University of Virginia Library

Normalisation: Lexical Types

Type 1. A (Copy-Text) and H (Hocclex) Agreement

• a) Perfect ca. 66%
• b) Abnormal (very rare): e.g., H 2 x wol, 137 x wole
• c) Indifferent e.g., A whylom, H 1 x whylom, 4 x whilom. A made, H 28 x maad, 39 x made

Type 2. Normalisation by Extant Paradigm

• a) Perfect (i.e., 100% H usage): e.g., his (480/0), is (746/0), it (596/0), noght (19/0), noon (108/0), right (125/0), the (1342/0), thee (359/0), they (219/0), with (189/0)
• b) Composite e.g., bid-dynge (1+1/0), coyn-worth (8/0), meen-est (11/0), vn-seur (10/0)
• c) Analogous Forms e.g., H o x -þ, H o x sch; scholde>sholde, rekkyþ>rekketh

Type 3. Non-Extant Forms

• a) Possible Analogies e.g., A another (H othir), A asyde (H syde), A feyntyd (H feynte)
• b) Inflectional/Suffix Analogies e.g., A noysaunce, perchaunce; H substance, penance
• c) Non-Extant with Fisher/Chancery English Analogies?
• d) Non-Extant e.g., A Chestre ynne

The first type, with A (Arundel) and H (Hoccleve holograph/HOCCLEX) agreement, is observable at three levels (the Perfect, the Abnormal, and the Indifferent); the second, with normalisation by extant paradigm, is observable at the levels of Perfect, Composite, and Analogous; and the third, with non-extant forms (in HOCCLEX), is observable at four levels (Possible Analogies, Inflectional/Suffix Analogies, Non-Extant with Fisher/Chancery Analogies, and Non-Extant proper.

1a. Perfect Agreement: This level is illustrated throughout the text and can be seen quite clearly in the trial normalisation of two stanzas printed below.

It is, in practical terms, equivalent to all the cases fitting level one on the Normalisation Model, together with various proportions of those on levels two, three, and four. In brief, the great predominance of this type means that most of the features of A will still remain in the edited version. So, for example, in the first stanza of the extract, O, whyle, I, in, wele, I, was, honoured, And, many, on, of, my, glad, And, now, I, am, on, and, loured, There, how, wo, I, be, bystad, O, lord, &, vnsad, world, nat, mannes, persone, ffor, sone, but, for, good, allone—or forty-two words in fifty-four occurrences—are retained untouched (in this case a ratio of over seventy-five percent retention).

1b. Abnormal Agreement: In this type, there is still a clear agreement between A and H, but the form in A, while extant in H, is demonstrably a statistical aberration. (This unusual case is parallel to the "lower" statistical end of level seven on the Normalisation Model.) For example, the copy-text reading wol does exist in HOCCLEX, but only twice, as against 137 examples of wole. In the face of such overwhelming auctorial preference, the wole form is selected.[18] This type, we should emphasise again, is very rare.

1c. Indifferent Agreement: This third sub-group (properly identical only to level fourteen in the Normalisation Model, but possibly occurring elsewhere depending on one's concept of, and statistical base for, "indifference") illustrates similar indifference in both copy-text and HOCCLEX. Where, for example, A reads whylom and HOCCLEX provides virtually equal occurrences of both whylom and whilom, then the copy-text form would be followed.

2a. Perfect Normalisation: Just as in the first main group, the "Perfect" sub-group in Type 2 is again the most common and the most easily documented. It corresponds to levels two to six in the Normalisation Chart. It is the first case in the Lexical Types chart of major normalisation (i.e. an editorial change in the accidentals of the copy-text). For, in all examples encountered in Type 1 (with the exception of the rare and relatively insignificant 1b), the copy-text reading was retained. As can be seen from both the normalised text and the Lexical Types Chart, the great majority of cases of normalisation fall into this sub-group. Quite simply, this type shows, for example, that Hoccleve spells the third person singular present tense of the verb "to be" is 746 times, and never spells it any other way (e.g., as ys). In the face of such statistics, it would be a shirking of editorial responsibility to ignore this auctorial preference.

