An Early Oligocene age for the oldest known monkeys and rodents of South America

Significance

The Paleogene dispersals of monkeys and rodents from Africa to South America permanently and profoundly altered the composition of mammalian communities in the Neotropics, but the timing of these colonization events remains uncertain. Through a combination of geochronological analyses (detrital zircon dating of the Santa Rosa fossil locality in eastern Perú) and phylogenetic biochronological analyses (tip-dating age estimation of caviomorph rodents), we demonstrate that the oldest known primates and rodents of South America are unlikely to be older than Early Oligocene in age and are not Eocene in age as previously postulated. There are no grounds for rejecting the possibility of an earlier arrival for either clade, but at present the fossil record provides no evidence for earlier dispersals.

Abstract

The Santa Rosa fossil locality in eastern Perú produced the first Paleogene vertebrate fauna from the Amazon Basin, including the oldest known monkeys from South America. This diverse paleofauna was originally assigned an Eocene age based largely on the stage of evolution of the site’s caviomorph rodents and marsupials. Here, we present detrital zircon dates that indicate that the maximum composite age of Santa Rosa is 29.6 ± 0.08 Ma (Lower Oligocene), although several zircons from Santa Rosa date to the Upper Oligocene. The first appearance datum for Caviomorpha in South America is purported to be the CTA-27 site in the Contamana region of Perú, which is hypothesized to be ∼41 Ma (Middle Eocene) in age. However, the presence of the same caviomorph species and/or genera at both CTA-27 and at Santa Rosa is now difficult to reconcile with a >11-My age difference. To further test the Middle Eocene age estimate for CTA-27, we ran multiple Bayesian tip-dating analyses of Caviomorpha, treating the ages of all Paleogene species from Perú as unknown. These analyses produced mean age estimates for Santa Rosa that closely approximate the maximum 29.6 ± 0.08 Ma composite date provided by detrital zircons, but predict that CTA-27 is much younger than currently thought (∼30 Ma). We conclude that the ∼41 Ma age proposed for CTA-27 is incorrect, and that there are currently no compelling Eocene records of either rodents or primates in the known fossil record of South America.

The Santa Rosa local fauna, derived from an isolated riverbank exposure in the Departamento de Ucayali, Perú (Fig. 1), was the first Paleogene vertebrate paleofauna discovered in the low-latitude lowland Neotropics of Amazonia (1). This paleofauna has proven to be extremely rich in terms of numbers of specimens and their diversity. Fish and crocodilian teeth occur in the thousands, whereas amphibians, turtles, lacertilians, birds, bats, notoungulates, xenarthrans, and anthropoid primates form minor components. The most abundant, speciose, and best-studied mammalian groups are the marsupials and caviomorph rodents. Though anthropoids are rare in the Santa Rosa paleofauna, they are of outsized importance, as they potentially represent the oldest records of that clade in South America, and their inferred age would constrain scenarios for the timing of the trans-Atlantic dispersal of stem platyrrhines from Africa (2, 3).

Fig. 1.

Map of Perú, showing the locations of the Santa Rosa, Shapaja, and Contamana areas.

Three major factors have hindered studies of the Santa Rosan vertebrates. First, it is mostly a microvertebrate paleofauna comprising primarily isolated teeth, the majority of which are only a few millimeters in size. Moreover, there are few mammalian maxillary or mandibular fragments with more than one tooth, which makes associating teeth of different loci within a species’ tooth row difficult. Second, all of the identified mammalian taxa present have been new to science at the species level, most at the generic level, many at the family level, and some possibly even at the ordinal level. This makes biogeographic and biostratigraphic correlation difficult, if not impossible, in most cases. Third, there have been no means to date the paleofauna accurately using stratigraphy. The fossil site (Fig. 2) is located in a remote, isolated corner of eastern Perú where access is solely by chartered aircraft and there has been little to no modern geologic work. The regional stratigraphy is poorly exposed and known only rudimentarily.

Fig. 2.

Wide-angle and close-up views of the Santa Rosa outcrop. (Upper) The entire outcrop, showing the isolated upstream exposures (#1 and #2) of the fossil-bearing conglomeratic horizons. The primary portion of the outcrop extends from #3 to #11, and it is ∼100 m in length, as shown here, but it is underwater during the rainy season. All of the labeled conglomeratic horizons in this outcrop produce fossils. (Middle and Lower) The upstream end of the conglomeratic horizons end abruptly at a fault, immediately to the left of #3 to #5. The numbered spots indicate various conglomeratic horizons that were sampled individually. The black arrows indicate fault lines. Displacement in this view is slight, but observe left of spot #7, above the arrow. The conglomeratic horizons are thickest at the upstream end (#3 to #5) and taper to disappearing downstream of #11. Zircon samples were collected from #2, #4, #7, and near #10. The bench at the top of the exposed sediment marks the contact, the Ucayali Unconformity, between the Paleogene deposits and the Mio-Pliocene Madre de Dios Formation.

