Rule of Iron

The "Rule of Iron" (règle de fer) was the name that Dom Henri Quentin used to describe his method for establishing a critical text of the Vulgate. Sadly, his description seems to have utterly baffled other textual critics, whose descriptions of it make no mathematical sense at all -- and I am not enough of a linguist to try to decipher his maths to see if they are sound. J. Burke Severs says that Quentin's method had no way to deal with conflation, and Vinton A. Deering agrees, so I suspect this is true and that this furthermore makes it likely that Quentin could not deal with mixture -- which, in a New Testament context, makes his method effectively useless.

In addition, Quentin's method of creating the stemma was so laborious that John M. Manly & Edith Rickert, The Text of the Canterbury Tales, volume II (University of Chicago Press, 1940), p. 17, say that it can only be applied to a small section of text -- a few hundred lines. This, they point out, means that it simply cannot work for block mixed manuscripts (although they do not use that term). And it's not that Manly and Rickert objected to work; their edition of the Canterbury Tales is by far the most detailed ever prepared, and took them decades. So if they say it's too much work, we can safely assume it's too much work! Computers would help here -- but given that we now have a much better tool in the form of cladistics, I don't see why anyone would bother implementing Quentin's classification method. Especially given that it has no method for dealing with mixture.

But there are in fact two parts to Quentin's method. In the first stage, he tries to classify manuscripts. His goal, in essence, is to find three basic types -- in simplest terms, to force a stemma with three branches, avoiding the Bédier Problem of two-branch stemma. (I find myself thinking that it's too bad Quentin didn't know anything about voting theory; he really should have looked at Condorcet cycles. But that's not really relevant to what he did.) Of course, forcing all stemma to have three branches is no better than forcing them to have two, which is what Bédier accuses other critics of doing; what we want is the correct stemma, not the one with some prescribed number of splits!

Quentin was also deeply concerned that Lachmann-type stemma could only reconstruct the archetypes, but that critics would act as if they were reconstructing the autograph. On this basis he argued that agreement in error was not sufficient to construct a stemma -- since the very thing we were reconstructing, the archetype, would contain errors. So agreement in error might be agreement with the archetype against the autograph! So Quentin set out to come up with a classification based on overall agreements.

Up to this point, Quentin was dead right, and Lachmann and everyone else was simply wrong. It was Quentin's method of classification, which forced all traditions into a three-branch stemma, that (appears to) fail. All we can really say for Quentin's classification method is that it was rigorous.

(I should add that it is perhaps not repeatable; the preface to his edition of Genesis is full of manuscript descriptions and has lists of sample readings but does not really have an algorithm to convert this data into his stemma; indeed, the edition doesn't show a stemma. It just lists groups. Perhaps this was explained in greater detail elsewhere, but it all looks very hand-wavy to me.)

The groupings weren't all of Quentin's method, though. Having come up with his classification, the actual Rule of Iron was that two beats one -- that is, if you have three primary witnesses or types, and two have a single reading and the third has another, that the reading found in the two are original.

This is a little oversimplified -- there are circumstances in which an outside influence can cause two types of text to be independently corrupted. But the basic rule is sound. Quentin's failure lay not in his final idea but in the methods he used to determine his three witnesses, which were both confusing and (from what I can tell) confused.

Quentin's task was to edit the Vulgate Old Testament, and he died long before the task was finished (as did everyone else initially involved in the project), so the Rule of Iron was used only in the Catholic critical edition and only in the Pentateuch. Quentin's three witnesses were the Turonensis type (headed by G, Paris National Library Lat 2334; also cav and tol); the Amiatinus type (headed by A, Amiatinus; Alcuin also files here), and Ottobonianus (headed by O, Vatican Ottobonis Latin 66; also Theodulf).

It is interesting to examine how well Quentin's method compares to that of his successors. We can at least attempt this by comparing the various volumes of the Vatican vulgate edition with the Stuttgart Vulgate, which was edited relatively consistently. (Note: I counted differences rather casually; the numbers may be a little off. And the sample may be imperfect -- e.g. I skipped Wisdom of Solomon because the text seemed utterly chaotic. But the results are indicative.)

