<div>Inferring genetic population structure in endangered, highly migratory species such as sea turtles is a necessary but difficult task in order to design conservation and management plans. Genetically discrete populations are not obvious in highly migratory species, yet require unique conservation planning due to unique spatial and behavioral life-history characteristics. Population structure may be inferred using slowly evolving mitochondrial DNA (mtDNA), but some populations may have diverged recently and are difficult to detect using mtDNA. In these cases, rapidly evolving nuclear microsatellites may better elucidate population structuring. Bayesian inference and ordination may be useful for assigning individuals to inferred populations when populations are unknown. It is important to carefully examine population inference results to detect hierarchical population structuring, and to use multiple, mathematically diverse methods when inferring and describing population structure from genetic data. Here I use Bayesian inference, ordination, and multiple genetic analyses to investigate population structure in Olive ridley sea turtles (ORs; Lepidochelys olivacea) nesting in northwestern Costa Rica (NWCR) and across the entire Eastern Tropical Pacific (ETP). Mitochondrial DNA did not show structure within NWCR, and existing data from prior studies are not appropriately published to compare NWCR to Mexican ORs. In NWCR, Bayesian inference suggested one population, but ordination suggested four moderately structured populations with high internal relatedness, and moderate to high levels of connectivity. In the ETP, Bayesian inference suggested a Mexican and Central American population, but hierarchical analysis revealed a third subpopulation within Mexico. Ordination revealed nine cryptic clusters across the ETP that primarily corresponded to Mexican and Central American populations but contained individuals from both populations, some from other, distant nesting sites. The subpopulation within Mexico was well-defined after ordination, and all clusters displayed high 10 internal relatedness and moderate genetic differentiation. Bottlenecks were detected in both putative populations, at seven Mexican and two Central American nesting beaches, and in six out of nine inferred clusters, including three out of four Mexican clusters. Bottleneck events may have played some role in cluster differentiation. Migration was significant from Mexico to Central America at multiple levels, but did not necessarily agree with potential migrants elucidated by ordination. Migration was generally lower between ordination-inferred clusters than between nesting sites or Bayesian-inferred clusters. Phylogenetic trees generally supported structuring by ordination, rather than by Bayesian inference. Structuring in ordination not tied to bottleneck events could be due to mating behaviors or patterns of nesting beach colonization dictated by environmental features. In this study, ordination provided a more practical and nuanced framework for defining MUs and DIPs in ETP ORs than did STRUCTURE. This may be due to hierarchical structuring within ETP ORs that may be present in other sea turtle populations and species. In the case of ETP ORs, hierarchical structure may be an artefact of recent population bottlenecks and subsequent recolonization of nesting beaches, or due to mating at foraging grounds or along migratory routes. Bayesian inference may not be the best method for population inference in highly migratory species such as sea turtles, which have a high potential for broad scale genetic connectivity, and therefore may display hierarchical population structuring not easily related to nesting sites. Future studies, and perhaps published studies, should incorporate Bayesian inference and ordination, as well as other measures of population divergence and descriptive statistics, when searching for population structure in highly migratory species such as sea turtles.</div>