2b. Composite Normalisation: This sub-group consists of cases where the complete copy-text word-form is not extant in HOCCLEX, but where we can identify the necessary elements in the word and reconstruct it according to observable auctorial patterns. It therefore corresponds to various levels (from eight to twelve) on the Normalisation Model, depending on the statistical evidence. Thus, while HOCCLEX does not list the word coyn-worth (it may, in fact, be a nonce-word), it does show both coyn and worth, and we can therefore

speculate with some confidence that Hoccleve would have very much preferred the "-th" to the "-þ" form. This case illustrates the principle which we can often employ in such examples: that while the apparent statistical base may be very small (theoretically non-existent for this compound word), patterns demonstrable elsewhere can reinforce the evidence considerably. That is, since we know (as mentioned earlier) that Hoccleve uses the thorn ("þ") only in the contracted form þt and on no other occasion, the evidence supporting the normalisation to coyn-worth is very much more substantial than this single case might suggest.

2c. Analogous Normalisation: This type is a development of the arguments used in the previous sub-group and therefore also corresponds to levels eight to twelve in the Normalisation Model. For example, although rekketh does not exist in HOCCLEX, and although it is not strictly speaking a composite form but an inflected one, the evidence on "þ" employed in reconstructing coynworth can also be used to substantiate the "-th" inflection in rekketh as auctorial. (See below, on the use of the reverse lexicon, for other examples of inflectional analogies.) Similar evidence can be cited to emend the "longsch" form to "longsh" in such words as scholde or any other orthographic cluster patterns which are similarly consistent in the holographs (and are accessible, as mentioned above, through HOCCLEX's ability to pick out any specific morphological series, not just the word-unit itself).

3a. Non-Extant with Possible Analogy: One question that frequently confronts the Hoccleve editors is—just how common are genuinely non-extant forms in HOCCLEX that have no analogy to the documented forms? For example, the word another does not occur in HOCCLEX (as an orthographically compound word), but the word othir does, spelled with the "-ir" form not the "-er". Should we regard the two words as analogous, and therefore normalise the another to anothir, or should we consider them as representing two separate lexical units in the author's mind, and treat them differently (presumably normalising all copy-text other to othir, but retaining copy-text another because of the "empty" entry in the HOCCLEX column)? That is, are we using the level one or the level fourteen rationales in the Normalisation Model? In this particular case, I think one could argue that the etymology, morphology (and semantic connection) of the two words are similar enough to suggest that another should be dealt with as a type 2b or 2c normalised form. In fact, most examples of apparently non-extant words can be fairly clearly defended as belonging to one of the Type 2 lexical models. This ambiguous Type 3a is entered here largely to emphasise both the conservative ideology of the normalisation process (i.e., that the editor should at first consider whether a genuine anomaly has been found, in appropriate "classical" Pergamanian fashion) and the equally necessary—and in the Hoccleve holographs frequently more successful—conjecturalist search for analogies, in, of course, a "classical" Alexandrian manner.

3b. Inflectional/Suffix Analogies: A similar question to that posed in 3a might also be asked here, for although the principle of inflectional analogy was defended and illustrated under type 2b, it might be debatable whether a specific suffix or inflection could possibly change the lexical identity of the word in the author's mind in a manner similar to that suggested for the another/othir pair. For example, is it significant that Hoccleve, when choosing between an "-ance" and "-aunce" ending to abstract nouns, greatly prefers

the "-ance" on occasions when it can be considered as a genuine suffix (e.g. in non-monosyllabic words, remembrance, variance, ignorance), as opposed to the much rarer examples of "-aunce" like daunce or chaunce? Or more particularly, can such predilections be used in determining the form of abstract nouns which do not occur in HOCCLEX, e.g. perchaunce, noysaunce? The answer to such paradigmatic questions, and the decision to regard this Type as belonging to level thirteen or one of the NORMLEX levels (eight to twelve) in the Normalisation Model will clearly depend on the statistical and morphological consistency that each suffix presents.

