Textual Evidence

Royse on
Scribal Habits:

P45 P46 P47 P66 P72 P75

Excerpt from: James R. Royse, Scribal habits in early Greek New Testament papyri ,
vol 36 NT Tools & Studies, (Brill, 2008)

Page Index

Royse on Scribal Habits: - P45  P46  P47  P66  P72  P75:
    Scribal Habits: Results and Rankings (pgs 719-720)
    Non-Singular Readings: (pg 724)
    Codex Vaticanus (B): the character of its scribe (pg 728)
    A New Canon of Transcriptional Probability: Prefer the Longer Reading - with exceptions.
    (Some) Concluding Remarks : 5. Extension to Non-Singular Readings...

Return to Index

Royse on
Scribal Habits


"Even Royse’s 1981 dissertation was considered groundbreaking and remained a standard work for twenty-seven years (until it was superseded by the monograph). To put this work into perspective, E.C. Colwell had urged that someone someday would publish a commentary on singular readings (and he was talking about the three papyri he had studied).

Royse came along, corrected and updated Colwell’s original work. Royse checked the readings against all available editions. His study was much more nuanced than an attempt to classify the text of these witnesses in broad terms, e.g., according to text-types. Royse's goal was to cast light of each scribe’s habits. This was necessary to be able to arrive at canons of criticism applicable to papyri. Hernandez said he was “electrified” by the 1981 study. He was given “access.” It was a gigantic how-to-do manual. It was permeated of transparency. Everything was available for scrutiny. His prior interest was in theological variation.

Royse, however, was not a friend of “theologizing.” Royse applied an extreme caution. He singled out only three variants that were theologically motivated (in P72). It was cristall clear for Hernandez that he would have to learn from Royse. He now had to put behind him notions of theological variation and start with the mundane, the facsimiles! Everything that he needed was there in Royse’s work, being the model."

- Excerpt: SBL Boston, Book Review of James Royse,
Scribal Habits in Early Greek NT Papyri, pt. 1 (Apr 20, 2009)

The following is excerpted for review purposes from: Scribal habits in early Greek New Testament papyri by James Ronald Royse, (Brill, 2008)

Headings have been added for clarity and navigation purposes.

Chapter 10

The Shorter Reading?

pgs 719-720

...to the total number of significant singulars [singular readings], ...we have: 58

P45 P46 P47 P66 P72 P75
Additions 29526/715/16/171411
Omissions 6016115/16202234
Ratio O/A - 2.71.2 -
Net Words Lost 90244/24529/3116/17/1826/2732
Significant Singulars 2104525310974106
Ratio NWL/SS .43.54.55 - .58.15 - .17.35 - .36.30

If we rank the papyri in terms of how much text they lose per significant singular, we have:
[i.e., best to worst]

P66.15 - .17 1.2 - 1.3
P75.30 3.1
P72.35 - .361.6
P47.55 - .582.1 - 2.7

The scribes thus differ considerably in their rate of loss of text, ranging from the extraordinarily careless scribes of P47 and P46 (who lose more than one word per 3 errors) and the careless scribes of P75, P72 and P45 (who lose roughly one word per 3 errors), to the comparatively careful scribe of P66.

If we look at the ratio of omissions to additions, we find they range from P75 and P46 (who omit about 3 times as often as they add), and P47 and P45 (who omit more than twice as often as they add), to P72 (who moits rather more often than he adds) and P66 (who omits onlyh slightly more often than he adds). The reader will doubtless find yet other ways to compare our scribes. 59

And, of course, these scribes differ greatly among themselves with respect to other patterns of error. Indeed, precisely because the six scribes differ in so many ways, are copying different portions of the NT, and are utilizing texts of different sorts, it would seem that their common habit of shortening the text is a general habit, and not an anomalous feature of one or two particular scribes.

To be sure, one could contend that all six scribes are anomalous, but, given their many differences, such a view would seem highly implausibile, and to be based on no evidence. 60 Naturally, we might eventually discover other early papyri that would force us to revise these conclusions, but we have to work with the available evidence.

And there seems in fact to be no reason to suppose that we just happened to have found manuscripts from the six scribes of antiquity who tended to shorten their text. On the contrary, it would seem that these six manuscripts should represent a fair sample (in so far as any sample of six could be fair 61) of the scribal activity involved in the copying of the NT in Egypt in the years from say, 175 to 300.

