Amino Acid Racemization

"An erroneously dated calibration sample will obviously yield erroneous racemization dates." — Jeffrey Bada, 1985, explaining in print why his own published ages for Californian skeletons were wrong by a factor of ten.  Bada invented the method.¹

"It is suggested that the apparent ages obtained by this method are unreliable and, with few exceptions, are much older than the real ages." — Duane Gish, Institute for Creation Research, 1975, nine years before accelerator radiocarbon proved him right about those particular ages.²

"Morphological, immunological and spectroscopic evidence of preserved tissues in dinosaurs and other fossils seems to be inconsistent with the observed levels of hydrolysis, dehydration and racemization in intracrystalline proteins from the fossil mollusc shell and eggshell." — Beatrice Demarchi and thirty-one co-authors, eLife, 2016.³

Summary

Amino acid racemization is a chemical clock.  Living things build proteins out of left-handed amino acids.  After death those molecules slowly flip to the right-handed form until the two are present in equal amounts.  Measure how far that has gone and, if you know the rate, you get a time.

The rate is the problem.  It doubles roughly every 3 °C.

1. The method failed publicly and its founder wrote the post-mortem himself. Jeffrey Bada dated the Del Mar skeleton from southern California to 48,000 years and the Sunnyvale skeleton to 70,000. Accelerator radiocarbon on the same bones returned about 5,000 years for both. That is an error of nine- to nineteen-fold, all of it in the direction of false antiquity. Bada co-authored the paper that overturned his own dates and then published the explanation in Annual Review of Earth and Planetary Sciences. A second calibration failure sits at the foot of the method, and that one was fraudulent: Bada's 1973 co-author Reiner Protsch was forced out of the University of Frankfurt in 2005 for falsifying dates, and his racemization ages for three German skulls match his own radiocarbon ages for the same skulls to within about five per cent, while both are wrong by up to a hundredfold.
2. The cause was not mainly temperature. It is tempting to say that a reaction this sensitive to heat could never work, and that is an overstatement. Bada's own Arrhenius fit says a 2 °C mistake in the assumed burial temperature inflates an age by 63 per cent, not by a factor of ten. The tenfold error needed two compounding mistakes: a calibration bone assumed to be 17,150 years old that is really about 5,000, and a rate constant that is not transferable between bones that have leached to different degrees. Both are Bada's own diagnoses.
3. He did not abandon the method, and he was not obliged to. In the same paper he wrote that racemization ages "seem generally to be reasonably accurate." Anyone who presents this episode as a retraction of racemization dating is describing something a reader can disprove in ten seconds.
4. The method was later rebuilt, and the rebuild works. Kirsty Penkman's group stopped analysing whole shells and started analysing only the protein sealed inside the mineral crystals. At 140 °C for 24 hours that fraction loses 1 per cent of its amino acids where a whole shell loses 40. Ranking 74 British sites by it reproduces the order given by river terrace heights, by two independent mammal biostratigraphic markers, and by the interglacial stratotypes. The authors are explicit that it is a relative ranking and not an age in years.
5. The chemistry that makes the rebuild work also sets a hard ceiling on protein survival, and this is where the young-earth case is strongest. Running Bada's published rate law forward, a fossil buried for 70 million years at any temperature at or above freezing should be racemic to six decimal places. Mainstream authors say the same thing in print, and in 2016 a mainstream paper conceded that reported dinosaur soft tissue "seems to be inconsistent with the observed levels of hydrolysis, dehydration and racemization" that its own laboratory measures.
6. The opposite young-earth argument, that low D/L ratios prove fossils are young, fails, and this article says so in its own voice — including in its two strongest published forms, which are not Truman's. Larry Helmick argued in 1976 for a kink in the racemization-versus-depth curve at the Flood boundary; the reversible rate law produces that kink on its own, and his one supporting core is one he himself reports two laboratories reading in opposite directions. R. H. Brown showed in 1985 that the apparent rate constant falls about 2,000-fold across the column and asked whether the ages should be shortened instead; but a constant fitted from a quantity that saturates must fall in inverse proportion to the assumed age, and that artefact alone is larger than the fall he measured. Royal Truman's Journal of Creation paper recalibrates published measurements to Flood-era ages using a rate constant measured at 49.5 °C, in the same paper in which he writes that "the claim that D/L ratios are reliable indicators of age is incorrect." A method too unreliable to give old ages is too unreliable to give young ones. And Bada measured Tyrannosaurus and Edmontosaurus bone himself, finding highly racemized amino acids, which is what deep time predicts and not what a 4,500-year-old bone would show.

The verdict.  The Californian episode is a genuine and instructive failure of a calibrated dating method, and the lesson generalises: the calibration sample carries the whole result, and an error there is invisible to any amount of internal consistency.  The intracrystalline rebuild is real and must be conceded.  The survival ceilings that racemization chemistry imposes on ancient protein are the strongest young-earth material in this topic.  Re-dating fossils young from D/L ratios is not an argument anyone should make, and this site will not make it.

Every source this article rests on is catalogued, with what it measured and which way it cuts, in a companion page: the literature.

What racemization is

Twenty amino acids build almost all life.  Nineteen of them come in two mirror-image forms, left-handed (L) and right-handed (D), like a pair of gloves.  Living cells use the L form almost exclusively.

Once an organism dies, the enzymes that maintained that purity stop working.  The chemical bond holding the hydrogen on the central carbon can break and re-form on the other side.  Each amino acid has an equal chance of flipping either way, so given enough time the population drifts to a 50-50 mixture.  The measured quantity is the ratio of D to L, written D/L, and it runs from about 0 in fresh tissue to 1 at equilibrium.

Aspartic acid racemizes fastest of the common amino acids and is the one used for bone.  Valine is among the slowest.  That spread is useful: a sample can be checked for internal consistency by asking whether the fast and slow amino acids give the same answer.

The rate law is a reversible first-order reaction.  Bada wrote it down in 1972 and restated it as Equation 2 of the 1973 paper that founded bone dating:⁴

ln[(1 + D/L)/(1 - D/L)]_t  -  ln[(1 + D/L)/(1 - D/L)]_0  =  2 k t

which rearranges to D/L(t) = tanh(k t + atanh(D/L at t = 0)).

Everything then depends on k, and k obeys the Arrhenius equation.  Bada's own least-squares fit, to heated modern bone, fossil bone, and living human teeth, is printed in the caption to Figure 4 of his 1985 review:¹

ln k_asp (yr^-1)  =  56.6  -  19.7 x 10^3 / T(K)

The slope is the activation energy over the gas constant, so Ea = 19,700 x 1.987 = 39.1 kcal/mol, or 163.8 kJ/mol.  On the facing page Bada gives the measured value separately as "about 38 ± 3 kcal mol⁻¹."¹

That is a steep number, and Bada spelled out what it means before anyone had to point it out to him:¹

"An activation energy of 35-40 kcal mol⁻¹ implies that variations of ± 1° in the temperature of the depositional environment of a fossil bone would effect the rate of racemization in the fossil by about ± 25%."

An age from this method is t = atanh(D/L) / k.  It scales as 1/k.  A rate error of a given size is an age error of the same size, exactly.  So an uncalibrated racemization age is not a measurement of elapsed time.  It is a measurement of integrated thermal history, and it only becomes a time when somebody supplies the temperature from outside the measurement.

The arithmetic, worked in the open

Every number in this section was computed for this article from Bada's own published Arrhenius fit, using the Deno TypeScript program at research/racemization-kinetics.ts.  Its full output is saved beside it as racemization-kinetics.out.txt, so a hostile reader can redo the whole thing.⁵

How long aspartic acid in bone takes to reach D/L = 0.5, starting from zero:

Read the first and last rows together.  The same reaction takes 46.9 million years in permafrost and 2,400 years in a warm soil.  That is a factor of twenty thousand across the range of temperatures at which bones are actually buried.

