Manuscript Collation

Contents: Introduction * Samples of Collations * Long and Short Collations * What we learn from collations * Collations in Other Languages * The physical task of collating

Introduction

The manuscript collation is perhaps the most fundamental of all the tool of textual criticism -- the essential source of the data of the discipline.

We should perhaps start with a warning -- "collation" means different things in different contexts. To a bibliographer or librarian, it means the arrangement of pages in a manuscript, which we will talk about below. But to a New Testament scholar, the term "collation" is an abbreviated record of the readings of a manuscript. Except when otherwise specified, that is what this article means by a collation.

The purpose of a collation is to transmit all basic information about the text of a manuscript without publishing the text of the manuscript in full. It does this by comparing a manuscript against a standard printed edition (usually the Textus Receptus) and noting all "significant" differences. The amount of space this can save is tremendous. The collation of 1739 by Lake and New, for instance, requires 24 pages to cover all of Paul, when printed in large print. The Nestle-Aland twenty-seventh edition, printed in small print, requires 179 pages for the same books. Even allowing for the space required for the the critical apparatus of the Nestle text, this is a savings of at least a factor of five. And this for a manuscript with a relatively large number of deviations from the Textus Receptus! A Byzantine manuscript of the same books would result in a much shorter collation.

There is, unfortunately, no universally recognized standard method of collation even for Greek New Testament manuscripts, let alone for texts of other works. Different transcribers use different base editions, and have different styles of collation. The problem of base editions is probably beyond solution; the edition generally regarded as standard (the 1873 Oxford edition of the Textus Receptus for the Greek New Testament) has been out of print for a very long time, and no new standard is emerging. (Latin scholars are slightly better off; the Clementine Vulgate is the sort-of standard.) Some have proposed collating Greek manuscripts against the United Bible Societies text, but this would mean that older and newer collations would be based on different texts -- a notion unfortunate enough that collations against the TR will probably continue for the foreseeable future. The TR also has the advantage of being a relatively Byzantine text, so that it takes relatively little space to collate Byzantine manuscripts against it (which also reduces the effort needed for the collation, which in turn probably reduces the number of errors). There are reportedly online versions of the Oxford text which can be used for collation; I cannot vouch for their accuracy.

The form of collations is somewhat more standardized, though not perfectly so. In general, a collation consists of a series of variations recorded in the following form: Chapter and verse number, lemma (the text of the proof edition), and the variant (the text of the manuscript). The text of the lemma and the variant are normally separated by a square bracket, thus: ]

So, for instance, the first variation in the Nestle-Aland apparatus occurs in Matthew 1:3. Here the majority of witnesses, including the Textus Receptus, read Ζαρα. In B, however, we read instead ΖΑΡΕ. So the collation of B against the Textus Receptus would read

1:1 Ζαρα ] ΖΑΡΕ

There are, of course, variations on the format of collations (far more than there should be, because everyone seems to think that their own way is "best" even though what is best would be standardization!); see the section on Samples of Collations. The most common variation in collation format involves omissions. For instance, in Mark 1:1, 28 (and several other manuscripts) omit the words υιου θεου. This may be noted in several ways, e.g.

1:1 υιου θεου ] OM. (I'd consider this the standard way), or
1:1 OM. υιου θεου or
1:1 - υιου θεου

It is also quite common to see changes in word order marked ~. Ideally (to prevent ambiguity) both the word order of the collation base and the reading of the manuscript should be noted. You may also see "+" or "add" for additions to the text and "-" for omissions.

If a manuscript has been corrected, these readings should be noted. The reading of the original hand should, of course, be marked with the asterisk (*). If there are multiple correctors, care should be taken to distinguish them. Some collations will include readings of the correctors in the body of the collation; others add them as comments. Which is more effective may depend on the frequency and nature of the corrections.

Editors disagree about the exact amount of detail to be recorded in a collation. Some, e.g., would include variations involving nu movable, while others would omit it. Most would exclude punctuation, since this is known not to be original. (Though some would include make notes of punctuation marks in manuscripts which rarely include it.) Itacisms are also frequently excluded (although if they are very frequent, it should probably be noted in a comment on the collation). The use of Nomina Sacra normally is not noted unless an abnormal form is used or in some way it affects the interpretation. But there are no hard and fast rules -- except two: First, a collation should announce at the beginning what features it does and does not include, and second, if a reading may or may not have textual significance (e.g. in the case of an itacism), it must be noted. To put it another way: When in doubt, note the variant.

In general, one should try to collate "whole variants" -- that is, if two consecutive words form a logical entity, one should record changes to both together, but if they are unrelated, treat them as two different variants.

Another difficulty arises with damaged texts. One needs a way to indicate both completely illegible letters (e.g. where there is a hole in the page containing a whole letter) and partially legible letters. The notation for the former is usually a dot (e.g. Λ . ΓΕΙ indicates a λ followed by a defect large enough for one letter, then γει If the defect is large enough for two letters, one uses two dots, etc (e.g. Λ . ΓΕΙ is ΛΕΓΕΙ with one letter missing, while "Λ . . ΕΙ" would be the same word with a gap of two letters, etc.). Gaps of more than a few letters are often marked in the margin (e.g. if a manuscript were defective for the first verse of John's gospel, we would say something like "N.B. MS. defective for εν... ην o λογος).

