Introduction

Testudo costarricensis was described and named in a lengthy article by ^{Segura (1944}) based on a single, ~80% complete fossil shell of a continental tortoise (family Testudinidae), purportedly from the Oligocene or lower Miocene of Costa Rica. It was widely and repeatedly cited as a Miocene Central American tortoise (e.g., ^{Auffenberg, 1971}; Coto and Acuña, 1986; ^{Alvarado, 1994}; ^{Cadena et al., 2012}), and became the basis of paleobiogeographic interpretations of the origin of South American continental tortoises (e.g., Auffenberg, 1971; ^{Coto and Acuña, 1986}). However, careful re-examination of the holotype shell of T. costarricensis in the collection of the Laboratorio de Paleontología de la Escuela Centroamericana de Geología, leads us to conclude that this specimen is a fossil of Oligopherus laticunea collected from the Eocene-Oligocene White River Group of the western United States.

History

The article that named “Testudo” costarricensis, ^{Segura (1944}), was originally published for a scientific and literary competition that was held by the Society of Natural Sciences and Pharmacy of the Republic of Guatemala. Apparently, the invitation to write the article came in 1942. Segura’s (1944) description of a new tortoise species, including three low resolution photographs of the holotype, won the competition, and Segura received a gold medal from the President of Guatemala, with the following inscription: “Empowers Natural Sciences and Pharmacy, Guatemala Scientific Contest, Anniversary 1918- 1943, 18 set. 1943 Honor of Merit Prof. Alfonso Segura Paguaga.”

Segura’s manuscript was published in the monthly journal of La Escuela de Farmacia (Journal of the School of Pharmacy) in Guatemala in 1944. Segura had visited the Smithsonian Institution and the Wyoming desert (USA) for paleontological training between October 1941 and September 1942 (^{Alvarado, 1989}, p. 279; Alvarado, 1994, p. 88-91), and later he sent a draft of the description and drawings in the article to Smithsonian paleontologist Charles W. Gilmore, who reviewed and accepted the proposed new taxon. The Costa Rican scientists Rómulo Valerio Rodriguez, Jorge León Arguedas, Vitalia Sáenz Bejarano, José Antonio Zavaleta, Eloisa de Segura and Surana Segura Paguaga also reviewed the manuscript.

According to ^{Segura (1944}; also see ^{Alvarado, 1994}), the turtle fossil was found by Adán Arce near Peralta de Limón in 1930. Subsequently, in December 1942, Mr. Ramón Ulloa (father and son) and Mr. Hugo Guzmán supposedly encouraged Mr. Arce to donate the fossil to the National Agricultural School in San José. Dr. César Dóndoli, head of the Section of Geology of the National Department of Agriculture (and considered the father of Costa Rican geology), then reportedly provided the turtle fossil to Segura for study.

The type description and locality of “Testudo” costarricensis were mentioned without critical review in several papers (e. g., ^{Auffenberg, 1971}, ¹⁹⁷⁴; Coto and Acuña, 1986; ^{Alvarado, 1994}; ^{Cadena et al., 2012}; ^{Lucas et al., 2007}; ^{Lucas, 2014}; ^{Lucas and Alvarado, 2017}). Alvarado (1989, 1994), Lucas et al. (2007), Lucas (2014) and Lucas and Alvarado (2017) emphasized the historical significance of this find and the role of Alfonso Segura Paguaga (1913-2002) as the first native born paleontologist in Costa Rica.

^{Auffenberg (1971}) and Coto and Acuña (1986) reclassified Segura’s taxon as Geochelone costarricensis. Later, Coto y Pritchard (verbal communication cited by ^{Alvarado, 1994}) suggested that the fossil might represent a new genus, possibly of late Eocene age. ^{Loveridge and Williams (1957}), based on the erroneous reconstruction of Segura (1944), mentioned “Geochelone costarricensis” as an example of a tortoise species with the humeral pectoral sulcus crossing the entoplastron, but make no further reference to it.

^{Segura’s (1944}) original identification and description of the tortoise was used by ^{Auffenberg (1971}) to support his hypothesis of the origins of South American and Antillean tortoises. Thus, Auffenberg considered “Geochelone” costarricensis not to belong to the genus Chelonoidis, but possibly to be a transitional form between Chelonoidis and basal tortoises that retained some features associated with Indotestudo. Unfortunately, these links were based on the incorrect reconstruction of the sutures and sulci by Segura (1944) as will be shown below (compare Figure 1A-C with D-F).