A check that the model is being run correctly.  Bada and Protsch stated in 1973 that aspartic acid in bone at 20 °C reaches D/L = 0.333 in "about 15,000-20,000 years."⁴  This model gives 13.9 kyr with no starting offset, and 11.1 kyr using their own D/L at t = 0 of 0.07.  That is the right order of magnitude and about 25 to 45 per cent low.  The discrepancy is honest and has a cause: the 1985 Arrhenius fit includes fossil-bone and living-tooth data points that the 1973 estimate did not.  Nothing below turns on the difference.

How much an error in the assumed temperature costs.  This is the number the whole dispute turns on, so here it is against a 10 °C base:

The 1 °C row reproduces Bada's own published figure of about ± 25 per cent, coming out at 28.  Across his stated uncertainty on the activation energy, 38 ± 3 kcal/mol, the 2 °C row runs from × 1.55 to × 1.67, so the conclusion does not depend on which activation energy is taken.

Say plainly what this does and does not show.  A 2 °C mistake moves an age by about 60 per cent.  It does not move it by a factor of ten.  Anyone arguing that temperature sensitivity alone sank the Californian dates is overstating the case, and the next section shows what actually did it.

And now the number that matters most in this article.  Run the same rate law forward for 70 million years, the age assigned to the youngest dinosaur-bearing rocks:

At any burial temperature at or above freezing, a genuinely 70-million-year-old sample should be racemic to the limit of measurement.  Not "quite racemized."  Racemic to six decimal places, with a thousandfold margin at 0 °C and a hundred-thousandfold margin at 10 °C.  Only permanently sub-freezing burial produces anything else, and that is an extrapolation into a regime with no liquid water, where a rate law measured in wet bone may not apply at all.  Bada makes the same caveat himself, and it points the same way.⁶

The inverse question, since the young-earth argument in section 3(b) below depends on it.  What age does a genuinely low D/L imply?

Look across any row.  The same measurement gives answers two orders of magnitude apart depending on the temperature fed in.  That is the free parameter, and it is available to anyone who wants an answer, in either direction.

The caveat that has to travel with all of it

The simple first-order picture above is not the whole chemistry, and the people who took it apart were not creationists.

Matthew Collins, Ed Waite and Adri van Duin published the demolition in the same 1999 issue of Philosophical Transactions of the Royal Society B that carried Bada's own survival paper.⁷  Their finding is that aspartic acid does not racemize freely in bone.  It has to pass through a strained ring-shaped intermediate, a succinimide, and that ring cannot form while the amino acid is locked inside an intact collagen triple helix.  So the D/L a laboratory measures in fossil bone "reflects the ratio of non-helical (high proportion of D-Asx) to triple helical (L-Asx) collagen."  It is a measure of how far the collagen has unwound, not a clock.⁷

Their verdict, printed on page 60:⁷

"Despite the large number of experiments which have been conducted to derive kinetic parameters for Asx racemization, when applied to fossil material the pattern and rates are unpredictable."

And from their numbered summary on the same page:⁷

"(i) estimating rates of Asx racemization by extrapolation from high temperatures is to be avoided. At low temperatures, Asx racemization will remain unpredictable due to the overwhelming influence of protein conformation of the rates of bound residues; … (iii) although they are the most widely used archaeological materials for Asx dating, bones and teeth are probably the worst materials to choose, due to the extreme conformational effect of the collagen triple helix; (iv) leaching distorts the kinetics by removing the peptides and free amino acids, which racemize more predictably."

This cuts against both sides with equal force, and that is why it is here rather than buried in a footnote.  It explains why the Californian bone dates failed.  It also means that any young-earth argument leaning on a precise extrapolated racemization half-life for bone is standing on the same sand.  The tables above are good for orders of magnitude.  They are not good for two-significant-figure claims about a particular bone, and this article does not use them that way.

The founding claim, and the assumption printed in its own abstract

Bone dating by racemization begins with one paper: Jeffrey Bada and Reiner Protsch, PNAS, May 1973.⁴

Their method was calibration by a known age.  Take one bone whose radiocarbon age you trust, measure its D/L, and solve for k.  Then use that k on every other bone from the same region.  The known bone was from the Naisiusiu Beds at Olduvai Gorge, with a collagen radiocarbon age of 17,550 ± 1,000 years and D/L(Asp) = 0.32, giving k = 1.48 ± 0.09 × 10⁻⁵ yr⁻¹.  A bone from the older Ndutu Beds at D/L = 0.72 then came out at 56,000 ± 3,500 years, and the Eyasi I hominid skull at D/L near 0.5 came out at 34,000.⁴

They did not hide the weak point.  It is the second sentence of their abstract:⁴

"The only assumption required with this approach is that the average temperature experienced by the 'calibration' sample is representative of the average temperature experienced by older samples."

And again in the body, with the numbers attached:⁴

"It should be emphasized that the only assumption concerning the temperature, and therefore the value of k_asp, made in this calculation is that the average temperature at Olduvai Gorge over the last 18,000 yr is representative of the average temperature over the last 56,000 yr."

They thought the assumption safe.  Their reason was that glacial-cycle temperature swings recorded in the tropical Atlantic and in a New Zealand cave came to "only about 0.5°" over 20,000 years compared with longer spans.

Twelve years later Bada published his own measurement of that same quantity at that same site, and it was not 0.5 °C.  Using the Holocene and Upper Pleistocene rate constants from his own Table 2, he calculated the difference in ground temperature between the two periods at Olduvai:¹

"we calculate a value of ΔT ≈ 7-8°C, which suggests that there has been a substantial difference in the ground temperatures at Olduvai throughout the Holocene and Upper Pleistocene"

By the sensitivity table above, 7 °C is a factor of about six in the rate.  The one assumption the founding paper said it required turned out, on the founder's own later measurement at the founding site, to be wrong by more than an order of magnitude in the quantity that matters.

The co-author was later found to have falsified dates

This has to be said plainly, because a skeptical reader will find it without help.  It also turns out to support what this article argues rather than to undermine it, which is why it is here in the body and not buried.

Bada's co-author on the 1973 paper was Reiner Protsch von Zieten of the University of Frankfurt.  On 18 February 2005 the university announced that Protsch had been forced to retire over what it called "falsehoods and manipulations."  Luke Harding reported it in The Guardian the next morning and named the man who ran the inquiry:⁸⁸

"'It's deeply embarrassing.  Of course the university feels very bad about this,' Professor Ulrich Brandt, who led the investigation into Prof Protsch's activities, said yesterday.  'Prof Protsch refused to meet us.  But we had 10 sittings with 12 witnesses.'"

The redating was not done by the inquiry.  It was done by Oxford:⁸

"But his university inquiry was told that a crucial Hamburg skull fragment, which was believed to have come from the world's oldest German, a Neanderthal known as Hahnhöfersand Man, was actually a mere 7,500 years old, according to Oxford University's radiocarbon dating unit.  The unit established that other skulls had been wrongly dated too."

"Another of the professor's sensational finds, 'Binshof-Speyer' woman, lived in 1,300 BC and not 21,300 years ago, as he had claimed, while 'Paderborn-Sande man' (dated at 27,400 BC) only died a couple of hundred years ago, in 1750."

Now the part that matters here, and it is not the scandal.  Protsch did not only radiocarbon-date those three skulls.  Racemization ages for the same three were published under his name, and the two sets agree with each other.  The creationist writer Peter Line set the two columns side by side in 2005, which is where this article met the point.⁹⁹  In the table below, the first two columns are Protsch's own published determinations as Line reports them, and the third is Harding's.  Harding independently prints the same two claimed radiocarbon ages, 21,300 and 27,400 years, so only the racemization column rests on Line alone.

Two methods that share no physics, agreeing to within about five per cent, and both wrong by up to a factor of a hundred.  Whether the racemization numbers were measured badly or written to match, nobody has established, and this article does not guess.  What can be said is the methodological point, and it is worth more than the guess:  agreement between two dating methods is not confirmation when both numbers came from one hand.  That is routinely offered as proof that both are right.  Here it is proof of nothing.

Three things follow.

First, what is not touched.  Protsch's conduct says nothing about Bada's chemistry.  The rate law, the Arrhenius fit, the Californian post-mortem and the whole intracrystalline rebuild are independent of him.

