidus (Desmarest), M. leniceps Thomas, M. obscurus (Wagner), and M. stimulax Thomas as valid species, but Emmons (1993b) showed that didelphoides and obscurus both belong in the genus Makalata, and Patton et al. (2000) subsequently described a new species, Mesomys occultus. Therefore, current usage would now recognize four valid species of Mesomys: M. hispidus (including ecaudatus Wagner, ferrugineus Günther, and spicatus Thomas as synonyms; after Woods, 1993), M. leniceps, M. occultus, and M. stimulax.

In order to identify our single Paracou voucher, we consulted the original descriptions of all nominal taxa currently referred to Mesomys, and we examined the holotypes of every named form except ecaudatus. In addition, we measured relevant series of specimens in the AMNH, USNM, and BMNH to assess geographic variation in morphometric characters. Below we explain why, despite this effort, we are still unable to confidently assign a specific epithet to our material.

Of all the material we examined, the most distinctive is the holotype of Mesomys leniceps from the Andean highlands of eastern Peru. This specimen (BMNH 26.8.6.61) is distinguished from all other congeners by its much longer tail (relative to head-and-body length), finer and denser tail hairs, finer and softer spines, more convergent toothrows, posteriorly constricted incisive foramina, and much smaller bullae. By contrast, Amazonian specimens of Mesomys (including all of the remaining nominal taxa in this genus) have relatively shorter tails, coarser spines and tail hairs, less convergent toothrows, posteriorly unconstricted incisive foramina, and larger bullae. Although we searched diligently for external and craniodental character variation among Amazonian samples of Mesomys, only maxillary toothrow length appears to offer any broadly useful morphological basis for taxonomic inference.

The first example of Mesomys to be reported from any of the Guianas was a Surinamese specimen that Husson (1978) identified as M. stimulax. We have examined five additional specimens (table 44) from the Guiana subregion, including our Paracou voucher, all of which essentially resemble Husson's material and appear to represent the same taxon despite substantial variation in some cranial dimensions (plausibly attributable to age differences) and in external proportions (perhaps attributable to measurement artifacts). Consistent with Husson's identification, these Guianan specimens have small toothrows (MTR = 5.9–6.6 mm) that fall within the range of variation exhibited by specimens referable to stimulax from southeastern Amazonia (right-hand column, table 44). By contrast with these small-toothed Guiana and Southeastern subregion samples, all western Amazonian Mesomys samples have mean toothrow lengths greater than 7 mm (e.g., the series measured by Patton et al., 2000: tables 56, 58). Small-toothed and large-toothed Mesomys are sympatric at Igarapé Amorim on the left (west) bank of the Rio Tapajos (specimens in AMNH), but elsewhere these morphometric classes appear to be allopatric. Although it is geographically part of the Guiana subregion of Amazonia, the Venezuelan state of Amazonas is inhabited by large-toothed Mesomys (e.g., nine USNM specimens with a mean maxillary toothrow length of 7.2 mm), an observation consistent with the presence of other western Amazonian taxa (e.g., Dasyprocta fuliginosa, Myoprocta pratti; see above) in that area.

Although western Amazonian Mesomys specimens have often been identified as M. hispidus (e.g., by Cabrera, 1961; da Silva and Patton, 1993; Patton et al., 1994, 2000; Voss and Emmons, 1996), this usage appears to be incorrect. The maxillary toothrow of the type of hispidus21 measures 6.4 mm, well outside the observed range of variation in this dimension for any known western Amazonian sample. Unfortunately, it is not known exactly where the type was collected. Desmarest (1817) gave the type locality as “l'Amérique méridionale”, but the specimen was almost certainly obtained by Alexandre Rodriguez Ferreira,22 whose known collecting itinerary (reproduced by Hershkovitz, 1987a: fig. 3) was largely confined to Amazonian Brazil. Tate (1935) proposed restricting the type locality of hispidus to Borba, a settlement on the right bank of the lower Rio Madeira where Natterer collected the type of ecaudatus in 1830 (Pelzeln, 1883). Tate's action was biologically arbitrary but served to justify his synonymization of hispidus and ecaudatus, the types of which he had not seen. Unfortunately, two species of Mesomys with divergent toothrow measurements are known to occur in the interfluvial region between the Madeira and the Tapajos (e.g., at Igarapé Amorim, see above) and it is not known whether the type of ecaudatus is a small-toothed animal (like hispidus) or not. Until the type of ecaudatus is examined and measured, the oldest available name based on a large-toothed Mesomys specimen is ferrugineus Günther (1876) from northeastern Peru.

Based simply on toothrow measurements, our Paracou voucher and other similar specimens from the Guiana subregion of Amazonia could justifiably be referred to Mesomys hispidus, of which stimulax is perhaps only a junior synonym. However, Patton et al. (2000) reported high cytochrome-b sequence divergence between small-toothed Mesomys specimens from opposite sides of the Amazon, a result that clearly indicates the insufficiency of toothrow length as a basis for taxonomic inference and brings into sharper focus the problem of exactly where the type of hispidus was collected (see footnote 22). For the moment, these are insoluble problems for which the only appropriate interim solution is to leave our voucher material unnamed.

Other Small-Toothed Specimens Examined from the Guiana Subregion: Brazil—Amapá, Serra do Navio (USNM 543283). French Guiana—Les Nouragues (V-924). Guyana— Potaro-Siparuni, 5 km SE Surama (ROM 103346); Upper Takutu-Upper Essequibo, Maipaima Creek in Kanuku Mountains (LHE 968, an uncataloged specimen to be deposited in UG).

Field Observations: Our single record of Mesomys from Paracou is based on a specimen taken in a Victor trap tied to a broad liana 1.9 m above the ground in swampy primary forest.

Proechimys J. A. Allen

The last comprehensive revision of the genus Proechimys was Moojen's (1948) monograph, but subsequent research has resulted in substantial modifications of his pioneering taxonomic synthesis. In particular, studies of well-sampled local faunas have proven crucial for distinguishing patterns of morphological, karyotypic, and molecular variation within and among sympatric species of Proechimys, especially in western Amazonia where four or more occur at some localities (Patton and Gardner, 1972; da Silva, 1998; Patton et al., 2000). To date, however, no equivalent studies of character variation within and among sympatric Proechimys species from other Amazonian subregions have been published.

Until quite recently, only a single widespread species of Proechimys, variously identified as P. guyannensis (E. Geoffroy, 1803) or as P. cayennensis (Desmarest, 1817), was recognized in the Guianas (e.g., by Tate, 1935; Cabrera, 1961). In 1978, however, A. M. Husson (at the Rijksmuseum van Natuurlijke Historie, Leiden) and F. Petter (at the Muséum National d'Histoire Naturelle, Paris) independently reported the occurrence of two sympatric forms in Surinam and in French Guiana, respectively. Although both authors recognized a common large species and a rarer small species, they gave different accounts of diagnostic nonmetrical characters and they used different nomenclature.

Husson (1978) identified his large Surinamese specimens as Proechimys guyannensis (E. Geoffroy, 1803), and his small specimens as P. warreni Thomas (1905). In Husson's opinion, measurements of the cheekteeth provided the best diagnostic criterion: 18 specimens of the large Surinamese species had maxillary toothrow lengths of 8.3–9.2 mm, whereas 6 specimens of the small species had cheektooth measurements of 7.2–7.7 mm. Husson also reported that his large species was more abundant, and had more rufous pelage, a relatively shorter tail, less appressed caudal hairs, and a wider mesopterygoid notch in the back of the hard palate than his small species (op. cit.: 429–438). Husson based his identification of the large species primarily on Geoffroy's (1803) description of pelage color in guyannensis because the skull of the type (not measured by Geoffroy) was thought to have been lost (Rode, 1945). Husson's identification of the small species, however, was based on direct external and craniodental comparisons with the type of warreni.

Petter (1978) reported the rediscovery of the type skull of Proechimys guyannensis and assigned this name to the smaller (and rarer) of the two forms found in French Guiana; the larger (and more abundant) form, apparently undescribed, was then named as a new species, P. cuvieri. However, whereas Husson found that the large and small Surinamese species were nonoverlapping in maxillary toothrow length, Petter reported overlapping variation in this measurement, with observed ranges of 7.0–8.5 mm for guyannensis and 8.1–9.3 mm for cuvieri. Petter did not mention the other distinguishing characters discussed by Husson, but he noted karyotypic differences (2N = 40 in guyannensis, 2N = 28 in cuvieri) and differences in the number of enamel islands of the lower cheekteeth (two per tooth in guyannensis, three in cuvieri). The karyotypes of guyannensis and cuvieri were subsequently described in greater detail by Reig et al. (1979).

In an important review of morphological variation in Proechimys, Patton (1987) identified several qualitative characters distinguishing clusters of populations assigned to his cuvieri and guyannensis species groups. However, Patton's study included no French Guianan material, so the diagnostic value of the characters he investigated is unknown for the local populations sampled by our study. Indeed, Guillotin and Ponge (1984: 287) doubted that cuvieri and guyannensis could be distinguished by “méthodes morphologiques classiques”, a conclusion based on bivariate and multivariate analyses that showed no obvious morphometric discontinuity among French Guianan specimens representing both of the karyomorphs reported by Petter (1978) and Reig et al. (1979). The apparent lack of diagnostic external characters between guyannensis and cuvieri has also been an impediment for ecologists unable to identify these species by nondestructive sampling in field studies (e.g., Forget, 1991).

The morphological characters of French Guianan populations of Proechimys cuvieri and P. guyannensis are important because both species are based on types collected in French Guiana (Petter, 1978). Additionally, the application of these names to the sympatric Surinamese forms reported by Husson remains to be evaluated. In order to identify our Paracou vouchers, and to provide a basis for future revisionary research with these species, we examined most of the Proechimys specimens from French Guiana currently held in American and European museums. Except as noted below, our results are based on measurements and qualitative character data obtained from fully adult but nonsenescent animals, herein defined as members of age categories 8 and 9 of Patton and Rogers (1983). The following results of our analyses broadly overlap those independently obtained by Catzeflis and Steiner (2000), who examined many of the same specimens.

Maxillary toothrow measurements of nonsenescent adult Proechimys from French Guiana form two distributions (fig. 76): a sparse cluster of small specimens with cheekteeth measuring 6.9–8.0 mm, and a denser cluster of larger specimens with longer toothrows (8.2–9.3 mm). Following Petter (1978), we associate the name guyannensis with the smaller-toothed form based on the toothrow dimensions of the holotype (MNHN 1995.1395, an old adult), which has an alveolar measurement of 7.8 mm and an estimated crown measurement of 7.3 mm (table 45). The larger-toothed specimens that we measured include the type of cuvieri, MNHN 1977.774, which has a crown-length toothrow measurement of 8.5 mm. To supplement this essentially univariate distinction, we examined qualitative variation in craniodental, external, and genitalic characters to determine whether or not Proechimys with small versus large cheekteeth differ in other respects.

We scored the septum that separates the right and left incisive foramina as “complete” or “incomplete” (fig. 77). Variation in this character among the specimens at hand is determined by the presence or absence of contact between the bony capsules containing Jacobson's organ and a median process of the maxillary bone. Initially, we distinguished complete septa formed by slender maxillary processes from complete septa formed by robust maxillary processes, but intermediate conditions made this additional refinement too arbitrary for confident scoring.

We scored the development of the canal transmitting the infraorbital nerve in the floor of the infraorbital foramen (fig. 78) using the numerical coding suggested by Patton (1987): no groove present (1), or groove moderately developed (2), or groove well defined (3). Scoring these conditions as well as intermediate states (1.5, 2.5) was accomplished by reference to exemplar specimens of Brazilian Proechimys guyannensis previously examined by J. L. Patton (personal commun.). Therefore, we presume that our trait-frequency data and his (Patton, 1987: table 3) are comparable.

We also followed Patton's (1987) numerical coding convention for the depth of the mesopterygoid fossa (fig. 77): not extending to the posterior margin of M3 (1), or extending to the posterior half of M3 (2), or extending to the anterior half of M3 (3), or extending to the posterior half of M2 (4), or extending to the anterior half of M2 (5). Because the angle formed by the posterior palatal margins of the mesopterygoid is correlated with the depth of penetration of the fossa between the toothrows (deeper fossae have more acutely angled palatal margins; Patton, 1987), we did not score the shape of the mesopterygoid notch as a separate character.

Finally, we recorded the number of internal folds on the second mandibular molar (corresponding to the “îlots d’émail” of Petter, 1978); compound (Y-shaped) folds were each counted as one-and-a-half folds. As noted by Patton (1987), internal fold number and morphology change with toothwear, so it is particularly important that this character be scored among individuals of approximately equivalent age.

Trait frequencies of these four qualitative characters differ significantly between our operationally defined samples of cuvieri (with MTR ≥ 8.2 mm) and guyannensis (with MTR ;le 8.0 mm) despite the small number of guyannensis available for scoring (table 46). The presence or absence of a complete incisive septum and the development of an infraorbital groove are the two qualitative characters most consistently correlated with maxillary toothrow length: all large-toothed specimens (cuvieri) have a complete incisive septum and either lack an infraorbital groove entirely or have a very weakly defined groove; by contrast, most small-toothed specimens have an incomplete incisive septum and well-defined infraorbital grooves. Although most specimens of both tooth-size classes have a moderately deep mesopterygoid fossa and 2½ folds on m2, no specimens of small-toothed rats in our sample have either a very shallow mesopterygoid fossa or m2s with 3 folds. It is noteworthy that the qualitative traits of the specimen with a maxillary toothrow length of 8.0 mm (MNHN 1981.48, which could be interpreted as an outlier of either tooth-size class) link it unambiguously with guyannensis, and that specimens with “atypical” states for one qualitative character (e.g., AMNH 267047, a small-toothed animal that lacks any trace of an infraorbital groove) are not atypical in other qualitative respects. Altogether, these results strongly support the hypothesis that the discrete tooth-size classes associated with the names cuvieri and guyannensis represent valid species, a conclusion that is further bolstered by external, genitalic, and karyotypic comparisons.