Moreover, there is now considerable confirmation of a wider scribal tendency to omit, notably from several independant studies undertaken since my dissertation. 62

Original Footnotes:

58. The figures here differ from those found both in my dissertation, 602, and in my "Scribal Tendencies", 246. The changes are mostly the result of the asterisking of readings because of new evidence. Apart from the details of the numbers, the rankings of the papyri according to various criteria have also shifted. What remains constant is that all the papyri tend to lose text, and that P66 is, in this respect, the most careful scribe.

59. Cf. Havet, Manuel de critique verbale, para 427.

60. Hoskier's comment on codex 181 cited in chtp. 2, n.11, is again relevant. After finding that almost half the singular readings in 181 are omissions, Hoskier says:

"This large number of omissions points to a great carelessness and not to 'a shorter text being preferable'."

The extreme application of the preference for shorter readings would lead us to see 181 as a very valuable text, and of course critics have sometimes held that shorter readings found in our papyri have claims to originality. But such a course forsakes the evidence in behalf of preserving a principle. Surely the evidence is that 181 is careless, and similarly that our six scribes are omitting through carelessness (if not deliberately on occasion).

61. Of course, we do not have (and doubtless will never have) a set of early MSS that would satisfy the requirements of a random sample as understood in statistics. We must take the MSS that history has given us, and with all due caution attempt to learn from them.

62. The work of Kevin James, The Corruption of the Word: The Failure of Modern NT Scholarship (Williamsburg, N.Mex.: Micro-Load Press, 1990), contains many interesting comments concerning scribal habits, and in particular reports (73-80) on a study of 13 important witnesses (not including any papyri) in John 1 - 11 with respect to the maxim of lectio brevior potior.

James's results indicate that omission is much more frequent than addition, and he concludes (76) that "the omission of words is a clear copyist habit (at least in some MSS), whether from mistake or by design." However, James's study is rendered problematic by the fact that his determination of the status of a reading was made (73-74) by comparison with "the words in each verse where King James and modern versions agree."

Moreover, his sources may not be the best. For instance, he states (66) that P66 omits the line over the nomen sacrum IC at John 5:8 and 6:10; but the omissions are only in the printed text by Martin in the editio princeps (duly noted by James [252] as his source), and the lines can be clearly seen in the photos of P66. But I have not attempted to check his results for the 13 MSS involved.

pgs 724f

NON-Singular Readings

[Royce here quotes Min:]

"The investigated papyri exhibit overall considerably more omissions than additions. That corresponds to the overal picture of the investigation of the extensive papyri, which likewise come from this early period."

( - Min then refers to my [1981] dissertation) ...

Turning to readings that are of more significance, Min observes that the omissions, whether short or long, may result from negligence, whereas longer additions require some cause, and harmonization is often that causal factor. 80 He then goes on to discuss the overall results concerning harmonizations, finding that harmonization to parallels is more common than harmonization to the context. 81

Now, in general I believe that the evidence based on non-singular readings is not as probative about scribal habits as is the evidence based on singular readings. Nevertheless, Min's work follows a carefully designed methodology, and it can hardly be doubted that his results are of significance for our understanding of the scribal habits found in these early MSS.

Original Footnotes:

80. Die fruheste Uberlieferung des Matthausevangeliums, 284.

81. Ibid., 284-288. Min contrasts this result with the result of my dissertation that harmonization to the context is more frequent than harmonization to parallels, and notes that the explanation of the difference may be that his study is limited to Matthew, where there are more parallels than in NT books generally.

pgs 728

The Characterization of B

Codex B: Even Hort granted that there is "a proneness on the part of the scribe of B to drop petty words not evidently required by the sense," and that in B "omissions due to clerical error, and especially to homoeoteleuton, naturally take place sometimes without destruction of sense." 98 But of course Hort was convinced that what might appear to be omissions by B are in fact often additions by other MSS. Many other scholars have not looked as favourably on B. Von Soden cites occasional omissions of letters or syllables, as welll as one or more words, and notes that the omissions sometimes occur through homoeoteleuton. 99

Original Footnotes:

98. Introduction para. 314 (pp. 236-237). [This entire section is a transparent attempt by Hort to dismiss the many singular readings of Codex Vaticanus (B), and disallow them as evidence for the purpose of characterizing the scribe of B. In fact however, singular readings are the only reliable evidence for characterizing the habits of the scribe of a manuscript.]