3c. Non-Extant with Fisher Analogies: As mentioned earlier, we had at one point hoped to be able to use the 90,000-word Fisher concordance of bureaucratic Middle English to widen the evidence for morphological extension. Ironically, however, the level of consistency found in Hoccleve seems in general to be even higher than that charted by Fisher, so that while the HOCCLEX concordance is inevitably smaller than the Fisher, it is very much more amenable to the establishment of accidental patterns usable in the normalisation of copy-text. To date, therefore, we have not been able to employ the Fisher concordance with a surety greater than that accorded the HOCCLEX.

3d. Genuine Non-Extant Forms: With all the various qualifications listed above, and with the evidence of the normalised text printed below, it should be clear that genuinely non-extant forms are comparatively rare. That is, it is very unusual for us to come across a word in the copy-text for which there is no paradigm available in HOCCLEX or NORMLEX, either by analogy, composition, or inflection. When such words do occur, they are frequently proper nouns (e.g. Chestre ynne, on the fifth line of the poem), and given the generally-accepted dubiousness of the proper noun in other textual theories (e.g., stemmatics), we are content to allow copy-text the primacy. These cases correspond to level thirteen or even fourteen in the theoretical Normalisation Model, where the "residual authority" of copy-text can be safely invoked without compromising the argument for normalisation.

As mentioned above, one of the potentially most useful features introduced into the HOCCLEX program was the reverse lexicon facility. This automatic rewriting of the basic text-data would allow us to gain access to the concordance not only through conventional alphabetisation, but also through word-endings, thereby making it possible to analyse the holograph morphology for inflections, suffixes, etc. Rather than simply re-store (and re-print) every occurrence of each word both backwards and forwards, the HOCCLEX program uses a code number system (as shown in the listings printed below) which can key the entry in the reverse lexicon back to the similarly numbered main entry in the concordance proper. A typical section of the reverse lexicon appears below.

• 3435) SI
• 3230) SIH
• 2273) SILLE
• 1826) SINIDIPUC
• 1922) SITIECED
• 6099) SIY
• 1934) SIDEED
• 5316) SIHT
• 4034) SIM
• 1806) SIORC
• 5400) SITNEMROT
• 4149) SIDEEN
• 1541) SIKRELC
• 867) SIMA
• 5658) SIOV
• 3451) SIWI
• 1399) SIEKRAC
• 5788) SIKREW
• 4169) SIN
• 5413) SIRUOT
• 6143) SIWY

This section of the lexicon, therefore, lists every form in the holographs ending in "-is" (including, obviously, the word is itself—entry code 3435). Admittedly, a reverse alphabet does not make for easy reading, and one has to "reform" each word mentally in order to give it its conventional value (although we could have the computer simply print the reverse examples in the orthodox forwards manner). Now, using the key-code, we can trace each form back to the main entry in the concordance and thus to the text itself.

• 3435) IS (There are 746 listings in the concordance for this entry. They are not given here.)
• 1934) DEEDIS 3861 (i.e., line number in HOCCLEX text) TOTAL 1
• 4149) NEEDIS 3863 TOTAL 1
• 3230) HIS (There are 480 listings in the concordance for this entry. They are not given here.)
• 5316) THIS (There are 416 listings for this entry in the concordance. They are not given here.)
• 1541) CLERKIS 2829 TOTAL 1
• 5788) WERKIS 2831 TOTAL 1
• 2273) ELLIS 5795 TOTAL 1
• 4034) MIS (29 listings in concordance)
• 867) AMIS (12 listings)
• 4169) NIS 464 TOTAL 1
• 1826) CUPIDINIS [now deleted in the lexicon] TOTAL 1
• 1399) CARKEIS 1065 TOTAL 1
• 1806) CROIS (13 listings)
• 5658) VOIS (7 listings)
• 5413) TOURIS 3325 TOTAL 1
• 1922) DECEITIS 6573 TOTAL 1
• 5400) TORMENTIS 5702 TOTAL 1
• 3451) IWIS 4414, 4726 TOTAL 2
• 6143) YWIS 3839 TOTAL 1
• 6099) YIS (14 listings)

Omitting the very common words (is, his, this, mis, amis, crois, vois, yis) for reasons which will become clear in a moment, we can produce the following lines from the text of the holographs:

