This is an unusually important post on how to solve the problem of the date of the Qur’an, by my colleague Stephen Shoemaker, connected with his earlier scintillating discussion based on his recent book Creating the Qur’an, which you can check out here: Creating the Qur’an: A Historical-Critical Study eBook : Shoemaker, Stephen J.: Books

The question is: can’t you just do a scientific dating of the Qur’an manuscripts and quickly solve the question: when were they produced?  The answer may surprise you.  It enlightened *me*


Radiocarbon Dating and the Origins of the Qur’an:

The Perils of Scientism and Internet Sensationalism


Bart invited me to make another post or two about studying the origins of the Qur’an from a historical-critical perspective, and right off the bat I knew that I needed to write something about attempts to radiocarbon date early Qur’anic manuscripts. It turns out that over the last ten years this topic has become the 800-pound gorilla in the room (to mix metaphors), and much like an actual 800-pound gorilla in any room, it has become a huge distraction and a hindrance to clear thinking. To borrow another trite metaphor, many scholars have invoked the radiocarbon analyses of certain early manuscripts as if these were a kind of silver bullet capable of bringing an end to all further discussion about the Qur’an’s origins. With the evidence of “science” now firmly in hand, the conversation is over, they maintain, and abracadabra, the Qur’an’s early history is, as it turns out, exactly what the Islamic tradition says it is.

Nevertheless, the problem is that most scholars who are eager to impose such scientific closure on this thorny and contested topic do not appear to understand fully the subtleties and limitations of the method in question. In the interests of brevity, I will not explain in detail how this method of radiocarbon analysis works, although interested readers may consult the third chapter of my Creating the Qur’an, where I explain the process, its basis, and its limitations in clear and basic terms. There too I address in a much more sophisticated manner the various points that I make in this post, as well as identifying even deeper problems with the ways that radiocarbon dating has been used in Qur’anic studies.

Radiocarbon dating is of course a remarkable tool for the historian when used properly, but in our case, it turns out to be more of a chainsaw than a chisel, when the latter is what we need. For instance, if one needs to date an object broadly, say to determine whether it was manufactured in either late antiquity or the late Middle Ages, then radiocarbon dating is your best friend. With this sort of range in view, its results are clear and decisive. So, for instance, when scholars used this method to date the Shroud of Turin, the results definitively identified this object as a medieval, rather than ancient, fabrication. Within a range of centuries, then, radiocarbon dating is rock solid.

The problem in our case, however, is that debates around the formation of the Qur’an hinge on a matter of a couple of decades, rather than centuries. There can be little question that the Qur’an as we now have it had come into existence by the early eighth century, which radiocarbon dating solidly confirms. Yet the more pressing question is whether the Qur’an’s definitive, canonical form was established by around 650, or 670, or 690. Such precision is more than radiocarbon dating can provide, as any archaeologist or historian familiar with this method will acknowledge. It is a lesson that scholars of biblical studies, and the Dead Sea Scrolls in particular, learned decades ago, and even still New Testament scholars struggle to identify the dates of the earliest fragments of the gospels and Paul’s letters. But back to the Qur’an.

In recent years samples were taken from several early Qur’anic manuscripts in order to determine the date of their production using radiometric analysis. Some of the initial results appeared to align very favorably with the canonical account of the Qur’an’s origins (which I outlined in my previous post), quickly creating a sense among more traditional scholars that at last science had proven the “skeptics” wrong. The first efforts to date a Qur’anic manuscript using this method involved a single folio that was stolen from Yemen and now is in the possession of Stanford University. Although these initial results were published in a scholarly journal, soon thereafter radiocarbon datings of early Qur’anic manuscripts in Birmingham, England and Tübingen were announced on library websites, only without full publication and analysis of the data. Encouraged by online articles breathlessly heralding discovery of “the oldest Qur’an,” social media (and Twitter especially) quickly declared that the matter of the Qur’an’s origins had been decided. Science had proven the veracity of the traditional Islamic account of the Qur’an’s formation, a verdict pronounced via Twitter threads rather than through the careful argument and evidence of a scholarly publication. Science had spoken, and science was infallible.

