Working notes for Shared Genetic Mistakes and Common Ancestry. Every source here is either cited in the article or explicitly marked not used with a reason. Library keys refer to sources/index.tsv.
The quotation ledger is in quotes.
Method note
Every quotation in the article was checked against the downloaded PDF bytes, not against a web summary or a search-engine snippet. Two-column PDFs were re-extracted with plain pdftotext rather than pdftotext -layout, because a sentence that spans a column break reads as absent under a naive phrase search. Three sources could not be obtained and their quotations are flagged as unverified in both this file and the article footnotes.
The creationist replies, in their own words
Added 2026-10-27 by the strengthening review. Before then the article cited no creationist source, which broke the standing rule that a creationist position is stated from its own primary source before it is priced. Each quotation below was re-read from the bytes this session.
Liu & Soper 2009, Answers Research Journal 2:97 (liu-2009). Used. Model, not measurement: orthologous ERVs "were created to occupy similar genomic loci in separate species by a single designer to carry out similar physiological functions" (section "The Problem of ERVs"); exogenous retroviruses "were derived from endogenous retroviruses" (Abstract); directed integration "may also partially explain" shared positions ("Regulation of ERVs by the cell"); the syncytin-1 locus is "especially challenging to creationists" ("A Tale of Two ERV 'Cousins'"). Direction: complicates. It is the fullest creationist proposal and gives no count or prediction. Verified: full (saved HTML, section headings from its h2/h3 tags).
Liu & Nguyen 2018, Proc. ICC 8:191 (liu-2018). Used for one sentence, p. 198: shared position is "more likely due to functional necessity rather than common descent." Laboratory work on HERV-K regulation in a cancer cell line; no count of shared loci. Folio anchored per page with pdftotext -layout: PDF p. 9 = 198. Verified: full.
Terborg 2013, Journal of Creation 27(3):105 (borger-2013; printed byline Peer Terborg). Used. The "VIGE-first hypothesis", p. 105 (abstract); the syncytins need "at least six independent gene-capture events" and "map to different genomic locations", p. 108. Direction: the first supports the creationist reversal of direction; the second, reported by a creationist, supports the control-case argument against it. Verified: full.
Wood 2006, Occasional Papers of the BSG 7:1 (wood-2006). Used. A creationist biologist's review of the chimpanzee genome. p. 9 (PDF p. 11, offset +2 confirmed by the printed folio): shared pseudogenes with identical lesions are "difficult to explain" without common ancestry. Direction: undercuts the creationist position, in a creationist's words. Verified: full.
The argument as its advocates state it
Theobald, 29+ Evidences for Macroevolution, sections 2 and 4 (theobald-2012c, theobald-2012d). Used. The clearest and most-cited online statement of the shared-marker argument, with explicit falsification criteria, which is what makes it usable. Prediction 2.3 gives the vitamin C case; 4.3, 4.4 and 4.5 give transposons, redundant pseudogenes and endogenous retroviruses. Two claims from 4.3 are load-bearing for the article: that transposons "insert fairly randomly," and that "deletions are never clean and usually part of the transposon sequence remains." The second is directly contradicted by van de Lagemaat et al. 2005, which is the sharpest engagement available anywhere in this literature.
Hillis 1999, PNAS 96:9979 (hillis-1999). Used heavily. The paper that named the "perfect character." It is far more balanced than its title, and it does four jobs at once: it reports the clean 20-insertion artiodactyl result that placed hippos with whales; it quotes Nikaido et al.'s claim that homoplasy is "for all practical purposes, zero" and that "there is no need for statistical analysis"; it observes that the same claim has been made and refuted for DNA–DNA hybridization, genome rearrangements and indels; and it gives the lineage-sorting probability bound (up to 0.3 for a single locus even when speciation events are separated by N generations). If only one source is read on this topic, it should be this one.
Salem et al. 2003, PNAS 100:12787 (salem-2003). Used. The steelman dataset: 133 Alu loci settling the human–chimpanzee–gorilla trichotomy, with exactly one conflicting locus (Ye5AH137, human + gorilla, absent in both chimpanzees), attributed to lineage sorting. Cited both for the strength of the result and for the conflicting locus.
