Holochilus (Smith and Patton, 1999)

Weksler, M., 2006, Phylogenetic Relationships Of Oryzomine Rodents (Muroidea: Sigmodontinae): Separate And Combined Analyses Of Morphological And Molecular Data, Bulletin of the American Museum of Natural History 2006 (296), pp. 1-149 : 78-89

Holochilus , angouya , and Amphinectomys + Nectomys ).

ARBOREAL AND SEMIAQUATIC SPECIALI- ZATIONS: Oecomys is the only oryzomyine taxon with obvious specializations for arboreal life. Although species belonging to other oryzomyine clades (such Oligoryzomys and the subflavus group of Oryzomys ) are sometimes reported to have some arboreal capacity ( Alho, 1982; Fonseca and Redford, 1984; Alho and Villela, 1985; Mares et al., 1986), most individuals of these taxa are collected on the ground ( Mares et al., 1989; Bonvicino et al., 2005). In contrast, various studies have shown that Oecomys specimens are found mostly, and sometimes exclusively, in trees, often as high as 15–20 m above the ground ( Hershkovitz, 1960; Mares et al., 1989; Patton et al., 2000; Voss et al., 2001). The main adaptations for arboreal life observed in Oecomys include: more robust and compact hindfeet, with lateral broadening of the metatarsus; first and fifth digits more powerful and opposable; and plantar pads modified for grasping ( Hershkovitz, 1960). These features are also observed in other arboreal sigmodontine rodents (e.g., Rhipidomys ).

In contrast, semiaquatic (amphibious) specializations are observed in seven oryzo- myine taxa: Amphinectomys , Nectomys , Lundomys , Holochilus , Sigmodontomys alfari , Pseudoryzomys , and in the palustris group of Oryzomys . The degrees of morphological adaptations of these taxa are proportionate to the amount of time the animals spend in water and to the type of activities performed there (i.e., foraging and nestbuilding; Stein, 1988). Specimens of Holochilus , Lundomys , Amphinectomys , and Nectomys , which are found almost exclusively alongside bodies of freshwater, display well-developed interdigi- tal webbing and natatory fringes, whereas Oryzomys palustris , Pseudoryzomys simplex , and Sigmodontomys alfari , which are less dependent on water bodies, have reduced interdigital webbing and do not display natatory fringes. The same pattern is observed in relation to the size of these rats: Holochilus , Amphinectomys , Nectomys , and Lundomys are the largest living oryzomyines, while Sigmodontomys , O. palustris , and Pseudoryzomys are all medium-sized rodents. This size gradient is in agreement with the hypotheses of larger body size for semiaquatic small mammals ( Wolff and Guthrie, 1985).

All semiaquatic oryzomyines are members of clade D, where they are divided into two lineages: (1) the palustris group, Pseudoryzomys , Lundomys , and Holochilus ; and (2) Amphinectomys , Nectomys , and Sigmodontomys alfari , together with the terrestrial S. aphrastus and Melanomys caliginosus . Reconstruction of ancestral states of the characters related to the semiaquatic habitus (fig. 45) indicates that such adaptations occurred at least twice within oryzomyines. Natatory fringes appeared in the last common ancestor (LCA) of Lundomys and Holochilus and in the LCA of Amphinectomys and Nectomys . Optimizations of interdigital webbing transformations, however, are ambiguous (fig. 45). In ACCTRAN optimization, interdigital webs evolved twice and were lost twice: small membranes (not extending to first interphalangeal joints) appeared in the LCA of the palustris group + Lundomys + Holochilus + Pseudoryzomys , and in the LCA of Amphinectomys + Nectomys + Melanomys + Sigmodontomys ; well-developed webs (extending to or beyond first interphalangeal joints) appeared in the LCA of Lundomys + Holochilus , and in the LCA of Amphinectomys + Nectomys ; webs were lost in the lineage leading to Oryzomys couesi and in the LCA of Melanomys + Sigmodontomys aphrastus . In DELTRAN optimization, webbing appeared five times and was not lost in any lineage: small webs appeared in the LCA of Lundomys + Holochilus + Pseudoryzomys , and independently in the lineages leading to O. palustris and Sigmodontomys alfari , whereas large webs appeared in the LCA of Holochilus + Lundomys and in the LCA of Nectomys + Amphinectomys . Note that in ACCTRAN optimization, the sequence of modifications of the two hindfoot specializations follow the same pattern in the two semiaquatic clades, with the appearance of small webs preceding the synchronous development of long webs and natatory fringes.