As noted by Malcolm (1992), the tail of Proechimys cuvieri is visibly hairier than that of P. guyannensis. This difference is caused by the individual caudal hairs, which curve outward from beneath each epithelial scale in P. cuvieri, where they can easily be seen standing away from the caudal surface (fig. 79, left). By contrast, the individual hairs are appressed to the caudal surface in P. guyannensis, where they are difficult to see without magnification; in consequence, the tail appears to be smooth and naked (fig. 79, right).

The dorsal body pelage of Proechimys cuvieri is, on average, redder (more saturated) than the generally drab (grayish or yellowish brown) fur of P. guyannensis, but there is sufficient overlap in color among the specimens at hand that this contrast is useful for field identification only in combination with size and other characters. This seems to be the only significant species color difference in the material we examined. For example, we did not see any diagnostically useful differences in ventral fur color (pure white in most specimens of both species), tail pigmentation (distinctly bicolored in most specimens of both species), or hindfoot markings.

The male genitalia differ strikingly in shape between Proechimys cuvieri and P. guyannensis. Whereas the penis of cuvieri is short and very broad, that of guyannensis is long and slender, an obvious contrast that is reflected in the highly divergent bacular morphologies illustrated by Patton (1987: figs. 5, 10) for members of his cuvieri and guyannensis species groups. Although this character cannot be used to identify most museum study skins (few of which have the penis attached), it is potentially useful for field identifications because the penis can be extruded by retracting the prepuce of live animals.

Karyotypic data recorded on skin tags of MNHN specimens indicate that different diploid counts are associated with the divergent morphological phenotypes described above and corroborate Petter's (1978) and Reig et al.'s (1979) taxonomic assignments. Thus, five French Guianan specimens that we examined with recorded karyotypes of 2N = 28 (MNHN 1972.639, 1974.263, 1974.266, 1981.36, 1998.315) represent the cuvieri morphotype, whereas two specimens that we examined with 2N = 40 (MNHN 1983.376, 1998.312) represent the guyannensis morphotype.

In the following species accounts we summarize diagnostic characters, comment on Husson's (1978) identifications of Surinamese material, and evaluate the probable geographic distribution of Proechimys cuvieri and P. guyannensis based on our assessment of character variation in the material at hand.

Proechimys cuvieri Petter Figures 77–80

Voucher Material: AMNH 266570–266575, 266578, 266580–266582, 266588, 266589, 266591, 266592, 266594, 267025–267030, 267032, 267034, 267039, 267041, 267045, 267599, 267601–267603, 269122; MNHN 1998.685–1998.699. Total = 46 specimens.

Identification: Based on the preceding synthesis of character data from French Guianan material, adult specimens of Proechimys cuvieri can be characterized as large rats that contrast with adults of the smaller sympatric species P. guyannensis by their longer (≥8.2 mm) maxillary toothrows (versus ;le8.0 mm in guyannensis), complete incisive septum (vs. septum usually incomplete in guyannensis), incisive foramina with larger posterolateral flanges (vs. foramina usually with smaller flanges in guyannensis), palatal bridge with better developed median keel (vs. palate unkeeled or weakly keeled in guyannensis), infraorbital groove absent or usually weakly developed (vs. groove usually well developed in guyannensis), mesopterygoid fossa often shallow and broad (vs. deeper and narrower in many guyannensis), lower molars usually with more than two internal folds (vs. lower molars often with two folds in guyannensis), tail visibly hairier (vs. apparently smooth and naked in guyannensis), pelage often reddish (vs. drab in most guyannensis), penis short and very broad (vs. longer and narrower in guyannensis), and diploid karyotype with 28 chromosomes (vs. 2N = 40 in guyannensis).

In our experience, fully adult Proechimys in French Guiana can be identified to species in the field with considerable confidence using external measurements (e.g., hindfoot length, table 45) and the qualitative external traits described above. However, taxonomic assignments of juvenile and subadult Proechimys (with unmolted or incompletely molted soft, gray, immature pelage) are always problematic. Tooth impressions (Malcolm, 1992) and molecular markers (Steiner et al., 2000) are potentially useful tools for identifying young animals that should be incorporated in future field studies. Otherwise, many individuals will inevitably remain unidentified in ecological research based on nondestructive sampling, especially during the rainy season, when a considerable fraction of the population consists of young animals (Guillotin, 1982).

Specimens that we examined document the sympatry of Proechimys cuvieri and P. guyannensis at several localities in French Guiana, including Arataye, Cayenne, Florida, Montsinéry, and St.-Eugène, as well as Paracou. We presume that these species are also co-distributed elsewhere, certainly in Surinam and perhaps throughout the Guiana subregion of Amazonia.

According to Patton et al. (2000), Proechimys cuvieri is widely distributed in Amazonia, including large parts of the Guianan, southeastern, and western subregions. Although cuvieri is said to be relatively uniform in morphological characters throughout this enormous range, significant mtDNA sequence divergence (7–9%) exists among several geographic clusters of populations sampled by those authors (op. cit.). We have not attempted to evaluate geographic variation among Amazonian populations of cuvieri-like spiny rats for this faunal report, but it is relevant to note that typical (French Guianan) cuvieri appears to differ significantly from western Amazonian material in some morphological traits. For example, whereas Patton et al. (2000) reported that the vomer (an element of the incisive septum) is exposed in most (25 out of 34) specimens from the Rio Juruá, this bone is not exposed in any of our 22 adult Paracou vouchers. If geographic patterns of variation in this and other morphological characters were found to be consistently correlated with mtDNA haplotype divergence, it would be reasonable to infer that two or more species could be represented among the samples currently referred to this species.

Remarks: Husson (1978) identified Surinamese material of this species as Proechimys guyannensis on the basis of Geoffroy's (1803) and Desmarest's (1817) color descriptions of the type, and on the assumption that the type of guyannensis probably represented the commoner of the two Guianan forms. Color alone, however, is not a reliable basis for species identification, and the other morphological details mentioned in Geoffroy's and Desmarest's descriptions are likewise insufficient to determine which local species they had in hand. Petter's (1978) rediscovery of the long-lost type skull of P. guyannensis finally resolved the identity of that taxon and indicated the necessity of naming the larger form as a new species.

Other Specimens Examined: French Guiana—Arataye (MNHN 1983.378; USNM 548450–548452), Cacao (MNHN 1981.107), Cayenne (MNHN 1970.223, 1974.263), Florida (MNHN 1981.46, 1981.50), Piste St.-Élie (MNHN 1982.523), St.-Eugène (MNHN 1995.3220–1995.3222, 1995.3224, 1998.314, 1998.315, 1998.1821), Saül (MNHN 1977.774 [holotype], 1981.23, 1981.24, 1981.26, 1981.29–1981.33, 1981.36), Station FRG near Montsinéry (MNHN 1986.1129), Trois-Sauts (MNHN 1981.54, 1981.56), no other locality data (MNHN 1972.639, 1974.266).

Field Observations: All of our unambiguous records of Proechimys cuvieri from Paracou are based on collected specimens. Of our 46 vouchers, 18 (39%) were taken in Sherman traps, 14 (30%) were shot, 7 (15%) were taken in Victor traps, 4 (11%) were taken in Tomahawk traps, 2 (4%) were taken in Conibear traps, and 1 was taken in a pitfall. Forty-four specimens (96%) were shot or trapped at ground level, but 2 specimens (4%) were taken in traps tied to lianas 0.7–1.0 m above the ground. Habitat data recorded for 45 specimens include 10 captures (22%) in well-drained primary forest, 3 captures (7%) in swampy primary forest, 10 captures (22%) in creekside primary forest, 2 captures (4%) in primary forest of unspecified character, and 20 captures (44%) in secondary vegetation. Microhabitat notes accompanying 28 specimens trapped at ground level record captures made under masses of fallen branches and other debris (8 specimens), under logs (5), in unsheltered sites in dense understory vegetation (5), beside logs (3), at the bases of trees (2), among stilt roots (1), on top of a log (1), inside a hollow log (1), on a smooth branch fallen over a stream (1), and under an overhanging stream bank (1).

Proechimys guyannensis (E. Geoffroy) Figures 77–80

Voucher Material: AMNH 266576, 266577, 266586, 266595, 267037, 267038, 267047; MNHN 1998.682–1998.684. Total = 10 specimens.

Identification: External and craniodental characters that distinguish this species from Proechimys cuvieri in French Guiana are discussed above and need not be repeated here.

Patton (1987) mapped the distribution of the guyannensis species group of Proechimys as extending throughout eastern Amazonia together with adjacent parts of the northern Venezuelan coast and the Brazilian Cerrado. Besides the type species, Patton listed the following taxa as group members: cherriei Thomas, roberti Thomas, vacillator Thomas, oris Thomas, warreni Thomas, boimensis Allen, arescens Osgood, riparum Moojen, and arabupu Moojen (see Cabrera, 1961, for bibliographic references). Patton believed that more than one valid species was represented by these names, emphasizing morphological differences between samples from the Guiana subregion (for which the oldest available name is guyannensis), and those from south of the Amazon (for which the oldest name is roberti).

The morphological trait frequencies reported herein for typical Proechimys guyannensis (from French Guiana) resemble those tabulated by Patton (1987) for guyannensis-group samples from north of the Amazon, which we provisionally regard as conspecific. By contrast, specimens from southeastern Amazonia have divergent trait frequencies (op. cit.) and also differ from north-bank samples in karyotypes and cytochrome-b sequences (Weksler et al., 2001). Apparently, all recently collected Proechimys from Venezuelan coastal rainforests (north of the Orinoco) are referable to other species groups (Aguilera and Corti, 1994; Aguilera et al., 1995; Corti and Aguilera, 1995). Therefore, P. guyannensis appears to be an Amazonian endemic largely, but perhaps not exclusively, distributed in the Guiana subregion.23

Remarks: Surinamese material of this species was identified as Proechimys warreni by Husson (1978), who used the name P. guyannensis for the larger species identified as P. cuvieri in this report (see the preceding account).

Although Proechimys guyannensis (Geoffroy, 1803) has long been recognized as a valid name (e.g., by Moojen, 1948; Hershkovitz, 1948b; Cabrera, 1961; Patton and Gardner, 1972; Patton, 1987), this epithet was rejected as unavailable by Woods (1993), who used the replacement name cayennensis Desmarest (1817) instead. In our opinion, Mus guyannensis and other names first published by Geoffroy (1803) are unambiguously available from that work for the reasons clearly explained by Hershkovitz (1955) and Holthuis (1963).

Other Specimens Examined: French Guiana—Arataye (MNHN 1983.381; USNM 548454–548456), Cayenne (MNHN 1983.374, 1983.376, 1995.1395 [holotype]), “Fleuve Oyapock” (MNHN 1983.375), Florida (MNHN 1981.48, 1982.601), Montsinéry (MNHN 1986.1124), Nancibo (MNHN 1986.1130), Petit Saut (MNHN 1998.312), St.-Eugène (MNHN 1994.128), no other locality data (MNHN 1981.88, 1981.103).

Field Observations: All of our definite records of Proechimys guyannensis at Paracou are based on collected specimens. Of our 10 vouchers, 3 (30%) were captured in Victor traps, 5 (50%) in Sherman traps, and 2 (20%) were shot. Eight of our vouchers (80%) were shot or trapped on the ground, but 2 (20%) were taken in traps tied to lianas 0.5–1.6 m above the ground. Recorded habitat data include 9 captures (90%) in well-drained primary forest and 1 (10%) in creekside primary forest. By comparison with Proechimys cuvieri, which we caught with almost equal frequency in primary forest and secondary growth, P. guyannensis was captured significantly more often in primary forest (table 47).

Sampling Results from Different Methods

None of the methods that we used to inventory the nonvolant mammal fauna at Paracou recorded all of the species known to occur in our study area. Instead, each method appeared to be maximally effective for some taxa and relatively ineffective for others. Below we summarize our principal sampling results method by method and explain some of the factors that affect our subsequent assessments of complementarity and completeness.

Conventional Trapping: Most conventional trapping at or near ground level (0–3 m by our convention) was accomplished in 1991 and 1992, when ten Victor/Sherman traplines were established at widely separated sites within our 3-km sampling radius. Whereas some Victor/Sherman traplines contained over 100 traps and extended for 2 km or more, most were shorter and contained fewer traps. The shortest interval that any Victor/Sherman trapline was operational was 8 nights, the longest interval was 14 nights, and the average interval was 12 nights. In total, we trapped on 71 dates in 1991 and 1992 with an average of 84 Victor/Sherman traps deployed per night. Supplementary ground-level trapping with other equipment (Tomahawks, Conibears, and legholds) was much more intensive in 1991 than in any subsequent year, but the numbers of supplementary traps deployed each night were not consistently recorded.

Combining results from conventional trapping with all commercially available equipment, we captured a total of 162 individuals representing 18 species of marsupials and rodents, including our only examples of Monodelphis brevicaudata and Mesomys sp. (table 48). Species accumulated quickly at first: we took 15 species in our first 74 captures, for an initial average rate of almost one new species per five individuals trapped (fig. 82). However, only three additional species were represented among the last 88 captures by this method, or about one new species per 30 individuals trapped.

