MAMMALIAN DIVERSITY AND MATSES ETHNOMAMMALOGY IN AMAZONIAN PERU

3 Introduction.3 Materials and Methods.5 Systematic Accounts.13 Family Didelphidae.13 Subfamily Caluromyinae.14 Subfamily Glironiinae.17 Subfamily Hyladelphinae.20 Subfamily Didelphinae.23 Tribe Marmosini.24 Tribe etachirini.41 Tribe Didelphini.46 Tribe Thylamyini.57 Discussion.69 Acknowledgments.75 References.76 Appendix 1: Gazetteer.82 Appendix 2: Expected species.84 Appendix 3: Specimens of Caluromys sequenced for cytochrome b 85 Appendix 4: Specimens of Metachirus sequenced for cytochrome b 86 Appendix 5: Primers used to amplify cytochrome b.87

The introductory report in this series (Voss and Fleck, 2011) summarized current knowledge about the physical geography and floristics of the Yavari-Ucayali interfluve (figs. 1, 2), provided background information about the indigenous cultures of the region, and analyzed taxonomic and ethnographic data on primates. Our second report (Voss and Fleck, 2017)  guage, which is rich in descriptive terminology for both mammals and local habitats (Fleck and Harder, 2000); these were subsequently trans¬ lated by D.W.F. and transcribed to loose-leaf field journals that are now bound and archived at the AMNH.

Ethnobiological Methods
Recorded Monologs: From May to July of 1998 we elicited monologs about the natural history of local mammals from hunters at four different Peruvian Matses villages, and we recorded these interviews on digital minidisk.
All monologs were recorded in the Matses lan¬ guage. To elicit these texts, informants were asked to talk about a single mammalian folk taxon (e.g., cheka bebedi, "four-eyed opos¬ sums"), which was mentioned only once by the interviewer (Fleck). Informants were asked to say as much as they liked about any topic relat¬ ing to the taxon in question (see Voss and Fleck, 2017: appendix 1 for a free English   FIG. 5. Two pages from the field notes of a Matses man employed to hunt for marsupials and rodents at Nuevo San Juan. On the left-hand page, which records observations made on 5 October 1999, the second entry reads, "10:10 [pm] I killed an opossum in primary forest on the lower slope of a hillside. It was up in a slender tree." The corresponding voucher is MUSM 15311, a specimen of Marmosa constantiae.
of Matses assistants, who named the palms while in the forest, prior to being collected.
Palm specimens were identified in the field using published identification guides (Hender¬ son, 1994;Henderson et al., 1995), and voucher material was subsequently deposited at the her¬ barium of the Museo de Historia Natural de la  (Spichiger et al., 1989(Spichiger et al., , 1990. Plants and trees in the proximity of Nuevo San Juan were also identi¬ fied by Fleck using identification keys in Gentry (1993).
Botanical taxa corresponding to Matses Measurements: For specimens measured in the field according to the American proto¬ col (Hall, 1962), we transcribed total length (nose to fleshy tail tip, TL) and length of tail Morphological Terminology: Most of the terms for external morphology in this report fol¬ low Voss andJansa (2003, 2009). An important exception (not defined in those references) are descriptors of ventral fur coloration. Following Tate (1933), we refer to hairs that are the same color from root to tip as "self-colored"; fur that is composed of such hairs is then referred to as selfwhitish, self-yellowish, etc. By contrast, "graybased" fur is composed of hairs that have gray bases and paler (e.g., whitish or yellowish) tips. Both kinds of fur (self-colored and gray-based) can occur together, and their distribution in the ventral pelage is an important taxonomic criterion for species identification in several genera.
Age Criteria: Age determination is crucial for many aspects of didelphid taxonomy. Unless otherwise noted below, we recorded measure¬ ments and scored qualitative morphological data from adult specimens only. Following Voss et al. (2001), a specimen was judged to be juve¬ nile if dP3 is still in place, subadult if dP3 has been shed but P3 and/or M4 is still incom¬ pletely erupted, and adult if the permanent upper dentition is complete. Because didelphid deciduous premolars are molariform, juveniles have only two premolariform teeth (PI and P2) between the upper canine and the first molari¬ form tooth (dP3), whereas adults have three fully erupted premolariform teeth (PI, P2, and P3) between the upper canine and the first molariform tooth (Ml). Note that "juvenile," "subadult," and "adult" as used herein are descriptors of skeletal rather than reproductive maturity (many didelphids are reproductively active before the permanent dentition has com¬ pletely erupted; Diaz and Flores, 2008). Molecular Data: We analyzed DNA sequence data to supplement our morphologybased taxonomic accounts for Caluromys and Metachirus. DNA was extracted from ethanolpreserved tissues or from fragments of dried skin obtained from museum specimens (listed in appendices 3 and 4, respectively) using methods described by Voss and Jansa (2009) and Giarla et al. (2010). To minimize risk of contamination, all extractions from dried tissues were performed in an isolated laboratory where mammalian poly¬ merase chain reaction (PCR) products were not present. We PCR-amplified the mitochondrial gene encoding cytochrome b (CYTB) using the primers listed in appendix 5 and the methods described by Gutierrez et al. (2010) (Posada, 2008). We conducted maximum-likelihood analyses of the aligned sequence matrices using the GTRGAMMA model imple¬ mented in RAxML ver. 8. 2.10 (Stamatakis, 2014), with 1000 replicates of rapid bootstrap¬ ping to evaluate nodal support. All phylogenetic analyses were implemented in the CIPRES Sci¬ ence Gateway (Miller et al., 2010). We estimated uncorrected genetic distances within and among species and haplogroups using MEGA7 (Kumar et al., 2016 At least one additional caluromyine species and another three didelphine species are known from localities north and south of our region and might also be expected to occur there as well (appendix 2). The following accounts follow the classification of Voss and Jansa (2009), who pro¬ vided a key and morphological diagnoses for the genera, tribes, and subfamilies of Recent opos¬ sums. Emmons' (1997) field guide contains illus¬ trations and descriptions of external characters for all the genera and some of the species men¬ tioned below, but recent revisionary work (e.g., Rossi et al., 2010; has resulted in numerous changes to Emmons' species-level taxonomy. Ethnobiology: The Matses classify opossums into three categories: (1) cheka, which includes everything except short-tailed opossums and the common opossum; (2) vama, which includes sev-NO. 432 eral local species of short-tailed opossums (.Monodephis spp.); and (3) mapiokos, the com¬ mon opossum (Didelphis marsupialis). Only the first category is divided into named subtypes: abuk cheka, the "woolly opossum" (Caluromys lanatus); cheka bebedi, "four-eyed opossums" (Metachirus and Philander spp.); and chekampi, "mouse opossums" (including all local species of small, long-tailed, black-masked didelphines plus Hyladelphys kalinowskii).