Other hindfoot characters also seem to be influenced by the transition to a semiaquatic mode of life (fig. 45). Hypothenar pads, interdigital pads, and ungual tufts, for example, are all reduced or lost in semiaquatic taxa. Changes in these characters are synchronous with ACCTRAN-optimized changes in known semiaquatic specializations. Interdigital pads and ungual tufts were reduced, and hypothenar pads were lost in the LCA of the Oryzomys palustris group + Lundomys + Holochilus + Pseudoryzomys , and in the LCA of Amphinectomys + Nectomys + Melanomys + Sigmodontomys . Ungual tufts were lost in the LCA of Lundomys + Holochilus and in the lineage leading to Nectomys apicalis . 12 Functional morphological research might shed light on the role (if any) of these features in semiaquatic locomotion.

HABITAT: Twenty-one oryzomyine clades, encompassing 61 % of all oryzomyine species, are found exclusively in forest environments, especially in ombrofilous forest. Five forest-dwelling clades ( Nectomys , Oecomys , angouya , nitidus , and megacephalus ) are also distributed in open vegetation biomes, such as the Cerrado and Llanos, but are mostly found in forest patches within open vegetation, such as gallery forests and ‘‘cerradão’’ ( Alho et al., 1986; Mares et al., 1986; Nitikman and Mares, 1987; Mares et al., 1989; Bonvicino et al., 1996, 1998; Talamoni and Dias, 1999; Lacher and Alho, 2001). Five other oryzomyine clades representing 10 % of oryzomyine species, however, are found only in open-vegetation biomes: Lundomys in the Pampas; Pseudoryzomys in

12 This is the only discrepancy between these characters, but other species of Nectomys also lack ungual tufts, and the primitive condition for the genus is probably absence of tufts. The condition in Amphinectomys is unknown, but if this taxon lacks ungual tufts, the pattern of concordance will also be perfect for this character.

the Cerrado, Caatinga, and Chaco; Zygodontomys in the Llanos and other northern South American and Central American savannas; and the xanthaeolus group of Oryzomys and Nesoryzomys in the dry coastal biomes of Western Peru, Ecuador, and the Galapagos Islands ( Patton and Hafner, 1983; Voss, 1991; Voss and Myers, 1991; Voss and Carleton, 1993; Dowler et al., 2000; Guabloche et al., 2002). Finally, species of five clades (28 % of oryzomyine species) are found in both forest and nonforest environments: Holochilus , Microryzomys , Oligoryzomys , palustris , and subflavus .

Forest-dwelling taxa occur in all oryzomyine lineages (fig. 46), whereas exclusively open-vegetation dwellers are found only in clades A and D. All members of clade B are exclusively sylvan taxa, while clades A, B, and D contain both types. Reconstruction of this feature in the recovered cladogram (fig. 46) indicates that the ancestral oryzomyine was a forest-dwelling taxon, and that invasions of open-vegetation environments occurred at least four times in oryzomyine evolution.

Three other significant modifications from the generalized oryzomyine bauplan are observed in restricted clades: the volelike appearance of Zygodontomys , Melanomys and Handleyomys species ( Allen, 1913; Voss, 1991; Voss et al., 2001); the spiny pelage of Scolomys and Neacomys ; and the hypsodonty of Holochilus molars (modified for intake of grass; Hershkovitz, 1955). Members of 11 lineages retained the primitive, unspecialized morphotype: S. aphrastus , O. polius , nitidus group, subflavus group, albigularis group, talamancae group, alfaroi group, chapmani group, melanotis group, megacephalus group, and yunganus group. As expected, all of these taxa, with the exception of S. aphrastus , are currently part of Oryzomys , highlighting, in a phylogenetic framework, the notion of the genus as a wastebasket group of unspecialized, mostly forest-dwelling, rats. As inferred by the reconstruction on the combined tree of size and the ecological adaptations discussed above, evolutionary niche shifts occurred de novo in different clades of these unspecialized oryzomyines. In turn, lineages that went through such evolutionary transitions are now among the most speciose oryzomyine clades assigned to genera: Oligoryzomys (16 spp.), Oecomys (15 spp.), Nectomys (8 spp.), and Neacomys (8 spp.). This indicates the conquest of different ecological roles in South American biomes that opened the possibilities for further diversification of oryzomyines.