In general, Victors and Shermans took smaller species than did Tomahawks, Conibears, and legholds (fig. 83), but there was some taxonomic overlap (table 50). For example, both classes of traps commonly took adult Proechimys, and juveniles of some large marsupials (e.g., Didelphis marsupialis, Metachirus nudicaudatus, Philander opossum) usually taken in Tomahawks and Conibears were also occasionally taken in Victors. However, Victors and Shermans never captured squirrels or any adult individuals of Didelphis, Metachirus, Philander, or Myoprocta. Similarly, Tomahawks, Conibears, and legholds never took Marmosops, Monodelphis, Neacomys, Oryzomys, or Mesomys at Paracou. Therefore, both kinds of equipment provided useful information about local diversity.

The unimpressive numbers of individuals captured by conventional trapping at Paracou is explained by a dramatic decline in trap success that we experienced over the course of our inventory. This phenomenon is best illustrated by the results obtained with Victor and Sherman traps, for which accurate counts of the numbers set each night are available. In effect, trap success (fig. 84, indicated by the slope of the solid line in the upper graph) changed abruptly when, after about 2600 trap-nights in 1991, traplines previously deployed for general collecting in productive habitats (e.g., well-drained primary forest) were shifted to less productive sites (e.g., along streams) and baited for certain target species expected to occur there. Furthermore, trap success in comparable situations was never as high in subsequent field seasons as it was in 1991. For example, we made 3.6 captures per 100 trap-nights using Victor and Sherman traps for general collecting in primary forest in 1991, but only 0.6 captures per 100 trap-nights using the same equipment in the same habitat in 1992 (an 83% decline). In total, we recorded only 128 captures in 5960 Victor/Sherman trap-nights at Paracou, for an overall success rate (captures per trap-night × 100) of 2.1%.

Species accumulation plotted against captures for Victor/Sherman trapping (fig. 84, lower graph) shows no sign of an asymptote despite the flattening terminal portion of the analogous graph for all conventional trapping combined (fig. 82). This paradox refects the late appearance in Victor traps of juvenile individuals of two species (Didelphis marsupialis, Myoprocta acouchy) previously taken only in Tomahawks, Conibears, or legholds, together with new taxa not taken by any other method (Monodelphis and Mesomys). Clearly, the apparent completeness of Victor/Sherman species sampling as assessed by trap-nights (fig. 84, dashed line in upper graph) is just an artifact of declining trap success.

Arboreal Trapping: We used arboreal platform traps only in 1993, when 25 trapping stations were established at approximately 20-m intervals along a transect through both well-drained and swampy primary forest. Trapping platforms were installed between 7.2 m and 19 m above the ground, with an average platform height of 13.7 m. From the date when the first arboreal trap station was operational (29 July) to the date when all platform trapping was discontinued (15 September), we trapped continuously for 47 nights with an average of 21 trapping stations operational per night. In total, we logged 1002 station-nights, which is equivalent to 2004 trap-nights if both Shermans and Tomahawks are counted. Seventeen individual marsupials and rodents were captured, for an overall success rate (captures per trap-night × 100) of only 0.8%.

Our 17 arboreal trap captures represent six species, none of which were taken exclusively by this method (table 48). The temporal pattern of captures (fig. 85, solid line in upper graph) suggests that traps were avoided for the first two weeks after installation (our first arboreal capture was made after 454 trap-nights) and that more prolonged trapping could have resulted in many more captures (trap success was higher in the last week of operation than in any previous interval). Although species accumulation plotted against captures (fig. 85, bottom graph) shows some indication of levelling, the number of individuals taken is too small to plausibly suggest any asymptotic value for taxonomic diversity obtainable by this method.

Pitfall Trapping: This method was used only in 1993, when five pitfall traplines were installed at widely separated locations in our study area: two in well-drained primary forest, one in swampy primary forest, and two in creekside primary forest. From the date when the first pitfall trapline was operational (21 July) to the date when all pitfalls were taken up (13 September), we trapped for 54 consecutive nights. Except for the first week of trapping (when the lines were still being installed), 55 buckets per night (11 in each line) were deployed as pitfalls throughout this interval. In total, we logged 2783 bucket-nights and captured 45 individual mammals, for an overall success rate (captures per bucket-night × 100) of 1.6%.

Our 45 mammalian pitfall captures24 represent 12 species of marsupials and rodents, of which two (Neacomys dubosti and Neusticomys oyapocki) were not taken by any other method (table 48). Most individual mammals captured in pitfalls were small, the great majority (84%) weighing less than 30 g (fig. 83). Pitfall captures accumulated rapidly in the first few weeks of trapping (fig. 86, solid line in upper graph), when we recorded about three captures per 100 bucket-nights on average; thereafter, our capture rate declined to a more-or-less steady rate of about 0.8 captures per 100 bucket-nights. Although species accumulation plotted against bucket-nights (fig. 86, dashed line in upper graph) suggests that faunal sampling with pitfalls was nearly complete, species accumulation plotted against captures (fig. 86, lower graph) does not show a convincing asymptote.

Diurnal Hunting: Because most of our sightings and collections of diurnal mammals resulted from chance encounters, sampling effort is hard to quantify for this method. On average, all inventory personnel spent at least several daylight hours in the forest per calendar date, so the total accumulation of time available for diurnal mammal observations was considerable: about 2500 hours (634 person-days × 4 hours/person-day) is a plausible estimate. However, much of the time that we spent in the forest by day was not conducive to observing cryptic species (e.g., sloths and pygmy squirrels), nor were all personnel alert to fleeting encounters with wary taxa locally persecuted as game (e.g., deer and monkeys). Given these and other problems with quantifying observational effort, field days are probably no worse than any alternative unit.

Diurnal mammal sightings accumulated at an initially rapid rate during our first several weeks at Paracou, but few new species were subsequently recorded by this method (fig. 87). Over the entire course of our inventory, we recorded 16 species from diurnal sightings (table 48). Although many of these were not found by us using other inventory methods, most were previously known to the local residents whom we interviewed. Oddly enough, the only species uniquely recorded by a diurnal observation was the usually nocturnal rodent Makalata didelphoides.

Nocturnal Hunting: Most of our observations and collections of nonvolant nocturnal mammals resulted from deliberate hunting (including sight censuses when few or no specimens were taken). Unfortunately, the duration of nightly hunts was not quantified in 1991 when this method was used most intensively. However, an estimate can be based on the fact that RSV and DPL both hunted every night on 58 consecutive dates in that year; assuming a typical hunt duration of 3.0 hours (the known average for hunts by the same personnel in 1992) yields a probable total effort of about 350 hours devoted to this method in 1991. Hunt durations were consistently recorded in subsequent field seasons: we devoted 96.7 hours to this method in 1992, 79.2 hours in 1993, and 64.4 hours in 1994. In total, we systematically hunted at night on 123 dates from 1991 to 1994 at Paracou, for a total cumulative effort of about 590 hours.

Nevertheless, chance observations made while bat netting contributed so many important records of nonvolant mammals (including our first of Caluromys philander, Hyladelphys kalinowskii, Oligoryzomys fulvescens, and Priodontes maximus) that it is misleading to ignore this source of nocturnal observations. Indeed, whereas our deliberate hunting effort per field season declined from 1991 to 1994, our bat netting effort increased in almost direct proportion (all inventory personnel were usually engaged in one activity or the other after dusk, seldom returning to camp before midnight), so the total time actually or potentially available to observe nonvolant nocturnal mammals per field date probably remained more-or-less constant throughout our inventory. It therefore makes sense to plot species accumulation by nocturnal observation as a function of calendar dates rather than dedicated search time (fig. 87, solid line).

Over the 202 dates that we worked at Paracou from 1991 to 1994, we recorded 31 species of nonvolant mammals by nocturnal observation, five of which were not recorded by any other method (Gracilinanus emiliae, Dasypus kappleri, Leopardus wiedii, Oligoryzomys fulvescens, Coendou melanurus; table 48). Although our rate of species accumulation by nocturnal observation was steepest in the first several weeks and decreased subsequently, no convincing asymptotic value is indicated. In total, we recorded 271 nocturnal observations that were unambiguously identifiable to species, but this sum does not include numerous unvouchered observations (mostly of common taxa) from 1991, nor did we record chance encounters with common species while bat netting. Therefore, only the smaller frequency classes for this method (table 48) represent accurate counts.

More interpretable frequency data were obtained from 1992 to 1994, when every identifiable observation (including sightings and distinctive sounds, except howler monkey vocalizations) were recorded in the course of 240.3 hours of systematic nocturnal hunting (table 51). Unidentified sightings and sounds (e.g., distant eyeshine, crashing noises in the undergrowth) were not consistently recorded. On average, we made 1.1 identifiable observations per hour of nighttime hunting from 1992 to 1994, a rate that did not differ substantially among the four persons principally involved in this activity (RSV, DPL, R. W. Kays, and L. H. Emmons).

Most identifiable nocturnal observations at Paracou represent a few common species with bright eyeshine, distinctive vocalizations, and/or noisy habits. Thus, records of Potos flavus, Cuniculus paca, Philander opossum, and Chironectes minimus account for over half of the data in table 51. By contrast, ten species represented by only one record each include some that may be genuinely uncommon, but others that are simply inconspicuous (silent, small, and/or with weak eyeshine), difficult to identify without specimens in hand, or usually diurnal (e.g., Gracilinanus emiliae, Marmosa murina, Nasua nasua, Neacomys paracou, Oecomys rutilus, Oryzomys megacephalus, Proechimys guyannensis, Myoprocta acouchy).

Interviews: Interviews resulted in 33 positive identifications of local species, of which 12 were not recorded by any other method (Bradypus tridactylus, Cabassous unicinctus, Cyclopes didactylus, Myrmecophaga tridactyla, Cebus apella, Saimiri sciureus, Speothos venaticus, Herpailurus yagouarondi, Leopardus pardalis, Puma concolor, Galictis vittata, Tapirus terrestris; table 48). Most of the species uniquely recorded from interviews are either uncommon throughout the Neotropics (e.g., Cabassous unicinctus, Galictis vittata) or are locally uncommon from overhunting (e.g., Cebus apella, Tapirus terrestris); others, however, are perhaps common but seldom seen because they are cryptic (e.g., Bradypus tridactylus) or wary (e.g., Leopardus pardalis). Only two relatively large and easily identified local species (Chironectes minimus, Coendou melanurus) were not familiar to our interviewees.

Although interviewees often estimated the relative frequency with which mammals were observed (e.g., see accounts for Bradypus tridactylus and Mazama gouazoubira, above), accurate counts of second-hand observations were only available for the rarest species. Interview records known to have been based on single observations include those for Cabassous unicinctus, Priodontes maximus, Pithecia pithecia, Saimiri sciureus, and Galictis vittata. Only two second-hand observations are known for Cyclopes didactylus and Nasua nasua; three for Speothos venaticus and Leopardus pardalis; and four for Myrmecophaga tridactyla, Panthera onca, and Puma concolor.

Summary: Although each inventory method that we used to sample nonvolant mammal diversity at Paracou produced a different list of species, some methods were clearly more productive and/or produced more distinctive lists than others. Relevant quantitative comparisons (table 52) include the total number of species recorded by each method, the number of unique species, and pairwise complementarity values. The latter reflect the extent to which two methods provide nonredundant diversity information: low complementarity values imply high redundancy, whereas high complementarity implies low redundancy (see table footnote for computational details).

Clearly, interviews produced more species records (33) than any other single method used in our nonvolant inventory, as well as the largest number (12) and proportion (36%) of unique species. Nocturnal hunting ranks next by these criteria, followed by conventional trapping. Pitfall trapping and diurnal hunting each recorded some unique species, but arboreal trapping did not and was also least productive in terms of total records.

The lowest complementarity value calculated from our sampling data (60%) corresponds to the comparison of diurnal hunting with interviews, methods that were substantially redundant for the obvious reason that almost all diurnally active mammals at Paracou were known to local inhabitants. Similarly, the second-lowest complementarity value we computed (64%) corresponds to the comparison of pitfall trapping with conventional trapping, methods that produced broadly overlapping lists of small marsupials and rodents. By contrast, other pairs of methods (such as pitfall trapping and interviews on the one hand, or arboreal trapping and diurnal hunting on the other) are 100% complementary because they produced no recorded species in common.

Overall, 22 species were each recorded by only one of the six methods we used to sample the nonvolant fauna at Paracou. If the two species uniquely taken in conventional traps by previous researchers (Nectomys melanius, Oryzomys macconnelli) are added to this total, then about 38% of the known nonvolant fauna (64 species) were recorded by a single method each. If miscellaneous records (e.g., scavenged material, occasional observations of distinctive spoor) are discounted for an additional four species (Marmosa murina, Panthera onca, Tayassu pecari, Coendou prehensilis; table 48), then a total of 28 species (44% of the known fauna) was not redundantly recorded. Of the 36 redundantly recorded species, 20 were recorded by two methods each, 14 by three methods, and 2 by four methods.

Estimating Completeness

Because none of our species accumulation graphs shows a convincing asymptote, it is reasonable to expect that more species could have been recorded with additional sampling effort by each method. Visual comparisons of these graphs, however, suggest that asymptotic values for species accumulation may have been more closely approached by some methods (e.g., conventional trapping, fig. 82) than by others (e.g., pitfall trapping, fig. 86). In order to assess sampling completeness by less subjective criteria, we compared observed species counts with predicted values obtained by nonparametric extrapolations.

