Mitochondrial Eve: measurements and quantitative checks

Howell, Neil et al.  “How rapidly does the human mitochondrial genome evolve?”  American Journal of Human Genetics 59(3):501–509.  1996.  Page 501:  “triplasmic descendants with three mitochondrial genotypes”.  Both mutations produced these descendants, which are heteroplasmic (carry more than one mitochondrial type), not fixed.  A pedigree study that opened the pedigree-rate literature.  It sequenced the D-loop (control region) of 45 members of a large matrilineal Leber hereditary optic neuropathy pedigree and found two germline mutations in 88 transmissions, normalised here as 2/88 over 1,122 bp of control region.  The article uses it as one row of its table of pedigree rates. ↩ ↩

Parsons, Thomas J. et al.  “A high observed substitution rate in the human mitochondrial DNA control region.”  Nature Genetics 15(4):363–368.  Apr 1997.  A genetic study that sequenced 610 base pairs of two hypervariable segments (HV1 and HV2) of mitochondrial DNA directly in 327 mother-to-child transmissions, not by RFLP analysis as one online rebuttal says.  It found a substitution rate of 2.5 per site per million years, about 20 times the rate predicted by phylogenetic analyses.  The article uses it because its authors worked out the age that this fast rate gives, about 6,500 years, and called it clearly incompatible with the known age of modern humans. ↩ ↩

Cavelier, Lucia et al.  “MtDNA substitution rate and segregation of heteroplasmy in coding and noncoding regions.”  Human Genetics 107(1):45–50.  2000.  The study reports:  “No homoplasmic nucleotide substitutions were detected in a total of 292 transmissions.”  The forgotten null result, and the first study to count fixed changes specifically.  The article uses it as the no-fixed-changes row of its table of pedigree rates. ↩ ↩

Sigurðardóttir, Sigrún et al.  “The mutation rate in the human mtDNA control region.”  American Journal of Human Genetics 66(5):1599–1609.  2000.  A pedigree study of the control region of mtDNA in 272 people from 26 large Icelandic family trees, covering 705 mother-to-child transmissions, which found 3 substitutions.  Howell 2003 re-scored the same dataset as 5 by counting newly arising heteroplasmies as well, which shows the arising-versus-fixing distinction inside a single dataset.  The article uses it as one row of its table of pedigree rates. ↩ ↩

Heyer, Evelyne et al.  “Phylogenetic and familial estimates of mitochondrial substitution rates: study of control region mutations in deep-rooting pedigrees.”  American Journal of Human Genetics 69(5):1113–1126.  2001.  The authors write that hypermutable sites.  “are expected to be of no or little informativity for the evaluation of mutational distances at the phylogenetic time scale.”  A pedigree study of control region mutations in deep-rooting French-Canadian family trees, which found 4 substitutions in 508 maternal transmissions over 673 bp.  Its more important result is the pooled one: of 18 substitutions in 1,729 transmissions, only 12 were at independent sites, because three positions had each mutated twice.  The article uses it as the direct empirical rebuttal to any argument that extrapolates a control-region pedigree count backwards. ↩ ↩

Forster, Lucy et al.  “Natural radioactivity and human mitochondrial DNA mutations.”  PNAS 99(21):13950–13954.  2002.  A pedigree study of 248 families in Kerala, covering 579 bp of the control region, compared with families on low natural radiation.  It counts new heteroplasmies explicitly and exclusively: 22 in 595 high-radiation transmissions against 1 in 200 low-radiation transmissions, and none of the 23 was fixed.  The article uses the low-radiation arm as its figure, and uses the study as its test of an environmental insult: a severe natural exposure raised arising mutations without fixing any. ↩ ↩ ↩

