The Byzantine Priority Hypothesis

Contents: Introduction * Critical Arguments for the Byzantine Text * Critical Arguments against the Byzantine Text * Testing the Byzantine Text * Summary * Addendum


The first printed New Testaments were all primarily Byzantine. Indeed, the Textus Receptus was, for too long, used as the standard for the Byzantine text (and even once it was challenged, it continued to be treated as if identical to the Byzantine text). In the nineteenth century, though, due to the works of scholars such as Lachmann and Hort, that changed. The key element of Hort's theory -- the one part still accepted after the rest was generally abandoned -- was his "proof" of the lateness of the Byzantine text. For most of the century following Hort, the uselessness of the Byzantine text was not only universally accepted, but nearly unquestioned.

In the late twentieth century, that has changed. A group of scholars -- mostly American and mostly conservative evangelicals -- have called for a return to the Byzantine (or at least the Majority) text.

One must be careful in assessing people who prefer the Byzantine text. Most such are not textual critics, and do not engage in textual criticism. Anyone who favours the King James Version or the Textus Receptus, or who claims providential preservation or some kind of divine sanction for a particular text, is not and cannot be a textual critic. It is unfortunate that these non-critics have infected the arguments about the Byzantine text, as their irrational, unreasonable, and uncritical arguments serve only to muddy what should be a reasonable and fruitful debate. It is even more unfortunate that some legitimate critics who support the Byzantine text have accepted their rhetoric. This argument, like all critical arguments, must be decided based on evidence and logic, not faith or claims of what "must" be so. The typical argument is "providential preservation" -- the claim that God must have preserved the original text in all its purity. But as Harry A. Sturz (who is about as sympathetic to the Byzantine text as anyone can be while not being a pure Byzantine-prioritist) notes, "Hills [the leading exponent of this sort of preservation] fails to show why the sovereign God must act in a particular way." [Harry A. Sturz, The Byzantine Text-Type & New Testament Textual Criticism (1984), p. 42. Italics added.] (For more on this subject, see the article on Theology and Textual Criticism.)

But while these non-critics (and non-critical thinkers) make up the majority of those who prefer Byzantine or Byzantine-like texts, they are not the entirety of the Byzantine-priority movement. There are genuine textual scholars who prefer the Byzantine text, and others who, without entirely approving it, would still give it a much greater place than Hort did.

Critical Arguments for the Byzantine Text

The major names in this movement are Harry A. Sturz, (who, in The Byzantine Text-Type & New Testament Textual Criticism, offers the case that the Byzantine type should be considered just as early as the Alexandrian and "Western" types) and the two sets of editors, Zane C. Hodges and Arthur L. Farstad (who published The Greek New Testament According to the Majority Text) and Maurice A. Robinson and William G. Pierpont (who published The New Testament in the Original Greek according to the Byzantine/Majority Textform).

Those who believe in Byzantine Priority on critical grounds usually offer three lines of argument: First, that Hort's proof that the Byzantine text is late is false; second, that the numerical preponderance of the Byzantine text is proof of its fundamental originality, and third, that the readings of the Byzantine text are superior to those of other types (by some standard or other). (Those such as Sturz who argue simply for Byzantine equality obviously pursue only the first line of argument.) Those wishing to see the claims of these authors should consult Sturz or the arguments presented by Pierpont & Robinson (whose introduction, presenting the main arguments of their case, is available here. My thanks to Dr. Robinson for making this material available in electronic form).

The claim that the sheer number of Byzantine manuscripts proves the originality of the type is most easily disposed of, since it is false on its face. This is the Fallacy of Number -- and it is a fallacy. By this argument, the predominant life on earth would be the anaerobic bacteria (now in fact nearly extinct, as they die on contact with oxygen in the air), and the human race would have originated in China or India. It is true that, if nothing intereferes with the transmission process (meaning that all manuscripts produce approximately equal numbers of descendents), then the text found in the majority of manuscripts would likely be the most original text. But there is no reason to think that the transmission process was absolutely smooth -- such things almost never are, in the real world; those who claim that the history of the New Testament text is smooth must present positive proof that it was smooth, rather than making unverifiable and improbable claims. There is, in fact, strong evidence that the course of transmission was not free of interference. The evidence is that different areas developed different local texts (the Alexandrian text in Egypt, the Byzantine in Constantinople and its vicinity, etc.). Of these areas, only Byzantium was still in Christian hands after the tenth century, when the main bulk of manuscripts were produced. Thus, no matter what the original text, we would expect manuscripts which contain the local text of Byzantium (seemingly what we call the Byzantine Text) to be the clear majority of surviving witnesses.

