Hi Renaud,

Thanks for your reply. My question is more philosophical than practical. I appreciate that variability inherent in different samples from the same stratigraphic depth will influence the estimated age, as will the application of different pre-treatment methods, but many researchers report dating results from multiple dating facilities without explaining these potential sources of error. Given that commercial radiocarbon dating facilities are expected to calibrate their mass spectrometers using universal standards of known age, would the use of more than one facility be beneficial, or would it be better to submit all samples to the same lab?

Thanks Renaud,

Your parting comment strikes a chord with me. I once had bones dated by a commercial lab that were just too good to be true so I had the same specimens redated by a research lab who found the bones were devoid of collagen! This made me suspicious of all commercial labs for a while, but the probability that the first set of results could be explained by chance were 1 in 5040! So what did the first lab date and were their ages somehow equivalent to the true age of the bones? I guess the bottom line is that researchers using dating results cannot behave like simple consumers but rather must have a good understanding of the dating and pre-treatment processes used as well as the implications they have on interpretation of dating results.

HI Renaud:

I can provide more answers based on my experience with luminescence dating. When sending samples to two or three labs (which happens more than one would suspect) you will not always get the same age BUT YOU SHOULD GET THE SAME AGE WITHIN ERROR. For example, suppose Lab A returned an age of 17 ± 2 ka, while Lab B returned an age of 15 ± 3 ka and Lab C returned an age of 19 ± 2 ka. You might think this is all too much scatter and you never got the same age from all three labs, until you realize that the labs did not return ages of 10 ka vs 20 ka or 5 ka vs 50 ka...they all returned answers easily within the error range (15-19 ka for Lab A, 12-18 ka for Lab B, and 17-21 ka for Lab C).

Some people might say, well that's OK for luminescence dating because they always have big errors on their ages, but even within the smaller errors for radiocarbon or Ar/Ar dating, there will be overlap. (Or at least we HOPE so!). And all labs, even luminescence ones, run standards so they can know when sampling procedures or machine tolerances are off. Yes, even luminescence labs have a standard provided by Riso National labs (a coastal sand from Denmark if I remember correctly).

And finally, research labs may be too busy to run samples they have not collected and therefore do not know if they were sampled from the "correct" deposit or in the "correct" way, but this can be easily solved by sending a student to the lab before hand or during a study. Research labs will often have summer teaching courses or otherwise jump at the chance to use student help in exchange for imparting the protocol they use for sampling and measurements.

Hi Shannon

Thank you for your contribution. This is exactly the type of discussion I was hoping my question would generate! Your point regarding ages with overlapping errors is well taken, though I do wonder why the error of OSL ages are reported as one standard deviation instead of two. Research labs do seem to take a long time to generate a result but I have always attributed this to the meticulous care they take in sample preparation. One would hope that since they are getting paid to provide an age that they would treat it as high priority (though this sounds naive even as I write it). I think collection procedures, particularly the in situ acquisition of background dose rate, are probably more important for OSL dating than other techniques commonly applied to Late Quaternary assemblages but given the expense of any dating method, I consider it common sense to at least look at a dating company/lab’s web page to ensure that you follow their recommended collection procedures.

Hi Matt

One standard deviation seems to have been the protocol and we just keep following protocol. I have reported ages as two sigma sometimes and the errors look pretty horrific, so maybe that is why people keep the one sigma. Also, as I understand it, two sigma is kept for isotopic decay methods that are measured as high precision on mass spectrometers...and, of course, luminescence is not high precision (yet!) and is not an isotopic decay method.

Generally OSL takes a long time to get back because of the number of measurements we have to make to get an equivalent dose. At least 50 measurements for one sample! The sample prep is actually pretty quick (less than one week for a batch of 20-nothing like the cosmogenic prep work). And, of course, there are too few labs for too many projects so we all get backed up and will keep being backed up as long as OSL is as popular as it is.

One time I actually calculated how much it should cost per age if we took into account machine time (or machine cost), lab space, lab consumables, and salary of techs and researchers. It would be $1,100 (US)!! So hopefully that $400 or $500 per sample you pay makes you feel better. Half the cost is absorbed by overhead costs I guess.

AND...yes...measuring the dose rate is MORE important than many realize.

Hi Shannon,

Your comments are congruent with my limited experience with OSL, though all of the OSL ages I’ve commissioned have been done using the single grain technique on hundreds of quartz grains which as you would know takes even longer but has the advantage of providing a population of ages and helps to detect partial bleaching and mixing of different aged sediment. In Australia, even at ‘mates rates’ were paying around AU$1200 per age which seems expensive but for assemblages that are beyond the reach of radiocarbon dating and lacking speleothems, OSL is pretty much the only option. This brings me to another question: What is your opinion on comparison between ages achieved using different dating techniques applied to different samples and indeed different substances from the same assemblage?

Thanks for the heads up! It would be interesting to apply your technique to fossils from Kelly Hill Cave on Kangaroo Island where I have excavated an assemblage and dated it using radiocarbon on bone, U/Th on speleothems and OSL on quartz.

It's also important to check how the uncertainities are being propagated and reported. An age result may only report internal measurement uncertainty and not propagate the uncertainty on a tracer or reference. That is fine if comparing between analyses made in the same session with the same tracer/reference, but the uncertainities will be underestimated if the age is being compared to data from other analytical sessions, other tracers or even other methods. Hence it could suggest that there is scatter of results between different labs whereas it is actually overly optimistic precision.

I was concerned about this Keith. Could you recommend a few papers that investigate this problem?

I'm not aware of anything published specifically discussing the uncertainity propagation issues. There was a recent article in Elements that gives a good description of precision and accuracy: http://dx.doi.org/10.2113/gselements.9.1.19

I have seen talks from BGS/NERC guys Dan Condon, Joe Hiess and Matt Horstwood that have included a very interesting flowchart illustrating sources of values (and hence uncertainities) in a U/Pb TIMS measurement all the way back to SI units, but I'm not sure if that has been published anywhere - I will ask!

Thank you Keith.

Dan Condon has let me know that there are some pre-prints about the tracer calibration experiment with the figure I described available on the EarthTime website:

http://www.earth-time.org/pdf/condon_tracer.pdf

http://www.earth-time.org/pdf/mclean_tracer.pdf

http://www.earth-time.org/pdf/calpos.pdf

Thank you for directing me to Condon et al.'s awesome work. Do you think the approach is applicable to any radiometric dating technique?

The concepts of traceability could certainly be adapted to any technique. Many techniques would actually build on this chart because they rely on TIMS U-Pb dating of their references. When time permits I may try to adapt the chart for SHRIMP/SIMS U-Pb analyses.