Howell, Neil et al.  “The pedigree rate of sequence divergence in the human mitochondrial genome: there is a difference between phylogenetic and pedigree rates.”  American Journal of Human Genetics 72(3):659–670.  2003.  It counts arising, not fixing, in the paper's own words:  “All newly arising substitutions, both heteroplasmic and homoplasmic, were included in our divergence-rate estimates.”  A pooled analysis of nine control-region pedigree studies, counting 28 mutations in 2,633 transmissions and counting every newly arising change, whether it sat in some of a person's mitochondrial copies or in all of them.  Its figure of 0.95 mutations per base pair per million years is a divergence rate, twice the mutation rate, which is verifiable from the paper's own table; halving it and keeping it inside the 1,122 base pairs of the control region gives the 4.72 × 10^-7 tabulated in the article.  It is the canonical pooled control-region figure, and the article relies on it.  Multiplying the published figure by 16,569 sites and doubling it, as Carter 2025 endnote 10 does, inflates the answer roughly thirty-fold. ↩ ↩ ↩

Santos, Cristina et al.  “Understanding differences between phylogenetic and pedigree-derived mtDNA mutation rate: a model using families from the Azores Islands (Portugal).”  Molecular Biology and Evolution 22(6):1490–1505.  2005.  The authors write:  “Eleven new point substitutions were found, all of them in heteroplasmy, from which five appear to be somatic mutations and six can be considered germinal.”  The last-named author is Aluja, María Pilar.  A pedigree study of Azores families, built as a model to explain why phylogenetic and pedigree mutation rates differ.  It is the first study to correct an arising count into a fixing rate explicitly, and zero reached homoplasmy.  The tabulated rate is the corrected figure, a model output; the raw data are 6 germinal mutations in 321 transmissions.  The article uses it as its model-corrected Azores row. ↩ ↩

Madrigal, Lorena et al.  “High mitochondrial mutation rates estimated from deep-rooting Costa Rican pedigrees.”  American Journal of Physical Anthropology 148(3):327–333.  2012.  A pedigree study of 19 deep-rooting genealogies from Costa Rica, covering 289 meioses (transmissions from parent to child), that found 7 mutations in hypervariable region I, the fast-changing part of the control region.  The paper uses a measured female generation interval of 28.3 years, which is worth noting against the 15 years the young-earth calculation requires; its abstract headline rate of 1.24 × 10^-6 per site per year could not be reproduced from its own Results or Table 2, which give 2.51 × 10^-6 (maximum) and 0.89 × 10^-6 (minimum).  The article quotes its authors' caution that the fast and slow estimates should both be considered. ↩ ↩

Guo, Yan et al.  “Very low-level heteroplasmy mtDNA variations are inherited in humans.”  Journal of Genetics and Genomics 40(12):607–615.  2013.  Their finding, in their words:  “Homoplasmic SNPs were inherited without exception within each family.”  A sequencing study of 26 mother-child pairs at a median depth of 3,981-fold, more than twenty times deeper than the Sardinian data the young-earth rate is built on.  It found zero fixed changes in 26 transmissions, where Jeanson's rate predicts 4.1; the article uses it as a test of that rate. ↩ ↩ ↩

Rebolledo-Jaramillo, Boris et al.  “Maternal age effect and severe germ-line bottleneck in the inheritance of human mitochondrial DNA.”  PNAS 111(43):15474–15479.  2014.  A study of 39 mother-child pairs, using blood and buccal (cheek) tissue sequenced at 20,000-fold coverage with a 1% detection threshold, that asks whether a mother's age changes the mitochondrial DNA she passes on.  It estimated the germline bottleneck at 30-35 mitochondrial genomes.  The article uses it for the bottleneck formula that separates mutations that arise from mutations that become fixed, and for its maternal-age numbers. ↩ ↩ ↩ ↩

Ding, Jun et al.  “Assessing Mitochondrial DNA Variation and Copy Number in Lymphocytes of ~2,000 Sardinians Using Tailored Sequencing Analysis Tools.”  PLoS Genetics 11(7):e1005306.  2015.  A study that sequenced lymphocytes (white blood cells) from about 2,000 Sardinians at roughly 180-fold coverage and reported how often mothers and children share mitochondrial variants, which is concordance, not a mutation rate.  It is the study Jeanson's whole-genome rate rests on, and the article uses it to show that his rate repurposes a concordance figure as a mutation count. ↩ ↩ ↩