The fact is that replicative processes (which include everything from the breeding of drug-resistant bacteria to the copying of manuscripts) generally do not follow straightforward reproductive paths. One cannot argue from the nature of transmission to the history of the text; the history of the text is too complex and peculiar for that. One can only argue from the history of the text to the nature of transmission (and, in fact, our knowledge of the history of the text is insufficient to allow us to argue in either direction).

If analogies from bacteria don't seem convincing, how about analogies from language? That languages come into existence, evolve, and decay cannot be denied. English exists today; it did not exist two thousand years ago. Latin was common two thousand years ago; today it is a dead language (though still widely known and remembered). These are facts. From this, we can reconstruct the languages from which other languages descended.

English and Latin both go back to proto-Indo-European. This language no longer exists, and, just like the New Testament archetype, must be reconstructed. This is an imprecise process, and the results are not assured. But consider what the argument of number says: It says that the preponderant weight of witnesses is the primary means of determining what is original.

Right now, English is the dominant Indo-European language. Does this mean that Indo-European is closer to English, which has hundreds of millions of native speakers, than to Sanskrit, which is a dead language? Sixteen hundred years ago, when Latin was dominant, was Indo-European more like Latin? We don't know the answer with certainty -- but we know that Indo-European was only one language, and was what it was. Numbers of later speakers don't affect the question.

We can also cite examples of how non-original texts can become dominant. This is more common in with non-Biblical texts, but there is at least one New Testament example: The Byzantine subgroup von Soden labelled Kr. As far as I know, all parties admit that this type is recensional, at least in the sense that it is carefully controlled and deliberately published -- the manuscripts agree very closely, the apparatus is unique, and the text is highly recognizable although definitely Byzantine. This type was created no earlier than the eleventh century. Yet, according to Von Soden, it constitutes the absolute majority of manuscripts copied in the final centuries of the manuscript era (and while this seems to be a slight exaggeration -- very many manuscripts of other types continued to be copied -- the type was certainly more common than any other textual group in late centuries). Had printing not been invented, Kr would almost certainly have become the dominant type. What, then, of a text-type at least seven centuries older than Kr? By all accounts, the Byzantine text was in existence by the fourth century. Certainly it could have become dominant whether original or not -- just as the majority of tuberculosis bacteria are now drug-resistant even though such bacteria were few and far between (if indeed they existed at all) a century ago.

We can offer another analogy from the manuscripts. The vast majority of surviving manuscripts from the third century and earlier are from Egypt. (Based on the table of early manuscripts in Aland and Aland, The Text of the New Testament, 94% of all such ancient manuscripts are Egyptian.) Does this mean that 94% of all early manuscripts which ever existed were written and used in Egypt? Of course not! This is simply another accident of history.

There is also the interesting case of the Peshitta Syriac (at least in the Old Testament). According to Ernst Würthwein, The Text of the Old Testament (translated by Erroll F. Rhodes), p. 87, prior to the ninth century, there was significant diversity in Peshitta manuscripts. But in the tenth century, a single monastery collected every Peshitta manuscript it could lay its hands on -- supposedly collecting 250 copies just in the year 932! This left very few copies of the Peshitta in circulation -- and it appears that all later copies were taken from a single ninth century copy which remained in Syria.

Thus we have many analogies to the descent of New Testament manuscripts: From biology. From linguistics. From manuscripts of secular authors. Even from subgroups of the New Testament tradition. In no case does number mean anything. It may be that the New Testament tradition is unique. But why should it be? God has not made Christianity the dominant world religion. God has not preserved theological purity. God has not given the human race good government. Why should God have done something special with New Testament manuscripts?