BIOGEOGRAPHY

The evaluation of the recovered phylogeny using methods of biogeographical analysis (e.g., Wiley, 1987; Bremer, 1992; Ronquist, 1994; Bremer, 1995; Ronquist, 1997; Hausdorf, 1998; Ronquist, 1998) is beyond the scope of this study because many oryzomyines species were omitted, obscuring potentially important patterns found in the internal clades ( Ronquist, 1996). Nevertheless, the present phylogeny provides a heuristic framework for assessing previous oryzomyine biogeographic scenarios, as well as a basis for future inquiries in oryzomyine biogeography. In this section, I also summarize the current knowledge on distributional patterns, fossil record, and molecular dating for the tribe.

The first step in biogeographic analysis is delimitation of the distributional areas that will serve as discrete entities in the elaboration of workable hypotheses ( Humphries and Parenti, 1999). Most oryzomyine genera and species groups can be classified into three general distribution patterns as delineated below.

TRANS- ANDEAN DISTRIBUTION (fig. 47): This category encompasses taxa primarily distributed in lower-montane and lowland habitats west of the Andes, such as trans- Andean lowland rainforests ( Voss and Emmons, 1996) and the arid coastal region of Peru and Ecuador (including the Galapagos Islands). Nine oryzomyine taxa are found in such trans-Andean landscapes: Melanomys , Sigmodontomys , Nesoryzomys , and six groups of Oryzomys ( palustris , talamancae , melanotis , alfaroi , chapmani , and xanthaeolus ). The chapmani and melanotis groups are restricted to forests of Central America and Mexico ( Goldman, 1918), whereas Nesoryzomys occurs only in the Galapagos Islands ( Dowler et al., 2000), and the xanthaeolus group occurs in both the Galapagos Islands and in the Pacific littoral zone of Peru and Ecuador ( Patton and Hafner, 1983). The remaining trans-Andean taxa are found in forests from northwestern South America into Central America, or in the case of the palustris group, in the United States ( Allen, 1913; Goldman, 1918; Hall, 1981; Musser and Carleton, 1993; Carleton and Musser, 1995; Voss and Emmons, 1996; Musser et al., 1998; Sanchez-H. et al., 2001).

ANDEAN DISTRIBUTION (fig. 47): This category encompasses taxa found only in Andean habitats such as montane forest and Paramos, usually on both sides of the cordillera and above 1500–2000 m. Six taxa are found exclusively or primarily in the Andes mountains: Microryzomys , Handleyomys , and four groups of Oryzomys ( albigularis , balneator , hammondi , and polius ). Microryzomys and the albigularis group are widely distributed in the Andes, from central Bolivia to northern Venezuela ( Carleton and Musser, 1989; Patton et al., 1990; Musser and Carleton, 1993; Percequillo, 2003). The four remaining taxa have restricted range: Handleyomys is found in cloud forest of the Colombian Occidental and Central Cordilleras ( Voss et al., 2002); O. balneator occurs in lower montane forest in both west and east slopes of the Andes in Ecuador and Peru ( Musser et al., 1998); O. hammondi is known from a single locality from Ecuadorian western Andes ( Thomas, 1913; specimens from UMMZ); and O. polius is known from a few localities from the eastern Peruvian Andes ( Osgood, 1913; specimens from AMNH and FMNH).

CIS- ANDEAN DISTRIBUTION (fig. 47): This category encompasses taxa primarily distributed in lowland or lower montane biomes east of the Andes, such as Amazon, Atlantic and coastal Venezuelan rainforests, Llanos, Cerrado, Chaco, Pampas, and Caatinga. This category includes 15 taxa: Nectomys , Amphinectomys , Holochilus , Lundomys , Pseudoryzomys , Neacomys , Oecomys , Scolomys , and five groups of Oryzomys ( megacephalus , yunganus , nitidus , angouya , and subflavus ). Lundomys , Pseudoryzomys , and the angouya and subflavus groups occur only in southeast South America (central and eastern Brazil, Bolivia, Paraguay, Uruguay, and Argentina) ( Voss and Myers, 1991; Musser and Carleton, 1993; Voss and Carleton, 1993; Percequillo, 1998; Langguth and Bonvicino, 2002; Bonvicino, 2003). Holochilus , Nectomys , Oecomys , and the megacephalus and nitidus groups are widely distributed in the Amazon basin, eastern Andean piedmont, and southeast South America ( Hershkovitz, 1944, 1955, 1960; Musser et al., 1998; Percequillo, 1998; Weksler et al., 1999; Patton et al., 2000). Neacomys and the yunganus group occur throughout Amazonia ( Musser et al., 1998; Patton et al., 2000; Voss et al., 2001), whereas Amphinectomys and Scolomys are found only in western Amazonia ( Malygin et al., 1994; Patton and da Silva, 1995; Gómez-Laverde et al., 2004).

