Text Types And Textual Kinship

Contents: Introduction * History of the Study of Text Types * Recent Efforts * Revelation * The Catholic Epistles * The Pauline Epistles * Acts * The Gospels * The Definition of a Text-Type * The Use of Text-Types in Textual Criticism * Appendix I: The Names and Descriptions of the Various Text-Types * Appendix II: Text-Types and their Witnesses * Appendix III: Von Soden's Textual System * Footnotes


All manuscripts, except autographs, are copied from other manuscripts. This means that some manuscripts are "descendents" of other manuscripts. Others manuscripts, though not descended from one another, are relatives -- both derived from some common ancestor. What's more, some are close relatives; others are distant. In this sense, manuscripts are like people, though they usually have only one parent (the exception is a manuscript which is mixed or block-mixed.) The study of textual kinship tries to make sense of these various relationships. Once this is done, the results can be used to try to trace the history of the text, and from there to seek the original text.

History of the Study of Text-types

The first New Testament textual critic to show interest in textual relationships seems to have been Johann Albrecht Bengel. In his 1725 essay on textual criticism, he notes that manuscripts need to be classified into "companies, families, tribes, [and] nations."[1]

Although all these levels of relationship exist, only two (the "family" and the "nation") have exercised the energy of textual critics to a significant degree.[*2] The highest level, Bengel's "nation," is what we now call a text-type.

Specific attempts to precisely define the term "text-type" will be described below. For now, it is most important to remember the general definition: The Text-Type is the loosest sort of kindred relationship between manuscripts that can be recognized short of the autograph. That is, a text-type consists of manuscripts which display some sort of relationship, but whose kinship is so loose that it cannot possibly be classified or described in detail. We cannot give a precise stemma for the various manuscripts of a text-type, showing all branches and lost intermediate links; at best, we can group them into families and clans.

Once the concept of text-types was firmly established, the obvious next step was to locate them and determine which manuscripts belong to which types. Bengel was the first to make the attempt; he defined the "African" and "Asiatic" text-types. Given the materials he had available, this is fairly impressive; the "Asiatic" type is what we now call Byzantine; the "African" is everything else -- what we would call "pre-Byzantine" (or at least "non-Byzantine"). Bengel not only correctly segregated these types, but he hypothesized that the Asiatic/Byzantine manuscripts, though far more numerous, contained a more recent, inferior text (a view held by most scholars ever since).

Bengel's system was refined by J. S. Semler, then further clarified by J.J. Griesbach. Griesbach's system, with minimal modifications, was followed by Westcott and Hort, and is still accepted by many textual critics today.

Griesbach saw three text-types, which he called "Byzantine," "Alexandrian," and "Western." The Byzantine text consisted of the mass of manuscripts, mostly late; it is generally a full, smooth text (a point usually admitted even by those who consider it superior; they simply believe that the shorter, harsher texts are the result of assorted accidents), and seems to be the type associated with Constantinople and the Byzantine empire. The Western text is largely Latin; it is found primarily in the Old Latin and in a few Greek/Latin diglot uncials (in the Gospels, D/05; in Acts, D/05 plus a few versions such as the margin of the Harklean Syriac; in Paul, D/06, F/010, G/012). The Alexandrian text, which in Griesbach's time was known only in a few witnesses such as L/019 and 33, was held to be the early text of Alexandria, and was already recognized by Griesbach as valuable.

The study of text-types reached a peak in the work of F. J. A. Hort and B. F. Westcott. Their classification was almost the same as Griesbach's; they retained the "Western" text exactly as they found it. The Byzantine text they also accepted, though they called it "Syrian." Their only real departure came in the area of the Alexandrian text.

Griesbach had known only late, badly mixed Alexandrian witnesses. Westcott and Hort had two very nearly pure witnesses available (B/03 and Aleph/01), as well as greater knowledge of the Coptic versions. They felt that Griesbach's Alexandrian text could be divided into two parts: The early part, represented by B+Aleph, and a later part, containing most other non-Western, non-Byzantine manuscripts. They called the early phase of this text "Neutral" (since they felt it to be substantially equivalent to the original text) and the later phase "Alexandrian."

But Hort was not content to look for text-types; he also looked at them. The "Western" text, Hort observed, was expansive and paraphrastic; he held it in very low esteem. (In defense of the "Western" text, it should be observed that Hort's observations were based primarily on Codex Bezae, D/05. This text is indeed very wild -- but there is no real reason to presume it is a representative example of the "Western" text. The "Western" text of Paul, for instance, is much less wild.)

The Byzantine/Syrian text, in Hort's view, is less extreme but also less valuable. It is full of clarifications, harmonizations, and (in Hort's view) conflations. It is also late; he held that the earliest father to show a clearly Byzantine text was Chrysostom (moderns sometimes list Asterius the Sophist as the earliest, but this hardly affects the argument. There are still no early witnesses to the Byzantine text -- though we should note that, if it is indeed the text used in Byzantium, there are no early witnesses surviving to the text used in that region). It was Hort's view that this text was compiled from the other extant types, with deliberate modification as well as comparison.

Hort's "Alexandrian" type was a much more slippery affair, since -- as he himself admitted -- none of the surviving manuscripts contained a pure Alexandrian text. Hort felt that this type is basically similar to the "Neutral" text, with a few "learned" corrections to improve the style. It exists (in a scattered, mixed form) in later manuscripts such as C/05, L/019, and 33.

The prize of the text-types, however, is the "Neutral" text. Represented primarily by B/03, with Aleph/01 as the second witness and some support from mixed manuscripts such as C/05, L/019, T/029, and 33, it represents almost without modification the original text. The text printed by Westcott and Hort is, in almost all instances, the Neutral text (the so-called "Western Non-Interpolations" represent one of the few major exceptions).

In the years since Westcott and Hort, almost all parts of their theory have been assailed. The "Alexandrian" text almost immediately disappeared; the consensus is that the "Neutral" and "Alexandrian" texts are one and the same, with the "Neutral" text being the earlier phase (or, perhaps, just the purer manuscripts of the type). The combined text-type is referred to by Griesbach's name "Alexandrian." (In recent years, however, Kurt and Barbara Aland have spoken of an "Egyptian" text that seems similar to the Westcott/Hort "Alexandrian" text. And it is unquestionably true that there are non-Byzantine readings which occur only in late Alexandrian witnesses. Thus we may well speak of "Egyptian" or late Alexandrian readings. The problem is that there is still no way to draw a line between the Alexandrian and Egyptian texts; they merge continuously into each other.)

The "Western" text has also had defenders, notably A. C. Clark and L. Vaganay. Clark, in particular, attempted to explain the Alexandrian text as an accidental shortening of the "Western" text. Although his observations on textual transmission can be useful (he is correct, for instance, in noting that the most common cause of variation is accidental scribal error), few scholars have accepted the pro-"Western" view.

The age of the text-types has also been questioned. Some -- e.g. the Alands -- hold that there were no text-types before the fourth century.[*3] Eldon J. Epp admits, "There is a continuing and genuine disagreement, if not contention, as to whether or not 'text-types' existed in the earliest centuries...."[4] The answer to this depends, in part, on the definition of text-types (covered below). But one can at least say that many of the text-types have early representatives -- e.g. something very close to the Alexandrian text of the gospels, held by some to be roughly contemporary with B, is found earlier in P66 and P75. The family 1739 text of Paul is close to the text of Origen. Aleph's text of the Apocalypse occurs also in P47. P46 and P72 (as well as the Sahidic version) attest to the B text in Paul and the Catholics respectively. This list could easily be expanded using the Fathers and versions. The vast majority of early manuscripts seem to show kinship with the text-types found in the later ones. This would seem to imply that the text-types are survivals from an earlier era.

Perhaps the greatest controversy, however, rose over the Byzantine text. Even in Hort's time, it had a staunch defender in Burgon. These Byzantine loyalists pointed out -- correctly -- that the conflations in which Hort placed so much confidence are very rare. The defenders of the Byzantine text did not, however, manage to convince scholars that Hort's other arguments were wrong; most still believe that the Byzantine text is full of harmonizations and explications, and that, as a text-type (i.e. a unified collection of readings), its earliest attestation comes from the fourth century.[*5]

Despite all attacks, the Westcott/Hort text and textual theory have remained strongly dominant into the twentieth century. The most important Greek text of this century, the United Bible Society edition (UBS3, UBS4, NA26, NA27), is essentially a Hortian text. (For a demonstration of this point, see the analysis of the text of Colossians. Every major edition since Von Soden, except Vogels, has at least half again as many agreements with WH as with the Byzantine edition of Hodges and Farstad, and in several cases the ratio approaches or even exceeds 2:1.)

Still, the twentieth century has seen some advances in textual theory. The basic goal has been to systematize the study -- to classify all manuscripts, not just a handful of the more important.

The last person to attempt to define text-types across the entire New Testament (assuming that they were the same in all parts) was H. von Soden. Von Soden deserves credit for several advances. First, he attempted to study the entire manuscript tradition. Second, he tried to establish degrees of textual kinship, just as Bengel had suggested nearly two centuries earlier.

Von Soden grouped the manuscripts into three text-types. One of these, the "H" (Hesychian) type, is essentially the same as the traditional Alexandrian/Neutral text. Curiously, von Soden made no attempt to subdivide this text, even though the Alexandrian text is ripe for division.

Von Soden did, however, work hard to subdivide the Byzantine text (which he called "K," for Koine). This was noteworthy; until this time, the Byzantine text had been treated as a monolithic unity (and not distinguished from its corrupt descendent, the Textus Receptus. There are in fact over 1500 places where the Textus Receptus differs from the Majority Text, some of them -- e.g. the placement of the Doxology of Romans -- quite significant).

