Study	Whole genome1–16,569 · 16,569 bp				Count / method	Trans­missions	Pedigrees	Sapiens age	All-three age baseline	Method and caveats
	Control region16,024–576 · 1,122 bp			Coding region577–16,023 · 15,447 bp
	HVR-I16,024–16,383 · 360 bp	HVR-II57–372 · 316 bp	Rest16,384–56, 373–576 · 446 bp
Howell 1996	0.02272 / 88 · 1,194 bp			—	DetectedNew germline mutations, both heteroplasmic (HVR-II nt 152 and 195), plus one somatic	88	1 LHON pedigree, 45 members	17 ka	29.5 ka	Opened the pedigree-rate literature: the “ten times faster” claim began with two events in one disease pedigree. The primary sequenced nt 16024–16569 and 1–648 (about 1,194 bp), the control region plus 72 adjoining coding bases. Howell 2003 restates 2/88; Sigurðardóttir 2000 says 2/81.
Bendall 1996as tabulated by Howell 2003	0.01114 / 360 · 360 bp	—	—	—	DetectedNewly arising substitutions, heteroplasmic or homoplasmic, as Howell pooled them	360	Not given in Howell’s table	17.7 ka	32.8 ka	Howell 2003 Table 2: 4/360, divergence rate 0.99 (99.5% interval 0.14–3.35) per site per Myr, computed over Howell’s 1,122-bp convention at 20-year generations. HVR-1 only. The exact window was not checked; the age uses the 360-bp HVR-I window.
Mumm 1997as tabulated by Howell 2003	0.01691 / 59 · 360 bp	—	—	—	DetectedNewly arising substitutions, heteroplasmic or homoplasmic, as Howell pooled them	59	Not given in Howell’s table	11.6 ka	21.5 ka	Howell 2003 Table 2: 1/59, divergence rate 1.51 (99.5% interval 0.00–11.71) per site per Myr, computed over Howell’s 1,122-bp convention at 20-year generations. HVR-1 only; one event. The exact window was not checked; the age uses the 360-bp HVR-I window.
Parsons 1997	0.030610 / 327 · ~610 bp		—	—	DetectedSubstitutions between close relatives; heteroplasmic and homoplasmic both counted, per Howell 2003	327	134 maternal lineages of close relatives	10.2 ka	14.4 ka	The load-bearing young-earth citation. Ten differences across about 610 bp of HVR-I and HVR-II, not the whole molecule. The ~610 bp and the 20-year generation are reverse-engineered from the published 2.5/site/Myr; exact endpoints are unverified, so the age uses a broad HVR-I+II profile. The abstract gives the count. Howell’s Table 2 books this laboratory’s data as Parsons and Holland 1998 (10/306) plus the Forensic Science Service set (1/32): 11/338 together, rows below.
Parsons 1997: Forensic Science Service setas tabulated by Howell 2003	0.03131 / 32 · 676 bp		—	—	DetectedNewly arising substitutions, heteroplasmic or homoplasmic, as Howell pooled them	32	Not given in Howell’s table	11.1 ka	15.6 ka	Howell 2003 Table 2: 1/32, divergence rate 2.76 (99.5% interval 0.01–20.43) per site per Myr, computed over Howell’s 1,122-bp convention at 20-year generations. HVR-1 and HVR-2. Howell’s footnote: “the analyses at the U.K. Forensic Science Service that were reported in Parsons et al. (1997).” One event. The age uses the 676-bp HVR-I+II window.
Soodyall 1997as tabulated by Howell 2003	00 / 108 · 676 bp		—	—	DetectedNewly arising substitutions, heteroplasmic or homoplasmic, as Howell pooled them	108	Not given in Howell’s table	> 12.5 ka*	> 17.6 ka*	Howell 2003 Table 2: 0/108, divergence rate 0.00 (99.5% interval 0.00–4.79) per site per Myr, computed over Howell’s 1,122-bp convention at 20-year generations. HVR-1 and HVR-2. The Tristan da Cunha study, founded by a handful of women in the nineteenth century; Gibbons 1998 reports the same zero in 108. A zero count does not establish a zero rate. The age uses the 676-bp HVR-I+II window.
Parsons and Holland 1998as tabulated by Howell 2003	0.032710 / 306 · 676 bp		—	—	DetectedNewly arising substitutions, heteroplasmic or homoplasmic, as Howell pooled them	306	Not given in Howell’s table	10.6 ka	14.9 ka	Howell 2003 Table 2: 10/306, divergence rate 2.91 (99.5% interval 0.99–6.44) per site per Myr, computed over Howell’s 1,122-bp convention at 20-year generations. HVR-1 and HVR-2; the same laboratory as Parsons 1997. The exact window was not checked; the age uses the 676-bp HVR-I+II window.