As expected in such simplifications, the geographic range of various taxa blurs the exact limits of the three categories. One species of the albigularis group, O. devius , is found in the highlands of Costa Rica and Panama ( Carleton and Musser, 1995; Percequillo, 2003). One species of Oecomys (Oe. bicolor ) and two species of Neacomys ( N. tenuipes and N. pictus ) are found in eastern Panama and western Colombia ( Hall, 1981; Musser and Carleton, 1993); another species of Oecomys (Oe. trinitatis ) reaches Costa Rica. Two species of Nectomys ( N. magdalenae and N. grandis ) are found in the inter- Andean valleys of the Magdalena and Cauca Rivers ( Hershkovitz, 1944; Bonvicino, in prep.). One species of the yunganus group ( Oryzomys tatei ) is restricted to the eastern Andean piedmont ( Musser et al., 1998) and could be considered an Andean taxon. Sigmodontomys and Melanomys occur in a few localities at the coastal Venezuelan forests east of the Andes, and Melanomys also occurs in some localities in the eastern Ecuadorian Andes piedmont ( Voss and Emmons, 1996). In each of these cases, however, taxa have restricted distributions outside their main geographic category, and I assume this is a result of secondary, recent range expansion.

In contrast, two taxa could not be effectively categorized into the patterns. Oligoryzomys is distributed from Mexico to Terra del Fuego, being found east, west, and at the Andes mountains ( Carleton and Musser, 1989; Patton et al., 1990; Carleton and Musser, 1995). Zygodontomys is widely distributed in open vegetation biomes of northern South America and eastern Central America, occurring east and west of the Andes (Voss, 1991). The species of these two genera are treated individually in the present analysis. Zygodontomys cherriei has trans- Andean distribution, whereas Z. brevicauda is distributed east of the Cordillera de Mérida (Voss, 1991). Among Oligoryzomys species , O. fulvescens is distributed from Mexico to the eastern Amazon basin in Brazil; remaining species are encompassed in the cis- Andean category, being restricted to southeastern South America.

The two major methodological categories currently in use in taxon biogeography (sensu Hovenkamp, 1997) can be divided in ‘‘projection-rule biogeography’’ and ‘‘vicariant biogeography’’ ( Seberg, 1988). These approaches are considered separately bellow.

VICARIANCE BIOGEOGRAPHY: No major vicariance scenario has been proposed for the biogeographic history of oryzomyines, or for any other sigmodontine tribe. The most obvious vicariance scenario for oryzomyines would be related to the Andean uplift, with the separation of trans- and cis-Andean lineages ( van der Hammen, 1974). Molecular clock estimates for the initial diversification of oryzomyines point to a period between 5 and 9 mybp. This time range is based on the bounds estimated for the diversification of all sigmodontine lineages except sigmodonts and ichthyomyines by Steppan et al. (2004) using four nuclear genes, as well as on estimates for the origin of the oryzomyine lineage by Engel et al. (1998) using 1340 bp of the mitochondrial genome. Smith and Patton (1999: fig. 10) also placed the origin of oryzomyines in this period in their analysis of cytochrome b sequences. This time boundary is also congruent with the time estimated by DNA- DNA hybridization for the divergence of Akodon and Oryzomys ( Catzeflis et al., 1993) . If these estimates are corroborated in future studies that use better calibration points and denser taxonomic sampling, then oryzomyines could have been affected by the final orogenic surge of the Andes between 3 and 5 mybp ( Irving, 1975; Simpson, 1979; Kellogg, 1984; Helmens and van der Hammen, 1994), a scenario proposed for other neotropical groups such as Heliconius butterflies ( Brower, 1996), sand flies ( Arrivillaga et al., 2002), tropidurine lizards ( Harvey and Gutberlet, 2000), and various bird groups ( Cracraft and Prum, 1988; Prum, 1988; Brumfield and Capparella, 1996).

