New Article: Quantifying Variants in the Apostolos

In the latest issue of NTS, Greg Lanier has an article expanding on my work estimating the number of variants. He takes a deep dive into the data from the ECM of the Catholic Letters and now Acts to try a way of quantifying the variants in the most important witnesses.

Here is the abstract:

This article interacts with Peter J. Gurry’s recent estimate of the total textual variants in the Greek New Testament (NTS 62 (2016)) by (i) employing a different (and complementary) method using data from the Editio Critica Maior and (ii) producing an estimate that is narrowly confined to the ‘key’ manuscript witnesses for Acts and the Catholic Letters (a mix of majuscules and minuscules, both Byzantine and non-Byzantine). The results prove more useful for framing the development and distribution of textual variants in this group of key witnesses.

I wrote before, “There is more that can be done with these data and hopefully I and others will explore some of those in the future.” While I stumbled over some of the things done with the data here, this is the kind of additional work I had in mind. In particular, I want to highlight the three points that stood out to me.

First, he uses places where there is no variation in the ECM volumes to help calculate manuscript agreement. This is important because the CBGM only uses places where there is some disagreement to calculate the overall agreement. From this, he can say that about 15% of the 12,356 “textual units” in the Apostolos (Acts + Catholic Letters) have no variation in the ECM. With a larger collation set (like Tommy’s work on Jude), this diminishes, of course, but this is valuable especially because CBGM comparisons only compare “textual units” where there is some variation (i.e., variant units). In other words, they leave out all the places where there is no variation in the collated witnesses. (There is a good reason for this in the CBGM, by the way, but that’s another subject.)

Second, Lanier tabulated how many variants are actually found in each variant unit. This is important because it offers one good metric for how much work editors have to do. The reason is that deciding between ten different readings is usually a lot more work than deciding between two. Lanier found that in the Apostolos, about 64% of all variant units have just two readings. Another 27% have three and the remaining 9% have four or more. In other words, the vast majority of decisions in the ECM are between three or fewer variants. Gerd Mink had run the same numbers for James in 2004 with similar results, but now we have them for the rest of the Apostolos. I think this may be one of the most helpful stats in the article

Finally, Lanier plotted the rate of variation per word across the 28 chapters of Acts and found that, among his 16 key witnesses, the rate increases slightly toward the end (see figure). He says, “The data, of course, cannot tell us why this is the case, but it is nevertheless helpful empirical confirmation in the aggregate (for sixteen witnesses) of what textual critics have hypothesised for Acts within individual witnesses” (p. 568). I would suggest that scribal fatigue is the most obvious explanation. As scribes got further into Acts, fatigue set in and they copied a little worse than they did at the start of the book. (This, of course, does not mean that they made no mistakes at the start of books as in, oh, I don’t know, say Mark 1.1 for example.)

Variation rate across Acts

Update: See Lanier’s blog post for a bit on how he crunched the data.

How Many Variants Make It into Your Greek New Testament?

Greek students sometimes get the wrong impression that their Nestle-Aland apparatus records all the variants for the New Testament. I certainly thought this at one point. And it’s not just students. I once heard a story about one of Kurt Aland’s colleagues who called him up in a state of distress because he had found a variant in the Syriac that was not in the NA apparatus! Aland had to break the news to him that this poor variant was not alone. The Nestle apparatus, like most apparatuses is selective.

But just how selective is your Greek New Testament? Since we now have complete collations for several small parts of the NT, and since I’ve done the boring exciting work of counting the variants in them, we can now get a sense of just how selective our standard editions are in their apparatuses.

Drawing on the data available for John 18, Philemon, and Jude, we can see just how many of the total number of variants the UBS4, NA28, and ECM apparatuses include. It’s not a perfect comparison because these editions all include variants from versions and fathers which the exhaustive collations don’t. They also don’t necessarily share the same variant units either. But for giving a general sense of the data, I think these comparisons are helpful.