Cavelier 2000: CR	00 / 292 · 1,122 bp			—	HomoplasmicHomoplasmic substitutions only	292	33 maternal lineages, 194 people	> 37.5 ka*	> 64.1 ka*	No homoplasmic substitutions detected; a zero count does not establish a zero mutation rate. Only length heteroplasmy, at three poly-C sites (65, 309, 16189). The zero result routinely left out of compilations on both sides. The limit assumes independent detectable events. Howell 2003 books the study as HVR-1 and HVR-2, 0/292.
Cavelier 2000: coding sample	—	—	—	00 / 292 · positions 5,550–6,050 · 500 bp	HomoplasmicHomoplasmic substitutions only	Same 292	Same	> 1.23 ka*	> 6.79 ka*	5550–6050 is 501 inclusive positions; the paper calls it 500 bp. The zero-count limit uses that window.
Sigurðardóttir 2000	0.004263 / 705 · 673 bp		—	—	HomoplasmicBase substitutions between relatives: the three homoplasmic changes, per Howell 2003	705	26 large Icelandic pedigrees, 272 people	81.1 ka	114 ka	Largest pre-genomic study, about 8 times slower than Parsons. The published HVR mask is 673 bp; the fully sequenced amplicon is nt 16012–394 (~952 bp), and inclusive endpoint arithmetic does not reproduce 673. Madrigal’s Table 3 pairs Howell’s count with Sigurðardóttir’s rate.
Sigurðardóttir 2000: rescored	0.007095 / 705 · 673 bp		—	—	Detected; rescoredHomoplasmic and heteroplasmic changes, Howell’s rescoring	Same 705	Same	48.7 ka	68.6 ka	Howell’s five-event rescoring of the same data, using the study’s published 673-bp HVR mask; not an independent study.
Heyer 2001	0.007874 / 508 · 673 bp		—	—	Pedigree differencesSubstitutions between relatives; pooled with earlier studies, 18 / 1,729 = 0.0104	508	Deep-rooting French-Canadian pedigrees (Quebec)	44 ka	62 ka	HVR-I 16024–16383 and HVR-II 58–370. Only 12 of the pooled 18 mutations sat at independent sites; three positions mutated twice. Fitted model: 40 fast sites (CI 27–54) at 274 per site per million generations. The pooled count overlaps other studies.
Forster 2002: low radiation	0.0051 / 200 · 579 bp			—	Detected heteroplasmyNew heteroplasmies only; none homoplasmic at either dose	200	248 Kerala families, 988 people, across 2 to 4 generations (795 transmissions in both groups)	68.2 ka	86.5 ka	16093–16390 and 35–315: parts of both HVRs plus adjoining bases. Kerala natural radioactivity, low-dose group.
Forster 2002: high radiation	0.03722 / 595 · 579 bp			—	Detected heteroplasmyNew heteroplasmies only; none homoplasmic	595	Same families	9.22 ka	11.7 ka	A 7.4-fold rise in heteroplasmies appearing, and zero fixations. The radiation-hit positions “mutate, on average, 12 times faster during evolution.” The exposure comparison is not a universal clock.
Howell 2003: own pedigreesas tabulated by Howell 2003	0.005411 / 185 · 1,122 bp			—	DetectedNewly arising substitutions, heteroplasmic or homoplasmic, as Howell pooled them	185	Not given in Howell’s table	71.3 ka	122 ka	Howell 2003 Table 2: 1/185, divergence rate 0.48 (99.5% interval 0.00–3.87) per site per Myr, computed over Howell’s 1,122-bp convention at 20-year generations. Control region nt 16024–16569 and 1–576 in five LHON families. One event; included in the pooled row below.
Howell 2003: pooled CR	0.010628 / 2,633 · pooled · 1,122 bp			—	Detected; pooledEverything newly arising, heteroplasmic or homoplasmic	2,633	5 LHON pedigrees (185 transmissions) plus nine earlier data sets	36.2 ka	61.9 ka	The canonical pooled rate, published as a divergence rate of 0.95/site/Myr (twice the mutation rate; 99.5% interval 0.53–1.57) at 20-year generations, treating every study as 1,122 bp. Multiplying it across 16,569 sites is Carter’s error. Regions and assays differ between the pooled studies. Table 2’s ten study rows sum to 2,643 transmissions, not the printed 2,633.
