Details of the methods used to collect these data were provided in our previous publications (Ishibashi & Saitoh, 2008a, 2008b; Ishibashi et al., 1998; Ishibashi et al., 1995). Because a few initial errors in parentage have been corrected in the present database, the descriptions provided in Ishibashi and Saitoh (2008b) are the most useful for understanding the procedure used to establish parentage. An outline of our methods follows. On 29 September 1992, 47 individuals (22 males and 25 females) from seven natural gray-sided vole populations were introduced into a 3-ha (200 m × 150 m) outdoor enclosure established in a secondary forest in Sapporo, Japan (42°59’03”N 141°23’14”E; Figure 1). These founders were chosen to maximize heterozygosity at three microsatellite loci. Until 27 April 1994, we conducted 3-day capture–mark–recapture surveys every two weeks using live traps set at 300 trap stations in a 20 × 15-grid pattern with 10-m spacing, except during periods of deep snow cover. Traps were set at 0900 h and checked at 1400 h and 2000 h every day; all traps were closed after the check at 2000 h. Through these surveys, we continuously monitored the location, weight, and reproductive status of each individual in the enclosure. Complementary trapping was also performed within the home ranges of breeding females to mark juveniles as early as possible (~20 days after birth). On first capture, each individual was marked by toe clipping for subsequent identification, and the clipped toes were used as DNA samples. For all individuals born within the enclosure, both parents were determined based on female reproduction history (i.e., timing of pregnancy and lactation), home range location, and genotypes at 3–5 microsatellite loci. After identifying parents, we comprehensively estimated dates of birth for all litters based on female reproduction history, the palpation of pregnant females on capture, and first capture dates and weights of juveniles. The day after the final capture was designated the estimated death date, unless the exact date of death was known. The potential predators 4 within the enclosure included snakes (Elaphe spp.), Ural owls (Strix urelensis), sables (Martes zibellina), red foxes (Vulpes vulpes), and cats (Felis catus). Parentage was genetically established for all captured voles, except for one immigrant and two litters derived from sisters with identical genotypes at five microsatellite loci. For these two litters, the mothers of the offspring (N = 8) were determined based on their weights and trapping locations at first capture. The male was considered an intruder from outside the enclosure because it had unique microsatellite allele combinations compared with all possible candidate mothers. No mismatch alleles were observed among all combinations of assigned parents and offspring (N = 920). The first and last litters from which juveniles were successfully weaned during the study period were estimated to have been produced on 19 October 1992 and 15 October 1993, respectively. Individuals born within the enclosure (N = 920) were derived from 215 litters, among which 51 litters had multiple paternity. The male intruder (M23032/DNA# 901) was excluded from the parentage estimation dataset (vole_data.csv) because it did not sire any offspring in the enclosure. Our previous study (Ishibashi & Saitoh, 2008b) verified our paternity estimation results using the CERVUS v3.0 program (Kalinowski et al., 2007). The CERVUS analysis supported our paternity assignment for almost all offspring, i.e., the assigned father had the highest score among all candidates. However, in 2.7% of all offspring (25/920), the males assigned as fathers were found to have positive but lower scores according to CERVUS, indicating that they were not the most likely fathers for these offspring. For the 25 inconsistent cases, an additional genotyping of five to six microsatellite loci was performed. The results revealed that in 23 of these cases, the most likely father, as designated by CERVUS, showed allele mismatches with the offspring, therefore confirming 99.8% (918/920) of our initial paternity estimates were correct. However, in both of the remaining cases, the offspring showed allele mismatches with the previously assigned fathers. Consequently, we adopted the CERVUS analysis results for these two offspring, which showed no mismatched alleles at any of the 10 loci examined. Thus, CERVUS analysis allowed us to verify and improve the accuracy of our parentage estimates.