Results

Here are the results of my comparison:

And in table form with percentages:

Section	Number of Variants (% of total)
	UBS4	NA28	ECM	TOTAL
John 18	10 (0.3%)	183 (6.0%)	–	3,058
Philemon	16 (1.4%)	55 (4.6%)	–	1,185
Jude	47 (2.8%)	145 (8.6%)	789 (46.6%)	1,694

From these numbers we can see just how selective the UBS and NA apparatuses are. In the UBS, we only get 0.3 percent of the total variants in John 18 and it’s still only 2.8 percent in Jude. The NA does better, of course, but it is still just a drop in the bucket of the total number. Somewhat surprisingly, even the ECM with its extensive apparatus comes in below 50 percent.

Signficance

I see at least two conclusions we can draw from these comparisons. The first is that unqualified claims that most of our variants were created early on are simply false. Given that the NA, UBS, and ECM apparatuses all lean heavily toward the earliest evidence, it follows that the bulk of our variants were created (or at least are first attested) later on. In Jude, for example, over 53 percent of variants don’t show up in an edition that represents the text’s first 1,000 years of development (i.e., the ECM). So we cannot blithely claim that most of our variants were created early. It may be true that most of our most interesting or most important variants were created early, but that would take us well beyond the time I have for this blog post.

Second, insofar as these editions succeed in presenting the most significant variants for exegesis and translation (I’m thinking now of the NA and UBS editions), we can conclude that most of our variants are pretty insignificant.

Now that phrase “insofar as” is a pretty important qualifier and I can already hear many astute blog readers protesting that the NA/UBS editions leave out too much of value. I hear your protest. But editors have to draw the line somewhere and I don’t envy them for that. Even knowing that editors leave out some readings we think should be there, my comparisons here still give us at least one concrete way to answer the question “How many variants matter?” If we adjust the question to “How many variants mattered enough that the editors included them in my Greek New Testament?” the answer is less than ten percent.

HT: Greg Lanier for suggesting the idea.

The Types of Singular Errors in the Book of James

I recently finished working through all the singular readings in James according to the Editio Critica Maior (ECM). There are 1,245 of them in my dataset. I picked James because it has the most manuscripts of all the books in the ECM of the Catholic Epistles.

One result from my data is the various types of changes that occur among the singular errors. Here “error” (Fehler) is any reading that didn’t make either logical or grammatical sense to the editors (p. 27*) and “singular” is defined relative to the Greek manuscripts in the ECM. There are 493 such singular errors (39.6% of all singulars).

Using the basic categories from Royse, the singular errors in James divide as follows:

As you can see, the vast majority are substitutions (84.5%). Unexpectedly, additions outnumber omissions almost 2:1 (9.6% vs. 5.0%). This is not true of singular readings as a whole in James where, as in Royse’s study, omissions outnumber additions. You could make an argument that singular errors provide us with greater “moral certainty” (to use Hort’s phrase) about their being scribally created than non-error singulars since nonsense is less likely to be copied than sense. But I find that there is a certain trade off in studying scribal change between the guarantee of purity in the data (“moral certainty”) and the volume of the data. At some point, more data that’s less pure is actually better (i.e., more representative) than less data that’s more pure. But more on that another time.

To give a flavor of some of these errors, here are the first two verses of James.

Error] correct reading	Error #	Address
δου] δουλος	f	1.1/14
τη διαασπορα] τη διασπορα	f1	1.1/26–28
φερη] χαιρειν	f	1.1/30
χαριαν] χαραν	f	1.2/4
πειρασμος] πειρασμοις	f	1.2/14
περιπεση] περιπεσητε	f1	1.2/16
περιπεσητω] περιπεσητε	f2	1.2/16
ποικιλλοι] ποικιλοις	f2	1.2/18

I should also note that I am counting cases of these changes such that if two words are replaced, this counts as one substitution. Likewise, a five word addition would count as one case of a “2+ addition.”

An interesting study would be to consider the non-singular errors alongside the singular errors to see if there are any patterns that might help explain why some were made more than once (assuming they are non-genetic). If anyone is interested, you can download the spreadsheet with all 493 singular errors.

How Many Variants Are There in the Greek New Testament?

(Not this many)

The short answer is no one knows because most of our manuscripts remain uncollated. But this has not stopped scholars from offering numbers since at least the publication of John Mill’s 1707 edition of the Greek New Testament.

That edition was said to have 30,000 variants (that number too is an estimate, by the way). In 1848, J. Scott Porter suggested 100,000 in his Principles of Textual Criticism (p. 11) and the number has been rising ever since. Most recently, Eldon Epp has given a “wild guess” as high as 750,000.

