Kurt Aland and Barbara Aland's

Manuscript Categories

Contents: * Introduction * The Categories * What the Categories Mean * Appendix I: How the Alands Classify Leading Minuscules * Appendix II: Testing the Classifications * Appendix III: A Rigorous Classification


In 1981, Kurt and Barbara Aland published Der Text des Neuen Testaments (English translation: The Text of the New Testament, translated by Erroll F. Rhodes, Second edition, Eerdmans/ E. J. Brill, 1989). The most noteworthy feature of this edition was its new classification of manuscripts. Based primarily on the "Thousand Readings in a Thousand Minuscules" project (the results of which are now being published in the series Text und Textwert der griechischen Handschriften des Neuen Testaments, K. Aland et al, 1987 and following), the Alands set out to place the vast majority of known manuscripts into "Categories."

As a classification scheme, their attempt was at once a success and a failure. A success, in that it has conveniently gathered data about how Byzantine the various manuscripts are. A failure, because it has not been widely adopted, and in any case does not succeed in moving beyond Byzantine/non-Byzantine classification.

The Categories

We may briefly outline their classification scheme as follows (excerpted from Aland & Aland, p. 106):

The Alands base their categorizations on a very simple set of statistics. All of a manuscripts's readings are broken up into "Type 1" readings (Byzantine), "Type 2" readings (readings which agree with GNT, i.e. almost without exception Alexandrian readings; some readings, which are both Alexandrian and Byzantine, are "Type 1/2"), and "Type S" readings, which belong to neither Type 1 nor Type 2.

It will thus be observed that the Alands have only one way to measure the nature of a manuscript: By its ratio of Type 1 (Byzantine) to Type 2 (Alexandrian) readings. The Type S readings are completely unclassified; they might be "Western," "Cæsarean" -- or anything else imaginable (including simple errors).

Thus in practice the Alands' categories become:

A handful of examples will demonstrate the imperfections of this system (note that these are not defects in the data, merely the results of the Alands' simplistic analysis which counts only Type 1 and Type 2 readings, rather than the rates of agreement between manuscripts which they also calculated):

The same problem occurs, to an even greater extent, among the Category III manuscripts. While almost every manuscript in this category is mixed, with Byzantine readings combined with other types, the nature of the mixture varies. We have Byzantine/"Western" mixes (629); Byzantine/"Cæsarean" mixes (family 1, family 13, 28, 565, 700), family 1739/Byzantine mixes (6, 323, 424**, 945, etc.), and a large number of Alexandrian/Byzantine mixtures (of which 104 and 579 are typical examples). Taking only Paul as an example, there are also at least two family groups which are heavily Byzantine but highly distinct: Family 1611 (family 2138): 1505, 1611, 2495, etc. and Family 330 (330, 451, 2492).

We should also note that the Alands fail to assign a category to many manuscripts. In general these are manuscripts with a small handful of non-Byzantine readings, but not enough to qualify as Category III. (In effect, one can treat unclassified manuscripts as another category.) This non-category Category has its own problems, however. For example, the leading manuscripts of the large and well-known Family Π -- Π itself and K -- are listed as Category V (which is fair enough, since this family is clearly Byzantine though obviously distinct from Kx and Kr). Of the minuscule members of the family, however, most are included among the Uncategorized.

We may also compare the results of the Alands' classifications with the results of the Claremont Profile Method in Luke. Wisse lists a total of 36 groups. Excluding Group B as a text-type rather than a legitimate group, we still find that in 19 of 35 cases the Alands reach no consensus as to the classification of the members of a group. That is, if we take all the members of one of Wisse's groups, we find that these members are classified by the Alands as being members of two categories -- sometimes even three!. In addition, we find in these groups that at least 25% of the members of the group fall into each of the leading two categories; only seven groups -- including the members of Kx and Kr -- are treated entirely consistently. For details see the entry on the Claremont Profile Method.) In some instances this is likely due to block mixture undetected by Wisse -- but one must also suspect that the Alands did not rigidly define their categories. This generally will not matter in practice -- but one should always allow for the possibility that a manuscript might need to "shift" a category following further examination.

What the Categories Mean

Thus as a genealogical description the Alands' categories fail. A manuscript simply cannot be described by the few statistics they use.

However, the Categorization should not be deemed a complete failure. It is, in fact, one of the most important results of recent years. For the first time, we have a nearly-comprehensive and, within its limits, accurate examination of the minuscules. If Categories II and III, as well as the unclassified manuscripts, contain an immense diversity of material, Category V is absolutely clear: It is the Byzantine text. Manuscripts found here are Byzantine, and manuscripts found in Categories III and higher are not -- at least not purely. In addition, the manuscripts in Category I (with the exception of the fragmentary early papyri, which are too short to classify this way, and 1175, which is block-mixed with the Byzantine text in Paul and the Catholic Epistles) are all very pure representatives of their types. As long as appropriate care is taken to correctly understand the manuscripts in Categories I, II, and III, and the arbitrary Category IV is ignored, the system can be very useful.

