Luminescence Dating: OSL, TL and the Young-Earth Objections

"If the dating methods say 2 million years, it may actually be a hundred years or just ten years." — Andrew Snelling, Answers in Genesis, 2018¹

"This sample gave an age of several thousand years when dated by conventional total bleach methods." — John Prescott and Rudolfs Purvinskis in Ancient TL, on an Australian dune whose time since last exposure to light "can scarcely exceed a decade"²

"They also show that correct burial ages can be obtained using TL methods, provided the sediments are bleached sufficiently before burial to reset the dating signal." — Richard Roberts and colleagues, 1998, on the site that worked, published the same year as the site that did not³

Summary

Luminescence dating puts a number on when a grain of sand last saw daylight.  Buried quartz sits in a weak radiation field from the uranium, thorium and potassium in the surrounding sediment.  That radiation knocks electrons loose and some get stuck in flaws in the crystal.  They pile up.  Sunlight empties the flaws in minutes, so the count resets every time the grain is exposed.  Dig it up in the dark, stimulate it with light or heat in a laboratory, and the photons it gives off measure how much radiation it soaked up since burial.  Divide that by how fast the sediment delivers radiation and the answer is a burial age.

Five things follow, and they do not all favour the same side.

1. The method is a dosimeter, not a decay clock, and this changes which objections are even available. Age equals accumulated dose divided by dose rate, and both numbers are measured on the sample itself. The dose comes from stimulating the actual grains. The dose rate comes from assaying the actual sediment. There is no parent-daughter ratio anywhere in it. So two of the three assumptions Answers in Genesis names against radiometric dating, the unknown starting inventory and contamination by daughter atoms, have nothing to attack here. That is not a weak objection. It is an objection pointed at the wrong mechanism. It is also that organisation's framework and not the movement's: four of the five young-earth authors who have written on this method never use the parent-daughter checklist at all.
2. Sunlight zeroes the clock at deposition, and that single fact puts accelerated decay in the Flood year out of reach. A dune laid down after the Flood was bleached by daylight after the Flood. Whatever charge those grains carried from an earlier accelerated-decay episode was wiped out at burial. So the dose the laboratory measures accumulated entirely within the last 4,350 years on a young-earth chronology. To turn a 32.5 ka reading into 4,350 years you need the decay constants of ²³⁸U, ²³²Th and ⁴⁰K to have averaged 7.5 times today's value after the Flood, in that particular dune, and then to have returned to normal leaving no trace. RATE's episodes all end with the Flood year,⁴ so on RATE's model nothing can do that. One young-earth author does put acceleration after the Flood: Michael Oard runs it through the Ice Age and says it "telescopes other Quaternary dating methods to the time after the Flood."⁵ Against him the timing argument does not apply, and the question becomes what his acceleration would cost, which he does not work out. None of the five young-earth sources on luminescence itself makes the claim, and the five are now enumerated rather than searched for.
3. Three obvious young-earth objections were computed for this article and they do not work. Accelerating the decay would not measurably deplete the uranium (5 parts per million of the ²³⁸U inventory, at any compression factor). It would not cook the sediment (239 J/kg, a temperature rise of 0.3 K). And it would raise the local radiation dose to only about five times natural background, which is survivable. The arithmetic is shown below and the script is in this article's research/ folder. This site does not print an argument its own arithmetic refutes.
4. The method's documented failures are real, large and worth knowing, and every single one runs in the same direction. Jinmium was published at 116-176 thousand years and corrected to under 10 thousand, a factor of at least twelve. A standard thermoluminescence protocol once returned "several thousand years" for an Australian dune less than ten years old. The people who built the modern protocol showed in 2016 that their own standard single-grain procedure, applied blind, would have underestimated four radiocarbon-controlled ages by more than 40%. All of these are over-estimates or procedure failures caught by the field itself, published in the field's own journals, usually within two years.
5. Verdict: on luminescence dating specifically, the young-earth position is weak, and it is weak for a reason that is easy to state and hard to escape. The errors that exist are factors of 1.5 to 12, not factors of a million. They all come from grains that never saw the sun, and unbleached grains make ages too old, which is the wrong direction for a young-earth argument that needs 32,000 years to be 4,000. The one mechanism that could compress ages by three orders of magnitude cannot reach a deposit that was bleached after the Flood unless it kept running after the Flood, and the one author who says it did gives it no size and concedes its radiation cost. The strongest things a young-earth writer can truthfully say here are that the method has produced published errors of an order of magnitude, that agreement between related methods has already fooled a good team once, and that published luminescence ages almost never carry the full systematic uncertainty budget a calibration laboratory would demand. All three are worth saying. None is evidence for a 6,000-year earth.

One correction to an earlier version of this page belongs in the summary, because it was a claim about the state of the evidence.  That version said the young-earth engagement with this method was "close to nil," that the fullest treatment was a magazine article, and that creation.com "is genuinely unaudited here."  All three were artefacts of broken search boxes.  ICR's search returns nothing for thermoluminescence while ICR hosts an article with that word in its title.  creation.com's archive is reachable through an open, undocumented search endpoint.  Five young-earth sources on this method are now enumerated and each is answered above, including a nine-page technical paper by a professor of medical physics whose working field is thermoluminescent dosimetry.  A search box with unmeasured recall can widen a search and can never close one, and every negative finding on this page has been rebuilt on enumeration and controls.

Every source this article rests on is catalogued, with what it measured and how it was obtained, in a companion page: the literature.

How the clock works

A quartz or feldspar grain sitting in sediment is being irradiated all the time, faintly, from every direction.  The uranium and thorium chains and the ⁴⁰K in the surrounding grains emit alpha particles, beta particles and gamma rays.  Cosmic rays add a little more.  That radiation ionises the crystal.  Most of the freed electrons fall straight back, but a few get caught in defects in the lattice and stay there, sometimes for hundreds of thousands of years.

Two things empty those traps.  Sunlight does it in minutes to days.  Heat above roughly 300 °C does it too.  Emptying the traps is called bleaching, and it is what sets the clock to zero.  A grain blown across a dune in daylight is bleached.  A pot fired in a kiln is bleached.  A grain that spalls off a cave wall in the dark and drops onto the cave floor is not, and that is the failure mode this whole article keeps coming back to.

The age is a ratio, and this is the equation the rest of the article works with:

age (ka)  =  equivalent dose D_e (Gy)  /  environmental dose rate (Gy/ka)

The numerator is measured on the grains.  In the laboratory the sample is stimulated with light (optically stimulated luminescence, OSL) or with heat (thermoluminescence, TL) and the emitted photons are counted.  Then the same grains are given known laboratory doses and stimulated again, building a calibration curve for those specific grains.  D_e, the equivalent dose, is the laboratory dose that reproduces the natural light level.  It is in grays.

The denominator is measured on the sediment.  The uranium, thorium, potassium and rubidium contents of the sediment the grains came out of are measured, usually by high-resolution gamma spectrometry or by ICP-MS, and converted to a dose rate with published conversion factors.  It is in grays per thousand years.

Nothing is inherited from a global calibration.  In potassium-argon, uranium-lead or rubidium-strontium dating, the rate of the clock is a decay constant measured somewhere else, on other material, and assumed to apply.  In luminescence, the rate of the clock is measured on this sediment, in this field, this year.  There is also no parent-daughter ratio.  The trapped charge is not a daughter product of anything inside the grain; it is a record of energy delivered from outside.  Section Two of the three assumptions have nothing to attack works out what that does to the standard young-earth critique.

Four variants matter here.

• TL, thermoluminescence. Heat-stimulated, in use since the 1960s. Its weakness is that different peaks in the glow curve bleach at very different rates, so a protocol that reads a slowly-bleaching peak reads charge that sunlight never removed. That is exactly what went wrong at Jinmium.
• OSL, usually blue or green light on quartz. Reads only the light-sensitive traps, so it is better zeroed than TL by construction.
• Single-grain OSL, the same measurement on one grain at a time. A population of badly bleached grains then shows up as a separate population instead of being averaged into the answer. This is the technique that caught Jinmium.
• IRSL and pIRIR on potassium feldspar. Feldspar saturates at a higher dose than quartz and so reaches further back in time, but it leaks charge and needs a measured correction.

ESR, electron spin resonance, reads the same trapped charge magnetically rather than optically and is applied to tooth enamel and quartz.  Snelling groups it with luminescence, and for the purposes of this article that grouping is fair.

The arithmetic, worked in the open

Everything below can be redone from scratch.  The conversion factors come from Guérin, Mercier and Adamiec's 2011 update in Ancient TL, which recomputed the energy released per disintegration in the ²³²Th, ²³⁸U and ²³⁵U chains and by ⁴⁰K and ⁸⁷Rb from the Brookhaven Evaluated Nuclear Structure Data Files.⁶  These are the numbers every luminescence age in the world is divided by.  The paper is open access under CC BY.