• 3861) A TALE EEK7[19] / WHICH I IN THE ROMAUN DEEDIS
• 3863) OF YOW MY LADYES AS I MOOT NEEDIS
• 2829) LADYES EEK7 COMPLEYNEN HEM ON CLERKIS
• 2831) IN WHICHE THEY LAKKEN WO6MENNES WERKIS
• 5795) HELP NE REFUYT IS THER FOR THEE NOON ELLIS
• 464) IT NIS NE 2HAL BEEN / AS YEE HAN SUPPOSID EXPLICIT EXPLA CUPIDINIS [now deleted in the lexicon]
• 1065) AND MY CARKEIS REPLEET WITH HEUYNE22E
• 3325) PALLE2ES / MANERS / CA2TELS GRETE AND TOURIS
• 6573) THAT HAD HIM TERNED WITH FALS DECEITIS

---

142




• 5702) BUT ABOUE ALLE KYNDES OF TORMENTIS
• 4414) GRANT MERCY 2EIDE THE 2HIPMAN IWIS
• 4726) A FAIR LADY HE FOND HANGYNGE IWIS
• 3893) VN TO THE HOLY LOND I WOLE YWIS

The next stage is to compare this group as a whole with other reverse entries in order to establish the comparative data from which possible patterns of auctorial usage might be discerned. Take, for example, the group of entries in the reverse lexicon for words ending in "-ys". The following listings appear:

• 4430) SYDARAP
• 4501) SYEP
• 4688) SYRP
• 5652) SYWNV
• 4422) SYELAP
• 1837) SYETRUC
• 4619) SYTNERP
• 4971) SYWTHGIR
• 3417) SYEN40I
• 2337) SYMENE
• 1010) SYUA
• 2804) SYENRUOF
• 1508) SYOHC
• 6033) SYW

Note, first of all, that there is no equivalent listing in the "-ys" group for the is, his, and amis recorded in the "-is" group, thereby establishing that there is 100% consistency in the form of these common words. This information could, however, have been gathered easily enough from a conventional alphabetised concordance, and the usefulness of the reverse lexicon is to measure the frequency and nature of the "-ys" group as a whole against the entries found in the "-is" group. Following the same procedure as before, we can trace the occurrences of all the "-ys" words to the text itself, via the main entry in the main concordance.

• 4430) PARADYS 1592, 5412 TOTAL 2
• 4422) PALEYS 1572, 3112, 4814 TOTAL 3
• 3417) I04NEYS 6491 TOTAL 1
• 2804) FOURNEYS 5459 TOTAL 1
• 4501) PEYS 1484 TOTAL 1
• 1837) CURTEYS 4528, 5770 TOTAL 2
• 2337) ENEMYS 5506 TOTAL 1
• 1508) CHOYS 5997 TOTAL 1
• 4688) PRYS (7 listings)
• 4619) PRENTYS 1036 TOTAL 1
• 1010) AUYS 3551, 3674, 3691, 3702 TOTAL 4
• 6033) WYS (9 listings)
• 5652) VNWYS 77, 2481, 5508 TOTAL 3
• 4971) RIGHTWYS 5587 TOTAL 1

These entries appear in the text as:

• 1592) PARADYS YATES OPNED BEEN BY THEE
• 5411) O GOOD LORD GOD / O LORD OF PARADYS
• 1572) PALEYS OF CRY2T / FLOUR OF VIRGYNITEE
• 3112) IN OUR PALEYS / WHER MANY A MILION
• 4814) HOOM VN TO HIS PALEYS THIS EMPEROUR
• 6491) FFORTH HIS I04NEYS THIS IONATHAS HEELD
• 5459) X. YEER IN A FOURNEYS BRENNYNGE THER7
• 1484) OR NAT BY IU2T PEYS MY 2ENTENCES WEYE
• 4528) CURTEYS IHU ME KEEPE NOW THIS HOUR
• 5770) AND CURTEYS LORD / OF THY BENIGNITEE