Fig.1: Geochelone costarricensis A-C, original reconstruction from ^{Segura (1944}); D-F, revised reconstructions. G-H, reconstruction of Oligopherus laticunea based on ^{Hutchison (1996}) A, D, G, dorsal view; B, E, lateral view; C, F, H, ventral view. Abreviations: Anal, anal scute; Abd, Abdominal scute; fem, Femoral scute; Epi, epiplastron; Ent, entoplastron; G, gular scute; Hum, humeral scute; Hipo, hypoplastron; Hyo, hyoplastron; L.C., Left Costal; V., neural; Pect, pectoral scute; Py , pygal; Spy, suprapygal; E.C., right costal; Xic, xiphiplastron. 

Locality

According to ^{Segura (1944}), the turtle fossil was found at a place called Mile 52, near Peralta de Limón in east-central Costa Rica, and he stated that the holotype shell of Testudo costarricensis was found loose on Eocene marine sediments and below Quaternary volcanic deposits. Thus, he concluded that an Oligocene or perhaps early Miocene age of the tortoise fossil is possible. When ^{Alvarado (1994}) interviewed Segura in 1987, Segura affirmed that the fossil was found in 1930, although in his original paper of 1944 he never mentioned the year of the discovery.

According to the geological map of ^{Madrigal (1985}), in the vicinity of Peralta, marine shale, sandstone and calcarenites of likely Oligocene age are exposed and overlain by Quaternary volcanic rocks. However, ^{Fernández (1987}) assigned these marine strata an older, Paleocene-middle Eocene age and identified them as turbidites rich in volcanic debris deposited on the proximal continental slope. There are no reports near Peralta of continental sediments of Paleogene or Miocene age; the oldest continental strata there are of Pliocene and Quaternary age. A visit to the area by one of us (GEA) found no rocks consistent with the matrix associated with the holotype of Testudo costarricensis. The volcaniclastic rocks in the surrounding area are Quaternary lahar and debris flow deposits, as well as alluvial and colluvial deposits composed by rounded to angular fragments. These rocks do not resemble the sandstone matrix associated with Testudo costarricensis (Fig. 2).

Fig. 2: A-C, Photographs of the holotype of Testudo costarricensis; and D, close-up of ventral view of the right epiplastron showing putative sulcus continuing onto the matrix A, ventral view; B, dorsal view; C, lateral view. 

The matrix inside the shell of the holotype of Testudo costarricensis is a medium-grained, clastic (not volcaniclastic) sandstone, which does not resemble the rocks that are exposed in the vicinity of Peralta de Limón. Therefore, the published locality and stratigraphic horizon of T. costarricensis seem unlikely, if not impossible, as the source of a Oligo-Miocene tortoise fossil.

Description and taxonomy

The holotype of Testudo costarricensis (CF 2509) (Fig. 1-2) is a small (216 mm carapace length), moderately domed tortoise shell. Its preserved portion of the carapace includes the complete neurals, costals, nuchal, pygal and suprapygals as well as the anterior two and posterior four peripherals. The plastron is less complete, missing the medial portion between the two lobes and the posterior most portion of the anterior lobe. The fossil was evidently damaged sometime between the taking of the photograph published in ^{Alvarado (1994}, fig. 11) and the present, so that part of the plastron is now missing (Fig. 2). The remainder of the shell appears unchanged since the photographs of ^{Segura (1944}).

The nuchal is broad and pentagonal, with a roughly square cervical scute. This sutures to the first neural at the center of the first vertebral scute. The neural formula is 6-6-4-8-46-6-6, with the first neural enlarged, and the shorter sides of the anterior hexagonal neurals (n1-n2) facing toward the posterior end and toward the anterior end in the posterior three neurals (n6-n8). Neurals 3 and 5 are roughly square and are crossed by the anterior and posterior sulci of the third vertebral, respectively. Neural 1 sutures to the nuchal, neural 2, and costals 1 and 2. Neural 2 sutures to neurals 1 and 3 as well as costals 2 and 3. Neural 3 sutures to costal 3. Neural 4 is octagonal and sutures to costals 3, 4, and 5. Neural 5 contacts costal 5. Neural 6 contacts costals 5 and 6. Neural 7 contacts costals 6 and 7. Neural 8 contacts costals 7 and 8 as well as the anterior suprapygal. Neural 7 is divided roughly in half by the anterior sulcus of the 5th vertebral scute.

The costals are strongly wedged, as in all crown tortoises. The first costals contact the nuchal, at least two peripherals and the anterior lateral edges of neural 1. Costal 2 is expanded laterally and has the pleural 1-pleural 2 sulcus dividing it roughly in half. Costal 3 is constricted laterally, particularly lateral to the vertebral-pleural sulcus. Costal 4 is expanded laterally and divided roughly in half by the marginal 2-marginal 3 sulcus. Costal 5 is constricted laterally, and its proximal end is divided roughly in half by the vertebral 3-vertebral 4 sulcus. Costal 6 is expanded laterally and divided roughly in half along most of its width by the marginal 4-marginal 5 sulcus. Costal 7 expands laterally, suturing to the anterior edge of the suprapygal. The insertion of the ribs into the peripherals is unclear. The costo-peripheral suture coincides with the pleural-marginal sulcus, as in most tortoises.