Howell 2003: coding counts	—	—	—	0.008664 / 462 · Subset	Pedigree eventsCoding-region events in partially sequenced LHON pedigrees	462	LHON pedigrees and Cavelier’s families	Not calculated	Not calculated	Mixed coverage: earlier LHON fragments, later coding sequences, Cavelier’s 500 bp. The 6,773 bp in Howell 1996 is not the common mask of all 462 transmissions. The raw count has no matched age.
Howell 2003: coding approximation	—	—	—	0.00927 model0.06/site/Myr divergence, halved, × 15,447 × 20 y · 15,447 bp	Full-region lower estimateThe same coding events, converted to a rate	Same	Same	112 ka	427 ka	Published 0.06/site/Myr divergence rate (99.5% CI 0.008–0.20), halved, then converted using the study’s 20-year generation. Treating partial coverage as complete coding coverage makes this conservative; it is not a directly observed whole-coding rate.
Santos 2005: germinal	0.01876 / 321 · ~610 bp		—	—	Detected; germinal classNew point substitutions, all heteroplasmic, classified somatic or germinal	321	Azores families; heteroplasmy in 72% of them	16.7 ka	23.6 ka	Six germinal events; 0.034 = 11/321 counting the five judged somatic. Counts from the abstract; exact windows are unverified, so the regional age is approximate.
Santos 2005: corrected	0.0037 modelmodel, corrected to fixation · ~610 bp		—	—	Transmission modelThe germinal events, corrected for loss before fixation	Same 321	Same	84.4 ka	119 ka	First pedigree study to apply the arising-to-fixing correction; its corrected 0.2415/site/Myr lands “in the upper end” of phylogenetic values. A model-corrected rate for the measured region, not a direct homoplasmic count; it must not be assigned to 16,569 bases.
Madrigal 2012: maximum	0.02567 / 273 · 360 bp	—	—	—	Pedigree differencesSubstitutions between relatives; heteroplasmy status not stated	273	19 deep-rooting Costa Rican pedigrees of at least 7 generations; 53 people typed	7.67 ka	14.2 ka	HVR-I, 360 bp. Generation time measured, 28.3 years over 289 meioses: at that interval multiply these ages by 1.132. The text reports 7 events over 299 meioses; its printed 71.1 × 10-6 per site per generation equals 7/273. Three of the mutated sites are known hot spots. The abstract’s 1.24 × 10-6 per site per year appears nowhere in the results.
Madrigal 2012: minimum	0.009092 / 220 · 360 bp	—	—	—	Pedigree differencesSubstitutions between relatives; heteroplasmy status not stated	220	Same	21.6 ka	40.1 ka	Excluding a pedigree and hot-spot sites as the paper does leaves 2/220. Different branch and hot-spot exclusions define the two counts.
Guo 2013	00 / 26 · 3,981× depth · 16,569 bp · < 2.78 × 10-7 per site per year				HomoplasmicHomoplasmic changes; heteroplasmies followed down to 0.1%	26	18 families, 44 people; mother–child pairs plus 9 sibling pairs	> 12.4 ka*	> 39.1 ka*	No discordant homoplasmic SNPs; 12 heteroplasmies shared by mother and child. A zero count does not exclude rare events, and 26 transmissions give only a weak bound. The mothers were childhood-cancer survivors with uterine irradiation, which biases the count upward. Implies a bottleneck near 200, the high outlier.
Rebolledo-Jaramillo 2014: arising	0.143 modelα = μ0/(2 ln 99); μ0 = 51/(39 × 16,569) observed · 16,569 bp · 3.45 × 10-7 per site per year				Arising-rate modelGermline point heteroplasmies at ≥ 1% in both blood and buccal tissue, fed into a transmission model	39	39 mother–child pairs, two tissues each	9.98 ka	31.5 ka	Fitted arising parameter α. The paper estimates μ0 as 51/(39 × 16,569 bp): the 51 are heteroplasmies observed above its 1% threshold, which include inherited variation; they are not 51 new mutations. 39% of heteroplasmies were tissue-specific, so somatic.
Rebolledo-Jaramillo 2014: retained	0.0045 model2.7 × 10-7 per site per generation × 16,569 ≈ 0.0045; bottleneck 32.3 · 16,569 bp · 1.09 × 10-8 per site per year				Transmission modelThe same heteroplasmies, divided by the measured bottleneck	Same 39	Same	317 ka	1 Ma	The paper that writes μ = α/N. Fitted μ after frequency and bottleneck modelling (IQR 8.3 × 10-8 to 8.2 × 10-7 per site per generation), not a direct count of homoplasmic changes. The authors call their estimate a lower bound. The best usable clock parameter from this study.