---

143




• 5506) THUS BODYES FREENDES BEEN MAAD ENEMYS
• 5997) IN OUR CHOYS AND ELECCOUN 1 TO GRETE FOOLES
• 1036) I DAR NAT 2EYN PRENTYS AND ARONDEL
• 3551) AS 5 * IT 2EEMETH TO MY 2YMPLE AUYS
• 3674) SHAL HAUE / NEEDITH GOOD AUYS 2OOTHLY
• 3691) SMERTITH THE FOOL FOR LAK7 OF GOOD AUYS
• 3702) THOMAS 2AUF BETTRE AUYS I HOLDE IT GOOD
• 77) THOW ART VNWYS / THOGH THOW THE WYS P3TENDE
• 2481) AND HE A CHAPMAN IS NAT 2O VNWYS
• 5508) THE MAN CONTINUELLY 1 YIT 2O VNWYS
• 5587) THE RIGHTWYS MAN / OR 2LEE HIM 2ODEYNLY

Certain patterns now begin to appear. At first glance, it might seem that the incidence of "-is" and "-ys" forms is virtually indifferent, for there are twenty-one entries in the reverse lexicon for "-is" and fourteen for "-ys". But there are several qualifications to be made to this apparent indifference, some of them directly helpful in the establishing of auctorial norms. Thus, it can be seen that in fact none of the fourteen entries in the "-ys" group (as observable in the thirty-seven actual occurrences in the text) are genuine inflections in "-ys". Some (paradys, paleys, fourneys, peys, choys, prys, prentys, auys, wys, vnwys, and rightwys) are simply part of the basic root of the word or its composite form, and while they might be useful in charting auctorial spelling preferences for these individual words—or the morphological types they represent—they are irrelevant to a determination of a preference between the "-is" and "-ys" as suffixes or inflections. Others (io4neys, enemys), while showing inflected forms (plurals), are based on a root which already ends in "y" and can therefore be similarly disregarded as evidence for a "-ys" / "-is" inflectional choice. When we compare this poor showing of the "-ys" group with the evidence of the "-is" group, we can first admit that, again, there are several "-is" forms which are part of the root of the word itself (is, his, this, ellis, amis, carkeis, crois, vois, and yis). There are also three contracted or composite forms (nis, iwis, ywis) which similarly have no inflectional significance. The Latin entry (cupidinis) is obviously of no value to the English accidence (and has in fact been deleted from the concordance), and the prefix mis, while clearly confirming an auctorial preference for the "is" spelling over the "ys", should not be cited in the inflectional evidence. However, that still leaves deedis, needis, clerkis, werkis, touris, deceitis, and tormentis, all of which are genuine plural forms in "-is" and suggest (unlike any form in the "-ys" group), a distinct auctorial pattern for this inflection.

This brief analysis of one very small variant as charted in the reverse lexicon is obviously both incomplete and constructed on such a small data-base as to make absolute surety difficult to demonstrate. It is offered here largely because the data-base is so small and can therefore be presented

SAMPLE NORMALISATION OF TEXT

                                                           

Copy-Text Transcript	Normalised with Annotations	Clear Text Edited Version
(101)	(101)	(101)
	(9/0)
O whyle I stode in welesemiinv. I was honoured	O whyle I stood in wele I was honoured[1]	O whyle I stood in wele I was honoured
	(15/0)
And many on of my compaygne glad	And many on of my conpaignie glad	And many on of my conpaignie glad
	(27/0) (2/0+33/4)
And now I am mylongslokyd on and loured	And now I am milongslookid on and [loured]	And now I am milongslookid on and loured
	(108/0)
There rekkyþ non; how wo I be bylongstad	There rekketh [2] noon how wo I be bylongstad	There rekketh noon how wo I be bylongstad
	(418/1) (50/1) (3/0) (746/0)
O lord thys worlde vnlongstabel ys & vnlongsad	5 O lord this world vnlongstable is & vnlongsad	5 O lord this world vnlongstable is & vnlongsad 5
	(418/0)
Thys world honnuryþ nat mannes perlongsone	This world hon[ury]th [3] nat mannes perlongsone	This world honnuryth nat mannes perlongsone
	(254/10) (41/0)
ffor hym longsylfsemiinv. longsone but for good allone	ffor him longself sone but for good allone	ffor him longself longsone but for good allone
(102)	(102)	(102)
	(17/0+9/1)
fful longsothe fynde I the word of longsalomon	fful longsoothe fynde I the word of longsalomon	fful longsoothe fynde I the word of longsalomon
	(23/0)
That to many obeyen al thynkys	That to many obeyen al thynges	That to many obeyen al thynges
	(8/0) (746/0) (4/0)
ffor that my coyn and coynworþ ys a gon	10 ffor that my coyn and coynworth is agoon	10 ffor that my coyn and coynworth is agoon
	(29/0+37/0) (1/0+1/0)
Contrarien they my wylle and my byddynge	Contrarien they my wil and my biddynge	Contrarien they my wil and my biddynge 10
	(189/0+45/0) (415/0)
That in my welthe wyth her flaterynges	That in my welthe with hir flaterynges	That in my welthe with hir flaterynges
	(189/0+45/0) (109/2 + 13/2)
Helden wyth me what þt I wroght or longseyde	Helden with me what þt I wroght or longseide	Helden with me what þt I wroght or longseide
Now dilongsobeyen they that thanne obeyde	Now dilongsobeyen they that thanne obeyde	Now dilongsobeyen they that thanne obeyde