Thus, although number certainly is not an argument against the Byzantine text, it is a very feeble argument indeed in its favour. If there is any real evidence against the Byzantine text, it will certainly overcome the evidence of number.

Andrew Lohr suggested another argument on behalf of the Byzantine text, this one geographical/historical:

Consider: where did the originals go? This is sometimes argued, and has to be, book by book. Take I Corinthians, though. I think most agree the original, call it g0 (generation 0), went to Corinth, and most likely (we cannot be certain) stayed there until it wore out. Let's call copies copied from the original g1, copies from a g1 copy g2, and so on. Where are most of the g1 copies likely to be? Near Corinth. When someone takes a g1 copy to a distance, say to Alexandria, the g2 copies where it's taken are likely to be taken from the copy, preserving its idosyncrasies, perhaps with local "corrections" (Alexandria had a tradition of textual criticism.) Maybe a copy that's going to a distance would be made with special care, maybe with a haste that makes errors likely (probably some cases of both; scribes had various individual tendencies.) When the g0 copy wears out, its neighborhood probably has a number of g1 copies that can be corrected from each other. But by the time a remote g1 copy wears out, it will probably have established its deviations in its neighborhood. So the most accurate copies will tend to be in the neighborhood of where the originals were.

And most of the NT originals, the g0s, were probably in the "Byzantine arc" from Jerusalem through Turkey and Greece to Rome. (Old Conybeare and Howson speculated that Hebrews might have gone to Alexandria.) So not necessarily the most accurate particular copies, but the most accurate tendency of text -- average sloppiness around a g0, rather than a set of deviations coming from a g1 or g2 or local editing -- is likely to be found in Byzantine areas.

Much of this is likely enough. (Indeed, I think that, if you grant a couple of assumptions, it's the best argument for Byzantine priority that I've heard.) Certainly it makes sense that the earliest copies would cluster around the archetypes. But there are several drawbacks. One is that we have no actual proof that the Byzantine text is the text from the area of the Byzantine arc. A second is that there are probably two archetypes of, say, 1 Corinthians: The copy sent by Paul to the Corinthians, and the copy he presumably kept -- and while the former would be in Corinth, the latter might be anywhere. A third difficulty is that the book would probably be more often copied in Corinth -- and so, although each individual copy might be better than a copy at an equal "generation depth" elsewhere, the net result might be a worse text simply because of more generations. A final difficulty, applying more to the epistles than the gospels, is that they were collected very early, and we don't know where the collection was made or on what textual basis -- but probably most later copies derive from that, not from better or worse local texts. Lohr's argument is like the argument from number: It has some theoretical validity, but there are too many things which might have gone wrong for it to allow us any certainty.

So any argument for the Byzantine text must lie on other grounds: On the basis of its readings. Can such an argument succeed? Or, to put it another way, do the arguments against the Byzantine text fail?

Critical Arguments against the Byzantine Text

This is where we return to Hort. Despite a century of further research and discoveries, despite a general turning away from Hort's near-absolute acceptance of the Alexandrian text, despite refusal to accept other parts of Hort's theory, his rejection of the Byzantine text is still widely considered final and convincing. What were Hort's arguments, and how well have they stood the test of time?

Hort offered three basic arguments against the Byzantine text (which he called the Syrian text):

Posterity of Syrian (δ) to 'Western' (β) and other (neutral, α) readings shown

(This rather simplifies Hort's list, as he uses other arguments in addition. Not all his arguments, however, are actually directed against the Byzantine text. Hort, e.g., has been accused of using genealogy against the Byzantine text, and it has been argued that this use is improper. If Hort had indeed done so, this would be a valid charge against him -- but Hort did not direct genealogy against the Byzantine text; he directed it against the fallacy of number. For this purpose, his hypothetical use of genealogy is perfectly valid; it's just that it's not an argument against the Byzantine text. It is simply an argument against the methods used by certain pro-Byzantine scholars. So we are left with the three basic arguments against the Byzantine text, which are also the most decisive if valid.)

