How Similar Are Human and Chimp DNA?

"This difference corresponds to ~3% of both genomes and dwarfs the 1.23% difference resulting from nucleotide substitutions" — The Chimpanzee Sequencing and Analysis Consortium, Nature, 2005¹

"The 'unaligned' chimp DNA is not too different; it's too similar." — Todd Wood, young-earth creationist biologist, 2011²

"There isn't one single way to express the genetic distance between two complicated living organisms." — Pascal Gagneux, quoted in Science, 2007³

Summary

How much of human DNA matches chimpanzee DNA?  Textbooks say 98.8 %.  Creationist papers say 84 %, 80 %, or 70 %.  A 2025 Nature paper is now quoted for "85 % identical" and "15 % different".  Most of these numbers are correct arithmetic.  They disagree because each divides a different numerator by a different denominator.  The two exceptions, 80 % and 11 %, subtract the same missing DNA twice.  This article gives the four definitions, the number each produces, and the one thing almost nobody reports:  what the same method gives when it compares two humans.

What the creationist critics got right.

1. The 98.8 % slogan overstated. It counted only single-letter substitutions inside the 2.4 billion bases that aligned one-to-one. The 2005 consortium said so itself. Insertions and deletions add about 3 %, and in the paper's own words that "dwarfs" the substitution figure.¹ Roy Britten had published 95 % in PNAS three years earlier.⁴
2. Richard Buggs' 84–85 % is honest arithmetic. Count every base of one human genome, charge everything that did not match one-to-one as different, and 84.38 % in 2018 or 85.1 % in 2025 is what comes out.⁵⁶
3. The old chimpanzee assemblies were human-guided, and the mainstream says so in print. Zev Kronenberg and colleagues wrote in Science in 2018 that this bias "has effectively 'humanized' other ape genome assemblies".⁷

What the creationist critics got wrong.

4. The 2025 data carry their own control, and every creationist commentary skipped it. In the table Casey Luskin cites for "16.11 percent" different, a second human genome scores 88.80 % against the human reference and chimpanzee scores 83.89 %.⁸ So 11.2 of Luskin's 16.1 points also separate two humans, and this method cannot say how much of them is species difference. What it can charge to species is the 4.9-point excess, which is a floor, not a ceiling. The same paper's count of human-specific structurally divergent DNA, which requires both human genomes to agree, comes to about 5 % of the genome.⁸
5. Gap divergence fails its own control. The two chromosome copies inside one gorilla differ by 13.8 % gap divergence. Human and chimpanzee differ by 13.3 %.⁸ A metric on which one animal differs from itself by more than two species differ from each other is not measuring species difference.
6. The two metrics that pass their controls give small numbers and put the apes in order. Single-nucleotide divergence: human-chimp 1.6 %, within one human 0.16 %, a 7.4-fold signal.⁸ Structural variants over 50 bases: human-chimp 2.65 % of the genome, human-human 0.37 %, a 7-fold signal.⁷ Both order the apes chimp, then gorilla, then orangutan.
7. "Unalignable" does not mean "different", and a young-earth creationist proved it. Todd Wood took the 157 Mb of chimpanzee DNA that would not align to human and ran Megablast on it. 42,570 of 48,975 fragments matched human at a median 97.4 % identity. Genuinely absent from human: 25.3 Mb, or 0.84 %.²
8. The 80 % and 11 % figures are double subtractions. Jeffrey Tomkins in 2018 subtracted Buggs' 4.06 % unaligned fraction from a number that already had it removed.⁹ Nathaniel Jeanson in 2017 did the same with Dennis Venema's 95 %.¹⁰
9. The 70 % figure was withdrawn by its own author. Tomkins found a BLASTN bug, published it, named his own affected paper, and revised the figure to 88 %.¹¹ The 88 % is mean identity inside alignments, with unalignable DNA left out, so it is not a whole-genome figure.

What a percentage cannot do.  It cannot decide ancestry in either direction.  Two random DNA sequences match at about 25 % by chance, so the scale runs from 25 to 100, not from 0 to 100.¹²  A high number is less impressive than it sounds, and a low number is less damaging.  What carries ancestry information is the pattern across species, not the size of the number.  Every metric that passes its control orders the apes chimp < gorilla < orangutan.  That is what common descent predicts.  Design accounts of nested similarity exist, Walter ReMine's in 1993 and Winston Ewert's in 2018, but neither yet predicts how far the shared DNA differs from each ape, and the young-earth biologist Todd Wood wrote in 2006 that "Creation biology needs an explanation of the pattern of similarities".¹³  A sequence percentage also does not measure how different the animals are:  in a 2007 survey, 6 to 8 % of the exons measured were spliced differently in the same tissues of the two species.¹⁴

Part 1: Four definitions, four numbers, one table

The word "similarity" hides a choice about what to divide by.  Four choices are in use.  The table gives each one, the human-versus-chimpanzee number it yields, and the number the same method yields when it compares two human genomes.  That last column is the control.  Without it a percentage says nothing, because no alignment method returns 100 % even for the same species.

Two other quantities circulate.  The alignable fraction is how much of one genome can be placed against the other at all;  Tomkins' 70 % and 88 % are versions of it, and it depends on the software more than the genomes.  The gene-complement difference is a count of genes, not bases;  Jeffery Demuth and colleagues reported 6.4 % in 2006.¹⁶

Definitions 1 and 4 separate species from within-species noise by about seven times.  Definition 3 separates them by 1.4 to 2.5 times, and on one of its sub-measures a single gorilla differs from itself by more than human differs from chimpanzee.  Every low creationist figure, 70 %, 80 %, 84 %, 85 %, comes from definition 3 or the alignable fraction.

Part 2: Where 98.8 % came from, and who published the corrections

The 2005 draft.  The Chimpanzee Sequencing and Analysis Consortium published the first chimpanzee genome in Nature on 1 September 2005.  It was a 3.6-fold shotgun draft of one male West African chimpanzee named Clint, assembled with a program called ARACHNE that "made limited use of human genome sequence (NCBI build 34) to facilitate and confirm contig linking".¹  The comparison covered "~2.4 gigabases (Gb) of high-quality sequence" in best-reciprocal alignments.¹  Inside that 2.4 Gb, "Single-nucleotide substitutions occur at a mean rate of 1.23% between copies of the human and chimpanzee genome, with 1.06% or less corresponding to fixed divergence between the species."¹  One hundred minus 1.23 is 98.77.  That is the slogan.