A notation is also needed for a partially legible letter (and such are common; suppose a page has lost a margin, and the last thing at the edge of the page is a vertical stroke |. Depending on how the scribe wrote, this stroke could be a portion of any number of letters, e.g. Γ Η Ι Κ Μ Ν Π Ρ. The standard notation in such cases has been for the collator to guess what the letter probably was, then mark it with a dot below the letter. As this for a long time was difficult or impossible to do in electronic formats, other solutions have been devised, such as placing the letter in parenthesis or in some sort of symbolic notation (the COLLATE program uses a tag pair, [ut]...[/ut]). This should be made clear in the introduction to the collation. (And it should be repeated that this information must be provided. Printing a reconstructed text without noting this fact is purely inexcusable. Indeed, if there is any real doubt about the letter in the manuscript, and multiple readings are possible, these should probably be noted in the margin.) Fortunately, it is now possible to use the unicode feature known as "combining diacritical marks" to put dots below letters. So if we wished to indicate that the second and third letters of λεγει are doubtful, we can show it as λε̣γ̣ει.

A good collation should probably also be prefaced by information about the manuscript -- e.g. a list of lacunae (even though these will also be noted in the body of the collations), characteristics of the scribe, description of non-Biblical materials included in the volume. This information may not be of significance for the text, but it may well indicate something about the history of the volume -- which, in turn, may provide clues about the text in the book.

It is possible to collate multiple manuscripts in one collation -- indeed, very many collations follow this format, as it saves space. One simply notes which manuscripts have which readings by listing them after the variant. So, for example, the first few lines of Clark's collations of 223, 876, 1022, 1799, 1960, 2401, 2412, and 2423 in 1 Thessalonians reads:

1:1 θεσσαλονικαιων 223, θεσσαλονικεων 1022      θεω + και 876

1:5 υμας(1) ] ημας 1960      -εν (3) 1022 2423**

Thus we see that, in 1 Thessalonians 1:1, 223 and 1022 have various (mis)spellings for θεσσαλονικαιων; the other manuscripts (876, 1960, 2401, 2412, and 2423; 1799 is defective here) agree with the reading of the Textus Receptus. Later in the verse, 876 has θεω και πατρι for the θεω πατρι of the other manuscripts and the TR. From there on, all the manuscripts agree with the TR until 1:5, where 1960 reads εις ημας for the εις υμας of the other manuscripts and the TR. Finally, where most of the manuscripts read και εν πνευματι αγιου, 1022 and the corrector of 2423 omit εν.

Beyond this, the only way to get a feeling for collations is to work with them. The following samples provide a very brief introduction to this process....

Samples of Collations

The table below shows several samples of collations, all taken directly from published and relatively widely available editions of manuscripts. The first column of the table shows the text of Ephesians 1:1-6 as found in the Textus Receptus. The next three columns show the texts of manuscripts 330, 1739, and 1799 (taken, respectively, from the collations published by Davies, Lake and New, and Clark). The differences from the TR text are shown in bold (with omissions being marked [--], in order to make the omissions obvious). This is followed by the actual text of the collations (sometimes with some extraneous material about other manuscripts removed), so that the reader can see how each of these three collators approached their task. (Of the three, the collation of 330 by Davies is much the most idiosyncratic.) The verse numbers of course do not occur in the manuscript.

Davies's collation of 330 (without the collations of 436, 462, 2344):
1. ~ χριστου α. ιησου.
   OM. ιησου².
6. ης / εν η.
   + υιω αυτου ρ. ηγαπμενω
N.B.: In this collation, / replaces ] and the lemma appears after rather than before the slash. (This takes a great deal of getting used to! -- which is a good demonstration of why collation formats should be standardized.) Also, the abbreviation α. is used for "before"; ρ. stands for "after." The symbol "~" is used here (as often elsewhere) for a change in word order.

Lake and New's collation of 1739:
i.1. om εν εφεσω
  2 om χαρις .... χριστου
  3 χριστω praem εν
  6 εν η ] ης
N.B.: The notation praem means "add before the lemma" or "is prefixed by." Similar Latin notations may be encountered elsewhere.

Clark's collation of 1799:
3. +αδελφοι [ ευλογητος
6 - εις επαινον δοξης της χαριτος αυτου
N.B.: This collation uses both [ and ]. [ indicates an insertion before the word listed in the lemma. Note also the use of + to indicate an addition and - for an omission

Note incidentally that the collations of both Davies and Clark were done with typewriters rather than being typeset. This will affect just what information can be included.

One other thing to consider in preparing a collation of more than one manuscript is what symbol to use for the manuscript. The obvious temptation is to use its Kurzgefaste Liste number, as is done in most critical editions. But this is not universal. British editions, in particular, have a tendency to refer to manuscripts by letter. This has the advantage of brevity, but it makes for a lot more work!

For example, here are the first five verses of Matthew as collated by Scrivener in A Full and Exact Collation of about Twenty Greek Manuscripts of the Holy Gospels:

CAPUT I. vv. 1,2 ἁβρααμ hlmnpqrsx alii.   v. 2. ιουδα y.   v. 3. - τον secund. y   vv 4, 5. σαλωμ c.   v. 5. ῥαχαμ cp. ιωβηδ bis cg. ωβηθ prim. r. οβηδ bis y.

Scrivener does not give manuscript numbers -- not even Scrivener/Burgon numbers! -- for his manuscripts, but as best I can tell, they are:
a=470
b=471
c=472
d=473
e=474
f=475
g=71
h=476
j=N (022) (British Library Cotton Collection fragment only)
k=478
l=479
m=201
n=480
o=481
p=482
q=483
r=484
s=485
t=486
u=487 (lost)
v=488 (lost)
x=𝓁¹⁸³
y=𝓁¹⁸⁴

So this collation, if done my way, would look like this (note: I'm not making up the accents on Αβρααμ in verses 1 and 2; those are straight out of Scrivener's edition of the Textus Receptus):