Units are grays per thousand years, for an infinite matrix in secular equilibrium.⁶  Guérin's own summary of how much the update moved things: "The overall effect of our update obviously depends on each case, but may reach a few percent on the final obtained age calculation."⁶

A Deno script, research/dose-rate.ts in this article's folder, takes those factors and computes dose rates for three realistic sediments, converts a measured dose to an age, and then asks what a young-earth chronology would require.  Run it with deno run html/unpublished-ai/luminescence-dating/research/dose-rate.ts.

Dose rates for three sediments

Dry infinite-matrix values first, then the dose an actual etched 90-125 µm quartz grain receives.  The alpha term drops out because the hydrofluoric-acid etch removes the outer rind of the grain, which is the only part alphas reach.  The beta term is attenuated for grain size, and beta and gamma are attenuated by pore water.  A cosmic-ray term is added for 2 m of burial.⁷

The check that this is behaving.  Roberts and colleagues measured 0.77 ± 0.09 and 0.91 ± 0.08 Gy/ka at Malakunanja II in northern Australia.³  Helena Alexanderson measured 1.90 ± 0.07 Gy/ka for a Greenland river sand.⁸  The computed range, 0.79 to 2.72, brackets both published values.

From dose to age

Take the middle sediment, 1.539 Gy/ka:

The bottom row is roughly the practical ceiling.  Murray and seven co-authors, in the 2021 Nature Reviews Methods Primers survey of the method, put it this way:⁹ laboratory dose response curves "often increase to large doses (hundreds of Gy), but natural doses either do not increase above ~200 Gy, or if they do, they often appear to be an underestimate."⁹

The 50 Gy row, 32.5 thousand years, is the worked example for the rest of this article.  It is an ordinary result for an ordinary dune.

Water content, the largest ordinary systematic

Water in the pore space absorbs radiation that would otherwise have reached the grain.  More water means a lower dose rate and therefore an older age.  Nobody can measure what the water content was over the last thirty thousand years, so it is estimated.  Here is what that estimate is worth, same sediment, same 50 Gy, only the assumed lifetime water content changed:

The computed slope is 0.96% of age per percentage point of water across 0-25%, which reproduces the field's own rule of thumb exactly.  Murray and colleagues state it as "a 1% increase/decrease in lifetime average water content typically leads to a 1% increase/decrease in derived age."⁹  This is a genuine weakness and it should be conceded plainly.  It is also bounded.  The physically extreme spread from bone dry to fully saturated is about ±10%, not a factor of a thousand.

The young-earth argument in its strongest published form

There is more of it than the first version of this page reported, and the correction is the first order of business.  That version said the fullest young-earth treatment "found anywhere in this research is a magazine article," and that ICR and creation.com had nothing.  That was an artefact of broken search boxes, and section How thin is the creationist engagement, exactly? sets out exactly how it happened.  Five sources carry the whole young-earth case against luminescence dating.  They are set out here oldest first.

Donald DeYoung 1994, the one worked case

Donald DeYoung, a physicist at Grace College, reviewed the age of Barringer Crater in Arizona for the Creation Research Society Quarterly.¹⁰  One of the four dating routes he examines is thermoluminescence, and his description of the physics is correct, including the reset:

"At the time of collision, the rocks were shock-heated, which 'reset the clock' by allowing large-scale electron migration to their original sites.  Upon cooling, the internal radiation once again began the process of electron removal."¹⁰

He reports Steven Sutton's measured crater age of 49,900 ± 2,900 years and rejects it:

"However, this result is very tentative for two reasons.  First, calibration of the technique is uncertain.  Its success, even with pottery, has been limited.  Second, the underlying physical mechanism of thermoluminescence is poorly understood.  The method is not a simple, straight forward dating technique."¹⁰

One of those two objections was answered six years later and one is still partly live.  Calibration is what the single-aliquot regenerative-dose protocol exists to do: after the natural signal is read, the same grains are given known laboratory doses and read again, so the calibration is built from the sample itself rather than imported, and a fixed test dose after each measurement corrects for the sensitivity change that reading the sample causes.⁹  The mechanism objection is fairer.  Murray and seven co-authors still write in 2021 that "the controls on luminescence saturation with dose are poorly understood."⁹  A method can work without its microphysics being fully understood — that is true of most of thermometry — but DeYoung was not wrong to say the microphysics was incomplete.

His third move should not be made by anyone.  Faced with a century of published crater ages converging on the same figure, he writes:

"Many recently-published crater ages converge around 50,000 years, perhaps an example of 'tracking' (DeYoung, 1976).  This is the tendency of published data to cluster around a particular value, similar to peer pressure, whether the value is correct or not."¹⁰

That explanation cannot be tested and it cannot be limited.  Any agreement between independent measurements can be relabelled peer pressure, including the agreements the young-earth side relies on — the radiocarbon dates on Dead Sea Scroll parchment, the tree-ring counts, the concordant helium-diffusion measurements in RATE.  A rule that dissolves the other side's convergences dissolves your own.

Tas Walker 2003, the canonical statement of the assumptions

The single most-read young-earth statement about these methods is a sidebar.  Tas Walker, a mechanical engineer and geologist with Creation Ministries International, published "The dating game" in Creation 26(1) in December 2003, on the Lake Mungo redating.¹¹  Attached to it is a box called "Dating methods" that treats radiocarbon, TL, OSL, ESR, Th/U and Pa/U in turn.  It is the closest thing the movement has to a standard account, and it names four assumptions:

"Thermoluminescence (TL) dates are obtained from individual grains of common minerals such as quartz.  When such grains are heated, they emit light, and this is related to the radiation 'stored' in the crystal structure.  It is assumed that the radiation was slowly absorbed from the environment, building up from zero at a certain time in the past (perhaps when the grain was last exposed to sunlight). ...  Unfortunately, there are many unknowns and many assumptions need to be made, including the amount of radiation 'stored' in the mineral at a certain time in the past, that the change in radiation has only been affected by the radiation in the environment, that the radiation in the environment has remained constant, and that the sensitivity of the crystal to radiation has not changed.  All these factors can be affected by water, heat, sunlight, the accumulation or leaching of minerals in the environment, and many other causes."¹¹

He then folds the other two methods into the same account:

"Optically-stimulated luminescence (OSL) dates are based on exactly the same principle as TL.  But instead of heating the grain, it is exposed to light to make it emit its 'stored' radiation.  The calculated date is based on the same assumptions, and affected by the same uncertainties, as for TL."¹¹

"Electron-spin resonance (ESR) dates are based on the same principles as TL and OSL.  However, the 'stored' radiation in the sample is measured by exposing it to gamma radiation and measuring the radiation emitted.  The measuring technique does not destroy the 'stored' radiation (as does TL and OSL), so the measurement can be repeated on the same sample."¹¹

Walker's account is accurate, and grouping ESR with luminescence is his framing before it is this article's.  Take his four assumptions one at a time.

• "The amount of radiation stored in the mineral at a certain time in the past." This is the residual-dose problem and it is real. It is also the one assumption that is measured rather than assumed, by running the protocol on modern sediment and by measuring grains one at a time. Sections Zero-age tests and Jinmium, and the correction that was itself twice too old give both a pass and a failure of that test.
• "That the change in radiation has only been affected by the radiation in the environment." Also real. It is violated by anomalous fading in feldspar, by thermal transfer, and by light leaking into a sample. Fading is measured as a g-value and corrected, imperfectly, and this article says so.
• "That the radiation in the environment has remained constant." This is the live objection, and it is the same one Cupps, Snelling and DeWerd all reach. The whole of section What compressing a luminescence age would actually require is the answer.
• "That the sensitivity of the crystal to radiation has not changed." This is Walker's best point and it deserves credit. Quartz really does change its luminescence sensitivity when it is heated and dosed, sometimes by a large factor, and an uncorrected sensitivity change corrupts the calibration. It is also the assumption that had already been retired when he wrote. The single-aliquot regenerative-dose protocol, published by Murray and Wintle in 2000, follows every measurement with a fixed test dose and divides by its response, so the calibration curve is built from sensitivity-corrected ratios rather than raw signals. Murray and colleagues put the reason plainly: "sensitivity changes from measurement to measurement have been observed for any regeneration of the signal in the laboratory, rendering interpolation inaccurate," and SAR "successfully overcomes this problem by using the luminescence response to a fixed test dose to correct for sensitivity changes."⁹ And the correction is large, which is the part that makes Walker's point a good one. In their own worked example the sensitivity-corrected dose is 3.8 Gy against 1.8 Gy uncorrected, and for a sedimentary quartz aliquot 25 Gy against 14 Gy.⁹ Sensitivity change is not assumed away. It is measured on every aliquot, it is nearly a factor of two, and it is divided out.