Although it is not possible to go into von Soden's results in detail here (an outline is found in Appendix III), let alone the minor modifications they were subjected to in the light of the Claremont Profile Method, we can note that he did find a variety of Byzantine groups. The most important of these, in his view, are as follows:
Soden's Group NameModern Name Leading representatives (according to von Soden)
KxKx (no uncials; hundreds of minuscules, mostly obscure; Erasmus's leading manuscript 2e is Kx)
KrKr(no uncials; no early minuscules; though there are hundreds of Kr manuscripts overall, only a relative handful of those known to Tischendorf, including 18, 35, 55, 66, 83, 128, 141, 147, 155, 167, 170, 189, 201, etc. belong to this group)
K1(Kx Cluster W) S V W
Ki(Kx Cluster W) E F G H
Ik (also Ka)Family P (A) K P Y

There are, of course, many other non-"Western" non-Alexandrian manuscripts and groupings, most of which Von Soden listed as "I" even though they are clearly primarily Byzantine; the student who wishes more information is referred to the work of Wisse on the Claremont Profile Method.

Outside of the Gospels, many of these groups disappear (or at least cannot be recognized). Kr, however, endures, and a new group, Kc, appears.

Von Soden's work on the Byzantine text has generally been accepted (often for lack of an alternative; no one wants to have to re-do his work). Some parts have been directly confirmed (e.g. Voss verified the existence of Kr, and various scholars studied Family P).

The most thorough study, however, has been that of Wisse and McReynolds, based on the already-mentioned Claremont Profile Method. They generally confirmed Von Soden's groups (though making many detailed modifications). However, Von Soden's Kx, Ki, and K1 may be too similar to be disinguished. [*6]

The chart below shows the frequency of occurrence of the basic types of the text, based on the evaluations of Wisse. The types shown are:

It might be noted in passing that the Textus Receptus belongs to none of these groups. It is Byzantine, but of no particular type (the base text, that of 2, is largely Kx in the gospels, but the influence of 1, of the vulgate, and of other texts has caused the TR to diverge from all these groups). This confirms Colwell's urgent entreaty (made also by Zuntz) that manuscript classification not be based on divergences from the Textus Receptus. But to return to Von Soden....

For all his work on the Byzantine text, though, von Soden's pride and joy was his "I" (Jerusalem) text-type. The "I" text, which von Soden discovered ("invented" might be a better word) was rather like the "Western" text on steroids. It included, naturally, all the "Western" witnesses (such as they are). It included what would later be called "Cæsarean" witnesses (e.g. Q/038, family 1, family 13, 28, 565, 700). In Paul, it included a number of witnesses that are actually mostly Alexandrian (e.g. family 2127). And it included many texts that are almost purely Byzantine (e.g. N/022, U/030). (For details on von Soden's system, with comments on most of his individual groups, see Appendix III: Von Soden's Textual System.)

Von Soden felt that his three text-types, I, H, and K, all went back to the original, and that their common ancestor was the original text. He therefore reconstructed a text that, with some exceptions (where he believed there were corruptions either caused by K or within K), followed the readings of two of the three text-types. Since he placed a much higher value on K than did Westcott and Hort, his resultant text was much more Byzantine than theirs.

Later scholars were not impressed with Von Soden's efforts. To begin with, it has been all but universally agreed that the "I" text does not exist. This obviously removes one prop from his proposed I-H-K text. In addition, with a few exceptions such as Sturz,[*10] scholars will not accept his contention that "H" and "K" are contemporary. Most scholars accept the Hortian view that the Alexandrian text-type predates the Byzantine; a few feel the reverse. And both camps agree that von Soden's use of the two was inaccurate and unacceptable.

Recent Efforts

Since von Soden's time, the emphasis has been on classifying the text-types of individual portions of the Bible. This "local" study has been much more fruitful, and has resulted in many modifications to the Westcott-Hort scheme of three basic (and undifferentiated) text-types.

Before proceeding to these recent studies, however, we should perhaps dispose of the work of Kurt and Barbara Aland.[11] The Alands have two rating systems, one for early manuscripts and one for late. Early manuscripts (from before the fourth century) are classified as "strict," "normal," or "free." Although this is on its face a rating of the degree of care practiced by the scribe, in effect it becomes a value judgment on the quality of the manuscript. Worse, the Alands apply this system to even such short fragments as P52, which are simply too small to classify. Of the early papyri, only the "big six" (P45 P46 P47 P66 P72 P75), plus perhaps P13, are extensive enough to analyse fully. (P74 is also extensive enough to classify, but is not an early papyrus; it dates from the seventh century.)

For later manuscripts, the Alands place manuscripts in "Categories" I-V. These categories are based solely on the Byzantine content of manuscripts, and are not objectively controlled. (Example: 0243 and 1739 are very close cousins, perhaps even sisters. But 1739 is "Category I" and 0243 is "Category II"). What is more, the Alands have a strong bias toward their own text. In addition, "Category IV" consists solely of Codex Bezae and a few fragments!

The Alands' classifications have some value; Category V manuscripts are Byzantine, and those in the other categories are something else. Category I manuscripts have texts which are entirely non-Byzantine (and largely Alexandrian); Categories II and III are mixed, and may belong to any text-type. But as an assesment of the type of text, as opposed to its fidelity to the Alexandrian and Byzantine groups, the Aland categories are useless.

Fortunately, most critics have sought more readily applicable results. Some of their findings are summarized below:


In the Apocalypse, the defining work has been that of Josef Schmid.[12] Schmid partly accepted the Hortian view that only two text-types (Alexandrian and Byzantine) have been preserved for this book. However, both groups must be subdivided. What had been called the Alexandrian text in fact includes two types. The best group is represented by A/02, C/04, the vulgate, and a handful of later minuscules such as 2053; this probably ought to be labelled the "Alexandrian" text. Distinctly inferior, despite its earlier attestation, is the group which contains Aleph/01 and P47. The Byzantine text falls into the "strict" Byzantine group (what the Nestle-Aland text calls MK, of which the earliest full representative is 046; this is the largest grouping, and has several subgroups) and the text found in Andreas of Caesarea's commentary (MA, representing perhaps a third of the total manuscripts, starting with P/025 and including 1r, the manuscript on which the Textus Receptus is based).

The Catholic Epistles

Perhaps the best work of all has been done on the Catholic Epistles. Here the dominant names are those of W. L. Richards,[*13] Jean Duplacy, and Christian-Bernard Amphoux.[14] All of these studies are slightly imperfect (Richards, in particular, is plagued by inaccurate collations and foolish assumptions), but between them they provide a diverse analysis. I would summarize their results as follows (with some amplification of my own): There are four text-types in the Catholics. They are (in order of their earliest known witnesses) the Alexandrian text, family 1739, family 2138, and the Byzantine text.

The Alexandrian text, as usual, consists of B/03, Aleph/01, and their followers. It appears to have several subgroups. The earliest of these consists of P72 and B, possibly supported by the Sahidic Coptic (it is possible that this group should be considered a separate text-type; the small amount of text preserved by P72 makes this difficult to verify). Next comes Aleph, which stands alone. Then comes a large group headed by A/02 and 33. Other key members of this group are 436 and the Bohairic Coptic. Most later Alexandrian manuscripts (e.g. Y/044 and 81) seem to derive from this text, although most have suffered Byzantine mixture.

Family 1739 falls into three subgroups. The oldest witness to the group, C/04, stands perhaps closer to the Alexandrian text than the others (It may be block-mixed; Richards regards it as Alexandrian, Amphoux as closer to 1739, and my numbers put it in between but leaning toward 1739. Stephen C. Carlson separates it from both groups but places it very close to the original, which would also explain the what we see). The next witness, 1739, is perhaps also the best; certainly it is the central witness. A number of manuscripts cluster around it, among which 323, 424c, 945, 1881, and 2298 are noteworthy. Finally, there is 1241 (and possibly 1243), which preserve the same general sort of text but which stand apart (perhaps as a result of casual copying; 1241 is a poorly-written, rather wild text). Amphoux views this family as "Cæsarean," and certainly it is close to Origen. In the author's opinion, its value is at least equal to the pure Alexandrian text. (It should be noted that my terminology here is rather poor. I have used "family 1739" to refer both to the smaller manuscript family which contains 1739, 323, 945, etc., and to the larger text-type which also contains C/04 and 1241. This shows our need for clearer terminology; perhaps we should refer to "family 1739" and "group 1739.")

Family 2138 also falls into several subgroups (e.g. 2138+1611, 2412+614, 1505+2495, 630+2200+206+429+522+1799). In general, however, these subgroups merely represent different sets of Byzantine corruptions. The oldest (though hardly the best) witness to this text-type is the Harklean Syriac; the earliest Greek witness is 2138 (dated 1072). Other witnesses include -- but are probably not limited to -- 206 429 522 614 630 1505 1518 1611 1799 2138 2412 2495. As it stands, this text-type has been heavily influenced by the Byzantine text; it is not clear whether this influence was present from the start. Amphoux considers it to be the remnants of the "Western" text; it should be noted, however, that it bears little similarity to the surviving Latin witnesses. The group bears certain "historical" links to the 1739 group (there are surprisingly many witnesses which show the 2138 type in the Acts or Catholics but go with 1739 elsewhere); Carlson thinks this may also be genealogical.

The fourth textual grouping is, of course, the Byzantine text. It has the usual subgroups, none of them being of particular note. It is interesting that, although we see Byzantine influence in the Syriac versions, the earliest purely Byzantine witnesses in the Catholics are the ninth century uncials K/018, L/020, and 049.

The Pauline Epistles

The Pauline Epistles also have a complex textual situation. Here, in particular, the classical system of Alexandrian/Byzantine/(Cæsarean)/"Western" breaks down.

In Paul, the great name is that of Zuntz,[15] who deserves credit as the first scholar to treat the papyri with real respect. Earlier experts had tried to fit the papyri into existing textual theory. Zuntz chose to start from the papyri. Focusing on 1 Corinthians and Hebrews, he discovered an affinity between P46 and B/03. (In fact this affinity extends throughout Paul, although P46 has a rather wild text in Romans.) Instead of two non-Byzantine texts of Paul (Alexandrian and "Western"), there were three: the Alexandrian, found in Aleph/01, A/02, C/04, 33, etc.; the "Western," in D/06, F/010, G/012, and the Latin versions; and the new text, which Zuntz called "proto-Alexandrian," found in P46, B, 1739, and the Coptic versions.[*16]

Sadly, later critics have paid little attention to Zuntz's classifications. They neither seek to refine them nor to use them in criticism.