Second, what is weakened.  The 1973 calibration rests on a single radiocarbon determination, UCLA-1695, and the paper says that age was obtained "by use of the method described by Protsch."  This article does not claim that particular date was faked.  Nobody has shown it, and asserting it would be the same move this article objects to elsewhere.  The narrower claim is enough:  the founding calibration is one unreplicated determination by a man later found to have falsified determinations, and it should be called unconfirmed rather than established.

Third, and this is why it belongs in an article about calibration.  It is a second, independent instance of the failure this whole article is about.  In California the calibration bone was honestly misdated, and the racemization ages came out nine to nineteen times too old.  Here the calibration was fraudulent, and the racemization ages came out up to a hundred times too old.  Different causes, same mechanism:  the calibration sample carried the whole result.

And note the direction, because it runs against the side this article is on.  Every error in this story is too old.  A young-earth argument from racemization needs the method to make old things look young.  Its documented failures, honest and dishonest alike, all make young things look old.

What went wrong in California

This is the best-documented self-correction in the whole of geochronology, and none of it is creationist work.  Physicists with a new machine caught it, and the man who invented the method wrote up why he had been wrong.

The claim

In 1974 and 1975 Bada and colleagues applied bone racemization to a set of Californian palaeoindian skeletons.  R. E. Taylor's 2009 review in Radiocarbon¹⁰ lists the results in one sentence:¹⁰

"AAR-based age estimates assigned ages of 70,000 yr to the Sunnyvale (northern California) skeleton, >50,000 to the Haverty or Angeles Mesa skeleton (southern California), 48,000 yr to the Del Mar skeleton (southern California) and, apparently confirming an earlier ¹⁴C value, 23,000 yr for the Yuha skeleton (southeastern California), although U-series-based determinations indicated an age of about 19,000 yr"

Bada's own later summary of what those numbers meant at the time:¹

"In 1974 and 1975, a series of aspartic acid racemization dates for several California paleoindians were determined, and the ages of skeletons from sites at Del Mar and Sunnyvale were estimated to be 40,000-60,000 yr (Bada et al 1974, Bada & Helfman 1975). These data seemed to provide some of the first direct evidence that human beings had migrated into the Americas considerably earlier than the generally accepted time of about 13,000 yr ago."

This mattered.  It was the leading physical evidence for a pre-Clovis peopling of the Americas.

The redating

Accelerator mass spectrometry made it possible to radiocarbon-date a few milligrams of purified collagen rather than grams of whole bone.  Taylor again, same page:¹⁰

"In each case, the redating determined that all of these human skeletal samples previously assigned Pleistocene ages were, in fact, of Holocene age. For example, the 48,000-yr AAR-deduced age of Del Mar was determined on XAD-purified collagen fractions to range from 5380 ± 390 BP (AA-2666) to 4830 ± 200 BP (AA-2665), while the AAR-deduced 70,000-yr Sunnyvale skeleton was assigned ages ranging from 6300 ± 400 BP [OxA-187] to 3600 ± 600 BP [UCR-1437A/AA-50]"

The arithmetic, done here from those published numbers:

Note who published the correction.  Bada was a co-author of the 1984 Nature paper that overturned his own dates.¹¹  Whatever else is true of this episode, the man did not defend a dead result.

Bada's own explanation

The post-mortem is in his 1985 review in Annual Review of Earth and Planetary Sciences¹, and it is worth quoting at length because it is unusually plain-spoken.

First, the finding:¹

"The AMS-based ¹⁴C dates indicate that all of the skeletons previously assigned an Upper Pleistocene age are in fact more likely to be Holocene in age."

Then the first cause.  The calibration bone was the Laguna skeleton, whose conventional radiocarbon age had been 17,150 years.  It is not:¹

"The original racemization dates for the Del Mar and Sunnyvale skeletons were based on the 17,150 ± 1450 yr age for the Laguna skeleton. However, as can be seen in Table 5, the AMS ¹⁴C age for this skeleton indicates that it is only ~5000 yr old."

And then the general statement of the lesson, which is the single most valuable sentence in this article:¹

"The racemization results of the California paleoindian skeletons demonstrate that one of the essential factors in determining the validity of the racemization date for any particular fossil bone is the reliability of the age estimate of the calibration sample. An erroneously dated calibration sample will obviously yield erroneous racemization dates. If the Laguna and Los Angeles Man skeletons had been originally assigned the Holocene ages recently determined using the AMS-based radiocarbon method, skeletons such as those from Del Mar and Sunnyvale never would have yielded racemization ages of great antiquity."

The second cause is leaching, and it is separate.  Bones do not hold their amino acids equally well, and the ones that have lost most racemize faster:¹

"in samples where the level of amino acid preservation is < 5%, the k_asp value is clearly greater than when the degree of amino acid preservation is > 5%. This suggests that when the collagen has undergone extensive hydrolysis, the peptide residues that remain have an aspartic acid racemization rate that is greater than that in bones in which the collagen is better preserved."

He is candid that the reasons are not settled:¹

"It is not known what factors have contributed to the difference in the extent of collagen preservation in the California paleoindian skeletons of approximately the same age, but they probably include leaching by groundwaters and/or heating effects resulting from either shallow burial depths or cultural burial practices. Also, many of these California paleoindian skeletons come from coastal shell middens in which the presence of carbonates in the burial soils may have accelerated the rate of collagen hydrolysis."

The rule he drew from it is a restriction on when the method may be used at all:¹

"The results in Figure 5 also indicate that a particular calibration constant should only be used to date other fossil bones having amino acid contents similar to that of the calibration samples."

How big was each cause?

This is where a careless article would overclaim, so here is the arithmetic.

Bada's revised calibration constant, averaged from the AMS ages of the Laguna and Los Angeles Man skeletons, was k = 6.043 × 10⁻⁵ yr⁻¹.  The original was about 1.5 × 10⁻⁵.  That is a factor of about 4 in the rate, and therefore a factor of about 4 in every age derived from it.

Inverting Bada's Arrhenius fit puts a scale on that.  A rate constant of 1.5 × 10⁻⁵ yr⁻¹ corresponds to a constant burial temperature of 17.8 °C, and 6.043 × 10⁻⁵ to 23.9 °C.  The whole correction is the size of a 6.1 °C error in assumed burial temperature.⁵

A factor of 4 is not a factor of ten.  The rest came from the calibration bone itself being misdated by more than a factor of three, 17,150 years assumed against about 5,000 measured.  Multiply the two and you land in the right range.

So the honest statement is: two compounding errors, not one.  A wrong calibration age, and a rate constant that is not transferable between bones of different preservation.  Temperature sensitivity is the reason both errors were amplified rather than damped, but temperature guessing on its own would not have produced a tenfold result.

What Bada did not do

He did not abandon the method.  Four pages after the sentence quoted above, in the same review:¹

"The ages derived from the extent of amino acid racemization in a fossil bone seem generally to be reasonably accurate, and this method provides an important geochronometer that can be used to estimate the age of fossil bones and teeth over a time period ranging from a few thousand to several million years."

That has to be quoted too.  Presenting Bada as having renounced racemization dating would be a misrepresentation that a skeptical reader can check in one search, and it would rightly cost this article everything else it says.

His position was that the chemistry was sound and the calibration had been bad.  A reader is free to think that too generous.  But it is his position, and this is what he wrote.

The 1975 creationist reading: right verdict, wrong mechanism

Duane Gish published a critique of racemization dating for the Institute for Creation Research on 1 May 1975.²  That is nine years before accelerator radiocarbon reversed the Californian dates.  It holds up better than most creationist geochronology of that decade, and it is worth grading carefully because the grade is split.

What Gish got right.  He identified temperature as the load-bearing assumption and gave a number close to the correct one:²

"The rate of racemization is highly temperature dependent. The study with bone indicated that an uncertainty of ±2° would yield an age with an error of ±50%."

Against this article's computed 63 per cent for 2 °C, and Bada's own 25 per cent for 1 °C, ±50 per cent is in the right place.  And his conclusion was vindicated within a decade:²

"There is no doubt that proteins in bone and shell and other fossil material undergo hydrolysis and that the amino acids contained in them suffer racemization with increasing age of fossil material. To use rates of racemization as a dating method, however, the entire history of the fossil material would have to be known, including temperature and the entire diagenetic process, especially the chemical environment that contributed to this process, and most especially the pH. Since all of these factors, most of which accelerate racemization rates, cannot be known, it is suggested that the apparent ages obtained by this method are unreliable and, with few exceptions, are much older than the real ages."