Subfamily Caluromyinae
Only a single species of caluromyine, the woolly opossum (Caluromys lanatus), is known from the Yavari-Ucayali interfluve, but a second species (the black-shouldered opossum, Caluromysiops irrupta) could reasonably be expected to occur in our region (appendix 2).
Identification: Caluromys lanatus has never been revised, and several subspecies are cur¬ rently recognized as valid (Gardner, 2008). To assess the taxonomic significance of the current trinomial classification, we analyzed sequence data from the mitochondrial gene encoding cytochrome b, and we examined specimens from throughout western Amazonia, where no fewer than three subspecies-C. /. nattereri (Matschie, 1917), C. I ochropus (Wagner, 1842), and C. /. ornatus (Tschudi, 1845)-are said to occur.
Unfortunately, representative sequence data are currently unavailable from the nominotypical subspecies (in Paraguay), C. /. cicur (Bangs, 1898; in transAndean Colombia), and C. 1. vitalinus (Miranda-Ribeiro, 1936;in southeastern Brazil), so the materials at hand are insufficient for a comprehensive revision of the species.  190249, USNM 546177), and others are almost uniformly brownish (e.g., MUSM 15291). The base of the tail is reddish in some specimens, but grayish in others. A patch of grayish hair between the shoulders (said to be diagnostic of C. derbianus; Caceres and Carmignotto, 2006) is present in several specimens among those we examined (e.g., MVZ 157608, 190249, 190250).
Most specimens have uniformly gray-based fur on the throat, chest, and upper abdomen as well as on the ventral surfaces of the forelimbs, but others have irregular patches of self-cream fur on the throat and in the forelimb axillae; the lower abdomen and groin are seemingly always 2 Our geographic samples of Caluromys philander are far too few to support taxonomic conclusions, but it should be pointed out that the haplogroups we recovered are impossible to reconcile with the current trinomial nomenclature for this species, nor do they correspond to the taxa recognized by Lopez-Fuster et al. (2008), who suggested, inter alia, that the Trinidadian form {trinitatis Thomas, 1894) is a distinct species. (outgroups used for rooting are not shown). Branch tips are sequenced specimens labeled by geographic origin and a sequence identifier (see appendix 3); numbers in parentheses refer to localities mapped in figure 8. Branch support values above selected branches are bootstrap frequencies. Asterisks mark specimens from the Yavari-Ucayali interfluve.
self-cream, but parous females have orangestained fur surrounding the pouch. Although some of this pelage variation might be geo¬ graphic, there are substantial differences in col¬ oration among sympatrically collected skins, and intermediates exist among most coat-color phenotypes. Cranial size and shape differences among the specimens we examined do not sug¬ gest anything more than individual and ontoge¬ netic variation. In effect, we concur with Fonseca and Astuas (2015)  Matses Natural History: The woolly opos¬ sum is reddish. It has a stripe going down its rostrum. Its tail is partly bare. Its ears are large.
The woolly opossum is arboreal, but it some¬ times descends to the ground to forage. It descends by climbing down a vine. It is found in all rainforest habitat types. It makes its nest in tree cavities that it lines with leaves. It also makes nests in the branches of dicot trees, or in leaf lit¬ ter that collects in palm crowns.
The woolly opossum is nocturnal. After forag¬ ing for dicot tree fruits and insects during the night, it returns to its nest at dawn. It climbs quickly on vines.
The woolly opossum is solitary. It gives birth to many young that it keeps in its pouch. The young suck milk inside the pouch. When the young get big, they leave their mother, one by one.
Margays and snakes prey on the western woolly opossum.
When the wooly opossum sees people, it opens its mouth and hisses, wanting to bite them. Remarks: Four of our six vouchers were shot from trees at night by Matses hunters, three in secondary forest (abandoned swiddens) and one in hilltop primary forest; heights were not recorded, but the specimen shot in primary for¬ est was said to be "very high up in a big tree" (English translation from Matses field notes).
One specimen was captured by hand in the late afternoon by several boys, who shook it from the crown of a small guaba tree (Fabaceae: Inga edulis) on the outskirts of the village. The specimen from Jenaro Herrera was trapped on a liana 18 m above the ground in disturbed primary forest.