See, however, Appendix II for some tweaks to the system.

Appendix I: How the Alands Classify the Leading Minuscules

The table below lists all the minuscules which are cited as "Constant Witnesses" in the Nestle-Aland 26th and 27th editions, along with their Aland categories in each of the five sections of the New Testament. The final column, Comments, shows the categorization I believe should be applied (where it differs from the Alands'), or gives further detail on their categorization.

Manuscripteap crComment
28III (Mk)
V (MtLk)
81IIIIII described as "at least Category II."
323IIIIIIII Actually probably Category V in Paul; block-mixed and so probably Category III in the Catholics
365VIIIV Member of Family 2127. Most members of this family are listed as Category III, although 2127 itself is Category II.
565III "the average is raised by Mark, with Matthew and Luke far lower." (John appears to be more Byzantine than Mark but less so than the other gospels.)
579II (MkLk) Although it is not explicitly stated, the manuscript is probably Category II in John and Category III in Matthew.
614IIIIIIIII Paul should be Category V, not Category III. Listed as a sister to 2412; the pair belong to Family 2138 in the Acts and Catholics but are Byzantine in Paul.
892II Portions of John from a later, much more Byzantine hand
1010V Listed as a possible member of Family 1424, but 1010 is much more Byzantine than the other members of that group and probably does not belong with it. (So also Wisse.)
1175II Probably should be Category I in Acts, II in Paul (except for Romans, which is Byzantine), perhaps III in the Catholics (there are some interesting readings in the earlier letters, but the Johannine Epistles are Byzantine)
1241IIIVIIII Probably should be Category II in Luke, III in the other gospels, V in Acts, I in the Catholics. In Paul, the basic run of the text is Category V. The manuscript has supplements, however (possibly a third of the total) which are clearly Category III
1424III (Mk)
V (MtLk)
1505VIIIIIIIII Pair with 2495. Member of Family 1611/Family 2138 in Acts, Catholics, Paul
1506VII Fragment in Paul, but clearly strongly Alexandrian. May be Category I in that corpus (based on unusual text which omits Romans 16!)
1611IIIIIIIIIIIMember of Family 1611/Family 2138 in Acts, Catholics, Paul
1739IIII Text of Acts is more Byzantine than in Paul or Catholics, but still stands at the head of an independent family, implying Category I
2030III Fragment (about six chapters); categorization must be considered tentative
2050II Fragment (about eight chapters); categorization must be considered tentative
2062I Fragment (about nine chapters); categorization must be considered tentative
2344IIIIIIIIClassification in Catholics perhaps questionable. Manuscript is badly water-damaged and often unreadable
2351III Fragment (about thirteen chapters); categorization must be considered tentative
2427I Mark only. The evidence is strong that it is a forgery.
2464IIIIII Classification is too high; probably should be Category III. Romans is Byzantine.
2495IIIIIIIIIIIIIII Listed as "Category III with reservations, but higher in the Catholic Epistles." In fact a sister or nearly of 1505, and should be classified accordingly.

Appendix II: Testing the Classifications

The descriptions above generally cover the intent of the Aland classifications. But the result needs to be tested -- we want to know how reliable are the classifications.

In an attempt to investigate this, I re-examined the data for some of the manuscripts. For this purpose, I took every manuscript, uncial and minuscule, whose statistics were listed in the second edition of The Text of the New Testament. I chose to use the gospels section as (I assumed) representative. (I'm not so sure this is true, now; it appears that the fraction of valuable manuscripts is much, much higher in the Acts and Epistles than in the Gospels.) I took every manuscript for which there were at least fifty sample readings. In a few cases, where the Aland categorized books individually, I did the same.

It turns out that the Alands gave statistics for 101 manuscripts in the Gospels: ℵ, A, B, C, D, E, F, G, H, K, L, M, N, S, U, V, W, X, Y, Γ, Δ, Θ, Λ, Π, Σ, Φ, Ψ, Ω, 047, 0211, 0233, 1, 5, 6, 13, 28, 33, 61, 69, 157, 180, 189, 205, 209, 218, 263, 330, 346, 365, 431, 461, 522, 543, 565, 579, 597, 700, 720, 788, 826, 828, 886, 892, 945, 983, 1006, 1010, 1071, 1241, 1243, 1251, 1292, 1319, 1342, 1359, 1398, 1409, 1424 (Mark), 1424 (Matthew+Luke), 1448, 1505, 1506, 1542b (Mark), 1563, 1573, 1582, 1642, 1678, 1704, 2127, 2147, 2193, 2200, 2374, 2400, 2427, 2492, 2495, 2516, 2523, 2542, and 2718.

