Tree-of-life quotations, audited

No verdict on this page was reached through a summarising tool.  Every entry was checked by downloading the source and searching its own text.  Where a source could not be opened, the entry says so and the quotation does not run.

Summary

Seven quotations in wide circulation about the "collapse" of the tree of life do not survive a check against the source.  Two are garbled.  Five are word-perfect and change meaning the moment the next sentence is restored — and one of those five is not the scientist's own wording at all, but a press release's compression of a sentence in his paper.  Two further quotations in the ledger below fail the same way and are simply less famous, which is why the counts table records nine defective passages and this paragraph names seven.

The worst of them is Ford Doolittle's "It is as if we have failed at the task that Darwin set for us."  The next sentence in the same paragraph is "But in fact, our science is working just as it should."

Some of the errors were made here.  This project's own working note file on the tree of life, /home/share/notes/eric/research/biology/orchard.md, is where several of these quotations were sitting, and it is the source of the misspelling in the Doolittle 1999 line, the garble in the Berra Corvette passage, and a paraphrase of a 2013 paper that the paper itself refutes.  Those are named below in their own section.  Owning them is the point of the page.

The page is not one-sided.  Plenty of quotations in this literature are accurate, in context, and useful to the young-earth and design side, and they are printed at full strength.  So are the ones that cut the other way — including Todd Wood writing "Evolution is not a theory in crisis," and Michael Behe writing "Common descent is true."

The verdicts

The counts

Sixty-five passages are on this page.  Thirty-eight of them came out of this project's note file and were audited quotation by quotation.  Twenty-seven more were read directly in the sources during research, and they are listed separately so that no passage has two homes.

How the counting works.  One row per passage, filed under its most serious defect.  Doolittle's 1999 sentence is counted as GARBLED even though it is also misleading in context, because a wrong word is the worse fault.  Several passages counted ACCURATE still carry a defect in the citation rather than in the words — no author named, a magazine credited instead of the journal, an unmarked splice across twenty printed pages — and each of those is named in its row.  Strip those out and the number of note-file entries with no defect at all drops to roughly half.  The counts therefore understate the problem rather than inflating it.

Four passages that this page once filed NOT VERIFIED have since been read in the source and moved:  Zuckerkandl and Pauling 1965, the second half of the Berra figure caption, Ridley 1985, and Michael Donoghue's remark in Wired.  Three remain, listed in "The three that do not run" below.

Two paraphrases in the note file are not quotations and are not counted.  Both are withdrawn; see "Two paraphrases that fail" below.

The seven that matter most

1. "It is as if we have failed at the task that Darwin set for us"

Circulates as:  "It is as if we have failed at the task that Darwin set for us: delineating the unique structure of the tree of life."

Credited to:  W. Ford Doolittle, 2000.

Verdict:  ACCURATE BUT MISLEADING IN CONTEXT.  This is the single most recycled quote-mine in the tree-of-life literature.

Real source:  Doolittle WF, "Uprooting the Tree of Life," Scientific American 282(2):90–95, February 2000.  The sentence is in the final paragraph, page 95, third column.

What surrounds it.  The whole closing paragraph, in order:

"Some biologists find these notions confusing and discouraging.  It is as if we have failed at the task that Darwin set for us: delineating the unique structure of the tree of life.  But in fact, our science is working just as it should.  An attractive hypothesis or model (the single tree) suggested experiments, in this case the collection of gene sequences and their analysis with the methods of molecular phylogeny.  The data show the model to be too simple.  Now new hypotheses, having final forms we cannot yet guess, are called for."

Two things go wrong at once.  Doolittle is reporting other people's reaction — "Some biologists find these notions confusing and discouraging" — not stating his own conclusion.  And he answers that reaction in the very next sentence.  The quotation as it circulates stops one sentence before the author's reply.

There is no honest way to deploy it.  The article uses it only as an example of the practice, with the rebuttal printed in the same breath.

Two independent checks were run on this passage from different copies.  One researcher read the 2006 University of Chicago Press reprint (Evolution: A Scientific American Reader, pages 87–98, the passage at page 97).  Another read the original February 2000 magazine offprint, which carries "Copyright 2000 Scientific American, Inc." in the running footer, and which pins the magazine page at 95.  The words agree exactly.

And from the same article, the sentence that settles the scope (magazine p. 94; 2006 reprint p. 96):

"What would a truer model look like?  At the top, treelike branching would continue to be apt for multicellular animals, plants and fungi."