Subfamily Glironiinae
The only species of this subfamily that is cur¬ rently recognized as valid, Glironia venusta, is known to occur in our region from a single specimen.
Identification: As currently understood, Glironia venusta ranges across much of Amazo¬ nia but remains known from only a few speci¬ mens (Barkley, 2008;Ardente et al., 2013).
Among the handful that we were able to examine for this report were the holotypes of two nominal species-aequatorialis Anthony, 1926, and criniger Anthony, 1926- Thomas (1912b). Measurements are in quotes because external dimensions were measured according to the British protocol (Lankester, 1904), and craniodental measurements may have been affected by preservational artifacts (Thomas, 1912b). by Anthony (1926) are nothing more than intra¬ specific variation.
Ethnobiology: The Matses have no definite knowledge of this species, and therefore do not have a name for it or any particular beliefs about it.
Matses Natural History: The Matses have no knowledge of the appearance or behavior of this species.
Remarks: According to Barkley (2008), LSU 28421 was captured at night in a mist net set for bats in primary forest. Local habitats at the cap¬ ture site, which is not subject to seasonal inunda¬ tion, were described by .

Subfamily Hyladelphinae
The only currently recognized species in the subfamily Hyladelphinae is known from our region on the basis of just two specimens from a single locality Hyladelphys kalinowskii (Hershkovitz, 1992) Figures 9A, 10A Voucher Material (total = 2): Nuevo San Juan (AMNH 276725; MUSM 11031).
Identification: This tiny marsupial, long unknown to science and only recently recog¬ nized as the sole survivor of an ancient lineage, is still represented by fewer than two dozen museum specimens despite its wide Amazonian distribution and apparent lack of habitat specific¬ ity (Hershkovitz, 1992;Voss et al., 2001;Jansa and Voss, 2005;Astua, 2006;Hice and Velazco, 2012;Catzeflis, 2018). Our two specimens con¬ sist of the damaged skull of a juvenile that retains the diagnostically reduced milk dentition (Voss et al., 2001: figs. 17, 18), and the fluid-preserved carcass and extracted skull of an adult female. Diaz (2014) erroneously reported that the latter specimen (AMNH 276725) was collected at Jenaro Herrera.
Hyladelphys kalinowskii is one of three superficially similar species in our region, all of which are very small (<30 g) opossums with long tails, black masks, and reddish-brown dorsal fur. Despite these resemblances, they are only distantly related to one another, and specimens in hand are easily identified by numerous integumental and craniodental dif¬ ferences (figs. 9, 10; table 4). Traits unique to H. kalinowskii among this trio of tiny didelphids include the posterior extent of its black¬ ish facial markings, possession of just four mammae, an indistinctly banded tail (due to paler skin over the vertebral articulations), lack of a premaxillary rostral process, absence of a posterior accessory cusp on Cl, a third upper premolar (P3) that is conspicuously smaller than P2, and an exceptionally short upper molar row (LM <4.6 mm). Hyladelphys kalinowskii additionally differs from Marmosa lepida by its exclusively self-white ventral fur and lack of postorbital processes. Hyladelphys kalinowskii additionally differs from Gracilinanus emiliae by having a much shorter tail, and by lacking a gular gland, palatine fenestrae, and secondary foramina ovales.
Both of our specimens conform in all quali¬ tative traits to the emended description of Hyladelphys kalinowskii provided by Jansa and Voss (2005). Measurements of our adult female specimen are a bit smaller than those of the adult female holotype from Cuzco department and the adult female paratype from Junin (table 5), but without any adequate population sample to assess individual (nongeographic) variation in this species, such differences are hard to interpret. Although Jansa and Voss (2005) discussed the possibility that multiple species of Hyladelphys might be represented among the specimens currently referred to H. kalinowskii, ours are so similar to Hershkovitz's (1992) original material that this identifi¬ cation would seem to be beyond dispute.  French Guiana-Paracou (AMNH 267003, 267338, 267339). Hacienda 4 Although the original phrasing of the Matses conversation that elicited this information was not recorded, the usual term for leaf litter is shapu, which can either refer to a thick layer of dead leaves on the ground or to an accumulation of dead leaves in the crown of a stemless palm (typically 1-2 m above the ground). Unfortunately, this ambiguity cannot now be resolved, but the latter interpretation would be more consis¬ tent with what is currently known about the nesting habits of Hyladelphys (Catzeflis, 2018).  Ethnobiology: The opossums that the Mat¬ ses call chekampi ("little opossums") are superfi¬ cially similar (small, black-masked, long-tailed, and pouchless), although some of our interview¬ ees acknowledged that chekampi come in differ¬ ent sizes and in different shades of gray and brown. Some Matses use mapiokosempi (the name of the common opossum with a diminu¬ tive suffix) as an alternative name.

Mouse opossums come into Matses houses
and eat their food. Sometimes they make nests in Matses houses, but they seldom stay long. The Matses say that mouse opossums usually enter their houses during heavy rains.
Matses Natural History: Mouse opossums have long tails and large ears. Some are gray, while others are reddish, and others are darkcolored. They are similar to four-eyed opossums, but much smaller.