Wei, Wei et al.  “Germline selection shapes human mitochondrial DNA diversity.”  Science 364(6442):eaau6520.  2019.  A genetic study that sequenced 1,526 mother-offspring pairs and counted the new heteroplasmies in each child.  Supplementary Table S2 lists 13 de novo D-loop variants, but the rate printed on the same row requires 131; every other row is exactly consistent with n = 1,526, and this one implies n = 151.  It is a dropped digit.  The article uses it as its main direct comparison of mutations appearing with mutations staying.  Mirror:  Repository.cam.ac.uk ↩ ↩ ↩ ↩

Zaidi, Arslan A. et al.  “Bottleneck and selection in the germline and maternal age influence transmission of mitochondrial DNA in human pedigrees.”  PNAS 116(50):25172–25178.  2019.  A genetic study of heteroplasmy, the mix of mutant and normal mitochondrial DNA inside one person, in 96 multigenerational families, with two tissues per person and a mean sequencing depth of about 11,400-fold.  It estimated a germline bottleneck of 10.3 segregating units (95% HPD 8.5–12.3) and a germline rate of 4.72 × 10^-7 per site per generation.  The heteroplasmy count is a measurement; the bottleneck correction that turns it into a fixation rate is a model.  The article uses it for the one measured lever on germline mutation, maternal age, and for the bottleneck that maternal age tightens. ↩ ↩ ↩ ↩

Connell, Jennifer R. et al.  “Pedigree derived mutation rate across the entire mitochondrial genome of the Norfolk Island population.”  Scientific Reports 12:6827.  2022.  A pedigree study that sequenced the complete mitochondrial genome of 225 individuals from 45 maternal lineages in the documented Norfolk Island genealogy, covering 345 mother-child transmissions, and found nine changes, seven of them heteroplasmic (a variant present in only some of a person's mitochondrial copies) and two fixed changes.  Counting all nine gives 0.058 mutations per site per million years; excluding the seven heteroplasmic changes gives 0.013, that is 1.30 × 10^-8 per site per year, against the phylogenetic 1.665 × 10^-8.  Jeanson's rate of 0.158 per generation predicts 54.6 fixed changes in 345 transmissions, and the probability of observing 2 or fewer is about 3 × 10^-21; this is the cleanest direct measurement of the whole-genome fixing rate in a deep human pedigree and the most important source in the article. ↩ ↩ ↩ ↩

Árnadóttir, Erla R. et al.  “The rate and nature of mitochondrial DNA mutations in human pedigrees.”  Cell 187(15):3904–3918.e8.  2024.  A study from deCODE Genetics/Amgen in Reykjavik of 64,806 Icelanders from 2,548 matrilines, with 116,663 mother-child transmissions and 8,199 mutations, which the article calls the largest such study ever run.  The article gives the rate as 0.0703 mutations per genome per transmission.  The article uses it as its direct measurement of the pedigree rate, divided into the measured depth of the human mitochondrial tree, which gives roughly 24,000 years.  Mirror:  Decode.com ↩ ↩ ↩ ↩ ↩