The statistic I adopted for my analysis is the ratio of distinctly non-Byzantine readings to Byzantine readings. That is, the Alands classify readings into four groups: Group 1, which is Byzantine, Group 1+2, which are Byzantine readings also found in the UBS edition, Group 2, which are reading of UBS not found in the Byzantine text, and Group S, which is readings not found in either the Byzantine text or UBS. Generally speaking, we may assume that Group 1 readings are uninteresting, Group 1+2 readings unhelpful, and Group 2 and Group S readings are valuable for classification purposes. So I calculated (Gr2 Rdgs + GrS Rdgs)/(Gr 1 Rdgs).

For a purely Byzantine manuscript, this ratio would work out to 0. Theoretically, a manuscript entirely free of Byzantine influence would have an infinite ratio, since it would have no Byzantine readings. In practice, of course, no manuscript will have an infinite ratio.

Though it turns out that very few have a ratio of even 1. Of the 102 test cases, only 12 -- B, 2427 (which we now know to be a forgery), ℵ, L, D, Θ, Ψ, 892, C, W, 1, and 33 -- have ratios of 1 or higher. The following table shows these manuscripts, with their Aland categories and their ratios:


At the other extreme, 461, 1251, and 1642 have a ratio of only 0.04. Of the manuscripts listed by the Alands, the sixteen with the lowest ratios are all Category V. The lowest ratio for a Type III manuscript is the 0.06 turned in by 1448. If we take the extremes for each tier, they are as follows:

CategoryHighest Ratio
for this category
ManuscriptLowest Ratio
for this category
I29.78B 11.52
III1.77Ψ 0.062495

The graph below shows the range of the ratios in each category:

Category Ratios

Thus it will be seen that every category except category I substantially overlaps the next category down. To some extent, to be sure, there is an explanation (e.g. the Alands call 579 a Category II only in Mark and Luke, so Matthew presumably has a higher fraction of Byzantine readings) -- but they do not break out the figures. In any case, the above result shows firmly the danger of relying on the Alands' subjective assessment rather than looking at the actual numbers -- or, better yet, the actual readings of the manuscript.

The median ratio for each category is:
Category I: 12.17
Category II: 1.54
Category III: 0.58
Category V: 0.09

So, in round numbers, a Category I manuscript is expected to be 90% non-Byzantine. A Category II is 60% non-Byzantine. A Category III is 30% non-Byzantine. And a Category V is 90% Byzantine.

But these are just the typical numbers. What we are interested in is the range. For the three categories that have enough manuscripts to allow meaningful samples (II, III, and V), then, let us look at the manuscripts within one (estimated) standard deviation -- i.e., in this case, the two-thirds of manuscripts closest to the mean.

CategoryHighest ratio
among middle 2/3
MS Lowest ratio
among middle 2/3
II1.91Θ 1.0333
Manuscript Categories

We can graph this data also. The graph at right plots the ratio of Byzantine to non-Byzantine readings of the manuscripts of Category III and Category V, counting the number of manuscripts in each block (grouping the manuscripts into blocks of .05, e.g. 0.000 to 0.049, 0.050 to 0.099, 0.100 to 0.149, etc.).

Both distributions follow a roughly normal curve, with that for the Category V manuscripts centered in the range 0.05-0.10 and that for the Category III manuscripts centered at 0.55-0.60, but we notice that the Category III curve is very flat and very spread out, and that there is a very large overlap between Category III and Category V -- confirming what we saw above in the graph of the extremes. That was not a fluke; the overlap between Category III and Category V is large; a better classification system would clearly have had a rigorous mathematical definition ("what the Alands think" is not a rigorous definition!) that would have drawn a clearer distinction.

Nonetheless, the idea behind distinction between Category III and Category V is clear (and those between Categories I, II, and III even clearer), even if the Alands' actual classification does not entirely conform to it. But this hazy distinction largely demonstrates the point I am are trying to make: The Category distinction is a grade distinction, not a clade distinction. That is, the Aland categories tell us effectively nothing about the actual ancestry of the manuscripts; they just tell us, within limits, how large is their Byzantine component. We can't tell if that Byzantine component is the result of direct descent from a Byzantine ancestor, or the result of mixture via correction, and we can't tell what other components, if any, the manuscript contains. This does not make the Categories useless -- but it does need to be kept in mind.