The cis- and trans-Andean patterns, however, do not show correspondence to the major oryzomyine lineages (fig. 48). Instead, each major lineage displays all or most of the described biogeographic patterns. Clades B and D contain Andean, cis-Andean, and trans-Andean taxa, whereas clade C has Andean and cis-Andean taxa; clade A has cis-Andean and trans-Andean taxa. Even within each lineage, the distribution categories are not grouped into monophyletic units (fig. 48). In clade A, the two cis-Andean species are paraphyletic. In clade B, two unrelated taxa, Handleyomys and the albigularis group, have Andean distribution; three trans-Andean taxa ( O. chapmani , O. alfaroi , and O. rostratus ) are grouped, but a fourth taxon, O. talamancae , is not included in this clade; and the cis-Andean taxa form a paraphyletic grade. In clade C, the two Andean taxa ( Microryzomys and O. balneator ) do form a clade, but the cis-Andean taxa are observed within three clades: one formed by Neacomys and two subclades of Oligoryzomys . Finally, cis- and trans-Andean taxa of clade D are dispersed over several lineages. Only two clades in the present analysis follow a cis–trans-Andean sister-group pattern: the palustris (trans) and tetralophodont (cis) clade, and the Melanomys + Sigmodontomys (trans) and Amphinectomys + Nectomys (cis) clade. A third clade that follows this pattern is recovered with moderate support by the IRBP analysis, containing the talamancae (trans) and nitidus (cis) groups.

It is clear from the above patterns (or lack of thereof) that any analytical methodology within the vicariance framework will have to include several cases of over-Andean dispersal within oryzomyine phylogeny.

PROGRESSION RULE BIOGEOGRAPHY: The center of origin concept was once ousted from historical biogeography ( Croizat et al., 1974; Nelson and Platnick, 1981), but several procedures ( Bremer, 1992, Ronquist, 1994, 1997; Hausdorf, 1998) have tried to reinstate it in cladistic framework under the label ‘‘ancestral area methodologies’’ ( Ebach, 1999; Crisci, 2001). The basic premise of these methods is the progression rule of Hennig (1966), which Bremer (1994: 255– 256) rephrased as ‘‘(1) areas positionally more plesiomorphic (present on ‘deep’ branches) in a cladogram of a particular group are more likely parts of the ancestral area for that group than are positionally more apomorphic areas and (2) areas represented on numerous branches of the cladogram are more likely parts of the ancestral area than are areas represented on a few branches.’’

This methodological formulation fits well to biogeography of sigmodontines, as scenarios for the group have relied heavily on the center of origin concept, with a strong dispersalist element. Almost all classical sigmodontine biogeographic scenarios agree that oryzomyines, as all sigmodontines, are descendants from proto-sigmodontine ancestors that invaded (i.e., dispersed to) South America from North America in the late Cenozoic ( Simpson, 1950; Hershkovitz, 1966b, 1969, 1972; Patterson and Pascual, 1968; Patterson and Pascual, 1972; Savage, 1974; Baskin, 1978 1986; Marshall, 1979; Reig, 1980; Simpson, 1980; Jacobs and Lindsay, 1984; Reig, 1984; Slaughter and Ubelaker, 1984; Reig, 1986; Czaplewski, 1987; Baskin, 1989). The South American fossil record strongly corroborates this hypothesis, as sigmodontine rodents suddenly appear in the well-known Argentinean fossil sequence at the Pliocene (Montehermosan and Chapadmalalan; Pardiñas et al., 2002).

The major disagreements between these dispersalist scenarios are the time of the dispersion and the place of the initial diversification for the sigmodontine groups (i.e., the place of the ancestral area of the group). Scenarios range from an early arrival (early or middle Miocene, 15–24 mybp) of a primitive stock of muroids in South America by waif dispersal or island hopping across the Bolivar Trough, and further radiation in the South American continent (e.g., Hershkovitz, 1966b; Reig, 1980, 1984), to a relatively recent entrance (Early or Middle Pliocene 2.5–4 mybp) of an already diversified sigmodontine stock through the Panama Isthmus (e.g., Patterson and Pascual, 1968; Patterson and Pascual, 1972; Baskin, 1986; Baskin, 1989; Czaplewski, 1987).