The same paper said the slogan was incomplete.  Two pages later:  "we estimate that the human and chimpanzee genomes each contain 40–45 Mb of species-specific euchromatic sequence, and the indel differences between the genomes thus total ~90 Mb.  This difference corresponds to ~3% of both genomes and dwarfs the 1.23% difference resulting from nucleotide substitutions".¹  The abstract counted "approximately thirty-five million single-nucleotide changes, five million insertion/deletion events, and various chromosomal rearrangements."¹  So the indel-inclusive figure, about 96 %, is on page 73 of the paper that produced 98.8 %.

The 1.06 % is the part nobody quotes.  About a fifth of the 1.23 % is polymorphism, positions where humans differ from each other or chimpanzees do.  The consortium's fixed divergence is "~1.06% or less".¹  That is the first within-species correction in this literature, and it is in the 2005 paper.

Britten, three years earlier.  Roy Britten aligned five chimpanzee BAC clones, 779 kb in total, against human with a purpose-written aligner that counted the full length of every gap.  His abstract:  "the old saw that we share 98.5% of our DNA sequence with chimpanzee is probably in error.  For this sample, a better estimate would be that 95% of the base pairs are exactly shared between chimpanzee and human DNA.  In this sample of 779 kb, the divergence due to base substitution is 1.4%, and there is an additional 3.4% difference due to the presence of indels."⁴  Five clones are not a genome, and Britten said so.⁴  Later work confirmed it.

Varki, in the same year as the consortium.  Ajit Varki was a consortium co-author.  Three months after the Nature paper he put this in the abstract of a Genome Research perspective:  "The difference between the two genomes is actually not ~1%, but ~4%—comprising ~35 million single nucleotide differences and ~90 Mb of insertions and deletions."¹⁵  He also gave the number that runs the other way:  "orthologous proteins are still extremely similar, with almost a third being identical, and the typical protein differing only by two amino acids between human and chimpanzees.  Thus, the oft-repeated '<1% difference' still applies to amino acid sequences."¹⁵

Cohen and Preuss carried it into print.  Jon Cohen's 2007 Science news piece, "Relative Differences: The Myth of 1%", relayed the consortium's arithmetic and its 3 % indel figure to a general readership.³  Todd Preuss wrote in PNAS in 2012 that "the total DNA sequence similarity between humans and chimpanzees is not 98% to 99%, but instead closer to 95% to 96%, although the rearrangements are so extensive as to render one-dimensional comparisons overly simplistic."¹⁷  Creationist quotations of Preuss usually stop before "closer to 95% to 96%".  He is endorsing definition 2, the consortium's own figure.  He is not endorsing 84 %.

So the "suppression" charge fails.  Tomkins and Jerry Bergman wrote in 2012 that "highly selective data-omission processes, driven by Darwinian dogma" produced the 98 % figure.¹⁸  Every number they use to correct it comes from the papers they accuse:  the 3 % is on page 73 of the Nature paper, the 4 % is in a Genome Research abstract by a consortium author, the 95 % is in PNAS from 2002.  A slogan outran the papers.  The papers hid nothing.

Part 3: The 84 % family, and why the same alignment also gives 96.7 %

Buggs 2018.  Richard Buggs is Professor of Evolutionary Genomics at Queen Mary University of London.  He is a theist who accepts common ancestry, and he is the source of the 84 % that the creationist literature now runs on.  In July 2018 he downloaded the UCSC LASTZ alignment of human hg38 against chimpanzee panTro6 and sorted every base of the human genome into bins:  "4.06% had no alignment to the chimp assembly / 5.18% was in CNVs relative to chimp / 1.12% differed due to SNPs in the one-to-one best aligned regions / 0.28% differed due to indels within the one-to-one best aligned regions … The percentage of nucleotides in the human genome that had one-to-one exact matches in the chimpanzee genome was 84.38%".⁵

The arithmetic has a component nobody quotes.  Add his four bins:  4.06 + 5.18 + 1.12 + 0.28 = 10.64.  One hundred minus 10.64 is 89.36, not 84.38.  The missing five points are DNA that had not been sequenced:  "roughly 5% of the human genome has not been fully assembled yet, so the best we can do for that 5% is predict how similar it will be to the chimpanzee genome."⁵  Buggs counted it as non-identical.  About a sixth of his 2018 "difference" was human DNA nobody had read.  The T2T-CHM13 assembly has since read it.

Read the bins and 98.6 % reappears.  SNPs plus indels inside the one-to-one aligned regions total 1.40 %, the 2005 consortium's number from a different aligner thirteen years later.  The largest bin is copy-number variation at 5.18 %.  A duplicated segment is nearly identical to its source but has no one-to-one partner, so it is charged as different.  Buggs' analysis confirms definition 1 and puts the rest into copy number, alignment failure and unread DNA.

Buggs' own retraction.  He had predicted in 2008 that the true figure would be near 70 %.  In 2018 he wrote:  "This prediction is not borne out by the more recent data above."  The main reason, he said, was that "I thought that the 2005 chimpanzee genome assembly was far more complete than it actually was."⁵  His verdict on the 95 % that Dennis Venema had called "genome wide":  "I would accept 95% as a prediction, but not as a statement of established fact."⁵

Seaman and Buggs 2020:  same alignment, 96.66 %.  Two years later Josiah Seaman and Buggs published a visualisation tool in Frontiers in Genetics and reported statistics for the same UCSC hg38-versus-panTro6 chain alignment.  Table 2, after removing 72,352,500 bp of manually annotated centromeres from a 2,937,639,113 bp reference length, leaving 2,865,286,613 bp:  "Alignment length/Reference length minus centromeres 97.98% / Identical bases/Reference length minus centromeres 96.66%".¹⁹  Unaligned human sequence stays in that denominator.  The 2.02 % shortfall from 100 to 97.98 is the unaligned DNA, and the 1.32 % drop from 97.98 to 96.66 is the mismatch inside alignments.

The reconciliation, computed for this article.  Start from 84.38 %.  Add back the three things the 2020 paper does not charge as difference:  about 5.00 % unassembled human N-bases, 5.18 % copy-number-variant regions, and 2.46 % centromeres.  84.38 + 5.00 + 5.18 + 2.46 = 97.02 %, against the 96.66 % that Seaman and Buggs report.  Same alignment file, same laboratory, twelve points apart, and the whole gap is the denominator.