Mouse opossums are arboreal and terrestrial. They are abundant in the forest and sometimes come into Matses swiddens. They make nests of dry leaves in thick vegetation up in trees, in cavi¬ ties in tree branches, or in leaf litter that accumu¬ lates in the crowns of palms. In Matses swiddens they make nests among plantain plants using dead plantain leaves. In houses they make nests in containers where clothes are kept.
Mouse opossums are strictly nocturnal. They give birth to many young. They always carry their young with them. Otherwise they are solitary.
Mouse opossums are eaten by margays and snakes. They make a high-pitched hiss when they are threatened.
Mouse opossums eat all sorts of things. They eat crickets/katydids; large cockroaches; and small birds, eggs, and hatchlings that they find at night in nests. They also eat dicot tree fruits, including those of the mannan tsipuis tree (Inga spp.; Mimosoideae) and the mesocarp of swamppalm (Mauritia flexuosa) fruits.
Other Interfluvial Records: None that can be confidently associated with this species.  Tate [1933: 83], who saw the type when it was less than 40 years old), the upper molars are smaller (LM = 8.1 mm), and the postprotocristae are short. Like Tate (1933), we restrict the application of Marmosa regina to the holotype, and we reject the application of this name to any of the Amazonian nominal taxa treated as subjective junior synonyms or subspecies by Gardner (1993Gardner ( , 2005. According to Patton et al. (2000), the nextoldest name for this species is Marmosa germana Thomas, 1904, which was treated as a synonym of M. regina by Gardner (1993), but as a valid subspecies of M. regina by Gardner (2005) and by Gardner and Creighton (2008).
However, the type of germana (BMNH 80.5.6.77, an adult female from eastern Ecua¬ dor; Thomas, 1904b) Thomas (1924) described Marmosa rutteri as a valid species, but Tate (1933: 81) considered rut¬ teri to be a subspecies of M. germana, and most subsequent authors have considered rutteri to be a junior synonym of germana (now usually ranked as a subspecies of regina; e.g., by Gardner, 2005;Gardner and Creighton, 2008). In fact, M. rutteri is a distinct species with the diagnostic traits listed in table 8, and analyses of DNA sequence data (Voss et al., in prep.) suggest that it is more closely Remarks: Of our four specimens from Nuevo San Juan, one was captured by hand inside a Matses house, one was trapped on top of a fallen log in primary upland forest, one was shot from a tree (at an unrecorded height) in primary upland forest, and one was taken by hand inside a hollow tree at the edge of the Rio Galvez.
Two subspecies of Marmosa lepida were recognized by Tate (1933), of which our mate¬ rial is unambiguously referable to the nominotypical form. Marmosa l. lepida occurs on both sides of the Amazon, from the base of the Andes to the Atlantic coast, with remarkably little evidence of geographic variation in either morphology or cytochrome b sequences (Tate, 1933;Gutierrez et al., 2010;Guimaraes et al., 2018). By contrast, M. 1. grandis Tate, 1931, is known only from Buenavista in Santa Cruz province, Bolivia. We examined the young adult female holotype of grandis (BMNH 26.12.4.94), which differs from other speci¬ mens of M. lepida in ventral pelage color (entirely gray-based buffy) and in lacking pos¬ terior accessory cusps on Cl and cl. Given the morphological uniformity exhibited by Ama¬ zonian specimens of M. lepida, we are not per¬ suaded that grandis is really conspecific.
Although Tate (1933: 205)   is also used to refer to the turnip-tailed gecko (Gekkonidae: Thecadactylus solimoensis) and to an unidentified arboreal rat that was for¬ merly used for black magic.7 The Matses do not consider vama to be a type of cheka.
7 Several generations ago, according to the Matses, men of evil intention would concoct a poison from yama rats, which are said to be found on leafless trees that have recently died. The poisoner would place a large clay pot containing ripe plantains at the base of such a tree at night and hide nearby to wait. When the rat entered the pot and began to eat the plantains, the poisoner would place a lid over the pot and take it to an isolated hut. There, without lifting the lid, he would place the pot over a fire until the contents were completely burned. Next, he would gather the burned remnants into a length of bamboo, which was used as a mortar to grind the contents to dust and ash. This pulverized substance was then sprinkled on the head of a sleeping victim, who would breathe in the rat ashes, become insane, and die shortly afterward.  . 6).
e Lingual to hypoconids of ml and m2, respectively. Monodelphis (Mygalodelphys) peruviana (Osgood, 1913) Voucher we found no external morphological differences between these taxa: whereas he reported that adusta has shorter dorsal fur than peruviana, our measurements indicate that both species have dor¬ sal fur that ranges from 3 to 4 mm long, and we were unable to distinguish the "ill-defined blackish area on the posterior dorsum and the rump" that was said to be present in adusta, but absent in peru¬ viana. Likewise, we observed no consistent qualita¬ tive craniodental differences between specimens collected on opposite banks of the Amazon. Sample sizes are too small for confident statistical compari¬ sons of craniodental measurements, but broadly overlapping ranges for most dimensions (table 12) suggest that morphometric differences, if any, are unlikely to provide a secure basis for species separa¬ tion. Although we are currently unable to provide diagnoses of the taxa currently associated with these names, it seems prudent to maintain current usage pending a comprehensive revision of the sub ¬ genus Mygalodelphys.   PB 9.1 ± 0.5 (8.3-9.6) 7 9.2 ± 0.2 (8.9-9.5) 5 9.3± 0.5 (8.8-9.9) 4 8.2 MTR 11.2 ± 0.6 (10.6-12.4) 7 11.2 ± 0.4 (10.8-11.8) 5 11.0 ± 0.2 (10.8-11.1) 4 10.7 LM 6.1 ± 0.3 (5.9-6.7) 7 6.1 ± 0.2 (5.8 -6.3) 6 5.9 ± 0.2 (5.6-6.0) 4 5.5 Ml-3 5.3 ± 0.2 (5.1-5.6) 7 5.3 ± 0.2 (5.0-5.5) 6 5. Monodelphis (Pyrodelphys) emiliae (Thomas, 1912)  nally reported by Thomas (1912a) and subse¬ quently noted by Pine and Handley (1984), Patton et al. (2000), and Pavan and Voss (2016: fig. 14) was also seen in our fresh material. The   Although Monodelphis emiliae was reported from the "Iquitos area" by Patton et al. (2000), all the specimens known to have been col¬ lected in northeastern Peru are from the right ("south") bank of the Amazon.9 Despite many decades of collecting near Iquitos-on the left 9 Note that, due to the river s convoluted course, the right bank of the Amazon near Iquitos is actually the east side and the left bank is the west side ( fig. 2, inset). To avoid confusion, how¬ ever, we use "south" and "north" with respect to the river's macrogeographic orientation.