Book  Pages in Stuttgart Vulgate  Differences  Differences per page  
Leviticus4149 1.20
Psalms LXX93240.26
1 Maccabees47290.62

What is clear is that the books edited by Quentin had more variants from the Stuttgart text than did the Writings or the Former Prophets. Things are less clear for the Later Prophets, but a very high fraction of the variants in Daniel are in the sections Jerome did not translate, and Hosea's text is chaotic even in Hebrew. It seems clear that Quentin's methods resulted in a text that diverged more from the critical norm than did the texts edited by those who followed him -- but not a lot more. In any case, this does not make his text inherently inferior. For that, we have to look at the text itself.

It might be worth looking at how all this works. To avoid the side effects of working at the beginning or end of a book or section, let's start with Exodus 4 and look at all the places where the Stuttgart Vulgate differs from the Vatican edition for the next twenty chapters (i.e. half the book, Exodus 4-23). Manuscripts, and manuscript groups, are cited by their Vatican edition symbols; the editions are 𝖀=Vatican, 𝖂=Stuttgart, 𝕮=Clementine, 𝕲=Gutenberg.

ReadingStuttgart text and supportersVatican text and supportersOther readings
4:27ei obviam BCMPTXΘΛΠΣΦΨΩBDF 𝖂obviam O 𝖀obviam ei AΨM 𝕮𝕲
5:14i(m)pletis rell 𝖂𝕮i(m)plestis BΠΣO** 𝖀𝕲
6:14aesrom McP2ΘHΦZΨB 𝖂esrom GXΘAG*MΛΣΦAGPRVΨDFΩSM 𝖀𝕲asrom ABCM*OP*TΠD; hesron 𝕮; esron ΘG**ΨMΩJ; asron ΠC
6:21isuar AMΣO 𝖂isaar ΘG**ΣT 𝖀𝕮isoar G*; isacaar ΛL; isacar Λ; isachar BOPTΘAG*HMΛHΠCΦAGPRZΨF; ysahar C; ysuar ΩJS 𝕲; isacar XΣM; issachar GcΦV; ysachar ΠDΨBDMΩM
6:24Abiasab ABGcMOTXΘAMΛΣOΦA*ΨBFΩSJ 𝖂𝕲Abiasaph PΘG**ΣMTΨD 𝖀𝕮Habiasaph C; habiasap ΘH; abisab G*vidΠΦA**GPRVZΨMΩM
7:15egredietur BCGMP2TcX ΘΛΠΣΦΨΩ 𝖂𝕮𝕲egreditur AOP*T* 𝖀
7:19ut AMPTXΘΛΠΣΦΨΩ 𝖂𝕮𝕲et GO 𝖀hut (sic.) C; omit B
12:29omne AcBG*MPTθΛΠDΖΦΨΩ 𝖂𝕮𝕲omnem A*GcOXΠC (C homnem) 𝖀
12:31auocatisque ABCG*OTXΘΛΠΣΦAGPRVZ**ΩM 𝖂uocauitque ΦZ* 𝖀uocatisque pharaoh GcPcΨΩJS (M farao) 𝕮𝕲
14:9in ahiroth ATΘAGMΠΦAGPRZ 𝖂in phiahiroth 𝖀in phihahiroth 𝕮; in phiairoth ΩJ 𝕲; in phiayroth CΩM; in phyairot ΩS; in ahioroth G; in phiahirot ΣM; in phiaroth BΛL; in fiahiroth ΣO; in fiairoth ΛH**; in fiaroth XΛH*; in fiaroht ΣT; in airoth MΘHΨB; in hairoth ΦV*; in hayroth ΨF; in hahiroth ΦV**; in ahirath O; in airoht P*; in aihroht Pc; in athiroth ΨD
15:17sanctuarium AB*MTΛLΦAGRVZ 𝖂sanctuarium tuum COPΘΛHΠΣΦPΨΩ 𝖀𝕮𝕲
15:23mara BMTΘA**HM**ΛLΣTΦΨ 𝖂𝕮marath APXΘA*GM*ΛHΠΣOΩ 𝖀𝕲mare O; marat CΣM
16:18congregarunt ABΘAGH*MΠΣOTΦΨMΩJS 𝖂congregauerunt CGMPTXΘH**ΛΣMΨBDFΩM 𝖀𝕮𝕲
16:32retro rell 𝖂𝕮𝕲uestras conjecture, in † 𝖀omit P2ΨB*
16:36oephi M*ΘA** 𝖂 (ephi BMcP2TcΛHΠΦA**Ω 𝕮𝕲)ephae TΘA*G*HMΦA*GRVZ 𝖀oephae AΛLΦPΨD; hoephae C; epha ΘG**; efi ΣO; oefi ΣMT; aephi ΨB**F; aephy Xc; ephie ΨM
19:16tertius dies BCGMTXΘΠΣMOΦΨBΩJ 𝖂𝕮𝕲dies tertius AOPΛΣTΨDFMΩMS 𝖀
19:21velint BCGMPCTXΘΠΣΦAPR**V**Z**Ψ 𝖂velit AOΛΩ 𝖀𝕮𝕲vellent P*; vellint ΦGR*V*Z*
20:1locutus quoque est ACΛΣOT 𝖂lo(cu)tusque est GMPTXΘΠΣMΦΨΩ 𝖀𝕮𝕲locutus est B; locutus quoque O
21:7consuerunt CGMPTΘAGMΠΣOΦGPRVZΨD 𝖂 consueuerunt ABOΘHΛMΣOTABFΩJS 𝖀𝕮𝕲 consueuerant ΩM; consueuerat ΨM; consueberunt XΛL**; consuberunt ΛL*
21:13in manu ABCXΛΣTΩM 𝖂in manus GMOTΘΣMOΦΨΩJS 𝖀𝕮𝕲in manum P; in manibus Π
21:35uulnerarit AGΠDΣT 𝖂uulnerauit 𝖀uulnerauerit CMP2TΘΛΣMOΦΨΩ 𝕮𝕲; uulnerari ΠC; uulneraberit BX; ulneraret P*
22:17consuerunt ABTPMΘΛLΣΦΨDM 𝖂consueuerunt CGOXΛH(**)ΠΨBFΩ 𝖀𝕮𝕲
22:25urgues B*GM*P*TXΘΛH*LΣTΦΨB*DFM 𝖂 (C hurgues)urguebis 𝖀 (conjecture?)urgebis McΛH**Ω 𝕮𝕲; urges ABcPcΠΣMΨB**; argues OΣO;