Unreliable, and skewed old.  That is exactly what the Del Mar and Sunnyvale skeletons turned out to be.

What Gish got wrong.  Read that quotation again and notice where he puts the emphasis: "most especially the pH."  His central mechanism is that acidity, not temperature, dominates racemization in real fossils.

Bada ran the direct experiment and reported it in the same 1985 review.  In heated modern bone, "in the range pH 3-8, racemization rates are not affected by pH," because "the bone minerals effectively buffer the system involved in racemization."  Bone is its own buffer.  Outside that range the mineral starts to dissolve, but as Bada notes, bones are not preserved in those environments.¹

So the creationist got the verdict right and the mechanism wrong, and the difference matters.  A right conclusion reached by a wrong argument is not a vindicated argument.  It is a coincidence that happens to have come out well, and treating it as a success is how a movement fails to learn anything.  The correct reading of Gish 1975 is that his instinct about calibration and unknowable histories was sound, his chemistry was not, and the part that was sound was the part that survived.

The rebuild, and it works

If the story stopped in 1985 it would be a clean case study in a failed method.  It does not stop there, and an article that left it there would be misleading.

The diagnosis

Bone is an open system.  Amino acids leach out through groundwater, contaminating modern amino acids leach in, and a leached sample racemizes at a different rate from an unleached one.  Collins showed why that is fatal to the kinetics, and Bada's Figure 5 showed it happening in the Californian bones.

Kirsty Penkman's group at York took the obvious next step: throw away everything that can move, and analyse only the protein sealed inside the mineral crystals themselves.

The measurement

Their 2008 paper in Quaternary Geochronology¹² tested whether the trapped fraction really is closed.  From the abstract:¹²

"Exposure of powdered mollusc shells to concentrated NaOCl for 48 h effectively reduced the amino acid content of the four taxa to a residual level, assumed to represent the intra-crystalline fraction. When heated in water at 140 °C for 24 h, only 1% of amino acids were leached from the intra-crystalline fraction of modern shells compared with 40% from whole shell. Free amino acids were more effectively retained in the intra-crystalline fraction, comprising 55% (compared with 18%) of the whole shell after 24 h at 140 °C. For fossil gastropods, the inter-shell variability in D/L values for the intra-crystalline fraction of a single-age population was reduced by 50% compared with conventionally analysed shells."

One per cent against forty.  That is not an assertion that the system is closed.  It is a measurement that it is closed, under conditions far harsher than burial, and it is the reason the rebuilt method deserves to be taken seriously.

The authors state their own limits.  The bleach treatment does not work on the bivalve Corbicula fluminalis, "possibly due to variability in shell ultrastructure."  And a stubborn bound fraction survives that "has never been satisfactorily explained," which is why "very few studies report true equilibrium D/L values of 1.0."¹²

The test

The 2011 Nature paper is where the rebuilt method faces independent data.  Penkman and seven co-authors measured 470 opercula, the little calcite trapdoors of the freshwater snail Bithynia, from 74 British sites spanning the whole Quaternary.  They combined aspartic acid, glutamic acid, serine, alanine and valine, plus serine dehydration, into a single index of Intra-crystalline Protein Decomposition.¹³

Then they did not convert it into years, and they say why:¹³

"Sites with independent geochronology can be used to calibrate IcPD results but, as only 7 of our pre-Holocene sites have associated dates, we have not done this here."

That sentence is the fairest thing in the paper and the one most often dropped when it is cited.  What the 2011 paper delivers is a ranking of 74 sites, oldest to youngest.  It then checks that ranking against four things measured by entirely different means:

1. The stratotypes. "All stratotypes yielding Bithynia have been analysed."
2. River terrace elevation. "Increasing protein decomposition within opercula is also consistent with increasing river terrace elevation (and therefore age)… Assemblages of amino acid values from the Thames sequence, the most complete British fluvial archive, clearly correspond with discrete terrace aggradations." The authors flag themselves that this is not fully independent.
3. Mammal biostratigraphy. Two markers, and they are the good kind, because they are mutually exclusive rather than gradational. "Opercula from sites yielding Arvicola show less protein degradation than those containing M. savini", tracking the water-vole turnover. And "post-Anglian sites with Hippopotamus show less protein breakdown than sites yielding Corbicula."
4. Radiocarbon, which it overturned. At Isleworth a radiocarbon date of about 43,000 years had been taken as the age of the deposit. The protein index there is higher than at a site correlated with a considerably older interglacial, so "Radiocarbon therefore provides only a minimum age for the Isleworth deposits."

Their summary: "The comparisons above demonstrate the remarkable consistency of our new method with independent lines of evidence."¹³

Grading it honestly

Concede this, and concede it fully.  Four orderings, derived from protein chemistry, from river terrace heights, from two mammal turnovers, and from a set of type sections, agree with each other across 74 sites.  Getting four independent orderings to agree by accident is hard, and no young-earth model on offer predicts that they would.  A young-earth reader should find this uncomfortable, and being told otherwise would be a disservice.

And state the limits with equal force.

• It is a relative ranking. It inherits the marine isotope stage timescale rather than testing it, and its own authors decline to convert it to years.
• It still rests on a thermal assumption, that "similarly small thermal gradients existed during past interglacials, when most of the racemization would have occurred."¹³ That is the same class of assumption that sank the bone method. The difference is that Britain is small and thermally homogeneous, and the authors argue the gradient is small, which is a much weaker demand than assuming Olduvai's temperature was constant for 56,000 years.
• It works on calcite opercula and eggshell, not on bone. It rehabilitates none of the old bone dates, and nobody claims it does.

The young-earth reply, and it is not silly

One creationist writer answered this paper, and answered it quickly.  Elizabeth Mitchell reviewed it for Answers in Genesis on 13 August 2011, under the title "A Snail's Tale of the Past."¹⁴  An earlier draft of this article said nobody in the creationist literature had engaged the 2011 result.  That was false, and the correction is printed here rather than quietly removed.

Her objection has two parts.

Part one:  the calibration is circular.¹⁴

"The problem with this new technique lies in its foundational assumptions.  All time-measuring techniques must be calibrated against a reliable standard.  Here, the protein-based chronometer is calibrated according to uniformitarian assumptions about earth's icy past.  In other words, each degradation ratio corresponds to a particular layer, and the dates claimed for that layer are then assigned to the degradation ratio."

Part two, and this is the interesting one:  the agreement is explained by shared temperature sensitivity.¹⁴

"Since oxygen ratios and amino acid racemization are both temperature sensitive, they correspond and vary according to the seasonal and climatic changes they were exposed to.  But neither tells us anything about the duration of those variations, and neither confirms the notion of multiple ice ages over millions of years."

Part one is granted, in the same words this article already used.  The IcPD scale is not a chronometer.  Penkman's own sentence is that only 7 of the pre-Holocene sites have independent dates and "we have not done this here."  Anyone reporting the 2011 paper as a set of ages in years is misreporting it, and Mitchell is right about that.  But she is answering a claim the paper does not make.  What is on the table is an ordering, and an ordering does not need a calibrated timescale to be a result.

Part two does not work, and the reason is in the chemistry set out above.

Racemization is a one-way approach to equilibrium.  D/L rises and never falls.  Cooling a shell slows the reaction;  it does not run it backwards.  The oxygen-isotope ratio in a marine core is the opposite kind of quantity — a state that goes up and down with ice volume, over and over, which is why the marine isotope stages are numbered as alternating warm and cold intervals instead of counted off in one direction.  One is a cumulative total, the other is an oscillation.  Making both sensitive to temperature does not make them correlate.  A car's odometer and its speedometer both respond to the accelerator, and they do not track each other.