Tribe Metachirini
Members of this tribe are commonly known as "brown four-eyed opossums" or "pouchless four-eyed opossums" to distinguish them from the grayish, pouched four-eyed opossums referred to Philander. The only included genus, Metachirus, was long thought to be monotypic, but molecular sequence analyses reported by Patton et al. (2000) and Patton and Costa (2003) recovered several mtDNA haplogroups, some of which were so highly divergent as to suggest the existence of multiple species.
Indeed, there is no lack of available names for brown four-eyed opossums: in the last synthe¬ sis of the taxonomic literature (Gardner and Dagosto, 2008), no fewer than 11 epithets were treated as synonyms or subspecies of M. nudicaudatus (Geoffroy, 1803), the type locality of which is in northeastern Amazonia (French Guiana). In the account that follows, we explain why we use a different binomen for the brown four-eyed opossums of the Yavari-Ucayali interfluve.
Other Interfluvial Records: Jenaro Her¬ rera (as Metachirus nudicaudatus; Pacheco, 1991;Pavlinov, 1994;Fleck and Harder, 1995), San Pedro (as M. nudicaudatus-, Valqui, 1999Valqui, , 2001.  13B). Additionally, the rostrum tends to be more robust and the zygomatic arches tend to be more rounded laterally in southwestern Amazonian material than in M. nudicaudatus, which typically has a longer, narrower rostrum NO. 432  . 14), and statistical comparisons of fenestral length and width confirm that these visually obvious differences are highly significant (p < 0.0001 by two-tailed Students t tests). Although subtle differences in the dentition also appear to distinguish southwest¬ ern Amazonian specimens from typical M.
Other Interfluvial Records: Nuevo San Juan (this report), Quebrada Pobreza (Pitman et al., 2015), San Pedro (Valqui, 1999), Santa Rosa (this report).  b Gardner and Dagosto (2008: 37) incorrectly spelled this epithet as "inbutus." c The lectotype designated by Pohle (1927). A specimen in the Berlin Museum (ZMB 2326) is incorrectly labelled as the type of myosuros: it is not the specimen designated as lectotype by Pohle and cannot have been part of the original material examined by Temminck (1824), who did not mention any specimens from Berlin.

Although central Amazonia (including the
Yavari-Ucayali interfluve) falls outside the mapped geographic range of Chironectes mini¬ mus in most standard references (e.g., Stein and Patton, 2008), recent observations and gene-sequencing results (Ardente et al., 2013;Brandao et al., 2015;Oliveira et al., 2016; suggest that the water opos¬ sum is continuously distributed from Central America to southeastern Brazil. Despite geo¬ graphic variation in craniodental morphology reported by Damasceno and Astua (2016) and Cerqueira and Weber (2017), this species seems to be genetically undifferentiated across most, if not all, of cis-Andean South America, and it shows minimal genetic divergence even across the Andes . Therefore, the "subspecies" of C. minimus cur¬ rently recognized by authors (Gardner, 2005;Stein and Patton, 2008)
Identification: Didelphis marsupialis, the socalled common or black-eared opossum of Ama¬ zonia and Central America, is externally unmistakable (Husson, 1978;Emmons, 1997), and although similar in most qualitative aspects of craniodental morphology to species of Philander (Voss and Jansa, 2009), it is so much larger in all dimensions that skulls, and even isolated teeth, are easily identified. There is remarkably little genetic variation in this species throughout its geographic range (e.g., <1.6% mean sequence divergence at the cytochrome b locus among samples from Costa Rica to central Amazonia;Patton et al., 2000: fig. 39). Measurements of the few adult specimens from our region (table 18) overlap the range of variation reported by Patton et al. (2000) for material collected along the Rio Jurua in west¬ ern Brazil, but our specimens are large by com¬ parison with their tabulated sample means.
Although the name Didelphis marsupialis has been used consistently for many years to refer to the black-eared Amazonian opossum, Gurgel-Filho et al. (2015) resurrected the obsolete synonym D. karkinophaga Zimmermann, 1780, for this species and proposed that Linnaeus's (1758) epithet be used for the white-eared species that has long been known  Dias et al. (2018), who thought that a specimen in Uppsala might be the lectotype designated by Thomas (1911 The common opossum is nocturnal and soli¬ tary. The female carries its young in its pouch and suckles them inside the pouch. It gives birth to many young.