Vernon Cupps 2016, the shortest and the cleanest

The Institute for Creation Research published a whole article on the method: Vernon Cupps, "Examining Thermoluminescence Dating," Acts & Facts 45(5):12, May 2016.¹²  Cupps holds a PhD in nuclear physics from Indiana University-Bloomington.  He states the method as a ratio and puts his objection on the denominator:

"Scientists measure the amount of light to get the total measured radiation dose (TMRD).  They divide this by an assumed radiation dose rate (RDR) to estimate the pottery's age."¹²

"However, problems arise from assuming a uniform radiation dose rate over any significant period of time and assuming that the TMRD resulted from the object or artifact being in a strictly constrained environment identical to that in which it was found.  Both assumptions become less realistic with the passage of time."¹²

Cupps is the fairest of the five, and the reason is what he leaves out.  He does not run the three-assumption parent-and-daughter checklist that section Two of the three assumptions have nothing to attack criticises.  He goes straight to the dose rate, which is the only assumption a dosimeter actually has.  Whatever else is wrong with his article, he has understood what he is criticising.

His specific reason for doubting the dose rate does not survive contact with the numbers:

"The constancy of the RDR is even more problematic because it's based on the uniformitarian assumption that the RDR has been constant.  However, it's well known among radiation physicists that RDRs vary with location, season, solar activity, and even time of day."¹²

His source for that sentence is the American Nuclear Society's public "Radiation Dose Calculator," a consumer tool that estimates a person's annual exposure.¹²  Those variations are real and they are variations in the external radiation field at the earth's surface — mostly cosmic rays and airborne radon.  A buried quartz grain is not exposed to that field.  Its dose comes from the uranium, thorium and potassium in the few tens of centimetres of sediment around it.  The article's own dose-rate table puts numbers on the difference: for the typical aeolian sand, beta and gamma from the sediment supply about 1.39 Gy/ka of the 1.54 Gy/ka total, and the cosmic-ray term is 0.15 Gy/ka, under 10%.  Solar activity modulates the cosmic-ray flux by of order 10-20% over the solar cycle, so it moves the total dose rate by one or two percent.  Time of day moves it by nothing at all: a grain two metres down does not know what time it is.

His second objection is a real laboratory concern aimed at the wrong object:

"Another problem with the TMRD is the calibration of the detector, since different crystals in an artifact can contain different amounts and/or types of luminescence material.  For example, a lithium fluoride crystal can preferentially respond to gamma thermal neutron, beta proton, or alpha particle radiation depending on whether it is constructed from ⁶Li or ⁷Li or a mixture of the two and what trace elements are included in its matrix."¹²

The materials he names are the wrong materials, and this is checkable from his own footnote.  His list of "certain crystalline materials" in pottery is "calcium fluoride (CaF₂), lithium fluoride (LiF₂), calcium sulfate (CaSO₄), lithium borate (Li₂B₄O₇), calcium borate (Ca₃(BO₃)₂), potassium bromide (KBr), and feldspars."¹²  That is the standard list of manufactured thermoluminescent dosimeter phosphors, the pellets a radiation worker wears on a badge.  Only feldspar is on both lists.  Archaeological and sediment luminescence dating measures quartz and potassium feldspar extracted from the sample, and it does not need an absolute detector calibration at all, because the natural signal and the calibrating signal are read from the same grains through the same optics.  The ⁶Li/⁷Li point is a genuine issue for badge dosimetry and has nothing to do with dating a dune.

Andrew Snelling 2018, the most-read version

The most-read young-earth statement on the subject is Andrew Snelling's "How Reliable Are Dates for Human Fossils?", published in Answers on 1 September 2018.¹  Snelling holds a PhD in geology from the University of Sydney and is Answers in Genesis's director of research, and his description of the physics is accurate.  He explains that the surrounding rock irradiates the mineral, that the trapped electrons accumulate, and that the stored energy is read out in the laboratory, comparing the grain to "a rechargeable battery."¹

He then applies the standard Answers in Genesis framework, which is the same three assumptions the organisation applies to every dating method:

"Assumption 1  The original number of unstable atoms can be known.  Scientists assume how many atoms appeared at the beginning based on how many parent and daughter atoms are left today.

Assumption 2  The rate of change was constant.  Scientists assume that radioactive atoms have changed at the same rate throughout time, ignoring the impact of creation or changes during Noah's flood.

Assumption 3  The daughter atoms were all produced by radioactive decay.  Scientists assume that no outside forces, such as flowing groundwater, contaminated the sample."¹

His specific questions about luminescence are these, verbatim:

"Do the minerals show when the fossil bone was buried, or did the minerals first get their stored energy somewhere else before being buried with the bone?  Did the mineral grains, typically quartz or potassium-feldspar, have impurities that could have affected the rate of storage of electrons?  Were these minerals totally reset to zero before being buried with the bone?"¹

And on ESR:

"Has the radiation dose been small enough so that the minerals could cope with it?  (If the radiation dose was too large, then the minerals' 'batteries' would have overfilled, the extra 'charge' would be lost, and thus the dates obtained would be highly erroneous.)"¹

The chronology he is fitting the dates into is stated in the same article: creation 6,000 years ago, "cataclysmic changes brought on by Noah's flood about 4,350 years ago," and radioactive elements that "decayed rapidly during and soon after the flood, so these 'early' dates must be calibrated to the biblical timeframe."¹  His conclusion is the pull quote at the top of this page.

Two of his three luminescence questions are exactly the field's own failure modes, and both have documented mainstream instances.  "Did the minerals first get their stored energy somewhere else" and "were these minerals totally reset to zero" describe Jinmium, and the team that corrected Jinmium says so in print in almost the same words.  The ESR saturation objection describes the ~200 Gy quartz ceiling, and Murray and colleagues not only agree it exists but add that "the controls on luminescence saturation with dose are poorly understood."⁹  Snelling is not making these up.

Larry DeWerd 2020, the fullest treatment that exists

This is the strongest young-earth engagement with luminescence dating anywhere, and it is by a long way the best qualified.  Larry A. DeWerd is Professor of Medical Physics at the University of Wisconsin-Madison and runs a radiation calibration laboratory; thermoluminescent dosimetry is his working field.  His nine-page paper "Dating Uncertainties with Thermoluminescence" appeared in the Creation Research Society Quarterly in Spring 2020.¹³  It opens by naming Cupps 2016 as the prior review,¹³ which is a small confirmation that these five sources are the whole literature and that its authors know it.

His method is the interesting part, and it is a good idea.  Published luminescence ages usually carry a standard deviation.  DeWerd points out that a standard deviation is a Type A uncertainty — the scatter of repeated measurements — and that a calibration laboratory would never report a result that way.  The National Institute of Standards and Technology's guide requires Type B terms as well: the systematic effects you have to estimate by judgement because you cannot measure them.¹³  So he builds a full NIST-style uncertainty budget for a thermoluminescence date.  That is a fair criticism and this article concedes it.  Most published luminescence ages do quote a combined counting-and-fitting uncertainty plus a dose-rate term, and do not formally propagate the judgement terms — the lifetime water content, the assumption of secular equilibrium, the assumption that the sediment's radionuclide content has not changed.  Murray and colleagues' own realistic accuracy floor of about 5% is a statement about the same gap from the other direction.

His budget, Table II, is eleven line items applied to quartz dating:

The arithmetic in that table is right.  The eleven values summed in quadrature give √1825 = 42.72%, and doubling for 95% coverage gives 85.44%.  His conclusion from it is this:

"Note that the conclusion would be, using these estimates, that an age of 20,000 years could really be an age between 2,900 to 10,000 years.  This discrepancy indicates that an honest evaluation of the uncertainty involved in the dating experiments must be done, rather than just claiming a date because it is what the scientist would expect.  All of the dates determined using TL are generally less than 10,000 years.  However, this age is probably an overestimate based on the uncertainties involved."¹³

And in his summary:

"If results are analyzed and all uncertainties taken into account, the ages of the TL dating of artifacts would be expected to be younger than 10,000 years."¹³

Why it does not work, in three steps

One: half of his own interval is missing, and it is the half that goes the wrong way.  Apply his 85.44% to 20,000 years.  The lower bound is 20,000 × (1 − 0.8544) = 2,912 years, which is the 2,900 he prints.  The upper bound is 20,000 × (1 + 0.8544) = 37,088 years, which he does not print.  He printed 10,000 instead.  There is no coverage factor anywhere in his paper that turns 20,000 into 10,000; that would be −50%, and −50% is not in Table II.  This is not a typographical slip, because his summary repeats the 10,000-year figure as a general conclusion.  An uncertainty band that reaches down to 2,900 years reaches up to 37,000 years by the same arithmetic, and 37,000 is older than the date he set out to shrink.

Two: an uncertainty is not a bias, and the paper needs it to be one.  A Type A/Type B budget describes how far a measurement might sit from the true value in either direction.  It cannot pull a population of ages systematically downward.  For "all of the dates determined using TL" to come out under 10,000 years, every one of eleven independent error terms would have to have pushed the same way, near its maximum, in every laboratory, on every sample, for fifty years.  That is not what a coverage factor means.  If DeWerd's terms were real and random, roughly half of all published TL ages would be too young, and luminescence would be returning Holocene ages for deposits that radiocarbon and stratigraphy put in the Pleistocene.  It does not: at Malakunanja II, TL, single-grain OSL and radiocarbon agree at 44-56 thousand years, and the near-surface control reads zero.