The oryzomyine fossil record is extremely poor and cannot provide, by itself, any evidence for the timing or the place of such initial radiation. The earliest oryzomyines in the South American fossil record are from the Pleistocene ( Steppan, 1996; Pardiñas et al., 2002). Steppan (1996) estimated the age of Holochilus primigenus from Bolivia at between 0.7 and 1 mybp. Forms of Holochilus , Lundomys , Nectomys , and Oligoryzomys were retrieved from the Ensenadense (Early- Middle Pleistocene of Argentina; Pardiñas et al., 2002). All these taxa are placed at advanced, or apomorphic, positions in the tree, suggesting that diversification of the tribe must have occurred before the Pleistocene. Oryzomyines (and Sigmodontines in general) are absent from the middle Miocene (13 mybp) Honda group of La Venta, the richest Miocene tropical fossil deposit in South America ( Kay and Madden, 1997). Sampling for small vertebrates at this site has been exemplary, as indicated by the recovery of examples of small mammals groups, such as echimyid rodents ( Walton, 1997) and didelphid marsupials ( Goin, 1997), that are often captured alongside oryzomyines in present-day tropical forest biomes ( Voss and Emmons, 1996). Thus, based on this strong negative evidence, the upper bound for oryzomyine absence in South America is about 13 mybp.

The North American record also does not provide clues for timing and place of oryzomyine diversification. Previous suggestions that Oryzomys is present in the Pliocene or even Miocene in North America ( Jacobs and Lindsay, 1984; Baskin, 1986, 1989) are unfounded. Oryzomys ? pliocaenicus from the Miocene (Hemphilian) of Kansas ( Hibbard, 1939) is ‘‘generically indeterminate, but may represent Bensonomys ’’ ( Baskin, 1986: 295; see also Hershkovitz, 1966b: 737).? Oryzomys from the Miocene (Hemphillian) of Oregon ( Shotwell, 1970) is more closely related to ‘‘ Peromyscus ’’ pliocenicus, putatively an ancestral to the fossil neotomine † Repomys (see May, 1981; Baskin, 1986). ‘‘ Oryzomys ’’ from the Early Pliocene (Blancan) of New Mexico (May in Jacobs and Lindsay, 1984; Repenning and May, 1986) is ‘‘close dentally to † Jacobsomys and † Symmetrodontomys and might belong to the genus † Jacobsomys ’’ ( Czaplewski, 1987: 194). † Bensonomys and † Symmetrodontomys were placed by Mc- Kenna and Bell (1997, following Musser, in lit.) in the Peromyscini tribe, whereas † Jacobsomys was tentatively assigned to sigmodontines sensu lato (i.e., sensu Carleton and Musser, 1984) as Sigmodontinae incertae sedis. The earliest confirmed oryzomyines present in the North American fossil record are referred to Oryzomys fossilis (5 Oryz- omys palustris ), from the Pleistocene of Florida, Georgia, Kansas, and Texas ( Hibbard and Taylor, 1960; Dalquest, 1962; Hibbard, 1963; Webb, 1974; Kurten and Anderson, 1980; Webb and Wilkins, 1984; Hulbert and Pratt, 1998). All fossil localities of O. palustris with a more precise dating are from middle (Rancholabrean; 0.3 mybp) and late (Sangamonian; 0.13 mybp) Pleistocene deposits ( Webb, 1974; Webb and Wilkins, 1984).

Fossils from other sigmodontine tribes, in turn, are present in both North and South American Pliocene. Fossils of Phyllotini and Akodontini are present in South America in the Montehermosan (4–5 mybp) and Chapadmalalan (3.5–4 mybp), respectively ( Pardiñas et al., 2002), whereas fossils of † Prosigmodon and Sigmodon are present in North America in the Hemphillian (6.8 mybp) and Blancan (3.3 mybp), respectively ( Martin, 1979; Czaplewski, 1987; Korth, 1994). Thus, fossil data alone do not provide much evidence for the ancestral area of oryzomyines.

Application of the progression rule to the current phylogeny suggests a South American ancestral area for oryzomyines. The most basal taxon for each lineage is cis-Andean or, in case of clade D, eastern Andean (fig. 48): Scolomys , nitidus group, Neacomys , and Oryzomys polius for clades A, B, C, and D, respectively. Central American (trans-Andean) taxa are always recovered deeply nested within the major lineages.