These arguments are of varying degrees of strength.

The argument based on conflations must be rejected. Hort listed only eight conflations in the Byzantine text -- by no means a sufficient sample to prove his point. And yet, these seem to be the only true instances of the Byzantine text conflating two other readings. (This should come as no surprise; even if one accepts the view that the Byzantine text is a deliberate creation -- and few would still maintain this point -- it still worked primarily by picking and choosing between points of variation, not conflating them.) What's more, we find conflations in many manuscripts. The conflations may be a black mark against the Byzantine text, but they are not proof of anything.

The argument about the age of the Byzantine witnesses has somewhat more validity. The earliest (almost-)purely-Byzantine Greek manuscript of the Gospels is A, of the fifth century; outside the Gospels, we have to turn to Ψ, from the eighth century or later. The earliest Byzantine version, in the Gospels, is the Peshitta Syriac; outside the Gospels, none of the important versions (Latin, Syriac, Coptic, Armenian, Georgian) is Byzantine. Among the Fathers, the earliest to show a Byzantine text (among those who give us enough text to clearly make the determination) is Chrysostom. Thus the direct evidence cannot take the Byzantine text back beyond the fourth century -- particularly as all of these early witnesses (A, Peshitta, Chrysostom) have relatively impure Byzantine texts, displaying an unusually high number of divergences from the textform that came to dominate in the minuscule era.

Byzantine apologists have gone to great lengths to try to explain this away. Sturz, for instance, offers fifteen pages (150 readings) where the Byzantine text opposes Westcott and Hort's text but has early support. This is a rather dubious procedure, based on a weak definition of the Alexandrian text (the fact that Westcott and Hort print a reading does not mean that it is the Alexandrian reading, or that any reading they do not print is non-Alexandrian; in any case, there is good reason to believe that Westcott and Hort did not know of all text-types), and attempts to refute a theory that no one fully accepts any more -- but even if Sturz's lists were entirely accurate, the results mean nothing. It is not enough to prove that individual Byzantine readings are old; it is universally agreed that most Byzantine readings are old. The only way to prove, using the manuscripts, that the Byzantine type is old is to find an old Byzantine manuscript. No one -- not Burgon, not Sturz, not Hodges, not Robinson -- has been able to do this.

This argument, however, is not strong. Arguments from silence never are. The presence of an early Byzantine witness would prove the Byzantine type to be early, but the absence of such a witness proves absolutely nothing. The "Cæsarean" type has no Greek witnesses older than the ninth century, but its antiquity was never questioned (though its existence remains subject to argument). Even the "Western" text cannot display a Greek witness prior to the fifth or sixth century. (It is true that older patristic evidence is claimed for the "Western" text -- though this is less decisive than sometimes claimed, since the text of Codex Bezae does not agree entirely with these witnesses.) It's worth noting that we don't have any early writings from the Byzantine area, where that text might be expected to be found. Thus, the absence of early Byzantine manuscripts proves very little except that the Byzantine text was not universal in early times. If anything, the Byzantine apologists' attempts to explain away the lack of early Byzantine witnesses is a case of "protesting too much"; their argument would look stronger if they didn't try to prove the unprovable.

Still, on this count as on the last, the matter must rest as "Case Unproved."

Thus the final verdict on the Byzantine test must rest upon the matter of internal evidence of lateness. Hort, interestingly, did not attempt to prove this point; he simply stated it, with some handwaving at conflations and the like. Later editors have presented examples of Byzantine readings which the internal evidence clearly convicts of being late -- enough such that the case against the Byzantine text seemed very strong. But all of these were based on isolated instances. We can certainly offer isolated counter-instances. Consider, for instance, the last word of Jesus in Matthew and Mark. Did he say, "ΗΛΙ ΗΛΙ κτλ" or "ΕΛΩΙ ΕΛΩΙ κτλ"? The following table shows the data (we'll ignore the variation in the other words):