The logic of extrapolating unobserved species richness from incomplete samples was recently reviewed by Colwell and Coddington (1994), and the application of certain nonparametric methods to our Paracou bat data was explained by Simmons and Voss (1998: 182–184). As in that study, we first consider the results of applying Chao's (1984) estimator, which is based on the total number of observed species, S_obs, the number of singletons (species recorded only once), a, and the number of doubletons (species recorded only twice), b. The expected total number of species, S^*, is then given by the expression

The results of such calculations (table 53) suggest that sampling by some inventory methods was indeed more complete than sampling by others, although confidence intervals are wide enough to include the observed number of species in every case. Of particular interest is the apparent near-completeness of conventional trapping on the one hand, and the relative incompleteness of pitfall trapping on the other, estimates that more-or-less coincide with our subjective interpretation of species accumulation graphs. On the other hand, the rather high completeness estimate for arboreal trapping is difficult to reconcile with our scant capture success using this method.

Given the inherent uncertainty of all extrapolation procedures, other estimates of S^* are of interest in order to bracket the range of plausible inferences that can be based on our sampling data. We therefore used three additional nonparametric species-richness estimators to assess inventory completeness (table 54). By each estimator, our conventional-trapping data appear to be the most complete (although not always by a large margin), and our pitfall-trapping data the least complete. The remaining data partitions (arboreal trapping, diurnal hunting, nocturnal hunting, interviews) are not consistently ranked inter se by completeness. Perhaps the best synthesis of these results is obtained by taking average completeness values across all four sets of extrapolation figures, which yields the sequence: conventional trapping > arboreal trapping and interviews > nocturnal hunting > diurnal hunting > pitfall trapping.

Pooling sampling results from all nonvolant inventory methods, eight species are each represented in our data by single records (Gracilinanus emiliae, Cabassous unicinctus, Saimiri sciureus, Galictis vittata, Neacomys dubosti, Coendou melanurus, Makalata didelphoides, Mesomys sp.) and four species are represented by just two records each (Priodontes maximus, Cyclopes didactylus, Nectomys melanius, Neusticomys oyapocki). Using Chao's (1984) extrapolation procedure (table 53), the predicted total nonvolant species richness at Paracou is 72 species, an estimate that suggests our nonvolant inventory is 89% complete. Other nonparametric estimators yield S^* values in the range of 69–74 species and corresponding completeness estimates of 86–93% (table 54).

Obviously, neither the presence of additional nonvolant species in the Paracou fauna nor their absence can be proven without additional fieldwork. However, it is noteworthy that the numbers of missing species implied by these statistical extrapolations (5–10, calculated as S^* minus S_obs) correspond closely with the number of additional species that could be expected to occur locally on the basis of known geographic and ecological distributions (appendix 1). Presumably, that short list (Didelphis albiventris, Marmosa lepida, Cebus olivaceus, Leopardus tigrinus, Oecomys bicolor, Oecomys rex, Rhipidomys leucodactylus, Echimys chrysurus, Isothrix sinnamariensis) includes the likeliest candidates for discovery by future inventory efforts at Paracou.

Taxonomic Composition and Biogeography

In terms of higher-level taxonomic composition, the nonvolant mammals of Paracou represent a typical Neotropical lowland rainforest fauna. Seven of the eight orders commonly found in Central and South American rainforests (Marsupialia, Xenarthra, Primates, Carnivora, Perissodactyla, Artiodactyla, Rodentia) are represented in our inventory; only Lagomorpha, which is never represented by more than a single species at any rainforest site, is absent. All of the families and genera of nonvolant mammals in the Paracou fauna are likewise widespread rainforest taxa.

In terms of species composition, Paracou clusters (fig. 89) with five other faunas from the Guiana subregion of Amazonia, next with a single southeastern Amazonian site, then with a discrete grouping of three western Amazonian localities, and lastly with a pair of Central American faunas. Pairwise measures of faunal resemblance (table 56) indicate that the Paracou nonvolant inventory is most similar to that from Arataye, another French Guianan locality only 136 km to the SSE, and least similar to that from La Selva, a Costa Rican locality almost 3400 km to the WNW. Overall, these results clearly indicate a pattern of increasing faunal resemblance with increasing geographic proximity, and that the Paracou fauna in particular is most similar to others from northeastern Amazonia.

Paracou nonvolant mammals can be sorted (table 57) into seven groups based on their pattern of distribution among the four Neotropical lowland rainforest regions recognized by Voss and Emmons (1996). The largest distributional group (pattern 1) includes 24 species that are endemic to Amazonia (A), or are even more narrowly restricted to the Guianan (G) subregion of Amazonia: Hyladelphys kalinowskii (A), Marmosops parvidens (A), Marmosops pinheiroi (A), Monodelphis brevicaudata (G), Dasypus kappleri (A), Saguinus midas (G), Ateles paniscus (G), Pithecia pithecia (G), Saimiri sciureus (A), Sciurillus pusillus (A), Sciurus aestuans (G), Neacomys dubosti (G), Neacomys paracou (G), Neusticomys oyapocki (G), Oecomys auyantepui (G), Oecomys rutilus (G), Oryzomys macconnelli (A), Oryzomys yunganus (A), Rhipidomys nitela (A), Coendou melanurus (G), Myoprocta acouchy (A), Mesomys sp. (G), Proechimys cuvieri (A), and Proechimys guyannensis (A). Therefore, a substantial fraction (38%) of the Paracou fauna is distinctively Amazonian or Guianan in character.

The next-largest distributional group (pattern 2) includes 18 species that (as currently recognized by taxonomists) occur in all four Neotropical rainforest regions. These essentially pan-Neotropical taxa include Chironectes minimus, Metachirus nudicaudatus, Dasypus novemcinctus, Myrmecophaga tridactyla, Speothos venaticus, Herpailurus yagouarondi, Leopardus pardalis, Leopardus wiedii, Panthera onca, Puma concolor, Eira barbara, Galictis vittata, Potos flavus, Mazama americana, Mazama gouazoubira, Pecari tajacu, Tayassu pecari, and Cuniculus paca. Pending future revisionary study (see below), this group of species, comprising about 28% of the fauna, is zoogeographically uninformative.

By contrast, other Paracou nonvolant mammals are neither Amazonian endemics (as currently recognized) nor pan-Neotropical, but have distributions that suggest a variety of inter-regional connections. Thus, nine species occur in both Amazonian and Coastal Venezuelan rainforests, but not in the other two regions (pattern 3: Gracilinanus emiliae, Bradypus tridactylus, Choloepus didactylus, Cabassous unicinctus, Priodontes maximus, Alouatta seniculus, Nectomys melanius, Oryzomys megacephalus, Makalata didelphoides), seven species occur in all rainforest regions except the trans-Andean (pattern 4: Caluromys philander, Marmosa murina, Micoureus demerarae, Tamandua tetradactyla, Tapirus terrestris, Coendou prehensilis, Dasyprocta leporina), and four species occur in all rainforests except the Atlantic region of southeastern Brazil (pattern 5: Didelphis marsupialis, Philander opossum, Cyclopes didactylus, Oligoryzomys fulvescens). One species (Nasua nasua) occurs in all rainforest regions except the Coastal Venezuelan (pattern 6), and another (Cebus apella) occurs only in Amazonia and the Atlantic rainforest region of southeastern Brazil (pattern 7).25

Of course, not all of the nonvolant species that belong to the Paracou fauna are restricted to rainforest habitats: many are eurytopic and occur in a wide range of other Neotropical biomes. For example, at least 16 of the 18 species with distributional pattern 2 are also known from the Llanos (Eisenberg et al., 1979; Ibáñez, 1981), the Caatinga (Willig and Mares, 1989), the Cerrado (Redford, 1983; Fonseca and Redford, 1984; Mares et al., 1989), or the Pantanal (Schaller, 1983). On the other hand, none of the taxa that we recorded at Paracou are consistently associated elsewhere with nonforest habitats. A few Paracou species (Marmosa murina, Saimiri sciureus, Oligoryzomys fulvescens) are perhaps never found in unbroken tracts of Guianan rainforest, but these can be characterized as inhabitants of the rainforest edge that probably occur along riverbanks, blow-downs, and other natural openings at undisturbed sites. Several unambiguously nonforest mammals might inhabit the coastal savannas and salt marshes just north of our study area (e.g., Odocoileus cariacou, Holochilus sciureus, Zygodontomys brevicauda), but none were encountered within our sampling radius. Therefore, although some nonvolant Paracou mammals are not rainforest specialists, the fauna as a whole includes only species that are known to occur in rainforested landscapes and appears to constitute an ecologically homogeneous assemblage.

Species Richness

Comparisons of nonvolant mammal species richness among Neotropical rainforest inventories are complicated by many factors. As documented in the preceding analyses of sampling results, different inventory methods effectively sample different sets of species, and increased effort with any method generally produces more species. Therefore, even in the absence of real intersite diversity differences, long-term inventories (or those using more methods) would be expected to obtain larger species lists than short-term inventories (or those using fewer methods). Differences in the ecological scope of inventory fieldwork can also affect species richness comparisons because sites with greater habitat diversity will tend to be richer than sites with fewer habitats. Finally, large mammals are likely to be more abundant, and therefore more easily recorded, at undisturbed localities than at sites which have been partially defaunated by hunting.

Published descriptions of inventory fieldwork (cited in the footnotes to table 55) document intersite differences in all of these potentially confounding factors: (1) Although fieldworkers at all sites used conventional trapping, diurnal hunting, and nocturnal hunting to sample nonvolant mammal diversity, other methods (pitfall trapping, arboreal trapping, and interviews) were less consistently applied, and no previous inventory has used all six methods in combination. (2) Overall sampling effort is hard to quantify for nonvolant mammal surveys, but the duration of inventory work at each site (a monotonic correlate of effort) ranged from less than a year at some sites to many years at others. (3) Whereas most sites were located on rivers or lakes that provide habitat for semiaquatic, riparian, and forest-edge species, others were in well-drained uplands remote from large bodies of water. (4) Subsistence hunting probably eliminated some primates and other large species prior to inventory fieldwork at several sites, whereas other sites were pristine.

As a consequence of such disparities, tabulated species counts are undoubtedly affected by sampling artifacts to a greater or lesser extent. The most that can be said from these data, without introducing additional assumptions or corrections, is that nonvolant species richness at Paracou more nearly resembles that from other well-sampled sites in the Guiana subregion of Amazonia (e.g., Kartabo, Arataye) than they do Central American sites (which have fewer species for most orders) or western Amazonian sites (which tend to have more species). To better exemplify regional and subregional differences in species richness, we focus our comparisons on two sites where prolonged fieldwork may have compensated in some degree for methodological incompleteness (table 58).

La Selva (with over 30 years of faunal observations) probably provides the best-sampled nonvolant mammal fauna from any Central American lowland rainforest, and Manu (with over 20 years) provides an appropriate western Amazonian counterpart. Although it is not possible to estimate sampling completeness at either of these sites by extrapolation (requisite frequency data are not available), surrogate estimates can be based on the number of rainforest species expected at each locality from geographic range data. Such calculations (see Voss and Emmons, 1996: table 10) suggest that the La Selva nonvolant inventory is about 95% complete and the Manu nonvolant inventory about 85% complete. By any of the extrapolations computed earlier in this report, the Paracou nonvolant fauna falls between La Selva and Manu in estimated completeness (86–93%). Therefore, known species richness and estimated completeness are inversely correlated among these three inventories.

Graphical comparisons (fig. 90) summarize the principal higher-taxonomic patterns of species richness among La Selva, Paracou, and Manu. Intersite differences in known species richness are most pronounced for Marsupialia, Primates, and Glires (Rodentia + Lagomorpha). For each of these clades, La Selva is the least diverse site; Paracou and Manu are equivalent in known marsupial diversity, but Manu is more diverse than Paracou for both primates and rodents. Paracou has the most known xenarthran species of any site, but the least number of known carnivore species; ungulate species richness is invariant. Overall, known nonvolant species richness increases by 12 from La Selva to Paracou, and by 15 from Paracou to Manu; from La Selva to Manu, the net increase in species richness (27 species) amounts to 52% of the former fauna. Given the previously noted inverse correlation between known species richness and estimated inventory completeness, the true diversity gradient from La Selva to Paracou to Manu is likely to be even steeper than these quantities imply.

Whereas Glires is the most diverse clade at each locality (accounting for an almost constant proportion, 33–35%, of the local nonvolant species), the remaining clades are not consistently ranked by species richness. At La Selva, Carnivora is the most speciose after Glires, distantly followed by Xenarthra; Marsupialia and Ungulata are next, and Primates is last. At Paracou, however, Marsupialia is more speciose than Carnivora, which is closely followed by Xenarthra; Primates is next, followed by Ungulata. At Manu, Carnivora is again the most speciose clade after Glires, but Primates is almost as diverse, followed closely by Marsupialia; Xenarthra is next and Ungulata is last.

Although some of these faunal differences are probably artifactual,26 the short lists of expected species for each locality (Voss and Emmons, 1996: appendices 2, 10; this report: appendix 1) suggest that future inventory work is unlikely to change the overall picture very much. For example, primates are likely to remain in last place at La Selva, as are ungulates at Paracou and Manu. A few clades may switch ranks at some sites (e.g., marsupials may prove to be more diverse than primates at Manu), but the larger diversity differences (among clades at each locality, and among localities for the same clade) are probably robust.