The Santa Rosa local fauna was tentatively assigned an Eocene age (1). This age assignment was based on the perceived stage of evolution of the components of the large caviomorph rodent assemblage (4) and the wide variety of primitive marsupials present in the paleofauna (5). Confidence in the Eocene age was reinforced by the discovery of an anthropoid primate, Perupithecus ucayaliensis, the single known tooth of which (an upper first molar) is very similar to that of the Late Eocene genus Talahpithecus from Libya (6). A Late Eocene or Early Oligocene age for the paleofauna was proposed based on study of the Santa Rosa cingulate xenarthrans (7), whereas Shockey et al. (8) suggested that the site’s notoungulates supported an “imprecise” late Paleogene age of Early to Late Oligocene. More recently, the surprising discovery of a parapithecid monkey of African origin (Ucayalipithecus perdita) at Santa Rosa (3) provided a Lower Oligocene estimate for the site (31.7 Ma), derived from Bayesian tip-dating (BTD) analysis.

Several years after the discovery of the Santa Rosa locality and the first descriptions of Paleogene vertebrates from lowland Amazonia (1), another international team began paleontological work in the region south of the town of Contamana (07°19’ S, 74°56’ W), specifically along the Quebrada Cachiyacu, a right bank tributary of the Ucayali River, where Kummel (9) had described a stratigraphic sequence extending from the Cretaceous to the Plio-Pleistocene. This work resulted in the identification of a number of Paleogene sites, which have been estimated to range in age from Middle Eocene to Upper Oligocene (10–15). If this dating scheme is correct, some of the Contamana sites have produced the oldest known caviomorph rodents from South America and constrain the caviomorph dispersal from Africa to have occurred by the late Middle Eocene (∼41 Ma). However, a number of the vertebrate taxa reported from these sites appear to be closely related to those that were previously identified in the Santa Rosa local fauna, particularly various species of caviomorph rodents (16) and genera of marsupials.

The Shapaja area of Perú, northwest of Contamana in the Departamento de San Martín, has also yielded remains of Paleogene caviomorphs and, unlike the oldest Contamana localities, fragmentary dental remains of anthropoids as well (17, 18). No geochronological data are available from the Shapaja localities (17). Boivin et al. (18) argued that the Shapaja localities are close in age to, but are likely younger than, Santa Rosa (p. 134). More recently, Antoine et al. (17) presented chemostratigraphies of the Shapaja sections based on δ¹³C analyses of dispersed organic matter and carbonate nodules. They found positive δ¹³C excursions in the section which they correlated with the Oligocene oxygen isotope event 1 (Oi-1) and Oi-1a events of the earliest Oligocene and used this evidence to argue that their oldest caviomorph-bearing locality (TAR-20) is Upper Eocene in age, the next-oldest (TAR-21) is close in age to the Eocene–Oligocene boundary, and all others are earliest Oligocene in age.

The confusing mix of conflicting biochronological signals from the Paleogene mammal-bearing sites of Perú would ideally be resolved through robust geochronological analyses. Here, we present data from detrital zircon U-Pb geochronology that indicate that the maximum age of the Santa Rosa site is Lower Oligocene (29.6 ± 0.08 Ma), with the youngest grain from the locality dating to 25 ± 1.67 Ma (Upper Oligocene). This evidence forces a reanalysis of the rodent-bearing localities of Contamana and Shapaja, several of which have strong biochronological ties to Santa Rosa but are purported to be Eocene in age. In the absence of reliable geochronological data from these sites, we provide an independent, unbiased, and objective test of the proposed ages of the Contamana and Shapaja caviomorph-bearing sites by employing BTD analyses (19–21) of an unmodified caviomorph morphological matrix (22) that was originally compiled by several of the researchers who have argued in favor of Eocene age estimates for some of the Contamana and Shapaja caviomorph-bearing sites. The results of these BTD analyses suggest that there are no known caviomorphs of Eocene age in the fossil record of South America, and that all of the caviomorph-bearing sites of Contamana and Shapaja are likely to be either Lower Oligocene in age or younger.

Results

Laser Ablation Inductively Coupled Plasma Mass Spectrometry (LA-ICP-MS) Detrital Zircon Analyses.