Three: most of the table is a badge-dosimeter budget carried across to a method that does not need it.  "Traceability to NIST (particulate radiation)," "Match of PMT to emission spectra," "Calibration Dose," "Reproducibility of reading" and "Energy response from different emissions" — 20%, his largest single term after supralinearity — are the terms you need when a detector must report an absolute dose in grays traceable to a national standard.  That is what DeWerd's own worked example, Table I, actually is: an uncertainty table for a brachytherapy dose-rate constant measured with LiF:Mg,Ti pellets.  Luminescence dating is not an absolute measurement.  It is a substitution measurement: the natural signal and the calibrating laboratory dose are read from the same grains, through the same filters, into the same photomultiplier, on the same night.  The photomultiplier's spectral match, its absolute efficiency and its traceability to NIST all divide out of the ratio.  His 25% supralinearity term is conditional in his own wording — "if supralinearity not accounted for" — and building a dose-response curve is precisely how it is accounted for.  Strip the terms that cancel in a substitution measurement and the surviving items are the environmental ones: 10%, 10%, 5%, 15%, which sum in quadrature to about 21%, or ±42% expanded.  That is a serious number and this article does not dismiss it.  It is also a factor of 1.4, and a young-earth chronology needs a factor of 7.5.

And one claim in the paper is simply false as a matter of published record.  "All of the dates determined using TL are generally less than 10,000 years" is contradicted by Aitken's Thermoluminescence Dating, which DeWerd cites in his own bibliography, by the 61 ± 10 and 45 ± 7 thousand-year TL ages at Malakunanja II, and by the 116-176 thousand-year TL ages published for Jinmium.  The Jinmium numbers were wrong, and they were TL, and they were nowhere near 10,000 years.

What should be kept from it.  DeWerd's demand that luminescence ages carry a full Type B budget is the single best point any young-earth author has made about this method, and the field has not fully met it.  His own summary also contains a sentence his side rarely quotes: "It would appear that the most accurate dating method for artifacts would be via the TL process."¹³  A young-earth medical physicist calling thermoluminescence the most accurate artefact-dating method available is worth more, as testimony, than any of the dismissals above.

Two of the three assumptions have nothing to attack

Here is where the framework misfires, and a skeptical reader will spot it in about four seconds.

Assumption 1, the original number of unstable atoms, does not apply.  Luminescence does not count parent atoms and does not need to know how many there were at the start.  It measures how much energy a crystal absorbed.  The starting condition it does need, that the traps were empty, is not assumed at all: it is tested, and section Zero-age tests gives the tests and one famous case where the test failed.

Assumption 3, contamination by daughter atoms, does not apply either.  There are no daughter atoms in the measurement.  Trapped electrons are not a decay product of anything inside the grain.  Groundwater carrying uranium into or out of the sediment is a real problem, but it is a problem for the dose rate, it is checked by gamma spectrometry, and it is not the problem Snelling's assumption 3 describes.

Assumption 2, a constant rate, is the one that bites, and it is the only one.  The rest of this article is about what it would take for it to bite hard enough.

This matters more than a debating point.  An argument aimed at the wrong mechanism is not a weak argument.  It is a misdirected one, and importing a parent-daughter assumption list wholesale onto a dosimeter costs the young-earth side credibility it cannot afford to spend.  It should be dropped.

And the criticism belongs to Answers in Genesis's framework, not to the movement.  This is a correction to the first version of this page, which had only Snelling and read his three assumptions as though they were the young-earth position.  They are not.  Walker's 2003 sidebar names four assumptions and every one of them is about the dosimeter: the initial stored dose, whether anything but environmental radiation changed the signal, whether the environmental dose rate was constant, and whether the crystal's sensitivity changed.  Cupps goes straight to the dose rate.  DeWerd builds an uncertainty budget.  DeYoung asks about calibration and mechanism.  Four of the five young-earth sources on this method never run the parent-daughter checklist at all, and the fifth is the one whose organisation applies the same three lines to every dating method it discusses.

What compressing a luminescence age would actually require

Age equals dose divided by dose rate.  Squeezing the age therefore means multiplying the dose rate by the same factor.  For the worked example, 50 Gy at 1.539 Gy/ka:

Over a few thousand years the number of parent atoms cannot change appreciably: ²³⁸U has a half-life of 4.468 Gyr, ²³²Th 14.05 Gyr, ⁴⁰K 1.248 Gyr.  Activity is the decay constant times the number of atoms, so the whole multiplier has to fall on the decay constant.  That is the RATE hypothesis, applied here.

Three young-earth objections that were computed and do not bite

These are reported because they fail, not because they are arguments to make.

Parent depletion does not constrain anything.  If the decay constants were 7.5 times higher, the parents were being consumed 7.5 times faster.  How much of the inventory would that use up?

Five parts per million.  And the figure is identical for all three scenarios, which is the interesting part: the total number of decays is pinned by the measured equivalent dose, so it does not matter whether they happened over 4,350 years or over 100.  A young-earth argument that says "the uranium would have run out" is simply wrong at these compressions, and nobody should make it.

Self-heating of the dated sediment does not constrain anything either.  One gray is one joule per kilogram.  The bulk absorbed dose over the compressed interval, using the full dry infinite-matrix rate of 7.367 Gy/ka for that sediment, is 239.3 Gy, which is 239.3 J/kg.  Quartz sand has a heat capacity around 800 J/(kg·K).  That is a temperature rise of 0.3 K.  The heat problem for accelerated decay is real, but it lives in the whole crust and in uranium-rich source rocks over billions of years of implied decay, not in a dune over a few thousand.  It is worked out for the crust in the hydroplate heat article, and it should not be imported here as though it applied to the dated sediment.

The biological dose objection bites, but only weakly.  World-average natural background to a human is about 2.4 mGy per year.¹⁴  The three required sediment dose rates are 11.5, 50 and 500 mGy per year, which is 4.8, 20.8 and 208 times background.  Five times background is a moderately radioactive region, not a death zone.  Only the 100-year scenario reaches a level where chronic-exposure effects would be obvious, and that scenario is not one Snelling proposes.  This is not an argument that can be pressed hard.

So the quantitative objections to a raised dose rate are weaker than they look.  The decisive objection is not arithmetic at all.

Timing decides it for the Flood-year models

Sunlight zeroes the clock at deposition.  An accelerated-decay episode that ended before deposition therefore leaves no trace whatsoever.  Whether that settles the question depends on when a young-earth author puts the acceleration, so the placements come first, each in its author's own words.

1. RATE puts no episode after the Flood. Russell Humphreys' chapter in RATE's first volume proposes "three episodes: one in the early part of the Creation week, another between the Fall and the Flood, and the third during the year of the Genesis Flood," and its Figure 2 shows the same three.⁴
2. Snelling, writing for Answers in Genesis in 2018, says radioactive elements "decayed rapidly during and soon after the flood."¹ That is a popular summary with no rates, and it does not say how long "soon after" lasted.
3. Michael Oard runs it through the Ice Age, and applies it to the Quaternary dating methods as a class. His 2021 Creation Research Society Quarterly paper on cosmogenic nuclides calls a nearby supernova a possible "cause for accelerated radiometric decay during the Flood and Ice Age," and says that "accelerated radiometric decay after the Flood telescopes other Quaternary dating methods to the time after the Flood."⁵ From aluminium-26/beryllium-10 burial ratios he concludes: "Therefore, accelerated radiometric decay continued for a while after the Flood (Table I)."⁵ He does not name luminescence, but it is a Quaternary dating method, so his claim reaches it.

An earlier version of this article called Snelling's phrase RATE's position and said nobody claims acceleration after the Flood.  Both were wrong, as the three placements show.

Now work the timing through for a sand dune emplaced after the Flood, which is where a young-earth chronology has to put almost every OSL-dated aeolian deposit on earth.

1. The grains in a post-Flood dune blew across the surface in daylight while the dune was building. Whatever charge they carried from any earlier episode was bleached out. The traps were empty at burial, and they were empty because of sunlight, not because of any assumption about the past.
2. So the dose the laboratory measures accumulated after the Flood, during the roughly 4,350 years this chronology allows.
3. On RATE's placement, no acceleration was running then. To read 32.5 thousand years, the decay constants of ²³⁸U, ²³²Th and ⁴⁰K in that particular dune must have averaged 7.5 times today's value over those 4,350 years, and RATE supplies no episode to do it. The same holds for any dune bleached after Snelling's "soon after" was over.
4. On Oard's placement, a dune bleached during the Ice Age was dosed while the acceleration was still running, and the timing argument does not reach it.