Mitchell's model also makes a prediction, and it loses.  If the protein state inside an operculum recorded warmth rather than accumulated time, then opercula from two equally warm intervals should read about the same.  Penkman's own faunal test supplies exactly that pair.  Hippopotamus and Corbicula both need warmth, and both mark warm intervals in post-Anglian Britain, and the two faunas do not occur together.  The measured result is that "post-Anglian sites with Hippopotamus show less protein breakdown than sites yielding Corbicula."¹³  Two warm-climate faunas, different protein states, separating cleanly.  A thermometer does not do that.  An accumulating quantity does.

And δ¹⁸O is only one of the four orderings, and the weakest of them for this purpose.  The other three are not temperature records at all.

• River terrace height records a valley cutting down. A terrace sits higher because the river had longer to incise below it. Nothing in that is thermal.
• The water-vole turnover, Mimomys savini giving way to Arvicola, is a faunal replacement. Water voles do not un-evolve when the climate cools, and the two forms are mutually exclusive rather than gradational, which is the whole reason the marker is worth anything.
• The Hippopotamus/Corbicula split works the same way, as above.

So Mitchell's move, taken at its strongest, dissolves the agreement with δ¹⁸O and leaves the agreement with superposition and with biostratigraphy standing.  Those are the orderings that carry the argument, and a shared thermometer cannot produce them.

What would actually answer Penkman, and what nobody has done.  Showing the calibration is circular is not enough, because the ranking survives without the calibration.  An answer has to explain why an irreversible chemical index, measured on 470 objects from 74 places, comes out in the same order as river-terrace incision and as two mammal turnovers.  That work has not been done by anyone.

The two young-earth arguments, and why they must be kept apart

Young-earth writers use racemization in two ways.  They are different claims, they rest on different evidence, and one is far stronger than the other.  Almost all of the confusion in this topic comes from running them together, so they are taken separately here and graded separately.

3(a). The protein-survival argument. This is the strong one.

The claim: amino acids and peptides should racemize to equilibrium and hydrolyse away long before the ages assigned to the rocks that contain them.  If intact protein turns up in a 68-million-year-old bone, something in the picture is wrong.

Its strength is that it is stated most forcefully by mainstream chemists, in mainstream journals, from their own measurements.  It does not need a creationist source at all.

Bada, Xueyun Wang and Healy Hamilton published the survival ceilings in Philosophical Transactions of the Royal Society B in 1999.⁶  On hydrolysis:⁶

"In the carbonate matrix, proteins are nearly completely hydrolysed to smaller peptides and free amino acids in about 10⁶ years on the ocean floor, and in around 10⁵ years in surface environments. In bones, hydrolysis of the main protein component, collagen, is even more rapid and little intact collagen remains after only 1-3 × 10⁴ years, except in bones in cool or dry depositional environments."

On racemization:⁶

"If no contaminants have been introduced into the system, based on the racemization half-lives determined from known age fossils, all amino acids should be totally racemized in <5-10 million years in cold depositional environments; in temperate regions a racemic mixture of all amino acids would be attained in <1 million years"

And in a figure caption, as compactly as it can be put:⁶

"Racemization should be complete in <1-5 million years in most environments on the Earth."

Those numbers are measured, not assumed, and this article's own independent computation from Bada's Arrhenius fit agrees with them: at 10 °C, k·t over 70 million years is 162, which is about a hundred and sixty times the amount of reaction needed to reach equilibrium.

Now the part young-earth writers usually skip, and it is decisive.

Bada measured dinosaur bone himself.  A Tyrannosaurus from Montana and an Edmontosaurus from the North Slope of Alaska.  Here is what he found:⁶

"In summary, the dinosaur eggshells and bones we analysed contain only trace amounts, if any, of their original amino acids, along with exogenous components. Any endogenous amino acids that are present have undergone extensive diagenesis as evidenced by the presence of highly racemized amino acids. The general lack of unaltered, original amino-acid components and the presence of contaminants in these dinosaur samples implies that the possibility is remote that other biomolecular information such as nucleic acid and protein sequences would be preserved."

Highly racemized.  That is what 70 million years predicts.  It is not what 4,500 years predicts.  A popular young-earth claim that dinosaur amino acids are found unracemized is contradicted by the only direct measurement of the question by the field's leading practitioner, and this site is not going to print an argument its own evidence refutes.

What survives is narrower, and it is real.  Bada's last clause in that quotation, written in 1999, is a prediction: recovering protein sequences from dinosaur material is remote.  Since 2007 several laboratories have reported exactly that.  Whether those sequences are endogenous is a separate fight with its own literature, and this site handles it at dinosaur soft tissue.  It is not re-litigated here.  What belongs here is one narrow point: a peptide sequence and an unracemized amino acid pool are different objects, and evidence for one is not evidence for the other.

The mainstream answer, and it must not be waved away

The best mainstream response is a specific mechanism with a computed prediction that came true, and it is in eLife in 2016.³

Beatrice Demarchi and thirty-one co-authors, including Penkman and Collins, recovered peptide sequences from ostrich eggshell at Laetoli, dated 3.8 million years, and Olduvai Gorge, dated 1.3 million years.  From the abstract:³

"Here, we demonstrate the role of mineral binding in preserving the protein sequence in ostrich (Struthionidae) eggshell, including from the palaeontological sites of Laetoli (3.8 Ma) and Olduvai Gorge (1.3 Ma) in Tanzania… Thermal age calculations demonstrate that the Laetoli and Olduvai peptides are 50 times older than any previously authenticated sequence (equivalent to ~16 Ma at a constant 10 °C)."

The mechanism is not a hand-wave.  They ran molecular dynamics on the eggshell proteins struthiocalcin-1 and struthiocalcin-2, computed which domain binds most tightly to the calcite crystal surface, and then found that this is the domain that survives.  A prediction was made from physics and confirmed by measurement.  Any young-earth writer who claims there is no proposed mechanism for deep-time protein survival is not describing the current literature.

The same paper also states the tension in print, which is why it appears at the top of this page:³

"Morphological (Towe and Urbanek, 1972; Bertazzo et al., 2015), immunological (Schweitzer et al., 2013) and spectroscopic (Bertazzo et al., 2015) evidence of preserved tissues in dinosaurs and other fossils seems to be inconsistent with the observed levels of hydrolysis, dehydration and racemization (Penkman et al., 2013) in intracrystalline proteins from the fossil mollusc shell (Sykes et al., 1995) and eggshell (Brooks et al., 1990)."

And here is the detail that neither side should drop.  Those surviving Laetoli peptides are racemized.  From the caption to Figure 2:³

"Linear increase of THAA Val D/L values with the log of thermal age… The average hydropathicity of the peptides identified remains stable up to Val D/Ls ~1. Note that Val D/Ls > 1.0 are unexpected and may be due to decomposition processes occurring in the closed system."

Valine is the slowest-racemizing amino acid in routine use.  A valine D/L of about 1 is full racemization.  So Demarchi 2016 says both of these at once: a short, tightly bound peptide sequence can survive to a thermal age around 16 million years at 10 °C, and it can do so while its amino acids are racemic.  "Racemized" and "no sequence information" are not the same claim, and this article will not treat them as one.

The thermal ages measured, from their Table 1, put the scale on it:

The honest form of the survival argument

Putting all of that together, here is what can be defended and what cannot:

• Defensible. Racemization to equilibrium within a few million years above freezing is measured, not assumed. Bada's own dinosaur samples show it. A genuinely 70-million-year-old fossil should show D/L near 1 for every amino acid, and this article's own computation from Bada's fit confirms it with a large margin.
• Defensible. Bulk collagen survival is a much shorter clock than racemization, on the order of tens of thousands of years in bone, and it is the one under real pressure from the reported dinosaur results.
• Defensible. The demonstrated survival mechanism reaches about 16 million years at 10 °C for a short, strongly bound peptide inside calcite. Nobody has demonstrated it to 70 million years, and nobody has demonstrated any mechanism for bulk collagen in bone at Hell Creek temperatures. That gap is roughly two orders of magnitude, and it is for a different kind of object.
• Not defensible. That dinosaur amino acids are unracemized. Bada measured them and they are not.
• Not defensible. That the gap proves a young earth. An unexplained preservation anomaly is an unexplained preservation anomaly. It is a genuine problem for the standard picture, it is worth pressing, and it is not a date.