Results of all zircon analyses are provided in SI Appendix, Tables S1–S4 and Cenozoic grains are summarized in Fig. 3. Of the 440 zircon grains analyzed, out of 110 zircon grains analyzed per sample, 114 date to the Cenozoic. Only 2 grains date from the Paleocene, and only 20 date from the Eocene. Of the other Cenozoic-aged zircon grains, 25 date to the Upper Oligocene, and 67 date to the Lower Oligocene.

Fig. 3.

Detrital zircon geochronology of the Santa Rosa deposit. (A–D) Kernel density estimates of the Cenozoic-aged detrital zircons from four sample sites at Santa Rosa, Perú, illustrate the concentration of the zircons to the Oligocene (Bandwidth = 1; Bin width = 1). For the complete dataset for the entire spectrum of zircons from the four sites, see SI Appendix, Tables S1–S4. The number of Cenozoic-aged zircons (out of 110 zircons analyzed per sample) is given for each plot. Note that of the 440 zircons analyzed, only 20 date from the Eocene and only 2 from the Paleocene. Of the other Oligocene-aged zircons, 25 date to the Upper Oligocene and 67 date to the Lower Oligocene. (E) Plot of zircon ages using IsoplotR.

Following the law of detrital zircon chronology (i.e., a sedimentary rock cannot be older than the youngest zircon grain it contains) the maximum depositional age for the Santa Rosa deposits is 25 ± 1.67 Ma, or Upper Oligocene, based on youngest single grain (YSG) analyses of the detrital zircons from the four samples analyzed (specifically from spot 4, see Fig. 2). However, as noted by Herriott et al. (23), detrital zircon analyses are always complicated by uncertainty, with YSG analyses often producing results indicating a younger age than results derived from analyses of the youngest statistical populations (YSP). To derive the age of the YSP, we used the program IsoplotR (24), which resulted in a composite age for the four samples of 29.6 ± 0.08 Ma (Fig. 3). Results for the YSP analyses for each of the four samples using IsoplotR are presented in SI Appendix, Figs. S3–S6.

The results from both types of analyses, YSG and YSP, falsify the postulated Eocene age assigned to the Santa Rosa local fauna by Campbell et al. (1). Given that there are no significant statistical differences among the four samples analyzed, we prefer to recognize the YSP age of 29.6 ± 0.08 Ma, or Lower Oligocene, as the best maximum age for the Santa Rosa deposit.

Ages of the Paleogene Caviomorph-Bearing Sites of Perú, as Inferred from BTD Analysis.

BTD analyses of the full caviomorph matrix of Marivaux et al. (22), with all species from the Santa Rosa, Contamana (CTA), and Shapaja (TAR) localities assigned a broad post-Paleocene (56 to 0 My; i.e., effectively “unknown”) age prior, returned mean ages of 29.5 Ma for Santa Rosa (n = 7 species), 30.3 Ma for CTA-27 (n = 4 species), 27.0 Ma for CTA-29 (n = 1 species), 28.7 Ma for the Shapaja localities (n = 6 species), 26.0 Ma for CTA-61 (n = 3 species), and 21.8 Ma for CTA-32 (n = 4 species) (SI Appendix, Fig. S7). The 29.5 Ma BTD estimate for Santa Rosa is remarkably close to both the 29.6 ± 0.08 Ma YSP estimate from detrital zircons and to the mean 30.25 Ma estimate derived for the two Santa Rosa anthropoids that were included in an entirely independent BTD analysis of a primate matrix (3). Our analyses correctly predicted that Pozomys ucayaliensis from CTA-29 is younger (27.0 Ma) than the rodents from the underlying level CTA-27 (mean of 30.3 Ma) and provide an independent age estimate for CTA-29 that is ∼16.5 My younger than that provided by “biotite” grains from the same layer (presumably reworked, see Discussion) that yielded a ⁴⁰Ar/³⁹Ar date of 43.44 ± 2.5 Ma.