It is the author's opinion that even Zuntz's classification leaves something to be desired. (Zuntz's method was centered wholly around P46, especially about its agreements. This is a commendable procedure in that it focuses on the manuscript itself, but by ignoring P46's disagreements and their nature, Zuntz was unable to see the full scope of the tradition. Witnessing a continuum from P46 to 1739 to Aleph to A, he assumed that this was a historical continuum; in fact it is genetic. A proper comparison must start by looking at all manuscripts.) First, the P46/B text, although it clearly comes from Egypt, is not the forerunner of the main Alexandrian text; it is a distinct text which simply shares many Alexandrian readings. Second, the Bohairic Coptic goes with Aleph/A/C/33, not P46/B/sa. And finally, 1739 and its relatives, although akin to P46/B, form a text-type in their own right, which in fact stands between the other three, having many readings in common with all three other early text-types. (Or so it appears; the difficulty is literally that the manuscripts of the 1739 type are, except for Byzantine mixture, so close together. They almost certainly derive from an Archetype not many generations prior to 1739. This family, plus Origen, form the 1739 type. The problem is that one family, plus one Father, make a very thin text-type, as do P46 and B....)

To summarize: In addition to the Byzantine text, there are four early text-types in Paul: P46/B/sahidic, the traditional "Alexandrian" text (Aleph/A/C/33/bohairic; later and inferior forms of this text are found in 81, 442, 1175, family 2127 (=256 365 1319 2127 etc.), and several dozen other manuscripts); the "Western" text (D/F/G/Old Latin); and family 1739 (1739, 0243/0121b, 0121a, 6, 424c, 630 (in part), 1881, etc.; this family is particularly close to the text of Origen). In addition, two families exist with more heavily Byzantine but seemingly independent texts: family 330 (330, 451, 2492) and family 1611 (the remnants of family 2138 of the Catholics: 1505 1611 2495 Harklean; 1022 in the Pastorals and Hebrews; also probably 2005. This family is much more Byzantine in Paul than in the Catholics). These latter two groups may be the remnants of earlier text-types.


Textual theory in the Acts has not advanced much since Hort. The two basic groups are still the Alexandrian (P74, Aleph/01, A/02, B/03, 33, 81, 1175, cop) and the "Western" (D, Old Latin, joined in part by the margin of the Harklean Syriac and some other versions, as well as by a handful of minuscules). It is interesting to note that, in the Acts as in the Catholics, there is a significant gap between B and A (with most of the later Alexandrian manuscripts orbiting about the latter and P74). Aleph stands between B and A; if it did not exist, there might be greater questions about the unity of the Alexandrian text. P45 possesses an independent text, but is too fragmentary to tell us much. The great questions revolve about the minuscule families, of which there are at least three important ones. The best-known of these is Family 2138 (which in Acts might best be called Family 614 after its best-known member). Its relationship to the "Western" text is widely assumed but needs to be examined. Family 1739, well-known from the epistles, exists and includes 1739, 323, 630, 1891, etc., but the basic study of the group, by Geer, simply verifies the existence of the type without in offering a useful analysis of its nature. It appears that it is somewhat weaker and much more Byzantine in Acts than the other epistles, and does not add much to our knowledge. (The theory that it is "Western" is, however, dubious; it agrees with B far more often than with D.) In addition, there is a third family, which we might call Family 36; this includes among others 36, 307, 453, 610, 1678 -- all commentary manuscripts, listed by Von Soden as being of the Andreas type and listed as Ia1. This family is rather more Byzantine than family 1739, but Geer tentatively links one of its leading members (453) to Family 1739. This point perhaps needs to be investigated more fully. Several groups are now studying the text of Acts; one may hope that they will soon be able to offer results.

The Gospels

If labours in the rest of the New Testament has been fruitful, the gospels seem to continue to resist progress. Years of work on the "Western" text have produced a number of theories but no general consensus.

The chief problem is that, after years of looking, Codex Bezae (D/05) remains the only Greek witness to the "Western" text. (P5 and 0171 have been offered as other examples of "Western" texts; this is certainly possible, since both have rather "wild" texts, but both are fragmentary, and neither is particularly close to D.) In addition, D shows signs of editing (especially in the gospel of Luke. The most obvious example is Luke's genealogy of Jesus in Luke 3:23-38, where D offers a modified form of Matthew's genealogy. D also has a very high number of singular readings, many of which have no support even among the Old Latins; these too may be the result of editing). This has led Kurt Aland to propose that the "Western" text is not a legitimate text-type. (In answer, one might point to the large number of Latin witnesses that attest to "Western" readings. In the author's opinion, the "Western" text exists. We merely should use the Latin texts, rather than D, as the basis for reconstructing it.) Others have sought to break off the Old Syriac witnesses, placing them in their own "Syriac" text-type. This is reasonable, but can hardly be considered certain until we have more witnesses to the type, preferably in Greek. Colwell's balanced conclusion is as follows: "The so-called Western... text-type is the uncontrolled, popular text of the second century. It has no unity and should not be referred to as the 'Western Text.'"[17]

But there can be no better illustration of the problems of gospel criticism than the history of the "Cæsarean" text.

The history of this text begins with Kirsopp Lake, who opened the twentieth century by announcing the existence of the textual family that bears his name (family 1, the "Lake Group"). In the following years he and his colleagues Blake and New discovered that this group could be associated with a number of other manuscripts (notably Q/038, family 13, 565, and 700). Then B.H. Streeter proposed that this group was a new text-type.[18] Since it seemed to be associated with those works of Origen written while he was in Cæsarea, Streeter dubbed the group "Cæsarean."

The problem with this text was its definition. Streeter, Lake, and their colleagues functionally defined the Cæsarean text as "any reading not found in the Textus Receptus. and supported by two or more 'Cæsarean' witnesses." Apart from its circularity, which is perhaps inevitable (and which could be controlled by proper statistical methods), this definition suffers severely by being dependent on the Textus Receptus, which simply is not a representative Byzantine text. Using it, Streeter was able to find vast numbers of "Cæsarean" witnesses (e.g. family P) that are in reality ordinary Byzantine witnesses that happen to belong to families rather remote from the Textus Receptus. Indeed, many of Streeter's "Cæsarean" readings are in fact purely Byzantine!

The real difficulty with the Cæsarean text, however, was the lack of a pure representative. Even the best witnesses to the text, Q/038, family 1, and the Armenian and Georgian versions, have suffered significant Byzantine mixture; it appears that only about half of their pre-Byzantine readings survive. (And, it need hardly be added, each manuscript has a different pattern of mixture, making their rates of agreement rather low.)

By the middle of the century, the Cæsarean text was already coming under attack. Hurtado applied what might appear to be the coup de gras in his 1973 thesis.[19] Hurtado showed, fairly conclusively, that the connection that Streeter and Kenyon had postulated between P45 and W/032 (the "pre-Cæsarean" witnesses) and the bulk of the "Cæsarean" text did not exist.

Hurtado's study, based on all variants in Mark found in Aleph/01, A/02, B/03, D/05, W/032, Q/038, family 13, 565, and the Textus Receptus, was interpreted as dissolving the "Cæsarean" text. In fact it did nothing of the kind. Streeter and Lake defined the text only in the non-Byzantine readings of the witnesses, but Hurtado looked at all readings. Thus Hurtado did not even address Streeter's definition of the text-type. And Streeter did have some basis for his opinions; there are many special readings shared by the so-called "Cæsarean" witnesses. (An obvious example is the reading Ihsoun (ton) Barabban in Matt. 27:16-17. This reading is found only in a subset of the "Cæsarean" witnesses: Q f1 700* arm geo2.) On the other hand, as is shown in the section Testing the Byzantine Text in the article on the Byzantine Priority Hypothesis, many other "Cæsarean" readings appear in fact to be harmonization. Thus both the case for and the case against the "Cæsarean" text leads to difficulties.

Which forces us, at last, to wrestle with a fundamental question: "What is a text-type?" Our answer to this has important implications -- and not just for the "Cæsarean" text. For example, we have already noted that B/03 and Aleph/01 have different text-types in Paul. There are hints that they differ in the Catholics as well. What about in the Gospels? It can be shown that both manuscripts are part of tighter families within the Alexandrian text (B is closely related to P75, T/029, L/019, and the Sahidic Coptic; Aleph goes with Z/035, probably the Bohairic Coptic, and certain of the mixed minuscules). Are these text-types, or merely clans within a text-type?[*20] And, whatever the answer, how can we use this information? These are among the great questions textual critics need to face.

The Definition of a Text-type

An analogy may help here: Think of the text as a crystal and text-types as its facets. If the crystal is subjected to pressure, it will usually separate along the lines of the facets. The behavior of the text is similar: if a text is subjected to the "pressure" of a variant reading, it will tend to break along the lines of text-types. This does not mean that it will always separate at all the facets, nor that all facets are equally likely break-points.[*21] But while this analogy describes the situation fairly well in general terms, we must have more precision..

Westcott and Hort, although they made extensive use of text-types, did not offer a clear definition. Most of their references are to "genealogy,"[*22] which is misleading, since it is rarely possible to determine the exact relationship between manuscripts.[*23] (Even such similar manuscripts as P75 and B are no closer than uncle and nephew, and are more likely cousins at several removes.) Similarly, B.H. Streeter describes "local texts" at length, but at no point offers a useful definition. Most of the standard manuals are no better. No wonder that, even today, many scholars will say that they "know a text-type when [they] see it."

The first attempt to create an automatic method for determining text-type was probably Hutton's "triple readings," proposed in 1911 in An Atlas of Textual Criticism. Hutton proposed to look at those readings where the Alexandrian, Byzantine, and "Western" texts all had distinct readings. This would allow a newly-discovered manuscript to be quickly classified.