Doolittle's uprooting is of the prokaryote tree and the base of the eukaryotes.  He leaves the animal, plant and fungal tree standing, in print, in the same article whose title is quoted against it.

2. "Molecular phylogenists will have failed to find the 'true tree'"

Circulates as:  "Molecular phylogenists will have failed to find the 'true tree,' not because their methods are inadequate or because they have chosen the wrong genes, but because the history of life cannot properly be represented as a tree."

Credited to:  W. Ford Doolittle, 1999.

Verdict:  GARBLED, and misleading in context on top of that.

Real source:  Doolittle WF, "Phylogenetic Classification and the Universal Tree," Science 284(5423):2124–2128, 25 June 1999.  The sentence is in the abstract, page 2124.

What is wrong with the words.  Doolittle wrote "phylogeneticists," not "phylogenists."  The misspelling sits in this project's own notes, and it is a useful tracer:  a copy carrying it did not come from the paper.

What surrounds it.  The abstract, with the quoted sentence in place:

"Although confidence in some of the tree's early branches has recently been shaken, new approaches could still resolve many methodological uncertainties.  More challenging is evidence that most archaeal and bacterial genomes (and the inferred ancestral eukaryotic nuclear genome) contain genes from multiple sources.  If 'chimerism' or 'lateral gene transfer' cannot be dismissed as trivial in extent or limited to special categories of genes, then no hierarchical universal classification can be taken as natural.  Molecular phylogeneticists will have failed to find the 'true tree,' not because their methods are inadequate or because they have chosen the wrong genes, but because the history of life cannot properly be represented as a tree.  However, taxonomies based on molecular sequences will remain indispensable, and understanding of the evolutionary process will ultimately be enriched, not impoverished."

The famous sentence is the "then" half of an "if."  The future perfect "will have failed" is doing conditional work.  It is not a report of a result.  The abstract then ends by refusing the pessimistic reading outright.

The author says so himself.  Sixteen years later, asked about that paper, Doolittle described the sentence as conditional:

"I think I got some notoriety with my 1999 Science review. ... In 1999, we didn't know very much yet.  There were many individual cases where people were inferring lateral gene transfer and some irritating inability to disprove lateral gene transfer in other cases.  I said, if it turns out to be as big as I think it's going to turn out to be, then we just have to wonder whether a tree is the appropriate model."

That is from Gitschier J, "The Philosophical Approach: An Interview with Ford Doolittle," PLoS Genetics 11(5):e1005173, 7 May 2015, page 7 of the article PDF.  It is open access under CC-BY, so anyone can check it.

The scope also matters.  Doolittle's argument is about archaeal and bacterial genomes and the ancestral eukaryotic nucleus.  It is not about animals, and the same paper affirms shared descent explicitly:  biologists after Darwin "knew the ordering principle to be shared descent from an ever more limited number of common ancestors ... back to the last common ancestor of all living things" (page 2124).

3. "We've just annihilated the tree of life"

Circulates as:  "We've just annihilated the tree of life.  It's not a tree any more, it's a different topology entirely."

Credited to:  Michael Syvanen, University of California, Davis, 2009.

Verdict:  ACCURATE BUT MISLEADING IN CONTEXT.  He said it.  It is about one sea squirt.

Real source:  Lawton G, "Uprooting Darwin's tree," New Scientist issue 2692, 24 January 2009, pages 34–39, published online as "Why Darwin was wrong about the tree of life."  The quotation is on page 39, inside a boxed sidebar headed "Natural born chimeras" — not in the main feature.

What surrounds it.  The sidebar reports Syvanen's comparison of 2,000 genes shared by humans, frogs, sea squirts, sea urchins, fruit flies and nematodes.  Sea-squirt genes gave contradictory answers.  His explanation is that tunicates are chimeras, formed by the fusion of an early chordate with an ancestor of the sea urchins about 600 million years ago.  "We've just annihilated the tree of life" is his description of what that would mean.  It is a claim about one lineage, offered as a minority hypothesis, in a box.

Three things the circulating version leaves out.

He published the analysis.  Syvanen M & Ducore J, "Whole Genome Comparisons Reveals a Possible Chimeric Origin for a Major Metazoan Assemblage," Journal of Biological Systems 18(2):261–275, 2010.  The magazine's "he failed" is not the end of the story.