Given the estimated time for initial diversification of oryzomyines based on molecular studies, the oryzomyine ancestor must have arrived in South America prior to the formation of the Panamanian land bridge at 3.5–4.0 mybp ( Coates et al., 1992; Ibaraki, 1997) by over-water dispersal (but see Coates et al., 2004, for new evidence pointing to a collision of the Central American arc with South America at 7.1 mybp). The capability of oryzomyines to undertake long-distance water dispersal is well known ( Carleton and Olson, 1999). Oryzomyines are known from volcanic islands without former subaerial connections to the South American plate, such as the Galapagos Islands ( Nesoryzomys and Oryzomys bauri ) and Fernando de Noronha († Noronhomys ). These islands are situated at 1000 km and 310 km, respectively, from South America, a distance far exceeding the one among the stepping-stone connections that existed between the North and South American plates since the Miocene ( Donnelly, 1992).

Thus, there is no phylogenetic indication that the immediate precursor of oryzomyines was in Central America as previously proposed ( Patterson and Pascual, 1968, 1972; Baskin, 1978, 1986; Marshall, 1979; Simpson, 1980; Jacobs and Lindsay, 1984; Czaplewski, 1987; Baskin, 1989; Engel et al., 1998). Such a hypothesis would involve several ad hoc events of dispersal and extinction. The principal argument used for a North American diversification of oryzomyines was the supposed presence of members of the tribe (among other sigmodontines) in the North American Tertiary fossil record, that is, before the formation of the Panamanian land bridge. As discussed above, no undisputed fossil oryzomyine is known from the Tertiary of North America.

Current oryzomyine diversity in Central and North America is more likely a product of independent colonizations made by the different clades within oryzomyines that are currently found in that area ( Zygodontomys , alfaroi - melanotis - chapmani , talamancae , Oligoryzomys , palustris , Melanomys , and Sigmodontomys ). This scenario is similar to the one proposed by Hershkovitz (1966b), who suggested nonsynchronous dispersion of various oryzomyine lineages from northern South America in a timeframe ranging from before the completion of the Panamanian bridge to recent times. The palustris group was included by Hershkovitz in his ‘‘Stratum III: Old South American Migrants in North American’’, which ‘‘returned over water routes to Middle America [during the Pliocene] and differentiated significantly [in situ]’’ (Hershkovitz, 1966: 733). Melanomys , Sigmodontomys , talamancae , alfaroi , Zygodontomys , and Oligoryzomys were considered members of ‘‘Stratum IV: Late South American Migrants in Middle America’’, which ‘‘spread over Panamanian land-bridge into Middle America [during Pleistocene]; [with] low grade subspeciation’’ (Hershkovitz, 1966: 737). Oryzomys melanotis and Sigmodontomys aphrastus were of doubtful position. Hershkovitz (1966) also included in the recent migrants group some taxa that were considered here as with borderline distribution in Central America, such as albigularis , Oecomys , and Neacomys .

A slightly different scenario is the concomitant range expansion of all the independent oryzomyine lineages into Central America after the completion of the Panamanian land bridge, without over-water dispersal events in the Pliocene. Such invasion would be second only to the Didelphidae dispersal into Central America as part of the ‘‘Great American Interchange’’ in terms of number of genera, and would be the most successful in terms of species diversity. Much of the oryzomyine radiation is restricted to tropical North America, with only one taxon extending into the Nearctic region, a pattern also observed in other groups of South American invaders such as Didelphidae , Echimyidae , and Xenarthra ( Webb and Wilkins, 1984). Some of the invading oryzomyine groups have experienced larger in situ diversification, such as the alfaroi - melanotis - chapmani clade, the palustris group, and Oligoryzomys ( Hershkovitz, 1966b; Hall, 1981; Carleton and Musser and 1995), whereas others appear to be in early stages of dispersion into Central America, especially those cis- Andean taxa with minimal distribution in Central America, such as Neacomys and Oecomys .

A more precise location of the oryzomyine ancestral area in South America requires delimitation of smaller units for analysis than the three general categories provided here. Two areas are likely candidates: the region of premontane forests of the northern Andes and the western Amazon lowland forests. All basal taxa within each lineage— Neacomys , polius , hammondi , the nitidus group, and Scolomys —are distributed in the western Amazon and/or submontane Andes. The few biogeographical analyses done with oryzomyine genera or species groups also placed western Amazon/eastern Andes taxa as basal to their own lineages ( Costa, 2003). In contrast, taxa restricted to other South American landscapes, such as southeastern South America ( Lundomys , Pseudoryzomys , subflavus and angouya groups), are situated farther from the oryzomyine root in the present phylogeny and in lower level analyses (Patton et al., 2001; Costa, 2003).