This method had two problems. First, it assumed the solution: Only three text-types were permitted, and the readings of those three were assumed to be already known. Second, even if one felt assured of the method, triple readings were too rare to be much help. Hutton had only about three hundred triple readings in the entire New Testament. This meant that there were no more than a few dozen in any given book. Comparison at a few dozen points of variation is simply not enough to produce assured results.

It was not until the mid-Twentieth century that E.C. Colwell offered the first balanced definition of a text-type.[*24] In one essay he gave a qualitative definition ("A Text-type is the largest group of sources which can be generally identified").[25] He adds the important qualification, "This definition is a definition of a text-type as a group of manuscripts [italics mine], not... a list of readings." Five years later, in an influential essay, Colwell went further. He attempted a quantitative definition. (Indeed, his method is frequently called the "quantitative method" -- a name that makes me cringe, since any statistical method is a "quantitative method.") His statement on the subject is perhaps the most-quoted statement on genealogy since Hort's time:

"This suggest that the quantitative definition of a text-type is a group of manuscripts that agree more than 70 per cent of the time and is separated by a gap of about ten percent from its neighbors."[26]

Colwell deserves immense credit for offering this definition (as well as for his other methodological studies; he is perhaps the greatest worker in this field in the twentieth century). This definition has the advantages of being clear, precise, and usable. Unfortunately, in the author's experience, it does not work. (It strikes me as almost tragic that Colwell's most-frequently-cited comment on text-types is also one of the few that is not entirely correct. It's worth noting that he rarely if ever refers back to this criterion.)

There are two reasons why the Colwell-Tune definition is imperfect. First, the percentage of agreements between manuscripts is entirely dependent on the sample. Second, the "gap" which Colwell refers to disappears when working with mixed manuscripts. Let us offer examples.[*27]

To take the first point first, consider the relationship between B/03 and Aleph/01 in chapter 2 of Colossians. The two manuscripts agree in only two of the seven variations cited in GNT4, or 29%. If we take the 29 variants cited in NA27 (excluding conjectures), we find that they agree in 18 of 29, or 62%. If we turn to the Munster Institute's New Testament "Auf Papyrus," and examine the variants supported by two or more uncials (excluding orthographic variants), we find that the two agree in 32 of 47, or 68%. But if we turn to the editia minor of Tischendorf8, we find agreement in 19 of 32 non-orthographic variants, or 59%. Even if we throw out the small GNT sample, we still have almost a ten percent variation between the three remaining sample sets, all of which form large and reasonable bases for comparison. Which one should we use in deciding whether B and Aleph belong to the same text-type? The 68% number, which places them on the fringe of qualifying? The 59% number, which isn't even close? Or something else?

All told, Aleph and B have 25 disagreements in this chapter (though some are scribal errors, usually in Aleph). How do we decide how many variants to spread these 25 differences out over to determine if there is 70% agreement?

A thought-experiment about mixed minuscules should be sufficient to demonstrate the non-existence of the "gap." Suppose X is an unmixed manuscript, Y is copied from X with five percent Byzantine admixture, Z is copied from Y with another 5% admixture, and so on. It follows that X can never have a ten percent gap; that space is occupied by Y, Z, and so on down the line. If that is not proof enough, one can present a concrete example based on B in the Gospels. Using a large (990 reading) sample and 39 Greek manuscripts, I found two documents (P75 and 2427, which we have since learned is a modern production heavily influenced by B) which, in their particular areas, agreed with B over 80% of the time. Below this was a gap -- but most manuscripts that are considered to belong with B (including Aleph, L, 33, and 892) are on the far side of the gap![*28] The next-closest manuscript was Xi/040 in Luke, at 68%. From there down to the final manuscript in the list (D/05, with 30% overall agreement), there was no gap larger than eight percentage points (and even this gap would have been filled had I included the Coptic versions). The median gap among non-Byzantine manuscripts was one, and even the arithmetic mean ("average") was under two. Colwell's "gaps" will simply not exist in large manuscript samples.

There is also a problem with the conceptual model of the Colwell system. Take a manuscript like L/019 of the gospels. It has a significant Byzantine component -- large enough that it will likely fail Colwell's 70% criterion for agreement with the pure Alexandrian witnesses. But -- where it is non-Byzantine -- it stands very close to B/03, and is one of the closest allies of that manuscript. Should we not be able to recognize L as a degenerate relative of B, and use it on that basis?

Some would propose to address the problem by adjusting the numbers -- e.g. by allowing a 60% instead of a 70% threshold. This may work in some cases, but cannot be guaranteed; any statistic will be dependent on its sample. It is possible that we could assign percentages if we could produce a "representative" list of variants -- but what is a "representative" variant reading?

Some such as Zuntz[*29] and Wisse[*30] are ready to throw the whole thing over and abandon statistics altogether. This is perhaps premature, but we definitely need to tighten up our methods.

Colwell's failure again leaves us seeking informal definitions. In 1995, Eldon Jay Epp offered this "working" definition: "A text-type may be defined as an established textual cluster or constellation of MSS with a distinctive textual character or complexion that differentiates it from other textual constellations."[31] He adds, "Such differentiations must not be based on general impressions or random samples but on full quantitative comparison...."[32] Unfortunately, Epp has little to add from there; he goes on to work with the Colwell definition. (Though he soon after admits that manuscripts are like a scattered "galaxy" or a "spectrum," thus implicitly denying the existence of the gap.[33]) Also, there is (at present) no hope of fully collating all New Testament manuscripts; we must work with samples.

Maurice Robinson, in private correspondence, has offered what is probably the best available informal definition: "[A text-type seems to be found in] a shared pattern of readings held in common in a significant degree by member MSS to the exclusion of the presence of competing patterns in a proportionally significant quantity." This is the sort of definition we need -- but it can be made useful only by supplying a definition of "pattern" and a way of determining a "proportionally significant quantity."

A different approach, attempting just this, and also arising from Colwell, is the Claremont Profile Method. The CPM attempts to determine textual affinities by looking at a "profile" of readings in selected chapters.[34]

The CPM offers distinct advantages. It allows manuscripts to be quickly and easily measured against known groups. Its defect is that it has no ability to define groups (it finds groups, but no definition is offered of what constitutes a group; if Wisse had not started from Von Soden, his results might have been completely different), and no way of measuring mixture. (The notorious example of this is that, in Luke, D/05 shows a profile that makes it a member of the Alexandrian text!) The reason for this is not hard to find: the CPM (commonly, but imprecisely, referred to as the "Profile Method"; this name should be avoided, as there are many other profile methods possible) takes a manuscript, finds its readings in a "profile" of selected passages, and looks for a match in its store of profiles. If it finds one, it is done. But if it fails to find one, it is also done, and writes off the manuscript as "mixed." No attempt is made, if the manuscript is mixed, to determine what the mixture is.[*35]

The result sees scholars still flailing around, trying new methodological tricks. For example, more and more scholars are classifying by pericopes -- that is, taking a particular incident and collating it.[36] If used properly, this has real advantages. Unlike the Aland system, it allows us potentially to check for block mixture, because it gives us detailed data at several points. It is faster than collating to the Aland readings, since there is no need to search for this reading, then that. It covers more ground than the CPM's chapters. It also (again, potentially) gives us enough data to work with, assuming we choose long enough pericopes (say, a dozen or more verses, with at least forty variant readings) and enough pericopes (say, one every three chapters). But these latter cautions are very important (the collator still needs to check a sufficient number of variants!), and this requirement is often ignored.

At this point, we would appear to have reached an impasse. Some scholars, such as Hurtado, swear by the Colwell-Tune definition. Others, such as Richards, find flaws but produce nothing better. Yet others, such as Wisse, move down to such a level of detail that they not only can't see the forest for the trees, they can't even see the trees for the blades of grass between them.

The problem ultimately goes back to a failure of terminology. It was Stephen C. Carlson who seems to have had the key insight: That a genetic text-type is not automatically a quantitative text-type. That is, two witnesses descended from a common ancestor may not have a high rate of agreements due to mixture, while two manuscripts which are not the direct descendents of a common ancestor may have a high rate of agreement due to mixture.

Take a specific example: Dabs1. This is a direct copy of D/06. D/06 is unquestionably "Western." So, genetically, Dabs1 is obviously "Western."

But before Dabs1 was copied, D/06 was heavily corrected toward the Byzantine text. So heavily, in fact, that most assessments of Dabs (based on standard lists of variants, as opposed to its peculiar readings) would say that it agrees with the Byzantine text. Genetically, Dabs1 is "Western." Quantitatively, it is Byzantine.

Does this matter? In a word, yes. We have no need of Dabs1 as a Byzantine text; there are plenty of others to choose from, and they aren't all mixed up with "Western" readings and places where the copyist misread a correction in D, producing nonsense. But if we didn't have D, Dabs1 would be a significant witness to the "Western" text; even though it's mostly Byzantine, its non-Byzantine readings go back to an early state of the "Western" and should be used to reconstruct that type.

But if we know that to be true of Dabs1, shouldn't it be equally true of 81, or 104, or 565, or any other mixed manuscript? These manuscripts don't lose value because their ancestors are lost; they gain. And, somehow, we need to find their components. At this point, Colwell/Tune, Claremont, Hurtado, and everyone else who insists on quantitative text-types fails. Their results are accurate, but they do not help us!

So now what? The task is to find a definition of text-types which somehow account for mixture.

This is an area where workers have been relatively few; not all critics recognize the need for it, and even had the need been recognized, it was hard to do much until the present generation -- partly due to lack of data and partly because the approaches proposed have all been computationally intensive; much of what follows is possible only due to the use of computers.

Let me start with my own personal approach, simply because I know it and know how I came to it. When I started, I had seen absolutely no research of this type. I was reading books like Metzger claiming that this manuscript was related to that. But I'm a physicist; I wanted numbers. I took the data from the UBS editions and stuffed it into a database, and started calculating rates of agreement. (This was fundamentally similar to the Munster "thousand readings," with the difference that I have some idea of what constitutes a meaningful sample.)