The creationist concession.  Robert Carter of Creation Ministries International, who holds a PhD in marine biology and genetics, phoned Seaman to ask about his methods.  Carter's summary:  "To reach the higher percentage similarity, they cut out everything humans and chimps do not share, including the centromeres, telomeres, copy number variations of many annotated genes, and hundreds of thousands of small insertions and deletions that must be included to align the two genomes.  This 'apples to apples' comparison is fine, as long as everybody acknowledges that the true similarity is necessarily less than 96.6%."²⁰  That is a defensible position.  It is also compatible with every mainstream figure in this article, which is why Carter immediately changes the subject to the mutation rate.  His description is right about centromeres and wrong about unaligned regions, which the 2020 paper keeps in the denominator.

Part 4: The 2025 telomere-to-telomere genomes carry their own control

What changed in 2025.  On 8 May 2025 DongAhn Yoo, Arang Rhie and about 140 co-authors published "Complete sequencing of ape genomes" in Nature.  They assembled chimpanzee, bonobo, gorilla, both orangutans and siamang from long reads, both chromosome copies of each animal resolved, 215 gapless chromosomes, and no step that mapped ape reads onto the human reference.  The scaffold objection does not apply to it.

The headline that creationists quoted.  "Overall, sequence comparisons among the complete ape genomes revealed greater divergence than previously estimated (Supplementary Notes III–IV).  Indeed, 12.5–27.3% of an ape genome failed to align or was inconsistent with a simple one-to-one alignment, thereby introducing gaps.  Gap divergence showed a 5-fold to 15-fold difference in the number of affected megabases when compared to single-nucleotide variants, which was due to rapidly evolving and structurally variant regions of the genome as well as technical limitations of alignment in repetitive regions."⁸  Note the last clause.  The authors say in the same sentence that gap divergence includes alignment failure.

Two quantities, defined by the authors.  "SNV divergence is defined as the fraction of positions in the target haplotype where the two haplotypes are in different nucleotide states.  Gap divergence is defined as the fraction of positions in the target haplotype that are not aligned to the other haplotype, which could be due to biological processes (e.g., gene loss/gain and insertions/deletions), missing data, or technical problems (e.g., alignment failure due to SVs, repetitive elements, etc.)."⁸

Buggs' 2025 sum.  Buggs read the human-versus-chimpanzee values off Supplementary Figures III.11 and III.12 and added them:  "the authors report gap divergence of 13.3% and single nucleotide variant divergence of 1.6% … Adding the average gap divergence and average single nucleotide variant divergence together gives a total difference of 14.9%.  Thus, as I understand it, for the latest assemblies, 85.1 % of the nucleotides in the human genome have one-to-one exact matches in the chimpanzee genome."⁶  The arithmetic is right and the figures are correctly read.

Luskin's "15 percent different".  Casey Luskin of the Discovery Institute went to Supplementary Tables III.19 and III.20, which report how much of the human reference each ape genome covers under a multi-species alignment:  "only 84.95 percent of the nucleotides in the chimp genome had an identical 1:1 correspondence with the human genome.  So that is basically a haploid genomic difference of 15.05 percent between humans and chimps."²¹  From the diploid table:  "the percent difference between humans and chimps grew to 16.11 percent, because only 83.89 percent of nucleotides in the diploid autosomal chimp genome showed an identical 1:1 correspondence to the human genome."²¹  He added that these findings "were buried deep in the 174-page-long Supplementary Data".²¹  Both of those figure sentences reverse the tables' direction.  The supplement introduces them with "The coverage on T2T-CHM13/hs1 is described in these tables", and each is headed "Alignment coverage of T2T-CHM13 (hs1)".  So 84.95 % is the share of the human genome matched one-to-one in chimpanzee, not the share of the chimpanzee genome.⁸  The figures are transcribed right.  The description is backwards.

The control rows are the first rows of the tables he cites.  Supplementary Table III.19 compares seven genomes against the T2T-CHM13 human reference.  Its first data row is GRCh38, the standard human reference, at 86.96 % identical one-to-one.  Chimpanzee is 84.95 %.⁸  Supplementary Table III.20 adds HG002, a fully telomere-to-telomere diploid human genome.  HG002 scores 88.80 %.  GRCh38 scores 87.87 %.  Chimpanzee scores 83.89 %.⁸  By Luskin's method, two complete human genomes are 11.2 % different from each other.  The shortfall that can be charged to species is 88.80 minus 83.89, which is 4.91 points.  On the haploid table it is 2.01 points.

The control sets a floor, not a ceiling.  The 11.2 % that fails between two humans is not shown to be identical between human and chimpanzee.  It is DNA this method reads badly, and the paper says where most of such DNA sits:  "centromeres, acrocentric short arms and subterminal heterochromatic caps, but also numerous gene-rich SD regions".⁸  Those are also the regions where the apes differ most.  So 4.91 points is what the table can charge to species with its own control, and on this definition the true species difference lies somewhere between 4.91 and 16.11 points.  A second measure in the same paper narrows the range from below.  The authors catalogued "structurally divergent regions", sequence that does not align or aligns below 85 % identity, and found "an average of 327 Mb of sequence (10%) per ape lineage".⁸  For the human lineage they counted only sequence "conserved between human haplotypes (i.e., CHM13 and HG002) but divergent from other primate haplotypes".  That count is 165.68 Mb, about 5.3 % of a 3.1 Gb genome, of which roughly 50 Mb is ordinary euchromatin and the rest centromeres, acrocentric arms and other heterochromatin.⁸  Two methods with a human control built in land near 5 %.  Neither can see a species difference inside DNA that already differs between the two humans, so both are floors.  The chimpanzee lineage carries its own 298.58 Mb, plus 56.23 Mb shared with the bonobo, much of it in subterminal caps that human chromosomes lack.⁸  That DNA sits in the chimpanzee genome, and a percentage that divides by the human genome never counts it.

The X and the Y.  Luskin:  "only 79.88 percent of nucleotides in the chimp X chromosome showed a 1:1 correspondence with the human X chromosome.  That amounts to a difference of 20.12 percent!"²¹  The human HG002 maternal X, in the same table, scores 85.34 %.  Species shortfall:  5.46 points.  For the Y:  "only 4.32 percent of nucleotides in the chimp Y chromosome had a 1:1 correspondence with the human Y chromosome.  That's not a typo — that's a difference of 95.68 percent!"²¹  In the same table a real human Y, HG002 paternal, scores 27.02 %, and GRCh38's human Y scores 8.87 %.  By this method two human Y chromosomes are 73 % different.  In Supplementary Table III.17, under a different aligner, the same HG002 Y scores 99.94 %.  The Y figures measure how one aligner handles palindromic repeats.  They do not measure the chromosomes.