99. Die Schriften, Untersuchungen, 907-908, 910-911. Von Soden also discusses the features of the common ancestor of Aleph and B (what he calls δ1-2), and presents (ibid., 900) instances of omission, many by leaps, allegedly made in that common ancestor.

See also Scrivener, A Plain Introduction, 1:120, and Lagrange, Critique Textuelle, 88, who dismisses the omissions as "lapsus du scribe."

Kubo, P72 and the Text of Codex Vaticanus, 17-19, briefly looks at the singular readings of B in 1st/2nd Peter and Jude, and finds that the "largest group of these is omissions" (17). Indeed, Kubo lists 12 omissions (18) but only 1 addition (19). (One should also include the omission at 1st Pet. 1:1, which Kubo [17] counts as a "scribal error".) These omissions (and the addition) are of one or two words, except for the omission of 1st Pet. 5:3, which is unexplained (despite Kubo's unconvincing attempt [18] to explain it by homoeoArcton).

pgs 735

A New Canon of
Transcriptional Probability

...[on the other hand] ("the longer reading is to be preferred") would hardly be justified.

Rather, what does seem to follow from this analysis is that the burden of proof should be shifted from the proponents of the longer text to the defenders of the shorter text.

There does not need -- at least within the earliest centuries of transmission -- to be a "reason" for an omission; rather, omission is a "natural" error for these early scribes. What does need to be stated is some list of possible exceptions to this general tendency. And the present stduty gives at least a preliminary indication of what the likely exceptions are. Obviously, harmonization to the context is a pervasive source of error, and it often results in additions. Other kinds of harmonizations and grammatical improvements (such as avoiding asyndeton 124 ) may also involve additions. 125

These considerations suggest, therefore,
that a Canon of Transcriptional Probability could be formulated as follows:

In general the longer reading is to be preferred, except where:

a) the longer reading appears, on external grounds, to be late; 126 or,

b) the longer reading may have arisen from harmonization to the immediate context, to parallels, or to general usage, or

c) the longer reading may have arisen from an attempt at grammatical improvement.

The frequency of omissions by scribal leaps and of omissions of certain inessential words such as pronouns must be kept in mind, and when such omissions may have occurred the longer reading should be viewed as even more likely.

Such a canon would, of course, be only one tool among many for investgating the NT text. And deciding whether one of the three listed types of exceptions has occurred would often be difficult. 127

This is all we should expect. As Hort and others have noted, and as we have seen amply confirmed in our six papyri, the sorts of errors made by scribes vary enormously. Hence, no simple rule will suffice for all or even most variations. What we can hope for, though, are guidelines that are based on the evidence of the documents. And with the judicious use of such guidelines may come insights into the history of the text and, from time to time, revisions of what would otherwise be accepted as the NT text. 128

124. Here our criteria may conflict: Did the scribes naturally omit an inessential conjunction, or did they add a conjunction for the sake of smoothness? THe scribes studied here do both.

125. Scrivener, Six Lectures, 115, cites only two specific examples in support of his version of the lectio brevior rule, but each is, as he emphasizes, an instance where the words from a parallel have been inserted.

126. External evidence must in any case be constantly considered, as we attempt to construct the history of the text. As Colwell says ("Hort Redivivus", 157):

"the crucial question for early as for late witnesses is still, 'Where do they fit into a plausible reconstruction of the history of the MS tradition?' "

Griesbach, Novum Testamentum Graece 1:lxiv, similarly qualified his principle of lectio brevior potior ("unless it lacks entirely the authority of the ancient and weighty witnesses" [Metzger/Ehrman The Text of the NT, 166]). Such a clause gives proper respect to the historical nature of textual criticism, and also allows that later scribes may have copied in ways quite different from earlier ones. See also Epp, "Eclectic Method", 228 = Epp and Fee, Studies, 153.

127. Greenlee, Introduction, 112 provides two rules that need careful balancing:

(b) the shorter reading is generally preferable if an intentional change has been made. The reason is that scribes at times made intentional additions to clarify a passage, but rarely made an intentional omission...

(c) The longer reading is often preferable if an unintentional change has been made. The reason is that scribes were more likely to omit a word or a phrase accidentally than to add accidentally.