The results didn't agree much with what everyone was saying. Either textual critics were insane, or more was needed to verify the claimed results.

Without prejudice to the theory that textual critics are insane, I decided to work on better tools. I spent about five years on this, working up better mathematics, and I never really finished; I was never able to produce an independent, verifiable, and non-sample-dependent definition of a text-type. But I reached certain conclusions which I believe incontestible. (But, of course, they are contested.)

The most important step, in my opinion, is the use of multiple statistics for comparison. Colwell's "quantitative method" work is based only on overall rates of agreements. The Claremont method uses classified agreements, but with very limited scope and no flexibility. Hutton used only special sorts of agreements.

Instead of using a single statistic, we should use multiple statistics. (The first to propose something along these lines was Colwell, but the first to publish a method of this sort was Bart D. Ehrman [*37] Ehrman classifies readings according to how important they are in studying the text-type (e.g. some readings are "characteristic" of the type). This is a distinct improvement in the sense that it gets at the nature of readings. If we knew with certainty the nature of all extant text-types, it would be effective. The defect, however, is the same as Hutton's: It assumes the solution. Ehrman can't find new text-types because his method forces him into the straightjacket of existing types. And if his list of witnesses is wrong, as Streeter's was, then his results are ruined.)

My own method generates profiles on the fly. This has the advantage that you need know nothing about the readings or the texts. It is based on four measures of relationship: Overall percentage of agrements, percentage of agreements in readings where both manuscripts are non-Byzantine (this measures the kinship of mixed manuscripts), percentage of agreements where the Byzantine text divides (this helps measure the Byzantine group to which a manuscript belongs), and "near-singular readings" (readings where the manuscript has the support only of a handful of witnesses. This statistic serves to find a document's immediate kin).

There is nothing magic about these statistics; presumably you could replace one or two of them with some other measure. But together they offer something that a mere comparison of overall agreements ever can: A picture of the component texts of a reading. If two manuscripts have high overall rates of agreement, then of course they are akin. (Though "high" in this context means more than 70% agreement!) But a low rate of agreement does not deny kinship; it may mean the manuscripts are unrelated, or that they are related but with different patterns of mixture. High rates of agreement in non-Byzantine, and especially unique, readings is what counts. This is the same as Ehrman's concept, but without pre-assuming text-types.

The use of multiple statistic methods, since they have never been formally tested, is beyond the scope of this article. Suffice it to say that the definition of text-types and the relationships between manuscripts is a field with much room for growth.

One such recent example is the work of Stephen C. Carlson. He has turned to the biological sciences for help, notably from the mathematical area known as "cladistics." For a brief overview of his results, the reader should consult the article on stemma. Carlson's work does not directly address the matter of text-types. Indeed, his stemma are often so complex that no true text-types can be discerned. This is surprising and disconcerting; the existence of text-types seems well-established, and if Carlson cannot find them, it implies a real need for examining either his results or our overall thinking. But doing so could well give us a whole new perspective on the matter -- for example, it reminds us of the point in footnote [*2] that the exact ancestor of a text-type probably never existed.

The Use of Text-Types in Textual Criticism

Different scholars evaluate the evidence of text-types differently. Westcott and Hort's text is based largely on the evidence of text-types, and remains the model New Testament text to this day (if it be noted that the UBS text has now supplanted WH, it should be noticed that UBS, like the texts of Bover and Merk, differs very little from WH). By contrast, von Soden's text, also based on a theory of text-types, is not treated with much respect.

The warning here, of course, is that text-types must be used accurately. If our textual theory is inadequate, the text based on it can only be inadequate. Work on text-types can only stop when all known manuscripts have been comprehensively examined.

In the meantime, we must decide how to use our provisional text-types. Some scholars continue to follow them slavishly (and inaccurately, since these scholars usually continue their allegiance to the Westcott-Hort theories). Others reject text-types altogether.

In the author's view, this is foolish. The way Hort dealt with text-types was subject to attack, because in his time only two early types were admitted, leaving us with no physical basis for deciding between the two. One could only choose between the types on internal grounds. Hort himself admitted this problem.[38] Today, however, with three or more non-Byzantine text-types for most Biblical books, we can do better. We cannot rely on a particular text-type absolutely, since all are subject to various defects. Still, if one accepts the Hortian theory that the Byzantine text is late, then a reading supported by all pre-Byzantine text-types can surely be regarded as original (or, at least, as the earliest recoverable text). A reading supported by a majority of early types may not be original, but the "presumption of originality" is in its favor. Such a reading should only be set aside if there is overwhelming internal evidence against it. Take, as an example, Jude 1. The UBS text reads tois en qeou patri. After tois, however, some two dozen witnesses, including (6) 322 323 424c 614 876 945 1241 1243 1505 1611 1739 1852 1881 2138 2412 2492 2495 sy arm, add eqnhsin. (The prejudices of the UBS comittee are clearly shown by the fact that they rate this variant an "A," meaning that they have no doubts. This presumably is because all the important uncials support their reading.) But look at the evidence: of the three non-Byzantine text-types in the Catholics (as found by Richards, or Amphoux, or, well, me), two (family 1739 and family 2138) add eqnhsin. Only the Alexandrian text (P72 Aleph A B Y 81 436) omits the word. Since there is no reason for the insertion (there is no similar passage in the New Testament), it is at least reasonable to add eqnhsin on the evidence of two of the three text-types. We might, of course, bracket it as questionable.[*39]

In addition, knowledge of text-types can sometimes affect how we assess a variant. Let us take 2 Pet. 2:13 as an example. The UBS text reads en tais apatais. This is in fact a triple variant:

Editors generally reject agapais as an assimilation to Jude 12. However, the readings agnoiais almost certainly derives from agapais. Since agnoias is supported by family 1739, an early text-type, it is much more reasonable to assume that the original reading is agapais, and that apatais and agnoias are both errors derived from this. (Eberhard Nestle also offered cogent internal reasons to adopt this reading.[40])

A final warning: All of the above is about classifying manuscripts. A description of a manuscript must consist of two parts: The manuscript's affinities and its peculiarities. Many manyscripts are unreliable in some way or other -- they exchange e and ai, they omit words, they misspell names, they otherwise render themselves unhelpful for certain variants. Knowing which manuscripts are related is no use if you don't know where you can trust them. Manuscripts must be treated as individuals and as members of a group.

Let's summarize: Textual criticism is based on internal and external criteria. But -- unless one is content to be a radical eclectic[*41] -- the only firm basis for criticism is actual manuscripts. And those manuscripts can only be used properly if their text-types are known and their relationships studied. Else how can we tell which readings are authentic to the manuscript's tradition and which are simply errors?

As has so often been the case, it is hard to make a better summary than Colwell's:

The program of textual studies requires that the critic take five steps. I, Begin with readings; II, Characterize individual scribes and manuscripts; III, Group the manuscripts; IV, Construct a historical framework; V, Make a final judgment on readings.[42]

Appendix I: The Names and Descriptions of the Various Text-Types

The following list shows the various names that different scholars have used for text-types. The first element in each list is what I consider the "proper" modern name; this is followed by a list of editors and the names they used.

Generally Acknowledged Text Types

Westcott-Hort -- Neutral+Alexandrian (also a)
Von Soden -- Eta (Hesychian) (H)
Kenyon -- B (b)
Lagrange -- B
Characteristics of the type: Conservative. Relatively free of harmonzation and paraphrase. Short. Willing to accept difficult readings.
Primary witnesses: P75 (gospels), B (except in Paul), Aleph, Coptic versions. Also A, C, 33 in Paul; A 33 in the Catholics; A C in the Apocalypse.

Westcott-Hort -- Syrian (also d)
Von Soden -- Kappa (Koine) (K)
Kenyon -- A (a)
Lagrange -- A
Characteristics of the type: Widespread. Usually regarded as far-removed from the original documents, but worthy of detailed study because of the influence it has had on mixed manuscripts. Marked by smooth and easy readings and by harmonizations, but rarely indulges in paraphrase or the major expansions seemingly found in the "Western" text. Widely regarded as derived from other text-types; it usually preserves the easiest reading. It rarely creates readings.[*43]
Primary witnesses: A E F G H K M S U V Y G P S etc. (gospels); H L P 049 056 0142 (Acts); K L 049 056 0142 (Paul, Catholics); P 046 (Apocalypse). Also found in the mass of minuscules; over 80% of manuscripts are purely Byzantine, over 90% are primarily Byzantine, and not more than 2% can be considered entirely free of Byzantine mixture.

Von Soden -- Iota (Jerusalem) (I), in part (most strong "Cæsarean" witnesses are found in Soden's Ia group, with family 1 being his Ih and family 13 being Ii.)
Kenyon -- Gamma (g)
Lagrange -- C
Characteristics of the type: Mildly paraphrastic, so as to give an appearance of falling between the Alexandrian and "Western" texts. Since no pure manuscripts are known, most other descriptions of the type have been conjectural. To date found only in the gospels (unless family 1739 is Cæsarean, which is unlikely).
Primary witnesses: Q, family 1, family 13, 565, 700, arm, geo (P45 and parts of W claimed by some; however, P45 is a wild text, and W's relationship to the group is questionable)
Note: The existence of the "Cæsarean" text has been questioned by many; see the discussion above.

Westcott-Hort -- Western (also b)
Kenyon -- D (d)
Von Soden -- Iota (Jerusalem) (I), in part
Lagrange -- D
Characteristics of the type: Marked by paraphrase, occasional expansion, and possible additions from oral sources. Fond of striking and abrupt readings. Reaches its most extreme form in D/05 (Codex Bezae); the "Western" text of Paul (found in D/06, etc.) is a much more restrained text.
Primary witnesses: D/05 (Gospels, Acts), Old Latin, D/06 (Paul) F/010+G/012 (Paul); occasional readings in the versions. Connected by some with family 2138 and with certain fragmentary papyri.