1:1 ’Αβραάμ ] ἁβρααμ 201 476 479 480 482 483 484 485 𝓁¹⁸³
1:2 ’Αβραὰμ ] ἁβρααμ 201 476 479 480 482 483 484 485 𝓁¹⁸³
    Ιουδαν ] ιουδα 𝓁¹⁸⁴
1:3 τον Ζαρα ] omit τον 𝓁¹⁸⁴
1:4 σαλμων ] σαλωμ 472
1:5 σαλμων ] σαλωμ 472
     ραχαβ ] ῥαχαμ 472 482
     ωβηδ ... ωβηδ ] ιωβηδ ... ιωβηδ 71 472; οβηδ ... ωβηδ 484; οβηδ ... οβηδ 𝓁¹⁸⁴

This is obviously a lot longer. Some of that is simply the more explicit collation format. But even if we took Scrivener's format and simply converted the letters to proper numbers, we'd get

CAPUT I. vv. 1,2 ἁβρααμ ἁβρααμ 201 476 479 480 482 483 484 485 𝓁¹⁸³ alii.   v. 2. ιουδα 𝓁¹⁸⁴.   v. 3. - τον secund. 𝓁¹⁸⁴   vv 4, 5. σαλωμ 472.   v. 5. ῥαχαμ 472 482. ιωβηδ bis 71 47. ωβηθ prim. 484. οβηδ bis 𝓁¹⁸⁴.

Thus using single letters rather than giving the full symbols saves about 25% of the size of the apparatus. Whether the extra effort it entails is worth the savings is, I suppose, something the reader must decide.

For a fuller sample of a collation, one is invited to examine the author's own PDF collation (or, rather, a collation I combined from various sources) of 0243 and fourteen other manuscripts of Hebrews.

Long and Short Collations

The descriptions above are of simple text collations. That is, all they contain is the comparison of two sources (usually a manuscript and a printed text). This we might label a short collation. Or it might be called a collation of the text -- as opposed to a collation of the manuscript. A collation of a manuscript is a description of the physical properties of the manuscript.

To fully describe a manuscript calls for some additional information -- data about the form of the manuscript, in addition to (or apart from) its textual content. Montague Rhodes James, who did an amazing job of cataloging most of the manuscripts in Cambridge University, its colleges, and some other libraries, lists the following information which should also be supplied:

1. Material -- i.e. papyrus, parchment, paper, or a mixture. It is also worth noting the arrangement of the material. Paper and parchment, if mixed at all, were generally mixed in a regular pattern, and parchment, if used throughout, will usually be arranged so hair side faces hair side and flesh faces flesh. If there is a disruption of this pattern, it may indicate missing material or a change in plan -- or even a combination of two different pre-existing manuscripts.
   If the material is paper, and the paper is recent enough to have watermarks (rare but not unknown for NT manuscripts), the watermarks should be listed, as they are indications of date and provenance. Note that every page needs to be checked for watermarks, as the paper may well come from different lots.
2. Description of the page -- the size (height and width), the manner of ruling, and the number of lines per page. Also the method used for ruling: a dry point (which leaves grooves but no marks) or a plummet, which leaves a faint brown line. One should note whether it is rules on both sides or just one, and whether both margines were ruled or just one (sometimes a scribe would rule two leaves at once, allowing one margin to serve both, and a ruling on one side of the page might be deep enough to show through on the other side.)
3. Binding -- both the modern binding and any indications of an earlier binding (e.g. if a book was once chained, but is so no longer, there will likely be traces of rust on the outer pages. If a book has been rebound and trimmed, marginal material may be cut off).
4. Number of (surviving) leaves. Also, if the leaves are numbered, the number which appear to have been initially present. Be sure to clarify whether leaves, quires, or folios are numbered -- folio numbering was the most common.
5. Arrangement of quires. Although most books use a four-sheet standard (meaning each quire has eight leaves of 16 sheets), it is the opinion of James that the majority of books have at least some irregularity -- leaves cut out, or an extra leaf inserted into a quire at some point other than the middle. In counting the size of quires, the first thing to look for is of course the string used to bind the quires into a volume. There may also be catch phrases in the extreme margin, showing the last word on the page and/or the first on the next page, to help the binder organize the quires -- but these are often cut off after binding, so they are not to be relied upon.
   The standard notation for a quire of a given length is xⁿ, where x is a letter denoting the quire, and n is the number of leaves (not pages or sheets) in the quire. So a description a⁸b⁶ means that we have a codex which now has two quires, the first of eight leaves (four sheets) and the second of six leaves (three sheets). If a quire is described as "(wants n)," it means that leaf n has gone; if it has afterward "(+n* text)" it means that a leaf has been added after leaf m.
   So here might be a typical example of this sort of collation:
   a⁸(wants 1) b⁸ c⁸ d⁸(wants 3) e⁸(+6* εαν) f⁴
   The above means that we have a codex of six quires, the first five of which (numbered a, b, c, d, e) originally had had eight leaves (four sheets, 16 pages) and the last of which (f) had four leaves (2 sheets, 8 pages). The first quire has lost its first leaf (not a rare occurrence). In other words, it has lost pages 1 and 2. Quires b and c are intact. Quire d has lost its third leaf -- in other words, the fifth and sixth pages of this quire (which would be pages 53 and 54 of the manuscript as a whole). The fifth quire, e, also originally had eight sheets, but after the sixth sheet, an additional leaf has been inserted, which begins with the word εαν. The final quire, f, is apparently intact but has only four leaves (two sheets). Thus the total manuscript started with 44 leaves (22 sheets, 88 pages), but two leaves have been lost and one additional leaf added.
6. History. Any indications of ownership or past location. This of course includes scribal colophons -- but also information in the binding or leaves at the front. Many owners will have written at least their name in the manuscript, often with a date and perhaps a place. These will generally be reliable (unlike colophons, which may be fake), and will give a latest possible date for a manuscript, as well as perhaps hinting at the place where the manuscript was written. It is suggested that all such information be recorded even if it does not appear to have much use; later scholars might have access to historical data the collator does not.
7. Contents. Although some manuscripts contain nothing more than the Gospels or Epistles or whatnot, most will contain at least some additional material -- prologues, Eusebian tables, something. All such elements should be listed even if their texts are not collated.
8. Illustrations. To truly catalog the illustrations in a manuscript requires a specialist, but some information should be given: The dimensions of the illustration (not necessarily in inches or centimeters, but as a fraction of the page -- e.g. full width, half height). The general subject of the illustration should be described if it can be determined, as well as characters in the picture and their clothing. Also, many illustrations will have some sort of background (white, red and white squares, blue and green lozenges). This is often characteristic of a particular school, and should be mentioned. If any lettering is found in the illustration, it should be noted.
   In New Testament manuscripts, the most common illustrations are of course of the Evangelists. But several things should still be noted -- e.g. whether they are shown in company with their symbolic representations (man, ox, lion, eagle), whether they are shown in contemporary garb or in what a scribe might think was Palestinian clothing, how they are writing (sitting or standing, at a desk or somewhere else), and whether they are alone or in company (e.g. since John is sometimes said to have dictated his gospel to Prochorus, it is not unusual to see him alongside another person). It seems to me in addition that halos are more common in late illustrations than early, although I have not tested this formally.
   Non-Biblical manuscripts have a greater variety of illustrations -- e.g. our only illustrations of Chaucer are in Chaucerian manuscripts, but generally he does not take up the whole width of the page; the image is shown in a sort of a sidebar, perhaps with the poet pointing to a particularly interesting passage. Information about non-stereotyped illustrations of this sort can be particularly helpful.