Rebolledo-Jaramillo 2014: D-loop retained	0.00168 model1.5 × 10-6 per site per generation × 1,122 · 1,122 bp			—	Transmission modelD-loop heteroplasmies, modelled separately	Same 39	Same	229 ka	391 ka	The paper’s separately modelled D-loop rate; its interquartile range, 4.8 × 10-7 to 4.7 × 10-6, spans tenfold.
Ding 2015: heteroplasmy gains	0.423141 / 333 · ≥ 4% MAF · ~180× · 16,569 bp · 1.02 × 10-6 per site per year				Detected discordanceHeteroplasmic discordance between mother and child, lymphocytes only	333	333 Sardinian parent–child trios	3.37 ka	10.7 ka	Child variants absent from the mother’s called class. Discordance between tissues sampled from mother and child is not automatically a new germline mutation; threshold, somatic origin and calling errors prevent identifying a retained rate.
Ding 2015: homoplasmy discordance	0.18963 / 333 · 35 gains (0.105) + 28 losses · 16,569 bp · 4.57 × 10-7 per site per year				Discordant callsHomoplasmic discordance between mother and child, either direction	Same 333	Same	Not calculated	Not calculated	The dataset behind Jeanson’s rate: 28 of his 63 discordances are variants the mother had and the child lost. The authors read the 35 child-not-mother calls as noise, and they were never re-examined at the read level. Gains and losses cannot both simply be counted as new inherited substitutions. Not a validated mutation rate.
Jeanson 2015 synthesis, for comparison	0.15863 / 398 · 16,569 bp · 3.82 × 10-7 per site per year				Proposed mutation clockHomoplasmic discordances in Ding 2015, pooled with Guo and Rebolledo-Jaramillo (both zero), treated as homoplasmic changes	398	Three studies pooled	9.02 ka	28.5 ka	Mostly Ding’s unvalidated discordances. Predicts 4.1 homoplasmic changes in Guo’s 26 transmissions (0 seen), 54.6 in Norfolk’s 345 (2 seen) and 242 near-homoplasmic events in Wei’s 1,526 (3 seen). Shown to test the proposal, not as a measured rate.
Wei 2019: detected	0.313477 / 1,526 · > 1% VAF · ~1,900× · 16,569 bp · 7.55 × 10-7 per site per year				Detected de novoNew heteroplasmies above 1% in the child’s blood	1,526	1,526 mother–offspring pairs, whole blood	4.57 ka	14.4 ka	614 maternal heteroplasmies lost against 477 gained. Offspring blood variants undetected in the mother; not all germline or destined to persist. Low-level against near-homoplasmic is a factor of 159 in one generation: a snapshot, not the eventual fate of the 477.
Wei 2019: control	0.0858131 / 1,526 · 1,122 bp			—	Detected de novoThe same events in the D-loop	Same	Same	4.49 ka	7.66 ka	131 D-loop events; the supplementary Table S2’s printed 13 is inconsistent with its rate and totals. Regional clock illustration only.
Wei 2019: coding	—	—	—	0.227346 / 1,526 · 15,447 bp	Detected de novoThe same events outside the D-loop (genes, rRNA, tRNA)	Same	Same	4.6 ka	17.4 ka	346 outside-D-loop events. Same tissue and threshold limitations.
Wei 2019: frequency weighted	0.0194 model477 × 0.062 / 1,526 · 16,569 bp · 4.68 × 10-8 per site per year				Neutral illustrationThe 477, each weighted by its heteroplasmic fraction	Same	Same	73.7 ka	233 ka	Σ-frequency approximation. Blood is not the germline; selection and undetected variants remain unmodelled.
Wei 2019: above 80%	0.001973 / 1,526 · 16,569 bp · 4.75 × 10-9 per site per year				Near-homoplasmicDe novo variants above 80% heteroplasmic fraction	Same	Same	726 ka	2.29 Ma	Two of the three events have ascertainment concerns. Above 80% is not complete homoplasmy, and no population-wide fixation estimate follows.
Zaidi 2019	0.00782 model4.72 × 10-7 per site per generation × 16,569 · ~11,400× · 16,569 bp · 1.89 × 10-8 per site per year				Transmission modelGermline heteroplasmies (in both blood and buccal) converted to a fixation rate through a bottleneck of 10.3	249	96 families, 345 people, 2 to 4 generations, at least 2 siblings per mother	183 ka	577 ka	Independent of Rebolledo-Jaramillo, with a bottleneck three times smaller; per transmission it is 1.7 times Rebolledo-Jaramillo’s retained rate. Nonsynonymous-to-synonymous occurrence 0.43. A model-inferred germline rate, not a count of observed homoplasmic transitions.