There is an interesting shift as we move into the Acts. In the Gospels, only 40 substantial manuscripts were Category III or higher. In Acts, despite a much smaller manuscript base, there are 58 substantial manuscripts of Category III or higher. There are 12 manuscripts the Alands call Category I or Category II, compared to nine in the Gospels. And, on the whole, these manuscripts appear to be better -- though this depends on the statistic you use. In the gospels, recall, the median Byzantine/non-Byzantine ratio for a Category II manuscript was 1.54; that for Category III was 0.58. In Acts, the median for Category II is still only 1.50 (statistically equivalent to the figure in the Gospels), and the median for Category III is 0.61. But if we take the table of extreme values, we find this:

CategoryHighest Ratio
for this category
ManuscriptLowest Ratio
for this category
I41.50B 2.1933

If we again look at the manuscripts within one (estimated) standard deviation -- i.e., in this case, the two-thirds of manuscripts closest to the mean -- we find

CategoryHighest ratio
among middle 2/3
MS Lowest ratio
among middle 2/3

The interesting observation is that the most Byzantine manuscripts of Acts actually show a more extreme fraction of Byzantine readings than those of the Gospels (though this may merely reflect on the readings the Alands chose), but the overall curve is clearly less Byzantine than in the Gospels.

In Paul, we have an astonishing 88 manuscripts of Category III or higher -- six of Category I (ℵ A B 33 1175 1739), ten of Category II (C D* F -- but not G! -- 81 256 1506 1881 1962 2127 2464), and 72 of Category III. However, it turns out that a number of these Category III manuscripts have very low ratios of non-Byzantine readings; it appears that the Alands classified them based on Acts and the Catholic Epistles and ignored the weaker text of Paul. Probably between eight and fifteen of them should be demoted.

Once again let's look at the extreme values for the manuscripts of each category:

CategoryHighest Ratio
for this category
ManuscriptLowest Ratio
for this category
I18.78B 1.321175

Again let's examine standard deviations:

CategoryHighest ratio
among middle 2/3
MS Lowest ratio
among middle 2/3

In the Catholics, we have 76 manuscripts of Category III or above -- nine (!) of Category I, 14 of Category II, and 53 of Category III. Here are the extreme values for each Category:

CategoryHighest Ratio
for this category
ManuscriptLowest Ratio
for this category
I90.0B 0.861175

And the manuscripts within one standard deviation:

CategoryHighest ratio
among middle 2/3
MS Lowest ratio
among middle 2/3

It's worth noting that, although B has the highest ratio of any manuscript in all four of these sections, the ratio varies by a factor of more than four from one section to another. It is unlikely that this is the result of any change in B; it is simply the nature of the Alands' (non-random) samples.

Appendix III: A Rigorous Classification

The idiosyncratic sample base described in the previous appendix, combined with the way the Alands present their numbers, makes it difficult to accurately classify a manuscript based on their data. There are really only two measures we have available to us. We can take the ratio of Type I to Type II readings, which is a prejudicial statistic because it assumes the UBS/GNT text is accurate (I flatly would refuse to touch such a statistic), or one which includes the Type S readings. The problem with Type S readings is that they include everything from scribal errors to readings of significant manuscript groupings. A Type S reading in a badly-copied manuscript like 28 may just be an error; a Type S reading in a good manuscript like 1739 is important for classification and may well be original. We simply cannot tell.

Still, the Aland data is what we have. We would like to get the best classification scheme we can based on it. A rigorous classification. For this purpose, what I will do is look at the ratio given above -- Byzantine to non-Byzantine readings -- and attempt a quick classification on this basis. Note that this is only a classification of independence from the Byzantine tradition; it makes no attempt to determine the actual nature of the manuscripts involved. What I have tried to do is find a natural gap in the data to roughly separate the four categories.


For the gospels, we have five manuscripts with a ratio greater than 3.25, and none between 1.91 and 3.25, so it seems obvious that manuscripts above 3.25 should be our "Category I." The five manuscripts involved are as follows (the figures in parenthesis are their Aland category and their ratio):

Mathematical Category I: ℵ (I: 11.52), B (I: 29.78), D (IV: 3.25), L (II: 3.63), 2427 (I: 12.17).

The gap between Category II and Category III is less obvious; we have a large gap from 1.33 to 1.74, a smaller one from 1.20 to 1.33, another from 1.08 to 1.20, and another from 0.91 to 0.81. Both the first and last gaps are tempting -- the first because it is so large, the latter because there really are manuscripts clumped above and below it. But if we chose the first gap, we would have only three Category II manuscripts. So I will choose a cutoff of 0.9, giving us this list instead:

Mathematical Category II: C (II: 1.33), W (III: 1.2), Δ (III: 0.97), Ψ (III: 1.78), Θ (II: 1.91), 1 (III: 1.08), 33 (II: 1.03), 565 (III: 0.91), 579 (II: 0.97), 892 (II: 1.74), 1342 (III: 0.93), 1582 (III: 0.96)