The common opossum is eaten by jaguars, ocelots, and margays. It vocalizes saying "chocod chocod." The common opossum eats spiny rats and birds, including tinamous, that it finds nesting on or near the ground at night. It enters coops to eat chickens, and it also eats pet guans, crickets/ katydids, cockroaches, and rotten meat.
Remarks: Of our seven specimens, two were trapped on the ground in a swamp-palm (Mauritia flexuosa) swamp, one was trapped on the ground in primary upland forest, one was trapped at a height of 9 m in a tree in primary upland forest, one was caught by hand 2 m above the ground on a sapling in primary upland for¬ est, one was trapped by a Matses boy (probably on the ground in secondary forest), and one was shot by a mestizo hunter in unrecorded circum-  Philander Brisson, 1762 Two species of gray four-eyed opossums, Philander mcilhennyi and P. pebas, are definitely known to occur in the Yavari-Ucayali interfluve, and a third species (P. canus) could be expected to occur there based on geographic range data (appendix 2). Specimens of both P. pebas and P. canus have long been misidentified as P. opos¬ sum (an eastern-Amazonian taxon; , so the identity of animals previously reported as P. opossum from the Yavari-Ucayali interfluve (Pavlinov, 1994;Fleck and Harder, 1995) is unclear. To facilitate the identification of specimens collected in the course of future fieldwork, we summarize qualitative traits and morphometric variation for all three species (tables 19-21).
The Matses do not consistently recognize more than a single kind of opossum with pale supra¬ ocular spots, so the name they apply to such ani¬ mals, cheka bebedi, could refer either to gray four-eyed opossums or to the superficially similar brown four-eyed opossum (Metachirus myosuros). Because species of Philander are more commonly encountered than M. myosuros, we summarize relevant ethnographic information here.
Ethnobiology: Cheka bebedi, the most fre¬ quently used name for these species, literally means "spotted-forehead opossums," but in some Matses villages they are called cheka dewisak "long-snouted opossums." They are sometimes also referred to as chekadapa "big opossum," but this is not considered a real name; rather it is a descriptive phrase used to distinguish these large species from mouse opossums. Some Matses rec¬ ognize that there is more than one local type of cheka bebedi, but they do not differentiate them linguistically.
The Matses do not eat these opossums, although children sometimes shoot them with arrows. One informant mentioned that a four¬ eyed opossum ate his pet dove.
They are arboreal and terrestrial. They are common in all rainforest habitats but seem to be more common along streams. Four-eyed opossums are nocturnal.
Four-eyed opossums have many young, which they carry around and suckle inside their pouches. The young leave the mother when they get about half the size of the mother. Other than mothers with young, they are solitary.
Four-eyed opossums are eaten by margays and ocelots.

Four-eyed opossums open their mouth very
wide and hiss aggressively when found by peo¬ ple, and try to bite them if they get close.
Among the birds they prey upon are tinamous, ground doves, and rusty-belted tapaculos.
Identification: Amazonian specimens of Philander with distinctly blackish middorsal fur belong to two strongly supported cytochrome b haplogroups (Patton and da Silva, 1997;Patton et al., 2000;. One haplogroup is represented by specimens collected north of the Amazon (in northern Loreto, eastern Ecua¬ dor, southeastern Colombia, southern Venezu¬ ela, and northwestern Brazil), but it also includes a sequence obtained from a paratype of P. andersoni (Osgood, 1913), which was col¬ giving some skins a distinctly shaggy appearance.
By contrast, typical material of P. andersoni has paler-grayish flanks, such that the black middor¬ sal stripe is more obvious; additionally, the mid¬ dorsal fur of andersoni is said to be shorter than that of mcilhennyi, and the ventral fur is either self-whitish or gray-based whitish (rather than dark gray). Lastly, P. andersoni is said to have a medial patch of pale fur at the base of the ear that is absent from the entirely blackish crown of P. mcilhennyi (see da Silva, 1997, 2008). Patton et al. (2000) remarked that Philander andersoni and P. mcilhennyi might be sympatric on the "lower Rio Javari in northeastern Peru," cit¬ ing a personal communication with D.W.F., who had observed both phenotypes at Nuevo San Juan (on the Rio Galvez). In fact, our voucher material includes specimens that distinctly resemble P. mcilhennyi (e.g., AMNH 272818, MUSM 13299) and others that closely resemble P. andersoni (e.g., AMNH 273055; fig. 19). Cytochrome b sequences that we obtained from specimens of both pheno¬ types, however, all belong to the mcilhennyi hap¬ logroup , so introgression rather than sympatry might be the more appropriate interpretation of pelage variation at this locality.
Nevertheless, we maintain current binomial usage pending further genetic analysis of our material (Jansa and Voss, in prep.).