virtually intact (with the exception of, for example, the total listings for the contexts of such common words as his, is, and this). It is intended merely to show the method for establishing paradigms via the reverse lexicon, and to chart a typical example of consistency of usage, resulting in a 100% figure, even though in this case that 100% represents only seven different words in one column, with a zero showing in the other possible choice. If this example were used as a means of determining a system of accidence for all plural nouns in Hoccleve, it would obviously be a misguided editorial procedure (for example, we would first have to add to the data-base all other possible plural forms which might compete with the recorded "-is" preference. However, it can tell us that, in this limited case, Hoccleve never uses the "-ys" form as a genuine plural inflection and, therefore, that a "-ys" plural in copy-text should be viewed with extreme suspicion, and very likely emended by reference to this "-is" group (and certainly if the copy-text produced a "-ys" for any of the seven words in the control group itself).

The first point to be emphasised is that this brief selection from the text of the Hoccleve Regement, in its copy-text, annotated, and clear-text forms, is intended to provide as wide a range as possible of the "levels" of normalisation and the various editorial procedures used to deal with these levels. The text has, therefore, been consciously "over-normalised", in the sense that any possibility for normalisation has at least been charted in the annotations. The text as a whole has, in general, a much lower level of editorial participation, although even this particularly problematical section does display the typically reassuring accumulation of evidence (from HOCCLEX, the reverse lexicon, and NORMLEX) upon which decisions to normalise can usually be made with confidence.

The first column is a diplomatic transcript of stanzas 101 and 102 of the Arundel manuscript. Word-division and punctuation in the manuscript are retained, and there is none of the conventional normalisation often occurring in otherwise conservative editions (e.g., initial "ff" is not normalised to capital "F", although it clearly has this function throughout the manuscript). The purpose of providing the copy-text transcript is merely to show the typical raw material with which the Hoccleve editors are faced, and, of course, to demonstrate that, despite the apparent complexity of the annotations in the normalised version, the rate of editorial normalisation per word of Arundel text is really quite modest.

The annotated version can be read as follows:

As is rather obvious, we are very doubtful that the published form of the normalised text of Hoccleve's Regement will have the luxury of being able to present all of this annotated material on the textual page, and we suspect that the famous "barbed wire" of editorial intervention would in this case severely interfere with the reader's access to Hoccleve's text. It is presented in this state here largely to demonstrate the type, range, and level of surety the normalisation process encounters. The explanatory material could no doubt work equally well in a conventional apparatus (distinguished, of course, from substantive emendation, which certainly occurs much less frequently), with appropriate lemma-keying to the stanzaic page.