The largest gap below this is from 0.30 to 0.41. This seems to be the obvious cutoff for Category III. So:

Mathematical Category III: 13 (III: 0.57), 28 (III: 0.58), 69 (V: 0.54), 205 (III: 0.81), 209 (III: 0.78), 346 (III: 0.45), 543 (III: 0.58), 700 (III: 0.61), 788 (III: 0.7), 826 (III: 0.55), 828 (III: 0.61), 983 (III: 0.56), 1241 (III: 0.63), 1424Mark (III: 0.66), 1424MtLk (V: 0.41), 1542bMk (III: 0.61), 2193 (III: 0.54), 2542 (III: 0.71)

We note that every manuscript in this group except 69 and 1424MtLk is shown as Category III by the Alands. It is interesting to observe, however, that some relatively important manuscripts -- A N X 157 1071 -- fall below this threshold. It appears that the truly pure Byzantine manuscripts have a ratio less than about 0.15. So I would suggest that we define a Category IV, unlike the Aland Category IV, of manuscripts clearly Byzantine but with a significant number of interesting readings also:

Mathematical Category IV: A (V: 0.22), N (V: 0.26), X (V: 0.17), Σ (V: 0.29), Φ (V: 0.22), 0211 (V: 0.17), 0233 (III: 0.17), 61 (V: 0.2), 157 (III: 0.24), 1071 (III: 0.3), 1243 (III: 0.16), 1506 (V: 0.18), 2200 (V: 0.17)

It perhaps tells us something about how the Aland did their classifications that the uncials in this group are mostly Category V, the minuscules mostly Category III.

Finally, here are the manuscripts for which the Alands give statistics which are clearly Byzantine, with very little non-Byzantine text -- what the Alands would call Category V. I will call them Category B, for Byzantine.

Mathematical Category B: E (V: 0.05), F (V: 0.07), G (V: 0.14), H (V: 0.05), K (V: 0.12), M (V: 0.09), S (V: 0.08), U (V: 0.11), V (V: 0.13), Y (V: 0.05), Γ (V: 0.07), Λ (V: 0.05), Π (V: 0.15), Ω (V: 0.06), 047 (V: 0.15), 5 (V: 0.09), 6 (V: 0.07), 180 (V: 0.08), 189 (V: 0.05), 218 (V: 0.11), 263 (V: 0.05), 330 (V: 0.08), 365 (V: 0.07), 431 (V: 0.05), 461 (V: 0.04), 522 (V: 0.07), 597 (V: 0.06), 720 (V: 0.09), 886 (V: 0.08), 945 (V: 0.07), 1006 (V: 0.12), 1010 (V: 0.06), 1251 (V: 0.04), 1292 (V: 0.05), 1319 (V: 0.12), 1359 (V: 0.05), 1398 (V: 0.09), 1409 (V: 0.05), 1448 (III: 0.06), 1505 (V: 0.06), 1563 (V: 0.12), 1573 (V: 0.14), 1642 (V: 0.04), 1678 (III: 0.1), 1704 (V: 0.13), 2127 (V: 0.1), 2147 (V: 0.09), 2374 (V: 0.05), 2400 (V: 0.1), 2492 (V: 0.1), 2495 (III: 0.06), 2516 (V: 0.09), 2523 (V: 0.07), 2718 (III: 0.12)

The general soundness of the Aland classification is shown by the fact that, of these 54 manuscripts, 50 are Category V in their system. But four of them managed to be classified Category III.


Without going into detail of the process of determining the groups, here are the equivalent categories for Acts. We might note that the dividing line between categories III, IV, and B was much more blurry in this case than in the gospels; the cutoffs I used were somewhat arbitrary (determined in part by what I knew of the manuscripts rather than the numbers. The categories are still determined solely by the ratios, but the dividing line were chosen in part to put the largest fraction of manuscripts in the groups where they seemed to belong).

Mathematical Category I: ℵ (I: 7.82), A (I: 7.7), B (I: 41.5), C (II: 3.77), 81 (II: 6.43), 1175 (I: 3.61)

Mathematical Category II: D (IV: 2.29), E (II: 1.22), 33 (I: 2.19), 36 (II: 1.5), 181 (III: 1.47), 453 (III: 1.36), 610 (III: 1.48), 945 (III: 1.27), 1678 (III: 1.32), 1739 (II: 1.59), 1884 (III: 1.26), 1891 (II: 1.47), 2344 (III: 1.32)