Of course, any good scientist will tell you that scientific data must always, just like any other evidence, be carefully interpreted, and while scientific measurement can be decisive in some instances, in others, the investigator is confronted with significant ambiguity. Within a range of centuries, radiocarbon data is decisive. But beyond this broad scope, more precision requires greater interpretation of the radiometric data. And once we begin to reach for individual decades, the data can longer be meaningfully interpreted. Ironically, in applying this method of dating to early Qur’anic manuscripts, scholars have seemingly created more problems than they have been able to solve. For example, some folios returned dates indicating that the manuscript in question was a hundred years or more earlier than Muhammad himself!

In other instances, a single page was analyzed independently by as many as four different labs. One would expect that if this method were accurate and reliable, the results would be consistent, regardless of where the analysis was done. But instead, the results were often all over the place: a folio dated to 611-660 by one lab was determined by another to date sometime between 433-599; another folio dated 590-660 by one lab was dated by another to 388-535. Now, these results are rather decisive if one wants to know whether the object in question is ancient or medieval. Yet given such widely varying datings of the same object, one obviously cannot seek to pinpoint a particular decade. The data are too skewed, and even by the standards of radiocarbon analysis, these results were more inconsistent than one generally expects from this method. Still, the Qur’anic sages of the Twitterverse were unshaken. Clearly – they proclaimed with neither evidence nor argument – the labs that returned the data that they didn’t like simply had botched the job (I am not making this up).

Eventually, some scholars instead began to take the scientific data seriously and to ask whether, for whatever reason, something was not working quite right with the radiocarbon dating of objects from the early medieval Near East. And so, they turned to early manuscripts whose dates were already well established, through scribal notes or the like, and had them dated using radiocarbon analysis. In each instance, the date that was returned was significantly off, by as much as a century in some cases, further calling into question the use of this method for dating early Qur’anic manuscripts within more than a century or two. Of course, these problems do not mean that radiocarbon dating is worthless or does not work. Far from it. Yet it does remind us that we must respect the method’s limitations. We cannot ask it to do more than it is capable of and must likewise allow for its constraints.

Even so, something still seems to be significantly off in our efforts to date manuscripts from this era using radiocarbon analysis. Presumably there are some underlying errors in the framework that we use to analyze the raw data obtained from these objects. Such inaccuracies in interpretation have been relatively common in the history of radiocarbon dating, and as the discipline has progressed, these have regularly been corrected. Presently we stand at a moment where significant correction is seemingly needed for dating objects from the early medieval Near East. The good news, however, is that scientists who specialize in this kind of analysis are aware of the constant need for such adjustments, and already new avenues are being pursued that will potentially yield more consistent results in seeking to date early Islamic artifacts. Yet even once such refinements are in place, it remains extremely unlikely that radiocarbon dating alone will fully resolve the mystery of the Qur’an’s origins, and historical-critical investigation of the text itself will remain the most vital set of tools for this endeavor.


Over $2 Million Donated to Charity!

We have two goals at Ehrman Blog. One is to increase your knowledge of the New Testament and early Christianity. The other is to raise money for charity! In fact, in 2022, we raised over $360,000 for the charities below.

Become a Member Today!






2023-07-06T11:24:11-04:00July 15th, 2023|Public Forum|

Share Bart’s Post on These Platforms


  1. TomTerrific July 15, 2023 at 9:05 am

    A very interesting article. What is causing the disparities in the data?

  2. GeoffClifton July 15, 2023 at 11:34 am

    I read a book called Catastrophe by David Keys, which argued that there had been a very significant disturbance to the Earth’s climate around the middle of the 6th century AD. It had been suggested that this was either a cometary impact or a volcanic eruption on a grand scale. I wonder whether this might have affected C14 readings for the next couple of centuries (thereby contaminating the Qu’ranic texts Stephen was talking about)?

  3. mini1071 July 15, 2023 at 5:07 pm

    Tried to prove a fraud by radiocarbon dating a ledger in evidence…. Plus or minus 50 years was little help!

  4. OmarRobb July 16, 2023 at 6:57 am

    Hi Stephen,

    This is a tough subject as you know, and I truly think that there are issues in your “analysis”. I did start to write my reply but realized that it covers the size of many comments. Therefore, I decided to put it as a linked pdf to reduce the space of comments.