What no one has done—so far as I’m aware—is give a reliable justification for their suggested number of variants. This despite the fact that the number continues to be a matter of genuine apologetic interest—not only to those wanting to defend the NT’s textual reliability but just as much to those wanting to oppose it.

Since a number of major collations have been published in recent years, I thought something could probably be done to set the question on firmer ground. In the latest issue of NTS you’ll find my best attempt at doing that. I won’t spoil it except to say that my results are larger than most other estimates (including Ehrman’s) but still lower than Epp’s “wild guess.”

In order to produce a good estimate, you need three things: a good data source from which to estimate, a method of extrapolating, and a clear definition of what you’re estimating. In the last case, I decided to exclude spelling differences. This was partly because two of my data sources didn’t include very many of them and partly because I just don’t find the number of spelling differences to be all that significant.

Arguably the most important question, however, is how to define “variant.” I don’t think it is always appreciated in our discipline that the term “variant” is necessarily relative. Something can only vary from something else. If you only have one of something, you have no variation. So the question was whether I should define “variant” in relation to the manuscripts or in relation to some fixed, printed text of the New Testament. I decided to go with the former. My estimate is thus an estimate about the number of cases where the manuscripts vary from one another. It is not an estimate of the number of cases where the manuscripts vary from any particular reconstruction of the original text.

What this means, is that my estimate assumes nothing about whether any of the estimated variants are are also original or authorial. But undoubtedly a great many of them are. Naturally, one question I’ve been asked about my estimate is “How many of your estimated variants are original?” The question is actually not hard to answer. If we assume that the original reading has survived in the manuscript tradition at each point of variation, then it’s simply a matter of counting the number of variation units and extrapolating from there. Thankfully, I did count the number of points of variation in my data.

Based on my sources, the number of “original variants” ranges from 17%–25% of the total number of extant variants. The percentage is much lower in the Text und Textwert volumes (only 9%) because the average number of variants per variant unit is much higher there. I suspect this has to do with the way the variant units were chosen in those volumes. But that’s a post for another day.

Besides variants and variant units, I also kept track of the number of singulars and nonsense readings in my data sources. If we add those in, we can graph the results in a way that gives us some perspective on the kind of variation an editor of a large collation must deal with. Obviously, most of these variants don’t (and probably shouldn’t) make their way in to our hand editions. But it’s still helpful to know what these percentages are.



The hard numbers for these charts are all in the article which is now online here. I’ve also put up the pre-pub version (which is basically the same) on my Academia page for those without access to NTS. There is more that can be done with these data and hopefully I and others will explore some of those in the future.

Distribution of New Testament Manuscripts by Century

The Alands have a nice chart showing the distribution of Greek New Testament manuscripts by century in their textual criticism handbook, but the data are almost 30 years old now. Thankfully, it’s possible to update their chart with data from the online Kurzgefasste Liste.

Here is the absolute distribution:

And here is a distribution as a percentage of all the manuscripts for each century:

Of course there are caveats. These are simply based on the Gregory-Aland categories with the problems that attend them. Second, I followed the Alands in listing a manuscript with its upper date range so that a 2nd/3rd century manuscript is listed in the 3rd here. There were just over 300 manuscripts that fall into that category. Lastly, I made sure my own data was clean before I produced my charts, but I don’t make any claims that the numbers are 100% accurate.

Feel free to correct these, use them in your lectures, or frame them for your children’s wall.

Islamic Awareness article

The Islamic Awareness website is a high-quality Muslim apologetics site. It seems to be run by a group of intelligent people, with boundless time and energy (they may tell me otherwise). I have been rather reluctant therefore to start any sort of debate with the bright folk there, knowing that they will have far more time to reply to anything I write than I will have to reply. For anyone who wants to look at the quantity and quality of material produced just check here. One particular article that has been drawn to my attention is the one of the reliability of the NT text. A full response would require an article of equal or perhaps greater length. Time forbids a comprehensive reply, but I wish here briefly to show some of the weaknesses and imbalances in this article by Saifullah et al.