3(b). The "racemization gives young ages" argument. This one fails.

The claim: measured D/L ratios in old-dated fossils are too low for the assigned ages, so the fossils are thousands of years old rather than millions.

The fullest published version is Royal Truman's Journal of Creation paper of 2023¹⁵, part 5 of a five-part series.  Parts 1 to 4 are about homochirality and the origin of life and belong elsewhere.  Part 5 is about dating, and its abstract states the goal without ambiguity:¹⁵

"We show that dates based on proportion of amino acid racemization can be reliably recalibrated to predict dates which do not exceed the age of the Genesis Flood."

His method has two steps.  First, take the Californian skeletons from Bada 1985's Table 5, pick the Yuha specimen as the calibration sample instead of Laguna, derive an average k of 1.95 × 10⁻⁴ yr⁻¹, and re-date everything with it, concluding that "if the new calibrated dates are correct, then most of these American Indians lived in the Christian era."  Second, for Pleistocene mollusc shells, adopt an alanine rate constant of 0.0116 yr⁻¹ measured by Steen and colleagues in seafloor sediment, and conclude that "according to figure 7, only about 150 years might have been necessary for the Pleistocene fossils to attain a D/L = 0.95 for alanine and not a million years."

That rate constant was measured at 49.5 °C.

Now name the contradiction, because it is in the same paper.  Truman also writes:¹⁵

"The claim that D/L ratios are reliable indicators of age is incorrect."

And:¹⁵

"Perhaps racemization slows down with degree of fossilization but by how much and when would be unknown. More accurate dates are unknowable without a complex, valid fossilization model."

And he quotes Collins approvingly:¹⁵

"I agree with Collins' assessment that 'Despite the large number of experiments which have been conducted to derive kinetic parameters for Asx racemization, when applied to fossil material the pattern and rates are unpredictable.'"

A method whose rate constants are unpredictable cannot then be recalibrated to yield reliable young dates.  The first half of Truman's paper destroys the second half.  This is said once and without gloating, because the first half is largely correct and well sourced, and the man did real work on it.

The specific arithmetic problem.  The 150-year figure comes from importing a rate constant measured at 49.5 °C into a shell supposed to have sat on a post-Flood shoreline.  Truman is candid that this is a supposition, writing that "49.5 °C might seem high, but this would be the effective and not average temperature."¹⁵  Candour does not repair it.  By the table computed above, a 2 °C move in assumed burial temperature changes an age by 63 per cent, and a 10 °C move changes it by a factor of ten.  A free parameter that ranges over 40 °C can produce any age anyone likes.

That is precisely the defect that made Bada wrong in 1974.  It does not become a sound method because the direction of the error is now convenient.

Two things Truman gets right, and they should be said.

He notices something real: across Bada's Table 5, the older the assigned age, the lower the fitted rate constant, and he writes that this is a pattern "which I have not found pointed out anywhere in the literature."  The pattern is genuinely there in that table, and it is what an over-fitted calibration looks like.  But it is not an unanswered anomaly.  Bada 1985 gives a mechanism for a correlation in the opposite direction, that worse-preserved bone yields higher rate constants, and Collins 1999 gives another, that leaching removes the fast-racemizing free amino acids and leaves the slow bound ones.  It is a fair observation with a mainstream answer.

And his critical treatment of the Californian episode, working from Bada's own tables, is accurate.  One citation defect to route around: his reference 15 gives Bada 1985 as Annual Review of Earth and Planetary Sciences volume 1.  It is volume 13.  Cite Bada directly.

Truman is not the strongest published version of this argument, and stopping at him would be dishonest.  Two earlier creationist papers make better cases.  Both were missed when this article was first written, and both are set out and answered here, because attacking the weakest form of an argument while a stronger one sits in an open archive is exactly the practice this site objects to when others do it.

Helmick 1976: a slope, not an age

Larry Helmick was Professor of Chemistry at Cedarville College and had published on base-catalysed hydrogen-deuterium exchange.  His 1976 paper in Creation Research Society Quarterly runs nine printed pages and 115 endnotes, almost all of them to Geochimica et Cosmochimica Acta, Earth and Planetary Science Letters, Science and Nature.¹⁶

He does not claim that racemization gives young ages.  His argument is about the shape of a curve.  Plot D/L against depth in a deep-sea sediment core.  He says the two histories predict different shapes, and prints them side by side as his Figure 4:¹⁶

"These effects of a worldwide flood would result in plots of racemization versus depth in marine sediments which are not linear.  Instead, it might be expected that the extent of racemization would increase with depth in present ocean sediments only for the period of time since the flood.  Sediments laid down rapidly during the flood, which includes most of the geologic column, might show erratic changes from one layer to the next, but little increase in the average degree of racemization with depth."

He then says the feature is already in the data:¹⁶

"This change in slope has actually been observed for one long sediment core (V23-110), and may be present in data from other cores."

This is a genuine prediction, and the answer given to Truman above does not touch it.  Truman needs the rate constant to be transportable so that he can re-date things.  Helmick needs nothing of the kind — a kink in a curve survives any amount of uncertainty about k.  So it has to be answered on its own terms.  There are three answers.

First, the curve he says his opponents predict is not the curve the chemistry predicts.  On the previous page Helmick writes that on the conventional view "nearly linear plots of racemization versus age would be expected (Figure 4,A)."¹⁶  They would not.  The rate law printed at the top of this article is a reversible first-order reaction, and its solution is D/L(t) = tanh(kt + c).  That curve climbs steeply and then flattens as it approaches equilibrium.  It has to.  A plot of D/L against depth that rises near the top of a core and goes flat lower down is what the ordinary kinetics produces by itself, with no Flood in it.  The straight lines in the literature are straight because they are drawn against the logarithmic transform ln[(1 + D/L)/(1 − D/L)], which unbends the tanh by construction.  Helmick's stated expectation for the conventional model is a near-linear rise in the extent of racemization, and the rate law does not give one.  That is the load-bearing error in the paper.

Second, the one core it rests on is a core he himself reports as irreproducible.  Five pages earlier, in his own words:¹⁶

"Furthermore, Wehmiller and Hare have reported that racemization in the free amino acid fraction of core V23-110 decreases with depth, rather than increases.  Yet Kvenvolden et al. reported increasing racemization with depth for the same core."

He presents that as a flat contradiction between two mainstream groups on the same core.  Whatever its resolution, a slope discontinuity read off that core is not a measurement, and his four other cores are offered only as "may be present."

Third, both models predict flattening, so flattening cannot separate them.  Helmick's model says D/L goes flat below the Flood boundary because those sediments were dumped fast and stirred.  The conventional model says D/L goes flat below some depth because the samples have reached equilibrium.  The conventional model is the one that can be pinned down:  it says what the flat value should be, namely equilibrium, and it says roughly where the flattening should start, from the sedimentation rate and the bottom-water temperature.  Helmick's boundary depth is a free parameter he does not compute.  "There is a kink somewhere" is not a strong prediction.

Where this leaves him.  Helmick's own conclusion is that meaningful sediment ages cannot be got out of a total-amino-acid D/L and a first-order rate constant.¹⁶  On that he agrees with Bada, with Collins, and with this article.

Brown 1985: the rate constant that falls 2,000-fold

The best young-earth paper on this subject is R. H. Brown's, published in Origins, the journal of the Geoscience Research Institute, in 1985 — the same year as Bada's post-mortem.¹⁷  It is better than Truman's, better argued than Gish's, and it reaches one of this article's own conclusions forty years early.  A site that grades both sides by the same standard has to say so.

His first argument is the protein-survival argument, in 1985:¹⁷

"Amino acids have been reported from fossils distributed throughout the geologic column (Florkin 1969).  Since detectable levels of many amino acids are expected to survive only a few million years, at best, these observations are an enigma (Abelson 1956, 1957)."

He names the fossils instead of gesturing at them:  residual amino acids in Silurian graptolites at a putative 400-430 million years, and protein- and peptide-bound amino acids in Jurassic shells at 135-180 million.  That is §3(a) of this article, built from the same materials, two decades before Schweitzer.