    I know that the comments in the blog here are not designed for debates, but for quick notes and for clarifying positions, however, I preferred to put this pdf reply here as a part of the “Opinion and the Opposite Opinion” for this sensitive subject.

    The pdf reply link:

    I need to clarify here that this is not an article, it is just a 5 points reply to your post.

    • stevenpirog July 17, 2023 at 9:09 pm

      I read most of this pdf reply- my strongest contention is with your point #2.

      How does dating the Qur’an to a certain time does not “increase” the probability of any other claims regarding the Qur’an? I can see how it can decrease the probability of alternative claims that require different dating. But can you explain how data that are consistent with potentially multiple claims increases the probability of one specific claim that is dependent on other independent evidence?

      Also, is there some criteria by which you label one dating a “statistical anomaly”? It seems the 388-535 dating overlap is just as consistent with the other sets as the 611-660 dating overlap is.

      • OmarRobb July 20, 2023 at 12:48 pm

        # This pdf reply started by a generalization that carbon-dating is not precise, but it can support some conclusions over the others. If this generalization is not accurate at the least then what is the use of it!

        # Conclusions shouldn’t be just claims, but they should be claims supported by some sort of data. The Islamic tradition claims are not based on thin-air but points 3# and 4# of this pdf reply provided some of these data. In point 3#, the ancient Quranic manuscripts can be dated by noting the used diacritics and the style of writing. From this we could conclude that the Sana’a lower-text is the oldest, then the Birmingham, then the Sana’a upper-text. The Birmingham and upper-text can be dated to the middle of the Umayyad reign, while the lower-text is probably much before. This can be concluded even before utilizing the carbon-dating data.

        # Point 4# highlighted the single origin claim for the Quran by noting the quick spread of the Muslims in the east and west then the civil wars later and the fact that all versions of the Quran in the east and west are almost identical.


      • OmarRobb July 20, 2023 at 12:49 pm


        But as we have three different “ancient” documents that are almost identical to the current Quran, then this could also be used to enforce the “single origin” claim.

        # Point 2# in this pdf reply highlighted two issues: All the lab tests do suggest that the skin that is used for the Birmingham manuscript was for an animal that died before 660AD. Doesn’t this support the previous claim related to the dating of this manuscript?!!

        # For any loop of tests that are influenced by assumptions then the test sets might be spread in the chart, or most of the tests are within a reasonable range while some others are far off, etc. For example, a skillful (though not perfect) archer hitting a target for 100 times, then I can expect that most of the hits would be closely around the target, but there might be some few that are far off that target.

        That what I suggested in the second part of point 2#: the last three data tests are within a good proximity while the first test is far off.

        • stevenpirog July 20, 2023 at 4:13 pm

          I really only am focused on claims in point 2, as they are more statistically oriented and that is something I have been trained in and taught myself. In your pdf you state the evidence “increases the probability” of a claim, when technically it supports it. It is a common to get the two conflated but there is a difference between data “supporting” or “increasing the probability” of a claim.

          As for throwing out a set of data because it is off- there are formal processes involved in justifying dismissing collected data, which require more than just saying it’s “off”. The information required to make that justification is not in either this original post by Dr. Shoemaker or your pdf reply. You would likely have to cite data from the statistical analysis of the original papers. You’re free to throw out the data however you want, but doing so is very susceptible to unjustified biases.

          • OmarRobb July 23, 2023 at 1:04 pm

            # Point 2 is not technical: The post implied that these datasets cannot provide useful data as there were skewed. I discussed this matter from a general perspective that these sets are useful as they can provide at least an upper limit. Also, a three sets are within a good proximity and one set is a bit off, therefore, this set could be ignored. This was a general discussion not a technical one.

            # I don’t see a big difference between E1: “an evidence supporting a claim”, and E2: “an evidence increasing the probability-of-accuracy for that claim”.

            E2 is the result of E1: If you managed to support a claim with an evidence (i.e. E1) then this evidence should lead to E2. Therefore, both expressions would lead (in a general discussion) to the same intended meaning.

            Also, to my understanding, “probability-of-accuracy” is not a technical term, it is a description. I think the technical term is the “confidence level”. But I think all would understand the meaning of the “probability-of-accuracy”.