Hedges et al. 2004, Genome Research 14:1068 (hedges-2004). Used. Supplies the mirror-image conflict (CS12, gorilla + chimpanzee, absent in human), the 10%-of-the-human-genome figure for Alu, and the twofold excess of recent human Alu insertions over chimpanzee.
The measured failure rates — the numbers that decide the question
Ray, Xing, Salem & Batzer 2006, Systematic Biology 55:928 (ray-2006). Used heavily. The single most important source for anyone arguing this topic in either direction, because it is the only one that reports rates across the whole primate literature. Catalogues four confounders across 11,000+ loci; finds precise-parallel insertion at 0.0005 events/insertion; re-audits Bashir et al. 2005 down from 23 claimed homoplasies to 13 among ~2,500 loci; and argues that precise excision is self-limiting because it needs intact flanking direct repeats. The title is a deliberate downgrade of Hillis's.
Han et al. 2011, Systematic Biology 60:375 (han-2011). Used heavily. The direct test. 66 TE insertions across 34 avian introns; two strictly homoplastic with hemiplasy ruled out; six insertion hot spots; verdict that TEs "should not be viewed as perfect characters exempt from homoplasy." And, cutting the other way, retention index 0.97 against 0.52–0.58 for sequence data on the same tree. Both halves are cited in the article, deliberately. This paper was also listed in agents/memory/phylogenetic-congruence.md as one of the eight gaps most likely to move the §5a article; obtaining it closes that gap.
Braun et al. 2011, BMC Evolutionary Biology 11:141 (braun-2011). Used. Same team, microinversions instead of insertions; same verdict; two hotspot loci; hemiplasy explicitly ruled out as the explanation.
Cantrell et al. 2001, Genetics 158:769 (cantrell-2001). Used heavily. Two SINE insertion hot spots in mys-9, each with two independent insertions at identical sites, distinguishable only by sequencing. Supplies the clearest single statement of the caution: "SINE insertions are not homoplasy-free phylogenetic markers."
van de Lagemaat et al. 2005, Genome Research 15:1243 (lagemaat-2005). Used heavily. Precise excision: deleted loci "indistinguishable from pre-integration sites"; 0.5–1% of apparent human–chimpanzee retroelement insertion differences are actually deletions; 19% of 200–500 bp deletions since the split are flanked by identical repeats of ≥10 bp. Refutes Theobald's stated exception.
Belshaw et al. 2007, Journal of Virology 81:9437 (belshaw-2007). Used. The equivalent problem for ERVs, which collapse to solo LTRs by LTR–LTR recombination. Rate ~200× higher for recent than for pre-divergence HERV-K(HML2) loci. Establishes the presence/absence asymmetry for ERV data specifically.
Mitchell et al. 2004, PLoS Biology 2:e234 (mitchell-2004). Used, narrowly. Family-specific integration-site preferences for ASLV, HIV and MLV. Cited only for the premise-level point that integration is non-uniform, not for any homoplasy claim, because the paper makes none.
Guiblet et al. 2021, Nucleic Acids Research 49:1497 (guiblet-2021). Used. ~13% of the human genome folds into non-B structures with elevated, structure-specific substitution frequencies. This is the peer-reviewed source behind a ScienceDaily item that the project research notes had been citing; the press release should be replaced with this everywhere.
Wang et al. 2007, Genome Research 17:1186 (wang-2007). Used (added 2026-10-27). Measurement: 40,569 HIV integration sites mapped in a human T-cell line after fifty independent infections; 41 pairs in opposite orientations at exactly the same base pair (p. 1188). That implies an effective pool of 1 to 2 x 10^7 sites, a few hundred times smaller than a uniform genome (see the article's collision-pool note). Direction: complicates. It kills the "one chance in three billion" version of the argument, and at its strength explains about 1% of shared human-chimpanzee ERV positions. One virus in one cultured cell line, not a germline element. Verified: full.