His second argument is the quantitative form of the claim this section has been rejecting, and it is new to this article.  Brown collected published racemization rate constants out of nineteen sources and plotted them against the assigned age of the fossil each came from.  They are not constant.  They fall:¹⁷

"If each amino acid could be described by a characteristic racemization rate constant as a component of fossil protein, the data points in figures (3) and (4) would cluster about a horizontal line.  The demonstrated clustering about a line which slopes downward indicates that the apparent racemization rate constant is actually not a constant, but is related to fossil age, diminishing as age increases."

The fall is 700-fold in his Figure 3 and about 2,000-fold in his Figure 4, across an age axis running from 100 years to 10 million.¹⁷  He then inverts it:¹⁷

"…the progressive decline of racemization constants in figures (3) and (4) might be taken as an indication that the fossil age assignments are progressively excessive.  … Many difficulties would be resolved, and much anomalous data would be reconciled if fossil ages were adjusted to make the data points in figures (3) and (4) scatter about a horizontal line, rather than a line sloping downwards.  While the data discussed in this paper provide an adequate scientific basis for such an adjustment, the extent to which it 'would undermine other conclusions' is a price very few members of the scientific community would be willing to pay."

His Table 2 prints what that adjustment would do:¹⁷

* Brown's own footnote to that cell: "Insufficient data for comparison."

And then he refuses to bank it:¹⁷

"The rough estimates given in Table 2 are presented here for their possible interest, without any claim as to their significance.  A more sophisticated data analysis would be required before such claims should be made."

Credit where it is due, and there is a good deal of it.  Brown tested the obvious mainstream explanation himself and killed it.  If older fossils showed lower rate constants because they had sat at lower temperatures, how much lower would those temperatures have to be?  He does the arithmetic.  A 160-fold rate reduction between assigned ages of 3,000 and 2,000,000 years needs an effective average storage temperature of −19 °C against a modern mean of 7 °C, and his specimens come from "arctic, desert, temperate, and ocean floor environments."¹⁷  That is the right test and he ran it against his own case.

Here is the answer, and it does not require doubting a single one of his numbers.

The apparent rate constant is computed from a quantity that has a ceiling, so it has to decline with assumed age whether or not anything is wrong.  A racemization "rate constant" for one fossil is got by taking its D/L, running it back through the rate law, and dividing by the assumed age.  D/L cannot go past equilibrium.  Once a sample has reached equilibrium — which, by the tables computed near the top of this article, happens within a few hundred thousand years at any temperature above freezing — the numerator stops growing and the fitted constant falls in exact inverse proportion to the age assumed for the sample.  Nothing in the chemistry changed.  The clock ran out.

Now put Brown's own numbers into that.  Between assigned ages of 3,000 and 2,000,000 years the assumed age rises 667-fold, so a saturated sample's fitted constant must fall 667-fold.  Brown measures a 160-fold fall over exactly that interval.  Across the whole span of his Figure 4, 100 years to 10 million, saturation alone would produce a hundred-thousand-fold decline;  he reports about 2,000.  The trend he found is not steeper than the artefact.  It is a great deal shallower.

That is an upper bound rather than a fitted model, and it should be said so:  samples young enough to sit far from equilibrium contribute no decline at all, so the artefact does not account for the shape of his curve point by point.  But an upper bound is what the argument needs.  Brown's inference requires the observed decline to be larger than anything the ordinary chemistry can produce.  It is smaller.  There is nothing left over for a revised chronology to explain.

A second and smaller point:  his figures pool bone, shell, coral and wood, and free against total amino acid fractions, from arctic to ocean floor.  Collins 1999 later showed that the "constant" is specific to the material and even to the protein's conformation, so a plot mixing all of those could not have shown a single value in any case.

Brown half-saw this himself.  He raised, and did not pursue, a model in which "as a fossil ages the breakdown of proteins continually renews the supply of L-molecules in locations with the higher values of k," which would hold the true average constant while the fitted one fell.¹⁷  He wrote that the community's course of preference was "to assign the apparent inverse relationship between effective average racemization rate constant and fossil age to some time-dependent factor that is not yet fully understood," and he treated that as an evasion.  Fourteen years later Collins, Waite and van Duin named the factor.  In bone, Asx racemization is gated by how far the collagen triple helix has unwound, so the measured D/L "reflects the ratio of non-helical (high proportion of D-Asx) to triple helical (L-Asx) collagen," and the "rate constant" is not a rate constant at all.⁷  The unknown factor Brown was asked to take on faith turned out to exist and to be measurable.

And notice what his own inversion delivers, on his own table.  Five thousand to eighteen thousand years.  Not 4,500.  Brown's adjustment does not produce a Flood chronology, he never says it does, and he explicitly declines to claim anything for it.  Anyone quoting his Table 2 for a young earth is quoting past the sentence printed under it.

How thin is the creationist engagement, exactly?

"Nobody has answered this" is a claim about the world, and it is easy to get wrong by searching badly.  An earlier draft of this article got it wrong about Penkman.  So here is what was counted, and where the counting stops.¹⁸

• Creation Ministries International. All 296 issue-contents pages of Journal of Creation volumes 1-40 and Creation volumes 2-47 were harvested, giving 6,741 items. Truman's five-part racemization series is filed under Paper, and no item in any Letters, Perspectives, Viewpoint or Forum column carries racemization or amino acids in its title. The series has never been disputed in print by its own venue.
• The International Conference on Creationism. The full text of all 457 papers from 1986 to 2023 was searched for racemi* and alloisoleucine. Five papers match. Two are origin-of-life chirality, one is a passing mention, and the remaining two use racemization in support of some other argument. No ICC paper has ever been devoted to amino acid racemization dating. Controls over the same corpus returned 37 papers for radiocarbon and 27 for zircon.
• Answers in Genesis. A site search for amino acid racemization returns six results. One is Mitchell 2011, answered above; one is Peter Line's 2005 piece on the Protsch fraud; the rest are protein-survival articles.
• The Creation Research Society. CRSQ returns 28 full-text hits for racemization, and almost all of them are origin-of-life homochirality. Helmick 1976 is the only dating paper among them, and the journal's own fiftieth-anniversary retrospective files it under chemical evolution rather than under dating.

The boundary of that search, stated plainly.  Only the ICC sweep read every word;  the other three ran over titles and indexes.  A paragraph inside an article whose title says nothing about racemization would not be found by them.  So this is a strong result about what the movement has published on the subject and a weak one about what it has said in passing.

What it adds up to.  Two substantial papers in fifty years, Helmick 1976 and Brown 1985;  one modern series, Truman's;  one news item on the 2011 rebuild;  and Gish 1975, which predates most of the evidence.  That is a thin record for a method the movement cites often.  It is also a record whose two best papers were both missed when this article was first written, which is a fair measure of how little they circulate.

A note on amber and tar

There is a category of genuine low-D/L anomaly, and it is worth naming because it is the place a young-earth writer will most want to look and the worst place to look.

Bada reports that insects entombed in amber "contain well-preserved unracemized amino acids, apparently because of the anhydrous nature of the amber matrix."⁶  Water is required both for hydrolysis and for the carbanion step by which racemization proceeds.  Take the water away and the clock stops.  Bada makes the same argument for sub-freezing burial, where "racemization rates could be considerably slower because the system may be analogous to anhydrous conditions."⁶

Amber and asphalt are therefore the two environments in which the mainstream model itself predicts low D/L at old assigned ages.  Finding low D/L there is a confirmation of the model, not a refutation of it.  A 1974 report of a La Brea tar-pit sabre-tooth bone with racemization barely above modern bone circulates in the creationist literature and would be the best measured datum for the young-age argument if it held up.  It was not obtained from a primary source for this article and is recorded in quarantine.md rather than printed.  Even if it is exactly as reported, asphalt is the one place it would be expected.

Common objections

"You are using one method's known failure to attack methods that do not share its assumptions."