A curious aspect of intraspecific variation in both Philander mcilhennyi and P. andersoni is sexual dimorphism in the middorsal fur, which is significantly longer, on average, in females than in males (table 22). Same-sex comparisons, however, support da Silvas (1997, 2008) description of P. mcilhennyi as longerfurred than P. andersoni, although there is spe¬ cies overlap in our measurements of fur length even when comparing males with males and females with females. To our knowledge, sexual dimorphism in fur length has not been reported from other didelphid species.  Ucayali interfluve by Pavlinov (1994)  Gracilinanus emiliae (Thomas, 1909)  Gracilinanus emilae is the only species of Gracilinanus expected to occur in the Yavari-Ucayali interfluve, although G. peruanus-formerly synonymized with G. agilis (see Semedo et al., 2015)-is known from scattered localities in the Recently reported specimens of Gracilinanus emiliae from scattered localities in Brazil (Silva et al, 2013;Brandao et al, 2014;Rocha et al, 2015) confirm that this species is very widely distrib¬ uted in Amazonia. Despite substantial (ca. 5%) divergence at the cytochrome b locus between our Peruvian voucher and the eastern Amazo¬ nian material analyzed by Rocha et al. (2015), 10 The specimen that Huamani et al. (2009) reported as Gra¬ cilinanus agilis from the Zona Reservada Pucacuro in north¬ ern Loreto (MUSM 24430) is a juvenile specimen of G. emiliae. Another specimen that Huamani et al. (2009) identi¬ fied as G. agilis (MUSM 14085) is an unidentifiable Marmosops with molar teeth worn almost to the roots. The specimens from Bellavista (in Cajamarca department [Stephens and Tray¬ lor, 1983]) that Tate (1933: 199) tentatively identified as "Mar¬ mosa agilis peruana" (MCZ 17057-17059) consist only of skins; although these are unambiguously identifiable as Gra¬ cilinanus, and are clearly not G. emiliae (the ventral fur is extensively gray-based), they are not identifiable to species without cranial material. this appears to be a phenotypically homogeneous taxon, without any obvious morphological differ¬ ences among the specimens we examined. Marmosops Matschie, 1916 Species of Marmosops superficially resemble species of Marmosa, but external and cranioden¬ tal traits that distinguish these distantly related taxa-first recognized as separate genera by Gardner and Creighton (1989)-were described and illustrated by Voss et al. (2004). Subsequent phylogenetic research revealed a deep dichotomy in the genus that was recently formalized by subgeneric nomenclature .
Eleven nominal taxa are currently regarded as synonyms of Marmosops noctivagus (e.g., by Voss and Jansa, 2009), but it is currently difficult to determine the application of names to most of the haplogroups discovered by , a problem that those authors discussed at were taken in secondary growth (abandoned swiddens), and 2 (8%) were captured in houses.
Of 12 specimens accompanied by substrate infor¬ mation, 9 (75%) were trapped on the ground or on fallen logs, 2 (17%) were trapped on lianas between 1.5 and 1.8 m above the ground, and one was taken by hand as it perched close to the ground on a small tree. A single specimen was taken in the daytime from a leaf nest in the crown of an Astrocaryum palm at an unrecorded height above the ground, and another specimen was found dead, but all the specimens shot or captured by hand while active were taken at night, and all of the specimens trapped by D.W.F. and R.S.V.
were found at dawn in traps that had been baited in the late afternoon of the previous day.  268216, 272709, 272760, 273050, 273078, 273151, 273189;MUSM 11040, 11046, 11047, 13285, 13286, 15298-15300, 15306, 15307) Emmons, personal commun.), of which the two anteriormost pairs are "thoracic" (sensu Tate, 1933), extending toward the lower chest from the 13 As explained by , the epithet impavidus has been used for at least two distinct species, of which one has traits that are inconsistent with the original descrip¬ tion, and the other is not known to occur near the type local¬ ity. The type is apparently lost, and it is not certain that the animal described by Tschudi was even a species of Marmosops.  Remarks: Of the 23 specimens for which we have capture information, 13 were trapped, shot, or captured by hand while climbing on saplings, lianas, logs, or fallen branches at estimated heights from 0.2 to 1 m above the ground; only five are definitely known to have been captured on the ground (of which three were taken in pitfalls). Most (15) specimens were taken in pri¬ mary upland forest, but four were in secondary forest (abandoned swiddens), two were in swampy primary forest, one was in primary floodplain forest, and one was captured in a house. Ten specimens were shot or captured by hand while active at night, and another 10 were found at dawn in traps that had been baited in the late afternoon of the previous day (three specimens found at dawn in pitfalls might have been taken at any time in the previous 24 hours). Marmosops (Sciophanes) bishopi (Pine, 1981)  Identification: The small species of Mar¬ mosops formerly associated with the name M. parvidens (e.g., by Pine, 1981) were referred to the subgenus Sciophanes by . Specimens from the Yavari-Ucayali inter¬ fluve were subsequently examined by , who referred them to M. bish¬ opi, a taxon originally described as a subspecies of M. parvidens on the basis of a female holotype collected in central Brazil. As recognized by   Voss and Fleck, in prep.). How¬ ever, it is possible that even more marsupial spe¬ cies might occur locally but were not recorded by us or by the other collectors and researchers who previously worked in the region. Because inven¬ tory completeness is a potential problem for fau-  (table 28) gives a very rough idea of species encounter rates, these data did not result from a uniform sampling process (sensu Colwell and Coddington, 1994), so they are not really suitable for diversity extrapolation or other forms of statisti-15 Unlike primates and large mammals, most marsupials can¬ not be reliably identified without specimens in hand (Voss and Emmons, 1996) (Voss and Emmons, 1996). By contrast, we are not aware of any com¬ pelling evidence for environmental "filtering"

Marmosops (Marmosops) soinii, new species
(sensu Kraft et al., 2015) or biotic interactions that might account for the absence of any of these gen¬ era at Amazonian rainforest sites where they are not known to occur.
