We use a Bayesian skyline plot to visually illustrate changes in Native American female effective population size (Ne
) over time. Bayesian skyline plots assume a single migration event, which makes the approach ideal for questions concerning the peopling of the Americas since it is generally agreed that there was a single migration 
. Our new skyline plot () strongly supports a large population expansion (~1.8 orders of magnitude, or 80-fold) that occurred ~16–12 thousand years ago (kya). This timing suggests an entry to the New World that was coincident with the retreat of the North American ice sheets, i.e. the opening of an ‘ice-free corridor’ ~17–14 kya 
. Immediately before this expansion, there is a small drop in effective population size ~17–16 kya (this is an insignificant change, as judged by the overlap in 95% confidence intervals at the beginning and end of the population decrease), possibly corresponding to a population bottleneck prior to entry to the Americas. Before 17 kya, the skyline plot is flat with no evidence of the early (~40 kya) population expansion we reported previously 
. The absence of an early expansion signal in the skyline plot may simply indicate that divergence of proto-Amerinds from the Asian gene pool was not accompanied by significant population growth. These results are highly consistent with our earlier analysis of only 20 Native American mitochondrial coding genomes 
, in support of theoretical expectations by Felsenstein 
that increasing sample size is an inefficient way to improve the accuracy of maximum likelihood estimations from coalescent analyses of population genetic data.
Bayesian skyline plot of 148 Native American mitochondrial coding genome sequences.
Our new analysis (with non-Native American sequences eliminated and more Native American sequences added) shows a larger population increase (80-fold vs 16-fold) over a smaller period of time (16–12 kya vs 16–9 kya) relative to our previous analysis that inadvertently included non-Native American sequences 
. The non-Native American sequences likely introduced additional variation that artificially increased Ne
prior to the expansion. Thus, we can estimate a new Ne
for the New World founding population of 1,800 (this number is multiplied by two since the skyline plot only estimates the female effective population size). This number is closer to our previous isolation-with-migration (IM)-based estimate of 1,200 
and thus reduces the variation around our estimate of the size of the founding population to ~1,000–2,000 effective individuals.
Prior to entry to the New World, we propose a period of isolation. A valid question remains - How long was the period of isolation? In the absence of a biphasic skyline plot, we can calculate first approximations of the time necessary to generate the defining variants for the New World mithochondrial founding haplogroups. All New World mitochondrial sequences cluster in five monophyletic clades, representing founding haplogroups that are differentiated from non-New World haplogroups by the presence of specific, defining genetic variants. The variants that occur on the branch leading to each New World founding haplogroup represent variation that evolved prior to expansion into the Americas whereas variation within each founding haplogroup, i.e. nucleotide diversity within a haplogroup, represents variation that evolved after entry to the Americas – we are interested in the variation that occurred prior to entry into the Americas. There is strong consensus on the number of New World founding haplogroup-defining variants, including both coding and non-coding hypervariable regions I and II (HVRI+II) variants 
. However, there is a wide range of substitution rates that have been estimated for both coding and non-coding variants 
. Fagundes et al. 
tend to favor the slower substitution rates whereas we generally favor the faster substitution rates, particularly for coding variants since a faster rate (~1.7×10−8
substitutions/site/year) has been confirmed using two independent approaches 
. However, to be complete since there is ongoing debate about the correct calculation of substitution rates most recently 
, we present a series of estimates based on coding and HVRI+II variants using both fast and slow substitution rates (). As is evident from the calculations, there is a wide range of estimates for the time necessary to generate the New World defining variants, i.e. averages range from ~6,000 to ~25,000 years. By averaging across coding and non-coding variants and including fast and slow substitution rates, we report a range of ~7–15 thousand years. This estimate suggests that Amerind ancestors may have experienced a period of isolation lasting at least 7–15 thousand years prior to their expansion into the Americas (see the blue box in ).
Estimates of time necessary to generate the mitochondrial genome variants that define New World founding haplogroups.