In addition, in describing a printed book, one should supply a complete description of the title page -- ideally including a photograph, but it should also include a complete version of the text, with formatting descriptions, to allow electronic searching. This is because it was not uncommon to see a book printed by one printer but published by several publishers, with no changes except to the title page. To properly sort through the editions, proper title page data is vital.

Description of the binding is also important. Sadly, many important volumes have lost their original bindings and even had their pages trimmed by re-binders, but if an ancient binding survives, it should be described in detail (including material (both type, such as leather, and source, such as the type of animal, if it can be determined), color, any tooling, clasps, inlays, ties, ornaments, gold leaf, and anything else that is in any way unusual).

Note that this sort of collation can be offered for printed books as well as manuscripts. It's meaningless for cheap paperbacks, with glued bindings, but most hardcover books still have sewn bindings, which can be informative. For example, my 1939 printing of the Loeb Classical Library edition of Tacitus's Agricola, Germania, and Dialogue on Oratory (translated by Sir William Peterson) appears at first examination (I did this casually) to have a 6-leaf opening quire, then 25 quires (labelled A-Z at the bottom of the first page, omitting U/V and W) of 8 leaves each, then a closing quire of 8 leaves. The first leaf of the first quire is pasted to the front cover, and the last leaf of the final quire is pasted to the back cover. The fifth leaf of the front quire has been cut out and a picture of a bust of Tacitus glued to the stub; similarly the second (seventh-from-the-end) leaf of the last quire has been cut out and replaced by a map. So this collation would be something like
a⁶(wants 5)(+5* image of Tacitus)b-y⁸z⁸(wants 2)(2* map)
However, since the all but the first and last quires have printed quire letters, we could respect that by giving the collation as
a⁶(wants 5)(+5* image of Tacitus)A-U⁸X⁸Y⁸Z⁸zz⁸(wants 2)(2* map)
We might also note the pastedowns, as well as mentioning that several of the leaves are blank flyleaves.
My personal copy was previously owned, and there are notations from a previous owner, as well as sale marks, so if for some reason we really wanted to describe this particular copy, we might note the bookplate in the front cover, an inscription in green ink, a seller's note, etc.

We certainly see this information in some accounts of New Testament manuscripts. For example, Clark gave a description of 1960 which seems to imply this collation (note that the quires are assigned Greek numerals):
lacks α-ιε, ις⁸ιζ⁸(lacks 8)ιη⁸(lacks 1, 8)ιθ-κγ⁸κδ⁶κε-κζ⁸(lacks κη)κθ⁸λ⁴

This description instantly makes clear that the volume is extremely defective, lacking roughly the entire first half plus some leaves after that. Of course, in this case, Clark's description makes that clear (I derived the information from Clark's description!) -- but having it set out in consistent form makes it easier to compare manuscripts.

What we learn from collations

It may seem that working with collations is a rather specialized task, and that the use of a critical apparatus is enough for the ordinary student. This is true in some instances, but much oversimplified. A collation can teach us a great deal about a manuscript that cannot be learned from the apparatus criticus.

The collation, unlike the apparatus, teaches us something about the nature of the manuscript itself (especially if it includes the format information above, but the statement is true even if we don't have the quire information). If we examine the collation of Hebrews, for instance, we observe that Codex Claromontanus (D) regularly confuses the endings -θε and -θαι, even when there is no variation in the other manuscripts. We learn, therefore, that Claromontanus has no authority when there are genuine variants of this type.

Most manuscripts have some such idiosyncrasies. ℵ, for instance, regularly confounds ΕΙ and Ι, while 056 and 0142 have a habit (derived probably from their common ancestor) of adding extra iotas. 1799 regularly inserts αδελφοι into texts (probably based on the lectionary), and so is unreliable for the handful of legitimate variants involving this word. And so forth. None of these facts can be learned from a critical apparatus, since it doesn't include such variants, and most are quickly obvious in a collation.