The ranges of mean estimates provided by our various BTD analyses (Table 1 and SI Appendix, Tables S5–S9) are largely in agreement with age estimates that had previously been proposed for the Shapaja and younger Contamana localities. The 29.1 to 28.3 Ma estimates for the Shapaja localities are consistent with the Lower Oligocene age favored by Boivin et al. (18), who argued that the sites were likely to be close in age to Santa Rosa but perhaps slightly younger. The 26.3 to 23.8 Ma (Upper Oligocene) age estimates for the younger Contamana site CTA-61 are consistent with the 26.56 ± 0.07 Ma date from a detrital zircon (at level CTA-08SA) that was collected 4 m below the fossil-bearing layer. The 22.7 to 20.8 Ma estimates for CTA-32 (i.e., Lower Miocene, but potentially very close to the Oligocene–Miocene boundary of 23.03 Ma) are younger than the Upper Oligocene age proposed by Boivin et al. (14), but they are also the estimates with the most uncertainty. Only the BTD estimates of 31.0 to 29.8 Ma and 27.4 to 26.4 Ma for the older Contamana localities (CTA-27 and CTA-29, respectively) are strongly at odds with the Middle Eocene ages previously proposed for these sites (10, 11).

Table 1.

Age estimates for early caviomorph-bearing localities of Perú based on BTD analysis of the morphological character matrix of Marivaux et al. (22)

Locality	Ages of all Peruvian localities unconstrained (“unknown”)		Age of Santa Rosa constrained to ∼29.6 Ma, other Peruvian localities “unknown”		All Peruvian localities further constrained	Previous age estimate	Revised age estimate
	Mean, full dataset	Mean, reduced dataset	Mean, full dataset	Mean, reduced dataset	Mean, full dataset
CTA-32	21.79 Ma	20.81 Ma	22.46 Ma	22.05 Ma	22.74 Ma	Upper Oligocene	Upper Oligocene or Earliest Miocene
	(n = 4)	(n = 2)	(n = 4)	(n = 2)	(n = 4)
CTA-61	26.02 Ma	24.29 Ma	26.33 Ma	24.79 Ma	23.77 Ma	Upper Oligocene	Upper Oligocene
	(n = 3)	(n = 2)	(n = 3)	(n = 2)	(n = 3)
Shapaja	28.71 Ma	28.26 Ma	29.0 Ma	29.13 Ma	28.58 Ma	Upper Eocene to Lower Oligocene	Lower to Upper Oligocene
	(n = 6)	(n = 6)	(n = 6)	(n = 6)	(n = 6)
Santa Rosa	29.47 Ma	29.26 Ma	—	—	28.16 Ma	Upper Eocene or Lower Oligocene	Lower to Upper Oligocene
	(n = 7)	(n = 7)			(n = 7)
CTA-27	30.27 Ma	30.23 Ma	30.86 Ma	30.98 Ma	29.78 Ma	Late Middle Eocene	Lower Oligocene
	(n = 4)	(n = 4)	(n = 4)	(n = 4)	(n = 4)

Note that the revised age estimates for Shapaja and Santa Rosa (Lower to Upper Oligocene) take into account the evidence from chemostratigraphy (Shapaja) and detrital zircon geochronology (Santa Rosa), which suggest the possibility of an Upper Oligocene age.

We estimated the marginal likelihoods of the two competing hypotheses for the age of CTA-27 and CTA-29 (Middle Eocene versus Lower Oligocene) by running two stepping-stone analyses in MrBayes 3.2.7 (25): one in which CTA-27 and CTA-29 were fixed to the youngest bound (40.9 Ma) of the age range favored by Antoine et al. (10), and one in which it was fixed to the mean age recovered by our BTD analysis of the full dataset (30.3 Ma). In both analyses, Santa Rosa, Shapaja, and the younger Contamana localities were fixed to the mean age estimates from our BTD analysis of the full dataset. The resulting estimated marginal likelihoods for the Middle Eocene (−10,005.28) and Lower Oligocene (−9,994.64) hypotheses for the age of CTA-27 and CTA-29 differ by 10.64 units, and the resulting 2 × log_e Bayes factor test statistic of 21.28 indicates “very strong” (26) evidence in favor of rejecting the Middle Eocene estimate.

Discussion

Questioning the Middle Eocene Age Estimate for the Earliest Caviomorph Rodents.

The first rodent-bearing Paleogene faunas from Perú that were potentially older than that of Santa Rosa were recovered from several localities along the Quebrada Cachiyacu near the town of Contamana. These sites produced a number of isolated rodent teeth and an assortment of fragmentary vertebrate fossils (10, 11). The first sites to be reported were interpreted to be Middle Eocene in age (10) based on three criteria: ⁴⁰Ar/³⁹Ar dates on biotites, mammalian biochronology, and an associated palynoflora “characteristic of the Middle Eocene to Early Oligocene.” Each of these justifications requires reassessment in light of the geochronological evidence from Santa Rosa and the results of our BTD analyses, which support a much younger (Lower Oligocene) age for the oldest Contamana localities (CTA-27 and CTA-29).