 Matthew 27:46Mark 15:34
ΗΛΙ ΗΛΙ A (D E Θ ηλει ηλει) F G K (L αηλι αηλι) W Y Δ Π
1 13 33 565 579 700 892 1424 1582
it am cav ful hub* lich sang
(D Θ 565 ηλει ηλει) 059 131
ΕΛΩΙ ΕΛΩΙ ℵ (B ελωει ελωει)
33 hub** harl val cop
ℵ A B C E F G H K L W Y Δ Ψ
(1 1582 ελωι ελωει) 13 28 579 700 892 1424 it vg

If we rearrange this list by text-types, we see the following:

 Reading in
Reading in
(ℵ B 33 cop)
(A E F G K pm)
(Θ 565)

Thus we see that the Byzantine text, and only the Byzantine text, is free from assimilation in one or the other reading. It doesn't really matter which reading is original; all the text-types except the Byzantine have a conforming reading in one or the other gospel.

Testing the Byzantine Text

Even as isolated instances, the readings mustered against the Majority text are probably enough to make us suspect that the Byzantine type is not the original text, but they are certainly not enough to make us declare it late. What is needed is a detailed test of a particular section of text, listing all differences between the Byzantine and other text-types (ignoring readings of individual manuscripts; also, the Textus Receptus must not be used to represent the Byzantine text). One the divergences are identified, they must be classified based on internal evidence. If the Byzantine text fails the test significantly more often than the other text-types, then and only then can it be judged late.

This is a difficult task to undertake casually. Properly, we need to test the Byzantine text in all five major Biblical sections (Gospels, Acts, Catholics, Paul, Apocalypse), and large enough samples to be meaningful (at least fifteen chapters for the Gospels, ten for Paul, and five for the other sections. Note that it is perfectly possible that the Byzantine text could be late in one corpus and early in another). To do the job well would probably require a doctoral thesis.

We can only offer some small samples. (The apparatus of Hodges & Farstad can be very helpful here in seeking variants, though the manuscript data is clearly inadequate; the apparatus of Nestle, which simply omits many Byzantine variants, is not sufficient.) The list below is taken from Mark, chapter 9. (A chapter chosen because it offers many gospel parallels. This is because assimilation of parallels is one of the few cases where internal evidence is consistently decisive: The harmonized reading is inferior unless the unharmonized reading is the result of clear scribal error.)

Note that this is not a critical apparatus of Mark 9; it lists only places where text-types (appear to) divide. To avoid bias, the Byzantine reading is always listed first, then the Alexandrian, then any others. This is followed by a comment about which is original. Note: Variants found only in the "Western" text are not listed, as there is only one Greek witness to this type and few claim this text as original. I do, however, note "Cæsarean"-only readings.

This is only a twenty verse sample, but it gives us a total of 37 readings. If we examine their nature, we find the following:

Reading TypeNumberPercent
Alexandrian clearly superior38%
Alexandrian marginally superior514%
Byzantine clearly superior25%
Byzantine marginally superior38%
Neither reading superior1027%
Alexandrian and Byzantine texts agree1438%

Given the small size of the sample (only 13 readings where one text shows superiority), we cannot draw any definite conclusions. We must have a larger sample. But in this sample at least, the Byzantine text obviously does not show the sort of massive inferiority implied by Hort. (Indeed, the truly bad text, with an extreme degree of assimilation, appears to be the "Cæsarean" text.)

If by some wild chance the above proportions are indicative, it would appear that the Alexandrian text is slightly better, but the Byzantine could not be considered secondary. It would have to be considered an independent text-type which simply hasn't endured as well as the Alexandrian. But, given the size of the sample, it is quite possible that if we gathered a truly large sample, we might find the Byzantine text equalling or surpassing the Alexandrian.

We should also note the presence of eight readings where the Byzantine text stands alone. This is a strong indication that the Byzantine text is not simply a combination of Alexandrian and Western (or even Alexandrian, Cæsarean, and Western) readings. It is either independent of the other three, or it includes contributions from some other unidentified ("proto-Byzantine"?) text-type.