In addition, a collation is a complete catalog of the readings of a manuscript, whereas a critical apparatus is always limited. As an example, consider the collation of Hebrews already cited above. This collation includes fifteen manuscripts, and shows 61 variants in Hebrews 1. The Nestle-Aland text, by contrast, cites only 21 variants, despite having 23 so-called "constant witnesses." Similarly, the collation of Colossians 1 below cites 115 variants; NA²⁸ cites 37, one of them a variant interpretation rather than a reading of the text. Thus, as a rule of thumb, NA²⁸ cites only about a third of the variants in its manuscripts (although it is more complete for manuscripts like B and ℵ). Most of the extra variants in the collation are, of course, trivial -- spelling mistakes and the like -- but by working with the critical apparatus rather than the collations, one forfeits the ability to decide which variants are important. In addition, most critical apparatus have an associated critical text. This critical text will, almost inevitably, bias the user toward its readings. Whereas a collation, since it is based on a non-critical text (the Textus Receptus), should not result in any pre-judgement of the readings.

As a final exercise, let's do a full critical apparatus as a collation rather than as an apparatus. I'm going to do Colossians chapter 1. The base text is the TR as represented in the apparatus of Hodges and Farstad. All variants are noted except itacisms and nu-movable (which one might collate those for one manuscript, but not for this many!). I will collate 𝔓⁴⁶ ℵ A B C D F G K L Ψ 049 056 0142 223 330 436 876 1960 2412 plus the UBS (𝖀) and Hodges and Farstad (𝕳) editions. Note that this is my style of collation, which is intended to be very clear about the various readings -- as a result, I often quote a longer lemma than many collators.

Note also the instance in 1:6 where I cite a few witnesses in support of the lemma rather than against it. This is to clarify the exact readings of those witnesses. In 1:12, I cite only 𝕳^txt in support of the lemma, and similarly 𝕳^marg in 1:14. Note that these are not the only witnesses supporting the lemma; all witnesses not cited against the lemma support it. But these witnesses are cited explicitly to make their readings clear.

This gives us a lot of information not as easily noticed in a standard critical apparatus. The kinship of F and G, for instance, is easily observed, and the curious sloppiness of 330, or its exemplar, is also evident. On the other hand, it is surprisingly difficult to detect Byzantine sub-types. For instance, 223 is believed to belong to von Soden's K^c group, and 1960 goes with K^r. But 223 has only five differences from the TR in this chapters, some of them shared with the mass of the Byzantine text, and 1960 has only five differences, one of them the work of a corrector. So there can be no instant classification of their sub-types. We simply need more text.

There is one other thing we can do with a collation, particularly a collation of many texts such as this one, and that is to get a picture of just how erratic a manuscript is, by looking at how many odd readings it has. These are readings that do not show up in the critical apparatus. The list below takes each of out manuscripts and lists how many readings it has that are singular, how many have just one supporter (with the supporter), and how many have just two supporters (with the supporters)

This sort of table lets us discover many things quickly. We note, for instance, the high number of singular and rare readings in 𝔓⁴⁶ and 330. The overwhelming evidence is that 𝔓⁴⁶ is simply a wild copy. But 330... unfortunately, we can't tell from this how many of its readings are truly singular and how many would be shared by its relatives 451 and 2492. But based on this evidence, we would be inclined to think it like 𝔓⁴⁶, although to a far lesser degree: valuable and interesting but unreliable. Ψ, which is mostly Byzantine, also stands out in this chapter for its high rate of unusual readings -- something that is not evident from the Nestle apparatus.

We can also see close kinship easily. Observe for instance 056 and 0142. There are four readings which they share only with each other (plus one they share with each other and 876), and no other rare readings at all except for 0142's two singular readings! Their low rare-reading totals make it clear that they are Byzantine, but they represent a peculiar type within the Byzantine text.

Even more obvious is the kinship between F and G, as demonstrated by their fourteen agreements only with each other plus the three agreements of D+F+G. You can't get this from the Nestle apparatus; while it marks all the D+F+G readings, only one of the F+G special readings (1:12) is marked. It's true that these readings are extremely unlikely to be original, and so aren't of great importance in determining the text, but the point is, the kinship of F and G is much easier to detect with the collations than from the critical apparatus.

Of course, one chapter does not a statistically significant sample make. For both 056/0142 and for Ψ, we should probably do other tests.

In the case of 056 and 0142, I will take all chapters of Colossians, not just chapter 1, and look for places where 056 and 0142 agree with no papyrus or uncial support, then I will look for other supporters among the minuscules. Similarly for Ψ. The readings of the papyri and uncials are taken from the relevant volume of Das Neue Testament auf Papyrus. (Note that the number of singular agreements here is smaller than that in the above table because some of the singulars have supporters not included in the table.)

Special Agreements of 056 0142 with no papyrus/uncial support

• Col. 1:2 πατρος και κυριου Ιησου Χριστου
• Col. 1:15 ος εστιν (omitting εικων) 056* 0142*
• Col. 1:18 ος εστιν απαρχη
• Col. 1:19 οτι ο εν αυτου 056* 0142
• Col. 1:23 ει γε και επιμενετε
• Col. 3:3 κεκρυπται εν τω Χριστω συν τω θεω
• Col. 3:23 και παν ο (ε)αν ποιητε
• Col. 4:7 συνδουλος εν Χριστου
• Col. 4:8 ημων (omitting και παρακαλεση τας καρδιας υμων -- h.t.?)
• Col. 4:9 εξ εστιν υμων παντα 056*^vid 0142

Readings of Ψ with no papyrus/uncial support

• Col. 1:7 συνδουλου υμων
• Col. 1:22 παραστησας υμας αγιους
• Col. 1:24 εν τω σωματι μου
• Col. 1:24 μου ο εστιν (omitting υπερ του σωματος αυτου)
• Col. 1:27 ο θεος γνωναι
• Col. 1:28 εν ημεις
• Col. 3:4 υμεις συν αυτω (omitting φανερωθησεσθε εν δοξη)
• Col. 3:15 εν σωματι ενι
• Col. 3:17 ονοματι του κυριου Ιησου [also 330 1799]
• Col. 3:20 τουτο γαρ ευαρεστου εστιν Ψ* (Ψ^c ευαρεστον)
• Col. 3:23 παν ο τι (ε)αν ποιητε
• Col. 4:7 και δουλος εν κυριω

In both these cases, there seems to be a higher rate of special readings in Colossians 1, but in each case, there are at least some more in later chapters. In the case of Ψ, it looks as if the scribe was somewhat incompetent, but was more alert at some times than others. Other scribes also have this problem; it should not surprise us. The special agreements of 056 0142 are much more likely to be genetic, and it might be that they are simply the readings of Œcumenius. But it is much easier to spot them in collations, which include all readings, than in an apparatus, which would have omitted most of these!