Arbeithuber, Barbara et al.  “Allele frequency selection and no age-related increase in human oocyte mitochondrial mutations.”  Science Advances 11(32):eadw4954.  Aug 6, 2025.  The oocytes came from 28 eggs of eight women aged 20-29, 25 from six aged 30-34, and 27 from eight aged 35-42.  The age result, from the abstract:  “with age, mutations increased in blood and saliva but not in oocytes … mtDNA in human oocytes is protected against accumulation of mutations with aging and having functional consequences.”  The mixed model gives Z-test P = 2.7 × 10^-3 for blood, 5.7 × 10^-2 for saliva and 0.688 for oocytes; the permutation test on old versus young oocytes gives P = 0.550 and a 0.99-fold decrease.  The paper's count is 52 de novo mutations at minor allele frequency >= 1% over.  “70 mother-to-oocyte transmissions … the germline mutation rate to be 3.95 × 10^-4 … and 1.95 × 10^-5 … mutations per site per generation for the D-loop and coding region, respectively.”  The per-genome figure quoted in this article is computed from those two published rates over the regions they cover: 3.95 × 10^-4 × 1,122 = 0.443 for the D-loop, and 1.95 × 10^-5 × 15,359 = 0.300 for what the paper calls the coding region (protein-coding 11,341 bp, rRNA 2,513 bp and tRNA 1,505 bp).  The 88 bp of non-coding sequence outside the D-loop is reported separately and omitted here.  Together they give 0.743 mutations per genome per generation over the 16,481 bp rated, or 4.5 × 10^-5 per site per generation and 1.8 × 10^-6 per site per year at 25 years.  The implied Eve age divides Behar's 57.1 root-to-tip by that figure.  A 2025 study that applied duplex sequencing, whose error rate is below 10^-7, to 80 single human eggs (oocytes) from 22 women aged 20 to 42, with blood and saliva from the same donors.  The article uses it as the only direct measurement of the quantity a young-earth rescue would need, and it points the wrong way for that rescue.  Mirror:  PMC ↩ ↩ ↩ ↩ ↩ ↩

Jeanson, Nathaniel T.  “A Young-Earth Creation Human Mitochondrial DNA 'Clock': Whole Mitochondrial Genome Mutation Rate Confirms D-Loop Results.”  Answers Research Journal 8:375–378.  2015.  A young-earth creationist article that builds a whole-genome mutation rate by pooling three whole-genome parent-child studies into one count of transmissions and mutations.  Its Supplemental Table 1, with formulas intact, reproduces every figure in Table 1 exactly when recomputed from its own inputs, so there is no arithmetic error in it.  The article uses it as the whole-genome step of Jeanson's clock argument.  Mirror:  Assets.answersresearchjournal.org ↩ ↩

Fu, Qiaomei et al.  “A revised timescale for human evolution based on ancient mitochondrial genomes.”  Current Biology 23(7):553–559.  2013.  A dating study that used ten radiocarbon-dated ancient mitochondrial genomes, spanning 40,000 years, as tip calibrations (samples of known age that anchor the clock), with no ape calibration anywhere in it.  It gives a whole-genome rate of 2.67 × 10^-8 per site per year (95% HPD, the range the model gives the rate with 95% probability, 2.16-3.16) and a mitochondrial TMRCA, the time back to the most recent common ancestor, of 157,000 years (120,000-197,000); the same data analysed by linear regression give 1.92 × 10^-8, a 39% difference, and the regression's own confidence interval spans a 2.3-fold range on nine points.  The article uses it for its calibrated rate in the comparison table and for the Cro-Magnon 1 redating quoted below. ↩

Brotherton, Paul et al.  “Neolithic mitochondrial haplogroup H genomes and the genetic origins of Europeans.”  Nature Communications 4:1764.  2013.  The last-named authors are Cooper, Alan and the Genographic Consortium.  A study of 39 complete haplogroup H genomes from Central European skeletons dated 5500-2000 BC.  Every calibration point sits inside the dendrochronologically anchored part of the radiocarbon curve, with no Pleistocene sample, no ape split and no Out-of-Africa node, and a date-randomisation test confirmed the dates carry enough signal.  The authors note their 2.4 × 10^-8 is 1.45x the ape-calibrated 1.66 × 10^-8.  The article uses it as its Holocene-only calibration row. ↩