        • stevenpirog July 24, 2023 at 12:17 am

          1) Ignoring data because simply because it is skewed is working with the assumption that skewed data is somehow invalid. That is not the case. Some data, like death rates per age group, are skewed by nature. Some only look skewed. But the data is the data. You do not get to ignore it because you feel it doesn’t match up

          2) Certainly a single piece of evidence E supporting claim C1 “could” by definition increase the probability of claim C1, but only if the alternative is “Not C1”. If evidence supports C1 as equally well as other claims C2, C3, C4, etc, then you really haven’t increased the probability of C1 over other claims. For instance, you mentioned yourself the RC data (the ones you didn’t ignore at least) show the animal used for parchment was likely killed before 660. Why couldn’t the animal have just been killed earlier (rough estimate 610 by my terrible averaging here), with all other details the same? It doesn’t change the overall narrative.

          I really don’t have a problem with the overall conclusions you make. I’m not knowledgeable enough to argue with them. But I do have problems with the demonstrated statistical methodology here.

          • OmarRobb July 26, 2023 at 2:33 pm

            # I totally agree with you but “from a technical perspective”. As I have said, my comment wasn’t technical, but it was from a general perspective. As I have said before: the post implied that these datasets cannot provide useful data as there were skewed, and I highlighted that the off data “can” be ignored. The post was presenting a general (and not “technical”) argument and my argument here wasn’t technical.

            # If we have 3 claims (A, B, C) and we have an evidence that support all of them, then I think it is misleading to say that “this evidence supports A”, as this is just part of the truth. However, this is a communication issue about how to describe things, and it is not a technical issue.

            Nonetheless, my comment was clear in this regard: {So, yes, carbon-dating doesn’t give a pinpoint date, but it does give a probability range, and with it, we can increase the probability-of-accuracy of some conclusions over others}, which I think is valid.

    • sshoema July 21, 2023 at 2:06 pm

      Hi Omar. I took a quick look at your points. My response is: why don’t you just read my book, rather than asking if I address this that or the other thing (I do). It is free!, and I think you would learn a lot.

      • OmarRobb July 22, 2023 at 4:27 am

        Hi Stephen,

        I looked briefly at few chapters, and I think that there are lots of unsupported claims (or at least in the parts that I have looked), and I do regard that point 5# in my pdf reply still hold.

        For example: you didn’t clarify in the review of the traditional narrative (chapter-1) that the committee at the time of Abu-Baker followed a clear criteria for collecting the Quran: the verses of the Quran need to be supported by one written document and two trusted witnesses (at least). This is a fundamental narrative that should have been mentioned.

        Also, you didn’t clarify the traditional view about Othman’s work. I told you in your previous post that Arabia was never united before Islam for at least 8000 years, and every tribe had some variations in their scripting (i.e. the way they wrote the words). Othman just standardized the script. If Othman standardized the pronunciation then he should have sent reciters to the east and west to teach “the official reciting”, but he didn’t, he just sent the official script of the Quran according to the Meccans’ scripting standard.


        • OmarRobb July 22, 2023 at 4:30 am


          You said that Ḥudhayfa reported to Abu-Baker “that significantly divergent versions of the Qur’an were in use in Syria and Iraq”. But this is misleading: there were variations in the way that people were writing the Quran, but the pronunciation was the same. Again: Othman didn’t standardize the Quran; he just standardized the script.

          Also, the issue of the COT (which I did discuss in your previous post) is vital here:

          [Note: COT: Chain-Oral-Tradition. AOT: Anonymous-Oral-Tradition].

          In the critical analysis of the NT, Scholars collected and analyzed the data that were based on AOT and made the criteria to filter the right stuff from it. But you decided to regard the COT to be unreliable, and then, you gave all Islamic narratives (the AOT, the weak COT, and the trusted COT) equal weights, therefore, it is not surprising that you came to the conclusion that: “The Islamic tradition instead reports a tangle of conflicting and disjointed memories about the origins of the Qur’an”. But this is not critical analysis; even in the analysis of the NT, you don’t give equal weights to the narratives, but you differentiate the narratives according to some established criteria.