Clearly, exemplar comparisons can be over-interpreted, and it would be unwise to make sweeping generalizations about geographic diversity patterns based on just these three sites. Nevertheless, the nonvolant mammal fauna at Paracou is probably representative of that throughout most of the Guiana subregion of Amazonia, which would appear to be richer in species (especially of marsupials and rodents) than Central American rainforest faunas, but less species-rich (especially in primates) than western Amazonian faunas. Although such regional and subregional diversity contrasts could have historical explanations, it is nevertheless appropriate to explore their consequences for the ecological structure of contemporary communities, which may suggest other causal factors.

Trophic Guilds and Other Topics

Several previous studies have tabulated the ecobehavioral attributes of nonvolant rainforest mammals at various Neotropical inventory sites (e.g., Eisenberg and Thorington, 1973; Janson and Emmons, 1990; Peres, 1999), but none have provided a comprehensive trophic classification. Indeed, there are many problems in attempting to do so. Whereas most rainforest bats can be sorted into guilds based on reasonably straightforward dietary and behavioral attributes (Bonaccorso, 1979; Kalko et al., 1996; Simmons and Voss, 1998), many nonvolant mammals are harder to characterize ecologically. Numerous species feed on fruit when it is seasonally abundant, but switch to different resources when fruit is scarce; for example, woody browse (Mazama americana; Bodmer, 1990), seeds (Dasyprocta leporina; Henry, 1999), or animals (many didelphids; Atramentowicz, 1988). The poorly documented or completely unknown food habits of some nonvolant taxa (e.g., Glironia, Sciurillus, many murids) is another obstacle to trophic classification.

Foraging substrate is an important factor in defining nonvolant guilds that introduces additional ambiguities because many species do not restrict their activities to one substrate type. Whereas some taxa are unambiguously terrestrial or arboreal, numerous trapping and observational studies (e.g., Janson and Emmons, 1990; Malcolm, 1991; Woodman et al., 1995; this study) suggest that others are scansorial—active both in trees and on the ground. Similarly, whereas some species that forage primarily in water are unambiguously semiaquatic (Chironectes, Lontra), others forage at or near the water's edge (Procyon, Nectomys, Hydrochoerus) and are perhaps better described as riparian. Unfortunately, it is difficult to classify species by such subtle (but perhaps important) behavioral distinctions in the frequent absence of relevant field studies.

Despite these and other problems, some guilds of nonvolant rainforest mammals have been widely and more-or-less consistently recognized (e.g., arboreal folivores; Montgomery, 1978), and others can readily be defined to label obvious clusters of ecologically similar species (e.g., terrestrial granivore/frugivores; Peres, 1999). By adopting or modifying some previously recognized ecobehavioral categories and recognizing a few new ones, we were able to sort most of the nonvolant taxa from La Selva, Paracou, and Manu into a reasonably small number of guilds (table 59). Obviously, some of these are very broadly defined—such as terrestrial animalivores—and could be further subdivided for more detailed analyses. Alternatively, many could be combined for the purposes of broader generalizations (e.g., about primary versus secondary consumers). In effect, we tried to achieve a balance between excessively fine distinctions on the one hand and insufficient resolution on the other while minimizing the number of arbitrary decisions necessary to assign taxa to different categories.

We primarily used natural history data summarized by Emmons (1997) to assign taxa to guilds, but we consulted other references for supplementary information. When two or more references suggested different guild assignments for the same taxon, we based our final assignment on the most detailed available field study or the most thorough literature review. Unavoidably, many guild assignments are arguable (how much more fruit in the diet should distinguish a frugivore from an omnivore?), and some are frankly speculative (e.g., dietary categories for most murids). However, insofar as possible we used consistent criteria from locality to locality, so the results are hopefully unbiased for comparative purposes.

Arboreal Folivores: This guild is represented by just three species at Paracou, two of which are diurnal (Bradypus tridactylus and Alouatta seniculus) and one nocturnal (Choloepus didactylus). The same number of arboreal folivores, belonging to the same genera, are present at most Neotropical rainforest sites except where a fourth taxon (Dactylomys) feeds in bamboo and other dense vegetation growing on rich alluvial soils (as at Manu; Emmons, 1997).

Although it is probable that howler populations had been depleted by hunting in our study area, it is unlikely that local sloth populations were significantly affected. For Choloepus, the only taxon for which intersite density comparisons are possible, our sighting rate by walked nocturnal census (0.08 individuals per 10 hours; table 51) is about one-eighth the rate reported by Emmons (1984)27 at Limoncocha, a western Amazonian locality with rich soils, and one-fourth the average sighting rate on Barro Colorado Island (Glanz, 1982). Although our infrequent views of Choloepus and our inability to obtain even a single sighting of Bradypus suggest that sloths are locally rare, these taxa were among the commonest mammals rescued from primary forest flooded by a hydroelectric dam at Petit Saut, about 28 km SSW of our study area (Vié, 1999), where their joint density was estimated to be about 2.6 individuals/km² (Taube et al., 1999).

Arboreal Frugivores: The most conspicuous intersite differences in trophic composition concern this guild, which contains only five species at Paracou, a mere 8% of the total nonvolant fauna. By contrast, six arboreal frugivores constitute 12% of the less speciose nonvolant fauna of La Selva, and 13 species constitute 16% of the more speciose fauna of Manu. The striking deficit of arboreal frugivores at Paracou is partly, but not entirely, a consequence of the previously noted faunal differences in primate diversity. Thus, although the Manu fauna includes several frugivorous primate genera not found at Paracou, it also includes additional nonprimate arboreal frugivores (Caluromysiops, Bassaricyon), one of which (Bassaricyon) also occurs at La Selva. Furthermore, whereas western Amazonian tamarins (members of the Saguinus fuscicollis and S. mystax groups, both of which occur at Manu) appear to be primarily frugivorous (Peres, 1993a), the single tamarin that occurs at Paracou, S. midas, is not (Pack et al., 1999). Because only one arboreal frugivore (Cebus olivaceus) is among the species that might have been locally extirpated by hunting prior to our inventory (or that might yet be discovered by future fieldwork; appendix 1), it seems unlikely that this guild is grossly underrepresented at Paracou as a consequence of sampling inadequacy.

All of the diurnal arboreal frugivores in our study area—Ateles paniscus, Cebus apella, Saimiri sciureus—were very uncommon, the first two probably as a consequence of persistent hunting. The only commonly encountered species belonging to this guild (Caluromys philander, Potos flavus) were both nocturnal. In fact, our sighting rate for kinkajous at Paracou (3.08 per 10 hours of walked nocturnal census) is comparable to the highest rate reported by Glanz (1982: table 4) from Barro Colorado Island, and is higher than any figures that Emmons (1984: table 4) reported from four Amazonian localities. This result seems anomalous in the context of other indications that the Paracou environment is not rich in fruit resources, but it is possible that kinkajous are locally abundant as a consequence of ecological release following the virtual extirpation of competing diurnal species. Our sighting rate for Caluromys philander (0.17 per 10 hours), however, is low by comparison with most reported rates for congeneric species on Barro Colorado Island (C. derbianus; Glanz, 1982: table 4) and in western Amazonia (C. lanatus; Emmons, 1984: table 4). It is noteworthy that trapping and sighting data for C. philander at Paracou are consistent with the results of other recent field studies (Charles-Dominique et al., 1981; Malcolm, 1991) in suggesting that this is one of the most exclusively arboreal of Amazonian marsupials.

Arboreal Granivore/Frugivores: This guild consists of nine species at Paracou, representing about 14% of the known nonvolant fauna. Roughly similar proportions of the La Selva and Manu faunas are arboreal granivore/frugivores, despite geographic differences in taxonomic composition. Among other contrasts, Central American faunas contain no granivorous monkeys, whereas squirrels are more speciose in both Central America and western Amazonia than they are in the Guiana subregion.

Members of this guild at Paracou include two diurnal species, one a primate (Pithecia pithecia) and the other a squirrel (Sciurus aestuans); the remainder, three murids (Oecomys auyantepui, O. rutilus, Rhipidomys nitela) and four caviomorphs (Coendou melanurus, C. prehensilis, Makalata didelphoides, Mesomys sp.), are all nocturnal. All of the members of this guild except Sciurus aestuans and the murids were very uncommon at Paracou. Most of our scant field observations for these taxa are more-or-less consistent with previously published natural history data (summarized by Emmons, 1997), but the diurnal sighting of Makalata didelphoides is an inexplicable oddity.

Arboreal Gummivores: No mammal in the Guiana subregion of Amazonia is known to subsist primarily on saps and gums, although it is possible that the unknown diet of Sciurillus pusillus includes some type of plant exudate (Emmons, 1997).

Arboreal Omnivores: Five Paracou species are assigned to this guild, constituting 8% of the known fauna. Proportions of arboreal omnivores in the La Selva and Manu faunas are similar.

At Paracou, arboreal omnivores consist of a single diurnal primate (Saguinus midas) and at least four nocturnal marsupials (Didelphis marsupialis, Marmosa murina, Micoureus demerarae, Philander opossum). Of the marsupials, our data suggest that Micoureus is the most consistently arboreal, followed by Didelphis and Philander; Paracou records are too scanty to place Marmosa murina in this ranking, but Charles-Dominique et al. (1981) suggested that this species is primarily active in the forest understory, like P. opossum. Most members of this guild were common in our study area, although numbers of D. marsupialis declined abruptly after 1991; only M. murina was consistently uncommon. Our nocturnal sighting rate for P. opossum (0.96 per 10 hours) is comparable to the highest rate reported for this species from Barro Colorado Island (Glanz, 1982: table 4) and is higher than either of the two previously reported rates from Amazonian localities (Emmons, 1984: table 4).

Arboreal Insectivores: This guild includes only two species at Paracou and at all other adequately sampled Neotropical rainforest localities. Our field data contribute no new natural history information about either Cyclopes or Tamandua.

Terrestrial Herbivores: The Paracou fauna includes no species assignable to this guild, nor (apparently) does any other known rainforest fauna from the Guiana subregion of Amazonia.

Terrestrial Granivore/Frugivores: Thirteen species belong to this guild at Paracou, where they constitute 20% of the nonvolant fauna. Similar proportions of terrestrial granivore/frugivores are present in the La Selva and Manu faunas despite differences in the taxonomic composition of this group from site to site.

At Paracou, terrestrial granivore/frugivores include four large diurnal species (Pecari tajacu, Tayassu pecari, Dasyprocta leporina, Myoprocta acouchy), one large nocturnal species (Cuniculus paca), and eight small nocturnal species (Neacomys dubosti, N. paracou, Oligoryzomys fulvescens, Oryzomys macconnelli, O. megacephalus, O. yunganus, Proechimys cuvieri, P. guyannensis). Both of the peccaries, locally hunted for meat, were uncommon or wary; only Dasyprocta and Myoprocta were regularly seen (or heard) by us, but neither was really common. By contrast, our sighting rate for pacas (1.1 individuals per 10 hours of walked nocturnal census) is about equal to the average sighting rate (1.3 per 10 hours) for five Amazonian inventories tabulated by Emmons (1984: table 4), and is similar to sighting rates from two of the three nocturnal census episodes recorded for Barro Colorado Island (Glanz, 1982: table 4). It is possible that pacas are relatively common at Paracou for the same reason that kinkajous are, because local populations of most competing species in the diurnal fauna have been depleted by hunting.

Small nocturnal members of this guild accounted for 68% of our captures using all conventional traps, and 80% of our captures using just Victor and Sherman traps. Interestingly, this is the only guild at Paracou where sympatry between congeneric forms is the rule rather than the exception: Neacomys, Oryzomys, and Proechimys are each represented by two or more species in the local fauna. Although most of the small nocturnal granivore/frugivores we recorded at Paracou occur syntopically in well-drained primary forest (the only exception is Oligoryzomys fulvescens, found exclusively in roadside secondary growth), our capture results suggest that some closely related species may differ in their use of other habitats. Thus, Oryzomys yunganus occurs more frequently than O. megacephalus in swampy and creekside forest, whereas Proechimys cuvieri occurs more frequently than P. guyannensis in secondary growth. To the best of our knowledge, ecological differences between O. megacephalus and O. yunganus have not been observed previously, but Malcolm's (1992) report of differential habitat use by P. cuvieri and P. guyannensis near Manaus agrees with our sampling results for these species. Neacomys paracou and N. dubosti occur syntopically at Paracou and are known from sympatric collections elsewhere in the Guiana subregion, but available information is too sparse to suggest how (or if) these congeners differ ecologically.

Terrestrial Folivore/Frugivores: Three ungulate species occupy this guild at all compared inventory sites, where they account for only 4–6% of each fauna. All of the Paracou species were persecuted by local hunters and only two were actually observed by us. Our fieldnotes contain no noteworthy natural history information about either Mazama or Tapirus, but a recent analysis of tapir diets in French Guiana (Henry et al., 2000) was based in part on material collected near Paracou.

Terrestrial Omnivores: Two members of this guild are present at Paracou, La Selva, and Manu, where they constitute an almost constant minor fraction of the known nonvolant fauna. Although both Paracou species (Eira barbara and Nasua nasua) are large, diurnal, and not locally hunted for food, neither was commonly observed in our study and no noteworthy natural history observations were recorded about them.