Belshaw et al. 2004, PNAS 101:4894 (belshaw-2004) and Chimpanzee Sequencing and Analysis Consortium 2005, Nature 437:69 (chimpanzee-consortium-2005). Used (added 2026-10-27) for the denominators of the collision arithmetic: 98,000 human ERV elements and fragments (Belshaw p. 4894, a HERVd figure), and 361 lineage-specific ERV insertions in the aligned genomes (Consortium Table 2, p. 75). Verified: full.
Hemiplasy — the largest effect, and the one most often misquoted
Avise & Robinson 2008, Systematic Biology 57:503 (avise-2008). Used. Names the phenomenon. Necessary to the article's argument because conflating hemiplasy with homoplasy is the commonest error on both sides of this debate.
Guerrero & Hahn 2018, PNAS 115:12787 (guerrero-2018). Used. Quantifies it: hemiplasy is as likely as or likelier than homoplasy across a wide parameter range, even at low discordance.
Churakov et al. 2009, Genome Research 19:868 (churakov-2009). Used heavily. The cleanest single illustration in the literature: 5 vs 9 vs 8 retroposon insertions supporting three mutually exclusive placental roots, from ~2 million screened loci, concluding a "soft" polytomy.
Doronina et al. 2017, Genome Research 27:997 (doronina-2017). Used heavily. Same result one node down, with four-way conflict: 14 / 8 / 9 / 12 markers for four mutually exclusive basal orders. "All possible evolutionary scenarios received considerable retrotransposon support."
Suh, Smeds & Ellegren 2015, PLoS Biology 13:e1002224 (suh-2015). Used, with a warning attached. 745 of 2,118 retrotransposon markers incongruent; 35% / 34% of autosomal and Z loci; "may neither resemble a fully bifurcating tree nor are they resolvable as such." The article flags explicitly that the authors state their markers "contain negligible homoplasy," so this cannot be cited against shared ancestry.
Kuritzin et al. 2016, PLoS Computational Biology 12:e1004812 (kuritzin-2016). Used, one mention. The statistical machinery the retroposon community built because their markers conflict.
Hallström & Janke 2010, MBE 27:2804 (hallstrom-2010). Used, one mention, for the title and the general finding. Its detailed argument is not worked into the text; it would be the natural next source if this section is expanded.
Pseudogenes that mislead
Thalmann et al. 2004, Molecular Ecology 13:321 (thalmann-2004). Used. Numts in gorillas render standard mtDNA analysis "impossible." The authenticity control was long-range PCR, which cannot amplify a nuclear insert — a clean methodological check, which is why the result stands.
Song et al. 2008, PNAS 105:13486 (song-2008). Used. Numts inflate barcoding species counts in grasshoppers and crayfish; filtering helps but does not fully work; numt distribution is lineage-specific and unpredictable a priori.
Álvarez & Wendel 2003, Mol Phyl Evol 29:417 (alvarez-2003). Used heavily. The pseudogene-specific centrepiece. 66% of plant phylogenies over five years used ITS, 34% used nothing else; rDNA pseudogenes and paralogy make homoplasy higher in ITS than in other data sets; recommends single- or low-copy nuclear genes instead. Heavily cited in botany and almost unknown in apologetics on either side.
Thomson et al. 2014, PNAS 111:E3659 (thomson-2014). Used, one mention. miRNA presence/absence data were promoted as an ideal marker and do not behave. Included to show the pattern is general across marker classes, not specific to transposons.
Hormozdiari et al. 2013, PNAS 110:13457 (hormozdiari-2013, new to the library 2026-10-27). Used. Measurement: about 49,000 non-reference great ape Alu and L1 insertions; among the 72 best genomes, 187 insertions conflict with the great ape tree, and 157 (84%) are flanked by SNPs with the same conflicting tree, the signature of lineage sorting (p. 13458). The 30 that fail are the honest residue; the paper assumes a 10% false-genotyping rate. Direction: supports the mainstream reading (lineage sorting is tested, not assumed), while fixing an upper bound on unexplained conflict. Verified: full (publisher PDF via the Bilkent repository).
Doronina et al. 2017, flanking-sequence test (doronina-2017, already cited for its counts). p. 1000: for nine of ten groupings the flanking DNA gives the same relationships as the insertions; "ILS, rather than homoplasy, is the predominant reason". Two-column page, read with a right-column crop. Verified: full.