So the argument is decisive against RATE and against any model that ends the acceleration with the Flood year.  It does not show that accelerated decay never happened.  It shows that acceleration confined to the Flood is unavailable as an explanation for luminescence ages of post-Flood sediment, because the clock was not running during the Flood.  It had not been started yet.

What Oard's placement would have to deliver

Against Oard the question is not whether the acceleration can reach the dune but what it would have to do.  He gives it no factor and no history of how the rate changed, and his paper is about beryllium-10, not luminescence.  Four requirements follow from the numbers in this article.

1. The factor has to rise with the published age. A 32.5 ka dune needs the dose rate averaged over the 4,350 years since the Flood to be 7.5 times today's. A dune near the quartz ceiling of 100 to 150 thousand years needs 23 to 34 on the same averaging. If the acceleration was confined to the Ice Age, the factor while it ran must be larger still, because each dune has spent part of its burial at today's rates. An acceleration strongest early and fading later could in principle give older dunes more of it. Nobody has written down that curve.
2. The computed objections above do not stop it at a factor of 7.5, but they start to bite as the window shrinks. Squeeze the same dose into a few centuries and the factor moves toward the 100-year row of the table, where the dose to living things is the obstacle. Oard names the same problem himself: accelerated decay "presents a problem of too much radiation during the Ice Age."⁵
3. Whatever mechanism does it must keep the uranium isotopes in lockstep. The uranium-series article shows that measured δ²³⁴U holds ²³⁸U and ²³⁴U to the same speed-up within 0.015 per cent, which no proposed barrier-changing mechanism delivers.
4. It must fit luminescence, uranium-series and burial ages at once. Those three methods need different factors on samples that all formed within the same few thousand years, and whether one decay-rate history can supply them, and leave tree rings and ice layers agreeing with the results, is worked through in Geochemical Clocks. No young-earth author has published such a history.

So on Oard's placement the timing argument gives way to a cost argument.  The cost is unpaid rather than shown to be unpayable, and that is the honest state of the question.

The young-earth model has exactly two levers here.  One is a higher past dose rate, and the section above takes it away for post-Flood deposits on RATE's placement and leaves it unpriced on Oard's.  The other is unbleached grains carrying an inherited signal, which is real, is documented, and is the subject of the next section.  But it points the wrong way.  Inherited charge makes ages too old, never too young.  It is a source of error, not a source of a thousandfold error, and it is the field's own literature that found it.

Where the method genuinely fails

Jinmium, and the correction that was itself twice too old

In December 1996 Richard Fullagar, David Price and Lesley Head published thermoluminescence ages from Jinmium, a sandstone rock shelter in the Northern Territory of Australia.  They reported 50-75 thousand years for quartz sands associated with buried pecked circular engravings, and 116-176 thousand years for the underlying artefact-bearing deposits.¹⁵  If that were right it would roughly have doubled the accepted antiquity of human occupation of Australia, and it was reported worldwide as such.

Nigel Spooner published a critique in Antiquity in March 1998.¹⁶  On 28 May 1998 Richard Roberts and ten co-authors published the redating in Nature,¹⁷ and their abstract is worth quoting in full because it contains both the numbers and the mechanism:

"Using conventional (multiple-grain) optical dating methods, we estimate that the base of the deposit is 22 kyr.  However, dating of individual grains shows that some have been buried more recently.  The single-grain optical ages indicate that the Jinmium deposit is younger than 10 kyr.  This result is in agreement with the late-Holocene ages obtained for the upper two-thirds of the deposit from radiocarbon measurements.  We suggest that some grains have older optical ages because they received insufficient exposure to sunlight before burial.  The presence of such grains in a sample will cause age overestimates using multiple-grain methods, whether using thermoluminescence or optical dating."¹⁷

The mechanism in plain words.  The shelter is cut into a sandstone bluff.  Quartz grains spalling off the wall fall straight onto the floor and get buried without ever having seen the sun.  Those grains still carry the charge they accumulated inside solid rock over geological time.  Average them in with the properly bleached wind-blown grains and the average is far too high.  Measure the grains one at a time and they show up as a separate, much older population, and can be rejected.

Two numbers should be printed, not one.  The headline correction is from 116-176 thousand years to under 10 thousand, a factor of at least twelve.  But the intermediate answer, from multiple-grain optical dating, which is a better method than the original TL, was 22 thousand years.  That is still more than twice too old.  The second number is the more instructive one.  An improved method, applied honestly, got it substantially wrong as well.  Anyone using Jinmium as an argument should quote both.

And the timeline is the other half of the story.  Claim published December 1996.  Critique in Antiquity March 1998.  Refutation in Nature 28 May 1998.  Eighteen months from claim to published refutation, entirely inside the mainstream literature, with two of the original authors, Fullagar and Rhys Jones, listed on the refuting paper.  The follow-up, with the full single-grain and multiple-grain data, appeared in Archaeometry the next year.¹⁸

Malakunanja II, which is the control, and must be read next to Jinmium

The same team published a second paper in 1998, in Ancient TL, on a different Australian rock shelter.³  If the Jinmium failure were a failure of thermoluminescence as such, this one should have failed too.  It did not.

Their own conclusion:

"They also show that correct burial ages can be obtained using TL methods, provided the sediments are bleached sufficiently before burial to reset the dating signal.  The Jinmium sediments failed to meet this requirement (Roberts, 1997; Spooner, 1998), whereas adequate bleaching at Malakunanja II is indicated by the modern TL age for the near-surface sample (Figure 1), the similarity of palaeodoses obtained from TL signals (at 325°C and 375°C) with markedly different susceptibilities to light (Figure 4), and the concordance of TL and ¹⁴C ages at this site (Figure 1)."³

Same team.  Same continent.  Same year.  Four lines of evidence that the sediment was properly bleached, three methods agreeing, and a near-surface control returning 0.2 ± 1.2 thousand years, which is zero.

Printing Jinmium without Malakunanja II would be exactly the selective quotation this article criticises elsewhere.  Jinmium shows that the method fails when the grains were never bleached.  Malakunanja II shows that it works when they were, and gives a test for telling the two cases apart.  Those are the same finding read from two sides.

Zero-age tests

The cleanest way to test a clock is to run it on something whose age is known to be zero.

Prescott and Purvinskis, 1991.²  This one is the young-earth argument's second-best exhibit, and it comes from the method's own journal.  Sample CA20S/1 came from Carlo Blow in South Australia, a dune actively encroaching on living vegetation, so that "the time since the sample was last exposed to light can scarcely exceed a decade."²  Under the then-conventional total-bleach TL protocol it "gave an age of several thousand years."²  A second sample, from a currently mobile dune surface at the Roonka archaeological site, "previously gave an age of some 400 years."²

That is a standard protocol returning thousands of years for a deposit less than ten years old, stated in the authors' own words in a mainstream journal.  A method's own journal publishing its own failure mode is worth more than any critic's summary of it.

It is also the paper that fixed the problem.  Under a selective-bleach protocol reading the rapidly-bleaching 325 °C peak, both samples returned modern age, and the authors state the sensitivity: "limited tests show that a true ED of about 0.1 Gy would be detectable."²  Their conclusion is that "the proposed selective bleach technique gives zero age for modern samples for which total bleach methods give non-zero ages."²  So the honest reading is that a protocol failed, the failure was measured, a numerical zero-age criterion was set, and the protocol was replaced.  Both halves belong in any fair account.

Alexanderson, 2007.⁸  Helena Alexanderson measured equivalent doses in modern sediment from four rivers on Jameson Land in East Greenland, which is close to the worst realistic bleaching environment there is: glacier-fed water under a high-latitude sun.  River-bed sand carried about 0.1 Gy.  Silt carried about 1.1 Gy.  A spring-flood sidebar carried about 1 Gy, "which corresponds to ca 500 years."⁸  A subaerial reference deposit was "completely zeroed (ED=0.00±0.07 Gy)," which at 1.90 Gy/ka is zero plus or minus 37 years.⁸

Her own framing of what that means: "Incomplete bleaching of this amount (1 Gy) is generally not a significant source of error for sediments of Pleistocene age."⁸  She is right, and the qualifier matters in both directions.  Half a millennium of spurious age is fatal to a Holocene date and negligible to a fifty-thousand-year one.

Note the size difference between these two failure modes, because they get conflated.  Ordinary incomplete bleaching in a river adds of order 0.1 to 1 Gy, which is hundreds of years.  Grains that spalled off a rock face in the dark carry a full geological dose and add tens of thousands of years.  Only the second produced Jinmium, and only single-grain measurement separates them.