Terrestrial Animalivores: This large and admittedly heterogeneous guild comprises a higher fraction (25%) of the fauna at Paracou than at either La Selva or Manu. Three terrestrial animalivores at Paracou are diurnal (Monodelphis, Speothos, Galictis), six are active both by day and at night (Myrmecophaga, Herpailurus, Leopardus pardalis, Panthera, Puma), and the remaining eight are nocturnal (Marmosops parvidens, M. pinheiroi, Metachirus nudicaudatus, Dasypus kappleri, D. novemcinctus, Cabassous, Priodontes, Leopardus weidii). All of the terrestrial-animalivorous carnivores at Paracou (felids, Speothos, and Galictis) were uncommon and most were only recorded by us through interviews. Monodelphis, Cabassous, Dasypus kappleri, Priodontes, and Myrmecophaga were likewise rarely collected, observed, or recorded from interviews. Only four local members of this guild were encountered often enough to provide noteworthy behavioral data or density estimates.

Based on their highly carnassialized dentitions, small size, and the absence of macroscopic seeds in feces, we assume that species of Marmosops are predominantly, if not exclusively, animalivorous. Although species of this genus have opposable halluces and prehensile tails, our nocturnal observations and trapping results suggest that M. parvidens and M. pinheiroi are predominantly terrestrial, seldom ascending slender lianas or narrow stems more than a meter or two above the forest floor; none were taken in arboreal platform traps. Malcolm (1991) likewise trapped M. parvidens only on the ground, despite his intensive program of arboreal platform trapping. Woodman et al. (1995) captured M. noctivagus equally often on the ground and above ground, but none of their above-ground traps were set higher than 2 m. Patton et al. (2000: appendix A) listed several species of Marmosops as having been taken in “canopy” traps at some collecting localities along the Rio Juruá in western Amazonia, but their text summaries of habitat data (op. cit.: 53–61) indicate that none were actually trapped at heights exceeding 2 m; instead, the majority (82%) of all captures of this genus on the Rio Juruá were on the ground. Because many unambiguously terrestrial mammals (including Oryzomys, Myoprocta, and Proechimys in our study) will occasionally ascend lianas to reach baited traps, infrequent captures of Marmosops in low vegetation is not inconsistent with our provisional guild assignment for this genus.

Metachirus has been variously characterized in standard references as an arboreal omnivore (Walker, 1964) or a semiarboreal frugivore (Hunsaker, 1977; Streilein, 1982), but these ecobehavioral descriptors appear to be unsupported by field data. Instead, our capture results agree with observations from many other trapping programs (e.g., Miles et al., 1981; Malcolm, 1991; Patton et al., 2000) in suggesting that this is one of the most exclusively terrestrial of all New World marsupials. Similarly, our examination of stomach contents from 10 Paracou specimens is consistent with evidence from other dietary studies (Mathews, 1977; Santori et al., 1996; Freitas et al., 1997) that Metachirus is almost entirely insectivorous. Our nocturnal sighting rate for Metachirus (0.42 per 10 hours) is within the range of values reported for this taxon by Emmons (1984: table 4) at other Amazonian sites.

Semiaquatic/Riparian Folivores: This guild is not represented at Paracou, where suitable habitat for capybaras (Hydrochoerus) is absent. Semiaquatic-folivorous mammals never represent more than a tiny fraction of any Neotropical rainforest fauna.

Semiaquatic/Riparian Omnivores: Only a single species assignable to this guild is present at Paracou and at Manu; no ecological equivalent is known from Central American rainforests. The Paracou species (Nectomys melanius) is apparently very uncommon and was not trapped by us.

Semiaquatic/Riparian Animalivores: This guild comprises a smaller fraction of the fauna at Paracou than at either La Selva or Manu, where large bodies of water support resident otter populations. At Paracou, this guild is only represented by two species, both of which are nocturnal. Water opossums (Chironectes minimus) were sighted frequently in our study area (0.7 individuals per 10 hours of nocturnal census) and are obviously common in local streams, but we recorded no other significant natural history information about this species. Our pitfall captures of Neusticomys oyapocki are the first obtained by this method, which might usefully be applied to detect the presence of congeneric species at other Amazonian localities.

Species Richness and Inventory Completeness

We recorded a total of 142 species of mammals at Paracou, of which 128 were captured or observed as result of our own efforts in the field from 1991 to 1994. The remaining 14 species were either recorded from interviews with local residents or were identified from museum voucher material collected in our study area by previous researchers. Because our total species accumulation results (fig. 91) do not indicate a convincing asymptote, it is reasonable to assume that we could have recorded more species at Paracou with additional fieldwork, perhaps including some species not previously observed or collected by others.

To explore the range of plausible estimates of true species richness for the entire Paracou mammal fauna, we summed the results of four nonparametric extrapolation methods applied separately28 to our sampling data for bats and nonvolant mammals (table 60). Given the inherent uncertainty of any extrapolation procedure, it is encouraging that the range of total richness values predicted by these methods is really quite narrow (155–168 spp.). The results from Chao's methods should probably be interpreted as lower bounds for plausible inference from our data because sampling effort was sufficiently intensive that most species were each observed three or more times (i.e., information in the larger frequency classes is “nonnegligible” sensu Chao [1984, 1987]). By contrast, the jackknife methods yield statistically consistent estimators (Burnham and Overton, 1979) that should converge on true species richness as sampling effort increases. In fact, the numbers of species for bats, nonvolant mammals, and total mammals predicted by the second-order jackknife (in the fifth column of table 60) are very close to ecogeographic expectations based on the known northern French Guianan source fauna.

Estimating the overall completeness of our mammal inventory clearly depends on which of these extrapolations of true species richness is used. Whereas Chao's methods suggest completeness values (observed/extrapolated species richness × 100) in the range of 90–92%, the jackknife methods suggest lower values of about 85–87%. These differences are not large, however, and even the lowest estimate (based on the second-order bootstrap) suggests that the mammal fauna at Paracou is at least as well sampled as those at Central American field stations with decades-long histories of biological research. For example, faunal sampling at La Selva is perhaps only 85% complete after more than 30 years of field research at that site, whereas faunal sampling near Barro Colorado Island is perhaps only 78% complete after more than 70 years (Voss and Emmons, 1996: table 10).

Bats constitute about 55% of the known Paracou mammal fauna, and our extrapolations suggest that similar proportions (54–56%) would have been obtained with additional sampling effort. In order of decreasing richness, rodents are clearly a distant second to bats, representing only 15% of the observed total species, followed by marsupials (8%), carnivores (7%), xenarthrans (6%), primates (4%), and ungulates (4%). The small list of expected nonvolant species (appendix 1) suggests that this rank ordering by relative diversity is unlikely to change significantly with additional fieldwork in our study area.

The taxonomic dominance of bats in Neotropical rainforest mammal faunas is evidenced by all large species lists obtained with modern collecting methods (table 61), but bats may be proportionately more diverse or less diverse at some localities than they are at Paracou. Wilson (1990) suggested that bats probably represent about 60% of the fauna at La Selva, a somewhat higher figure than is supported by observed species counts from that site, but approximately what could be expected there based on ecogeographic range data (Voss and Emmons, 1996: appendix 2). Similar proportions will perhaps be found to characterize other Central American faunas, where most nonvolant clades (notably marsupials, primates, and rodents) are conspicuously less diverse than in Amazonia. By contrast, bats may constitute a smaller proportion of western Amazonian mammal faunas, where many nonvolant clades are maximally diverse. Bat communities have not been intensively sampled over multiple years at any western Amazonian site, however, with the result that existing species counts (e.g., from Balta, Manu, and Cuzco Amazónico; table 61) probably underestimate relative bat diversity. Based on expected species lists (Voss and Emmons, 1996: appendices 9–11) it seems likely that bats will eventually be found to comprise about 50% of the total mammal fauna at most western Amazonian localities.

Although a few more mammalian species are currently known from Paracou than from any other Neotropical rainforest site for which published faunal lists are currently available (table 61, last column), this result is hard to interpret due to inventory differences in many of the confounding factors previously discussed in our comparative analyses of bat and nonvolant species richness. Among such factors, sampling methods and effort seem likely to be the most important: because we used many inventory methods over multiple field seasons at Paracou, our species list could be artifactually larger than those from more diverse sites where faunal sampling was substantially less complete.

Species accumulation graphs from our first (1991) field season at Paracou and from methodologically comparable fieldwork on the lower Rio Xingu (in southeastern Amazonia) and at Balta (in western Amazonia) tend to support this interpretation (fig. 92). For any specified level of sampling effort after the first several person-weeks at each site, the number of species recorded at Balta is consistently higher than that recorded on the Rio Xingu, which in turn is consistently higher than that recorded at Paracou. Therefore, the large species list we eventually obtained at Paracou is plausibly explained by our prolonged total sampling effort (634 person-days), and by our subsequent use (after 1991) of more methods to capture elusive species.

It is noteworthy that the rank-ordering of inventory sites by species richness implied by figure 92 is the same as that expected from geographic range data, which predict maximal mammalian diversity in the western part of Amazonia, intermediate diversity in the southeast, and minimal diversity in the Guiana subregion (Voss and Emmons, 1996). Although three data points are not sufficient to establish a significant correlation between expected and observed diversity, this result is consistent with other lines of evidence suggesting that the Paracou mammal community is a typically Guianan assemblage whose essential characteristics are largely determined by its biogeographic context. That context now requires more precise definition before its implications for community phenomena other than species richness can be assessed.

Biogeography

Our previous biogeographic analyses of Paracou bats and nonvolant mammals suggested that these groups show essentially similar spatial patterns of faunal relationships with other rainforest inventories. In both analyses, Paracou first clustered with other inventory sites from the Guiana subregion of Amazonia, next with sites from other Amazonian subregions, and lastly with Central American sites (Simmons and Voss, 1998: fig. 76; this report: fig. 89). Obviously, faunal similarity is correlated with geographic proximity in both datasets, but bats and nonvolant mammals differ in other quantitative biogeographic phenomena that merit further analysis.

To obtain strictly comparable biogeographic data for bats and nonvolant mammals, we computed matrices of pairwise faunal similarity (scaled as percentages in table 62) and airline distances (table 63) among all ten inventory sites at which presence/absence data for both groups were obtained within the same study area (La Selva, Barro Colorado, Imataca, Paracou, Arataye, Cunucunuma, Xingu, Balta, Manu, Cuzco Amazónico).29 Among this common set of geographic samples, bat faunal similarity ranges from a minimum value of 31% (for the comparison Barro Colorado-Xingu) to a maximum of 68% (Balta-Manu), with a mean of 46% and a standard deviation of 9%. By contrast, nonvolant faunal similarity ranges from 10% (for La Selva-Xingu) to 78% (Paracou-Arataye), with a mean of 34% and a standard deviation of 16%. At this spatial scale (airline distances among these sites range from 136 to 3788 km with a mean of 2038 km and a standard deviation of 954 km), bat faunas are therefore more similar to one another (on average) and are less variable in composition than nonvolant mammal faunas.

Matrix correlations (assessed for statistical significance by random permutations; Smouse et al., 1986) indicate that the inverse relationship between faunal similarity and airline distance is equally strong for bats (r = ;ms0.81, p = 0.001) and nonvolant mammals (r = ;ms0.82, p = 0.001), but faunal similarity declines with distance at different average rates for the two groups (figs. 93, 94). Linear regression suggests that faunal divergence averages about 14% per 1000 km for nonvolant mammals, almost twice the estimated rate of about 8% per 1000 km for bats. In addition, the residual variance in faunal similarity values—the variation “unexplained” by airline distance—is substantially greater for nonvolant mammals than it is for bats, with an F-ratio (computed from the mean-squared-error estimates of the respective regressions) of about 2.8.

The magnitude of the residual variation in nonvolant faunal similarity is aptly illustrated by comparisons between Paracou and two almost equidistant inventory sites. Although Imataca is slightly further from Paracou (1023 km) than is Xingu (994 km), percent nonvolant faunal similarity between Paracou and Imataca is 61%, whereas nonvolant faunal similarity between Paracou and Xingu is only 36%. Of course, faunal similarity values estimated from incomplete inventory data are subject to sampling error, which plausibly accounts for some differences between the Paracou and Xingu species lists (many ubiquitous carnivore taxa, for example, were not recorded by USNM fieldworkers during their short visit to the Xingu site; Voss and Emmons, 1996: appendix 8). An additional factor of obvious importance, however, is the role of the lower Amazon as a barrier to nonvolant faunal exchange: at least 23 of the mismatches (species present in one fauna but not the other) between the Paracou and Xingu lists are taxa with eastern Amazonian ranges that are wholly or largely restricted to either the north side of the Amazon (e.g., Bradypus tridactylus, Saguinus midas, Alouatta seniculus, Ateles paniscus, Pithecia pithecia, Oecomys auyantepui, Oecomys rutilus, Coendou melanurus, Proechimys guyannensis) or to the south side (e.g., Bradypus variegatus, Saguinus niger, Alouatta belzebul, Aotus infulatus, Callicebus moloch, Oecomys paricola, Oryzomys emmonsae, Oxymycterus amazonicus, Proechimys goeldii, Proechimys oris). By contrast, bat faunal similarity does not differ dramatically between Paracou-Imataca (56%) on the one hand and Paracou-Xingu (49%) on the other, and only two bat species appear to have range limits defined by the lower Amazon (Tonatia schulzi and Lasiurus atratus, both known only from the north side). The same set of inventory comparisons therefore suggests that major rivers and other large-scale habitat discontinuities are less effective as dispersal barriers for bats than for nonvolant mammals.