The published split is not the one he gave the reporter.  He told New Scientist "Roughly 50 per cent of its genes have one evolutionary history and 50 per cent another."  The peer-reviewed paper says "One group, containing about 40 % of the proteins, supports the classical assemblage of the tunicate with vertebrates, while the remaining group places the tunicate outside of the chordate assemblage."  Cite 40/60.

The rest of his own dataset is a tree.  The paper states that with Ciona removed, the remaining eight taxa give the same topology from both groups of genes.

A hostile reader will also point out that Ciona intestinalis has one of the fastest-evolving genomes among deuterostomes and is the textbook long-branch-attraction case, and that the mainstream answer since Delsuc and colleagues in 2006 places tunicates as the sister group of vertebrates.  Syvanen is a documented case of one well-supported result flatly contradicting another, with the field split on which is the artefact.  He is not evidence that phylogenetics has failed in general.

4. "The tree of life is being politely buried"

Circulates as:  "The tree of life is being politely buried, we all know that."

Credited to:  Michael Rose, University of California, Irvine, usually by way of The Guardian, January 2009.

Verdict:  ACCURATE BUT MISLEADING IN CONTEXT, with a second-hand attribution on top.

Real source:  Lawton, New Scientist 2692, page 39.  The Guardian was reporting the New Scientist cover, and its version silently drops ", we all know that."

What surrounds it.  Four paragraphs earlier, in a sentence that begins on page 38 and finishes on page 39, the same feature says:

"Nobody is arguing – yet – that the tree concept has outlived its usefulness in animals and plants.  While vertical descent is no longer the only game in town, it is still the best way of explaining how multicellular organisms are related to one another – a tree of 51 per cent, maybe.  In that respect, Darwin's vision has triumphed: he knew nothing of micro-organisms and built his theory on the plants and animals he could see around him."

And two paragraphs later, still on page 39:

"Both he and Doolittle are at pains to stress that downgrading the tree of life doesn't mean that the theory of evolution is wrong – just that evolution is not as tidy as we would like to believe."

Rose is also introduced as the outlier.  Lawton's own words are "Others, however, don't think it is time to relax ... Rose goes even further."  Quoting him as a spokesman for the field inverts the article's presentation of him.

Two more facts belong beside this cover, because a reader will find them anyway.  The editorial in the same issue predicted the creationist use and denied it in advance:  "None of this should give succour to creationists ... Expect to find excerpts ripped out of context and presented as evidence that biologists are deserting the theory of evolution en masse.  They are not."  And a letter in issue 2696, 18 February 2009, page 24, signed by Daniel Dennett, Jerry Coyne, Richard Dawkins and Paul Myers, concedes more than it means to:  "Of course there's a tree; it's just more of a banyan than an oak at its single-celled-organism base."

Citing the cover itself is fair.  A mainstream science weekly did run "DARWIN WAS WRONG — Cutting down the tree of life" in Darwin's bicentenary year, and every named quotation inside it is genuine.  Citing it against animal or human phylogeny contradicts the text underneath the cover.

5. Berra's Corvette

Circulates as:  "If you compare a 1953 and a 1954 Corvette, side by side, then a 1954 and a 1955 model, and so on, the descent with modification is overwhelmingly obvious.  This is what paleontologists do with fossils..."

Credited to:  Tim Berra, 1990.

Verdict:  GARBLED, in two places.

Real source:  Berra TM, Evolution and the Myth of Creationism: A Basic Guide to the Facts in the Evolution Debate, Stanford University Press, 1990, page 117.  Read from the Internet Archive lending scan evolutionmythofc0000unse, leaf 139, with the leaf-to-page map taken from the item's own book-reader data.  Five other Internet Archive items hold the same book; one of them, B-001-000-890, needs no loan at all.

What Berra actually wrote:

"If the australopithecines, Homo habilis and H. erectus, were still alive today, and if we could parade them before the world, there could be no doubt of our relatedness to them.  It would be like attending an auto show.  If you look at a 1953 Corvette and compare it to the latest model, only the most general resemblances are evident, but if you compare a 1953 and a 1954 Corvette, side by side, then a 1954 and a 1955 model, and so on, the descent with modification is overwhelmingly obvious (Figure 41).  This is what paleoanthropologists do with fossils, and the evidence is so solid and comprehensive that it cannot be denied by reasonable people."