Straw MenIn the opening paragraph the authors state their purpose as being to counter four false arguments for the integrity of the NT text: the first, that the NT text is reliable because a large number of manuscripts survives; the second, that most of the NT text could be reconstructed by quotations from early church Fathers; the third, that the quantitative integrity of the NT text can accurately be described by statistics suggesting integrity of 95-99.9%; the fourth, that some manuscripts of the NT are extremely early.

Of course, we should observe that even if all these arguments were to fall that would not mean that there was no ground for believing the NT to be true. After all, Aquinas, Calvi, Luther, etc., did not use any of these relatively modern arguments. To my knowledge the second argument is not one that is commonly used, and the fourth (claims by C.P. Thiede and J. O’Callaghan that there are first-century NT manuscripts) is not one that is advanced by most Christians with knowledge in the area. The four counter-arguments are presented by Saifullah et al. as if they were replies to mainstream Christian scholars, and yet how many authors can they quote who would use all four arguments that they attribute to Christians?[1]

On the first argument, that the mere number of NT manuscripts suggests that the NT text has been well transmitted, the authors present us with a rather crude version of an argument that appears in some Christian apologetic literature. As they note, the argument goes back at least to F.F. Bruce, but it can hardly be suggested that Bruce thought that mere number was sufficient to indicate reliability.

The third argument, of statistics for textual integrity, is one that is used by Christians, and the authors rightly point out some of its weaknesses, though, as we will see, they suggest replacing it with a far more misleading set of statistics.

StatisticsPercentages of textual certainty are generally misleading, but then so are the percentages of uncertainty suggested by the authors. If you define certainty one way you can get an answer in the high nineties. On the other hand, if you prefer to talk of ‘variant free verses’ as in this article you get a lower percentage. The authors say,

The percentage agreement of the verses when all the four Gospels are considered is 54.5%. This is very close to the probability that a tail (or head) appears when a coin is tossed once (i.e., the probability that a tail or head appears when a coin is tossed is 50%!). It is still a mystery to us from where exactly the evangelicals pick-up [sic] such fantastic ‘agreements’ between the Greek manuscripts of the New Testament.

However, they have purposely chosen the Gospels rather than Epistles and the unit of the verse rather than a more meaningful smaller unit in order to achieve their low statistic. If two adjacent verses contain fifty words and one of them has two words in a varying order then according to Saifullah et al. we have 50% uncertainty of the text, when in fact only 4% of the text is affected! Now I’m willing to admit that there are problems with the statistics of Christian apologists, but it seems to me that they are not wider from the mark than the statistics of this article. Moreover, we must remember here that not every variant makes the text uncertain; not every variant has a reasonable claim to be original. Here they are comparing seven different editions of the NT. Yet sometimes these editions have decided to print texts with minimal or no Greek support. For instance, in Matthew 1:16 von Soden printed a Greek text without the support of any Greek manuscripts. Thus their statistics for variation include variants that were in von Soden’s mind as well as real variants in Greek manuscripts. The variant in Matthew 1:16 arose because von Soden, despite linguistic evidence,[2] began with naturalistic assumptions that sought to explain Jesus’ birth without recourse to a miracle and accordingly chose to print what he thought must be the original reading. Ironically, in the article by Saifullah et al. they typically build on naturalistic assumptions within textual criticism. Likewise the belief that there are ‘primitive errors’ within the NT text (i.e. errors that precede all our manuscripts) is a supposition that is open to challenge rather than a bare fact.

A key thing to realize when making any useful statistics on this issue is that in general the more manuscripts you have the more known variations there will be in the entire collection of manuscripts. Therefore, Saifullah et al. are in the position that the more evidence the Christian finds, the more evidence for NT unreliability Saifullah et al. will believe that they have. If Christians had twice the number of manuscripts they presently have almost certainly there would be more variants than are presently known. Thus, almost in principle, Saifullah et al. are committed to the view that the more evidence there is, the less certain something is. This has the air of making a vice out of a virtue. This is where their argument unravels. Although simple numerical comparison between NT texts and Classical texts cannot assure us of the reliability of the former, NT texts are not only more numerous than Classical texts, they are also generally earlier, more widely cited, and more widely translated. Obviously this produces a wealth of material attesting the NT, but with that wealth comes evidence of variety. Those sceptical of Christianity, whether Muslims, secularists or any other group, can of course latch onto the variety and then make the lazy assumption that it is not possible to apply critical assessment to the variety nor to argue that much of the variation has no claim to be original.