The obvious observation is just how minor most of these differences are. Before I started this, I thought that I would try to classify the types of variants, but if we look at it, most of them are cases of changes of verb tense where a very simple scribal error likely produced the variant. To try to figure out which edition is correct would require an extremely deep dive into the Hebrew that likely won't produce much. Plus some of the differences involve names, which always get fractured by scribes, and some seem to be editing errors in the Vatican edition (the Stuttgart text is now largely free of this because it has gone through so many editions). The bottom line is that the Rule of Iron probably produced a text with more variations from the Stuttgart edition than would be produced if two different eclectic editors did the editing, but the divergences are of very little significance.

To be sure, a lot of this is because the variants in the Vulgate are relatively small, at least in the early centuries. Compared to the situation in the New Testament, the Rule of Iron had little to do. It might also be interesting to try Rule of Iron methods in the New Testament.

Of course, this raises the question of what manuscript groupings Quentin would have come up with for the New Testament. It would take someone who knows the Rule of Iron better than I to figure that out, but odds are that the groupings would be B and its allies, D, and the Byzantine Text. What manuscripts would represent each group? For the Gospels, I would propose a trial based on the following:

Group 1 (Alexandrian): B plus ℵ L 33 579 892. The group reading is that of B as long as at least one other supports it; if B is alone, then the reading of ℵ plus at least one other; if both stand alone, then the reading of the other four if they unite; if they don't, then the reading of B.