Error one.  Berra wrote "paleoanthropologists."  The passage sits in a chapter on human evolution and is about hominid fossils, not about the fossil record at large.

Error two, and it is the one that matters.  The circulating version cuts the analogy's opening clause:  "If you look at a 1953 Corvette and compare it to the latest model, only the most general resemblances are evident, but..."  That clause is the argument.  Berra's point is that widely separated forms look unrelated, and that only the finely spaced series shows continuity.  Removing it changes what he claimed.

Why this one is worth getting right.  The design objection to Berra is strong and does not need a garbled quotation to make it.  Corvettes are manufactured, not begotten.  A nested series of designed artifacts is exactly the pattern Berra offers as proof of unguided descent.  Phillip Johnson named this "Berra's blunder" in Defeating Darwinism by Opening Minds (1997), and that book could not be obtained — the only Internet Archive scan is not available to borrow and the Access Research Network page is a 404 — so his wording is not quoted here.  The observation stands without him, and the main article measures it:  designed objects with a copying history do produce strong congruence.

The caption checks out in full, and it runs across two pages.  Figure 41 is a two-page photo spread.  Page 118 carries the 1953 and 1962 cars and the caption as far as "The point is that the"; page 119 carries the 1968 and 1978 cars and finishes it:  "The point is that the Corvette evolved through a selection process acting on variations that resulted in a series of transitional forms and an endpoint rather distinct from the starting point.  A similar process shapes the evolution of organisms.  (Photos courtesy of General Motors Corporation.)"  The widely circulated remainder is word-perfect.  The archive's full-text index does truncate the paragraph, which is what defeated the first six probes; a second, unrestricted scan of the same book (B-001-000-890) publishes the whole OCR, and both figure pages were then read off page images.

6. Sharks are more like humans than like fish

Circulates as:  "We were very surprised to find, that for many categories of proteins, sharks share more similarities with humans than zebrafish."

Credited to:  Michael Stanhope, Cornell University, 2013.  This project's notes carried it with no source line at all.

Verdict:  ACCURATE BUT MISLEADING IN CONTEXT — severely.  It is not a phylogenetic result.

Real source:  Hodes C, "Shark, human proteins are surprisingly similar," Cornell Chronicle, 3 December 2013, syndicated by ScienceDaily on 5 December.  Both pages agree word for word.  The underlying paper is Richards VP, Suzuki H, Stanhope MJ & Shivji MS, "Characterization of the heart transcriptome of the white shark (Carcharodon carcharias)," BMC Genomics 14:697, 2013.

What was actually measured.  The team sequenced the transcriptome of one tissue — a heart — from one confiscated shark.  They annotated 21,019 open reading frames, 19,277 of them with Gene Ontology terms.  Then they ran Fisher tests asking which functional categories were over- or under-represented compared with the zebrafish and human transcriptomes.  Their sentence:  "Fisher tests showed the white shark transcriptome to be significantly enriched for numerous metabolic GO terms compared to the zebra fish and human transcriptomes, with white shark showing more similarity to human than to zebra fish (i.e. fewer terms were significantly different)."  The same sentence begins "While acknowledging the limitations of our single tissue transcriptome."

So the measured quantity is how many functional categories differ in proportion in one heart.  That is not a phylogenetic character.  It never enters a tree, so it cannot conflict with one.

And the author gives the cause in the same interview.  Stanhope:  "We think this has to do with a rare trait in fish called regional endothermy."  White sharks are warm-bodied.  So are mammals.  A convergent-physiology explanation, from the scientist, in the same press release the quotation comes from.

Using this line as evidence of phylogenetic incongruence is a category error.  It is a good example of how press-release language gets read as a tree claim when it is a physiology claim.

7. "Vanishingly small"

Circulates as:  a block quotation attributed to Todd Oakley:  "The probability of complex organs evolving multiple times with similar trajectories should be vanishingly small."

Verdict:  ACCURATE BUT MISLEADING IN CONTEXT.  The wording is a press officer's paraphrase.  The claim is Oakley's own, and his paper states it more clearly.