However, it is not only lazy, but it is also ignorant to conclude that every variant in a Greek manuscript has a claim to be original. When one gets familiar with a Greek manuscript one will often find that it makes a recurrent type of error. Once you factor out the errors that are peculiar to that manuscript these variations no longer play a role in identifying the text of the NT. Unfortunately, within the discipline of NT textual criticism it has been common to present textual witnesses, such as manuscripts or translations, prior to undertaking an analysis of what factors within that witness are almost certainly secondary. This is something that I have had reason to criticize in relation to early translations on a number of occasions.[3] Essentially, these critical editions of the Greek NT serve up for us information in an undigested form. They often assume that variation in a translation means that there was variation in the Greek original from which that translation was made. They likewise do not adequately consider the mistakes of individual scribes within the manuscripts. As a consequence many modern editions serve us up with an apparatus full of variations that, to the unsuspecting eye, probably look like evidence that there is a high degree of uncertainty about the biblical text. To the extent that Saifullah and others are building on a false assumption that all variants bring the original into question, their work is inadequate.

The Age of ManuscriptsIt takes some skill to present an argument that makes the fact that the NT exists in earlier manuscripts than the overwhelming majority of Classical texts into a weakness for the NT, but this is what happens in Table VI. A clear strong point of NT transmission (relative to all texts of its age) is argued to be a weakness.

The statement following Table VI that 1 and 2 Timothy and 3 John only have ‘very late manuscripts’ is rather misleading. Here comparison with Classical works is relevant. Scholars have a great deal of confidence in the integrity of the text of Classical works though the manuscripts are generally much later than those of NT works. If 1 and 2 Timothy and 3 John are only in ‘very late manuscripts’ then most Classical works are only in ‘very, very late manuscripts’. The argument presented by Saifullah et al. is that the witnesses are late, and yet their argument is also subject to attrition. One century or two centuries ago the earliest manuscripts were considerably later than the ones we now have. It would be better to say that the manuscripts are early, and are getting earlier all the time. Moreover, their table ignores the fact that sometimes manuscripts in languages other than Greek exist prior to Greek manuscripts. For instance, there is a Coptic manuscript in London, namely Crosby-Schøyen Codex Ms 193, which contains all of 1 Peter, which has been dated to the late second century or early third century by Roberts (whose datings the authors accept elsewhere in this article), though other scholars put this manuscript later in the third century (J. Goehring, The Crosby-Schøyen Codex Ms 193 [CSCO 521, Louvain, 1990]). Thus their entry for 1 Peter might be amended to show this earlier witness. Now the point is this: most scholars would accept that if 1 Peter was being translated into Coptic by this stage then so were the Gospels. Christian literature was very widespread early on and spread across language boundaries. It would not be possible for some conspiracy to exist to change all the Bible manuscripts at any one stage because they were too widespread. What is not observed by Saifullah and friends, and which would need to be worked out at greater length, is that the amount of variation amongst extant NT texts is probably too little for it to be plausible that there was widespread change of these texts early on.

Certainty and Diversity
The article rather lacks precision when dealing with the issue of certainty. The article seems to confuse lack of certainty amongst a range of scholars with lack of certainty of the issue. I doubt whether the authors would accept an argument of the structure: a range of scholars are uncertain about the status of Islam, therefore the status of Islam is uncertain. Yet their argument is very much of the structure that if a range of scholars are uncertain about the NT text therefore the NT text is itself uncertain. The article does not consider that ‘lack of certainty’ is a feature of many areas of scholarship. You could take many an issue from global warming to economic theory and find a range of opinions and therefore declare the issue uncertain. This, however, would not mean that another scholar could not rationally reach a much higher degree of certainty about the text than was available within the academy as a whole. They do not show that an individual scholar could not rationally be 99% or more sure of what the original text was.