The single-grain procedure itself, which is the strongest mainstream concession found

In 2016 Kristina Thomsen, Andrew Murray, Jan-Pieter Buylaert, Mayank Jain, Jakob Helt-Hansen and Thierry Aubry tested single-grain quartz OSL against four sediment samples from the Bordes-Fitte rock shelter at Roches d'Abilly in France with radiocarbon AMS control in the 22-48 thousand year range.¹⁹  Murray co-authored the single-aliquot regenerative-dose protocol that the whole field now runs on, and Thomsen the post-IR IRSL method for feldspar.  These are not outsiders taking a shot at the technique.  From the abstract:

"It is shown that had these samples been analysed in the absence of other age control, standard modelling decisions based on the shape of single-grain dose distributions would have led to significant misinterpretation of results and a corresponding >40% underestimation in age.  If we instead ignore this standard decision process and apply weighted average and mixing models then the most likely results deviate from the expected ages by >10%.  Finally, we show that by careful consideration of the luminescence characteristics of individual grains, we are able to obtain good agreement with the independent age control by applying alternative rejection criteria but this is at the cost of reducing the accepted grain population by more than an order of magnitude, with the corresponding inevitable decrease in precision."¹⁹

Their conclusion goes further than the abstract does:

"we conclude that, for these quartz samples, with doses in the range ~60 to 110 Gy, multi-grain analysis gives the most accurate ages; there is no advantage in the use of single grains.  Indeed, had these samples been analysed in the absence of other age control, the application of standard single-grain methods would have led to significant misinterpretation of results and a corresponding inaccuracy in ages."¹⁹

And they flag the awkwardness themselves: "Given the very large fraction of single-grains that must be rejected to provide accurate single-grain dose estimates, it is of course surprising that multi-grain dose estimates (based on the sum of signals from acceptable as well as unacceptable individual grains) provide accurate dose estimates without any further data selection."¹⁹

What this quotation says and what it does not.  It says that the standard decision procedure for single-grain analysis, applied blind, would have been out by more than 40% on these four samples.  It does not say luminescence ages are generally 40% wrong, and it does not say the method failed here: the multi-grain quartz, the feldspar pIRIR₂₉₀ and the radiocarbon all agreed.  Quoting the 40% without the surrounding sentences would be a misuse of it, and this article says so in advance because the temptation is obvious.

The rest of the honest failure list

Saturation.  Traps fill up.  Once they are full, more dose produces no more signal and the age is only a lower bound.  Murray and colleagues put natural doses as not increasing above roughly 200 Gy, and add that "the controls on luminescence saturation with dose are poorly understood."⁹  At 1.539 Gy/ka that caps quartz at about 130 thousand years; in low-activity sediment at 0.79 Gy/ka it reaches about 253 thousand.  So the honest working range for quartz OSL is roughly a few decades to 100-150 thousand years.  Feldspar pIRIR exists because feldspar saturates later and reaches further back.  Snelling's ESR objection is a correct description of this effect and the field agrees with him about it.

Anomalous fading in feldspar.  Feldspar leaks charge.  Svenja Riedesel's 2025 review in Ancient TL²⁰ states that fading "has been shown to be ubiquitous in feldspars, affecting both thermoluminescence as well as optically (infrared) stimulated luminescence signals," that the accepted mechanism is "quantum mechanical tunnelling of trapped electrons from the ground state of the electron trapping centre to a nearby recombination centre," and that "fading causes an underestimation of the luminescence age."²⁰  The review notes that measurements "all over the globe have shown the ubiquitous nature of this process ... questioning the validity of the term 'anomalous' fading."²⁰  The correction is the g-value, "the percentage of luminescence lost due to fading per decade of time," measured in the laboratory by giving the same dose repeatedly and reading it after different storage delays.²⁰  The two standard correction models are Huntley and Lamothe 2001 and Kars 2008, and the review shows the first going negative at long times for large g, which is a known pathology.  So a feldspar age carries a modelled correction whose model is imperfect.  Quartz OSL, the workhorse, does not fade and needs no such correction.

Radon escape.  ²²²Rn has a half-life of 3.8 days and is a gas, so it can diffuse out of porous sediment, taking its daughters' contribution to the dose rate with it.  Guérin's tables give both "Total" and "Pre-Rn total" rows precisely for this reason.  Under total radon escape the natural-uranium alpha term falls from 2.795 to 1.267 Gy/ka per ppm, a 55% swing.⁶  This is why gamma spectrometry is used to check whether the chains are in secular equilibrium.  Roberts and colleagues report exactly that check at Malakunanja II, "which showed that a condition of secular equilibrium presently exists in the ²³⁸U and ²³²Th chains."³

Microdosimetry.  Individual grains sit at different distances from individual radioactive mineral inclusions, so the beta dose varies from grain to grain even in uniform-looking sediment.  Murray and colleagues concede the consequence: "inherent dosimetry uncertainties likely prevent the accurate dating of deposits subject to significant post-depositional mixing."⁹

What the validation record actually shows

Three separate kinds of test bear on whether the method works.

Zero-age controls.  A modern subaerial deposit in Greenland returned 0.00 ± 0.07 Gy.  A near-surface sample at Malakunanja II returned 0.2 ± 1.2 thousand years.  Both are clean passes.  Against those, a modern dune returned "several thousand years" under a 1980s TL protocol.  All three are in the preceding section, and all three are mainstream measurements.

Cross-method agreement.  At Malakunanja II, TL, multi-grain OSL, single-grain OSL and radiocarbon agree.  At Bordes-Fitte, multi-grain quartz OSL, feldspar pIRIR₂₉₀ and AMS radiocarbon on ultrafiltered bone collagen agree across 22-48 thousand years.¹⁹  At Jinmium, single-grain OSL agrees with radiocarbon while TL and multi-grain OSL disagree with both, and it was that disagreement that exposed the error.

Compilations against independent age control.  Murray and colleagues summarise the whole record: the performance of sand-sized quartz OSL "to date samples from a variety of deposition environments, each with independent age control, has been evaluated by others.  These data sets are combined in Fig. 11b, for 249 ages there is now a considerable body of data suggesting that OSL dating is accurate over a time span of more than four orders of magnitude, and that the precision and accuracy are largely in the range originally predicted."⁹  They also give the realistic floor in the same paper: improvements should "help improve luminescence age accuracy to the probable minimum of ~5%."⁹

Two caveats on that last paragraph, and this article will not skip them.  The three compilations underneath the 249-age figure are Murray and Olley 2002, Madsen and colleagues 2005, and Rittenour 2008, and none of them was obtained for this article.  The commonly repeated claim that OSL is accurate to about 10% traces back to Murray and Olley 2002, and that paper was chased through the Geochronometria archive, CORE and BASE without success.²¹  So the accuracy claim is reported here at second hand and flagged as such.  Second, the validation studies quoted throughout this article come from a small number of institutions: the Risø and Aarhus laboratories in Denmark, and the Roberts and Olley group in Australia.  That is not a conspiracy, it is a small field, but the independent-age-control literature is narrower than "249 ages" makes it sound.

So the fair summary is 5-10% accuracy in favourable material, worse when bleaching, saturation or water history are in doubt, and a documented failure record that is real but bounded.

How thin is the creationist engagement, exactly?

Thinner than the mainstream literature, and not as thin as the first version of this page said.  That version reported that ICR's search URLs 404 and that creation.com "is genuinely unaudited here."  Both statements were true about the search boxes and both were wrong about the literature.  ICR had published a whole article on thermoluminescence dating.  creation.com had a canonical assumption sidebar on TL, OSL and ESR that has been in print since 2003.  The Creation Research Society Quarterly had a nine-page paper on the subject by a professor of medical physics who does thermoluminescent dosimetry for a living.  All three were found by enumeration after the search boxes had said nothing.

That failure is worth printing, because it is a general one.  A search box whose recall is unknown can widen a search.  It can never close one.  Any claim on this site of the form "the young-earth literature has not engaged this" has to rest on an enumeration — a sitemap, a harvested contents index, or a grep of downloaded full text — and the negatives below have been rebuilt on that basis.

The proof that a search box cannot close a search

icr.org's search endpoint is https://www.icr.org/home/submitsearch/?f_keyword_all=<term>.  Queried on 30 August 2026 it answered "Your search returned no results" for thermoluminescence, and for the control term zircon as well.²²  Meanwhile https://www.icr.org/sitemap.xml returns 29,086 URLs, of which 19,013 are article slugs, and one grep of that list for luminesc returns three hits: two on bioluminescence in animals, and examining-thermoluminescence-dating.  The article the search box could not find has the search term in its title.

What the enumerations actually found

The ICC negative is now a real negative, established at body-text level.  All 457 ICC paper PDFs from the nine conferences were downloaded and converted to text.  The pattern luminescen|thermolumin|electron spin resonance matches six of the 457, and not one of the six is about dating.²⁵  Four are cathodoluminescence, an imaging technique for looking at growth zoning in quartz and diamond.  One is bacterial bioluminescence in pony fish.  The sixth is Emil Silvestru's 2003 paper on caves,²⁶ which contains the parenthesis quoted below and one thermoluminescence reference in its bibliography.  The controls behave: radiocarbon matches 37 of the 457 papers and zircon matches 27.  So in thirty-seven years the young-earth movement's technical conference has never had a paper on luminescence dating, and that is now a countable fact rather than the result of a search.