Real source:  the paper.  Pankey MS, Minin VN, Imholte GC, Suchard MA & Oakley TH, "Predictable transcriptome evolution in the convergent and complex bioluminescent organs of squid," PNAS 111(44):E4736–E4742, 2014.  The first sentence of its Significance statement reads:  "Unless there are strong constraints, the probability of complex organs originating multiple times through similar trajectories should be vanishingly small."  The paper records that Pankey and Oakley wrote it.  What circulates is not that sentence but a paraphrase of it, in Cohen J, University of California, Santa Barbara press release, 21 October 2014, carried by ScienceDaily, which prints it unquoted with the tag "noted Oakley" — a reported statement, not a transcription.

What the paragraph actually says:

"The team's results demonstrate that the evolution of overall gene expression underlying convergent complex traits may be predictable.  This finding is unexpected and could indicate unusually strong constraints: The probability of complex organs evolving multiple times with similar trajectories should be vanishingly small, noted Oakley.  Yet the team's novel bioinformatic approaches indicate the evolution of convergent phenotypes is associated with the convergent expression of thousands of genes."

There are direct Oakley quotations elsewhere in the same release, and the writer chose not to put quotation marks round this one — because it is his own compression of the paper, not Oakley's spoken words.  Quote the paper, not the release.  The paper's wording differs:  "originating" for "evolving," "through" for "with," and the governing clause "Unless there are strong constraints" restored to the front.

And it argues the opposite of the use it is put to.  The paragraph's structure is:  convergence should be improbable, yet it happened, and the gene-expression route to it turned out predictable.  The finding is that convergence is more constrained and more repeatable than expected.  That makes convergence a cheaper explanation for conflicting characters, not a more expensive one.  Quoting the "vanishingly small" clause and stopping inverts the paragraph.

Where this project's own notes are the source of the error

The file is /home/share/notes/eric/research/biology/orchard.md — 106 KB, 803 lines, last edited 3 August 2026.  It is a working note file, not a publication, and it is more honest than most quotation collections on this subject:  it marks seven items "unverified" itself, and at line 729 it has already struck out a whale-phylogeny quotation with the annotation "(source doesn't check out!)".  That is exactly the right instinct.  It just was not applied to the rest.

Transcription errors.

• "phylogenists" for "phylogeneticists" in the Doolittle 1999 line (line 564).
• "paleontologists" for "paleoanthropologists" in the Berra passage, plus the dropped opening clause of the Corvette analogy.
• "1167 6 83" for "1167 ± 83" in a Graur and Martin passage in the sister file molecular clocks.txt. That is a text-extraction artefact for the plus-or-minus sign, and it is a tell: the passage was pasted out of a PDF dump and never re-read.
• Two typos in the Jenner citation — "teh" for "the" and "PRocess" for "Process." Those are optical-character-recognition artefacts, and they show the text was read off a screenshot posted on X rather than off the paper.

Truncations and unmarked splices.

• The Doolittle 2000 excerpt ends one sentence before Doolittle's own reply.
• The Dávalos bat block joins a passage from the Introduction to a passage from the Discussion with a bare ellipsis. Those are about twenty printed pages apart.
• The Doolittle and Brunet 2016 block runs the one-sentence abstract into the opening paragraph of the body and drops a sentence without an ellipsis (lines 374 and 585).
• An O'Malley and Koonin 2011 block is spliced across several hundred words of the Conclusions and omits the sentence "The TOL is very much alive in this sense, and we predict its demise is far from imminent."
• The Rokas and Carroll homoplasy sentence drops the clause carrying the number, and the sentence before it, which names lineage sorting as the cause.
• A Dávalos quotation has "[genetic]" inserted into "different sets of characters." The insertion is wrong. The paper means characters in general, morphological included — the whole study is morphology against molecules. Deleting the bracket makes the quotation stronger.

Attribution errors.

• The Oakley sentence is set as a block quotation (line 290). The transcription is faithful — the note file kept the press officer's tag "noted Oakley" and gave the ScienceDaily URL — but the wording is the release's paraphrase. Cite Pankey et al. 2014 instead.
• The Rose and Bapteste quotations are credited to The Guardian (line 704). The Guardian was reporting New Scientist.
• The placental-mammal editorial is quoted with a journal and year but no author, from a salvomag.com page — the giveaway is a #sthash…dpuf share-button artefact left on the URL. The authors are Emma Teeling and S. Blair Hedges.
• The Doolittle 2015 passages are credited to sandwalk.blogspot.com. Sandwalk is Larry Moran's blog reproducing a PLoS Genetics interview. Cite the interview.
• The Stanhope quotation has no source line at all.