Gap divergence fails its control.  Supplementary Figure III.12 reports mean autosomal gap divergence for every pair, including each animal's two chromosome copies against each other.  Human versus chimpanzee:  13.30 %.  Within one human:  3.48 %.  Within one chimpanzee:  8.09 %.  Within one gorilla:  13.82 %.⁸  The two haplotypes of a single gorilla differ by more gap divergence than a human differs from a chimpanzee.  The metric is 2.3 times its within-species floor, and it does not order the apes:  with human as the target genome in both cases, human-orangutan comes out at 15.47 %, below human-gorilla at 17.88 %.  With gorilla as the target, gorilla-versus-human rises to 27.32 %, about 9 points higher.  Supplementary Figure V.15 counts roughly 300 Mb of gorilla-specific subterminal cap sequence, which human DNA cannot match, enough to account for most of that asymmetry.⁸  Gap divergence is mostly structural polymorphism and alignment failure.  Adding it to SNV divergence, as Buggs does, produces a sum on which one chimpanzee is 91.6 % identical to itself and one human is 96.4 % identical to itself.

SNV divergence passes its control.  Supplementary Figure III.11 gives mean autosomal SNV divergence.  Human versus chimpanzee:  1.60 %.  Human versus gorilla:  1.99 %.  Human versus orangutan:  3.64 %.  Within one human:  0.16 %.  Within one chimpanzee:  0.27 %.⁸  The species signal is 7.4 times the mean within-species floor, and the apes come out in order.  The 1.6 % is higher than 2005's 1.23 % because the T2T assemblies reach the fast-mutating repeats the 2005 alignment could not.  More complete data showed more SNV difference, as creationists predicted, by a fraction of a point rather than by fifteen.

A typo in the supplement that both sides should avoid.  Supplementary Information page 25 states human-chimpanzee autosomal SNV divergence as "0.15-0.16%".⁸  Figure III.11, two pages later, plots it at 0.0146 to 0.0160, which is 1.46 to 1.60 %.  The sentence is off by a factor of ten for every human-versus-ape pair.  Buggs cited the figure and got 1.6 %.  Anyone who cites the sentence will report a species difference equal to the within-human value.

The computation, run for this article.  The script research/divergence-controls.ts beside this article puts every headline number above next to the control from the same source:

One more number, computed here and reported nowhere else:  the raw gap-plus-SNV difference minus the mean of the two within-species sums.  Human-chimp 14.90 − 6.00 = 8.9 %.  Human-gorilla 19.87 − 9.02 = 10.9 %.  Human-orangutan 19.11 − 5.33 = 13.8 %.  The ordering is right and the values are far below the raw figures.

Part 5: The creationist numbers, one at a time

The young-earth literature on this question is mostly the work of Jeffrey Tomkins of the Institute for Creation Research, in four papers in Answers Research Journal and one review with Jerry Bergman.  Two of the four contain creditable work.  All four turn "did not align" into "is different".

2013:  70 %.  Tomkins sliced the chimpanzee assembly into fragments, of a length chosen per chromosome, and ran BLASTN against the matching human chromosome.  "The definition of similarity for each chromosome was the amount (percent) of optimally aligned chimp DNA."²²  Result:  "Genome-wide, only 70% of the chimpanzee DNA was similar to human under the most optimal sequence-slice conditions."²²  The parameter list is in the methods:  "-word_size 11, -evalue 10, -max_target_seqs 1, -dust no, -soft_masking false, -ungapped."²²  The last flag is the problem.  An ungapped alignment stops at the first insertion or deletion.  Human and chimpanzee differ by about five million indels, so in a 300-base slice one is likely, and everything past it is scored as unaligned.  What the pipeline measures is the average distance to the next indel.  The objection is visible in the parameter list itself.

2015:  the bug, and 88 %.  Tomkins then found that BLASTN versions after 2.2.25+ silently dropped queries in large batches:  "This claim potentially affects the results of past research done by this author, particularly in relation to human-chimp DNA similarity."¹¹  He re-ran the analysis with the older version and two other aligners:  "The analyses with BLASTN v 2.2.25+ and nucmer indicate that the alignable portions of the current chimpanzee genome assembly are 88% similar on average to human.  The LASTZ algorithm only returned an average overall similarity of 73%."¹¹  He found the bug that destroyed his own headline result and published it.  Note the switch of quantity:  70 % was the share of chimpanzee DNA that aligned at all, and 88 % is identity inside whatever aligned.  The reanalysis kept -ungapped, and Tomkins gives a reason:  with gapping allowed, version 2.2.25+ "resulted in a continuous loop effect returning no hits", so he ran every version ungapped to compare them on equal terms.¹¹  That is a fair reason for comparing BLAST versions, not for measuring two genomes.  His second aligner answers that better:  nucmer allows gaps, needs no slicing, and returned the same 88 %.  But both figures are mean identity inside alignments, with unalignable sequence excluded, and Tomkins reports that nucmer's aligned length came out longer than the chromosomes because its alignments overlapped.¹¹  The same paper reports three aligners on the same two genomes giving 73 % and 88 %, against 98.6 % inside Buggs' one-to-one blocks.  The spread is the software, not the genomes.

2016:  contamination.  Tomkins compared 101 batches of raw chimpanzee sequencing reads.  The early batches, which fed the 2005 draft, matched human about 7 % better than later ones.  He read this as human contamination and concluded that "the chimpanzee genome is approximately 85% identical overall to human."²³  The worry is real;  contamination in early-2000s databases is documented.  The inference has no control.  Sanger read quality improved sharply across those years, and better reads align better to everything.  The check that would decide it, running the same batches against the chimpanzee genome, is not reported.  The paper also states that "all of the short DNA sequences produced by the DNA sequencing machinery (known as trace reads) have been assembled onto the human genome—using it as a framework scaffold or reference sequence (Mikkelsen et al. 2005 …)".²³  Mikkelsen et al. 2005 is the consortium paper, and it says the assembler "made limited use of human genome sequence (NCBI build 34) to facilitate and confirm contig linking".¹  Contig sequence came from chimpanzee reads.  Human ordered and oriented them.