Mathematical Category III: Ψ (III: 0.9), 5 (III: 0.41), 88 (III: 0.57), 94 (III: 0.87), 180 (III: 0.98), 307 (III: 0.52), 322 (III: 0.61), 323 (III: 0.61), 429 (III: 0.61), 431 (III: 0.9), 436 (III: 0.37), 441 (III: 0.65), 467 (V: 0.46), 522 (III: 0.54), 614 (III: 0.45), 621 (III: 0.49), 623 (III: 0.62), 629 (III: 0.89), 630 (III: 0.95), 915 (III: 0.43), 1292 (V: 0.36), 1409 (II: 1.08), 1505 (III: 0.44), 1611 (III: 0.42), 1642 (III: 0.96), 1704 (III: 1.1), 1751 (III: 0.74), 1838 (III: 0.36), 1842 (III: 0.51), 1875 (III: 1.09), 2138 (III: 0.55), 2200 (III: 1.00), 2298 (III: 0.71), 2412 (III: 0.43), 2495 (III: 0.47), 2718 (III: 0.51)

Mathematical Category IV: 6 (V: 0.27), 61 (V: 0.23), 69 (V: 0.21), 103 (V: 0.23), 104 (V: 0.28), 189 (V: 0.16), 206 (V: 0.3), 209 (V: 0.15), 218 (V: 0.17), 326 (III: 0.24), 459 (V: 0.22), 1243 (III: 0.18), 1319 (V: 0.27), 1359 (V: 0.17), 1718 (III: 0.23), 1735 (III: 0.31), 1852 (III: 0.33), 1877 (V: 0.18), 2147 (V: 0.26), 2544 (V: 0.19), 2652 (V: 0.28)

Mathematical Category B: H (V: 0.06), L (V: 0.06), P (V: 0.03), 049 (V: 0.09), 056 (V: 0.04), 0142 (V: 0.05), 1 (V: 0.03), 205 (V: 0.1), 254 (V: 0.04), 256 (V: 0.06), 263 (V: 0.05), 330 (V: 0.03), 365 (V: 0.1), 378 (V: 0.07), 424* (V: 0.04), 424c (V: 0.11), 451 (V: 0.04), 642 (V: 0.08), 911 (V: 0.04), 917 (V: 0.12), 1241 (V: 0.01), 1251 (V: 0.14), 1398 (V: 0.03), 1424 (V: 0.01), 1448 (V: 0.11), 1524 (V: 0.07), 1563 (V: 0.15), 1573 (V: 0.07), 1841 (V: 0.06), 1845 (III: 0.08), 1854 (V: 0.07), 1874 (V: 0.13), 2127 (V: 0.13), 2400 (V: 0.06), 2492 (V: 0.07), 2516 (V: 0.12), 2523 (V: 0.05), 2541 (V: 0.1)


Turning to Paul, the best cutoffs seemed to give the groups shown below. Note the large number of manuscripts with ratios above 3.0, giving us a very large class of Category I manuscripts. Nor is there much doubt that this is the location where the dividing line should be located, since the weakest of these manuscripts (which is, believe it or not, ℵ) has a ratio of 3.04, and the next manuscript (1881) has a ratio of 1.81. Paul probably qualifies as the one section of the New Testament where you could construct a fairly adequate text by looking only at Category I manuscripts. The flip side is that Category II is relatively small (and I was tempted to make it even smaller and draw the line at 1.5, which would have put only 1881 and 1506 in Category II).

Mathematical Category I: ℵ (I: 3.04), A (I: 10.32), B (I: 18.78), C (II: 4.17), D* (II: 3.63), F (II: 3.7), G (III: 3.5), 33 (I: 3.34), 81 (II: 3.86), 1739 (I: 4.6)

Mathematical Category II: P (III: 1.44), 256 (II: 1.09), 1175 (I: 1.32), 1506 (II: 1.75), 1881 (II: 1.81), 1962 (II: 1.04), 2127 (II: 1.11)

Mathematical Category III: D** (III: 0.57), Ψ (III: 0.58), 0150 (III: 0.87), 6 (III: 0.77), 104 (III: 0.70), 263 (III: 0.82), 365 (III: 0.87), 424** (III: 0.76), 436 (III: 0.60), 441 (III: 0.66), 442 (III: 0.97), 459 (III: 0.63), 467 (III: 0.56), 621 (III: 0.59), 1319 (III: 0.71), 1573 (III: 0.79), 1910 (III: 0.72), 1912 (III: 0.61), 1942 (III: 0.67), 1959 (III: 0.58), 2005 (III: 0.58), 2464 (II: 0.95)