Two claims are sourced to a chat with a language model.  Line 185 links a ChatGPT conversation about microRNA conflicts.  Line 524 links a ChatGPT share link and a Gemini link about Eugene McCarthy's human-pig hybrid claims — and the note itself records that the model "couldn't access macroevolution.net and sourced claims from other sites instead."  Neither is a source.  Neither may ever appear in an article on this site, in any form, for any claim.  A language model's output is not evidence about what a paper says.  It is a guess about what a paper says.

Several links were already dead before the audit.  The ge.tt file-drop links at lines 17, 688, 730, 738, 741 and 743 no longer resolve, and a talk cited at line 383 is noted in the file itself as gone.  Anything whose only citation is one of those is currently unverifiable.

One thing the file had and never used.  Syvanen's published paper sits at line 151.  The New Scientist claim that "he failed" is quoted at lines 195 and 743.  The two were never connected.

Two paraphrases that fail

"When we remove the data that doesn't fit well our phylogenetic tree looks better."  This is the note file's gloss on Aberer AJ, Krompass D & Stamatakis A, "Pruning Rogue Taxa Improves Phylogenetic Accuracy," Systematic Biology 62(1):162–166, 2013, reached by way of a Journal of Creation article rather than the paper.  The paper refutes it in the same paragraph as the sentence being quoted.

Three specific problems.  The "true tree" is not conventional wisdom — it is a simulated tree, known by construction because the sequences were generated along it, across 400 datasets.  A random-pruning control was run:  "pruning a taxon set of equal size (as the calculated dropset) at random had a negative effect on the congruence with the true tree."  And a rogue taxon is not "data that doesn't fit" — the paper defines rogues as taxa "which assume varying and often contradictory positions in the tree set," identified by their instability across bootstrap replicates of the same data, not by disagreement with an expected answer.

The paraphrase is withdrawn.  A narrower criticism survives, and the paper states it in its own words:  rogues "are mostly simply excluded (pruned) in current studies," so the published tree carries no answer for exactly the taxa whose placement was disputed.  That is a real transparency problem.  It is not "they threw out inconvenient data."

"At every level the amount of discordance is about 10–35 %."  This is the note file's reading of Figure 2 in Rokas A & Carroll SB, "Bushes in the Tree of Life," PLoS Biology 4(11):e352, 2006.  It is wrong, and it is wrong in the direction of understating the paper.  Panel A gives 24 % of informative characters conflicting at human, chimp and gorilla.  Panel B is total non-resolution among elephant, sirenian and hyrax.  Panel C has about 300 informative characters splitting evenly three ways among tetrapod, coelacanth and lungfish — about two-thirds discordance against whichever answer is true.  Panel D has two genome-scale studies supporting contradictory topologies, each with high support.  Describe the panels; do not average them.

The ledger

A. Quotations that were in this project's notes

Sorted by the year of the source, oldest first.

B. Quotations read directly in the sources

These were opened during research rather than inherited from the notes.  They are listed separately so that no passage is recorded twice.  They are grouped by argument rather than by date, because several of them only make sense in pairs.

The three that do not run

These are the NOT VERIFIED rows collected in one place, with what was tried.  A quotation that cannot be traced does not appear in the article.  "Could not find it" is not "does not exist," so each of these is a job left open, not a verdict of fabrication.

• Jenner 2000, Evolution & Development 2(4):208–221 — Unpaywall says closed, Europe PMC has the abstract only, scholar.archive.org returned HTTP 500, ResearchGate 403, and Wiley 403s every header set tried.
• Graur & Martin 2004, Trends in Genetics 20:80–86 — not attempted.
• The New Scientist editorial, January 2009 — not obtained from the magazine. Three independent reproductions posted within five days of publication agree word for word: P. Z. Myers on Pharyngula, 28 January 2009, who says the magazine's Roger Highfield asked him to circulate it; a commenter on Larry Moran's Sandwalk, 23 January 2009; and James Hrynyshyn on Island of Doubt, 26 January 2009. The wording is settled; the provenance is not primary.

Two further items were chased and could not be had, and neither is quoted:  Phillip Johnson's "Berra's blunder" passage in Defeating Darwinism by Opening Minds (1997), and the body text of Teeling & Hedges 2013 beyond its abstract.