Group 2 (Western): D. What else is there, since Quentin does not seem to use the versions?

Group 3 (Byzantine): Here the question of which manuscripts to use is acute, and I have no way to tell what Quentin would have decided. What I intend to do is take the consensus of A (earliest member of Family Π), E (earliest member of Kx/Ki), and M (best-known member of the M groups).

The sample is of Luke 5. Only places where the groups disagree will be cited. The group reading, as defined by the above rule, is shown in bold. Data is taken from IGNTP Luke (except where I consulted Swanson because the IGNTP apparatus made the readings ridiculously confusing, e.g. 5:5a). Where there is a true three-way split between the groups (e.g. 5:5d), the reading of the Alexandrian group is adopted, since I assume (perhaps wrongly) that Quentin would have detected its superiority; when there is a three way split but a substantial portion of one of the groups defects, I go with the reading supported by one group and part of another (e.g. 5:1c, 5:5a). Where the Rule of Iron reading is uncertain, it is marked with a red question marks (as, e.g., with a three-way split), ?. Several editions are also cited:

(I omit Vogels because it is not a critical edition and Tischendorf because Tischendorf didn't really have a critical principle.)

VerseRule of Iron readingOther readings
5:1aτου ακουειν 33 D E-Mκαι ακουειν B-ℵ-L-579-892 A ℬℳ𝒮𝒰𝒲
5:1bκαι αυτος ην εστως B-ℵ-L-33-579-892 A-E-M ℬℋℳ𝒮𝒰𝒲αυτον D
5:1c?γεννησαρετ? B-ℵ-L-33-579-892 A ℬℋtxtℳ𝒮𝒰𝒲γεννησαρεθ E-M ● γεννησαρεδ D ● γενησαρετ ℋmarg
5:2aειδε(ν) B-ℵ-L-892 E-M ℬℋℳ𝒮𝒰𝒲ιδε(ν) 33-579 A D
5:2bδυο πλοια ℵc-L-33 D E-M ℋℳ𝒰𝒲margπλοια δυο B-579-892 𝒲txt ● πλοια ℵ* ● δυο πλοιαρια A ● πλοιαρια δυο ℬ𝒮 (!)
5:2cαπ αυτων αποβαντες B-ℵc-L-33-579-892 D ℬℳ𝒮𝒰𝒲αποβαντες απ αυτων A-E-M ℋ ● αυτων αποβαντες ℵ*
5:2dεπλυνον B-892 D ℬ𝒰𝒲txtαπεπλυναν 33 A-E-M ℋ𝒮 ● επλυναν ℵ-L-579 ℳ𝒲marg
5:3aεμβας B-ℵ-L-33-579-892 A-E-F ℬℋℳ𝒮𝒰𝒲ενβας D
5:3bτων πλοιων B-ℵ-L-33-579-892 A-E-M ℬℋℳ𝒮𝒰𝒲πλοιον D
5:3cσιμωνος B-ℵ-L-33-579 D ℬℳ𝒰𝒲του σιμωνος 892 A-E-M ℋ[𝒮]
5:3dαπο της γης επαναγαγειν B-ℵ-L-33-579-892 A-E-F ℬℋℳ𝒮𝒰𝒲επαναγαγειν απο της γης D