Function — whether "mistake" is the right word
Pink et al. 2011, RNA 17:792 (pink-2011). Used. Review of transcribed pseudogenes, siRNA processing, microRNA decoys.
Chuong, Elde & Feschotte 2016, Science 351:1083 (chuong-2016). Used. The strongest function result available: CRISPR deletion of ERV-derived enhancers impairs interferon-induced gene expression, including AIM2 inflammasome activation. Note carefully that the paper describes these as lineage-specific ERVs dispersed independently — the article says so, because otherwise the citation would be misleading.
ENCODE 2012, Graur et al. 2013, Doolittle 2013 (encode-2012, graur-2013, doolittle-2013). Used, one mention each, to point at the live dispute over what fraction of the genome is functional without relitigating it here. That fight belongs in the functional-DNA articles, which already carry it.
Non-random mutation
Cunningham et al. 1997, MBE 14:113 (cunningham-1997). Used. Already in the library from the phylogenetic-congruence work. Nine independent T7 deletions, seven with identical breakpoints, from a 13-bp repeat; explicit warning against giving undue phylogenetic weight to indels.
Shapiro 2011, Evolution: A View from the 21st Century, p. 82. Used, one sentence. Not held in the source library; located through Google Books search-inside. Cited only for the general claim that no genome-change operator is truly random. Should be replaced with a peer-reviewed source or dropped — it is the weakest citation on the page.
Crow 2000, Nature Reviews Genetics 1:40. Used, but UNVERIFIED. The achondroplasia figures (153 of 154 cases at codon 1138; 150 of 153 the same G→A transition) are carried from the project research notes. Closed at Nature with no open-access location. Flagged in the article footnote and in quotes. This is the best single illustration of recurrent identical mutation available and it is worth the effort to obtain properly.
Case studies
Drouin, Godin & Pagé 2011, Current Genomics 12:371 (drouin-2011). Used heavily. The vitamin C review. Anthropoid primates and guinea pigs lost different exons, so those losses are independent; bat GLO has been reactivated; Passeriformes require three or four reacquisitions on the accepted phylogeny. Note for future work: the project research notes cite "PMC3145266" for the bat reactivation quotation as though it were Cui et al.; PMC3145266 is this Drouin review, so the attribution in the notes should be corrected.
Brawand, Wahli & Kaessmann 2008, PLoS Biology 6:e63 (brawand-2008). Used. Vitellogenin pseudogenes; multiple independent inactivation events including VIT1 in therians and monotremes.
Meredith et al. 2009, PLoS Genetics 5:e1000634 (meredith-2009). Used. ENAM decay tracking enamel loss in the fossil record. Cited as a genuine strength of the evolutionary case, with the caveat that convergent loss under relaxed selection is expected, so the nesting of particular lesions is what carries the argument.
Zhang et al. 2010, Genome Biology 11:R26 (zhang-2010b). Used. 76 unitary pseudogenes in the human lineage; 11 polymorphic; two are pseudogenes in non-human primates, implying resurrection in the human lineage.
Oda et al. 2002, MBE 19:640 (oda-2002). Used, verified from the PDF. The urate oxidase codon-107 stop in gorilla, orangutan and gibbon but not human or chimpanzee, which the authors say "requires multiple origins" (p. 643); four of the five CGA codons converted to TGA (p. 643); and the Appendix II legend (p. 648) that codon 107 is still polymorphic in orangutans. Folios anchored on the running heads at PDF p. 4 = 643 and PDF p. 9 = 648. Direction: supports the creationist reply on point lesions. With Inai 2003, one of the two best documented cases of shared damage that is not shared inheritance.
Inai, Ohta & Nishikimi 2003, J Nutr Sci Vitaminol 49:315 (inai-2003). Used (added 2026-10-27). Measurement: human and guinea pig GULO remnants, lost independently, compared with rat over 647 nucleotides; the identical substitution at 47 of the 129 human-changed positions (36%), p = 1.84 x 10^-12 under uniform rates (Fig. 2 caption, p. 317; the text layer loses the superscript minus, which the next sentence fixes). Direction: supports the creationist reply that shared point changes in a broken gene can arise independently. Verified: full. Limit: point substitutions only; it does not bear on shared exon deletions or insertion positions.