Connell 2022: all	0.02619 / 345 · Sanger · 16,569 bp · 6.30 × 10-8 per site per year				Detected; mixed classHomoplasmic changes plus heteroplasmies above 20% MAF	345	45 maternal lineages, 225 people, from the last four generations of an 11-generation island pedigree	54.7 ka	173 ka	Norfolk Island. Seven heteroplasmies and two homoplasmic changes; retaining or excluding heteroplasmy changes the estimand. Its control-region rate, 0.469/site/Myr, matches Howell 2003’s pooled 0.475. Population-specific pedigree.
Connell 2022: homoplasmic	0.00582 / 345 · 16,569 bp · 1.40 × 10-8 per site per year				HomoplasmicThe two homoplasmic transitions only	Same 345	Same	246 ka	778 ka	The best direct retained-change count here, but only two events, so its interval is broad. No bottleneck model anywhere in the derivation; its published 1.30 × 10-8 per site per year is 0.78 times the phylogenetic rate. At the paper’s 26.9-year interval, multiply ages by 1.076. Detection and family structure still require modelling.
Connell 2022: control, all	0.01455 / 345 · 1,122 bp			—	Detected; mixed classThe nine events, by position	Same 345	Same	26.6 ka	45.4 ka	The same nine-event study, partitioned by the positions in its Table 1. HVR-II ends at 371 here. One or zero events give weak constraints. The count-derived CR rate differs slightly from the printed Table 2 estimate.
Connell 2022: control, homoplasmic	0.00291 / 345 · 1,122 bp			—	HomoplasmicThe two homoplasmic transitions, by position	Same 345	Same	133 ka	227 ka	The same nine-event study, partitioned by the positions in its Table 1. HVR-II ends at 371 here. One or zero events give weak constraints. The count-derived CR rate differs slightly from the printed Table 2 estimate.
Connell 2022: HVR-I, all	0.01164 / 345 · 360 bp	—	—	—	Detected; mixed classThe nine events, by position	Same 345	Same	17 ka	31.4 ka	The same nine-event study, partitioned by the positions in its Table 1. HVR-II ends at 371 here. One or zero events give weak constraints.
Connell 2022: HVR-I, homoplasmic	0.00291 / 345 · 360 bp	—	—	—	HomoplasmicThe two homoplasmic transitions, by position	Same 345	Same	67.9 ka	126 ka	The same nine-event study, partitioned by the positions in its Table 1. HVR-II ends at 371 here. One or zero events give weak constraints.
Connell 2022: HVR-II, all	—	0.00291 / 345 · 315 bp	—	—	Detected; mixed classThe nine events, by position	Same 345	Same	51.6 ka	42.8 ka	The same nine-event study, partitioned by the positions in its Table 1. HVR-II ends at 371 here. One or zero events give weak constraints.
Connell 2022: HVR-II, homoplasmic	—	00 / 345 · 315 bp	—	—	HomoplasmicThe two homoplasmic transitions, by position	Same 345	Same	> 17.2 ka*	> 14.3 ka*	The same nine-event study, partitioned by the positions in its Table 1. HVR-II ends at 371 here. One or zero events give weak constraints.
Connell 2022: rest of CR, all	—	—	00 / 345 · 447 bp	—	Detected; mixed classThe nine events, by position	Same 345	Same	> 4.48 ka*	> 19.5 ka*	The same nine-event study, partitioned by the positions in its Table 1. HVR-II ends at 371 here. One or zero events give weak constraints.
Connell 2022: coding, all	—	—	—	0.01164 / 345 · 15,447 bp	Detected; mixed classThe nine events, by position	Same 345	Same	89.9 ka	341 ka	The same nine-event study, partitioned by the positions in its Table 1. HVR-II ends at 371 here. One or zero events give weak constraints.
Connell 2022: coding, homoplasmic	—	—	—	0.00291 / 345 · 15,447 bp	HomoplasmicThe two homoplasmic transitions, by position	Same 345	Same	360 ka	1.36 Ma	The same nine-event study, partitioned by the positions in its Table 1. HVR-II ends at 371 here. One or zero events give weak constraints.
Árnadóttir 2024: all	0.07038,199 / 116,663 · 16,569 bp · 1.70 × 10-7 per site per year				Detected; branch inferredEvery mutation reaching 5% of reads, with one carrier at 10% or more	116,663	64,806 living Icelanders in 2,548 matrilines (deCODE genealogy)	20.3 ka	64.2 ka	Seventy-five times larger than anything before it. Hypermutable positions dominate the count; the control region runs 8.3 times the coding rate per site. At deCODE’s 29.3-year interval, multiply ages by 1.172. A tight bottleneck is not a universal divisor converting this count into retained substitutions.