2018:  84 %, and the double subtraction to 80 %.  Tomkins took 18,000 contigs from a new chimpanzee assembly built without human guidance and ran BLASTN, this time with gaps allowed, against human, against the older panTro4 chimpanzee assembly, and against panTro5.  Table 1:  human 84.39 %, panTro4 91.91 %, panTro5 100.0 %.  "The main finding of significance to the issue of alleged common ancestry between humans and chimpanzees is the fact that the average alignment identity was only 84% (table 1)."⁹  He then noticed Buggs' 84.38 % and wrote:  "The identities of the alignable DNA in this study and that obtained from LASTZ alignments are both 84% and in perfect agreement.  Furthermore, neither this study nor the LASTZ matched nucleotides account for the unalignable regions of the genome between human and chimpanzee.  From the LASTZ alignment data, Buggs calculated, '4.06% had no alignment to the chimp assembly.'  If this non-alignable sequence data is combined with the known alignment data (Buggs 2018), a reasonably accurate estimate of overall human-chimpanzee genome similarity would be close to 80%".⁹

The two 84s measure opposite ends of the pipeline.  Tomkins' 84.39 % is mean identity inside BLASTN alignments, with unaligned contig excluded.  He says so himself:  "the 84.4% nucleotide identity of the alignments is not an indicator of overall genome similarity because it does not include the regions of the contigs that are so different that they are non-alignable."⁹  That last inference has a setting behind it.  His run used max_target_seqs 1, max_hsps 1, which reports one local alignment per contig.⁹  When an alignment stops at a large gap, the rest of the contig may align as a second piece that the run never reported.  So his finding that "only about one-third of each chimpanzee contig on average could be aligned to the human genome as the best hit"⁹ measures one alignment per contig, not how much of the contig is alignable.  Buggs' 84.38 % is the share of the whole human genome with a one-to-one exact match, with the 4.06 % non-aligning, 5.18 % CNV and 5 % unassembled already subtracted, in Buggs' own itemisation.  "Neither … account for the unalignable regions" is false of Buggs' figure.  Tomkins subtracted the 4.06 % a second time.  The 80 % is a double subtraction, and it is the number most often quoted from this paper.

Tomkins' own control row refutes his conclusion.  He called the panTro5 run a "strong control measure" because it returned 100 %.⁹  But panTro5 was built from the very contigs he was querying, as he states in the same paper, so finding them in it tests nothing.  The real control is the panTro4 row.  PanTro4 was assembled from reads of the same animal, Clint, and Clint's own contigs match it at only 91.91 %.  Tomkins reads this as proof that "the humanization of the chimpanzee genome was extremely severe".⁹  Taken literally, with panTro4 modelled as a mixture of chimpanzee and human sequence, 91.91 = (1 − h) × 100 + h × 84.39 gives h ≈ 0.52:  panTro4 would have to be about half human DNA.  Nobody claims that.  The plainer reading is that whatever costs 8 points between a chimpanzee and its own earlier assembly also operates in the human comparison.  Under Tomkins' own pipeline the species-attributable difference is at most 91.91 minus 84.39, about 7.5 points.

2012:  the review.  Tomkins and Bergman went through the earlier literature and, for each study, added back whatever sequence its authors had not aligned.  Their verdict:  "it is safe to conclude that human–chimp genome similarity is not more than ~87% identical, and possibly not higher than 81%."¹⁸  Half of this is legitimate.  Where a paper reported substitutions only, they added indel bases and got Britten's answer:  "Their similarity estimate on the final alignment, omitting indels and non-aligned areas, was 98.9%.  Including indels, we derived a value of 95.6% for the alignments, similar to Britten's research."¹⁸  The other half assumes what has to be shown.  On Ebersberger et al. 2002:  "One-third of the chimp sequence would not align to the human genome and was discarded. …  A more realistic sequence similarity based on the researchers' own numbers for discarded data in the alignments alone is not more than 65% (table 1)."¹⁸  That treats every fragment that failed to place uniquely against a 2001-era draft as wholly different.  Part 6 tests that.

Jeanson's 11 %.  Nathaniel Jeanson of Answers in Genesis argued in 2017 that Venema's 95 % left out the DNA that never aligned:  "in Table S8 of the same paper, the authors report that around 6% of the human DNA sequence found no match to chimpanzee DNA."¹⁰  "If you add the 6% difference … to the 5% difference … you arrive at a total of 11% genetic difference between humans and chimpanzees."¹⁰  His citation is accurate.  Table S8 of the consortium's supplement was checked for this article, and its total row gives 6 % of non-gap human bases un-aligned on placed contigs.  The inference is not.  The same table gives 20 % un-aligned for the X and 38 % for the Y, and those two chromosomes were sequenced at half the depth of the autosomes.¹  Half the coverage, triple the "difference".  The un-aligned fraction in Table S8 tracks sequencing depth, not evolutionary distance.  And Venema's sentence was "one finds that the human and chimpanzee genomes are indeed about 95% identical, genome wide".²⁴  Whether Venema was entitled to "genome wide" is a fair fight, and Buggs said he was not.  But Jeanson recast it as a within-alignment figure in order to subtract the unaligned 6 % from it again.  This is the same double subtraction as Tomkins 2018.

The scaffold complaint, tested.  The creationist objection that early chimpanzee assemblies were built on the human genome is correct as a description, and Tomkins quotes the admission accurately.  Kronenberg and colleagues in the Eichler laboratory wrote in Science in 2018:  "the higher quality human genome assemblies have often been used to guide the final stages of nonhuman genome projects, including the order and orientation of sequence contigs and, perhaps more importantly, the annotation of genes.  This bias has effectively 'humanized' other ape genome assemblies, minimizing potential structural and transcript differences observed between the species."⁷  The same paper is the test of the objection.  "We next scaffolded the chimpanzee and orangutan genomes without guidance from the human reference genome."⁷  The result:  "We observed a modest elevation in SNV divergence compared to previous genome comparisons".⁷  Yoo 2025 removed the last of the bias with an ape pangenome and got 1.46 to 1.60 % SNV divergence depending on which genome is the denominator, up from 1.23 %.⁸  The prediction was right in direction.  The predicted collapse toward 80 % did not happen.