Mathematical Category IV: 075 (III: 0.39), 5 (III: 0.22), 61 (III: 0.38), 69 (III: 0.34), 88 (III: 0.32), 103 (V: 0.17), 181 (III: 0.28), 218 (III: 0.35), 326 (III: 0.34), 330 (III: 0.36), 451 (III: 0.46), 623 (III: 0.22), 629 (III: 0.50), 630 (III: 0.49), 886 (V: 0.19), 915 (III: 0.37), 917 (III: 0.24), 1241 (III: 0.42), 1243 (III: 0.17), 1398 (III: 0.44), 1505 (III: 0.37), 1524 (V: 0.17), 1611 (III: 0.35), 1678 (III: 0.16), 1735 (III: 0.17), 1751 (III: 0.22), 1836 (III: 0.28), 1838 (III: 0.49), 1852 (III: 0.25), 1874 (III: 0.35), 1875 (III: 0.28), 1877 (III: 0.37), 1908 (III: 0.38), 2110 (III: 0.43), 2138 (III: 0.20), 2197 (V: 0.19), 2200 (III: 0.49), 2344 (III: 0.20), 2400 (V: 0.43), 2492 (III: 0.50), 2495 (III: 0.25), 2516 (III: 0.34), 2523 (III: 0.33), 2544 (III: 0.31)

Mathematical Category B: K (V: 0.12), L (V: 0.05), 049 (V: 0.03), 056 (V: 0.07), 0142 (V: 0.06), 0151 (V: 0.09), 1 (V: 0.03), 94 (III: 0.15), 180 (V: 0.04), 189 (V: 0.03), 205 (V: 0.05), 206 (V: 0.10), 209 (V: 0.06), 254 (V: 0.14), 322 (III: 0.08), 323 (III: 0.08), 378 (V: 0.03), 398 (V: 0.02), 424* (V: 0.04), 429 (V: 0.10), 431 (V: 0.04), 522 (V: 0.04), 614 (III: 0.04), 642 (V: 0.06), 720 (V: 0.12), 911 (V: 0.02), 918 (V: 0.04), 945 (V: 0.04), 1251 (V: 0.11), 1292 (V: 0.04), 1359 (V: 0.10), 1409 (V: 0.04), 1424 (V: 0.05), 1448 (V: 0.04), 1523 (V: 0.15), 1563 (III: 0.15), 1642 (V: 0.08), 1704 (V: 0.03), 1718 (III: 0.14), 1841 (V: 0.00), 1845 (III: 0.09), 1846 (III: 0.08), 1854 (V: 0.03), 1891 (V: 0.04), 2147 (V: 0.04), 2298 (V: 0.05), 2374 (V: 0.04), 2412 (III: 0.03), 2541 (V: 0.03), 2652 (V: 0.02), 2718 (III: 0.08)

We might note, incidentally, the danger that simple categorization causes. An example is 630. For Paul as a whole, e.g., 630 shows up in Category IV. But in fact it is block mixed (or progressively mixed, or something). In the early part of Paul, it is weak Family 1739, which would surely make it Category III. From about Ephesians on, it is purely Byzantine. So, properly, we should list it as III/B. No doubt there are other instances of this as well; we simply cannot tell from the Aland numbers.

Catholic Epistles

Finally, here is how things appear to break down for the Catholic Epistles. We again have a very large number of Category I witnesses, but there really isn't much doubt about this dividing line, since the weakest of these witnesses, 33, is at 2.95 and the next-best witness, 323, is at 2.37.

It might be worth noting that, even within Category I, there appears to be a bit of a gap: B is at 90.00 (!), 1739 at 5.40, and then the other nine witnesses I've grouped here are between 4.17 and 2.95. Thus B and 1739 stand far away from the pack. It is worth noting that, although we have about as many Category I witnesses here as in Paul, they do not represent the full range of manuscripts nearly as well. The members of Family 2138 -- a very distinct and important group -- are all mixed enough that none of them reaches Category I status. Indeed, it is arguable that none of them deserve Category II status. There were two possible gaps to define Category II: Between 81 (1.97) and 1505 (1.69), or between 2138 (1.42) and 1067 (1.27). The former gap is larger, but it would leave only four manucripts in Category II (81, 322, 323, 2344), so I chose the latter gap (which had the secondary effect of putting several Family 2138 manuscripts, including 1505, 2138, and 2495, in Category II). But this is arbitrary; if you're willing to allow more than four non-Byzantine classses, there could be a cut between 81 and 1505.

Mathematical Category I: ℵ (I: 3.43), A (I: 4.17), B (I: 90.00), C (II: 3.44), Ψ (II: 3.24), 33 (I: 2.95), 1241 (I: 4.12), 1243 (I: 3.18), 1739 (I: 5.40), 1852 (II: 4.00), 1881 (II: 3.38)

Mathematical Category II: 81 (II: 1.97), 322 (II: 2.37), 323 (II: 2.37), 1505 (III: 1.69), 1735 (II: 1.60), 2138 (III: 1.42), 2298 (II: 1.57), 2344 (I: 2.33), 2464 (II: 1.53), 2495 (III: 1.51)