Silvestru's parenthesis is the whole of it:

"There exists a solid and well-documented literature regarding the issue of reliability of 'absolute dating' and alternative methods (thermoluminescence, fission tracks, dendrochronology etc.).  Their main point is that they should not be taken as infallible."²⁶

That is one clause in a list, in a paper about cave art.  It contains no analysis of the method.

creation.com is no longer unaudited.  Its search is an unauthenticated Meilisearch endpoint that indexes full article bodies and needs no browser: https://backend.creation.com/meilisearch/search?q=<term>&types=article&locale=en&hitsPerPage=40&page=1&hybrid=false.²⁷  Run on 30 August 2026 it returns 25 articles containing luminescence, 12 containing thermoluminescence, 11 containing the phrase "optically-stimulated luminescence" and 13 containing "electron spin resonance", against controls of 99 for zircon and 398 for radiocarbon.  Of those 25, exactly one contains a methodological account of the method, and it is Tas Walker's "Dating methods" sidebar, quoted above.  The other 24 cite an OSL, TL or ESR date in passing while reporting some mainstream disagreement — Denisova Cave, Homo floresiensis, the White Sands footprints, the Namib sand sea, Sterkfontein.  None of them makes an argument this article has not answered.

One of the 24 deserves a sentence, because it is a fairness point that cuts against the usual caricature.  An undated creation.com news item, "Age of Aboriginal Art Disputed," reported the Jinmium correction as it happened and reported it accurately, by quoting a mainstream luminescence specialist rather than dismissing the method:  "A luminescence dating expert at the Research School of Pacific and Asian Studies, Dr Nigel Spooner believes that the dates could be out by a factor of 10!  He claims that the maximum age for the artefacts is 20,000 years, and 'may be quite a bit less than that'."²⁸  Whatever else is true of the movement's engagement, its response to the field's most famous failure was to cite the field's own correction.

Separately, a harvest of 6,741 creation.com and Journal of Creation contents-page entries — including 150 Letters to the Editor, 637 Perspectives, 106 Viewpoints and 49 Forum items — contains exactly one title matching luminescen, and it is about fireflies.²⁹  No luminescence-specific letter, perspective or forum piece exists in forty volumes of Journal of Creation.  One caveat belongs beside that: the Creation Research Society Quarterly's online repository types all its items as feature articles and appears not to ingest its letters column at all, so a reply to DeWerd could exist there unseen.  On the evidence that can be enumerated, nobody inside the movement has publicly disputed anything anybody in the movement has said about this method.

The two books

Marvin Lubenow, Bones of Contention, revised edition, 2004.  Obtained and searched.  It mentions thermoluminescence three times, all of them reporting mainstream dates or listing references, and "electron spin resonance" once, in an endnote.³⁰  Its chapter on dating devotes itself to a "coverage gap":

"This is because there is a coverage gap involving the dating methods.  This gap involves the methods evolutionists believe are the most reliable—radiocarbon and potassium argon (K-Ar).  This coverage gap occurs during the very heart of the time period most crucial to human evolution—the period from about 40,000 years ago to about 200,000 years ago on the evolutionist's time scale."³⁰

The window Lubenow says cannot be dated is precisely the window luminescence and ESR were built to fill.  Cupps says the same thing from the other side, and treats it as the method's motive: "So a substantial gap exists between dating objects less than 50,000 years old and more than one million years old.  The relatively new luminescence dating technique attempts to fill this gap."¹²  Lubenow names radiocarbon and potassium-argon, argues that neither reaches 40-200 thousand years, and does not name TL, OSL or ESR anywhere in that chapter.  That is a real omission in the most widely read young-earth book on hominin dating, and it is the reason the "coverage gap" argument reads as dated: quartz OSL works from decades to 100-150 thousand years, which covers the whole of the gap he describes.

John Woodmorappe, The Mythology of Modern Dating Methods (Master Books, 1999).  Still not obtained, and now with the routes recorded.  Open Library holds a catalogue record for it with no Internet Archive scan attached; an Open Library full-text search for the title returns only other books citing it, including John Reed's Rocks Aren't Clocks; and archive.org's own search finds no copy.  It appears never to have been scanned.  Its chapter on non-radiometric methods therefore remains unread, and that is the one real hole left in this survey.

What the enumerations do not settle

Three things.  First, ICR's search box cannot be used to establish that ICR has nothing further, and this article's enumeration only covers what is in the sitemap; print-only Acts & Facts material from before the web archive may exist.  Second, the Journal of Creation body index reaches article text but the pre-2000 CEN Technical Journal back catalogue is only partly digitised.  Third, Woodmorappe.  None of those three gaps is likely to change the picture, because the four sources that were found agree with each other and none of them proposes a mechanism that could compress an age by more than about a factor of two.  But they are gaps, and this article does not pretend otherwise.

Common objections, and answers

"The dose rate you measure today is not the dose rate over the whole burial history."  Correct, and it is the strongest form of the constant-rate objection.  Three things bound it.  Radon escape is the main way the dose rate can have differed, and gamma spectrometry tests for it directly by checking whether the ²³⁸U and ²³²Th chains are in secular equilibrium today.  Water content is the other, and it moves the age by at most about ±10% across the physically possible range.  Beyond those two, changing the dose rate by the factor a young-earth chronology needs, 7.5 at minimum, means changing the decay constants themselves, which brings the timing problem in the next objection.

"Production of trapped charge assumes decay rates that are the thing in dispute."  This is the only one of Snelling's three assumptions that applies, and it is a fair objection in principle.  In practice it turns on timing.  The trap population was emptied by sunlight at deposition.  For a post-Flood dune, that means the entire accumulated dose was laid down after the Flood.  So it is not enough for decay to have been accelerated during the Flood; it has to have been accelerated afterwards, by an average factor of 7.5 over the 4,350 years since, in that specific sediment, before returning to today's value.  RATE's episodes end with the Flood year,⁴ so they cannot do it.  Snelling's "during and soon after the flood" reaches only a dune bleached inside that window.¹  Oard's acceleration "during the Flood and Ice Age" does reach an Ice Age dune,⁵ and a writer who takes that route has to supply what Oard does not:  a decay-rate history that gives each dune the dose its published age implies, at a radiation cost Oard himself calls "a problem."⁵

"Water content is guessed, and it scales the age directly."  Also correct, and the field says the same thing in the same terms.  The relationship is close to one for one: a percentage point of lifetime average water is a percent of age.  The honest range of the guess is the difference between bone dry and saturated, and that is about ±10%.  It is a real error term.  It is four orders of magnitude short of what a young-earth chronology needs.

"Jinmium shows the method can be wrong by a factor of ten."  It does, and the article prints the number rather than minimising it.  Three things belong beside it.  The correction came from inside the field, in eighteen months, with two of the original authors on the refuting paper.  The failure had a specific, identified cause, quartz spalling off a shelter wall in the dark, and a test for that cause, single-grain measurement, which is now standard.  And Malakunanja II is the same team's control from the same year, where TL, single-grain OSL and radiocarbon all agree and the near-surface sample reads zero.  A method that can be wrong by a factor of twelve under an identifiable condition, and that can detect that condition, is not thereby a method that can be wrong by a factor of a million.

"Thomsen 2016 shows the standard procedure is 40% wrong."  It shows that the standard single-grain decision procedure, applied blind, would have been more than 40% wrong on four samples.  On those same four samples multi-grain quartz OSL, feldspar pIRIR and radiocarbon all agreed.  The authors' own conclusion is that multi-grain analysis was the more accurate method there.  It is a serious result and a genuine embarrassment to a fashionable technique.  It is not a demonstration that luminescence ages are worthless, and quoting it as one would be dishonest.

"Concordance between methods proves nothing, because they can be concordantly wrong."  This is the best version of the skeptical case and it has a documented example.  At Lake Mungo in 1999 four apparently independent lines, ESR, two uranium-series methods and OSL, agreed on 62 ± 6 thousand years for the Mungo 3 skeleton.³¹  In 2003 a series of 25 optical ages put both burials at 40 ± 2 thousand years, and the older figure was wrong by 55%.³²  The reason is that ESR, Th/U and Pa/U were all reading the same uranium in the same open-system bone and shared a single uranium-uptake assumption, so they were not independent at all.  That case is worked out in radiometric concordance and is not re-derived here.  What it establishes is that concordance is evidence, not proof, and that the strength of the evidence depends on whether the methods share an assumption.  What it does not establish is any error larger than a factor of two.