The laws of textual disclosure might indeed suggest that the traces of our editorial handiwork with regard to accidentals should be made available to the critical reader in the apparatus, so that the complete textual features—surface and otherwise—of the copy-text could be recreated

out of this apparatus. As mentioned earlier, such a desideratum would involve a lemma-entry in the apparatus for roughly one-third of the individual words in the copy-text. Given that only one witness—Arundel itself—would need to be cited (compared to the more than forty witnesses whose readings would have to be listed in a complete historical collation of the rarer substantive variants), this level of editorial openhandedness would not necessarily result in an apparatus more unwieldy than those already produced for other mediaeval works, particularly those with multiple authority. However, it is debatable whether all of this material would be required by the typical reader (historian, political scientist, literary critic) of the stanzaic page, and it is likely that the apparatus reconstructing the accidentals of Arundel will occur in a separate microform addendum to the edition, along with the complete historical collation. It might be argued that, since Arundel is only a vehicle (albeit a relatively consistent and reliable one) for the reconstruction of auctorial accidentals, then neither editor nor reader should be greatly concerned with its surface features per se, and that even a microform accidentals apparatus is thus an editorial luxury. However, we do not want the normalisation to be entirely silent, and therefore the microform addendum can at least function as a record of what we have done, demonstrating we trust the validity and coherence of our editorial decisions.[20]

The confidence with which normalisation may proceed obviously depends upon our following a procedure which is careful, consistent, and conservative. When the normalised form is created, it must be based on a rigorous application of formal principles of documentary evidence, and all dubious cases must be emended only as far as that documentary evidence and its analogies will allow. In our case, these principles translate into the following practices:

The specific conditions of the Hoccleve oeuvre, with a substantial body of holographs of minor works, together with the text of a major work available only in many scribal copies, are no doubt so unusual that it would be difficult to find immediately useful theories or practices derived from the Hoccleve editing that could be incorporated into the editorial principles of scholars working either in mediaeval or other periods. With the exception of the Ormulum and possibly the The Equatorie of the Planetis mentioned earlier (see footnote 9)—neither of which is really parallel to the Hoccleve problem—there are virtually no examples in Middle English literature of a document purporting to represent auctorial preferences in accidentals. Where holographs do exist in other periods, they are, of course, most often versions of the very work being edited, and may either be rejected as copy-text in favour of a printed edition or used for accidentals according to the principles enunciated by Greg and his followers. And when putative holographs of works other than that being edited do occur, their possible value as paradigms for accidentals or other auctorial preferences (e.g., of style, imagery etc.) has more often been the source of editorial contention than consensus. One thinks inevitably of the various attempts to determine the content and style of the entire Shakespeare canon from the characteristics of the famous "Hand D" in the fragment of Sir Thomas More. The issue may simply be that the desired consistency in accidentals is not demonstrably clear enough in the limited data available, and that therefore the admittedly scribal/compositorial features of a non-auctorial copy (printed or otherwise) are less problematical than an attempt to resuscitate auctorial preferences from such holographs, especially where a copy-text reasonably close to the author's time and language may be selected.

But the Hoccleve holographs do not have such limitations. They contain a comparatively large body of material (in fact, nearly 7,000 lines of verse, a slightly larger lexicon-base than the poem of 5,000 lines for which

they provide the comparative data). They display a quite remarkable degree of consistency—more so than any previous scholars of fifteenth-century English had anticipated; and while their lexicon is obviously not completely identical to that created by the necessities of the different subject-matter of the Regement, the lexical overlap is so practically useful that there are very few occasions where the editor is left without some paradigm or other. And finally, unlike the Chaucer Equatorie, the Hoccleve holographs are in the same medium (verse) as the rest of the oeuvre, and usually employ the same metrical and strophic patterns. For all of these reasons, the Hoccleve holographs may not be denied their status as valuable evidence in the editorial process: to do so would be to accept an acknowledged non-auctorial condition of the text in favour of one which can, with careful scholarship, be shown to rely upon the demonstrable preferences of an author who clearly cared very much about the exact form of his language.[21] We do not deny the apparent novelty of our methods, but claim that in fact they fall well within the accepted definitions of editorial theory and practice: as suggested several times in this essay, we have simply combined the "classical" Lachmannian orthodoxy in selection of a "base-text" for substantives with a logical extension of the "modern" Gregian orthodoxy in the selection of a "copy-text" for accidentals. The paradox of Thomas Hoccleve is that these two documents, base-text and copy-text, happen to be of different works, but that their combination will result in a final text which is closer to auctorial intentions than the use of either document independently would have brought us.