It should be noted that the above table does not prove that 056 0142 are immediate kin. Although one possible explanation for that data would be that they were sisters or cousins, it would be equally possible that the two were unrelated manuscripts that were imperfectly corrected toward a common standard with an unusual text (as 424* is unrelated to 1739 but has been corrected toward it; if 424-as-corrected had been copied, it would show a pattern of agreements with 1739 that might look like that of 056 0142, but which would not in fact be the result of direct kinship). In fact 056 0142 do seem to be very close, but more investigation is still probably needed. Von Soden listed all of the following as having Œcumenius's praxapostolos commentary: 056, 0142, 82, 91, 94, 101, 250, 314, 327, 424, 441/442, 454, 468, 605, 607, 617, 621, 641, 911, 918, 1360, 1839, 1862, 1871, 1934. (Not a very distinguished list of manuscripts, is it?). All of these, according to the KListe, are in fact commentary manuscripts and include Paul except for 327 424 468 918 1360 1839 1871 (not listed as commentary manuscripts, though of course they could have the text and simply omit the commentary). By comparing the manuscipts this list we could attempt to determine whether the common text of 056 0142 is simply the Œcumenius text or whether the two are closely related even within that type.

We can at least make a preliminary check, by looking at the Text und Textwert data for these manuscripts. I don't have the Paul volume to hand, but I can look at the Acts data for those of the above manuscripts included in it. I've listed the five closest substantial manuscripts (at least 80 readings) for each. I list the manuscript, the percent agreement, and the number of variants for which both exist -- e.g. 056's closest relative is 0142 (97%, 104). That means that, yes, 0142 is the closest manuscript to 056. They share 97% of the readings for which both exist, and there are 104 readings for which both exist.

* 441 and 1839 each have only 78 readings, so I've looked at MSS. which have at least 65 of their readings.

This would obviously be much easier if T&T had profiled more of these manuscripts, but on the evidence, it really looks as if the kinship of 056 0142 is more fundamental than just the fact that they're Œcumenius manuscripts.

Collations in Other Languages

Greek is not the only language for which we need collations, of course. Any text existing in multiple copies calls for collation of these copies. This includes even printed works, such as the First Folio of Shakespeare, as long as corrections were made during printing, but of course it's most important for non-Greek manuscripts. Which may show the same sort of variety as we see in the Greek witnesses.

Let's take a couple of examples from the Vulgate. The following is a proper collation of Matthew 6:7-15. The text on the left is the Clementine Vulgate; that on the right is the text of Codex Lichfeldensis (as given by Hopkins-James). Divergences between the two are shown in bold in the Lichfield text so you can readily refer back between collation and text. The collation follows the text.

Collation of Lichfeldensis against the Clementine Vulgate:

6:7 nolite ] omit
    ethnici ] ci (sic.)
    quod ] qui
6:8 assimilari ] adsimillare
    quid ] quibus (scribe initially wrote quid then corrected it)
6:9 orabitis ] orabitis tur nomentuu
    sanctificetur ] scifice (i.e. sanctifice)
6:10 adveniat ] et ueniet
6:11 supersubstantialem ] cotidianum
     hodie ] odie
6:12 dimittimus ] demittimus
6:13 in tentationem ] intemptemtationem
     Amen ] omit
6:14 dimiseritis ] demisseritis
     dimittet ] demittet
6:15 non dimiseritis ] nondemisseritis
     vester] uester [i.e. vester] caelestis
     dimittet ] demittet

This collation is a fairly standard collation format (as you'll see, it's not exactly how Hopkins-James did it!), similar to that for the Greek texts above. But, in Latin collations as in Greek, that doesn't mean it's always followed! Just to show the possible variations, here are samples of the marginalia to this passage in several Latin editions. I have in my library of this passage one publication of a Latin Gospel manuscript (Lemuel J. Hopkins-James, The Celtic Gospels, the edition of Codex Lichfeldensis used to create the above collation of that manuscript) and three critical editions apart from the standard Stuttgart Vulgate: The smaller Wordsworth-White, Merk, and the Nestle Greek/Latin/English triglot. None of these are collations except Hopkins-James, but they all show apparatus formats. Let's show a handful of variants to show how Latin collations and critical editions are sometimes done (for the symbols used for the manuscripts, see the section on the Vulgate in the article on the Versions). As a sample, let's reproduce the text and apparatus of all four volumes for Matthew 6:7-13, then do comparisons side by side for several readings.

Hopkins-James

Merk

Nestle

Wordsworth/White (editio minor)

Other examples of the various styles:

Mark 12:29. The Clementine text reads Dominus Deus tuus; this has the support of Dublinensis, Sangermanensis, Vallicellanus, and others; Amiatinus and other early manuscripts read Dominus Deus noster (compare the Greek). Our authorities describe the variant as follows:

Luke 2:2. The Clementine text reads Cyrino, supported by the large majority of manuscripts. The Wordsworth/White text reads Quirino on the basis of harleianus (and the historical name Quirinius). Our authorities describe the variant as follows:

Non-Latin Collations

At least Latin is widely read and has relatively stong standards. In less-common languages, things can get even more difficult. The following shows the opening of two editions of the Old English poem The Dream of the Rood. Both are based on the same manuscript (the Vercelli Book), though with different orthographic styles. I parallel the first ten lines of the poem as presented by (1) Bruce Dickens and Alan S. C. Ross, The Dream of the Rood, Methuen's Old English Library, 1963; and (2) John C. Pope, Seven Old English Poems, Norton, 1981.