The full failure logs live next door.  The quarantine page records every source behind this article that could not be read, with the exact routes that failed and what to try next.  The division is simple:  this page gives verdicts on wording and context for passages that were reached, and quarantine records what was not reached, including many sources that never produced a quotation at all.  Where both pages cover the same item — the Berra caption, the Jenner sentence, the venom count, the placental-mammal editorial — quarantine has the longer account.

One place the two pages disagree, stated rather than smoothed.  Quarantine's section D4 records that Behe's own statement accepting common descent could not be reached, because three Internet Archive copies of Darwin's Black Box refused to lend.  It could:  a separate dossier read it out of copies already in this project's source library, a PDF of the 2006 edition and an EPUB of The Edge of Evolution, and both passages are quoted with their locators at entry 65 above.  Two researchers reached opposite conclusions about the same book on the same evening because one searched the open web and the other searched the shelf.  That is the standing rule in miniature:  "could not find it" is not "does not exist," and the second route is the one worth trying.

How the checking was done

No quotation on this page was established through a summarising tool.  Every verdict came from downloading the source and searching its own text.  That rule exists because a tool that converts a page and then answers a question about it hands back a small model's paraphrase, which can invent a page number or smooth a wording without saying so.

Four practical findings are recorded here, because each produced a false "missing" result before it was understood.

Multi-column PDFs defeat a naive search, not the PDF.  Searching a pdftotext -layout dump of the Dávalos bat paper for nine test phrases finds two.  Searching a reading-order pdftotext dump finds three.  Normalising whitespace and joining hyphens across line breaks in the same reading-order dump finds all nine.  The Doolittle 2000 closing sentence behaves identically:  invisible in both raw dumps, whole once the whitespace is normalised.  Column cropping and publisher XML both work too, and both are slower.  Normalise first;  and never call a quotation missing until this has been ruled out.

Try every scan of a book, and check whether a loan is needed at all.  Six Internet Archive items hold Berra 1990.  One of them, B-001-000-890, is not lending-restricted and publishes its full OCR as a downloadable _djvu.txt;  it is what finally settled the Figure 41 caption.  Two further traps sit behind the ones that "refuse to lend."  The Internet Archive allows only one browse-only loan at a time, so while one book is open every other book answers "This book is not available to borrow at this time" — the same message an exhausted copy gives.  Return the loan and ask again before recording a refusal.

Leaf numbers are not page numbers.  The archive's reader indexes by leaf, and the offset varies by book.  The map comes from the item's own reader data:  for Berra, leaf 139 is printed page 117.  For ReMine's Biotic Message the offset is 8, verified on three hits whose printed pages were known independently.

Publisher-deposited abstracts are often reachable when the paper is not.  That route produced the whole Teeling and Hedges editorial, and the Salichos, Scally and Jenner abstracts, without fighting a paywall.

One more thing is worth recording, because it is a check nobody planned.  Two researchers audited the three Doolittle quotations separately, from different hosted copies, without comparing notes.  They reached the same verdicts on all three.  One had the 2006 book reprint; the other had the original magazine offprint, which pinned the page numbers the first could not map.  Agreement reached that way is worth more than agreement reached by two people reading the same file.

What this page changes

Three quotations leave the argument.  The Doolittle 2000 line, the Stanhope shark line and the Oakley line cannot be used as evidence at all, because their own context reverses them.  The Berra figure caption stays:  it now checks out in full, and it is the strongest piece of design-side material on the page, because Berra picked a series of manufactured objects to illustrate unguided descent.

Three change shape.  The Doolittle 1999 line survives only with the "If … then" and the closing "enriched, not impoverished."  The Rose line survives only beside "a tree of 51 per cent … Darwin's vision has triumphed."  The Syvanen line survives only as a claim about tunicates, at 40/60, from a published paper.

Several get stronger.  The best material in the note file was second-hand, and the primary papers say more.  Salichos and Rokas state 1,070 of 1,070 in their own abstract.  Jarvis and colleagues give 63 %, 66 % and 80 % average topological distance across 48 bird genomes.  Scally and colleagues give the gorilla 30 %.  Dávalos and colleagues write that "No special type of data can guarantee the lack of systematic bias."  Numbers from an abstract beat adjectives from a news site every time, and they are harder to argue with.

And one lesson generalises.  Of the roughly 150 quotation-bearing lines in this project's note file, a large share cite a press release, a magazine, a blog, a forum or an apologetics site rather than the paper.  Every defect found here entered by that route.  Nothing entered by somebody reading a journal and copying it out wrong.