Árnadóttir 2024: control	0.026 model2.38 × 10-5 per site × 1,092 callable · 1,092 bp			—	Detected; regionalThe same events in the control region	Same	Same	14.4 ka	24.6 ka	Callable CR denominator: 1,092 bp after exclusions. Published per-site parameter; the age uses the broader CR profile. Threshold and hypermutable sites limit its use as a clock.
Árnadóttir 2024: coding	—	—	—	0.0443 model2.87 × 10-6 per site × 15,443 callable · 15,443 bp	Detected; regionalThe same events in the coding region	Same	Same	23.5 ka	89.2 ka	16,535 attempted sites minus 1,092 CR sites gives 15,443. The age uses the broader coding profile; selection and detection thresholds affect the connection to a long-run clock.
Arbeithuber 2025: total	0.74352 / 70 · ≥ 1% VAF · duplex sequencing, error < 10-7 · 16,569 bp · 1.79 × 10-6 per site per year				Detected in oocytesDe novo mutations at 1% or more of a single oocyte’s mtDNA copies, counted before any transmission	70	80 single oocytes from 22 women aged 20–42, with blood and saliva from each donor	1.92 ka	6.07 ka	The fastest human rate measured and the purest arising rate: counted in the germ cell itself, before bottleneck, drift or transmission. No rise with maternal age in oocytes (P = 0.688) against a rise in blood (P = 0.0027). Mature oocytes have already experienced selection; this is not a retained-lineage rate.
Arbeithuber 2025: D-loop	0.443 model3.95 × 10-4 per site per generation × 1,122 · 1,122 bp			—	Detected in oocytesThe same mutations in the D-loop	Same	Same	0.869 ka	1.48 ka	The authors’ own regional split. Surviving egg variants are not equivalent to homoplasmic descendant sequences.
Arbeithuber 2025: coding genes	—	—	—	0.3 model1.95 × 10-5 per site per generation × 15,359 · 15,359 bp	Detected in oocytesThe same mutations in genes, rRNA and tRNA	Same	Same	3.46 ka	13.1 ka	Protein, rRNA and tRNA genes total 15,359 bp; the 88 bp of other non-coding sequence are reported separately and are not in this denominator. The age uses a coding-region approximation.

Study	Whole genome1–16,569 · 16,569 bp				What was compared	Calibration assumed	Authors’ age of Eve	Sapiens age	All-three age baseline	Method and caveats
	Control region16,024–576 · 1,122 bp			Coding region577–16,023 · 15,447 bp
	HVR-I16,024–16,383 · 360 bp	HVR-II57–372 · 316 bp	Rest16,384–56, 373–576 · 446 bp
Horai 1995: control	0.002 model1,143 bp · 7.00 × 10-8 per site per year			—	3 human and 4 ape complete genomes	Orangutan–African ape split at 13 Myr; the human–chimp 4.87 Myr is an output	143 ka ± 18 (synonymous and D-loop); 450 ka ± 70 (nonsynonymous and RNA sites, same paper)	196 ka	335 ka	The 7.00 × 10-8 D-loop rate is itself derived from the 3.89 × 10-8 synonymous rate. 1,143-bp control-region convention; the regional distance is approximate. Other site classes give different dates.
Ingman 2000	—	—	—	0.00656 model15,426 bp · 1.70 × 10-8 per site per year	53 human genomes and 1 chimpanzee; D-loop excluded	Human–chimp split at 5 Myr	171.5 ka ± 50	159 ka	602 ka	Uses the coding distance, not the 57.1 whole depth.
Mishmar 2003	—	—	—	0.00487 model15,447 bp · 1.26 × 10-8 per site per year	104 human genomes, coding region	Human–chimp split at 6.5 Myr	198 ka ± 19	214 ka	813 ka	Selection and calibration uncertainty remain relevant; the paper argues for climatic selection while dating with a strict clock.
Kivisild 2006: synonymous	—	—	—	Not determined modelSubset · 3.50 × 10-8 per site per year	277 genomes, synonymous transitions only	6 Myr species split; 6.5 Myr mtDNA coalescence	160 ka (SD 22)	Not calculated	Not calculated	The rate is per synonymous site. No matching synonymous-site count and root profile was established here, so neither a per-transmission figure nor an implied age is calculated.