5:3eολιγον B-ℵ-L-33-579-892 A-E-F ℬℋℳ𝒮𝒰𝒲οσον οσον D
5:3fκαι καθισας 33-579-892 D A-E-Fκαθισας δε B-ℵ-L ℬℳ𝒮𝒰𝒲
5:3gεδιδασκεν εκ του πλοιου L-33-579-892 A-E-M ℬℋεκ του πλοιου εδιδασκεν B ℳ𝒮𝒰𝒲 ● εν τω πλοιω ℵ D
5:4ως δε B-ℵ-L-33-579-892 A-E-M ℬℋℳ𝒮𝒰𝒲οτε δε D
5:5a?και αποκριθεις ο σιμων ειπεν αυτω? 33-579-892 A-E-M ℬℋ (𝒮 και αποκριθεις [ο] σιμων ειπεν [αυτω])και αποκριθεις σιμων ειπεν B-ℵc ℳ𝒰𝒲 ● και αποκριθεις ειπεν σιμων ℵ* ● και αποκριθεις σιμων ειπεν αυτω L ● ο δε σιμων αποκριθεις ειπεν αυτω D
5:5bεπιστατα B-ℵ-L-33-579-892 A-E-M ℬℋℳ𝒮𝒰𝒲διδασκαλε D
5:5cτης νυκτος 579-892 D A-E-M ℋ[𝒮]νυκτος B-ℵ-L-33 ℬℳ𝒰𝒲
5:5d?χαλασω τα δικτυα? B-ℵ-L-579-892 ℬℳ𝒰𝒲χαλασω το δικτυον 33 A-E-M ℋ𝒮 ● ου μη παρακουσομαι D
5:6aτουτο ποιησαντες B-ℵ-L-33-579-892 A-E-M ℬℋℳ𝒮𝒰𝒲ευθυς χαλασαντες τα δικτυα D
5:6bπληθος ιχθυων B-ℵ-L-33-892 A-E-M ℬℋℳ𝒮𝒰𝒲ιχθυων πλυθος 579 D
5:6c?διερρηγνυτο δε το δικτυον? A-E-M ℋ𝒮διηρησεττο δε τα δικτυα B*-L-892 ℬℳ𝒰𝒲 ● διηρρησεττο δε τα δικτυα Bc-ℵ ● διηρρησεττο δε το δικτυον 33 ● διερρηγνυτο δε τα δικτυα 579 ● ωστε τα δικτυα ρησσεσθαι D
5:7aκατενευσαν B-ℵc-L-(33 [....]σαν)-579-892 A-E-M ℬℋℳ𝒮𝒰𝒲κατενευον D ● κατενευσεν ℵ*
5:7bεν τω ετερω B-ℵ-L-579 D ℬℳ𝒰𝒲τοις εν τω ετερω 33-892 A-E-M ℋ[𝒮]
5:7cσυλλαβεσθαι B-(συνλαμβανεσθε ℵ*)-(συνλαβεσθαι ℵc)-L-579-892 A-E-M ℬℋℳ𝒮𝒰𝒲 (33 uncertain)βοηθειν D
5:7dκαι ηλθον B-(και ελθαν ℵ-L BW)-33-579-892 A-E-M ℋℳ𝒮𝒰ελθοντες D
5:7eωστε B-ℵ-L-33-579-892 A-E-M ℬℋℳ𝒮𝒰𝒲ωστε παρα τι D
5:7fαυτα B-ℵ-L-579-892 A-E-M ℬℋℳ𝒮𝒰𝒲 (33 uncertain)omit D
5:8aιδων δε σιμων πετρος B-ℵ-L-579 A-E-M ℬℋℳ𝒮𝒰𝒲 (33 uncertain but has πετρος)ιδων δε ο σιμων 892 ● ο δε σιμων D
5:8b?τοις γονασι(ν) του ιησου? L-33-579 A-Mmarg𝒮τοις γονασι(ν) ιησου B-ℵ-892 E ℬℋtxtℳ𝒰𝒲 ● αυτου τοις ποσι D
5:8cλεγων B-ℵ-L-579-892 A-E-M ℬℋℳ𝒮𝒰𝒲 (33 uncertain)λεγων παρακαλω D
5:9aκαι παντας τους συν αυτω B-ℵ-L-33-579-892 A-E-M ℬℋℳ𝒮𝒰𝒲omit D
5:9bη συνελαβον ℵ-L-33-892 A-E-M ℬℋℳ𝒮𝒲margων συνελαβον B D 𝒰𝒲txt ● ην συνελαβον 579
5:10a?ιακωβον και ιωαννην υιους? L-33-579 A-E-M ℬℋℳ𝒮𝒰ιακωβον και ιωανην υιους B 𝒲 ● ιακωβος και ιωαννης οι υιοι ℵ-892 ● (rewrite entire verse) D
5:10b?σιμωνα ο ιησους? ℵ-33-579-892 A-E-M ℬℋℳ𝒮𝒰σιμωνα ιησους B-L 𝒲 ● (rewrite entire verse) D