Despite the generic-level homogeneity of Amazonian marsupial faunas, species turnover (beta diversity) along the SW/NE axis is substan¬ tial. Of the 25 marsupial species represented in the combined faunal lists from Nuevo San Juan and Paracou, 20 species (80%) are represented at one site or the other, but not at both. What dis¬ persal barriers or environmental gradients might account for such numerous replacements of one congeneric species by another across the 2500 km that separate these inventory sites?
The short answer is, we do not yet know.
c Unpublished sequencing results (Voss et al., in prep. are unknown from any left-bank counterparts despite intensive recent collecting between Iqui¬ tos and Nauta (Hice and Velazco, 2012;Diaz, 2014). Therefore, dispersal of somewhat more than half of the regional marsupial fauna seems to be limited by the Amazon to one degree or another. Additionally, it seems noteworthy that few of the species or haplogroups that replace one another on opposite sides of the river are sister clades, a phenomenon that has previously been noted in studies of other organisms sepa¬ rated by Amazonian rivers (Naka and Brum¬ field, 2018).

Matses Ethnomammalogy
Matses knowledge of marsupials is very incomplete by comparison with their detailed knowledge about primates and large mammals (Voss andFleck, 2011, 2017  a Also Thecadactylus solimoensis (Squamata: Gekkonidae) and an unidentified rodent.
exception of the common opossum (Didelphis marsupialis)-which enters villages to eat chick¬ ens-they are seldom encountered. Because opossums are culturally unimport¬ ant (inedible and not dangerous), the Matses seldom need to talk about them, so their mar¬ supial lexicon is correspondingly underdifferen¬ tiated with respect to zoological taxa (table 32). Didelphis marsupialis has its own proper name, and short-tailed opossums (Monodelphis spp.) are also recognized, albeit collectively, as belonging to a distinct folk taxon, but all the rest are treated as kinds of cheka and labelled according to arboreality, size, or markings. The water opossum, unknown to most of the Mat¬ ses, is not named. There is no Matses name, nor, apparently, even a covert category (sensu Berlin et al., 1968) that includes all local species of didelphids, although the variant term for chekampi implies some recognition of kinship between mouse opossums and the common opossum.
Matses inattention to opossum natural history and taxonomic diversity is consistent with their scant knowledge of (and rudimentary lexicons for) other small and culturally unimportant ani¬ mals (e.g., bats; Fleck et al., 2002). It makes sense that the large, chicken-thieving, and externally unmistakable common opossum should have its own name, but vama is something of an enigma. As noted earlier, this name applies not only to species of Monodelphis, but also to a lizard and to an unidentified rat formerly used for sorcery. It is not obvious what these animals have in com¬ mon, but the supernatural powers of vama rats (which are not limited to their former use for magical murder) suggests that it might be some long-forgotten occult attribute.

APPENDIX 2 Expected Species
Caluromysiops irrupta: The black-shouldered opossum is an elusive arboreal species known to occur along the left (north) bank of the Amazon in Loreto department and in Madre de Dios (Emmons, 2008). The few published observations of C. irrupta accompanied by definite habitat information (e.g., Janson et al., 1981) suggests that it occurs in whitewater floodplains; therefore, if it occurs in our region, it is perhaps to be found in the still poorly sampled riparian forests along the lower Ucayali and the right bank of the Amazon.
Philander canus: The Yavari-Ucayali inter¬ fluve occupies a wide gap in the known distribution of this species   fig. 9), which is known to occur north of our region in Venezuela and eastern Colombia (where it was formerly known as P. mondolfii Lew et al., 2006) and south of our region in Peru, Bolivia, Paraguay, central Brazil, and northern Argentina. The closest speci¬ men personally examined by us is from Moyobamba (6°03'S, 76°58'W; the type locality) in San Martin department, but two specimen that Diaz (2014) identified as P. olrogi Flores et al., 2008 (a junior synonym of P. canus;  were collected on the adjacent left (north) bank of the Amazon about 40 km SW of Iquitos. The geo¬ graphic distribution of collection localities for this species suggests that it occurs in upland habitats, but we have not found any explicit description of the Amazonian habitats in which it has been taken.
Marmosa rubra: This western Amazonian species is known from just 12 localities (Rossi et al., 2010), of which 11 are north of the Yavari-Ucayali interfluve (in southeastern Colombia, eastern Ecuador, and northern Loreto) and one is to the south (in Cusco). The specimens closest to our region were collected by the Olallas in 1925 at a locality they called "Boca Rio Curaray," about 180 km NW of Iquitos (Wiley, 2010: fig 2).
Nothing is apparently known about the ecologi¬ cal circumstances in which this species occurs.
Marmosops caucae: This is the oldest available name for the species that Patton et al. (2000) called M. neblina Gardner, 1989. Based on specimens sequenced by   fig. 1), this species-or species complex-occurs north, south, east, and west of the Yavari-Ucayali interfluve; the closest to our region are from the Rio Jurua (e.g., at Igarape Nova Impresa: 6°48'S, 70°44'W) in western Brazil, where they were taken in "varzea forest or disturbed river-edge areas" (Patton et al., 2000: 57).
Other specimens collected not too far from the Yavari-Ucayali interfluve include two (FMNH 87119,87120) from Santa Elena (ca. 4°50' S, 74°13' W) on the Rio Samiria. We have not examined the material that Diaz (2014) identified as M. neblina from the left bank of the Amazon south of Iquitos, but no specimens referable to M. caucae were among the many small didelphids previously col¬ lected from the same general area by Hice and Velazco (2012).