Endicott & Ho 2008: external	—	—	—	0.000687 model7,198 bp · 3.82 × 10-9 per site per year	177 human and 2 chimpanzee genomes; 1st and 2nd codon positions (7,198 bp) analysed alone	Human–chimp split, lognormal, 95% within 5–7.5 Myr	162 ka (122–213) from the concatenated 11,940 bp at 6.81 × 10-8; 310 ka (158–508) from these positions alone	Not calculated	Not calculated	Table 3’s separately analysed PC1+2 rate (95% HPD 2.44–5.06 × 10-9). Matching root counts for these positions were not established, so no age is calculated here.
Endicott & Ho 2008: internal	—	—	—	0.00234 model7,198 bp · 1.30 × 10-8 per site per year	The same 177 genomes and software, calibration swapped	Haplogroup P in Sahul at 40–55 ka, plus H1 and H3 at about 18 ka. No ape	108 ka (82–134) from the concatenated 11,940 bp at 9.66 × 10-8; 79 ka (60–106) from these positions alone	Not calculated	Not calculated	Table 3’s separately analysed PC1+2 rate (1.01–1.58 × 10-8), 3.4 times the external one. Not a direct pedigree measurement.
Soares 2009	0.0069 model16,569 bp · 1.67 × 10-8 per site per year				Tree of 2,196 complete genomes, corrected for purifying selection	Human–chimp split at 6.5 Myr plus 0.5 Myr lineage coalescence	192.4 ka (Figure 6)	207 ka	654 ka	The benchmark every pedigree rate is compared against; a model output, not a count. The table uses its summary whole-molecule rate (one mutation every 3,624 years), not the full age-dependent function. The coding-region cross-check gives a very similar 1.708 × 10-8.
Soares 2009: HVS-I	0.0015 model350 bp · 1.71 × 10-7 per site per year	—	—	—	The same tree, HVS-I 16051–16400 (350 bp)	Scaled against the synonymous clock	Same tree	128 ka	236 ka	The paper’s average of 1.62 × 10-7 and 1.784 × 10-7: “1 mutation every 16,677 years” in the segment. Its window differs from HVR-I 16024–16383; the age uses the HVR-I profile.
Fu 2013	0.0111 model16,569 bp · 2.67 × 10-8 per site per year				10 dated ancient and 311 modern genomes	Radiocarbon-dated tips; Bayesian relaxed clock	157 ka (120–197)	129 ka	408 ka	Dates, site corrections and shared ancestry matter.
Brotherton 2013	0.00994 model16,569 bp · 2.40 × 10-8 per site per year				39 Neolithic and Bronze Age genomes	Holocene-only tip calibration	—	144 ka	454 ka	A published genome-rate summary rather than a raw homoplasmic count.
Rieux 2014	0.00888 model16,569 bp · 2.14 × 10-8 per site per year				26 retained ancient and 320 modern genomes; damage screening	Tip, node and root calibration compared	143 ka (112–180)	161 ka	508 ka	Its fitted mtEve age and the standardized depth division here are different calculations.
Fu 2014	0.0105 model16,569 bp · 2.53 × 10-8 per site per year				The Ust’-Ishim femur and 311 modern genomes	A single dated tip	—	136 ka	430 ka	The specimen date is part of the inference.
Posth 2016	0.0113 model16,569 bp · 2.74 × 10-8 per site per year				66 dated ancient and 311 modern genomes	Dated tips; Bayesian skyline, strict rate	—	126 ka	397 ka	Shares some comparison data with earlier work.
Cabrera 2021: youngest bin	0.0179 model16,569 bp · 4.33 × 10-8 per site per year				423 dated and 784 modern genomes	Direct mutation counts by age bin, no tree	—	79.6 ka	251 ka	The same data, binned by age, give a rate that falls 2.3-fold from the youngest bin to the oldest. A bin-specific rate is not automatically a rate for the complete tree.
Cabrera 2021: oldest bin	0.00791 model16,569 bp · 1.91 × 10-8 per site per year				423 dated and 784 modern genomes	Direct mutation counts by age bin, no tree	—	180 ka	570 ka	The same data, binned by age, give a rate that falls 2.3-fold from the youngest bin to the oldest. A bin-specific rate is not automatically a rate for the complete tree.
Översti & Palo 2022: U4	0.0365 model16,569 bp · 8.80 × 10-8 per site per year				42 Holocene U4 genomes, no moderns	Root-to-tip regression on dated tips	—	39.2 ka	124 ka	The fastest defensible number in the mainstream literature, and the best single figure available to the young-earth side. It is also 42 sequences from one twig of the tree with R² = 0.37, and its own x-intercept still puts that twig’s root at 12,723 years. Not an established universal mtDNA rate.