Peripheral 1 contacts costal 1, the nuchal and peripheral 2. The contacts of costal 2 are unclear beyond the anterior edge, which contacts costal 1 and peripheral 1. The fourth peripheral from the posterior end contacts costals 4 through 6. The third peripheral from the posterior end contacts costals 6 and 7. The second peripheral from the rear contacts costals 7 and 8. The posteriormost peripheral contacts costal 8 and the anterior suprapygal. The pygal is small, only contacting the posterior suprapygal and the posteriormost pair of peripherals.

The epiplastra are about half the length of the anterior lobe of the plastron. They are significantly eroded on the left side, removing much of the surface of the bone. There is no significant anterior gular projection, as seen in many male tortoises. The entoplastron is diamond shaped, with the posterior point slightly abbreviated. The more anterior points join the epiplastron posterior sutures and the midline suture anteriorly. The entoplastron is overlapped by the gular scutes at its anterior extremity. The hyoplastron and hypoplastron are largely missing, preserving no evidence of their suture. The abdominal scutes extend into the posterior lobe of the plastron, just anterior to the suture with the xiphiplastron. Posterior to this, the posterior lobe is divided into two-thirds femoral and one-third anal scutes by two concave, posteriorly-oriented sulci. Two-thirds of the femoral scutes overlie the xiphiplastron and the rest overlie the hypoplastron.

Testudo costarricensis should be considered a junior synonym of Oligopherus laticunea (^{Hutchison, 1996}) based on their osteological similarities, including a similar neural formula, the restriction of costal 1 to only contact neural 1, and the lack of anterior elongation of costal 1. Furthermore, the shallow concavity of the sulci outlining the femoral and anal scutes in T. costarricensis is significantly different than is seen in Stylemys (^{Hay, 1908}; Hutchison, 1996). The posterior emargination of the plastron is deeper than in Stylemys. As stated by ^{Auffenberg (1971}), T. costarricensis lacks the derived traits of Chelonoidis, most importantly the absence of a cervical scute. Chelonoidis includes all South American and Caribbean tortoises. All of this together gives G. costarricensis distinct North American affinities. T. costarricensis is essentially identical to Gopherus laticunea (Hay, 1908; Hutchison, 1996), and we conclude that these taxa are synonymous.

Previous phylogenetic analysis

Coto and Acuña (1986), in their phylogenetic analysis of the taxonomic placement of Geochelone costarricenesis, differentiated this taxon from other tortoises based on a matrix of 18 characters. These characters, when present in T. costarricensis, are also present in Oligopherus laticunea:

The results of Coto and Acuña (1986) appear to have been biased by their limited taxon inclusion and unclear characters, such as character 4 (the depression of the nuchal) that vary in degree of development, but are coded by them as a simple presence/absence. In fact, the nuchal depression of Testudo costarricensis curves at this point to a degree very similar to turtles of the genera Oligopherus and Gopherus. Given the many similarities and the lack of any clear differences, we thus consider Testudo costarricensis to be a synonym of Oligopherus laticunea.

Discussion

Based on the above, we conclude that the holotype of Testudo costarricensis is a specimen of continental Oligopherus laticunea, a common tortoise in the Eocene-Oligocene White River Group of the western USA, and that its published provenance is false. We add to this two observations: (1) the matrix of the holotype of T. costarricensis has been varnished in a manner similar to that commonly done by commercial collectors of White River Group fossils; and (2) the preservation of the specimen, as a three-dimensional, nearly complete carapace and plastron filled with matrix, is also a characteristic preservation of White River Group tortoise fossils. So, how did a White River Group tortoise make its way to Costa Rica to be described by ^{Segura (1944}) as a new species of tortoise?

Segura received a scholarship to study at the National Museum of Natural History of the Smithsonian Institution from October 1941 to September 1942, under the tutelage of Wetmore, and with the collaboration of C. W. Gilmore, C. L. Gazin and G. Sternberg. During this time, Segura participated in fossil-collecting trips to Wyoming (Segura, 1944; ^{Alvarado, 1994}, and letters in the archives of Alvarado). Segura returned from the USA to Costa Rica in September 1942, and the turtle was donated to Dóndoli in December 1942, when Adán Arce presumably found the fossil. Thus, it seems most likely that Testudo costarricensis was a scientific hoax or an imported fossil that a non-Costa Rican person lost or gave as a gift to the local people along the railroad to the Caribbean Sea. The fact that this supposed Costa Rican tortoise remained in the scientific literature for more than 70 years can only be attributed to the fact that nobody critically re-examined the holotype prior to our research.

Conclusions