Rieux, Adrien et al.  “Improved calibration of the human mitochondrial clock using ancient genomes.”  Molecular Biology and Evolution 31(10):2780–2792.  2014.  A recalibration of the human mitochondrial clock using 30 high-quality ancient whole mitochondrial genomes and 320 contemporary ones, of which 26 ancient genomes were kept after damage screening.  It gives a whole-genome rate of 2.143 × 10^-8 per site per year and a mitochondrial time to the most recent common ancestor of 143,000 years (112,000-180,000).  One error to note: Table S3 lists the six Blätterhöhle samples' ages in calibrated BC while labelling the column "Age (CalBP)," so each is exactly 2,000 years too young.  The article uses it for the tip-calibrated clock and its tests of rate constancy, damage and calibration. ↩ ↩

Fu, Qiaomei et al.  “Genome sequence of a 45,000-year-old modern human from western Siberia.”  Nature 514(7523):445–449.  2014.  A genome study of one femur from Ust'-Ishim in western Siberia, dated by two independent radiocarbon measurements of 41,400 ± 1,300 and 41,400 ± 1,400 BP, which calibrate to 46,880-43,210 cal BP.  Its mitochondrial rate is 2.53 × 10^-8 per site per year (95% HPD 1.76-3.23), and the abstract's rounded interval of 1.8 to 3.2 × 10^-8 is not a point estimate; the paper's own supplement gives the rate of Fu 2013 as 2.47 × 10^-8, where the published value is 2.67.  The article uses it as the single dated tip of its rate table and as its answer to the claim that fossil DNA is too degraded to use. ↩

Posth, Cosimo et al.  “Pleistocene mitochondrial genomes suggest a single major dispersal of non-Africans and a Late Glacial population turnover in Europe.”  Current Biology 26(6):827–833.  2016.  The last-named author is Krause, Johannes.  A study that recovered 35 new mitochondrial genomes from Late Pleistocene and early Holocene European hunter-gatherers.  It used 66 radiocarbon-dated ancient genomes as tip calibrations alongside 311 modern ones, spanning 3,600 to 46,880 cal BP, with roughly a third of the calibration points Holocene.  Four model combinations give 2.62, 2.67, 2.74 and 2.77 × 10^-8; the preferred Bayesian-skyline strict-rate model gives 2.74 × 10^-8 (2.44-3.01 HPD).  The article uses it as its table row for that model. ↩

Cabrera, Vicente M.  “Human molecular evolutionary rate, time dependency and transient polymorphism effects viewed through ancient and modern mitochondrial DNA genomes.”  Scientific Reports 11:5036.  2021.  A direct count of mutation differences between 423 radiocarbon-dated ancient genomes and 784 modern ones, binned into ten age periods, with no phylogenetic model at all.  Mean differences rise with age, almost monotonically (one small dip, from 1.63 to 1.51, at the fourth period): 0.63 at 1,119 BP, 3.28 at 8,504, 11.07 at 29,638, 14.39 at 40,160, which is arguably the cleanest single table in the subject.  Flagged discrepancy: the rates in its Table 2 cannot be reproduced from Table 1 by the paper's own stated formula, and reverse-engineering implies mutually inconsistent sequence lengths that are never stated.  The direction and monotonicity are not in doubt; the exact rate figures are not reproducible from published numbers, and are tabulated here with that caveat.  The article uses its youngest and oldest bins as two rows of its rate table. ↩ ↩

Översti, Sanni, and Jukka U. Palo.  “Variation in the substitution rates among the human mitochondrial haplogroup U sublineages.”  Genome Biology and Evolution 14(7):evac097.  2022.  A dating study of 234 ancient human mitochondrial genomes from haplogroup U, a branch of the maternal family tree, each dated by radiocarbon (calibrated years before present, cal BP).  Of these, 212 are Holocene and 84 are younger than 4,500 cal BP.  The authors ran twelve root-to-tip regressions (tests of how many mutations build up with time) on non-clock trees run in TempEst, all with positive slopes and R² from 0.10 to 0.84.  Their haplogroup U4 regression on 42 Holocene-only samples gives 8.80 × 10^-8 per site per year, the fastest defensible figure in the literature and the best number available to the young-earth side; even it implies a U4 coalescence beyond 12,000 years, for one small twig of the tree.  The article uses it for the largest lineage-to-lineage rate difference in human mitochondrial DNA, about twofold between U2/U5a and U5b. ↩