Part 6: What "unalignable" means, tested by a creationist

Every low figure in Part 5 converts "did not align" into "is different".  In January 2011 Todd Wood, then at Bryan College's Center for Origins Research and a young-earth creationist, tested that conversion directly.  He was answering Fazale Rana of Reasons to Believe, who had argued from the unaligned fraction:  "I have to admit that this reasoning sounds good, and I began this analysis wondering if Rana might be on to something.  Unfortunately for Rana, my preliminary investigation of this unalignable DNA indicates that the alignment problem is not because the chimp DNA is too different from the human DNA.  It's too similar."²

The experiment.  Wood downloaded the chimpanzee sequence that Ensembl could not place on a chromosome:  "The 'unalignable' DNA consists of 48,975 fragments totaling 157.0 million nucleotides (Mb).  That's about 5.2% of a 3 billion nucleotide (Gb) genome."²  He ran Megablast against the human genome without requiring a unique hit.  "The Megablast results indicated that 42,570 of these 48,975 'unalignable' sequences aligned to some region of the human genome.  How similar are those fragments?  The median percent identity is 97.4%.  So they're just as similar as everything else in the chimp and human genomes.  The only difference is that you can't specify which precise region of the human genome the 'unalignable' chimp fragments should align to."²  What remained:  "my preliminary estimate is that around 25.3 Mb of chimp DNA truly has no similar match in the sequenced human genome.  That's about 0.84% of the length of the human genome."²

Why it matters.  Wood had no motive to defend a high figure and expected the opposite result.  The "unalignable" reservoir that Rana, Tomkins, Bergman, Jeanson and Luskin all treat as unmeasured difference turned out, when measured, to be 97.4 % identical and merely repetitive.  Tomkins and Bergman's 2012 review says Wood's earlier work "did little to support creationist claims"¹⁸ and does not engage this result.  No creationist reply to it has been found.  Venema, on the other side, called Wood's scholarship "entirely trustworthy".²⁴

The limits.  These are preliminary numbers from an unpublished project on a 6x-era assembly, and the full publication Wood promised has not been located.  But the point was confirmed nine years later from the other side.  Seaman and Buggs found a stretch of human protein-coding DNA that the whole-genome alignment had left uncovered:  "This result is caused by a 200bp segment of protein coding DNA in Hg38 that is not covered by the PanTro6 alignment.  BLAST searches for this sequence in Pan troglodytes returns hits at the expected 98.66% identity, so it is safe to conclude that the sequence is present but the whole genome alignment is imperfect.  In contrast, olfactory receptor 2T10 is genuinely missing from Pan troglodytes but present in Gorilla gorilla and Pongo abelii."¹⁹  Real species-specific losses exist and are small.  Alignment failure is not one of them.

Part 7: What a percentage can and cannot say about ancestry

The scale does not start at zero.  Jonathan Marks, a biological anthropologist hostile to creationism, made the point in 2002.  Two unrelated DNA sequences match at about a quarter of positions by chance, because there are four letters:  "if you compare the DNA of a human and a daffodil, the 25% mark is actually the zero mark, and since humans and daffodils do share a common ancestry, you would expect them to be generally more similar than 25%.  Say, about 35%."¹²  A 98.8 % is therefore a point most of the way up a scale whose bottom is 25.  It is not "1 % different" in the sense a listener hears.  But the same fact cuts the creationist way, and harder.  If the floor is 25 %, an 85 % is not "15 % different" in that sense either.  Marks again:  "If you were to compare human and chimpanzee structure in the context of a three-way comparison to, say, a snail, it would be patently obvious that humans and chimps are over 99% identical in practically every way you could compare all three animals."¹²  A creationist who quotes Marks against 98 % has quoted a book whose thesis is that humans and chimpanzees are obviously nearly identical and the argument is over what that means.

Which genome is the denominator.  A chimpanzee cell carries roughly a tenth more DNA than a human cell, largely in heterochromatin caps on the chromosome ends.  A percentage that divides by one genome silently decides which extra DNA to ignore, which is why gap divergence is asymmetric:  human-versus-chimp 13.30 %, chimp-versus-human 12.47 %.⁸  The choice of denominator moves the answer by nearly a point.

Gene counts are a different thing again.  Demuth and colleagues reported in 2006 that "the human genome contains 1,418 genes—6.4% of all genes—that do not have orthologs in the chimpanzee genome".¹⁶  This is a count of genes, and a duplicated gene that is 99.9 % identical to its source scores as a whole difference.  It also shrank with the data:  when chimpanzee coverage went from 4x to 6x, "the difference in gene complement between humans and chimpanzees decreases by ~1.5 fold (~2,382 in v32-39 to 1,418 in v41)".¹⁶  The authors' own floor was 2 %:  "there would still be a 2% difference in the gene complement between humans and chimps (~400 genes)."¹⁶  Nobody has re-run this on the 2025 assemblies.  Anyone quoting 6.4 % today is quoting a 6x draft.

Indels cut against the people who lean on them.  Britten, who put indels into the difference figure in the first place, also said what they are good for:  "The gaps should be useful markers for distinguishing the evolutionary relationship among closely related species, because once a deletion occurs, it would be very unlikely that the missing region sequence would be reconstructed de novo, unless it was repeat that could be copied."⁴  A deletion shared by human and chimpanzee but absent in gorilla is close to unforgeable as a marker of shared history.  The creationist papers on the percentage use indels to raise the difference figure and never confront what shared indels imply.  One young-earth creationist did.  Todd Wood, in 2006, on pseudogenes broken by the same substitution or deletion in human and chimpanzee:  "While common design could be a reasonable first step to explain similarity of functional genes, it is difficult to explain why pseudogenes with the exact same substitutions or deletions would be shared between species that did not share a common ancestor."¹³  The sibling article on shared ERVs and pseudogenes takes up the creationist replies, including mutation hotspots.  One caution runs the other way.  A single shared indel is evidence about one stretch of DNA, not about the species tree.  The gorilla genome paper of 2012 found that "In 30% of the genome, gorilla is closer to human or chimpanzee than the latter are to each other",²⁵ and the 2025 genomes put the share of the autosomes affected at an average of 39.5 %.⁸  Mainstream genetics explains this as incomplete lineage sorting:  the ancestral population carried several versions of many stretches of DNA, and different daughter species kept different ones.  On that account a discordant indel is still inherited, just sorted unevenly.  It does mean that the chimp-gorilla-orangutan order below is a genome-wide average, not a rule every locus obeys.

What carries ancestry information is the ordering.  A single percentage measures distance.  It does not measure descent, and no value of it, high or low, decides the question.  What does carry information is whether the distances nest.  Kronenberg's structural-variant counts, with two human genomes as built-in controls, come out human-human 0.37 %, chimpanzee 2.65 %, gorilla 3.00 %, orangutan 6.62 %.⁷  The paper reports "the number of SVs increasing as a function of evolutionary distance from human".⁷  Yoo's SNV divergence comes out chimpanzee 1.60 %, gorilla 1.99 %, orangutan 3.64 %.⁸  Two independent metrics, both about seven times above their within-species floor, give the same order:  chimp, gorilla, orangutan.  That is the order common descent predicts from a branching tree.