        • OmarRobb July 22, 2023 at 4:32 am


          For example, Al-Tabarani is a trusted Scholar, but he didn’t do any chain-analyses, he just recorded the COT that he heard. Therefore, you will find lots of contradictions in his collection; because he collected the weak COT with the trusted COT. So, to use a narrative from Al-Tabarani, then we need first to conduct a proper chain-analysis for it.

          But it is unfair and misleading to prove the contradiction in the Islamic tradition by pointing to the Book of Al-Tabarani. This is the same as using Ibn-Ishaq to prove the contradiction in the Islamic tradition, as Ibn-Ishaq is based on AOT.

          So, you didn’t establish a clear criteria and you regarded all COT to be unreliable, and you gave all narratives equal weights. So, we have here a fundamental issue with the way you conducted your research.

          Also, You said that the lower-text of Sana’a was erased because it was nonstandard version, but I doubt it, not with just maximum of 180 different words. But it is clear that the scribe of the lower-text wasn’t professional (due to his style). Isn’t this a valid reason?!

          I think if I dig more deeply, I might find even more issues.

  5. fishician July 16, 2023 at 10:21 am

    I am currently reading through the Qur’an, a modern English translation. I am not surprised that it mentions Biblical figures like Abraham and Moses, but I was surprised when I read a couple of things that sounded like they came from the Infancy Gospel of Thomas and the Proto-gospel of James. Is it likely that Muhammad would have known those gospels? Is it more likely that such details were added by later followers?

    • BDEhrman July 20, 2023 at 4:57 am

      Yup, it’s one of the fascinating features of the Qur’an in relation to Xty. As to the two optoins, my view is: neither. I think stories like these were in circulation among Christians and were known outside of Christian circles as well.

      • sshoema July 21, 2023 at 2:13 pm

        Bart is right. These were widely known among Christians, especially the Protevangelium. I’d wager that Christians were much more familiar, in fact, with the Protevangelium’s content than with Revelation or James or Third John for a LONG time. Muhammad’s new religious movement clearly had a strong connection to Christianity, somehow – there is a lot of “Christian” content in the Qur’an, and even the “Old Testament” material seems to have come from Syriac Christian sources.

        But the question of whether Muhammad knew these writings is a big one. Not only were he and his communities in Mecca and Medina nonliterate according to the most up to date scholarship, but there is no evidence at all for any Christian presence anywhere near either of those two places. So where did it come from? The possbility that many of these traditions were added to the corpus of the Qur’an’s sacred tradition by Muhammad’s later follwers is strong. I write about this in the book.

    • OmarRobb July 22, 2023 at 4:53 am

      Hi Stephen,

      I am not sure if you would regard the following to be logical, but as I have mentioned in point 4# (in the pdf reply), the quick expansion of the Muslims to the east and west, then the civil wars and the defragmentation would all suggest a single origin for the current Quran.

      So, what do you suggest here: do you think that Othman when he standardized the script (about 20 years after the death of the Prophet) heard about the Christian stories and said: these are very good stories, let us add them to the Quran, and all the Muslims said yeah!

      It is so easy to make a claim, but an academic claim needs (at least) to be within a reasonable model, then there should be efforts to try and prove or refute this model.

      But here you are presenting a claim without a reasonable model:

      # How (and when) these narratives have been added to the Quran after Muhammed?

      # What was the Muslim reactions? Did they object/fight or they just accepted it?

      # If these narratives have been added after the quick expansion then how this could fit with the logic presented previously?

      # Etc.

  6. RICHWEN90 July 18, 2023 at 10:26 am

    I wonder whether this might be an issue in manuscript dating: you have the material the manuscript was composed on, written on, but the age of the material might be quite different than the age of the actual writing– is it possible for a parchment sheet, or papyrus, or whatever to be decades older than the date of its use, which would be the actual origin of the manuscript as a composition expressing various ideas, etc.?

  7. moldfield July 23, 2023 at 2:48 am

    I think there may be an error in the text (above):

    “And once we begin to reach for individual decades, the data can NO?? longer be meaningfully interpreted.” I studied dating techniques in grad school, and I agree with the main point of this piece RE specificity of radiocarbon dating.

Leave A Comment