Horai, Satoshi et al.  “Recent African origin of modern humans revealed by complete sequences of hominoid mitochondrial DNAs.”  PNAS 92(2):532–536.  1995.  A research paper that sequenced the complete mitochondrial DNA of three humans, three African apes and an orangutan to estimate how fast hominoid mtDNA changes and when modern humans shared a common ancestor.  The calibration is two steps removed from anything datable: the orangutan and African ape split is fixed at 13 million years, and the human-chimpanzee split of 4.87 ± 0.23 million years (Myr) comes out of that.  The paper's own data give two irreconcilable dates for the human ancestor: 143,000 ± 18,000 years from synonymous sites and the D-loop, and 450,000 ± 70,000 from nonsynonymous and RNA-gene sites.  Only the first is ever quoted.  The article uses it in its table of rate estimates, where it gives both dates. ↩

Ingman, Max et al.  “Mitochondrial genome variation and the origin of modern humans.”  Nature 408(6813):708–713.  2000.  “has not evolved at a constant rate across all human lineages”.  “consequently less suitable for dating evolutionary events”.  A genetic study of 53 complete mitochondrial genomes, the whole molecule, from people around the world.  The mean pairwise difference over the whole molecule was 61.1 for all humans (n=53), 38.5 for non-Africans (n=32) and 76.7 for Africans (n=21), and the authors left out the D-loop, a control region, because it does not evolve at a constant rate.  The article uses it for the rate that depends on an assumed 5-million-year human-chimpanzee split, and for its whole-genome sequences, which first re-established the African root. ↩

Mishmar, Dan et al.  “Natural selection shaped regional mtDNA variation in humans.”  PNAS 100(1):171–176.  2003.  The last-named author is Wallace, Douglas C.  A study of 104 complete mtDNA genomes from all global regions, arguing that natural selection, with climate as one influence, shaped regional mtDNA variation.  The measured quantity is a ratio: human tree depth to the human-chimpanzee ancestor, 24.88 against 818.05 × 10^-4 substitutions per nucleotide, or 1 : 32.8, and the 198 ka date comes from fixing the denominator at 6.5 Myr.  Its separate claim of climate-driven selection on ATP6 rests on counts as small as 3 versus 7 nonsynonymous changes and was not reproduced by Kivisild 2006.  The article uses it for its 198 ka row, the date that the human-chimpanzee calibration gives. ↩

Kivisild, Toomas et al.  “The role of selection in the evolution of human mitochondrial genomes.”  Genetics 172(1):373–387.  2006.  A 2006 study of complete coding-region sequences from 277 people on five continents, arranged as a family tree of mitochondrial lineages.  Of the 1,788 mutations in the tree, 1,758 occurred at 1,292 variable coding-region sites, and the surplus of amino-acid-changing mutations is a general feature of the young branches, so the ratio of amino-acid-changing to silent variation declines with the age of the branch.  It did not replicate the climate-based regional differences in amino-acid-changing mutations that Ruiz-Pesini et al. 2004 claimed, although Mishmar et al. 2003 is among the studies it cites for the claim that natural selection has significantly shaped mtDNA evolution.  Its coalescent date for the tree is 160,000 years (SD 22,000).  The article reads the observed decline as variants carried by healthy living people still being removed, and uses it for the claim that mitochondrial sites are not neutral. ↩

Endicott, Phillip, and Simon Y. W. Ho.  “A Bayesian evaluation of human mitochondrial substitution rates.”  American Journal of Human Genetics 82(4):895–902.  2008.  A Bayesian phylogenetic analysis of 177 human and two chimpanzee mitochondrial genomes that dated the human mitochondrial ancestor twice on the same data, once with a human-chimpanzee calibration and once with dates of human settlement inside the tree.  The swap moved the answer by half, which is why the article treats every ape-calibrated date as conditional on its calibration. ↩ ↩ ↩