Different average dispersal abilities of bats on the one hand and nonvolant mammals on the other could explain the disparate results of regressing faunal similarity on airline distance for these groups. Because widely separated localities will usually have more intervening barriers between them than adjacent localities, faunas composed of taxa with lower average dispersal abilities should diverge more rapidly with distance (on average) than faunas composed of more vagile taxa. However, because barriers sometimes do occur between nearby sites (as those on opposite river banks), the residual variance should also be larger for faunas composed of taxa with lower average dispersal abilities.

Patterns of species distribution among Neotropical rainforest regions separated by high mountains, major rivers, and xeromorphic vegetation (fig. 95) are consistent with the hypothesis that large-scale habitat discontinuities are less effective as faunal barriers for bats than they are for nonvolant mammals. Thus, most Paracou bats (42 species, representing 54% of the local chiropteran fauna; table 64) occur in all four Neotropical rainforest regions, another 19 bat species (24%) occur in three out of four rainforest regions, and an additional 11 species (14%) occur in two regions. Altogether, 92% of Paracou bats are apparently undifferentiated across one or more major landscape features delimiting the Neotropical rainforest biota. Only six species (about 8% of the local chiropteran fauna) are Amazonian endemics.

Nonvolant Paracou mammals show contrasting patterns of species membership in these distributional categories, especially the first and last. Only 18 species (comprising just 28% of the nonvolant fauna) occur in all four Neotropical rainforest regions, whereas 24 species (fully 38% of the nonvolant fauna) are Amazonian endemics. To facilitate statistical testing of these frequency differences, we dichotomously classified all Paracou mammals as either “widespread” (present in two or more rainforest regions) or “endemic” (restricted to Amazonia). The hypothesis that Paracou bats and nonvolant mammals do not differ in relative endemism (and implied dispersal ability) can be confidently rejected based on the resulting 2 × 2 contingency analysis (table 65).

To explore the geographic structure of endemism in rainforest bats and nonvolant mammals, we applied biogeographic parsimony methods (as originally described by Rosen, 1988, 1992) to identify repeated patterns of species-sharing among inventory sites. Parsimony analyses of endemism (PAE) were implemented with heuristic search algorithms in PAUP* (version 4.0b3a; Swofford, 2000), which were applied separately to the presence/absence data for bats and nonvolant mammals; an all-zero operational geographic unit was added to both datasets to root the resulting networks. Relative support for different locality groupings was estimated by bootstrap resampling with 1000 replicates using 100 random-addition sequences per replicate.

The geographic patterns recovered with ≥50% bootstrap support by PAE are a subset of those previously obtained from UPGMA clustering by Jaccard's coefficient for both datasets (figs. 96, 97). It is also noteworthy that no PAE grouping that received more than 50% bootstrap support in one dataset conflicts with any PAE grouping with equivalent support in the other. Instead, a Central American/Amazonian faunal dichotomy is indicated for both bats and nonvolant mammals, as is a well-supported grouping of three adjacent inventory sites in southwestern Amazonia. However, PAE based on the nonvolant data supports additional intra-Amazonian faunal relationships that are not consistently recovered by PAE from the bat data.

A spatially compact grouping of four nonvolant inventories from the coastal watershed of the Guianan subregion of Amazonia (Imataca, Kartabo, Arataye, Paracou) is recovered by PAE with moderately strong (71%) bootstrap support, but this endemicity set is immediately joined with similar (74%) support by the nonvolant inventory from Manaus, a site deep in the continental interior near the confluence of the Rio Negro and the Amazon. Although the next nonvolant inventory site to join is Cunucunuma (another deep-continental site), this grouping has only marginal (57%) bootstrap support. Finally, PAE recovers an even more inclusive grouping composed of these six sites plus the nonvolant inventory from the Rio Xingu, which joins the others with moderate (68%) bootstrap support despite its isolated position south of the Amazon.

Of course, some of this recovered structure is due to the absence of inventory data from many geographically intermediate localities. Inevitably, the addition of new faunal lists with novel combinations of species will reduce the distinctiveness of some currently well-supported groupings in future analyses. A more general criticism of these results could justifiably invoke the inappropriateness of hierarchical models for analyzing data that perhaps lack a natural hierarchical structure.30 Certain spatial patterns recovered in our analyses, however, are strongly supported by external evidence and seem likely to sustain meaningful biogeographic interpretation. One such pattern is the cluster of inventories that includes Paracou and most, but not all, of the other analyzed sites from the Guiana subregion of Amazonia.

Although the area enclosed by the interconnected waters of the Orinoco, the Rio Negro, and the lower Amazon has traditionally been recognized as a convenient unit for biogeographic analysis (e.g., by Tate, 1939; Hoogmoed, 1979; Mori, 1991; Voss and Emmons, 1996), the Guiana subregion of Amazonia is actually inhabited by two distinctly different rainforest mammal faunas (fig. 98). One of these consists of an apparently allochthonous (non-Guianan) assemblage that is represented by collections from Cunucunuma, Esmeralda, Boca Mavaca, Neblina Base Camp, and several other localities east of the Casiquiare in southern Venezuela. The other fauna, distinctively Guianan in taxonomic composition, is represented by collections from Imataca, Kartabo, Paracou, Arataye, Manaus, and numerous other sites in the Guianas (e.g., Dadanawa) and Guianan Brazil (e.g., Faro, Serra do Navio).

The fauna that occurs east of the Casiquiare in southern Venezuela includes at least 19 species, all of which have more-or-less extensive western Amazonian distributions (table 66). Most of these do not range much farther into the Guiana subregion, but a few have been collected or observed as far eastward as the right (west) bank of the Rio Branco, which apparently constitutes a significant geographic limit to this fauna in Brazil: Ateles belzebuth (see Nunes et al., 1988), Callicebus torquatus (Hershkovitz, 1990), Sciurus igniventris (Emmons, 1997: map 147), and Oecomys concolor (G. G. Musser, personal commun.). In Venezuela, some allochthonous taxa appear to reach their eastern range limits at or near the Río Caura (e.g., Philander andersoni, Ateles belzebuth) or the Río Caroní (e.g., Aotus trivirgatus, Callicebus torquatus; Linares, 1998). Apparently, only two members of this fauna reach the left bank of the Essequibo or its upper tributaries in Guyana: Bassaricyon gabbii (based on the single historical record discussed by Tate, 1939) and Caluromys lanatus (see Emmons, 1993a).

A few other non-Guianan mammals are known from scattered localities on the north bank of the lower Amazon. These could be members of a distinctive whitewater-floodplain biota (e.g., as mapped by Salo and Räsänen, 1989: fig. 1), the taxonomic composition of which might include both western Amazonian elements (e.g., Glironia venusta; da Silva and Langguth, 1989) and central Amazonian endemics that occur on both sides of the river (e.g., Makalata grandis; Emmons, 1997, personal commun.). The apparently narrow distributions of Dactylomys dactylinus and Bradypus variegatus along the north bank of the lower Amazon may also be restricted to floodplain habitats.

By contrast, the remainder of the Guiana subregion (east of the Río Caroní-Rio Branco and north of the Amazonian floodplain) is inhabited by a strikingly homogeneous and unmistakably autochthonous fauna, at least 17 members of which have sufficiently congruent range limits to usefully define a Guianan center of mammalian endemism (table 67). Although a few taxa that we consider Guianan endemics are known to occur west of the Río Caroní (e.g., Monodelphis brevicaudata, Chiropotes satanas chiropotes, Proechimys guyannensis) or along the south bank of the lower Amazon (Marmosops parvidens, Marmosops pinheiroi, Bradypus tridactylus, Myoprocta acouchy), range overlap is most extensive in the core area whose periphery is suggested by the closed circles in figure 98. Other typically Guianan but nonendemic species that have similar distributional limits within the subregion include Philander opossum (replaced by P. andersoni in southern Venezuela), Cebus olivaceus (replaced by C. albifrons in southern Venezuela), Sciurus aestuans (replaced by S. gilvigularis in southern Venezuela), Dasyprocta leporina (replaced by D. fuliginosa in southern Venezuela), Myoprocta acouchy (replaced by M. pratti in southern Venezuela), and a small-toothed form of Mesomys with distinctive mtDNA sequences (replaced by a large-toothed congener with divergent mitochondrial genes in southern Venezuela).

This geographic segregation of distinct rainforest mammal faunas in the Guiana subregion strikingly resembles the avifaunal patterns previously mapped by Cracraft (1985: fig. 3). In particular, his concept of an “Imeri” avifauna that extends from northwestern Amazonia into southern Venezuela is consistent with our interpretation of the southern Venezuelan mammal fauna as essentially non-Guianan.31 Likewise, his “Guyanan” center of avian endemism appears to be geographically identical with our similarly named mammalian center. Such congruence invites yet wider organismal comparisons to test the generality of these results.

Only a few relevant compilations of distributional data are available for other higher taxa, but several show similar geographic patterns. In particular, mapped range limits for rainforest lizards (Dixon, 1979: fig. 9:8), snakes (Dixon, 1979: fig. 9:9), and Lecythidaceae (Mori, 1991: fig. 1) suggest that most Guianan endemics in these groups likewise do not occur west of the Río Caroní and the Rio Branco. Possibly, the broad lower reaches of both rivers—together with the mostly savanna-covered highlands from which they arise—have been historically effective barriers to faunal and floral exchange between the rainforested Imeri and Guyana lowlands. Alternatively (or additionally), these rivers might approximate the geographic limits of some distinctive combination of soils and climate to which the endemic Guianan rainforest biota is uniquely adapted. Whatever historical and/or contemporary-ecological factors might explain this phenomenon, however, it seems clear that the Guianan center of mammalian endemism is part of a general pattern of biotic differentiation shared with other sympatric groups of rainforest organisms.

The Paracou mammal fauna is a typically Guianan assemblage that includes at least 14 of the 17 endemics listed in table 67 together with many nonendemic but typically Guianan species (e.g., Philander opossum, Ametrida centurio, Sciurillus pusillus, Dasyprocta leporina). Although we recorded a few taxa at Paracou that were previously unknown as elements of the Guianan fauna (Hyladelphys kalinowskii, Saccopteryx gymnura, Micronycteris homezi, M. schmidtorum), these are more plausibly interpreted as widespread but elusive species than as biogeographic anomalies. By contrast, most of the other mammals we recorded at Paracou are known from numerous additional Guianan localities from Imataca or Kartabo to Manaus and the Serra do Navio.

The biogeographic character of the Paracou fauna is equally apparent, however, in our failure to record many taxa that are widespread in western and/or southeastern Amazonia. In fact, the shared absence of such species as Glossophaga comissarisi, Carollia castanea, Rhinophylla fisherae, Enchisthenes hartii, Platyrrhinus infuscus, Sphaeronycteris toxophyllum, Sturnira magna, and Uroderma magnirostrum more readily characterizes Guianan bat inventories (and accounts for their cohesion in faunal cluster analyses) than does the shared presence of such rare endemics as Tonatia schulzi and Lasiurus atratus. Among nonvolant Amazonian inventories, the shared absence of Callithrix, Cebuella, Cacajao, Callicebus, Lagothrix, Bassaricyon, Microsciurus, giant squirrels (Urosciurus), Scolomys, and Dactylomys is likewise uniquely Guianan.

Of course, mammalian faunal composition is not constant within the Guianan center. For example, some species that have been recorded near Kartabo (e.g., Neacomys guianae; table 19, footnote b) or Manaus (Saguinus bicolor; Hershkovitz, 1977: fig. X15) are not known to occur in French Guiana. Similarly, a few species that we collected at Paracou are currently known only from additional sites in French Guiana, Amapá, or eastern Surinam (Molossus barnesi, Neacomys dubosti, Neusticomys oyapocki). Such narrowly distributed taxa suggest the existence of subcenters of mammalian endemism, perhaps corresponding in location to several of the “refugia” that Prance (1982: fig. 11.9) postulated to explain distributional phenomena in the Guianan rainforest flora. In particular, Prance's “East Guianan” endemics—rainforest plants with small ranges centered on French Guiana and Amapá (e.g., Corythophora amapaensis; see Mori and Prance, 1987: fig. VI-1)—may represent elements of the same narrowly endemic biota that we sampled at Paracou and share the same history of geographic isolation and/or local adaptation. Much future collecting, however, will be required to convincingly document geographic range limits for East Guianan endemic mammals, which do not appear to account for more than a small fraction (about 2%) of the Paracou fauna.

Community Structure

Using the information previously summarized in our separate guild analyses of bats and nonvolant species, we classified all Paracou mammals by diel activity (nocturnal, diurnal, both), substrate use (aerial, arboreal, terrestrial, semiaquatic), and trophic role (primary consumers, secondary consumers, omnivores) to examine basic aspects of community-wide patterns of resource use (table 68). Obviously, a more complete ecological representation of the entire fauna could be obtained by cross-classifying species using the same criteria (e.g., to count diurnal-arboreal primary consumers, nocturnal-terrestrial secondary consumers, etc.) and by taking other physiologically significant traits (such as body weight) into account. However, faunal analyses based on such comprehensive treatments are dauntingly complex (Skalli and Dubost, 1986; Dubost, 1987) and beyond the scope of this report. For many ecological researchers, the appropriate units of community analysis represent biomass (e.g., kg/km²; Eisenberg and Thorington, 1973; Janson and Emmons, 1990; Peres, 1999) rather than species, but taxonomic diversity (not energy flow) is the subject of interest in the following discussion.