While granting the general plausibility of these principles and their justification, I would prefere to make the points with less appeal to psychological states.

Perhaps here I can reply briefly to some further interesting comments by Silva, who in his "The Text of Galatians", 23 and more fully in his "Internal Evidence in the Text-Critical Use of the LXX", 157-161, criticizes some of my earlier remarks about the shorter reading. Silva correctly observes (159) that further discrimination among additions and omissions is possible and even necessary, and that many of the omissions are either the results of scribal leap or of what Silva calls "empty words", such as conjunctions, articles, prepositions, and pronouns. (Cf. Havet, Manuel de critique verbale, para 420-427).

It is true that if one removes those omissions from the counts, the proportions will look much different. But this is tantamount to holding that the scribes tend to add, assuming that one ignores most of the places where they tend to omit. [!].

I would certainly accept Silva's reminder that Griesbach's formulation of the lectio brevior potior principle is far from a simple preference for the shorter reading, and that its correct application requires a sensitivity to the many exceptions and conditions that Griesbach notes. (Silva returns to this point in his "Response" in Rethinking NT Textual Criticism, ed. D.A. Black, 145.) But once one emphasizes all the exceptions, should really continue to accept the principle? And we may recall that, as seen earlier in this chapter, most statements of the principle are far less restricted than is Griesbach's. See also my "Scribal Tendencies," 243-244.

128. Some results along these lines may be found in my "Scribal Leaps". Readings that have such a possible cause are, of course, particularly suspect according to the canon stated above. See also, for example, Aasgaard, "Brothers in Brackets? A Plea for Rethinking the Use of [ ] in NA/UBS", as cited in chapter 5, n. 560.

pgs 741

(Some) Concluding Remarks

5. Non-Singular Readings

A major difficulty in the isolation of readings created by a particular scribe is the possibility of coincidental agreement in error with other scribes. Colwell was aware of this problem, and suggested that one study a particular manuscript with reference to its textual type, thus assuming that readings supported by MSS from other textual traditions may nevertheless be creations of the manuscript's scribe.

This is certainly plausible when the MS does belong to a clearly defined textual type (P75 is the best example), but there remains the formidable task of deciding when an agreement across textual traditions is accidental and when it is due to contamination ['mixture']. 23

It seems, though, that it is possible to gain at least some information on the likelihood of coincidental agreement by the study of MSS whose textual antecedents are clearly established. 24

For instance, one might take a MS of Family 1, a MS of Family 13, a MS of Family Π, and an early papyrus codex, and see whether there are any agreements among some of these against other witnesses, and particularly against other members of the same families. 25

Such studies would give some firm data on the frequency of coincidence in error, and, as a consequence, would permit more precise determinations of what errors are likely or unlikely to occur without genealogical relation. Such a determination would indeed be essential for the eventual extension of the results obtained by an analysis of singular readings to the majority of important textual variants, where two or more readings are non-singular. 26 [26]


The goal of clarifying the history of the transmission of the NT text and perhaps even penetrating to the original words of the writers of the NT provides our incentive to ponder the habits of scribes of the NT MSS. Despite their errors, to which we are all liable, we must be grateful to them for preserving the text of the NT. May the present work serve to illuminate their efforts.

23. Cf. the comments by Epp concerning Fee's notion of "sub-singular reading" in "Toward the Clarification", 162-163 = Epp & Fee, Studies, 53-54.

24. Colwell's example ("Scribal Habits", Studies, 123-4,item V) of agreements between singular readings of P66 and the RSV falls into this category, since the textual basis for the RSV is known not to have included P66. The modern versions, whose Greek Vorlagen are more or less precisely known, furnish clear evidence of such coincidence; see a few examples cited in my " P66 and the Ethiopic Version", and e.g., in ch.7, n. 328.

25. ...At Mark 9:13 von Soden reports that παντα, which he says is harmonized to Matt.7:12, is read by 565 1071; the microfilms confirm that all 3 MSS add παντα. Here a genetic relationship seems at least very unlikely.

26. Related to such an investigation would the sort of analysis undertaken by Fee in his P75, P66, and Origen, 42-44 = Epp & Fee, Studies 270-272 where he attempts to characterize the tendencies of textual traditions. His results are very interesting, and could be made even more persuasive if one know how to account adequately for the possibility of coincidental agreement in error. ...

Return to Index