Proposed Text-Types

P46+B (Paul)
Zuntz -- Proto-Alexandrian
Characteristics of the type: Generally possessed of very rough, unpolished readings which give strong evidence of being original. Forceful. Few witnesses are known, so the type is difficult to reconstruct.
Primary witnesses: P46, B, Sahidic

family 1739 (Acts, Paul, Catholics)
Zuntz -- Proto-Alexandrian
Characteristics of the type: Stands midway between the other types. It shares readings with P46/B, the "Alexandrian" text, and the "Western" text. Close to but not identical with Origen. Its readings are generally conservative; it will make occasional clarifications but no major changes. Arguably the best text-type in Paul.
Primary witnesses: 1739. In Paul, also 0243/0121b (which appears to be a cousin of 1739). 1881 is the third witness here. In the Catholics, the core witnesses are C, 1241, and 1739, with most of the lesser manuscripts clustered around 1739.

family 2138 (Acts, Paul, Catholics)
Vaganay -- "Western"
Characteristics of the type: Heavily Byzantine (especially in Paul, where the type almost disappears), but with a large number of independent readings. Often has striking variants which, however, do not appear to be related to the Latin. Therefore the type does not appear to be "Western."
Primary witnesses: 2138 (except in Paul), 1611, 1505+2495, Harklean, 2412+614 (except in Paul), 630+2200+1799+429+522 (Catholics only)

Appendix II: Text-Types and their Witnesses

Primary witnesses are shown in bold (e.g. P75); witnesses with only scattered readings of a type are enclosed in parentheses. Subgroups within the larger group are joined by plus signs (+). Note that this list is not comprehensive.[*44] Also, some of the groups (e.g. the witnesses to the "Cæsarean" text) are based on standard lists, and have not been tested by modern methods.








P66 P75+B+T Aleph+Z C L (X) D (Mark) Xi (Y) 33 579 892 1241 sa bo
P74 Aleph A B C 33 81 1175 vg? sa bo
Aleph A C I (P) 33 81 (104) (436) 442 1175 (1241supp) 1506* 1962 fam 2127 2464 bo
P72+B Aleph A+33+436 Y 81 vg sa bo
A C vg 1006 2050 2053 2062 2344? bo


(also includes most minuscules)

(A) E F G H K M (N) (P) (Q) S U V Y G L S F 047 (E) H L P Y 049 056 0142 1241 K L (Y) 049 056 0142 (33 1175 2464 in Romans)
K L 049 056 0142
(1175 in Johannines)
MK: 046 429 522 2138
MA: P 051 1 181


Q f1 f13 22 28(Mk) 565 700 arm geo


D Old Latin Syrsin? Syrcur? D (E) Syrhark-marg saG67 D F G Old Latin (not r) (629) (goth)


P13 P46 B sa


P47 Aleph 2344?

family 1739

1739 630 945 1891 2200 2298 1739 0243/0121b 0121a 6 424c 1881 (630 in Romans-Galatians) C 1241 1739 6 322 323 945 1881 2298 1243+2492?

family 2138

614+2412 383? 1505+2495 1611 2138 Syrhark 1505+2495 1611 Syrhark 2005? (1022) 614+2412 630+1799+2200 1505+2495 1611 2138 Syrhark 206 429 522 1799

family 330

330+451 2492


f1 = family 1 = 1, 118, 131, 205, 209, 1582. This is the family known as the "Lake Group" (l).

f13 = family 13 = 13, 69, 124, 174, 230, 346, 543, 788, 826, 828, 983, l547 etc. This is the family known as the "Ferrar Group" (f).

family 2127 = 256, 263?, 365, 459, 1319, 1573, 2127 (perhaps also arm) This family was called "family 1319" by the followers of von Soden.

It also appears likely that we should define a "family 630," consisting of, at minimum 630 2200, and probably also 206 429 522. The curious thing about this group is its shifting nature. In Acts it goes with 1739. In Paul it goes with 1739 in the early epistles, then turns Byzantine. In the Catholics it goes with 2138. There is a hint here of a relationship -- historical rather than textual -- between family 1739 and family 2138 that might be worth investigating.

Appendix III: Von Soden's Textual System

The following lists summarize Von Soden's system in the various portions of the New Testament. For the H and I types, all manuscripts of the type cited by von Soden are listed (except for occasional fragments. Gregory notation is used throughout); for the Byzantine (K) types, only a handful of manuscripts are included.

It should be noted that von Soden treated commentary manuscripts as a separate type with a separate history; with the exception of manuscripts of the Apocalypse (where there is a separate Andreas type), they are not treated here.

It should be remembered that Von Soden did not cite manuscripts in the order given here, nor in numerical order. Students wishing to use his edition will have to consult it, or one of the related works, to use his apparatus.

To summarize Von Soden's textual theory, there are three types, I, H, and K. The first of these is, very roughly, the "Western" and "Cæsarean" texts (with a lot of extraneous material thrown in); the second is the Alexandrian text, and the third the Byzantine. Von Soden sought the original text in the consensus of these.

(It should be added that, with only the most minor exceptions, von Soden does not allow the possibility of mixture. This is one of the major defects in his classification of the I groups.)

The Gospels


For an overall view of Von Soden's system in the Acts, Paul, and the Catholic Epistles, see the Summary following the section on the Catholic Epistles.


For an overall view of Von Soden's system in the Acts, Paul, and the Catholic Epistles, see the Summary following the section on the Catholic Epistles.

Catholic Epistles

For an overall view of Von Soden's system in the Acts, Paul, and the Catholic Epistles, see the Summary following the section on the Catholic Epistles.

Summary of Von Soden's work on the Acts, Paul, Catholic Epistles

It has become customary to ignore Von Soden's groupings outside the Gospels, and with good reason; many of the manuscripts he classified simply do not show the features he attributes to them, and manuscripts shift groups more than his system allows. And yet, if we look at the overall results for the Acts and Epistles, von Soden's results bear a striking resemblance to the results outlined in this document. The "H" group is the Alexandrian text (von Soden cannot be faulted for failing to realize the existence of the P46/B type in Paul; a text-type can only be recognized when two witnesses exist, and von Soden did not know P46). Ia is the "Western" text. Ib is Family 1739. Ic is Family 2138. And the "K" text is the Byzantine text. If von Soden is to be faulted, it is for not clearly identifying the boundaries of the types. In other words, though Von Soden did not realize it, he too was struggling with the definition of a text-type, just as we have done. In addition, von Soden included many irrelevant witnesses in his groups (often, it appears, by assuming that a manuscript had the same type in all three sections unless it was known to undergo a shift). This, combined with the rather sloppy way witnesses were cited, makes it hard to perceive the broad accuracy of its groupings (e.g. it's hard to realize that Ib is Family 1739 in Paul when von Soden places 1739 and all its kin in H!).


Von Soden's textual theory in the Apocalypse has received even less attention than his work in other areas, having been completely eclipsed by the work of Schmid.[12] The outline which follows is, therefore, less detailed than those which preceded. Note that the following list does not agree, even approximately, with the citation order in Merk or Bover! Von Soden in these books has a bad habit of putting manuscripts in multiple categories -- e.g. 051 is listed as an Andreas manuscript (An2) with a text-type of H. The information here is as interpreted in the Kurzgefasste Liste. Note that not all the manuscripts listed under the Andreas type actually have Andreas's commentary; the manuscripts listed here are listed by von Soden as having the Andreas-type text, but some (e.g. 1611) have no commentary at all.


1. English translation from Metzger, The Text of the New Testament, 2nd/3rd edition (Oxford, 1992), page 112. [back]

2. Almost the only exception to this is E.C. Colwell, who carefully defines all four levels and gives examples of each. A family, in his terminology, is a group for which an accurate stemma can be constructed. By this definition, he felt that that family 1 is a true family but family 13 is not. The worst offender against this system is probably B. H. Streeter, who called the Cæsarean text-type "Family Q."
Because other textual critics have not used the intermediate levels, no widely-accepted terminology exists. Even Colwell had trouble with this; at various times he referred to the intermediate levels as "tribes," "clans," and "sub-text-types." (His formal suggestion, in "Genealogical Method: Its Achievements and Its Limitations," reprinted in Studies in Methodology in New Testament Textual Criticism, Eerdmans, 1969, p. 82, is to use the "clan" for one of the intermediate levels.) For this reason I have used "family" for all levels of kinship. I know better, but I have no other language available. A logical approach might be to speak of, in ascending order, the "family" (a group of related manuscripts for which a detailed stemma can be constructed), a "super-family" (for which one can sketch a stemma without being able to offer full details), a "sub-text-type" (closer than a text-type, but too loose for any stemmatic work to be done), and the full-fledged text-type. On this basis, P75-B-T in the gospels would, I believe (in the absence of certain evidence either way), be a sub-text-type. Family 1739 in Paul would be a super-family. So would Family 2138 in the Catholics. In the Catholics, the "tight" form of family 1739 (excluding C 1241) would be a super-family; the larger family (including those manuscripts) would be a text-type. In Paul, 330 and 451 form a family; adding 2492 creates a super-family. Family 2127 is a super-family.
Note that it is possible to determine the ancestral text of a family, and perhaps even a super-family, precisely. These groups presumably derive from some one examplar. This is not true of the higher levels (especially of text-types). One cannot construct a text and say, "This is the is the Alexandrian text." There never was such a thing; no manuscript ever had all the readings we call "Alexandrian." But we can determine many, perhaps most, of the readings characteristic of the type, and use these to help us determine the original text. [back]

3. See, e.g. Kurt and Barbara Aland, The Text of the New Testament (English translation by Erroll F. Rhodes, Eerdmans, 1989). On p. 56, in discussing text-types, they say "In the fourth century a new era begins." On p. 65, the claim is even more forceful: "The major text-types trace their beginnings to the Diocletianic persecution and the Age of Constantine which followed."
I must admit to finding this a very curious view; it seems to imply that text-types were somehow something that late scribes wanted to have, and so developed starting from the fourth century. But clearly the Alexandrian text, at the very least, existed before the fourth century, since P75 is effectively identical to B. So either P75 is a part of the Alexandrian text, or it is the archetype of the text. Either way, the Alexandrian text-type was in existence in the third century, even if not widespread. [back]