Mathematical Category III: 5 (III: 0.94), 436 (III: 1.19), 442 (II: 1.19), 614 (III: 0.93), 621 (III: 0.96), 623 (III: 1.04), 630 (III: 1.00), 945 (III: 1.24), 1067 (II: 1.27), 1175 (I: 0.86), 1292 (II: 1.00), 1409 (II: 1.02), 2200 (III: 1.00), 2412 (III: 1.04), 2541 (III: 0.90)

Mathematical Category IV: P (III: 0.73), 6 (III: 0.60), 36 (III: 0.48), 61 (III: 0.50), 69 (V: 0.22), 88 (III: 0.42), 94 (III: 0.36), 104 (III: 0.46), 181 (III: 0.24), 206 (III: 0.53), 218 (III: 0.40), 254 (III: 0.47), 307 (III: 0.52), 378 (III: 0.56), 398 (III: 0.28), 424** (III: 0.48), 429 (III: 0.60), 431 (III: 0.29), 453 (III: 0.49), 467 (V: 0.30), 522 (III: 0.70), 629 (III: 0.80), 642 (III: 0.41), 720 (V: 0.32), 915 (III: 0.45), 918 (III: 0.48), 1359 (III: 0.45), 1448 (III: 0.47), 1524 (III: 0.48), 1563 (V: 0.43), 1678 (III: 0.52), 1718 (III: 0.47), 1751 (III: 0.24), 1838 (III: 0.48), 1842 (III: 0.31), 1845 (III: 0.53), 1875 (III: 0.23), 2147 (III: 0.64), 2197 (III: 0.51), 2374 (III: 0.59), 2492 (III: 0.47), 2544 (V: 0.29), 2652 (III: 0.64), 2718 (III: 0.34)

Mathematical Category B: K (V: 0.10), L (V: 0.10), 049 (V: 0.05), 056 (V: 0.14), 0142 (V: 0.14), 1 (V: 0.03), 103 (V: 0.11), 180 (V: 0.13), 189 (V: 0.07), 205 (V: 0.06), 209 (V: 0.1), 256 (V: 0.07), 263 (V: 0.06), 330 (V: 0.1), 365 (V: 0.15), 424* (V: 0.01), 451 (V: 0.10), 610 (V: 0.08), 911 (V: 0.04), 917 (V: 0.11), 1251 (V: 0.07), 1319 (V: 0.1), 1398 (V: 0.07), 1424 (V: 0.06), 1573 (V: 0.05), 1642 (V: 0.07), 1704 (V: 0.02), 1841 (III: 0.04), 1854 (V: 0.05), 1874 (V: 0.13), 1877 (V: 0.10), 1891 (V: 0.06), 2127 (V: 0.11), 2400 (V: 0.04), 2516 (V: 0.03), 2523 (V: 0.14)

The lists above, of course, include only the few hundred manuscripts for which the Alands supply data. They either do not supply data for the remaining manuscripts, or the manuscripts are too fragmentary for the data to be meaningful. The manuscripts for which they did not supply data are generally either unclassified or Category V. The above data shows that there is some overlap between what should be Category III and Category V (e.g., in the Catholics, there are thee manuscripts in Category IV which the Alands make Category V, and one in Category B which they list as Category III). But their accuracy rate is on the order of 85%, and it is very rare for them to miss by more than one category (except in the handful of cases where they apply one category to a manuscript which belongs in different categories in different sections). Thus it seems likely that the manuscripts they list in Category V can be safely ignored and represented by a sample. The trick remains to choose between the several hundred manuscripts of Category IV and higher.

I must stress that this is not the final word. The Aland samples are too small to be entirely reliable, especially if a manuscript is block-mixed, and the classifications above are based on only a single statistical measure, which is imperfect because of the difference between meaningful and meaningless non-type-1/2 readings. But it is at least a measure of Byzantine-ness based solely on mathematics.

A footnote: Some may object to my seeking gaps to define the differences between categories, pointing out -- correctly -- that I have elsewhere denied the existence or significance in gaps of percentage agreements.

The situations, however, are not parallel. It is, yet again, a distinction between grades and clades. A text-type, as I define the term, is a clade, so percentages and gaps are not relevant.

But two of the bare handful of assured results of New Testament TC is the existence of the Byzantine Text and of mixture. This makes it meaningful to attempt to assess the degree of Byzantine mixture in a manuscript -- and, while the Aland data does not allow us to really determine genetic ancestry, it is generally enough to determine degree of Byzantine influence.

This is a grade distinction, nothing more: All we are seeking is percent of Byzantine readings. In that context, we need dividing lines between categories. We could of course be arbitrary; there is in this case no real problem with that. But since there are gaps (at least some of them), placing our category divisions within those gaps makes the distinctions between categories more distinctive. So I tried to find suitable gaps.