The Paracou fauna does not appear to be unusual with respect to patterns of either diel activity or substrate use if allowance is made for the undersampling of bats in most Neotropical rainforest inventories on which previous conjectures about whole-community structure have been based (e.g., by Fleming, 1973; Bourlière, 1989). The Paracou fauna itself may be undersampled for arboreal species, however, which comprise a somewhat smaller proportion of the fauna at this site than at other Neotropical rainforest localities previously analyzed for substrate use (e.g., Kay and Madden, 1997: table 30.4). Based on the list of additional nonvolant species that might yet be recorded in our study area (appendix 1), future inventory work could plausibly bring the numbers of arboreal and terrestrial species in the local community into closer conformance with ratios observed in other Neotropical faunas.

Apparently, the only distinctive community-wide aspect of resource use at Paracou concerns trophic structure. Whereas most species of mammals in both New World and Old World tropical rainforests are said to be primary consumers (feeding primarily on fruits, seeds, nectar, leaves, or other plant tissues; Bourlière, 1973, 1989), species of secondary consumers (faunivores) outnumber primary consumers by a substantial margin at Paracou. Contingency tests suggest that this difference is highly significant for some pairwise comparisons with other well-known Amazonian faunas. At Manu, for example, the numbers of primary and secondary consumers are almost inversely proportional to those observed at Paracou (table 69), and it therefore seems appropriate to consider alternative hypotheses that might account for the preponderance of secondary consumers in our study.

A partial explanation clearly involves sampling artifacts. Primary consumers constitute most of the mammalian biomass in all Neotropical rainforest communities studied to date (Eisenberg and Thorington, 1973; Janson and Emmons, 1990; Peres, 1999), where frugivores and granivores in particular tend to be larger-bodied and/or to maintain higher population densities than most faunivores. In addition, some standard inventory methods are known to produce trophically biased samples of rainforest mammal communities; mistnets, for example, are primarily effective for capturing frugivorous bats (Fleming et al., 1972; LaVal and Fitch, 1977; Kalko et al., 1996; Simmons and Voss, 1998), and conventional traps are most effective for capturing frugivorous-granivorous rodents (Voss and Emmons, 1996; this report). For these reasons, short-term and/or methodologically limited inventories probably tend to overestimate the ratio of primary to secondary consumers in local faunas. Given the inverse correlation between known relative primary consumer diversity and cumulative sampling effort at Paracou (fig. 99), trophic comparisons with less intensively worked sites are obviously problematic.

Indeed, future fieldwork at Manu will probably add more secondary than primary consumers to the species list from that site because 33 out of the 48 additional species that could still be expected there (Voss and Emmons, 1996: appendix 10) are faunivores. However, at least part of the difference in trophic structure between the mammal communities at Manu and Paracou is not artifactual. Many taxa of primary consumers that occur at Manu are widespread in western Amazonia but have no known ecological equivalents at Paracou or in other core-Guianan faunas (e.g., Caluromysiops, Callithrix, Aotus, Callicebus, Lagothrix, Bassaricyon, Microsciurus, giant squirrels [Urosciurus], Dinomys, Dactylomys, Sylvilagus), and several taxa of primary consumers common to both inventories are consistently more speciose at western Amazonian than at Guianan localities (e.g., Carollia, Platyrrhinus, Uroderma, Saguinus, Proechimys). By contrast, no taxon of primary consumers in the Guianan fauna seems to lack an ecological equivalent in western Amazonia, and no taxon of primary consumers common to both faunas is known to be consistently more speciose at Guianan than at western Amazonian localities. Because secondary consumers appear to be about equally diverse in both Amazonian subregions, at least part of the observed trophic divergence between the Paracou and Manu communities seems to reflect a biogeographic gradient that principally affects the species richness of primary consumers.32

Revisionary Taxonomy

The importance of continued revisionary taxonomic research for mammalian biogeographic studies in the rainforested Neotropical lowlands can hardly be overemphasized. Estimates of both local richness and faunal complementarity (endemism) rely crucially on the ability of investigators to distinguish taxa, an ability that is compromised by our currently inadequate knowledge of species limits in many Neotropical clades. Just how much remains to be learned is suggested by the taxonomic results of this study, wherein we applied the most basic of all species criteria (morphological diagnosability) to assign names to voucher material collected at Paracou.

Based on character differences described or referenced in Simmons and Voss (1998) and this report, the following 23 species (newly described or resurrected from synonymy) should be recognized as valid: Marmosa quichua, Marmosops bishopi, Marmosops juninensis, Marmosops pinheiroi, Monodelphis glirina, Monodelphis palliolata, Centronycteris centralis, Peropteryx trinitatis, Micronycteris brosseti, Micronycteris homezi, Micronycteris microtis, Mimon cozumelae, Eptesicus andinus, Eptesicus chiriquinus, Molossops paranus, Molossus barnesi, Saguinus niger, Neacomys dubosti, Neacomys paracou, Nectomys melanius, Oecomys auyantepui, Coendou melanurus, and Mesomys ferrugineus. Of course, we were not the first to recognize many of these species as valid. Most of the resurrected taxa in this list were originally named as full species and were long recognized as such until synonymized (often with little or no explicit justification) by uncritical advocates of the “polytypic” species concept. More recently, Handley (1976), Brosset and Charles-Dominique (1990), and other cited authors anticipated some of our taxonomic conclusions. However, none of these species were recognized as valid by Wilson and Reeder (1993), whose checklist serves as an adequate summary of the prevailing taxonomic consensus at the time we began our inventory work. By contrast, our results suggest that only two species that Wilson and Reeder recognized as valid should be synonymized: Choeroniscus intermedius (with C. minor), and Nectomys parvipes (with N. melanius).

The net change in taxonomic diversity resulting from research with Paracou voucher material, 21 species, represents an increase of almost 2% over the total Neotropical mammal fauna recognized prior to this study (1145 species according to Patterson, 1994). Because Guianan mammals are relatively well known as a consequence of long accessibility to European naturalists, similarly detailed analyses of voucher material collected in more remote areas (e.g., in western Amazonia; Patton et al., 2000) seem certain to result in substantially larger diversity increments.

Predictably, the improved taxonomic resolution resulting from such research will contribute to a much-needed assessment of biogeographic congruence between mammalian and other biotic distributional data. The close correspondence between spatial patterns of mammalian and avian endemism in the Guiana subregion of Amazonia, for example, is documented herein by numerous distributional data previously obscured by synonymy. Based on our preliminary assessment of many extralimital alpha-taxonomic problems, we anticipate that future research will provide strong mammalian support for many of the endemic patterns first remarked and clearly delimited for Amazonian birds by Cracraft (1985). Many of the molecular results summarized by Patton et al. (1997, 2000) tend to support the same conclusion, in addition to suggesting other patterns of endemism that may exist within recognized avian centers. Plausibly, future revisionary research incorporating both morphological and molecular data will afford the strongest basis for assessing the generality of biogeographic congruence across major clades of Amazonian organisms.

Mapping Guianan Endemicity

Much additional collecting is needed before the geographic range limits of mammals identified herein as Guianan endemics can be mapped with greater confidence. Notably useful would be thoroughly vouchered faunal surveys from northern Brazil, especially (1) in Amazonas and Roraima states between the Rio Negro and the Rio Branco (where most Guianan endemics are apparently absent), (2) in Pará state between the Trombetas and the Jari (an area from which no large faunal lists are apparently available), and (3) from additional sites in Amapá. Special attention should be devoted to the potential existence of a distinctive mammalian fauna in the Amazonian white-water floodplain vis-á-vis the Guianan terra firme in each of these interfluvial zones.

Faunal surveys along the south bank of the Amazon are equally important to test the hypothesis that putative Guianan endemics are not widely distributed there. Indeed, the efficacy of the lower Amazon as a barrier to faunal dispersion should be tested by replicated transects through várzea and terra firme habitats on opposite banks (after Patton et al., 2000). Such transects should devote comparable effort to sampling bat and nonvolant mammalian faunas in order to test the conjecture that riverine barriers are not equally effective for these groups.

Species distributions in the transitional area between the western Amazonian (“Imeri”) fauna of southern Venezuela and the Guianan fauna of eastern Venezuela and western Guyana likewise merit targeted fieldwork and careful mapping. Of special interest are the geographic boundaries between such replacing-species pairs as Philander andersoni/P. opossum, Bradypus variegatus/B. tridactylus, Ateles belzebuth/A. paniscus, Cebus albifrons/C. olivaceus, Dasyprocta fuliginosa/D. leporina, and Myoprocta pratti/M. acouchy. Because most faunal transitions in the Amazonian biota are currently thought to coincide abruptly with very large riverine barriers, the apparently gradual (or stepwise) loss of western Amazonian taxa and their replacement with Guianan forms across hundreds of kilometers of more-or-less continuous Venezuelan lowland forest (Huber and Alarcon, 1982) is of exceptional biogeographic interest.

Historical Connections with Other Areas of Endemism

Phylogenetic analyses of mammalian clades that include endemic Guianan taxa will contribute to an historical assessment of the isolation and assembly of this well-marked center of endemism. Maximally informative analyses with this objective should include morphological and/or molecular terminals from as many areas of Amazonian endemism as possible in order to clearly distinguish among alternative biogeographic scenarios. Minimally, taxa inhabiting each of the Napo/Imeri, Inambari, and Rondônia/Pará/Belém centers (after Cracraft, 1985) would appear necessary for any analysis to be informative about the historical biogeography of Guianan endemics.

Unfortunately, the few available phylogenetic datasets that approximate these conditions give contradictory biogeographic indications. The most impressive, Canavez et al.‘s (1999) analysis of β₂-microglobulin sequences from Saguinus species, suggests that the sister taxon to the endemic Guianan lineage S. bicolor + S. midas is S. niger from southeastern Amazonia (Pará/Belém), a result that implies relatively recent vicariance or dispersal across the lower Amazon. A similar scenario is implied by several of the cytochrome-b sequence analyses reported by Patton et al. (2000: e.g., figs. 171A, 171C), but sister-group relationships between Guianan and western Amazonian haplotypes in other analyses by the same researchers (op. cit.: e.g., figs. 172B, 172D) imply relatively more recent faunal connections across the Rio Negro. Although it is crucial to draw the distinction between haplotypes and taxa in these mitochondrial-gene studies, the prima facie implication of such incongruence is that not all Guianan taxa share the same biogeographic history. Whether any one historical pattern better characterizes the biogeography of Guianan endemics than another will require much more phylogenetic research to convincingly establish.

Causal Explanations for Community Composition

The hypothesis that the mammalian community at Paracou is broadly representative of the Guianan fauna with respect to species richness and trophic structure should be tested with methodologically similar inventory fieldwork at other Guianan sites. Given the essential taxonomic homogeneity of this center of endemism, however, together with the marked tendency for inventory results to converge on geographic expectations with sufficient sampling effort (Voss and Emmons, 1996: fig. 24), it seems probable that such will prove to be the case. If so, then appropriate explanations for these community characteristics should be based on historical and/or ecological circumstances common to the Guianan fauna as a whole rather than on the environmental peculiarities of any particular study site.

Certain paleoenvironmental reconstructions, for example, have suggested that Guianan rainforests were simultaneously isolated and reduced in area during the Pleistocene: by an enormous freshwater lake (e.g., Frailey et al., 1988: fig. 8), or by savannas and other arid vegetation (e.g., Clapperton, 1993: fig. 8). Both scenarios lack convincing independent support (Colinvaux, 1996), but either might explain the lower mammalian species richness of modern Guianan rainforests (as residual insular effects). Neither areal reduction nor isolation, however, plausibly accounts for the conspicuous trophic bias observed in our comparisons of Guianan with western Amazonian mammal communities.

The ancient soils produced by in-situ chemical weathering of the Guiana Shield are routinely characterized as exceptionally nutrient-poor (Jordan, 1985; Sanchez, 1989), especially by comparison with the younger alluvium (mechanically weathered from the Andes) that forms the mineral substrate throughout much of western Amazonia. Poor soils have been hypothesized to negatively affect the species richness of rainforest mammals by constraining plant productivity (Emmons, 1984; Gentry and Emmons, 1987), and it seems intuitively obvious that plant productivity should more directly impact primary consumer than faunivore diversity. Geology therefore offers a superficially plausible explanation for two of the most distinctive contrasts between Guianan and western Amazonian mammal faunas.

The Guianan center of endemism is not, however, coextensive with the Guiana Shield, which extends hundreds of kilometers west of the Rio Negro to the base of the Andes (Gibbs and Barron, 1993), so geology cannot be the whole story. Nor do the climatic maps that we have examined suggest any sharp change in the amount or seasonality of annual rainfall that coincides with the hypothesized limits of the Guianan center of endemism. Although multifactorial models of plant productivity might reveal some ecologically significant combination of edaphic and climatic variables that broadly characterizes Guianan versus other Amazonian landscapes, it does not seem plausible that ecological gradients alone could result in sharply defined areas of endemism in the absence of historical barriers to dispersal. Given that gradients exist, however, and that taxa differ in their ecological requirements, it would be interesting to explore the possibility (with computer simulations) that semipermeable dispersal barriers might attract species boundaries over many generations in much the same way that they appear to attract step clines in gene frequencies (Endler, 1977). Alternative scenarios for the historical assembly of Amazonian biotas have tended to view major rivers as effective dispersal barriers (e.g., Wallace, 1852; Cracraft, 1985; Ayres and Clutton-Brock, 1992), or as mere hydrological boundaries between ecologically distinctive terrains (Tuomisto and Ruokolainen, 1997). The nature of faunal transitions between mammalian centers of endemism, however, suggests that a more interactive conception of biotic assembly might better fit the facts at hand.