First, site CTA-29—which, importantly, yielded the caviomorph species Pozomys ucayaliensis (13) that has also been identified at Santa Rosa (16)—was dated to 43.44 ± 2.5 Ma based on ⁴⁰Ar/³⁹Ar step heating of single biotite grains (n = 3) derived from a “tuffaceous silt located 47 m above the fossiliferous level” comprising CTA-27 (10). However, in defiance of the Law of Superposition (i.e., in undeformed stratigraphic sequences, the oldest strata will always be at the bottom of a sequence), and without explanation, locality CTA-27 was assigned an age of 41.6 to 40.94, or late Middle Eocene. We infer that the authors were possibly taking into account the large SD about the mean (±2.5 My) in the ⁴⁰Ar/³⁹Ar date to make the argument that the underlying stratum was younger. There is no evidence that overturning of the beds exposed along the Quebrada Cachiyacu has taken place, and on their depiction of the Cachiyacu Anticline of Kummel (9) [Maquia Anticline of Antoine et al. (10), their figure S5], the authors did not indicate any overturned beds.

The biotite grains from CTA-29 that provided the 43.44 ± 2.5 Ma ⁴⁰Ar/³⁹Ar date were derived from a “reddish-brown tuffaceous silt” (10)—which, by definition, means they were derived from a silt, not a tuff. In recognizing this, the silt must have been a water-deposited sediment and not a consolidated volcanic ash, or tuff, and therefore the probability that the silt and its contained biotite grains comprised reworked sediments sourced from older deposits is quite high (by comparison, note for instance that there are 19 reworked detrital zircons of Middle and Late Eocene age at Santa Rosa; SI Appendix, Tables S1–S4). Antoine et al. (10) cited Kummel (9) as describing the tuffaceous silt they dated as a tuff. However, Kummel (ref. 10, p. 1258) mentioned only one ash and no tuffaceous silts in the Tertiary portion of the section along the Quebrada Cachiyacu; that ash was a 2-m-thick, massive, nodular, light-gray to greenish buff-colored tuff in the Chambira Formation. This deposit was located 294 m below the contact with the Ipururo Formation and 255 m above the contact with the Yahuarango Formation (9). Therefore, the biotites from the reddish-brown silt dated by Antoine et al. (10), which they cite as coming from the top of the Yahuarango Formation, could not have come from the ash Kummel (9) described. Later, Boivin et al. (14) described the sediments at CTA-27 and CTA-29 as scattered, inframetric, channelized “sands” and sandstone lenses deposited within a low-energy small stream, as opposed to a tuffaceous silt (10). Located in a foreland basin receiving sediments from the Andes, it can safely be assumed that any sands containing biotites were reworked many times before being deposited in the lowlands of the Contamana region. Any radiometric dates derived from reworked biotites do not necessarily, and in all probability do not, reflect the age of deposition of the sandstone lenses.

Antoine et al. (ref. 10, electronic supplementary material) correlated CTA-29 with CTA-08SA, the latter a deposit comprising “a coarse altered tuffaceous sand with a mixture of quartz, feldspars, biotite, and white micas, and a carbonate matrix” located on the opposite, or southwest, side of the Cachiyacu Anticline from CTA-29. Antoine et al. (11) later described both CTA-29 and CTA-08SA as “tuffs,” a clear departure from their earlier descriptions; no estimate for the percentage of tuffaceous material present in the sands of the sites was given. Regarding the age of CTA-08SA, Antoine et al. (10) reported “no consistent data” from attempts to date biotites from that site.

However, of importance for the dating of the stratigraphy, Antoine et al. (11) later reported ID-TIMS U/Pb (²⁰⁶Pb/²³⁸U) dates on five detrital zircon grains from CTA-08SA. Two of the zircon grains produced ages older than the Cenozoic (264.4 Ma and 903.3 Ma); these are significant only in indicating that the deposit is detrital in origin and definitely not a tuff. The other three zircon grains yielded dates of 34.91 ± 0.20, 30.04 ± 0.08, and 26.56 ± 0.07 Ma, or latest Upper Eocene to Lower Oligocene. By themselves, these three dates do not comprise a sample sufficiently large enough to be treated statistically as a population, nor do they indicate the age of the deposit, but they do suggest a possible maximum depositional age of the deposit. It is therefore striking that the three youngest zircon dates all fall within the range of dates of zircon grains from Santa Rosa, which do form a significant population (Fig. 3).