The physical task of collating

For the reasons described above, it is strongly suggested that every student undertake a collation or two. It will give you a feeling for manuscripts that you simply won't get any other way. It need not be of an actual manuscript (though this is best). But for your first try, you can take one printed or electronic text and compare it against another. (Printed texts are probably better than electronic, since a true collation will normally involve a physical manuscript.) Ideally it should be an actual manuscript text, but if worst comes to worst, one can (say) collate the UBS text against the Textus Receptus.

The results can be educational and humbling, especially if you are able to compare the result with an existing collation of the manuscript. Collation is a difficult and stressful occupation, even with the best manuscripts (generally the easiest are the better-preserved uncials) When dealing with a more difficult manuscript (e.g. 6, which is written in such a small hand that some people need to magnify it to read it; or 33 or 2344, damaged by damp; or a palimpsest; or any of the hundreds of manuscripts written by scribes with bad handwriting), the task becomes even more daunting.

Plus, as L. C. Hector once wrote, "There is some truth in the paradox that legibility in a manuscript document consists chiefly in the reader's prior knowledge of what it contains. Certainly the most difficult documents to read are by no means necessarily those in the most outlandish hands: they are far more likely to be those which are cast in a form outside the reader's experience" (L. C. Hector, The Handwriting of English Documents, second edition, 1966, p. 26).

To give a personal example: The collation of Hebrews mentioned above was based entirely on already-extant transcriptions, so eyestrain was not a factor. (Fortunately for me, as I have very weak eyes!) It was not, for obvious reasons, checked by anyone else, and I myself checked only the non-orthographic variants. The result is only about a dozen pages long, even in large print. And even so, it took me dozens of hours (spread over several months) to compile. And there are doubtless several errors even so -- for which I can only apologize.

At that, the task is easier than it used to be. Today, we can collate with a computer, either in a word processor or using a program such as COLLATE. Older collations were done on paper or other non-electronic form. Moorman, for instance, suggested index cards, with the lemma typed at the top of each card and the lines on the card representing the various manuscripts to be collated.

Let's give an example from 1 Peter 1:1, in case you ever have to deal with such a thing. I'm going to use the UBS text as the base, and include the readings of 𝔓⁷² ℵ A B

For a second example, here is my collation of several texts of an old folk song, "Boney on the Isle of Saint Helena." Moorman's cards would look like

It will be evident that to do a long work will take a lot of cards, and that it is only possible to collate a limited number of witnesses, and sometimes you will have problems with variants that run across cards. Hence the advantage of doing all this electronically.

One important point to remember in preparing a collation is that, if you are ollating a classical manuscript, your collation will likely be used in preparing a stemma, and it may be so used even if it is a Biblical manuscript (if the manuscript belongs to a family such as Family 13 or K^r). You should operate on this basis, e.g. by highlighting readings which are likely to have genealogical significance.

The task being what it is, careful preparation is required to create a good collation. Experts make the following recommendations for accurate collations:

• All manuscripts should be "twice checked." Two collations should be made, without reference to each other, and compared. For a collation which is to be published or used for a major critical apparatus, it should not only be double-checked and compared, but the work should be done independently. That is, the initial collations should be done by two different individuals, and the results compares by a third individual, who will make reference to the manuscript where the two collations differ.
• Before beginning a collation, one should familiarize one's self with the manuscript. The best way to do this is to collate a chapter or two (preferably not the first chapters one will collate "for real"), then throw this collation away. Its only purpose is to make the collator familiar with the manuscript -- the handwriting style, the scribal peculiarities, the organization of the pages.
  Even before making the collation, you might want to consider making an actual transcription of a chapter or two. Copy a substantial portion of the manuscript, text for text, punctuation for punctuation, abbreviation for abbreviation. This will force you to learn the orthography as well as the script.
• Before making a collation (or publishing a diplomatic edition), one should also practice the script of the manuscript. Make a table of the alphabet, and write it in the style of the scribe. So, e.g., if it's an uncial, and it uses the lunar sigma C rather than Σ, practice writing the C. Figure out if the scribe is starting the stroke from the top or the bottom (it will usually be the top for C. But what about Δ or Α?) By getting an actual feel for writing the letters, you will have a better grasp on how they can be distorted, which will help you with damaged or difficult sections of the text.
• A tip from my own work with Paris Vulgates, where the text is extremely small: If possible, you might want to start by making an enlarged photograph or scan of the manuscript. That way, when you start, you won't be trying to decipher ab acharon usƺ geth Ꜫ ťminos suos lib’avit de manu philistinoʮ but can see that it reads ab acharon usƺ geth Ꜫ ťminos suos lib’avit de manu philistinoʮ, i.e. ab Acharon us(que) Geth et terminos suos liberavit de manu philistinorum, 1 Samuel 7:14). There will be time when the scan will be imperfect and you'll have to look at the manuscript -- but there will also be times when the manuscript simply can't be read at actual size and an enlargement will help. For the Paris Vulgates, I liked to blow the image up about two and a half or three times, and work from the enlarged version while regularly referring to the actual manuscript. If the manuscript is highly legible, you'll probably want to do it the other way around, and work with the manuscript while occasionally referring to the photo. Also, if by some chance you are collating a manuscript with extremely large print (for example, a choral text, on which the letters can be several centimeters tall and perhaps a little spluttery), make a reduced scan, so that it's easier to see a word in context, which can make it easier to interpret.
• One should maintain a proper schedule. Spending too much time collating will result in poor quality work, and may lead to eyestrain as well. Ideally, one should not collate for more than two hours a day, and one should not allow more than four hours of work under any circumstances. One should take regular breaks, and assure that there is proper lighting and working conditions. Distractions such as phone calls should be avoided if at all possible. I probably shouldn't have to point out that you should not try to multitask (no texting while collating!), but I'll say it anyway, just in case.
• Even if you cannot bring in someone else to check your collations, use as many cross-checks as possible. The method you use depends on your techniques. You might, for instance, read back your collation (aloud) to see if you have transcribed the words correctly. You might "collate back" into the Textus Receptus, and read that back, or compare it with your original manuscript.
• Never assume, because a word has the right beginning and ending, that it is the "correct" word. Variant spellings, some of which could represent different words or word forms, are very common.
• Before beginning the collation, examine the corrections to the manuscript. Try to determine how many correctors there have been, and perhaps their habits.
• If you are working as part of a project, and so are called upon to follow a particular collation format, study the format in detail before beginning. Look over other collations in this format. Practice using the format. (This is distinct from practicing with the manuscript. Don't try both at once; you're likely to lose track of one or the other.)