4. Eldon J. Epp, "Decision Points in New Testament Textual Criticism," printed in Epp and Gordon D. Fee, Studies in the Theory and Practice of New Testament Textual Criticism (Studies and Documents 45, Eerdmans, 1993), p. 37. The essay goes on to marshal arguments on both sides. [back]

5. In fairness, it should be pointed out that there are two sorts of supporters of the Byzantine text, with variations in each group. Without going into detail, since their views remain in the minority, they are:

A summary of the arguments of the pro-Byzantine scholars, showing evidence that the Byzantine text is at least better than Hort claimed, can be found in the article on Byzantine Priority. [back]

6. Frederick Wisse (Frederick Wisse, The Profile Method for Classifying and Evaluating Manuscript Evidence (Studies & Documents 44, Eerdmans, 1982), p. 94), reports that "Ki and Kik are not distinguishable from K1, and K1 could not maintain itself as an independent group and is treated as a Kx cluster." As a partial defense of Von Soden, however, we might note that Wisse's data indicates a historical if not a textual distinction between Cluster W and the rest of Kx; most early Kx manuscripts belong to Cluster W, and the type seems to have died out by the end of the twelfth century, when Kx proper becomes dominant. As evidence we offer this list of early Kx manuscripts (consisting of all purely Kx manuscripts listed by Wisse as of the tenth century or earlier, plus all pure Cluster W manuscripts of any age):
CenturyKx Cl W Not Cl W
VIII and before E --NONE--
IX V W 461 1080 1295 2142 047 2224 2500
X S 151 344 364 584 1077 1281 2563 2722 G H G 14 29 135 144 274 435 478 564 568 669 875 1055 1078 1172 1203 1225 1351 1662 2195 2414
XI 65 123 143 271 277 699 1045 1470 1691 2176 2287 2442 2571 2637 (nearly 100)
XII 471 667 688 1083 2702 (Hundreds)
and after

Observe that in the eighth and ninth centuries Cluster W is dominant; in the tenth Kx proper is taking over, and after the eleventh Cluster W was dying out. [back]

7. cf. Wisse, pp. 103-105. [back]

8. Ibid, pp. 92-94. [back]

9. According to Wisse, 734 of 1385 Gospel manuscripts tested belonged to Kx in whole or in part. Ibid, pp. 16-17. [back]

10. See Harry A. Sturz, The Byzantine Text-Type & New Testament Textual Criticism (Nelson, 1984). Sturz's findings are based on Hort's three text-types, but with the Byzantine text upheld as early. Thus, unlike von Soden (who felt that K was the worst of the three text-types), his text is eclectic but perhaps more Byzantine than anything else. His method is shown by the names of some of his chapters: "Distinctive Byzantine Readings Are Found in Early Papyri" (true enough, but many -- such as Colwell -- believe that a text-type consists of manuscripts, not readings; in any case, not all distinctive Byzantine readings have early attestation); "The Silence of the Fathers Is Explainable and Therefore Is Not a Proof of Lateness" (Sturz points out that Chrysostom, generally regarded as the earliest Byzantine witness, is also the earliest writer from the Antiochene region. A legitimate argument, but if the Byzantine text were original, would its readings not be found outside Byzantium and Syria?); "The 'Conflate' or Longer Readings Are Not a Proof of Lateness" (true, but most moderns accept that conflate readings should not be used as arguments against the Byzantine text; they are too few); "The Composite Nature of the Byzantine Text Attests the Early Existence of Its Readings Where Its Strands Unite" (contradictory on its face; what Sturz means is that the great breadth of the Byzantine text indicates that it is much older than its witnesses. This can be conceded -- but it should be noted that, except in the Gospels, the purest Byzantine witnesses come from the ninth century; even if their archetype is much earlier, it need not be early. Also, the Byzantine text, compared to the other known text-types, shows relatively little variation, meaning that the witnesses need not be far removed from the earliest examples of the text-type); "The Byzantine Text Is Unedited in the Westcott-Hort Sense" (now widely conceded, but not relevant to the argument. It can be unedited and still be late). Sturz devotes most of his efforts to disproving the theories of Westcott & Hort (theories which, it should be noted, are no longer accepted in detail by anyone); he also offers extensive lists of Byzantine readings which are found in early manuscripts. He cannot, however, offer any proof that the Byzantine text as a whole predates the fourth century. Sturz is also guilty of some logical fallacies -- e.g. on pp. 91-92 he uses an argument of Silva New's, which really applies to Codex Alexandrinus, to demonstrate that family P predates A's date in the fifth century. It is true that an ancestor of the two must predate A -- but not that the fully-developed family P text must do so.
The above may sound like a blanket indictment of Sturz. It is not; in fact, Sturz has a good deal of truth in his case (see the article on Byzantine Priority). It's just that Sturz's methodology is invalid (what he showed, he showed despite himself), and he has been reduced to an invalid form of argument by the absurd and insupportable claims of all parties in the argument. [back]

11. Kurt Aland and Barbara Aland, The Text of the New Testament (translated by Erroll F. Rhodes; Eerdmans, 1989). The manuscript statistics occupy most of pp. 83-158. [back]

12. Josef Schmid, Studien zur Geschichte des griechischen Apokalypse-Textes (Munich, 1955-1956). [back]

13. W. L. Richards, The Classification of the Greek Manuscripts of the Johannine Epistles (SBL Dissertation Series 35, Scholars Press, 1977). See especially pp. 137-141. Among Richards's more perverse assumptions is his belief that "Mixed" qualifies as a text-type (! -- see in particular pp. 176-178).
In defence of Richards, Clinton Baldwin points out to me that the bad terminology used by Richards does not preclude these manuscripts forming a text-type (an obvious truth); indeed, it is possible that this type arose by mixture. The problem is the way Richards expresses things: Calling a text-type "mixed" implies that it arose by mixture -- or even, possibly, that it is the result of a bunch of manuscripts being independently mixed. The former is possible but prejudicial (the existence of the mixture needs to be demonstrated, not asserted); the latter is flatly impossible. Thus this group, if verified, needs another name. [back]

14. Most of Duplacy's and Amphoux's works are available only in French. Brief English summaries are found in Leon Vaganay and Christian-Bernard Amphoux's An Introduction to New Testament Textual Criticism (English translation by Amphoux and Jenny Heimerdinger, Cambridge, 1991), pp. 23-24, 97, 103-105; also 70, 106-116, etc. [back]

15. G. Zuntz, The Text of the Epistles; A Disquisition upon the Corpus Paulinum (Schweitz Lectures, 1953). [back]

16. Zuntz's words are "We may describe this group -- P46 B 1739 sah boh Clem Orig -- as 'proto-Alexandrian'." (op. cit., page 156). Additional, if partial, confirmation of this is found confirmation of this is found in M. Silva's essay on P46, Aleph, A, and B in Galatians, where he found a clear kinship between P46 and B and another between Aleph and A. See "The Text of Galatians: Evidence from the Earliest Greek Manuscripts," in David Alan Black, ed., Scribes and Scriptures: New Testament Essays in Honor of J. Harold Greenlee (Eisenbrauns, 1992). [back]

17. "Method in Establishing the Nature of Text-Types of New Testament Manuscripts," reprinted in Studies in Methodology(Eerdmans, 1969), p. 53. [back]

18. B. H. Streeter, The Four Gospels: A Study of Origins (Macmillan, 1924, 1927). Textual problems are covered in pp. 26-148, 565-600. [back]

19. Published as Larry W. Hurtado, Text-Critical Methodology and the Pre-Cæsarean Text: Codex W in the Gospel of Mark (Studies and Documents 43, Eerdmans, 1981) [back]

20. I have not personally seen any writings which claim that the P75/B and Aleph textual groups belong to separate text-types. R. Kieffer, however, is reported to have found two Alexandrian texts in a portion of John. (See David C. Parker, "The Majuscule Manuscripts of the New Testament," printed in Ehrman & Holmes, The Text of the New Testament in Contemporary Research (Studies & Documents 46, Eerdmans, 1995), pp. 34-35.) [back]

21. For example, I know of only one instance in Paul where all the text-types have clearly separate readings. The reading is 1 Cor. 14:39 (UBS reads to lalein mh kwluete glwssais). The variants are as follows:

Less good, because certain witnesses depart their type, is Romans 4:1 eurhkenai Abraam ton propatora hmwn. Here the readings break down as:

As for how often the witnesses divide, it can be shown that the three text-types P46/B, Alexandrian, and family 1739 are all closer to each other than they are to the Byzantine text, and that the "Western" text is even more distinct. Does this mean that the P46/B, the Alexandrian text, and family 1739 all form one text-type? That has been the view of most scholars, but it need not be so. Just as a crystal can be more likely to break at one facet than at another, text-types can be more or less distinct. We can just as well account for the facts of the case by assuming that the P46/B, Alexandrian, and family 1739 types were simply truer to the original text than was the "Western" group; this would make them just as much alike as if they were genetically related. To me, this appears to be the actual situation in Paul. [back]

22. Hort's basic statement is found in The New Testament in the Original Greek, Introduction [and] Appendix, p. 57, paragraph 73, "The proper method of genealogy consists... in the more or less complete recovery of successive ancestors by analysis of their respective descendants, each ancestral text so recovered being in its turn used... for the recovery of a yet earlier common ancestor." In the same paragraph Hort admits that the number of manuscripts preserved rarely permits real genealogical work -- but he still believes in the method, i.e. in reconstructing one Alexandrian text and one Western text -- and reconstructing the "original" text on this basis. Moderns hold out no such hope; even though we have access to more and earlier manuscripts than Hort, we have no reason to believe that text-types ever existed in a single manuscript. Thus almost all modern critics agree that Hort's use of B as the basis of the "Neutral" text, and the "Neutral" text as substantially equivalent to the original text, must be set aside and a more eclectic method substituted. If nothing else, more attention needs to be paid to the other representatives of the Alexandrian text, so that the history of the text-type can be studied. [back]