Antoine et al. (11) described undated zircon grains from CTA-08SA as generally subrounded and having appearances suggesting that they are mostly detrital in origin, as opposed to a more pristine appearance that would occur in grains from a tuff. The combination of sediments described for CTA-08SA by Antoine et al. (10) are again consistent with a detrital origin for the deposit, and the authors unequivocally state that the five analyzed zircon grains are detrital in origin, which conflicts with the description of their having come from a tuff. Based on the dates from the zircon grains, Antoine et al. (11) retracted the correlation of CTA-29 with CTA-08SA proposed by Antoine et al. (10). Instead, they proposed a correlation of CTA-29 on the NE flank of the Cachiyacu Anticline with CTA-67 on the SW flank of the Cachiyacu Anticline, the latter being lower in the stratigraphic section than CTA-08SA. CTA-67 was described as comprising a tuff, but no mention was made of attempts to secure a radiometric date from the site. A late Middle Eocene age was assigned to CTA-67 based on unspecified “all available dating proxies,” presumably meaning miscellaneous fossil content.

Similarly, Boivin et al. (14) reported locality CTA-61 on the SW flank of the anticline as dating to the Upper Oligocene by radioisotopy when its age was apparently extrapolated from mammalian biochronology and from the dates for detrital zircons derived from CTA-08SA, a deposit stratigraphically lower in the section but on the same flank of the anticline.

Unfortunately, it will probably never be possible to secure an accurate date for CTA-29 and CTA-27 because Boivin et al. (14) reported that both localities, among others along the Quebrada Cachiyacu, were irremediably lost because of landslides, erosion, or silting caused by flooding events.

Regarding the mammalian biochronology used by Antoine et al. (10) to support a late Middle Eocene age for CTA-27, it is notable that, at the time of their publication, no vertebrate fossil from CTA-27 was affirmatively assigned to a previously known species. Subsequently, however, one of the caviomorph species from CTA-27 (Cachiyacuy kummeli) was identified at Santa Rosa, as was Pozomys ucayaliensis (16), which was first described from the Contamana region at the stratigraphically higher CTA-29 level that yielded the 43.44 ± 2.5 Ma ⁴⁰Ar/³⁹Ar date. The late Middle Eocene age (i.e., 41.6 to 40.94 Ma) estimated for the mammalian paleofauna from CTA-27 by Antoine et al. (10) appears to have been initially based primarily on their interpretation of the stage of evolution of fragmentary vertebrate specimens, and it might have been significantly different had the Cachiyacuy kummeli and Pozomys ucayaliensis specimens from Santa Rosa been known at the time. Similarly, further justification provided by Antoine et al. (10) for CTA-27 being older than Santa Rosa was their interpretation that the caviomorph rodents from CTA-27 were remarkably small compared with those of Santa Rosa (the assumption being that the smaller the individuals comprising a species, the older the species). However, Arnal et al. (16) have since documented that this was not true; some specimens from Santa Rosa are actually as small as, or smaller than, the smallest rodents from Contamana. It is now clear that all of the genera and two of the species recognized from Contamana also occurred in Santa Rosa, thereby demonstrating considerable taxonomic overlap, and thus a similarity in ages, of the sites (16). Boivin et al. (14) acknowledged these similarities but took the position that they probably reflected biogeographic affinities more than contemporaneity.

Regarding the third line of evidence used to support a Middle Eocene age for sites such as CTA-27 (i.e., a continental palynoflora), there is, in fact, no well-dated palynological zonation for western lowland Amazonia, a vast region with many distinct subregions. Although it is possible to develop a zonation for a given site without a well-dated profile from a region near the profile site, if not actually at the site, interpreting an accurate age for the palynological zones is not possible. In the case of the Contamana sites, Antoine et al. (10) correlated what they perceived the pollen zonation of the Contamana sites to be with that of the Llanos of Colombia, thousands of kilometers to the north (27). Furthermore, of the 83 palynomorphs identified, 52 were represented by only a single specimen, only 15 were identified to species, and an unspecified number were considered “novel.” Nonetheless, the age inferred by Antoine et al. (10) from the pollen assemblage ranged from 44.8 (Middle Eocene) to 30.9 Ma (Early Oligocene), a span of 13.9 My.

Age of the Shapaja Localities.

Our BTD results are also inconsistent with the latest Eocene and earliest Oligocene ages that Antoine et al. (17) have proposed for the Shapaja localities. Though our analyses do not correctly predict the sequence of the Shapaja localities, the mean estimates for individual species nevertheless range from 31.6 to 20.7 Ma, with composite means for the section ranging from 29.1 to 28.3 Ma. The composite mean estimates for the Shapaja section are all very close to those for Santa Rosa, differing by only 0.42 to 1.0 My across analyses with both broad (effectively “unknown”) and more restricted age priors. These results are consistent with the close correlation of the Shapaja and Santa Rosa sections originally proposed by Boivin et al. (18).