So what does this tell us? Logically we would expect the Rule of Iron text to be closest to von Soden, since his I-H-K hypothesis resembles the rule we were using. Is it so? The table below shows the agreements with the Rule of Iron text for the leading witnesses and editors (note: this always follows the first hand of the manuscripts and the main text reading of the editions).

Note: there are 35 points of variation in our sample. In 19 of them, all the editions agree, so our real sample is the other 16 readings. Keep that in mind in assessing the rates of agreement below.


This is not as informative as we might like; all the editions have relatively similar numbers except for the dramatically low total for Westcott and Hort. There is, strikingly, no special kinship with Von Soden; rather, the closest of the critical editions is Bover. Hodges and Farstad exceed every critical edition except Bover. Thus it is clear that the Rule of Iron, if applied as I have outlined (and, of course, I don't know if it would have been), would have produced a more Byzantine text, and a much less Alexandrian text, than most of the modern editions.

Another vital point is to realize just how strongly a Rule of Iron method will depend on the three types chosen to use in reckoning the majority. Let's take Paul as our example this time: The standard post-Hortian model maintains that there are Alexandrian, Byzantine, and "Western" types, and that B is mostly Alexandrian but with "Western" mixture. Zuntz, Carlson, and I all find, in some way or another, that B is not an Alexandrian/Byzantine mix but that 𝔓46 and B form a separate type, not immediately akin to ℵ A C 33 etc.

So let's take three scenarios: Assume

We'll do Galatians 2. This time, I won't do the orthographic variants. Since we're doing three Rule of Iron types, I won't try to classify the witnesses; I'll just list them and then do the three types:
Zu = the Zuntzian reading
MH = Modified Hort
BW = B is Western

We'll also mark the groups where they disagree with the edition reading, e.g. Zu is the reading that the Rule of Iron would produce in the Zuntz scenario, but Zu1 is the reading of the first Zuntz group (𝔓46 B 1739) where it goes against Zu, Zu2 is the second Zuntz group (ℵ A 33); Zu3 is the third Zuntz group (D F G).

In cases where the Rule of Iron can't decide (e.g. 2:6, 2:14), I've put the reading I think Quentin would have chosen under each set of assumptions in parentheses, red, un-bold, e.g. (Zu).

So our full list of witnesses will be 𝔓46 ℵ A B D F G K L 056 33 1739, with each text based on three types based on three witnesses, so each text is ultimately based on nine of these twelve.