As an alternative to the above procedure, we might look for variants where one reading is clearly, obviously, and undeniably easier than the other. Examples of this would be readings such as Mark 1:2 (Byz add/Alex omit Ησαια) and James 5:7 (Byz add/Alex omit υετον). Such readings, however, are very rare. (Readings where internal evidence favours a particular reading are not rare, but absolutely decisive cases such as the two listed above are highly unusual.) But not all such readings favour the Alexandrian text; consider 1 Corinthians 13:3, where only the Byzantine reading καυθησωμαι can be said to explain the others (since, if it were original, it would invite the two other readings; if either of the other readings were original, there would be no reason for a variant to arise). That being the case, we must find all such readings, which is probably not practical.

Summary and author's expression of opinion:

When I started this article, I expected the Byzantine text to come off as clearly and significantly inferior to the other text-types. I was wrong. While I believe additional tests are needed, I cannot help but suspect that Hort was in error, and the Byzantine text has independent value. This does not make me a believer in Byzantine priority, but I am tempted toward a "Sturzian" position, in which the Byzantine text becomes one of the constellation of text-types which must be examined to understand a reading.

The basic difficulty, and the reason this issue remains unresolved, is the matter of pattern. It is not sufficient to do as Sturz did and show that some Byzantine readings are early; this does not mean that the type as a whole is early. But it is equally invalid to do as Hort did and claim, because some Byzantine readings are late, that the type as a whole is late. The only way to demonstrate the matter as a whole is to examine the Byzantine text as a whole. One must either subject all the readings in a particular passage to the test, or one must use a statistically significant sample of randomly selected readings. It is not sufficient to use readings which, in some manner, bring themselves forward (e.g. by having the support of a papyrus). It's like taking a political poll by asking all registered Democrats to reveal their presidential preference. It may comfort the candidate (if he's stupid enough), but it really doesn't tell us much.

There seems to be a strong desire among scholars to make textual criticism simple (as opposed to repeatable or mechanical; although these may seem like the same thing, they are not). Hort made TC simple by effectively excluding all text-types but the Alexandrian. The Byzantine prioritists make TC simple by excluding all text-types but the Byzantine. One wishes it could be so -- but there is no reason to believe that TC is simple. If it were simple, we could have reduced it to a machine algorithm by now. But no one has yet succeeded in so doing -- and probably won't until we make some methodological breakthrough.


The above was my opinion as of mid-2002. Since that time, I have become aware of a major project by Wieland Willker which included an attempt to prove the very point described above.

It's somewhat difficult to assess Dr. Willker's work for this purpose, because what he engaged in was a full-fledged textual commentary -- a very useful item, far better than the UBS commentary, as it includes more readings and a more complete assessment of internal and external evidence.

What's more, his assessment at several points appears very cogent, agreeing with much of what I have found. Examples:

Regarding the "Cæsarean" text: The main concern of its editor was to harmonize. This explains the heavy editing in Mk. Unfortunately all witnesses of the group underwent subsequent Byzantine correction to a different degree. We have no pure witness. Θ is the best we have. Full collations of all remotely Caesarean witnesses might be in order to clear up the kinship.

Regarding the "Western" text: Is D a singular idiosyncracy? If "D+it" ever was a Greek texttype is questionable. Do all or most of the Old Latin witnesses go back to one single translation?

Dr. Willker classifies readings according to a scale similar to the above (i.e. Byz or UBS clearly or slightly superior), save that he is more interested in the readings of the UBS edition than those of particular text-types. But he does include an appendix looking at the particular types. The display is graphic rather than tabular, but it appears that the results are roughly as follows:

Percentage of Secondary readings, By Text-Type

Text-Type% Secondary Readings

Hort, obviously, would be thrilled with these results.

I must emphasize that these are not my results, and the material I have from Dr. Willker does not permit me to directly verify the assessments of readings based on internal evidence. I suspect, looking at his commentary, that the data set includes many readings I would not have considered decisive. But we must give him credit: if his results can be verified, and stand up under statistical examination, they would appear to deliver nearly the final blow to the Byzantine text; while the type is not entirely bad, it has little claim to stand on its own.