Also, unless you are collating a manuscript in your home city, you will have only a limited time to prepare the collation. So prepare: Often it will be possible to obtain photographs, or a microfilm. Get those, even if you have to pay for them, and study them before you visit the manuscript. Look at the handwriting. Look at the corrections. Can you tell if they are all in the same hand? What can you tell about the pen and ink and the surface on which they are written. Prepare a list of questions to be resolved when you first physically encounter the manuscript.

Another suggestion, this one personal: Don't create your first collation by working in a foreign language! Start by comparing two texts in your own language. A good place for this is in collections of old folk songs or modernized editions of an ancient text. This lets you practice the physical task of collation without having to worry about understanding a foreign language as well.

Charles Moorman, Editing the Middle English Manuscript, p. 46, gives another warning: "The editor cannot afford to become himself a scribe if he can avoid it." Moorman was writing in the 1970s, before personal computers, so some of his reasons are rather irrelevant. But the basic argument is sound: If you are transcribing the manuscript in full, you are not collating, and you are making errors of your own. If there already exists an edition of the work you are collating, collate against that, no matter how obscure, just to reduce the error rate. This is not always possible for classical documents, but in the New Testament, most important sources have been printed by someone. So take advantage.

For those who wish to have something to work from, and whose native language is English, here are two transcriptions of a fifteenth century English text, "The Agincourt Carol." (This should, incidentally, put to rest the notion that "carols" are Christmas songs; they are a particular form of religious ballad.) The first is from Chappell's Popular Music of the Olden Time (also variously known as Old English Popular Music, etc.); the second is from Percy's Reliques. The Percy text was transcribed from a manuscript copied from the manuscript used by Chappell. That is, the genealogy is this:

      Archetype
          |
    ---------------
    |             |
Chappell        [Copy]
                  |
                Percy

Deo gracias anglia,
Redde pro victoria

1 Owre kynge went forth to normandy,
  With grace and myght of chyvalry:
  Ther god for him wrought mervelusly.
  Wherfore englonde may calle and cry
                        Deo gracias....

2 He sette a sege the sothe for to say,
  to harflu toune with ryal aray;
  that toune he wan, and made afray,
  that fraunce shal rywe tyl domesday.
                        Deo gracias....

3 Than went owre Kynge with alle his oste,
  thorwe fraunce for all the frenshe boste:
  he spared no drede of leste ne most,
  tyl he come to agincourt coste.
                        Deo gracias....

4 Than forsoth that knyght comely,
  in agincourt feld he faught manly:
  thorw grace of god most myghty,
  he had bothe the felde and the victory.
                        Deo gracias....

5 Ther dukys and erlys, lorde and barone,
  were take and slayne, and that wel sone,
  and some were ladde into Lundone
  with ioye and merthe and grete renone
                        Deo gracias....

6 Now gracious god he save owre Kynge,
  his peple, and all his wel wyllynge:
  gef him gode lyfe and gode endynge,
  that we with merth mowe savely synge,
                        Deo gracias....

Deo gratias Anglia redde pro victoria!

1 Owre kynge went forth to Normandy,
  With grace and myyt of chivalry;
  The God for hym wrouyt marvelously,
  Wherfore Englonde may calle, and cry
                        Deo gratias:
  Deo gratias Anglia redde pro victoria.

2 He sette a sege, the sothe for to say,
  To Harflue toune with ryal aray;
  That toune he wan, and made a fray,
  That Fraunce shall rywe tyl domes day.
                        Deo gratias &c.

3 Than went owre kynge, with alle his oste,
  Thorowe Fraunce for all the Frenshe boste;
  He spared 'for' drede of leste, ne most,
  Tyl he come to Agincourt coste.
                        Deo gratias &c.

4 Than for sothe that knyyt comely
  In Agincourt feld he faught manly:
  Thorow grace of God most myyty
  He had bothe the felde and the victory.
                        Deo gratias &c.

5 Ther dukys, and erlys, lorde and barone,
  Were take, and slayne, and that wel sone,
  And some were ledde in to Lundone
  With joye, and merthe, and grete renone.
                        Deo gratias &c.

6 Now gracious God he save owre kynge,
  His peple, and all his wel wyllynge,
  Gef him gode lyfe, and gode endynge,
  That we with merth mowe savely synge
                        Deo gratias &c.

(We note incidentally that, using these texts, we can detect the loss of an obsolete letter, just as Homeric scholars can detect the fact that Homer used the digamma. The Middle English text of this song clearly used the yogh, ȝ. In Chappell, this was replaced -- as is now fairly normal -- by gh; the Percy text substitutes y.)