Antoine et al. (17) note that their placement of the Shapaja localities near the Eocene–Oligocene boundary implies that these faunas include 14 new first appearances in South America (either at the genus, family, or superfamily level), with most of the taxa in question otherwise first known from Deseadan (Upper Oligocene) localities. With the geochronological evidence now available from Santa Rosa, and the biochronological evidence for a close correlation of the Shapaja and Santa Rosa sections, we propose that a more parsimonious explanation for the occurrence of otherwise Deseadan taxa in the Shapaja localities is that those localities are, in fact, Deseadan in age. Given such an age estimate, we suggest that the two positive δ¹³C excursions identified in the Shapaja chemostratigraphies would be better correlated with those that occur at ∼26.5 and ∼25.5 Ma in the generalized global record (ref. 28, their figure 28.11). We further propose that the positive δ¹³C values identified in the lowest parts of the Shapaja chemostratigraphies more likely represent a zone dating to ∼31 to 30 Ma (magnetochrons Cn10-11) in the generalized global record (28). The Upper Oligocene ages that our proposed chemostratigraphic correlation requires for the stratigraphically higher Shapaja localities are younger than the composite estimates provided by our BTD analyses. However, the same can be said of the multiple zircons of Upper Oligocene age at Santa Rosa (one as young as 25 ± 1.67 Ma) that are younger than that site’s composite mean estimate derived from our BTD analyses. Surprisingly, given the much older age estimates that have been proposed for both Santa Rosa and Shapaja, it may actually be the case that these localities are not only Deseadan in age but also date to the latter half of the Deseadan, and our BTD analyses are overestimating their ages.

Boivin et al. (29) recently described new rodents from the Balsayacu area of the Departamento de San Martín that they interpreted as being intermediate in age between the CTA-27 locality (here considered to be ∼31 to 30 Ma) and those of Santa Rosa and Shapaja (here considered to be younger than ∼30 Ma and possibly Upper Oligocene in age). Given all of the evidence presented here, we anticipate that geochronological and/or phylogenetic biochronological evidence will ultimately place the Balsayacu localities (TAR‐55/TAR‐55bis, TAR‐76, and TAR‐77) in the Oligocene as well.

Implications for the Timing of Anthropoid and Caviomorph Dispersals to South America.

Given available evidence, we see no compelling justification for the assignment of a Middle, or even an Upper, Eocene age for the oldest rodent-bearing localities of Contamana and Shapaja. However, this conclusion should not be taken to mean that caviomorphs were not present in South America during the Eocene. In fact, a Middle Eocene origin for crown Caviomorpha is supported by our BTD analyses of Marivaux et al.’s (22) matrix, which provided mean estimates for the origin of that clade that extend back to 40.2 to 39.4 Ma (i.e., late Middle Eocene, Barrancan SALMA). These estimates are remarkably congruent with the independent molecular estimate of 39.9 Ma that Upham and Patterson (30) recovered for crown Caviomorpha when CTA-27 was excluded as a calibration point in their analyses. These results predict that caviomorphs should eventually be found in the currently very meager Late Eocene fossil record of South America.

There are no such ancient constraints on the timing of the primate dispersals from Africa to South America, as recent molecular estimates for the origin of crown Platyrrhini generally favor a Late Oligocene or Early Miocene origin of that clade (31–33). Given the absence of anthropoid fossils from CTA-27 and CTA-29, and uncertainty about the precise age of the Santa Rosa and Shapaja anthropoid fossils, we can only state that we see no evidence for this clade having been present in South America prior to ∼30 Ma. However, we also have no reason to rule out an earlier dispersal, for instance near the Eocene–Oligocene boundary (3).

Materials and Methods

Collection of Detrital Zircons.

In a brief reconnaissance trip to the Santa Rosa locality in 2016 to determine if the site still existed after an 18-y collecting hiatus, we sampled ∼160 kg of matrix for microvertebrate processing. The matrix came from 11 sites spaced across the face of the ∼100-m-long outcrop. Samples of the fine-grained (i.e., less than 1.0 mm) portion of the matrix from two sites (#2 and near #10) at opposite ends of the outcrop (Fig. 2) were submitted for detrital zircon analysis via LA-ICP-MS. However, as these samples were not collected with detrital zircon analysis in mind, in a subsequent (2019) field season, we carefully collected two additional samples from the center of the outcrop (Fig. 2; sites #4 and #7). We submitted the latter samples directly to the laboratory, thereby ensuring that there was no contamination during processing. As the objective of the detrital zircon dating was to determine the maximum age for the deposit, we targeted for dating those zircon crystals that looked to be youngest.