Soares, Pedro et al.  “Correcting for Purifying Selection: An Improved Human Mitochondrial Molecular Clock.”  American Journal of Human Genetics 84(6):740–759.  2009.  “one mutation every 3624 years”.  “a severe shortage of such uncontroversial episodes”.  A study that recalibrated the human mitochondrial clock for the whole molecule, using more than 2,000 complete mitochondrial genomes.  Its rate of 1.665 × 10^-8 substitutions per nucleotide per year is calibrated on an assumed 6.5–7 million-year human-chimpanzee split, so it is a model output rather than a measurement, and the young-earth circularity critique is correct about this paper specifically.  The paper also concedes that there are few uncontroversial episodes available for archaeological calibration.  The article uses it for the phylogenetic rate that the young-earth argument has to overturn, and for the size of the selection correction, which the paper's own curve puts at 1.50 times at 0–1 thousand years falling to 1.13 times at 50–100 thousand years. ↩ ↩

Mean and maximum pairwise substitutions by region for 54 living humans (the revised Cambridge Reference Sequence plus Ingman's 53 genomes), 36 Neanderthal records representing 35 individuals, and 7 Denisovan records, all on reference coordinates with the length-variable tracts 303–315, 16,184–16,193 and 8,281–8,289 left out.  The code reads the pairwise matrices and the sequence inventory and writes the panels. ↩ ↩ ↩ ↩

Behar, Doron M. et al.  “A 'Copernican' Reassessment of the Human Mitochondrial DNA Tree from its Root.”  American Journal of Human Genetics 90(4):675–684.  2012.  The root lies between L0 and L1'2'3'4'5'6, the paper says they are.  “separated from each other by 14 coding and four control-region mutations.”  A reassessment of the human mitochondrial DNA family tree that proposes a new reference sequence rooted on Neanderthal mitogenomes, the Reconstructed Sapiens Reference Sequence.  The abstract says the phylogeny was refined by adding information from 8,216 modern mitogenomes, while the mean distance below is reported over a larger set of 18,843, so the two numbers refer to different things.  The mean distance from the reconstructed root to a modern tip is 57.1 substitutions (standard deviation 5.9, range 41-77), and the paper also reports statistically significant molecular-clock violations in M and L2, which is a reason for caution about every published interval on either side.  The article uses it for the 57.1-substitution mean depth of the tree, its 41-77 range, and the bifurcation at the root. ↩

Kim, Hie Lim, and Stephan C. Schuster.  “Poor Man's 1000 Genome Project: Recent Human Population Expansion Confounds the Detection of Disease Alleles in 7,098 Complete Mitochondrial Genomes.”  Frontiers in Genetics 4.  2013.  A survey of 7,098 complete human mitochondrial genomes that reports how much people's mitochondrial DNA differs on average.  The mean pairwise difference across the whole molecule is 39.8 worldwide, with a maximum of 123; within Africa it is 64.0 (maximum 123), among non-Africans 35.4 (maximum 71), in Asians 30.1 and in Europeans 30.6.  The control region differs by 9.9 over 1,122 bp and the rest of the molecule by 29.9 over 15,447 bp, a 4.6-fold difference in per-site diversity that the young-earth clock, whose D-loop and whole-genome rates differ by only 1.13-fold, does not reconcile.  The paper is number 13 in volume 4.  Jeanson's stated non-African targets of 39.8 and 39.3 are in fact the global figures.  The article uses it as the observed comparator for every clock calculation in the article. ↩

Trifunovic, Aleksandra et al.  “Premature ageing in mice expressing defective mitochondrial DNA polymerase.”  Nature 429(6990):417–423.  2004.  A laboratory study in which mice were engineered to carry a proofreading-deficient version of the mitochondrial DNA polymerase.  It is the mtDNA mutator mouse, and the only measurement of what a broken mitochondrial polymerase actually does to the mutation rate.  PubMed ID 15164064.  The article uses it because it measured the effect in exactly the system the critic names. ↩