23. Observe Colwell's comment, "[Hort] used genealogical method very little and that the basic element in his method was judgement of readings is now widely recognized" (made in "Method in Grouping New Testament Manuscripts," reprinted in Studies in Methodology, p. 2. This essay, although not as well-known as the 1963 essay listed below, is probably the best statement of how to deal with text-types -- and how not to deal with them -- ever written). In "Genealogical Method: Its Achievement and Its Limitations" (Studies in Methodology, p. 65) Colwell makes the interesting observation that, although Hort diagrams a manuscript stemma (p. 54), it is artificial. The manuscripts shown do not exist. Streeter (op. cit., p. 26) diagrams both his own and Hort's theories, but in both diagrams the manuscripts are offered more as examples of a type than as actual products of genealogy.
There are a few manuscripts for which we can trace exact genealogy -- but they are few. In Kurt Aland's 1963 edition of the Kurzgefasste Liste der Griechischen Handschriften des Neuen Testaments we find the following:

In addition, there are certain manuscripts that are so close that they can be confidently listed as descended from an close common ancestor though their exact relationship is uncertain (e.g. Fp and Gp, 1 and 1582, 205 and 209, 614 and 2412, 630 and 2200, 0243 and 1739; probably also 1739 and 1891 in Acts; we might also list 2495, slightly corrupted from 1505; and 0121, descended from 1739 with some Byzantine mixture).
Finally, Wisse lists roughly a third of Kr manuscripts as "perfect," i.e. agreeing exactly with the group profile. Chances are that some of these sixty manuscripts, if examined very carefully, would prove to be closely related.
This out of a nominal list of 2972 Greek manuscripts! It is likely that there are additional undiscovered copies (since so few manuscripts have properly been cross-compared), but available evidence indicates that they are few. Clearly true genealogy has little place in NT studies.
For some slight background on how genealogy is used (in its true form), see the article on Non-Biblical Textual Criticism and the item on Stemma. [back]

24. Indeed, Colwell was one of the first to plead exclusively for the use of the word "text-type" in this context. See ibid, p. 9. [back]

25. Ibid. [back]

26. Ernest C. Colwell and Ernest W. Tune, "Method in Establishing Quantitative Relationships between Text-Types of New Testament Manuscripts," reprinted in Studies in Methodology, p. 59. [back]

27. Ironically, it was Colwell himself who first pointed out the defect in his method -- four years before he proposed his definition! In "Method in Locating a Newly-Discovered Manuscript" (Studies in Methodology, page 33), he wrote "Weak members of a Text-type may contain no more of the total content of a text-type than strong members of some other text-type may contain. The comparison in total agreements of one manuscript with another manuscript has little significance beyond that of confirmation, and then only if the agreement is large enough to be distinctive." R. H. Rouse, in "The Transmission of the Text" (published in Richard Jenkins, ed., The Legacy of Rome: A New Appraisal, p. 39) puts this rather in reverse: "if survivors are few, the stemma perforce brings into proximity manuscripts that, historically, were widely separated in time and place." In other words, if we have only a small fraction of the manuscripts, we may find textual links which are not in fact genealogical! This phenomenon has been frequently found in biology, where it is known as "long branch assimilation." [back]

28. If someone objects that comparisons across the gospel corpus are not valid, let me simply add that I examined individual books, and even sections of books, and the results were the same within the margin for error. At times the leading manuscripts (especially W) shifted slightly, but the general picture never did. So I present overall statistics because they are simpler.
The actual percentages of agreement with B, for those interested, are as follows (note that these should not be considered definitive; again, statistics depend on the sample used! But because the sample is large, the relative values are likely to be close to correct):
Sorted by manuscriptSorted by percent
MSPercent AgreementMSPercent Agreement
Aleph589/990=59% Xi56/83=67%
W314/979=32% f1410/981=42%

29."Before you can apply statistics you must have exact and complete figures -- which in this field do not exist. In fact, they never will nor can exist. None but commensurable entities can be reduces to figures, and no two variants are strictly commensurable. Readings of all shades between good and bad; slips of the pen and intentional alterations; attestation by anything between one and a thousand witnesses: what is their common denominator?" (Zuntz, The Text of the Epistles, page 58.) It will be evident that I do not entirely agree with his wholesale abandonment of statistics -- but I do agree that statistics, like manuscripts, must be weighed and not counted. [back]

30. "[Mixed] manuscripts could never meet the Colwell-Tune standard of 70%. Agreements expressed in percentages will tend to wash out the characteristics of the group to which the mixed MS belongs. Nothing can offset this drawback of statistical analysis." (Wisse, p. 31). It should be noted that Wisse's own Profile method is in fact statistical -- merely less blatantly so, and based on different statistics! [back]

31. Eldon Jay Epp, "The Papyrus Manuscripts of the New Testament," printed in Ehrman & Holmes, p. 16. [back]

32. Ibid, pp. 16-17. [back]

33. Ibid., p. 18. [back]

34. For the CPM, see especially the work of Wisse cited above. [back]

35. Another "thought-experiment" will demonstrate this point. Let us consider a typical "profile" for a hypothetical "Ephesian" text-type. (In this example I am using the Claremont methodology rather loosely, but it gets its point across.) Let us draw profiles, as Wisse does, with Xs for non-Byantine group readings (and Os for plain old Byzantine readings). So in a sample of six readings, the Ephesian profile would be
Original Text
. . . . . .

Now let's take two manuscripts of this text-type, and arbitrarily mix in three Byzantine readings in each. So we get two profiles that look like this:
MS 1
X . . X . X
. O O . O .

MS 2
. X . X X .
O . O . . O

Not only do the profiles not look particularly "Ephesian," they bear no resemblance to each other! (For the record, there are many more ways to mix three Byzantine readings into six Ephesians readings than the two shown above -- a total of 20, out of 64 possible arrangements of readings -- but they all average out to a mere 50% agreement between the resulting texts: 25% in "Ephesian" readings and 25% in Byzantine readings.) So much for the ability of the CPM to handle mixture. [back]

36. The most recent example of this known to me is Tommy Wasserman's "The Patmos Project: An Investigation of the Patmos Family of New Testament MSS and Its Allies in the Pericope of the Adulteress and Beyond," Th.M. Exam, 2001, now available in a different form in volume 7 of the online digest TC. [back]

37. For Colwell's discussion, see "Method in Locating a Newly-Discovered Manuscript" (op. cit., p. 39). Colwell writes, "In conclusion I suggest that the location of a manuscript within the tradition should use Multiple Readings to find the related group, Distinctive Readings to demonstrate the kinship, and total comparison to confirm the relationship." This is not the list of statistics I offer, and in my opinion is inferior (since "Multiple Readings" assume the solution) -- but it is, obviously, a multiple-statistic method.
For Ehrman's initial publication, see Didymus the Blind and the Text of the Gospels. [back]

38. See Hort's discussion in op. cit., paragraphs 71-72, pp. 56-57 (referring to the diagram on p. 54); also (more summarily), paragraph 50, p. 42. [back]

39. Of course, there are instances where internal evidence outweighs the majority of text-types. A good example of this is Matt. 27:16-17; although the Alexandrian and "Western" types both read "Barabbas" and only the Cæsarean reads "Jesus Barabbas," we should accept the latter reading on internal grounds. [back]

40. Eberhard Nestle, Introduction to the Textual Criticism of the Greek New Testament (English translation by William Edie, Putnam, 1901), pp. 325-326. [back]

41. That is, to work in the manner of Kilpatrick and Elliot, who gather variants from the manuscripts but then judge them based only on internal criteria. Colwell, in commenting on this overuse of internal criteria, quotes a clever remark from A. E. Housman: "[These editors use manuscripts] as drunkards use lampposts--, not to light them on their way but to dissimulate their instability." (Quoted in Studies in Methodology, p. 153). The irony is that Housman chose to do his chief work on Manilius at least in part because it afforded more than the usual scope for conjectures. [back]

42. Colwell, "Hort Redivivus: A Plea and a Program," reprinted in Studies in Methodology, p. 160. [back]

43. So Zuntz: "...it seems to me unlikely that the Byzantine editors ever altered the text without manuscript evidence. They left so many hopelessly difficult places unassailed! Their method, I submit, was selection rather than conjecture." (The Text of the Epistles, p. 55; quoted in part by Colwell in Studies in Methodology, p. 49). [back]

44. Scholars who wish to find more related witnesses may wish to consult K. Aland et al, Text und Textwert der griechischen Handschriften des Neuen Testaments (1987 and following). This is unquestional the best -- often the only -- gathering of data available for most biblical manuscripts. Students should, however, be aware of the difficulties in using this edition. First, it does not sample enough variants to allow complete classification of mixed witnesses (in Paul, e.g., there are fewer than 300 readings, rather than the 800 I would like to see. This means that it can be used to classify relatively pure manuscripts, but is not sufficient to deal with mixed manuscripts). Second, it is difficult to use; the data is scattered throughout the volumes, and there is no simple way to look at the data for an entire corpus of books. This makes it easier to examine the data for particular books, but almost impossible to use the data over large areas. Third, the summaries of results (which show the most closely related manuscripts) are almost unreadable, as they consistently show manuscripts which are extant for only one or two variants at the top, leaving the user helplessly struggling to find a manuscript's real relatives. The Alands have already used the data to make one useful determination: They have given us a fairly definitive list of Byzantine manuscripts in their list of "Categories" (though it does not classify the manuscripts within the Byzantine tradition). But the student who wishes to do more, though well-advised to start with T&T, should be prepared to have to do much further analysis. Frankly, someone with some genuine math skills and a vast amount of free time could do the world a great favor and take T&T and convert the results into a useful single volume of data from which actual analysis could proceed. [back]