"All these corrections came after the fact.  The method is unfalsifiable."  The opposite is closer to true.  Prescott and Purvinskis published a numerical zero-age criterion in 1991: a protocol must return zero on a modern sample, and 0.1 Gy is detectable.  That is a falsification test with a threshold, published by practitioners, and the total-bleach protocol failed it and was dropped.  Jinmium was falsified by radiocarbon and by single-grain measurement.  The 62,000-year Lake Mungo figure was displaced by a new series of 25 optical ages, published in Nature by a team that included Nigel Spooner, an author of the 1999 paper it corrected.³²

"A full uncertainty budget shows a 20,000-year TL date could really be 2,900 years."  This is DeWerd 2020 and it is the best-argued young-earth objection there is.  It fails on its own arithmetic.  His expanded uncertainty is 85.44%, which on 20,000 years gives a band of 2,912 to 37,088 years.  He prints the lower bound and replaces the upper bound with 10,000.  Beyond that, an uncertainty band is symmetric by construction and cannot bias a whole literature downward; and most of the eleven terms in his table are absolute-dosimetry terms that cancel in a substitution measurement, because the natural signal and the calibrating dose are read from the same grains through the same optics.  Keep only the environmental terms and the honest expanded uncertainty is about ±42%.  That is a factor of 1.4.  A young-earth chronology needs 7.5.

"The environmental radiation dose rate varies with location, season, solar activity and even time of day."  This is Cupps 2016, and it is true of the external radiation field a person walks around in.  It is not true of a buried grain.  For the worked example in this article, beta and gamma radiation from the surrounding sediment supply about 1.39 of the 1.54 Gy/ka, and the cosmic-ray term is 0.15 Gy/ka.  Solar modulation moves the cosmic-ray flux by 10-20% over the eleven-year cycle, which is one or two percent of the total.  Season and time of day move it by nothing.  The one large variation in a buried dose rate is radon escape, and that is checked directly by gamma spectrometry.

"The crystal's sensitivity to radiation may have changed."  This is Walker 2003, and it is the best of his four assumptions.  Quartz does change sensitivity, and an uncorrected change corrupts the calibration.  It is also solved, and was solved three years before he wrote.  The single-aliquot regenerative-dose protocol follows every measurement with a fixed test dose and divides the signal by the test-dose response, so what is fitted is a sensitivity-corrected ratio.  Sensitivity change is measured on each aliquot and divided out rather than assumed away.

"Nobody has answered this because the creationist literature has not engaged it."  Not true, and this article said something close to it in its first version before finding out otherwise.  Five young-earth sources address the method: DeYoung 1994, Walker 2003, Cupps 2016, Snelling 2018 and DeWerd 2020.  Two of them are by authors with directly relevant doctorates.  They are answered above, individually, on their arguments.  What is true is that the movement's technical conference has never had a paper on the method in thirty-seven years, and that no Letter, Perspective, Viewpoint or Forum item in forty volumes of Journal of Creation carries a luminescence title.  Whether anyone answered DeWerd inside the Creation Research Society Quarterly cannot be said either way, because that journal's online repository ingests feature articles and not its letters column.

"You are relying on secular sources for the accuracy claims."  Yes, and deliberately.  The point of citing Murray, Thomsen, Roberts and Prescott for the method's failures is that they have no motive to invent them.  Where their evidence has not been obtained, as with Murray and Olley 2002, this article says so instead of repeating the number on trust.

Where the young-earth case is strongest

It is strongest as a case against uncritical trust in a published date, and it is not strong as a case for a young earth.  Three exhibits carry it, all mainstream, all verbatim, and none of them creationist in origin.

1. Jinmium. 116-176 thousand years in Antiquity, revised to under 10 thousand in Nature. A factor of at least twelve on a site that made international news. And the first correction, multi-grain OSL at 22 thousand years, was itself more than twice too old, which is the part most retellings leave out.
2. Prescott and Purvinskis 1991. A then-standard TL protocol returned "several thousand years" for a dune less than a decade old, and about 400 years for a modern dune surface. Published in the field's own journal, in the authors' own words.
3. Thomsen and colleagues 2016. The authors of the single-aliquot regenerative-dose protocol and of feldspar pIRIR reporting that their own standard single-grain decision procedure, applied blind, would have underestimated four radiocarbon-controlled ages by more than 40%.

Add Lake Mungo, where four apparently independent methods agreed to within 2% and were then displaced by a figure 55% lower.^{31 32}

And add one argument that is creationist in origin and is a good one.  Larry DeWerd's 2020 complaint that published luminescence ages report a Type A standard deviation and almost never a full NIST-style Type B budget is correct, and the judgement terms he wants propagated — lifetime water content, secular equilibrium, whether the sediment's radionuclide content has changed — are exactly the terms a critic should ask about.  His own arithmetic does not support the conclusion he draws from it, for the reasons set out above, but the demand itself is legitimate and the field has not fully met it.

That is a real, citable body of evidence that luminescence ages can be badly wrong, that a published number with error bars is not the same as a true number, and that agreement between related methods has already deceived a competent team once.  Anyone who tells a young-earth audience that geochronology never gets caught out is not telling the truth, and these four cases are the answer to him.

Where the young-earth case fails

It fails on the size and the direction of the errors, and on the mechanism.

Every documented failure listed above is an overestimate, and every one has the same cause: charge that sunlight never removed.  A young-earth argument needs published ages to be too old by three to four orders of magnitude.  The documented errors are factors of 1.5 to 12.  There is no case in this literature of a luminescence age being wrong by a factor of a thousand, and no proposed mechanism that would produce one.

Every one of those failures was caught by the same method or a sister method, usually within two years, and published by the people whose careers depend on the technique.  A field that publishes its own zero-age failures in its own journal, and then publishes the fix and a numerical threshold, is not behaving like a field hiding a systematic error.

And the mechanism a young-earth model would need is unavailable on RATE's placement and unpriced on the post-Flood placements in print.  Accelerated decay confined to the Flood year cannot reach a dune deposited after it, because sunlight emptied the traps at deposition.  To reach it, the acceleration has to keep running after the Flood.  Oard says it did, through the Ice Age,⁵ but he gives it no factor and no history, applies it to beryllium-10 rather than to any luminescence age, and concedes that it "presents a problem of too much radiation during the Ice Age."⁵  Until someone supplies those numbers it is a placement, not a model.

The framework Answers in Genesis applies is also wrong for this method, and that is a separate failure.  Two of its three assumptions describe a parent-daughter system.  Luminescence is not one.  Repeating the three-assumption list against a dosimeter tells a knowledgeable reader that the writer has not understood what he is criticising, and it costs more than it gains.  That failure belongs to the framework, not to the movement: Walker, Cupps, DeWerd and DeYoung all go straight to the dose rate, which is the only assumption a dosimeter has.  Their arguments fail for other reasons, worked through above, but none of them fails for that one.

Verdict

Luminescence dating is a real measurement and, on this evidence, it works.  Both terms in the age equation are measured on the sample itself rather than borrowed from a global calibration.  Modern sediment returns zero to within about 0.1 Gy.  Independently dated deposits are recovered across four orders of magnitude.  OSL, TL, feldspar pIRIR and radiocarbon agree at Malakunanja II and at Bordes-Fitte.  Realistic accuracy is 5-10% in favourable material.

The method's real failure modes are close to the ones the five young-earth authors name.  Grains that never saw daylight inflate ages, sometimes by more than a factor of ten.  Traps saturate near 200 Gy, which caps quartz around 100-150 thousand years.  Feldspar leaks and needs a modelled correction with known pathologies.  Water history is estimated rather than measured and moves the age by up to ±10%.  Radon escape can swing the alpha dose rate by 55%.  The standard single-grain procedure has been shown to fail by more than 40% on samples with good control.

None of that scales to a young earth, and the reason is structural rather than arithmetical.  The errors run in the wrong direction, they are two to three orders of magnitude too small, and the one lever large enough to matter cannot reach a sediment whose clock was set to zero by sunlight after the Flood unless the acceleration kept running after the Flood.  RATE's model has no such episode.  Oard's has one, through the Ice Age, and against it the timing argument no longer decides the case;  what stands there instead is that nobody has turned his placement into rates that fit the dated dunes, the uranium series and the radiation budget together.

So the honest verdict is that on luminescence dating specifically the young-earth position is weak.  Saying otherwise would require pretending that failures which all run one way, and which the field finds and fixes itself, are evidence of a systematic inflation by six orders of magnitude.  They are not.  The young-earth writer who wants to use this literature honestly has two true and useful things to say with it — that published geochronological ages have been wrong by an order of magnitude and that concordance is not proof, and that published luminescence ages report a Type A uncertainty where a calibration laboratory would demand a full Type B budget — and he should say those and stop there.

Companion pages

• The literature, with every source, what it measured, how it was obtained, and which way it cuts.
• Quarantine, the numbers and locators this article declined to print as fact, with what was tried for each.
• Radiometric concordance, which works out Lake Mungo and the uranium-uptake problem that made four methods agree on a wrong answer.
• Hydroplate heat, for the accelerated-decay heat problem as it applies to the crust rather than to a dune.
• research/dose-rate.ts in this article's folder, which computes every number in the arithmetic sections from the Guérin et al. 2011 conversion factors.

Sources