LA-ICP-MS Detrital Zircon Analyses.

The LA-ICP-MS methods followed to separate zircons from sample material, and to then produce and process the zircon U/Pb data, are similar to those presented in Bradley et al. (34), Hults et al. (35), and Moore et al. (36). GeoSep Services performed all zircon laboratory and analytical procedures. Zircons were separated from the four samples collected from the outcrop using a combination of both standard separation procedures along with specific customized procedures described by Donelick et al. (37). This particular process 1) enhances the recovery of all possible grain sizes while minimizing the potential loss of smaller grains through the use of water table devices and 2) ensures the preservation of complete grains by minimizing grain breakage and/or fracturing associated with traditional methods of isolating individual grains from solid rock samples or, in this case, highly consolidated sediment.

Isotopic analyses were performed with a New Wave UP-213 laser ablation system in conjunction with an Agilent 7700× quadrapole inductively coupled plasma mass spectrometer (LA-ICP-MS) in the GeoAnalytical Lab at Washington State University. For all laser analyses, the beam diameter was 20 μm and the frequency was set at 5 Hz, yielding ablation pits ∼10 to 12 μm deep. He and Ar gas were used to deliver the ablated material into the plasma source. Each analysis of 32 cycles took ∼30 s to complete and consisted of a 6-s integration on peaks with the laser shutter closed to secure background measurements, followed by a 24-s integration with the shutter open and the laser ablating zircon material. A 25-s delay occurred between analyses. The isotopes measured included 202 Hg, 204 (Hg + Pb), 206 Pb, 207 Pb, 208 Pb, 232 Th, 235 U, and 238 U. Additional details about the LA-ICP-MS methods appear in the SI Appendix.

BTD Analyses.

Tip-dating analyses employed the morphological character matrix of Marivaux et al. (22) and were run in MrBayes v. 3.2.7 (25). No modifications were made to the matrix, and the same character ordering scheme was used. The clockratepr prior was calculated using an R script first published in Gunnell et al. (38), which used as input the “allcompat” consensus tree derived from a nonclock Bayesian analysis and the means of the upper and lower bounds of each uniform age prior; extant taxa were fixed at 0 Ma. So as not to bias the calculation of the clockratepr prior, we used the Middle Eocene estimate favored by Antoine et al. (10) for species from CTA-27 and CTA-29, including Cachiyacuy kummeli and Pozomys ucayaliensis. For two of the tip-dating analyses reported here, the Contamana, Shapaja, and Santa Rosa localities were assigned a broad post-Paleocene (56 to 0 My) age prior, whereas two other analyses employed the YSP age of 29.6 ± 0.08 Ma for Santa Rosa. For each set of age priors, we ran analyses with 1) only the taxa that Marivaux et al. included in their parsimony analysis and 2) all of the caviomorph species in their matrix, including species that were excluded from their parsimony analysis (i.e., Cephalomys arcidens, Cephalomys bolivianus, Litodontomys chubutomys, Loretomys minutus, Plesiosteiromys newelli, Soriamys gaimanesis, and Ucayalimys crassidens). We ran an additional analysis of the full dataset with more constrained age priors, including a prior for Santa Rosa that disallowed ages older than 30 Ma, but allowed ages as young as the Oligocene–Miocene boundary (SI Appendix, Table S9 for details); this analysis yielded much more constrained 95% highest posterior density (HPD) intervals for each site than those calculated with the broad 56 to 0 My age prior for the Peruvian localities. Analyses were run for 50 million generations, with the first 25% discarded as burn-in. Resulting minimum effective sample sizes for all parameters were >750, and the average SDs of split frequencies in the final generations of each analysis were <0.009. Mean tip ages and 95% HPD for the taxa of interest were retrieved using FigTree (39). Stepping-stone analyses were also run in MrBayes for 35 million generations. Other details of the analyses can be found in the original input and output files that have been made available in the Dryad Digital Data Repository (http://datadryad.org).

Data Availability

Phylogenetic data matrices and results have been deposited in Dryad Digital Data Repository (https://doi.org/10.5061/dryad.9s4mw6mg8) (40). Previously published data were used for this work (https://doi.org/10.1098/rspb.2019.2806).