Stewart, James B. et al.  “Strong purifying selection in transmission of mammalian mitochondrial DNA.”  PLoS Biology 6(1):e10.  2008.  A study of how randomly generated mtDNA mutations pass from mother to offspring in the Trifunovic mutator mice.  It is the direct experimental test of what happens to a raised mutation rate as it passes through the germline.  The article uses it because the germline removes the protein-changing mutations that the polymerase adds.  Mirror:  PMC ↩

This project's scan of a 206-gene panel of DNA-repair and replication genes in the Neanderthal exomes and the Altai genome, run against the published Neanderthal change catalogues in two independent passes:  research dossier, catalogue scan and analysis code.  The catalogues hold single-nucleotide changes only, and the allele-frequency lookup is the archived one, not a fresh census of current gnomAD. ↩

This project's analysis of a 19-gene core fidelity panel in the high-coverage Denisova 3 genome:  final core summary, coverage and variant panel.  These are coding and coverage results, not functional assays. ↩

UniProt.  DNA polymerase subunit gamma-1 (P54098, DPOG1_HUMAN).  n.d.  The annotated sites are the exonuclease active site at Asp198; exonuclease motifs Exo I 196-200, Exo II 267-275 and Exo III 395-403; and polymerase motifs Pol A 887-896, Pol B 943-958 and Pol C 1134-1141.  Residue 579 is annotated as an RNA-primer binding residue and 593 as a template-strand binding residue.  The three natural variants UniProt records between residues 570 and 600 are R574W (autosomal-recessive progressive external ophthalmoplegia, dbSNP rs774474723), R579W (the same condition, rs556925652) and P587L (that condition plus mitochondrial DNA depletion syndromes 4A and 4B, rs113994096).  The Neanderthal change itself is dbSNP rs1463396518.  The protein database entry for the main subunit of the human mitochondrial DNA polymerase, which copies mitochondrial DNA.  It lists the enzyme's 1,239 residues, its active sites and motifs, and the natural variants recorded in each region.  The article uses it for the point that the stretch between residues 570 and 600 sits beside working parts of the enzyme and carries only recessive disease variants. ↩

Cosmic-ray dose at altitude, calculated by this project from the altitude formulas UNSCEAR prints in Annex B (paragraphs 61–66), with z in kilometres:  the directly ionising component scales as E(z) = E(0) × [0.21 exp(−1.649z) + 0.79 exp(0.4528z)], and the neutron component as E(z) = E(0) × b × exp(az), with b = 1 and a = 1 below 2 km and b = 2 and a = 0.7 above.  The site altitudes are the only other inputs.⁴² ↩

Arnold, Lee J. et al.  “Luminescence dating and palaeomagnetic age constraint on hominins from Sima de los Huesos, Atapuerca, Spain.”  Journal of Human Evolution 67:85–107.  2014.  A dating study of the Sima de los Huesos deposit using two physical methods, luminescence dating and palaeomagnetism.  It gives a minimum age of 427,000 ± 12,000 years for the deposit, from methods sharing no assumptions with the molecular estimate.  The molecular and physical dates agree to within 7%, and no young-earth response to this result appears in the literature.  The article uses it as the physical check on the mitochondrial date. ↩

UNSCEAR.  Sources and Effects of Ionizing Radiation, UNSCEAR 2008 Report Volume I, Annex B: Exposures of the public and workers from various sources of radiation.  United Nations, New York.  2008 (edition of 2010).  Paragraph 61–66.  The annex to the UNSCEAR 2008 report that covers radiation exposures of the public and of workers from various sources.  Its paragraphs explain how cosmic-ray dose rates rise with altitude, including a variation by a factor of about 4 between sea level and 4,000 m in Tibet, as measured in China.  The article uses it for the altitude dose figures behind its comparison with Baishiya Cave, while noting that dose quantities and environments are not perfectly interchangeable. ↩