ReadingText of one or another groupOther variants
2:1παλιν ανηβην 𝔓46 ℵ A B 33 1739 K L 056 Zu MH BWανηβην παλιν D F G Zu3 MH2 BW2
2:4καταδουλωσουσιν ℵ A B D Zu MH BWκαταδουλωσωσιν 𝔓46 F G 33 1739 Zu1; καταδουλωσωνται K 056 MH3 BW3, καταδουλωσονται L
2:6προσωπον θεος ανθρωπου B K L 1739 (Zu) (BW) Zu1 MH3 BW3προσωπον ο θεος ανθρωπου 𝔓46 ℵ A 33 (MH) Zu2 MH1 BW1θεος ανθρωπου προσωπον D (F G) Zu3 MH2 BW2; προσωπον δε θεος ανθρωπου 056
2:9Ιακωβος και Κηφας ℵ B K L 056 33 1739 Zu MH BWΠετρος και Ιακωβος D(*) F G Zu3 MH2 BW2; Ιακωβος και Πετρος 𝔓46; Ιακωβος A
2:10των πτωχων ινα (hiat. 𝔓46) ℵ A B K L 056 33 1739 Zu MH BWινα των πτωχων D F G Zu3 MH2 BW2
2:11ηλθε(ν) Κηφας (hiat. 𝔓46) ℵ A B 33 1739 Zu MH1 BW1ηλθε(ν) Πετρος D F G K L 056 Zu3 MH BW
2:12ηλθεν 𝔓46 ℵ B D* F G 33 Zu MH BWηλθον A Dc K L 056 1739 MH3 BW3
2:13αυτων τη υποκρισει 𝔓46 (ℵ υποκρισε) A B K L 056 (1739) Zu MH BWτη υποκρισει αυτων D F G 33 Zu3 MH2 BW2
2:14τω Κηφα 𝔓46 ℵ A B 33 1739 Zu MH1 BW1τω Πετρω D F G K L (056 τωι Πετρωι) MH BW Zu3
2:14και ουχι Ιουδαικως ζης ℵc B (F G ουκ) Zu (Zu1) (Zu3) (BW) BW2ζης και ουχι Ιουδαικως D* (Dc L 056 ουκ) (K ουχ) MH BW3και ουχ Ιδαικως ℵ* A 33 Zu2 MH1 BW1ζης 𝔓46; ουκ Ιουδαικω ζης 1739
2:14πως τα εθνη 𝔓46 ℵ A B D F G 33 1739 Zu MH BWτι τα εθνη K L 056 MH3 BW3
2:16ειδοτες 𝔓46 A Dc K 056 33 1739 Zu BW MH3ειδοτες δε ℵ B D* F G L MH Zu3 BW2
2:16πιστεως Ιησου Χριστου 𝔓46 ℵ D F G K L 056 1739 Zu MH BWπιστεως Χριστου Ιησου A B(* πιστες Χριστου Ιησου) 33 Zu2 MH1 BW1
2:16εις Χριστουν Ιησουν 𝔓46 ℵ A D F G K L 056 Zu MH BWεις Ιησουν Χριστουν B 33 1739 Zu1
2:16νομου οτι 𝔓46 ℵ A B D* F G 33 1739 Zu MH BWνομου διοτι Dc K L 056 MH3 BW3
2:16εξ εργων νομου ου δικαιωθησεται 𝔓46 ℵ A D F G 33 1739 Zu MH BWου δικαιωθησεται εξ εργων νομου K L 056 MH3 BW3εξ εργων νομου δικαιωθησεται B
2:18συνιστανω 𝔓46 ℵ A B D* F(*) G 33 1739 Zu MH BWσυνιστημι Dc K L 056 MH3 BW3
2:20του θεου και χριστου 𝔓46 B D* F G Zu MH2 BW2του υιου του θεου ℵ A Dc K L 056 33 1739 Zu2 MH BW

Observe that all I have really done in this exercise is to reclassify B -- either lumping it with another text-type, Alexandrian or Western (which forces us to include the Byzantine text to grant us a third group) or splitting it off into its own type (which allows, or forces, us to drop the Byzantine text to keep the number of types at three). Observe that this one change in our assumptions resulted in six places in Galatians 2 where we would have had a different resultant text. Thus the Rule of Iron is extremely dependent on the exact data fed into it. Whether this is a defect is of course something the user must decide, but it is definitely an inherent aspect of the method.

If we look at the readings at those seven places, here is what we find:

You'll note that I produced a pure Zuntz text, while UBS waffled a bit. In the four instances where UBS is unequivocal, it agrees three times with Zu, once with MH, twice with BW. This tells us two things: One is to remind us, again, that in the Rule of Iron, it matters greatly just how our witnesses are classified. The other is to make it clear that the Rule of Iron isn't much influenced by internal criteria. Personally, I would say that, in a situation where one has a three-branch stemma, the Rule of Iron works fairly well -- but if you have a three-branch stemma, you can use classical Lachmann stemmatics. If you don't have a three-branch stemma, then the Rule of Iron is a classification disaster waiting to happen. The rigour of the idea breaks down in the face of the methodological weakness.