What design accounts say about the order.  The plain "common design" reply does not reach it, and the most careful young-earth biologist on the subject said so.  Todd Wood wrote in 2006:  "Creation biology needs an explanation of the pattern of similarities, not merely an ad hoc appeal to a common designer."¹³  He named one attempt, Walter ReMine's 1993 message theory, which holds that life was given a nested pattern so that it would read as the work of one designer.  Wood called it "the only recent attempt to explain the pattern of organismal similarity" and judged that it "does not explain animals that are very similar to humans."¹³  The most developed design model since is Winston Ewert's dependency graph, published in the intelligent-design journal BIO-Complexity in 2018.  He grants the premise:  "biological data resembles a nested hierarchy sufficiently well to require an explanation."  His proposal borrows from software, where programs share libraries, and "postulates that different biological species share modules related by a dependency graph."²⁶  Fitted to nine gene-family databases, his model beat a tree on every one, by "over 10,000 bits" even on the least favourable.²⁶  That is a real, testable design account, and the older claim that no design model of nested similarity exists is out of date.  But it is a model of which gene families a species has, not of how many letters differ in the genes they share, and Ewert says so:  "Addressing the gene family sizes or exact sequences is left to future research."²⁶  The ordering in this article is substitution divergence, most of it outside genes.  Module reuse gives no obvious reason why that DNA should differ least from chimpanzee, more from gorilla, more again from orangutan.  No design model yet predicts that number for each ape, and none is a similarity percentage.

A sequence percentage does not measure how different the animals are.  This is the creationist point with the best secular support, and it is older than the genome.  Sean Carroll summarised Mary-Claire King and Allan Wilson's 1975 Science paper:  "They concluded that the small degree of molecular divergence observed could not account for the anatomical or behavioral differences between chimps and humans.  Rather, they proposed that evolutionary changes in anatomy and way of life are more often based on changes in the mechanisms controlling the expression of genes than on sequence changes in proteins."²⁷  The measurements since bear out the size of the regulatory layer.  John Calarco and colleagues in Benjamin Blencowe's laboratory profiled alternative splicing, the cutting and joining of a gene's message into different proteins, in the frontal cortex and heart of both species in 2007:  "Despite an overall sequence identity between human and chimpanzee genomic coding regions of 98%–99%, we observe that 6%–8% of surveyed alternative exons display pronounced splicing level differences."¹⁴  Their "pronounced" means a difference of more than 25 points in how often an exon is included, and about 80 % of exons differed by less than 15 points.¹⁴  Todd Wood drew the young-earth conclusion in 2006:  "Even at 99% identity, however, the biological and behavioral differences between chimpanzees and humans indicate that the source of these differences is not likely to be found entirely in the genome sequences."¹³  The skeptic's reply is fair and should be stated.  Expression and splicing differences are caused by sequence differences, they are counted in a different unit, and they show the same pattern as the DNA.  Philipp Khaitovich and Svante Pääbo's group compared five tissues in 2005 and found "that the patterns of differences in gene expression and gene sequences are markedly similar", in a way "consistent with a model of neutral evolution with negative selection."²⁸  So regulation answers the question "how different are we?" much better than a percentage does.  It does not answer "related or not?" any better.

Carter's handoff.  The strongest creationist move is Carter's, and it is not about similarity:  "if the percent similarity is much less than 99%, there is no way, mathematically, to explain how so many millions of difference arose in the (imagined) 6.5 million years since our last common ancestor."²⁰  That depends on the pedigree mutation rate, generation time and neutral fixation, and it is the subject of the sibling article on the human-chimp split and the mutation rate.  One thing should be said here.  The 2025 SNV divergence of 1.6 % is higher than the 1.23 % the rate arguments were built on.  The T2T data therefore make the creationist rate objection stronger and the creationist similarity objection weaker, at the same time.

Objections and responses

Skeptic:  "So the creationists were wrong.  98.8 % stands." No.  98.8 % is substitutions only, inside the 2.4 Gb that aligned one-to-one in a 2005 draft.  The consortium wrote that indels "dwarf" it.  With indels the figure is about 96 %.  On a whole-genome one-to-one basis, subtracting the within-human floor gives 91 % to 95 % depending on the method, and that is an upper bound:  the floor subtraction cannot see species differences inside DNA that already differs between two humans.  The slogan overstated, and the people who said so first were Britten, Varki and the consortium itself.

Creationist:  "The 2025 Nature paper itself says 12.5 to 27.3 % failed to align." It does, and the same sentence attributes that to "rapidly evolving and structurally variant regions of the genome as well as technical limitations of alignment in repetitive regions."  The paper's own supplement shows the two haplotypes of one gorilla failing to align to each other by 13.8 %.

Creationist:  "Luskin's tables are from the paper.  The 84.95 % and 83.89 % are real." They are real, and they are correctly transcribed.  The first rows of those same tables are human genomes at 86.96 % and 88.80 %.  Luskin did not quote them.  The shortfall that the tables can charge to species is 2 to 5 points.  That is a floor:  the 11 points that also separate two humans are unmeasured by this method, not shown identical.  The paper's human-specific structurally divergent DNA, counted with both humans as a control, comes to about 5 %.

Creationist:  "Human-to-human comparisons in the 80s mean the aligners are useless, not that chimps are close." Carter says this:  "The fact that her ungapped human-to-human BLAST searches yield a percent similarity in the 80s is highly suspicious.  If this is the case, BLAST is generally useless".²⁰  BLASTN with -ungapped is useless for this purpose;  that is the objection.  It stops at the first indel, and two humans differ by millions of indels.  Drop the flag, as Tomkins did in 2018, and his own control row gives a 7.5-point species difference.

Creationist:  "The chimpanzee assemblies were humanized.  Kronenberg admits it." Kronenberg admits it, and then tests it.  Assembling chimpanzee with no human guidance produced "a modest elevation in SNV divergence".  Yoo 2025 removed the last of the bias and got 1.6 %.⁸  The prediction that de-humanised assemblies would show 15 to 30 % difference was tested twice and failed twice.