Project: U-subclade regression	0.0107 model16,519 bp · 2.59 × 10-8 per site per year				229 dated ancient genomes	Root-to-tip, per-subclade intercepts	—	133 ka	421 ka	An in-house descriptive calculation. Independent-error intervals omit phylogenetic covariance and date uncertainty.

Panel	Whole genome1–16,569 · 16,569 bp				What is measured	Pairs	Sample	Average pair’s ancestor at Soares’s rate	Notes
	Control region16,024–576 · 1,122 bp			Coding region577–16,023 · 15,447 bp
	HVR-I16,024–16,383 · 360 bp	HVR-II57–372 · 316 bp	Rest16,384–56, 373–576 · 446 bp
All living humans54 genomes	59.40mean of all pairs · max 106 · control region 15.88 (max 33)				Pairwise differences within the group	1,431	rCRS plus Ingman’s 53 globally sampled genomes	108 ka	Africa-heavy by design, which is why its numbers are higher than Kim and Schuster’s. Half the average pair is a pair-specific branch length, not the root-to-tip depth.
	8.19max 19	5.76max 16	1.92max 5	43.52max 87
Behar root to tip18,843 genomes	57.1mean · range 41–77				Substitutions from the reconstructed root (RSRS) to each living genome	—	18,843 complete genomes	207 ka	The sapiens-age column’s whole-molecule anchor; median 56, SD 5.9. A 6,000-year root needs 5.74 × 10-7 per site per year: 26.8 times Rieux 2014’s rate and 34.5 times Soares 2009’s. The second line is this project’s allocation of the 57.1 by the RSRS profile of the 54 genomes (scaled by 57.1/50.241), not a figure Behar published.
	7.87allocated	5.98allocated	1.56allocated	41.69allocated
Kim and Schuster7,098 genomes	39.8Π · max 123				Nucleotide diversity Π: differences between two random genomes	—	7,098 complete genomes, 10% African	72.1 ka	A Europe- and Asia-heavy panel. Its control-region and coding figures are Π over 1,122 and 15,447 bp; no HVR split is published.
	9.9max 32			29.9max 98
All Neanderthals36 records, 35 individuals	31.49mean of all pairs · max 118 · control region 10.40 (max 29)				Pairwise differences within the group	630	Every near-complete Neanderthal mitogenome in GenBank at the time of the relative-rate test	57.1 ka	About half the modern human spread.
	5.84max 20	3.57max 10	0.98max 6	21.09max 93
All Denisovans7 records	42.76mean of all pairs · max 86 · control region 13.76 (max 25)				Pairwise differences within the group	21	Seven mitogenomes	77.5 ka	Seven records only, so the maximum means little. Wider than the Neanderthal spread despite the smaller sample. Denisova 19 and 21 have identical mtDNA.
	8.33max 14	4.67max 9	0.76max 2	29.00max 61
Human with Neanderthalbetween groups	199.35mean of all pairs · max 225 · control region 35.06 (max 45)				Mean pairwise differences between the two groups	1,944	The two panels above	361 ka	Spread over 6,000 years at 25-year generations, it needs 0.415 retained changes per lineage per generation, with tips treated as contemporaneous.
	20.50max 30	9.03max 16	5.52max 9	164.29max 182
Human with Denisovanbetween groups	360.83†mean of all pairs · max 391 · control region 44.42 (max 56)				Mean pairwise differences between the two groups	378	The two panels above	654 ka	Spread over 6,000 years at 25-year generations, it needs 0.752 retained changes per lineage per generation, with tips treated as contemporaneous.
	25.44max 32	8.70max 16	10.27max 14	316.42†max 341
Neanderthal with Denisovanbetween groups	349.79mean of all pairs · max 378 · control region 52.62† (max 57)				Mean pairwise differences between the two groups	252	The two panels above	634 ka	Spread over 6,000 years at 25-year generations, it needs 0.729 retained changes per lineage per generation, with tips treated as contemporaneous.
	29.13†max 34	9.92†max 12	13.58†max 17	297.17max 322
Human with Sima de los Huesosbetween groups	279.33mean of all pairs · max 296 · control region 20.15 (max 27)				Mean pairwise differences between the groups	54	54 living humans against the one Sima sequence	506 ka	One sequence only. Not used by the all-three column, which takes the three groups with several sequences each.
	7.17max 12	7.98max 13	5.00max 7	259.19max 269
All three pooled97 genomes	154.17mean of all pairs · max 391 · control region 27.44 (max 57)				Pairwise differences across all 97	4,656	The three panels above	279 ka	Uncorrected pairwise figures, not model split dates. Half the pooled average underestimates the root, because many pairs share younger ancestors.
	15.55max 34	7.29max 16	4.60max 17	126.73max 341