Fair, and this article does not do that.  Racemization dating failed in California for a reason specific to calibrated chemical clocks: the rate constant is imported from a sample of assumed known age, and the import carries every error in that assumption.  Radiometric methods do not work that way.  A U-Pb concordia age is not calibrated against a bone somebody dated by another means, and an isochron tests its own assumptions internally.  The transferable lesson is narrower and it is real: wherever a method's answer is set by a calibration sample, the calibration sample carries the whole result, and an error there is invisible to any amount of internal consistency in the measurements themselves.  That is worth knowing.  It is not a general argument against geochronology, and using it as one would be dishonest.

"Penkman rebuilt the method.  The 1974 failure is ancient history."

Half right, and the half that is right is conceded above without reservation.  The intracrystalline method is a different measurement on a different material with a demonstrated closed system, and its 74-site ranking agrees with four independent orderings.  But three things follow that the objection glosses over.  The rebuild does not work on bone, so no old bone date is rehabilitated.  It produces a ranking, not years, and its authors say so.  And the history is not irrelevant, because the failure mode was not exotic.  It was a calibration sample everyone trusted, and it stood for a decade, and it took a new instrument rather than a new argument to catch it.

"Protein survival to 16 million years is already conceded, so extending it to 70 million is a difference of degree, not of kind."

This is the strongest objection in the topic and it deserves a straight answer.  Three things separate the two cases, and none of them is rhetorical.  First, the object.  Demarchi recovered a short peptide bound to a calcite surface inside a closed crystal.  The dinosaur claims are for bulk collagen fibrils and cellular structures inside porous, groundwater-flushed bone.  Second, the mechanism.  Mineral binding was demonstrated with a computed binding energy and a confirmed prediction about which domain survives.  No such mechanism has been demonstrated for collagen in bone.  Third, the size of the gap.  Sixteen million to seventy million is a factor of four in thermal age, which sounds small, but the decomposition is exponential, so a factor of four in time is an enormous factor in surviving material.  The gap is real and unexplained.  It is also not an age measurement, and this article does not present it as one.

"Racemization kinetics measured at 100 °C cannot be extrapolated down to burial temperatures."

This is a fair objection and it cuts against both sides, which is why it is answered rather than deflected.  Collins, Waite and van Duin say exactly this in print, as their first numbered conclusion: "estimating rates of Asx racemization by extrapolation from high temperatures is to be avoided."⁷  Their reason is mechanistic, that the succinimide route is controlled by protein conformation at low temperature in a way that high-temperature experiments do not reproduce.

So the extrapolation cannot bear a two-significant-figure age for a particular bone.  Three points survive it.  Bada's own Arrhenius fit is not purely a high-temperature extrapolation: it includes fossil-bone points and a living human tooth point at 37 °C, which is why it appears in this article at all.  The conclusion about a 70-million-year-old fossil does not need precision, because the margin at 10 °C is a factor of about 160 over what equilibrium requires, and a rate constant would have to be wrong by two orders of magnitude to save it.  And the objection is far more damaging to the young-age argument than to the survival argument, because Truman's 150-year figure imports a rate constant measured at 49.5 °C and needs it to be accurate, whereas the survival argument only needs an order of magnitude.

"Bada blamed the calibration, not the chemistry.  So the method was never really broken."

That is his position and it is quoted in full above.  Two things qualify it.  Bada's own restriction, that a calibration constant should only be applied to bones of similar amino acid content, is not a small caveat.  It says the method cannot be transported between sites without a matched calibration sample at each, which removes most of what made it attractive.  And Collins in 1999 went further than Bada did, concluding that bones and teeth "are probably the worst materials to choose."⁷  The field's own answer was to stop using bone, which is a stronger verdict than Bada's.

"Creationists only cite the 1974 failure and never the 2011 success."

Largely true, with one exception now on the record:  Elizabeth Mitchell reviewed the 2011 Nature paper for Answers in Genesis within days of its publication, and her review is quoted and answered above.  Otherwise this article is the correction.  Gish 1975 predates the rebuild and cannot be faulted for it.  Truman 2023 does not.  The intracrystalline literature is not obscure, it is in Nature, and any treatment of racemization dating that stops in 1985 is out of date by four decades.

Where the young-earth case is strongest

The survival ceilings, stated by mainstream chemists from their own measurements, and the gap between them and reported dinosaur biomolecules.

There is no need to dispute the chemistry, quote a creationist, or assume anything about the past.  Bada's published rate law, run forward, says a fossil buried for 70 million years anywhere above freezing should be racemic to the measurement limit.  Bada's published survival figures say bulk collagen is gone in tens of thousands of years in bone.  Demarchi's paper says the demonstrated survival mechanism reaches about 16 million years at 10 °C for a short bound peptide inside calcite, and states in its own introduction that reported dinosaur soft tissue "seems to be inconsistent with the observed levels of hydrolysis, dehydration and racemization" measured in closed-system biominerals.

Two orders of magnitude, an unexplained mechanism, and a mainstream concession in print.  That is a real problem for the standard picture, and pressing it requires no special pleading.

And the transferable lesson from California, which is genuinely important.  A dating method carried by a calibration sample is only as good as that sample, an error there is systematic rather than random, it will not show up as scatter, and internal consistency will not catch it.  This one stood for a decade and was caught by a new machine.  That is worth remembering the next time a dated chronology is said to be confirmed by its own internal agreement.  The Protsch episode is a second instance of the same lesson with the calibration falsified rather than merely mistaken, and it adds a corollary:  two methods agreeing closely is not confirmation when both numbers came from the same analyst.

Where the young-earth case fails

The attempt to recalibrate racemization to Flood-era ages fails, and this site says so without hedging.

It fails three ways.  It uses a free temperature parameter to produce a predetermined answer, which is the exact defect that made Bada wrong in 1974.  It is argued in the same paper that says racemization rates in fossil material are unpredictable, which is self-refuting.  And its factual premise about dinosaurs is wrong: the amino acids in dinosaur bone were measured by the field's leading practitioner and found highly racemized, which is what deep time predicts.

Its two better predecessors fail too, for different reasons, and they are the ones worth refuting.  Larry Helmick's 1976 slope discontinuity is a feature the reversible rate law produces by itself, with no Flood in it, and it rests on one sediment core that he himself reports two laboratories reading in opposite directions.  R. H. Brown's 1985 finding that the apparent rate constant falls about 2,000-fold across the column is real, but a constant fitted from a saturating quantity has to fall in inverse proportion to the assumed age, and that artefact alone is larger than the decline he measured.  Both are quoted at length and answered above.

The Penkman result still has no young-earth answer, though it does now have a reply.  Elizabeth Mitchell answered it for Answers in Genesis on 13 August 2011, arguing that the framework is circularly calibrated against the Milankovitch chronology and that the D/L and δ¹⁸O records agree only because both respond to temperature.¹⁴  The first half is right, and this article concedes it above in the same words.  The second half fails, because racemization is irreversible while δ¹⁸O oscillates, and because three of the four orderings the paper is checked against are not temperature records at all.  Four independent orderings across 74 sites agreeing with each other is not something a single-year Flood plus a few thousand years of post-Flood deposition predicts, and no one has yet explained why they agree.

Saying all of that is not a concession.  It is the reason a reader should take the survival argument seriously.  An encyclopedia that discards its own side's weak argument earns the right to press its strong one.  An encyclopedia that keeps both has neither.

Every source, with what it measured and which way it cuts, is catalogued at the literature.  Two claims that could not be verified from a primary source are recorded in quarantine.md rather than printed here, along with a note on the third, which was released from quarantine on 30 August 2026 once a contemporaneous source for it was obtained.

Sources

Access routes are recorded where they matter, because several of the primary papers in this topic are paywalled and the free replacements are not obvious.  Four sources central to the Californian episode were not obtained directly and nothing in this article rests on them: Bada, Schroeder and Carter 1974 in Science; Taylor et al. 1985 and Bada 1985 in American Antiquity; and Bada et al. 1984 in Nature.  Every claim about what those papers said is carried instead by two free sources that reproduce their numbers, Bada's 1985 Annual Review paper and Taylor's 2009 Radiocarbon review, and the footnotes below say which.  No quotation in this article was established with an AI page-summarising tool; every one was read out of downloaded bytes.