Creationist:  "Todd Wood's result is a blog post." It is, and it is preliminary, and this article says so.  It has stood unanswered for fifteen years, Seaman and Buggs confirmed the same thing at a locus in a peer-reviewed paper, and the T2T data now make the same point with finished assemblies.

Creationist:  "Similarity only shows common design.  A designer reuses what works." Reuse explains why two species share genes.  It does not by itself explain why the shared DNA differs least from chimpanzee, more from gorilla and more again from orangutan, and Todd Wood called the bare appeal "ad hoc" in 2006.¹³  The best answer on the design side is Winston Ewert's dependency graph, which fits gene-family presence and absence better than a tree.²⁶  It has not yet been applied to substitution divergence, and Ewert says that is future work.²⁶

Creationist:  "Whatever the percentage, a human is not 98.8 % chimpanzee." Correct, and the mainstream has said so since King and Wilson in 1975.²⁷  Splicing differs at 6 to 8 % of the exons measured, against 1 to 2 % sequence difference in coding DNA.¹⁴  But expression and splicing differences track the sequence differences and show the same pattern,²⁸ so they measure how far apart the species are, not whether they are related.

Both:  "Which number should I use?" Say which question you are asking.  Substitutions in aligned DNA:  98.4 to 98.8 %, against a within-human 99.84 %.  Substitutions plus indels:  about 96 %.  Whole-genome one-to-one identity:  84 to 89 %, against a within-human 89 to 96 %.  Structural-variant load:  2.65 % different, against a within-human 0.4 %.  Any of these is honest with its control attached.  None is honest without it.

Issues and unanswered questions

Computations that would settle what remains, none of which has been run.

1. A human-versus-human BLASTN with Tomkins' exact 2013 parameters. Run GRCh38 against T2T-CHM13, or HG002 maternal against paternal, with -word_size 11 -evalue 10 -max_target_seqs 1 -dust no -soft_masking false -ungapped on 300-base slices. If two humans come back in the 80s, every Tomkins figure before 2018 measures the aligner. This needs one server and a weekend, and Carter has called the expected result "highly suspicious" without running it.
2. Wood's Megablast test on the 2025 assemblies. Take every window of the chimpanzee T2T assembly with high gap divergence against T2T-CHM13 and BLAST it against the whole human genome without requiring a unique hit. Report the fraction with any hit above 90 % identity and the median identity. Wood's 2011 answer was 87 % of fragments and 97.4 %.
3. The gene-complement figure on complete genomes. Re-run the Demuth gene-family analysis on the Yoo 2025 annotations. The figure fell from 2,382 to 1,418 genes between 4x and 6x coverage. Whoever runs it first defines the number.
4. A tree test on shared structural variants. Take the 17,789 fixed human-specific structural variants from Kronenberg 2018 and test whether their presence and absence across the six T2T apes is tree-like. No creationist-facing test has been published.

Weaknesses in this article's own case.  The "excess over control" subtraction in Part 4 is crude, and is offered only because the alternative in print is no correction at all.  It gives a floor on the species difference, not the difference itself:  DNA that fails to align between two humans may differ far more between human and chimpanzee, and no method in print measures how much.  The ordering argument rests on genome-wide averages, while 30 to 40 % of the genome sorts the three African apes some other way;  incomplete lineage sorting explains that on the mainstream account, but it is a modelled explanation, not a direct observation.  Wood's numbers are preliminary and unpublished.  The Y-chromosome argument shows that the Progressive Cactus figures are an artefact, but it does not supply a corrected human-chimp Y figure.  Marks' 25 % floor is a statement about random sequence, and real genomes are not random.

Sources not obtained.  King and Wilson's 1975 Science paper is cited through Sean Carroll's 2005 summary.  The 2012 vertebrate splicing survey by Nuno Barbosa-Morais and colleagues was read in abstract only, which does not name the species pair behind its "within 6 million years", so it is not used.  Tomkins' 2011 ARJ paper reporting 86–89 % is known only through its quotation in the 2012 review.  Carter's Facebook post giving 3.6 % within one human is not fetchable;  the number is reproduced here from Yoo's figures.  The Ebersberger 2002 paper was not checked against its own methods.

Verdict

The 98.8 % slogan overstated, and the creationist critics were right to say so.  It counted substitutions only, inside the aligned part of a draft genome.  The consortium said in 2005 that indels dwarf it, and about 96 % is the honest figure on that definition.  Buggs' 84 to 85 % is correct arithmetic for a whole-genome one-to-one count.  The old assemblies were human-guided and the Eichler laboratory said so in Science.

The creationist numbers below 84 % are errors, and the 84 to 85 % is misread.  Seventy per cent came from an ungapped aligner and a software bug, and its author withdrew it.  Eighty per cent and 11 % are double subtractions.  The 2025 tables that give 15 to 16 % different give 11 % different for two human genomes in their first rows.  What those tables can charge to species is about 5 points, as a floor, and the paper's own count of human-specific structurally divergent DNA agrees at about 5 %.  The true whole-genome figure on this definition lies between that floor and the raw 16 points, and no method in print yet closes the gap.  Gap divergence, the quantity underneath every one of these figures, fails its own control:  one gorilla differs from itself by more than human differs from chimpanzee.

The metrics that pass their controls give small numbers and put the apes in order.  SNV divergence is 1.6 %, seven times the within-human floor.  Structural-variant load is 2.65 %, seven times the within-human floor.  Both order the apes chimp, gorilla, orangutan.  Todd Wood, a young-earth creationist, showed in 2011 that "unalignable" chimpanzee DNA is 97.4 % identical to human and merely repetitive, and nobody on his side has answered him.

What a percentage can say.  It can measure distance on a scale whose floor is 25 %, and it can say nothing about descent by itself.  What speaks to descent is the ordering, and every metric that survives its control gives the ordering common descent predicts.  Design models of nested similarity exist, ReMine's and Ewert's, but none yet predicts substitution divergence, and Todd Wood's 2006 verdict that creation biology still needs "an explanation of the pattern of similarities" stands.  Nor does a percentage measure how different the animals are:  6 to 8 % of the exons measured are spliced differently, though expression differences follow the same pattern as the sequence.  The honest young-earth position on this page is Carter's:  the ceiling is 96.6 %, the true figure is below it, and the real argument is about the mutation rate, not the percentage.

Companion pages

• The literature — every source consulted, with what it measured, whether it was read in full, and which way it cuts.
• Quarantine — the quotations and figures that circulate on this topic and could not be verified from bytes.
• Chromosome 2 Fusion — the structural difference that is counted in events, not percent.

Sources