Fig. 3. Admixture graph model summarizing key findings.

Maximum |Z-score| = 2.6 for a difference between observed and expected f-statistics (|Z| = 2.7 restricting the analysis to transversions). The model presented fits only after adding small proportions of deeply diverging ancestry into PuntaSantaAna_6600BP and Kawéskar_800BP (splitting before the radiation of Native Americans), which we hypothesize reflects not real ancestry but rather technical artifacts due to these samples being shotgun sequenced and not UDG treated, causing them to be attracted to the outgroup (without modeling these edges, shown in Supplementary Fig. 6, the maximum |Z-score| is 5.1, but this drops to 3.3 with only transversions). Dashed lines indicate admixture between two different lineages with percentages being the admixture proportions. Numbers on solid lines are genetic drift with units of F_ST × 1000. Z-scores were determined from standard errors obtained from jackknife resampling.