Lavialle et al. 2013, Phil Trans R Soc B 368:20120507 (lavialle-2013). Used (added 2026-10-27). The syncytins as the control case: primate and muroid syncytins "are not syntenic", so independent captures sit at different loci (p. 6). Direction: supports common descent for the shared human-chimpanzee ERVs, against the created-in-place reading. Verified: full.
Sought and not obtained
- Rokas & Holland 2000, TREE 15:454, "Rare genomic changes as a tool for phylogenetics." The founding paper for the whole marker class. Closed at Elsevier, no OA location. Its own caveats section would be the ideal opening quotation for this article. Cited only at second hand, through Han et al. 2011's summary, and the article says so.
- Bensasson, Zhang, Hartl & Hewitt 2001, TREE 16:314, "Mitochondrial pseudogenes: evolution's misplaced witnesses." Perfect title for the numts section. Closed at Elsevier, no OA location. Not used — Thalmann 2004 and Song 2008 cover the same ground and are on disk.
- Ohta & Nishikimi 1999, BBA 1472:408, the primate GULO sequences. Closed at Elsevier. Not used directly; the article reaches the same material through Theobald's summary and Drouin et al. 2011.
- Crow 2000: used but unverified, and flagged. (Oda et al. 2002 was obtained and verified; see Case studies.)
- Cui et al. 2011, MBE, "Progressive pseudogenization: vitamin C synthesis and its loss in bats." Not fetched; the Drouin review reports its result and is open access. Worth obtaining if the vitamin C section is expanded.
- Springer & Gatesy 2016, "The gene tree delusion," Mol Phyl Evol 94:1. Abstract only, closed at Elsevier. Not used here; it is already noted as an outstanding gap for the phylogenetic-congruence article.
- Nikaido, Rooney & Okada 1999, PNAS 96:10261. The companion paper Hillis is commenting on. Not fetched — its two load-bearing claims are quoted inside Hillis, which is on disk, and quoting them through Hillis is honest because Hillis gives page numbers for both. Should be obtained if the artiodactyl case is expanded.
Deliberately left out
- Jeffrey Tomkins's GULO work (
tomkins-2014). Cited once (2026-10-27), only as the creationist statement that the GULO exon losses were independent in each species. Its similarity figures and argument are tested in full in the sibling articlehuman-chimp-ervs-and-pseudogenes, which finds the figures came from a BLASTN method Tomkins himself later documented as faulty (tomkins-2015b). Not repeated here. - Richard Dawkins's video remarks on pseudogenes recorded in the research notes. A transcribed spoken remark is a weak citation when the same claim is available in Theobald's written and versioned form.
- The human chromosome 2 fusion. A different argument with a different structure; it belongs in the human-and-chimpanzee-genomes article, not here.
- Endogenous retrovirus counts between humans and chimpanzees. Filled 2026-10-27:
belshaw-2004(98,000 human ERV elements and fragments, p. 4894, a HERVd figure) andchimpanzee-consortium-2005Table 2 (361 lineage-specific ERV insertions in the aligned genomes, p. 75). Theobald's "at least seven" is still not used. - Considered 2026-10-27 and not used. Gilad et al. 2003 olfactory receptor table (
gilad-2003): the sibling article counts identical disruptions from Table 1 by script with a judgement call on undetermined cells; this article links there rather than repeat the count. Jeanson's Replacing Darwin (jeanson-2017d, YAML only) and Cserhati (cserhati-2019d, on snake legs): nothing on shared insertions on disk. Carter 2009 (carter-2009c) and Sternberg 2010 (sternberg-2010c): general function-of-junk-DNA claims already covered by Pink 2011, Chuong 2016 and the ENCODE dispute. Terborg 2009 (terborg-2009_design-of-life-part-4-variation-inducing-genetic-elements): the VIGE proposal is cited through the later 2013 paper, which states it directly.