Muslims and Christians live in an academic environment where the overwhelming number of scholars in the world’s most distinguished universities do not believe that God has given a record of his words to humans. This is an environment in which a believer in God, whether Christian or Muslim, should begin with an approach which does not uncritically accept that trends in academic scholarship necessarily represent the truth. There must be a rigorous attention to evidence and argument. The article in question could, I think, do more to distinguish between bare fact and the consensus opinion of scholars working in a largely secular academic environment.

In my publications cited, as well as in others found on my webpage, I have sought to document large numbers of textual variants within the Bible which are only believed to have existed due to uncritical presuppositions which have flourished in a more secular age. They do not really exist. These pseudo-variants are open to critique on scholarly and linguistic grounds, though I have found that a sceptical attitude towards secular attitudes to scripture has also been helpful in exposing some of these errors. I suspect, though I cannot prove, that if more scholars were to have a similar scepticism and apply it in their area of study that the text of the NT would appear in an ever better light.

There is obviously plenty to discuss with the Islamic Awareness folk and if the rumour is true that some of them are in Cambridge, England, then perhaps I will get a chance to meet them and chat as I'm usually down there a couple of times a year. I would not even deign to think that these few ramblings are a match in quality for their far more thoroughly researched article.

[1] Clearly the authors are trying to persuade us that this is how many Christians argue. The first argument is said to be used ‘invariably’. The second argument is introduced as ‘The Christian apologists’ second line of defence…’ Assuming the apostrophe to be correctly placed, the authors are asserting that this is generally the line that is taken by Christian apologists.
[2] I refute von Soden’s reconstruction of Matthew 1:16 in P.J. Williams, Early Syriac Translation Technique and the Textual Criticism of the Greek Gospels (Texts and Studies 3:2; Piscataway: Gorgias Press, 2004) pp. 240-44.
[3] For my criticism of this procedure in the textual apparatus see, P.J. Williams, ‘On the Representation of Sahidic within the Apparatus of the Nestle-Aland Novum Testamentum Graece’, Journal of Coptic Studies 8 (2006) 123-25; ‘Bread and the Peshitta in Matthew 16:11-12 and 12:4’, Novum Testamentum 48 (2001) 331-33; ‘Some Problems in Determining the Vorlage of Early Syriac Versions of the NT’, New Testament Studies 47 (2001) 537-43; ‘“According to all” in MT and the Peshitta’, Zeitschrift für Althebraistik 12 (1999) 107-109. I understand that my conclusions will be taken into consideration in future versions of the Nestle-Aland edition.

Number of possible variants

Mark D. Roberts is blogging here on Gospel reliability with particular reference to Misquoting Jesus. In the second section (here) he calculates the approximate number of possible variants in the Gospels and comes up with the figure 64 million. Allowing for the high estimate of 200,000 variants in Gospel manuscripts he comments that this is not very many.

Such use of numbers is something we have reflected on before (here). Statistics have long been in use in estimates by textual critics of the reliability of the NT transmission. Numbers of variants were important to Bentley and percentages to Westcott and Hort. More recently Ehrman has found the number of variants in the NT (200-400 thousand?) to be significant, but from a different perspective Norman Geisler would cite 99.5% accuracy for NT transmission.

One reaction could be to run from all use of numbers, but if we are not to be too dismissive we may at least ask what legitimate ways there are of using them.

Unique variants per manuscript page

In a recent message to Wieland Willker’s textual criticism list (here), Daniel Wallace estimates the number of pages of extant Greek NT manuscript at 2.5 million.

In a Kenneth W. Clark memorial lecture of 1997, Bart Ehrman (here) says:

‘No one knows for sure how many differences there are among our surviving witnesses, simply because no one has yet been able to count them all. The best estimates put the number at around 300,000, but perhaps it’s better to put this figure in comparative terms. There are more differences among our manuscripts than there are words in the NT.’

What is the basis of Ehrman’s calculation? There is no further reference.

If Wallace’s and Ehrman’s estimations are both correct then it strikes me that the overall rate of unique variants per manuscript page is incredibly low. This is only just above one unique variant per 10 ‘pages’ (which I presume in Wallace’s original context, i.e. of photography, means either a single side or the facing sides of an open codex). Obviously the rate of unique variants per page is not the same as the rate of errors since the same error could occur independently in two manuscripts and then only count as one unique variant. Nevertheless, I still find the ratio rather low. Has Ehrman been too conservative?